ARTICLE IN PRESS
`
`IASR
`
`PAINÒ xxx (2013) xxx–xxx
`
`PAIN®
`
`w w w . e l s e v i e r . c o m / l o c a t e / p a i n
`
`Associations between KCNJ6 (GIRK2) gene polymorphisms
`and pain-related phenotypes
`Stephen Bruehl a,⇑, Jerod S. Denton a, Daniel Lonergan a, Mary Ellen Koran b, Melissa Chont a,
`
`Christopher Sobey a, Shanik Fernando a, William S. Bush b, Puneet Mishra a, Tricia A. Thornton-Wells b
`a Department of Anesthesiology, Vanderbilt University School of Medicine, Nashville, TN, USA
`b Center for Human Genetics Research, Department of Molecular Physiology & Biophysics, Vanderbilt University School of Medicine, Nashville, TN, USA
`
`Sponsorships or competing interests that may be relevant to content are disclosed at the end of this article.
`a r t i c l e
`i n f o
`a b s t r a c t
`
`Article history:
`Received 19 July 2013
`Received in revised form 21 August 2013
`Accepted 22 August 2013
`Available online xxxx
`
`Keywords:
`Chronic pain
`Genetics
`GIRK
`KCNJ3
`KCNJ6
`Pain
`Polymorphism
`Postsurgical Pain
`Potassium channel
`
`1. Introduction
`
`G-protein coupled inwardly rectifying potassium (GIRK) channels are effectors determining degree of
`analgesia experienced upon opioid receptor activation by endogenous and exogenous opioids. The impact
`of GIRK-related genetic variation on human pain responses has received little research attention. We used
`a tag single nucleotide polymorphism (SNP) approach to comprehensively examine pain-related effects of
`KCNJ3 (GIRK1) and KCNJ6 (GIRK2) gene variation. Forty-one KCNJ3 and 69 KCNJ6 tag SNPs were selected,
`capturing the known variability in each gene. The primary sample included 311 white patients
`undergoing total knee arthroplasty in whom postsurgical oral opioid analgesic medication order data
`were available. Primary sample findings were then replicated in an independent white sample of 63
`healthy pain-free individuals and 75 individuals with chronic low back pain (CLBP) who provided data
`regarding laboratory acute pain responsiveness (ischemic task) and chronic pain intensity and unpleas-
`antness (CLBP only). Univariate quantitative trait analyses in the primary sample revealed that 8 KCNJ6
`SNPs were significantly associated with the medication order phenotype (P < .05); overall effects of the
`KCNJ6 gene (gene set-based analysis) just failed to reach significance (P = .054). No significant KCNJ3
`effects were observed. A continuous GIRK Related Risk Score (GRRS) was derived in the primary sample
`to summarize each individual’s number of KCNJ6 ‘‘pain risk’’ alleles. This GRRS was applied to the repli-
`cation sample, which revealed significant associations (P < .05) between higher GRRS values and lower
`acute pain tolerance and higher CLBP intensity and unpleasantness. Results suggest further exploration
`of the impact of KCNJ6 genetic variation on pain outcomes is warranted.
`Ó 2013 International Association for the Study of Pain. Published by Elsevier B.V. All rights reserved.
`
`Genetic influences on human pain perception and risk for
`chronic pain are likely to be polygenic [8]. Numerous single nucle-
`otide polymorphisms (SNPs) have been identified in human studies
`as potential contributors. SNPs in genes encoding for the mu opioid
`receptor (OPRM1), the beta-2 adrenergic receptor (ADRB2), and cat-
`echol-O-methyltransferase (COMT) have all been demonstrated to
`influence acute pain sensitivity [7,9,10,13,16,38,49], chronic pain
`intensity [11,19,28,34], and risk for development of chronic pain
`conditions [6,9,12,15,19,29,39,43]. Prior work also suggests that
`pain-related SNPs (eg, A118G SNP [rs1799971] of the OPRM1 gene)
`may influence responses to opioid analgesics, although the degree
`of this influence remains debatable [45].
`
`⇑ Corresponding author. Address: Vanderbilt University Medical Center, 701
`
`Medical Arts Building, 1211 Twenty-first Ave S, Nashville, TN 37212, USA. Tel.: +1
`615 936 1821; fax: +1 615 936 8983.
