Content uploaded by Concetta Dagostino
Author content
All content in this area was uploaded by Concetta Dagostino on Feb 19, 2019
Content may be subject to copyright.
Downloaded from https://journals.lww.com/pain by BhDMf5ePHKav1zEoum1tQfN4a+kJLhEZgbsIHo4XMi0hCywCX1AWnYQp/IlQrHD3be73+7zLk/QlBRHAHE26NPUF3qWQ5/Miuaa2hRD6vVg= on 02/19/2019
Downloadedfromhttps://journals.lww.com/pain by BhDMf5ePHKav1zEoum1tQfN4a+kJLhEZgbsIHo4XMi0hCywCX1AWnYQp/IlQrHD3be73+7zLk/QlBRHAHE26NPUF3qWQ5/Miuaa2hRD6vVg= on 02/19/2019
Research Paper
Genetic pathway analysis reveals a major role for
extracellular matrix organization in inflammatory
and neuropathic pain
Marc Parisien
a
, Alexander Samoshkin
a,b
, Shannon N. Tansley
a
, Marjo H. Piltonen
a
, Loren J. Martin
c
,
Nehme El-Hachem
a
, Concetta Dagostino
d,e
, Massimo Allegri
f,g
, Jeffrey S. Mogil
a,h
, Arkady Khoutorsky
a,i
,
Luda Diatchenko
a,
*
Abstract
Chronic pain is a debilitating and poorly treated condition whose underlying mechanisms are poorly understood. Nerve injury and
inflammation cause alterations in gene expression in tissues associated with pain processing, supporting molecular and cellular
mechanisms that maintain painful states. However, it is not known whether transcriptome changes can be used to reconstruct
a molecular pathophysiology of pain. In the current study, we identify molecular pathways contributing to chronic pain states through
the analysis of global changes in the transcriptome of dorsal root ganglia, spinal cord, brain, and blood in mouse assays of nerve
injury– and inflammation-induced pain. Comparative analyses of differentially expressed genes identified substantial similarities
between 2 animal pain assays and with human low-back pain. Furthermore, the extracellular matrix (ECM) organization has been
found the most commonly regulated pathway across all tested tissues in the 2 animal assays. Examination of human genome-wide
association study data sets revealed an overrepresentation of differentially expressed genes within the ECM organization pathway in
single nucleotide polymorphisms most strongly associated with human back pain. In summary, our comprehensive transcriptomics
analysis in mouse and human identified ECM organization as a central molecular pathway in the development of chronic pain.
Keywords: Chronic pain, Next-generation sequencing, Differentially expressed genes, Extracellular matrix, GWAS,
Bioinformatics
1. Introduction
Chronic pain is a debilitating condition that serves no obvious
biological function, in contrast with acute pain triggered at injury,
protecting the organism from further damage by activating
withdrawal behaviors. More than 20% of the population worldwide
suffer from chronic pain, making it the leading cause of disability in
humans.
56
The mechanisms underlying the development of chronic
pain are not well understood. Accordingly, available medications for
chronic pain target the symptoms of the disease and not the
underlying pathology.
66
This treatment strategy results in inadequate
pain relief under most circumstances and is accompanied by a high
prevalence of side effects.
23,49,69
Two major types of pain states are inflammatory pain (IP),
caused by tissue inflammation or injury, and neuropathic pain (NP),
caused by damage to or dysfunction of the nervous system itself.
Inflammatory pain and NP share several underlying neuronal
mechanisms, including increased excitability of primary afferents,
reduced spinal inhibitory tone, and the involvement of higher brain
centers. Despite the progress in uncovering neuronal alterations
along the pain pathway, the genetic programs and molecular
mechanisms supporting these changes are not well known.
Preclinical studies in animal models have shown that pain
states caused by inflammation or nerve injury are accompanied
by massive changes in gene expression along the pain pathway,
including in the dorsal root ganglion (DRG), spinal cord (SC), and
supraspinal brain areas.
25,29,30,72
Long-lasting changes in gene
expression are believed to underlie the development of a sus-
tained pain state by mediating the molecular and cellular
mechanisms causing sensitization of peripheral and central pain
circuits and the generation of ectopic discharge, both hallmarks
of chronic pain.
17, 43, 47
Complex pathophysiology of chronic pain
assumes a model of multigene contribution to the reorganization
of the somatosensory system in persistent pain states. Identifi-
cation of these complex gene interactions and subthreshold
contributions to pain phenotypes, as well as more global gene
expression patterns, require unique bioinformatics approaches
and the generation of high-quality gene expression data sets of
different pain models and tissues.
Sponsorships or competing interests that may be relevant to content are disclosed
at the end of this article.
M. Parisien and A. Samoshkin contributed equally to this work.
a
Alan Edwards Centre for Research on Pain, McGill University, Montré
al, QC, Canada,
b
School of the Clinical Medicine, University of Cambridge, Cambridge, United Kingdom,
c
Department of Psychology, University of Toronto, Mississauga, ON, Canada,
d
Department of Medicine and Surgery, University of Parma, Parma, Italy,
e
Study In
Multidisciplinary Pain Research (SIMPAR), Parma, Italy,
f
ItalianPainGroup,Milan,Italy,
g
Pain Therapy Service, Policlinico Monza Hospital, Monza, Italy, Departments of
h
Psychology and,
i
Anesthesia, McGill University, Montré
al, QC, Canada
*Corresponding author. Address: Alan Edwards Centre for Research on Pain
Genome Building, Room 2201, 740 Dr. Penfield Avenue, Montreal, Quebec,
Canada H3A 0G1. Tel.: 514-398-2878; fax: 514-398-8900. E-mail address:
luda.diatchenko@mcgill.ca (L. Diatchenko).
Supplemental digital content is available for this article. Direct URL citations appear
in the printed text and are provided in the HTML and PDF versions of this article on
the journal’s Web site (www.painjournalonline.com).
PAIN 00 (2019) 1–13
©2019 International Association for the Study of Pain
http://dx.doi.org/10.1097/j.pain.0000000000001471
Month 2019·Volume 00 ·Number 00 www.painjournalonline.com 1
Copyright © 2019 by the International Association for the Study of Pain. Unauthorized reproduction of this article is prohibited.
The current understanding of gene expression signatures in
pain states is largely based on studies using a single pain assay
and is limited to 1 or 2 tissues within the pain pathway. Recent
single-cell profiling or cell type–specific transcriptomic studies
provide important insights into changes in distinct cell types
under pathological conditions; however, these studies are lacking
a relevance to dynamics of pathological changes.
24,51,61
Here,
we undertook a more comprehensive approach to study the gene
expression landscape in pain states by performing transcriptomic
analysis of 4 tissues (DRG, SC, brain, and blood) in 2 mouse pain
assays modeling inflammatory or NP. Our deep RNA sequencing
(.50 M reads) at tissue-wide levels allowed for detection of lowly
expressed genes and quantitation of minute differential gene
expression fold changes even at the mixed-cell–type sample. A
pathway analysis of differentially expressed genes (DEGs)
identified the extracellular matrix (ECM) organization as the most
commonly regulated category across pain assays and tissues.
2. Results
2.1. Transcriptomics of mouse pain models
To study the pathophysiology of pain states at the molecular level,
we performed next-generation deep sequencing of RNAs
extracted from tissues relevant to pain processes in 2 mouse
pain assays (Fig. 1A). Two well-characterized assays were used:
complete Freund’s adjuvant (CFA) as an assay of IP and spared
nerve injury (SNI) as an assay of NP. The tissues were extracted 3
days after CFA injection and 7 days after SNI. These time points
were selected based on behavioral data (Fig. S1, available at
http://links.lww.com/PAIN/A710; legend available at http://links.
lww.com/PAIN/A724): CFA injection–induced hypersensitivity
peaks at day 3 (Fig. S1A, available at http://links.lww.com/
PAIN/A710; legend available at http://links.lww.com/PAIN/
A724), whereas SNI-induced hypersensitivity develops to its
maximal extent at day 7 (Fig. S1B, available at http://links.lww.
com/PAIN/A724). From each mouse (n 53/assay), 4 tissues
were extracted after injury: DRG (L3-L5), dorsal horn of the
lumbar SC, whole brain, and whole blood. The poly(A) mRNA was
purified and sequenced using the Illumina HiSeq2000 sequencing
platform. Quality of deep sequencing data was assessed in a 4-
fold manner: (1) byRNA fragment mapping rates on the mouse
genome (Fig. S2, available at http://links.lww.com/PAIN/A710;
legend available at http://links.lww.com/PAIN/A724); (2) by
triplicates for within-group gene expression correlations (Fig.
S3, available at http://links.lww.com/PAIN/A710; legend avail-
able at http://links.lww.com/PAIN/A724); (3) by gene expression
clustering (Fig. 1B); and (4) by principal component analyses
(PCAs) (Fig. 1C). All approaches confirmed the high quality of
RNA purification (RIN 9.3 60.5, minimum 7.8) and sequencing
(mapped reads 77.4 M 624.8 M, minimum 45.7 M). Gene
expression was visualized in a pileup plot (Fig. S4A, available at
http://links.lww.com/PAIN/A710; legend available at http://links.
lww.com/PAIN/A724). Differential gene expression was assessed
in a volcano plot (Fig. S4B, available at http://links.lww.com/
PAIN/A710; legend available at http://links.lww.com/PAIN/
A724), where p-values represent P-value-weighted fold
changes.
