Meta-analysis of microarray gene expression studies on intracranial aneurysms

Neuroscience 201 (2012) 105–113

META-ANALYSIS OF MICROARRAY GENE EXPRESSION STUDIES ON INTRACRANIAL ANEURYSMS C. RODER,a H. KASUYA,b A. HARATI,c M. TATAGIBA,a I. INOUEd AND B. KRISCHEKa*

The rupture of intracranial aneurysms (IAs) and resulting acute subarachnoid hemorrhage (SAH) is one of the most devastating acute neurological conditions known to date. Although technical advances have changed the treatment of IAs dramatically in the last years, the outcome of patients still has a poor prognosis. The formation of IAs is assumed to be caused by diverse exo- and endogenous factors such as inflammatory mechanisms, hypertension, smoking, excessive alcohol intake and genomic factors (van Gijn et al., 2007; Kataoka and Aoki, 2010). Affected intracranial vessels are known to have a unique histological structure within the human body. They consist of a tunica intima with an endothelial layer and smooth muscle cells (SMCs), a tunica media with the internal elastic lamina and SMCs, and a tunica adventitia composed of loose connective tissue. IAs usually develop at branching points within the circle of Willis. At these branching points the continuity of the tunica media shows gaps, also called raphés, which are thought to be a predisposing location for the development of aneurysmatic pouches. Microscopic findings such as the lack of an outer elastic lamina, disorganization of extracellular matrix, atherosclerotic and inflammatory changes caused by complement activation, a decrease in reticular fibers, and many others, in combination with the raphés, continuous blood pressure, and shear stress at the points of bifurcation, seem to play an important role in the development of IAs (Krischek and Tatagiba, 2008; Tulamo et al., 2010). A precise histological analysis of these changes was published by Frösen et al. (Frösen et al., 2004). As the etiology appears to be a very complex process, many approaches such as macroscopic, microscopic, or molecular techniques were aimed to identify factors leading to the development of IAs. Known association with other genetic diseases such as polycystic kidney disease or Ehlers–Danlos syndrome, as well as reported familial cases, indicate possible genomic causes (Ronkainen et al., 1997; Krischek and Tatagiba, 2008). Large efforts have been put into the research of genomic factors throughout the last years, but the comprehension of interaction between multiple genomic and/or environmental factors contributing to the development of intracranial aneurysms appears to be a challenging task (Kataoka and Aoki, 2010). Genetic research on IAs pursued four major approaches throughout the last years: Linkage studies, candidate gene association studies, genome-wide association studies (GWAS), and gene expression profiling. Linkage studies are used to identify cosegregation of possible IAcausing genomic markers in affected families. By performing candidate gene association studies genes that are

a Department of Neurosurgery, University of Tübingen, Hoppe-SeylerStr. 3, 72076 Tübingen, Germany b

Department of Neurosurgery, Medical Center East, Tokyo Women’s Medical University, 2-1-10 Nishiogu, Arakawa-ku, Tokyo 116-8567, Japan c

Department of Neurosurgery, Klinikum Vest Recklinghausen, Ruhr University Bochum, Dorstener Stra␤e 151, 45657 Recklinghausen, Germany d

Division of Molecular Life Science, School of Medicine, Tokai University, Shimokasuya 143, Isehara, Kanagawa, Japan

Abstract—The rupture of intracranial aneurysms (IAs) is one of the most devastating neurological conditions known to date. Although treatment has changed dramatically throughout the last decades, the outcome of patients still has a poor prognosis. Besides environmental factors, genomics seem to be a very important factor in the genesis of this disease. Different approaches to decrypt genomic causes were pursued throughout the last years. Microarray gene expression studies comparing aneurysmal and healthy tissue seem to be one of the most promising approaches. However, large amounts of data created with each study, make a comparison or interpretation of results difficult. We analyzed microarray gene expression studies on IAs (vs. control tissue) and compared lists of genes with altered expression provided by the authors. Additionally functional pathway analysis was performed. We identified five microarray gene expression studies analyzing a total of 60 samples of IA tissue (30 ruptured IA, 30 unruptured IA). A total of 507 genes with altered expression were listed, of which 57 showed differences in more than two studies and seven in more than three studies (BCL2, COL1A2, COL3A1, COL5A2, CXCL12, TIMP4, TNC). The metaanalysis of five microarray gene expression studies on IAs revealed seven genes that are very likely to be involved in the genesis of IAs. Further analysis of these genes might provide valuable information on mechanisms causing this disease. © 2011 IBRO. Published by Elsevier Ltd. All rights reserved. Key words: genetics, gene expression, intracranial aneurysm, subarachnoid hemorrhage, stroke. *Correspondence to: B. Krischek, Division of Neurosurgery, Toronto Western Hospital, University of Toronto, 399 Bathurst Street, Toronto, ON, Canada M5T 2S8. Tel: ⫹1-416-301-9927; fax: ⫹1-416-913-1131. E-mail address: [email protected] (B. Krischek). Abbreviations: AVM, arteriovenous malformations; BP, GO Biological Process; CC, GO Cellular Component; ECM, extracellular matrix; GO, gene ontology; GWAS, genome-wide association studies; IAs, intracranial aneurysms; KEGG, Kyoto Encyclopedia of Genes and Genomes; MF, GO Molecular Function; MMA, middle meningeal artery; MMP, matrix metalloproteinase; RA, ruptured aneurysm; SAGE, serial analysis of gene expression; SAH, subarachnoid hemorrhage; STA, superficial temporal artery; TNC, tenascin C; TIMPs, tissue inhibitors of MMPs; UA, unruptured aneurysm. 0306-4522/12 $36.00 © 2011 IBRO. Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.neuroscience.2011.10.033

