Abstract
Published observational studies have revealed the connection between
neurodegenerative disorders and inflammatory bowel disease (IBD),
whereas the causal association remains largely unclear. Our study aims
to assess the causality and identify the shared genetic architecture
between neurodegenerative disorders and IBD. Two-sample Mendelian
randomization analyses were performed to assess the causality between
IBD and neurodegenerative disorders (amyotrophic lateral sclerosis
[ALS], Alzheimer’s disease [AD], Parkinson’s disease [PD], and multiple
sclerosis [MS]). Shared genetic loci, functional interpretation, and
transcriptomic profiles were further investigated in ALS and IBD. We
identified that genetic predisposition to IBD was suggestively
associated with lower odds of ALS (odds ratio [OR] 0.96, 95% confidence
interval [CI] 0.94 to 0.99). In contrast, IBD was not genetically
associated with an increased risk of AD, PD, or MS (and vice versa).
Two shared genetic loci (rs6571361 and rs7154847) were derived, and
SCFD1, G2E3, and HEATR5A were further identified as novel risk genes
with enriched functions related to membrane trafficking. G2E3 was
differentially expressed and significantly correlated with SCFD1 in
patients with ALS or IBD. Our study reveals the suggestively protective
role of IBD on ALS, and does not support the causality of AD, PD, or MS
on IBD (and vice versa). Our findings indicate possible shared genetic
architecture and pathways between ALS and IBD. These results provide
insights into the pathogenesis and therapeutics of IBD and
neurodegenerative disorders.
Keywords: Amyotrophic lateral sclerosis, Neurodegenerative disorder,
Inflammatory bowel disease, Mendelian randomization, Shared genes
__________________________________________________________________
Inflammatory bowel diseases (IBD), mainly comprised of Crohn’s disease
(CD) and ulcerative colitis (UC), are a group of chronic and
relapsing-remitting disorders of the intestine with an undefined
etiology. Numerous extraintestinal and co-morbid conditions are
frequently accompanied by IBD [[45]1]. Neurodegenerative disorders are
debilitating diseases characterized by progressive and selective loss
of function or structure of neuronal systems [[46]2]. Neurodegenerative
disorders include various intractable diseases such as amyotrophic
lateral sclerosis (ALS), Alzheimer’s disease (AD), Parkinson’s disease
(PD), and multiple sclerosis with no curative therapy [[47]3]. These
disorders share an insidious onset and exacerbate irreversibly
throughout the disease course, principally occurring in the aging
population [[48]4].
The gut-brain axis consists of the bidirectional communications between
the intestine and the central nervous system, and such crosstalk
indicates the comorbidities of intestinal inflammation and neurological
degeneration [[49]5]. Researchers identified that IBD increases the
risk of PD in a Danish nationwide cohort involving 7.5 million
individuals during a 37-year follow-up [[50]6]. Another study based on
the American cohorts demonstrates that a higher incidence of PD is
observed among IBD patients [[51]7]. A study based on the Taiwanese
National Health Insurance Research Database indicates that IBD is
associated with an increased risk of the subsequent development of
dementia [[52]8]. The increased prevalence of MS among IBD patients is
also verified by a recent meta-analysis [[53]9]. An increasing amount
of real-life evidence from epidemiology supports their correlations. In
addition, shared mechanisms, for example, autophagy, might be involved
in the pathogenesis of IBD and neurodegenerative diseases including ALS
[[54]10, [55]11].
However, the previously reported linkages between IBD and
neurodegenerative disorders were mainly observational, while the
causality remains largely unexplored. Therefore, it is critical to
comprehensively investigate the causal effects of IBD on
neurodegenerative disorders using a Mendelian randomization (MR)
design.
In this study, the two-sample MR approach was utilized with large-scale
genome-wide association study (GWAS) data to evaluate the potential
causal relationship between liability to IBD and neurodegenerative
disorders, with a main focus on ALS, AD, PD, and MS. Potential genetic
links were explored to further elucidate their correlation, and
transcriptomic profiles of the shared risk loci were further evaluated.
