Strategies towards safer opioid analgesics - a review of old and upcoming targets

Abstract

Opioids continue to be of use for the treatment of pain. Most clinically used analgesics target the mu opioid receptor (MOR) whose activation results in adverse effects like respiratory depression, addiction, and abuse liability. Various approaches have been used by the field to separate receptor mediated analgesic actions from its adverse effects. These include biased agonism, opioids targeting multiple receptors, allosteric modulators, heteromers, and splice variants of the MOR. This review will focus on the current status of the field and some upcoming targets of interest which may lead to a safer next generation of analgesics.

Introduction

Chronic pain is defined as pain that persists beyond the normal healing time (>3 months), and affects over 100 million Americans. The total economic and social cost of chronic pain is estimated to be at least $600 billion. There are currently no universally effective treatments for chronic pain, particularly neuropathic pain, which is associated with poor health outcomes including anxiety, depression, poor sleep, inability to work, and increased rates of suicide. Neuropathic pain is caused by dysfunction or damage to the somatosensory nervous system that often results from stroke, diabetes, multiple sclerosis, spinal cord injury, or from peripheral nerve damage or dysfunction, and has a worldwide prevalence of 7–10% and is associated with a poor quality of life. The majority of chronic pain sufferers do not experience complete pain relief with first line treatments such as calcium channel inhibitors (e.g. gabapentin, pregabalin), antidepressants, and/or anticonvulsants, and patients are often unable to tolerate effective doses because of serious adverse effects. Second-line treatments include opioids that target the mu-opioidreceptor(MOR) such as morphine, tramadol, and fentanyl. However, they rapidly cause tolerance and have side-effects such as respiratory depression and addiction. Moreover, misuse of prescription opioids has led to an opioid abuse crisis with opioid overdoses killing over 93,000 people in the US in 2020 according to recent CDC data. This highlights the urgent unmet need for identifying new therapeutic targets to not only treat chronic pain effectively, but to do so without causing side effects such as addiction.

One promising approach over the last 10 to 15 years has been biased signaling at the MOR which was proposed to differentiate analgesic signaling (e.g. Gα_I/O) from side-effect signaling (e.g. βarrestin2). Combined with other molecular discoveries, including negative feedback systems upregulated during opioid treatment, receptor heterodimerization, and allosteric modulators, these findings suggest new drug discovery approaches that are fundamentally different from previous target-based strategies, since they simultaneously engage multiple targets or activate their receptor targets in new ways. These novel approaches hold the potential to develop opioids without adverse effects that we have been pursuing for more than a century. The approaches profiled in this review are summarized in Figure 1, and exemplary novel ligands created to exploit these new approaches are shown in Figure 2, Figure 3 and Figure 4. Please note, receptor nomenclature has been adjusted to conform to BJP’s Concise Guide to Pharmacology (Alexander et al., 2019).

Figure 1: Summary of Novel Approaches to Improve Opioid Therapy.

A summary scheme for each novel approach profiled in this review is shown. Each includes exemplary novel ligands created to exploit these strategies, all of which result in enhanced analgesia with reduced or maintained side effects. MOR = mu opioid receptor; NOP = nociceptin orphanin/FQ receptor; KOR = kappa opioid receptor; DOR = delta opioid receptor; HSP90 = heat shock protein 90; βarr2 = βarrestin2; MDOR = mu/delta opioid receptor heterodimer; 6TM-MOR = 6 transmembrane segment MOR splice variant. Figure created with BioRender.com.

Figure 2.

Chemical structures of multifunctional opioids and dual agonists at MOR-KOR and MOR-NOP.

Figure 3.

Chemical structures of dual agonists at KOR-DOR, MOR agonist-DOR antagonists and dual agonists at MOR-DOR.

Figure 4.

Chemical structures of biased MOR agonists, heteromer selective ligands, allosteric modulators and ligands targeting truncated splice variants of MOR.

Multi-Functional Opioids

One strategy utilized to overcome the undesirable effects of receptor mediated on target side effects is to target multiple opioid receptors simultaneously. The four members of the opioid receptor family are the mu opioid receptor (MOR), kappaopioidreceptor(KOR), deltaopioidreceptor(DOR), and nociceptinopioidreceptors(NOP); all 4 of these receptors have been shown to modulate pain. Mixed opioid receptor agonists and mixed agonist-antagonists may be a viable strategy to develop analgesics with reduced side-effects. Mixed agonists could maintain antinociceptive effects, associated with activation of an individual subtype, while the antagonism and/or agonism of the other subtype may be used to dial out the adverse effects. This approach was first suggested by the observation that when the MOR system is chronically stimulated, other negative feedback systems are also upregulated to compensate, which can drive side effects. Several of these feedback systems have been identified, and when targeted, have been shown to alleviate opioid side effects.

In this section we will review promising newer mixed target profiles like MOR-KOR agonism, MOR-NOP agonism, MOR-DOR dual agonism, MOR agonism-DOR antagonism, and KOR-DOR agonism. While some of these profiles have been explored for years, their results have not yet been translated to human testing.

MOR/KOR Dual Agonists

As opposed to MOR activation, which besides antinociception causes respiratory depression, inhibition of gut motility, and induces reinforcing effects, KOR activation induces antinociception without respiratory depression and inhibition of gut motility. Because KOR agonists often induce dysphoria they are generally considered not to have abuse liability. Prior evidence from preclinical studies suggest that dual activation of MOR and KOR may produce synergistic analgesia, while offsetting respective liabilities associated with individual subtypes. For example, U50,488h shows no respiratory depression on its own and has been reported to reduce the respiratory depression mediated by DAMGO administered supraspinally (Haji & Takeda, 2001). Another dual agonist, nalbuphine, with moderate efficacy MOR partial agonism and high efficacy KOR partial agonism in animals shows negligible respiratory depression (Schmidt, Tam, Shotzberger, Smith, Clark & Vernier, 1985). Dezocine is a MOR/KOR partial agonist that shows antinociception with reduced side effects, and is in clinical use in China (Wang et al., 2018b). Older mixed agents include 8-Carboxamidocyclazocine (8-CAC) which is a full agonist at both KOR and MOR, butorphan which is a partial agonist at both MOR and KOR, and agents like nalbuphine, pentazocine, nalmefene, nalorphine, and butorphanol, which are KOR agonist-MOR partial agonist/antagonists. These older ligands and their literature are reviewed in (Paton, Atigari, Kaska, Prisinzano & Kivell, 2020). More recent examples in this category include PPL-101/103 and MP1207/08.

