RESEARCH ARTICLE
Gene-le el associa ion analysis o sys emic
scle osis: A compa ison o A ican-Ame icans
and Whi e popula ions
Olga Y. Go lo a
1
*, Ya ang Li
1
, I an Go lo
1
, Jun Ying
2
, Wei V. Chen
3
, She in Assassi
2
,
John D. Re eille
2
, F ank C. A ne
2
, Xiaodong Zhou
2
, La a Bossini-Cas illo
4
, Elena Lopez-
Isac
5
, Ma ialbe Acos a-He e a
5
, Pe e K. G ege sen
6
, Anne e T. Lee
6
, Vi ginia D. S een
7
,
Ba i J. Fessle
8
, Dinesh Khanna
9
, Elena Schiopu
9
, Richa d M. Sil e
10
, Je y A. Moli o
11
,
Daniel E. Fu s
12,13,14
, Suzanne Ka aja
12
, Robe W. Simms
15
, Robe A. La ya is
16
,
Pa icia Ca ei a
17
, Ca men Pila Simeon
18
, I an Cas ell i
19
, Emma Bel an
20
,
No be o O ego
21
, Ch is ophe I. Amos
1
, Ja ie Ma in
5
, Mau een D. Mayes
2
1Depa men o Biomedical Da a Science, Geisel School o Medicine, Da mou h College, Lebanon, NH,
Uni ed S a es o Ame ica, 2Depa men o In e nal Medicine, Di ision o Rheuma ology, Uni e si y o Texas
McGo e n Medical School, Hous on, TX, Uni ed S a es o Ame ica, 3Depa men o Bios a is ics, UT MD
Ande son Cance Cen e , Hous on, TX, Uni ed S a es o Ame ica, 4Wellcome T us Sange Ins i u e,
Hinx on, Uni ed Kingdom, 5Ins i u e o Pa asi ology and Biomedicine Lo
´pez-Ney a, IPBLN-CSIC, G anada,
Spain, 6Robe S. Boas Cen e o Genomics and Human Gene ics, Feins ein Ins i u e o Medical Resea ch,
Manhasse , NY, Uni ed S a es o Ame ica, 7Di ision o Rheuma ology, Geo ge own Uni e si y Medical
Cen e , Washing on, D.C., Uni ed S a es o Ame ica, 8Di ision o Rheuma ology, Uni e si y o Alabama—
Bi mingham, Bi mingham, AL, Uni ed S a es o Ame ica, 9Di ision o Rheuma ology, Uni e si y o Michigan,
Ann A bo , MI, Uni ed S a es o Ame ica, 10 Di ision o Rheuma ology, Medical Uni e si y o Sou h Ca olina,
Cha les on, SC, Uni ed S a es o Ame ica, 11 Di ision o Rheuma ic and Au oimmune Diseases, Uni e si y
o Minneso a, Minneapolis, MN, Uni ed S a es o Ame ica, 12 Di ision o Rheuma ology, Uni e si y o
Cali o nia—Los Angeles, Los Angeles, CA, Uni ed S a es o Ame ica, 13 Uni e si y o Washing on, Sea le,
WA, Uni ed S a es o Ame ica, 14 Uni e si y o Flo ence, Flo ence, I aly, 15 Di ision o Rheuma ology,
Bos on Uni e si y, Bos on, MA, Uni ed S a es o Ame ica, 16 Uni e si y o Pi sbu gh, Pi sbu gh, PA, Uni ed
S a es o Ame ica, 17 Depa men o Rheuma ology, Hospi al Uni e si a io, Mad id, Spain, 18 Depa men o
In e nal Medicine, Valle de Heb o
´n Hospi al, Ba celona, Spain, 19 Hospi al de la San a C eu i San Pau,
Ba celona, Spain, 20 Hospi al Uni e si a io y Poli e
´cnico La Fe, Valencia, Spain, 21 Uni e si y Hospi al San
Cecilio, G anada, Spain
*Olga.Y.Go lo a@da mou h.edu
Abs ac
Gene-le el analysis o ImmunoChip o genome-wide associa ion s udies (GWAS) da a has
no been p e iously epo ed o sys emic scle osis (SSc, scle ode ma). The objec i e o
his s udy was o analyze gene ic suscep ibili y loci in SSc a he gene le el and o de e mine
i he de ec ed associa ions we e sha ed in A ican-Ame ican and Whi e popula ions, using
da a om ImmunoChip and GWAS geno yping s udies. The Whi e sample included 1833
cases and 3466 con ols (956 cases and 2741 con ols om he US and 877 cases and 725
con ols om Spain) and he A ican Ame ican sample, 291 cases and 260 con ols. In bo h
Whi es and A ican Ame icans, we pe o med a gene-le el analysis ha in eg a es associa-
ion s a is ics in a gene possibly ha bo ing mul iple SNPs wi h weak e ec on disease isk,
using Ve sa ile Gene-based Associa ion S udy (VEGAS) so wa e. The SNP-le el analysis
was pe o med using PLINK .1.07. We iden i ied 4 no el candida e genes (STAT1,
FCGR2C,NIPSNAP3B, and SCT) signi ican ly associa ed and 4 genes (SERBP1,PINX1,
TMEM175 and EXOC2) sugges i ely associa ed wi h SSc in he gene le el analysis in
PLOS ONE | h ps://doi.o g/10.1371/jou nal.pone.0189498 Janua y 2, 2018 1 / 12
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Ci a ion: Go lo a OY, Li Y, Go lo I, Ying J, Chen
WV, Assassi S, e al. (2018) Gene-le el associa ion
analysis o sys emic scle osis: A compa ison o
A ican-Ame icans and Whi e popula ions. PLoS
ONE 13(1): e0189498. h ps://doi.o g/10.1371/
jou nal.pone.0189498
Edi o : Masa aka Kuwana, Keio Uni e si y, JAPAN
Recei ed: Augus 31, 2017
Accep ed: No embe 27, 2017
Published: Janua y 2, 2018
Copy igh : ©2018 Go lo a e al. This is an open
access a icle dis ibu ed unde he e ms o he
C ea i e Commons A ibu ion License, which
pe mi s un es ic ed use, dis ibu ion, and
ep oduc ion in any medium, p o ided he o iginal
au ho and sou ce a e c edi ed.
