Abstract
Vertigo is a leading symptom of various peripheral and central
vestibular disorders. Although genome-wide association studies (GWASs)
have identified multiple risk variants for vertigo, how these risk
variants contribute to the risk of vertigo remains unknown. Discovery
proteome-wide association study (PWAS) was first performed by
integrating the protein quantitative trait loci from the dorsolateral
prefrontal cortex (DLPFC) in the Banner Sun Health Research Institute
dataset (n = 152) and GWAS summary of vertigo (n = 942 613), followed
by replication PWAS using the protein quantitative trait loci from the
DLPFC in Religious Orders Study or the Rush Memory and Aging Project
dataset (n = 376). Transcriptome-wide association studies (TWASs) were
then performed by integrating the same GWAS datasets of vertigo (n =
942 613) with mRNA expression reference from human fetal brain, and
DLPFC. Chemical-related gene set enrichment analysis (GSEA) and Gene
ontology/Kyoto Encyclopedia of Genes and Genomes pathway enrichment
analyses were finally conducted to further reveal the pathogenesis of
vertigo. Permutation-based empirical P values were calculated in PWAS,
TWAS, and GSEA. By integrating the GWAS of vertigo and two independent
brain proteomes from human DLPFC, three genes were identified to
genetically regulate protein abundance levels in vertigo, and were not
previously implicated by GWAS, including MTERFD2 (P[Banner] = 0.045,
P[ROSMAP] = 0.031), MGST1 (P[Banner] = 0.014, P[ROSMAP] = 0.018), and
RAB3B (P[Banner] = 0.045, P[ROSMAP] = 0.035). Compared with TWAS
results, we identified overlapping genes RAB3B (P[TWAS] = 0.017) and
MTERFD2 (P[TWAS] = 0.003) that showed significant associations with
vertigo at both proteome-wide and transcriptome-wide levels.
Chemical-related GSEA identified multiple chemicals that might be
associated with vertigo, such as nickel (P = 0.007), glycidamide (P =
0.005), and proanthocyanidins (P = 0.015). Our study provides novel
clues for understanding the biological mechanism of vertigo, and
highlights several possible risks and therapeutic chemicals for
vertigo.
Keywords: vertigo, proteome-wide association study, transcriptome-wide
association study, chemical, gene set enrichment analysis
__________________________________________________________________
Cheng et al. performed proteome-wide association studies to identify
genes whose cis-regulated protein abundance changes in the human brain
were associated with vertigo, followed by chemical-related gene set
enrichment analysis. They provide novel genes (MTERFD2, MGST1, and
RAB3B) into the pathogenesis of vertigo, and highlight promising
chemicals for further therapeutics research.
Graphical Abstract
Graphical abstract.
[38]Graphical abstract
[39]Open in a new tab
Introduction
Vertigo is a subtype of dizziness defined as the illusion of motion
caused by the asymmetrical intervention of the vestibular system.^[40]1
Vestibular central lesions affecting the pons, medulla, or cerebellum
may cause vertigo, severe ataxia, vomiting, nausea, and other
neurological signs.^[41]1 Conventionally, the diseases causing vertigo
can be classified into three broad categories: otological vertigo,
central vertigo, and psychogenic dizziness.^[42]2,[43]3 Common causes
of vertigo include benign paroxysmal postural vertigo, migraine,
vestibular neuritis, Ménière's syndrome, adverse drug effects, and
disturbed blood pressure regulation.^[44]4 The prevalence of vertigo is
6.5% and increases with age, with approximately 65% of patients being
female^[45]5 and a lifetime prevalence of about 20–30%.^[46]6 Vertigo
can come on suddenly and last for a few seconds or maybe constant for
several days, which is a major risk factor for falls and bone
fractures, placing a huge burden on the healthcare system.^[47]7
Several common vertigo syndromes are known to be genetically
heterogeneous and have been identified in patients with isolated
recurrent attacks of vertigo, genetic deafness syndromes, and
neurological disorders.^[48]8 Genetic studies have identified several
mutations in the KCNA1 and CACNA1A genes associated with recurrent
vertigo, suggesting that voltage-gated channels and solute carriers in
the plasma membrane of neurons play a key role in recurrent
vertigo.^[49]9 A genome-wide association study (GWAS) also uncovered
six sequence variations associated with vertigo risk.^[50]10 In
addition, the central nervous system (CNS) plays an important role in
the pathogenesis of vertigo.^[51]11 Interference with protein–protein
interaction networks in neurons, glia, and other cell types has also
been found to be associated with multifactorial neurological
disorders.^[52]12 Although previous research works have focused on
transcription as a central regulator of protein expression, it is now
increasingly recognized that the CNS relies on efficient updating of
the protein landscape.^[53]13 However, the characteristics of brain
proteins and genetics of vertigo remain unclear.