PPL-101 and PPL-103 (Khroyan et al., 2017) are based on the morphine template and contain a chiral alpha methyl N-substituent leading to a unique combination of high binding affinities at both MOR and KOR and partial agonist activities at all three opioid receptors (MOR, KOR and DOR) with lowest efficacy at MOR and highest efficacy at KOR. They did not produce conditioned place aversion or preference, increased locomotor activity, or cause locomotor sensitization in mice.

MP1207 and MP1208 are based on the 6β-amidomorphinan template (Uprety et al., 2021), both of which displayed MOR/KOR Gi - partial agonism with diminished arrestin signaling (bias factors of 8 and 22, respectively at KOR and no arrestin signaling at MOR), while retaining efficacious analgesia when administered icv without inducing respiratory depression, hyperlocomotion (as compared to morphine at equianalgesic doses), or conditioned place preference/aversion in mice. One caveat: given that both classes of ligands are partial agonists, it is possible that partial agonism as opposed to bias plays a critical role in the improved in vivo profile observed for these compounds.

MOR/NOP Dual Agonists

Agonists targeting the fourth opioid receptor subtype, the nociceptin/orphanin FQ (N/OFQ) peptide (NOP) receptor, have been shown to have antinociceptive properties on their own. The roles of NOP in antinociception are complicated depending on the routes of administration in rodents (Kiguchi & Ko, 2019; Toll, Bruchas, Calo, Cox & Zaveri, 2016). However, in non-human primates (NHPs) NOP agonists exert antinociception supraspinally, spinally, as well as systemically (Kiguchi & Ko, 2019). More importantly NOP agonists do not show respiratory depression, itch, physical dependence, or addiction and in turn beneficially modulate the pain-relieving and addictive properties of MOR agonists (Toll, Bruchas, Calo, Cox & Zaveri, 2016) at spinal and systemic levels (Cremeans, Gruley, Kyle & Ko, 2012; Hu, Calo, Guerrini & Ko, 2010; Kiguchi, Ding & Ko, 2016). Taken together this has led to the development of bifunctional NOP/MOR agonists as a viable approach to develop non-addictive antinociceptive agents. Several molecules based on this hypothesis are now known like Cebranopadol (Lambert, Bird & Rowbotham, 2015), AT121 (Ding et al., 2018), BU08028 (Ding et al., 2016) and BU10038 (Kiguchi et al., 2019).

Cebranopadol was the first molecule of the non-morphinan template in this series and was discovered in 2014 (Linz et al., 2014). The compound showed high binding affinity at MOR, KOR, and NOP, and was found to be a full agonist at both MOR and NOP in [³⁵S]GTPγS functional assays (Linz et al., 2014; Tzschentke, Linz, Koch & Christoph, 2019). In cell lines cebranopadol was found to have 12-fold selectivity for MOR over NOP with comparable maximal stimulation of 89% and 103% at NOP and MOR compared to appropriate control agonists at each subtype. In rodents, cebranopadol showed anti-nociceptive effects in acute, chronic, and neuropathic pain models. These in vivo actions were consistent with combined treatment of NOP and MOR agonists when administered systemically. In rodents, cebranopadol exerted potent anti-allodynic effects irrespective of route of administration (spinal, supraspinal, or systemic). In conditioned place preference (CPP) and in drug discrimination assays in rodents it behaved similarly to morphine, which may be a consequence of its full agonist actions at MOR (de Guglielmo, Matzeu, Kononoff, Mattioni, Martin-Fardon & George, 2017; Tzschentke & Rutten, 2018). In NHPs it showed antinociceptive actions spinally as well as systemically without causing itch, which is commonly associated with MOR-based antinociceptive agents (Kiguchi & Ko, 2019).

In clinical studies, cebranopadol administered orally showed antinociception as well as typical opioid side-effects like respiratory depression and miosis, though these side-effects were present for shorter intervals compared to the drug’s antinociceptive actions. These findings suggest that cebranopadol exhibits a safer pharmacological profile over other clinically used MOR analgesics (Tzschentke, Linz, Koch & Christoph, 2019) like morphine though a head to head comparison with morphine or fentanyl was not made in the same trial. However, in the past Dahan and colleagues reported respiratory depression with MOR agonists like morphine and in comparison cebranopadol fared favorably (Dahan et al., 2017; Dahan, Romberg, Teppema, Sarton, Bijl & Olofsen, 2004). Cebranopadol has also been evaluated for lower back pain, diabetic neuropathy, and in cancer pain patients (Eerdekens et al., 2019). In all cases the drug was found to have analgesic actions with constipation, nausea, and dizziness in patients suffering from low back pain (Christoph, Eerdekens, Kok, Volkers & Freynhagen, 2017) seen in <10% cases. The compound is expected to be in phase III clinical trials soon. Whether the rewarding properties seen in rodents translates to humans still remains to be seen.

AT121 (Ding et al., 2018) is a partial agonist of NOP and MOP in comparison to cebranopadol which is a full agonist at both receptors. AT121 activated both MOR and NOP equipotently. Maximal efficacy in a [³⁵S]GTPγS assay was found to be 41% and 14% for NOP and MOR respectively. In NHPs, AT-121 showed antinociceptive effects comparable with morphine while showing no respiratory depression, reinforcing behavior, itching, opioid-induced hyperalgesia, or physical dependence. In addition, this NOP/MOR partial agonist attenuated the reinforcing effects of oxycodone without disrupting food intake. Its antinociception was antagonized by the NOP antagonist J-13397 and MOR antagonist naltrexone.

The Husbands group reported an analog of buprenorphine with a similar NOP/MOR partial agonism profile to AT121. BU08028 (Ding et al., 2016) had 13 fold higher selectivity for MOR over NOP. Its maximal efficacy at NOP was 48% compared to 21% at MOR. BU08028 showed antinociception in NHPs mediated in vivo by NOP as well as MOR. Similar to AT121 this drug showed no reinforcing behavior, respiratory depression, itching, or physical dependence. Another naltrexone based analog, BU10038 (Kiguchi et al., 2019), was recently characterized. BU10038 had a 14 fold MOR to NOP selectivity in binding assays, similar to BU 08028. Efficacy at the receptors was also similar, with 18% at MOR and 34% at NOP. Spinal administration of BU10038 showed antinociception without itching or analgesic tolerance when compared to morphine, while systemically it showed antinociception from both NOP and MOR while not showing respiratory depression, abuse liability, or physical dependence.

