Genetic Study of Intracranial Aneurysms
Background and Purpose—Rupture of intracranial aneurysms (IAs) causes subarachnoid hemorrhage, leading to immediate death or severe disability. Identification of the genetic factors involved is critical for disease prevention and treatment. We aimed to identify the susceptibility genes for IAs.
Methods—Exome sequencing was performed in 12 families with histories of multiple cases of IA (number of cases per family ≥3), with a total of 42 cases. Various filtering strategies were used to select the candidate variants. Replicate association studies of several candidate variants were performed in probands of 24 additional IA families and 426 sporadic IA cases. Functional analysis for the mutations was conducted.
Results—After sequencing and filtering, 78 variants were selected for the following reasons: allele frequencies of variants in 42 patients was significantly (P<0.05) larger than expected; variants were completely shared by all patients with IA within ≥1 family; variants predicted damage to the structure or function of the protein by PolyPhen-2 (Polymorphism Phenotyping V2) and SIFT (Sorting Intolerance From Tolerant). We selected 10 variants from 9 genes (GPR63, ADAMST15, MLL2, IL10RA, PAFAH2, THBD, IL11RA, FILIP1L, and ZNF222) to form 78 candidate variants by considering commonness in families, known disease genes, or ontology association with angiogenesis. Replicate association studies revealed that only p.E133Q in ADAMTS15 was aggregated in the familial IA cases (odds ratio, 5.96; 95% confidence interval, 2.40–14.82; P=0.0001; significant after the Bonferroni correction [P=0.05/78=0.0006]). Silencing ADAMTS15 and overexpression of ADAMTS15 p.E133Q accelerated endothelial cell migration, suggesting that ADAMTS15 may have antiangiogenic activity.
Conclusions—ADAMTS15 is a candidate gene for IAs.
The overall prevalence of intracranial aneurysms (IAs) is estimated at 3.2% in the general population.1 The rupture of an IA is one of the most devastating neurological conditions known.2 Multiple risk factors, such as cigarette smoking, hypertension, and alcohol consumption, are known risk factors for the formation and possible rupture of IAs.3–5 The familial occurrence of IAs suggests that genetic factors are involved in disease susceptibility.6,7 Although several genome-wide association studies have been performed worldwide8–13 because of their limited power to detect rare variants that are thought to have larger effect sizes, the genetic predisposition of IAs is largely unknown.
In accordance with the hypothesis that less common variants (minor allele frequency [MAF] <0.05) may contribute to IA development, we performed whole exome sequencing of 42 cases with a definite phenotype of IA from 12 families. Further replicate association studies of several candidate variants in additional familial and sporadic IA cases and biological investigation were performed.
In cooperation with hospitals in the Western part of Japan, we established a large cohort of familial and sporadic IA cases.14 These cases were diagnosed by angiography or subarachnoid hemorrhage with aneurysm rupture or in some cases, unruptured IAs were confirmed during intracranial surgery. Individual information and lifestyle data were collected by interview. With respect to familial IA, these families showed high IA aggregation (≥2 first- to third-degree relatives affected by IA) with a supposed strong genetic component. Twelve families with ≥3 definite IA cases (a total of 42 cases) were selected for whole exome sequencing (Figure 1). The probands of another unrelated 24 IA families were used for further replication study. Then, 426 unrelated sporadic IA cases were used for further association study of the selected candidate variants. Written informed consent was obtained from all participants.
For the control population, we used exome data from the Japanese genetic variation consortium database (a reference database of genetic variations in the Japanese population that contains genetic variations determined by exome sequencing of 1208 individuals and genotyping data of common variations obtained from a cohort of 3248 individuals [http://www.genome.med.kyoto-u.ac.jp/SnpDB/]). This study was approved by the Institutional Review Board and Ethics Committee of Kyoto University School of Medicine, Japan.
Exome Sequencing, Mapping, Variant Calling, and Prioritization
Genomic DNA was extracted from peripheral blood lymphocytes using the QIAamp DNA Blood Mini Kit (Qiagen, Hilden, Germany). Whole exome sequencing was performed by Riken Genesis Co, Ltd in Japan. Exon capture was performed using the SureSelect 70.4 Mb Human All Exon V4+UTR Kit (Agilent Technologies, Santa Clara, CA), and sequencing was done with using the Illumina HiSeq 2000 platform (Illumina Inc, San Diego, CA). Sequence mapping and variant detection were performed using the Burrows-Wheeler Aligner 0.6.2 and Genome Analysis Toolkit software. Details of sequencing and mapping are available in the Methods in the online-only Data Supplement.
