Targeted Genotyping Identifies Susceptibility Locus in Brain-derived Neurotrophic Factor Gene for Chronic Postsurgical Pain

POSTOPERATIVE pain is an inevitable consequence of surgery. As wound healing is completed with time, it is anticipated that pain will be reduced accompanied by a return of function. However, some patients continue to suffer persistent pain over the surgical wound, which could not be explained by disease recurrence.^1,2  This disease entity, currently known as chronic postsurgical pain, has been estimated to affect 9.2 to 80.0% of patients having a variety of surgery.^3–8  Considering that more than 230 million surgeries are performed each year worldwide,⁹  the data would imply that millions of patients will continue to suffer wound pain, months to years after their surgery.

Although peripheral nerve injury, severe acute postoperative pain, and psychosocial factors are thought to increase the risk for chronic postsurgical pain,^5,10  considerable efforts have been made to identify the genetic determinants of pain.¹¹  For instance, single-nucleotide polymorphisms (SNPs) of catecholamine-O-methyltransferase (COMT) altered pain sensitivity in experimental settings.¹²  Similarly, using quantitative trait locus mapping, two variant alleles of the melanocortin-1 receptor gene (MC1R) were associated with greater analgesia from κ-opioid in women.¹³  Other investigators have found that genetic polymorphism of interleukin-1β is involved in allodynia and possibly postsurgical neuropathic and inflammatory pain.¹⁴  Multiple polymorphisms within GTP cyclohydrolase (GCH1) have also been found to predict chronic back pain syndrome.¹⁵  In addition, our recent study found that patients with AA genotypes at polymorphisms, rs2070697 and rs2236742, in cathepsin G (CTSG) gene significantly reduced the risk for chronic postsurgical pain.¹⁶  We therefore hypothesized that genetic variations are associated with the development of chronic postsurgical pain. In the current study, we analyzed the association of 638 SNPs located in 54 pain-related genes and explored the risk factors associated with chronic postsurgical pain in the first 1,873 patients enrolled in our Persistent Pain after Surgery Study. We also investigated the functional role of an identified SNP (brain-derived neurotrophic factor, BDNF) in an experimental nociception model using knock-in mice.

The Persistent Pain after Surgery Study was an observational cohort study to evaluate the epidemiology and risk factors of chronic postsurgical pain. The study objectives and protocol are summarized in the Chinese Clinical Trial Registry (ChiCTR-ONC-10001099). The Clinical Research Ethics Committee approved the study, and written informed consent was obtained from all patients. Between January 2011 and July 2013, we recruited a total of 1,870 patients into the study from two tertiary referral hospitals. After 1,300 patients had completed 12-month follow-up, we noted that the rate of chronic postsurgical pain was 2.5-fold higher than we anticipated (21.4% vs. 8%). Recognizing that we have sufficient events, we decided to conduct an exploratory genetic association analysis in the first 1,152 patients (discovery cohort). These patients were also included in our previous report showing the association between CTSG gene polymorphism and risk of chronic postsurgical pain.¹⁶  With the limited sample size, we then selected patients with and without chronic postsurgical pain in a 1:1 ratio, from the remaining patients, for a matched case-control study (validation cohort) to verify the results in the discovery cohort.

We enrolled patients aged 18 yr or older who had in-patient and ambulatory surgery that included a skin incision. Patients were excluded if they were scheduled for a pure endoscopic or radiologic procedure or were previously enrolled in the study. Patients who were not expected to be available for or to cooperate with interviews were also excluded.

Before surgery, research staff interviewed all patients. Baseline patient characteristics, including age, sex, level of education, history of prior pain syndromes, employment status, and smoking habits were recorded. The types of surgery performed and the anesthetic techniques used were noted. After surgery, patients reported their experience of pain over the surgical wound using an 11-point pain analog scale (0 points = no pain to 10 points = worst imaginable pain). The average pain score and the total amount of opioid consumed, expressed as morphine equivalents, during the first 3 postoperative days were recorded.

