Common variant near the endothelin receptor type
A (EDNRA) gene is associated with intracranial
aneurysm risk
Departments of aNeurosurgery, bNeurobiology, and cGenetics, Yale Program on Neurogenetics, Yale Center for Human Genetics and Genomics, and
s
Department of Internal Medicine and tHoward Hughes Medical Institute, Yale University School of Medicine, New Haven, CT 06510; dHuman Genome Center,
Institute of Medical Science, University of Tokyo, Tokyo 108-8639, Japan; eDepartment of Neurosurgery, Helsinki University Central Hospital, Helsinki, FI-00029
HUS, Finland; fDepartment of Neurology, Rudolf Magnus Institute of Neuroscience, University Medical Center Utrecht, 3584 CX Utrecht, The Netherlands;
g
Department of Public Health, School of Medicine, Chiba University, Chiba 260-8670, Japan; hService de Neurochirurgie, Department of Clinical Neurosciences,
Geneva University Hospital, 1211 Geneva 4, Switzerland; iDepartment of Neurosurgery, Medical Center East, Tokyo Women’s University, Tokyo 116-8567,
Japan; jDepartment of Neurosurgery, Kuopio University Hospital, Kuopio FI-70211, Finland; kKlinik und Poliklinik für Neurochirurgie, Universitätsklinikum Carl
Gustav Carus der Technischen Universität Dresden, 01307 Dresden, Germany; lDepartment of Neurology, Goethe University, 60596 Frankfurt am Main,
Germany; mDepartment of Neurosurgery, University of Bonn, D-53105 Bonn, Germany; nDepartment of Neurosurgery, University of Tübingen, 72076
Tubingen, Germany; oYale Center for Genome Analysis, Orange, CT 06477; pDivision of Human Genetics, National Institute of Genetics, Shizuoka 411-8540,
Japan; qWellcome Trust Sanger Institute, Hinxton, Cambridge CB10 1HH, United Kingdom; and rUnité Mixte de Recherche Institut National de la Santé et de la
Recherche Médicale S937, University Pierre and Marie Curie, Paris 6, France
Contributed by Richard P. Lifton, October 18, 2011 (sent for review February 14, 2011)
The pathogenesis of intracranial aneurysm (IA) formation and rupture is complex, with significant contribution from genetic factors.
We previously reported genome-wide association studies based on
European discovery and Japanese replication cohorts of 5,891 cases
and 14,181 controls that identified five disease-related loci. These
studies were based on testing replication of genomic regions that
contained SNPs with posterior probability of association (PPA)
greater than 0.5 in the discovery cohort. To identify additional IA
risk loci, we pursued 14 loci with PPAs in the discovery cohort between 0.1 and 0.5. Twenty-five SNPs from these loci were genotyped using two independent Japanese cohorts, and the results
from discovery and replication cohorts were combined by metaanalysis. The results demonstrated significant association of IA with
rs6841581 on chromosome 4q31.23, immediately 5′ of the endothelin receptor type A with P = 2.2 × 10−8 [odds ratio (OR) = 1.22, PPA =
0.986]. We also observed substantially increased evidence of association for two other regions on chromosomes 12q22 (OR = 1.16, P =
1.1 × 10−7, PPA = 0.934) and 20p12.1 (OR = 1.20, P = 6.9 × 10−7, PPA =
0.728). Although endothelin signaling has been hypothesized to
play a role in various cardiovascular disorders for over two decades,
our results are unique in providing genetic evidence for a significant
association with IA and suggest that manipulation of the endothelin
pathway may have important implications for the prevention and
treatment of IA.
subarachnoid hemorrhage
| stroke | genetic risk loci
I
ntracranial aneurysms (IAs) are balloon-like dilations of cerebral arteries and affect 2–5% of the population (1). Although
most of these lesions are clinically silent, their rupture and
consequent subarachnoid hemorrhage usually occurs between
ages 40 and 60 without prior warning, resulting in substantial
morbidity and mortality (2, 3).
