University of Groningen
Systematic identification of trans eQTLs as putative drivers of known disease associations
Westra, Harm-Jan; Peters, Marjolein J.; Esko, Tonu; Yaghootkar, Hanieh; Schurmann,
Claudia; Kettunen, Johannes; Christiansen, Mark W.; Fairfax, Benjamin P.; Schramm,
Katharina; Powell, Joseph E.
Published in:
Nature Genetics
DOI:
10.1038/ng.2756
IMPORTANT NOTE: You are advised to consult the publisher's version (publisher's PDF) if you wish to cite from
it. Please check the document version below.
Document Version
Publisher's PDF, also known as Version of record
Publication date:
2013
Link to publication in University of Groningen/UMCG research database
Citation for published version (APA):
Westra, H-J., Peters, M. J., Esko, T., Yaghootkar, H., Schurmann, C., Kettunen, J., Christiansen, M. W.,
Fairfax, B. P., Schramm, K., Powell, J. E., Zhernakova, A., Zhernakova, D. V., Veldink, J. H., Van den Berg,
L. H., Karjalainen, J., Withoff, S., Uitterlinden, A. G., Hofman, A., Rivadeneira, F., ... Hoen, P. A. C. .
(2013). Systematic identification of trans eQTLs as putative drivers of known disease associations.
Nature
Genetics
,
45
(10), 1238-1243. https://doi.org/10.1038/ng.2756
Copyright
Other than for strictly personal use, it is not permitted to download or to forward/distribute the text or part of it without the consent of the
author(s) and/or copyright holder(s), unless the work is under an open content license (like Creative Commons).
The publication may also be distributed here under the terms of Article 25fa of the Dutch Copyright Act, indicated by the “Taverne” license.
More information can be found on the University of Groningen website: https://www.rug.nl/library/open-access/self-archiving-pure/taverne-
amendment.
Take-down policy
If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately
and investigate your claim.
Downloaded from the University of Groningen/UMCG research database (Pure): http://www.rug.nl/research/portal. For technical reasons the
number of authors shown on this cover page is limited to 10 maximum.
1238 VOLUME 45 | NUMBER 10 | OCTOBER 2013 Nature GeNetics
L E T T E R S
Identifying the downstream effects of disease-associated
SNPs is challenging. To help overcome this problem,
we performed expression quantitative trait locus (eQTL)
meta-analysis in non-transformed peripheral blood samples
from 5,311 individuals with replication in 2,775 individuals.
We identified and replicated trans eQTLs for 233 SNPs
(reflecting 103 independent loci) that were previously
associated with complex traits at genome-wide significance.
Some of these SNPs affect multiple genes in trans that are
known to be altered in individuals with disease: rs4917014,
previously associated with systemic lupus erythematosus
(SLE)
1
, altered gene expression of C1QB and five type I
interferon response genes, both hallmarks of SLE
2–4
. DeepSAGE
RNA sequencing showed that rs4917014 strongly alters the
3′
UTR levels of IKZF1 in cis, and chromatin immunoprecipitation
and sequencing analysis of the trans-regulated genes implicated
IKZF1 as the causal gene. Variants associated with cholesterol
metabolism and type 1 diabetes showed similar phenomena,
indicating that large-scale eQTL mapping provides insight into
the downstream effects of many trait-associated variants.
Genome-wide association studies (GWAS) have identified thou-
sands of variants that are associated with complex traits and diseases.
However, because most variants are noncoding, it is difficult to iden-
tify causal genes. Several eQTL-mapping studies
5–8
have shown that
disease-predisposing variants often affect the gene expression levels
of nearby genes (cis eQTLs). A few recent studies have also identified
trans eQTLs
5,9–13
, showing the downstream consequences of some
variants. However, the total number of reported trans eQTLs is low,
mainly owing to the multiple-testing burden. To improve statisti-
cal power, we performed an eQTL meta-analysis in 5,311 peripheral
blood samples from 7 studies (EGCUT
14
, InCHIANTI
15
, Rotterdam
Study
16
, Fehrmann
5
, HVH
17–19
, SHIP-TREND
20
and DILGOM
21
) and
replication analysis in another 2,775 samples. We aimed to ascertain
to what extent SNPs affect genes in cis and in trans and to determine
whether eQTL mapping in peripheral blood could identify down-
stream pathways that might be drivers of disease processes.