`E-mail address: Stephen.Bruehl@vanderbilt.edu (S. Bruehl).
`
`One commonality between OPRM1 and COMT SNPs targeted in
`prior work is that each can potentially influence the magnitude
`of opioid inhibition upon activation of opioid receptors by endog-
`enous or exogenous opioid agonists [1,20,49]. The degree of opioid
`inhibition upon receptor activation is also influenced by numerous
`effectors, including G-protein coupled inwardly rectifying potas-
`sium (GIRK) channels of the Kir3.X family [25–27]. GIRK channels
`are activated by the b and c subunits of heterotrimeric Gi/o proteins
`after stimulation of opioid receptors by endogenous or exogenous
`opioids. The ensuing efflux of potassium ions hyperpolarizes the
`membrane potential, dampens neuronal excitability, and limits
`nociceptive transmission [14]. Several
`studies
`in animals
`document that both the KCNJ3 (GIRK1) and KCNJ6 (GIRK2) genes
`can influence pain and opioid analgesic responses [17,25,27,42].
`Indeed, the possibility of direct pharmacological manipulation of
`GIRK channel activity has been suggested as one avenue for devel-
`oping novel analgesic medications [2,21,32,44].
`Surprisingly, human work examining whether GIRK-related
`genetic variation influences pain responses has been sparse. Only
`
`0304-3959/$36.00 Ó 2013 International Association for the Study of Pain. Published by Elsevier B.V. All rights reserved.
`http://dx.doi.org/10.1016/j.pain.2013.08.026
`
`Please cite this article in press as: Bruehl S et al. Associations between KCNJ6 (GIRK2) gene polymorphisms and pain-related phenotypes. PAINÒ (2013),
`http://dx.doi.org/10.1016/j.pain.2013.08.026
`
`

`

`2
`
`S. Bruehl et al. / PAINÒ xxx (2013) xxx–xxx
`
`ARTICLE IN PRESS
`
`2 studies have explored this topic, both examining the pain-related
`impact of a small number of SNPs in the KCNJ6 gene. In patients
`undergoing major abdominal surgery, homozygous carriers of the
`A allele of the A1032G SNP (rs2070995) required rescue pain med-
`ication more frequently than those with the G allele, although no
`associations with postsurgical acute pain ratings were observed
`[33]. Other work found that compared to individuals with the G al-
`lele, homozygous carriers of the A allele required more methadone
`yet had fewer withdrawal symptoms in methadone substitution
`therapy patients, and required marginally higher opioid doses for
`pain control in chronic pain patients [24]. No human studies to
`date have examined the potential influence of KCNJ3 gene variants
`on pain-related outcomes, although such influence is suggested by
`animal work. For example, genetic deletion or pharmacological
`inhibition of KCNJ3-containing channels increases thermal noci-
`ception and blunts the analgesic response to opioids [26,27].
`The current study used a tag SNP approach to explore possible
`associations between a comprehensive array of SNPs in the KCNJ3
`and KCNJ6 genes and a postsurgical pain phenotype (oral opioid
`analgesic medication orders) in a large informatics-based sample.
`Findings were then replicated in an independent sample combin-
`ing data from 3 previously published studies using similar entry
`criteria [3–5] with regard to measures reflecting acute laboratory
`pain responsiveness and chronic low back pain intensity
`phenotypes.
`
`2. Methods
`
`2.1. Design
`
`This study used a correlational design to examine the impact of
`a comprehensive array of KCNJ3 and KCNJ6 SNPs on oral analgesic
`medication orders in a large clinical postsurgical primary sample,
`with replication of the resulting pain-relevant SNPs on acute labo-
`ratory pain and chronic back pain phenotypes in an independent
`sample.