105

106

C. Roder et al. / Neuroscience 201 (2012) 105–113

thought to be involved in the development of IAs are analyzed (totally or partially) and compared with healthy controls. GWAS aim to identify disease-related genetic variants by investigating single nucleotide polymorphisms across the whole genome based on a case-control study design. Gene expression studies are capable of identifying differences in transcription of thousands of genes on a genome-wide scale by using microarrays. Under the assumption of a multi-factorial and/or multi-genetic etiology of IAs, the latter technique may indicate involved genes and altered molecular pathways. One of the most crucial aspects of this kind of study-design is what kind of controltissue is used for comparison. As tissue of intracranial arteries cannot be harvested from healthy patients, researchers have used superficial temporal artery (STA), middle meningeal artery (MMA), or feeding arteries of arteriovenous malformations (AVM) (Krischek et al., 2008; Li et al., 2009; Shi et al., 2009; Marchese et al., 2010; Pera et al., 2010). One of the first genome-wide comparisons of gene expression was performed by Peters et al. (Peters et al., 2001) in 2001 who used the “Serial analysis of gene expression” (SAGE-Lite) technique with tissue specimen of one 3-year-old female patient comparing aneurysmatic with STA tissue. To date, five microarray mRNA profiling studies comparing aneurysmatic with control tissue in a genome-wide expression analysis have been published identifying possible disease causing pathways and each focusing on different hypotheses concerning the development of IAs (Krischek et al., 2008; Li et al., 2009; Shi et al., 2009; Marchese et al., 2010; Pera et al., 2010). Combination and comparison of these studies may have the potential to substantiate and filter the results of each single study and may provide further insight into the pathogenesis of IAs. Two general approaches for the integration of different microarray studies are known: re-analysis of primary data sets by merging data from multiple studies or the comparison of published results/gene-lists (Cahan et al., 2007). The re-analysis of raw-data has several limitations, such as the use of different platforms, different gene nomenclature, and the use of different control tissues. But one of the main shortcomings is the limited amount of data deposited in public databases so far (Griffith et al., 2006; Cahan et al., 2007). However, comparable gene-lists can be found in almost every publication on gene expression studies of intracranial aneurysms and therefore provide the possibility to compare and analyze results. In this study we perform a meta-analysis of all published whole-genome microarray gene-expression studies based on gene-lists provided by the respective authors (Krischek et al., 2008; Li et al., 2009; Shi et al., 2009; Marchese et al., 2010; Pera et al., 2010). Our article includes a short compendium of each studies’ results, the collection of the published genes, comparative analysis of over- and under-expressed genes as well as functional pathway analysis with gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG).

EXPERIMENTAL PROCEDURES Data collection We queried Pubmed and analyzed literature for microarray gene expression studies comparing IA wall with control tissue. The following key words (and their combinations) were used: “intracranial aneurysm, gene expression, genetics, review.” We analyzed the literature that has been published between 2000 and June 2011. Non-human studies were excluded, as well as candidategene or candidate-region expression studies not following a whole genome microarray approach. All genes were listed alphabetically in a Microsoft Excel sheet along with information on the original publication (Supplemental data with a complete gene list is available upon request).

Comparative analysis Genes with over-expression were labeled with 1, under-expression 2, equal-expression ↔. Names of genes with common differences in two or more studies were labeled with 11 for common over-expression, 22 for common under-expression, ‡ for differing results. If no information on a specific gene was available, no symbols were used. In case of availability of fold-change rates of expression, these were included in the chart: [0, 0.2] as “---”, (0.2, 0.5] as “--”, (0.5, 0.8] as “-”, [1.2, 2) as “⫹”, [2, 5) as “⫹⫹”, and [5, ⬁) as “⫹⫹⫹”. Genes with altered expression levels in more than two studies were listed in an additional chart including the chromosomal position (Table 1). Genes with altered expression in three or more studies were labeled with an asterisk.

Functional classification Functional classification was performed by using the Genecodis 2.0 online platform (Carmona-Saez et al., 2007; Nogales-Cadenas et al., 2009). The settings “homo sapiens” for organism, “GO Biological Process (BP),” “GO Molecular Function (MF),” “GO Cellular Component (CC),” and “KEGG Pathways” for annotations, and “lowest level” for GO Levels were used. For functional classification analysis, gene listings were separated into the following two sub-groups: 1) Genes with altered expression 2) Genes with altered expression in two or more studies Genes with altered expression in three or more studies were not analyzed separately due to small numbers. In Table 2 the top three results of each sub-group analysis are listed.