MATERIALS AND METHODS
Data sources
This study relied on publicly available, de-identified and
summary-level data mainly from four large-scale cohorts: studies from
International Inflammatory Bowel Disease Genetics Consortium (IIBDGC)
[[56]12], Neuron [[57]13], European Alzheimer & Dementia Biobank (EADB)
[[58]14], International Parkinson's Disease Genomics Consortium (IPDGC)
[[59]15], and the UK Biobank [[60]16]. The data for IBD (including UC
and CD) from the IIBDGC study were based on 25,042 IBD cases, 12,194 CD
cases, 12,366 UC cases, and approximately 35,000 control subjects. The
ALS GWAS summary data from the study published in Neuron contained
20,806 cases diagnosed by EI Escorial criteria and 59,804 controls. The
data for AD, which contained 39,106 cases and 401,577 controls, were
derived from EADB. The data for PD was based on 33,674 cases and
449,056 control subjects. The data for MS were derived from the UK
Biobank and included 456,348 individuals (775 cases and 455,573 control
subjects). The descriptions of the studies are provided in
[61]Supplementary Table 1.
Statistical analyses
Mendelian randomization analysis
Two-sample MR was performed using the TwoSampleMR package [[62]17]. The
instrumental variables were chosen based on the arbitrary P value
cut-off. A group of single nucleotide polymorphisms (SNPs) with GWAS
significance (P < 5 × 10^-8) associated with each trait were selected.
The SNPs were clumped by linkage disequilibrium (LD) with an r^2 <
0.001 and distance (kb) = 5 000 to ensure that the instruments for the
exposure were independent. F statistics for each instrument were
estimated by F = β^2/SE^2 [[63]18].
For subsequent analysis, inverse variance weighted (IVW) regression was
mainly selected for the inference of causality based on three
assumptions: 1) variants are associated with the exposure; 2) variants
are independent of confounding factors; 3) variants do not directly
affect the outcome [[64]19].
The reverse causality was assessed to evaluate whether neurogenerative
disorders were causally associated with IBD.
Sensitivity analysis
For the traits with an IVW P < 0.05 and SNP number > 2, heterogeneity
tests were performed to evaluate the viability of the effects using
Cochran's Q test. Heterogeneity was considered to exist when Cochran’s
Q test’s P < 0.05 and I^2 > 50%. MR-Egger and weighted-median (WM)
tests were additionally performed to assess the causal effects.
MR-Egger regression is based on the assumption that the pleiotropic
associates are independent, while it could be inaccurate and largely
affected by outlying genetic variants [[65]20]. The WM estimates can
provide valid estimates when ≥ 50% of the weight in the analysis comes
from the SNVs that are valid instrumental variables [[66]20]. MR test
with weighted mode-based estimate (WMBE) was also performed.
Mendelian Randomization Pleiotropy Residual Sum and Outlier (MR-PRESSO)
test was performed to identify horizontal pleiotropic outliers
[[67]21]. Leave-one-out analysis was used for traits with multiple
instruments to assess whether the causality was driven by one single
variant.
The causality was accepted following the criteria described in the
previous publications: IVW was significant and one of the following
assumptions was met: 1) no detected heterogeneity with MR-Egger, WM,
and WMBE in the same direction; 2) heterogeneity was detected, while it
was corrected by MR-PRESSO (<50% of the instruments were considered
outliers); 3) heterogeneity existed with MR-PRESSO test detecting >50%
of outliers, while MR Egger and WM were significant with the same
direction of effect, and WMBE was in the same direction [[68]21,
[69]22].
We used the P value threshold with Bonferroni correction after dividing
0.05 by the number of tests performed. The Bonferroni correction
assumes that independent tests were performed. A P value < 1.25×10^-2
(0.05/4, with Bonferroni correction) was considered statistically
significant and a P value between 1.25×10^-2 and 0.05 was considered
suggestively significant in the MR analyses.
Genetic correlation analysis and genomic control
Genetic covariance analyzer (GNOVA), which can calculate genetic
covariance and estimate genetic correlation according to genetic
covariance and heritability, was used to evaluate the genetic
correlation between IBD and ALS [[70]23]. The reference data originated
from the 1000 Genomes Project European population using default
parameters. GNOVA was considered more powerful than conventional
cross-trait linkage disequilibrium score regression (LDSC) and adopted
as the method for evaluating genetic correlation in our analysis
[[71]23, [72]24]. LDSC was also used to calculate genetic correlations
for the selected traits. ANNOVAR was utilized to annotate SNPs in genic
and intragenic regions [[73]25]. PLINK-clump was used to prune the SNPs
in LD (r^2 > 0.2 within 250 kb) [[74]26].
