ARTICLE
GWAS and colocaliza ion analyses implica e
ca o id in ima-media hickness and ca o id
plaque loci in ca dio ascula ou comes
No a F anceschini1, Claudia Giamba olomei e al.
#
Ca o id a e y in ima media hickness (cIMT) and ca o id plaque a e measu es o subclinical
a he oscle osis associa ed wi h ischemic s oke and co ona y hea disease (CHD). He e, we
unde ake me a-analyses o genome-wide associa ion s udies (GWAS) in 71,128 indi iduals
o cIMT, and 48,434 indi iduals o ca o id plaque ai s. We iden i y eigh no el suscep -
ibili y loci o cIMT, one independen associa ion a he p e iously-iden ified PINX1 locus, and
one no el locus o ca o id plaque. Colocaliza ion analysis wi h nea by ascula exp ession
quan i a i e loci (cis-eQTLs) de i ed om a e ial wall and me abolic issues ob ained om
pa ien s wi h CHD iden ifies candida e genes a wo po en ially addi ional loci, ADAMTS9 and
LOXL4. LD sco e eg ession e eals significan gene ic co ela ions be ween cIMT and plaque
ai s, and bo h cIMT and plaque wi h CHD, any s oke sub ype and ischemic s oke. Ou
s udy p o ides insigh s in o genes and issue-specific egula o y mechanisms linking a he -
oscle osis bo h o i s unc ional genomic o igins and i s clinical consequences in humans.
No a F anceschini, Claudia Giamba olomei e al.
#
h ps://doi.o g/10.1038/s41467-018-07340-5 OPEN
Co espondence and eques s o ma e ials should be add essed o J.L.M.B. (email: [email p o ec ed])
o o C.J.O. (email: [email p o ec ed]).
#
A ull lis o au ho s and hei a filia ions appea s a he end o he pape .
NATURE COMMUNICATIONS | (2018) 9:5141 | h ps://doi.o g/10.1038/s41467-018-07340-5 | www.na u e.com/na u ecommunica ions 1
1234567890():,;
A he oscle osis is cha ac e ized by an accumula ion o
lipid- ich and inflamma o y deposi s (plaques) in he sub-
in imal space o medium and la ge a e ies. Plaque
enla gemen leads o blood flow limi a ion, o gan ischemia, and/
o issue nec osis. Plaque up u e can lead o ab up ascula
occlusion, which unde lies clinical ca dio ascula e en s,
including myoca dial in a c ion and ischemic s oke. Co ona y
hea disease (CHD) accoun s o one in se en dea hs, and s oke
accoun s o one in 20 dea hs in he US1. Because a he oscle osis
has a long p e-clinical phase, ea ly de ec ion o a he oscle osis
using non-in asi e me hods may help iden i y indi iduals a isk
o a he oscle o ic clinical e en s2, and p o ides an oppo uni y
o p e en ion. Subclinical a he oscle osis can be de ec ed by B-
mode ul asound measu emen o common ca o id a e y in ima-
media hickness (cIMT) o ca o id plaques1.
Subclinical and clinical a he oscle osis has known gene ic
componen s3. Genome-wide associa ion s udies (GWAS) o
subclinical a he oscle osis ha e p e iously iden ified h ee loci
significan ly associa ed wi h cIMT a ZHX2,APOC1, and PINX1,
and wo loci associa ed wi h common ca o id a e y plaque a
PIK3CG and EDNRA4. An exome-wide-associa ion s udy iden-
ified significan associa ions o he APOE ε2 allele wi h cIMT and
co ona y a e y calcifica ion5. The APOE single nucleo ide poly-
mo phism (SNP) s7412 is in linkage disequilib ium (LD) wi h
he APOC1 a ian , hus ep esen ing he same signal. Addi ional
GWAS-iden ified associa ions we e epo ed o ca o id plaque a
he 9p21 and SFXN2 loci6, and o cIMT a he CFDP1-
TMEM170A locus7. Howe e , hese p io s udies we e o limi ed
sample size and genomic co e age, and ailed o in es iga e he
e iological ole ha subclinical a he oscle osis may ha e on
a he oscle o ic clinical e en s.
He ein, we pe o m a la ge me a-analysis o GWAS o sub-
clinical a he oscle osis by analyzing 1000 Genomes impu ed
geno ype da a ob ained om collabo a ions be ween he Coho s
o Hea and Aging Resea ch in Genomic Epidemiology
(CHARGE) conso ium8and he Uni e si y College London-
Edinbu gh-B is ol (UCLEB) conso ium9. One o he g ea es
challenges in he ansla ion o GWAS findings o biological
unde s anding is ela ed o he limi ed access o RNA exp ession
da a om disease- ele an issues. Consequen ly, we sough o
eliably iden i y he issue-specific gene egula o y unc ions
esponsible o he GWAS signals by p io i izing candida e genes
o es ablished and no el loci o cIMT and ca o id plaque using
s a is ical me hods o colocaliza ion10. These me hods in eg a e
iden ified loci wi h exp ession quan i a i e loci (eQTLs) in e ed
om ca dio ascula disease- ele an gene ics o RNA exp ession,
he S ockholm-Ta u A he oscle osis Re e se Ne wo k Engi-
nee ing Task (STARNET) s udy, whe e a e ial wall and
me abolic- ela ed RNA samples we e collec ed om up o 600
pa ien s wi h CHD11. We also e alua e he ela ionships o cIMT
and ca o id plaque wi h clinically appa en CHD and s oke using
summa y da a om wo la ge conso ia. In summa y, ou s udy
sequen ially assesses he gene ic epidemiology and issue-specific
pa e ns o gene egula ion in ol ed in he o ma ion o sub-
clinical a he oscle osis ai s ac oss ca dio ascula disease- ela ed
issues.
Resul s
S udy desc ip ion. The s udy design is shown in Fig. 1.We
unde ook me a-analysis o GWAS in indi iduals o Eu opean
ances y o cIMT (up o 71,128 pa icipan s om 31 s udies) and
ca o id plaque (up o 48,434 pa icipan s om 17 s udies; 21,540
wi h defined ca o id plaque) (Supplemen a y Table 1). cIMT and
plaque we e e alua ed using high- esolu ion B-mode ul a-
sonog aphy and eading p o ocols as p e iously epo ed4. Ca -
o id plaque was defined by a he oscle o ic hickening o he
common ca o id a e y wall o he p oxy measu e o luminal
s enosis g ea e han 25% (Supplemen a y Table 2). Each coho
pe o med associa ion analyses using s anda dized p o ocols
(Me hods) o a ian s impu ed based on he 1000 Genomes
P ojec (1000G) phase 1 3 e e ence. Ex ensi e quali y con ol
(QC) was applied o da a, and he e was li le e idence o
popula ion s a ifica ion in any o he s udies o ei he ai
(Supplemen a y Table 3). The s udy-specific esul s we e com-
bined using fixed-e ec me a-analyses, gi en he low he e o-
genei y ac oss s udies (0% he e ogenei y)12.
GWAS me a-analyses o cIMT and ca o id plaque. Fo cIMT,
11 loci had a leas one SNP associa ion ha eached he genome-
wide associa ion h eshold (p<5×10
−8), o which eigh we e
newly desc ibed and h ee ha e been p e iously epo ed
(Table 1). The closes genes o he eigh loci we e: 1q32.2
in e genic ( s201648240), ATP6AP1L ( s224904), AIG1
( s6907215), PIK3CG ( s13225723), MCPH1 ( s2912063),
SGK223 ( s11785239), VTI1 ( s1196033), and CBFA2T3
( s844396). Fo h ee loci p e iously epo ed, he closes genes
we e ZHX2 ( s148147734), PINX1 ( s200482500), and APOE
( s7412).
