Professional Documents
Culture Documents
Role of The Gut-Brain Axis in The Shared Genetic Etiology Between Gastrointestinal Tract Diseases and Psychiatric Disorders A Genome-Wide Pleiotropic Analysis
Role of The Gut-Brain Axis in The Shared Genetic Etiology Between Gastrointestinal Tract Diseases and Psychiatric Disorders A Genome-Wide Pleiotropic Analysis
Supplemental content
IMPORTANCE Comorbidities and genetic correlations between gastrointestinal tract diseases
and psychiatric disorders have been widely reported, with the gut-brain axis (GBA)
hypothesized as a potential biological basis. However, the degree to which the shared genetic
determinants are involved in these associations underlying the GBA is unclear.
OBJECTIVE To investigate the shared genetic etiology between gastrointestinal tract diseases
and psychiatric disorders and to identify shared genomic loci, genes, and pathways.
DESIGN, SETTING, AND PARTICIPANTS This genome-wide pleiotropic association study using
genome-wide association summary statistics from publicly available data sources was
performed with various statistical genetic approaches to sequentially investigate the
pleiotropic associations from genome-wide single-nucleotide variation (SNV; formerly
single-nucleotide polymorphism [SNP]), and gene levels and biological pathways to
disentangle the underlying shared genetic etiology between 4 gastrointestinal tract
diseases (inflammatory bowel disease, irritable bowel syndrome, peptic ulcer disease, and
gastroesophageal reflux disease) and 6 psychiatric disorders (schizophrenia, bipolar disorder,
major depressive disorder, attention-deficit/hyperactivity disorder, posttraumatic stress
disorder, and anorexia nervosa). Data were collected from March 10, 2021, to August 25,
2021, and analysis was performed from January 8 through May 30, 2022.
MAIN OUTCOMES AND MEASURES The primary outcomes consisted of a list of genetic loci,
genes, and pathways shared between gastrointestinal tract diseases and psychiatric
disorders.
RESULTS Extensive genetic correlations and genetic overlaps were found among 22 of 24 trait
pairs. Pleiotropic analysis under a composite null hypothesis identified 2910 significant
potential pleiotropic SNVs in 19 trait pairs, with 83 pleiotropic loci and 24 colocalized loci
detected. Gene-based analysis found 158 unique candidate pleiotropic genes, which were
highly enriched in certain GBA-related phenotypes and tissues, whereas pathway enrichment
analysis further highlighted biological pathways primarily involving cell adhesion, synaptic
Author Affiliations: Department of
structure and function, and immune cell differentiation. Several identified pleiotropic loci Biostatistics, School of Public Health,
also shared causal variants with gut microbiomes. Mendelian randomization analysis further Cheeloo College of Medicine,
illustrated vertical pleiotropy across 8 pairwise traits. Notably, many pleiotropic loci were Shandong University, Jinan,
Shandong, China (Gong, Guo, L. Liu,
identified for multiple pairwise traits, such as 1q32.1 (INAVA), 19q13.33 (FUT2), 11q23.2 Yan, S. Liu, Wang, Xue, Yuan);
(NCAM1), and 1p32.3 (LRP8). Institute for Medical Dataology,
Shandong University, Jinan, China
CONCLUSIONS AND RELEVANCE These findings suggest that the pleiotropic genetic (Gong, Guo, L. Liu, Yan, S. Liu, Wang,
determinants between gastrointestinal tract diseases and psychiatric disorders are Xue, Yuan); School of Medicine,
extensively distributed across the genome. These findings not only support the shared Cheeloo College of Medicine,
Shandong University Jinan, China
genetic basis underlying the GBA but also have important implications for intervention (Li); Department of Biostatistics,
and treatment targets of these diseases simultaneously. University of Michigan, Ann Arbor
(Zhou); Center for Statistical
Genetics, University of Michigan,
Ann Arbor (Zhou).
Corresponding Author: Zhongshang
Yuan, PhD, Department of
Biostatistics, School of Public Health,
Cheeloo College of Medicine,
Shandong University, 44,
Wenhua West Road, Jinan, Shandong
JAMA Psychiatry. doi:10.1001/jamapsychiatry.2022.4974 250012, China (yuanzhongshang@
Published online February 8, 2023. sdu.edu.cn).
(Reprinted) E1
T
he comorbidities and associations between gastroin-
testinal tract diseases and psychiatric disorders have Key Points
been widely reported,1-3 which was likely to be regu-
Question To what extent are shared genetic determinants in
lated by the gut-brain axis (GBA). The GBA is characterized the comorbidities and associations between gastrointestinal tract
by bidirectional interactions between the gastrointestinal tract diseases and psychiatry disorders involved in the gut-brain axis?
and the central nervous system (CNS) and would link intesti-
Findings In this genome-wide pleiotropic association study
nal dysfunction and inflammation with brain function and psy-
using genome-wide association summary statistics from publicly
chiatric disorders.4,5 Various biological mechanisms were in- available data sources, pervasive genetic correlations and genetic
volved in the GBA, such as inflammatory immune responses, overlaps between gastrointestinal tract diseases and psychiatric
the autonomic nervous system, and enteric nervous system, disorders were found. The pleiotropic genetic determinants
where the role of the composition of gut microbiota and re- between them were extensively distributed across the genome.
lated metabolites has been particularly highlighted.4-7 Under- Meaning These findings not only support the shared genetic basis
lying the conceptual framework of the GBA, the shared ge- underlying the gut-brain axis but also have important implications
netic etiology might be involved in the associations between for intervention and treatment targets of these 2 types of diseases
gastrointestinal tract diseases and psychiatric disorders. simultaneously.
