ABCB1 and OPRM1 single-nucleotide polymorphisms collectively modulate chronic shoulder pain and dysfunction in South African breast cancer survivors

The term ‘breast cancer survivors’ (BCSs) is used to describe individuals with cancer ‘from the point of diagnosis through the balance of his/her life’ [1–3]. In BCS, 30–50% of survivors develop postoperative chronic shoulder pain and dysfunction, leading to reduced function and poor quality of life that can persist for 6–8 years after breast cancer (BC) treatment [4–8]. Chronic pain is multifactorial with biological, psychological, social and cultural contributors [9,10]. Risk factors for chronic postoperative pain in BCS include age at surgery, axillary lymph node dissection (ALND), adjuvant therapies, preoperative and severe acute postoperative pain and genetics [11,12]. Opioids are clinically used to treat both acute and chronic pain, and some of the common drugs administered for pain control in BCSs are morphine, codeine, oxycodone, tramadol, methadone and fentanyl, among others [13]. To date, several genes have been implicated to contribute to the variable response to opioids, specifically with regard to the different phases in drug pharmacology [14]. This study focused on the ABCB1 and OPRM1 genes that are well described to be essential in the distribution and signaling of opioids [14]

The ABCB1 gene, also known as MDR1, is involved in cellular homeostasis and the ATP-dependent translocation of substrates [15,16]. The gene encodes a P-glycoprotein (P-gp) present in a variety of human tissues and is the major determinant of opioid bioavailability into the brain [17,18]. P-gp is a ubiquitous membrane transporter, with an affinity for several substrates, including opioids (e.g., morphine, methadone and fentanyl), all clinically used to treat moderate to severe pain. These substrates also have a high binding affinity to the OPRM1 receptor, thus activating pain inhibition via the G-protein-coupled receptor (GPCR) cascade.

At the blood–brain barrier (BBB), P-gp is found in the endothelial cells surrounding the lumen of the brain vasculature. P-gp function is to export opioids from the blood into the endothelial cells, and also diffusing it back into circulation. This mechanism ensures the protection against neurotoxicity but also the distribution of opioids to the central nervous system (CNS) [19,20]. Additionally, knockout studies have shown that analgesic effects vary between mdr1(-/-) and mdr1(+/+ and +/-) mice [21,22].

The rs1045642 G>A, rs2032852 T>G/A and rs1128503 G>A polymorphisms remain the most frequently studied ABCB1 polymorphisms in pharmacogenetic studies [23–25]. Using immunohistochemistry and Western blot analysis, Hoffmeyer et al. [26] found the ABCB1 rs1045642 (A/A) genotype correlated with a twofold reduction in P-gp intestinal expression. To test the effects of the rs1045642 single-nucleotide polymorphism (SNP) on P-gp activity, the authors measured duodenum plasma concentrations before and after administration of digoxin and found a greater than fourfold difference between (A/A) carriers (higher levels) and (G/G) carriers (lower levels), [26,27]. The results indicating that the ABCB1 rs1045642 (G>A) polymorphism influences both P-gp expression and activity. Subsequent acute pain-related studies have shown the rs1045642 (A) allele to be associated with neuropathic, cancer and postoperative pediatric pain [28–31]. However, the consensus on these associations and functions of the P-gp remain inconclusive [26,32].

The OPRM1 gene, is a GPCR that facilitates the pharmacological effects of opioids through the inhibition of cell signaling [18]. The gene encodes the mu-opioid receptor protein (MOR-1), which is widely distributed in the brain and spinal cord and acts as the primary binding site for both endogenous and exogenous opioids [28,33]. OPRM1 rs1799971 (A118G), the most studied polymorphism, involves an asparagine to aspartate substitution (Asn40Asp), resulting in the loss of an N-linked glycosylation site in the extracellular receptor region [34–36]. The substitution thereby effects the protein stability, expression and signaling efficiency [28,37]. OPRM1 rs540825 involves a histidine to glutamine (His > Glu) substitution within the C-terminal, intracellular domain of the receptor [38].

To date, two non cancer-related studies have shown an association between the ABCB1 and OPRM1 genes and risk for chronic pain [39,40]. No studies have explored the relationship between these polymorphisms and shoulder pain/disability. Understanding the genetic contribution of these polymorphisms could assist in the identification of potential predictive markers for the development of chronic postoperative shoulder pain and dysfunction.

The present study aimed to assess nongenetic and genetic risk factors for chronic shoulder pain and disability in a South African cohort of mixed-ancestry BCSs. The study objectives were to determine the genotype/allele frequency distributions for the polymorphisms in ABCB1 and OPRM1, analyze inferred haplotypes between the polymorphism for each gene and investigate the gene–gene interactions between ABCB1 and OPRM1.

