Identification of A₃ adenosine receptor agonists as novel non‐narcotic analgesics

Abstract

Chronic pain negatively impacts the quality of life in a variety of patient populations. The current therapeutic repertoire is inadequate in managing patient pain and warrants the development of new therapeutics. Adenosine and its four cognate receptors (A₁, A_2A, A_2B and A₃) have important roles in physiological and pathophysiological states, including chronic pain. Preclinical and clinical studies have revealed that while adenosine and agonists of the A₁ and A_2A receptors have antinociceptive properties, their therapeutic utility is limited by adverse cardiovascular side effects. In contrast, our understanding of the A₃ receptor is only in its infancy, but exciting preclinical observations of A₃ receptor antinociception, which have been bolstered by clinical trials of A₃ receptor agonists in other disease states, suggest pain relief without cardiovascular side effects and with sufficient tolerability. Our goal herein is to briefly discuss adenosine and its receptors in the context of pathological pain and to consider the current data regarding A₃ receptor‐mediated antinociception. We will highlight recent findings regarding the impact of the A₃ receptor on pain pathways and examine the current state of selective A₃ receptor agonists used for these studies. The adenosine‐to‐A₃ receptor pathway represents an important endogenous system that can be targeted to provide safe, effective pain relief from chronic pain.

Abbreviations

ADA

adenosine deaminase

ADK

adenosine kinase

CCI

chronic constriction injury

CIPN

chemotherapy‐induced peripheral neuropathy

CPP

conditioned place preference

ENT

equilibrative nucleoside transporter

PN

peroxynitrite

RVM

rostral ventromedial medulla

SO

superoxide

TLR4

toll‐like receptor 4

Tables of Links

TARGETS
GPCRs a	Enzymes c
A1 receptor	ADA
A2A receptor	Adenylyl cyclase (AC)
A2B receptor	ADK
A3 receptor	Ecto‐5’‐nucleotidase
Catalytic receptors b	ERK1
TLR4	ERK2
Transporters d	GAD65
CNTs	p38 MAPK
ENT1	PKA
ENT2	S‐adenosylhomocysteine hydrolase
GAT1	TrkB
GLT1 (EAAT1)
KCC2

LIGANDS
ABT‐702	IL‐1β
Adenosine	IL‐10
ADP	Inosine
AMP	MRS1191
ATP	MRS1220
BDNF	MRS1523
cAMP	MRS5698
CCL2	Nitric oxide (NO)
GABA	NMDA
IB‐MECA	TNF‐α

These Tables list key protein targets and ligands in this article which are hyperlinked to corresponding entries in http://www.guidetopharmacology.org, the common portal for data from the IUPHAR/BPS Guide to PHARMACOLOGY (Pawson et al., 2014) and are permanently archived in the Concise Guide to PHARMACOLOGY 2015/16 (^a,b,c,dAlexander et al., 2015a, 2015b, 2015c, 2015d).

Introduction

Chronic pain afflicts an estimated 10% of the world's adult population (Goldberg and McGee, 2011). The current therapeutic approaches for chronic pain include but are not limited to the use of nonsteroidal anti‐inflammatory drugs (NSAIDs), antidepressants, anticonvulsants and opioid pain relievers; however, these strategies are frequently either inadequate or are associated with side effects that reduce quality of life or result in the discontinuation of therapy (Goldberg and McGee, 2011; Pizzo and Clark, 2012). The search for new therapeutic targets is therefore of great importance. Adenosine and two of its associated adenosine receptor subtypes, the A₁ and the A_2A receptor, have been investigated in the field of pain with varying degrees of success; however, these agents lack a useful therapeutic index due to cardiovascular side effects. In response, the focus of research has turned to the previously overlooked A₃ receptor, which displays both preclinical antinociceptive properties (Yoon et al., 2004; Janes et al., 2015, 2014b; Ford et al., 2015; Little et al., 2015) and, in trials for non‐pain conditions including psoriasis, hepatitis, rheumatoid arthritis, and glaucoma, offers a therapeutic index and tolerability that would be suitable for the treatment of chronic pain (Fishman et al., 2012). The aim of this review is to summarize the existing literature on adenosine and its receptors in the context of pain with a particular emphasis on the A₃ receptor and its prospect as a novel solution to the problem of chronic pain management.

Adenosine production and metabolism

Adenosine is an endogenous purine nucleoside that regulates a number of physiological processes and is an important neuromodulator in the CNS (Fredholm et al., 2011). Concentrations of adenosine are generated by almost all cell types (Zimmermann, 2000), and accordingly, the physiological generation and neurobiology of adenosine has been thoroughly reviewed elsewhere (Hu et al., 2014). Here, we will provide a brief, contextual overview of adenosine homeostasis in nervous tissues.

In the context of pain, the most notable function of adenosine in the CNS is its role as a neuromodulator for neurotransmitter systems including glutamate, GABA, ACh and dopamine. In these systems, adenosine limits the extent of neuroexcitability and also regulates neuroplasticity (Sebastiao and Ribeiro, 1996). However, CNS adenosine is not restricted to neuronal synapses nor is it a directional signalling molecule., Adenosine plays a regulatory role in glial activation state and function, such that adenosine can be said to impact the entire synaptic unit (Cunha, 2008; Dias et al., 2013).