Genome-wide association studies (GWAS) have identi-
fied multiple genetic variants (ie, single-nucleotide varia-
tions [SNVs]; formerly single-nucleotide polymorphisms the underlying shared genetic etiology. Of note, under
[SNPs]) associated with gastrointestinal tract diseases and the framework of pleiotropic analysis, we first performed
psychiatric disorders. 8-10 Genetic correlations have been the SNV-level analysis to detect pleiotropic variants and loci,
suggested between these 2 types of diseases using linkage followed by pairwise colocalization analysis to determine
disequilibrium (LD) score regression (LDSC).10-14 However, colocalized loci and gene-level analysis to identify candidate
it remains unclear whether the overall genetic correlation pleiotropic genes, based on which we further performed par-
would be attributed to a few loci or across the genome.15 In- allel phenotype and tissue-specific enrichment analysis
deed, there would be genetic overlap even without any ge- to characterize the phenotype and tissue specificity as well
netic correlation. Although previous studies have investi- as additional gene-level analysis to identify the tissue-
gated genetic overlap,16 shared susceptibility genes,17,18 and specific and cell type-specific pleiotropic genes. We also
causal relationships19-21 between these 2 types of diseases, they highlighted the role of gut microbiomes in interpreting
mainly focused on inflammatory bowel disease (IBD) and the shared genetic etiology, followed by mendelian random-
psychiatric disorders with limited sample sizes. Recently, ization analysis to evaluate pairwise causal associations
2 studies10,22 have conducted GWAS of specific gastrointesti- and partly characterize different types of pleiotropy (vertical
nal tract diseases as well as systematic post-GWAS analyses and pleiotropy or horizontal pleiotropy).
pointed out the necessity to explore the shared genetic risk
across traits to improve the understanding of the disordered
brain-gut interactions. Therefore, it is of great importance to
further seek out the specific genomic variants or loci account-
Methods
ing for genome-wide genetic correlation and to deeply probe GWAS Data Sets
into the shared genetic etiology between these 2 types of dis- We sought GWAS summary statistics from publicly available
eases. Shared genetic etiology also indicates the potential data sources with European ancestry, owing to the limited
pleiotropy, which often acts as genetic confounding of the as- availability of well-powered GWAS with other ancestries, and
sociations between trait pairs.23-25 Cross-trait analysis has been selected GWAS with sample sizes larger than 50 000 to en-
proposed to investigate the pleiotropic genetic variants or loci sure statistical power. GWAS for GERD, IBD, and PUD were ob-
among multiple traits by leveraging the correlation of GWAS tained from the same gastrointestinal tract GWAS based on
signals,23,26-29 where the pleiotropic loci could serve as inter- 456 327 individuals from UK Biobank (UKB).10 GWAS for IBS
vention targets with the potential to simultaneously prevent were obtained from a larger meta-analysis with 486 601 indi-
or treat these diseases. viduals (53 400 cases and 433 201 controls).22 GWAS for the
In this genome-wide pleiotropic association study, using 6 psychiatric disorders were from the Psychiatric Genomics
large-scale GWAS summary data, we performed a genome- Consortium, including schizophrenia,30 BIP,31 MDD,32 ADHD,33
wide pairwise trait pleiotropic analysis between 4 gastroin- PTSD,34 and AN.35 In addition, GWAS for early age-related
testinal tract diseases (IBD, irritable bowel syndrome [IBS], macular degeneration (AMD)36 and cataract37 were obtained
peptic ulcer disease [PUD], and gastroesophageal reflux dis- to serve as a common set of negative controls for both gastro-
ease [GERD]) and 6 psychiatric disorders (schizophrenia, intestinal tract diseases and psychiatric disorders, given these
bipolar disorder [BIP], major depressive disorder [MDD], 2 disorders are relatively limited to the pathological lesions of
attention-deficit/hyperactivity disorder [ADHD], posttrau- intraocular contents, and previous studies also showed no sig-
matic stress disorder [PTSD], and anorexia nervosa [AN]) nificant genetic correlations between AMD and MDD38 as well
through various statistical genetic approaches to sequentially as among cataract, GERD, and psychiatric symptoms.39 All
investigate the pleiotropic associations from genome-wide, GWAS were approved by relevant ethic committees, and writ-
SNV, and gene levels and biological pathways to disentangle ten informed consent was obtained from all participants, with
details provided in the eTable 1 in Supplement 1. Data were col- based on 2 different reference panels, Genotype-Tissue
lected from March 10, 2021, to August 25, 2021, and analyzed Expression project (GTEx)47 and the Encyclopedia of DNA
from January 8 through May 30, 2022. This genome-wide pleio- Elements project (ENCODE).48 We declared the significance
tropic association study followed the Strengthening the with a nominal threshold (2-sided P < .05) for these 2 parallel
Reporting of Genetic Association Studies (STREGA) reporting enrichment analyses. In addition, we used E-MAGMA (expres-
guideline. sion quantitative trait loci [eQTL]–informed MAGMA)49 and
transcriptome-wide association study analysis using joint-
Statistical Analysis tissue imputation50 to further investigate the tissue-specific
All analyses were performed after excluding SNVs in the genes, parallelized with H-MAGMA (Hi-C–coupled MAGMA)51
major histocompatibility complex region (chromosome 6: to indicate the cell-type specificity, in which we declared the
25-35 megabase [Mb]) due to its complex LD structure and significance with locus-specific Bonferroni correction. Gene
restricted to biallelic SNVs with minor allele frequency larger set enrichment analysis was also performed to identify poten-
than 0.01. Details of these methods are provided in Figure 1 tial biological pathways using the clusterProfiler package.52
and eMethods in Supplement 1. Significantly enriched pathways were declared with normal-
Both LDSC40 and high-definition likelihood41 were used ized enrichment score greater than 2 and adjusted P < .05.
to assess genome-wide genetic correlations for 24 pairwise Multitrait colocalization analysis with HyPrColoc (hypoth-
traits. The intercept from LDSC could also indicate potential esis prioritization for multi-trait colocalization)53 incorporat-
sample overlap between 2 GWAS. Given that genetic correla- ing host-microbiome GWAS was performed to highlight the
tion only reflects the overall correlation across the genome, we critical role of gut microbiome.