All patients were contacted again at 12 months after surgery with a telephone interview. Patients were asked to rate the severity of pain, and its interference with daily activities was scored using the brief pain inventory.¹⁷  For each item, 0 points represents no pain or interference, and 10 points indicates worst imaginable pain or completely interferes with activities. The characteristic of the pain was scored using the neuropathic pain questionnaire.¹⁸  In addition, patients reported their current health status using the EQ-5D scale.¹⁹  Use of analgesics and other pain management therapies during the 12 months after the index surgery were also noted. The primary outcome was pain over the surgical site that has persisted for 12 months after the index surgery.^20,21 

Genomic DNA was extracted from 5 ml of peripheral blood collected at induction of anesthesia according to a standard phenol-chloroform (Applied Biosystems by Life Technologies, USA) procedure. A total of 768 SNPs within 65 genes (Supplemental Digital Content 1, https://links.lww.com/ALN/B564) were selected based on (1) SNP tagging of Han Chinese in Beijing through the HapMap database (ftp://ftp.ncbi.nlm.nih.gov/hapmap/; accessed October 31, 2017); (2) the Pain Genes database (http://www.jbldesign.com/jmogil/enter.html; accessed October 31, 2017); and (3) microarray data from our previous study.¹¹  SNP genotyping by GoldenGate genotyping assay following the manufacturer’s instructions (Beadstation 500, Illumina).

The probability of successful genotyping was predicted based on the validation status of the SNP and the minor allele frequencies from published studies using the assay design tool software (Illumina). The results were expressed as a design ability rank score, ranging from 0 to 1.1. For the 768 SNPs, the mean (± SD) design ability rank score was 0.84 ± 0.21 (range, 0.5 to 1.1). A score of 0.4 or higher was considered as having good probability for successful genotyping. We genotyped 1,140 samples with 12 samples left as blank control. To facilitate clustering of genotype, the assay algorithm was trained using one Centre de'Etude du Polymorphism Humain trio in four replicates (NA12878: daughter, NA12891: father, and NA12892: mother). We expressed the performance of the genotyping assay using the GenCall score (0 to 1), which is the ratio of separation between homozygote and heterozygote clusters for any SNP.

Patients in the validation cohort were selected from the subsequent 503 patients recruited to the study. We identified 103 patients with chronic postsurgical pain (cases). Matched controls were selected from the first patient who had surgery within the month of the case and who was in the same 5-yr age stratum, sex, and type of surgery. Genotyping of the validation cohort was performed using TaqMan assays (Life Technologies) on 384-well polymerase chain reaction plates with TaqMan genotyping master mix (Life Technologies). Based on the incidence of chronic postsurgical pain in the discovery cohort (more than 20%), we estimated that a validation cohort of 206 patients would provide 83% power to identify SNP with minor allele frequency of more than 20% for an odds ratio of more than 1.2.

This animal experiment was approved by the institutional animal care and use committee of Shandong University (Shandong, China). Adult male mice (20 to 30 g) of genotypes BDNF^Val/Val and BDNF^Met/Met were obtained from the Shandong Provincial Key Laboratory of Mental Disorders, Shandong University. Chronic postsurgical pain was studied using a plantar incision model as previously described.²²  In this model, a 10-mm longitudinal incision was made into the plantar surface of the hind paw. Muscle fibers under the wound were then elevated from surrounding tissues and stretched 10 times using curved forceps. The wound was closed with 6-0 nylon sutures and sterilized with iodine solution. The incision model produced consistent mechanical allodynia for more than 6 days. We studied 15 mice with each BDNF^Val/Val and BDNF^Met/Met genotype. Mechanical allodynia was measured with a series of 10 von Frey filaments (bending force of 0.008, 0.02, 0.07, 0.16, 0.4, 0.6, 1, 1.4, 2, and 4 g) at 6 h and then on days 1, 3, 5, and 7 after surgery. The filaments were applied to an area adjacent to the incision. A positive response was defined as withdrawal of the tested paw related to the application of stimulus. Paw withdrawal threshold was determined by Dixon’s up-and-down method.²³  The threshold was calculated as the average of at least three measurements recorded 5 min apart. Investigators who performed surgery and determined the paw withdrawal threshold were blinded to the BDNF genotype.

The categorical association for the disease trait (case vs. control) was analyzed using chi-square test or Fisher exact test, as appropriate. Student’s t test was used for continuous variables. PLINK-1.07 (Center for Human Genetic Research, Massachusetts General Hospital Broad Institute of Harvard and MIT, Cambridge, Massachusetts) software was employed for allelic test.²⁴  Bonferroni correction was used to adjust the P values to reduce type I errors due to multiple testing. The Hardy–Weinberg equilibrium statistics and cluster plots of each significantly associated SNP were individually examined. We determined the association between SNPs by haplotype analysis using the CI method (Haploview 4.2, Daly Lab at the Broad Institute, Cambridge, Massachusetts).^25,26  We calculated the normalized coefficient of linkage disequilibrium (D′) and Pearson correlation coefficient (r²), such that D′ and r² values of 1 indicate perfect linkage disequilibrium between SNPs.

