Effect of Perioperative Opioids on Cancer-Relevant Circulating Parameters: Mu Opioid Receptor and Toll-Like Receptor 4 Activation Potential, and Proteolytic Profile

The possibility that opioids can modulate tumor growth and metastasis is of intense interest. Opioids have been documented either to promote or to prevent tumor growth and metastasis. This dichotomy presumably arises from the use of different models; the use of a vast range of opiate doses and concentrations in chronic or acute regimens; and the recent observation that some effects of opioids are not opioid receptor (OR)–mediated (1).

We have shown that morphine alters the circulating proteolytic profile of tumor-bearing and tumor-free mice (2, 3), and downregulates the proinvasive and proangiogenic interaction between breast cancer cells and macrophages in vitro (4, 5). Morphine decreased matrix metalloprotease-9 (MMP-9) and increased tissue inhibitor of metalloproteases-1 (TIMP-1) in multiple experimental models (2–4), with consequences on tumor cell migration and invasion (3, 4).

Studies of the OR-independent actions of opioids have identified that the toll-like receptor 4 (TLR4) can respond to opioids (6, 7). Opioids or their metabolites can activate both the μ-opioid receptor (MOR) and TLR4, which are expressed on cancer cells as well as tumor-associated cells, and regulate metastasis signaling pathways (8–10). Our recent in vitro results confirm that opioids per se activate TLR4, but they prevent lipopolysaccharide (LPS)-induced TRL4 activation (7). We have further shown that plasma from morphine-administered mice could activate MOR and TLR4, and prevent LPS-induced TLR4 activation (7).

The present study was designed to investigate the potential interplay between opioid analgesia and tumor metastasis through modulation of MOR and TLR4 activation, and matrix degradation potential, in patients undergoing surgery. The circulating parameters measured help to understand whether and how perioperative opioids can influence the long-term disease outcome. We deliberately studied a noncancer patient cohort in order to avoid interference from the effect of cancer on immunomodulation, serum constituents, and receptor activation.

Plasma and serum samples were collected from a cohort of 60 patients undergoing elective lower limb joint replacement preoperatively and at 3, 6, and 24 hours after the surgery. Plasma and serum were also collected from 10 healthy volunteers. Exclusion criteria included cancer, pregnancy, and systemic inflammatory disease such as rheumatoid arthritis and sero-negative arthritides. Data collection included patient characteristics (operation, age, and gender), details of opioid administration (agents, route, dose, and timing), and visual analog and numerical rating pain scores (recorded at the time of blood sample collections). Opioid administration was converted into morphine IV equivalents using the conversion table of the Australia and New Zealand College of Anesthetists Faculty of Pain Medicine (11), with the exclusion of remifentanil due to its short half-life. The study was approved by The University of Queensland Medical Research Ethics Committee (Approval Number: 2014000678) and Mater Health Services Human Research Ethics Committee (Mater reference number: HREC/14/MHS/18). Informed consent was obtained from each subject.

Plasma samples were collected from morphine-treated mice (1 or 10 mg/kg) 10 minutes after intraperitoneal injection in a study previously described (3, 7) and approved by The University of Queensland Animal Ethics Committee (AEC Approval Number: PHARM/381/13/ANZCA).

The following reagents were purchased from Invitrogen (Life Technologies): DMEM, trypsin-EDTA, penicillin/streptomycin, TNFα, and G418 sulfate solutions. Normocin, HEK-Blue Selection, Zeocin, Hygromycin B Gold, LPS from the photosynthetic bacterium Rhodobacter sphaeroides (LPS-RS), and QUANTI-Blue were purchased from Jomar Life Research. The Alphascreen cAMP assay Kit was from PerkinElmer. Forskolin (Fsk) and 3-Isobutyl-1-methylxanthine (IBMX) were obtained from Sapphire Bioscience. The following reagents were supplied by Sigma-Aldrich: FBS, LPS from Escherichia coli, Hanks' balanced salts, Tween 20, and 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES).