further applied genetic analysis incorporating pleiotropy and Last, we conducted bidirectional mendelian randomiza-
annotation (GPA)29 to explore the overall genetic overlap be- tion analysis for 24 pairwise traits between gastrointestinal
tween traits. The Bonferroni-corrected significant threshold tract diseases and psychiatric disorders to investigate the po-
was set at P < 2.08 × 10−3 (.05/24). In addition, negative con- tential causal trait pairs (ie, vertical pleiotropy), along with
trol analysis was performed through LDSC between AMD and negative control analysis using AMD and cataract. We used the
cataract and the 4 gastrointestinal tract diseases as well as the inverse-variance weighted method54 as main analysis and in
6 psychiatric disorders, with Bonferroni-corrected significant total 6 alternative mendelian randomization methods with dif-
threshold set at P < 2.5 × 10−3 (.05/20). ferent model assumptions as additional analyses for further
For the union set of pairwise traits with significant validation. For main analysis, we chose a false discovery rate
genetic correlation or genetic overlap, we used pleiotropic approach for multiple testing correction with the signifi-
analysis under composite null hypothesis (PLACO) to identify cance threshold being false discovery rate–adjusted P < .05,
potential pleiotropic SNVs.26 Single-nucleotide variations given that the commonly used Bonferroni correction is often
with P < 5 × 10 −8 for PLACO were considered significant too stringent for multiple nonindependent tests. The nomi-
pleiotropic variants. Functional mapping and annotation of nally significant threshold (P < .05) was used for alternative
genetic associations (FUMA)42 was applied to characterize mendelian randomization methods. The Latent Heritable
potential pleiotropic loci, based on which a bayesian colocal- Confounder Mendelian Randomization (LHC-MR) method,
ization analysis43 was performed to further identify shared which could account for sample overlap, was used for pair-
causal variants in each pleiotropic locus. We declared a colo- wise traits with potential sample overlap to further validate
calized locus with posterior probability of H4 (PP.H4) larger the mendelian randomization results.55
than 0.7.
Based on PLACO results, we further explored the shared Genomic Loci Characterization
biological mechanisms of these pleiotropic loci. We per- and Functional Annotation
formed gene-level multimarker analysis of GenoMic annota- For significant pleiotropic SNVs from PLACO, we applied FUMA
tion (MAGMA)44 analysis on the genes located in or over- to identify independent variants, characterize genomic risk
lapped with the pleiotropic loci based on both PLACO outputs loci, and annotate the functions of variants using LD informa-
and single-trait GWAS to identify candidate pleiotropic genes, tion from the 1000 Genome Project phase 3 reference panel
with the significance declared at the locus-specific Bonferroni- of European population.42 We characterized independent
corrected 2-sided P < .05 for MAGMA analysis on PLACO re- SNVs with r2 less than 0.6000 and lead SNVs with r2 less than
sults and 2-sided P < .05 for MAGMA analyses based on origi- 0.1000 within 1 Mb. Genomic risk loci were defined by merg-
nal single-trait GWAS. Further, we performed phenotype ing genomic regions if the physical distance between lead
enrichment analysis based on the Mouse Genome Informat- SNVs was less than 250 kilobase.42
ics platform45 to characterize the phenotype specificity of these Functional annotations, including ANNOVAR software
pleiotropic genes against that of nonpleotropic genes by ex- tool categories (Bioinformatics), combined annotation-
amining the differences in the proportions of genes associ- dependent depletion (CADD) scores, and RegulomeDB scores,
ated with certain phenotypes in the pleiotropic gene group were also provided by FUMA. Single-nucleotide variations with
against that in the nonpleotropic gene group using the Fisher a CADD score larger than 12.37 were considered a potentially
exact test. Then, we performed tissue-specific enrichment deleterious variant. 42 Single-nucleotide variations with
analyses to illustrate the tissue specificity of these pleiotro- P < 5 × 10−8 in each single-trait GWAS were also annotated by
pic genes using the deTS tissue-specific enrichment method46 FUMA for comparison.
..
Trait 1 Trait 2
..
E[z1*z2] .. PM10 PM11 PM01
20 14 20
12
15 15
10
-log10 P value
8
10 10
6
4
5 5
2
0 0 0
Chr 1 2 3 4 5 6 7 8 9 10 11 12 13 15 17 19 22 Chr 1 2 3 4 5 6 7 8 9 10 11 12 13 15 17 19 22 Chr 1 2 3 4 5 6 7 8 9 10 11 12 13 15 17 19 22
GWAS of GIT disease PLACO GWAS of PSY disorder
P .adjust
Asymmetric synapse .0067
Dendritic spine .0159
Running enrichment score
D Bidirectional associations
Confounder
Mendelian randomization
We conducted the comprehensive pleiotropic analysis between of GenoMic annotation (MAGMA); H-MAGMA, Hi-C–coupled MAGMA;
4 gastrointestinal tract (GIT) diseases and 6 psychiatric (PSY) disorders HyPrColoc, hypothesis prioritization for multi-trait colocalization; LD, linkage
from different perspectives. COLOC indicates colocalization; GSEA, gene disequilibrium; PLACO, pleiotropic analysis under composite null hypothesis;
set enrichment analysis; GWAS genome-wide association study; SNV, single-nucleotide variation; and TWAS-JTI, transcriptome-wide association
E-MAGMA, expression quantitative trait loci–informed multi-marker analysis study analysis using joint-tissue imputation.
Table 1. Genetic Correlation and Genetic Overlap Estimations Between 24 Pairwise Traitsa
Figure 2. The Overall Landscape of the Pleiotropic Associations Across 4 Gastrointestinal Tract Diseases and 6 Psychiatric Disorders
DENND1B
DYNLT5
A1S
RAB5B
ERBB3
SGIP1
NCAM1
SOUX
FOXP
CRB1
CACN
CAM M6
RBM
RB
1
WD AS1
KV
5
AL R9
KIF VA
R82
CA 21B
S
A1
TL
INA
FO
CN
XP
MS N1 V
GI SP1
MO AMK
HE 1
T1 A
AC GYF
Ba
UF
CY
R
cte
+ 11q23
–12q13.2
1p31.3 +
P
2
1p31.1 +
roi
A 17
L1
1q31.3–
da
5q33 +
–3p2
BO S3M A1
L2
les
9P2 .3–
1q32.1
F3 1
+3p13
2–
BC
10q 3 +
RC T IR CAF
–6q 1.2
21.
+3
1. 31
.2
S7
1q 23.3–
S A
5q 24+ +
16.
p2
AV
4q 2.1
+7
IN 21B
1
q3
22 1–
1q
+3
3
N I F
+
K
7q 12.