RESULTS We identified five studies that analyzed a total of 60 aneurysms, 30 were ruptured (RA) and 30 unruptured (UA) aneurysms (Krischek et al., 2008; Li et al., 2009; Shi et al., 2009; Marchese et al., 2010; Pera et al., 2010). Short compendium of the studies’ results In 2008 Krischek et al. (Krischek et al., 2008) analyzed the gene expression of six ruptured and four unruptured IAs compared to four AVM feeder-arteries using the Agilent RNA 6000 Nano LabChip. The results showed differences in expression of 521 genes, 263 were over-expressed, 258 under-expressed. GO analysis showed that antigen processing was the most significantly associated term within these genes (P⫽1.64E-16). Functional network-based analysis (ingenuity pathway analysis) revealed highest association (over-expression) of major histocompatibility

C. Roder et al. / Neuroscience 201 (2012) 105–113

107

Table 1. Genes with altered expression in two or more studies. Genes with over-expression were labeled with 1, under-expression 2, equalexpression ↔. Names of genes with common differences in two or more studies were labeled with 11 for common over-expression, 22 for common under-expression, ‡ for differing results. If no information on a specific gene was available, no symbols were used. Genes with altered expression in three or more studies were labeled with an asterisk. In case of availability of fold-change rates of expression, these were included in the chart: [0, 0.2] as “---”, (0.2, 0.5] as “--”, (0.5, 0.8] as “-”, [1.2, 2) as “⫹”, [2, 5) as “⫹⫹”, and [5, ⬁) as “⫹⫹⫹” Author:

Marchese et al. (2010)

Pera et al. (2010)

Li et al. (2009)

Shi et al. (2009)

Krischek et al. (2008)

Specimen:

RA vs. MMA/STA

RA vs. MMA

UA vs. STA

RA/UA vs. RA/UA STA

RA/UA vs. AVMf

Platform:

Affymetrix

Affymetrix

Affymetrix

Illumina

Agilent

Number of specimen:

12 RA vs. 5 control MMA/STA

8 RA vs. 5 MMA

6 UA vs. 5 MMA

3 UA vs. 3 STA

3RA/3UA vs. 3RA/3UA STA

6 RA, 4 UA vs. 4 AVMf

2(--)

2(---)

2

Chromosome:

Gene:

21q21.2 17q21 10q11.2 18q21.3 22q11.1 1p13.3-p11 7q31.1 5q31.1 17p13 19p13.3 17q21.33 7q22.1 2q31 2q35-q37 9q34.2-q34.3 2q14-q32 8p22 20q13 10q11.1 5q31 2q21 2q35 1q12-q23 5q23-q31 1q21.2-q21.3 Xq28 2q34 6p21.3 6p21.3 6p21.3 6p21.3 21q22.11 12q23.2 4q21 7p21 12q13 5q23.2 12q15 10q23.2 8q24.21 3q21 13q13.3 12q15-q21 19q13.2 6p21.3 2p24

ADAMTS1‡ AOC322 ALOX511 BCL2*22 BID11 CASQ222 CAV122 CD1411 CD6811 CFD22 COL1A111 COL1A2*11 COL3A1*‡ COL4A4‡ COL5A111 COL5A2*11 CTSB‡ CTSZ11 CXCL12*‡ CXCL1422 CXCR411 DES22 DPT22 FBN2‡ FCGR1A11 FLNA22 FN111 HLA-B‡ HLA-C‡ HLA-DRB1‡ HLA-G‡ IFNGR211 IGF122 IGJ‡ IL622 KRT1811 LOX‡ LYZ11 MMRN222 MYC22 MYLK2 POSTN11 PPP1R12A‡ PPP1R15A‡ PSMB8‡ ROCK2‡

UA vs. MMA

1(⫹⫹⫹)

2

2(--) 1(⫹⫹)

2 2(---)

2(---) 1(⫹⫹) 2

1(⫹⫹) 2(--) 2(--)

1 1

1 1(⫹⫹) 1(⫹)

2(--) 1(⫹⫹⫹) 1(⫹⫹⫹) 2(---) 1(⫹⫹⫹) 1(⫹⫹) 1(⫹⫹⫹) 1(⫹⫹) 1(⫹⫹⫹) 1(⫹⫹⫹) 1(⫹⫹⫹)

1(⫹⫹) 2 2

2(--) 2(--)

2(---) 2(-)

1(⫹)

1(⫹⫹)

2(--)

2(--) 2(---) 1(⫹⫹⫹) 2(---)

2(--)

2(--)

1 1 1 1 1 ↔ 1 1 1

1 2(--) 1(⫹⫹)

1(⫹)

1 1 1 1 1 1

2 (---) 1 (⫹⫹⫹)

1 2 1

2 2 2 2 2(--)

1

2 1 2

2(--)

1

2 2

1(⫹⫹)

1(⫹⫹⫹) 2(-) 1(⫹)

2 2

1 1 2 2 1

1 2(---) 2(---) 2(--) 2(--)

2 2(--)

2 1 ↔ 1 1 ↔

108

C. Roder et al. / Neuroscience 201 (2012) 105–113

Table 1. Continued Author:

Marchese et al. (2010)

Pera et al. (2010)

Li et al. (2009)

Shi et al. (2009)

Krischek et al. (2008)

Specimen:

RA vs. MMA/STA

RA vs. MMA

UA vs. STA

RA/UA vs. RA/UA STA

RA/UA vs. AVMf

Platform:

Affymetrix

Affymetrix

Affymetrix

Illumina

Agilent

Number of specimen:

12 RA vs. 5 control MMA/STA

8 RA vs. 5 MMA

6 UA vs. 5 MMA

3 UA vs. 3 STA

3RA/3UA vs. 3RA/3UA STA

6 RA, 4 UA vs. 4 AVMf

2(--) 2(--) 2(--) ↔ 2(-)

2(--) 2(---) 2(---) 1(⫹) 2(-)

2(--)

2(--)

Chromosome:

Gene:

8q22 22q12.2 4q35.1 6p21.3 1p34-p33 17q25 3p25 6q27 9q33 6p21.3 2p23.3

RUNX1T1‡ SMTN2 SORBS22 TAPBP‡ TIE1‡ TIMP2‡ TIMP4*22 THBS211 TNC*22 TNF‡ TP53I311

2(--) 2(--)

UA vs. MMA

↔ 2 2 2

2

1(⫹⫹⫹) 2(---) 1(⫹) 1(⫹)

↔ 1 2 1 2 ↔ 1

RA, ruptured aneurysm; UA, unruptured aneurysm; MMA, middle meningeal artery; STA, superficial temporal artery; AVMf, arteriovenous malformation feeding artery.