`
`2.2. Subjects
`
`2.2.1. Primary sample
`The primary sample used to initially identify pain-relevant
`KCNJ3 and KCNJ6 SNPs was a large clinical postsurgical sample with
`electronic medical record data available in whom an informatics
`approach could be applied. To focus on patients with a comparable
`degree of tissue injury, the primary sample was drawn from a pool
`of 881 patients seen at Vanderbilt University Medical Center since
`2002 who displayed a CPT code of 27447 (total knee arthroplasty;
`TKA), who had undergone a unilateral TKA, and who had DNA sam-
`
`Table 1
`Characteristics of the study samples.a
`
`ples available in BioVU, the Vanderbilt biobank of de-identified
`DNA samples obtained for research purposes from discarded blood
`[36,37]. For this study, the selected BioVU DNA samples were
`linked in a de-identified manner to pain-relevant phenotypes via
`matching to the electronic inpatient medication order database
`at Vanderbilt (Wizorder). Routine DNA sampling and electronic
`medication records were implemented over differing time periods
`resulting in only a subset of patients in the potential subject pool
`with information available from both sources. The key phenotype
`targeted in the primary informatics sample was total number of
`oral opioid analgesic medication orders entered during each given
`patient’s inpatient hospital stay after TKA. For this portion of the
`study, patients included in the primary sample were limited to
`white patients with BioVU DNA samples who had the necessary
`medication order information available in Wizorder to permit
`characterization of this phenotype (n = 311). The decision to re-
`strict the final sample to white patients (the largest single racial
`group) was made to reduce potential confounds related to popula-
`tion substructure. To validate the oral analgesic medication order
`phenotype, postsurgical pain intensity data available in a subset
`of 82 patients from this larger pool were manually extracted from
`the Synthetic Derivative database, the Vanderbilt de-identified
`electronic medical records database.
`
`2.2.2. Replication sample
`To maximize statistical power in the replication sample, the
`current study combined data from 3 similar studies previously
`conducted in our laboratory, in which DNA samples were obtained
`in chronic low back pain (CLBP) subjects and healthy pain-free sub-
`jects [3–5]. Both groups contributed data regarding laboratory
`acute pain response phenotype (ischemic pain threshold and toler-
`ance), with the CLBP group also providing data regarding chronic
`pain phenotype (chronic back pain intensity and unpleasantness).
`For the acute pain phenotype, only those subjects experiencing
`the ischemic task in the absence of study drugs or other experi-
`mental manipulations that might alter pain responses were in-
`cluded in replication analyses. The current sample was restricted
`to white subjects for comparability with the primary sample and
`to minimize the potential influence of population substructure.
`All subjects met basic study medical eligibility criteria that were
`similar across the 3 studies. These criteria were: age between 18
`and 55 years, current normotensive status (resting blood pressure
`<140/90 mmHg), not pregnant, no history of cardiovascular dis-
`ease, hypertension, liver or kidney disorders, or opiate depen-
`dence; no current daily use of opioid analgesics, and no current
`use of antihypertensive medications. Absence of recent opiate
`use was confirmed via urine opioid screen in 66 of the subjects
`(all subjects participating in studies by Bruehl and colleagues
`
`Characteristic
`
`Primary sample
`
`Replication sample
`
`TKA medication order phenotype
`
`CLBP
`
`No. of subjects
`Gender (% female)
`Age (y)
`Total oral analgesic medication orders
`Ischemic acute pain threshold (s)
`Ischemic acute pain tolerance (s)
`Chronic pain duration (median, mo)
`VAS chronic pain intensity (0–100)
`VAS chronic pain unpleasantness (0–100)
`GRRS
`
`311
`63.3
`62.8 ± 10.48
`4.6 ± 2.96
`–
`–
`–
`–
`–
`8.0 ± 2.64
`
`49
`53.1
`36.4 ± 10.37*
`–
`48.5 ± 68.47
`260.0 ± 76.40
`70.0
`43.5 ± 15.43
`48.1 ± 17.28
`7.5 ± 2.01
`
`Healthy
`
`63
`57.1
`31.9 ± 9.25
`–
`39.9 ± 60.63
`250.9 ± 78.04
`–
`–
`–
`7.7 ± 1.86
`
`TKA, total knee arthroplasty; CLBP, chronic low back pain; VAS, visual analog scale; GRRS, GIRK Related Risk Score.
`a Summary statistics are presented as percentages or mean ± SD.
`* P < .05 for age difference between CLBP and healthy replication subsamples.
`
`Please cite this article in press as: Bruehl S et al. Associations between KCNJ6 (GIRK2) gene polymorphisms and pain-related phenotypes. PAINÒ (2013),
`http://dx.doi.org/10.1016/j.pain.2013.08.026
`
`

`

`ARTICLE IN PRESS
`
`S. Bruehl et al. / PAINÒ xxx (2013) xxx–xxx
`
`3
`
`[3,4]). Additional inclusion criteria for the CLBP group were chronic
`daily low back pain of at least 3 months duration with an average
`past month severity of at least 3 of 10. The final replication sample
`size was n = 112, including 46 subjects from Bruehl et al. [5], 11
`subjects from Bruehl et al. [4], and 55 subjects from Bruehl et al.
