Chronic pain is a global burden that promotes disability and unnecessary suffering. To date, efficacious treatment of chronic pain has not been achieved. Thus, new therapeutic targets are needed. Here, we demonstrate that increasing endogenous adenosine levels through selective adenosine kinase inhibition produces powerful analgesic effects in rodent models of experimental neuropathic pain through the A3 adenosine receptor (A3AR, now known as ADORA3) signalling pathway. Similar results were obtained by the administration of a novel and highly selective A3AR agonist. These effects were prevented by blockade of spinal and supraspinal A3AR, lost in A3AR knock-out mice, and independent of opioid and endocannabinoid mechanisms. A3AR activation also relieved non-evoked spontaneous pain behaviours without promoting analgesic tolerance or inherent reward. Further examination revealed that A3AR activation reduced spinal cord pain processing by decreasing the excitability of spinal wide dynamic range neurons and producing supraspinal inhibition of spinal nociception through activation of serotonergic and noradrenergic bulbospinal circuits. Critically, engaging the A3AR mechanism did not alter nociceptive thresholds in non-neuropathy animals and therefore produced selective alleviation of persistent neuropathic pain states. These studies reveal A3AR activation by adenosine as an endogenous anti-nociceptive pathway and support the development of A3AR agonists as novel therapeutics to treat chronic pain.
Chronic pain is an enormous unmet medical need with a multi-billion dollar impact on society (Pizzo and Clark, 2012). The most successful pharmacological approaches for the treatment of chronic pain rely on engagement of endogenous circuits involving opioid, adrenergic, and calcium channel mechanisms (Millan, 2002). However, drugs exploiting these pathways are often associated with intolerable side effects that result in discontinued use, inadequate pain relief, and diminished quality of life. The purine nucleoside adenosine functions in extracellular signalling within the peripheral nervous system and CNS at four G protein-coupled adenosine receptor (AR) subtypes: A1, A2A, A2B, and A3 (Fredholm et al., 2011). With a short physiological half-life (<10 s), adenosine is released to signal locally following conditions such as tissue trauma and pain (Fredholm et al., 2011). While adenosine is reported to provide potent and long-lasting pain suppression in both preclinical animal models and human subject studies, targeting this endogenous pathway for pain management has not yet been achieved (Zylka, 2011). The analgesic effect of adenosine has been attributed to A1AR and A2AAR activation (Zylka, 2011); however, these conclusions were made without examining the contribution of other receptor subtypes. As a consequence, a decade of preclinical and clinical development efforts to treat chronic pain have focused on A1AR/A2AAR agonists, but have failed to yield a viable therapeutic approach due to several undesirable actions, specifically A1AR/A2AAR-mediated cardiovascular side effects (Zylka, 2011; Boison, 2013). Discovery of a way to safely utilize the endogenous adenosine system for analgesia could offer meaningful relief for pain sufferers. A3AR is highly expressed in many inflammatory cells including glial cells (Abbracchio et al., 1997) and found in peripheral sensory neurons (Ru et al., 2011) and neurons in various areas of the CNS (Lopes et al., 2003). Endogenous adenosine signalling through A3AR has been demonstrated to be neuroprotective (Boison et al., 2010; Fishman et al., 2012); however, it is unknown whether this endogenous pathway affects nociceptive processing. In contrast to the limited therapeutic utility of A1AR and A2AAR agonists, the A3AR agonist IB-MECA [N6-(3-iodobenzyl)-adenosine-5′-N-methyluronamide] and its chlorinated counterpart Cl-IB-MECA have advanced to clinical trials for non-pain indications and show a good safety profile (Fishman et al., 2012). Very little is known about the roles of the A3AR in pain and the mechanisms and sites of action of A3AR agonists. One report has examined the effects of IB-MECA in the formalin test in mice (Yoon et al., 2005); intrathecal delivery of IB-MECA attenuated the inflammatory component, Phase 2 but not Phase 1. We recently reported that IB-MECA is also effective in models of neuropathic pain induced by diverse chemotherapeutic agents (paclitaxel, oxaliplatin and bortezomib) and constriction of the sciatic nerve (Chen et al., 2012). The effects of IB-MECA were corroborated with the highly selective A3AR agonist MRS5698 (Tosh et al., 2012) in a model of oxaliplatin-induced neuropathic pain (Janes et al., 2014). In this study, we extend our previous findings and demonstrate for the first time that activation of A3AR at spinal and supraspinal sites is key in the anti-nociceptive effects mediated by endogenous adenosine and provide pharmacological evidence that identifies a selective A3AR agonist as a potent non-narcotic agent that produces persistent pain relief without altering the protective actions of acute physiological pain.