HEK 293 (human embryonic kidney 293) cells stably transfected with MOR (HEK-MOP cells) were generated as described previously (12) and cultured in DMEM containing 100 U/mL penicillin, 100 μg/mL streptomycin, and 500 μg/mL G418 sulfate. All HEK-Blue cell lines were purchased from Jomar Life Research. HEK-Blue-hTLR4 cells [HEK 293 cells stably transfected with the human TLR4, coreceptor MD-2 and CD14, as well as secreted embryonic alkaline phosphatase (SEAP) reporter genes], HEK-Blue hMD2-CD14 cells (HEK 293 cells overexpress human MD-2 and CD14 and the SEAP reporter genes), and HEK-Blue Null2 cells (HEK 293 cells harboring the SEAP reporter gene only) were maintained in DMEM plus 50 U/mL penicillin, 50 μg/mL streptomycin, 1X HEK-Blue Selection antibiotics (HEK-Blue-hTLR4 cells), 200 μg/mL of Hygromycin B Gold and 100 μg/mL of Zeocin (HEK-Blue hMD2-CD14 cells), 100 μg/mL of Zeocin (HEK-Blue Null2 cells) to maintain the stability of the transgenes, and 100 μg/mL Normocin (a mix of two antibiotics active against mycoplasma, Gram-positive and Gram-negative bacteria, and one antifungal; all cells). No routine mycoplasma testing was performed. Identity of the cells was verified by regular inclusion of positive controls activating transgene products. All cells were grown for less than 10 passages after thawing, in medium supplemented with 10% (v/v) FBS at 37°C with 5% CO₂.

The μ receptor activation was assessed using AlphaScreen cAMP Assay and HEK293 cells overexpressing the μ receptor or control HEK-Blue Null2 cells that do not express the μ receptor. The assay measures endogenous cAMP production by FSK-stimulated cells in a competition-based assay. Activation of MOR downregulates cellular cAMP production. As previously described (7), 4.5 μL of stimulation buffer (1x HBSS, 1%BSA, 0.5 mmol/L IBMX, and 5 mmol/L HEPES) and 0.5 μL patient plasma samples with or without Fsk (30 μmol/L final concentration) were added to each well of 96-well half-area plates containing 5 μL/well of a combination of HEK-MOP cells (∼25,000 cells) and anti-cAMP acceptor beads (0.4 units/μL), and incubated with shaking for 30 minutes at room temperature in the dark. Biotinylated cAMP and streptavidin donor beads in 15 μL lysis buffer (0.3% Tween-20, 0.1% BSA, and 5 mmol/L HEPES) were added to the reaction system for incubation overnight, with shaking in the dark. The fluorescent signal was analyzed using the Perkin EnVision-Alpha Reader 2101 (PerkinElmer).

The determination of TLR4 activation relies on the detection of SEAP under the control of NF-κB and AP-1 by cells stably transfected with hTLR4 and MD-2/CD14 coreceptor genes. Activation of TLR4 activates NF-κB and AP-1, which induce the production of SEAP in the culture medium. This may be determined using a second-step QUANTI-Blue colorimetric enzyme assay. As previously described (7), HEK-Blue-hTLR4 cells were seeded in 96-well plates (10,000 cells/well) and incubated for 48 hours. The cells were then treated with patient plasma samples [5% (v/v)] with or without LPS in serum-free medium and incubated for 12 hours at 37°C in 5% CO₂. In control experiments, LPS was replaced with TNFα as a TLR4-independent NF-κB activator (7). Cell culture supernatant of SEAP-expressing cells (20 μL) was added into 96-well plates containing 180 μL/well of prewarmed QUANTI-Blue aliquots, and incubated at 37°C in 5% CO₂ for 4 hours. The absorbance of QUANTI-Blue was read at a wavelength of 655 nm with a spectrophotometer (Bio-Rad Laboratories Inc.).