Concentrations of adenosine in CNS tissue are reported to persist at a basal level of 25–250 nM (Dunwiddie and Masino, 2001). Adenosine is generated as a metabolic intermediate in both the intracellular and extracellular spaces and passively exchanged along its concentration gradient via the ubiquitously expressed equilibrative nucleoside transporters (ENTs), ENT1 and ENT2 (Brundege and Dunwiddie, 1998; Peng et al., 2005) or through concentrative nucleoside transporters (CNTs) along the concentration gradient of sodium (Bonan, 2012; Choi and Berdis, 2012). The extracellular function of adenosine is regulated by (i) changes in the ratio of intracellular: extracellular adenosine generation and consequently its driving gradient across passive transport systems and (ii) the local expression profile of adenosine receptors mediating its response (Deussen et al., 1999; Zimmermann, 2000). Together, these two elements are modified in the pathological cellular environment in order to facilitate the action of adenosine as an anti‐inflammatory, inhibitory neuromodulator.

Intracellular adenosine generation in the CNS is predominantly a resultant from the dephosphorylation of AMP by soluble 5′‐nucleotidases (Latini and Pedata, 2001). Unlike other physiological tissues, adenosine generated from S‐adenosylhomocysteine hydrolysis does not significantly contribute to adenosine concentrations in the CNS. In the extracellular space, ATP is released as a co‐transmitter or in response to cellular insult (e.g. inflammation, cellular stress or excitotoxicity) (Ballarin et al., 1991; Engler, 1991; Latini and Pedata, 2001) and can be dephosphorylated by ectonucleoside triphosphate diphosphohydrolases (CD39 family) and either ecto‐5′‐nucleotidase (CD73) (Robson et al., 2006; Bonan, 2012) or by tissue‐nonspecific alkaline phosphatase (Sebastian‐Serrano et al., 2015) to form extracellular adenosine (Figure 1). Adenosine generation is therefore tightly coupled to the availability of ATP and ADP in the intracellular and extracellular spaces.

Figure 1.

Adenosine synthesis and metabolism. ATP can be released from various cell types in response to cell excitation or insult. ATP can be deophosphorylated in sequence to form ADP, AMP and finally adenosine. In the extracellular space, ectonucleotidases (CD39 and CD73) facilitate formation of adenosine. Adenosine can act on its cognate receptors (A₁, A_2A, A_2B and A₃) or be removed from the extracellular space by metabolic enzymes (adenosine deaminase, ADA) or by transport back into the cell via equilibrative nucleoside transporters (ENTs) or concentrative nucleoside transporters (CNTs). In the intracellular space, adenosine can be converted to AMP (by adenosine kinase, ADK) which in turn is catalysed to ADP and then ATP. Intracellular adenosine can also be generated from AMP by 5′‐nucleotidase.

A second factor limiting the ratio of intracellular: extracellular adenosine is the metabolic inactivation of adenosine through adenosine kinase (ADK) phosphorylation (Spychala et al., 1996) or, to a lesser extent, the activity of adenosine deaminase (ADA) (Blackburn and Kellems, 1996). The function of these enzymes limits the physiological half‐life of adenosine to <1 s (Moser et al., 1989). Inhibition of ADK is the most effective strategy for increasing the extracellular concentration of adenosine and occurs by potentiating intracellular concentrations of adenosine and supporting an outward driving gradient through passive ENTs (Keil and DeLander, 1992; Zhang et al., 1993). Pharmacological blockade of ADK activity in the CNS results in adenosine‐mediated inhibition of spinal nociceptive transmission via the ENT‐dependent release of adenosine (Otsuguro et al., 2015), and inhibitors of ADK are efficacious in rodent experimental neuropathic pain models (Kowaluk et al., 2000; McGaraughty et al., 2005). It is important to note that ADK expression shifts from neurons to astrocytes during postnatal development, and accordingly, astrocytes are central to this aspect of adenosine homeostasis (Studer et al., 2006).

Adenosine receptors

As mentioned previously, the abundance of adenosine receptor subtypes is the second critical factor mediating the effects of extracellular adenosine, and much remains to be understood regarding the dynamics of adenosine receptor expression during cellular insult and pathological nociception. The extracellular actions of adenosine are mediated by its four cognate GPCRs: the A₁, A_2A, A_2B and A₃ receptor.

In the CNS, the A₁ receptor is highly expressed both pre‐synaptically and post‐synaptically on neurons as well as on glia in the brain (cortex, cerebellum and hippocampus) and specifically in the superficial laminae of the spinal cord dorsal horn (Gessi et al., 2011). It is known that glial expression of the A₁ receptor can be depressed in multiple sclerosis patients (Johnston et al., 2001), and expression of the A₁ receptor in the lumbar dorsal horn is decreased in a model of post‐operative pain. However, expression of the A₁ receptor is increased following sciatic nerve constriction (Yamaoka et al., 2013). These data suggest that the A₁ receptor is responsive to neuroinflammatory and nociceptive pathology, and that A₁ receptor antinociception may involve differential expression of the receptor post‐injury. However, in the context of off‐target effects, the presence of the A₁ receptor in cardiovascular tissue—particularly the atrioventricular node—is responsible for high‐grade atrioventricular block mediated by A₁ receptor agonists (Kiesman et al., 2009) and represents an unfortunate hurdle to the therapeutic exploitation of the A₁ receptor