Materials and methods
Male Sprague Dawley rats, wild-type C57BL/6 mice, A3AR−/− mice (Salvatore et al., 2000), and female BALB/cfC3H mice were used for all experiments. A total of 301 rats and 63 mice were employed. Procedures for the maintenance and use of animals were in accordance with the International Association for the Study of Pain (Seattle, MD), NIH (Bethesda, MD) guidelines on laboratory animal welfare, United Kingdom Home Office in compliance with the UK Animals (Scientific Procedures) Act 1986, and conformed to the regulations of the SLU, AU, UCL and SUN Committees on Animal Research. All observations were performed with the operators blind to the identity of the test substances administered.
Persistent neuropathic pain models were performed as previously described including: chronic constriction injury (CCI) (Little et al., 2012), spared nerve injury (Decosterd and Woolf, 2000), spinal nerve ligation (Kim and Chung, 1992), chemotherapy-induced peripheral neuropathy (Chen et al., 2012), and cancer-induced bone pain (Lozano-Ondoua et al., 2013).
Behaviour for pain models
Mechano-allodynia was assessed using calibrated von Frey filaments. Neurological function and motor coordination were evaluated by Rotarod motor test. Normal nociception was assessed by tail flick and hot-plate latency tests. Spontaneous and affective aspects of spinal nerve ligation-induced pain were assessed using conditioned place preference as described elsewhere (King et al., 2009). In animals with cancer-induced bone pain, spontaneous flinching and guarding behaviours were monitored (Lozano-Ondoua et al., 2013).
Rostral ventromedial medulla (RVM) cannulation, intrathecal catheterization, and subsequent injections including bilateral RVM microinjections were performed as previously described (Little et al., 2012).
In vivo electrophysiology
Extracellular recordings of spinal cord wide dynamic range neurons (L4-L5 segmental levels) were conducted on post-spinal nerve ligation Days 14–18 in isoflurane-anaesthetized rats.
Significant differences were defined as a value of P < 0.05. Behavioural data and mechanical and thermal coding were analysed by mixed-model two-way repeated measures ANOVA with Bonferroni comparisons. Electrophysiology data were analysed using a paired Student’s t-test. Sphericity was tested with Mauchly’s test, Greenhouse-Geisser corrections were used where required. Conditioned place preference difference scores were analysed using the paired t-test. Statistical analyses were performed using SPSS v21 (IBM) and Graphpad Prism v6.02.
Details of all procedures and reagents are described fully in the Supplementary material.
The adenosine kinase inhibitor ABT-702 reverses neuropathic pain
We first explored the contribution of A3AR signalling to anti-nociception during persistent pain using a rodent model of CCI (Bennett and Xie, 1988). Adenosine kinase is a key intracellular enzyme regulating intra- and extracellular concentrations of adenosine and its inhibition effectively potentiates extracellular adenosine concentration and signalling (Boison, 2013). Systemic administration of the selective non-nucleoside adenosine kinase inhibitor ABT-702 (Jarvis et al., 2000) at peak CCI-induced pain (post-injury Day 7) reversed established hypersensitivity to mechanical stimuli (mechano-allodynia) as measured by changes in paw withdrawal threshold (g) (Fig. 1A). The anti-allodynic effects of ABT-702 were partially attenuated by pretreatment or post-treatment with the selective A3AR antagonist MRS1523 (Li et al., 1998) and in A3AR−/− mice (Salvatore et al., 2000) with CCI (Fig. 1A–C). MRS1523 alone had no effect on paw withdrawal threshold. Partial attenuation is consistent with the reported contribution of A1AR to the effects of ABT-702 (Kowaluk et al., 2000). Additionally, the long-lasting effects of ABT-702 also agree with previous reports where its anti-allodynic action in a neuropathic pain model was sustained up to 11 h (Kowaluk et al., 2000). ABT-702 did not alter paw withdrawal thresholds in unaffected contralateral paws (Supplementary Fig. 1A). To examine the importance of A3AR signalling in persistent neuropathic pain of a different aetiology, we tested the effects of ABT-702 in chemotherapy (i.e. paclitaxel)-induced peripheral neuropathy (Chen et al., 2012), a common dose-limiting toxicity and cause of dose reduction for many chemotherapeutic agents (Farquhar-Smith, 2011). ABT-702 reversed peak (Day 25) chemotherapy-induced mechano-allodynia and mechano-hyperalgesia in an A3AR-dependent manner as anti-nociception was partially blocked or reversed by MRS1523 (Supplementary Fig. 1B–E).