Serum samples from patients were analyzed for matrix metalloproteinases (MMP) and TIMPs using previously described gelatin zymography and reverse zymography, respectively (3). A polyacrylamide gel [11% (w/v)] containing 1% (w/v) gelatin was used to quantify gelatinases (MMP-2 and MMP-9), and a 12% (w/v) polyacrylamide gel containing 1% (w/v) gelatin and 20% (v/v) NIH3T3-conditioned medium as a source of gelatinase was prepared to quantify the level of TIMPs. Equal amounts of protein from the samples were separated using SDS-PAGE electrophoresis. The gels were incubated and rinsed with the following solutions: renaturing solution [5 mmol/L CaCl₂, 50 mmol/L Tris, and 2.5% (v/v) Triton X-100] overnight, solution containing 50 mmol/L Tris–HCl and 5 mmol/L CaCl₂ at 37°C for 3 hours, staining solution [0.25% (w/v) Coomassie Blue R-250, 45% (v/v) methanol, 10% (v/v) glacial acetic acid] overnight. The gels were destained using an aqueous solution of 25% (v/v) methanol and 10% (v/v) glacial acetic acid. Hydrolysis of gelatin by proteinases is reflected as clear bands on a blue background of undegraded gelatin after destaining, whereas TIMPs activity inhibit the proteolytic activity of MMPs, resulting in dark blue bands against the background where gelatin is degraded. The gels were scanned using high-resolution flatbed scanning, and band intensities were measured using ImageJ software.

Samples collected from 60 patients at 4 time points and 10 healthy volunteers were randomized into 3 different set-ups for distribution in 96-well plates for 3 independent measurements of MOR or TLR4 activation potential. Similarly, samples from 60 patients at 2 time points (preoperatively and at 24 hours after the surgery) and healthy volunteers were randomized into three different set-ups for distribution in zymography and reverse zymography gels. To normalize data, a mixture of all samples collected from the 10 healthy volunteers (referred to as mini mix) and a mixture of all samples (referred to as super mix) were tested in every plate of cAMP assay and TLR4 signaling assay, and in every gel of zymography and reverse zymography. Within each independent experiment, data were normalized using super mix/mini mix for comparison of results across multiple plates or gels.

Statistical analysis was performed using GraphPad Prism software (v. 7.00) and R statistical software (13). Data are shown as mean ± SEM of patient groups as detailed in the figure legends. The comparison between three or more groups defined by one factor was performed using one-way ANOVA followed by post hoc Dunnett multiple comparison test. The comparison between individual groups of interest was carried out using Student t tests or Mann–Whitney test as specified. The Pearson correlation coefficient was calculated between pairwise predictors. Linear mixed models were fitted with time and clinical variables as fixed effect, and subjects as random effect using lme4 R package (14). P values from t tests via the Satterthwaite approximation are reported, with a significance level set to 0.05.

Of the 60 patients (35 males and 25 females), 45 received some form of regional and/or neuraxial anesthesia, whereas 15 did not (Table 1). Fentanyl represented the majority (∼80%) of systemic opioids administered during the course of the study (Fig. 1A). Preoperative opioid administration was minimal, with 2 patients receiving 200 mg each of slow release tramadol. The highest morphine equivalents (per hour) were administered in the early postoperative period (first 3 hours; Fig. 1B). Those patients receiving any form of regional and/or neuraxial anesthesia were administered less opioids than those patients without regional anesthesia during and for 24 hours after surgery: cumulative IV morphine equivalents mg [median (first, third quartile): 52 (20.67, 217.3) vs. 100.3 (55, 259.3), respectively; (P: 0.0027, Mann–Whitney test)].

Numerical pain scores and visual analog pain scores were recorded. Both were significantly reduced at 3 and 6 hours (Fig. 1C). Pairwise correlation showed that numerical and visual analog pain measures were highly correlated [3 hours: correlation 0.83 (P: 1.41 × 10⁻¹⁵); 6 hours: correlation 0.67 (P: 6.81 × 10⁻⁰⁹); 24 hours: correlation 0.64 (P: 2.97 × 10⁻⁰⁸)], and clinical pain score was therefore used for all subsequent analyses.

The ability of the plasma samples to activate MOR was examined via cAMP production in an MOR-overexpressing cell line (Fig. 2A). There was significant MOR activation (i.e., decreased cAMP production) at all three time points after the surgery. To verify that decreased cAMP in the HEK-MOP cells was not due to MOR-independent cellular signaling elicited by factors present in plasma, the experiment was repeated using control HEK cells that do not overexpress the MOR (HEK-Blue Null2 cells; Fig. 2B). No decrease in cAMP production was observed in the control cell line. Therefore, decreased cAMP production by HEK-MOP cells is due to MOR-active molecules.

