The dichotomous role of epiregulin in pain

1. Introduction

Establishment of chronic pain is often a result of the body’s inability to restore physiological homeostasis following acute pain.²⁵ Both acute and chronic pain states have large genetic component which we have now started to identify.^32,52 The Orofacial Pain Prospective Evaluation and Risk Assessment (OPPERA) study was designed to examine and identify biopsychosocial, environmental and genetic factors that contribute to the onset and chronicity of orofacial pain.²⁷ In the OPPERA cohort, case-control association analysis that focused on a common orofacial pain condition, temporomandibular disorders (TMD), using a panel of 358 pain-relevant candidate genes revealed that the genes encoding for the epidermal growth factor receptor (EGFR) and its ligand epiregulin (EREG) had the highest association with TMD risk.³¹ However, genes of other EGFR ligands and receptors have not been tested for their association with human pain phenotypes.

Epiregulin (EREG), a member of the epidermal growth factor (EGF) family of peptide growth factors, plays important roles in angiogenesis and vascular remodeling. It is a potent mitogen with direct and indirect proinflammatory effects.³⁸ EREG binds to EGFR (ErbB1) and ErbB4 (HER4), but also stimulates signaling of ErbB2 (HER2/Neu) and ErbB3 (HER3) through ligand-induced heterodimerization with a cognate receptor. Blocking EGFR with pharmacologically available small molecules and monoclonal antibodies produces analgesia in animals³¹ and chronic pain patients^19,20,22. For the current study, we hypothesized that the EREG-EGFR pathway uniquely contributes to the development and persistence of pain.

For the first time, we systematically screened single-nucleotide polymorphisms (SNPs) in all gene loci belonging to EGFR-family receptors (namely, EGFR, ERBB2, ERBB3 and ERBB4) and their ligands (namely, AREG, BTC, EGF, EPGN, EREG, HBEGF, MUC4, NRG1, NRG2, NRG3, NRG4 and TGFA) for their association with reported clinical pain in the OPPERA cohort. We chose to use characteristic pain intensity (CPI) as an outcome measure because of its clinical significance.¹¹ Our analysis indicated that from the 16 genes screened, only EREG gene SNPs were associated with CPI. Next, we characterized the association between EREG variants and other pain severity phenotypes, namely, acute pain intensity and the number of other chronic painful comorbidities in OPPERA. The same EREG variant that was protective for chronic pain intensity, increased risk for acute pain intensity in OPPERA. We then validated the dichotomous effect of EREG using an independent cohort from the UK biobank (UKB). We also demonstrated the direction of the genetic effect of the identified SNPs on corresponding mRNA expression level through subsequent cis-eQTL analyses from two independent studies. Lastly, the dichotomous role of EREG for pain phenotypes was tested using mouse models of chronic and acute pain sensitivity.

2. Methods

2.1. Cohorts description

The OPPERA cohort was used as a discovery cohort for this study. The study methods have been described in detail elsewhere.² In summary, the prospective cohort study enrolled 3,263 participants between May 2006 and November 2008 at four USA study sites: Baltimore, Maryland; Buffalo, New York; Chapel Hill, North Carolina; and Gainesville, Florida, from which 3,161 were genotyped. To be eligible for enrollment, the participants had to satisfy the selection criteria determined during telephone screening and at the baseline clinical visit. The facial pain characteristics were collected using the OPPERA Comprehensive Pain Symptom Questionnaire (CPSQ), and TMD was diagnosed by trained examiners using the Research Diagnostic Criteria for TMD (RDC/TMD).¹⁰ Participants were followed at quarterly (3-monthly) intervals after the baseline visit with questionnaires and clinical visits. The project’s protocol was approved by the institutional review boards at each OPPERA study site and at McGill University. Written informed consent was obtained from each participant prior to their enrollment.

To replicate our findings from OPPERA, data from UKB was used. Described in detail elsewhere,⁴⁵ the UKB study is a large prospective multicenter study of people living in the United Kingdom that had recruited 503,325 individuals between 2006 and 2010. Follow-up data were collected after 2012. Ethics approval for the UK Biobank study was obtained from the North West Centre for Research Ethics Committee (11/NW/0382), and all participants provided written informed consent. Their participation involved completing questionnaires, undergoing an interview with a trained nurse during which a range of physical measures was collected, and donating samples of blood, urine, and saliva.

2.2. Outcome measures

The OPPERA study used the Research Diagnostic Criteria to define TMD cases.¹⁰ According to this criteria, an individual was deemed as a chronic TMD case if an examiner had confirmed pain in the orofacial area for at least 5 days a month for ≥ 6 months and either ≥ 3 temporomandibular muscle groups or ≥ 1 temporomandibular joint painful to palpation or jaw movement. In addition, the OPPERA study used the CPSQ questionnaire, a self-report instrument, to assess the presence of multiple pain symptoms and associated characteristics.⁴¹

CPSQ asked the following screening question:

“Have you ever had pain in your face, jaw, temple, in front of the ear or in the ear, not including toothache or ear infection?”

If the participant answered “yes” to the aforementioned question, s/he was asked the following:

1. “How many years or months ago did your facial pain begin?”

2. “How would you describe the duration of your facial pain?”

3. “How would you rate your facial pain at the present time, i.e., right now?”

4. “In the past 6 months, how intense was your worst facial pain?”

5. “In the past 6 months, on average, how intense was your facial pain?”