Da a A ailabili y S a emen : Gene ic da a is
a ailable om dbGaP eposi o y (h ps://www.ncbi.
nlm.nih.go /p ojec s/gap/cgi-bin/s udy.cgi?s udy_
id=phs000357. 1.p1). Addi ional da a con ains
po en ially iden i ying pa icipan in o ma ion and is
es ic ed by he E hics Commi ee o Ins i u o de
Pa asi ologı
´a y Biomedicina. In e es ed, quali ied
esea che s may eques he da a by con ac ing
Comi e de E ica del CSIC a [email p o ec ed].
All o he ele an da a a e wi hin he pape and i s
Suppo ing In o ma ion iles.
Whi e pa ien s. As an explo a o y analysis we compa ed he esul s on Whi es wi h hose
om A ican Ame icans. O p e iously es ablished suscep ibili y genes iden i ied in Whi es,
only TNFAIP3 was signi ican a he nominal le el (p = 6.13x10
-3
) in A ican Ame icans in
he gene-le el analysis o he ImmunoChip da a. Among he op sugges i e no el genes
iden i ied in Whi es based on he ImmunoChip da a, FCGR2C and PINX1 we e only nomi-
nally signi ican in A ican Ame icans (p = 0.016 and p = 0.028, espec i ely), while among
he op no el genes iden i ied in he gene-le el analysis in A ican Ame icans, UNC5C (p =
5.57x10
-4
) and CLEC16A (p = 0.0463) we e also nominally signi ican in Whi es. We also
p esen he gene-le el analysis o SSc clinical and au oan ibody pheno ypes among Whi es.
Ou indings need o be alida ed by independen s udies, pa icula ly due o he limi ed
sample size o A ican Ame icans.
In oduc ion
Sys emic scle osis (SSc, scle ode ma) [MIM 181750] is an au oimmune disease cha ac e ized
by h ee key ea u es: (1) ib osis o skin and in e nal o gans, (2) a asculopa hy, and (3) au o-
an ibody p oduc ion. I is a mul io gan sys em disease wi h conside able pheno ypic he e oge-
nei y, esul ing in a b oad spec um o disease se e i y. Se e al genome-wide, ImmunoChip,
and ollow-up associa ion s udies we e conduc ed o iden i y SNPs associa ed wi h SSc isk [1–
8]. All published s udies implemen ed SNP-le el analysis meaning ha each SNP was analyzed
sepa a ely and hose wi h genome wide le el o s a is ical signi icance we e deemed isk associ-
a ed. SNP-le el analysis is e ec i e o iden i ica ion o SNPs wi h s ong indi idual e ec s,
howe e , i is unde powe ed o de ec genes ca ying mul iple SNPs in he same gene o small
o medium e ec size [9–11]. In he la e case, gene-le el analysis can be bene icial because i
will de ec genes wi h mul iple small e ec size SNPs as signi ican e en i hese genes do no
ha bo any indi idual SNPs signi ican a he genome-wide le el. Howe e , a gene-le el analy-
sis has ne e been applied o SSc.
In his s udy we pe o med a gene-le el analysis ocusing on he da a gene a ed by he
ImmunoChip pla o m. We compa ed esul s om he gene-le el analysis wi h he esul s gen-
e a ed by adi ional SNP-le el analysis. We also pe o med a gene-le el analysis o Immuno-
Chip and genome-wide associa ion s udy (GWAS) da a on o A ican-Ame ican SSc pa ien s.
Al hough based on a ela i ely small g oup o pa ien s, his s udy ep esen s he i s epo o
gene ic analysis o A ican Ame icans wi h SSc. The esul s o he gene-le el analysis o SSc
clinical pheno ypes (limi ed SSc (lcSSc) and di use SSc (dcSSc)) as well as au oan ibody sub-
se s (an i-cen ome e au oan ibodies (ACA) and an i-DNA opoisome ase I (ATA) au oan i-
bodies) among Whi es a e also p esen ed.
Ma e ials and me hods
The s udy has been app o ed by he Ins i u ional Re iew Boa ds o he pa icipa ing US ins i-
u ions, namely Bos on Uni e si y; Geo ge own Uni e si y; Medical Uni e si y o Sou h Ca o-
lina; Uni e si y o Alabama, Bi mingham; Uni e si y o Cali o nia Los Angeles; Uni e si y o
Michigan; Uni e si y o Minneso a; Uni e si y o Washing on; Uni e si y o Pi sbu gh; Uni-
e si y o Texas Heal h Science Cen e , Hous on, Uni e si y o Texas MD Ande son Cance
Cen e , Hous on, Geisel School o Medicine, Da mou h College; and e hics commi ees o he
pa icipa ing o eign ins i u ions, namely Ins i u e o Pa asi ology and Biomedicine Lo
´pez-
Gene-le el associa ion analysis o sys emic scle osis
PLOS ONE | h ps://doi.o g/10.1371/jou nal.pone.0189498 Janua y 2, 2018 2 / 12
Funding: Funding was p o ided o MDM by he
Na ional Ins i u es o Heal h (NIH) he Na ional
Ins i u e o A h i is, Musculoskele al and Skin
Diseases (NIAMS h ps://www.niams.nih.go /)
Cen e s o Resea ch T ansla ion (CORT) P50-
AR054144, NIH g an N01-AR-02251 and R01-AR-
055258, and he Depa men o De ense (DD)
Cong essionally Di ec ed Medical Resea ch
P og am (h p://cdm p.a my.mil/) W81XWH-07-1-
011 and WX81XWH-13-1-0452 o he collec ion,
analysis and in e p e a ion o he da a. The unde s
had no ole in s udy design, da a collec ion and
analysis, decision o publish, o p epa a ion o he
manusc ip .