Materials & methods

Study design

A cross-sectional study was conducted in accordance with the STREGA reporting recommendations for the reporting of genetic association studies [41]. The study forms part of a larger ongoing project that aims to investigate the association between genetic polymorphisms within the opioid metabolic pathway and the clinical phenotype of chronic shoulder pain and dysfunction in BCSs.

Participants & setting

A total of 252 women were randomly recruited from a tertiary hospital in South Africa. Volunteers were considered eligible if they were >18 years, diagnosed with unilateral BC, had undergone BC surgery ≥1 year before recruitment and self-identified as mixed race. The mixed-ancestry population in South Africa is a unique population characterized by its diversity descending from Africa, Asia and European populations [42]. Volunteers with any history of shoulder/neck pathology before diagnoses and treatment for BC, any connective tissue disorders, renal insufficiency, diabetes mellitus, hypercholesteremia, diagnoses of local recurrences or lymphedema were excluded.

Study procedure

After receiving informed consent, relevant demographic and clinical data for each participant was obtained from the medical records and participants completed the Shoulder Pain and Disability Index (SPADI). Blood samples were drawn (10 ml) from the unaffected arm using appropriately labelled EDTA vacutainer tubes and stored at -20°C. The standard DNA extraction protocol described by Lahiri and Nurnberger [43] was employed, followed by DNA quantitation using the Take 2 plate with the BioTek HT Synergy^TM (Agilent Technologies, CA, USA) multi-plate reader. DNA purity and concentrations ranged from 1.2 to 2.165 (A₂₆₀/A₂₈₀ ratio) and 24 ng/μl to 389 ng/μl, respectively. Deidentified DNA samples were stored for genotyping analysis at -20°C.

Study outcome measures

Self-reported outcome measure: SPADI

The SPADI index, a self-reported outcome measure, was used to assess pain and disability associated with musculoskeletal pathologies of the shoulder. The index has two domains – namely, pain and disability, each with five and eight items (total of 13 items), that describe daily activities. For each domain, the items are scored on a scale of 0 (no pain/difficulty) to 10 (worst pain/difficulty), then added and converted to a percentage as a representation of pain or disability. Furthermore, for each domain the end score (in %) was used to categorize participants into no-low (scores >30%) and moderate-high (scores ≥30%) groups. Previous literature has shown that SPADI scores of >30% can affect the activities of day-to-day living of individuals and are associated with cases presenting moderate-high pain on the visual analogue scale (VAS) scale, and for that reason, we used these parameters to categorize our groups [44–46]. Reliability and validity studies show that the SPADI index have excellent reliability (test–retest ICC: ∼0.89) with factor analysis indicating the items to represent both pain and disability characteristics observed in shoulder pathologies [45,47,48]. In this study, we evaluated pain, disability and combined (pain and disability) related to the shoulder in BCSs.

SNP selection & genotyping

The candidate genes ABCB1 (rs1045642 G>A, rs1128503 G>A) and OPRM1 (rs1799971 A>G, rs540825A>T), which have previously been associated with the opioid signaling (pain inhibition) pathway, were included in this study [17,31,49]. Samples were genotyped for the polymorphisms in 96-well plates using the TaqMan^TM SNP genotyping assays (Thermo Fisher Scientific, Applied Biosystems, CA, USA) according to the manufacturer’s instructions. Accordingly, ABCB1 TaqMan^TM SNPs are classified as drug metabolizing enzyme (DME) assays, whereas OPRM1 SNPS are classified as standard assays (non-DME). Non-DME SNP: each sample well contained 0.2 μl of TaqMan specific primer (final concentrations of 1x), 4 μl of TaqMan genotype master mix, 2.8 μl of dH2O and 1 μl of DNA template (template [DNA] 1–10 ng) to a final volume of 8 μl. DME SNP: In these wells, each sample contained 0.4 μl of TaqMan specific primer (final concentrations of 1x), 4 μl of TaqMan genotype master mix, 2.6 μl of dH2O and 1 μl of DNA template (template [DNA] 1–10 ng). Negative controls (no DNA template) and technical replicates were used for quality control purposes. The standard PCR conditions for the OPRM1 SNPs were activation at 95°C for 10 min (1 cycle), followed with denaturation at 92°C for 15 s and annealing/extend/acquisition at 60°C for 60 s (40 cycles), and holding at 60°C for 10 min. PCR conditions for the ABCB1 SNPs were activation at 95°C for 10 min (1 cycle), followed with denaturation at 95°C for 15 s and annealing/extend/acquisition at 60°C for 90 s (50 cycles), and holding at 60°C for 10 min. The PCR reactions were performed using the Quant studio 3 real-time PCR (Thermo Fisher Scientific, Applied Biosystems, CA, USA) system and genotypes were automatically called and analyzed using the Thermo Fisher Cloud genotyping analysis Software version 3.3.0-SR2-build 21. Successful genotyping was considered if the sample was amplified for all polymorphisms, except when failing to amplify after two PCR reruns. A success rate of 99% was observed for ABCB1 rs1045642 G>A, rs1128503 G>A and OPRM1 rs1799971 A>G, with 94% call for OPRM1 rs540825 A>T, including the repeat typing. All laboratory work was conducted at HPALS, University of Cape Town (South Africa).