The responses of the above-mentioned question (i) were numerical (years and months) The responses to the question (ii) were collected as either “persistent”, “recurrent” or “one time”, whereas, the responses to the questions (iii) to (iv) were collected using a numerical rating scale (NRS) where 0 was marked as ‘no pain’ and 10 was ‘pain as bad as it could be’. Measuring pain intensity on a numerical rating scale (NRS) is a valid and clinically meaningful measure of pain severity.¹³ CPI is an arithmetic mean of the three NRS ratings, namely, pain right now, worst pain in six months and average pain in six months. Contrary to NRS alone, CPI is temporally stable³⁴, provides a more reliable estimation of pain severity,^11,48 and has been demonstrated to be a significant predictor of TMD chronicity.¹² Hence, we chose CPI as our primary outcome measure. Out of 3,161 genotyped OPPERA participants, 399 participants were excluded due to missing or poor-quality phenotype data. Out of the remaining 2,762 participants, 1,626 never had pain in the facial region. CPI scores were not calculated for these 1,626 participants. 894 out of the remaining 1,136 participants had facial pain for more than 3 months. We restricted our analysis to the participants with either persistent (n = 124) or recurrent (n = 264) facial pain to comply with the latest definition of chronic pain according to the international Association for the Study of Pain (IASP).⁴⁶ The CPI scores of these 388 OPPERA participants at baseline were considered as the chronic pain intensity (Figure 1A). Follow-up CPI scores of the controls with a CPI at baseline of zero were considered as an acute pain intensity marker (n = 213) (Figure 1A) and participants with acute CPI > 0 were considered as acute facial pain cases (n = 112). Other phenotypes from OPPERA included the number of comorbid pain conditions present from fibromyalgia, chronic fatigue syndrome, irritable bowel syndrome, interstitial cystitis, arthritis, and chronic pelvic pain, and, TMD caseness, as described earlier.

Figure 1:

Selection of acute and chronic pain phenotypes in the (A) OPPERA and (B) UKB cohorts.

As part of the UKB data collection framework, participants were asked: “In the last month, have you experienced any of the following that interfered with your usual activities?” (UKB data-field ID: 6159). Participants could choose all that apply from the following options: headache, facial pain, neck or shoulder pain, back pain, stomach or abdominal pain, hip pain, knee pain, pain all over the body, none of the above, and prefer not to answer. We generated a quantitative trait ranging from 0 to 8, corresponding to the number of sites reported as painful. Those reporting to have “pain all over the body” were assigned the maximum score of 8. If the site was painful for not more than a month, it was counted as an acute pain site. If a bodily site remained painful on follow-up after two or more years, then it was counted as a chronic pain site.

Caseness for participant’s pain sites were established as follows in UKB (Figure 1B):

1. Participants who reported no pain sites at baseline and follow-up were treated as controls.

2. If a participant reported pain at a particular body site for 1 month but not for 3 months, s/he was classified as an acute pain case for that particular site.

3. If a participant reported pain at a particular site for 1 month at baseline and at the same site for more than 3 months at follow-up, s/he was classified as a chronic case for that particular site.

2.3. Genotyping

Peripheral venous whole blood was collected at each OPPERA site into 5 mL polyethylene tubes containing ethylene-diamine tetra-acetic acid (EDTA) (Vacutainer; Beckton Dickinson and Company, Franklin Lakes, New Jersey), and the tubes were stored in −80°C freezer.⁴² Genomic DNA was purified utilizing the protocols of Qiagen™ extraction kits. Samples were genotyped using the Illumina HumanOmni2.5Exome-8v1A array (Illumina, Inc., San Diego, California) at the Center for Inherited Disease Research (Johns Hopkins University, Baltimore, Maryland). The details of genotyping and QC procedures have been described elsewhere.⁴³ Genotyping results were returned for 3,221 unique samples, representing the study participants. All the genotyped SNPs with minor allele frequency (MAF) greater than 5% in gene loci belonging either to EGF-family receptors (namely, EGFR, ERBB2, ERBB3 and ERBB4) or ligands (namely, AREG, BTC, EREG, EGF, EPGN, EREG, HBEGF, MUC4, NRG1, NRG2, NRG3, NRG4 and TGFA) were chosen for the association analyses (n = 2,407).

UKB’s genetic data for 488,288 participants was used. As described in detail elsewhere,⁸ blood samples were collected from participants on their visit to a UKB assessment center and the samples were stored at the UKB facility in Stockport, UK. Over a period of 18 months, samples were retrieved, DNA was extracted, and shipped to Affymetrix Research Services Laboratory for genotyping. A subset of 49,940 participants was genotyped using the Applied Biosystems UK BiLEVE Axiom Array by Affymetrix (now part of Thermo Fisher Scientific).⁴⁹ Remaining 438,348 participants were genotyped using the closely related Applied Biosystems UK Biobank Axiom Array that shares 95% marker content with the UK BiLEVE Axiom Array. Routine quality checks were carried out during the process of sample retrieval, DNA extraction, and genotype calling.

2.4. Mouse Subjects

Male adult (7 to 9 weeks of age) CD-1 [Crl:CD1(ICR)] mice were acquired from Charles River Laboratories (Saint Constant, QC) and used for all experiments. All mice were housed in groups of 4 upon arrival and procedures were conducted in accordance with the animal care standards set forth by the Canadian Council on Animal Care (CCAC) and were approved by the University of Toronto’s Biosciences Panel on Laboratory Animal Care. All animals were maintained within a temperature-controlled environment (23 ± 1° C) with a 12 :12 h light: dark cycle. A compressed cotton nesting square and crinkled paper bedding were provided in each cage as a source of environmental enrichment. All mice had access to food (Harlan Teklad 8604) and water ad libitum.

2.5. Anti-EREG monoclonal antibody (mAb)

A blocking/neutralizing EREG monoclonal antibody (mAb) (NBP2–21992, Novus Biologicals, Oakville, ON) was diluted in phosphate buffered saline (PBS) and administered directly into the tail vein (5 μg/5 μl). Control mice were injected with an equivalent volume of phosphate buffered saline.

2.6. Mouse Behavioral Assays

von Frey tests.