`[3]. Of the final replication sample, 63 (56.3%) were healthy pain-
`free controls (Pain-Free) and 49 (43.7%) were individuals with
`CLBP. Characteristics of both the primary and replication samples
`are summarized in Table 1.
`
`2.3. Procedures
`
`The Vanderbilt University Institutional Review Board (IRB) ap-
`proved all procedures in this study. Patients providing data in
`the primary postsurgical sample were all given the opportunity
`to opt out of DNA collection in accordance with IRB guidelines.
`All laboratory study subjects (replication sample) were volunteers
`who provided written informed consent prior
`to study
`participation.
`
`2.3.1. Primary sample procedures
`Data on inpatient oral opioid analgesic medication orders en-
`tered after TKA into the Wizorder electronic database were used
`to define the oral analgesic medication order phenotype. For each
`patient, an automated total count of any oral opioid analgesic med-
`ication orders entered was derived using SPSS syntax language
`(96.4% of orders were for oral immediate release oxycodone). Data
`on post-TKA intravenous analgesic orders were also available but
`were deemed inappropriate for analysis as a result of inadequate
`variability (more than 50% of patients had only a single intrave-
`nous analgesic order entered).
`To validate the oral analgesic order phenotype, standardized
`postsurgical pain ratings (0–10 scale, anchored with ‘‘No Pain’’
`and ‘‘Worst Possible Pain’’) obtained during inpatient physical
`therapy in the 3 days after the TKA procedure were extracted in
`a subset of 82 patients with available data. Pain ratings at rest
`and during activity were averaged over the 3 days for use as the
`overall postsurgical pain intensity measure.
`
`2.3.2. Replication sample procedures
`Detailed procedures for each laboratory study are provided
`elsewhere [3–5]. In brief, after providing informed consent, labora-
`tory study subjects completed a packet of demographic and psy-
`chometric questionnaires. For CLBP subjects, this packet included
`a visual analog scale measure of past month overall chronic back
`pain intensity (VAS Intensity; anchored with ‘‘No Pain’’ and ‘‘Worst
`Possible Pain’’), as well as a parallel scale assessing the affective
`component of chronic pain (VAS Unpleasantness; anchored with
`‘‘Not Unpleasant at All’’ and ‘‘The Most Unpleasant Possible’’).
`These measure were used to define the chronic pain phenotype
`for replication analyses. Both CLBP and Healthy subjects also
`participated in a standardized ischemic forearm acute pain task,
`a laboratory measure of acute pain sensitivity. Ischemic task proce-
`dures in all 3 laboratory studies were based on those described by
`Maurset et al. [30]. In brief, subjects were first asked to raise their
`dominant forearm over their head for 30 s followed by 2 min of
`dominant forearm muscle exercise using a hand dynamometer at
`50% of his or her maximal grip strength (as determined prior to
`beginning the laboratory procedures). Immediately after this, a
`BP cuff was inflated on the participant’s dominant bicep to
`200 mmHg. The cuff remained inflated until participants indicated
`that their pain tolerance had been reached, up to a maximum of
`5 min (for ethical requirements). Pain threshold was defined as
`the number of seconds elapsed between task onset and the sub-
`ject’s report that the task had become ‘‘painful.’’ Pain tolerance
`was defined as the number of seconds elapsed between task onset
`
`and the subject’s expressed desire to terminate the task. These
`measures comprised the acute laboratory pain responsiveness
`phenotype.
`
`2.4. Genetic assays
`
`Genetic samples were obtained via blood drawn from an
`indwelling venous cannula [3,5] or via buccal sampling [4]. DNA
`was extracted using the Gentra Systems AutoPure automated
`DNA extraction system in the Vanderbilt University DNA Resources
`Core.