1.1
KC GEF
p2
.1
–1
N +
1.
–
GIG J13
2q
10 33 X
31
3. TO
24
YF2 + 8 q24
Ba
1 5
F36
.1
9q
r ne
-8 q24 .32
1
DPY q 1
.1+ ZN
sie
D .3
– + 12. 32 + CB
lla
PDE 2q 4q24 1 1q 31.3 + PRK
4B
AN
3 2 12.1
q
MDD
+2 7.1
q .2–
D
8q
BIP
AD
–1p 33.1 A
INAV
PTS
q21 +
HD
21. 10 2.2 1B
KIF2
3
Z
+1p 1
SC
31. 16 p
3 HD
SC AD .1+ IP 2
6 K
NCAM1
+4q2
4
Z 1q32 IHO1
BIP
+11q23 1+ MST1
KLF7 .2 PTSD 3p21.3
+2q33.3 GE D AN COP1
Sutterella
MDD RD IB 1q25.1+
PSCA 5q14.3+ TMEM161B
-8q24.3
Roseburia
AN 11q23.2+
+4q31.22 MDD
MD
DD NCAM1
SLC10A7 13q14.3+
eae
P IB
OLFM4
D
.2 BIP
+ 11q12
PU
TMEM258
S
nadac
14q21.1
PTS 16p12 +
FADS2 D
18q2 .2+ LRFN5
4.3
FADS1 –8q2 .1
yromo
1.2+
HD
Porph ia
23 SC
+ 8p 2.2 AD
ter
Z 1p3
PSCA –3p
2 2.3+ TCF4
bac
Z
SC
.33 1p2
teo
q13 1 LRP
SD
1q .3– 8
ADH
pro
–19 3q 23.3
D
.33
PT
CPT
AN
MD
21 2 + 2
ma
2q 34.1 5q 2.3+
BIP
BIIP
P
+1 MA
GOH
ae
TR 9q 33.1
Gam
p
MS
1
ce
+ 33 DP
MA FUT2 3. 3
22 + YD
ira
.3
q1 14.
sp
9 1+ PC
6
11 q25 .31–
no
+1 11p C
Ru Ba
SP
p2 4.3
ST B
q2 –
11 q24
DU
ch
m cte
PP AG1
+
q2
1.
12
4.
ino ro
La
3 e
AL
–8
a
2+
e P2
co ide
3G c R3
–1
.2
22
ra FA
ST
cc s
TR pi A
31
23
P M
q1
ac
1p3
MS IP1 os
FA HF2 120
+11q 4.3
22.
1q
5q14
ea
hn
MA RAS T2
3
23.2
1p32.3
5.1
a
1q23.3+
A
–8p23.1
c
.2+
e
3p12.1–
i M
4.2
+1
+7q21.11
q1
La cter
4q27–
FU
–9q
12
–10q2
ba
+13
.3–
0A
–
PS M6
eo
VS SAM
rot OS
E
LY CA
1G
E
+
TH
ap
6K
h
PI HO 4
Alp
MP L6I
TP SP
ria
obacte
AR
NM H9
aprote
Gamm
RA 3H7B
2
1
ZC 300
AM
P
NG
EP
FAM OS
A
SZT I
NC
AP
120
A
HY
120
TME
1
1
2
LRP8
NCAM
P1
MAGO
CZIB
MSRA
CADM2
PCLO
FAM
IL2
BLTP1 IL21
XPNPE
M
161B
H
A circular dendrogram included 4 gastrointestinal tract diseases (inner circle, microorganisms (gray dots, PP>0.7, with the specific microorganism annotated
including gastroesophageal reflux disease [GERD], inflammatory bowel disease on the lines). A total of 196 significant pleiotropic genes (158 unique) were
[IBD], irritable bowel syndrome [IBS], and peptic ulcers disease [PUD]) and further identified by multimarker analysis of GenoMic annotation (MAGMA).
6 psychiatric disorders (second circle; attention-deficit/hyperactivity disorder For the trait pairs with more than 3 pleiotropic genes, we only showed the top
[ADHD], anorexia nervosa [AN], bipolar disorder [BIP], major depressive 3 pleiotropic genes according to the prioritization of candidate pleiotropic
disorder [MDD], posttraumatic stress disorder [PTSD], and schizophrenia genes, which were shown with the statistical prioritization attenuation
[SCZ]), resulting in 24 trait pairs. A total of 83 pleiotropic loci were identified manifested by clockwise direction (fourth circle), the genes with the tissue
across 19 trait pairs (third circle; no pleiotropic loci were identified for IBS-PTSD, specificity in at least one of gastrointestinal tract or brain tissues and in other
PUD-MDD, and PUD-PTSD), in which 53 of 83 top single nucleotide variations tissues identified by E-MAGMA were also highlighted (dark blue dots). Detailed
(SNVs) showed concordant associations (marked by a plus sign) with a certain information for 83 pleiotropic loci (eTable 6 in Supplement 1), colocalization
pair of traits while the remaining top SNVs showed discordant associations results (Table 2), multitrait colocalization results (eTable 17 Supplement 1),
(marked by a minus sign); 24 pleiotropic loci were colocalized for pairwise traits significant pleiotropic genes with annotations of tissue specificity (eTable 15 in
(orange dots, posterior probability of H4 [PP.H4]>0.7); and 11 were colocalized Supplement 1) was also provided.