Our genome-wide analysis identified cis eQTLs for 44% of all tested
genes (6,418 genes at probe-level false discovery rate (FDR) <0.05 and
4,690 genes with a more stringent Bonferroni multiple-testing correc-
tion; Table 1, Supplementary Figs. 1–3 and Supplementary Tables
1–3). Our trans-eQTL analysis focused on 4,542 SNPs that have been
implicated in complex disease or traits (derived from the Catalog of
Published GWAS; see URLs). In the discovery data set, we detected
trans eQTLs for 1,513 significant trans eQTLs that included 346 unique
SNPs (FDR <0.05; 8% of all tested SNPs; Table 1, Supplementary Fig. 4
and Supplementary Table 4) affecting the expression of 430 different
genes (643 trans eQTLs, including 200 unique SNPs and 223 different
genes with a more stringent Bonferroni correction).
We used stringent procedures for trans-eQTL detection
(Supplementary Note) and various benchmarks to ensure reliability:
for 26 trans-eQTL genes, the eQTL SNP affected multiple probes
within these genes (Supplementary Table 5), always with consistent
allelic directions, suggesting that our probe-filtering procedure was
effective in preventing false-positive trans eQTLs. Trans eQTLs showed
similar effect sizes across the various cohorts (Supplementary Fig. 5).
Systematic identification of trans eQTLs as putative
drivers of known disease associations
Harm-Jan Westra
1,40
, Marjolein J Peters
2,3,40
, Tõnu Esko
4,40
, Hanieh Yaghootkar
5,40
, Claudia Schurmann
6,40
,
Johannes Kettunen
7,8,40
, Mark W Christiansen
9,40
, Benjamin P Fairfax
10,11
, Katharina Schramm
12,13
,
Joseph E Powell
14,15
, Alexandra Zhernakova
1
, Daria V Zhernakova
1
, Jan H Veldink
16
, Leonard H Van den Berg
16
,
Juha Karjalainen
1
, Sebo Withoff
1
, André G Uitterlinden
2,3,17
, Albert Hofman
3,17
, Fernando Rivadeneira
2,3,17
,
Peter A C ’t Hoen
18
, Eva Reinmaa
4
, Krista Fischer
4
, Mari Nelis
4
, Lili Milani
4
, David Melzer
19
, Luigi Ferrucci
20
,
Andrew B Singleton
21
, Dena G Hernandez
21,22
, Michael A Nalls
21
, Georg Homuth
6
, Matthias Nauck
23
,
Dörte Radke
24
, Uwe Völker
6
, Markus Perola
4,8
, Veikko Salomaa
8
, Jennifer Brody
9
, Astrid Suchy-Dicey
25
,
Sina A Gharib
26
, Daniel A Enquobahrie
25
, Thomas Lumley
27
, Grant W Montgomery
28
, Seiko Makino
10
,
Holger Prokisch
12,13
, Christian Herder
29
, Michael Roden
29–31
, Harald Grallert
32
, Thomas Meitinger
12,13,33,34
,
Konstantin Strauch
35,36
, Yang Li
37
, Ritsert C Jansen
37
, Peter M Visscher
14,15
, Julian C Knight
10
,
Bruce M Psaty
9,38,41
, Samuli Ripatti
7,8,39,41
, Alexander Teumer
6,41
, Timothy M Frayling
5,41
, Andres Metspalu
4,41
,
Joyce B J van Meurs
2,3,41
& Lude Franke
1,41
A full list of authors affiliation appears at the end of the paper.
Received 24 December 2012; accepted 14 August 2013; published online 8 September 2013; doi:10.1038/ng.2756
npg
© 2013 Nature America, Inc. All rights reserved.
Nature GeNetics VOLUME 45 | NUMBER 10 | OCTOBER 2013 1239
L E T T E R S
We did not find evidence that trans eQTLs were driven by differences
in age or blood cell counts between individuals (Supplementary Fig. 6,
Supplementary Table 6 and Supplementary Note). However,
we cannot exclude this possibility entirely because FACS analyses
on individual cell types had not been conducted. We also detected
previously reported blood trans eQTLs
5
in this study (Supplementary
Fig. 7, Supplementary Table 7 and Supplementary Note).