We constructed a parsimonious model of independent risk factors for chronic postsurgical pain using the baseline characteristics and types of anesthetic techniques. Genetic factors (SNPs) were also included in our multivariable logistic regression model to test the predictors for chronic postoperative pain. Only factors that scored a P value of less than 0.05 in the univariate analysis were incorporated in the multivariable model. A forced simultaneous entry was adopted to avoid overfitting of the model.^27,28  Analysis was performed using Stata 13 (StataCorp LP, USA). We also determined the population attributable risk (PAR) for chronic postsurgical pain to indicate the proportion of outcomes that are attributable to the risk factor concerned.²⁹  PAR was calculated based on our multivariable regression analysis using the Interactive Risk Attributable Program (IRAP version 2.2, National Cancer Institute, Rockville, Maryland, 2002).

Factorial analysis of variance with repeated measures was used to detect any differences in paw withdrawal thresholds among groups. Significant differences at each of the time points were evaluated using a Bonferroni post hoc test. These analyses were performed using Stata 13.

Figure 1 shows the study chart, and table 1 reports patient characteristics in the discovery and validation cohorts. Overall, half of the patients were male with a mean age of 62 yr. About 11% of patients reported preexisting pain syndrome before surgery. Among these patients, 88.7% received regular nonsteroidal antiinflammatory drugs, 74.0% had gabapentinoid, 7.8% had tramadol, and 5.9% had other opioids. The majority of patients (83.4%) received general anesthesia, whereas 5.2% had regional or neuraxial block alone. The remaining 11.4% of patients had combined general and regional anesthesia.

At 1-yr follow-up, 246 patients (21.4%) in the discovery cohort (n = 1,152) reported pain over the wound site from the index surgery during the interview. Among these patients, 81 (7.0%) had severe pain with a pain score of 5 points or higher. The median (interquartile range) pain scores for patients reporting mild-to-moderate and severe pain were 2 (1 to 3) and 6 (5 to 8) points, respectively. In the validation cohort, 68 patients (of 103) had mild-to-moderate pain with a median (interquartile range) pain score of 2 (2 to 4), and the remaining 35 patients rated their pain severe with a pain score of 6 (5 to 8) points.

Chronic postsurgical pain affected all types of surgery (table 1). Tables S1, S2, and S3 in Supplemental Digital Content 2 (https://links.lww.com/ALN/B565) show the characteristics and impact of chronic postsurgical pain on daily activities and quality of life. Most patients who had chronic pain reported symptoms attributing to neuropathic pain. During the follow-up period, patients with pain reported difficulties with general activity, mood, walking, relations with others, sleep, and enjoyment of life. General health status was also adversely affected by pain. Age less than or equal to 65 yr of age, male sex, receiving less than 12 yr of education, reporting prior history of pain elsewhere, and reporting severe postoperative pain during the first 3 days after surgery were associated with higher risk for chronic postsurgical pain (table 2).

Among 768 SNPs, the minor allele frequencies of 130 SNPs were less than 5%. The remaining 638 SNPs in 54 genes had an average call rate of 99.2% (median frequency 97.8%). In 59 samples, the call rate was less than 97%. The allelic frequencies of all process controls, as integrated in the GenomeStudio software (Illumina, USA), were found to be within the reported range. In an association analysis of the discovery cohort, a total of 42 SNPs from 12 candidate genes were found to be associated with chronic postsurgical pain (table 3) with 23 SNPs (54.8%) increasing the risk of pain. After Bonferroni correction (P value < 0.05/638 or < 7.84E-5), two SNPs (rs6265 and rs1491850) in the BDNF gene, and one (rs6985606) in the κ-opioid receptor gene (OPRK1) remained significantly associated with chronic postsurgical pain (Supplemental Digital Content 1, https://links.lww.com/ALN/B564). Association was still significant under all modes of inheritance (Supplemental Digital Content 3, https://links.lww.com/ALN/B566).