GWAS seeks to strongly link genetic loci to diseases and other
heritable traits, but the statistical power is limited.^[54]14 Linkage
disequilibrium (LD) and population stratification make it convoluted
for GWAS to identify exact causal variants.^[55]15 The
transcriptome-wide association study (TWAS) tests whether the phenotype
being studied correlates with the level of gene expression predicted
from genetic variants, usually requiring a target tissue for
association tests.^[56]16 The TWAS model simulates gene expression,
while the proteome-wide association study (PWAS) model simulates
protein abundance, and the principle is completely orthogonal to gene
abundance signals.^[57]16 As a new method to detect gene-phenotype
associations mediated by changes in protein abundance, PWAS aggregates
all variants’ signals that jointly affect protein-coding genes,
reducing the burden of multiple tests and providing more interpretable
discoveries.^[58]17 For example, Liu et al. performed PWAS to identify
genes associated with changes in cis-regulatory protein abundance in
the human brain and found several high-confidence risk proteins
(including CNNM2 and CTNND1) for schizophrenia and depression.^[59]18
Another study performed PWAS by integrating depression GWAS and human
brain proteomics, followed by Mendelian randomization, and identified
19 genes consistent with depression causality.^[60]19
Human brain development is disrupted by chemical exposure, leading to
irreversible damage to the nervous system.^[61]20 Conversely, some
chemicals have positive effects on the CNS.^[62]21 Recent studies have
found that chemical exposure also affects epigenetic
inheritance.^[63]22 For example, prenatal exposure to
endocrine-disrupting chemicals might lead to poor behaviour and
cognitive dysfunction in children.^[64]23 The contribution of chemicals
to brain development is multifaceted and complex, and chemicals
associated with the endocrine activity are suspected to interfere with
neurodevelopment.^[65]20 In addition, neural tube and axial defects
were frequently found when evaluating the developmental effects of
chemicals in laboratory animal studies.^[66]24 Therefore, it is
necessary to identify the chemicals related to the pathogenesis of
vertigo.
In this study, PWASs were performed on vertigo by integrating a
large-scale GWAS and two independent human brain protein quantitative
trait loci (pQTL) datasets. Discovery PWAS was first conducted using
the pQTL from the dorsolateral prefrontal cortex (DLPFC) in the Banner
dataset. Replication PWAS was then performed using the pQTL from DLPFC
in the ROSMAP dataset. To validate the results of PWASs, we also
performed TWASs by integrating the same GWAS datasets and mRNA
expression reference from the human fetal brain, and DLPFC in which
gene expression was quantified with RNA-seq and RNA-seq splicing,
respectively. Chemical-related gene set enrichment analysis (GSEA) and
Gene ontology (GO)/Kyoto Encyclopedia of Genes and Genomes (KEGG)
pathway enrichment analyses were finally conducted to further reveal
the pathogenesis of vertigo. Our study identified risk genes whose
protein abundance is associated with vertigo, providing novel insights
into further mechanistic studies and the development of new therapies.
Materials and methods
GWAS summary data of vertigo
Genome-wide association summary level data used in this study was
derived from a recent publicly available large-scale GWAS of human
vertigo.^[67]10 Briefly, this dataset had 48 072 vertigo cases and 894
541 controls in the vertigo meta-analysis, all of whom were verified as
being of white origin.^[68]10 The deCODE was used to perform the sample
preparation and the whole-genome sequencing (WGS) in Iceland
samples.^[69]25 The Affymetrix chip UK BiLEVE Axiom and the Affymetrix
UK Biobank Axiom array were used to genotype the UK Biobank
samples.^[70]26 NovaSeq Illumina was used for WGS in US samples, and
Illumina Global Screening Array chips were used for genotyping. FinnGen
ThermoFisher Axiom array was used to genotype FinnGen samples. The
logistic regression assuming an additive model was applied and tested
for association between sequence variants and vertigo using deCODE
software.^[71]25 The fixed-effects inverse variance method based on
effect estimates and standard errors was used to combine vertigo GWAS
summary results from Iceland, the UK, Finland, and the US. This method
assumes that all study groups have a common OR, but allows for
different population frequencies for alleles and genotypes.^[72]27 In
total, 62 056 310 variants in the meta-analysis had imputation
information above 0.8 and MAF > 0.01%. The genotyping, quality control,
imputation, and association analysis of the dataset have been described
in detail elsewhere.^[73]10
Human brain pQTL datasets for protein reference weights
Human brain proteomes datasets were obtained from a recent publicly
available study.^[74]28,[75]29 Briefly, a proteome analysis was
performed using brain DLPFC of 376 European descent subjects collected
by the Religious Orders Study (ROS) or the Rush Memory and Aging
Project (ROSMAP dataset). After quality control (QC), 8356 proteins
were included in proteomic profiles for pQTL analysis. By integrating
proteomic data and SNP genotypes, 1475 proteins were significantly
heritable to genetic variation, and their protein weights were used in
our PWAS. Besides, another proteome analysis was performed using brain
DLPFC of 152 European descent subjects from the Banner Sun Health
Research Institute (Banner dataset).^[76]28 After QC, 8168 proteins
were included in proteomic profiles for pQTL analysis. By integrating
proteomic data and SNP genotypes, 1139 proteins were significantly
heritable to genetic variation, and their protein weights were used in
our PWAS. The brain protein weights from the ROSMAP and Banner brain
proteome can be downloaded from
[77]https://doi.org/10.7303/syn23627957.^[78]28 Detailed information
for the brain protein weights has been described in the study of Wingo
et al.^[79]28
PWAS for vertigo
The PWAS analysis of vertigo was performed by using FUSION software
([80]http://gusevlab.org/projects/fusion/).^[81]16 Using the
pre-computed reference weights of the ROSMAP and Banner brain proteome
together with GWAS summary data of vertigo, FUSION is capable to
estimate the associations of each protein abundance gene with vertigo.