Aside from the well-established risk factors, such as hypertension, smoking, female sex (4), and high shear stress imposed
on the cerebrovasculature (5), there is evidence for significant
genetic contribution to IA pathogenesis (6). As is the case for
other multifactorial diseases, both common and rare variants are
thought to contribute to IA. In an attempt to identify common
variants that confer risk of IA, we previously completed two
genome-wide association studies of IA (6, 7). The larger, second
www.pnas.org/cgi/doi/10.1073/pnas.1117137108
genome-wide association study (7) implemented a discovery and
a replication phase using samples of European and Japanese
descent, respectively. Using a discovery cohort of 2,780 cases and
12,515 controls, we analyzed 831,532 genotyped and imputed
autosomal SNPs for association with IA. We used a Bayesian
measure of the strength of association—the posterior probability
of association (PPA)—to prioritize SNPs by calculating to what
extent the data supports association with IA (7). This analysis
revealed five loci with PPA > 0.5. Following replication genotyping using two independent Japanese cohorts and combining
the discovery and replication cohort results, all five loci surpassed the genome-wide significance level of 5 × 10−8 (observed
P values ≤ 2.5 × 10−9) with PPA ≥ 0.998, suggesting that each
locus contains a variant that confers risk of developing IA. The
five loci were on chromosomes 8q12.1 (SOX17), 9p21.3 (CDKN2A/
CDKN2B), 10q24.32 (CNNM2), 13q13.1 (KL/STARD13), and
18q11.2 (RBBP8) (7).
Because these five loci explained only ∼5% of the IA genetic
risk and the number of SNPs showing P values < 0.001 was
greater than that expected by chance alone (7), we hypothesized
the presence of additional true IA risk loci among a range of
SNPs showing weaker evidence of association in the discovery cohort. We tested this hypothesis using the two Japanese
replication cohorts.
Results
Analysis of Previously Uninvestigated Intervals. The statistical
analysis of the discovery cohort was described in detail previously
(7). Following strict sample and SNP quality control (QC)
measures, we matched cases and controls of the same sex based
Author contributions: R.P.L. and M.G. designed research; K.Y., M.B., S.-K.L., K.B., E.G.,
Y.M.R., M.N., A.H., P.B., H.K., J.E.J., D.K., G.A., M.S., B.K., A.K.O., S.M., G.J.E.R., H.S., J.H.,
K.S., H.Z., I.I., A.P., F.C., Y.N., and M.G. performed research; K.Y., M.B., S.-K.L., K.B., Y.N.,
R.P.L., and M.G. analyzed data; and K.Y., R.P.L., and M.G. wrote the paper.
The authors declare no conflict of interest.
Freely available online through the PNAS open access option.
1
To whom correspondence may be addressed. E-mail: richard.lifton@yale.edu or murat.
gunel@yale.edu.
This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.
1073/pnas.1117137108/-/DCSupplemental.
PNAS | December 6, 2011 | vol. 108 | no. 49 | 19707–19712
MEDICAL SCIENCES
Katsuhito Yasunoa,b,c, Mehmet Bakırcıoğlua,b,c, Siew-Kee Lowd, Kaya Bilgüvara,b,c, Emília Gaála,b,c,e, Ynte M. Ruigrokf,
Mika Niemeläe, Akira Hatag, Philippe Bijlengah, Hidetoshi Kasuyai, Juha E. Jääskeläinenj, Dietmar Krexk,
Georg Auburgerl, Matthias Simonm, Boris Krischekn, Ali K. Ozturka,b,c, Shrikant Maneo, Gabriel J. E. Rinkelf,
Helmuth Steinmetzl, Juha Hernesniemie, Karl Schallerh, Hitoshi Zembutsud, Ituro Inouep, Aarno Palotieq,
François Cambienr, Yusuke Nakamurad, Richard P. Liftonc,s,t,1, and Murat Günela,b,c,1
on inferred genetic ancestry to eliminate potential confounding
because of population stratification and sex. We then tested for
association of 831,529 QC-passed SNPs with IA and evaluated
the strength of association using PPA (7). In addition to the five
previously investigated loci with PPA > 0.5, there were 15 additional loci with PPA > 0.1 (observed values between 0.1 and
0.31) (Fig. 1 and Table S1). One of these intervals, on chromosome 7 (PPA = 0.31), was detected only by imputed SNPs
without support from nearby genotyped SNPs in linkage disequilibrium (LD), suggesting an imputation error (8). Therefore,
we pursued the two-stage follow-up genotyping for the remaining
14 loci (Table 1).