Compe ing in e es s: The au ho s ha e decla ed
ha no compe ing in e es s exis .
Ney a, IPBLN-CSIC, G anada, Spain; Uni e si y o Flo ence, Flo ence, I aly; Hospi al Uni e -
si a io, Mad id, Spain; Valle de Heb o
´n Hospi al, Ba celona, Spain; Hospi al de la San a C eu i
San Pau, Ba celona, Spain; Hospi al Uni e si a io y Poli e
´cnico La Fe, Valencia, Spain; Uni-
e si y Hospi al San Cecilio, G anada, Spain. All clinical in es iga ion has been conduc ed
acco ding o he p inciples exp essed in he Decla a ion o Helsinki. W i en in o med consen
has been ob ained om he pa icipan s.
Race o he s udy pa icipan s was sel - epo ed, and we used p incipal componen analyses
o emo e ace ou lie s as desc ibed below. De ails on he Whi e popula ion, geno yping and
quali y con ol can be ound in ou p e iously published manusc ip [7]. The Whi e sample
included 956 cases and 2741 con ols om he US and 877 cases and 725 con ols om Spain,
a e exclusion o indi iduals based on quali y con ol (QC) (low call a es, non-Eu opean
ances y, o ela edness). The e we e 1087 (59%) Whi e pa ien s wi h lcSSc, 574 (31%) wi h
dcSSc, 671 (37%) ACA-posi i e (ACA+), and 347 (19%) ATA+ pa ien s (no all pa ien s could
be classi ied in o wo dis inc pheno ypes o had ei he ACA o ATA an ibodies).
The A ican Ame ican sample, a e quali y con ol measu es, included 291 cases (56 men
and 235 women) and 260 con ols (72 men and 188 women). In line wi h [12], he dis ibu ion
o clinical pheno ypes and au oan ibody subse s in he A ican Ame ican pa ien popula ion
was ma kedly di e en om ha in Whi es. The e we e 82 (28%) pa ien s wi h lcSSc, 201
(69%) wi h dcSSc, 21 (7%) ACA+, and 69 (24%) ATA+ among he A ican Ame ican cases (as
wi h Whi es, no all pa ien s could be classi ied in o wo dis inc pheno ypes o had ei he
ACA o ATA an ibodies wo pa ien s we e no es ed o ACA and ou o ATA).
ImmunoChip analysis: Geno yping was done by Illumina In inium single-nucleo ide poly-
mo phism (SNP) mic oa ay–ImmunoChip. Geno ype calling was done using he Illumina
iScan Sys em and he Geno yping Module ( .1.8.4) o he GenomeS udio Da a Analysis so -
wa e. We applied he ollowing c i e ia o QC: (1) indi iduals wi h call a e <90% we e
excluded, (2) ma ke s wi h call a es 90% we e excluded, and (3) ma ke s wi h allele dis i-
bu ions de ia ing om Ha dy-Weinbe g equilib ium (HWE) in con ols (p <1×10
−5
) we e
also excluded. A o al 126,270 ma ke s (101,692 o hem wi h a MAF >0.1%) passed QC and
we e included in he analysis.
The same ImmunoChip pla o m was used o Whi e and A ican Ame ican popula ions.
The geno yping a e was 0.988 in he A ican Ame ican sample while he geno yping a e was
0.998 among Whi es.
GWAS analysis: We also pe o med a genome-wide geno yping o bo h A ican Ame icans
and Whi es. A ican Ame icans we e geno yped on he Illumina Omni2.5 BeadChip ha ea-
u es ~2.5 million ma ke s cap u ing a ian s down o MAF 2.5% and co e s, in pa icula ,
A ican gene ic di e si y. The same exac indi iduals ha we e success ully geno yped on
ImmunoChip we e success ully geno yped on his pla o m. The geno yping a e a e QC was
0.993 in A ican Ame ican popula ion. The quali y con ol o A ican Ame icans also
included p incipal componen analysis as implemen ed in SNP & Va ia ion Sui e .7 (Golden
Helix). The i s h ee p incipal componen s we e de i ed o each indi idual om he A ican
Ame ican sample along wi h HapMap Phase 2 samples as e e ence popula ions. Indi iduals
de ia ing o mo e han 6 SDs om he A ican ances y clus e cen oid we e disca ded om
u he analysis. We also excluded indi iduals de ia ing mo e han 4 s anda d de ia ions om
he clus e cen oid. Finally, we excluded duplica e and closely ela ed samples (PIHAT 0.5).
The genome wide geno yping o he Whi e popula ions has been desc ibed p e iously in
Rads ake e al (2010) [3]. In b ie , Hap550K-BeadChip was used o US Whi es and Illumina
HumanCNV370K BeadChip in Spanish Whi es.