for certain gastrointestinal tract diseases, psychiatric disorders, and gut
ANNOVAR category annotation illustrated that 34 of 83 in- in Supplement 1). Of note, 5 mRNA exonic variants (4 unique)
dex SNVs (41%) were intronic variants and 27 of 83 (33%) were were located in 3 loci and mapped to 3 genes. Specifically, the
intergenic variants. Only 7 of 83 index SNVs (8%) (6 unique) index SNV rs20551 at 22q13.2 locus (PLACO P = 5.07 × 10−11 for
were exonic variants, including 5 messenger RNA (mRNA) ex- IBS-schizophrenia) was a significant eQTL for EP300 (OMIM
onic variants and 2 noncoding RNA exonic variants (eTable 5 602700) encoding p300 protein, which plays an important role
Trait pair Top SNV Locus boundarya Region Nearest gene PP.H3 PP.H4 Best causal SNV.PP.H4
IBD-MDD rs12118513 1:197342380–197781198 1q31.3 DENND1B 0.0395 0.7372 rs12118513b 0.2022
IBD-MDD rs60689680 5:158827769–158856513 5q33.3 AC008703.1 0.0614 0.7034 rs60689680b 0.1184
IBD-SCZ rs905634 1:200874229–201027055 1q32.1 INAVA 0.0813 0.9139 rs905634b 0.4835
IBD-SCZ rs492430 7:100219167–100523241 7q22.1 EPO 0.0801 0.9116 rs492430b 0.1452
IBD-BIP rs56073120 8:59800835–59925249 8q12.1 TOX 0.0435 0.9044 rs56073120b 0.0632
IBD-BIP rs7090073 10:64387108–64441247 10q21.2 ZNF365 0.0257 0.9719 rs7090073b 0.1992
IBD-BIP rs196001 16:23892887–23962504 16p12.2 PRKCB 0.0656 0.8724 rs196001b 0.0811
IBS-MDD rs3099439 5:87514778–87822672 5q14.3 TMEM161B 0.0536 0.9428 rs3099439b 0.2131
IBS-MDD rs4937872 11:112826867–112912811 11q23.2 RP11-629G13.1 0.0845 0.9030 rs4937872b 0.1163
IBS-MDD rs1862743 16:60665658–60743834 16p12.2 GNPATP 0.0108 0.9578 rs1862743b 0.4157
IBS-SCZ rs12031155 1:53658317–53752134 1p32.3 LRP8 0.0518 0.9368 rs12031155b 0.1577
IBS-SCZ rs7542202 1:163616199–163766672 1q23.3 RP4-640E24.1 0.0550 0.9299 rs7542202b 0.0691
IBS-SCZ rs1280622 3:135807609–136673157 3q22.3 RP11-731C17.1 0.1145 0.8386 rs7432375 0.0965
IBS-SCZ rs11604175 11:124619407–124624854 11q24.2 VSIG2 0.0011 0.7717 rs11604175b 0.5730
IBS-SCZ rs12277680 11:134576216–134595774 11q25 RP11-469N6.2 0.0106 0.7331 rs12277680b 0.4057
IBS-BIP rs5177 1:53658317–53752134 1p32.3 LRP8 0.0553 0.8665 rs5177b 0.1081
IBS-BIP rs2345964 1:163582980–163768927 1q23.3 RP4-640E24.1 0.0254 0.9684 rs2345964b 0.1043
IBS-BIP rs13239217 7:82387493–82583609 7q21.11 PCLO 0.0939 0.8493 rs13239217b 0.1919
PUD-SCZ rs681343 19:49103447–49254955 19q13.33 FUT2 0.0951 0.8927 rs681343b 0.2834
GERD-MDD rs1263674 2:208017033–208088987 2q33.3 AC007879.1:AC007879.2 0.0567 0.8596 rs62188630 0.0638
GERD-PTSD rs13107325 4:102938709–103438709 4q24 SLC39A8 0.0111 0.7904 rs13107325b 0.7793
GERD-SCZ rs1892346 1:66304167–66333877 1p31.3 PDE4B 0.0210 0.8141 rs1892346b 0.0854
GERD-SCZ rs13107325 4:102702364–103387161 4q24 SLC39A8 0.0020 0.9966 rs13107325b 0.8466
GERD-AN rs1873914 12:56368708–56478658 12q13.2 RAB5B 0.0484 0.8134 rs1873914b 0.1042
Abbreviations: ADHD, attention-deficit/hyperactivity disorder; AN, anorexia SCZ, schizophrenia; SNV.PP.H4, posterior probability of the best causal variant.
nervosa; Best causal, the candidate causal single-nucleotide variation (SNV); a
Locus boundary of each pleiotropic genomic risk locus was denoted as
BIP, bipolar disorder; GERD, gastroesophageal reflux disease; IBD, inflammatory “chromosome: start-end” defined by FUMA for the corresponding trait
bowel disease; IBS, irritable bowel syndrome; MDD, major depressive disorder; pair.
PP.H3, posterior probability of H3; PP.H4, posterior probability of H4; b
The top SNV in this locus was also identified as a candidate causal SNV.
PTSD, posttraumatic stress disorder; PUD, peptic ulcer disease;
in cell proliferation and differentiation. The index SNVs PP.H3 larger than 0.7000, indicating there might be different
rs681343 (PLACO P = 1.36 × 10−12 for PUD-schizophrenia) and causal variants in these loci (eTable 8 in Supplement 1).
rs601338 (PLACO P = 5.07 × 10−11 for PUD-ADHD) at 22q13.2
were in high LD (r2 = 0.996) and had almost identical eQTL Prioritization of Candidate Pleiotropic Genes
regulation information (eTable 7 in Supplement 1), regulating and Characterization of Phenotype and Tissue Specificity
the expression of FUT2 (OMIM 182100). The index SNV MAGMA analysis based on 295 potential pleiotropic genes that
rs13107325 at 4q24 locus (PLACO P = 1.78 × 10−14 for GERD- located in or overlapped with 83 pleiotropic loci identified 196
schizophrenia; PLACO P = 4.19 × 10−8 for GERD-PTSD) was a significant pleiotropic genes (158 unique), in which 38 genes
significant eQTL for SLC39A8 (OMIM 608732), which en- were detected in 2 or more trait pairs (eTable 9 in Supple-
codes ZIP8 metal cation transporter. In addition, 7 index SNVs ment 1). For example, NCAM1 was identified as a significant
(6 unique) were identified with CADD scores larger than 12.37, pleiotropic gene in 5 pairs of traits, followed by INAVA,
in which 2 mRNA exonic variants had higher CADD scores, in- CACNA1S (OMIM 114208), and KIF21B (OMIM 608322) in 4 pairs
cluding rs601338 (CADD score: 52; mapped gene: FUT2 [OMIM of traits and BANK1 (OMIM 610292), CAMKV (OMIM 614993),
182100]) and rs13107325 (CADD score: 34; mapped gene: DPYD (OMIM 612779), MON1A (OMIM 611464), MST1R (OMIM
SLC39A8). 600168), PSCA (OMIM 602470), RBM5 (OMIM 606884), RBM6
Further colocalization analysis identified 24 of 83 poten- (OMIM 606886), and SLC39A8 in 3 pairs of traits.