11
Data for differential expression of genes for all tissues
and assays can be found in Table S1 (available at http://
links.lww.com/PAIN/A711; legend available at http://links.lww.com/
PAIN/A724), Table S2 (available at http://links.lww.com/PAIN/
A712; legend available at http://links.lww.com/PAIN/A724),
Table S3 (available at http://links.lww.com/PAIN/A713; legend
available at http://links.lww.com/PAIN/A724), Table S4 (available
at http://links.lww.com/PAIN/A714; legend available at http://
links.lww.com/PAIN/A724), Table S5 (available at http://links.
lww.com/PAIN/A715; legend available at http://links.lww.com/
PAIN/A724), Table S6 (available at http://links.lww.com/PAIN/
A716; legend available at http://links.lww.com/PAIN/A724),
Table S7 (available at http://links.lww.com/PAIN/A717; legend
available at http://links.lww.com/PAIN/A724), and Table S8
(available at http://links.lww.com/PAIN/A718; legend available
at http://links.lww.com/PAIN/A724).
Unsupervised clustering of RNA-Seq samples based on gene
expression correlation first distinguished the nervous system
tissue samples from the whole blood (Fig. 1B). In the nervous
Figure 1. Overview of transcriptomics of mouse pain assays. (A) Experimental workflow. A total of 9 mice were split into 3 groups: control (CTR), treated with
complete Freund’s adjuvant (CFA), and underwent spared nerve injury (SNI). For each mouse, the following 4 tissues were collected: whole brain (BRN), spinal
cord (SC), dorsal root ganglion (DRG), and whole blood (BLD). From each tissue, next-generation sequencing was performed on Illumina HiSeq2000. (B)
Unsupervised clustering analyses of mouse transcriptome (M.mus.). The clustering was made based on one minus Pearson’s correlation coefficient in gene
expression between the samples. The clustering tree shows first branching of the nervous system (NS) tissues from the blood (BLD). Under NS, the clustering tree
shows branching of the tissues from the central nervous system (CNS) from the peripheral nervous system (PNS). Similar clustering is observed in humans (H.sap.,
grayed inset). (C) Tissue-based principal component analyses (PCA) of mouse transcriptome. CTR (green circles), CFA (yellow boxes), SNI (blue triangles).
2M. Parisien et al.·00 (2019) 1–13 PAIN
®
Copyright © 2019 by the International Association for the Study of Pain. Unauthorized reproduction of this article is prohibited.
system branch, the clustering distinguished samples of the
central nervous system (CNS: brain and SC) from the peripheral
nervous system (PNS: DRG). In the CNS sub-branch, all SC
samples clustered together, as did all brain samples. We verified
that the first 2 clustering branches, nervous system vs blood, and
CNS vs PNS were also present in the gene expression data of the
same 4 tissues in humans (Fig. 1B, inset; microarray data from
Ref. 5). Using the top 50 most DEGs, we performed within-tissue
PCAs to assess that the 3 pain states—control, CFA, and
SNI—were distinguishable in each tissue (Fig. 1C). Brain, SC,
and DRG presented clear pain model distinctions with only 2
principal components, whereas blood presented a more complex
pain profile, requiring higher dimensionality PCAs for better
separation. In all, the first 2 PCAs can explain only 17% of within-
tissue gene expression variance in the brain, while up to 32% of
the variance was explained in the DRG.
2.2. Replication of published data
Next, we evaluated the replication of DEG patterns observed in our
study with DEGs in various tissue-assay pairs from already-published
data sets (Fig. 2; Table S9, available at http://links.lww.com/PAIN/
A719; legend available at http://links.lww.com/PAIN/A724; and
Table S10, available at http://links.lww.com/PAIN/A720; legend
available at http://links.lww.com/PAIN/A724). We assessed all
related genome-wide experiments for which data are available
in NCBI’s Gene Expression Omnibus and ready for prompt
analyses with geo2r.
15,19
Expression fold changes and
associated P-values were converted into p-values for com-
parison with our work. We found 3 such data sets (Fig. 2). Data
set GSE18803
13
compares RNA expression data in SNI vs
sham at day 7 after SNI in the SC of rats (Fig. 2A). Data set
GSE15041
67
compares expression data in SNI vs sham in DRG
at day 7 after SNI in rats (Fig. 2B). Finally, data set GSE38859
57
Figure 2. Correlation between the DEGs in current data set with previously published genome-wide rodent studies, and correlation between mouse and human
transcriptomes. (A–C)DEG is measured as p-value. Eachdot is a gene. Pink line obtainedfrom linear regression. Correlation quantified usingrank-based Spearman’s
coefficient (r)withaccompanyingP-value (P). Colorcoding suggests genedensity. The corresponding tissuesof current data set has been compared with (A) ratspinal
cord 7 days SNI vs sham (GEO set GSE18803), (B) rats DRG ipsilateral 7 days SNI vs sham (GEO set GSE15041), and (C) rats DRG 3 days CFA vs sham (GEO set
GSE38859). (D) Clustering of mouse (Mmus) and human (Hsap) transcriptomes. The clustering is based on Spearman’s correlation coefficient of gene expression
between pairs of transcriptomes. Unsupervised clustering leads to first branching by species, suggesting strong influence of species-specific transcriptomes (left
panel). Removal of specie-specific transcriptomes leads to pairings by tissue types instead (right panel). (E) Transcriptome gene expression correlation by tissue type
betweenhuman and CTR mouse. For eachtissue, rank-basedSpearman’s coefficient (r)andP-value (P) are shown. Each dot isa gene. Colour codingsuggests gene
density. Pink line obtained from linear regression. Mouse RNA-Seq gene expression has been log2 transformed to match human microarray log-based gene
expression intensity (I). CFA, complete Freund’s adjuvant; DEG, differentially expressed gene; DRG, dorsal root ganglion; SNI, spared nerve injury
Month 2019·Volume 00 ·Number 00 www.painjournalonline.com 3
Copyright © 2019 by the International Association for the Study of Pain. Unauthorized reproduction of this article is prohibited.
compares expression data in CFA vs sham at day 3 in rats (Fig. 2C).
Because these data sets were compiled using rats, we expected
limited replication of DEGs because of interspecies gene expression
differences (Table S9, available at http://links.lww.com/PAIN/A719;
legend available at http://links.lww.com/PAIN/A724; Fig. 2). Further-
more, different quantified platforms were used, ie, RNA-Seq for
mouse and microarray for rat. In all 3 genome-wide replication
attempts, nevertheless, rank-based Spearman correlation coeffi-
cients ranged from 0.11 to 0.20, with accompanying P-values
between 10
217
and 10
2142
.
Analysis of summary data from a previously published large meta-
analysis of DEGs from multiple microarray studies of pain
30
was then
used to further quantify replication between our findings and previously
published data (Table S10, available at http://links.lww.com/PAIN/
A720; legend available at http://links.lww.com/PAIN/A724). The
meta-analysis compiled studies that were performed using both mice
and rats. It identified a total of 79 genes with significant expression fold
change in at least 4 independent microarray experiments. Details of
neuropathic and inflammation pain assays were also slightly differing
(eg, SNI vs chronic constriction injury). Replication of the surveyed
genes yielded a total of 65 committed fold change calls, for which
a total of 57 calls agree on the direction of the fold change, thus
generating 8 mismatched calls. The P-value for this level of success
rate, assuming probability to make a correct fold change direction call
is one-half (1/2), was evaluated using the hypergeometric test that
yielded P53.2 310
210
. Thus, our deep-sequencing data were of
high-quality and replicated previous studies.
2.3. Relevance of mouse transcriptome study for humans
We next examined the relevance of using mouse transcriptome
data to gain knowledge about humans. A data set containing
human gene expression data in the same 4 tissuesprobed herein
5
allowed for a direct comparison of transcriptomes between control
mice and humans, on a per-tissue basis (Fig. 2). Unsupervised
clustering of the transcriptomes yielded a dendrogram in which the
tissues did not cluster by type, but rather by species (Fig. 2D, left
panel). Removal of species-specific expressed genes and retaining
of genes expressed in both species yielded an unsupervised
clustering in which tissues clustered by identity (Fig. 2D,right
panel). In all cases, blood was distinguishable from nervous
tissues, and CNS (brain andSC) clustered togetherapart from PNS
(DRG). Comparison of transcriptomes on a per-tissue basis was
also performed (Fig. 2E). All rank-based correlations of gene
expression were statistically significant, with P-values below an
estimated 2.2 310
216
. Spearman correlation coefficients varied from
0.48 in DRG to 0.67 in the brain, indicating strong correspondences
between human and mouse transcriptomes, in each tissue. Once
again, because the transcriptomes have been quantified using
different technologies (RNA-Seq for mouse and microarray for
human), it is expected that at least some differences in gene
expression arose from the varying accuracies and signal-to-noise
ratios of the platforms, notwithstanding species-specific gene
expression.
2.4. Relevance of differential gene expression in mouse pain
models to humans with clinical condition
Our mouse data were further correlated with a complex human
pain state, namely back pain.As it has been shown previously that
back pain patients manifest symptoms with both neuropathic and
inflammatory components,
20
we compared DEGs in blood of
humans with back pain to DEGs in blood in both neuropathic and
inflammatory mouse pain models (Fig. 3). Very significant
correlations were observed for both comparisons; however, in
the mouse NP model, this correlation was stronger (Fig. 3A;
Spearman r50.47, P55310
2110
)thaninthemouse
inflammation pain model (Fig. 3B; Spearman r50.33, P533
10
239
). When genes were grouped by pathways, 80% of
differentially expressed pathways in humans displayed correlation
coefficients with SNI equal to or greater than 0.35, whereas only
40% of CFA differentially expressed pathways showed correlation
coefficients equal or greater than 0.35 (Fig. 3C). Thus, we observed
both a neuropathic and an inflammatory component at the level of
DEGs and differentially expressed pathways in the human back
pain profile. However, in this back pain cohort, the neuropathic
component was stronger (Kolmogorov–Smirnov P,2310
216
).