(MHC) class I and II related genes, which belong to the GO pathway of antigen presentation. Confirmation studies with RT PCR and immunohistochemical staining underlined the role of MHC II, as well as CD68 (macrophages, monocytes) and HLA I molecules. Shi et al. (Shi et al., 2009) compared the gene expression of aneurysm domes vs. tissue of their STAs of six IA patients (three ruptured, three unruptured) using the Illumina Human WG6-v2 Microarray Analysis platform. In all, 326 (172 increased, 154 decreased) genes showed differences in expression. The authors performed functional analysis by using GO and KEGG and divided genes into five categories: Focal adhesion (KEGG), ECM–receptor interaction (KEGG), cell communication (KEGG), inflammatory response (GO), and apoptosis (GO). In the focal adhesion pathway 53 genes were significantly differentially expressed: 28 genes were up-regulated, 25 were down-regulated. In the ECM (extracellular matrix)–receptor interaction pathway, the data showed that 39 genes were significantly differentially expressed: 28 genes were up-regulated and 11 were down-regulated. In the cell communication pathway, the findings demonstrated 38 significantly differentially expressed genes: 22 genes were upregulated and 16 were down-regulated. In the inflammatory and immune response pathway 58 genes with significantly altered gene expression were found: 40 were up-regulated, 18 down-regulated. In the apoptosis pathway, 129 genes were differentially expressed, 78 up-regulated, 51 down-regulated. The expression patterns of three unruptured IAs compared to three STAs from patients free of vascular diseases were analyzed by Li et al. (Li et al., 2009) in 2009 with Affymetrix HU133 Plus 2.0 microarrays. 1.160

genes showed significant differences in expression comparing IA and STA tissue: 164 genes were up-regulated, whereas 996 genes showed down-regulation. For analysis the gene ontology database (GO) was used. Within the group of GO biological processes a significant overrepresentation was found for up-regulated genes related to developmental processes. Genes with importance for components of the ECM were significantly up-regulated within the group of GO cellular components. In the group of GO molecular functions an up-regulation of genes for enzyme regulator activity and transporter activity was found. Further analysis of genes related to ECM showed up-regulation of 12 and downregulation of 7 genes. Biological pathway analysis was performed for genes related to immune or inflammatory response. The results showed different expression for 51 genes, whereas 41 were down-regulated, 10 were up-regulated. KEGG analysis was performed showing overrepresentation of four pathways: Focal adhesion (down-regulation), type 1 diabetes mellitus pathway (down-regulation), antigen processing and presentation pathway (down-regulation), and complement and coagulation cascades (four genes down-, five up-regulated). In 2010 Pera et al. (Pera et al., 2010) collected samples of eight ruptured (RA) and six unruptured (UA) aneurysm domes, as well as five middle meningeal arteries (MMA) control specimen and performed expression analysis with the Affymetrix GeneChip Human Gene ST 1.0. The authors reported up-regulation in eight and downregulation in 123 genes in both RA and UA, as well as RA-specific down-regulation in two genes and UA-specific up-regulation in 26 genes, all compared with controls. Analysis of functional classification of differentially

C. Roder et al. / Neuroscience 201 (2012) 105–113

109

Table 2. Top three results of functional analysis of genes with altered expression, over-expression, and under-expression. Some genes were not listed in the functional analysis tool used. Therefore the number of genes analyzed decreased from 507 to 467 Group:

Name:

Rank:

GO BP:

GO MF:

GO CC:

KEGG:

1)

Genes with altered expression (467):

1

Signal transduction (95/467) Immune response (56/467) Cell adhesion (56/ 467)

Protein binding (254/467)

Extracellular region (141/467) Cytoplasm (135/ 467) Plasma membrane (127/467)

1

Cell adhesion (8/57)

Protein binding (35/57)

2

Immune response (8/57)

3

Collagen fibril organization (7/57) Signal transduction (44/184)

Extracellular matrix structural constituent (8/57) Protein homodimerization activity (5/57) Protein binding (106/184)

Focal adhesion (60/ 467) Pathways in cancer (47/467) Cytokine–cytokine receptor interaction (45/ 467) Focal adhesion (16/ 57) ECM–receptor interaction (9/57)

2 3

2)

1)

Genes with altered expression in more than two studies (57):

Genes with overexpression (184):

1

2

1)

Genes with overexpression in more than two studies (18):

Genes with underexpression (261):

Receptor activity (28/184)

2

Collagen fibril organization (4/18)

3

Inflammatory response (4/18) Signal transduction (50/261) Cell adhesion (29/ 261) Multicellular organismal development (23/ 261) Positive regulation of cell proliferation (4/18) Response to drug (3/18) Cell adhesion (3/18)