tial pleiotropic loci (29%) with PP.H4 larger than 0.7, in which Further parallel phenotype and tissue-specific enrich-
22 top SNVs of corresponding loci were identified as candi- ment analysis suggested both higher phenotype and tissue
date shared causal variants (Table 2, Figure 2, and eFigure 3 specificity (Figure 3). Specifically, we obtained in total 19 326
in Supplement 1). Interestingly, the 19q13.33 locus, which was genes with available phenotype annotations resorting to Mouse
identified as pleiotropic loci for 2 pairs of traits, was only co- Genome Informatics platform, in which 144 of 158 unique pleio-
localized between PUD and schizophrenia (PP.H4 = 0.8927) tropic genes were included. These pleiotropic genes were as-
rather than PUD and ADHD (PP.H4 = 0.3552), with the same sociated with several GBA-related phenotypes with the analy-
potential shared causal variant rs681343 identified (mapped ses performed on each phenotype (Figure 3A) and were
gene: FUT2). In addition, 7 pleiotropic loci were identified with associated with the 2 phenotypes simultaneously (behavior-
Phenotype
MP:0005381 digestive/alimentary P = 1.88×10–4
0 10 20 30 40 50 60 70
No. of genes
neurological phenotype and digestive-alimentary pheno- specific genes were also provided (eFigure 4 and eTables 11 and
type), contrasting with that of nonpleotropic genes with 12 in Supplement 1).
less association with these phenotypes (7.64% vs 3.61%; In addition, E-MAGMA and joint-tissue imputation analy-
P = 1.80 × 10−2) (Figure 3B). Details of these genes were pro- sis also confirmed the tissue specificity not only in gastroin-
vided in eTable 10 in Supplement 1. Tissue-specific enrich- testinal tract and brain tissues but also in other GBA-related
ment analysis using deTS based on GTEx and ENCODE refer- tissues. In total, 52 pleiotropic genes (22 unique) across 8
ence panels suggested higher tissue specificity (Figure 3) pleiotropic loci were significantly detected in multiple
in several gastrointestinal tract and brain tissues, such as GBA-related tissues in at least 2 trait pairs, for example, PSCA
stomach, transverse colon, terminal ileum of small intestine, (8q24.3) for 3 pairwise traits (PUD-schizophrenia, PUD-BIP, and
Peyer’s patch, and diencephalon. The hypothalamic-pituitary- schizophrenia-AN), LPR8 (1p32.3) for 2 pairwise traits (IBS-
adrenal axis is an important component of the GBA and could schizophrenia and IBS-BIP), and FUT2 (19q13.33) for 2 pair-
provide primary biological response to stressful stimuli, with wise traits (PUD-schizophrenia and PUD-ADHD). H-MAGMA
adrenal gland especially highlighted. Details of these tissue- analysis further suggested the cell-type specificity of these
pleiotropic genes. Details of these gene-level analyses were also variants, pleiotropic genes, biological pathways, and a poten-
provided (eTables 13-15 in Supplement 1). tial genetic basis associated with gut microbiome. All these
findings supported the role of GBA in the shared genetic eti-
Synaptic and Immune-Related Mechanisms Shared Between ology underlying these 2 types of diseases.
Gastrointestinal Tract Diseases and Psychiatric Disorders Using multiple methods with different model assump-
Gene set enrichment analysis identified 127 significantly tions could provide complementary evidence and allow deep
enriched pathways (normalized enrichment score >2 and ad- investigation of the underlying pleiotropic associations from
justed P < .05), including 99 Gene Ontology (GO) terms and different perspectives. The GPA identified substantial ge-
28 Kyoto Encyclopedia of Genes and Genomes (KEGG) path- netic overlap between several trait pairs even without any
ways (eTable 16 in Supplement 1). These enriched pathways significant genetic correlation. The mixed directions across the
are mainly involved in cell adhesion, synaptic structure and genomic risk loci, which indicate the existence of concordant
function, and immune cell differentiation. For example, and discordant pleiotropy, also help explain the polygenic over-
homophilic cell adhesion via plasma membrane adhesion mol- lap despite nonsignificant genetic correlation. PLACO analy-
ecules (GO:0007156), which plays important roles in neural sis further determines the pleiotropic variants, followed by
and immune mechanisms, was identified in 5 pairs of traits. colocalization analysis to identify potential shared causal
A γ-aminobutyric acid (GABA)–modifying synapse (GO: variants in each pleiotropic locus. Mendelian randomization
0098982), which typically functions as an inhibitory syn- analysis could detect causal trait pairs and partially reflect
apse using GABA as neurotransmitter, was significantly en- vertical pleiotropy. Of note, we conducted LDSC to examine
riched in 3 pairs of traits. Notably, both T helper 1 (TH1) and potential sample overlap and performed pleiotropic analysis
TH2 cell differentiation (KEGG has04658) and TH17 cell differ- using PLACO as well as mendelian randomization analysis using
entiation (KEGG has04659), which involve in immunoinflam- LHC-MR to alleviate sample overlap issue. The LDSC and men-
matory responses, were highlighted in 2 pairs of traits. delian randomization analyses were of comparable consis-
tency with previous studies, with detailed comparisons pro-
Multitrait Colocalization Analysis to Pinpoint vided (eTables 21-23 in Supplement 1).