Risk loci identification
Conditional false discovery rate (cFDR) method was used to identify the
risk loci showing a strong association with IBD and ALS [[75]27]. cFDRs
are characterized as the probability that a certain SNP is falsely
positively correlated with the phenotype that the P values for both
phenotypes (principal and conditional) ≤ the observed P values.
[MATH: cFDRpIBD|p<
/mi>ALS=<
mi>Pr(H<
mrow>0IBD|PIBD≤p<
/mi>IBD,P<
/mi>ALS≤p<
/mi>ALS) :MATH]
As shown above, p[IBD] indicates the observed significance that a SNP
is associated with IBD, and p[ALS] indicates the observed strength of
association that the same SNP is associated with ALS.
[MATH:
H0IBD :MATH]
demonstrates the null hypothesis that a SNP is not correlated with IBD.
The SNPs with FDR < 0.01 were considered significant. Conjunctional
false discovery rate (conjFDR), which was defined as the maximum of the
two cFDR statistics, was further calculated to identify loci that were
associated with both IBD and ALS [[76]27].
Genotyping of the variants and disease risk
The variants were genotyped by the UK Biobank Axiom Array or the UK
BiLEVE Axiom Array [[77]28]. The information was coded as 0, 1, and 2
for noncarriers, heterozygous carriers, and homozygous carriers of the
minor allele, respectively. We explored the associations between the
risk variants and the outcomes (7,400 cases for IBD and 577 cases for
ALS) using logistic regression. The risk estimates were adjusted for
age, sex, Townsend deprivation index, ethnicity, alcohol consumption,
smoking status, metabolic equivalent of task, and body mass index. The
results were presented as adjusted odds ratios and 95% confidence
intervals (CIs).
Functional evaluation
The Genotype-Tissue Expression (GTEx) database was used to evaluate the
normalized effect size (NES) of the single-tissue cis-expression
quantitative trait loci (eQTL) in human tissues [[78]29]. The NES was
computed as the effect of the alternative allele compared to the
reference allele [[79]29]. The NESs were evaluated in different tissues
including adipose, breast, brain, colon, esophagus, heart, lung,
muscle, pancreas, skin, spleen, stomach, thyroid, whole blood, artery,
nerve, etc. Brain eQTL almanac (Braineac), an online dataset containing
data from 10 brain regions obtained from 134 control individuals, was
used to investigate eQTL in brain regions, including the cerebellum,
frontal cortex, hippocampus, medulla, occipital cortex, putamen,
substantia nigra, temporal cortex, thalamus, and white matter [[80]30].
Enrichr ([81]http://amp.pharm.mssm.edu/Enrichr/) was used to assess the
shared pathways associated with IBD and ALS, and 3 ontologic terms
(biological process, cellular component, and molecular function) were
analyzed based on the ontology developed by the Jackson Lab using their
MGI-MP browser [[82]31]. The P value was computed by Fisher's exact
test or the hypergeometric test, and the adjusted P value was
calculated based on the Benjamini-Hochberg method [[83]31]. The details
for the creation of Gene Ontology gene-set libraries for Enrichr are
demonstrated in Lists2Networks [[84]32]. The background gene sets
include the shared risk genes identified using the conjunctional FDR
and eQTL analyses, and genes located within the human leukocyte antigen
(HLA) region were excluded because of the complex LD patterns.