Taken together pharmacological evaluation of cebranopadol, AT121, BU08028, and BU10038 suggest that balanced partial agonist activation of NOP and MOR may lead to safer pain relievers. Challenges towards this hypothesis remain, e.g. the ratio of NOP:MOR efficacy and potency that leads to analgesia but not CPP is less clear. This is best illustrated by compounds from the Zaveri group (Khroyan, Polgar, Jiang, Zaveri & Toll, 2009; Khroyan et al., 2007). SR16435 (Khroyan et al., 2007) with a nearly equipotent/equi-efficacious ratio between MOR (EC₅₀=29 nM, E_Max=30%) and NOP (EC₅₀=28 nM, E_Max=45%) as well as partial agonism at both subtypes did show antinociception mediated by MOR but also showed CPP in spite of an in vitro profile similar to AT121 described above. SR14150 (NOP: EC₅₀=20nM, E_Max=54%; MOR: EC₅₀=99 nM, E_Max=23%) mediated analgesia in rodents through MOR and showed no CPP. Another compound, SR16507, had equipotency at both subtypes and was a full agonist at NOP and partial agonist at MOR (E_Max=47%) and showed MOR mediated antinociception while also showing CPP. In a chronic pain model however, both SR14150 and SR16435 showed potent anti-allodynic activity and in contrast to thermal nociception, this effect was completely blocked by the NOP receptor antagonist SB 612111 (Khroyan et al., 2009) but was not altered by naloxone. This suggests that NOP activation plays a bigger role in rodents in chronic pain models while MOR activation mediates in vivo actions in acute pain models.

Some of this lack of agreement between compounds could simply be in vitro results not translating to in vivo outcomes and/or differences in translatability between rodents and non-human primates (Cremeans, Gruley, Kyle & Ko, 2012). Further possibilities include meaningful differences in compound pharmacokinetics driving different outcomes, or aspects of pharmacodynamics that are not measured could impact compound performance, such as intrinsic efficacy. The antagonism of MOR mediated antinoception by NOP activation seen in rodents as well as antinociception by NOP activation is route dependent (Kiguchi, Ding, Kishioka & Ko, 2020; Kiguchi & Ko, 2019). It is hoped that with newer NOP/MOR ligands which are characterized in NHPs may lead to a better understanding of how NOP co-activation modulates MOR analgesia and reinforcing behavior and lead to a safer next generation of pain-relieving agents.

KOR/DOR Dual Agonists

Both KOR and DOR agonists are known to have antinociceptive actions. Complicating this anti-nociceptive activity, activation of DOR by some agonists is known to cause convulsions (Jutkiewicz, Baladi, Folk, Rice & Woods, 2006), although not all DOR agonists cause convulsions (Saitoh et al., 2018), leading to some confusion and controversy as to the role of the DOR in epileptic activity. On the other hand, activation of KOR is known to have anti-convulsant activity (Zangrandi, Burtscher, MacKay, Colmers & Schwarzer, 2016) in addition to blocking the addiction potential of drugs of abuse like cocaine and morphine (e.g. (Mello & Negus, 2000)). In addition, intrathecally administered KOR agonist U50,488H and DOR agonist DPDPE have been shown to have a synergistic anti-nociceptive effect (Miaskowski, Taiwo & Levine, 1990), suggesting that co-activation of DOR along with KOR in a balanced fashion may lead to fewer side-effects. This hypothesis has led to the synthesis of a few dual KOR-DOR agonists.

A chemical entity named MP1104 (Che et al., 2018) had dual KOR/DOR agonist actions in mice (Varadi et al., 2015a). MP1104 had antinociceptive actions in a radiant heat tail flick assay as well as the formalin assay (Ulker, Toma, White, Uprety, Majumdar & Damaj, 2020). The antinociceptive actions of MP1104 were intact in MOR KO mice and not blocked by the selective MOR antagonist β-funaltrexamine. The antinociception was blocked partially in KOR KO mice and by selective KOR and DOR antagonists, norBinaltorphimine(norBNI) and naltrindole (NTI) respectively. A combination of norBNI and NTI or KOR KO in the presence of NTI completely blocked the antinociception mediated by MP1104. Consistent with dual KOR/DOR agonist actions, MP1104 did not show typical KOR-like conditioned place aversion (CPA) in mice or rats (Atigari, Uprety, Pasternak, Majumdar & Kivell, 2019; Varadi et al., 2015a) while also not showing seizures in mice. Blockage of DOR receptors with NTI resulted in MP1104 showing CPA, suggestive that activation of DOR may blunt the aversive phenotype associated with KOR agonists. MP1104 also blocked cocaine CPP in mice and cocaine self-administration in rats while showing no anxiogenic, pro-depressive phenotypes in rats.

Two additional compounds which share this mixed DOR/KOR dual agonism include KDA 16 (Tang, Yang, Lunzer, Powers & Portoghese, 2011) and a diimidazodiazepine peptidomimetic, 2065–14 (Eans et al., 2015). When administered supraspinally, diimidazodiazepine, similar to MP1104, showed neither CPP nor CPA, supporting the hypothesis that DOR co-activation blunts KOR-mediated dysphoria. Taken together these probes demonstrate the promise of a dual-action DOR and KOR agonist to achieve pain relief with low side effects in vivo, suggesting that additional study of this topic is warranted in the near future.

MOR Agonist/DOR Antagonist Compounds

Chemical entities with MOR agonism and DOR antagonism have been heavily investigated with numerous probes now available. These ligands rest on the hypothesis that the DOR mediates at least some of the negative side effects of MOR activation, particularly tolerance and dependence/withdrawal. Support for this hypothesis came from early studies in rodents co-administered with the MOR agonist morphine with the DOR antagonist NTI, which demonstrated attenuated analgesic tolerance and physical dependence when compared to morphine alone (Abdelhamid, Sultana, Portoghese & Takemori, 1991). Furthermore, antisense knock-down of DOR produces a similar phenotype, suggesting a role for DOR in the development of morphine tolerance and physical dependence (Kest, Lee, McLemore & Inturrisi, 1996). Similarly, DOR antagonists are known to block morphine-induced reinforcing behavior (e.g. (Moron, Gullapalli, Taylor, Gupta, Gomes & Devi, 2010)). Together these studies suggested that a chemotype with MOR agonist and DOR antagonist activity would produce pain relief with reduced side effects.