Clinical risk factors

Known clinical risk factors for chronic postoperative shoulder pain and disability were assessed for associations [46,50]. These included participants' age at the time of surgery, surgical date and type, lymph node surgery type, adjuvant therapies and BC pathology

Statistical analysis

Sample size requirement for this study was calculated using the QUANTO v1.2.4.49 software, with an assumed, literature inferred, baseline risk of 40% for chronic pain (∼36%) and disability (∼30%) in BCSs [7,51–58]. A total sample of 150 participants was calculated as an effective size for the detection of an odds ratio ≥2 with an 80% power for allele frequencies of 0.1–0.5 in the dominant model (Supplementary Table 1). For the log additive and recessive models, our sample size was adequate to reach odds of 2–2.5 and 2.5 with 80% power for the minor allele frequencies of 0.1–0.5 and >0.4, respectively.

Comparative analysis of demographic and clinical data was assessed using Statistica V13.5.0.17 [59]. Basic descriptive statistics were analyzed using independent sample t-test, Pearson’s χ2 and Fisher’s exact tests (if n < 10) for pain, disability and combined (pain and disability) between the no-low and moderate-high groups. The R studio V1.3.1056 running R v4.0.4 language and programming environment (www.r-project.org) was used to analyze all genotype data [60]. Differences in genotype/allele and inferred haplotype frequency distribution were compared using Pearson’s χ2 and Fisher’s exact tests. By using the ‘genetics’ v1.3.8.1.3 package, Hardy–Weinberg equilibrium (HWE) probabilities and linkage disequilibrium (LD) was determined [61]. Logistic regression analysis evaluating the association between the genotype and pain/disability characteristics was assessed using the ‘SNPassoc’ v2.0.2 packages [62].

To evaluate the ABCB1 (rs1045642 G>A and rs1128503 G>A) and OPRM1 (rs1799971 A>G and rs540825 T>A) genetic intervals, inferred haplotypes were constructed for each gene using the individual genotype data. In addition, as a proxy for gene–gene interactions, five stepwise inferred allele–allele combination constructs were generated between the ABCB1 and OPRM1 genes. Analysis of inferred haplotype/allelic combination frequency distribution and association between no-low and moderate-high groups was determined using the ‘haplo.stats’ v1.8.6. package [63,64]. Furthermore, genotype effects were evaluated on the clinical variables, age, time since surgery, total number of nodes involved and examined, invasive ductal carcinoma (IDC), tumor grade, surgery types (mastectomy [MTX] vs wide local excision [WLE]) and lymph node surgery (axillary lymph node dissection [ALND] vs sentinel lymph node biopsy [SLNB]).

Bioinformatic analysis was conducted to explore gene-associated networks and interactions between and for ABCB1 and OPRM1 using the GeneMANIA and Enrichr platforms. Also, using an online tool, Software for Statistical Folding of Nucleic Acids and Studies of Regulatory RNAs (SFOLD), the 2D RNA structures for the genetic intervals containing the ABCB1 and OPRM1 polymorphisms were predicted. All software and online programs used in this study can be downloaded from the respective sites reported in the reference list. Data are presented as either means ± standard deviation, median (interquartile range [IQR]), or as percentage (n values). All logistic regression analyses were covaried for age at surgery (including odds ratios [OR] and CIs at 95%) and statistical significance was accepted a p < 0.05.

Results

Demographical & clinical characteristics for pain, disability & combined (pain & disability)

Table 1 summarizes the comparative analysis for demographic and clinical data with 27.0%, 19.0% and 22.0% of individuals reporting moderate-high pain, disability and combined (pain and disability), respectively. Age differed significantly within pain (p = 0.002), disability (p = 0.007) and combined (pain and disability) (p = 0.002) groups. Younger participants were noted to report moderate-high pain, disability and combined (pain and disability). Younger participants also reported to having fewer number of nodes involved in the disability (p = 0.027) and combined (pain and disability) (p = 0.025) groups. No significant difference was observed for breast and lymph node surgery, or adjuvant treatments between no-low and moderate-high groups for pain, disability or combined (pain and disability), p > 0.05 (Supplementary Table 3).

Genotype effects of the ABCB1 & OPRM1 polymorphisms on quantitative & categorial clinical characteristics

Genotype effect was evaluated for all clinical variables and noted associations for ABCB1 rs1045642 G>A (p = 0.022), ABCB1 rs1128503 G>A (p = 0.022) and OPRM1 rs540825 T>A (p = 0.015; Supplementary Tables 5 & 7).