2.5. Genome-wide analysis of differentially expressed genes
in pain assays
We then established the full list of genes differentially expressed in
each of the 4 tissues, and for each of the 2 pain assays (Fig. 4;
Table S11, available at http://links.lww.com/PAIN/A721; legend
available at http://links.lww.com/PAIN/A724). In the CFA assay,
the number of DEGs varied from 1123 in the brain to 2642 in SC
(Fig. 4A). A large proportion of DEGs are tissue-specific, and
there were only 48 genes that are differentially expressed in all
tissues in the CFA assay. In SNI, the number of DEGs varied from
Figure 3. Comparison of differential gene expression in human clinical pain conditions and mouse pain assays. (A and B) Gene expression profile in the blood of
human subjects (Hsap) with back pain (BP) were contrasted (p) against those with no pain (CTR). Gene expression profiles in the blood of mice (Mmus) under a pain
assay were contrasted (p) against those in the control group (CTR). Each dot is a gene in GO’s biological process pathway (GO:0008150). Colour coding reflects
gene density. Pink line obtained from linear regression. Rank-based Spearman’s correlation coefficient rho (r) and associated P-value (P) are shown. (A) Spared
nerve injury pain assays. (B) Complete Freund’s adjuvant pain assay. (C) Cumulative fraction of all GO’s biological process pathways as a function of their
Spearman’s r. Kolmogorov–Smirnov test P-value between the 2 curves is shown. Black vertical line indicates r50, while gold vertical line r50.35. CFA,
complete Freund’s adjuvant; SNI, spared nerve injury.
4M. Parisien et al.·00 (2019) 1–13 PAIN
®
Copyright © 2019 by the International Association for the Study of Pain. Unauthorized reproduction of this article is prohibited.
968 in brain up to 3194 in DRG (Fig. 4B), but only 18 genes were
differentially expressed in all 4 tissues.
Differential expression of genes was found conserved between
tissues. For CFA, all regression slopes were positive, and percent
variance explained (PVE) varied between 1% and 11% (Fig. 4C).
For SNI, we also found positive regression slopes (except for one),
while PVE varied between 0% and 8% (Fig. 4D). Most correlated
tissue pair in CFA was between DRG and blood (PVE 11%), while
in SNI between DRG and SC (PVE 8%).
Furthermore, there was a large overlap in DEGs between CFA
and SNI. The percentage of shared DEGs was found to range
from 24% in the brain up to 44% in SC (Fig. 4E). Unexpectedly,
these DEG overlaps within tissues for different pain assays were
substantially higher in magnitude than between tissues within one
pain model. The large number of shared DEG points to possible
common pathways that are activated by inflammation (CFA) and
nerve injury (SNI). All tissues considered, as much as 49% of
DEGs, were shared between CFA and SNI.
In addition, of the 800 genes already characterized in the context of
pain research (see Materials and Methods for source details), a total of
452 known pain genes (56%) of the total 8687 DEGs in our assays
overlapped (Fig. 4F; Table S12, available at http://links.lww.com/
PAIN/A722; legend available at http://links.lww.com/PAIN/A724).
This represents a 3.1-fold overrepresentation of observed known
differentially expressed pain genes over what would be expected by
chance alone, with associated enrichment P-value of 2 310
2129
.
2.6. Tissue-specific pathways involved in distinct
pain processes
We next developed a star plot that tracks differential expression in
each mouse pain assay, for each gene, in a tissue-based manner
Figure 4. Differentially expressed genes in pain assays and in pain-relevant tissues. (A and B) Four-way Venn diagram showing shared DEGs between tissues in
CFA (A) or SNI (B) pain assays. (C and D) Linear regression of DEGs for all tissue pairs. Shown are percent variance explained (Pearson’s r
2
) rounded to nearest
integer (upper right corner), and regression’s slope (m) with slope’s sign emphasized (lower left corner), for CFA (C) or SNI (D) pain assays. (E) Two-way Venn
diagrams showing shared DEGs between CFA and SNI pain assays, for each tissue, and all tissues combined (ALL). Counts at intersection expressed as
percentage of the smallest values between the 2 tissue-specific DEGs counts. (F) Two-way Venn diagram showing overlap of DEGs in any tissue or pain mouse
assay vs those known to contribute to pain states. Shown at the bottom is observed to expected enrichment ratio (x) of number of DEG among known pain genes,
with accompanying P-values. CFA, complete Freund’s adjuvant; DEGs, differentially expressed genes; SNI, spared nerve injury
Month 2019·Volume 00 ·Number 00 www.painjournalonline.com 5
Copyright © 2019 by the International Association for the Study of Pain. Unauthorized reproduction of this article is prohibited.
(Fig. 5). We defined each star plot by the 4 axes representing the
correlation between DEGs in SNI and CFA assays (Fig. 5A;andFig.
S5, available at http://links.lww.com/PAIN/A710; legend available at
http://links.lww.com/PAIN/A724). The star plot helped highlight genes
that were coregulated between CFA and SNI, again accentuating
shared pathophysiology; rank-based Spearman correlations varied
from 0.13 in the brain to 0.39 in DRG. The lists of genes identified from
the star plots allowed for gene set enrichment analyses (GSEA) (Table
S13, available at http://links.lww.com/PAIN/A723; legend available at
http://links.lww.com/PAIN/A724). The pathway that showed the most
significant enrichment P-value is reported for each of the 4 axes, in
each tissue-based star plot (Fig. 5B).
Figure 5. Tissue-specific pathways contributing to distinct pain processes. (A) Star plots depict DEGs between CFA and SNI pain assays, in a tissue-specific
manner. Each dot is a gene. Four axes are defined and colored accordingly: CFA-only DEGs (axis 1, green), DEGs correlated between CFA and SNI models (axis 2,
pink), SNI-only DEGs (axis 3, purple), and DEGs anticorrelated between CFA and SNI models (axis 4, orange). (B) For each tissue’s axes, the pathway with the best
FDR-corrected GSEA P-value is shown in a bar plot (left). Vertical red lines mark statistical significance at FDR 10% level. Also shown is the ECM organization
pathway (GO:0030198, right). (C) Bar plot showing top 10 differentially expressed pathways, summed over all axes of all tissues. Vertical red lines mark statistical
significance at FDR 10% level. (D) Heatmap of contribution of each tissue (top) and pain models (bottom) to the top 10 differentially expressed pathways. CFA,
complete Freund’s adjuvant; DEG, differentially expressed gene; ECM, extracellular matrix; GESA, gene set enrichment analyses; SNI, spared nerve injury.
6M. Parisien et al.·00 (2019) 1–13 PAIN
®
Copyright © 2019 by the International Association for the Study of Pain. Unauthorized reproduction of this article is prohibited.
Further analyses for pathway enrichment yielded a list of top 10
differentially expressed pathways, summed over all axes of all
tissues (Fig. 5C). The ECM organization pathway emerged as the
top differentially expressed pathway (GO:0030198). The ECM
organization pathway is a biological process that manages the
assembly, maintenance, and disassembly of the ECM. We find
this pathway to be enriched in 6 different axes (of 16) (Fig. 5B): in
brain in CFA; in SC, axes 1, 2, and 3; and in DRG, axes 1 and 2.
Importantly, when ECM organization seemed significantly
enriched in 2 different axes of the same tissue, 2 nonoverlapping
subsets of genes of these axes contribute to it (Table S13,
available at http://links.lww.com/PAIN/A723; legend available at
http://links.lww.com/PAIN/A724).
The top 10 differentially expressed pathways were then re-
examined for their impact on tissues and pain assays (Fig. 5D).
Tissues of the nervous system (brain, SC, and DRG) showed the
greatest contributions from the ECM organization pathway, in
both CFA and SNI.
2.7. Gene network in the extracellular matrix
organization pathway
We next identified the top 10 most DEGs in the ECM organization
pathway from theirimpact, defined as the sum of |p|valuesfromall
assays and all tissues of differential expression (Fig. 6). Proteins
encoded by these genes were found to be in a highly-connected
network of protein–protein interactions (Fig. 6A,STRINGdata-
base
59
). The proposed network is quality-controlled, built on
experimental evidence, text mining, and knowledge transfer
between species for orthologous pairs.
70
In the network, all
proteins displayed moderate-to-strong evidence of interaction to
at least one other protein in the network. We found differential
expression of these genes to be tissue- and assay-specific,
although it could reflect timing of their differential expression (Fig.
6B). Dorsal root ganglion and SC tissues displayed the most
differential expression of these genes. Clustering of the genes
suggested that Thbs1,Cd36,andComp were mostly down-
regulated, whereas the other 7 proteins were mostly upregulated.
We observed that the brain featured the least number of DEGs,
whereasthe highest numberof DEGs was in the DRG. Interestingly,
a search in the GeneRIF database
40
for co-occurrences of the
gene’s name and “pain” revealed that many of the ECM
organization genes in the network are already characterized in
the context of pain research: Sparc,
39,60
Col1a1,
1,33
Mmp13,
64
Ctss,
8,10,11,34,76,77
Elane,
68,78
and Comp.
16,28
2.8. Human genome-wide association study results on back
pain are enriched for genes in extracellular matrix
organization pathway
We next examined the correlation of our mouse DEGs data with
genome-wide association study (GWAS) pain data sets. We
tested for overrepresentation of single nucleotide polymorphisms
(SNPs) from GWAS data sets within human orthologues of mouse
DEGs in the set of SNPs with strongest association with back pain
(Fig. 7; and Fig. S6, available at http://links.lww.com/PAIN/A710;
legend available at http://links.lww.com/PAIN/A724).