For the first stage, we genotyped 25 SNPs (SI Methods) located
within these 14 intervals using the larger of the two Japanese
cohorts (JP2), comprising 2,282 cases and 905 controls (Table
S2). All of these SNPs passed QC filters. Association tests
revealed that three of these loci, on chromosomes 4q31.23,
12q22, and 20p12.1, supported association with IA [i.e., Bayes
Factor (BF) > 1] (Table 1). Although the data also supported
association with IA at SNPs on chromosomes 1p36.31 and
2q33.1, the risk alleles were different from those found in the
discovery cohort (Table 1). Further study of these latter loci will
be needed to determine whether this might be because of allelic
heterogeneity between European and East Asian populations.
In the second stage, using the JP1 cohort, we genotyped 13
SNPs in a total of nine loci that showed BF > 0.5 in the JP2
cohort with the same risk allele as the discovery cohort (Table 1
and Table S2). Two of the genotyped SNPs (rs2282652 and
rs1132274) failed to pass the QC filters and were excluded from
further analysis, leading to coverage of eight of the nine loci. JP1
data supported association with IA at SNPs located in two
intervals, 4q31.23 and 12q22 (Table 1 and Table S2).
Meta-analysis of JP1 and JP2 cohorts revealed that the combined replication data strengthened the association with IA at
SNPs located within two of the eight loci on chromosomes
4q31.23 and 12q22 (rs6841581 and rs6538595, respectively) by
increasing the odds of association 85.1- and 24.0-fold, respectively (P = 0.00042 and 0.0017) (Table 2 and Table S2). One
more SNP, rs1132274, for which the JP1 cohort data were not
available, showed P = 0.021 and BF = 4.9 in the JP2 cohort.
Combined Results. We performed meta-analysis to combine the
results from the discovery and replication cohorts (Fig. 2 and
Table S2). The new genotyping results substantially increased the
strength of the evidence of association for three loci on chromosomes 4q31.23, 12q22, and 20p12.1 compared with the discovery data (Figs. 2 and 3, Table 2, and Table S2).
The strongest association was detected at rs6841581, located
at 148,620,640 base pairs on chromosome 4q31.23, with P =
2.2 × 10−8 [odds ratio (OR) = 1.22, PPA = 0.986] (Figs. 2 and 3).
Only a single gene, Endothelin Receptor Type A (EDNRA), is
located within the LD interval containing this SNP (Fig. 2). SNP
rs6841581 lies only 1,129 bases from the 5′ end of the EDNRA
transcript (NM_001957.3), located within an interval predicted
to have regulatory functions (University of California at Santa
Cruz genome browser, http://www.genome.ucsc.edu). The encoded protein, EDNRA, plays an important role in the cerebrovascular physiology (see below). An examination of the publicly
available eQTL databases (http://eqtl.uchicago.edu) did not reveal any SNPs that significantly alter EDNRA expression.
The second strongest association was at rs6538595 on chromosome 12q22 (OR = 1.16, P = 1.1 × 10−7, PPA = 0.934) (Figs.
2 and 3). A cluster of SNPs strongly correlated with rs6538595 is
associated with IA and is mapped within the introns of the
FYVE, RhoGEF, and PH domain-containing 6 (FGD6) gene (Fig.
2). Three other genes, NADH dehydrogenase (ubiquinone) 1 alpha subcomplex 12 (NDUFA12), nuclear receptor subfamily 2,
group C, member 1 (NR2C1), and Vezatin, Adherens Junctions
Transmembrane (VEZT), are located within the same LD interval as rs6538595.
Finally, although the JP1 cohort data were not available,
a missense variant, rs1132274, within the Ribosome Binding Protein
1(RRBP1) gene located on chromosome 20p12.1 showed moderate evidence of association with IA (OR = 1.20, P = 6.9 × 10−7,
PPA = 0.728) (Figs. 2 and 3). Another gene, Destrin (DSTN), is
also contained in the same LD interval.
There was no evidence for a two-locus interaction that was
consistent across all cohorts between various SNPs that were
found to be associated with IA (Table S3).
Cumulative Effect. Analysis of the cumulative effect of the seven
IA risk loci including the two SNPs (rs6841581 and rs6538595)
replicated here, as well as the previously identified five SNPs (see
Methods), explains 6.1%, 4.4%, and 4.1% of the familial risk in
the Finnish, European, and Japanese cohorts, respectively (Table S4). The ORs of developing IA between the top and bottom
1% risk groups representing the tails of the distribution of genetic profiles ranged between 5.74 and 8.39 for the Japanese,
European, and Finnish cohorts analyzed (Table S4).