`We used a tag SNP approach to avoid redundancy in genotyping
`of variants that were expected to be in high linkage disequilibrium
`with each other. We selected tag SNPs from candidate genes KCNJ3
`and KCNJ6 based on the HapMap CEU reference population with
`the goal of capturing at least 80% of the variation in each gene
`while reducing the need for genotyping every variant. For KCNJ3,
`41 tag SNPs were selected to capture 100% of the allelic variation
`in 181 SNPs across the gene with a mean r2 value of 0.949. For
`KCNJ6, 69 tag SNPs were selected to capture 100% of the allelic var-
`iation in 301 SNPs across the gene with a mean r2 value of 0.952.
`Supplementary Tables 1 and 2 provide the full list of tag SNPs for
`each gene and the alleles they capture.
`Genotyping was performed using Sequenom MassARRAY
`(Sequenom, Inc., San Diego, CA) and TaqMan OpenArray (Applied
`Biosystems, Foster City, CA) platforms. Four Sequenom pools were
`designed that incorporated all but 3 of the selected tag SNPs (one
`that needed to be in a pool by itself and 2 that failed assay design;
`all from KCNJ6). Direct genotyping of 3 remaining KCNJ6 tag SNPs
`was conducted using premade TaqMan SNP genotyping assays.
`Negative controls (no template) and positive controls (DNA
`samples with known genotypes from Coriell Institute for Medical
`Research, Camden, NJ) were included for assay validation. Inter-
`and intraplate experimental duplicates and HapMap controls were
`run on each assay plate to serve as positive controls for examining
`genotyping accuracy. Individuals who were blinded to clinical
`study data and hypotheses conducted semi-automated genotype
`calling with manual inspection of intensity clusters. Genotyping
`call rates and tests of Hardy Weinberg Equilibrium (HWE) were
`calculated for all genotyped SNPs.
`
`2.5. Statistical analysis
`
`All genetic association analyses in the primary sample were
`conducted using PLINK, Version 1.07 (http://pngu.mgh.har-
`vard.edu/purcell/plink/) [35]. Demographic and replication sample
`analyses were conducted using the IBM SPSS Statistics Version 20
`statistical package (IBM SPSS Statistics, Inc., Chicago, IL). All analy-
`ses used the maximum number of cases available for each
`phenotype.
`Univariate analyses were conducted assuming an additive mod-
`el for each SNP, in which having 2 copies of the coded allele was
`expected to modify risk by twice as much as having a single copy.
`For the oral analgesic medication order phenotype, a quantitative
`trait (QT) analysis was conducted using linear regression. A 2-
`tailed probability value of P < .05 was used as the criterion for sta-
`tistical significance in univariate analyses in the primary sample.
`Probability values were not corrected for multiple comparisons
`in univariate analyses as a result of the exploratory nature of this
`study. However, to provide a means of addressing potentially ele-
`vated familywise error rate due to examination of multiple SNPs
`within each gene, we also conducted gene set-based analyses for
`each gene using PLINK. For these analyses, all tagged SNPs within
`each gene were considered in the gene set, and the average of
`the single-marker (QT) test statistics was computed as the gene-
`set test statistic. Permutation testing was then used to determine
`
`Please cite this article in press as: Bruehl S et al. Associations between KCNJ6 (GIRK2) gene polymorphisms and pain-related phenotypes. PAINÒ (2013),
`http://dx.doi.org/10.1016/j.pain.2013.08.026
`
`

`

`4
`
`S. Bruehl et al. / PAINÒ xxx (2013) xxx–xxx
`
`ARTICLE IN PRESS
`
`the empirical P value for the experimental gene-set statistic [31]. In
`the current study, results of these set-based analyses reflected the
`overall influence of the given gene on the oral analgesic medication
`order phenotype.
`Replication sample analyses examined associations between
`the GIRK-Related Risk Score derived in the same manner as in
`the primary post-TKA informatics sample (GRRS; detailed below)
`and the acute and chronic pain phenotypes. Associations with
`the chronic pain intensity and unpleasantness measures were
`examined using Pearson correlational analyses. Because the distri-
`bution of ischemic pain task tolerance times was truncated as a
`result of 61.9% of subjects reaching the maximum permitted task
`duration (300 s), analyses of the acute pain phenotype used 2
`complementary approaches. Pearson correlations were used to
`examine associations between GRRS values and the continuous
`pain threshold and pain tolerance values, and t-tests were used
`to compare GRRS values between subjects who tolerated the full
`5-min ischemic task and those who did not. Because of the direc-
`tional nature of the confirmatory hypotheses in the replication
`sample, a 1-tailed P < .05 value was used as the criterion for signif-
`icance in replication analyses to maximize statistical power.