The PIK3CG is a newly desc ibed locus o cIMT, bu has been
p e iously epo ed in a GWAS o ca o id plaque4. The wo
cIMT/Plaque, AOR/MAM cis-eQTLs: ou genes a h ee loci
ch 3:63561280-65833136 (ADAMTS9)
ch 10:99017729-101017321 (LOXL4)
ch 7:105299372-107743409 (CCDC71L, PRKAR2B)
Colocaliza ion
Subclinical GWAS and cis-eQTLs in AOR/MAM
Subclinical ai GWAS: cIMT and ca o id plaque
11 cIMT loci: 8 no el
ch 1:208953176 -indel, ch 5:81637916, ch 6:143608968,
ch 7:106416467, ch 8:6486033, ch 8:8205010,
ch 10:114410998, ch 16:88966667, ch 8:123401537-indel,
ch 8:10606223-indel, ch 19:45412079
Disco e y CHARGE and UCLEB conso ia
LD sco e eg ession
Subclinical GWAS and CHD/s oke GWAS
Sha ed gene ic basis
Genome-wide, wi h clinical ou comes
a
bLocal, wi h gene exp ession
cIMT/plaque and CHD cIMT and any s oke
cIMT and ischemic s oke
CARDIoGRAMPlusC4D
Signi ican co ela ions
5 plaque loci: 1 no el
ch 19:11189298, ch 4:148395284-indel,
ch 7:106411858, ch 9:22072301,
ch 16:75432686-indel
MEGASTROKE
Fig. 1 O e all s udy design. aGWAS me a-analyses o cIMT and ca o id plaque o gene disco e y. bLocal and genome-wide sha ed gene ic basis using
gene exp ession and clinical ou comes GWAS da a
ARTICLE NATURE COMMUNICATIONS | h ps://doi.o g/10.1038/s41467-018-07340-5
2NATURE COMMUNICATIONS | (2018) 9:5141 | h ps://doi.o g/10.1038/s41467-018-07340-5 | www.na u e.com/na u ecommunica ions
signals on ch omosome 8 nea MCPH1 ( s2912063) and SGK223
( s11785239) we e confi med o be independen h ough
condi ional analysis (Supplemen a y Table 4). A he PINX1
locus, he lowes associa ion p- alue a ian ( s200482500) was
no in LD wi h he p e iously epo ed associa ed a ian in he
egion ( s6601530, 2=0.0, Table 1), hus ep esen ing an
independen signal a his locus. Two addi ional loci o cIMT
had an SNP ha eached sugges i e e idence o associa ion (p<
1.0 × 10−7) including an SNP nea by APOB ( s515135) and an
in onic low equency a ian a ATG4B ( s139302128, mino
allele equency [MAF] =0.03) (Supplemen a y Table 5).
The GWAS me a-analysis o ca o id plaque iden ified fi e loci,
o which one has no been p e iously desc ibed (nea by gene
LDLR) (Table 1). A ou known loci associa ed wi h ca o id
plaque (nea by genes EDNRA,PIK3CG,CFDP1-TMEM170A, and
a he 9p21 egion), he mos significan ly associa ed a ian s
we e in LD wi h he p e iously epo ed SNPs (Table 1)4,6,7,
indica ing ha hese SNPs ma k he same associa ion a each
locus. Two sugges i e loci (p<10
−7) we e also iden ified nea by
he genes TMCO5B and STEAP2-AS1 (Supplemen a y Table 5).
Condi ional analyses confi med he p esence o a single
independen signal a each locus. Manha an and QQ plo s om
he me a-analysis o cIMT and ca o id plaque a e shown in
Supplemen a y Figu e 1 and egional plo s in Supplemen a y
Figu e 2. Fo es Plo s o all loci a e shown in Supplemen a y
Figu e 3.
Regula o y anno a ions o GWAS SNPs o cIMT/ca o id
plaque. To be e define po en ially causal a ian s wi hin he
iden ified gene ic isk loci, we join ly analyzed he GWAS da a
wi h unc ional genomic in o ma ion such as anno a ions on
ac i e ansc ip ion si es o open ch oma in egions (i.e., pe -
o med a fine-mapping unc ional genome-wide associa ion
analysis using GWAS13). Only a ian s in he PINX1 egion we e
ound o ha e a high p obabili y ha i s associa ion wi h cIMT is
d i en by SNPs ha all wi hin ansc ip ion si es in adipose-
de i ed mesenchymal s em cells a a DNaseI-hype sensi i e si e
(Supplemen a y Figu e 4), a finding ha p o ides a down-s eam
mechanis ic explana ion o he cIMT signal in he PINX1 locus.
To u he explo e he egula o y unc ions o a ian s in he
iden ified loci o cIMT and ca o id plaque, we in es iga ed
whe he he iden ified lead SNPs we e also eQTLs using ascula
RNAseq da a om GTEx (ao a, co ona y and ibial a e ies,
hea a ial appendage, and hea le en icle) and om he
co ona y a e y disease coho o STARNET (i.e., om he
a he oscle o ic-lesion- ee in e nal mamma y a e y [MAM] and
a he oscle o ic ao ic oo [AOR]). Lead SNP associa ed wi h
cIMT and ca o id plaque ( s13225723) in he PIK3CG locus was
ound o be ascula -specific eQTLs o CCDC71L and PRKAR2B
in GTEx ao a as well as in STARNET AOR and MAM issues
(Table 2, Fig. 2), sugges ing ha he gene ic egula ion o hese
wo genes a e esponsible o isk a ia ion in cIMT and ca o id
plaque de elopmen in his locus.
Colocaliza ion analysis o GWAS da a and STARNET eQTLs.
To iden i y u he candida e genes in issues a ec ed by a he -
oscle osis ha had s ong e idence o sha ing he same a ian o
cIMT and ca o id plaque as ound in ou GWAS, we conduc ed
pai wise colocaliza ion analysis o hese gene ic a ian s wi h cis-
eQTLs in he STARNET s udy10.
The pai wise colocaliza ion analysis is based on coloc, a
Bayesian s a is ical me hodology ha es s pai wise colocaliza ion
o SNPs in GWAS wi h eQTLs and, in his ashion, gene a es
pos e io p obabili ies o each locus weigh ing he e idence o
compe ing hypo hesis o ei he no colocaliza ion o sha ing o a
dis inc SNP a each locus10. We used summa y s a is ics om all
SNPs wi hin a 200-kb window a ound each gene co e ed by he
eQTL da ase s (N=18,705, see Me hods), and analyzed each
Table 1 Loci significan ly associa ed wi h cIMT and plaque GWAS
SNP Ch :posi ion Nea es coding
gene
Alleles (e ec /
o he )
E ec allele
eq.
Be a (SE) pN
Newly iden ified loci o cIMT
s201648240 1:208953176-indel LINC01717 −/AA 0.83 −0.0062 (0.0011) 4 × 10−954,752
s224904 5:81637916 ATP6AP1L C/G 0.95 −0.0088 (0.0016) 5 × 10−868,962
s6907215 6:143608968 AIG1 T/C 0.60 −0.0040
(0.0007)
5×10
−864,586
s13225723 7:106416467 PIK3CG A/G 0.22 0.0052 (0.0009) 3 × 10−968,070
s2912063 8:6486033 MCPH1 A/G 0.71 0.0045 (0.0008) 9 × 10−967,401
s11785239 8:8205010 SGK223 T/C 0.65 −0.0043
(0.0008)
9×10
−967,107
s11196033 10:114410998 VTI1A A/C 0.48 0.0042 (0.0008) 4 × 10−857,995
s844396 16:88966667 CBFA2T3 T/C 0.30 −0.0051 (0.0009) 6 × 10−950,377
Newly iden ified loci o plaque
s200495339 19:11189298-indel LDLR −/G 0.11 −0.1023 (0.0179) 1 × 10−836,569
Known loci o cIMT
s148147734a8:123401537-indel ZHX2 −/G 0.54 0.0050 (0.0007) 3 × 10
−11
58,141
s200482500a8:10606223-indel PINX1 −/GTACC 0.52 0.0056 (0.0008) 7 × 10−12 58,141
s7412a19:45412079 APOE T/C 0.08 −0.0119 (0.0015) 1 × 10−14 44,607
Known loci o plaque
s11413744b4:148395284-indel EDNRA −/T 0.86 −0.1586 (0.0253) 4 × 10
−10
39,577
s17477177b7:106411858 PIK3CG T/C 0.79 −0.1305 (0.0197) 4 × 10−11 47,863
s9632884b9:22072301 9p21 C/G 0.48 0.1127 (0.0163) 5 × 10−12 45,943
s113309773b16:75432686-indel CFDP1- TMEM170A −/C 0.46 −0.1259 (0.0194) 9 × 10−11 37,104
p=p- alues o associa ion om linea eg ession analysis, N= o al numbe in me a-analyses
aPublished cIMT SNP in LD wi h ou mos significan SNP: s11781551 ( 2=0.95 wi h s148147734), s6601530 ( 2=0 wi h s200482500), and s445925 ( 2=0.60 wi h s7412)
bPublished plaque SNP in LD wi h ou mos significan SNP: s1878406 ( 2=0.98 wi h s11413744), s17398575 ( 2=0.8 wi h s17477177), s9644862 ( 2=0.79 wi h s9632884), and s4888378 ( 2
=0.94 wi h s113309773)
NATURE COMMUNICATIONS | h ps://doi.o g/10.1038/s41467-018-07340-5 ARTICLE
NATURE COMMUNICATIONS | (2018) 9:5141 | h ps://doi.o g/10.1038/s41467-018-07340-5 | www.na u e.com/na u ecommunica ions 3
eQTL-GWAS da ase pai (Supplemen a y Table 6). A pos e io
p obabili y o ≥75% was conside ed s ong e idence o he issue-
specific eQTL-GWAS pai influencing bo h he exp ession and
GWAS ai a a pa icula egion. Resul s o his analysis a e
shown in Table 3and Supplemen a y Figu e 5. The s onges
e idence o an e ec on gene exp ession wi hin he egions
iden ified in ou s anda d GWAS me a-analysis was o he
CCDC71L and PRKAR2B genes a he p e iously desc ibed
ch omosome 7 cIMT locus (PIK3CG in Table 2, Fig. 2). These
genes showed e idence o colocaliza ion o bo h cIMT and
ca o id plaque in AOR and MAM issues (Table 3, Fig. 3).