A series of filters were used to prioritize variants. Variants were given higher priority if they were (1) predicted to affect protein-coding sequences (including missense, nonsense, read through, splice site variants and indels in the consensus coding sequence region); (2) less common in reference databases (MAF<0.05 in 1000 genome databases [ftp://ftp.1000genomes.ebi.ac.uk/vol1/ftp/release/20110521/] and the Japanese genetic variation consortium database [JGVCD]); (3) completely shared by all the affected individuals within ≥1 family; (4) observed minor allele count in 42 familial cases larger than the 95% confidence interval of the expected values calculated based on family structure (expected value is calculated by simulation; detailed Methods are available in the online-only Data Supplement); (5) damaging, as predicted by protein prediction programs (Polymorphism Phenotyping V2 [PolyPhen-2; http://genetics.bwh.harvard.edu/pph2] and Sorting Intolerant From Tolerant [SIFT; http://sift.bii.a-star.edu.sg/]) or cause nonsense mediated decay, predicted by SIFT indel (http://sift.bii.a-star.edu.sg/www/SIFT_indels2.html; Figure 2).
Replicate Association Study in Additional Familial and Sporadic IA Cases
The selection of candidate variants for association study was based on the commonness and functional relevance of IA or other known diseases. In terms of functional relevance, we chose a key word, which describes the pathological process of IA, from published meta-analysis. Candidate variants were directly sequenced using the Sanger method. Further genotyping of these selected variants in additional IA cases was performed using TaqMan or restriction fragment length polymorphism methods. Direct sequencing was performed for unclear genotyping results. Primers and detailed polymerase chain reaction conditions are shown in Table I in the online-only Data Supplement.
Human Umbilical Venous Endothelial Cell Culture, RNA Interference Experiment, Transfection of ADAMTS15 p.E133Q and Western Blot Analysis
Human umbilical venous endothelial cells (HUVECs) were obtained from Lonza (C2517A; Walkersville, MD) and cultured following a published method.15 Transfection of small interfering RNA (siRNA) was conducted as previously reported.16 We obtained ADAMTS15 siRNA (sc-37491; Santa Cruz Biotechnology, Santa Cruz, CA) and FILIP1L siRNA (sc-94184; Santa Cruz Biotechnology) with control siRNA-A (sc-37007; Santa Cruz Biotechnology). After 24 hours transfection, HUVECs were washed with PBS and lysed.
Expression ready, untagged ADAMTS15 wild-type expression plasmid (Human ADAMTS15 Gene cDNA Clone [full-length open reading frame Clone]; Cat. No. HG11912-G-N) and empty vector (pCMV/hygro-negative control vector, untagged; Cat. No. CV001) were purchased from Sino Biological Inc (Beijing, P.R. China). The E133Q mutation was introduced using a QuickChange kit (Agilent Technologies). All vector constructs were verified by sequencing. Plasmids were transfected into cells using an Amaxa Nucleofector Device (Lonza), following the manufacturer recommendations as previously reported.16
Western blots were detected using anti-ADAMTS15 (sc-68425; Santa Cruz Biotechnology), anti-FILIP1L (sc-102493; Santa Cruz Biotechnology), and anti–β-tubulin (sc-9104; Santa Cruz Biotechnology) antibodies, following a previously reported method.16 HUVECs were cultured in endothelial cell basal growth medium 2 with supplements containing vascular endothelial growth factor (Cat. No. CC-4147; Lonza, Cambridge, MA). Cell migration assay was assessed after 8 hours by scraping the cell monolayer in a 24-well plate following a previously reported method.15 Tube formation was assessed as described previously.16 HUVECs (40 000 cells per well) were seeded onto 96-well plates coated with Matrigel (BD Biosciences, San Jose, CA). After incubation for 12 hours at 37°C, digital images were captured of tubes that had formed. For quantification, the tube area, total tube length, and number of tube branches were calculated using ImageJ software. Parameters for assessing the tube formation function were obtained from 3 independent tube formation assays.
Statistical Power Calculations
Given the prevalence of IA in the general population of 0.03 and the significance level of 0.05, the statistical power under an arbitrarily assumed sample size for further association analysis was calculated using the CaTS Power Calculator across a range of relative risks and MAFs (http://www.sph.umich.edu/csg/abecasis/CaTS/; Figure I in the online-only Data Supplement.17
For selected candidate variants, the association study with IA was analyzed using STATISTICA 64-bit (StatSoft, Tulsa, OK). The corresponding genotype count information in the Japanese genetic variation consortium database was used as the population control. By assuming an autosomal-dominant model of inheritance, logistic regression analysis was used to assess the association between candidate variants and IA phenotype. The P value and odds ratio with 95% confidence interval were calculated. The odds ratio was calculated with respect to the risk allele. The Bonferroni correction was used to assess the significance level of the association. We applied the Bonferroni correction on the basis of 78 independent effective tests, and the Bonferroni threshold significance level was P=0.05/78=0.0006. A P<0.001 was considered suggestive of an association.