Overall bifunctional peptides/peptidomimetics/small molecules with mixed MOR agonist/DOR antagonist activity displayed reduced tolerance and physical dependence, which was attributed to DOR blockade. Results from these molecules thus broadly and consistently support the hypothesis. Examples in this class include DIPP-NH2[Ψ] (a peptide) (Schiller et al., 1999); KSK103 (peptidomimetic) (Mosberg et al., 2014); and SRI-22138 (Ananthan et al., 2012), SRI-39067 (Vekariya et al., 2020), UMB425 (Healy et al., 2013), AAH8 (Anand et al., 2018), and mitragynine pseudoindoxyl (Chakraborty & Majumdar, 2021; Varadi et al., 2016) (non-peptide small molecules). While compounds with this activity remain moderately studied, mitragynine pseudoindoxyl, AAH8 and SRI-39067 notably demonstrated no conditioned place preference and less tolerance and withdrawal than morphine. Together, these studies validate a rational approach to target DOR with antagonists in order to attenuate side effects associated with MOR agonists.

These studies have generated a number of compounds from different labs which all show the same general benefits, providing strong support to the hypothesis. Outside of LP1 which has been characterized in carrageenan and chronic constriction injury chronic pain models (Parenti et al., 2013), however, the compounds were all studied using acute pain assays, mostly tail flick, with the side effects measured in naïve mice. The performance of this ligand class is thus understudied in chronic pain models, especially neuropathic pain, including the side effect profile of these ligands in chronic pain models. This limitation provides a clear path for future investigation.

MOR/DOR Dual Agonists

As discussed above, multifunctional MOR agonist/DOR antagonist compounds have been suggested to produce anti-nociception with reduced side effects by multiple groups. It is thus surprising and seemingly paradoxical that MOR/DOR dual agonists have also been described as beneficial, by producing anti-nociception with some of the same side effects also reduced. The data supporting this finding is more limited than that for the agonist/antagonist compounds, but still clear.

Perhaps the first dual agonist compound described was the peptide biphalin, first created by Lipkowsky (Lipkowski, Konecka & Sroczynska, 1982). This peptide was shown to produce potent anti-nociception, with reduced dependence liability (Shen & Crain, 1995). This peptide then served as a scaffold for the creation of numerous cyclized and linear derivatives, mainly headed by Adriano Mollica’s research group. These daughter compounds also showed potent anti-nociception, and in some cases reduced side effect liability (e.g. (Stefanucci et al., 2019)).

Another peptide ligand with a similar profile was MMP-2200 (a.k.a. lactomorphin) created by the Robin Polt research group. This highly potent dual agonist showed highly efficacious anti-nociception with reduced tolerance, dependence, and reward liability (Lowery, Raymond, Giuvelis, Bidlack, Polt & Bilsky, 2011; Mabrouk, Falk, Sherman, Kennedy & Polt, 2012; Stevenson et al., 2015). Notably, this peptide was glycosylated, which the Polt group has shown permits systemic administration with stable pharmacokinetics and blood-brain barrier penetration (Mabrouk, Falk, Sherman, Kennedy & Polt, 2012). Beyond peptides, small molecule dual agonists also showed similar benefits. A series of fentanyl-based analogs with dual agonist activity showed potent anti-nociception in multiple pain models with reduced motor liability (Podolsky et al., 2013). Our lab developed 6β-aminomorphinan compounds and a carfentanyl amide (MP102; (Gutridge et al., 2020; Varadi et al., 2015b)) with dual agonist activity, potent anti-nociception, and reduced respiratory depression (Varadi et al., 2013). A mitragynine derivative (MGM-16) showed potent anti-nociception in neuropathic pain (Matsumoto et al., 2014). Together these findings provide a firm foundation for the hypothesis that dual mu/delta agonism produces anti-nociception with reduced side effects. One caveat however is that abuse liability for these compounds has not been studied, with the exception of MMP-2200 above; this liability must thus be examined during further development.

What then accounts for the seeming paradox of agonist/antagonists and dual agonists producing some of the same benefits? One thing to keep in mind is that most of the studies above as well as the agonist/antagonist studies tested antinociception in the same way. The acute tail flick test was used for most anti-nociceptive testing, and chronic compound was only given to naïve mice for tolerance/dependence testing. Chronic dosing in chronic pain models was not often used, and the roles of the MOR vs. DOR were not teased out. We thus recently reported the results of our comprehensive study of the small molecule dual agonist SRI-22141 (Lei, Vekariya, Ananthan & Streicher, 2020). Like many of the studies above, we found this compound produced potent acute anti-nociception in the tail flick test; we also showed potent anti-nociception in 2 different models of chronic neuropathic pain. However, we also gave the drug chronically to naïve as well as chronic pain mice, showing reduced tolerance and dependence in the context of those pain states. When we used a selective delta opioid antagonist, we found that the delta receptor contributed strongly to neuropathic pain relief, but did not contribute to tail flick pain relief. Importantly, we also found that the dual agonist produced an anti-inflammatory effect in the spinal cords of mice suffering from neuropathic pain, which was reversed by delta antagonist.

We thus hypothesize that the benefits produced by dual agonism are acting via a different mechanism. The delta activity is probably not contributing to acute pain relief in naïve mice in no pain or that are uninjured via tail flick test or similar, while it does contribute strongly to pain states with an inflammatory component like neuropathic pain. The benefits to tolerance and dependence are thus mediated by the anti-inflammatory effect of delta opioid activation. In support of this hypothesis, it has been long established that the delta receptor is made competent for pain relief by inflammatory stimulation (Brackley, Gomez, Akopian, Henry & Jeske, 2016). Stimulation of the delta receptor has also been shown to produce anti-inflammatory effects in several models, including our study above (Shrivastava et al., 2017). Lastly, neuroinflammation has been shown to play a key role in side effects like opioid tolerance and dependence (Li, Csakai, Jin, Zhang & Yin, 2016). Testing this hypothesis fully may help resolve this basic science paradox, and determine the mechanisms by which dual agonists and agonist/antagonists both produce effective pain relief with reduced side effects. These studies may also indicate which pain types both ligand classes could be most effective in treating; this knowledge will be needed as these compounds advance to clinical testing.