Discussion
In this study, we demonstrated the association of SNPs located
within three intervals on chromosomes 4q31.23, 12q22, and
20p12.1 with IA. Among these, rs6841581, located on chromosome 4q31.23 near the ENDRA gene, showed the most significant association, with a P value of 2.2 × 10−8 (PPA = 0.986). The
data for another SNP, rs6538595, on chromosome 12q22 also
supported association with IA; replication data increased the
probability of association from 0.114 to 0.934 (P = 1.1 × 10−7).
The evidence of association for the third SNP, rs1132274, on
chromosome 20p12.1 was less. Although the replication data
from the JP2 cohort increased the support for association with
Fig. 1. Distribution of the PPA along the genome The PPAs for SNPs analyzed in ref. 7 are plotted along the genomic coordinates [National Center for
Biotechnology Information (NCBI) build 36]. We omitted 767,877 SNPs with PPA < 0.0001 from a total of 831,532 SNPs. Fifteen regions analyzed in this article
(i.e., those regions containing a SNP with 0.1 < PPA < 0.5 but no SNPs with PPA > 0.5) are shaded in gray.
19708 | www.pnas.org/cgi/doi/10.1073/pnas.1117137108
Yasuno et al.
Table 1. Cohort-wise association test results for 14 representative SNPs
Chromosome
1p36.31
1p22.2
1q21.3
2q33.1*
4q31.23
5q23.2
8p23.2
8q24.23*
11q22.2
12p13.31
12q22
19q13.12
20p12.1
22q12.1
JP2 (Replication 1)
JP1 (Replication 2)
SNP
Position
RA
P
PPA
RA
P
log10(BF)
RA
rs1876848
rs1725390
rs905938
rs787994
rs6841581
rs2287696
rs2045637
rs1554349
rs2124216
rs728342
rs6538595
rs1688005
rs1132274
rs133885
6,876,262
91,031,160
153,258,013
197,931,366
148,620,640
122,488,231
2,963,188
139,604,536
101,644,113
5,577,633
94,030,754
40,340,205
17,544,155
24,489,289
G
A
T
T
G
A
A
A
A
G
A
G
A
G
2.0E-05
2.0E-05
1.7E-05
2.1E-05
1.1E-05
1.1E-05
8.6E-06
7.9E-05
9.1E-06
1.2E-05
1.8E-05
1.6E-05
1.5E-05
1.6E-05
0.1128
0.1011
0.1252
0.0988
0.1750
0.1760
0.2139
0.0349
0.1963
0.1601
0.1136
0.1244
0.1435
0.1230
A
A
T
C
G
A
G
A
A
G
A
T
A
G
0.016
0.14
0.62
8.6E-05
0.0066
0.27
0.36
0.15
0.20
0.33
0.0051
0.16
0.012
0.67
0.61
−0.11
−0.18
2.51
0.93
−0.32
−0.38
−0.12
−0.24
−0.30
1.01
−0.069
0.69
−0.42
NA
A
T
T
G
NA
NA
G
G
G
A
NA
NA†
NA
P
log10(BF)
0.59
0.82
0.070
0.023
−0.40
−0.15
0.19
0.53
0.33
0.65
0.56
0.13
−0.28
−0.43
−0.33
−0.012
Genomic positions were based on the human genome build 36. NA, the SNP was not genotyped. RA, risk allele aligned to
the forward strand of the reference genome.
*For these loci, the lead SNPs were not genotyped (see Methods).
†
Genotyping of rs1132274 in the JP1 cohort failed.
IA, we could not reliably genotype the JP1 cohort using the
available platforms, thereby limiting the evidence for replication
at this locus to a single Japanese cohort. Thus, we considered
only rs6841581 and rs6538595 on chromosomes 4q31.23 and
12q22, respectively, as previously undetected risk loci for IA.
Although the addition of these two loci remarkably increased the
difference in the odds of disease between the highest and lowest
risk groups to 5.7- and 8.4-fold in Japanese and Finnish cohorts,
respectively, this only slightly improved the predictability of the
disease risk (Table S4).