The ability of all plasma samples to activate TLR4 was tested using a reporter cell line overexpressing TLR4 and its two coreceptors MD2 and CD14. There was no increase in TLR4 activation potential in the postoperative period (Fig. 2C). On the contrary, at 3 and 6 hours, a slight decrease was detected. Opioids are known to elicit weak activation of TLR4, but to strongly inhibit TLR4 LPS-induced activation (15). We examined whether the preoperative plasma samples contained any TLR4-activating molecule in comparison with samples from the 10 healthy controls (Supplementary Fig. S1). Preoperative plasma from the patients indeed had significantly more TRL4 activation potential than healthy control plasma (P < 0.0001). We tested the ability of the plasma samples to modulate TLR4 activation induced by 1 ng/mL LPS. There was statistically significant inhibition at all three time points (Fig. 2D). To ensure that this effect was TLR4-dependent, we tested the ability of the plasma samples to prevent TNFα-induced NF-κB–mediated SEAP production. The results showed no effect of patient plasma samples on TNFα-induced SEAP expression by either HEK-Blue hTLR4 cells (Fig. 2E) or HEK-Blue hMD2-CD14 cells (Fig. 2F), confirming that the inhibition of LPS-induced activation is indeed TLR4-dependent.

We analyzed the samples collected at 24 hours for MMP-2 and MMP-9 activity via gelatin zymography and for TIMP activity using reverse gelatin zymography (Fig. 3). Analysis was restricted to the preoperative and 24-hour samples due to the delayed effect of morphine administration on the proteolytic profile in mice (3). There was no change at 24 hours in the MMP-2 and MMP-9 activities when compared with preoperative values. However, there was a decrease in the MMP-9–TIMP complex detected at 129 kDa (Fig. 3A). There was no overall change in TIMP-2 activity, but an approximate 4-fold increase in TIMP-1 was measured (Fig. 3B). This further translated into a significant decrease of the ratio MMP-9/TIMP-1 calculated for each sample (Fig. 3B).

Pairwise correlation coefficients were calculated between the variables measured at each postoperative time point. Pain correlated negatively with HEK-MOP cAMP production [correlation –0.28 (P = 0.036)], signifying that pain correlated positively with MOR activation, at the 3-hour time point only. LPS-induced TLR4 activation correlated with cAMP production [correlation 0.35 (P = 0.007)], which signifies that the ability of the samples to inhibit ligand-induced TLR4 activation correlated with MOR activation, at the 24-hour time point only. There was no correlation between cAMP production or opioid intake and any of the variables reflecting proteolytic potential that were altered postoperatively (namely, MMP-9–TIMP-1 complex, TIMP-1, MMP-9/TIMP-1 ratio). There was also no correlation between pain scores and opioid administration at each time point. However, there was a significant correlation between cumulative postoperative opioid intake and average pain score over the 24-hour postsurgery [correlation 0.4628 (P: 0.0003)].

Six linear mixed models were evaluated to examine the effect of the clinical variables measured at four time points to model either MOR activation, morphine intake, cumulative morphine intake, pain, TLR4 activation potential, or inhibition of LPS-induced TLR4 activation, and revealed the following: cAMP production had a significant effect to predict LPS-induced TLR4 activation (coefficient 0.25, P = 0.038; Table 2) and TLR4 activation had a significant effect to explain pain scores (coefficient 0.31, P = 0.038; Table 3). The ability of MOR activation to explain pain scores (P = 0.053) and TLR4 activation (P = 0.052) was borderline significant.

The most striking finding of our study is that the ability of plasma samples to activate MOR had a significant effect on their ability to inhibit LPS-induced TLR4 activation, which was observed in the linear mixed model, and the pair-wise correlation between these two variables at 24 hours. This is the first demonstration of the ability of opioids to modulate TLR4 activation in a clinical setting. The ability of several opioids to moderately activate TLR4, but to significantly prevent LPS-induced TLR4 activation has been demonstrated in vitro (7, 15, 16) and in silico (16, 17). In addition, we have shown that the plasma of morphine-treated mice could inhibit TLR4 activation by LPS at time points where opioids (morphine and morphine-3 glucuronide) were detectable and measured in the circulation (7). Analysis of plasma samples from mice administered with 0, 1, or 10 mg/kg morphine collected 10 minutes after intraperitoneal morphine injection (7) exposes a correlation (Supplementary Fig. S2) similar to that observed in the current clinical study, further strengthening our findings.