We applied this methodology to analyze GWAS results for
human back pain, the most common human musculoskeletal
pain condition. We used data from the UK Biobank project, for
which genotyping and medical records were collected for about
500,000 people.
2,58
Enrichment analyses showed that SNPs
within the genes that are differentially expressed in CFA/SNI are
overrepresented in top differentially expressed pathways (Fig. 7).
A heatmap shows which tissues and pain assays exhibit the
highest association with the human back pain phenotype. For the
ECM organization pathway, genes differentially expressed in the
brain and in the SC in an inflammation profile, and in DRG in
a neuropathic profile, showed marked enrichment of top GWAS
SNPs in these genes.
3. Discussion
To further our understanding of pain processes at the molecular
level, and to assess in an unbiased fashion the potential roles of
Figure 6. Protein–protein interaction network of the ECM organization pathway.
(A) Top 10 genes mostdifferentially expressed in the ECM organizationpathway,
organized in a protein–protein interaction (PPI) network. Edge confidence, from
highest to lowest, is proportional to experimental evidence for interacting protein
pair as provided in the STRING database. Genes are CD36 molecule
(thrombospondin receptor)—Cd36; collagen, type I, alpha 1—Col1a1; collagen,
type V, alpha 3—Col5a3; cartilage oligomeric matrix protein—Comp;cathepsin
S—Ctss; elastase, neutrophil expressed—Elane; matrix metallopeptid ase 13
(collagenase 3)—Mmp13; secreted protein, acidic, c ysteine-rich (osteonec tin)—
Sparc;thrombospondin1—Thbs1; and von Willebrand factor—Vwf.(B)
Heatmap of expression fold change of the top 10 genes across all tissues
(top) and pain models (bottom). Expression fold change is measured in p-value
units and dichotomized:increased expression compared withcontrol is in green,
while decreased is in red. ECM, extracellular matrix
Month 2019·Volume 00 ·Number 00 www.painjournalonline.com 7
Copyright © 2019 by the International Association for the Study of Pain. Unauthorized reproduction of this article is prohibited.
contributing biological processes, we performed genome-wide
next-generation transcriptomics analyses of pain-relevant tissues
in 2 mouse pain assays (Fig. 1). This approach provides a deep
analysis of transcriptomics of pain states. An alternative single-
cell sequencing approach generates a detailed picture of
changes in various cell types at the transcriptome level. However,
a compromise is made between the number of samples and the
sequencing depth or coverage. Typical results of single-cell
analyses report expression data for only about 3500 genes (with
large SDs),
65
a direct consequence of the sequencing depth of
about 2 310
5
mapped reads per sample.
24
The deep
sequencing obtained in our study (10
7
mapped reads compared
to 10
5
) allows for detecting low expressed genes, but most
importantly, to assign statistical significance of minute differences
in expression fold changes even in the mixed-cell–type popula-
tion. Once identified, genes of interest can be checked for
expression in specific cell types in single-cell sequencing
resources, in DRG,
65
or in SC.
24
We performed pathway analysis of DEGs across tissues and
assays. We identified several pathways known to be involved in
pain mediation, such as response to lipids, neuron differentiation,
and innate immune response. Strikingly, the most regulated
pathway in our analysis was ECM organization. The top 10 DEGs
within the ECM organization category encode proteins with
diverse functions including structural proteins such as collagen
and cartilage components (Col5a3,Comp,Col1a1, and Sparc),
proteins involved in cell-to-cell and cell-to-matrix interactions
(Vwf,Thbs1, and Cd36), and enzymes involved in ECM
remodelling (Ctss,Elane, and Mmp13). The observed diversity
in DEGs within the ECM suggests that identification of the ECM
organization as the most commonly regulated pathway in pain
assays reflects a central role of this biological domain as a whole
and is not related to a specific protein or group of proteins.
After identifying the top 10 biological pathways associated with
pain states in mouse assays, we attempted to correlate these
data with human GWAS of back pain (Fig. 7). We observed
overrepresentation of the genes corresponded to the top SNPs of
the back pain GWAS results within the ECM organization
pathway. The statistical enrichment was observed for the SNPs
situated within the genes identified in the inflammatory assay,
detected both in the brain and in the SC, and in the neuropathic
assay, detected in the DRG. Thus, we obtained evidence for
a major contribution of the ECM organization pathway to the
molecular pathophysiology of back pain and identified critical
elements of this contribution at multiple levels: (1) the top
associated SNPs, from the human GWAS, (2) the contributing
genes in which these SNPs reside, (3) the contributing pathways,
as genes are organized in modules, (4) the tissues in which
transcriptional changes and responses occur, and (5) the type of
contributing pain profile, inflammation, or neuropathic. Although
the association analysis was restricted to 10 pathways, it
recognized a new genetic pathway contributing to human back
pain and was able to identify combined impacts of individual
SNPs despite their modest effect sizes. Most of the DEGs within
the ECM organization pathway in the mouse assays of pain state
were detected in the dorsal horn of the SC and DRG tissues, with
some DEGs also observed in the brain. We thus anticipate that
ECM remodeling detected from expression profiling would
impact behavior at the whole organism level.
12
The ECM is composed of a variety of structural proteins, creating
the extracellular scaffold surrounding cells in the tissue.
7,42
It also
includes molecules mediating cell–cell and cell–matrix interaction,
and enzymes involved in ECM remodeling. In the nervous system,
ECM components provide not only the structural support for
neuronal and nonneuronal cells, but also regulate synapse
formation and function, and modulate neuronal excitability.
4,46,52
In the cortex, hippocampus, and amygdala, the ECM has been
demonstrated to restrict neuroplasticity by inhibiting activity-
dependent structural reorganization of presynaptic and post-
synaptic compartments,
55
suppressing the lateral mobility of
AMPA receptors in the membrane
21
and modulating neuronal
excitability.
4
Accordingly, disruption of the ECM enhanced synaptic
plasticity, acquisition of memories, and cognitive flexibility and
extinction.
46,50,71,74
Despite this progress inthe understanding key
roles of ECM in synaptic plasticity and memory formation, the role
of the ECM in regulation of nociceptive circuits and the de-
velopment of chronic pain is not well known.
Removal of inhibitory constraints is a paramount mechanism to
increase the excitability of nociceptive circuits in pathological pain
conditions.
14,45
Repeated or intense activation of sensory
neurons in response to peripheral tissue injury leads to a long-
lasting increase in the excitability of SC circuits (a process known
as central sensitization), resulting in a significant amplification of
peripheral inputs. Although central sensitization results from
a combination of mechanisms, the most prominent one is
a reduction of inhibitory tone. Several mechanisms have been
proposed to mediate the reduction of inhibitory tone in the SC in
pathological pain conditions, including (1) decreased expression
of neuron-specific K
1
-Cl
2
cotransporter-2 (KCC2) in spinal
neurons, leading to a change in chloride gradient, and thereby
attenuated inhibitory synaptic currents in response to GABA/
glycine
14
; (2) decrease in the activity of spinal inhibitory
neurons
62
; and, (3) reduced descending inhibition.
44
In light of
our results and previous studies in the cortex and hippocampus,
disinhibition of ECM-mediated suppression of spinal and
peripheral pain circuits might be a novel potential mechanism
promoting painful hypersensitive states. Modulation of ECM
composition and density could impact nociception not only by
direct effect on neuronal activity but also through T-cell and
macrophage infiltration, as well as nerve and blood vessel
sprouting, all factors potentially contributing to chronic pain.
26
We noted a number of limitations of this study. A single mouse
sex was used in our study to simplify interpretation and reduce
heterogeneity, given well-known sex differences in pain process-
ing.
41
Female mice were chosen since women represent the clear
majority of chronic pain patients.
41
Furthermore, we compared
transcriptomics of mice subjected to chronic inflammation and
nerve injury with matching home cage controls and not vehicle/
sham groups. This experimental design is suboptimal for animal
preclinical studies, but because we were comparing mouse data
Figure 7. Enrichment of SNPs in back pain GWAS for the ECM organization
pathway. Heatmap shows an enrichment of SNPs with strongest association
with human back pain condition in loci of genes whose orthologs in mice are
differentially expressed in experimental pain assays. The results are presented
for the top 10 differentially expressed pathways in a tissue-specific fashion
(top) and in a pain-specific fashion (bottom). ECM, extracellular matrix; GWAS,
genome-wide association study.
8M. Parisien et al.·00 (2019) 1–13 PAIN
®
Copyright © 2019 by the International Association for the Study of Pain. Unauthorized reproduction of this article is prohibited.
sets with human transcriptomics and GWAS, we reasoned that
this approach would generate data that better match human pain
studies, where healthy individuals serve as controls. Finally, we
used adult mice (18 weeks) to better mimic the situation in human
population where the disease is more common in adults.
However, further aging mice may display yet unique molecular
processes in pain assays, reflecting aging human aging pro-
cesses. Recognizing our limitations, we stressed that we
consider our results as a discovery that is needed to be followed
up and replicated in different populations, including males and
older mice, and different conditions, including better differentiat-
ing between pain and injury.
In summary, our catalog of genes differentially expressed in 2
mouse pain assays will be a valuable resource for the pain
research field, enabling study of the molecular basis underlying
pain processing. We showed that changes in transcriptome in
neuropathic and IP states in animal models display substantial
overlap between each other and with human LBP. Thus, these
results capture shared molecular pathophysiological mecha-
nisms between neuropathic and IP states. Furthermore, these
results validate mouse pain assays for studying human pain
conditions, although this does not exclude or quantify species
specificity in pain mechanisms. Finally, and most importantly, we
identified the ECM organization pathway as a major contributor to
both pain etiologies, including its major contribution to risk of
developing back pain at the level of human genetic variability.