The ABCB1 rs1045642 G>A polymorphism was associated with ‘time since surgery’, where the median (IQR) time that had passed were 3.0 (2.0–4.0) years for (G/G) and (A/A) carriers compared with 2.0 (2.0–4.0) years for (A/G) carriers (Supplementary Table 5). The ABCB1 rs1128503 G>A polymorphism was associated with ‘age at surgery’, where the median (IQR) age was 56 (49–63) years for (A/G) carriers, compared with 54 (47–61) and 54 (48–2) years for (G/G) and (A/A) carriers respectively, rendering a possible sample bias (Supplementary Table 5).

The OPRM1 rs540825 T>A polymorphism was associated with lymph node surgery, where the (A/A) (3.6% and 3.2%) genotype was less frequently observed in the ALND and SLNB surgeries, compared to the (T/T) (59.2% and 51.6%) and (A/T) (37.3% and 45.2%) genotypes, respectively (Supplementary Table 7).

No associations were observed between OPRM1 rs1799971 A>G and the clinical variables (Supplementary Tables 5 & 7).

Genotype & allele frequency distribution of ABCB1 & OPRM1 polymorphisms

Table 2 summarizes the differences in the genotype and allele frequency distribution patterns between the no-low and moderate-high pain, disability and combined (pain and disability) groups.

ABCB1

In the evaluation of pain, no associations for ABCB1 rs1045642 G>A (p = 0.367) and rs1128503 G>A (p = 0.783) were noted (Table 2). For disability, significant frequency differences were noted for ABCB1 rs1045642 G>A where the (A/A) genotype (p = 0.028; OR = 0.21; 95% CI: 0.05–0.93) was overrepresented in the no-low (14.9%) group compared with the moderate-high (4.3%) group (Table 2). Logistic regression showed the rs1045642 (A/A) genotype (dominant: p = 0.022; OR: 0.46; 95% CI: 0.24–0.90; recessive: p = 0.045; OR: 0.28; 95% CI: 0.06–1.21), was associated with reduced likelihood of reporting disability. In addition, the ABCB1 rs1045642 (A) allele (p = 0.015; OR: 0.52; 95% CI: 0.29–0.89) was overrepresented in the no-low group (37.9%), compared with the moderate-high group (23.9%), and associated with reduced likelihood of reporting disability (Table 2). ABCB1 rs1128503 G>A showed no differences in genotype or allele frequency distribution between the no-low and moderate-high groups (p = 0.812; Table 2).

Evaluation of combined (pain and disability) showed significant differences for ABCB1 rs1045642 G>A where the (A/A) genotype (p = 0.011; OR: 0.25; 95% CI: 0.07–0.89) was overrepresented in the no-low (15.0%) group compared with the moderate-high (5.7%) group (Table 2). Logistic regression showed the (A/A+A/G) (dominant; p = 0.004; OR: 0.40; 95% CI: 0.21–0.75) genotypes were associated with a reduced likelihood of reporting moderate-high combined (pain and disability). Once more, the ABCB1 rs1045642 (A) (p = 0.003; OR: 0.50; 95% CI: 0.29–0.83) was overrepresented in the no-low (38.5%) group compared with the moderate-high (23.6%) group and was associated with a reduced likelihood of reporting moderate-high combined (pain and disability) (Table 2). No differences were detected for ABCB1 rs1128503 G>A, between the no-low and moderate-high combined (pain and disability) (p = 0.552) groups.

OPRM1

Evaluating pain, we noted no significant differences in the genotype/allele frequency distribution for OPRM1 rs1799971 A>G (p = 0.199) and rs540825 T>A (p = 0.990) between the no-low and moderate-high groups (Table 2). Although the OPRM1 rs1799971 (G) allele (p = 0.064; OR: 0.58; 95% CI: 0.30–1.04) showed a trend toward an association.

Furthermore, when disability and combined (pain and disability) were evaluated, no significant associations were noted for OPRM1 rs1799971 A>G (p = 0.497 and p = 0.587) and rs540825 T>A (p = 0.275 and p = 0.435; Table 2). All groups were in HWE for all polymorphisms when pain, disability and combined (pain and disability) were evaluated (Table 2).

Inferred haplotypes for ABCB1 & OPRM1 polymorphisms

Haplotypes were inferred and constructed for the ABCB1 (rs1045642 G>A, rs1128503 G>A) and OPRM1 (rs1799971 A>G, rs540825 T>A) genes, using the individual genotype data for the polymorphisms, in the pain, disability and combined (pain and disability) categories.