Expression of the A_2A receptor in the brain is most notable in the striatum on post‐synaptic neurons, to a lesser extent pre‐synaptically in areas of the hippocampus and cerebral cortex and on glia (Svenningsson et al., 1997; Rebola et al., 2005). A_2A receptor expression is known to increase in response to insults such as hypoxia, spinal cord injury, streptozotocin‐induced diabetes and in other circumstances such as chronic behavioural stress and ageing (Janes et al., 2014b). Of relevance to chronic pain, pro‐inflammatory mediators including IL‐1β and TNF‐α are also known to enhance expression of the A_2A receptor in monocytes (Morello et al., 2006) such as microglia. Lastly, vasodilator A_2A receptors expressed on the epithelium of coronary blood vessels is of note with respect to the cardiovascular side effects of A_2A‐specific agents (Jacobson and Gao, 2006; Gao and Jacobson, 2007; Fredholm et al., 2011). The lower‐affinity A_2B receptor is also expressed in the CNS—particularly on immune‐related cells—and in the cardiovascular system, but because of the low‐expression profile and micromolar affinity of adenosine for the A_2B receptor, we will not extensively discuss the A_2B receptor in this review (Fredholm et al., 2000; Aherne et al., 2011).

For many years, the A₃ receptor was largely overlooked in CNS tissue due to a reported low profile of expression; however, it is now known that the A₃ receptor can be found in high levels on many immune cell types, including glial cells (Abbracchio et al., 1997; Poulsen and Quinn, 1998; Ochaion et al., 2009), as well as on both peripheral (Ru et al., 2011) and central neurons (Jacobson et al., 1993; Lopes et al., 2003; Giannaccini et al., 2008; Zhang et al., 2010) in the brain and spinal cord (Borea et al., 2015; Haeusler et al., 2015). Both mRNA and protein for the A₃ receptor have been documented in the lumbar spinal cord and in supraspinal areas including the rostral ventromedial medulla (RVM): indeed, A₃ receptors at the level of the peripheral afferent, the spinal cord and the RVM are functionally relevant in pain as selective A₃ receptor agonists administered via i.d., intrathecal (i.t.) or intra‐RVM routes dose‐dependently attenuate neuropathic pain behaviours (Little et al., 2015). It can therefore be concluded that the A₃ receptor is functionally expressed at multiple levels of pain processing (Figure 2).

Figure 2.

A₃ receptor's multiple sites of action. Studies employing selective A₃ receptor agonists have uncovered the efficacy of A₃ receptor activation at several sites important for pain processing. In the ascending pathway (red), a stimulus is conducted from the periphery to the spinal cord where it is then transported to the brain to be interpreted. The brain is able to modulate these events via the descending pathway (green). The i.d. administration of IB‐MECA (3–60 nmol; Salvemini unpublished results) dose‐dependently attenuated the reduction in paw withdrawal thresholds (PWTs) associated with chronic constriction injury (CCI). These results were extended by MRS5698 administration via surgically implanted catheters placed into the intrathecal (i.t.; 3–60 nmol) space or rostral ventromedial medulla (RVM; 0.3–3 nmol). Data are mean ± SD for n = 5 animals per group and analysed by two‐way ANOVA with Bonferroni comparisons. #P < 0.05 vs. D0; *P < 0.05 vs. D7. The i.t. data are reprinted with permission from J Neuro (Ford et al., 2015) . Intra‐RVM data are reprinted with permission from Brain (Little et al., 2015).

The A₃ receptor is unique among the adenosine receptors in that species disparities exist for both the receptor structure and distribution; the highest expression of the A₃ receptor in rats exists in testis and mast cells, whereas the highest levels in human are found in the liver and lung (Borea et al., 2015). Importantly, while no direct evidence has been found for the existence of the A₃ receptor in cardiomyocytes, several studies have demonstrated A₃ receptor‐mediated cardioprotection from ischaemic injury (Tracey et al., 1997; Thourani et al., 1999a, 1999b; Cross et al., 2002; Harrison et al., 2002; Headrick and Peart, 2005) and doxorubicin‐induced cardiotoxicity (Shneyvays et al., 1998; Shneyvays et al., 2001). Additionally, high A₃ receptor expression has been documented in the human coronary and carotid arteries (Hinze et al., 2012; Grandoch et al., 2013).

Given the distribution of adenosine receptors, it is also important to note the coupling mechanisms of these receptors. The A_2A and A_2B subtypes are G_s‐coupled GPCRs which stimulate AC and produce elevations in intracellular cAMP and associated signalling cascades (Fredholm et al., 2001). A₁ and A₃ receptors differ both in their coupling to the opposing G_i cascade which inhibits AC and in the ability to respond to inosine as a partial agonist (Boison et al., 2010; Fredholm et al., 2011). It is important to note that while these receptors have, at face value, opposing intracellular effects, the expression of these receptors on differing cell types can be responsible for commonalities in their activation in the CNS.

Adenosine and pain

Adenosine plays an integral role in CNS processing of pain through its regulation of excitatory neurotransmission, persistent neuronal signalling and regulation of glial activation and proliferation. Accordingly, the ability of adenosine and its analogues to inhibit pain behaviour has been documented in models of various aetiologies, including neuropathic pain associated with spinal cord injury, spinal nerve ligation and in the mustard oil, formalin and carrageenan pain models (Dickenson et al., 2000). Indeed, in the clinical setting, i.t. adenosine administration has been shown to provide sustained relief of chronic neuropathic pain lasting for several hours up to months in some patients (Hayashida et al., 2005). Adenosine therapy has also been employed for the prevention of post‐operative pain with mixed results: prophylactic i.v. administration of adenosine prior to surgical procedures conferred persistent pain relief in several studies (Hayashida et al., 2005; Gan and Habib, 2007), while another similar clinical trial did not observe a prophylactic effect (Habib et al., 2008). Unfortunately, the i.v. administration of adenosine is associated with undesirable cardiac side effects (Zylka, 2011) that limit its utility and route of administration (e.g. i.t.) in patients. Thus, isolating the antinociceptive qualities of adenosine from its cardiovascular side effects by evaluating receptor involvement has become an important focus for the development of adenosine‐based therapeutics in pain.