Mechanosensitivity was measured using the SUDO up-down method with von Frey hairs to estimate the 50% withdrawal threshold in pressure units (g/mm²).⁷ Mice were placed on a perforated metal floor (with 5-mm diameter holes placed 7 mm apart) within small Plexiglas cubicles (9 × 5 × 5 cm high) and a set of 8 calibrated von Frey fibers (Stoelting Touch Test Sensory Evaluator Kit no. 2 to no. 9; ranging from ~0.015 g to ~1.3 g of force) were applied to the plantar surface of the hind paw until the fibers bowed and then held for 3 seconds. The threshold force required to elicit withdrawal of the paw (median 50% withdrawal) was determined twice on each hind paw (and averaged) for all measurements, with sequential measurements separated by at least 20 minutes.

Complete Freund’s Adjuvant (CFA):

CFA (50%; Sigma-Aldrich) was injected intraplantar in a volume of 20 μL into the left hind paw using a 100-μL microsyringe with a 30-gauge needle. Mice were tested for mechanical thresholds of the injected hind paw using the von Frey test as described above, before and at selected time-points following CFA injection. The EREG mAb or vehicle control was injected 1 day or 3 days following CFA injection.

Spared nerve injury (SNI):

SNI, an experimental nerve injury designed to produce neuropathic pain, was performed under isoflurane/oxygen anesthesia as described previously^6,9. Briefly, using an operating microscope (X40), 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). Mice recovered on a heating pad – carefully monitored to prevent overheating – until ambulatory as per standard operating procedures. Mice were tested for mechanical sensitivity before and 14-days after surgery using the von Frey test as described above, except that the “spared” sural region was targeted for SNI by applying the fibers to the hind paw. Following von Frey mechanical testing on day 14, mice were injected (i.v.) with the EREG mAb or vehicle control and then tested for mechanical sensitivity 16-, 19-, 21- and 23-days post-surgery.

Capsaicin assay:

Mice were placed on a tabletop within Plexiglas cylinders (30 cm high; 30 cm diameter) and allowed to habituate for 15 min. Mice then received a subcutaneous injection of capsaicin (2.5 μg; Sigma-Aldrich) into the plantar left hindpaw (20 μl) and were digitally videotaped for 10 min. Video files were later scored for the total duration (s) of licking/biting (i.e. nocifensive behavior) of the injected paw. Two hours following capsaicin behavior, mechanosensitivity was measured using the SUDO-method (as described above). Care was taken to avoid the capsaicin injection site when testing mechanosensitivity. In these experiments, mice were pre-treated with the EREG mAb or vehicle control 2 days prior to capsaicin injection. Withdrawal thresholds for the uninjected paw were also measured to determine whether the EREG mAb altered mechanical thresholds per se.

2.7. Antibody measurements

The concentration of the epiregulin mAb antibody conjugated to Alexa-647 (NBP2–21992AF647, Novus Biologicals) was determined using the Cytation 5 Cell Imaging Multi-Mode Reader (Biotek). In brief, either PBS or the epiregulin mAB-Alexa-647 was injected i.v. and after 2-, 5-, or 7-days mice were euthanized to collect blood. Blood was centrifuged at 5000 rpm for 20 min at 4°C to separate plasma and kept at −80°C until analysis. Plasma samples (100 μl/well) along with known standard concentrations of the epiregulin mAb were loaded into a 96-well microplate for fluorescence-based intensity measurement (Invitrogen). Using the multi-mode plate reader, fluorescence intensity of Alexa-647 was measured with a bandwidth of 20 nm (640 nm excitation, 681nm emission). A standard curve was generated based on the fluorescence intensity values from the known standard concentrations, which was then used to calculate the concentration of the epiregulin mAB in the plasma samples.

2.8. Data analyses

The additive model of inheritance was assumed for all genetic analyses. The family-wise error rate was controlled using the Benjamini-Hochberg’s False Discovery Rate (FDR) method⁵ at 5% threshold. Pain phenotypes were considered as dependent variables and minor allele counts of SNPs were independent variables for Poisson and logistic regression models for count and binary outcome measures, respectively. For initial screening of all the 2,407 SNPs in EGFR family of receptors and ligands against chronic pain intensity, multivariate linear regression was conducted using PLINK (Broad Institute, Cambridge MA), version 1.09.³⁶ Haplotype analyses were carried out using Haplo.stats v1.7.7 (R-package)²⁴ which implements an expectation-maximization–derived score to test for a statistically significant association between haplotypes and outcome measurements. The statistical methods implemented in this R package assume that all subjects are unrelated and that haplotypes are ambiguous (due to unknown linkage phase of the genetic markers), while also allowing for missing alleles. Hence, unrelated participants from the OPPERA and UKB cohorts were selected using a 2^nd degree relatedness threshold as implemented in Kinship-based INference for Genome-wide association studies (KING).²⁹ The effect of all rare haplotypes with the estimated frequency > 5% in OPPERA and > 2% in UK biobank were compared against the effect of one ancestral haplotype. Generalized linear modelling was used to test for an association between genotypes and phenotypes. As the OPPERA participants were recruited in four study sites, recruitment sites were introduced as covariates in the regression models. Age, gender and the first three principle components (the markers of ancestry) were also included as covariates to adjust for population stratification. Similarly, the UKB data analyses were corrected for gender, age, ethnic background, and genotyping platforms. R v3.5.2 was used as the language and environment for statistical computation. Haplotype structure of EREG was analyzed using Haploview v 4.2.⁴

For eQTL analysis, two large-scale datasets, namely, the Framingham Heart Study (FHS),¹⁸ and the Genotype-Tissue Expression (GTEx) project¹⁴ (version 7) were used. The FHS data are available at dbGaP under the accession numbers phs000342 and phs000724. The GTEx data are also available at dbGAP under the accession number phs000424.v7.p2.

For mouse experiments, von Frey data were analyzed using repeated-measures ANOVA, followed where appropriate by Tukey’s honest significant difference (HSD) post-hoc test. Capsaicin nocifensive behavior was analyzed using independent t-tests.