To ensure reproducibility of the detected trans eQTLs, we replicated
trans eQTLs from our discovery meta-analysis in 2 independent studies
of peripheral blood gene expression: 52% in KORA F4 (n = 740
samples)
22
and 79% in BSGS (n = 862 samples)
23
(FDR < 0.05;
Supplementary Fig. 8). Irrespective of significance, 91% and 93%
of all 1,513 significant trans-eQTL SNP-probe combinations showed
consistent allelic direction in these replication cohorts compared with
in the discovery analysis. A meta-analysis of the two replication studies
improved replication rates: 89% of the 1,513 trans eQTLs were sig-
nificantly replicated (FDR <0.05), with 99.7% showing a consistent
allelic direction. Irrespective of significance, 97% of the trans eQTLs
showed a consistent allelic direction in this replication meta-analysis
(Supplementary Fig. 8). We found that some trans eQTLs could be
detected in three cell type–specific data sets (283 monocyte samples
9
,
282 B cell samples
9
and 608 HapMap lymphoblastoid cell line (LCL)
samples
24
; Supplementary Figs. 9 and 10). Despite the different tis-
sues analyzed in these three studies, we were able to significantly
replicate 7%, 4% and 2% of the trans eQTLs (FDR <0.05), respectively.
As 95% of the trans-eQTL SNPs explained less than 3% of the total
expression variance (Supplementary Fig. 11 and Supplementary
Table 6), we lack statistical power to replicate most trans eQTLs in
these smaller replication cohorts.
We subsequently confined further analyses to 2,082 different SNPs
that have been found to be associated with complex traits at genome-
wide significance (trait-associated SNPs; reported P < 5 × 10
−8
;
out of 4,542 unique SNPs that we tested). These 2,082 SNPs showed
a significantly higher number of trans-eQTL effects compared with
the 2,460 tested SNPs with reported disease associations at lower
significance levels (P = 8 × 10
−22
; Supplementary Fig. 12 and
Supplementary Note): 254 of these 2,082 SNPs showed a trans-
eQTL effect in the discovery analysis (reflecting 1,340 SNP-probe
combinations; 1,201 of these were significantly replicated in blood,
reflecting 233 different SNPs and 103 independent loci). For 671 of
these 1,340 trans eQTLs (50%), the trait-associated SNP (or a SNP
in strong linkage disequilibrium, LD) was the strongest trans-eQTL
SNP within the locus or was unlinked to the strongest trans
-eQTL
SNP (Supplementary Table 8 and Supplementary Note). The 2,082
trait-associated SNPs were 6 times more likely to cause trans-eQTL
effects than were randomly selected SNPs (matched for distance to
the gene and allele frequency; P = 5.6 × 10
−49
; Supplementary Fig. 13
and Supplementary Note). SNPs associated with (auto)immune or
hematological traits were twice as likely to underlie trans eQTLs com-
pared with other trait-associated SNPs (P = 5 × 10
−25
; Supplementary
Note). Trait-associated SNPs that also caused trans eQTLs affected
the expression levels of nearby transcription factors in cis more fre-
quently than trait-associated SNPs that did not affect genes in trans
(Fisher’s exact P = 0.032; Supplementary Note), suggesting that some
trans eQTLs arise owing to altered cis gene expression levels of nearby
transcription factors.
We examined the genomic properties of the trans-eQTL SNPs (and
their perfect proxies identified using data from the 1000 Genomes
Project
25,26
): these SNPs were significantly enriched for mapping
within microRNA (miRNA) binding sites (Fisher’s exact P < 0.05;
Fig. 1a). They mapped to regions in K562 (myeloid) and GM12878
(lymphoid) cell lines showing enrichment of histone enhancer signals
(fold change >2.5; Fig. 1b) compared to the signals observed in six
non-blood cell lines. Enhancer enrichment in myeloid and lymphoid
cells supports the validity of our blood-derived trans eQTLs. These
results suggest that trans-eQTL effects are tissue specific, a notion that
is supported by our inability to replicate a trans eQTL that was previ-
ously identified in adipose tissue
13
for SNP rs4731702, associated with
both type 2 diabetes (T2D) and lipid levels.