A3AR activation by MRS5698 reverses persistent neuropathic pain without tolerance
We next investigated the potential utility of a highly selective, orally bioavailable A3AR agonist MRS5698 to modulate persistent pain; MRS5698 (Fig. 2A inset) has a high affinity for A3AR (≤3 nM) and excellent selectivity (≥104-fold over human, rat, and mouse A1AR or A2AAR) (Tosh et al., 2012). Subcutaneous MRS5698 administration during peak CCI-induced neuropathic pain in rats reversed mechano-allodynia in a dose-dependent manner (ED50 =0.35 mg/kg or 0.6 µmol/kg at 1 h; Fig. 2A) comparable to morphine at peak effect (ED50 = 0.5 mg/kg or 1.8 µmol/kg) and to IB-MECA (ED50 = 0.2 mg/kg or 0.4 µmol/kg; n = 6) with a fast onset of action (<30 min) and full efficacy within 1 h post-dosing. MRS5698 lacked effect on paw withdrawal threshold in contralateral paws (Supplementary Fig. 2A). Further characterization of MRS5698 revealed a wide therapeutic index, inferred from a lack of sedation or motor deficits in animals tested on a Rotarod at 30 mg/kg (i.e. 100× the anti-allodynic ED50; Supplementary Fig. 2B) and the non-lethal toleration of oral doses up to 150 mg/kg in rats (i.e. 400× the anti-allodynic ED50). MRS5698 was effective when given by multiple systemic routes of administration (i.e. intraperitoneal, subcutaneous, and intravenous) and protective in other well-established rat neuropathic pain models including spared nerve injury (Decosterd and Woolf, 2000) and spinal nerve ligation (Kim and Chung, 1992). Indeed, therapeutic administration of MRS5698 (1 mg/kg) reversed maximal mechano-allodynia within 1 h by 82 ± 15% (mean ± SD, n = 5) and by 98 ± 3.3% (n = 7) in spared nerve injury and spinal nerve ligation, respectively. MRS5698 was similarly efficacious in a more complex model of persistent pain with multiple aetiologies: cancer-induced bone pain (Lozano-Ondoua et al., 2013). Noteworthy, MRS5698 administered on Day 10 post-cancer-induced bone pain induction provided dose-dependent relief of spontaneous pain behaviours (flinching, guarding) (ED50 = 0.1 mg/kg or 0.2 µmol/kg at 1 h; Fig. 2B). As A3AR agonists are in clinical trials as anticancer agents (Fishman et al., 2012), the potential dual pharmacological properties (anticancer effects and pain-relieving properties) of an A3AR agonist may offer a significant therapeutic advantage.