Transcriptomic evaluation of the risk genes
The whole blood expression profiles of ALS patients, IBD patients, and
controls were evaluated for the risk genes derived from the above
analysis using the datasets [85]GSE112680 and E-MTAB-11349 [[86]33,
[87]34]. The [88]GSE112680 dataset contained a transcriptome-wide
analysis of whole blood samples derived from 164 ALS cases and 137
control subjects [[89]33]. The E-MTAB-11349 dataset included 323 blood
samples from IBD patients and 267 blood samples from control subjects
(data released on March 1, 2022) [[90]34]. The Bioconductor lumi
package (v2.44.0) was used for background correction, variance
stabilizing transformation, normalization, and quality control of the
data [[91]35]. Wilcoxon Rank-Sum test with adjustment by
Benjamini-Hochberg method was used to compare the expressions of risk
genes between the diseased group and the control group. The ggstatsplot
package was used for the evaluation of correlation between gene
expression and data visualization [[92]36]. The data were analyzed
using R version 4.1.0 (R Project for Statistical Computing, Vienna,
Austria) or Python 2.7 (Python Software Foundation, Wilmington, US).
RESULTS
Instrumental variable selection
After the clumping process, LD-independent SNPs for IBD were derived,
and the following conditions were applied to further exclude the listed
SNPs: 1) during the extraction of SNPs from the outcomes (ALS, AD, PD,
and MS), a certain requested SNP was not identified and a proxy in LD
was not able to be found from the outcome; 2) no correction could be
performed for ambiguous SNPs or palindromic SNPs with ambiguous
strands. Consequently, the SNPs selected as IV for further analysis
would be included in those listed in [93]Supplementary Table 2. F
statistics for each IV-exposure association were larger than 10
(ranging from 29.86 to 500.60), and therefore the possibility of weak
instrumental variable bias was small in our study.
Two-sample Mendelian randomization analysis for causality of IBD on
neurodegenerative disorders
The causal effects of IBD on the four major neurodegenerative disorders
of interest, including ALS, AD, PD, and MS were explored. [94]Figure 1
and [95]Supplementary Figure 1 summarize the estimates of causal
effects of IBD on the four neurodegenerative disorders, and [96]Table 1
demonstrates the detailed information in addition to the MR-PRESSO
outlier test and the assessment of heterogeneity. Genetically predicted
IBD was suggestively and negatively associated with ALS (IVW [95%
confidence interval [CI]]: 0.96 [0.94-0.99], P = 0.03; all results from
the methods were directionally consistent), and no outliers were
detected by MR-PRESSO ([97]Fig. 1, [98]Supplementary Fig. 1A, [99]Table
1). The estimates of causal effect were also demonstrated in scatter
plots ([100]Fig. 2A). No evidence of confounding heterogeneity was
found by Cochran’s Q test (Q = 113.56, P > 0.10; [101]Table 1) and
leave-one-out test ([102]Supplementary Fig. 2A). Funnel plots, which
provide a visual indication, also demonstrated no heterogeneity for the
causal effect of IBD on ALS ([103]Supplementary Fig. 3A). Our results
indicated no significant evidence of horizontal pleiotropy for the
causality of IBD on ALS (MR-Egger intercept = -0.002, SE = 0.004, P =
0.63; MR-PRESSO global test P = 0.21; [104]Supplementary Table 4).
Figure 1.
[105]Figure 1.
[106]Open in a new tab
The causal effects of inflammatory bowel disease (IBD) on amyotrophic
lateral sclerosis (ALS), Alzheimer’s disease (AD), Parkinson’s disease
(PD), and multiple sclerosis (MS). Error bars represent the 95%
confidence intervals (CIs) for the estimates. CI: confidence interval;
IVW: inverse variance weighted.
Table 1.
Mendelian randomization (MR) analysis for the causality of inflammatory
bowel diseases (IBD) on neurodegenerative disorders.