CCDC71L had he highes p obabili y (>95%) o colocaliza ion
o cIMT, and MAM and AOR issue eQTLs, and o ca o id
plaque, and MAM and AOR issue eQTLs. We ound a low
p obabili y o colocaliza ion o he SNP wi h he PIK3CG gene
exp ession (<1%).
Table 2 Gene exp ession esul s o significan SNPs in GTEx and STARNET issues
SNP eQTLa(Gene, p) GTEx eQTLa(Gene, p) STARNET issues
AORbHEART (ATR/VEN)cAOR MAM
s201648240 CAMK1G, 0.0094
AL031316.1, 0.0040
CD34,0.00532
TRAF3IP3, 0.0097
s6907215 AL023584.1, 0.005384704 (VEN) ENSG00000217648,
0.00046
ENSG00000217648, 0.8 × 10−5
s13225723 AC005050.1,1×10
−10
ENSG00000177820.5, 7.0 × 10−5
CCDC71L,5×10
−6
PRKAR2B,4×10
−8
PIK3CG, 10 × 10−3
CCDC71L,6×10
−36
PRKAR2B,7×10
−7
SYPL1, 0.0043
CCDC71L,3×10
−33
PRKAR2B,6×10
−8
NAMPT,6×10
−6
s2912063 MCPH1, 0.0041 ENSG00000271743.1, 0.0093
(VEN)
MCPH1-AS1, 0.0020
s11785239 AC022784.1, 0.0078 (VEN) ERI1, 0.0069 PPP1R3B, 0.0036
s844396 ENSG00000141012.8, 0.003
AC092384.2, 0.001
CBFA2T3,1×10
−7
ZNF469, 0.004 (ATR)
AC092384.3,5×10
−6(ATR)
AC092384.1, 0.002 (ATR)
CBFA2T3, 0.0004 (ATR)
ZNF469, 0.002 (VEN)
AC138028.4, 0.001 (VEN)
ENSG00000224888.3, 0.009
(VEN)
PIEZO1, 0.0004 (VEN)
GALNS, 0.004 (VEN)
RPL13, 0.0024
ZNF276, 0.0070
TRAPPC2L, 0.0091
TRAPPC2L, 0.0040
ZNF276, 0.0059
s200495339 ENSG00000267105.1, 0.0005
(VEN)
s148147734 DERL1, 0.0082
s200482500 AF131215.6, 0.005
AF131215.5, 0.001
AF131215.5, 0.002 (ATR)
AF131215.6, 0.003 (VEN)
AF131215.5, 0.004 (VEN)
s7412 ENSG00000267163.1, 0.007
s11413744 PRMT9, 0.004
s17477177 ENSG00000267052.1,6×10
−11
ENSG00000177820.5,5×10
−6
CCDC71L,4×10
−7
PRKAR2B,2×10
−8
BCAP29, 0.002 (ATR) CCDC71L,2×10
−37
PRKAR2B,6×10
−7
SYPL1, 0.0091
CCDC71L,1 × 10−33
PRKAR2B,2×10
−8
NAMPT,1×10
−5
s9632884 DMRTA1, 0.007 (ATR) CDKN2B,2×10
−3CDKN2B,2×10
−3
s113309773 BCAR1,6×10
−11
ENSG00000261783.1,2×10
−16
GABARAPL2, 0.004
ENSG00000261783.1,1×10
−5
(ATR)
ENSG00000166822.8, 0.005
(ATR)
ENSG00000261783.1, 0.0003
(VEN)
ZFP1,4×10
−4
AC009078.2, 0.002
BCAR1,3×10
−12
CFDP1, 0.002
TMEM170A, 0.009
p=p- alues o associa ion om linea eg ession analysis
aThe lead SNP om GWAS is conside ed an eQTL i he cis-associa ion has a nominal p- alue o associa ion <0.01. Mul iple bu no all lead SNPs each genome-wide significance (p<10
−4).
bThis includes ao a (AOR)
cThis includes hea a ial (ATR) and hea le en icle (VEN)
ADAMTS9
LOXL4
PRKAR2B
CCDC71L
CIMT−AOR
CIMT−MAM
CIMT−SF
CIMT−SKLM
CIMT−VAF
CIMT−LIV
CIMT−Blood
PLAQUE−AOR
PLAQUE−MAM
PLAQUE−SF
PLAQUE−SKLM
PLAQUE−VAF
PLAQUE−LIV
PLAQUE−Blood
−1.0
−0.5
0.0
0.5
1.0
Pos e io
p obabili y
Fig. 2 Pai wise colocaliza ion esul s o genes iden ified o cIMT and
ca o id plaque GWAS me a-analysis wi h STARNET exp ession da ase s.
Red indica es a high pos e io p obabili y o colocaliza ion and blue a high
p obabili y o no colocaliza ion o he same SNP wi h issue eQTLs
ARTICLE NATURE COMMUNICATIONS | h ps://doi.o g/10.1038/s41467-018-07340-5
4NATURE COMMUNICATIONS | (2018) 9:5141 | h ps://doi.o g/10.1038/s41467-018-07340-5 | www.na u e.com/na u ecommunica ions
The eQTL associa ions a wo addi ional loci (ADAMTS9,
LOXL4) in MAM o AOR showed e idence o colocaliza ion wi h
cIMT o ca o id plaque, al hough GWAS associa ion p- alues a
hese loci did no mee he genome-wide significance h eshold
(Table 3, Supplemen a y Figu e 5). Albei wi h weake magni-
udes, he exp ession o hese wo genes we e also associa ed wi h
he op colocalizing SNPs as de ec ed in RNAseq da a in GTEx
ao a ( s17676309, ch 3:64730121, ADAMTS9,p=0.0003 and
s55917128, ch 10:100023359, LOXL4,p=0.0005).
Colocaliza ion o CHD and s oke GWAS and STARNET
eQTLs. We nex assessed i he ou genes (CCDC71L, PRKAR2B,
ADAMTS9, LOXL4) iden ified h ough colocaliza ion o cIMT/
ca o id plaque wi h issue-specific eQTLs also showed e idence
o colocaliza ion wi h CHD and s oke ai s (Supplemen a y
Da a 1 and Supplemen a y Figu e 6). We used GWAS summa y
da a o CHD (CARDIoGRAMPlusC4D), and s oke sub ypes
(MEGASTROKE) and AOR and MAM STARNET issue eQTLs
o hese analyses. CCDC71L and PRKAR2B had sugges i e e i-
dence o sha ing he same a ian wi h la ge essel disease s oke
in bo h AOR and MAM issues (p obabili y o colocaliza ion
≥20%, Supplemen a y Da a 1). In con as , he e was s ong
e idence (≥75%) o ejec a sha ed a ian o CHD and eQTLs a
his locus, hus sugges ing he e is a he oscle o ic ou come spe-
cifici y a ascula le el o his locus (Supplemen a y Figu e 5).
Th ee o hese genes, CCDC71L,PRKAR2B, and ADAMTS9,
showed e idence o sha ed gene ic influences o cIMT o ca o id
plaque on CHD/s oke ou comes when es ing he join associa-
ion using moloc, a mul iple- ai ex ension o coloc14 (Supple-
men a y Table 7). We also highligh he exp ession o KIAA1462
gene in MAM, ca o id plaque/cIMT, and CHD, which we e
posi i ely co ela ed (Supplemen a y Figu e 7). This gene has
sugges i e e idence o pai wise colocaliza ion wi h ca o id plaque
(67% o p obabili y o sha ed a ian be ween ca o id plaque and
eQTL in MAM), as well as a high p obabili y o sha ed a ian
be ween MAM eQTL exp ession o his gene, GWAS ca o id
plaque o cIMT, and CHD ai s (Supplemen a y Table 7). We
no e, howe e , ha he GWAS signal o ou comes ac oss he
da ase s did no each genome-wide significance and la ge
sample sizes may be needed o s eng hen he e idence o
in ol emen in disease ou comes.