ABCB1 (rs1045642 G>A-rs1128503 G>A)

The inferred ABCB1 rs1045642 G>A-rs1128503 G>A haplotype, yielded four combinations (G-G, A-A, G-A and A-G) (Figure 1A–C). For pain (p = 0.798), no significant differences were noted between the no-low and moderate-high groups (Figure 1A). For disability (p = 0.027), the (A-G) haplotype (p = 0.029; OR: 0.00; 95% CI: 0.00–0.00) was noted to be absent in the moderate-high (0.0%) group compared with the no-low group (6.9%) and significantly associated with reduced likelihood of reporting moderate-high disability (Figure 1B). Evaluating combined (pain and disability) (p = 0.019), the (A-A) haplotype was overrepresented in the no-low (30.9%) group compared with the moderate-high (22.5%) group and was associated with reduced likelihood (p = 0.029; OR: 0.63; 95% CI: 0.37–1.06) of reporting moderate-high combined (pain and disability) (Figure 1C).

OPRM1 (rs1799971 A>G-rs540825 T>A)

The inferred OPRM1 rs1799971 A>G-OPRM1 rs540825 T>A haplotype analyses generated four combinations, (A-T, A-A, G-T and G-A) (Figure 1D–F). For pain (p = 0.040) the inferred (G-T) haplotype (p = 0.019; OR: 0.33; 95% CI: 0.14–0.75) was overrepresented in the no-low (16.4%) group compared with the moderate-high (6.9%) group and was associated with a reduced likelihood for reporting moderate-high pain (Figure 1D). On the contrary, no differences in haplotype frequency distribution patterns were noted for disability (p = 0.173) or the combined (pain and disability) (p = 0.239; Figure 1E & F).

Stepwise allele-allele interaction between ABCB1 & OPRM1 polymorphisms

Individual genotype data for the ABCB1 (rs1045642 G>A-rs1128503 G>A) and OPRM1 (rs1799971 A>G-rs540525 T>A) polymorphisms were used to construct allele–allele combinations for ABCB1 and OPRM1 as a proxy for potential gene–gene interactions.

ABCB1 (rs1045642 G>A-rs1128503 G>A)–OPRM1 (rs1799971 A>G-rs540525 T>A) generated 16 combinations, of which eight yielded frequencies >3%. Evaluation of the allele–allele combination frequencies noted no significant differences between the no-low and moderate-high groups for pain, disability and combined (pain and disability) (p > 0.05; Supplementary Figure 1). Next, the ABCB1 (rs1045642 G>A-rs1128503 G>A)–OPRM1 (rs540525 T>A) generated eight allele–allele combinations of which six combinations yielded frequencies >5%; however, no significant differences were noted between the no-low and moderate-high pain, disability and combined (pain and disability) groups (Supplementary Figure 2).

ABCB1 (rs1045642 G>A)–OPRM1 (rs1799971 A>G-rs540825 T>A) generated six (G-A-T, A-A-T, G-G-T, G-A-A, A-A-A and A-G-T) allele combinations with frequencies >5% (Figure 2). No significant differences in the frequency distribution patterns were noted for these allele–allele combinations when pain scores were evaluated (p = 0.106; Figure 2A). In the disability (p = 0.053) and combined (pain and disability) scores (p = 0.027), the (A-A-T) combination was overrepresented in the no-low (24.4% and 24.4%) compared with the moderate-high (16.0% and 16.7%) groups, respectively (Figure 2B & C). Furthermore, the (A-A-T) combination was associated with a reduced likelihood of reporting moderate-high disability (p = 0.026; OR: 0.60; 95% CI: 0.30–1.18) and combined (pain and disability) scores (p = 0.029; OR: 0.58; 95% CI: 0.18–1.45).

Evaluation of ABCB1 (rs1045642 G>A) – OPRM1 (rs1799971 A>G) generated four combinations, (G-A, G-G, A-G and A-A) (Figure 3A–C). No significant differences were noted for either pain (p = 0.244) or disability (p = 0.077) (Figure 3A & B). For combined (pain and disability) (p = 0.028), the (G-A) combination was underrepresented in the no-low (49.3%) compared with the moderate-high (65.2%) group (Figure 3C). The (A-A) combination was also overrepresented in the no-low (31.8%) compared with the moderate-high (17.6%) group (Figure 3C). Moreover, the (G-A) (p = 0.005; OR: 1.00) and (A-A) (p = 0.008; OR: 0.44; 95% CI: 0.24–0.80) combinations were associated with equal odds and reduced likelihoods of reporting moderate-high combined (pain and disability), respectively.