G-Protein Biased Agonism

The development of opioid agonists biased against βarrestin2 recruitment has been a major area of focus in the opioid field. The first piece of the story for this approach was published by Laura Bohn in the lab of Marc Caron in 1999 when she discovered that βarrestin2 knockout in mice enhanced the anti-nociceptive efficacy of morphine (Bohn, Lefkowitz, Gainetdinov, Peppel, Caron & Lin, 1999). This finding was consistent with the known role of βarrestin2 in desensitizing/internalizing G Protein Coupled Receptors (GPCR). While interesting, this finding would not have been translationally meaningful if side effect efficacy was similarly boosted. Soon after the same group published a study showing that morphine tolerance was also blocked, consistent with a desensitizing role of βarrestin2, but not physical dependence (Bohn, Gainetdinov, Lin, Lefkowitz & Caron, 2000). Bohn and colleagues published another study showing that morphine reward learning was actually enhanced by βarrestin2 knockout; this finding raises safety concerns about biased agonists and reward that have never been fully addressed (Bohn et al., 2003).

True excitement for βarrestin2 knockout was generated when Laura Bohn’s group published more comprehensive studies of opioid side effects in her own independent lab. A 2005 study showed that morphine constipation and respiratory depression were all blocked by βarrestin2 knockout (Raehal, Walker & Bohn, 2005). A more comprehensive study in 2011 further found that tolerance and dependence/withdrawal were blocked by βarrestin2 knockout (Raehal & Bohn, 2011). One very important caveat to this study, however, is that high doses of morphine were able to produce equivalent side effects, and βarrestin2 knockout had no impact on side effects caused by fentanyl, methadone, and oxycodone. While collectively these studies did suggest that βarrestin2 knockout could improve the therapeutic index of opioids, these caveats and limitations, especially the enhanced reward seen with βarrestin2 knockout, must be kept in mind.

These studies suggested that stimulation of the opioid receptor without recruitment of βarrestin2 could result in improved analgesia with reduced side effects. A way to accomplish this goal came from the parallel field of biased agonism, also called functional selectivity. These studies, performed in other receptor systems apart from the opioid receptors, suggested that different ligands could activate the same receptor to produce different signal transduction cascades and thus different functional outcomes (reviewed in (Kenakin, 2007)). This concept was applied to the opioid field to screen for novel opioid agonists which could activate the mu opioid receptor (MOR) without recruiting βarrestin2.

Early reports identified biased compounds such as herkinorin, providing proof-of-concept, even if those particular drugs had poor drug-like properties (Groer et al., 2007; Tidgewell et al., 2008; Xu et al., 2007). Notably, to our knowledge, all identified biased agonists have been characterized for arrestin recruitment using one of several in vitro assays, which have poor sensitivity, and may not reflect arrestin recruitment in vivo. Perhaps due to this limitation, a recent study has found herkinorin to be unbiased at MOR (Manglik et al., 2016). PZM21 was identified fortuitously as a biased agonist as part of a drug design project with other aims; this unusual compound displayed decreased opioid side effects, but was also active in some pain assays but not others (Manglik et al., 2016). TRV130/Oliceridine was discovered by the pharmaceutical company Trevena; early in vivo reports showed decreased side effects in rodent models (Chen et al., 2013; DeWire et al., 2013). Based on these results TRV130/Oliceridine was advanced to clinical trials, which demonstrated rapid onset analgesia with potentially reduced respiratory depression (Soergel et al., 2014a; Soergel et al., 2014b; Viscusi et al., 2016). Although initially denied, Oliceridine was recently approved by the FDA for human use as a therapeutic with potentially improved safety, providing clinical validation for the biased agonist approach. It must be noted that TRV130 in preclinical studies has been reported to be addictive and recent literature argues it is a partial agonist in vivo as well (Faouzi, Varga & Majumdar, 2020; Singleton, Baptista-Hon, Edelsten, McCaughey, Camplisson & Hales, 2021). Laura Bohn’s group also published a recent paper with an entire new series of opioid biased agonists, and further demonstrated a linear relationship between bias factor and therapeutic index (Schmid et al., 2017).

These basic science, translational, and clinical studies do provide support for the use of biased agonists as improved pain therapeutics. However, recent studies have cast doubt on these findings (Cuitavi, Hipolito & Canals, 2021; Faouzi, Varga & Majumdar, 2020). A MOR knock-in mutant mouse with no C-terminal phosphorylation sites, and thus no desensitization or arrestin recruitment, was recently created (Kliewer et al., 2019). This mouse did have enhanced opioid anti-nociception and reduced tolerance, as expected from the known roles of βarrestin2, and in agreement with earlier studies. However, this mouse also had unchanged or worsened respiratory depression, constipation, and withdrawal, in direct contradiction to the side effect studies above. A replication study was also performed using a newly derived βarrestin2 knockout line on a backcrossed and clean genetic background, performed by 3 different labs in different countries (Kliewer et al., 2020). This βarrestin2 knockout mouse had morphine-induced constipation and respiratory depression that was indistinguishable from wild type. A similar negative replication study has also recently been published (Bachmutsky, Wei, Durand & Yackle, 2021). These findings, along with other recent studies, have cast doubt on the basic science foundation of this approach.

At the same time, recent studies have cast doubt on the biased agonist findings. Using PZM21 newly synthesized as well as obtained from the original reporting lab, a recent study found that this ligand recruited βarrestin2 in vitro and caused respiratory depression (40 mg/kg i.p. PZM21 showed equivalent respiratory depression to 10 mg/kg i.p. morphine) and tolerance in vivo (Hill et al., 2018). These findings are in direct contradiction to the results reported earlier (Manglik et al., 2016). A recent study (Gillis et al., 2020) also found that low intrinsic efficacy, instead of arrestin recruitment, could explain the observed benefits of biased agonists; this study directly tested TRV130/Oliceridine, PZM21, and SR-17018, the lead compound reported in (Schmid et al., 2017). The same study also found that a higher dose of PZM21 (100 mg/kg i.p.) displayed about half the respiratory depression in comparison to morphine at 10 mg/kg i.p., suggesting some separation of analgesia from respiratory depression. A recent preprint from Laura Bohn’s group also contests the findings that benefits seen with PZM21, SR17018 and TRV130 are due to low intrinsic efficacy and not biased signaling (Stahl & Bohn, 2020).