Gene ic co ela ions o cIMT/ca o id plaque and clinical ou -
comes. To p o ide e iological insigh s in o he ole o measu es o
subclinical a he oscle osis and majo a he oscle o ic disease
ou comes such as CHD and ischemic s oke, we quan ified he
gene ic co ela ion using c oss- ai LD sco e eg ession, a
me hod ha es ima es gene ic co ela ion ac oss di e en ai s
using summa y le el da a15. We used summa y s a is ics be ween
cIMT/ca o id plaque wi h CHD and s oke me a-analysis o
GWAS. Bo h cIMT and ca o id plaque had posi i e significan
gene ic co ela ions wi h CHD (all p< 0.05 a e adjus ing o
mul iple es ing), hough he magni ude o he co ela ion was
wice as s ong o ca o id plaque (0.52) as o cIMT (0.20)
(Table 4). The e was also e idence o gene ic co ela ions
be ween cIMT wi h any s oke and ischemic s oke sub ype.
Pa hway analysis and d uggabili y. Gene On ology (GO) ana-
lyses o genes iden ified in he loci o cIMT and ca o id plaque
acco ding o ou me a-analysis o GWAS (Table 1and Supple-
men a y Table 5) and in he colocaliza ion analyses (Table 3,
Supplemen a y Table 7) showed ha cIMT genes a e en iched in
lipop o ein- ela ed e ms and choles e ol e flux, whe eas ca o id
plaque genes a e en iched in e ms associa ed wi h fib oblas
apop osis (Supplemen a y Figu e 8). Analysis o he cIMT genes
using a GO Slim addi ionally iden ified se e al o he genes ha
we e associa ed wi h e ms desc ibing ca dio ascula de elop-
men , cell adhesion, and immune p ocesses, p ocesses al eady
conside ed ele an o a he oscle osis. Specifically, he e is co -
obo a ing e idence om GO ha CCDC71L, PRKAR2B, and
TWIST1 a e associa ed wi h cIMT/ca o id plaque as hey a e
in ol ed in lipid me abolism, wi h simila suppo ha
ADAMTS9, CDH13, and KIAA1462 a e associa ed wi h cIMT o
ca o id plaque isk as hey a e all in ol ed in cell adhesion and,
oge he wi h TWIST1, in ca dio ascula sys em de elopmen
(Supplemen a y Da a 2).
F om he loci associa ed wi h cIMT and ca o id plaque, we
iden ified se en genes (ATG4B, ALPL, LDLR, APOB, EDNRA,
APOE, and ADAMTS9) whose encoded p o eins a e a ge s a
a ious s ages o he d ug de elopmen p ocess (Supplemen a y
Tables 8 and 9). ADAMTS9 gene encodes a p o ein likely o be
d uggable16.ATG4B, ALPL, and LDLR a e p o eins being a ge ed
by compounds in p e-clinical phase ( ie 2), while APOB and
EDNRA a e p o eins a ge ed by d ugs in clinical phase o
licensed ( ie 1). APOB is he a ge o an app o ed FDA d ug o
ea men o amilial hype choles e olemia. EDNRA gene encodes
o endo helin A ecep o , agains which se e al an agonis s ha e
been de eloped o he ea men o pulmona y a e ial
Table 3 Colocaliza ion o cIMT and plaque wi h eQTLs in issues om pa ien s wi h CHD in STARNET issues o genes/ issues
combina ions ha ha e mo e han 75% p obabili y o sha e he same associa ed a ian
Region (ch :s a -s op) T ai Gene SNP wi h bes join p obabili y p, BETA (SE), Tissue pos e io p obabili y (PPA)aDi ec ion o e ec GWAS/eQTL
cIMT /plaque GWAS AOR eQTL MAM eQTL
ch 3:63561280-65833136 cIMT ADAMTS9 s17676309 (T/C) 2 × 10−6,
-0.0035 (0.0007)
2×10
−25,
−0.65 (0.06)
PPA=0.93
1×10
−23,
−0.61 (0.06)
PPA=0.89
−/−
ch 10:99017729-101017321 cIMT LOXL4 s55917128 (T/C) 5 × 10−7,
0.0037 (0.0007)
6×10
−8,
0.33 (0.06)
PPA=0.79
+/+
ch 7:105299372-107743409 cIMT CCDC71L
PRKAR2B
s12705390 (A/G) 5 × 10−9,
0.0049 (0.0008)
2×10
−37,
0.81 (0.06)
PPA=0.97
6×10
−7,
0.34 (0.07)
PPA=0.93
1×10
−33,
0.755 (0.06)
PPA=0.97
2×10
−8,
0.368 (0.06)
PPA=0.96
+/+
+/+
Plaque CCDC71L
PRKAR2B
s12705390 (A/G) 4 × 10−8,
0.12 (0.022)
2×10
−37,
0.80 (0.06)
PPA=0.97
6×10
−7,
0.33 (0.07)
PPA=0.93
1×10
−33,
0.75 (0.06)
PPA=0.97
2×10
−8,
0.37 (0.06)
PPA=0.96
+/+
+/+
PPA pos e io p obabili y o sha ing same SNP highe han 75%, cIMT common ca o id a e y in ima-media hickness, AOR ao a, MAM mamma y a e y
aThis signal eaches genome-wide significance in cIMT/plaque, and eaches a high p obabili y o being media ed by he genes in AOR and MAM
NATURE COMMUNICATIONS | h ps://doi.o g/10.1038/s41467-018-07340-5 ARTICLE
NATURE COMMUNICATIONS | (2018) 9:5141 | h ps://doi.o g/10.1038/s41467-018-07340-5 | www.na u e.com/na u ecommunica ions 5

hype ension o which a e in ad anced clinical phase de elop-
men o non-small cell lung cance and diabe ic neph opa hy.
Discussion
We p o ide esul s o a la ge me a-analysis o GWAS o sub-
clinical a he oscle osis and we in eg a e ou esul s wi h issue-
specific gene exp ession da a using eQTLs om bo h he ea ly
(MAM) and la e ad anced (AOR) a he oscle o ic a e ial wall
om he STARNET s udy o enable eliable disco e y o genes
wi h biological e idence o an inc eased p obabili y o con e ing
inhe i ed isk o a he oscle osis de elopmen . Ou disco e y
app oach using GWAS me a-analyses iden ified 16 loci sig-
nifican ly associa ed wi h ei he cIMT o ca o id plaque, o which
nine a e no el.
The in eg a ion o GWAS and issue-specificcis-eQTLs o he
join analyses o issue-specific eQTLs om CHD pa ien s iden-
ified wo po en ially addi ional loci colocalizing wi h cIMT o
ca o id plaque: ch 3:63561280-65833136 (ADAMTS9),
ch 10:99017729-101017321 (LOXL4). ADAMTS9 is a me allo-
p o einase in ol ed in h ombosis and angiogenesis and has been
associa ed wi h ca diome abolic ai s (wais - o-hip a io, wais
ci cum e ence, and ype 2 diabe es) in GWAS, and wi h co ona y
a e y calcifica ion in a gene-by-smoking in e ac ion GWAS17,18.
LOXL4 encodes a lysyl oxidase in ol ed in c osslinks o collagen
and elas in in he ex acellula ma ix. This amily o p o eins a e
in ol ed in he de elopmen o elas ic essels and mechanical
s eng h o he essel wall, and hei inhibi ion was associa ed
wi h he de elopmen o abdominal ao ic aneu ysms and mo e
se e e a he oscle osis in expe imen al models19.