Evaluation of the ABCB1 (rs1045642 G>A) – OPRM1 (rs540825 T>A) identified four allele–allele combinations (G-T, A-T, G-A and A-A) (Figure 3D–F). No significant differences in allele-frequencies were noted between the no-low and moderate-high groups for pain (p = 0.684; Figure 3D). For disability (p = 0.026) the (A-T) combination (p = 0.019; OR: 0.62; 95% CI: 0.33–1.16) was overrepresented in the no-low (28.8%) group compared with the moderate-high (19.3%) group and associated with reduced likelihoods of reporting disability (Figure 3E). The (G-A) combination (p = 0.021; OR: 1.57; 95% CI: 0.30–3.10) was underrepresented in the no-low (12.9%) compared with the moderate-high (21.4%) group and associated with increased likelihood for reporting disability (Figure 3E). For combined (pain and disability), the frequencies for the (A-T) (p = 0.014; OR: 0.62; 95% CI: 0.35–1.10) and (G-A) (p = 0.030; OR: 1.50; 95% CI: 0.78–2.86) allele combinations in the no-low (28.9% and 12.7%) and moderate-high (19.8% and 20.7%) groups were comparable to the frequencies observed in the disability category (Figure 3F). Once more, the (A-T) and (G-A) combinations were associated with a reduced and increased likelihoods of reporting combined (pain and disability).

Bioinformatic analysis

GeneMANIA analysis for ABCB1 and OPRM1, found no direct associated networks; however, we identified three secondary gene-associated networks (Supplementary Table 8) [65]. The analysis showed the ABCB4, GNB1 and SLC22A2 genes each shared co-expressed, predicted, shared protein domains, genetic interactions, pathway and physical interaction networks with ABCB1 and OPRM1 (Figure 4).

The Enrichr web-based application was used to screen ABCB1 and OPRM1 against several libraries of gene sets for transcriptional and regulatory factors, biological processes, pathways, diseases/drugs and phenotypes [66]. In the transcription library, the results showed ABCB1 and OPRM1 are both predicted targets for the microRNA-875-5p (Targetscan microRNA 2017; Supplementary Table 9). In the pathway library, the ABCB1 and OPRM1 gene products are associated with T-cell receptor regulation of apoptosis (Bioplanet 2019 Pathways) and Epilepsi (Elsevier Pathways Collection; Supplementary Table 9). Gene ontology (GO) libraries showed both genes are associated with the regulation of response to stress (GO0080134, Biological Process 2021; Supplementary Table 9). In addition, in the library for diseases and drugs, the genes are shown to be associated with Huntington’s disease, a neurodegenerative disease (HDsigDB Human 2021; Supplementary Table 9).

SFOLD analysis of ABCB1 rs1045642 G>A indicated that compared to the (G) allele, the (A) allele resulted in the loss of a hydrogen bond at nucleotide position 44 in the 5′-3′ direction and resulted in a predicted increased multibranch loop (Supplementary Figure 3). In addition, the substitution noted a 3-point (-20.10 to -17.10) change in the energy reaction representing protein stability. Analysis of the rs1128503 SNP containing structure indicated no distinct changes, however, the reaction showed a changed in energy between the (G) and (A) alleles (Supplementary Figure 3). SFOLD analysis of OPRM1 rs1799971 A>G showed that compared with the (A) allele, the (G) allele resulted in the loss of an internal loop (Supplementary Figure 4). Furthermore, the A>G substitution resulted in predicted 5-point change (ΔG○₃₇ = -33.20 to -38.10) in the energy reaction, with the (G) allele noting a more negative ΔG○₃₇, compared with the (A) allele. The secondary predicted changes for OPRM1 rs540825 T>A, indicated no structural differences between these two alleles; however, a change in the energy reaction was noted (ΔG○₃₇ = -22.70 to -19.70; Supplementary Figure 4).

Discussion

This study is the first to evaluate associations between polymorphisms in genes involved in the opioid metabolism and pain pathway, and chronic shoulder pain and dysfunction in BCSs of mixed ancestry from South Africa. The findings of our study confirms that age is a risk factor and infers that polymorphisms in the ABCB1 and OPRM1 genes may play a role in the development of chronic shoulder pain and dysfunction in BCSs.

Younger age was associated with an increased likelihood for reporting moderate-high pain, disability and combined (pain and disability) (Table 1). This result is in alignment with previous research reporting the association between age and persistent pain in BCSs [12,67]. It is proposed that changes in pain perception is age related; compared with older adults (>65 years), younger adults (18–44 years) report greater pain due to psychological stress [68]. Younger adults with chronic pain suffer physiological and mental health issues, and the disruption of daily routines, including the ability to work full time that can lead to a financial burden [69,70].

With more than 70% of participants having received an ALND, it was noteworthy that no significant association was found between ALND and shoulder pain, disability or combined (pain and disability). ALND is the surgical removal of lymph nodes in the axilla, which is a secondary measure to the detection of a tumor positive sentinel lymph node [71,72]. Moreover, it has been described as a risk factor for the development of persistent pain post-surgery for BC [12].