A₁ and A_2A receptors and pain

It is not our intent to discuss in detail the A₁ and A_2A receptors, because these targets have been the topic of many excellent reviews (Sawynok, 1998; Fredholm et al., 2011; Zylka, 2011; Chen et al., 2013), but these receptors have played an important role in the understanding of adenosine antinociception. The antinociceptive properties of adenosine were long attributed to the activation of the A₁ and A_2A receptor subtypes (Zylka, 2011; Sawynok, 2013). Genetic deletion of the A₁ receptor produces thermal hypersensitivity and exacerbates neuropathic behavioural responses to cold and heat (Fedorova et al., 2003; Wu et al., 2005), and it is well documented in the literature that A₁ receptor activation leads to antinociception in a range of pain models. A₁ receptor activation produces beneficial outcomes in preclinical models of acute and chronic pain including nerve injury‐induced pain (Cui et al., 1997; Gong et al., 2010), peri‐operative pain (Gan and Habib, 2007), inflammatory pain (Sowa et al., 2010), central pain following spinal cord injury (Sjolund et al., 1998) and painful diabetic neuropathy (Vincenzi et al., 2014; Katz et al., 2015). Additionally, a spinally administered A₁ receptor agonist reduces non‐evoked spontaneous pain behaviours resulting from a surgical model of pain (Zahn et al., 2007). The preclinical robustness of A₁ receptor pain relief resulted in clinical trials for multiple A₁ receptor agonists and an A₁ receptor allosteric enhancer; however, these drug trials were discontinued due to limited efficacy, presumably driven by a low therapeutic index and dose limitation (Romagnoli et al., 2010; Gessi et al., 2011).

The findings regarding the A_2A receptor have been controversial. Mice lacking the A_2A receptor demonstrate depressed responses to acute pain stimuli as measured by the hot‐plate and tail‐flick tests (Ledent et al., 1997). Peripheral administration of an A_2A receptor agonist is associated with nociceptive behaviours (Taiwo and Levine, 1990). However, very low doses of A_2A receptor agonists administered spinally have been shown to promote the sustained reversal of nerve injury‐induced pain in rats for weeks after a single i.t. injection (Loram et al., 2009). In models of post‐surgical pain (Zahn et al., 2007) and inflammatory pain (Poon and Sawynok, 1998), i.t. administration of A_2A receptor agonists had only transient antinociceptive effects that were minimal‐to‐negligible. However, an i.c.v. injection of an A_2A‐targeted antibody with agonist‐like activity produces antinociceptive effects in naïve mice (By et al., 2011). At the clinical level, a phase II trial of an A_2A receptor agonist BVT‐115959 in the treatment of diabetic neuropathy was completed in 2008, but the trial has been since abandoned (Swedish Orphan Biovitrum, 2014). The differing observations of A_2A receptor agonists in pain highlight an apparent dichotomy of peripheral versus central A_2A receptors in pain processing.

For the past decade, a narrow therapeutic focus on the A₁ and A_2A receptor has failed to harness adenosine antinociception effectively and without cardiovascular side effects (Zylka, 2011; Boison, 2013). In response, we anticipate that a greater emphasis on the A₃ receptor is a better utilization of adenosine antinociception to provide safe, effective pain relief at the clinical level.

A₃ receptors and pain

Our understanding of A₃ receptor signalling in pain has evolved greatly since the human A₃ receptor was first cloned in 1993 (Salvatore et al., 1993). Early investigations reached conclusions that were not necessarily correct due to the use of A₃ receptor‐targeted compounds with poor specificity such as N ⁶‐benzyl–NECA (Sawynok et al., 1997; 1999) and the A₃ receptor^−/− mouse (Wu et al., 2002). A 1997 publication examining the contribution of the A₃ receptor in pain reported that s.c. administration of N ⁶‐benzyl–NECA into the rodent hindpaw produced dose‐dependent flinching behaviour that was blocked by administration of a histamine H₁ receptor antagonist or a 5‐HT₂ receptor antagonist, but not by antagonists of A₁ or A₂ receptors (Sawynok et al., 1997). These studies lead to the incorrect hypothesis that the A₃ receptor was likely to mediate the pro‐nociceptive, pro‐inflammatory effect of N ⁶‐benzyl–NECA via the induction of mast cell degranulation. A subsequent study clarified that N ⁶‐benzyl–NECA nociception was not influenced by an A₃ receptor antagonist (MRS1191) but was in fact abolished by blockade of the A_2B receptor, a subtype previously implicated in inflammation (Feoktistov and Biaggioni, 2011).

The first definitive characterizations of the A₃ receptor as antinociceptive resulted from the use of a more specific A₃ receptor agonist, IB‐MECA (N ⁶‐(3‐iodobenzyl)‐adenosine‐5′‐N‐methyluronamide): IB‐MECA is 50‐fold selective for the A₃ receptor over the A₁ or A_2A receptor, as compared with the 14‐fold selectivity of N ⁶‐benzyl–NECA (Gallo‐Rodriguez et al., 1994; Jacobson, 1998). In 2005, a study demonstrated that systemically administered IB‐MECA exerts a significant antinociceptive effect during the second phase of the formalin test without altering protective nociceptive responses (i.e. response to noxious thermal or mechanical stimuli) (Yoon et al., 2005). i.t. administration of an A₃ receptor antagonist (MRS1220) prevented the antinociceptive actions of adenosine in the second phase of the formalin test, supporting a role for spinal A₃ receptors in the effect of adenosine (Yoon et al., 2006).