The most significant locus, 4q31.23, contains a single gene,
EDNRA, which has been of great interest in various cardiovascular pathologies. Indeed, the endothelin system has been implicated in the pathogenesis of cardiovascular disorders, including
pulmonary and primary hypertension (9). EDNRA, along with
EDNRB, are G protein-coupled receptors for endothelins, with
the 21-aa endothelin-1 (EDN1) being the predominant isoform
(10, 11). EDN1 is produced primarily by the vascular endothelium and smooth-muscle cells and is involved in maintaining vasomotor control and vascular homeostasis (10). EDNRA is found on
vascular smooth-muscle cells, including the cerebrovasculature,
along with the heart, kidney, and neuronal cells (12), and mediates the vasoconstriction and mitogenic effects of EDN1 (13, 14).
On the other hand, EDNRB reside on both smooth-muscle and
endothelial cells, with downstream signaling through nitric oxide
(15). Thus, EDN1 and its receptors, EDNRA and EDNRB, play
key roles in the maintenance of the vasculature by controlling
the balance between vasoconstriction and vasodilation in response to hemodynamic stress.
EDNRA may play a role in IA pathogenesis in two distinct
ways, depending on whether the EDNRA-mediated signaling is
up- or down-regulated by the causal variant captured by the
established SNP rs6841581. Increased EDNRA-mediated signaling might predispose to progression and rupture of IA through
a mechanism analogous to the one that has been implicated for
the development of atherosclerosis (16). Endothelin signaling
has been shown to be activated at the site of vascular injury (17).
Immediately after the injury, this might be beneficial with recruitment of endothelial cells to repair the damage (18). Following
this initial response, however, prolonged, excessive endothelin signaling might be harmful by leading to atherosclerosis. EDNRA
mediates this vascular mitogenic effect of EDN1 by promoting cell
cycle progression and proliferation, both of which might play a
role in IA progression and rupture (13, 14, 19). Consistent with
this role of endothelins, increased EDN1 and EDNRA levels have
been reported following rupture of IAs (20–22). In addition, hypertension and smoking, both well-established risk factors of IA
pathogenesis, have been shown to alter the expression of endothelins (23).
Alternatively, if EDNRA-mediated signaling were attenuated,
the risk allele might predispose to the formation of IA because of
the failure of the repair mechanism mentioned above (17, 18).
Table 2. Summary of results for SNPs located in three unique genomic intervals on chromosomes 4q31.23, 12q22, and 20p12.1
SNP
Position
Gene
RA
Cohort
P value
log10 (BF)
PPA
4q31.23
rs6841581
148,620,640
EDNRA
G
G
rs6538595
94,030,754
NDUFA12/NR2C1/
FGD6/VEZT
A
1.1E-05
0.00042
2.2E-08
1.8E-05
3.33
1.93
5.84
3.11
0.1750
12q22
Discovery
Replication
Combined
Discovery
Replication
Combined
Discovery
Replication (JP2)
Combined
0.0017
1.1E-07
1.5E-05
0.012
6.9E-07
1.38
5.15
3.22
0.69
4.43
Chromosome
A
20p12.1
rs1132274
17,544,155
RRBP1
A
A
0.9857
0.1136
0.9343
0.1435
0.7279
OR (95% CI)
1.25
1.20
1.22
1.16
(1.13–1.39)
(1.08–1.32)
(1.14–1.31)
(1.08–1.24)
1.16
1.16
1.22
1.16
1.20
(1.06–1.28)
(1.10–1.23)
(1.11–1.33)
(1.03–1.30)
(1.11–1.28)
Genomic positions were based on the human genome build 36. OR, the per-allele odds ratio of the risk allele; RA, risk allele aligned to the forward strand
of the reference genome. Replication and combined results are based on the fixed-effects model.
Yasuno et al.
PNAS | December 6, 2011 | vol. 108 | no. 49 | 19709
MEDICAL SCIENCES
Discovery
Decreased signaling might interfere with the repair process in
response to vascular injury, limiting the recruitment of vascular
progenitor cells to the site of the damage with resultant defective
repair, which in turn might result in the formation of arterial
aneurysms. In support of this hypothesis, the use of at least one
EDNRA antagonist in primates has been shown to be associated
with the formation of preaneurysmal coronary artery lesions,
characterized by fragmentation of the internal elastic lamina and
loss of the medial smooth muscle (24).