Evaluation of the genetic contribution of ABCB1 gene polymorphisms in modulating shoulder disability and movement-related pain in this BCS cohort showed that the ABCB1 rs1045642 (A/A) genotype and (A) allele were significantly associated with 80/75% and 48/50% reduced likelihood of reporting moderate-high disability/combined (pain and disability), respectively. Haplotype analyses further highlighted the ABCB1 genetic locus in modulating pain and disability where the ABCB1 (rs1045642 G>A – rs1128503 G>A) inferred (A-G) haplotype was associated with reduced likelihood of reporting disability. The (A-G) haplotype was absent in the moderate-high group, which inferred a reduced (100%) likelihood for reporting disability. The findings of our study are correlative to other studies reporting lower pain, although there are few data on postoperative pain [28,73,74]. The functional rs1045642 (A) and rs1128503 (A) alleles have steadily been associated with requiring fewer opioids compared with (G) allele carriers and with rs1045642 (A) carriers regarded as ‘good responders’ [17,29,35,75,76]. It is reported that the rs1045642 (A/A) genotype and (A) allele is associated with decreased protein functions and are at greater risk of opioid-related side effects that includes sweating, muscular tension, stress and sedation, compared with (G/G) and (A/G) carriers [26,77,78]. Furthermore, given its implication in opioid requirements and response studies, it is hypothesized that the (A/A) genotype have lower P-gp expression that may result in less efflux of opioids at the BBB level and therefore individuals with that genotype profile, may need fewer opioids to control pain. This hypothesis is supported by the findings of Hoffmeyer et al. [26], who showed a reduction in P-gp expression and altered activity related to the (A/A) genotype. More recently, allelic-specific expression analysis of liver autopsy samples noted the rs1045642 (A) allele was associated lower mRNA levels compared with the (G) allele [79].

Analyses of the OPRM1 genetic locus showed that although no independent associations were noted for the individual polymorphisms investigated, the inferred haplotypes implicate the genetic interval spanning these polymorphisms in modulating chronic shoulder pain and disability. The OPRM1 (rs1799971 A>G – rs540825 T>A) inferred (G-T) haplotype was associated with a 67% reduced likelihood of reporting pain (Figure 2A). Research data for the OPRM1 (rs1799971 A>G – rs540825 T>A) locus in postoperative pain is limited. A study conducted by De Gregori et al [80] analyzing seven OPRM1 polymorphisms, including both rs1799971 A>G and rs540827 T>A, observed that haplotype carriers containing the (G) and (T) alleles required more postoperative analgesia (POA) than haplotype carriers containing the (A) and (A) alleles, respectively. This observation suggests that the rs1799971 (G) and rs540825 (T) allele, in combination, required more POA may be in response to experiencing greater pain. It has been shown that the rs1799971 (G) allele increases the receptor binding affinity threefold for endogenous opioids yet reduces the expression of the receptor and leads to high morphine requirements [81]. Also, OPRM1 rs1799971 and rs540825 polymorphisms have both been investigated in pain and opioid studies because both have been described to modify the mu-receptor properties, although inconsistencies remain [30,36,40,49,76,80,82,83].

Evaluation of the ABCB1-OPRM1 allele combinations highlighted a combined genetic contribution to modulating disability and movement-related pain in this BCS cohort. Several associations were noted; however, the most interesting finding was for ABCB1 (rs1045642 G>A) – OPRM1 (rs540825 T>A) where the (A-T) allele combination was associated with reduced (38% and 38%) likelihood of reporting moderate-high disability and combined (pain and disability) (Figure 3E). Coupled with the alternate combination, (G-A), being associated with increased (57% and 50%) likelihood of reporting moderate-high disability and combined (pain and disability) (Figure 3F). Research data for pain studies surrounding the gene–gene interactions between the ABCB1 and OPRM1 genes, are limited. On the basis of these findings, we hypothesize that the ABCB1 rs1045642 (A) allele may act as a potential important regulator of opioid distribution. The OPRM1 rs540825 (T) allele is reported to require more opioids and, when inherited with the ABCB1 rs1045642 (A) allele, resulted in reduced likelihood of reporting pain and disability [80]. This is further supported by the alternate allele combinations whereby the OPRM1 rs540825 (A) allele that is reported to require fewer opioids are inherited with the ABCB1 rs1045642 (G) allele, shown to have increased P-gp expression, an increased likelihood of reporting pain and disability was noted. This could be explained as nociceptive sensitization by exposure to opioids, leading to opioid-induced hyperalgesia (OIH), the paradoxical effect that may intensify preexisting pain [84]. Furthermore, it is reported that upregulation of the P-gp membrane protein may contribute to opioid tolerance [85].

Another noteworthy association was the inferred (A-A) (ABCB1 rs1045642 G>A – OPRM1 rs1799971 A>G) allele combination associated with a 66% reduction in the likelihood of reporting combined (pain and disability) (Figure 3C). It was interesting to note in a previous study that haplotype carriers containing the ABCB1 rs1045642 (A) and rs1128503 (G) alleles in combination with the OPRM1 rs1799971 (A) allele were observed to have lower methadone dosage and plasma concentrations than the alternate ABCB1 alleles [86], allowing the conclusion that ABCB1 (lower) and OPRM1 (higher) polymorphisms were associated with opposing dose requirements [86]. This supports the findings of the present study that the ABCB1 rs1045642 (A) allele, when inherited with the OPRM1 rs1799971 (A), modulates pain scores and, in the previous study, opioid requirements.