Fo he SNP-le el analysis he associa ion s a is ics was compu ed ia logis ic eg ession
including sex as a co a ia e o each da ase . Fo he Whi e samples, me a-analysis combining
Gene-le el associa ion analysis o sys emic scle osis
PLOS ONE | h ps://doi.o g/10.1371/jou nal.pone.0189498 Janua y 2, 2018 3 / 12
odds a ios (OR) and s anda d e o s (SE) o indi idual da ase s (US and Spanish, so ha he
con ols in each se we e om he same coun y as cases) was pe o med by means o he
in e se- a iance me hod unde he assump ion o a ixed e ec as implemen ed in PLINK
.1.07. [13].
Fo he gene-le el analysis we used Ve sa ile Gene-based Associa ion S udy (VEGAS) [14].
We used VEGAS because i ou pe o ms simila me hods by sensi i i y and speci ici y om
simula ion s udies [15]. VEGAS can be applied o he da a gene a ed by any GWAS designs,
including amily-based GWAS, me a-analyses o GWAS and DNA-pooling-based GWAS. The
es uses in o ma ion om he comple e se o ma ke s wi hin a gene. To accoun o linkage
disequilib ium be ween ma ke s VEGAS uses simula ions om he mul i a ia e no mal dis i-
bu ion. VEGAS assigns SNPs o au osomal genes acco ding o posi ions on he UCSC Genome
B owse hg18 assembly. In o de o cap u e egula o y egions and SNPs in LD, he gene
bounda ies a e de ined as ±50 kb o 5’ and 3’ UTRs. VEGAS assigned SNPs geno yped by
ImmunoChip o 11,501 genes. Assuming independence o he gene le el es s, he h eshold
o s a is ical signi icance in he analysis o ImmunoChip in Whi es was se o be 4.35x10
-6
,
and a 2.8x10
-6
o he analysis o GWAS da a [14]. Howe e , since his h eshold is likely o be
conse a i e gi en he o e lap be ween genes, we epo indings wi h p- alues <10
−5
. Also,
since he sample size o A ican Ame icans was limi ed, o his popula ion we p esen ind-
ings wi h he p- alue below 10
−3
, acknowledging ha his is a s udy limi a ion and ha he
analysis is explo a o y and is in need o u he alida ion. We use he e m “nominal signi i-
cance” o deno e p- alues in he ange o 0.05 o 10
−3
, in e p e ing hem as weak e idence o
associa ion. We excluded HLA egion om he analysis because i is uni e sally signi ican .
We used Pa hwayS udio [16] o build a pa hway o known and no el SSc isk-associa ed
genes. The Pa hwayS udio uses ex mining o iden i y epo ed in e ac ions be ween genes
and build a ne wo k based on he known in e ac ions.
Resul s
ImmunoChip analysis
Table 1 shows esul s om he gene-le el and SNP-le el analyses o 19 non-HLA genes p e i-
ously shown o be associa ed wi h SSc in Whi es [1–8,17,18]. Ou o 19 known SSc genes, all
excep o SCHIP1,IRF8, and CD247we e nominally signi ican in he gene-le el analysis in
Whi es. IRF5,STAT4, and TNPO3we e signi ican in bo h he SNP- and gene-le el analyses
among Whi es.
In he gene-le el analysis o clinical pheno ypes (S1 and S2 Tables) and an ibody subse s
(S3 and S4 Tables), STAT4 was signi ican o lcSSc and ACA+ pa ien s and TNPO3 in ATA
+ pa ien s. O he 19 genes examined, only TNFAIP3 was nominally signi ican in A ican
Ame icans in he gene-le el analysis.
Table 2 shows non-HLA genes wi h he p- alues below 10
−5
in he gene-le el analysis o
Whi es (excluding hose al eady es ablished in Whi es), wi h he addi ion o p- alues o hese
genes in A ican Ame icans. The genes wi h he p- alues below 4.35x10
-6
a e shown in bold.
One gene ou o ou signi ican a his le el in Whi es, namely FCGR2C, was also nominally
signi ican in A ican Ame icans. PINX1, which was only bo de line signi ican in Whi es, also
showed a nominal signi icance in A ican Ame icans. Addi ionally, nominally signi ican
SNPs whe e obse ed in STAT1 and SCT genes and in he bo de line signi ican EXOC2 in he
analysis o A ican Ame icans (Table 2), e en hough hese genes did no each signi icance in
he gene-le el analysis in ha popula ion.
The op genes iden i ied o clinical pheno ypes and au oan ibody subse s a e shown in
S1–S4 Tables, in he le po ion o he gene-le el analyses based on ImmunoChip. FCGR2C,
Gene-le el associa ion analysis o sys emic scle osis
PLOS ONE | h ps://doi.o g/10.1371/jou nal.pone.0189498 Janua y 2, 2018 4 / 12
STAT1, and FCGR3Bwe e signi ican in lcSSc, al hough FCGR2Cand FCGR3Bsha ed he
mos signi ican SNP s455499 and he gene-le el p- alue o FCGR2Cwas mo e signi ican
(S1 Table). In dcSSc, h ee genes (IL34,ABBA-1, and VAC14) we e iden i ied as signi ican ,
al hough hey sha ed he mos signi ican SNP s11640251 and IL34 had he bes p- alue in he
gene-le el analysis (S2 Table). In ACA+ pa ien s, in addi ion o STAT4, six genes (FCGR2C,
Table 1. Resul s o he gene-le el analyses o known genes associa ed wi h isk o SSc in Whi es and A ican Ame icans, based on ImmunoChip.