Finally, endothelins, specifically EDNRA-mediated signaling,
have been implicated in the pathogenesis of cerebral vasospasm,
the pathologic vasoconstriction of the cerebral blood vessels that
can often result in delayed ischemic neurologic deficits following
IA rupture (22, 25). Several animal studies suggested a beneficial
effect of EDNRA inhibition in treating vasospasm, leading to
human clinical trials (26). The use of clazosentan, a selective
EDNRA antagonist with a predilection for the central nervous
system, has demonstrated reduction in angiographically demonstrated vasospasm, even though there was no improvement in
clinical outcome (27, 28). Given the potentially biphasic role of
endothelin signaling in IA pathogenesis, therapeutic strategies
involving EDNRA have to be considered cautiously. As mentioned above, the use of at least one EDNRA antagonist in
primates has been shown to be associated with the formation of
preaneurysmal coronary artery lesions (24).
The discovery of a significant association of IA with a risk allele
at immediate proximity to EDNRA within a predicted regulatory
region is unique in providing genetic evidence linking endothelins
to IA pathogenesis. Further studies will be needed to understand
the effects of this newly discovered risk variant on EDNRA-mediated signaling, and how these effects ultimately predispose to IA.
Once accomplished, this understanding can lead to pharmacological interventions that may have a potential therapeutic value in
the treatment of aneurysms before rupture.
Methods
Subjects. Consent was obtained from all study participants. The study cohorts
were described in detail elsewhere (7). Briefly, these cohorts included a genetically and sex-matched Finnish (FI) cohort of 808 cases and 4,393 controls,
and a combined European (CE) cohort of 1,972 cases and 8,122 controls. The
latter cohort consisted of three subcohorts based on the centers that
ascertained the case samples: the Netherlands (NL), Germany (DE), and
a pan-European (AN: @neurIST) cohort. The replication cohorts included two
independent Japanese case-control samples (JP1 and JP2). JP1 consisted of
829 cases and 761 controls; JP2 consisted of 2,282 cases and 905 controls.
Replication Strategy. We used a two-stage design to follow-up 14 candidate
intervals. In the first stage, we analyzed all of these loci using the JP2 cohort. For
the second stage, we chose SNPs that showed BF > 0.5 in the JP2 cohort with the
same risk allele as the discovery cohort, and analyzed them using the JP1 cohort.
Genotyping and QC. For SNPs reported in Table 1 and Table S2, we performed
genotyping of the JP1 cohort using either the MassARRAY (Sequenom) assay
or the Taqman (Applied Biosystems) platform. JP2 cases were genotyped
using the multiplex PCR-based Invader assay (Third Wave Technologies); JP2
controls were genotyped using the Illumina platform (29). We excluded SNPs
if any of the following three conditions were met in either cases or controls:
fraction of missing genotypes > 0.1, P value of the exact test of Hardy–
Weinberg equilibrium < 0.001, or minor allele frequency < 0.01.
Fig. 2. Regional plots for associated regions. For each chromosomal interval, −log10 P values for association test are plotted against the genomic
coordinates (NCBI build 36) (Upper); the recombination rates obtained from
the HapMap database and the RefSeq genes (hg18) within the regions
(Lower). The rs identifier of the SNP listed in Table 2 is shown for each
chromosomal interval, and its position is indicated by the gray vertical line
(Upper). Dark and light blue dots represent results of the genotyped and
imputed SNPs for the discovery cohort, respectively. Orange squares represent the association result from the replication cohort using the JP1 plus JP2
(rs6541581 and rs6538595) or JP2-only (rs1132274); combined P values of the
discovery and replication cohorts based on the fixed-effects model are
shown by red diamonds.
19710 | www.pnas.org/cgi/doi/10.1073/pnas.1117137108
Statistical Analysis. We tested for association between each SNP and IA by
fitting a logistic regression model with an additive effect of allele dosage and
a sex covariate. For multilocus analysis, we combined genotypes from JP1 and
JP2 and adjusted for the cohort label and sex, and analyzed the discovery
cohort using the conditional logistic regression (7). We performed metaanalysis to combine the cohort-wise results (SI Methods). We report the
results based on fixed-effects model in Fig. 2 and Table 2.
In addition to calculating the test P values, we also quantitatively measured the strength of the association using the BF and PPA, where the latter
provides a probabilistic measure of the strength of the evidence (30). We
regarded the association between an SNP and IA as replicated if BF > 10 in
Yasuno et al.