In the bioinformatic analysis, the GeneMANIA results showed that ABCB1 and OPRM1 share functional and associated networks that interact. Additionally, enrichment analysis generated form the libraries in Enrichr, showed ABCB1 and OPRM1 share common pathways related to apoptosis, immune response and opioid signaling. Using online mRNA structure prediction tools, we examined the predicted 2D RNA structure obtained by SFOLD for the regions containing the ABCB1 and OPRM1 SNPs. It is reported that SNPs can give rise to varying forms of mRNA structures, that consequently may affect mRNA stability and potentially alter translation efficiency of the protein and thereby affect protein function [87-90].

For the ABCB1 rs1045642 G>A polymorphism, the predicted secondary structure for the (A) allele showed an increased multibranch loop at that position compared with the (G) allele form. In addition, the (A) allele showed an increase in the ΔG○₃₇ score indicating a less stable mRNA structure compared with the (G) allele form, which supports the findings observed in the study conducted by Wang and Sadee [79].

We hypothesize that this polymorphism could therefore potentially modulate expression and availability of the transporter, through such structural changes. Although the ABCB1 rs1128503 (G) and (A) allele forms showed no differences in the predicted secondary mRNA structures, changes in their melting domains were noted as reflected in the energy reaction of the ΔG○₃₇ scores, indicating that this polymorphism may still affect the transport through other processes; for example, it may change interactions/efficiencies of the transporter protein within the cellular environment [87–90]. For the OPRM1 rs1799971 A>G polymorphism, the predicted secondary mRNA structure for the (G) allele, noted the loss of one (of three) internal loop at that position. Compared with the (A) allele, the (G) allele forms a hydrogen bond with the opposite nucleotide at that position. The predicted secondary structure for the (G) allele also noted a decrease in ΔG○₃₇ scores, which suggests a more stable mRNA transcript compared with the (A) allele form, and therefore this polymorphism could potentially modulate the expression and availability of the mu-opioid receptor through such structural changes. For the OPRM1 rs540825 T>A polymorphisms, no structural differences between the (T) and (A) allele forms were noted. However, like ABCB1 rs1128503 G>A, the differences in ΔG○₃₇ scores suggests that the polymorphism may still affect the receptor’s binding capacity through mRNA-structure-dependent processes [87–90].

With nearly 50% of BCSs suffering chronic shoulder pain and disability, it is imperative to understand and identify risk factors influencing susceptibility [7]. Current research in chronic postoperative pain phenotypes have described severe acute postoperative pain as a risk factor [12]. Moreover, genetic association studies report the relationship between candidate gene polymorphisms and pain phenotypes are vital in the understanding of the mechanisms underlying these differences in pain relief/perception [87]. We, therefore, hypothesize that inadequate pain relief during the acute postoperative period may influence pain severity thereby increasing the likelihood of developing chronic pain. The results found in this study show the complex nature of functional genetic polymorphisms with the potential to alter the protein structure and function.

Limitations

There are several limitations to this study. The sample was powered at <80% to detect effects sizes of OR = 1.5 (Supplementary Table 2). The study followed a hypothesis approach in which both ABCB1 and OPRM1 polymorphisms were previously implicated in the pain phenotype. More than two genetic polymorphisms were assessed (family-wise error rate) and, taken together with the small and underpowered (<80%) sample size, multiple testing was not adjusted for. Logistic regression analysis was applied to investigate gene–gene interactions. However, it is reported that this tends to be difficult to detect because many multilocus genotype combinations may have few or no data points, thereby hindering the outcome [88]. In our analysis, allele frequencies >3% were used to describe the gene–gene interaction between ABCB1 and OPRM1. Increasing the sample size may increase the power to detect significant differences in clinically relevant characteristics with genotype/allele frequencies and allow us to consider clinically relevant confounders in the regression analyses. In addition, in this cohort ethnicity was self-reported and therefore does not hold the same strength as genomic estimates, which may undermine the population stratification of our cohort.

Conclusion

In conclusion, this study adds to the consensus that age is a risk factor for pain and that lymph node surgery could potentially identify individuals that may be susceptible to pain and dysfunction in BCSs. Furthermore, this study provides evidence of an association between the genetic polymorphisms of the ABCB1 and OPRM1 genes and chronic shoulder pain and disability in BCSs. Future studies incorporating bigger genetic intervals and larger population sizes are required to further scrutinize and elucidate the role of the ABCB1 and OPRM1 genes in chronic shoulder pain and dysfunction.