Firstly, the reference expression weights were calculated using the
prediction models in FUSION. The calculated expression weights were
then combined with GWAS results to impute association statistics
between protein abundance gene expression levels and vertigo. Bayesian
sparse linear mixed model (BSLMM) was used to compute the
SNP-expression weights in the 1-Mb cis loci of the gene for a given
gene.^[82]30 The association test statistics between predicted gene
expression and vertigo were calculated as Z[PWAS] =
W′Z/(W′SW)^1/2.^[83]16Z denotes the scores of vertigo, while W denotes
the weights. S denotes the SNP-correlation covariance matrix. In this
study, a total of 5000 permutation tests in FUSION were implemented to
control the potential impact of multiple test problems. The
permutation-based empirical P value was calculated for each gene within
the ROSMAP and Banner brain proteome. Significant protein abundance
genes were considered at permutated P < 0.05.
TWAS for vertigo
The TWAS analyses were performed by using FUSION software
([84]http://gusevlab.org/projects/fusion/).^[85]16 Using the
pre-computed gene expression weights of different tissues together with
GWAS summary data of vertigo, FUSION is capable to estimate the
associations of each gene with vertigo in different tissues. In this
study, TWAS analysis was performed based on the GWAS summary statistics
of vertigo and pre-computed expression reference panel in TWAS included
the splicing-level (splicing QTL, sQTL) and mRNA-level (eQTL) weights
from CommonMind Consortium (CMC) in DLPFC profiles,^[86]31 and fetal
brain expression (eQTL) weights.^[87]32 The only difference between
PWASs and TWASs is the different external presentation reference
panels. PWAS is based on protein abundance, while TWAS is based on mRNA
expression level. Briefly, the gene expression weights were calculated
using the prediction models of FUSION, respectively. The calculated
expression weights were then combined with GWAS summary statistics to
impute association statistics between gene expression levels and
vertigo. BSLMM was used to compute the SNP-expression weights in the
1-Mb cis loci of the gene for a given gene.^[88]30 Multiple testing
problem of each gene was computed using 5000 permutation tests. Genes
with significant correlation signals were identified at permutated P <
0.05.
Colocalization analysis
Colocalization analysis was performed using Fusion software with
parameter—coloc P 0.05, which indicated that only genes with P < 0.05
were included to perform colocalization analysis.^[89]16 Colocalization
of GWAS and pQTL (ROSMAP and Banner datasets),^[90]28 eQTL (CMC and
fetal brain dataset),^[91]32 and sQTL (CMC dataset)^[92]31 signals were
performed to explore the same risk variants in vertigo.
Statistical analysis
The chemical-related gene set was obtained from the public Comparative
Toxicogenomics Database (CTD) which consisted of 1 379 105
chemical–gene interactions ([93]http://ctdbase.org/).^[94]33 CTD is an
innovative digital ecosystem that relates toxicological information for
chemicals, genes, phenotypes, diseases, and exposures to advance the
understanding of human health.^[95]34 The GSEA (4.2.3) was implemented
to evaluate the functional relevance between 10 103 chemicals and
vertigo.^[96]35 Briefly, GSEA was first applied to the PWAS-level data
of vertigo to determine whether vertigo-associated protein abundance
genes were significantly enriched in the chemical-related gene
sets.^[97]35 Following the standard GSEA approach, we set GS[m] as the
PWAS statistic. All genes were then ordered by ranking GS[m] from the
largest to the smallest, set as GS[m] = (GS[m1], GS[m2], …GS[mN]). We
set G[i] as the ith gene in chemical-related gene set C with N[C]
genes. A weighted Kolmogorov-Smirnov-like running sum statistic was
used to calculate the enrichment scores (ES[S]) of each analyzed
chemical.^[98]35
[MATH: ESS=max1≤j≤<
/mo>N{∑Gi∈C,i≤j<
/munder>|GSm
i|HN
R
−∑Gi∉C,i≤j<
/munder>1N−
NC}<
/mo> :MATH]
where N[R] defined as
[MATH: NR=∑Gi∈
C|GSm
i|H :MATH]
; N as the total number of genes; H as the parameter that gives higher
weights to genes with extreme statistics. We followed the
recommendation that using H = 1 for the original GSEA algorithm. The
TWAS-level data were then utilized to perform the same GSEA algorithm.