3. Results

3.1. Among all EGFR receptors and ligands, only EREG was associated with pain.

Our primary outcome measure, CPI, was well characterized in the OPPERA study. Hence, we used OPPERA as our discovery cohort. A total of 2,407 genotyped SNPs with MAF > 5% situated within the 16 genes of EGFR family receptors (i.e. EGFR, ERBB2, ERBB3, and ERBB4) and their ligands (AREG, BTC, EGF, EPGN, EREG, HBEGF, MUC4, NRG1, NRG2, NRG3, NRG4, and TGFA) were screened for an association with CPI in OPPERA (Figure 2, Supplementary Table 1). Only 7 SNPs passed the FDR threshold of 5% after correcting for age, gender, recruitment site and the first three principle components (Table 1). All significantly associated SNPs were located in the EREG gene with their minor alleles associated with less pain. Therefore, EREG was chosen as our primary candidate gene for further investigation.

Figure 2:

Quantile-quantile plot of 2,407 SNPs in genes coding for EGFR superfamily receptors and ligands, showing a significant association (FDR < 0.05) between seven EREG SNPs and chronic characteristic pain intensity (CPI) in OPPERA cohort.

Table 1:

Linear regression analyses of all the genes in EGFR receptor family and its ligands, and, chronic characteristic pain intensity (CPI) in OPPERA cohort, corrected for age, gender and the first three principal components. Only significant results (FDR ≤ 5%) are presented.

Gene	SNP	Ref	Min	Base pair location	β	P-val	FDR
EREG	rs10518126	G	A	75243119	−11.09	0.0000078	0.00944**
EREG	rs57839099	G	A	75243813	−11.09	0.0000078	0.00944**
EREG	rs200889776	G	A	75240770	−10.41	0.00002	0.011*
EREG	rs57933408	G	A	75243828	−10.51	0.00002	0.011*
EREG	rs201835071	G	A	75237587	−9.99	0.00003	0.011*
EREG	rs72859363	A	G	75246112	−10.24	0.00003	0.011*
EREG	rs6836436	A	C	75230930	−8.84	0.00013	0.041*

SNP: single nucleotide polymorphism (rs ID); Ref: reference allele; Min: minor allele; β: slope of least-squares line; FDR: false discovery rate;

^‘**’:FDR ≤ 0.01

^‘*’:0.05

3.2. EREG gene has two functional minor haplotypes.

Out of 2,407 tested SNPs in the EGFR family of receptors and ligands, seven SNPs within EREG were found to be significantly associated with CPI. The association results of the EREG SNPs were visualized along with their local linkage disequilibrium (LD), recombination patterns, and genomic region position. This regional plot of EREG (Figure 3A) uncovered substantial LD structure between EREG SNPs, with one LD block within the gene (Figure 3B, Supplementary Table 2). Furthermore, haplotype analysis identified two minor haplotypes, herein referred to as H2 and H3, with frequencies 17.3% and 5.8% respectively, in OPPERA. All seven significant SNPs were markers for the H3 haplotype of EREG (Figure 3C), while we’ve previously reported the association of the H2 haplotype of EREG, marked by the functional SNP, rs2367707 with TMD³¹, the H3 haplotype of EREG was not detected in our earlier analysis due to its relatively low frequency. A marker of the H3 haplotype, rs6836436, was deemed potentially functional as it was located in the 5’ UTR region of EREG. Lastly, the presence of reference allele (A) at rs1993665 exclusively marked the major haplotype (herein referred to as H1). Hence, rs1993665, rs6836436, and rs2367707 were chosen as the markers of H1, H2 and H3 haplotypes of EREG for haplotype association analyses in both, OPPERA and UKB cohorts. Their minor allele counts and frequencies in OPPERA and UK biobank are shown in Table 2. Haplotype frequencies based on the 3 marker SNPs in EREG as derived using expectation-maximization (E-M) algorithm were 67.33 % and 74.65 % for H1, 20.39 % and 19.91 % for H2, and, 7.61 % and 2.77 % for H3 in OPPERA and UKB, respectively (Table 3). eQTL databases, namely, FHS and GTEx were scanned for the H2 (rs2367707) and H3 (rs6836436) SNP markers. eQTL analyses revealed that both minor alleles at rs2367707 and rs6836436 were associated with decreased mRNA levels of EREG in the peripheral blood (Table 4).

Figure 3:

The EREG gene has two minor haplotypes. (A) Regional plot of EREG. (B) Illustration of the 16 SNPs in EREG Linkage Disequilibrium plot, numbers inside each cell indicate r² values, color reflects D’ value, ranging from white to red, (i.e. 0 to 1). (C) The sequence of three haplotypes with frequency > 5% within EREG gene locus. Major and minor alleles of EREG SNPs genotyped in OPPERA, SNPs significantly associated with CPI from Figure 2 are highlighted in green. Marker SNPs, namely, rs1993665, rs2367707, and rs6836436, for haplotypes H1, H2, and H3, respectively, are in highlighted in yellow.

Table 2:

Minor allele counts of the marker SNPs in OPPERA and UK biobank cohorts:

MAC	Cohort	H1 marker SNP (rs1993665)		H2 marker SNP (rs2367707)		H3 marker SNP (rs6836436)
		n	Frequency	n	Frequency	n	Frequency
0	OPPERA	1,361	49.4 %	1,672	60.7 %	2,368	86.0 %
1		1,082	39.3 %	956	34.7 %	349	12.7 %
2		312	11.3 %	127	4.6 %	38	1.4 %
0	UK biobank	267,521	58.5 %	281,970	61.7 %	432,835	94.7 %
1		163,181	35.7 %	154,109	33.7 %	23,650	5.2 %
2		26,568	5.8 %	21,191	4.6 %	785	0.2 %

MAC: minor allele counts; n: number of participants.

Table 3:

Haplotype frequencies of EREG as estimated through Expectation – Maximization (E-M) Algorithm:

Haplotype*	OPPERA (n = 2,755)	UK biobank (n = 473,879)
H1	67.33 %	74.65 %
H2	20.39 %	19.91 %
H3	7.61 %	2.77 %
Log-likelihood	−4400.5	−563884.5
lr stat for no LDϮ	3404.9	588766.6
p-value	8.63 × 10−08	4.99 × 10−23

^*haplotypes with frequencies > 5 % in OPPERA and > 2 % in UK biobank;

^Ϯlikelihood ratio test statistic contrasting the log-likelihood for the estimated haplotype frequencies versus the log-likelihood under the null assuming that the alleles from all the three loci are in linkage equilibrium.