Whi es A ican Ame icans
Gene Ch N
SNPs
P- alue Top SNP Top SNP P- alue N
SNPs
P- alue Top SNP Top SNP P- alue
STAT4 2 113 2.00E-09 s10174238 4.65E-13 108 0.105 s1551443 0.0106
TNPO3 7 102 3.00E-08 s17340351  2.39E-10 105 0.314 s1495461
§
0.0574
IRF5 7 49 2.00E-06 s12534421 7.42E-10 53 0.211 s1495461
‡
0.0574
TNIP1 5 124 4.90E-05 s10463312 4.87E-06 156 0.189 s2277940
§
0.00742
IL12RB2 1 69 5.40E-05 s2201584 1.29E-06 84 0.921 s11209045  0.0418
BLK 8 176 1.43E-04 s2409781 2.99E-08 172 0.552 s11250139  0.0244
ITGAM 16 53 1.96E-04 s1143683
†
5.70E-05 67 0.337 s4561481  0.00771
IL12RB1 19 42 0.001144 s2305743 7.23E-05 48 0.215 s426132 0.0733
TNFAIP3 6 45 0.00145 s892999  0.0103 69 0.00613 s10223636 0.00203
IL12A 3 110 0.00248 s4679867 8.79E-05 141 0.741 s4679867 0.0508
TYK2 19 60 0.00363 s2304256
†
1.61E-04 63 0.758 s8108709 0.1514
ATG5 6 54 0.0076 s11758079  0.00276 62 0.272 s4946731  0.0083
CSK 15 8 0.0129 s6495122
§
1.96E-05 9 0.787 s8033381 0.299
JAZF1 7 96 0.0325 s6971086 5.64E-04 112 0.844 s3823946 0.0122
PXK 3 180 0.0397 s7626140 0.00699 187 0.755 s9870786 0.0177
DNASE1L3 3 84 0.0464 s4681786  3.09E-04 93 0.597 s9843169  0.0177
CD247 1 5 0.0821 s2056626 0.00541 8 0.467 s10918695 0.211
IRF8 16 24 0.0986 s10863202  0.0165 28 0.518 s11642456  0.129
SCHIP1 3 81 0.174 s1675497 0.00108 77 0.449 s17753641 0.0543
In onic SNP
in e genic SNP
†
coding SNP
§
3’ downs eam SNP
‡
5’ ups eam SNP
h ps://doi.o g/10.1371/jou nal.pone.0189498. 001
Table 2. Compa ison o no el candida e genes (a p<4.35x10
-6
indica ed in bold) and sugges i e genes (4.35x10
-6
<p<10
−5
) de ec ed in Whi es o he co esponding
s a is ics in A ican Ame icans in he gene le el analysis, based on ImmunoChip.
Whi es A ican Ame icans
Gene Ch N
SNPs
P- alue Top SNP Top SNP p- alue N
SNPs
P- alue Top SNP Top SNP p- alue
STAT1 2 60 <1.00E-06 s11893432 4.01E-11 58 0.581 s16833197 0.0195
FCGR2C 1 26 3.00E-06 s4554699  2.72E-08 24 0.0158 s17411858  0.00497
NIPSNAP3B 9 3 4.00E-06 s3780540 7.79E-04 4 0.551 s3780540 0.31
SCT 11 34 4.00E-06 s4963128 6.72E-05 52 0.691 s10902178 0.0456
SERBP1 1 48 6.00E-06 s3790569 5.61E-05 61 0.855 s11807749  0.0682
PINX1 8 200 8.00E-06 s17152571 1.46E-05 210 0.0283 s17152345 6.61E-04
EXOC2 6 15 8.00E-06 s908026  2.79E-05 21 0.382 s7761186 0.0396
In onic SNP
in e genic SNP
h ps://doi.o g/10.1371/jou nal.pone.0189498. 002
Gene-le el associa ion analysis o sys emic scle osis
PLOS ONE | h ps://doi.o g/10.1371/jou nal.pone.0189498 Janua y 2, 2018 5 / 12

SRCAP,PHKG2,LOC90835,RNF40, and FCGR3B) eached signi icance in he ImmunoChip-
based gene-le el analysis, bu he middle ou genes sha ed he mos signi ican SNP s71889
27 and SRCAP showed he bes gene-le el p- alue (S3 Table), and FCGR2Cand FCGR3Balso
sha ed he mos signi ican SNP s455499. O hese wo genes, FCGR2Cshowed a mo e signi i-
can gene-le el p- alue like in he case o he lcSSc pheno ype. Among ATA+ pa ien s, in addi-
ion o TNPO3,C16o 68,P2RX1,C3o 25,IFT122, and MBD4 eached signi icance in he
gene-le el analysis, wi h he las h ee genes sha ing he same mos signi ican SNP s2307293
(S4 Table).
In he e e se app oach we selec ed op non-HLA genes mos signi ican in A ican Ame i-
cans (p<10
−3
) based on he gene le el analysis (Table 3; 13 genes bu only 9 independen
egions, due o he gene o e lap).
O hese genes, only UNC5C (p = 5.57x10
-4
) and CLEC16A(p = 0.0463) we e nominally sig-
ni ican in he gene-le el analysis in Whi es, and bo h hese genes and PHF19 ha bo ed a nom-
inally signi ican SNP in Whi es (Table 3).