Extracellular matrix structural constituent (5/18) Heparin binding (4/18)

1

1 2 3

2)

Genes with underexpression in more than two studies (18):

Nucleotide binding (47/ 467)

Immune response (30/184) Inflammatory response (29/184) Cell adhesion (4/18)

3 2)

Receptor activity (52/467)

1

2 3

Peptidase activity (18/ 184) Protein binding (13/18)

Protein binding (142/261) Nucleotide binding (38/ 261) ATP binding (33/261)

Extracellular region (25/57) Plasma membrane (12/57) Extracellular space (11/57) Extracellular region (68/184) Plasma membrane (56/184) Cytoplasm (49/ 184) Extracellular region (8/18) Plasma membrane (7/18)

Pathways in cancer (8/57) Cytokine–cytokine interaction (25/ 184) Pathways in cancer (23/184) Focal adhesion (23/ 184) Focal adhesion (6/ 18) ECM–receptor interaction (6/18)

Collagen (4/18)

N/A

Cytoplasm (83/ 261) Nucleus (68/261)

Focal adhesion (35/ 261) Pathways in cancer (27/261) Cytokine–cytokinereceptor interaction (29/ 261) Focal adhesion (6/ 18)

Extracellular region (67/261)

Protein binding (10/18)

Extracellular region (9/18)

Protein homodimerization activity (3/18) N/A

Extracellular space (5/18) Cytosol (5/18)

Pathways in cancer (4/18) Hypertrophic cardiomyopathy (3/18)

GO, Gene Ontology; BP, biological processes; MF, molecular function; CC, cellular component; KEGG, Kyoto Encyclopedia of Genes and Genomes; ECM, extracellular matrix; N/A, not applicable.

expressed genes (IA vs. controls) revealed two major groups (muscle system and cell adhesion) were downregulated and one (immune/inflammatory system) was up-regulated. Marchese et al. (Marchese et al., 2010) examined aneurysmal tissue from a group of 13 patients with ruptured IAs and a group of 14 patients with unruptured IAs using the Affymetrix U133A GeneChip in 2010. They used STA and MMA tissue from a group of four patients treated for pathologies other than aneurysm or vascular malformation. It was shown that expression with a fold change of three and more regarding ruptured aneurysms was upand down-regulated in pathways concerning ECM components, especially in members of the matrix metalloprotei-

nase (MMP) family and in apoptosis-related genes. Particularly MMP-2 and -9 were up-regulated, as well as the pro-apoptotic genes FAS, BAX, and BID, whereas the anti-apoptotic genes Bcl-X(L) and Bcl-2 were down regulated in ruptured aneurysms. Genes with altered expression Analysis of gene-lists provided by the authors revealed a total of 507 genes, whereas 180 were over-expressed, 259 under-expressed, 32 showed equal expression, 36 showed inconsistent results between the studies (Table 1, Supplemental data with a complete gene list is available upon request).

110

C. Roder et al. / Neuroscience 201 (2012) 105–113

Fig. 1. Flow chart showing numbers of genes with differences in expression.

Genes with altered expression in more than two studies In all, 57 genes showed differences in more than two studies, whereas 18 were under-expressed, 18 over-expressed, 21 showed differing results (Fig 1, Table 1). Genes with altered expression in more than three studies The expression of seven genes showed differences in more than three studies: Over-expression: collagen type 1 alpha 2 (COL1A2) (up-regulation in four of five studies) (Krischek et al., 2008; Li et al., 2009; Shi et al., 2009; Marchese et al., 2010), collagen type 5 alpha 2 (COL5A2) (up-regulation in three of five studies) (Krischek et al., 2008; Li et al., 2009; Shi et al., 2009). Under-expression: B-cell lymphoma 2 (BCL2) (downregulation in three of five studies) (Li et al., 2009; Shi et al., 2009; Marchese et al., 2010), tissue inhibitor of metalloproteinase 4 (TIMP4) (down-regulation in three of five studies) (Krischek et al., 2008; Marchese et al., 2010; Pera et al., 2010), tenascin C (TNC) (down-regulation in three of five studies) (Krischek et al., 2008; Li et al., 2009; Shi et al., 2009). Differing expression: collagen type 3 alpha 1 (COL3A1) (up-regulation in three, down-regulation in one of five studies) (Krischek et al., 2008; Li et al., 2009; Shi et al., 2009; Marchese et al., 2010), chemokine (C-X-C motif) ligand 12 (CXCL12) (down-regulation in two, up-regulation in one of five studies) (Krischek et al., 2008; Li et al., 2009; Shi et al., 2009). GO and KEGG analysis For GO and KEGG analysis the Genecodis 2.0 online platform was used (Carmona-Saez et al., 2007; NogalesCadenas et al., 2009). The top three results of the pathway analysis are shown in Table 2. In all, 467 genes with

altered expression in any study (subgroup 1) and 57 genes with altered expression in more than two studies (subgroup 2) were used for the pathway analysis: GO biological processes (BP): In subgroup 1) 95 of 467 genes with altered expression belong to the GO domain signal transduction, in subgroup 2) 8 of 57 genes belong to the GO domain cell adhesion. GO molecular function (MF): In subgroup 1) 254 of 467 genes with altered expression belong to the GO domain protein binding, in subgroup 2) 35 of 57 genes belong to the GO domain protein binding. GO cellular component (CC): In subgroup 1) 141 of 467 genes with altered expression belong to the GO domain extracellular region, in subgroup 2) 25 of 57 genes belong to the GO domain extracellular region. KEGG: In subgroup 1) 60 of 467 genes with altered expression belong to the KEGG pathway focal adhesion, in subgroup 2) 16 of 57 genes belong to the KEGG pathway focal adhesion. Additional pathway analysis was performed after the sub-groups 1 and 2 were separated into groups of genes with under-expression and with over-expression. The results are shown in Table 2.