Genetic work in the A₃ receptor knockout mice (A₃AR)^−/− mouse has followed a similar pattern of evolution. In 2002, a study of carrageenan‐induced peripheral inflammatory pain in the A₃AR^−/− mouse noted a mild increase in the development of thermal hyperalgesia as compared with wild‐type controls (Wu et al., 2002). There was no alteration in the protective (i.e. non‐pathological) nociceptive responses of A₃AR^−/− animals to noxious heat and mechanical stimuli, implicating the A₃ receptor in pathological rather than protective pain states. This report was partially contradicted by a later study that observed a decrease in the protective hot‐plate but not tail‐flick responses of A₃AR^−/− mice (Fedorova et al., 2003). Consequently, it was unclear whether the A₃ receptor affected normal protective nociception or was solely implicated in pathological pain states.

There were no studies published between 2006 and 2012 that examined the contribution of A₃ receptors in pain. In 2012, our laboratory revisited the A₃ receptor hypothesis using the agents IB‐MECA and Cl‐IB‐MECA in neuropathic pain (Chen et al., 2012; Little et al., 2015). Importantly, we have observed no impact of these agents and others on baseline nociceptive thresholds, and we have therefore concluded that A₃ receptor agents do not alter normal protective nociception. In neuropathic pain, both IB‐MECA and Cl‐IB‐MECA blocked the development of mechano‐allodynia following chronic constriction injury (CCI) in a manner prevented by an antagonist of the A₃ receptor but not antagonists of the A₁ or A_2A receptor (Chen et al., 2012). Low doses of IB‐MECA given in combination with morphine, gabapentin or amitriptyline increased the potency of these agents as analgesics (Chen et al., 2012). The antinociceptive effects of IB‐MECA and Cl‐IB‐MECA have since been corroborated by better‐selective A₃ agonists including MRS1898 [>100‐fold over A₁ or A_2A receptors (Gao et al., 2009)] and more recently MRS5698 [>10 000‐fold over A₁ or A_2A receptors (Tosh et al., 2012)] in rodent CCI, spared nerve injury and spinal nerve ligation models (Chen et al., 2012; Ford et al., 2015; Little et al., 2015). The specificity of a newer generation A₃ receptor agonists has been corroborated by the attenuation of MRS5698 antinociception in the A₃AR^−/− mouse, and alternatively in the presence of a specific A₃ receptor antagonist. (Little et al., 2015). In CCI, A₃ receptor agonists administered via i.d. (ipsilateral paw to nerve injury) injection (IB‐MECA, 3–60 nmol), i.t. cannula (MRS5698, 3–60 nmol) or RVM cannula (MRS5698, 0.3–3 nmol) dose‐dependently attenuate mechanical allodynia (Little et al., 2015). Antinociception conferred via systemic administration of the CNS‐permeant MRS5698 is attenuated with i.t. or intra‐RVM delivery of an A₃ receptor antagonist (Little et al., 2015). However, systemic administration of a peripherally restricted A₃ receptor agonist also reverses CCI‐induced peak mechanical allodynia, and the effect is not reversed by administration of an i.t. A₃ receptor antagonist (Paoletta et al., 2013). It can therefore be concluded that the A₃ receptor produces antinociceptive input at central and peripheral levels, which is an important characteristic of successful analgesic agents (e.g. opioids). Further studies are warranted to explore the relationship between peripheral and central A₃ receptors in pain.

A₃ receptor agonists have also been validated in a number of cancer‐related pain states. In models of chemotherapy‐induced peripheral neuropathy (CIPN), IB‐MECA (Chen et al., 2012; Janes et al., 2014b) and MRS5698 (Janes et al., 2015; Little et al., 2015) blocked the development of neuropathic pain without interfering with the antitumour effects (Chen et al., 2012). In a rat model of breast cancer bone metastasis, treatment with Cl‐IB‐MECA reduced tumour growth and the related bone pain (Varani et al., 2013), and MRS5698 had similar antinociceptive effects in a mouse model of breast cancer bone metastasis (Little et al., 2015). High expression of the receptor is detected on many malignant cell types, and A₃ receptor agonists have been shown to produce direct anticancer effects on their own and have been documented to enhance the actions of several widely used chemotherapeutics and attenuate the associated myelosuppression (Fishman et al., 2002; Fishman et al., 2009; 2012). Indeed, a successful phase I/II clinical trial of Cl‐IB‐MECA as an anticancer agent in hepatocellular carcinoma was recently completed by Can‐Fite BioPharma. Therefore, the use of A₃ receptor agonists may provide dual benefits in the treatment of a variety of cancer‐related pain states.