Mendelian randomization Heterogeneity
Exposure Outcome Method SNPs OR LL UL P Q Q_P
IBD ALS IVW 102 0.96 0.94 0.99 0.03 113.56 0.19
MR Egger 102 0.98 0.92 1.05 0.66 113.30 0.17
Weighted median 102 0.97 0.93 1.01 0.13
Weighted mode 102 0.98 0.92 1.04 0.41
MR PERSSO 102 0.97 0.94 0.99 0.03
AD IVW 79 1.01 0.99 1.02 0.41 54.51 0.98
MR Egger 79 0.99 0.96 1.03 0.60 53.42 0.98
Weighted median 79 1.00 0.98 1.03 0.97
Weighted mode 79 1.00 0.96 1.03 0.85
MR PERSSO 79 1.01 0.99 1.02 0.41
PD IVW 89 1.01 0.98 1.05 0.50 113.84 0.03
MR Egger 89 1.05 0.97 1.15 0.24 112.59 0.03
Weighted median 89 1.01 0.95 1.08 0.69
Weighted mode 89 1.02 0.95 1.09 0.65
MR PERSSO 89 1.01 0.98 1.05 0.50
MS IVW 101 1.04 0.94 1.16 0.43 124.27 0.05
MR Egger 101 0.80 0.62 1.03 0.08 117.98 0.09
Weighted median 101 1.01 0.87 1.17 0.94
Weighted mode 101 0.99 0.78 1.25 0.91
MR PERSSO 101 1.04 0.94 1.16 0.43
[107]Open in a new tab
AD: Alzheimer’s disease; ALS: amyotrophic lateral sclerosis; IBD:
inflammatory bowel disease; IVW: Inverse variance weighted; LL: lower
limits of odds ratio; OR: Odds ratio; PD: Parkinson’s disease; MS:
multiple sclerosis; SNP: single-nucleotide polymorphism; UL: upper
limits of odds ratio.
Figure 2.
[108]Figure 2.
[109]Open in a new tab
Scatter plots of the causal effect of inflammatory bowel disease (IBD)
on neurodegenerative disorders. (A) IBD on amyotrophic lateral
sclerosis (ALS); (B) IBD on Alzheimer’s disease (AD); (C) IBD on
Parkinson’s disease (PD); (D) IBD on multiple sclerosis (MS). The slope
of each line indicates the estimation of effects by each method. SNP:
single nucleotide polymorphism.
In contrast, no causality of IBD on AD (IVW [95% CI]: 1.01 [0.99-1.02],
P = 0.41; [110]Fig. 1, [111]Supplementary Fig. 1B), PD (IVW [95% CI]:
1.01 [0.98-1.05], P = 0.50; [112]Figure 1, [113]Supplementary Fig. 1C)
or MS (IVW [95% CI]: 1.04 [0.94-1.16], P = 0.43; [114]Fig. 1,
[115]Supplementary Fig. 1D) was found. Scatter plots were shown in
[116]Fig. 2B-D. Forest plots, leave-one-out test and funnel plots were
illustrated in [117]Supplementary Fig. 1B-D, [118]2B-D and [119]3B-D,
respectively.
Figure 3.
[120]Figure 3.
[121]Open in a new tab
Functional interpretation and transcriptomic analysis of shared risk
loci. (A) Expression quantitative trait loci (eQTL) of shared genetic
loci and normalized effect size (NES) of eQTLs in different tissues by
GTEx analysis. Blue: negative NES. Red: positive NES. (B) Enriched
pathways related to eQTLs. Deeper color indicates a more significant P
value. (C) Expressions of SCFD1, G2E3, and HEATR5A in different brain
regions by Braineac database. (D) mRNA expressions of G2E3 in patients
with inflammatory bowel disease (IBD) and control subjects (left
panel). Correlations between SCFD1 expression and G2E3 expression in
IBD patients (middle panel) and control subjects (right panel) from
[122]GSE112680. (E) mRNA expressions of G2E3 in patients with
inflammatory bowel disease (IBD) and control subjects (left panel).
Correlations between SCFD1 expression and G2E3 expression in ALS
patients (middle panel) and control subjects (right panel) from
E-MTAB-11349. ALL, average of all regions; ALS: amyotrophic lateral
sclerosis; COPII: coat protein complex II; CRBL, cerebellar cortex; ER:
endoplasmic reticulum; FCTX, frontal cortex; HIPP, hippocampus; IBD:
inflammatory bowel disease; MEDU, medulla; OCTX, occipital cortex;
PUTM, putamen; SNIG, substantia nigra; TCTX, temporal cortex; THAL,
thalamus; WHMT, intralobular white matter. * P < 0.05, ** P < 0.01, ***
P < 0.001.
Two-sample Mendelian randomization analysis for causality of CD and UC on
neurodegenerative disorders
The two major subtypes of IBD, including CD and UC, were further
evaluated for their causal roles in neurodegenerative disorders.