Critical Gut Microbiomes Overall, pleiotropic variants between gastrointestinal tract
Multitrait colocalization analysis from HyPrColoc high- diseases and psychiatric disorders were extensively distrib-
lighted 15 pleiotropic loci (11 unique) that were colocalized to uted, with several loci especially highlighted between cer-
share a causal variant, supporting the significant role of gut tain trait pairs, such as 1q32.1 (INAVA), 19q13.33 (FUT2), 11q23.2
microorganisms (Figure 2 and eTable 17 in Supplement 1). (NCAM1), and 1p32.3 (LRP8). Several loci previously identi-
Seven of 11 unique loci were identified to be colocalized among fied to be associated with gastrointestinal tract diseases were
schizophrenia, gastrointestinal tract diseases, and gut micro- illustrated to be potential pleiotropic loci shared with psychi-
organisms. For example, 19q13.33 was particularly colocal- atric disorders and vice versa (eDiscussion in Supplement 1).
ized between PUD and schizophrenia not only with the preva- The pleiotropic genes were more likely to be enriched in
lence of Ruminococcaceae species but with the abundance of GBA-related phenotypes, such as behavioral, neurological, and
Bacteroides species, where the index SNV rs681343 (mapped digestive phenotypes, as well as GBA-related tissues, espe-
gene: FUT2) was also identified as shared causal variant. cially gastrointestinal and brain tissues.
Shared genetic determinants also reflect common biologi-
Mendelian Randomization and Associations Between cal pathways, among which the TH1, TH2, and TH17 cell differ-
Gastrointestinal Tract Diseases and Psychiatric Disorders entiation pathways were highlighted between IBD and schizo-
Bidirectional mendelian randomization analyses using inverse- phrenia, BIP, and AN. The TH17 cells play important roles in the
variance weighted method showed 11 significant positive development of inflammatory responses and autoimmune
associations (eFigure 5 and eTable 18 in Supplement 1). diseases.56-58 Previous studies suggested the significant role of
In addition, no significant bidirectional associations were dysfunction of TH17 cell differentiation and the accumulation
detected in negative control analysis (eTable 19 in Supple- of TH17-related cytokines in the pathological process of IBD.59,60
ment 1). The results from several alternative mendelian ran- In addition, TH17 cells and related cytokines would provoke
domization methods were largely consistent with that from CNS neuroinflammation and neurotoxicity under pathologi-
main analysis (eTables 18 and 19 in Supplement 1). Notably, cal status and disrupt the blood-brain barrier and thus alter the
among 5 pairwise traits with potential sample overlap, LHC-MR permeability of the blood-brain barrier.61,62 Bacterially pro-
further validated the results (eTable 20 in Supplement 1). duced bile acid metabolites have been shown to inhibit TH17 cell
differentiation, which may be related to the pathophysiology
of IBD.63 Therefore, TH17 cell differentiation and function play
a vital role in the regulation of brain-gut-microbiome axis.
Discussion
In this genome-wide pleiotropic association study, we found Limitations
extensive genome-wide genetic correlations and genetic over- Our study is not without limitations. First, we were unable to
laps between gastrointestinal tract diseases and psychiatric assess the causal effects of gut microbiome on gastrointesti-
disorders. Further comprehensive analyses highlighted the nal tract diseases or psychiatric disorders through mendelian
pleiotropic genetic variants and loci, potential shared causal randomization analyses since SNVs with associated gut mi-
croorganisms were hard to obtain due to the limited sample disorders, we found that pleiotropic genetic variants, loci, and
size of gut microbiome GWAS. Therefore, caution should be genes were extensively distributed across the genome, with
used on the interpretation of the role of gut microorganisms. higher phenotype and tissue specificity. We highlighted some
Second, our study was restricted to European ancestry and may genetic determinants previously associated with gastrointes-
not generalize to other ancestries. tinal tract diseases that were shared with psychiatric disor-
ders and vice versa. More importantly, shared biological mecha-
nisms concerning immune response, synaptic structure and
function, and potential gut microbiome were identified. Our
Conclusions findings not only support the shared genetic basis underly-
Given strong evidence of genetic correlation and genetic over- ing GBA but provide novel insight into the intervention and
lap between gastrointestinal tract diseases and psychiatric treatment targets of these diseases.
randomization between complex traits and metabo-psychiatric origins for anorexia nervosa. identify risk genes using genome-wide association
diseases. Nat Genet. 2018;50(5):693-698. Nat Genet. 2019;51(8):1207-1214. study summary statistics. Bioinformatics. 2021;37
25. Sivakumaran S, Agakov F, Theodoratou E, et al. 36. Winkler TW, Grassmann F, Brandl C, et al. (16):2245-2249.
Abundant pleiotropy in human complex diseases Genome-wide association meta-analysis for early 50. Zhou D, Jiang Y, Zhong X, Cox NJ, Liu C,
and traits. Am J Hum Genet. 2011;89(5):607-618. age-related macular degeneration highlights novel Gamazon ER. A unified framework for joint-tissue
26. Ray D, Chatterjee N. A powerful method for loci and insights for advanced disease. BMC Med transcriptome-wide association and mendelian
pleiotropic analysis under composite null Genomics. 2020;13(1):120. randomization analysis. Nat Genet. 2020;52(11):
hypothesis identifies novel shared loci between 37. Watanabe K, Stringer S, Frei O, et al. A global 1239-1246.
type 2 diabetes and prostate cancer. PLoS Genet. overview of pleiotropy and genetic architecture in 51. Sey NYA, Hu B, Mah W, et al. A computational
2020;16(12):e1009218. complex traits. Nat Genet. 2019;51(9):1339-1348. tool (H-MAGMA) for improved prediction of
27. Hackinger S, Zeggini E. Statistical methods to 38. Amare AT, Vaez A, Hsu YH, et al. Bivariate brain-disorder risk genes by incorporating brain
detect pleiotropy in human complex traits. Open Biol. genome-wide association analyses of the broad chromatin interaction profiles. Nat Neurosci. 2020;
2017;7(11):170125. depression phenotype combined with major 23(4):583-593.
28. Lee CH, Shi H, Pasaniuc B, Eskin E, HanB. depressive disorder, bipolar disorder or 52. Yu G, Wang LG, Han Y, He QY. clusterProfiler:
PLEIO: a method to map and interpret pleiotropic schizophrenia reveal eight novel genetic loci for an R package for comparing biological themes
loci with GWAS summary statistics. Am J Hum Genet. depression. Mol Psychiatry. 2020;25(7):1420-1429. among gene clusters. OMICS. 2012;16(5):284-287.