Among he 63 op genes (27 independen egions) selec ed based on SNP le el analysis in
A ican Ame icans (S5 Table; genes wi h bes SNP p- alue<10
−3
in A ican Ame icans), 16
genes (12 independen egions) we e also nominally signi ican in Whi es a he gene le el and
42 genes ha bo ed a leas nominally signi ican SNPs in Whi es (25 di e en SNPs due o
assignmen o some SNPs o se e al genes a once). Since unde he null hypo hesis he expec ed
numbe o nominally signi ican SNPs in Whi es is ~1 (27x0.05 = 1.3), he esul s sugges some
o e lap in gene ic suscep ibili y loci o SSc be ween Whi es and A ican Ame icans.
GWAS da a analysis
We pe o med simila analyses also based on he GWAS geno yping which was, howe e , pe -
o med on di e en pla o ms in Whi es and A ican Ame icans, and his made he esul s less
Table 3. Genes mos signi ican (a p<10
−3
) in A ican Ame icans in he gene-le el analysis and he esul s o hese genes in Whi es, based on ImmunoChip.
A ican Ame icans Whi es
Gene Ch N
SNPs
P- alue Top SNP Top SNP p- alue N
SNPs
P- alue Top SNP Top SNP p- alue
SCN4B 11 4 1.20E-05 s868344
ǁ
4.04E-03 4 0.544 s671111
§
0.291
SCN2B 11 4 1.50E-05 s868344
§
4.04E-03 3 0.687 s671111
§
0.291
MRPL28 16 1 6.20E-05 s743961
‡
6.72E-05 1 0.0863 s3848368  0.0856
TMEM8 16 1 6.20E-05 s743961 6.72E-05 1 0.105 s3848368  0.0856
NME4 16 1 6.90E-05 s743961
‡
6.72E-05 1 0.0869 s3848368  0.0856
DECR2 16 1 7.50E-05 s743961
‡
6.72E-05 1 0.0836 s3848368  0.0856
UNC5C 4 8 1.22E-04 s7697199 6.20E-05 6 5.57E-04 s17381177 7.84E-04
RAB11FIP3 16 2 2.96E-04 s743961
‡
6.72E-05 1 0.358 s3785301 0.338
PLCG2 16 5 4.97E-04 s4325546 3.15E-03 8 0.601 s3936112 0.0874
CLEC16A 16 329 5.29E-04 s1646066  5.39E-04 355 0.0463 s16957854 0.00169
ABCB4 7 24 5.77E-04 s17149601 2.16E-04 17 0.271 s17149512
ǁ
0.083
TCERG1 5 1 7.61E-04 s10056189  7.31E-04 1 0.135 s10056189  0.171
PHF19 9 74 8.09E-04 s1008382 1.32E-04 61 0.721 s388040 0.0426
ǁ
3’UTR SNP
§
3’ downs eam
‡
5’ ups eam
in onic SNP
in e genic SNP
h ps://doi.o g/10.1371/jou nal.pone.0189498. 003
Gene-le el associa ion analysis o sys emic scle osis
PLOS ONE | h ps://doi.o g/10.1371/jou nal.pone.0189498 Janua y 2, 2018 6 / 12
compa able. The esul s a e p esen ed in S6–S11 Tables. In b ie , among he genes p e iously
iden i ied in Whi es, in addi ion o TNFAIP3,ATG5 showed a nominal signi icance in he
gene-le el analysis in A ican Ame icans, po en ially due o a dense co e age o his gene (81
s 62 SNPs) on he 2.5 M Omni pla o m han on ImmunoChip (S6 Table).
Beyond he genes p e iously es ablished as associa ed wi h SSc in Whi es, only one gene,
TMEM175, showed a bo de line signi ican associa ion in he gene-le el analysis in Whi es
(p = 3.0x10
-6
) in GWAS. I s mos signi ican SNP s2290405 was sha ed wi h wo o he genes
(SLC26A1 and DGKQ; S7 Table). The analysis o he co esponding genes in he GWAS da a
in A ican Ame icans did no de ec gene-le el signi icance o hese genes bu TMEM175 ha -
bo ed a nominally signi ican SNP s11946340 unlike he o he wo neighbo ing genes.
In he analysis o clinical pheno ypes and au oan ibody subse s, ou o al eady es ablished
SSc suscep ibili y genes, IRF5,TNPO3, and IRF8 we e signi ican in lcSSc pa ien s. The e we e
also se en newly iden i ied genes, o which ou (DGKQ,IDUA,TMEM175, and SLC26A1)
sha ed he same mos signi ican SNP s11724804; o hese, DGKQ showed he bes gene-le el
p- alue. IRF4,CCDC104,and TLR10we e also signi ican . No ably all hese genes we e a leas
nominally signi ican in he ImmunoChip-based gene-le el analysis, excep o CCDC104
which is no on ImmunoChip (S1 Table). In dcSSc, in addi ion o IRF5 and TNPO3,CPSF4
and ATP5J2 showed signi ican p- alues in he gene-le el analysis; hey sha ed he same mos
signi ican SNP s10235235, and CPSF4 was mo e signi ican in he gene-le el analysis. Excep
o IRF5 and TNPO3, no gene eached s a is ical signi icance in he ACA+ subse in he
GWAS-based gene-le el analysis. In he ATA+ subse , TLR10 and TLR1, sha ing he same
mos signi ican SNP s10024216, we e signi ican (bo h eached only nominal signi icance in
he ImmunoChip-based gene-le el analysis) (S4 Table).