cIMT Plaque
eQTL in ao a eQTL in SF
106.5106.4
7:106410777
7:106416467
7:106241491
7:106410777
106.3
Posi ion on ch 7 (Mb)
106.2106.1
106.5106.4106.3
Posi ion on ch 7 (Mb)
106.2106.1
0
10
20
–Log10 (exp ession in AOR P- alue)
–Log10 (exp ession in SF P- alue)
30
0
2
4
6
8
10
0
2
4
6
8
10
–Log10 (cIMT P- alue)
–Log10 (Plaque P- alue)
0
2
4
6
8
10
a
c
b
d
40
0.2
0.4
0.6
0.8
0.2
0.4
0.6
0.8
0.2
0.4
0.6
0.8
2
0.2
0.4
0.6
0.8
106.5106.4106.3
Posi ion on ch 7 (Mb)
106.2106.1
106.5
106.4106.3
Posi ion on ch 7 (Mb)
106.2106.1
CCDC71L CCDC71L
CCDC71L CCDC71L
2
2
2
Fig. 3 Associa ion esul s a he CCDC71L locus (ch omosome 7), showing a high pos e io p obabili y o a sha ed a ian o cIMT and ca o id plaque in
AOR and MAM eQTLs. −log10(p) SNP associa ion p- alues o cIMT (plo A) and ca o id plaque (plo B), and eQTL in AOR (plo C) and eQTL in SF (plo
D). Associa ion esul s in SF issue ha e a low p obabili y o a sha ed signal wi h cIMT and ca o id plaque, possibly indica ing a di e en mechanism in his
issue. eQTLs in MAM a e iden ical o AOR and no shown. The p- alues we e calcula ed by fi ing a linea eg ession model wi h cIMT o plaque as
dependen a iable and impu ed SNPs as independen a iables. Each do is an SNP and he colo indica es linkage disequilib ium ( 2) wi h he bes hi (in
pu ple)
ARTICLE NATURE COMMUNICATIONS | h ps://doi.o g/10.1038/s41467-018-07340-5
6NATURE COMMUNICATIONS | (2018) 9:5141 | h ps://doi.o g/10.1038/s41467-018-07340-5 | www.na u e.com/na u ecommunica ions
Some loci iden ified in ou me a-analysis o GWAS include
genes in known pa hways o a he oscle osis, including LDLR,
which is ela ed o lipid pa hways and CHD, and iden ified o
associa ions wi h ca o id plaque in ou s udy. Fo mos o he loci,
howe e , he unde lying gene implica ed in signals a e unknown.
Ou colocaliza ion app oach ound bo h CCDC71L and
PRKAR2B as he mos likely genes a he ch omosome 7 locus,
whe e PIK3CG was p e iously he sugges ed gene. This finding is
in ag eemen wi h a a ge ed sequencing s udy o subclinical
a he oscle osis15. An addi ional SNP ( s342286) a his locus has
been associa ed wi h pla ele s olume and eac i i y, and ca di-
o ascula ai s. Howe e , s342286 is no in LD wi h ou mos
significan SNP and i is no associa ed wi h cIMT o ca o id
plaque in ou s udies (p=0.49 and 0.01, espec i ely). O in e es ,
he a ian we iden ified in his s udy showed e idence o
colocaliza ion wi h cIMT/ca o id plaque and la ge essel disease
s oke bu no CHD, he e o e showing issue and ou come-
specifici y. CCDC71L has unknown unc ion. PRKAR2B codes o
one o he se e al egula o y subuni s o cAMP-dependen p o-
ein kinase and i s exp ession is ubiqui ous. In i o s udies ha e
shown ha adenosine-induced apop osis o a e ial smoo h
muscle cells in ol es a cAMP-dependen pa hway20.
Measu es o cIMT and ca o id plaque eflec ascula pa ho-
physiologic and a he oscle osis p ocesses, espec i ely, wi h ca -
o id plaque mo e s ongly eflec ing a he oscle o ic clinical
e en s. An impo an con ibu ion o his s udy is he suppo ing
e idence o o e all gene ic co ela ions o CHD and s oke (any
cause and ischemic s oke) wi h subclinical a he oscle osis ai s,
es ima ed using LD sco e me hods. Fu he highligh ing he
po en ial biological ele ance o ou findings, he gene ic co e-
la ions es ima es o CHD we e s onge o ca o id plaque han
o cIMT. Howe e , cIMT and ca o id plaque GWAS we e co -
ela ed, and he gene ic co ela ions es ima es wi h s oke we e
simila o cIMT o ca o id plaque, and no significan o ca o id
plaque. The colocaliza ion analyses p o ided addi ional insigh s
in he ela ionships be ween subclinical a he oscle osis, clinical
ou comes, and issue-specific egula ion a specific genomic
egions. Fo example, ou sugges i e op gene associa ion in
mul i- ai colocaliza ion o KIAA1462 included MAM eQTLs,
ca o id plaque, and CHD, suppo ing he sha ed gene ic e ec s a
his locus o a he oscle osis in ca o id and co ona y a e ies.
KIAA1462 has been p e iously epo ed in he same locus iden-
ified by GWAS o CHD21. This gene encodes a p o ein in ol ed
in cell–cell junc ions in endo helial cells22, which was ecen ly
shown o be in ol ed in pa hologic angiogenic p ocess in in i o
and in i o expe imen al models23. These findings sugges ha
he e may be impo an di e ences in ascula bed egula ion a
dis inc i e egions o a he oscle o ic ca dio ascula and s oke
ou comes ha may help o iden i y genes and specific a ge s o
CHD o s oke p e en ion and ea men .
Addi ional s udies in di e se and la ge samples ac oss he
mul iple da ase s a e needed o explo e hese esul s u he . As
mo e summa y s a is ics become a ailable o o he clinical end-
poin s beyond s oke and CHD (bo h in e ms o la ge sample
size and iche genome co e age), and as u he efinemen s in
clinical pheno ypes eme ge (e.g. om CHD o acu e co ona y
synd ome sub-componen s), s a egies o in eg a e his knowl-
edge using me hods such as moloc10 and eCAVIAR24 will con-
inue o be essen ial o ha nessing genome-wide findings in he
d ug-disco e y p ocess.
In summa y, ou s udy is a la ge GWAS me a-analysis o cIMT
and ca o id plaque. Th ough a sequen ial app oach o disco e y
and colocaliza ion s udies, we p o ide deepe insigh s in o disease
causal genes o subclinical cIMT and ca o id plaque o ma ion.
We confi med h ee loci and iden ified nine no el loci in he
me a-analyses o cIMT and ca o id plaque. Addi ionally, we
p o ide s ong e idence o he ole o h ee no el genes om ou
in eg a i e analysis o GWAS and eQTL da a. Mo eo e , he
iden ified co ela ions wi h CHD and s oke highligh no el
biological pa hways ha me i u he assessmen s as no el a -
ge s o d ug de elopmen .
Me hods
E hics s a emen . All human esea ch was app o ed by he ele an ins i u ional
e iew boa ds o each s udy, and conduc ed acco ding o he Decla a ion o
Helsinki. All pa icipan s p o ided w i en in o med consen .
Popula ions and pheno ypes. The disco e y GWAS in his s udy consis s o a
collabo a ion be ween he CHARGE 8and he UCLEB conso ia9, o gene ic
s udies o cIMT and ca o id plaque among indi iduals o Eu opean ances y
(Supplemen a y No e 1). All s udies ollowed s anda dized p o ocols o pheno ype
asce ainmen and s a is ical analyses. The desc ip i e cha ac e is ics o pa ici-
pa ing s udies a e shown in Supplemen a y Table 1.
cIMT and ca o id plaque measu es we e e alua ed using high- esolu ion B-
mode ul asonog aphy and eading p o ocols as p e iously epo ed4. We used
da a om he baseline examina ion o he fi s examina ion in which ca o id
ul asonog aphy was ob ained. cIMT was defined by he mean o he maximum o
se e al common ca o id a e y measu emen s, measu ed a he a wall o he nea
wall. Fo mos s udies, his was an a e age o mul iple measu emen s om bo h he
le and igh a e ies. We also examined a ca o id plaque pheno ype, defined by
a he oscle o ic hickening o he ca o id a e y wall o he p oxy measu e o
luminal s enosis g ea e han 25% (Supplemen a y Table 2).
Geno yping, impu a ion, and s udy-le el quali y con ol. Geno yping a ays and
QC p e-impu a ion a e shown in Supplemen a y Table 3. Each GWAS s udy
Table 4 Gene ic co ela ion be ween CHD and s oke ai s wi h cIMT and plaque, and cIMT wi h plaque using LD sco e and
me a-GWAS
Ca dio ascula disease ai Subclinical a he oscle osis ai Gene ic co ela ion SE zp
CHDacIMT 0.20 0.05 4.1114 4 × 10−5
Any s oke cIMT 0.30 0.07 4.2301 2.3 × 10−5
Ischemic s okebcIMT 0.31 0.07 4.646 3.4 × 10−6
Ca dio-embolic s okebcIMT 0.10 0.09 1.0729 0.28
Small essel disease s okebcIMT 0.33 0.18 1.8728 0.06
CHDaCa o id plaque 0.52 0.08 6.4263 1.3 × 10−10
Any s okebCa o id plaque 0.28 0.10 2.7097 0.007
Ischemic s okebCa o id plaque 0.27 0.10 2.6578 0.008
Ca dio-embolic s okebCa o id plaque 0.06 0.14 0.4684 0.64
Small essel disease s okebCa o id plaque −0.03 0.24 −0.1344 0.89
Plaque cIMT 0.40 0.10 3.9667 7.3 × 10−5
aCARDIoGRAMPlusC4D
bMEGASTROKE conso ium. Unable o es ima e he gene ic co ela ions wi h la ge essel disease
NATURE COMMUNICATIONS | h ps://doi.o g/10.1038/s41467-018-07340-5 ARTICLE
NATURE COMMUNICATIONS | (2018) 9:5141 | h ps://doi.o g/10.1038/s41467-018-07340-5 | www.na u e.com/na u ecommunica ions 7
conduc ed genome-wide impu a ion using a Phase 1 in eg a ed (Ma ch 2012
elease) e e ence panel om he 1000G Conso ium using IMPUTE225 o MaCH/
minimac26, and used Human Re e ence Genome Build 37. Sample QC was pe -
o med wi h exclusions based on call a es, ex eme he e ozygosi y, sex dis-
co dance, c yp ic ela edness, and ou lying e hnici y. SNP QC excluded a ian s
based on call a es ac oss samples and ex eme de ia ion om Ha dy–Weinbe g
equilib ium (Supplemen a y Table 3). Non-au osomal SNPs we e excluded om
impu a ion and associa ion analysis.