Following the normalization procedure, statistical test was performed
to calculate the null distribution of ES[S], with SNP labels of vertigo
GWAS randomly shuffled and then used for ES calculation.^[99]35 The
null distribution of ES^n (ES^n1, ES^n2, ES^n3, …, ES^nt) was obtained
after t time permutations. To adjust the bias caused by different sizes
of chemical-related gene sets, the observed ES[S] was normalized by the
standard deviation (Sd) and mean of permutated ES^n:
[MATH:
NESS=ESS
−mean(ESSn)SD(ESSn) :MATH]
Each enrichment analysis implemented 20 000 permutations. The
permutation-based empirical P values were finally calculated using the
NES for each chemical-related gene set. Significant enrichment was
identified at P < 0.05. Detailed information of permutation and
statistical analysis were described elsewhere.^[100]36
Gene ontology and pathway enrichment analysis
To determine biological features and enriched pathways for vertigo, the
gene ontology (GO) and pathway enrichment analysis were performed on
DAVID (Dec.2021),^[101]37 STRING (11.5),^[102]38 and Reactome (Version
81)^[103]39 using the significant genes identified by PWAS and TWAS
(P[PWAS] < 0.05 or P[TWAS] < 0.05), including pQTL PWAS using Banner
datasets,^[104]28 eQTL TWAS using CMC and fetal brain dataset,^[105]32
and sQTL TWAS using CMC dataset.^[106]31 All enrichment analyses were
false discovery rate (FDR) controlled using the Benjamini-Hochberg
method.^[107]40
Data availability
Detailed description of the datasets used in our study are shown in
Supplementary Material. The datasets generated and/or analyzed in the
current study are available from the corresponding authors upon
reasonable request.
Results
PWASs identified three genes that regulate protein levels in the brain and
were linked to risk of vertigo
Discovery PWAS was first performed by integrating the Banner human
brain pQTL data and GWAS of vertigo and identified 22 significant genes
for vertigo ([108]Table 1). To validate these results, replication PWAS
was then conducted using the ROSMAP human brain pQTL dataset and the
same GWAS of vertigo and identified 13 significant genes for vertigo
([109]Table 1). Importantly, three common vertigo risk genes showed
proteome-wide significant signals in both discovery and replication
stages, including MTERFD2 (P[Banner] = 0.045, P[ROSMAP] = 0.031), MGST1
(P[Banner] = 0.014, P[ROSMAP] = 0.018), and RAB3B (P[Banner] = 0.045,
P[ROSMAP] = 0.035) ([110]Table 1). The original PWAS results based on
the Banner and ROSMAP human brain proteomes are presented in [111]Fig.
1.
Table 1.
Proteome-wide significant genes for vertigo
Gene CHR Banner ROSMAP
Z[PWAS] P [PERM] Z[PWAS] P [PERM]
ASPHD1 16 4.967 2.00 × 10^−4
PPP2R4 9 −2.321 4.00 × 10^−4
MCEE 2 −2.427 0.005
PTGR1 9 −2.285 0.007
TMEM132E 17 −3.173 0.007
SHISA7 19 −2.441 0.010
SLC27A1 19 2.491 0.010
RIMS3 1 2.430 0.011
MSTO1 1 −1.97 0.011
ACOT11 1 2.393 0.0119
NOSTRIN 2 2.838 0.013
CCDC91 12 7.986 0.013
CDH7 18 2.376 0.014
MGST1 ^[112]a 12 2.279 0.014 2.256 0.018
S100A1 1 2.779 0.019
C12orf73 12 −2.90 0.019
DHRS4 14 2.638 0.020
SPRYD4 12 1.971 0.023
AGA 4 −2.357 0.025
C2orf76 2 2.181 0.026
MTERFD2 ^[113]a 2 −2.999 0.045 −2.999 0.031
PAK4 19 2.147 0.032
GPX1 3 −2.267 0.035
RAB3B ^[114]a 1 2.184 0.045 2.233 0.035
COQ10A 12 4.018 0.039
EXOG 3 2.067 0.0393
IPCEF1 6 −2.335 0.0396
PPFIA3 19 −2.583 0.0396
IFIT2 10 −2.044 0.042
ADSL 22 2.165 0.044
[115]Open in a new tab
CHR = chromosome; PERM = permutation; PWAS = proteome-wide association
study; ROSMAP = Religious Orders Study and Rush Memory and Aging
Project.
^a
Genes that reached proteome-wide significant level in both the Banner
and ROSMAP brain proteomes.
Figure 1.
[116]Figure 1
[117]Open in a new tab
Manhattan plots for the vertigo PWASs in the Banner and ROSMAP human
brain proteomes. (A) Manhattan plot for the vertigo PWAS integrating
the vertigo GWAS with the Banner proteomes (n = 152). (B) Manhattan
plot for the vertigo PWAS integrating the vertigo GWAS with the ROSMAP
proteomes (n = 376). Each dot on the x-axis represents a gene, and the
association strength on the y-axis represents the –log[10](original P
value) of PWAS. The heat-map scale bar indicates the distribution
density of analyzed protein regulated genes on the chromosome. Genes
that were proteome-wide significant using permutation test (MTERFD2,
MGST1, and RAB3B) in Banner and ROSMAP brain proteomes are shown with
mark. Chr = chromosome; PWAS = Proteome-Wide Association Study; ROSMAP
= Religious Orders Study and Rush Memory and Aging Project.
TWASs identified eight genes that regulate gene levels in the brain and were
linked to risk of vertigo
The eQTL-based TWAS identified 99 cis-regulated expression genes
associated with vertigo, such as RAB3B (P[TWAS] = 0.017), TMEM132E
(P[TWAS] = 0.002), and TCEA3 (P[TWAS] = 0.037) ([118]Supplementary
Table 1). The sQTL-based TWAS identified 148 genes associated with
vertigo, such as MTERFD2 (P[TWAS] = 0.003), C12orf73 (P[TWAS] = 0.027),
and SNRPC (P[TWAS] = 0.003) ([119]Supplementary Table 2). Notably,
eight candidate common genes were shared in eQTL-based and sQTL-based
TWAS, such as PIGG (P[eQTL] = 0.011, P[sQTL] = 0.009), ZSWIM7 (P[eQTL]
= 0.023, P[sQTL] = 0.017), and PMS2P5 (P[eQTL] = 0.033, P[sQTL] =
0.031) ([120]Supplementary Tables 1 and 2).