These trans eQTLs can provide insight into
the pathogenesis of disease. Although RNA
microarray studies have identified dysregu-
lated pathways for many complex diseases,
it is often unclear whether associated SNPs
first cause defects in the pathways whose
Average fold enrichment
Average fold enrichment
Fraction of trans-eQTL SNPs
0
1
2
3
4
5
6
7
8
a b
Fraction of trans-eQTL SNPs in category
Average fold enrichment compared to random data
Significant enrichment (Fisher’s exact P value < 0.05)
Enrichment in permuted data
Cell line
Enrichment in real data
*
0
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0
miRNA binding site (Sanger)
miRNA binding site (miRanda)
Splice enhancer or silencer
Transcription factor binding sites
Nonsynonymous
Copy number polymorphisms
GM12878
K562
HUVEC
HMEC
NHEK
HSMM
HepG2
H1
NHLF
1
2
3
4
5
6
7
8
9
10
*
Figure 1 Trans-eQTL SNPs are enriched for
functional elements. We investigated whether
trans-eQTL SNPs are enriched for certain
functional elements using the online tools
SNPInfo, SNPNexus and HaploReg that rely on
data from, among others, the ENCODE Project.
(a) Trans-eQTL SNPs are enriched for mapping
within miRNA binding sites. (b) Trans-eQTL
SNPs show strong enrichment (as annotated
using HaploReg) for enhancer regions that
are present in K562 (myeloid) and GM12878
(lymphoid) cell lines (error bars, 1 s.d.).
Table 1 Results of cis- and trans-eQTL mapping analyses
Cis-eQTL analysis Trans-eQTL analysis
FDR <0.05
significance
Bonferroni
significance
FDR <0.05
significance
Bonferroni
significance
Number of significant
unique SNP-probe pairs
664,097 395,543 1,513 643
Number of significant
unique eQTL SNPs
397,310 266,036 346 200
Number of significant
unique eQTL probes
8,228 5,738 494 240
Number of significant
unique eQTL-regulated
genes
6,418 4,690 430 223
Number of significant
unique eQTL probes not
mapping to genes
636 326 35 13
npg
© 2013 Nature America, Inc. All rights reserved.
1240 VOLUME 45 | NUMBER 10 | OCTOBER 2013 Nature GeNetics
L E T T E R S
dysregulation ultimately leads to disease or vice-versa. One example
of this type of complex disease is SLE, which is an autoimmune dis-
ease causing inflammation and tissue damage. Individuals with SLE
have increased type I interferon (IFN-α) levels, increased expression
of IFN-α response genes
4,27,28
and decreased expression of the C1Q
complement genes. We observed that four common SLE-associated
variants affected IFN-α response genes in cis (IRF5, IRF7, TAP2 and
PSMB9; Supplementary Table 3). As most SLE-associated SNPs do
not map near complement or IFN-α response genes, we assessed
whether SLE-associated SNPs affect these genes in trans. This was
the case for rs4917014, for which the SLE risk allele (rs4917014[T];
showing genome-wide significance in Asian populations and nominal
significance in European populations
1,24
) not only increased expres-
sion of five different IFN-α response genes (HERC5, IFI6, IFIT1,
MX1 and TNFRSF21; Fig. 2) but also decreased expression of three
different probes in CLEC10A. We also observed a nominally signifi-
cant association of rs4917014[T] with decreased expression of C1QB
(P = 5.2 × 10
−6
; FDR = 0.28), encoding a subunit of the C1q com-
plement complex, which has a protective role in lupus: complete
deletion of the genes encoding the C1q subunits practically ensures
the development of SLE
29,30
. CLEC10A and CLEC4C belong to the
C-type lectin family, which includes mannose-binding lectins (MBLs).