We further validated the association between the three SNPs and chronic postsurgical pain in a matched cohort of 206 surgical patients by TaqMan genotyping assay. Table 4 shows the allelic frequencies of the SNPs. The genotypic distribution of the three SNPs were in Hardy–Weinberg equilibrium. Consistent with the discovery cohort, patients carrying allele A of rs6265 in the BDNF gene were associated with a lower risk for chronic postsurgical pain compared with those carrying allele G. In contrast, the effect of the SNP rs1491850 in the BDNF gene was opposite in the validation cohort compared with the discovery cohort. The association of SNP (rs6985606) in the OPRK1 gene on chronic postsurgical pain could not be confirmed in the validation cohort.

Based on the discovery cohort, a haploblock was identified in the BDNF gene (fig. 2). We found that both rs6265 and rs1491850 were in block with the highest strength of linkage disequilibrium (D′ = 0.899; r² = 0.74). Similar findings were obtained with a haplotype analysis of all the SNPs in the BDNF gene reported in the literature (Supplemental Digital Content 4, https://links.lww.com/ALN/B567). Both SNPs in the BDNF gene (rs6265 and rs1491850) and its haploblock remained associated with chronic postsurgical pain when adjusted for clinical risk factors. The genetic factors had the largest population attributable risk (table 2).

The data presented so far suggested that both rs6265 and rs1491850 in BNDF gene were associated with chronic postsurgical pain. Although the latter is in the promotor region, rs6265 is a nonsynonymous SNP located in exon 2 of BDNF gene. To determine the functional role of rs6265, mechanical allodynia was measured in mice expressing the corresponding alteration of amino acid in BDNF. Concordant with our human data, mechanical allodynia was significantly decreased in BDNF^Met/Met mice compared with the BDNF^Val/Val group (P = 0.003, factorial analysis of variance with repeated measures; fig. 3).

Our study showed that chronic postsurgical pain was common. In a relatively unselected group of adult patients, more than 20% of patients reported persistent pain in the wound for 1 yr after surgery. Among these patients, 32.9% rated their pain as severe. The incidence is similar to a recently reported cohort of patients having abdominal hysterectomy (25.1%) and thoracotomy (37.6%).⁷  Our data also confirmed that chronic postsurgical pain affects quality of life, including social function, physical activities, emotion, and mental health.

In an attempt to investigate the genetic factors for chronic postsurgical pain, we evaluated a total of 638 SNPs from 54 pain-related genes. After rigorous screening, we found that SNP rs6265 in BDNF gene was associated with chronic postsurgical pain in two independent cohorts. BDNF is a secretory protein in the nervous system that binds to tyrosine-related kinase B (TrkB) and regulates synaptic plasticity by activating several signaling cascades.³⁰  In an experimental nociception model, de novo synthesis of BDNF protein activates postsynaptic excitatory receptor (e.g., N-methyl-d-aspartate receptor) and suppresses inhibitory receptors (e.g., γ-aminobutyric acid receptors).³¹  Each of these processes contributes to central sensitization in different parts of nociceptive pathways. Interestingly, antagonism of BDNF or TrkB alleviated nociceptive response in animals.^32,33  Recently, it has been shown that BDNF levels correlate with pain intensity. For instance, an elevated BDNF level was found in patients with frequent and severe abdominal pain because of chronic pancreatitis and irritable bowel syndrome.³⁴ 

The SNP rs6265 is located in the 5′-prodomain of immature BDNF protein, leading to a methionine (Met) substitution for the normal valine (Val) at codon 66 and hence also known as Val66Met.^34,35  In our study, patients who carried the G allele (Val/Met and Val/Val) were associated with higher risk of chronic postsurgical pain. In this regard, rs6265 (Val66Met) is one of the most widely studied SNPs in the BDNF gene. A number of genetic association studies have evaluated the role of Val66Met variants in nervous system function or disorders, such as memory formation, anxiety, and substance abuse. Recently, it was reported that the Val66Met variant may also affect cortical pain processing during experimental pain stimuli.³⁶  In that study, the Met allele augmented cortical pain processing in chronic pain patients and reduced pain processing in the pain-free controls. Similarly, in cultured hippocampal neurons, BDNF protein carrying the Met variant exhibited lower depolarization-induced secretion.³⁷  In Met variant BDNF-transgenic mice, secretion of BDNF from neurons was abolished, accompanied by an increased stress-related anxiety.^38,39  These data suggested Val66Met is a functional variant that reduces BDNF-dependent activity. In the nociceptive pathway, activity-dependent release of endogenous BDNF has been reported in nociceptors⁴⁰  and can be enhanced during inflammation.⁴¹  These studies confirmed the hypothesis that the Met allele impairs the BDNF release from nociceptors and thus reduces pain. Our findings from the clinical cohort study and transgenic BDNF^Met knock-in mice are consistent with this hypothesis. In mice, Met carriers were less likely to develop mechanical allodynia after plantar incision surgery. Previous studies suggested that BDNF affected chronic pain development via modulating ion channel activity and gene expression.^31–37  BDNF induces long-term potentiation in N-methyl-d-aspartate receptor, and this may last for more than 8 h in the central nociceptive neurons, such as those in the spinal dorsal horn lamina II neurons and therefore contributes to the production of neuropathic pain. On the other hand, it has been reported that the insufficient BDNF release and therefore the lack of late long-term potentiation in postsynaptic neurons prevented neuropathic pain. BDNF also causes transcriptional changes in spinothalamic neurons and contributes to the long-lasting central sensitization in the nociceptive pathway.^40,41 