4. Materials and methods
4.1. Statistics
Statistical tests used are indicated where the tests are performed.
Correction for multiple testing is applied whenever multiple tests
were involved, along with a statement indicating which method of
correction is used, based on a fair assessment of how correlated
the multiple tests are (eg, Bonferroni, Benjamini–Hochberg or FDR,
etc). Threshold for statistical significance is P-value #0.05 for each
test, unless specifically indicated at the location of the test.
4.2. Animals
For whole transcriptome studies,nine 18-week-old BALB/c female
mice were randomly assigned into 3 groups of 3 mice: (1) control
(CTR, naive), (2) IP induced with CFA, and (3) SNIfor modellingNP.
For chondroitinase ABC studies, both male and female mice were
used. All animals used in this experiment were bred on site, and
littermates were introduced into the study as soon as they matured
to 8 weeks of age. All mice were housed in standard polycarbonate
cages in groups of 3 or 4 same sex, in a temperature-controlled (20
61˚C) environment (14:10-hour light/dark cycle; lights on at 07:00
hours); tap water and food (Harlan Teklad 8604; Teklad Diets,
Madison, WI ) were available ad libitum.
4.3. Mouse pain assays
4.3.1. Complete Freund’s adjuvant assay
Complete Freund’s adjuvant (50%; Sigma) was injected sub-
cutaneously in a volume of 20 mL into the left and right plantar hind
paws using a 100-mL microsyringe with a 30-gauge needle.
Three days after CFA, mice were decapitated, and blood was
collected along with the brain and cerebellum, DRG (ipsilateral
L3-L5), and ipsilateral dorsal horn of SC. Blood was collected in
RNAprotect Animal Blood tubes (Qiagen, Germantown, MD) and
stored according to manufacturer’s recommendations. Blood
and tissue samples were kept at 280˚C until RNA extraction.
4.3.2. Spared nerve injury assay
Spared nerve injury, an experimental nerve injury designed to
produce NP, was performed under isoflurane/oxygen anesthesia
as described previously.
54
Briefly, using an operating microscope
(340), the 3 terminal branches of the sciatic nerve (tibial, sural,
and common peroneal) were exposed. The tibial and common
peroneal nerves were cut, after tight ligation with 6.0 silk,
“sparing” the sural nerve. The incisions were closed in layers
using interrupted sutures (6-0 Vicryl) and wound clips. The animal
recovered on a thermostatically controlled heating pad (carefully
monitored to prevent overheating) until ambulatory as per
standard operating procedures. To reduce the overall number
of mice required, the SNI surgery was performed in a bilateral
fashion (ie, left and right side). Seven days after the surgery, blood
and other tissues (L3-L5 DRG, dorsal horn of the SC, and brain)
were collected and stored as for CFA-treated mice.
4.3.3. Von Frey testing
The up–down method of Dixon
9
was used to estimate 50%
withdrawal thresholds using nylon monofilaments (Stoelting
Touch Test), calibrated monthly. All experiments took place
during the light cycle, between 9:00 and 16:00 hours. Mice were
placed in custom-made Plexiglas cubicles (5 38.5 36 cm) on
a perforated metal floor and were permitted to habituate for at
least 1 hour before testing. Filaments were applied to the plantar
surface of the hind paw for 1 second, and responses were
recorded. Two consecutive measures were taken on each hind
paw at each time point and averaged. Curves were plotted using
GraphPad Prism v.7.
4.4. RNA extraction
Total RNA from nervous tissues was isolated using RNeasy Lipid
Tissue Mini Kit, whereas RNA from blood cells was isolated using
the RNeasy Mini Kit, including DNaseI treatment (all from Qiagen),
according to the manufacturer’s instructions. Total RNA was
quantified using the NanoDrop 2000 (Thermo Scientific), and the
RNA quality was assessed with the 2100 Bioanalyzer (Agilent
Technologies).
4.5. Whole transcriptome sequencing (RNA-Seq)
All RNA-sequencing procedures were performed by Genomics
Platform facility (Institute for Research in Immunology and
Cancer, Montreal, Canada). Transcriptome libraries were gener-
ated from 1 mg of total RNA using the Kapa RNA–stranded
Sample Prep Kit (KK8400; KAPABiosystems) following the
manufacturer’s protocols. Briefly, poly-A mRNA was purified
using poly-T oligo-attached magnetic beads using 2 rounds of
purification. During the second elution of the poly-A RNA, the
RNA was fragmented and primed for cDNA synthesis. During
cDNA synthesis, dUTP was incorporated in the second-strand
synthesis, where dUTP-containing strand was selectively de-
graded. Adenylation of the 39ends and ligation of adapters were
performed after the manufacturer protocol. Enrichment of DNA
fragments that have adapter molecules on both ends was
performed using 10 cycles of PCR amplification using the KAPA
PCR mix and Illumina-adapted primers cocktail.
Month 2019·Volume 00 ·Number 00 www.painjournalonline.com 9
Copyright © 2019 by the International Association for the Study of Pain. Unauthorized reproduction of this article is prohibited.
4.6. Sequencing
Paired-end (2 3100 bp) sequencing was performed using the
Illumina HiSeq2000 machine running TruSeq v3 chemistry. Details of
the Illumina sequencing technologies can be found at https://
www.illumina.com/techniques/sequencing/rna-sequencing.html.
4.7. Processing of RNA-Seq data
RNA-Seq data have been trimmed with Trimmomatic v0.32,
6
then mapped on UCSC’s mouse genome version mm10
(grabbed from ftp://ussd-ftp.illumina.com/Mus_musculus/
UCSC/mm10/) using tophat v2.0.11
27
and bowtie v1.0.0.
32
Differential gene expression detected at gene level from
experimental triplicates using cuffdiff v2.2.1.
63
The RNA-Seq
data have been deposited in NCBI’s Gene Expression Omnibus
19
and are accessible through GEO accession number GSE111216.
We used a level of statistical significance at alpha 0.05 to define
DEGs |p|$critical value of 1.11.
73
Gene names from HUGO
Gene Nomenclature Committee (HGNC).
75
RNA fragment raw counts ranged from 61 M to 195 M, with an
average of 93 M reads (SD of 30 M) (Fig. S2, available at http://
links.lww.com/PAIN/A710; legend available at http://link-
s.lww.com/PAIN/A724). After adapter trimming, the 100 nts
paired-end reads had a mean length of 94 nts (SD of 13 nts),
retaining the specificity of paired-end sequencing. The average
ratio of mapped to trimmed reads was 89% (SD 4%). Within-
group gene expressions was all strongly correlated: P-values for
statistical significance for all correlated pairs were estimated to lie
below 2.2 310
216
, whereas squared Pearson’s correlation
coefficients r
2
ranged from 0.79 to 0.98 (Fig. S3, available at
http://links.lww.com/PAIN/A710; legend available at http://link-
s.lww.com/PAIN/A724). Tissues that display highest correlated
gene expression were DRG (mean r
2
0.98), followed by SC (0.97)
and whole brain (0.96). Less correlated is whole blood (0.84). The
CFA pain model showed marginally more robustly correlated
gene expression (mean r
2
0.95) than SNI (0.93).
4.8. Replication of publicly available data
Replication of differential expression was attempted using
genome-wide data sets, available in the GEO database,
19
as well
using a survey of differentially expressed, but pain-centric genes.
30
A total of 79 genes wereidentified through significant fold change in
at least 4 different microarray experiments. Of the study’s initial list
of 79 genes, 65 were matched to HGNC’s official gene names. For
each gene surveyed, information about expression fold change
(FC) direction (up, down) has been compiled in 2 tissues, namely
SC and DRG, and for 2 pain models, namely IP and NP. Here, we
have been careful to interpret absence of change direction as not
necessarily evidence for absence of change, therefore leading to
a two-state model of expression fold change per gene. For
replication, we first determined in our data the direction of change
of expression, up or down, based on the sign of the change and
from associated significance P-value P,0.05 (the fold change
was unassigned if P-value isinsignificant). For instance, wefind the
gene cathepsin S, Ctss, upregulated in SNI in DRG (log
2
FC 5
11.2, P55310
25
). Then, for each gene, we compiled the
number of times (up to 4; 2 tissues in 2 pain models) that the survey
and our results committed to call a fold change direction in
corresponding tissue-pain model pairs. For instance, the gene
Ctss has 3 corresponding committed calls: DRG and SC in NP,
and SC in IP. We also tracked fold change mismatches, or
conflicts, in committed calls. For instance, gene Tac1 (tachykinin
precursor 1) has previously been found to be upregulated in DRG in
NP but is found to be downregulated in our work.
4.9. Human and mouse transcriptomes comparison
Human gene expression data from the naive tissues studied in the
current data set were downloaded from
5
: http://xavierlab2.mgh.
harvard.edu/EnrichmentProfiler/download.html. Blood RNA
samples from human subjects diagnosed with low-back pain
were collected in University of Parma hospital at the first visit and
at the 3-month follow-up from the acute episode. The enrolled
subjects reported pain for less than 8 weeks and did not have
other pain episodes within the last 6 months. The subjects with
previous history of cancer or vertebral fracture were excluded.
Blood was collected by venipuncture into Tempus Blood tubes
(Applied Biosystems, Beverly, MA). Total RNA was extracted
using Maxwell 16 LEV simplyRNA Blood Kit (Promega, Madison,
WI) and quantified using the NanoDrop 1000 (Thermo Scientific,
Waltham, MA). The RNA quality was assessed with the 4200
TapeStation (Agilent Technologies, Santa Clara, CA). Blood
transcriptomics data for control subjects were from GEO set
GSE90081.