`
`3. Results
`
`3.1. Preliminary analyses
`
`Inspection of genotyping results from positive controls and
`experimental duplicates confirmed assay validity and concordance
`of genotype calls. Genotyping efficiency exceeded 91% for all SNPs,
`
`Table 2
`SNPs in the KCNJ6 gene significantly associated with the oral opioid analgesic
`medication order phenotype in quantitative trait analyses.
`
`SNP
`
`RA
`
`RA frequency
`
`Association with oral medication orders
`R2
`
`Beta
`
`P
`
`0.754
`0.466
`C
`rs1543754
`0.728
`0.701
`T
`rs858035
`0.573
`0.421
`C
`rs9981629
`0.550
`0.481
`A
`rs928723
`0.648
`0.820
`A
`rs2835925
`0.556
`0.219
`T
`rs2211843
`0.458
`0.516
`G
`rs1787337
`0.510
`0.232
`A
`rs2835930
`KCNJ6 set-based analysis (all tag SNPs)
`
`0.034
`0.027
`0.020
`0.018
`0.017
`0.013
`0.013
`0.012
`
`SNP, single nucleotide polymorphism; RA, risk allele.
`
`<.001
`.002
`.009
`.013
`.016
`.032
`.037
`.040
`.054
`
`with a median efficiency of 99%. Five SNPs were flagged as being
`out of Hardy-Weinberg equilibrium (P < .01) in the complete BioVU
`pool of 881 patients but were not removed from the analysis.
`
`3.2. KCNJ3 and KCNJ6 SNPs and the analgesic medication order
`phenotype
`
`Mean and standard deviation of the oral analgesic medication
`order count in the TKA sample are reported in Table 1. Validity
`of this key study phenotype was supported by the fact that it
`was correlated significantly with pain ratings obtained during
`postsurgical rehabilitation that were available in a subset of 82 pa-
`tients (r = 0.26, P = .01), in a direction indicating that more oral
`analgesic medication orders were entered for patients reporting
`greater post-TKA pain intensity. Table 2 summarizes the significant
`univariate associations between GIRK-related SNPs and the oral
`analgesic medication order phenotype. Eight KCNJ6 SNPs exhibited
`significant effects, with no significant effects for KCNJ3. Fig. 1 por-
`trays the chromosomal position of the 8 significant KCNJ6 SNPs. In
`the set-based analysis, which addressed possible familywise error
`rate inflation due to testing multiple SNPs in univariate analyses,
`the overall influence of the KCNJ6 gene on the oral analgesic med-
`ication order phenotype just failed to reach the criterion for statis-
`tical significance (empirical P = .054). The gene-set based analysis
`of the overall influence of the KCNJ3 gene was not significant
`(empirical P = 1.0).
`
`3.3. Derivation of the GIRK-related risk score
`
`To provide a simple means of summarizing the univariate re-
`sults, a GIRK-Related Risk Score (GRRS) was derived on the basis
`of the oral analgesic medication order phenotype in the primary
`sample. This GRRS included the 8 KCNJ6 SNPs showing significant
`univariate associations with the oral medication order phenotype
`(rs1543754,
`rs1787337,
`rs2211843,
`rs2835925,
`rs2835930,
`rs858035, rs928723, rs9981629). SNPs were coded for number of
`risk alleles present (0,1,2), such that more copies of the risk allele
`were associated with a greater number of oral analgesic medica-
`tion orders. Mean number of oral medication orders by risk allele
`status for these 8 KCNJ6 SNPs are presented in Table 3. Values were
`then summed across all 8 SNPs for a given individual, yielding a
`continuous GRRS ranging from 0 to 15 in the primary sample
`(Table 1). Within the post-TKA sample in which it was derived, this
`GRRS was correlated positively with number of oral analgesic or-
`ders entered into the medical record [r = 0.25, P < .001].
`
`rs9981629 +
`
`• rs928723
`
`3'
`
`Exon 4
`
`KCNJ6 (GIRK2)
`
`rs2211843
`
`•
`
`rs1787337
`
`•
`
`rs2835930
`
`•
`
`rs2835925 +
`rs1543754 +
`
`I
`Exon 3
`
`5'
`
`Exon 2
`
`Exon 1
`
`38950000
`
`39000000
`
`39100000
`39050000
`Chromosome 21 basepair position
`
`39150000
`
`39200000
`
`Fig. 1. Location of KCNJ6 SNPs exhibiting significant associations with the oral analgesic medication order phenotype. Gene diagram and base pair positions are from NCBI
`build 37.p10 primary assembly.