These contradictory findings assault both the basic science foundation and translational application of biased agonists for improved pain therapy. It still remains unclear if biased agonism can lead to less addictive drugs. While PZM21 has not shown CPP by multiple labs (Kudla et al., 2019; Manglik et al., 2016), it was found to produce reinforcing behavior in non-human primates (Ding et al., 2020). Clearly more study will be required to determine the true role of βarrestin2 in promoting analgesia vs. side effects, and whether G-protein biased agonists with intrinsic efficacy similar to the prototypic ligand DAMGO (in less amplified systems) are good clinical candidates. New studies should utilize refined methods to transcend the limitations of tools like germline knockout; the MOR phosphorylation mutant cited above is a good example of such an approach (Kliewer et al., 2019). New approaches should also be developed to assess ligand bias; as mentioned above, current in vitro assays of arrestin recruitment are very insensitive and may poorly reflect arrestin recruitment in neurons. Such studies are strongly needed to resolve this controversy and chart a new course forward for the field.

Opioid Receptor Heterodimers

The vast majority of studies that have been used to assign a particular role to a receptor or to determine the pharmacodynamic profile of a particular ligand use simple approaches that treat the receptor as an isolated unit. Studies on ligand selectivity and functional profile overwhelmingly use in vitro systems with a single overexpressed receptor, without other relevant receptors present. When those ligands are then used to assess the role of a particular receptor in vivo, the assumptions and limitations of those in vitro models are carried over to interpreting the in vivo results. However, it’s now clear that receptors do not function as isolated units; they engage in higher levels of organization within the cell, including receptor homo- and heterodimerization. A particular receptor in vivo could engage in multiple signaling types, including as an isolated receptor, as a homodimer pair, a heterodimer pair, or even larger units of organization that are still being described. A ligand targeting that receptor could thus evoke different clusters of signaling outputs, depending on how that receptor is organized and how that ligand interacts with these different organizational configurations. However, that in vivo complexity is missed when the assumptions from simple in vitro studies are used to interpret in vivo ligand results. Even other approaches like genetic knockdown/knockout do not avoid this problem, since a knocked out protein will be knocked out in all configurations, from single unit receptors to higher order heterodimer clusters. Appreciation for and investigation into these higher orders of complexity in receptor signaling is now being actively investigated in the opioid and pain field, with implications for drug discovery and development.

Perhaps the best described opioid heterodimer is the mu-delta opioid receptor heterodimer (MDOR). The MDOR has been unambiguously shown to exist in multiple in vitro cell culture models co-expressing both receptors (e.g. (Ananthan et al., 2012)). Using these models, the MDOR was suggested to engage in preferential signaling vs. the monomers, including βarrestin2 recruitment and ERK MAPK signaling (Rozenfeld & Devi, 2007). The chaperone protein RTP4 was also suggested to mediate the process of heterodimerization, and was further suggested to increase this activity in response to chronic morphine (Decaillot, Rozenfeld, Gupta & Devi, 2008; Fujita, Yokote, Gomes, Gupta, Ueda & Devi, 2019).

While useful, these models all involve overexpression in simplified cell lines, and could thus be an artifact of the culture model. There is some evidence of the presence and function of the MDOR in vivo, but the data is less clear, and has generated some controversy, in part due to the lack of MDOR-selective tools. Lakshmi Devi and colleagues used an MDOR-selective antibody generated by a screening approach to demonstrate broad expression of the MDOR in brain, which further increased in expression in response to chronic morphine (Gupta et al., 2010). This increase in response to morphine links with the RTP4 results above (Fujita, Yokote, Gomes, Gupta, Ueda & Devi, 2019), suggesting a mechanism by which MDOR could increase with morphine treatment. Jennifer Whistler and colleagues used a cocktail of drugs to stabilize MDOR expression, which led to a decrease in anti-nociceptive potency of opioid drugs (Milan-Lobo, Enquist, van Rijn & Whistler, 2013). Combined with the above findings, this suggests the MDOR could act as an anti-opioid negative feedback loop, stimulated by chronic opioid, which would then act to dampen the opioid system. In support of this hypothesis, another study found that the MDOR promoted opioid anti-nociceptive tolerance (Beaudry, Gendron & Moron, 2015). Further studies suggested specific signaling evoked by the MDOR in the CNS (Kabli, Fan, O’Dowd & George, 2014; Schuster, Metcalf, Kitto, Messing, Fairbanks & Wilcox, 2015), while another suggested that MDOR activation could have anti-depressive and anti-anxiety effects (Kabli, Nguyen, Balboni, O’Dowd & George, 2014). However, these approaches in general used indirect methods with unclear MDOR selectivity, which has generated some controversy as to the presence and role of the MDOR in vivo, abetted by other studies suggesting no functional MDOR interaction in vivo (Wang et al., 2018a). These limitations point out the need to develop specific tools to investigate the MDOR.

Along these lines, some pharmacological tools have been developed. The small molecule agonist CYM51010 was found to be MDOR-selective, and evoked anti-nociception with reduced tolerance (Gomes et al., 2013). However, this ligand was only 5–6 fold selective for the MDOR over MOR and DOR, and thus could cause a mix of heterodimer and monomer activity. We recently developed the compound MP135, with a higher MDOR agonist selectivity; we showed that this compound produced anti-nociception, but also significant opioid-like side effects including respiratory depression and abuse liability (Faouzi et al., 2020). MDAN21 was developed by the Portoghese group, which is a bivalent linked ligand with mu agonist and delta antagonist pharmacophores. This ligand caused highly potent anti-nociception in mice, with reduced tolerance, dependence, and reward (Daniels, Lenard, Etienne, Law, Roerig & Portoghese, 2005; Lenard, Daniels, Portoghese & Roerig, 2007). While this ligand did have a U-shaped linker length to potency relationship, suggesting heterodimer activity, the molecular pharmacology of this compound at the MDOR was not defined. In addition, it is not clear what pharmacodynamic effect this ligand would have at the MDOR with a mix of agonist and antagonist pharmacophores; it could even potentially cause MDOR antagonism or disruption. The results of these studies thus do not definitively suggest a MDOR role in vivo, although they do suggest this activity profile could be a promising therapeutic approach if a drug-like molecule could be developed. Lastly, we developed a bivalent linked compound with dual antagonist pharmacophores, deemed D24M. We showed this compound was ~90 fold selective for the MDOR over either monomer, and reduced morphine withdrawal in mice (Olson, Keresztes, Tashiro, Daconta, Hruby & Streicher, 2018). We further showed that D24M suppressed opioid anti-nociception via suppression of Src and CaMKII signaling in the brain, also supporting this hypothesis (Keresztes et al., 2021). These results are more clear for the role of the MDOR, since they use a ligand with a high selectivity and defined molecular pharmacology. These results further support the hypothesis that the MDOR acts as an anti-opioid negative feedback loop. Taken together, these studies all suggest that the MDOR could be targeted as a therapeutic candidate, most likely using a selective antagonist, to reverse the anti-opioid negative feedback loop caused by chronic morphine treatment. This should enhance anti-nociception while reducing side effects like tolerance and withdrawal. However, studies on the MDOR are still in the early stages, and much more work remains to be done on the basic science and translational potential of this target. This is especially true for abuse liability, which has not been definitively tested for the MDOR.