The clinical samples did not exhibit TLR4 activation despite the presence of opioids, and this may be due to a number of reasons. First, not all opioids activate TLR4 significantly (7), and many opioids only weakly activate TLR4 while they potently prevent LPS-induced activation (15). It could be that the opioids present in the postoperative plasma samples are either qualitatively or quantitatively inadequate to elicit measurable activation. Second, the observed decrease in TLR4 activation instead of the posited increase (Fig. 2C) suggests that the preoperative patient samples may contain activators of TLR4. To examine this possibility, we compared their abilities to activate TLR4 with those of samples from healthy controls. Indeed, activation of TLR4 was significantly more prominent in preoperative patient samples than in healthy control samples (Supplementary Fig. S1B), which may be inherent to the pathology underlying the need for surgery (18) and in line with the increased MMP-9 and increased TIMP-1 observed in the preoperative patient samples compared with the healthy controls (Supplementary Fig. S3).

The net effect on tumor growth and metastasis resulting from the modulation of TLR4 activation by opioids is unknown at present and is possibly context-dependent: expression or upregulation of TLR4 has been documented in tumor cells (19, 20), and although not every time the case, activation of TLR4 is mostly reported to promote cancer cell–aggressive behavior (8, 9, 21) and sustain a tumor-promoting inflammatory response (22). In contrast, activation of immune cell TLR4 is identified to be protective in the context of cancer, including in preclinical and clinical studies; a synthetic TLR4 agonist was shown to reduce adrenergic stress-induced metastasis in rodents (23). Furthermore, TLR4 is required for the immune response against dying tumor cells and dictates the efficacy of antitumor therapy in humans (24). Future studies linking modulation of TLR4 activation to disease outcome will benefit from this assay, performed with plasma samples in both the absence and presence of a TLR4 agonist such as LPS.

Our results did not find a correlation between opioid administration and plasma MOR activation potential, or that opioid intake could predict MOR activation. To eliminate the possibility that this could be due an inability of the cAMP assay to reflect the amount of opioid present in the circulation, we determined the relationship between (i) cAMP production and morphine dose administered to mice (Supplementary Fig. S4A) and (ii) cAMP production and morphine concentration added to spiked plasma (Supplementary Fig. S4B). A strong inverse correlation was found between morphine concentration and cAMP production, indicating that cAMP production is a valid measure of the ability of a plasma sample to activate the MOR. The results from the clinical samples shown above could be due to the grouping of opioid administration over a period of time in our analysis, especially between the 6 and the 24-hour time points, which does not take into account the time, route of administration, half-life, or duration of action of each opioid. It could also be due to differences in opioid metabolism between patients (25, 26), to the IV administration of an MOR antagonist (naloxone, 2 patients, three samples in total), or to a systemic contribution of neuraxial opioids. However, spinal doses were extremely low when compared with systemic opioid doses, and only 1 patient was administered epidural fentanyl. Although opioid intake imperfectly reflects overall exposure through a period of time, the cAMP production assay offers a measurement of actual exposure at a given time point (when the blood sample is drawn) and presents the advantage of measuring the resulting action of combined endogenous and exogenous agonists, antagonists, and their metabolites, some of which are MOR-active (e.g., morphine-6-glucuronide or normorphine). The correlation between MOR activation and pain at the 3-hour time point may signify that patients with more pain received more opiates, but this is not substantiated by the relationship between opioid intake during the first 3 hours and the pain scores at 3 hours (P: 0.418).