2021;108(1):36-48. 39. Choquet H, Melles RB, Anand D, et al; 23andMe 53. Foley CN, Staley JR, Breen PG, et al. A fast and
29. Chung D, Yang C, Li C, Gelernter J, Zhao H. GPA: Research Team. A large multiethnic GWAS efficient colocalization algorithm for identifying
a statistical approach to prioritizing GWAS results meta-analysis of cataract identifies new risk loci and shared genetic risk factors across multiple traits.
by integrating pleiotropy and annotation. PLoS Genet. sex-specific effects. Nat Commun. 2021;12(1):3595. Nat Commun. 2021;12(1):764.
2014;10(11):e1004787. 40. Bulik-Sullivan BK, Loh PR, Finucane HK, et al; 54. Burgess S, Butterworth A, Thompson SG.
30. Pardiñas AF, Holmans P, Pocklington AJ, et al; Schizophrenia Working Group of the Psychiatric Mendelian randomization analysis with multiple
GERAD1 Consortium; CRESTAR Consortium. Genomics Consortium. LD score regression genetic variants using summarized data. Genet
Common schizophrenia alleles are enriched in distinguishes confounding from polygenicity in Epidemiol. 2013;37(7):658-665.
mutation-intolerant genes and in regions under genome-wide association studies. Nat Genet. 2015; 55. Darrous L, Mounier N, Kutalik Z. Simultaneous
strong background selection. Nat Genet. 2018;50 47(3):291-295. estimation of bi-directional causal effects and
(3):381-389. 41. Ning Z, Pawitan Y, Shen X. High-definition heritable confounding from GWAS summary
31. Mullins N, Forstner AJ, O’Connell KS, et al; likelihood inference of genetic correlations across statistics. Nat Commun. 2021;12(1):7274.
HUNT All-In Psychiatry. Genome-wide association human complex traits. Nat Genet. 2020;52(8):859- 56. Yang J, Sundrud MS, Skepner J, Yamagata T.
study of more than 40 000 bipolar disorder cases 864. Targeting TH17 cells in autoimmune diseases. Trends
provides new insights into the underlying biology. 42. Watanabe K, Taskesen E, van Bochoven A, Pharmacol Sci. 2014;35(10):493-500.
Nat Genet. 2021;53(6):817-829. Posthuma D. Functional mapping and annotation 57. Korn T, Bettelli E, Oukka M, Kuchroo VK. IL-17
32. Howard DM, Adams MJ, Clarke TK, et al; of genetic associations with FUMA. Nat Commun. and TH17 cells. Annu Rev Immunol. 2009;27(1):
23andMe Research Team; Major Depressive 2017;8(1):1826. 485-517.
Disorder Working Group of the Psychiatric 43. Giambartolomei C, Vukcevic D, Schadt EE, et al. 58. Zhao J, Lu Q, Liu Y, et al. TH17 cells in
Genomics Consortium. Genome-wide Bayesian test for colocalisation between pairs of inflammatory bowel disease: cytokines, plasticity,
meta-analysis of depression identifies 102 genetic association studies using summary and therapies. J Immunol Res. 2021;2021:8816041.
independent variants and highlights the statistics. PLoS Genet. 2014;10(5):e1004383.
importance of the prefrontal brain regions. 59. Rovedatti L, Kudo T, Biancheri P, et al.
44. de Leeuw CA, Mooij JM, Heskes T, Posthuma D. Differential regulation of interleukin 17 and
Nat Neurosci. 2019;22(3):343-352. MAGMA: generalized gene-set analysis of GWAS interferon gamma production in inflammatory
33. Demontis D, Walters RK, Martin J, et al; data. PLoS Comput Biol. 2015;11(4):e1004219. bowel disease. Gut. 2009;58(12):1629-1636.
ADHD Working Group of the Psychiatric Genomics 45. Blake JA, Baldarelli R, Kadin JA, Richardson JE,
Consortium (PGC); Early Lifecourse & Genetic 60. Fujino S, Andoh A, Bamba S, et al. Increased
Smith CL, Bult CJ; Mouse Genome Database Group. expression of interleukin 17 in inflammatory bowel
Epidemiology (EAGLE) Consortium; 23andMe Mouse Genome Database (MGD): knowledgebase
Research Team. Discovery of the first genome-wide disease. Gut. 2003;52(1):65-70.
for mouse-human comparative biology. Nucleic
significant risk loci for attention deficit/ Acids Res. 2021;49(D1):D981-D987. 61. Kebir H, Kreymborg K, Ifergan I, et al.
hyperactivity disorder. Nat Genet. 2019;51(1):63-75. Human TH17 lymphocytes promote blood-brain
46. Pei G, Dai Y, Zhao Z, Jia P. deTS: tissue-specific barrier disruption and central nervous system
34. Nievergelt CM, Maihofer AX, Klengel T, et al. enrichment analysis to decode tissue specificity.
International meta-analysis of PTSD genome-wide inflammation. Nat Med. 2007;13(10):1173-1175.
Bioinformatics. 2019;35(19):3842-3845.
association studies identifies sex- and 62. Huppert J, Closhen D, Croxford A, et al.
ancestry-specific genetic risk loci. Nat Commun. 47. Consortium GT; GTEx Consortium. Cellular mechanisms of IL-17-induced blood-brain
2019;10(1):4558. The Genotype-Tissue Expression (GTEx) project. barrier disruption. FASEB J. 2010;24(4):1023-1034.
Nat Genet. 2013;45(6):580-585. doi:10.1038/ng.2653
35. Watson HJ, Yilmaz Z, Thornton LM, et al; 63. Paik D, Yao L, Zhang Y, et al. Human gut
Anorexia Nervosa Genetics Initiative; Eating 48. Consortium EP; ENCODE Project Consortium. bacteria produce TH17-modulating bile acid
Disorders Working Group of the Psychiatric An integrated encyclopedia of DNA elements in the metabolites. Nature. 2022;603(7903):907-912.
Genomics Consortium. Genome-wide association human genome. Nature. 2012;489(7414):57-74.
study identifies eight risk loci and implicates 49. Gerring ZF, Mina-Vargas A, Gamazon ER,
Derks EM. E-MAGMA: an eQTL-informed method to