53
RNA-Seq data were mapped using STAR
aligner,
18
and differential expression of genes was assessed
using FeatureCount
35
followed by Deseq2.
36
4.10. Species-specific gene expression
Identification of species-specific gene expression was performed
by counting in each species how many tissues expresses the
gene (from N 50-4), then computing the bias |N
mouse
-N
human
|;
a bias of 2 or more defines a gene that is species-specific.
4.11. Star plot
In the star plot, the 4 axes were defined: axis 1 identified genes
that were exclusively differentially expressed in the CFA pain
model (green; |p
CFA
|$2 and |p
SNI
|#1); axis 3, genes in SNI only
(purple; |p
SNI
|$2 and |p
CFA
|#1); axis 2, genes in both CFA and
SNI with same 1/1or 2/2fold change sign (pink; |p
CFA
|$2or
|p
SNI
|$2, and 2.0 $p
CFA
/p
SNI
$0.5); and finally axis 4, genes in
both CFA and SNI, but this time with opposite 6or 2/1fold
change sign (orange; |p
CFA
|$2or|p
SNI
|$2, and 2.0 $-p
CFA
/
p
SNI
$0.5). Because each axis required statistical significance at
the alpha 0.01 level with |p|$2,
73
identified genes represented
a confident set to perform further analyses.
4.12. Gene set enrichment analyses
Gene Ontology’s (GO) biological processes pathways have been
used for analysis.
3,22
For each star plot, the most significant
pathway was reported for each of the 4 axes. Once a pathway
was reported, genes that were members of that pathway were
removed from the list of genes and the GSEA was repeated; this
reduced the likelihood that pathways with similar or related
functions be reported. Because we aimed for biological
specificity, we limited analyses to pathways with 1000 genes or
less. In GO, pathways are organized in a hierarchical fashion; we
discarded pathways that have more than 5 generations of
children, again favoring pathways with more defined biological
functions. Finally, we required that a selected pathway has no
more than 50% overlap in gene composition with all those
pathways previously selected, promoting functional diversity, but
more importantly, preventing closely related pathways to be
overrepresented during the selection process.
10 M. Parisien et al.·00 (2019) 1–13 PAIN
®
Copyright © 2019 by the International Association for the Study of Pain. Unauthorized reproduction of this article is prohibited.
4.13. Human genome-wide association studies
We performed GWAS on data from the UK Biobank project
(application number 20802).
2,58
Back pain was selected as
a phenotype of choice because it is present in a large proportion
of the population (about 8%). To define cases, we used the field
3571, where subjects answered “no” to the question “Have you
had back pains for more than 3 months?,” while answering “yes”
to the “back pain” category in field 6159 where the question was
“In the last month have you experienced any of the following that
interfered with your usual activities?” (n 532,209). The limited
time extent is in line with the time span between pain intervention
in the mice and their sacrifice. Control individuals were those that
answered “none of the above” for the question in field 6159 (n 5
163,825). First-hand standard genotyping quality control was
performed by the UK BioBank consortium, and is fully docu-
mented on their web portal (http://biobank.ctsu.ox.ac.uk/crystal/
refer.cgi?id5155580). Genome-wide association study was
performed using genotypic data only, using the 500K cohort
version. People were discarded based on: failed genotyping QC
(as per UK BioBank–heterozygosity rate, genotyping rate, etc.),
genetic vs declared sex mismatch, voluntary retraction, and of
non-“white British” ancestry. Among individuals with first- and
second-degree relatives, we retained the one individual who
displayed the best genotyping rate. Kinship was estimated using
KiNG v2.1.
37
We used PLINK version v1.90b4.6 64-bit (August
15, 2017) to perform logistic regression,
48
using age, sex,
genotyping platform, and the first 5 genetic principal components
as covariables. Post-GWAS SNP QC featured at least 1% minor
allele frequency, genotyping rate better than or equal to 90%, and
P-value for departure from Hardy–Weinberg equilibrium no less
than 10
26
. PLINK was also used to extract linkage disequilibrium
information between SNPs and select tag SNPs based on r
2
#0.5; enrichment of SNPs in genes was evaluated using these
tag SNPs. Genes that span several haploblocks will be
represented by each block’s tag SNP; this might seem to skew
analyses as longer genes will contribute to more SNPs, but usage
of tag SNPs guarantee uncoupled GWAS results within long
genes as tag SNPs are, by definition, in low LD with one another.
4.14. Genes commonly implicated in pain pathways
The list of genes commonly implicated in pain pathways has been
composed from several sources: for human pain genes, we used
a combination of Algynomics’/Cogenics Pain Research Panel
V2
38
and Pain Research Forum’s pain gene resource (https://
www.painresearchforum.org/), and Amigo’s Gene Ontology term
GO:0019233, whereas for mouse pain genes, we used the Pain
Genes Database.
31
We also used Ensembl BioMart service to
provide for a correspondence between human genes and mouse
ones. The pain genes list comprises a total of n 5800 pain genes
(Table S12, available at http://links.lww.com/PAIN/A722; legend
available at http://links.lww.com/PAIN/A724), with matching
human/mouse homologs. All mouse pain genes have human
homologs, whereas a few human pain genes have no mouse
homologs (n 521).
4.15. Protein interaction network
We considered data from the STRING database, a repository of
interacting proteins.
59
Only interactions in the human taxon
(9606) were considered because they were more numerous than
that of the mouse (10090) ones.
4.16. Study approval
All mice experimental procedures were approved by the Animal
Care Committee at McGill University under protocol number 7869
and were performed in full agreement with the ethical guidelines
of the Canadian Council of Animal Care and the guidelines of the
Committee for Research and Ethical Issues of the International
Association for the Study of Pain.
The human low-back pain cohort is part of a larger study
(PainOmics), approved by the Ethical Committee at the hospital
university of Parma, protocol number 43543 version 8, and
registered on clinicaltrials.gov (NCT02037763). All patients
signed a written informed consent before the enrolment.
Conflict of interest statement
The authors have no conflict of interest to declare.
The accession number for the data sets reported in this article
is GEO set GSE111216.
Acknowledgments
The authors thank Dr. Samar Khoury for help with data extraction
from the UK Biobank. The current study was conducted under UK
Biobank application number 20802.
Funding for this work is kindly provided by the Canadian
Excellence Research Chairs (CERC) Program (www.cerc.gc.ca)
grant CERC09 (to L.D.), and the Rita Allen Foundation and the
American Pain Society Award in Pain (to A.K.).
Author contributions: L. Diatchenko, J. S. Mogil, M. Parisien,
A. Samoshkin, A. Khoutorsky, M. Allegri, N. El-Hachem, and
S. N. Tansley designed the analytical plan and experiments.
A. Samoshkin, S. N. Tansley, M. H. Piltonen, and L. J. Martin
performed animal experiment, collected, and processed mouse
tissues. M. Allegri and C. Dagostino were responsible for human
subject collection and samples processing. M. Parisien per-
formed all bioinformatics analyses, except human blood mRNA
processed by N. El-Hachem. M. Parisien, A. Samoshkin,
A. Khoutorsky, and L. Diatchenko wrote the manuscript.
A. Khoutorsky, J. S. Mogil, M. Allegri and L. Diatchenko supervised
the project. All the authors read and edited the final manuscript.
Appendix A. Supplemental digital content
Supplemental digital content associated with this article
can be found online at http://links.lww.com/PAIN/A710,
http://links.lww.com/PAIN/A711, http://links.lww.com/PAIN/A712,
http://links.lww.com/PAIN/A713, http://links.lww.com/PAIN/A714,
http://links.lww.com/PAIN/A715, http://links.lww.com/PAIN/A716,
http://links.lww.com/PAIN/A717, http://links.lww.com/PAIN/A718,
http://links.lww.com/PAIN/A719, http://links.lww.com/PAIN/A720,
http://links.lww.com/PAIN/A721, http://links.lww.com/PAIN/A722,
http://links.lww.com/PAIN/A723, and http://links.lww.com/PAIN/A724.
Article history:
Received 22 October 2018
Received in revised form 6 December 2018
Accepted 14 December 2018
Available online 3 January 2019
References
[1] Abdelaziz DM, Abdullah S, Magnussen C, Ribeiro-da-Silva A, Komarova SV,
Rauch F, Stone LS. Behavioral signs of pain and functional impairment in
a mouse model of osteogenesis imperfecta. Bone 2015;81:400–6.
Month 2019·Volume 00 ·Number 00 www.painjournalonline.com 11
Copyright © 2019 by the International Association for the Study of Pain. Unauthorized reproduction of this article is prohibited.
[2] Allen NE, Sudlow C, Peakman T, Collins R; Biobank UK. UK biobank data:
come and get it. Sci Transl Med 2014;6:224ed4.
[3] Ashburner M, Ball CA, Blake JA, Botstein D, Butler H, Cherry JM, Davis
AP, Dolinski K, Dwight SS, Eppig JT, Harris MA, Hill DP, Issel-Tarver L,
Kasarskis A, Lewis S, Matese JC, Richardson JE, Ringwald M, Rubin GM,
Sherlock G. Gene ontology: tool for the unification of biology. The Gene
Ontology Consortium. Nat Genet 2000;25:25–9.
[4] Balmer TS. Perineuronal nets enhance the excitability of fast-spiking
neurons. eNeuro 2016;3. doi: 10.1523/ENEURO.0112-16.2016.