DISCUSSION Genes with altered expression in more than three studies Seven genes showed altered expression in more than three studies: BCL2, COL1A2, COL3A1, COL5A2, CXCL12, TIMP4, TNC (Table 1). COL1A2, COL3A1, COL5A2, TIMP4, and TNC are known to modulate processes in the formation of the ECM, which have been described in association with IAs before (Hsia and Schwarzbauer, 2005; Li et al., 2009). Collagens in the development of IAs Collagen type 1 fibers together with collagen type 3 fibers are known to represent up to 90% of the total arterial collagen and are therefore important for the strength of arterial walls (Ruigrok and Rinkel, 2008). Up-regulation of different types of collagens was described in most gene expression studies reviewed (Table 1). Increase of expression could lead to larger amounts of collagen, which does not seem to explain a weak arterial wall as seen in IAs and SAH. However, increased expression of collagens can be explained by compensational mechanisms due to a higher collagen degradation caused by increased gelatinase activity, inflammatory cells, or higher activity of MMPs as shown in other studies (Chyatte and Lewis, 1997; Peters et al., 2001). An increased metabolism of collagen may also confirm the results of a study published by Mimata et al. (Mimata et al., 1997) who performed immunohistochemical studies on collagens in aneurysmal wall tissue. The authors pointed out that all types of collagens were grossly preserved, but with differences in distribution within the walls of IAs.

C. Roder et al. / Neuroscience 201 (2012) 105–113

Matrix metalloproteinases in the development of IAs Matrix metalloproteinases (MMPs), as well as their endogenous inhibitors the “Tissue Inhibitors of MMPs (TIMPs),” are multifunctional proteins which, amongst others, modulate ECM components and inflammatory processes (Koskivirta et al., 2006; Li et al., 2009). Previous studies showed that MMP2 and MMP9 were up-regulated in IAs (Li et al., 2009). However, our meta-analysis did not substantiate these findings, but we rather found that TIMP4 showed decreased expression in three of five studies (Table 1). Koskivirta et al. (Koskivirta et al., 2006) investigated the role of TIMP4 in inflammatory processes of cardiovascular pathologies using immunohistochemical staining and found increased levels of TIMP4. In vitro studies showed that high intraluminal pressure and mechanical strain increase activity of MMP2 and MMP9. Therefore Koskivirta et al. argued that increased levels of TIMP4 might be due to a compensatory mechanism, because TIMP4 is known to be capable of inhibiting the activity of all MMPs (Koskivirta et al., 2006). Therefore, decreased expression of TIMP4 might result in increased activity of MMPs (especially MMP2 and MMP9, which are known to show increased expression in IAs), resulting in uncontrolled ECM remodeling and subsequent development of IAs. Further research on the role of TIMP4 in IAs is needed to elucidate these pathways. Tenascin C in the development of IAs Tenascin C (TNC) is an oligomeric glycoprotein found in the ECM. Previous research showed that TNC antagonizes cell attachment of fibroblasts to fibronectin and that it is sharply up-regulated in tissues undergoing remodeling processes seen during wound repair, neovascularization, or inflammation. It is also expressed during neural, skeletal, and vascular morphogenesis in embryos (Hsia and Schwarzbauer, 2005). Shear stress with subsequent remodeling, inflammation, and micro-injury followed by repair processes are suspected to contribute to the formation and rupture of IAs (Paszkowiak and Dardik, 2003; Frösen et al., 2004; Krischek and Tatagiba, 2008). Therefore, malfunctioning TNC might lead to the development of IAs by disturbance of repair processes or pathological development of stress fibers as it has been shown in vitro after adding TNC to a matrix of fibrin, fibronectin, and fibroblasts (Hsia and Schwarzbauer, 2005). Functional analysis Functional analysis revealed overrepresentation of the GO domains and KEGG pathways as listed above. We are aware of the limited validity of these results due to selection bias by the pre-selected published gene lists; therefore, we do not discuss these results in detail. Nevertheless we believe that the analysis of pathways may provide valuable information on the genesis of IAs (Table 2). Limitations of gene-expression studies Gene-expression profiling enables the analysis of differences in expression levels of thousands of genes. Espe-