Fetal brain-based TWAS identified 20 cis-regulated expression genes
associated with vertigo, such as MXRA7 (P[TWAS] = 0.006), MSH3 (P[TWAS]
= 0.018), and DHFR (P[TWAS] = 0.038) ([121]Supplementary Table 3).
Interestingly, no common gene in fetal brain-based TWAS was shared with
eQTL and sQTL-based TWAS, which might highlight the differences for
vertigo mechanism in RNA level between adult and fetus.
The comparison of PWAS and TWAS highlighted two risk genes for vertigo
PWAS and TWAS highlighted several promising candidate genes for
vertigo, and a series of comparisons thus to be performed between PWAS
and TWAS results to explore the overlapped risk genes at the protein
and RNA level, including eQTL-based TWAS, sQTL-based TWAS, and fetal
brain-based TWAS. Proteome-wide candidate risk gene RAB3B was
significantly overlapped in eQTL-based TWAS (P[TWAS] = 0.017). MTERFD2
was statistically significant between PWAS and sQTL-based TWAS result
(P[TWAS] = 0.003). However, no proteome-wide significant genes were
supported by fetal brain-based TWAS results.
Colocalization analysis of common vertigo risk genes between PWAS and TWAS
Colocalization analysis based on the PP4 hypothesis which represents
the posterior probability for coloc colocalization analysis hypothesis
4, that is, eQTL and GWAS signals are driven by the same common causal
variant. The results indicated that the GWAS and pQTL signals were
driven by the same risk variants, including RAB3B (PP4[banner] = 0.007,
PP4[ROSMAP] < 0.001, PP4[RNAseq] = 0.007), and MTERFD2 (PP4[banner] =
0.049, PP4[ROSMAP] = 0.058, PP4[splicing] = 0.006). In particular,
RAB3B was detected to be associated with vertigo in two independent
PWASs and eQTL-based TWAS, which strongly suggested that RAB3B might be
a vital risk gene for vertigo.
Chemical-related GSEA implicates potential pathogenesis for vertigo
As most chemicals have pathogenic effects in disease, we explored
whether the chemicals in CTD can serve as potential pathogenic factors
for vertigo. For the Banner PWAS gene list, 15 candidate chemicals were
observed, such as glycidamide (P = 0.005), enzalutamide (P = 0.004),
and dihydrotestosterone (P = 0.007). For the ROSMAP PWAS gene list,
seven candidate chemicals were observed, such as nickel (P = 0.007),
proanthocyanidins (PAC; P = 0.015), and irinotecan (P = 0.025)
([122]Table 2). For the TWAS gene list, 22, 32, and 31 candidate
chemicals were significantly enriched in fetal brain, RNA-seq and
splicing data, respectively. Notably, three common chemicals overlapped
in RNA-seq and splicing TWAS results, including fipronil,
phenylpropanolamine, and fenbuconazole ([123]Supplementary Table 4).
Table 2.
Chemical-related gene set enrichment analysis for vertigo based on PWAS
results
Chemical name Banner ROSMAP
NES P NES P
Glycidamide 2.482 0.005 −1.926 0.974
Aldehyde 1.619 0.049 0.552 0.296
Menthol 1.725 0.041 0.423 0.336
Nimesulide 1.709 0.040 −0.183 0.574
Sirolimus 1.779 0.038 1.228 0.115
Enzalutamide 2.397 0.004 −0.160 0.573
Tunicamycin 1.744 0.041 −0.395 0.649
Vitallium 1.910 0.027 1.354 0.089
Dimethyloctadecan 1.739 0.041 −0.180 0.570
Belinostat 1.908 0.026 −0.096 0.535
Carcinogens 2.108 0.012 0.303 0.388
Carmustine 1.762 0.039 0.563 0.287
Clothianidin 2.130 0.015 0.392 0.351
DEET 2.100 0.011 0.766 0.228
Dihydrotestosterone 2.362 0.008 0.648 0.260
Gasoline −0.020 0.507 1.743 0.040
Proanthocyanidins −0.191 0.574 2.160 0.015
GSK-J4 1.012 0.156 2.180 0.015
Irinotecan −0.652 0.738 1.954 0.025
Nickel 0.001 0.497 2.421 0.007
Oxazolone 0.086 0.475 1.806 0.032
Benzo(a)pyrene −1.623 0.949 1.672 0.047
[124]Open in a new tab
Dimethyloctadecan = 2-amino-14,16-dimethyloctadecan-3-ol; PWAS =
proteome-wide association study; ROSMAP = Religious Orders Study and
Rush Memory and Aging Project; NES = Normalized Enrichment Score.