Finally, it is important to note that because the antinociceptive effects of IB‐MECA and other selective A₃ receptor agonists are not dependent upon endogenous opioid or endocannabinoid pathways (Ford et al., 2015; Little et al., 2015), studies have evaluated whether A₃‐specific antinociception lacks classical tolerance and inherent reward properties. In preclinical studies, A₃ receptor agonists are not subject to analgesic tolerance: in rats and mice, the effect of A₃ receptor agonists persisted following six repeated daily injections (compared with morphine, which demonstrated tolerance following repeated injections for 6 days) or when administered as a continuous infusion for 7 days (Little et al., 2015). These findings are interesting as repeated or continuous exposure to adenosine receptor agonists usually results in diminished responses and receptor down‐regulation known as ‘desensitization phenomenon,’ a response that is characteristic for all adenosine receptor subtypes (Klaasse et al., 2008). However, similar findings have been reported in animal models of autoimmune disorders and cancer, wherein chronically administered A₃ receptor agonists do not lose their anti‐inflammatory/anticancer effects in spite of A₃ receptor protein down‐regulation (Madi et al., 2003). It was demonstrated that down‐regulation of the receptor is associated with downstream inhibition of key regulatory proteins involved in inflammation/tumour growth, such that receptor down‐regulation in fact represents receptor functionality in these cases (Fishman et al., 2006). It is possible that a similar mechanism exists for the action of IB‐MECA and other A₃ receptor agonists in pain, but this hypothesis requires further investigation.

In order to evaluate the inherent reward and therefore abuse potential of A₃ receptor agonists, the conditioned place preference (CPP) method was combined with the induction of nerve injury pain in rats to evaluate spontaneous pain behaviours. Results from these studies indicate that A₃ receptor agonists such as MRS5698 produce CPP in nerve‐injured but not sham rats (unlike opioids and other drugs of abuse, which can elicit CPP from both naïve and injured animals), suggesting that A₃ receptor activation attenuates spontaneous pain behaviours without inherent reward (Little et al., 2015). Taken together with the observation that A₃ receptor agonists selectively modify pathological but not protective pain, it can be hypothesized that A₃ receptor agonists may circumvent the classical complications of tolerance and abuse potential associated with opioid therapy.

Mechanisms of A₃ receptor‐induced antinociception

Due to the recent emergence of the A₃ receptor as a valid target for pain relief, much remains to be explored regarding the specific mechanisms of action downstream of receptor activation. To date, it is known that in the CCI model of neuropathic pain, the effects of A₃ receptor agonists are mediated independently of the opioidergic and cannabinoid systems, but do act supraspinally to recruit the activation of 5‐hydroxytryptaminergic and noradrenergic bulbospinal circuits, and reduce the excitability of wide dynamic range spinal neurons (Little et al., 2015). Here, we will discuss the mechanisms by which A₃ receptor activation in other disease states alters processes involved in the development of central sensitization and pain, including protein kinase activity, glutamatergic neurotransmission, ion conductance and neuroinflammation. A summary of this discussion can be found in Figure 3.

Figure 3.

Potential mechanisms of A₃ receptor (A₃AR)‐mediated antinociception. Several pathways are known to be important in the establishment of pain states including impairment of GABAergic neurotransmission, enhanced neuroinflammation characterized by increased glial hyperactivation and TLR4 signalling, increased glutamatergic signalling and heightened production of nitroxidative species. The A₃ receptor has been shown to modulate these pathways through the use of selective agonists, potentially explaining A₃ receptor's antinociceptive actions.

The GABAergic system is an important component regulating correct nociceptive transmission. Following the release of GABA from interneurons within the CNS, the neurotransmitter GABA can act upon its receptors to dampen neuronal excitability and reduce nociceptive signalling. During pathological pain, dysfunction of the GABAergic interneuron system liberates excitatory neuronal signalling and ultimately results in a state of system hyperexcitability (Zeilhofer et al., 2012). An abnormality in GABA signalling can occur through a variety of mechanisms, including reduced GABA synthesis by the enzyme GAD65 (Stiller et al., 1996; Eaton et al., 1998), enhanced expression of the GABA reuptake transporter GAT‐1 (Eaton et al., 1998; Moore et al., 2002) and impairment of the K⁺‐Cl⁻ cotransporter KCC2, which maintains the anion gradient required for the inhibitory action of Cl⁻ through GABA_A channels (Coull et al., 2003; Price et al., 2005). It was recently demonstrated that the A₃ receptor agonist MRS5698 reverses CCI‐induced pain via a GABA‐mediated mechanism, wherein the CCI‐related dephosphorylation of GAD65 and GAT‐1 and the phosphorylation of KCC2 are ameliorated by treatment with an A₃ receptor agonist (Ford et al., 2015). These results indicate that restoration of the GABAergic inhibitory system contributes to the reversal of neuropathic pain following A₃ receptor activation. In addition to modifying phosphorylation of these enzymes, the A₃ receptor may also exert its effects on GABA‐mediated inhibition indirectly through the attenuation of brain‐derived neurotrophic factor (BDNF) signalling. Glial‐derived BDNF–TrkB signalling has been shown to reduce GABAergic signalling in the CCI model of neuropathic pain (Biggs et al., 2010; Ferrini and De, 2013; Smith, 2014). A₃ receptor activation is associated with the attenuation of astrocyte reactivity, neuroinflammatory response (Janes et al., 2015) and reactive microglial chemotaxis (Choi et al., 2011), such that A₃ receptor agonists may reduce BDNF associated with glial hyperactivation and free the GABAergic system to function properly.

A₃ receptor agonists have demonstrated neuroprotection in animal models that is potentially mediated through the induction of pro‐survival RhoA‐PLD signalling pathways (Jacobson, 1998; Fredholm et al., 2011). For example, A₃ receptor agonists prevent the decrease in PLD activity that occurs in response to prolonged reactive oxygen species exposure during apoptosis in cardiomyoctytes (Lee et al., 2001; Asemu et al., 2005). Accordingly, the protection of proper PLD function can then go on to increase the production of choline, which leads to activation of α7 nicotinic ACh receptors (Lee et al., 1993). This effect is known to be both neuroprotective and antinociceptive in chronic neuropathic pain (Feuerbach et al., 2009).