`
`Please cite this article in press as: Bruehl S et al. Associations between KCNJ6 (GIRK2) gene polymorphisms and pain-related phenotypes. PAINÒ (2013),
`http://dx.doi.org/10.1016/j.pain.2013.08.026
`
`

`

`ARTICLE IN PRESS
`
`S. Bruehl et al. / PAINÒ xxx (2013) xxx–xxx
`
`5
`
`Table 3
`Number of oral medication orders in the primary TKA sample by risk allele status for
`KCNJ6 SNPs exhibiting significant effects.
`
`SNP
`
`No. of risk alleles
`
`rs1543754
`rs858035
`rs9981629
`rs928723
`rs2835925
`rs2211843
`rs1787337
`rs2835930
`
`0
`
`3.95 ± 2.61
`3.50 ± 1.95
`3.94 ± 2.53
`4.04 ± 2.37
`3.41 ± 2.09
`4.17 ± 2.65
`3.99 ± 2.61
`4.31 ± 2.93
`
`1
`
`4.18 ± 2.72
`4.31 ± 2.85
`4.74 ± 2.78
`4.56 ± 2.83
`4.24 ± 2.72
`4.81 ± 2.93
`4.46 ± 3.01
`4.60 ± 2.56
`
`2
`
`5.29 ± 3.10
`4.78 ± 2.97
`4.82 ± 3.51
`4.95 ± 3.34
`4.61 ± 2.94
`5.62 ± 4.19
`4.92 ± 2.77
`4.91 ± 3.48
`
`TKA, total knee arthroplasty; SNP, single nucleotide polymorphism.
`
`3.4. Replication of the GRRS in the laboratory study sample
`
`Application of the same GRRS scoring method to the combined
`replication samples resulted in GRRS values ranging from 2 to 12
`(Table 1). Associations between GRRS values and the 2 measures
`of acute laboratory pain responses were examined in the combined
`replication subsamples. In line with the direction of effects in the
`primary sample, subjects with longer ischemic pain tolerance
`times (ie, relatively less pain sensitive) were found to have signif-
`icantly lower GRRS values [r(109) = 0.21, P = .01]. Consistent with
`these correlational findings, subjects reaching the maximum
`allowable pain tolerance on the ischemic pain task were found to
`have significantly lower GRRS values (ie, fewer risk alleles) than
`those not reaching maximum tolerance [less than maximum toler-
`ance: 8.1 ± 1.80; maximum tolerance: 7.4 ± 1.96; t(109) = 1.80,
`P = .04]. The association between ischemic pain threshold and
`GRRS values was not significant (P = .45).
`Replication regarding the chronic pain phenotype was con-
`ducted within the CLBP replication sample only. Subjects with
`higher GRRS values were found to report significantly greater past
`month chronic low back pain intensity [r(46) = 0.29, P = .02]. Asso-
`ciation between GRRS values and the affective component of
`chronic pain (ie, past month chronic low back pain unpleasantness)
`was of similar magnitude [r(46) = 0.29, P = .02]. Overall, results for
`both acute laboratory pain tolerance and the chronic back pain
`phenotype in the replication sample are in a direction supporting
`the validity of the KCNJ6 effects noted in the primary post-TKA
`sample regarding the oral analgesic medication order phenotype.
`Comparison of GRSS scores between the pain-free and CLBP repli-
`cation samples did not reveal significant differences (P > .10;
`Table 1).