[123]Supplementary Table 3 summarizes the estimates of causal effects
of CD and UC on the four neurodegenerative disorders. Genetically
predicted CD was negatively associated with ALS (IVW [95% CI]: 0.97
[0.95-0.99], P = 0.02; all results from the methods were directionally
consistent), and no outliers were detected by MR-PRESSO
([124]Supplementary Table 3). No evidence of confounding heterogeneity
was found by Cochran’s Q test (IVW Q = 77.55, P > 0.10;
[125]Supplementary Table 3). Our results indicated no significant
evidence of horizontal pleiotropy for the causality of CD on ALS
(MR-Egger intercept = -0.008, SE = 0.005, P = 0.12; MR-PRESSO global
test P = 0.46; [126]Supplementary Table 4). In addition, the causality
of UC on MS was also suggestively significant (IVW [95% CI]: 0.88
[0.78-0.99], P = 0.03; [127]Supplementary Table 3)
In contrast, no causality of UC on ALS, or CD on MS was identified. No
causal relationships were found for CD or UC on AD or PD. The risk
estimates were demonstrated in [128]Supplementary Table 3.
Two-sample Mendelian randomization analysis for causality of
neurodegenerative disorders on IBD
The causal effects of ALS (IVW [95% CI]: 1.00 [0.89-1.12], P > 0.05),
AD (IVW [95% CI]: 0.99 [0.94-1.05], P > 0.05), PD (IVW [95% CI]: 1.03
[0.99-1.08], P > 0.05), or MS (Wald ratio [95% CI]: 0.94 [0.86-1.03], P
> 0.05) on IBD were not significant ([129]Supplementary Table 5). No
causal relationships were found for the above neurodegenerative
disorders on UC or CD ([130]Supplementary Table 5).
Genetic correlation between IBD and ALS and identification of shared risk
loci
A significantly negative genetic correlation between IBD and ALS was
identified (genetic correlation = -0.326, P < 1.00×10^-3 by GNOVA;
genetic correlation = -0.158, P = 4.10×10^-2 by LDSC; [131]Table 2).
Conjunctional FDR analysis was further performed to explore genetic
variants associated with IBD conditional on ALS. Two risk loci were
further identified, namely rs6571361 (P = 2.54×10^-7, FDR = 3.48×10^-2;
SCFD1) and rs7154847 (P = 4.46×10^-7, FDR = 2.70×10^-2; G2E3)
([132]Supplementary Table 6). Compared with variant noncarriers, the
homozygous variant carriers had higher risks of developing ALS
(adjusted OR 1.42 [1.16-1.69], P = 0.03 for rs6571361; adjusted OR 1.48
[1.09-1.77], P = 0.01 for rs7154847; [133]Supplementary Table 7).
Table 2.
Genetic correlation between inflammatory bowel disease and
neurodegenerative disorders.
Genetic Covariance Analyzer
Disease ρ (SE) P Genetic Correlation
IBD-ALS -0.0472 (0.0057) <0.0001 -0.3255
IBD-AD -0.0031 (0.0029) 0.2786 -0.0279
IBD-PD 0.0019 (0.0022) 0.4023 -0.0242
IBD-MS -0.0011 (0.0022) 0.6103 -0.0346
Linkage Disequilibrium Score Regression
Disease Genetic Correlation (SE) P Z-score
IBD-ALS -0.1576 (0.0772) 0.0412 -2.0416
IBD-AD 0.0147 (0.0461) 0.7495 0.3193
IBD-PD 0.0138 (0.0404) 0.7321 0.3424
IBD-MS -0.1624 (0.2798) 0.5616 -0.5804
[134]Open in a new tab
AD: Alzheimer’s disease; ALS: amyotrophic lateral sclerosis; IBD:
inflammatory bowel disease; PD: Parkinson’s disease; SE: standard
error.
Exploration of potential function of shared risk loci
To interpret the function of shared risk loci identified by the
conjunction FDR method, cis- eQTL was evaluated in GTEx database. The
pleiotropic risk loci affect Sec1 Family Domain Containing 1 (SCFD1)
and G2/M phase-specific E3 ubiquitin-protein ligase (G2E3) in tissues
from both GTEx and Brain eQTL almanac (Braineac) ([135]Fig. 3A,
[136]Table 3). In addition, the pleiotropic risk loci can affect HEAT
repeat-containing protein 5A (HEATR5A) in brain tissues ([137]Table 3).