Previous studies have shown that between 30 and 76% of pain behaviors in animals are due to heritability factors.^32–41  In our study, we found that 9.77 and 7.36% of chronic postsurgical pain could be attributed to genetic variants at rs6265 and rs1491850 in BNDF, respectively. Aside from genetic factors, other risk factors have been proposed to increase the risk of chronic postsurgical pain. In our study, we found that younger patients (65 yr of age or less) were associated with increased risk of chronic pain after surgery, which is consistent with a previous study.⁴  Our study also found that male patients are at higher risk for chronic postsurgical pain than females. Importantly, we found the SNPs and its haplotype (PAR: 7.36 to 11.7%) had larger contributions than other clinical factors (PAR: 2.90 to 5.93%) for the development of chronic postsurgical pain.

Taken together, our data suggest that the variant of BDNF rs6265G>A was associated with lower risk of chronic postsurgical pain. It is propitious that routine genotyping of the SNP may help to identify high-risk surgical patients for more intense management of postsurgical pain.

Matthew T. V. Chan, M.B., B.S., Ph.D. (co-chair), Tony Gin, M.D. (co-chair), William K. K. Wu, Ph.D., Christopher H. K. Cheng, Ph.D.

Benny C. P. Cheng, M.B., B.S.

Matthew T. V. Chan, M.B., B.S., Ph.D., Tony Gin, M.D., Yuanyuan Tian, M.B., Ph.D., Zhaoyu Meng, Ph.D., Angel Mou, B.A., Ming Wai Tsang, B.A., Matthew W. M. Tsang, M.Sc., Raymond Kwan, M.Sc., Eliza Wong, B.Sc., Kevin Yau, B.Sc., Annie Lok, Eva Lee, M.Sc.

Benny C. P. Cheng, M.B., B.S., Carmen K. M. Lam., M.B., B.S., Sharon Tsang, M.B., Ch.B., Ph.D., Beaker Fung, M.Sc., Leo Lam, B.Sc.

Yuanyuan Tian, M.B., Ph.D., Xiaodong Liu, Ph.D., Yujian Shi, Ph.D., Shanglong Kou, Ph.D., Idy H. T. Ho, Ph.D., Sharon Tsang, M.B., Ch.B., Ph.D., Jun Yu, Ph.D., Christopher H. K. Cheng, Ph.D., Sunny H. Wong, M.B., Ch.B., D.Phil., Chee Sam Chan, William K. K. Wu, Ph.D., Matthew T. V. Chan, M.B., B.S., Ph.D.

Peter Lichtner, Ph.D.

Xiaodong Liu, Ph.D., Shanglong Kou, Ph.D., William K. K. Wu, Ph.D.

Hui Yu, Ph.D., Mingzhong Jia, M.Phil., Zheyu Chen, Ph.D.

This study was coordinated by the Department of Anesthesia and Intensive Care, Chinese University of Hong Kong, Prince of Wales Hospital, Hong Kong Special Administrative Region, China.

Supported by project grant No. 464212 from the General Research Fund, Research Grant Council (Hong Kong Special Administrative Region, China); project grant No. 11121051 from the Health and Medical Research Fund (Hong Kong Special Administrative Region, China); and project grant No. LTP001/14 and academic enhancement grant No. AEG13/001 from the Anesthesia and Pain Medicine Foundation, Australian and New Zealand College of Anaesthetists (Melbourne, Victoria, Australia).

The authors declare no competing interests.