There was a significant change in proteolytic profile in the 24-hour postsurgery plasma samples, with slight decreases in the MMP-9–TIMP complex and dramatic increases in TIMP-1. Our preclinical data showed a decrease in promigratory and proinvasive profile in the circulation of morphine-treated mice, associated with a decrease in MMP-9 and an increase in TIMPs (3). Surprisingly, in the present study, there was no correlation between opioid (intake or MOR activation) and the changes in proteolytic profile. This could be due to the fact that the time point at which we quantified MMP-9 and TIMP-1 in the circulation is not suitable to measure opioid-mediated changes, which in mice were seen after 3 days of morphine treatment (3)—a clinical study collecting blood at a later time than 24 hours could potentially answer this question. It is possible that the change in TIMP-1 we observed at 24 hours is due to effects of surgery rather than to opioid administration, as there are reports that TIMP-1 concentration is significantly increased following surgical trauma and the resulting inflammation (27). A study of patients undergoing colon cancer resection showed that after an initial decrease of 1 to 2 hours after surgery, TIMP-1 peaked at day 1 after surgery (28). Last, the changes due to opioids may not be apparent because the preoperative samples already have altered proteolytic profiles when compared with healthy subjects (Supplementary Fig. S3).

Another notable finding of our study is that TLR4 activation potential of the plasma samples had a significant effect to explain increasing pain scores, as revealed by the linear mixed model. TLR4 has been implicated as a mediator of pain in a number of preclinical models including neuropathic pain (29), bone cancer pain (30), surgical pain (31), ischemia-reperfusion–induced inflammatory pain (32), and opioid-induced hyperalgesia (6). One explanation for our finding is that surgery-induced tissue injury results in the release of damage-associated molecular pattern molecules capable of activating TLR4, such as HMGB1. Indeed, circulating HMGB1 was found to correlate positively with the length of surgery, blood loss, and fluid intake at days 1, 2, and 3 after major abdominal surgery (33). Clinical studies attempting to evaluate TLR4-mediated inflammation via circulating markers usually employ ELISA quantitation of alarmins such as HMGB1, TLR4 mRNA expression in whole blood or mononuclear cell subsets, or mononuclear cell isolation and TLR4 activation ex vivo; increased TLR4 expression and ex vivo activation on a subset of mononuclear cells 24 to 48 hours after major abdominal surgery were proposed to predict systemic inflammatory syndrome (34). In contrast, ex vivo TLR4 activation of whole blood cell cultures was shown to be suppressed in trauma ICU patients (35). The measurement of the ability of plasma samples to activate TLR4 on a reporter cell line, which we employ for the first time in a clinical study, presents the advantage of representing the combined effect of agonists and antagonists of the receptor that may be present in the circulation at a given time point, and provides valuable information when used with the appropriate controls (control cells lacking the TLR4 activation in the presence of TLR4-independent NF-κB–inducing factors; ref. 7).

Despite these limitations, evaluating the effect of perioperatively administered opioids on circulating parameters likely to affect the biology of cancer cells and other prominent tumor-associated cells represents a novel and promising approach to understanding whether perioperative analgesia of cancer surgery patients can influence the risk of long-term metastasis or recurrence.

No potential conflicts of interest were disclosed.

The sponsor had no involvement in study design, in collection, analysis and interpretation of data, in the writing of the report, and in the decision to submit the article for publication.

Conception and design: N. Xie, N. Matigian, P.J. Cabot, P.N. Shaw, K.-A. Lê Cao, D. Sturgess, M.-O. Parat

Development of methodology: N. Xie, P.J. Cabot, P.N. Shaw, M.-O. Parat

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): N. Xie, T. Vithanage, K. Gregory, Z.D. Nassar, D. Sturgess

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): N. Xie, N. Matigian, T. Vithanage, C.M.J. Kirkpatrick, K.-A. Lê Cao, D. Sturgess, M.-O. Parat

Writing, review, and/or revision of the manuscript: N. Xie, N. Matigian, P.N. Shaw, C.M.J. Kirkpatrick, K.-A. Lê Cao, D. Sturgess, M.-O. Parat

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): K. Gregory

Study supervision: D. Sturgess, M.-O. Parat

We thank Rau Walker for her technical assistance with blood collection. This study was supported in part by Pfizer funds granted by the Anaesthesia and Pain Medicine Foundation, Australian and New Zealand College of Anaesthetists (Project Grant 14/020, to D. Sturgess, M.-O. Parat, P.J. Cabot, and P.N. Shaw).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.