`
`4. Discussion
`
`Genetic influences on pain are polygenic [8], with SNPs in the
`OPRM1, COMT, and ADRB2 genes previously shown to influence
`acute and chronic pain intensity [7,9–11,13,16,19,28,34,38,49] or
`risk for development of chronic pain [6,9,12,15,19,29,39,43]. Both
`OPRM1- and COMT-related genetic influences on pain may involve
`opioid mechanisms [1,20,49]. GIRK channels are important effec-
`tors that can determine the degree of opioid inhibition occurring
`upon opioid receptor activation [14], and therefore, variations in
`GIRK-related genes provide another potential opioid-related path-
`way by which pain responses may be genetically influenced. Ani-
`mal studies confirm the relevance of both KCNJ3 and KCNJ6 genes
`to pain outcomes [17,25,27,42]. However, to date, only 2 human
`studies have explored this issue [24,33], with both limited to test-
`ing a relatively small number of KCNJ6 SNPs. The current study em-
`ployed a tag SNP approach to examine a comprehensive array of
`polymorphisms capturing known variability in the KCNJ3 and
`
`KCNJ6 genes as they relate to an informatics-based postsurgical
`pain phenotype (oral opioid analgesic medication orders after
`TKA), with subsequent replication of significant pain-related
`effects regarding acute and chronic pain phenotypes in an indepen-
`dent laboratory-based sample.
`Univariate quantitative trait analyses revealed that 8 KCNJ6
`SNPs were significantly associated with the oral analgesic medica-
`tion order phenotype. Gene set-based analysis indicated that the
`effect of variation in the KCNJ6 gene overall on this postsurgical
`pain phenotype just
`failed to reach statistical significance
`(P = .054). A pain-related influence of KCNJ6 was not unexpected,
`given that the only 2 prior human studies examining GIRK-related
`genetic variation on pain outcomes showed effects for KCNJ6. Both
`previous studies reported that the A1032G SNP (rs2070995) of the
`KCNJ6 gene showed significant effects on opioid analgesic re-
`sponses, although Nishizawa et al. [33] did not find statistically
`significant effects on acute (postsurgical) pain responses. In con-
`trast to the latter study, which found A1032G SNP effects on post-
`surgical rescue medication requirements, the current study did not
`find significant effects of the A1032G SNP (tagged by rs858003 in
`this study; r2 = 1.0 based on HapMap CEU population) on the post-
`surgical medication order phenotype examined. Nonetheless, a
`number of other KCNJ6 SNPs not examined in prior work were
`associated in the current study with the postsurgical oral medica-
`tion order phenotype. Whether
`the KCNJ6 SNPs
`showing
`pain-related effects in the current study influence opioid analgesic
`responses, as in Nishizawa et al. [33] and Lötsch et al. [24], could
`not be directly tested as a result of limitations of the informatics
`data available. This possibility remains to be examined in future
`work.
`Findings in the primary sample documenting pain-related ef-
`fects of several KCNJ6 SNPs are strengthened by results of cross-
`validation in an independent sample. A continuous GIRK-related
`risk score (GRRS) derived for each individual to summarize KCNJ6
`SNPs that exhibited significant pain-related effects in the primary
`sample was found to be associated in the same direction with both
`acute and chronic pain phenotypes in the laboratory-based replica-
`tion sample. Specifically, higher GRRS values were associated with
`lower pain tolerance to a standardized acute laboratory pain task
`and higher chronic low back pain intensity and unpleasantness.
`Taken together, these findings underscore the likely pain-relevance
`of variation in the KCNJ6 gene.
`Although prior work had examined pain-related KCNJ6 influ-
`ences in a limited way, no previous human study had examined
`variation in the KCNJ3 gene as it relates to pain phenotypes. Results
`of the current work did not reveal any significant KCNJ3 effects on
`the postsurgical analgesic medication order phenotype in the large
`primary sample. Nonetheless, positive findings in past animal
`studies [26,27] suggest that it may yet be worthwhile investigating
`possible impact of KCNJ3 SNPs as they relate to other pain-relevant
`phenotypes.
`GRRS values that captured significant pain-related KCNJ6 influ-
`ences in the primary sample, and were replicated vis-à-vis acute
`and chronic pain-related phenotypes in the laboratory sample,
`nonetheless did not display significant differences between the
`CLBP and pain-free groups in the replication sample. The effect size
`for observed GRRS differences across CLBP and pain-free groups
`was very small (eta squared = 0.003), suggesting that it is unlikely
`that inadequate power alone can explain the absence of significant
`GIRK-related chronic pain risk differences in this study. However,
`given the limited pain phenotype examined in the primary sample
`used to derive the GRRS and that this is the first study examining a
`comprehensive array of KCNJ3 and KCNJ6 polymorphisms, further
`investigation may be warranted. Previous cross-sectional studies
`document that variability in the