Table 4:

cis – eQTL in blood for the marker SNPs of minor haplotypes in EREG:

Cohort	n	SNP	Haplotype	β	P-value
FHS	2,770	rs2367707	H2	−0.14	1.2 × 10 −16
		rs6836436*	H3	NA	NA
GTEx	369	rs2367707	H2	−0.09	2.4 × 10 −02
		rs6836436	H3	−0.27	1.4 × 10 −04

FHS: Framingham Heart Study; GTEx: Genotype-Tissue Expression project; n: number of participants; SNP: single nucleotide polymorphism; β: slope of least-squares line.

^*H3 marker SNP, rs6836436, was not genotyped in Framingham Heart Study cohort.

3.3. H3 and H2 haplotypes of EREG protects from chronic pain

We hypothesized that genetic variability within the EREG locus may affect chronic pain intensity. From the OPPERA cohort, chronic pain intensity at baseline, TMD case status, and the number of chronic pain comorbidities were chosen as the chronic pain phenotypes for this study. Haplotype association analyses confirmed our previously reported protective role for the H2 haplotype for TMD case status (n = 2,755, OR = 0.84, FDR = 0.032, Figure 4A).³¹ Furthermore, the H3 haplotype of EREG was associated with lower chronic pain intensity (n = 388, β = −8.06, FDR = 0.033, Figure 4B) and was marginally protective against the number of chronic pain comorbidities (n = 2,748, β = −0.07, FDR = 0.08).

Figure 4:

H3 and H2 haplotypes of EREG protects from chronic clinical pain. (A-B) OPPERA cohort. (A) Bar plot of average minor allele counts of rs1993665, rs2367707 and rs6836436, (markers for haplotypes H1, H2 and H3, respectively) among chronic TMD cases and controls and (B) Plot of mean chronic pain intensity at baseline for minor allele counts of rs1993665, rs2367707 and rs6836436, (markers for haplotypes H1, H2 and H3, respectively) in the OPPERA cohort. (C-D) UKB cohort. (C) Bar plot of average minor allele counts of rs1993665, rs2367707 and rs6836436, (markers for haplotypes H1, H2 and H3, respectively) among chronic pain cases (at least one chronic pain site) and controls and (D) Plot of mean number of chronic pain sites for minor allele counts of rs1993665, rs2367707 and rs6836436, (markers for haplotypes H1, H2 and H3, respectively) in the UKB cohort. Symbols represent mean ± SEM; False Discovery Rates (FDR) were derived by generalized linear modelling for haplotype association; *FDR < 0.05; **FDR < 0.01.

For validation of these results, we used the UKB cohort. Pain intensity was not collected in the UKB cohort. Hence, we used the number of reported chronic pain body sites as a substitute for chronic pain intensity. Although pain severity in terms of intensity and anatomical extent are different phenotypes, they are correlated.^3,50 Painful sites were considered chronic if the pain persisted at the same site for at least two years and the number of reported chronic pain body sites was used as a substitute for chronic pain intensity. Additionally, the substantial size of the UKB cohort allowed us to consider pain at each of the eight reported body sites as individual chronic pain phenotypes for this analysis. H2 was protective for the report of at least one chronic pain site (n = 196,534, OR = 0.95, FDR = 0.031, Figure 4C) in UKB, while the presence of the H3 haplotype was associated with a decrease in the number of chronic pain sites (n = 196,534, β = −0.21, FDR = 0.003, Figure 4D) and was protective for chronic hip pain (n = 191,669, OR = 0.66, FDR = 0.028). The results of all chronic pain phenotypes are summarized in Table 5. Together, both the H2 and H3 haplotypes showed protective properties towards chronic pain with H3 haplotype displaying a stronger effect size consistent with its eQTL strength (defined by the slope, β, of the eQTL analysis, Table 4).

Table 5:

Haplotype association to analyze the relationship between chronic pain phenotypes and EREG haplotypes:

Cohort	Chronic pain phenotype	n	prevalence	Haplotype II		Haplotype III
				ESTIMATE	FDR	ESTIMATE	FDR
OPPERA	Chronic pain intensity	388	88.66 % Ϯ	β = −3.11	0.245	β = −8.06	0.033*
	TMD case status	2,755	31.87 %	OR = 0.84	0.032*	OR = 0.79	0.083
	No. of chronic pain comorbidities	2,748	78.93 %Δ	β = −0.04	0.174	β = −0.07	0.080
UK biobank	Chronic pain all over the body	190,866	0.05 %	OR = 1.08	0.660	OR = 0.95	0.913
	Chronic stomach or Abdominal pain	191,111	0.18 %	OR = 1.09	0.340	OR = 0.50	0.055
	Chronic headache	191,859	0.56 %	OR = 0.91	0.112	OR = 0.96	0.804
	Chronic neck or shoulder pain	192,507	0.90 %	OR = 0.96	0.309	OR = 0.80	0.082
	Chronic back pain	192,922	1.11 %	OR = 0.98	0.521	OR = 0.83	0.081
	Chronic hip pain	191,669	0.47 %	OR = 0.98	0.700	OR = 0.66	0.028*
	Chronic knee pain	192,796	1.05 %	OR = 0.94	0.138	OR = 0.99	0.966
	Chronic facial pain	190,832	0.03 %	OR = 0.77	0.305	OR = 0.27	0.204
	No. of chronic pain sites	196,534	2.93 %Δ	β = − 0.031	0.107	β = − 0.15	0.003*
	At least one chronic pain site			OR = 0.95	0.031*	OR = 0.90	0.131

CPI: characteristic pain intensity; n: valid number of participants;