Many studies implicate spinal glia in the development and maintenance of chronic pain, and a variety of pain states can be prevented or attenuated with agents that disrupt the stimulation of glial cells (Milligan and Watkins, 2009; Gwak et al., 2012; Old et al., 2015). During an enhanced response state, glial cells can release pro‐inflammatory cytokines and nitroxidative species, which can then sensitize the neurons in the dorsal horn leading to pain (Cao and Zhang, 2008; Milligan and Watkins, 2009). These glial‐derived pro‐inflammatory mediators act not only on neurons but also on glial cells leading to an amplification loop that is potentially responsible for the long‐lasting hypersensitivity underpinning some chronic pain states (Bradesi et al., 2001). A₃ receptor agonists have been shown to impart their beneficial actions at least partly through modulation of spinal neuroinflammatory processes, as IB‐MECA treatment results in a reduction of hyperactive astrocytes and reduced production of pro‐inflammatory/neuroexcitatory cytokines in models of CIPN (Janes et al., 2014a; Janes et al., 2015). Interestingly, A₃ receptor activation also enhances formation of the anti‐inflammatory cytokine IL‐10 (Hasko et al., 1996; Janes et al., 2014a; Janes et al., 2015) and the production of glial‐derived neuroprotective substances including CCL2 (Wittendorp et al., 2004). Both in vitro and in vivo studies have revealed that A₃ receptors produce these effects by inhibiting the p38 MAPK and NF‐κB signalling pathways (Madi et al., 2007; Varani et al., 2010; Varani et al., 2011; Janes et al., 2014a). A better understanding of whether this mechanism occurs in A₃ receptor‐mediated antineuroinflammatory pain relief is critical to the development of an A₃ therapeutic strategy in pain.

Increased formation of nitroxidative species including superoxide, NO and their highly pro‐nociceptive reaction product peroxynitrite (Salvemini and Neumann, 2010) have been shown to play an important role in the development and maintenance of pain of several aetiologies including acute and chronic inflammation (Ndengele et al., 2008), orofacial pain (Yeo et al., 2008), opiate‐induced hyperalgesia and antinociceptive tolerance (Muscoli et al., 2007), nerve injury‐induced pain (Rausaria et al., 2011) and CIPN (Doyle et al., 2012; Janes et al., 2013). In a model of CIPN, the A₃ receptor agonist IB‐MECA attenuated the spinal activation of NADPH oxidase, a source of superoxide as a precursor to peroxynitrite formation (Poderoso et al., 1996; Janes et al., 2014a). In prostate cancer cells, IB‐MECA treatment has been shown to inhibit NADPH oxidase activation through inhibition of intracellular cAMP/PKA (Jajoo et al., 2009) and by reducing the expression of NADPH oxidase subunits through inhibition of ERK1/2 activity (Jajoo et al., 2009). As such, the downstream effects of the A₃ receptor on PKA activation and ERK phosphorylation could underlie the observed effect of IB‐MECA on NADPH oxidase in the CNS during pain.

Given the role of adenosine in limiting excitatory neurotransmission, it is unsurprising that A₃ receptor activation impacts glutamatergic signalling. A₃ receptor agonists protect against the neurotoxic P2X7 receptor‐mediated (Zhang et al., 2006) or the glutamate‐ and NMDA‐mediated rises in intracellular Ca^{2 +} and neuronal excitability in vitro (Zhang et al., 2010). Alterations in glutamatergic neurotransmission and increased neuronal excitability are widely observed in models of chronic pain (Hansson and Ronnback, 2004; Latremoliere and Woolf, 2009; Gwak et al., 2012). A₃ receptor activation is associated with the inhibition of the post‐translational nitration and inactivation of the glutamate transporter GLT‐1 and glutamate synthase (Janes et al., 2014a), together largely responsible for maintaining proper glutamatergic signalling and termination (Mao et al., 2002). In addition, A₃ receptor agonists in neurons inhibit signalling through presynaptic metabotropic glutamate receptors (normally involved in reducing neurotransmission at glutamatergic synapses) (Macek et al., 1998), which have been shown to be involved in the induction of several pain states (Fundytus, 2001).

Finally, activation of the innate immune receptor toll‐like receptor 4 (TLR4) expressed on glial cells has been implicated in the development of neuropathic pain (Watkins et al., 2009; Li et al., 2014). Stimulation of the A₃ receptor with IB‐MECA has been documented to decrease the TLR4‐induced release of pro‐inflammatory mediators including TNF and macrophage inflammatory protein‐1α as well as to increase the production of the anti‐inflammatory IL‐10 (Hasko et al., 1996; Sajjadi et al., 1996; Szabo et al., 1998). Suppression of pro‐inflammatory mediators following TLR stimulation is lost in A₃ receptor knockout mice, suggesting that the A₃ receptor has a critical role in suppressing TLR4 responses (Salvatore et al., 2000).