P e-me a-analysis GWAS s udy-le el QC was pe o med using EasyQC
so wa e27. This QC excluded ma ke s absen in he 1000G e e ence panel; non A/
C/G/T/D/I ma ke s; duplica e ma ke s wi h low call a e; monomo phic SNPs and
hose wi h missing alues in alleles, allele equency, and be a es ima es; SNPs wi h
la ge e ec es ima es o s anda d e o (SE) ≥10; and SNPs wi h allele equency
di e ence >0.3 compa ed o 1000G e e ence panel. The e was a o al o 9,574,088
SNPs o he cIMT me a-analysis and 8,578,107 SNPs o he ca o id plaque me a-
analysis.
S a is ical analyses. Wi hin each s udy, we used linea and logis ic eg ession o
model cIMT and ca o id plaque, espec i ely, and an addi i e gene ic model (SNP
dosage) adjus ed o age, sex, and up o 10 p incipal componen s. We combined
summa y es ima es om each s udy and each ai using an in e se a iance
weigh ed me a-analysis. Addi ional fil e s we e applied du ing me a-analyses
including impu a ion quali y (MACH 2< 0.3 and IMPUTE in o <0.4), a mino
allele equency (MAF) <0.01, and SNPs ha we e no p esen in a leas ou
s udies. The genome-wide significance h eshold was conside ed a p< 5.0 × 10−8.
To assess he e idence o independen associa ions a each locus a aining
genome-wide significance, we pe o med condi ional analysis in a 1-Mb genomic
in e al flanking he lead SNP using GCTA28. This app oach uses summa y me a-
analysis s a is ics and a LD ma ix om an ances y-ma ched sample o pe o m
app oxima e condi ional SNP associa ion analysis. The es ima ed LD ma ix was
based on 9713 un ela ed indi iduals o Eu opean ances y om he ARIC s udy,
which was geno yped using an A yme ix 6.0 a ay and impu ed o he 1000G
panel using IMPUTE225.
Gene exp ession analysis using GTEx. GTEx Analysis V6 (dbGaP Accession
phs000424. 6.p1) eQTL esul s we e downloaded om GTEx po al o 44 issues,
and hen mapped o SNPs lis ed in Table 1. We used a alse disco e y a e (FDR) o
≤0.05.
Colocaliza ion analyses using eQTLs. We in eg a ed ou GWAS esul s wi h cis-
eQTL da a using a Bayesian me hod (coloc)10. This me hod e alua es whe he he
GWAS and eQTL associa ions bes fi a model in which he associa ions a e due o
a single sha ed a ian (summa ized by he pos e io p obabili y). We used gene
exp ession da ase s om mul iple issues om pa ien s wi h CHD o he STAR-
NET s udy, including blood, MAM, AOR, subcu aneous a (SF), isce al a
(VAF), skele al muscle (SKLM), and li e (LIV) ob ained om 600 pa ien s du ing
open hea su ge y11. Pai wise colocaliza ion was es ed be ween hese exp ession
disease issue da ase s and GWAS esul s om ou cIMT/ca o id plaque GWAS
me a-analysis. We used GWAS and eQTL summa y s a is ics o SNPs wi hin a
200-kb window a ound each gene co e ed by he eQTL da ase s. A pos e io
p obabili y o colocaliza ion ≥0.75 was conside ed a s ong e idence o a causal
gene. Nex , we epo ed he gene(s) in he STARNET da ase s ha had he
s onges e idence o sha ing he same a ian wi h cIMT o ca o id plaque
genome-wide. In an al e na i e analysis, we also es ed loci wi h an SNP ha
eached a h eshold o significan o sugges i e genome-wide significance o cIMT
o ca o id plaque ( epo ed in Table 1, Supplemen a y Table 5). Fo each egion
200kb a ound he SNP wi h he lowes associa ion p- alue, we epo he gene wi h
he highes p obabili y o being esponsible o he GWAS signal (Supplemen a y
Table 6).
Pai wise colocaliza ion o hese genes was also es ed o publicly a ailable
GWAS o CHD case-con ols (CARDIoGRAMPlusC4D) and s oke case-con ols
(MEGASTROKE conso ium). The MEGASTROKE da ase uses geno ypes
impu ed o he 1000G phase I haplo ype panel. The Eu opean ances y sample used
o gene a e hese esul s consis ed o 40,585 s oke cases and 406,111 con ols om
15 coho s and wo conso ia: he METASTROKE and CHARGE conso ia29. The
pheno ypes used in his analysis we e any s oke (n=39,067 cases, o al n=
442,142), ischemic s oke (IS, n=32,686 cases, o al n=423,266), and e iologic
s oke sub ypes:ca dioembolic s oke (CE, n=6,820 cases, o al n=314,368), la ge
essel disease (n=4,113, o al n=202,263), and small essel disease (SVD, n=
4,975, o al n=242,250). To explo e mul i- ai colocaliza ions, we used moloc14
wi h p io p obabili ies o 10−4 o GWAS/GWAS/eQTL, 10−6 o GWAS+eQTL/
GWAS o GWAS+GWAS/eQTL, and 10−7 o colocaliza ion o all h ee
associa ion signals.
Func ional anno a ion and epigene ic en ichmen analyses. F om he Epigen-
ome Roadmap P ojec 30,31, we ob ained egula o y in o ma ion using b oad classes
o ch oma in s a es (n=127 issues) cap u ing p omo e -associa ed, ansc ip ion-
associa ed, ac i e in e genic, and la ge-scale ep essed and epea -associa ed s a es.
F om ENCODE32, we ob ained ch oma in s a es, uni o mly p ocessed ansc ip-
ion ac o (TF) Chip assays and DNaseI Hype sensi i i y si es (DHS) o nine cells
lines. F om FANTOM533, we used in o ma ion om exp ession o enhance s in
each issue (n=112), and enhance s ha a e posi i ely di e en ially exp essed
agains any o he issue (n=110).
We used GWAS13 o iden i y genomic anno a ions ha a e en iched wi hin he
cIMT esul s and o selec he a ian s wi h suppo o a unc ional ole based on
he mos in o ma i e anno a ions. We only conside ed cIMT o hese analyses
because o he small numbe o iden ified loci o ca o id plaque. We fi s es ima ed
he en ichmen pa ame e s o each anno a ion indi idually and iden ified he se
o anno a ions wi h significan ma ginal associa ions. We hen applied 10- old
c oss- alida ion likelihood and o wa d selec ion o iden i y he se o anno a ions
ha significan ly imp o e he model fi , and e e se selec ion o each anno a ion
included in he model, as sugges ed in he GWAS wo kflow. We epo ed he
model wi h he highes c oss- alida ion likelihood and SNPs ha ha e egional
pos e io p obabili y o associa ion (PPA) >0.9 and di ec ly o e lap he genomic
anno a ions conside ed.
O e all gene ic co ela ion analysis. Gene ic co ela ion be ween cIMT/ca o id
plaque, CHD, and s oke ai s we e calcula ed using LD sco e eg ession app oach
LD-sco e, which uses GWAS summa y s a is ics and is no a ec ed by sample
o e lap. This me hod elies on he ac ha he χ2associa ion s a is ic o a gi en
SNP includes he e ec s o all SNPs ha a e in LD wi h i and i calcula es gene ic
co ela ion by pa i ioning he SNP he i abili ies15. Gene ic co ela ions be ween
s oke ai s (IS, CE, la ge essel disease, and SVD) and cIMT and ca o id plaque
we e calcula ed using so wa e a ailable a h p://gi hub.com/bulik/ldsc wi h
GWAS summa y s a is ics o ou cIMT/ca o id plaque GWAS, CARDIO-
GRAMPlusC4D da a, and s oke GWAS. We used he LD-sco es15, which a e
based on he 1000 Genomes Eu opean popula ion and es ima ed wi hin 1-cM
windows. Based on en es s pe o med ( wo subclinical ai s and fi e ou comes),
we se he significance h eshold o p=0.005.