111

cially under the assumption of a multifactorial genesis of IAs, such a broad overview might provide valuable information on genes and pathways of interest (Krischek and Tatagiba, 2008). However, this technique reaches its limits if asking whether the findings cause the pathology or are a reaction to another process. Also the acquisition of control tissue is problematic, as samples of walls of intracranial arteries from healthy subjects cannot be extracted. Therefore compromises in terms of taking samples from extracranial arteries (i.e. STA), intracranial arteries which may be affected by other pathologies (i.e. feeding arteries of arterio-venous-malformations), or autopsy specimen of intracranial arteries must be accepted and weighted against differences in gene expression caused by a different anatomy or non-IA–related pathological processes. Kurki et al. (Kurki et al., 2011) recently published a study comparing the gene expression of tissue of seven ruptured vs. eight unruptured intracranial aneurysms. Studies like these might bring new valuable information on the evaluation of risk for the rupture of known aneurysms and subsequently might be helpful for treatment decisions. Possibly future research on IAs can combine and analyze different forms of research as it has been done before: Weinsheimer et al. (Weinsheimer et al., 2007) combined results of a microarray expression study on IAs with results from a linkage study and were able to show that the adherens junction, MAPK, and notch signaling pathways may be involved in the formation of IAs. Limitations of our study design In this study we reviewed published gene lists of five gene expression studies on IAs and analyzed common differences in gene expression (Krischek et al., 2008; Li et al., 2009; Shi et al., 2009; Marchese et al., 2010; Pera et al., 2010). Four limitations of this study design should be pointed out: 1) The studies were performed on different gene expression analysis platforms meaning that genes with altered expression in one study may not even be part of the analysis on the used platform of another study. The number of genes analyzed in each study was limited and did not cover all genes known to date. 2) Our study depended on gene lists provided by the authors. If the authors did not attach importance to certain genes with altered expression and did not list them, we were not able to include these in our study. Both facts potentially limited the number of genes found with altered expression. However, these facts do not limit the validity of the seven genes found with differences in expression in three or more studies. 3) All studies included in this analysis were based on small numbers of specimen and are therefore of limited statistical power. 4) The goal of this analysis was to investigate differences in gene expression between control tissue vs. aneurysmatic intracranial arterial tissue to understand the genetic mechanisms in the genesis of IAs. A subgroup analysis of ruptured and unruptured IAs was not performed. As the molecular mechanisms of the development of IAs remain widely unknown, the study design was based on the assumption that RAs and UAs at least partly share the same etiology. As far as the information was

112

C. Roder et al. / Neuroscience 201 (2012) 105–113

available in the studies that were included in our analysis, only few differences in gene expression between RAs and UAs (Krischek et al., 2008; Shi et al., 2009; Marchese et al., 2010; Pera et al., 2010) had been found. Although a recent study by Kurki et al. analyzed gene expression of RAs vs. UAs and identified significant differences (Kurki et al., 2011), the comparison between RAs/ UAs vs. healthy tissue addresses the issue of understanding the etiology of IAs, whereas the comparison between RAs vs. UAs aims to identify factors leading to the rupture of IAs. For this reason the latter study was not included in our analysis. Impact of genetic findings on the treatment of intracranial aneurysms Genomic research of possible causes for disease development is one of the most rapidly growing fields of modern times: the speed of analysis increases steadily, whereas the research cost decreases at the same time. Large amounts of data are created but only little information can be translated to the treatment in everyday clinical routine. However, if disease-causing genomic variants are identified this may change therapy modalities as well as diagnostic procedures. An example is the management of breast cancer in which the identification of BRCA1 and BRCA2 mutations changed disease-screening routines and treatment, including prophylactic mastectomy and salpingo-oophorectomy (Nathason and Domcheck, 2011). As suspected in many diseases, intracranial aneurysms seem to be caused by multiple genomic and environmental factors (Feigin et al., 2005; Krischek and Tatagiba, 2008). The concluding analyses and understanding of the complex relations will take further large efforts. As genomic knowledge on IAs progresses, several clinical applications are imaginable such as: a bedside test for patients to determine a possible risk of later IA development or the evaluation of the rupture risk of patients that harbor an IA (Krischek and Tatagiba, 2008).

CONCLUSIONS By performing a meta-analysis of five genome-wide microarray gene expression studies with a total of 60 tissue samples of intracranial aneurysms, we identified 507 genes with altered expression, including 57 genes with an altered expression in more than two studies. Seven genes had an altered expression in three or more independent studies (BCL2, COL1A2, COL3A1, COL5A2, CXCL12, TIMP4, TNC). Further sequencing and functional analysis studies of these genes may provide additional insight into genomic factors involved in the complex process of the genesis of IAs. Due to the overall small number of specimen and the limited number of genes that were further analyzed, the available gene expression studies that have been studied so far are of limited power. Further studies with larger sample sizes, separate and combined analysis of UAs and RAs, and a more detailed (and possible functional) analysis of found target genes are needed for a better understanding of the etiology of this disease. IAs are

very likely to be caused by a multifactorial etiology with genomics playing an important role.

SUPPLEMENTS Supplemental data including the complete gene list is available upon request to the author.