A₃ receptor‐specific tools for the study of pain

For the benefit of future investigations of the A₃ receptor, we have provided a table of preclinically useful adenosine and A₃AR modulators, their pharmacological characteristics (Table 1) and their structures (Figure 4). Adenosine itself is a native, nonselective adenosine receptor agonist, while its metabolite inosine, generated following the action of ADA, weakly activates the A₃ receptor (Gao et al., 2011). Preclinically successful A₃ receptor agonists (compounds 1–6) have been generated through substitutions at the C2, N⁶ and 5′ positions, most favorably including: N ⁶‐benzyl (compounds 1–6) or small alkyl (compound 6); C2‐arylethynyl (compounds 4–6) or aryltriazolyl substitutions. IB‐MECA (compound 1) and Cl‐IB‐MECA (compound 2) are widely used in pharmacological probes of nM affinity with CI‐IB‐MECA being more selective for the A₃ receptor (K _D 2.4 nM versus 1.5 nM respectively), and both of these agents have been employed in clinical trials, such that they represent clinically available avenues for targeting the A₃ receptor. Compounds 3–6 contain a conformationally constrained bicyclic (N‐methanocarba) ring in place of ribose, which adds to the A₃ receptor selectivity (Tosh et al., 2012; 2014; 2015), and these agents are specific for the A₃ receptor with a selectivity of 0.7–3.5 nM. MRS5481 (compound 4) is a peripherally restricted agonist that is highly A₃‐selective (K _D 1.9 nM) (Paoletta et al., 2013). MRS5698 (compound 5) is balanced in affinity at the human and mouse A₃ receptors (K _D 3 nM) with insignificant activity at A₁ and A_2A receptors at 10 μM. Compound 6 displays a long duration of action in vivo when administered p.o.

Table 1.

A₃ receptor‐selective agents for use in the study of A₃‐mediated antinociception

Compound	K i or K D, nM (or % inhibition at 10 μM)
	Human			Mouse			Rat
	A1	A2A	A3	A1	A2A	A3	A1	A2A	A3
Agonists
1 – IB‐MECA	700 ± 270	6200 ± 100	2.4 ± 0.5	5.9	~700	0.087	54	56	1.1
2 – Cl‐IB‐MECA	220 ± 20	5400 ± 2500	1.5 ± 0.2	35	~10 000	0.18	820	470	0.33
3 – MRS1898	136 ± 22	784 ± 97	1.51 ± 0.23				83.9 ± 10.3	1660 ± 260	0.17 ± 0.04
4 – MRS5841	16%	7%	1.90 ± 0.03	15%	1%	11.3 ± 1.9
5 – MRS5698	6%	41%	3.49 ± 1.84	16%	27%	3.08 ± 0.23
6 – MRS5980	6%	24%	0.70 ± 0.11	38%	7%	36.1 ± 4.7
Antagonists
7 – MRS1523	>10 000	3660 ± 930	18.9	8000	>10 000	731	15 600	2050	113
8 – MRS1191	>10 000	>10 000	31.4				40 100	>10 000	1850

Sources: Jacobson et al., 1992; Ge et al., 2006; Melman et al., 2008; Franchetti et al., 2009; Liang et al., 2010; Tosh et al., 2012; Paoletta et al., 2013; and unpublished data.

Figure 4.

Pharmacological agents useful for the study of A₃ receptor‐mediated antinociception. Preclinically useful selective A₃ receptor agonists (1–6), A₃ receptor antagonists (7–8) and adenosine modulators (9–12).

An alternative strategy for the study of the A₃ receptor is the use of compounds that enhance extracellular adenosine concentrations such as inhibitors of ADA (e.g. compound 9, pentostatin) and inhibitors of adenosine kinase (e.g. compound 10 5‐iodotubercidin and compound 11 ABT‐702) used in tandem with selective A₃ receptor antagonists to elucidate A₃‐mediated effects (Figure 4). As an ADK inhibitor, compound 11 decreases both chronic and acute pain with peripheral or central administration (Kowaluk et al., 2000), and compound 12 (TRR496) is a selective A₁ receptor allosteric enhancer that suppresses pain in a manner comparable and additive with morphine in formalin and writhing tests and has anti‐allodynic effects in the streptozotocin‐induced model of diabetic neuropathic pain (Vincenzi et al., 2014). An effective and moderately selective A₃ receptor antagonist for use in mouse and rat is compound 8 (MRS1523) (Li et al., 1998), although species differences in affinity should be taken into consideration when using selective adenosine receptor ligands, as heterocyclic antagonists of the A₃ receptor often have much higher affinity at the human than the murine A₃ receptor.

Concluding remarks

Although early studies investigating the A₃ receptor presumed a peripheral pro‐nociceptive role in pain—in large part due to the use of nonselective agonists—the development of selective pharmacological tools targeting the A₃ receptor has now uncovered the robust antinociceptive properties of A₃ receptor agonists in a variety of pathological pain states. Emerging evidence suggests that harnessing the endogenous antinociceptive A₃ receptor pathway yields effective pain relief without altering normal protective nociception and without producing reward effects associated with abuse potential. Noteworthy, the cardiovascular side effects marring the usefulness of adenosine‐targeted therapies is reduced in A₃ receptor‐mediated strategies, and A₃ receptor agonists do not display complications in on‐going phase II/III clinical trials for non‐pain conditions. We propose that A₃ receptor activation may be a safe and successful strategy for exploiting the potent analgesic actions of adenosine to provide a breakthrough non‐opioid treatment for patients suffering from chronic pain.

Author contributions

All authors contributed to writing and reviewing sections of the manuscript and approved the final version.

Conflict of interest

The authors declare no conflicts of interest.

Janes, K. , Symons‐Liguori, A. , Jacobson, K. A. , and Salvemini, D. (2016) Identification of A₃ adenosine receptor agonists as novel non‐narcotic analgesics. British Journal of Pharmacology, 173: 1253–1267. doi: 10.1111/bph.13446.