[5] Benita Y, Cao Z, Giallourakis C, Li C, Gardet A, Xavier RJ. Gene
enrichment profiles reveal T-cell development, differentiation, and
lineage-specific transcription factors including ZBTB25 as a novel NF-
AT repressor. Blood 2010;115:5376–84.
[6] Bolger AM, Lohse M, Usadel B. Trimmomatic: a flexible trimmer for
illumina sequence data. Bioinformatics 2014;30:2114–20.
[7] Bonnans C, Chou J, Werb Z. Remodelling the extracellular matrix in
development and disease. Nat Rev Mol Cell Biol 2014;15:786–801.
[8] Cattaruzza F, Lyo V, Jones E, Pham D, Hawkins J, Kirkwood K, Valdez-
Morales E, Ibeakanma C, Vanner SJ, Bogyo M, Bunnett NW. Cathepsin S
is activated during colitis and causes visceral hyperalgesia by a PAR2-
dependent mechanism in mice. Gastroenterology 2011;141:
1864–74.e1861–1863.
[9] Chaplan SR, Bach FW, Pogrel JW, Chung JM, Yaksh TL. Quantitative
assessment of tactile allodynia in the rat paw. J Neurosci Meth 1994;53:
55–63.
[10] Clark AK, Wodarski R, Guida F, Sasso O, Malcangio M. Cathepsin S
release from primary cultured microglia is regulated by the P2X7 receptor.
Glia 2010;58:1710–26.
[11] Clark AK, Yip PK, Malcangio M. The liberation of fractalkine in the dorsal
horn requires microglial cathepsin S. J Neurosci 2009;29:6945–54.
[12] Cobos EJ, Nickerson CA, Gao F, Chandran V, Bravo-Caparr ´os I,
Gonz ´alez-Cano R, Riva P, Andrews NA, Latremoliere A, Seehus CR,
Perazzoli G, Nieto FR, Joller N, Painter MW, Ma CHE, Omura T, Chesler
EJ, Geschwind DH, Coppola G, Rangachari M, Woolf CJ, Costigan M.
Mechanistic differences in neuropathic pain modalities revealed by
correlating behavior with global expression profiling. Cell Rep 2018;22:
1301–12.
[13] Costigan M, Moss A, Latremoliere A, Johnston C, Verma-Gandhu M,
Herbert TA, Barrett L, Brenner GJ, Vardeh D, Woolf CJ, Fitzgerald M. T-
cell infiltration and signaling in the adult dorsal spinal cord is a major
contributor to neuropathic pain-like hypersensitivity. J Neurosci 2009;29:
14415–22.
[14] Coull JA, Boudreau D, Bachand K, Prescott SA, Nault F, S´
ık A, De
Koninck P, De Koninck Y. Trans-synaptic shift in anion gradient in spinal
lamina I neurons as a mechanism of neuropathic pain. Nature 2003;424:
938–42.
[15] Davis S, Meltzer PS. GEOquery: a bridge between the gene expression
Omnibus (GEO) and BioConductor. Bioinformatics 2007;23:1846–7.
[16] Denning WM, Woodland S, Winward JG, Leavitt MG, Parcell AC, Hopkins
JT, Francom D, Seeley MK. The influence of experimental anterior knee
pain during running on electromyography and articular cartilage
metabolism. Osteoarthritis Cartilage 2014;22:1111–9.
[17] Dib-Hajj SD, Waxman SG. Translational pain research: lessons from
genetics and genomics. Sci Transl Med 2014;6:249sr244.
[18] Dobin A, Davis CA, Schlesinger F, Drenkow J, Zaleski C, Jha S, Batut P,
Chaisson M, Gingeras TR. STAR: ultrafast universal RNA-seq aligner.
Bioinformatics 2013;29:15–21.
[19] Edgar R, Domrachev M, Lash AE. Gene Expression Omnibus: NCBI gene
expression and hybridization array data repository. Nucleic Acids Res
2002;30:207–10.
[20] Freynhagen R, Baron R. The evaluation of neuropathic components in
low back pain. Curr Pain Headache Rep 2009;13:185–90.
[21] Frischknecht R, Heine M, Perrais D, Seidenbecher CI, Choquet D,
Gundelfinger ED. Brain extracellular matrix affects AMPA receptor lateral
mobility and short-term synaptic plasticity. Nat Neurosci 2009;12:
897–904.
[22] Gene_Ontology_Consortium. Gene ontology consortium: going forward.
Nucleic Acids Res 2015;43:D1049–1056.
[23] Goshua A, Craigie S, Guyatt GH, Agarwal A, Li R, Bhullar JS, Scott N,
Chahal J, Pavalagantharajah S, Chang Y, Couban R, Busse JW. Patient
values and preferences regarding opioids for chronic noncancer pain:
a systematic review. Pain Med 2017;19:2469–80.
[24] Haring M, Zeisel A, Hochgerner H, Rinwa P, Jakobsson JET, Lonnerberg
P, La Manno G, Sharma N, Borgius L, Kiehn O, Lagerstrom MC,
Linnarsson S, Ernfors P. Neuronal atlas of the dorsal horn defines its
architecture and links sensory input to transcriptional cell types. Nat
Neurosci 2018;21:869–80.
[25] Hu G, Huang K, Hu Y, Du G, Xue Z, Zhu X, Fan G. Single-cell RNA-seq
reveals distinct injury responses in different types of DRG sensory
neurons. Sci Rep 2016;6:31851.
[26] Ji RR, Chamessian A, Zhang YQ. Pain regulation by non-neuronal cells
and inflammation. Science 2016;354:572–7.
[27] Kim D, Pertea G, Trapnell C, Pimentel H, Kelley R, Salzberg SL. TopHat2:
accurate alignment of transcriptomes in the presence of insertions,
deletions and gene fusions. Genome Biol 2013;14:R36.
[28] Kluzek S, Bay-Jensen AC, Judge A, Karsdal MA, Shorthose M, Spector
T, Hart D, Newton JL, Arden NK. Serum cartilage oligomeric matrix
protein and development of radiographic and painful knee osteoarthritis.
A community-based cohort of middle-aged women. Biomarkers 2015;
20:557–64.
[29] Korczeniewska OA, Husain S, Khan J, Eliav E, Soteropoulos P, Benoliel
R. Differential gene expression in trigeminal ganglia of male and female
rats following chronic constriction of the infraorbital nerve. Eur J Pain
2018;22:875–88.
[30] LaCroix-Fralish ML, Austin JS, Zheng FY, Levitin DJ, Mogil JS. Patterns of
pain: meta-analysis of microarray studies of pain. PAIN 2011;152:
1888–98.
[31] Lacroix-Fralish ML, Ledoux JB, Mogil JS. The pain genes database: an
interactive web browser of pain-related transgenic knockout studies.
PAIN 2007;131:3.e1–4.
[32] Langmead B, Trapnell C, Pop M, Salzberg SL. Ultrafast and memory-
efficient alignment of short DNA sequences to the human genome.
Genome Biol 2009;10:R25.
[33] Legerlotz K, Jones ER, Screen HR, Riley GP. Increased expression of IL-6
family members in tendon pathology. Rheumatology (Oxford) 2012;51:
1161–5.
[34] Leichsenring A, Backer I, Wendt W, Andriske M, Schmitz B, Stichel CC,
Lubbert H. Differential expression of Cathepsin S and X in the spinal cord
of a rat neuropathic pain model. BMC Neurosci 2008;9:80.
[35] Liao Y, Smyth GK, Shi W. featureCounts: an efficient general purpose
program for assigning sequence reads to genomic features.
Bioinformatics 2014;30:923–30.
[36] Love MI, Huber W, Anders S. Moderated estimation of fold change and
dispersion for RNA-seq data with DESeq2. Genome Biol 2014;15:550.
[37] Manichaikul A, Mychaleckyj JC, Rich SS, Daly K, Sale M, Chen WM.
Robust relationship inference in genome-wide association studies.
Bioinformatics 2010;26:2867–73.
[38] Meloto CB, Bortsov AV, Bair E, Helgeson E, Ostrom C, Smith SB, Dubner
R, Slade GD, Fillingim RB, Greenspan JD, Ohrbach R, Maixner W,
McLean SA, Diatchenko L. Modification of COMT-dependent pain
sensitivity by psychological stress and sex. PAIN 2016;157:858–67.
[39] Millecamps M, Tajerian M, Sage EH, Stone LS. Behavioral signs of
chronic back pain in the SPARC-null mouse. Spine (Phila Pa 1976) 2011;
36:95–102.
[40] Mitchell JA, Aronson AR, Mork JG, Folk LC, Humphrey SM, Ward JM.
Gene indexing: characterization and analysis of NLM’s GeneRIFs. AMIA
Annu Symp Proc 2003:460–4.
[41] Mogil JS. Sex differences in pain and pain inhibition: multiple explanations
of a controversial phenomenon. Nat Rev Neurosci 2012;13:859–66.
[42] Mouw JK, Ou G, Weaver VM. Extracellular matrix assembly: a multiscale
deconstruction. Nat Rev Mol Cell Biol 2014;15:771–85.
[43] Niederberger E, Resch E, Parnham MJ, Geisslinger G. Drugging the pain
epigenome. Nat Rev Neurol 2017;13:434–47.
[44] Ossipov MH, Morimura K, Porreca F. Descending pain modulation and
chronification of pain. Curr Opin Support Palliat Care 2014;8:143–51.
[45] Petitjean H, Pawlowski SA, Fraine SL, Sharif B, Hamad D, Fatima T, Berg
J, Brown CM, Jan LY, Ribeiro-da-Silva A, Braz JM, Basbaum AI, Sharif-
Naeini R. Dorsal horn parvalbumin neurons are gate-keepers of touch-
evoked pain after nerve injury. Cell Rep 2015;13:1246–57.