GO and pathway enrichment analysis
GO annotation and pathway analysis were accomplished to further reveal
the pathogenesis of vertigo. A total of 15 significant terms were
identified by using significant genes from PWAS and TWAS. For example,
CNS (P[FDR] = 2.84 × 10^−7) and brain stem (P[FDR] = 0.039) in CMC
splicing dataset-based TWAS, mitochondrion (P[FDR] = 3.9 × 10^−4) in
Banner dataset-based PWAS, cognitive function measurement (P[FDR] =
0.023) and serine metabolism (P[FDR] = 0.012) in fetal brain
dataset-based TWAS ([125]Supplementary Table 5).
Discussion
In this study, we performed PWASs for vertigo by integrating human
brain pQTL data and genome-wide associations. We identified three genes
that regulate protein levels in the human brain that were associated
with risk of vertigo by combining the results from the discovery and
replication PWASs. Among the three candidate genes, RAB3B and MTERFD2
were also validated in TWAS, strongly emphasizing the consistent
findings at mRNA and protein levels. Chemical-related GSEA and
GO/pathway enrichment analysis highlights that the pathogenesis of
vertigo may be related to neuronal exocytosis and CNS development.
Our study suggesting that RAB3B and MTERFD2 are highly risk genes whose
expression and protein abundance are significantly associated with
vertigo. RAB3B encodes a low molecular weight GTP-binding protein,
which is involved in the exocytosis of synaptic vesicles and secretory
granules in the CNS and the anterior pituitary cells.^[126]41 In
addition, RAB3B may be involved in polarized transport of basolateral
and tight junctional membrane proteins to the plasma membrane and the
regulation of synaptic plasticity.^[127]42,[128]43RAB3B has been
confirmed to be involved in neurotransmitters and synapses, pituitary
exocytosis, and immune function. In mouse models, Rab3B is required for
long-term depression of inhibitory synapses in the hippocampus,
short-term plasticity, and normal reverse learning.^[129]44 As a key
intracellular signalling molecule, RAB3A is highly expressed in the
brain and can control the downstream exocrine secretion of other
calcium-dependent processes in anterior pituitary cells.^[130]45 RAB3B
immunoreactivity plays an important role in human pituitary
adenoma,^[131]46 possibly reflecting the number of secretory granules
and exocytosis activity.^[132]47 In addition, GO and KEGG pathway
analysis also found that vertigo was involved in the biological
function of neurons. For example, regulation of short-term neuronal
synaptic plasticity in Banner PWAS, cell proliferation in the midbrain,
commissural neuron axon guidance in RNA-seq TWAS, neurotransmitter
secretion, and GABAergic synapse in splicing TWAS.
MTERFD2, also known as MTERF4, is another vertigo risk gene identified
in this study that belongs to the component of the mitochondrial
transcription termination factor (MTERF) family. Colocalization of
EGFP-MTERFD2 fusion protein suggested that MTERFD2 targeting
mitochondria exhibit a dynamic expression pattern during embryogenesis
and might play an important role in organ differentiation.^[133]48
MTERFD2 has been shown to recruit ribosomal RNA through NSUN4 to
regulate ribosomal biogenesis.^[134]49 It was found that MTERFD2 also
plays a vital role in mitochondrial regulation and neurodegenerative
disease. Ye et al. found that the overexpression of MTERFD2 in SH-SY5Y
cells partially alleviated 1-methyl-4-phenylpyridinium induced
mitochondrial dysfunction, and proposed that MTERFD2 might be the
trigger factor of the pathogenesis of Parkinson’s disease induced by
environmental toxins.^[135]50 Wang et al. emphasized that MTERFD2 plays
a key role in the pathogenesis of Alzheimer’s disease by inhibiting
ADAM10 in HEK293-APPswe cells and promoting amyloidogenic processing of
β-amyloid precursor protein.^[136]51 Besides, GO annotation also found
that vertigo is involved in the biological function of mitochondrial
regulation, such as mitochondrion in Banner PWAS. Among the three genes
overlapped in Banner and ROSMAP PWAS, RAB3B and MTERFD2 were also
verified by TWAS to be associated with vertigo at mRNA-level,
emphasizing that genetic risk variants likely confer risk of vertigo by
regulating messenger RNA expression and protein abundance of these
genes.
In particular, MGST1 and SPRYD4 were only recognized by the PWAS rather
than TWAS, highlighting the additional insights provided by focusing
brain proteins directly. MGST1 protein is ubiquitous in human tissues
and cell lines and is located in the endoplasmic reticulum and outer
mitochondrial membrane, playing a protective role against membrane
oxidative stress.^[137]52 It was found that Mgst1 deletion in mice
resulted in embryonic lethal and impaired haematopoietic function,
emphasizing that Mgst1 is essential for embryonic development and
haematopoietic function in vertebrates.^[138]53SPRYD4 has not been
widely explored, but it has been recently found to have anti-cancer
effects, such as acute myeloid leukemia^[139]54 and hepatocellular
carcinoma.^[140]55 In our GO annotation, glutathione peroxidase
activity was also found to involve in vertigo by Banner PWAS. Notably,
all three proteome-wide significant genes were not specified by the
original vertigo GWAS, highlighting the ability of PWAS in exploring
vertigo risk genes.
Chemical-related GSEA identified multiple neurotoxic chemicals that may
be associated with vertigo risk, such as nickel, glycidamide, fipronil.