The specificity of MRS5698 at A3AR was confirmed with pharmacological and genetic studies in the CCI model: its anti-nociceptive effects were blocked by MRS1523 and lost in A3AR−/− mice (Fig. 2C and D). MRS5698 anti-nociception was independent of endogenous opioid and cannabinoid pathways. Naloxone (non-selective µ-opioid antagonist) (Jurna, 1988), rimonabant (cannabinoid receptor 1 antagonist) (Rinaldi-Carmona et al., 1995), or SR144528 (cannabinoid receptor 2 antagonist) (Rinaldi-Carmona et al., 1998) pretreatment did not attenuate the effects of MRS5698 (Fig. 2E). Repeated injections of MRS5698 in rats did not lead to tolerance or a loss of efficacy since consecutive daily injections of MRS5698 on Days 8–15 reversed mechano-allodynia as effectively as on Day 7 (Fig. 2F). Continuous anti-nociception was observed with a 6-day subcutaneous mini-pump infusion of MRS5698 started on Day 7 post-CCI; acute exposure to MRS1523 reduced these effects, confirming A3AR signalling (Fig. 2G). In contrast, morphine lost its anti-nociceptive effects when given over a similar time course (Fig. 2F).
MRS5698 does not alter normal nociception
MRS5698 tested at the highest effective dose had no effect in tests that measure the acute thermal nociceptive component of physiological pain: tail flick and hot-plate (Fig. 2H and I).
MRS5698 produces conditioned place preference in nerve-injured rats
In addition to evoked pain behaviours, patients with neuropathic conditions commonly experience ongoing (spontaneous) pain that corresponds to the affective/motivational aspects of neuropathic pain. Relief of ongoing pain behaviours can engage the mesolimbic reward circuit (King et al., 2009; Navratilova et al., 2012) and be unmasked in animals using conditioned place preference. During ongoing pain, pairing an anti-nociceptive treatment that is not inherently rewarding, such as peripheral nerve block, with a context (chamber) can elicit conditioned place preference (King et al., 2009). Ideally, a pain therapeutic offers pain relief without abuse potential from inherent reward. Preconditioning baseline for total time spent in the chambers was not significantly different between vehicle and MRS5698 paired chambers for all animals (Supplementary Fig. 2C). We found that systemic MRS5698 (1 mg/kg) was associated with enhanced chamber preference at Day 14 post-spinal nerve ligation [difference score in seconds (i.e. post-conditioning minus preconditioning): 132.2 ± 42.8 s], but not sham surgery (difference score: −24.6 ± 46.7 s) (Fig. 2J). These results suggest that A3AR activation reverses spontaneous pain behaviours without inherent reward.
Spinal and supraspinal A3AR contribute to the reversal of mechano-allodynia
To explore sites and neurobiological mechanisms underlying the anti-nociceptive effects of A3AR activation, we examined A3AR in CNS regions related to nociceptive processing. We found A3AR/Adora3 mRNA transcript and protein in the spinal cord and RVM (Fig. 3A and B), key regions for nociceptive processing and modulation (Ossipov et al., 2010). A3AR/Adora3 mRNA was expressed in the lower lumbar (L4–6) spinal cord (mean Ct = 27.8 ± 0.3; n = 6) and micropunched RVM (mean Ct = 28.8 ± 0.2; n = 6) tissues (Fig. 3A); for comparison, Hprt1 was used as an endogenous control gene in both the spinal cord and RVM (mean Ct = 20.8 ± 0.4 and mean Ct = 21.5 ± 0.2, respectively). Similarly, western blot revealed A3AR/ADORA3 protein was expressed in these same tissues (Fig. 3B). Consistent with these A3AR expression patterns, spinal or bilateral intra-RVM microinjections of MRS1523 transiently abrogated ABT-702 anti-nociception, suggesting A3AR activation by endogenous adenosine at spinal and supraspinal sites (Fig. 3C). Furthermore, intrathecal or intra-RVM administration of MRS1523 blocked the ability of systemically administered MRS5698 to reverse mechano-allodynia (Fig. 4A and B). These findings confirm A3AR expression in pain-related CNS areas and demonstrate the functional relevance of A3AR to adenosine-mediated anti-nociception.