SLE risk allele rs4917014[T]
affects IKZF1 in cis and many
genes in trans
cis eQTLs
–10 100
Direction of
effect (z score)
FDR threshold of 0.05
Increases expression (FDR <0.05)
Decreases expression (FDR <0.05)
Significant replication (FDR <0.05)
Non-significant replication (FDR >0.05)
NS
*
IKZF1
HT12v3 probe
4390086
deepSAGE tag
7_50472431
50,275 50,400
0
0.2
0.4
0.6
0.8
1.0
0
20
40
Chromosome 7 postion (hg18) (kb)
Recombination rate (cM/Mb)
LD (R
2
)
IKZF1
SLE SNP MCV SNP
MCV
SNP rs12718597
Exon
Intron
P = 6.29 × 10
–6
P = 8.14 × 10
–13
AC010679.1, ALDH5A1, AP2S1, B4GALT3, C19orf62, C1orf128, C22orf13, C5orf4, CCBP2,
CSDA, E2F2, EIF2AK1, EIF3S9, FAM104A, FBXO7, GCAT, GPR146, HAGH, HEMGN, HK1,
HPS1, KCNH2, KLC3, KRT1, LGALS3, MAP2K3, MARCH8, MCOLN1, MSI2, OSBP2, PDLIM7,
PFDN5, PLEK2, PPP2R5B, PTMS, RAP1GAP, RIOK3, RP11-529I10.4, RPIA, SESN3, SIAH2,
SLC38A5, SLC6A8, SLC7A5, STOML2, TFDP1, TGM2, TMEM86B, TSTA3, VWCE
Genes involved in hemoglobin and erythrocyte metabolic processes
5′ 3′
EGCUT
SHIP–TREND
Fehrmann HT12
Fehrmann H8v2
Rotterdam Study
DILGOM
InCHIANTI
HVH HT12v3
HVH HT12v4
Meta–analysis
Genes involved in complement
Type I interferon response genes
Enrichment of IKZF1 binding (Wilcoxon P = 0.05)
Discovery cohorts
Replication cohorts
B Cells
Monocytes
NS
NS
NS NS
NS
NS
NS
NS NS
NS
NS
NS
NS
NS
NS
NS
NSNSNSNS
Peripheral blood (KORA F4)
Peripheral blood (BSGS)
* *
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
**
–10 100
CLEC10A
Chr: 17 (Probe 1)
6,918,761
–10 100
CLEC10A
Chr: 17 (Probe 2)
6,918,692
–10 100
CLEC10A
Chr: 17 (Probe 3)
6,920,160
–10 100
C1QB
Chr: 1
22,860,401
–10 100
IFI6
Chr: 1
27,867,392
–10 100
CLEC4C
Chr: 12
7,790,248
–10 100
HERC5
Chr:4
89,646,249
–10 100
IFIT1
Chr: 10
91,153,394
–10 100
TNFRSF21
Chr: 6
47,307,480
–10 100
MX1
Chr: 21
41,752,945
–1.37
–2.89
–3.46
–3.56
–3.40
–1.89
–2.47
–0.79
–0.82
–7.07
–3.45
–0.25
–2.15
1.89
–2.30
–1.49
–3.92
–2.43
–3.35
–3.80
–0.38
–0.72
–6.98
–2.59
–3.45
–2.83
–0.38
–1.67
–2.76
–4.70
–3.23
–1.13
–2.21
–1.68
1.04
–1.06
–6.51
–2.70
–0.32
–1.47
–2.21
–1.34
–2.20
–3.81
–0.71
–0.85
–1.28
–1.31
0.08
0.71
–4.56
–3.41
2.66
2.95
2.61
1.77
2.51
–0.30
–0.55
0.44
5.06
2.78
4.13
3.57
3.48
3.41
3.17
0.46
0.12
–0.47
2.23
5.14
1.51
–0.13
5.37
3.08
4.11
4.06
0.03
1.05
2.35
–0.47
–0.01
0.44
5.99
2.93
4.13
3.60
4.10
2.96
4.52
4.31
0.44
1.77
2.35
–1.41
–0.12
0.80
6.30
2.42
4.03
3.46
4.44
2.60
2.98
2.61
1.10
2.53
5.14
1.19
6.71
2.18
1.90
4.37
4.31
–0.77
3.65
4.17
4.31
0.75
1.66
2.07
0.84
0.08
0.61
7.14
3.41
3.83
3.96
4.13
0.38
1.78
–0.22
1.42
0.55
0.57
0.12
MCV trans-eQTL effects
SLE trans-eQTL effects
Figure 2 Independent trans-eQTL effects emanating from the IKZF1 locus. SNP rs4917014, associated with SLE, and unlinked SNP rs4917014,
associated with MCV, both affect the expression of IKZF1 in cis. rs12718597 affects 50 genes in trans (mostly involved in hemoglobin metabolism),
and rs4917014 affects 8 different genes in trans: the rs4917014[T] risk allele is associated with increased expression of genes involved in the type I
interferon response. At a somewhat lower significance threshold (FDR = 0.28), rs4917014[T] is associated with decreased complement C1QB
expression. Both processes are hallmark features of SLE.