Based on GTEx analysis, the NES of SCFD1 was generally higher in brain
tissues, whereas that of SCFD1 was significantly lower in the digestive
tract, including esophageal mucosa and colon ([138]Fig. 3A). The NES of
G2E3 was only significantly lower in three tissues, including
non-sun-exposed skin, tibial artery, and tibial nerve ([139]Fig. 3A).
Pathway enrichment analysis was further performed to identify
biological pathways identified by cis-eQTL analysis. Twenty pathways of
biological processes were significantly enriched, mainly including
vesicle-mediated transport and autophagy-related pathways (P < 0.05,
[140]Fig. 3B, [141]Supplementary Table 8). For pathways of cellular
components, the Golgi-associated pathways were significantly enriched
(P < 0.01, [142]Fig. 3B, [143]Supplementary Table 8). Syntaxin binding
pathway was significantly enriched for the pathway of molecular
function (P < 0.01, [144]Fig. 3B, [145]Supplementary Table 8).
Furthermore, the expressions of SCFD1, G2E3, and HEATR5A in brain
regions were evaluated using the Braineac database. The expression of
SCFD1 was highest in putamen, while that of SCFD1 was lowest in white
matter (Log2 fold change between expression in putamen/white matter =
1.4, P = 3.0×10^-12; [146]Fig. 3C). G2E3 expression was highest in
putamen, whereas G2E3 expression was lowest in substantia nigra (Log2
fold change between expression in putamen/substantia nigra = 1.9, P =
5.9 × 10^-24; [147]Fig. 3C). The expression of HEATR5A was highest in
white matter. In contrast, the expression of HEATR5A was lowest in
cerebellar cortex (Log2 fold change between expression in putamen/white
matter = 3.0, P = 7.8 × 10^-74; [148]Fig. 3C).
Table 3.
Expression quantitative trait loci (eQTL) indicating the functional
effects of shared risk single nucleotide polymorphisms (SNPs) in human
brain tissue.
Gene Symbol Position Chr start stop P
G2E3 31199109 chr14 31028362 31089250 <0.01
G2E3 31183168 chr14 31028362 31089250 0.01
HEATR5A 31183168 chr14 31760322 31889973 0.01
SCFD1 31183168 chr14 31091515 31223811 0.02
HEATR5A 31183168 chr14 31760322 31889973 0.02
HEATR5A 31183168 chr14 31760322 31889973 0.03
HEATR5A 31183168 chr14 31760322 31889973 0.04
HEATR5A 31199109 chr14 31760322 31889973 0.04
[149]Open in a new tab
Chr: chromosome; SNP: Single nucleotide polymorphism.
Differential expression of the risk genes and correlation of gene expressions
The mRNA expressions of G2E3, SCFD1, and HEATR5A in blood samples of
patients with IBD or ALS were further evaluated. The expressions of
G2E3 in blood samples of patients with IBD were significantly lower
than that of the control (P = 0.0036; [150]Fig. 3D), and positive
correlations between G2E3 expression and SCFD1 expression were observed
in both patients with IBD (Spearman ρ [95% CI] = 0.31 [0.20-0.40], P =
1.92×10^-8[;] [151]Fig. 3D) and control subjects (Spearman ρ [95% CI] =
0.42 [0.31-0.51], P = 1.42×10^-12; [152]Fig. 3D). In contrast, G2E3
expressions in ALS patients were significantly higher than those of the
control group (P = 0.0038; [153]Fig. 3E). Positive correlations between
the expression of G2E3 and SCFD1 were identified in both the ALS
diseased group (Spearman ρ [95% CI] = 0.41 [0.27-0.54], P =
4.1×10^-8[;][154]Fig. 3E) and the control group (Spearman ρ [95% CI] =
0.41 [0.27-0.54], P = 4.9×10^-2[;] [155]Fig. 3E). The expressions of
SCFD1 or HEATR5A were not significantly altered, or slightly altered in
blood samples of patients with IBD or ALS ([156]Supplementary Figure
4A-D). The correlations between expressions of G2E3 and HEATR5A, or
SCFD1 and HEATR5A, were not identified in patients with IBD or ALS, or
controls ([157]Supplementary Fig. 4E-L).