Characteristics of the Study Participants
Characteristics of the study participants are shown in Table 1. There was no significant difference in the age at diagnosis between familial and sporadic IA.
Whole Exome Sequencing Analysis
Whole exome sequencing generated 8.1 to 20.0 billion bases for 42 affected individuals. After mapping to the human reference genome (UCSC Genome Browser hg19), we obtained 6.9 to 16.5 Gb effective bases mapped to the genome, with a mean sequencing depth between 64- and 135-fold. The mean percentage of exomes covered to a read depth of ≥8-fold was >94.8%. After alignment and a series of quality controls steps, we identified 557 188 to 1 660 842 single nucleotide variants and 28 014 to 88 129 indels in the 42 individuals (Table II in the online-only Data Supplement and JGVCD).
This study focused on the analysis of less common deleterious variants in the consensus coding sequence region. After filtering against 2 reference databases with MAF<0.05, there were 7338 nonsynonymous single nucleotide variants and indels with a read depth ≥8. There were 975 variants that were completely shared by all patients with IA within ≥1 family. After simulation as described in the Methods in the online-only Data Supplement, there were 452 variants retained. Furthermore, we removed benign or tolerated variants as predicted by PolyPhen-2 and SIFT (or SIFT indel). Finally, 78 candidate variants were retained (part of the result is shown in Table 2, and the complete list of single nucleotide variants is presented in Table III in the online-only Data Supplement). Of these variants, none, except p.Y193F in GPR63 and p.R142H in C10orf122, were completely shared by patients with IA in >1 family. C10orf122 is a chromosome 6 open reading frame gene; it has not been assigned a function to date. Thus, we did not perform the subsequent replication study.
Replicate Association Study of Several Candidate Variants
In the candidate list, variants in genes that are common in patients with IA (p.Y193F in GPR63) are known as disease genes, or that have relevance to IA biology were selected. For a key word related to functional relevance, we searched for ontology based on functional relevance from meta-analysis articles18–20 and found angiogenesis. Thus, we chose genes in ontologies related to angiogenesis, including extracellular matrix integrity, inflammatory mediators, blood coagulation, and vascular endothelium maintenance. Finally, we selected genes important for extracellular matrix integrity (p.E133Q in ADAMTS15 and p.E631D in FILIP1L), inflammatory mediators (p.G105S in IL10RA and p.R261H in IL11RA), blood coagulation–related genes (p.R85C in PAFAH2 and p.R473H in ZNF222), and vascular endothelium maintenance (p.D486Y in THBD), and those that are deleterious or relate to certain diseases (p.V401M, p.R5224H in MLL2). All selected genes were given higher priority for further association study. Replicate association study of 10 variants was performed, and Results are presented in Table 3.
We found that p.E133Q in ADAMTS15 was aggregated significantly in the familial IA cases (P=0.0001<0.0006) after the Bonferroni correction, whereas p.E631D in FILIP1L was associated with familial IA with suggestive significance (0.0006<P=0.0007<0.001). However, p.D486Y in THBD, p.R261H in IL11RA, p.R85C in PAPAH2, and p.R473H in ZNF222 showed higher MAF in sporadic cases when compared with the general population (all nominal, P<0.05) but did not reach the suggestive significance level (P=0.001).
Effects of Silencing ADAMTS15 and FILIP1L on EC Function
Although ADAMTS15 p.E133Q and FILIP1L p.E631D are deleterious by bioinformatics prediction and they were classified as genes in the ontology of angiogenesis, it is uncertain whether ADAMTS15 and FILIP1L have an angiogenic function. Thus, we investigated whether they have angiogenic activity by silencing ADAMTS15 or FILIP1L. As shown in Figure 3A and 3B, silencing these genes accelerated EC migration, suggesting that they indeed have antiangiogenic activity.