REFERENCES Cahan P, Rovegno F, Mooney D, Newman JC, St. Laurent G III, McCaffrey TA (2007) Meta-analysis of microarray results: challenges, opportunities, and recommendations for standardization. Gene 401:12–18. Carmona-Saez P, Chagoyen M, Tirado F, Carazo JM, Pascual-Montano A (2007) GENECODIS: a web-based tool for finding significant concurrent annotations in gene lists. Genome Biol 8:R3. Chyatte D, Lewis I (1997) Gelatinase activity and the occurrence of cerebral aneurysms. Stroke 28:799 – 804. Feigin VL, Rinkel GJ, Lawes CM, Algra A, Bennett DA, van Gijn J, Anderson CS (2005) Risk factors for subarachnoid hemorrhage: an updated systematic review of epidemiological studies. Stroke 36:2773–2780. Frösen J, Piippo A, Paetau A, Kangasniemi M, Niemelä M, Hernesniemi J, Jääskeläinen J (2004) Remodeling of saccular cerebral artery aneurysm wall is associated with rupture: histological analysis of 24 unruptured and 42 ruptured cases. Stroke 35:2287–2293. Griffith OL, Melck A, Jones SJ, Wiseman SM (2006) Meta-analysis and meta-review of thyroid cancer gene expression profiling studies identifies important diagnostic biomarkers. J Clin Oncol 24:5043–5051. Hsia HC, Schwarzbauer JE (2005) Meet the tenascins: multifunctional and mysterious. J Biol Chem 280:26641–26644. Kataoka H, Aoki T (2010) Molecular basis for the development of intracranial aneurysm. Expert Rev Neurother 10:173–187. Koskivirta I, Rahkonen O, Mäyränpää M, Pakkanen S, Husheem M, Sainio A, Hakovirta H, Laine J, Jokinen E, Vuorio E, Kovanen P, Järveläinen H (2006) Tissue inhibitor of metalloproteinases 4 (TIMP4) is involved in inflammatory processes of human cardiovascular pathology. Histochem Cell Biol 126:335–342. Krischek B, Kasuya H, Tajima A, Akagawa H, Sasaki T, Yoneyama T, Ujiie H, Kubo O, Bonin M, Takakura K, Hori T, Inoue I (2008) Network-based gene expression analysis of intracranial aneurysm tissue reveals role of antigen presenting cells. Neuroscience 154:1398 –1407. Krischek B, Tatagiba M (2008) The influence of genetics on intracranial aneurysm formation and rupture: current knowledge and its possible impact on future treatment. Adv Tech Stand Neurosurg 33:131–147. Kurki MI, Häkkinen SK, Frösen J, Tulamo R, von und zu Fraunberg M, Wong G, Tromp G, Niemelä M, Hernesniemi J, Jääskeläinen JE (2011) Upregulated signaling pathways in ruptured human saccular intracranial aneurysm wall: an emerging regulative role of Tolllike receptor signaling and NF-[kappa] B, HIF1A and ETS transcription factors. Neurosurgery 68:1667–1676. Li L, Yang X, Jiang F, Dusting GJ, Wu Z (2009) Transcriptome-wide characterization of gene expression associated with unruptured intracranial aneurysms. Eur Neurol 62:330 –337. Marchese E, Vignati A, Albanese A, Nucci CG, Sabatino G, Tirpakova B, Lofrese G, Zelano G, Maira G (2010) Comparative evaluation of genome-wide gene expression profiles in ruptured and unruptured human intracranial aneurysms. J Biol Regul Homeost Agents 24:185–195. Mimata C, Kitaoka M, Nagahiro S, Iyama K, Hori H, Yoshioka H, Ushio Y (1997) Differential distribution and expressions of collagens in the cerebral aneurysmal wall. Acta Neuropathol (Berl) 94:197–206.

C. Roder et al. / Neuroscience 201 (2012) 105–113 Nathason KL, Domcheck SM (2011) Therapeutic approaches for women predisposed to breast cancer. Annu Rev Med 62:295–306. Nogales-Cadenas R, Carmona-Saez P, Vazquez M, Vicente C, Yang X, Tirado F, Carazo JM, Pascual-Montano A (2009) GeneCodis: interpreting gene lists through enrichment analysis and integration of diverse biological information. Nucleic Acids Res 37:W317–W322. Paszkowiak JJ, Dardik A (2003) Arterial wall shear stress: observations from the bench to the bedside. Vasc Endovasc Surg 37:47–57. Pera J, Korostynski M, Krzyszkowski T, Czopek J, Slowik A, Dziedzic T, Piechota M, Stachura K, Moskala M, Przewlocki R, Szczudlik A (2010) Gene expression profiles in human ruptured and unruptured intracranial aneurysms: what is the role of inflammation? Stroke 41:224 –231. Peters DG, Kassam AB, Feingold E, Heidrich-O’Hare E, Yonas H, Ferrell RE, Brufsky A (2001) Molecular anatomy of an intracranial aneurysm: coordinated expression of genes involved in wound healing and tissue remodeling. Stroke 32:1036 –1042.

113

Ronkainen A, Hernesniemi J, Puranen M, Niemitukia L, Vanninen R, Ryynänen M, Kuivaniemi H, Tromp G (1997) Familial intracranial aneurysms. Lancet 349:380 –384. Ruigrok YM, Rinkel GJ (2008) Genetics of intracranial aneurysms. Stroke 39:1049 –1055. Shi C, Awad IA, Jafari N, Lin S, Du P, Hage ZA, Shenkar R, Getch CC, Bredel M, Batjer HH, Bendok BR (2009) Genomics of human intracranial aneurysm wall. Stroke 40:1252–1261. Tulamo R, Frösen J, Junnikkala S, Paetau A, Kangasniemi M, Peláez J, Hernesniemi J, Niemelä M, Meri S (2010) Complement system becomes activated by the classical pathway in intracranial aneurysm walls. Lab Invest 90:168 –179. van Gijn J, Kerr RS, Rinkel GJ (2007) Subarachnoid haemorrhage. Lancet 369:306 –318. Weinsheimer S, Lenk GM, van der Voet M, Land S, Ronkainen A, Alafuzoff I, Kuivaniemi H, Tromp G (2007) Integration of expression profiles and genetic mapping data to identify candidate genes in intracranial aneurysm. Physiol Genomics 32:45–57.

(Accepted 18 October 2011) (Available online 26 October 2011)