^Ϯprevalence of TMD cases among n with valid chronic pain intensity scores at baseline; β: slope of least – squares line; OR: Odds Ratio;

^Δ:prevalence of at least one chronic pain comorbidity / chronic pain site among total number of participants; FDR: False Discovery Rate;

^‘**’:FDR < 0.001

^‘*’:0.05

3.4. H3 haplotype of EREG is a risk for acute pain.

Next, we tested the association of functional EREG haplotypes with acute clinical pain. To study the effects of minor haplotypes of EREG on acute pain, the first quarterly follow-up CPI scores measured in controls with baseline CPI = 0 were chosen as acute pain phenotype from the OPPERA cohort. Participants with no facial pain at baseline but CPI > 0 after three months were considered acute facial pain cases (Figure 5A). The number of reported acute (not more than 3 months) painful body sites was used as a marker of acute pain severity in UKB. In addition, participants with at least one reported acute painful site (n = 137,852) were contrasted against participants with no reported pain at all (n = 333,921). Haplotype H3 was strongly associated with acute pain but, unexpectedly, in the opposite direction compared to chronic pain. The presence of minor haplotype H3 was associated with an increase in acute pain intensity at follow-up (n = 213, β = 8.68, FDR = 0.039, Figure 5B) in the OPPERA cohort. In the UKB, H3 was a risk factor for self-reported acute pain of at least one site (n = 471,773, OR = 1.34, FDR = 0.0002, Figure 5C) and the total number of acute pain sites (n = 471,773, β = 0.028, FDR = 0.003, Figure 5D). Moreover, haplotype H3 was associated with increased odds of having acute pain all over the body (n = 335,565, OR = 1.33, FDR = 0.0003). No significant association was detected between acute pain phenotypes and haplotype H2. The results of acute pain phenotypes are summarized in Table 6.

Figure 5:

H3 haplotype of EREG is a risk for acute clinical pain. (A-B) OPPERA cohort. (A) Bar plot of average minor allele counts of rs1993665, rs2367707 and rs6836436, (markers for haplotypes H1, H2 and H3, respectively) among acute facial pain cases and controls and (B) Plot of mean of acute pain intensity at follow-up in controls for minor allele counts of rs1993665, rs2367707 and rs6836436, (markers for haplotypes H1, H2 and H3, respectively) in the OPPERA cohort. (C-D) UKB cohort. (C) Bar plot of average minor allele counts of rs1993665, rs2367707, and rs6836436, (markers for haplotypes H1, H2, and H3, respectively) among acute pain cases (at least one acute pain site) and controls. (D) A plot of the mean number of acute pain sites for minor allele counts of rs1993665, rs2367707, and rs6836436, (markers for haplotypes H1, H2, and H3, respectively) in the UK biobank (UKB) cohort. Symbols represent mean ± SEM; False Discovery Rates (FDR) were derived by generalized linear modelling for haplotype association; *FDR < 0.05; **FDR < 0.01.

Table 6:

Haplotype association to analyze the relationship between acute pain phenotypes and EREG haplotypes:

Cohort	Acute Pain Phenotype	n	prevalence	Haplotype II		Haplotype III
				ESTIMATE	FDR	ESTIMATE	FDR
OPPERA	Acute pain intensity	213	52.58%Ϯ	β = −1.83	0.517	β = 8.68	0.0390*
	Acute facial pain			OR = 0.76	0.289	OR = 1.99	0.121
UK biobank	Acute pain all over the body	335,565	0.87 %	OR = 1.01	0.909	OR = 1.33	0.0003**
	Acute stomach or Abdominal pain	352,780	5.34 %	OR = 0.99	0.261	OR = 0.94	0.066
	Acute headache	388,126	13.96 %	OR = 0.99	0.246	OR = 1.00	0.771
	Acute neck or shoulder pain	369,264	9.57 %	OR = 1.02	0.083	OR = 1.01	0.611
	Acute back pain	373,432	10.58 %	OR = 0.99	0.672	OR = 0.98	0.447
	Acute hip pain	346,291	3.57 %	OR = 0.99	0.820	OR = 0.98	0.758
	Acute knee pain	356,768	6.40 %	OR = 1.00	0.810	OR = 1.03	0.408
	Acute facial pain	338,577	1.37 %	OR = 0.97	0.405	OR = 0.95	0.541
	No. of acute pain sites	471,773	29.22 %Δ	β = −0.001	0.752	β = 0.028	0.003*
	At least one acute pain site			OR = 1.01	0.909	OR = 1.34	0.0002**

CPI: characteristic pain intensity; n: valid number of participants/cases and controls; number of controls (no pain at all) is 333,936 for all the UK biobank phenotypes;

^Ϯprevalence of participants with acute pain intensity at follow-up > 0 among n with valid acute acute pain intensity scores at follow-up; β: slope of least – squares line; OR: Odds Ratio;

^Δprevalence of at least one acute pain site. FDR: False Discovery Rate;

^‘**’:FDR < 0.001

^‘*’:0.05

3.5. EREG has a dichotomous role in pain behavior in mice

Since our human genetic association results indicated that the H2 and H3 haplotypes may have a protective role against chronic pain and that these haplotypes were associated with lower EREG mRNA in the blood (Table 4), we hypothesized that blocking EREG would reduce pain hypersensitivity in mouse models of chronic pain. For these experiments, we blocked EREG by peripheral administration of an EREG mAb (5 μg) directly into the tail vein. Strikingly, mice injected with a single intravenous injection of the EREG mAb recovered quicker than control mice when the mAb was administered 3 days post-CFA (i.e. peak of allodynia) (two-way ANOVA, treatment x repeated measures: F_3,42=5.53, p=0.003, Figure 6A). Since mice recovered from CFA-induced hypersensitivity by day 7 (t₁₄=1.35, p=0.2), it was difficult to determine whether the EREG mAb had long-lasting effects on allodynia or was specific for inflammatory pain. Thus, we assessed whether the EREG mAb reversed mechanosensitivity using the SNI model. For the SNI experiments, we administered the mAb 14 days following surgery to ensure that post-operative inflammation had resolved and that chronic mechanosensitivity had been established. As shown in Figure 6B, administration of the EREG mAb reversed mechanical hypersensitivity produced by SNI for up to 1-week following administration (two-way ANOVA, treatment x repeated measures: F_5,84=10.49, p<0.001). In a subset of mice, we tracked the levels of the EREG mAb (conjugated to Alexa 647) by measuring fluorescence intensity in the blood plasma 2-, 5-, and 7- days following mAb administration. The concentration of the EREG mAb was significantly elevated at all time-points post-injection compared with control mice (two-way ANOVA, treatment x day post-injection: F_2,17=4.4, p=0.027, Figure 6C).