Multiple less common deleterious variants were identified in the familial IA cases after exome sequencing and filtering (Table III in the online-only Data Supplement). Meta-analysis of >116 000 individuals identified 19 single nucleotide polymorphisms associated with IA that were mainly related to the vascular endothelium and extracellular matrix.19 In this study, we identified several less common or rare variants of extracellular matrix genes (ADAMTS15, FILIP1L, TTN, ITGB6, CRELD1, LRP4, and LRP5; Table 2; Table III in the online-only Data Supplement) shared by all patients with IA within ≥1 family. A replication study of p.E133Q in ADAMTS15 demonstrated that it was aggregated in familial IA cases significantly, even with Bonferroni correction (odds ratio, 5.96; 95% confidence interval, 2.40–14.82; P=0.00013). ADAMTS15 is a disintegrin and metalloproteinase with thrombospondin motifs 15, and previous gene expression studies of IAs demonstrated abnormal transcription of matrix metalloproteinases, indicating a possible role in predisposition to IA development.20 Another variant, p.E631D in FILIP1L, was also suggested to be involved in angiogenesis as previously predicted although it had weak evidence on association with IA.21,22
Because of the low allele frequencies for ADAMTS15 p.E133Q and FILIP1L p.E631D, it is difficult to prove that the sporadic cases would have the same variant as the familial IA cases. Alternatively, biological investigation was conducted to test whether these genes, and not individual variants, are involved in angiogenesis. The experiments clearly demonstrated that silencing these genes enhanced migration of HUVECs, indicating that the genes have antiangiogenic activity. Given that enhanced EC migration is associated with increased surface integrin expression when stimulated by vascular endothelial growth factor23 in endothelial cell basal growth medium 2, ADAMTS15 and FILIP1L are postulated to downregulate surface integrin. Furthermore, investigation into the effects of p.E133Q in ADAMTS15 revealed that overexpression of the mutation accelerated endothelial migration, suggesting that this is a loss-of-function mutation because it has the same effect as silencing ADAMTS15. Taken together, we considered that IA may be mediated by modifying the expression of integrin on ECs.
The limitations of this study should be considered. First, although the exome sequencing results might represent several candidate genes, we performed replicate association studies for only 10 of 78 candidate variants; the other candidate variants have not yet been validated. Given such a biased selection, we adhered to stringent statistical criteria, to minimize the chance of false-positives. Although proving pathogenic association of IA with rare family-specific variants (MAF<0.001) is difficult, even when using large sample sizes, gene-based association studies (such as a burden test) or multiple family replication studies might be useful. Second, exome sequencing has its own limitations. It cannot capture structural or noncoding variants, such as copy number variations, promoters, or enhancers, which may predispose to IA susceptibility. Third, because large definitely diagnosed IA-free controls were not available, we used the general population as the control in the replicate association analysis. This may limit the independence of the replication from the discovery study and perturb the associate results. However, such an approach could be justifiable for rare variants that cannot be detected without large samples and may not perturb the results significantly for familial IA if at all. Furthermore, given that founder variants are identified in the Japanese population, it is not known whether these variants are also associated with IAs in different ethnic populations. This should be explored in future studies.
There was good analytic strength in this study. We conducted genetic analysis for 42 selected IA-definite members in 12 IA-aggregated families. The sharing of a rare variant by all patients within IA families gave an indication of causality. In addition, although we selected 10 genes subjectively based on review of the literature, we obtained a significant candidate, ADAMTS15 p.E133Q, which showed significant association with IA even after the Bonferroni correction and was proven to be loss-of-function biologically. In conclusion, ADAMTS15 may warrant further replication studies and biological investigation in relation to IA.
We thank all participants for providing samples and also those involved in sample collection. Dr Yan is a postdoctoral fellow at Central South University (No. 149946), supported by the Postdoctoral International Exchange Plan in China.
Sources of Funding
This project was supported primarily by a grant from the Takayama Red Cross Hospital in Japan.
The online-only Data Supplement is available with this article at http://stroke.ahajournals.org/lookup/suppl/doi:10.1161/STROKEAHA.114.007286/-/DC1.
- Received September 3, 2014.
- Accepted December 26, 2015.
- © 2015 American Heart Association, Inc.
- Teunissen LL,
- Rinkel GJ,
- Algra A,
- van Gijn J
- Feigin VL,
- Rinkel GJ,
- Lawes CM,
- Algra A,
- Bennett DA,
- van Gijn J,
- et al
- Yamada S,
- Koizumi A,
- Iso H,
- Wada Y,
- Watanabe Y,
- Date C,
- et al
- Yasuno K,
- Bakircioğlu M,
- Low SK,
- Bilgüvar K,
- Gaál E,
- Ruigrok YM,
- et al
- Low SK,
- Takahashi A,
- Cha PC,
- Zembutsu H,
- Kamatani N,
- Kubo M,
- et al
- Foroud T,
- Koller DL,
- Lai D,
- Sauerbeck L,
- Anderson C,
- Ko N,
- et al
- Yamada S,
- Utsunomiya M,
- Inoue K,
- Nozaki K,
- Inoue S,
- Takenaka K,
- et al
- Milner R
- Guo S,
- Lok J,
- Hayakawa K,
- Leung W,
- Xing C,
- Ji X,
- et al
- Hitomi T,
- Habu T,
- Kobayashi H,
- Okuda H,
- Harada KH,
- Osafune K,
- et al
- Hussain I,
- Duffis EJ,
- Gandhi CD,
- Prestigiacomo CJ
- de Wit MC,
- Kros JM,
- Halley DJ,
- de Coo IF,
- Verdijk R,
- Jacobs BC,
- et al
- Elicceiri BP