DISCUSSION
By leveraging large GWAS datasets, our study indicates that the
causality of AD, PD or MS on IBD (and vice versa) is unfounded, whereas
the genetic liability to IBD is suggestively protective for ALS. Our
findings provide reassurance for patients suffering from
neurodegenerative disorders and IBD.
ALS is a rapidly progressive neuromuscular disease characterized by
dysfunction of both upper and lower motor neurons, and most ALS
patients die within 3 to 5 years due to respiratory failure [[158]37].
The pathogenesis of ALS is multifactorial and involves complex
interactions among diverse environmental and genetic factors. Defects
in vesicular trafficking and altered neuronal functions are indicated
as one of the pathogenic mechanisms in ALS. For IBD, altered morphology
of vesicles in colonic mucosa cells increases susceptibility to
experimental colitis, and autophagy is regarded as a central issue in
IBD development [[159]10, [160]38]. The shared risk genes could have
functional relevance to both IBD and ALS. SCFD1 plays important roles
in mediating vesicle transport and membrane-fusion events, as well as
autophagy as indicated by the pathway enrichment results [[161]39].
Recent studies have also identified that SCFD1 is one of the most
significant genes that mediate the risk of ALS [[162]40]. Our findings
indicate that the expression of G2E3 is reduced in patients with IBD,
whereas that of G2E3 is increased in patients with ALS, and the
expression of G2E3 in the two diseases was positively correlated with
the expression of SCFD1, which could simultaneously exert its
functions. Therefore, the decreased expression in patients with IBD
might serve as a protective factor for neurodegeneration, contributing
to the casual association of IBD with a decreased risk of ALS.
Currently, little is known about G2E3, and its functions in
inflammatory bowel disease or neurodegenerative disorders remain to be
explored.
Two recent studies tried to evaluate the causality between IBD and
neurodegenerative disorders [[163]41, [164]42]. However, in-depth
evaluation with shared genes was not performed [[165]41, [166]42].
Intriguingly, although previously published research demonstrated
potential causal effect of MS on IBD [[167]43], our study does not
support the causality. Our study also does not support a causal
relationship between IBD and AD or PD, which is contradictory to
previous observational studies [[168]6-[169]9]. Because a high
prevalence of anxiety and depression is observed in patients with IBD,
and unsubstantiated concerns would tremendously aggravate their
psychological comorbidities, which in turn worsen disease outcomes,
unnecessary worries should be avoided for patients [[170]44].
The major strength of our study is the two-sample MR design, which
limits the confounding and reverse-causality bias in observational
studies, with the use of strong instruments. Our results are robust and
with no significant evidence of horizontal pleiotropy. In addition, we
identified novel SNPs associated with ALS and IBD, which provide
plausible explanations for their correlation and deepen current
understandings of the disorders. The shared risk genes were validated
by transcriptomic analysis. Our study has several limitations. Our
study merely involves the European population, which limits the
universality of our findings to other ancestries. Due to the lack of
publicly available GWAS summary data or rarity in nature, the causality
of IBD on other neurodegenerative disorders was not assessed. Moreover,
although the significance of the causality for IBD on ALS is only
suggestive, Bonferroni correction can be considered overly conservative
[[171]45], and further investigations with larger and more powerful
datasets are warranted.
In summary, the causality of AD, PD, or MS on IBD (and vice versa) is
not supported by current evidence, whereas IBD is suggestively
protective to ALS. We suggest the possible genetic architecture shared
between ALS and IBD and indicate that modulation of membrane
trafficking-related pathways and dysregulation of G2E3 might contribute
to their pathogenesis.
Supplementary Materials
The Supplementary data can be found online at:
[172]www.aginganddisease.org/EN/10.14336/AD.2022.1209. GWAS summary
statistics are available from the original manuscript of each study in
Supplementary Table 1 and GWAS Catalog
([173]https://www.ebi.ac.uk/gwas/). Code used in this study is
available from the corresponding authors upon reasonable request.
[174]AD-14-4-1349-s.pdf^ (1.3MB, pdf)
Acknowledgements