[46] Pizzorusso T, Medini P, Berardi N, Chierzi S, Fawcett JW, Maffei L.
Reactivation of ocular dominance plasticity in the adult visual cortex.
Science 2002;298:1248–51.
[47] Price TJ, Inyang KE. Commonalities between pain and memory
mechanisms and their meaning for understanding chronic pain. Prog
Mol Biol Transl Sci 2015;131:409–34.
[48] Purcell S, Neale B, Todd-Brown K, Thomas L, Ferreira MA, Bender D,
Maller J, Sklar P, de Bakker PI, Daly MJ, Sham PC. PLINK: a tool set for
whole-genome association and population-based linkage analyses. Am J
Hum Genet 2007;81:559–75.
[49] Riediger C, Schuster T, Barlinn K, Maier S, Weitz J, Siepmann T. Adverse
effects of antidepressants for chronic pain: a systematic review and meta-
analysis. Front Neurol 2017;8:307.
[50] Romberg C, Yang S, Melani R, Andrews MR, Horner AE, Spillantini MG,
Bussey TJ, Fawcett JW, Pizzorusso T, Saksida LM. Depletion of
12 M. Parisien et al.·00 (2019) 1–13 PAIN
®
Copyright © 2019 by the International Association for the Study of Pain. Unauthorized reproduction of this article is prohibited.
perineuronal nets enhances recognition memory and long-term
depression in the perirhinal cortex. J Neurosci 2013;33:7057–65.
[51] Rozenbaum M, Rajman M, Rishal I, Koppel I, Koley S, Medzihradszky KF,
Oses-Prieto JA, Kawaguchi R, Amieux PS, Burlingame AL, Coppola G,
Fainzilber M. Translatome regulation in neuronal injury and axon
regrowth. eNeuro 2018;5. doi: 10.1523/ENEURO.0276-17.2018.
[52] Senkov O, Andjus P, Radenovic L, Soriano E, Dityatev A. Neural ECM
molecules in synaptic plasticity, learning, and memory. Prog Brain Res
2014;214:53–80.
[53] Shchetynsky K, Diaz-Gallo LM, Folkersen L, Hensvold AH, Catrina AI,
Berg L, Klareskog L, Padyukov L. Discovery of new candidate genes for
rheumatoid arthritis through integration of genetic association data with
expression pathway analysis. Arthritis Res Ther 2017;19:19.
[54] Shields SD, Eckert WA III, Basbaum AI. Spared nerve injury model of
neuropathic pain in the mouse: a behavioral and anatomic analysis. J Pain
2003;4:465–70.
[55] Sorg BA, Berretta S, Blacktop JM, Fawcett JW, Kitagawa H, Kwok JC,
Miquel M. Casting a wide net: role of perineuronal nets in neural plasticity.
J Neurosci 2016;36:11459–68.
[56] de Souza JB, Grossmann E, Perissinotti DMN, de Oliveira Junior JO, de
Fonseca PRB, Posso IP. Prevalence of chronic pain, treatments,
perception, and interference on life activities: Brazilian population-
based survey. Pain Res Manag 2017;2017:4643830.
[57] Strong JA, Xie W, Coyle DE, Zhang JM. Microarray analysis of rat sensory
ganglia after local inflammation implicates novel cytokines in pain. PLoS
One 2012;7:e40779.
[58] Sudlow C, Gallacher J, Allen N, Beral V, Burton P, Danesh J, Downey P,
Elliott P, Green J, Landray M, Liu B, Matthews P, Ong G, Pell J, Silman A,
Young A, Sprosen T, Peakman T, Collins R. UK biobank: an open access
resource for identifying the causes of a wide range of complex diseases of
middle and old age. PLoS Med 2015;12:e1001779.
[59] Szklarczyk D, Franceschini A, Wyder S, Forslund K, Heller D, Huerta-
Cepas J, Simonovic M, Roth A, Santos A, Tsafou KP, Kuhn M, Bork P,
Jensen LJ, von Mering C. STRING v10: protein-protein interaction
networks, integrated over the tree of life. Nucleic Acids Res 2015;43:
D447–452.
[60] Tajerian M, Alvarado S, Millecamps M, Dashwood T, Anderson KM,
Haglund L, Ouellet J, Szyf M, Stone LS. DNA methylation of SPARC and
chronic low back pain. Mol Pain 2011;7:65.
[61] Thakur M, Crow M, Richards N, Davey GI, Levine E, Kelleher JH, Agley
CC, Denk F, Harridge SD, McMahon SB. Defining the nociceptor
transcriptome. Front Mol Neurosci 2014;7:87.
[62] Torsney C, MacDermott AB. Disinhibition opens the gate to pathological
pain signaling in superficial neurokinin 1 receptor-expressing neurons in
rat spinal cord. J Neurosci 2006;26:1833–43.
[63] Trapnell C, Hendrickson DG, Sauvageau M, Goff L, Rinn JL, Pachter L.
Differential analysis of gene regulation at transcript resolution with RNA-
seq. Nat Biotechnol 2013;31:46–53.
[64] Tsarouhas A, Soufla G, Katonis P, Pasku D, Vakis A, Spandidos DA.
Transcript levels of major MMPs and ADAMTS-4 in relation to the
clinicopathological profile of patients with lumbar disc herniation. Eur
Spine J 2011;20:781–90.
[65] Usoskin D, Furlan A, Islam S, Abdo H, Lonnerberg P, Lou D, Hjerling-
Leffler J, Haeggstrom J, Kharchenko O, Kharchenko PV, Linnarsson S,
Ernfors P. Unbiased classification of sensory neuron types by large-scale
single-cell RNA sequencing. Nat Neurosci 2015;18:145–53.
[66] Vardeh D, Mannion RJ, Woolf CJ. Toward a mechanism-based approach
to pain diagnosis. J Pain 2016;17(9 suppl):T50–69.
[67] Vega-Avelaira D, Geranton SM, Fitzgerald M. Differential regulation of
immune responses and macrophage/neuron interactions in the dorsalroot
ganglion in young and adult rats following nerve injury. Mol Pain 2009;5:70.
[68] Vicuna L, Strochlic DE, Latremoliere A, Bali KK, Simonetti M, Husainie D,
Prokosch S, Riva P, Griffin RS, Njoo C, Gehrig S, Mall MA, Arnold B, Devor
M, Woolf CJ, Liberles SD, Costigan M, Kuner R. The serine protease
inhibitor SerpinA3N attenuates neuropathic pain by inhibiting T cell-
derived leukocyte elastase. Nat Med 2015;21:518–23.
[69] Volkow N, Benveniste H, McLellan AT. Use and misuse of opioids in
chronic pain. Annu Rev Med 2018;69:451–65.
[70] von Mering C, Jensen LJ, Snel B, Hooper SD, Krupp M, Foglierini M,
Jouffre N, Huynen MA, Bork P. STRING: known and predicted protein-
protein associations, integrated and transferred across organisms.
Nucleic Acids Res 2005;33:D433–437.
[71] Wang D, Fawcett J. The perineuronal net and the control of CNS
plasticity. Cell Tissue Res 2012;349:147–60.
[72] Wu S, Marie Lutz B, Miao X, Liang L, Mo K, Chang YJ, Du P, Soteropoulos
P, Tian B, Kaufman AG, Bekker A, Hu Y, Tao YX. Dorsal root ganglion
transcriptome analysis following peripheral nerve injury in mice. Mol Pain
2016;12. doi: 10.1177/1744806916629048.
[73] Xiao Y, Hsiao TH, Suresh U, Chen HI, Wu X, Wolf SE, Chen Y. A novel
significance score for gene selection and ranking. Bioinformatics 2014;
30:801–7.
[74] Xue YX, Xue LF, Liu JF, He J, Deng JH, Sun SC, Han HB, Luo YX, Xu LZ,
Wu P, Lu L. Depletion of perineuronal nets in the amygdala to enhance the
erasure of drug memories. J Neurosci 2014;34:6647–58.
[75] Yates B, Braschi B, Gray KA, Seal RL, Tweedie S, Bruford EA.
Genenames.org: the HGNC and VGNC resources in 2017. Nucleic
Acids Res 2017;45:D619–25.
[76] Zhang X, Wu Z, Hayashi Y, Okada R, Nakanishi H. Peripheral role of
cathepsin S in Th1 cell-dependent transition of nerve injury-induced
acute pain to a chronic pain state. J Neurosci 2014;34:3013–22.
[77] Zhao P, Lieu T, Barlow N, Metcalf M, Veldhuis NA, Jensen DD, Kocan M,
Sostegni S, Haerteis S, Baraznenok V, Henderson I, Lindstrom E,
Guerrero-Alba R, Valdez-Morales EE, Liedtke W, McIntyre P, Vanner SJ,
Korbmacher C, Bunnett NW. Cathepsin S causes inflammatory pain via
biased agonism of PAR2 and TRPV4. J Biol Chem 2014;289:27215–34.
[78] Zhao P, Lieu T, Barlow N, Sostegni S, Haerteis S, Korbmacher C, Liedtke
W, Jimenez-Vargas NN, Vanner SJ, Bunnett NW. Neutrophil elastase
activates protease-activated receptor-2 (PAR2) and transient receptor
potential vanilloid 4 (TRPV4) to cause inflammation and pain. J Biol Chem
2015;290:13875–87.
Month 2019·Volume 00 ·Number 00 www.painjournalonline.com 13
Copyright © 2019 by the International Association for the Study of Pain. Unauthorized reproduction of this article is prohibited.