During neuropathic pain, spinal nociceptive processing is enhanced by the increased spontaneous activity of wide dynamic range pain projection neurons, which are modulated by both spinal and supraspinal (e.g. RVM) mechanisms (Ossipov et al., 2010). Thus, the integrated effect of systemic MRS5698 was tested to determine if A3AR activation inhibits nociceptive processing by reducing the responsiveness of spinal wide dynamic range neurons. As previously shown, baseline neuronal responses to a variety of stimuli were similar in sham and neuropathic animals since the model involves a loss of two-thirds of the afferent input to the spinal cord; yet the wide dynamic range neurons in neuropathic animals maintain higher than predicted levels of activity (Chapman et al., 1998). Subcutaneous MRS5698 administered to nerve-injured rats during maximal mechano-allodynia significantly reduced evoked wide dynamic range neuron responses to non-noxious and noxious mechanical, thermal, and electrical stimulation with peak effects at 1 h post-dosing (Fig. 4C), but had no effect in sham-operated rats (Supplementary Fig. 2D and E), further supporting that A3AR activation produces state-dependent anti-nociception. The effects were clear and prolonged and tended to be larger for the mechanical stimuli (Fig. 4C). The drug effects were apparent at 30 min and maximal at 1 h. There was an incomplete recovery by 3 h. For comparable levels of neuronal firing, MRS5698 reduced the low intensity thermal response (42°C) by 65% compared to 84% for the 8 g mechanical stimulus. The corresponding values for the medium stimuli were 48% and 66%, and 32% and 54% for the highest intensities. Even with the highest mechanical and thermal stimuli the neuronal responses were still significantly reduced. Next, we evaluated the contribution of the RVM to the anti-allodynic effects of A3AR activation. The RVM is a primary source of descending inhibition of spinal nociception by engaging serotonergic and noradrenergic bulbospinal circuits (Ossipov et al., 2010); therefore, we investigated whether A3AR activation in the RVM employs these circuits. Intra-RVM MRS5698 dose-dependently reversed CCI mechano-allodynia in an A3AR-dependent manner as intra-RVM pretreatment of MRS1523 prevented anti-nociception (Fig. 4D). Intrathecal delivery of the serotonin receptor antagonist, methysergide (Hoyer et al., 1994) or the α2 noradrenergic receptor antagonist, yohimbine (Goldberg and Robertson, 1983) (Fig. 4E), also attenuated MRS5698’s anti-nociceptive effects; these data imply that A3AR activation in the RVM engages bulbospinal inhibitory circuits to suppress spinal nociception.
Despite extensive research efforts, chronic pain remains a large unmet medical need and novel therapies are required as alternative options when the standards of care are not sufficient or effective. Our results substantially extend our previous findings exploring A3AR in pain (Chen et al., 2012) by identifying an endogenous analgesic A3AR pathway within key regions of the CNS that is distinguished by its potent efficacy and state-dependent nature, functioning to suppress only pathological pain without altering the normal pain threshold or activating reward centres in normal rats. Cardiovascular side effects are the major limitation of therapeutic approaches that enhance adenosine signalling (Fredholm et al., 2011; Zylka, 2011; Boison, 2013); however, A3AR agonists in clinical trials, IB-MECA and Cl-IBMECA, have no reported serious side effects (Fishman et al., 2012). Yet, while these prototypical A3AR agonists were anti-nociceptive in our preclinical studies (Chen et al., 2012), there may be therapeutic advantages using highly selective A3AR agonists such as MRS5698. These are anticipated to potentially allow aggressive dose escalation in difficult or complicated clinical settings; their high degree of selectivity should confer an advantage as spill over effect on A1AR and A2AAR is significantly reduced. It has long been appreciated that harnessing the potent analgesic effects of adenosine could provide a breakthrough step towards an effective treatment for chronic pain. Our findings suggest that this goal may be achieved by focusing future work on the A3AR pathway as its activation provides robust anti-nociception across several models of neuropathic pain, and in the more complex pain state of cancer-induced bone pain, which includes inflammatory and neuropathic features.
We are grateful to Dr Gary Bennett for his invaluable advice throughout the course of these studies and for critically reviewing our work.
This work was supported by grants from the National Cancer Institute (RO1CA169519), NIH predoctoral fellowship (5T32GM008306), NIDDK Intramural Research Program, FIRB “Futuro in Ricerca” MIUR, Italy (RBFR126IGO_001) and with additional support from the Saint Louis Cancer Centre.
Supplementary material is available at Brain online.
adenosine A3 receptor, now known as ADORA3
chronic constriction injury
rostral ventromedial medulla