Figure 6:

The effects of systemically administering an EREG monoclonal antibody (mAb, 5 μg) in mouse models of pain. (A) Mice injected with the EREG mAb 3 days following CFA, as indicated by the arrow have higher paw withdrawal thresholds (g/mm²) compared with control mice 5 days post-CFA; n = 8/group. (B) The EREG mAb or vehicle control was administered 14 days following SNI surgery, as indicated by the arrow. A single administration of the mAb reverses mechanical allodynia for up to one week. (C) The concentration of EREG mAb in the blood plasma of mice following a single tail vein administration; n = 4/group. (D) Mice injected with the EREG mAb 1 day following CFA, as indicated by the arrow have lower paw withdrawal thresholds (g/mm²) than control mice 7 days post-CFA; n = 8/group. (E) Pre-treatment with the EREG mAb, 2 days before testing increases nocifensive behavior in the intraplantar capsaicin test of acute pain; n = 18–20/group. (F) A subset of mice from E were tested for mechanosensitivity following intraplantar capsaicin injection; n = 8/group. Mice injected with the EREG mAb have lower paw withdrawal thresholds (g/mm²) in the capsaicin injected paw, but not the uninjected paw when compared with controls. BL: baseline.*p < 0.05; **p < 0.001; ***p < 0.001 compared with vehicle at the indicated time points. †p < 0.05; ††p < 0.001 compared with capsaicin injected paw in F.

Considering that the human genetic association results indicated that lost-function H2 and H3 haplotypes were a risk for acute pain, we further hypothesized that blockade of EREG during the development of pain states may prolong or enhance hypersensitivity. Administration of the EREG mAb 1-day following CFA (i.e. during the development of hypersensitivity) delayed the natural recovery time-course of mechanosensitivity (two-way ANOVA, treatment x repeated measures: F_2,28=5.4, p=0.001, Figure 6D). This effect appeared to be independent of an inflammatory state per se, as levels of white blood cells were not different in mice that received the EREG mAb or vehicle control (Supplementary Table 3). Next, we injected capsaicin as a model of acute pain (measuring both nocifensive and mechanical withdrawal thresholds). Pre-treatment with the EREG mAb 2 days before capsaicin increased nocifensive behavior (t₁₄=2.54, p=0.02, Figure 6E) and decreased mechanical pain thresholds in the injected, but not uninjected paw (two-way ANOVA, treatment x paw: F_1,28=4.91,p=0.03, Figure 6F).

4. Discussion

In this study, we report the results of genetic screening within 16 genes of the EGFR family of receptors and ligands for their association with acute and chronic pain states. First, we identified EREG as the strongest sole contributor with two functional genetic variants and discovered a new haplotype H3 of EREG marked by the presence of a minor allele at SNP rs6836436. Second, and more surprisingly, we found that EREG has a dichotomous role in the pathophysiology of pain with its loss-of-function variants associated with decreased chronic pain severity but increased acute pain severity. We validated the results of this association analysis using mouse models of pain, where we found that neutralizing EREG with a mAb either reversed or enhanced pain behavior in chronic versus acute pain models, respectively.

Together, our results combined with previous reports,^23,31 suggest that EREG mitigates pain during the early stages of its development but eventually contributes to the establishment of chronic pain. Therefore, in addition to current pharmacotherapy of chronic pain conditions with non-steroidal anti-inflammatory drugs (NSAIDs), opioids, corticosteroids, anxiolytics, muscle relaxants, antidepressants, anticonvulsants and benzodiazepines, inhibition of EREG-EGFR complex formation could serve as a novel strategy to control chronic pain. EREG-targeted therapy would not only be efficient in managing chronic pain but may provide a safer alternative to currently available drugs for EGFR inhibition, as EGFR inhibitors have side effects like skin rash.²¹ Nonetheless, more definitive studies such as functional assays for rs6836436, preclinical experiments to further explore the role of EREG in acute and chronic pain, and, clinical trials for EREG inhibitors as analgesics are required to substantiate the role of EREG in the pathogenesis of pain.

Like other chronic diseases, early intervention is associated with better outcomes with chronic pain. Hence, there is a need to identify potential chronic pain patients during the acute stage for timely and optimal disease management. On one hand, a risk biomarker indicates the potential for developing a disease in an individual who does not currently have an identifiable clinical disease. Being associated with increased chronic pain severity and risk for developing chronic TMD, the presence of a major allele at rs6836436 or rs2367707 in acute pain patients might serve as a risk biomarker of chronic pain development. Nevertheless, it is important to recognize that pain is a highly polygenic trait and the contribution of each allele to the appreciable minor allelic frequency is expected to be modest. That is why we do not suggest rs6836436 or rs2367707 by themselves will act as sole predictors of pain states but could be useful inclusions into a screening panel of genetic markers for pain profiling. Conversely, a response biomarker could identify individuals who are more likely to experience a favorable or unfavorable effect from drug-treatment. The current findings suggest that the presence of a major allele at rs6836436 or rs2367707 may serve as a favorable response biomarker for EREG-EGFR-based pharmacotherapy of chronic pain. Thus, studying EREG gene polymorphism could accelerate the development of personalized pain medicine.