Other opioid heterodimers have also been found in the literature, and some could be exploited for improved pain therapy. Kelly Berg and colleagues have described a kappa-delta opioid heterodimer (KDOR) in peripheral nociceptors that can produce anti-nociception without the typical side effects of central mu or kappa ligands (Berg et al., 2012; Jacobs, Pando, Jennings, Chavera, Clarke & Berg, 2018; Jacobs et al., 2019). Some ligands have also been developed which may engage this target. A bivalent linked ligand KDAN18 has been developed and suggested to be KDOR-selective, and was also shown to produce anti-nociception (Daniels, Kulkarni, Xie, Bhushan & Portoghese, 2005). One caveat to this ligand though is that it has mixed kappa agonist and delta antagonist pharmacophores, and thus like MDAN21, the molecular pharmacology of this ligand at the KDOR is not clear. The small molecule 6’-guanidinonaltrindole (6’-GNTI) was also suggested to be KDOR-selective by the Berg and Whistler groups (Jacobs, Pando, Jennings, Chavera, Clarke & Berg, 2018; Waldhoer et al., 2005), and does produce anti-nociception. However other groups have reported that this ligand is a potent G-protein biased agonist at kappa opioid monomers, calling this selectivity into question (Rives, Rossillo, Liu-Chen & Javitch, 2012; Schmid, Streicher, Groer, Munro, Zhou & Bohn, 2013). Further work will thus be needed to develop proof-of-concept selective KDOR agonists and evaluate their therapeutic profile.

Among the opioid heterodimers observed, the MDOR and KDOR above are the best established. Other pairs have been described however, and in some cases linked to antinociception, including mu opioid-vasopressin1b, mu opioid-kappa opioid, mu opioid-GPR139, and more. These newly observed heterodimers have been recently reviewed by Zhang and colleagues (Zhang, Zhang, Hang & Liu, 2020). In general, these pairs are described by only one or a few studies, and some have not been validated in vivo. There are also no selective ligands for these targets, meaning that future basic science and translational work must be done to establish these heterodimers as valid therapeutic targets and to create new selective ligands that have the potential to modulate these targets to improve pain therapy. Much more work also has to be performed to study the abuse liability interactions of these heterodimer pairs, including the MDOR and KDOR, which in general has not been studied.

Opioid Receptor Allosteric Activation

Most MOR, KOR and DOR ligands described to date target the orthosteric binding site. In contrast, allosteric ligands bind to sites distinct from the orthosteric binding pocket and may have some advantages in terms of spatial-temporal control of receptor activation. Allosteric modulators, in particular positive allosteric modulators (PAMs), can enhance affinity, potency, and efficacy of exogenous drugs or endogenous opioid peptides. Crucially, a pure PAM will only potentiate receptors which already have an agonist bound, such as an endogenous peptide. This means that unlike an exogenous orthosteric agonist, not all receptors in the body will be activated, only those currently activated by endogenous systems. Negative allosteric modulators (NAMs) on the other hand can inhibit binding of orthosteric ligand agonists and/or decrease ligand functional activity. Silent allosteric modulators (SAMs) do not affect binding or activity of the orthosteric site agonists directly but instead block the activity of NAMs or PAMs. An ideal allosteric modulator will enhance the in vivo agonist effect of the endogenous peptides while at the same time reducing the negative adverse effects that accompany opioid receptor activation, since receptors in those circuits may not be activated by endogenous systems.

Two promising MOR PAMs are now known, BMS-986121 and BMS-986122, both of which enhance G-protein and βarrestin-2 signaling of orthosteric site MOR ligands (Burford et al., 2013) though G-protein signaling is enhanced more (discussed later). In addition, BMS986122 increased the potency but not efficacy of MOR full agonists and only enhanced the efficacy of partial agonists like morphine without effecting potency. As expected it did nothing to antagonists (Livingston & Traynor, 2014). Another probe, MS1, showed an increase of MOR βarrestin-2 over G-protein signaling, suggestive that this ligand can be used in conjunction with orthosteric ligands to induce biased agonism as a pharmacological tool, albeit not as a clinical candidate (Bisignano et al., 2015). Other PAMs of interest include BMS 986187, a biased allosteric modulator of DOR. BMS 986187 showed reduced receptor phosphorylation and low receptor internalization, compared to the full agonist SNC80 (Burford et al., 2015; Stanczyk, Livingston, Chang, Weinberg, Puthenveedu & Traynor, 2019).

Recent studies from the Traynor group have also suggested that BMS-986122 allosterically disrupts the Na⁺ binding site of MOR (Livingston & Traynor, 2014). Studies dating back to the 1970s showed that incubation of MOR with NaCl inhibits agonist binding and function while not affecting antagonist binding. The Na+ site is located in the middle of the 7-TM bundle and involves coordination with D114^2.50, N150^3.35, N328^7.45, S329^7.46, and S154^3.39, along with a number of highly organized water molecules across TM domains II, III, VI and VII (Katritch, Fenalti, Abola, Roth, Cherezov & Stevens, 2014). Importantly, the Na⁺ ion is absent in the structures of active state GPCRs like MOR (Huang et al., 2015). It is believed that expulsion of the Na⁺ from the pocket pushes TM6 outwards, along with an inward shift of TM7 at the level of the conserved NPxxY motif, thus facilitating G-protein binding and activation.

BMS-986122 and Na⁺ ions oppose each other’s action, and BMS-986122 binds at a distinct site from Na⁺ to allosterically disrupt the binding of Na⁺. While Na⁺ favors the inactive state of the receptor, the PAM BMS-986122 stabilizes the active state of the receptor, explaining the decreased potency of Na⁺ ions to inhibit binding of the agonist DAMGO in the presence of BMS-986122. A molecular dynamics simulation study proposes that BMS986122 stabilizes TM7 by engaging with Trp7.35 and prevents Na⁺ from disrupting agonist actions (Bartuzi, Kaczor & Matosiuk, 2016). Confirmation for this proposed model will require validation by other biophysical methods like NMR, and possibly atomic structures using X-ray or cryoEM of allosteric modulators bound to opioid receptors.