Nickel is a potential risk chemical for vertigo identified by our
ROSMAP PWAS and has also been shown to cause vertigo, with toxicity
manifesting in affecting T cell systems and inhibiting the activity of
natural killer cells.^[141]56 Thus, environmental and occupational
exposure to nickel is a potential risk factor for human brain
dysfunction and neurological symptoms.^[142]57 In addition, GSEA based
on TWAS results of RNA-seq and splicing jointly identified a potential
risk chemical, fipronil, which is a member of the phenylpyrazole class
of insecticide and has broad spectrum activity.^[143]58 Fipronil
interferes with the GABAergic system, affecting brain growth and neural
development.^[144]59 Lassiter et al. found that fibronil caused
oxidative stress in undifferentiated neuronal PC12 cells and
subsequently inhibited DNA and protein synthesis.^[145]60 Ki et al.
subsequently indicated that fipronil-induced neuronal apoptosis was
mediated by reactive oxygen species production.^[146]61
Conversely, our chemical-related GSEA have also identified some
chemicals that might play a protective role in vertigo, such as PAC and
piroxicam. PAC was a potential therapeutic chemical for vertigo
identified by our ROSMAP PWAS and was also widely distributed in common
foods including cereals, fruits, vegetables, and wines, and belonged to
a universal group of plant polyphenols with their powerful antioxidant
capacity and possible protective effects on human health.^[147]62 Chen
et al. suggested that ethanol causes cognitive impairment and that the
protective effect of PAC on ethanol-induced cognitive impairment may be
due to its antioxidant and anti-inflammatory activities.^[148]63 Gong
et al. found that lotus seedpod proanthocyanidins (LSPC) could prevent
hippocampal neuron damage and reduce the expression of P53 protein in
the hippocampus, suggesting that LSPC might be used to treat
Alzheimer’s disease.^[149]64 Piroxicam was detected to be significantly
enriched in fetal brain-based TWAS for vertigo, which could be
converted into novel CNS stimulants or inhibitors through
desulphurization, methylation, dehydrogenation, carboxylation, and
carbonylation.^[150]65 Animal toxicity studies showed that Piroxicam
significantly reversed oxidative stress induced by the mycotoxin
3-nitropropionic acid (3-NP) in mice, suggesting that Piroxicam
protected mice from 3-NP-induced oxidative stress and behavioural
changes in the brain.^[151]66 Mazumder et al. recommended the use of
Piroxicam as the gold standard for the prevention of cerebral ischaemia
(CI) neurodegeneration, since Piroxicam inhibits N-methyl-D-aspartate
(NMDA) receptor, which are also associated with CI-induced
neurodegeneration.^[152]67
Our study has several strengths. First, this study examined the mRNA
and protein levels in vertigo through PWAS and TWAS. Second, GSEA
provided potential risk chemicals and possible treatments for vertigo.
There are several limitations of our study. First, the pre-computed
reference weights of the ROSMAP and Banner brain proteome included both
older and neurological disorder individuals, which may result in slight
bias in our results. Besides, although tissue differences were
controlled in the TWAS analysis, the weak anatomical correlation
between vertigo and DLPFC could still lead to potential bias. Third,
our method is a theoretical analysis of the sequencing data of European
ancestors. Further studies of other genetic backgrounds and annotation
analyses are needed to confirm our findings and reveal the underlying
biological mechanisms of identified genes and chemicals in the
development of vertigo.
In conclusion, we identified three high-confidence genes (RAB3B,
MTERFD2, and MGST1) showed proteome-wide significant associations in
the human brain with vertigo in two independent brain proteomes,
strongly highlighting the underlying pathogenesis of these proteins in
vertigo. Our study provides new insights into the genetically regulated
protein abundance in vertigo and also highlights promising chemicals
for further pathogenesis investigations and therapeutics research.
Supplementary Material
fcac313_Supplementary_Data
[153]Click here for additional data file.^ (102.6KB, zip)
Abbreviations
3-NP =
3-nitropropionic acid
Banner dataset =
Banner Sun Health Research Institute dataset
BSLMM =
Bayesian sparse linear mixed model
CI =
cerebral ischaemia
CMC =
CommonMind Consortium
CNS =
central nervous system
CTD =
Comparative Toxicogenomics Database
DLPFC =
dorsolateral prefrontal cortex
eQTL =
expression quantitative trait loci
ESs =
enrichment scores
GO =
gene ontology
GSEA =
gene set enrichment analysis
GWAS =
genome-wide association study
LD =
linkage disequilibrium
LSPC =
lotus seedpod proanthocyanidins
NMDA =
N-methyl-D-aspartate
PAC =
proanthocyanidins
pQTL =
protein quantitative trait loci
PWAS =
proteome-wide association study
QC =
quality control
ROSMAP =
Religious Orders Study or the Rush Memory and Aging Project
Sd =
standard deviation
sQTL =
splicing quantitative trait loci
TWAS =
transcriptome-wide association study
WGS =
whole-genome sequencing
Contributor Information
Bolun Cheng, Key Laboratory of Trace Elements and Endemic Diseases,
Collaborative Innovation Center of Endemic Disease and Health Promotion
for Silk Road Region, School of Public Health, Health Science Center,
Xi’an Jiaotong University, No. 76 Yan Ta West Road, Xi’an 710061,
China; Key Laboratory for Disease Prevention and Control and Health
Promotion of Shaanxi Province, Health Science Center, Xi'an Jiaotong
University, Xi’an 710061, China; Key Laboratory of Environment and
Genes Related to Diseases of Ministry of Education of China, Health
Science Center, Xi'an Jiaotong University, Xi'an 710061, China.