Although we used a TMD-centric cohort (OPPERA) as a discovery cohort, our results do not suggest that the association between EREG and pain phenotypes are specific to TMD or orofacial pain. We identified an association between EREG SNPs and a number of chronic pain conditions, independent of body site. This suggests that EREG contributes to TMD through mechanisms overlapping with other chronic pain conditions.²⁸ Symptoms of chronic TMD such as generalized pain sensitivity, sleep, concentration difficulties, depression, bowel complaints, and headaches often overlap with those of more generalized chronic pain conditions like fibromyalgia and chronic fatigue syndrome.¹ However, our conclusions may be limited based on the statistical power of the discovery cohort, OPPERA. With a TMD incidence rate of 8% and MAF of EREG’s H2 and H3 haplotype at 17.3% and 5% in the OPPERA follow-up cohort, respectively, the data lacked sufficient power to analyze the association between EREG and the onset of TMD in OPPERA or other associated pain phenotypes. Moreover, due to limited acute CPI at follow-up data, we could not confirm if captured patients would resolve TMD in the OPPERA cohort or whether they would remain chronic.

It is important to recognize the absence of a good replication cohort for our discovery findings. We used CPI as a pain intensity marker in OPPERA, but this phenotype is rarely collected in large community cohorts like UKB. Thus, we used the number of painful sites as markers of acute and chronic pain severity in UKB. Although pain severity in terms of intensity and anatomical extent are correlated,^3,50 these are clearly different phenotypes with potentially overlapping pathophysiology. Furthermore, the original discovery cohort was TMD-centric, while the UKB subjects reported pain across all body sites. Here, we viewed orofacial pain as an idiopathic pain condition and assumed that EREG contributes to it through molecular mechanisms shared by other chronic pain conditions, as suggested in our previous work.³¹ Overall, our second analysis in the UKB cohort was a validation of the primary findings from OPPERA rather than a true replication. It, however, unambiguously supported the dichotomous nature of EREG’s contribution to pain.

We previously showed that both EGFR and EREG displayed a genetic association with chronic TMD where EREG showed the strongest association.³¹ The current study confirms the association of haplotype H2 with chronic TMD in a different subset of OPPERA subjects (Table 5) and also demonstrated that the same haplotype H2 was associated with the presence of at least one chronic pain site in UKB. Although haplotype H3 was not identified in our earlier studies – due to its low frequency – it has a stronger effect than H2, as is evident from the strength of associations with pain phenotypes (Table 5 and 6) and eQTL analysis (Table 4). However, both haplotypes, H2 and H3, are loss-of-function variants. Even though we do not know the exact molecular mechanisms through which haplotypes H2 and H3 control EREG mRNA levels, we have shown earlier that rs2367707 (marker of H2) reduces stability of the mRNA and 5’UTR location of rs6836436 (marker of H3) suggests control of transcription. The absence of association of haplotype H2 with acute pain phenotypes (Table 6) is likely a reflection of its effect size rather than evidence for a unique contribution of H3, but not H2 to acute pain. An inverse relationship exists between the effect sizes and the allele frequencies for all phenotypic traits.³⁵ It is also possible that state-dependent stimuli regulating EREG transcription and mRNA stability contribute equally to chronic pain, but only regulation of transcription contributes to acute pain.

Our current results are also in line with our previous report on the effect of EREG in animal pain models. We previously showed that administration of EREG but not other EGFR ligands to mice in the late phase of the formalin test increased pain sensitivity.³¹ Mouse experiments in the present study assessed the impact of blocking EREG in different pain models at different time-points and were designed to support and compliment the human genetic analysis. Mice treated with the EREG mAb during peak CFA allodynia (i.e. 3-days post-CFA) recovered quicker than control mice; SNI-induced allodynia was also reversed for up to one-week post-EREG mAb administration. These results provide generalizability across chronic pain assays and suggest that EREG neutralization may offer a novel analgesic strategy for established chronic pain. Furthermore, the EREG mAb delayed recovery from CFA when administered during the development of CFA-induced allodynia (i.e. 1-day post-CFA). EREG neutralization also enhanced nocifensive pain behavior and acute mechanosensitivity in mice injected with capsaicin. Broadly, these data support the findings from the human genetic analysis, where the H3 haplotype was found to be protective for chronic pain, but a risk marker for acute pain.

While the signaling mechanisms of EREG on acute and chronic pain have yet to be discovered, a recently published independent study has reported similar dichotomous effects of EREG, where application of EREG onto the spinal dorsal nerve roots of rats reduced evoked c-fiber responses but increased spontaneous activity in spinal dorsal horn neurons.²³ Furthermore, immune system dysfunction including elevated level of pro-inflammatory cytokines,^39,44,51 allergic and autoimmune disease comorbidities are common among chronic pain conditions.^16,30,37,47 Since EREG is temporally¹⁵ and causally^17,33 associated with activation of the immune system and inflammation,³⁸ EREG may contribute to pain through a systemic process. For instance, EREG is involved with the production of proinflammatory cytokines in macrophages⁴⁰ and EREG is increased during cutaneous wound inflammation and healing.²⁶ Thus, EREG production may be necessary for the resolution of inflammation and the natural recovery from pain; however, EREG may trigger expression of signaling cascades in primary afferent nerves or in the dorsal horn that promote long-term changes in neuronal excitability.^15,31,33 Nevertheless, a full understanding of the role of EREG in modulating pain severity through the immune and /or nervous system will require simultaneous study of EREG, immune cells, inflammation and pain responses, both preclinically and clinically.

In conclusion, the present study confirms the previously reported role of EREG in the pathogenesis of human chronic pain and preclinical pain models.³¹ In addition, this study discovered an analgesic role for EREG during the early stages of pain, while an opposite-pronociceptive role in establishing chronic pain. Explicitly, this study is an example of a human → mouse translational research that further affirms EREG’s potential as a biomarker of chronic pain, demystifies EREG-mediated pathogenesis of pain and suggests a novel, non-opioid therapy for chronic pain.