MEDICAL SCIENCES
Fig. 3. Consistency of association across cohorts. Forest plots are shown for meta-analysis of SNPs listed in Table 2. Squares and horizontal segments represent
estimated per-allele ORs and 95% confidence intervals (CIs) for individual cohorts. Diamonds represent the summary OR estimates and 95% CIs for the metaanalyses of six cohorts (fixed- and random-effects models). A log10(BF) > 0 supports association with IA, and a log10(BF) < 0 supports no association with IA.
ACKNOWLEDGMENTS. We thank the participants who made this study
possible, and for providing clinical information and biological samples
collected during the @neurIST project: Juan Macho (Hospital Clinic, Barcelona, Spain); Tamás Dóczi (University of Pècs Medical School, Pècs, Hungary);
James Byrne and Paul Summers (John Radcliffe Hospital, University of Oxford, Oxford, United Kingdom); Roelof Risselada and Miriam C. J. M.
Sturkenboom (Erasmus University Medical Center, Rotterdam, The Netherlands); Umang Patel, Stuart Coley, and Alan Waterworth (Royal Hallamshire
Hospital, Sheffield, United Kingdom); Daniel Rüfenacht (Swiss Neuro Institute, Clinic Hirslanden, Zürich, Switzerland); and regional collaborators
for the German case cohort, Beate Schoch in Essen, Bernhard Meyer in Bonn,
and Andreas Raabe in Bern, Switzerland. This study was supported by the
Yale Center for Human Genetics and Genomics and the Yale Program on
Neurogenetics; by National Institutes of Health Grants R01NS057756 and
R01NS067023 (to M.G.); and The Howard Hughes Medical Institute (to
R.P.L.). The @neurIST project was funded by the European Commission, VI
Framework Programme, Priority 2, Information Society Technologies, a European Public Funded Organization (Grant IST-FP6-027703). The Frankfurt case
cohort collection was supported by the German Federal Ministry of Education and Research (Grant 01GI9907).
1. Rinkel GJ, Djibuti M, Algra A, van Gijn J (1998) Prevalence and risk of rupture of intracranial aneurysms: a systematic review. Stroke 29:251–256.
2. International Study of Unruptured Intracranial Aneurysms Investigators (1998) Unruptured intracranial aneurysms—Risk of rupture and risks of surgical intervention.
N Engl J Med 339:1725–1733.
3. Bederson JB, et al. (2000) Recommendations for the management of patients with
unruptured intracranial aneurysms: A Statement for healthcare professionals from
the Stroke Council of the American Heart Association. Stroke 31:2742–2750.
4. Nahed BV, et al. (2007) Genetics of intracranial aneurysms. Neurosurgery 60:213–225;
discussion 225–226.
5. Nixon AM, Gunel M, Sumpio BE (2010) The critical role of hemodynamics in the development of cerebral vascular disease. J Neurosurg 112:1240–1253.
6. Bilguvar K, et al. (2008) Susceptibility loci for intracranial aneurysm in European and
Japanese populations. Nat Genet 40:1472–1477.
7. Yasuno K, et al. (2010) Genome-wide association study of intracranial aneurysm
identifies three new risk loci. Nat Genet 42:420–425.
the combined cohort of JP1 and JP2 (i.e., 10-fold increase in the odds of
association after observing replication data) (7).
We tested for deviation from a linear model in which two SNPs combine to
increase the log-odds of disease in an additive fashion. We also evaluated
potential clinical implications of the genetic profiles of the IA risk loci following the approach described by Clayton (31). See SI Methods for a detailed
description of the statistical analysis.
Yasuno et al.
PNAS | December 6, 2011 | vol. 108 | no. 49 | 19711
8. Wellcome Trust Case Control Consortium (2007) Genome-wide association study of
14,000 cases of seven common diseases and 3,000 shared controls. Nature 447:661–678.
9. Humbert M, Sitbon O, Simonneau G (2004) Treatment of pulmonary arterial hypertension. N Engl J Med 351:1425–1436.
10. Yanagisawa M, et al. (1988) A novel potent vasoconstrictor peptide produced by
vascular endothelial cells. Nature 332:411–415.
11. Inoue A, et al. (1989) The human endothelin family: Three structurally and pharmacologically distinct isopeptides predicted by three separate genes. Proc Natl Acad Sci
USA 86:2863–2867.