npg
© 2013 Nature America, Inc. All rights reserved.
Nature GeNetics VOLUME 45 | NUMBER 10 | OCTOBER 2013 1241
L E T T E R S
Although, to our knowledge, CLEC10A and CLEC4C have not been
studied in the context of SLE, the role of MBLs is similar to that of
the C1q complex, and MBLs are a risk factor for the development
of autoimmunity in humans and mice
3
. The rs4917014 trans eQTLs
replicated well in the peripheral blood and monocyte replication data
sets and reinforce the role of altered expression of the IFN-α pathway,
C-type lectin and C1Q genes in SLE. Individuals without SLE but
who carry the rs497014[T] risk allele show these pathway alterations,
indicating that these affected pathways are not solely a consequence
of SLE but could precede SLE onset.
We investigated the underlying mechanisms of the effects exerted
by rs4917014. IKZF1 is the only gene overlapping the rs4917014
locus. As this gene encodes a transcription factor (Ikaros-family
zinc finger 1), cis-regulatory effects of rs4917014 on IKZF1 and con-
sequent altered IKZF1 protein levels could constitute a mechanism
for the detected trans-eQTL effects. However, because our meta-
analysis did not initially detect a cis eQTL on the Illumina probe
for IKZF1 located near the 5′ UTR of the gene, we investigated the
3′ UTR using Deep Serial Analysis of Gene Expression (DeepSAGE)
next-generation RNA sequencing data from 94 peripheral blood
samples
31
. The variant rs4917014[T] allele increased expres-
sion levels of the 3′ UTR of IKZF1 (Spearman’s correlation = 0.45;
P = 6.29 × 10
−6
). Using Encyclopedia of DNA Elements (ENCODE)
Project
32
chromatin immunoprecipitation and sequencing (ChIP-
seq) data, we observed significantly increased IKZF1 protein
binding within genomic locations corresponding with trans eQTL–
upregulated genes compared with all other genic DNA (Wilcoxon
P value = 0.046) and with SLE cis eQTL–upregulated genes outside of
the IKZF1 locus (Wilcoxon P value = 4.3 × 10
−4
), thereby confirm-
ing the importance of IKZF1 in SLE. IKZF1 is also important for
other phenotypes: rs12718597, an unlinked intronic variant within
IKZF1, is associated with mean corpuscular volume (MCV)
33
and
affects the expression of Illumina probe 4390086 near the 5′ end of
IKZF1 in cis. Ikzf1 knockout mice show abnormal erythropoiesis
34
,
suggesting a causal role for human IKZF1 in MCV as well. However,
although rs12718597[A] was associated in trans with the upregula-
tion of 31 genes and with the downregulation of 19 genes, none of the
SLE trans-regulated genes overlapped with the MCV trans-regulated
genes. The latter were mainly involved in hemoglobin metabolism
and did not show increased IKZF1 binding (Wilcoxon P value = 0.35).
In summary, these results indicate that IKZF1 has multiple functions
and that different SNPs near IKZF1
elicit function-specific effects.
We identified other trans
eQTLs showing similar phenomena.