PATHWAY ANALYSES.Me hods o GO Slim: The Ensembl iden ifie s o all
p o ein-coding genes iden ified as in LD wi h he 12 a ian s o cIMT and 15
a ian s o ca o id plaque (including a ian s om main and sugges i e signals,
Table 1and Supplemen a y Table 5), and fi e genes o which he e is s ong
e idence o colocaliza ion (Table 3), we e mapped o UniP o accession numbe s,
using he UniP o ID mapping se ice (h p://www.unip o .o g/uploadlis s/). A GO
Slim analysis was pe o med on his lis using QuickGO (www.ebi.ac.uk/QuickGO)
and he Gene ic GO Slim. The GO e ms used in he final slim analysis we e
u he efined by adding/ emo ing GO e ms o p o ide mo e de ailed
in o ma ion abou he p ocesses co e ed.
Me hods o GO e m en ichmen analysis: The VLAD gene lis analysis and
isualiza ion ool (h p://p o o.in o ma ics.jax.o g/p o o ypes/ lad/) was used o
pe o m a GO e m en ichmen analysis on he same UniP o accessions as lis ed
o he GO Slim. The backg ound anno a ion se was ob ained om he
goa_human.ga file (da ed 21 No embe 2017, downloaded om p:// p.ebi.ac.uk/
pub/da abases/GO/goa/HUMAN/) and he on ology da a was ob ained om he
go-basic.obo file p o ided in he VLAD ool (analysis un 28 No embe 2017).
The LD block a ound op SNPs associa ed wi h cIMT and ca o id plaque was
cons uc ed using LD in o ma ion om he 1000 Genomes panel, as p e iously
ou lined in Finan e al.16. B iefly, he bounda ies o he LD egion we e defined as
he posi ions o he a ian s u hes ups eam and downs eam o a GWAS SNP
wi h an 2 alue o ≥0.5 and wi hin a 1-Mbp flank on ei he side o he GWAS
a ian . Associa ed a ian s ha we e no p esen in he 1000 Genomes panel ha
we e no in LD wi h any o he a ian s we e gi en a nominal flank o 2.5 kbp on
ei he side o he associa ion. Gene anno a ions using Ensembl e sion 79 we e
hen o e lapped o he LD egion.
D uggable genes. We examined he d uggabili y s a us o he nea es coding
genes iden ified in ou GWAS analysis on cIMT and ca o id plaque, including
significan (no el and eplica ed) and sugges i e ones, as well as genes iden ified
h ough colocaliza ion analysis. The d uggable gene se was calcula ed using he
p e iously desc ibed c i e ia: no el a ge s o fi s -in-class d ugs licensed since
2005; he a ge s o d ugs cu en ly in la e phase clinical de elopmen ; p e-clinical
phase small molecules wi h p o ein binding measu emen s epo ed in he
ChEMBL da abase; and genes encoding sec e ed o plasma memb ane p o eins ha
a e a ge s o monoclonal an ibodies and o he bio- he apeu ics16.Wedefined
h ee ie s o d uggable gene se s based on hei d ug de elopmen . In Tie 1, 1427
genes we e a ge s o app o ed small molecules and bio he apeu ic d ugs and
clinical-phase d ug candida es. Tie 2 comp ised 682 genes encoding a ge s wi h
known bioac i e d ug-like small molecule binding pa ne s and hose wi h sig-
nifican sequence simila i y o app o ed d ug a ge s. Tie 3 con ained 2370 genes
encoding sec e ed o ex acellula p o eins, p o eins wi h mo e dis an simila i y o
app o ed d ug a ge s, and d uggable genes no included in Tie 1 o 2 such as
GPCRs, nuclea ho mone ecep o s, ion channels, kinases, and phosphodies e ases.
URLs. Fo GTEx, see h p://g expo al.o g/. Fo Coloc, see h ps://c an. -p ojec .
o g/web/packages/coloc/coloc.pd . Fo , Moloc, see h ps://gi hub.com/clagiamba/
ARTICLE NATURE COMMUNICATIONS | h ps://doi.o g/10.1038/s41467-018-07340-5
8NATURE COMMUNICATIONS | (2018) 9:5141 | h ps://doi.o g/10.1038/s41467-018-07340-5 | www.na u e.com/na u ecommunica ions
moloc/blob/mas e /man/moloc-package.Rd. Fo CARDIoGRAMPlusC4D, see
www.ca diog amplusc4d.o g/. Fo LD sco es, www.b oadins i u e.o g/~bulik/
eu _ldsco es/. Fo UniP o ID, www.unip o .o g/uploadlis s/. Fo QuickGO,
www.ebi.ac.uk/QuickGO. Fo VLAD ool, see h p://p o o.in o ma ics.jax.o g/
p o o ypes/ lad/.
Da a a ailabili y
All ele an summa y s a is ics da a ha suppo he findings o his s udy ha e
been deposi in he da abase o Geno ypes and Pheno ypes (dbGaP) unde he
CHARGE acquisi ion numbe (h ps://www.ncbi.nlm.nih.go /p ojec s/gap/cgi-bin/
s udy.cgi?s udy_id=phs000930. 6.p1; accession phs000930. 6.p1). GWAS da a o
mos US s udies a e al eady a ailable in dbGAP.
Recei ed: 19 Decembe 2017 Accep ed: 24 Sep embe 2018
Re e ences
1. Moza a ian, D. e al. Hea Disease and S oke S a is ics-2016 Upda e: a
epo om he Ame ican Hea Associa ion. Ci cula ion 133, e38–e360
(2016).
2. F ieden, T. R. & Be wick, D. M. The Million Hea s ini ia i e--p e en ing
hea a acks and s okes. N. Engl. J. Med. 365, e27 (2011).
3. O’Donnell, C. J. & Nabel, E. G. Genomics o ca dio ascula disease. N. Engl. J.
Med. 365, 2098–2109 (2011).
4. Bis, J. C. e al. Me a-analysis o genome-wide associa ion s udies om he
CHARGE conso ium iden ifies common a ian s associa ed wi h ca o id
in ima media hickness and plaque. Na . Gene . 43, 940–947 (2011).
5. Na a ajan, P. e al. Mul ie hnic Exome-Wide Associa ion S udy o Subclinical
A he oscle osis. Ci c. Ca dio asc. Gene . 9, 511–520 (2016).
6. Po , J. e al. Genome-wide me a-analysis iden ifies no el loci o plaque bu den
in ca o id a e y. A he oscle osis 259,32–40 (2017).
7. Ge ow, K. e al. Iden ifica ion o he BCAR1-CFDP1-TMEM170A locus as a
de e minan o ca o id in ima-media hickness and co ona y a e y disease
isk. Ci c. Ca dio asc. Gene . 5, 656–665 (2012).
8. Psa y, B. M. e al. Coho s o Hea and Aging Resea ch in Genomic
Epidemiology (CHARGE) Conso ium: design o p ospec i e me a-analyses o
genome-wide associa ion s udies om 5 coho s. Ci c. Ca dio asc. Gene . 2,
73–80 (2009).
9. Shah, S. e al. Causal ele ance o blood lipid ac ions in he de elopmen o
ca o id a he oscle osis: Mendelian andomiza ion analysis. Ci c. Ca dio asc.
Gene . 6,63–72 (2013).
10. Giamba olomei, C. e al. Bayesian es o colocalisa ion be ween pai s o
gene ic associa ion s udies using summa y s a is ics. PLoS Gene . 10, e1004383
(2014).
11. F anzen, O. e al. Ca diome abolic isk loci sha e downs eam cis- and ans-
gene egula ion ac oss issues and diseases. Science 353, 827–830 (2016).
12. P ei e , R. M., Mi chell, H. G. & Pee, D. On combining da a om genome-
wide associ ion s udies o disco e disease-associa ed SNPs. S a . Sci. 24,
547–560 (2009).
13. Pick ell, J. K. Join analysis o unc ional genomic da a and genome-wide
associa ion s udies o 18 human ai s. Am. J. Hum. Gene . 94, 559–573
(2014).
14. Giamba olomei, C. e al. A Bayesian amewo k o mul iple ai colo-
caliza ion om summa y associa ion s a is ics. Bioin o ma ics 34, 2538–2545
(2018).
15. Finucane, H. K. e al. Pa i ioning he i abili y by unc ional anno a ion using
genome-wide associa ion summa y s a is ics. Na . Gene . 47, 1228–1235
(2015).
16. Finan, C. e al. The d uggable genome and suppo o a ge iden ifica ion
and alida ion in d ug de elopmen . Sci. T ansl. Med. 9, eaag1166 (2017).