12. Yu JC, Pickard JD, Davenport AP (1995) Endothelin ETA receptor expression in human
cerebrovascular smooth muscle cells. Br J Pharmacol 116:2441–2446.
13. Alberts GF, Peifley KA, Johns A, Kleha JF, Winkles JA (1994) Constitutive endothelin-1
overexpression promotes smooth muscle cell proliferation via an external autocrine
loop. J Biol Chem 269:10112–10118.
14. Suzuki E, et al. (1999) Molecular mechanisms of endothelin-1-induced cell-cycle progression: involvement of extracellular signal-regulated kinase, protein kinase C, and
phosphatidylinositol 3-kinase at distinct points. Circ Res 84:611–619.
15. Miyauchi T, Masaki T (1999) Pathophysiology of endothelin in the cardiovascular
system. Annu Rev Physiol 61:391–415.
16. Kowala MC, et al. (1995) Selective blockade of the endothelin subtype A receptor
decreases early atherosclerosis in hamsters fed cholesterol. Am J Pathol 146:819–826.
17. Wang X, et al. (1996) Expression of endothelin-1, endothelin-3, endothelin-converting enzyme-1, and endothelin-A and endothelin-B receptor mRNA after angioplastyinduced neointimal formation in the rat. Circ Res 78:322–328.
18. Douglas SA, Ohlstein EH (1993) Endothelin-1 promotes neointima formation after
balloon angioplasty in the rat. J Cardiovasc Pharmacol 22(Suppl 8):S371–S373.
19. Lerman A, et al. (1991) Circulating and tissue endothelin immunoreactivity in advanced atherosclerosis. N Engl J Med 325:997–1001.
19712 | www.pnas.org/cgi/doi/10.1073/pnas.1117137108
20. Suzuki H, et al. (1990) Endothelin immunoreactivity in cerebrospinal fluid of patients
with subarachnoid haemorrhage. Ann Med 22:233–236.
21. Kraus GE, Bucholz RD, Yoon KW, Knuepfer MM, Smith KR, Jr. (1991) Cerebrospinal
fluid endothelin-1 and endothelin-3 levels in normal and neurosurgical patients: A
clinical study and literature review. Surg Neurol 35:20–29.
22. Itoh S, Sasaki T, Asai A, Kuchino Y (1994) Prevention of delayed vasospasm by an
endothelin ETA receptor antagonist, BQ-123: Change of ETA receptor mRNA expression in a canine subarachnoid hemorrhage model. J Neurosurg 81:759–764.
23. Xu CB, Sun Y, Edvinsson L (2010) Cardiovascular risk factors regulate the expression of
vascular endothelin receptors. Pharmacol Ther 127:148–155.
24. Stephan-Gueldner M, Inomata A (2000) Coronary arterial lesions induced by endothelin antagonists. Toxicol Lett 112-113:531–535.
25. Asano T, Ikegaki I, Suzuki Y, Satoh S, Shibuya M (1989) Endothelin and the production
of cerebral vasospasm in dogs. Biochem Biophys Res Commun 159:1345–1351.
26. Nirei H, et al. (1993) An endothelin ETA receptor antagonist, FR139317, ameliorates
cerebral vasospasm in dogs. Life Sci 52:1869–1874.
27. Macdonald RL, et al.; CONSCIOUS-1 Investigators (2008) Clazosentan to overcome
neurological ischemia and infarction occurring after subarachnoid hemorrhage
(CONSCIOUS-1): Randomized, double-blind, placebo-controlled phase 2 dose-finding
trial. Stroke 39:3015–3021.
28. Kramer A, Fletcher J (2009) Do endothelin-receptor antagonists prevent delayed
neurological deficits and poor outcomes after aneurysmal subarachnoid hemorrhage?: A meta-analysis. Stroke 40:3403–3406.
29. Kamatani Y, et al. (2009) A genome-wide association study identifies variants in the
HLA-DP locus associated with chronic hepatitis B in Asians. Nat Genet 41:591–595.
30. Stephens M, Balding DJ (2009) Bayesian statistical methods for genetic association
studies. Nat Rev Genet 10:681–690.
31. Clayton DG (2009) Prediction and interaction in complex disease genetics: Experience
in type 1 diabetes. PLoS Genet 5:e1000540.
Yasuno et al.