In he e e se analysis, conside ing he op genes iden i ied in he gene-le el analysis in
A ican Ame icans (12 non-HLA genes bu 11 egions due o he gene o e lap, S8 Table),
none o he co esponding genes was e en nominally signi ican in Whi es bu 6 genes ha -
bo ed nominally signi ican SNPs. Among he genes iden i ied in he GWAS SNP-le el analy-
sis on A ican Ame icans as ha bo ing mos signi ican SNPs (p<10
−3
; 488 such genes bu
only 308 independen egions because o he gene o e lap), 311 genes (255 independen
egions) also ha bo ed a leas nominally signi ican SNPs in Whi es. Eigh genes (SLC2A13,
NRG3,SLC10A7,MKL1,DZIP1L,C8o 58,KIAA1967, and HDAC1; se en independen
egions, C8o 58and KIAA1967 ep esen ing he same egion) we e nominally signi ican in
bo h Whi es and A ican Ame icans in he gene-le el analysis. The esul s a e p esen ed in S9
Table (a). Fi e SNPs— s2994241 in C10o 27/ADAMTS14, s6025407 in BMP7, s6796265 in
OSBPL10, s7734699 in MRPS27, and s6075784 in STK35 –we e nominally signi ican in bo h
Whi es and A ican Ame icans. These i e SNPs a e ma ked in g een in S9 Table (a), and hei
isk e ec s a e shown in S9 Table (b).
We also ca alogued genes iden i ied in A ican Ame icans ei he in GWAS o ImmunoChip
gene-le el analysis (S10 Table), o by he op SNP p- alue (wi h p<10
−3
) (S11 Table). In he
ew cases whe e he same SNP was op in bo h GWAS and ImmunoChip analyses, a sligh p-
alue a ia ion is explained by he QC p ocedu es ha elimina ed di e en numbe o indi id-
uals om he ImmunoChip e sus GWAS analysis.
Genes o SSc and o he au oimmune diseases a e en iched by he immune esponse genes
[19–21]. One can expec , he e o e, ha SSc genes will be o en in ol ed in di ec in e ac ions.
We used Pa hwayS udio o build an in e ac ion ne wo k o known as well as 6 no el candida e
genes (bo h signi ican and sugges i e) (Table 3) iden i ied by he gene-le el analysis. Such ne -
wo ks may be use ul by p o iding guidance o explo e biological mechanisms unde lying SSc
isk. We ound ha wo sugges i e candida es, namely EXOC2 and PINX1, in e ac wi h
Gene-le el associa ion analysis o sys emic scle osis
PLOS ONE | h ps://doi.o g/10.1371/jou nal.pone.0189498 Janua y 2, 2018 7 / 12
known genes associa ed wi h isk o SSc. The EXOC2 p o ein has been shown o bind LST1
[22], and PINX1 and STAT1 show p o ein/p o ein in e ac ion [23] (S1 Fig).
Discussion
We iden i ied 4 no el candida e genes (STAT1,FCGR2C,NIPSNAP3B, and SCT) signi ican ly
associa ed and 4 genes (SERBP1,PINX1,TMEM175 and EXOC2) sugges i ely associa ed wi h
SSc in a gene le el analysis in Whi es. Some o hese genes ha e been shown o be di ec ly
in ol ed in immune esponse. Fo example, FCGR2Cencodes a membe o low-a ini y immu-
noglobulin gamma Fc ecep o s. FCGR2C is ound on he su ace o many immune esponse
cells. The gene encodes a ansmemb ane glycop o ein in ol ed in phagocy osis and clea ing
o immune complexes. A sugges i e no el gene, SERBP1, encodes a B-cell an igen, shown o
p edic an i- umo immune esponse [24]. Ano he sugges i e gene, EXOC2, is associa ed wi h
inna e immuni y and has been shown o play a ole in suscep ibili y o C ohn’s disease [25].
We no e ha he sugges i e signal a EXOC2 o e laps wi h he signal a IRF4 p e iously
desc ibed in he c oss-disease me a-GWAS o SSc and heuma oid a h i is [26], which poin s
a he impo ance o his egion in au oimmune condi ions. This gene ha bo ed a SNP s908
026 wi h a ela i ely s ong s a is ical e idence o isk associa ion (P = 2.8x10
-5
). Risk-associ-
a ed SNPs we e obse ed in o he gene-le el candida es as well. Fo example, s11893432
(STAT1 gene; p = 4.01x10
-11
) and s4554699 (FCGR2Cgene; p = 2.7x10
-8
) we e signi ican a
he GWAS le el. The mos signi ican SNPs in o he no el candida es we e: s2290405 in
TMEM175(p = 1.82x10
-6
), s17152571 in PINX1 gene (P = 1.5x10
-5
), s3790569 in SERBP1
gene (P = 5.6x10
-5
), s4963128 in SCT gene (P = 6.7x10
-5
), and s3780540 o NIPSNAP3B
gene (P = 7.8x10
-5
). We admi ha a u he e alida ion in an independen s udy o SSc in
Whi es is necessa y.
Ou o he 19 genes ha we e p e iously iden i ied as ha bo ing SSc suscep ibili y SNPs in
Whi es, only TNFAIP3 was nominally signi ican in he gene-le el analysis in A ican Ame i-
cans. P e iously, we showed ha SNPs o TNFAIP3 had a s ong associa ion wi h exp ession o
ma ix me allop o einase 1 and 3 in ib oblas s o e hnically di e se pa ien s in esponse o sil-
ica pa icle s imula ion [27].