17. Zeggini, E. e al. Me a-analysis o genome-wide associa ion da a and la ge-
scale eplica ion iden ifies addi ional suscep ibili y loci o ype 2 diabe es.
Na . Gene . 40, 638–645 (2008).
18. Pol us, L. M. e al. Genome-wide associa ion s udy o gene by smoking
in e ac ions in co ona y a e y calcifica ion. PLoS ONE 8, e74642 (2013).
19. Remus, E. W. e al. The ole o lysyl oxidase amily membe s in he
s abiliza ion o abdominal ao ic aneu ysms. Am. J. Physiol. Hea Ci c.
Physiol. 303, H1067–H1075 (2012).
20. Peyo , M. L. e al. Ex acellula adenosine induces apop osis o human a e ial
smoo h muscle cells ia A(2b)-pu inocep o . Ci c. Res. 86,76–85 (2000).
21. E dmann, J. e al. Genome-wide associa ion s udy iden ifies a new locus o
co ona y a e y disease on ch omosome 10p11.23. Eu . Hea J. 32, 158–168
(2011).
22. Akashi, M., Higashi, T., Masuda, S., Komo i, T. & Fu use, M. A co ona y
a e y disease-associa ed gene p oduc , JCAD/KIAA1462, is a no el
componen o endo helial cell-cell junc ions. Biochem. Biophys. Res. Commun.
413, 224–229 (2011).
23. Ha a, T. e al. Ta ge ed dis up ion o JCAD (Junc ional P o ein Associa ed
Wi h Co ona y A e y Disease)/KIAA1462, a co ona y a e y disease-
associa ed gene p oduc , inhibi s angiogenic p ocesses in i o and in i o.
A e ioscle . Th omb. Vasc. Biol. 37, 1667–1673 (2017).
24. Ho mozdia i, F., Kos em, E., Kang, E. Y., Pasaniuc, B. & Eskin, E. Iden i ying
causal a ian s a loci wi h mul iple signals o associa ion. Gene ics 198,
497–508 (2014).
25. Howie, B. N., Donnelly, P. & Ma chini, J. A flexible and accu a e geno ype
impu a ion me hod o he nex gene a ion o genome-wide associa ion
s udies. PLoS Gene . 5, e1000529 (2009).
26. Howie, B., Fuchsbe ge , C., S ephens, M., Ma chini, J. & Abecasis, G. R. Fas
and accu a e geno ype impu a ion in genome-wide associa ion s udies
h ough p e-phasing. Na . Gene . 44, 955–959 (2012).
27. Winkle , T. W. e al. Quali y con ol and conduc o genome-wide associa ion
me a-analyses. Na . P o oc. 9, 1192–1212 (2014).
28. Yang, J., Lee, S. H., Godda d, M. E. & Vissche , P. M. GCTA: a ool o
genome-wide complex ai analysis. Am. J. Hum. Gene . 88,76–82 (2011).
29. Malik, R. e al. Mul iances y genome-wide associa ion s udy o
520,000 subjec s iden ifies 32 loci associa ed wi h s oke and s oke sub ypes.
Na . Gene . 50, 524–537 (2018).
30. Roadmap Epigenomics, Conso ium e al. In eg a i e analysis o 111 e e ence
human epigenomes. Na u e 518, 317–330 (2015).
31. Zhu, L. J. In eg a i e analysis o ChIP-chip and ChIP-seq da ase . Me hods
Mol. Biol. 1067, 105–124 (2013).
32. ENCODE P ojec Conso ium. An in eg a ed encyclopedia o DNA elemen s
in he human genome. Na u e 489,57–74 (2012).
33. Ande sson, R. e al. An a las o ac i e enhance s ac oss human cell ypes and
issues. Na u e 507, 455–461 (2014).
Acknowledgemen s
The wo k was suppo ed by he ollowing g an s: Na ional Ins i u e o Heal h g an s:
R21HL123677, R21-HL140385, DK104806-01A1, R01-MD012765-01A1 (NF), Na ional
Ins i u es o Heal h awa ds R01HG009120, R01HG006399, U01CA194393,
T32NS048004 (CG), he Ame ican Hea Associa ion G an #17POST33350042 (PV),
he B i ish Hea Founda ion (RG/13/5/30112) and he Na ional Ins i u e o Heal h
Resea ch Uni e si y College London Hospi als Biomedical Resea ch Cen e (RCL and
RPH), he B i ish Hea Founda ion FS/14/55/30806 (JCH), he Ge man Fede al Minis y
o Educa ion and Resea ch (BMBF) in he con ex o he e:Med p og am (e:A he -
oSysMed), he DFG as pa o he CRC 1123 (B3), and he FP7/2007-2103 Eu opean
Union p ojec CVgenes@ a ge (g an ag eemen numbe Heal h-F2-2013-601456). We
hank Li-Ming Gan o assis ance wi h he STARNET s udy and Jon Whi e o assis ance
wi h UCLEB analyses. Addi ional acknowledgemen s a e included in Supplemen a y
No e 2.
Au ho con ibu ions
N.F., C.G., J.C.B., M.K., C.D., M.S., S.K.M., U.S., W.P., A.B.Z., A.H., A.T., A.G.U., A.B.N.,
B.W.P., C.D.L., E.T., F.R., H.V., I.J.D., L.J.L., M.D., O.R., O.H.F., R.S., R.M., T.B.H., T.L.,
U.F., W.P., A.D., A.S., A.H., C.M.D., D.A.L., D.O.M.-K., D.W.B., H.S., J.F.W., J.G.W., J.I.
R., J.M.W., M.L., M.K.E., S.E.H., U.V., V.G., A.D.H., J.P.C., and C.J.O. con ibu ed o
s udy concep and design. C.D., M.S., S.K.M., U.S., W.P., A.I., A.C., A.B.Z., A.T., A.G.U.,
A.W., A.J.S., A.B., A.R., B.H., B.W.P., C.F., D.B., D.H.O., D.T., D.K., E.T., E.M., E.G., E.B.,
E.E.S., E.I., F.R., F.B., G.H., H.C., H.V., H.S.M., I.J.D., J.W.J., J.G., J.P., J.T., J.E., K.D.T., L.
K., L.L., L.J.L., L.H., M.D., M.S., M.K., M.K., M.A.N., O.M., O.R., P.H.W., P.G., P.A., R.R.,
R.B., R.H., R.S., R.M., R.W.M.,. S.G.W., S.M.L., S.T., S.K., S.R.H., S.R., T.B.H., T.L., T.G.,
T.S., U.F., V.P., W.R., W.P., X.Z., A.S., A.H., B.M.P., C.M.D., D.A.L., D.O.M.-K., D.W.B.,
H.S., J.F.W., J.G.W., J.I.R., J.C.H., J.M.W., J.D., J.H., M.K.E., S.E.H., U.V., V.G., A.D.H., J.
L.M.B., J.P.C., and C.J.O. con ibu ed o acquisi ion o geno yping o pheno ypic da a. N.
F., C.G., C.F., J.C.B., R.P.H., R.C.L., S.M.T., T.W.W., M.G., M.K., C.D., A.V.S., E.H., E.M.
L., I.M.N., L.L., M.S., M.S., N.P., O.F., P.K.J., R.N., R.E.M., S.-J.H., S.K.M., U.S., A.I., A.T.,
K.R., A.J.C., B.S., C.D.L., C.W., F.V., G.C., H.S., J.P., J.L., K.G., L.M.R., M.T., M.A.N., O.
M., P.R., P.A., Q.W., R.J.S., S.H., S.S., S.M.L., T.S., X.Z., X.Z., X.G., Y.S., and L.-.P.L.
con ibu ed o s a is ical analysis and in e p e a ion o he da a. N.F., C.G., P.S.V., J.C.B.,
M.K., S.K.M., A.D.H., and J.P.C. con ibu ed o d a ing o he manusc ip . All au ho s
con ibu ed o he c i ical e ision o he manusc ip .
Addi ional in o ma ion
Supplemen a y In o ma ion accompanies his pape a h ps://doi.o g/10.1038/s41467-
018-07340-5.
Compe ing in e es s: C.F. ecei ed a ee o speaking a a cou se by Sp inge Heal hca e/
Malesci. E.I. is a scien ific ad iso o P ecision Wellness, Cellink and Olink P o eomics
NATURE COMMUNICATIONS | h ps://doi.o g/10.1038/s41467-018-07340-5 ARTICLE
NATURE COMMUNICATIONS | (2018) 9:5141 | h ps://doi.o g/10.1038/s41467-018-07340-5 | www.na u e.com/na u ecommunica ions 9