Peilin Meng, Key Laboratory of Trace Elements and Endemic Diseases,
Collaborative Innovation Center of Endemic Disease and Health Promotion
for Silk Road Region, School of Public Health, Health Science Center,
Xi’an Jiaotong University, No. 76 Yan Ta West Road, Xi’an 710061,
China; Key Laboratory for Disease Prevention and Control and Health
Promotion of Shaanxi Province, Health Science Center, Xi'an Jiaotong
University, Xi’an 710061, China; Key Laboratory of Environment and
Genes Related to Diseases of Ministry of Education of China, Health
Science Center, Xi'an Jiaotong University, Xi'an 710061, China.
Xuena Yang, Key Laboratory of Trace Elements and Endemic Diseases,
Collaborative Innovation Center of Endemic Disease and Health Promotion
for Silk Road Region, School of Public Health, Health Science Center,
Xi’an Jiaotong University, No. 76 Yan Ta West Road, Xi’an 710061,
China; Key Laboratory for Disease Prevention and Control and Health
Promotion of Shaanxi Province, Health Science Center, Xi'an Jiaotong
University, Xi’an 710061, China; Key Laboratory of Environment and
Genes Related to Diseases of Ministry of Education of China, Health
Science Center, Xi'an Jiaotong University, Xi'an 710061, China.
Shiqiang Cheng, Key Laboratory of Trace Elements and Endemic Diseases,
Collaborative Innovation Center of Endemic Disease and Health Promotion
for Silk Road Region, School of Public Health, Health Science Center,
Xi’an Jiaotong University, No. 76 Yan Ta West Road, Xi’an 710061,
China; Key Laboratory for Disease Prevention and Control and Health
Promotion of Shaanxi Province, Health Science Center, Xi'an Jiaotong
University, Xi’an 710061, China; Key Laboratory of Environment and
Genes Related to Diseases of Ministry of Education of China, Health
Science Center, Xi'an Jiaotong University, Xi'an 710061, China.
Li Liu, Key Laboratory of Trace Elements and Endemic Diseases,
Collaborative Innovation Center of Endemic Disease and Health Promotion
for Silk Road Region, School of Public Health, Health Science Center,
Xi’an Jiaotong University, No. 76 Yan Ta West Road, Xi’an 710061,
China; Key Laboratory for Disease Prevention and Control and Health
Promotion of Shaanxi Province, Health Science Center, Xi'an Jiaotong
University, Xi’an 710061, China; Key Laboratory of Environment and
Genes Related to Diseases of Ministry of Education of China, Health
Science Center, Xi'an Jiaotong University, Xi'an 710061, China.
Yumeng Jia, Key Laboratory of Trace Elements and Endemic Diseases,
Collaborative Innovation Center of Endemic Disease and Health Promotion
for Silk Road Region, School of Public Health, Health Science Center,
Xi’an Jiaotong University, No. 76 Yan Ta West Road, Xi’an 710061,
China; Key Laboratory for Disease Prevention and Control and Health
Promotion of Shaanxi Province, Health Science Center, Xi'an Jiaotong
University, Xi’an 710061, China; Key Laboratory of Environment and
Genes Related to Diseases of Ministry of Education of China, Health
Science Center, Xi'an Jiaotong University, Xi'an 710061, China.
Yan Wen, Key Laboratory of Trace Elements and Endemic Diseases,
Collaborative Innovation Center of Endemic Disease and Health Promotion
for Silk Road Region, School of Public Health, Health Science Center,
Xi’an Jiaotong University, No. 76 Yan Ta West Road, Xi’an 710061,
China; Key Laboratory for Disease Prevention and Control and Health
Promotion of Shaanxi Province, Health Science Center, Xi'an Jiaotong
University, Xi’an 710061, China; Key Laboratory of Environment and
Genes Related to Diseases of Ministry of Education of China, Health
Science Center, Xi'an Jiaotong University, Xi'an 710061, China.
Feng Zhang, Key Laboratory of Trace Elements and Endemic Diseases,
Collaborative Innovation Center of Endemic Disease and Health Promotion
for Silk Road Region, School of Public Health, Health Science Center,
Xi’an Jiaotong University, No. 76 Yan Ta West Road, Xi’an 710061,
China; Key Laboratory for Disease Prevention and Control and Health
Promotion of Shaanxi Province, Health Science Center, Xi'an Jiaotong
University, Xi’an 710061, China; Key Laboratory of Environment and
Genes Related to Diseases of Ministry of Education of China, Health
Science Center, Xi'an Jiaotong University, Xi'an 710061, China.
Funding
This study was supported by the Natural Science Basic Research Plan in
Shaanxi Province of China (2021JCW-08), and the Fundamental Research
Funds for the Central Universities.
Competing interests
The authors report no conflicts of interest.
Supplementary material
[154]Supplementary material is available at Brain Communications
online.
References