For example, rs174546 (located in the 3′ UTR of FADS1 and asso-
ciated with metabolic syndrome
35
and with low-density lipopro-
tein (LDL) and total cholesterol levels
36,37
) affected the expression
of TMEM258, FADS1 and FADS2 in cis and the expression of
LDLR in trans (Supplementary Fig. 14). LDLR encodes the LDL
receptor and contains common variants that are also associated with
lipid levels
37
. LDLR gene expression levels correlated negatively
(P < 3.0 × 10
−4
) with total, high-density lipoprotein (HDL) and LDL
cholesterol levels in the tested cohorts (Rotterdam Study and EGCUT;
Supplementary Table 9), indicating that peripheral blood is a use-
ful tissue for gaining insight into the downstream effects of lipid-
regulating SNPs.
For 21 different complex traits, at least 2 unlinked variants that
have been associated with these diseases affected exactly the same
gene in trans (compared with 1 complex trait similarly affected by
variants from equally sized but permuted lists of trans eQTLs; Table 2,
Supplementary Fig. 15 and Supplementary Table 10). Although most
of these traits are hematological (for example, mean platelet volume or
serum iron levels), we also observed this convergence for blood pres-
sure, celiac disease, multiple sclerosis and type 1 diabetes (T1D).
rs3184504 (located in an exon of SH2B3) and its proxy rs653178
(located in an intronic region of ATXN2 on chromosome 12) have
been associated with several autoimmune diseases, including T1D
38,39
and the production of autoantibodies therein
38,39
, celiac disease
8,40
,
hyperthyroidism
41
, vitiligo
42
and rheumatoid arthritis
40
, as well as
with other complex traits such as blood pressure
43,44
, chronic kidney
disease
45
and eosinophil counts
46
. We observed a cis-eQTL effect for
this SNP on SH2B3 (FDR < 0.05) and trans-eQTL effects on 14 genes
(FDR < 0.05; Fig. 3), all of which are highly expressed in neutrophils.
Because the trans-eQTLs effects could be explained by known effect
of rs3184504 on differences in cell count proportions
46
, we correlated
Table 2 Complex traits where multiple unlinked SNPs affect the
same downstream genes
Trait type Complex trait
Genes affected by at least two
unlinked trait-associated SNPs
Immune-related
traits
T1D GBP4, STAT1
T1D autoantibodies GBP4, STAT1
Celiac disease CXCR6, FYCO1
Multiple sclerosis CD5
Blood pressure
traits
Diastolic blood pressure LOC338758
Systolic blood pressure LOC338758
Hematological
traits
Hemoglobin ALAS2
Hematological parameters FBXO7
F cell distribution ESPN, PHOSPHO1, GNAS,
TSPAN13, VWCE,
Hematocrit ALAS2
Serum markers of iron
status
ALAS2
Red blood cell traits ALAS2
Serum iron levels ALAS2
Glycated hemoglobin levels ALAS2
Hematology traits ALAS2
Serum hepcidin ALAS2
β-thalassemia PHOSPHO1, VWCE, TSPAN13,
ESPN
Hematological and
biochemical traits
AL109955.37-3, RBM38, TRIM58
Mean corpuscular
hemoglobin
ALAS2, C18orf10, DNAJB2, ESPN,
HBM, KEL, PDZK1IP1, PIM1,
PRDX5, RAP1GAP, UBXN6,
VWCE, XK
Mean corpuscular volume ALAS2, B4GALT3, C18orf10,
C1orf128, C22orf13, C5orf4,
CCBP2, CSDA, DNAJB2,
EIF2AK1, ESPN, FBXO7, HAGH,
HBM, HPS1
, KEL, KLC3, KRT1,
LGALS3, MARCH8, MCOLN1,
OSBP2, PDZK1IP1, PHOSPHO1,
PIM1, PLEK2, PPP2R5B,
PRDX5, PTMS, RAP1GAP,
RIOK3, TGM2, TSTA3, UBXN6,
VWCE, XK
Mean platelet volume ABCC3, AL353716.18, AQP10,
C19orf33, C6orf152, CABP5,
CTDSPL, CTTN, CXCL5, ESAM,
F13A1, GNB5, GNG11, GP9,
GUCY1A3, ITGA2B, ITGB5,
LIMS1, LY6G6F, MMRN1, MPL,
NRGN, PARVB, PRDX6, PTCRA,
RAB27B, RBPMS2, SAMD14,
SH3BGRL2, TSPAN9, VCL
npg
© 2013 Nature America, Inc. All rights reserved.