Se e al ac o s could ha e con ibu ed o he absence o an associa ion in A ican Ame i-
cans o genes ound in Whi es. Fi s , he dis ibu ion o clinical pheno ypes is ma kedly di e -
en in he wo popula ions, wi h a conside ably highe p opo ion o he di use pheno ype
among A ican Ame icans (69%) as compa ed o Whi es (31%). Ou p e ious publica ion [6]
shows di e ences in he gene ic a chi ec u e o SSc clinical pheno ypes. Thus clinical pheno-
ype-speci ic analyses by e hnic g oup would be mos meaning ul, because hey would allow
o mo e accu a e acial compa isons. Un o una ely, he limi ed numbe o A ican Ame ican
pa icipan s p ecluded he pheno ype o au oan ibody subse analyses in he cu en s udy.
Second, he powe o he analysis in he A ican Ame icans was limi ed because o he sam-
ple size. Mo eo e , he powe o he analysis depends no solely on he sample size bu also on
he isk allele equency. S9 Table (b) exempli ies ha he e is a conside able a ia ion in he
allele equencies be ween he wo popula ions, which could ha e con ibu ed o he in e -e h-
nic di e ences. Thi d, e en i he e ec s o causal SNPs a e simila ac oss e hnici ies, GWAS-
iden i ied agging SNP alleles can be in he opposi e linkage phases in wo gi en e hnic g oups.
This will esul in he opposi e e ec s o he agging SNPs iden i ied as signi ican in bo h A i-
can Ame icans and Whi es, and he da a in S9 Table (b) sugges exac ly ha : some SNPs iden-
i ied as nominally signi ican ha e e y simila equencies bu he opposi e di ec ion o he
e ec in A ican Ame icans and Whi es. I is also possible ha he causal SNPs a e di e en in
di e en e hnici ies al hough he suscep ibili y genes a e he same. A gene-le el analysis o
Gene-le el associa ion analysis o sys emic scle osis
PLOS ONE | h ps://doi.o g/10.1371/jou nal.pone.0189498 Janua y 2, 2018 8 / 12
dense geno yping da a, such as ou s, should be able o cap u e he suscep ibili y genes e en in
case o he e hnic he e ogenei y o causal alleles, unless a gi en e hnici y lacks causal a ian s
in a po en ial suscep ibili y gene, in which case he gene will no be associa ed wi h disease in
ha e hnic g oup. The genes lis ed in Table 2, excep o NIPSNAP3B, a e densely SNP-geno-
yped in bo h Whi es and A ican Ame icans. Thus i is no e y likely ha an indi idual SNP
being poly- e sus monomo phic has led o he loss o an associa ion. Fo NIPSNAP3B, he op
SNP in bo h popula ions was he same, s3780540, and he MAF was ac ually highe in A ican
Ame icans (0.116) han in Whi es (0.0179), ye i was only signi ican in Whi es.
An in e ac ion ne wo k buil using Pa hway S udio de ec ed a la ge numbe o in e ac ions
be ween genes associa ed wi h SSc isk. Genes wi h he la ges numbe o in e ac ions include
ITGAM,AIF1,STAT1,IL12RB2; hese genes o m hubs o he ne wo k and a e likely o be
mas e genes in biological con ol o SSc isk. Two sugges i e candida es, EXOC2 and PINX1,
a e also pa o his ne wo k.
As men ioned be o e, limi a ions o his s udy a e (1) a ela i ely small sample size o A i-
can Ame icans, which p e en s us om d awing any de ini e conclusion conce ning he hi h-
e o un esol ed issue whe he he same genes/SNPs in luence SSc isk in di e en e hnic
g oups, and (2) he absence o independen alida ion coho s. We acknowledge, he e o e,
ha ou analyses should be conside ed explo a o y. Ne e heless, we ca ied ou such analyses
because he da a a e unique and hei analysis may be impo an o he unde s anding o he
ole o e hnici y in he gene ic a chi ec u e o SSc.
The esul s o ou explo a o y analysis migh sugges ha he e exis bo h ans- acial and
ace-speci ic suscep ibili y loci o SSc, bu u he alida ion by independen s udies, in pa ic-
ula a p ope ly powe ed SSc GWAS in A ican Ame icans ha allows subse analyses, is neces-
sa y o answe his ques ion.
Conclusions
A gene-le el analysis ocusing on he da a gene a ed by he ImmunoChip pla o m was pe -
o med on Whi e and A ican Ame ican SSc pa ien s. This s udy ep esen s he i s epo o
gene ic analysis o A ican Ame icans wi h SSc. The gene-le el analysis iden i ied ou no el
candida e genes (STAT1,FCGR2C,NIPSNAP3B, and SCT) signi ican ly associa ed wi h SSc in
Whi es. As an explo a o y analysis we compa ed he esul s in Whi es wi h hose gene a ed
om A ican Ame icans. The e was weak e idence o exis ence o SSc suscep ibili y loci ha
showed e ec s in bo h Whi es and A ican Ame icans. Ou indings need o be alida ed by
independen s udies, pa icula ly due o he limi ed sample size o A ican Ame icans. The
clinical pheno ype and au oan ibody subse analyses o Whi es a e also p esen ed, bu u u e
s udies should compa e he pheno ype- and au oan ibody-s a i ied analyses in Whi es and
A ican Ame icans.
Suppo ing in o ma ion
S1 Table. A ca alogue o genes iden i ied o limi ed SSc in Whi es as ha bo ing SNPs a
p<10
−5
in ei he ImmunoChip o GWAS da a.
(XLSX)
S2 Table. A ca alogue o genes iden i ied o di use SSc in Whi es as ha bo ing SNPs a
p<10
−5
in ei he ImmunoChip o GWAS da a.
(XLSX)
Gene-le el associa ion analysis o sys emic scle osis
PLOS ONE | h ps://doi.o g/10.1371/jou nal.pone.0189498 Janua y 2, 2018 9 / 12