BMS-986122 was recently evaluated in vitro, ex vivo, and in vivo as well. In cell lines it increased G-protein activation of the endogenous opioid peptide Met-enkephalin to a greater extent than for β-arrestin2. This was confirmed in mouse brain homogenates as well as in electrophysiological assays for GABA-release in the periaqueductal gray, where it also increased G-protein activity of Met-enkephalin. In mice it had antinociceptive actions mediated through the MOR while showing reduced constipation, conditioned place preference, and respiratory depression. Taken together the allosteric modulation approach represents a highly innovative strategy to enhance the body’s ability to fight pain while not recruiting adverse effects associated with orthosteric site binders, potentially including abuse liability (Kandasamy et al., 2021).

Targeting Opioid Receptor Splice Variants

The mu opioid receptor gene (OPRM1) undergoes extensive alternative splicing to produce multiple splice variants that are conserved in rodents and humans. These splice variants can be classified under three categories: a) full length seven transmembrane (7TM) domain C-terminal variants, b) truncated six transmembrane domain variants (6TM) missing the first transmembrane of MOR, and finally c) truncated single TM variants that only have the first TM.

The significance of 7TM C-terminal variants have been demonstrated by seminal work from Pasternak and Pan. These include differences in G-protein signaling, phosphorylation, and internalization, as well as differences in the pharmacological actions of morphine in mice (reviewed in (Pasternak, Childers & Pan, 2020)). Recent evidence also suggests that these C-terminal variants may lead to differential bias for a variety of MOR ligands (Narayan, Hunkele, Xu, Bassoni, Pasternak & Pan, 2021). Given the complexity involved with these variants it is unlikely these variants can be individually targeted for drug development. The single TM variants do not bind any opioids; instead they are believed to act as chaperones increasing the receptor expression of MOR at the cell surface and aid in increasing the analgesic potency of morphine in mice (Xu et al., 2013). This also makes the single TM variants a poor choice for drug development.

In contrast, the 6TM-MOR variants are of high interest for drug development based on the actions of the 6TM-selective agonist 3’-iodobenzoyl-6β-naltrexamide (IBNtxA) and its analog IBNalA (Grinnell et al., 2014; Majumdar et al., 2011b; Majumdar et al., 2012). In mice as well as rats, IBNtxA was found be an analgesic in the radiant heat tail flick and hot plate assays. In both species, IBNtxA showed no respiratory depression, physical dependence, or cross-tolerance with morphine. Furthermore in mice IBNtxA showed no reward behavior (Islam et al., 2020; Majumdar et al., 2011b) and was found to be a potent analgesic in neuropathic as well as inflammatory pain assays (Wieskopf et al., 2014).

The mechanism of action of IBNtxA is unique. In mice lacking the full length MOR gene (Oprm1 exon 1 KO mice) IBNtxA retained its analgesic actions, while morphine’s analgesic actions were totally lost. In a 6TM variant KO mice (Oprm1 exon 11–KO mice) which retains full length 7TM MOR variants, the analgesic actions of IBNtxA were lost while the analgesic actions of morphine remained intact. IBNtxA analgesia was independent of 7TM MOR, KOR and DOR, as mice lacking genes for 7TM MOR, KOR, and DOR (triple opioid KO mice preserving 6TM variants, (Majumdar et al., 2011b)) retained the analgesic actions of IBNtxA while the actions of morphine, U50,488h, and DPDPE was lost in these mice. The 6TM-dependent mechanism of action was further confirmed using lentivirus-mediated delivery of 6TM variant MOR-1G in mice lacking both 6TM as well as 7TM MOR variants (exon 1 and exon 11 Oprm1 KO mouse). While both morphine as well as IBNtxA were inactive in the parent mutant animal, the analgesic action of only IBNtxA and not morphine was rescued when 6TM variant MOR 1G was expressed in mice again (Lu, Xu, Rossi, Majumdar, Pasternak & Pan, 2015; Lu et al., 2018).

While these results are exciting and interesting, several unanswered questions still remain about the mechanism of 6TM action in vivo. Besides the actions of the synthetic drug IBNtxA, the 6TM variants also mediate the analgesic actions of KOR, DOR and α2 adrenergic agents like clonidine; however, none of these agents have any binding or functional activity at the 6TM receptor expressed alone in HEK cells (Marrone et al., 2016a). Morphine in Exon 11 MOR KO mice lacking 6TM variants also shows reduced tolerance, hyperalgesia, and locomotor effects while retaining its analgesic actions (Marrone et al., 2017). Other MOR agonists like buprenorphine, endomorphins, and DAPP (Dmt-D-Ala-Phe-Phe-NH2) (Marrone et al., 2016b) in the same mice show attenuated actions in mice lacking 6TM (Oprm1 exon 11–KO mice) as well as the classical 7TM KO mice (Oprm1 exon 1–KO mice). Taken together it appears 6TM MOR agonists can be classified under three categories: 1) 6TM MOR dependent/7TM MOR independent (IBNtxA), 2) 6TM dependent-7TM MOR dependent- (buprenorphine, peptides like DAPP and endomorphin), and 3) 6TM dependent/non MOR-KOR-DOR and α2 agents.

One hypothesis proposed to explain the complexity of 6TM actions is dimerization with a GPCR or non-GPCR protein. [¹²⁵I]IBNtxA binding was seen in triple opioid KO mice (mice lacking 7TM MOR, KOR and DOR) similar to wild type membranes, while lost in membranes from 6TM KO mice (Oprm1 exon 11–KO mice) mice. Binding of [¹²⁵I]IBNtxA was seen in membranes co-expressing 6TM variant MOR 1G along with ORL1 and not in MOR 1G alone (Majumdar et al., 2011a); this suggests that a 6TM variant may dimerize or physically associate with another non-opioid GPCR to initiate its actions, although the exact identity of the binding site has not yet been determined (Majumdar et al., 2011a). A recent manuscript reported co-expression of MOR 1G along with full length 7TM MOR led to enhancement of MOR expression at the cell surface, leading to functional 7TM MOR at the protein level. MOR-1G physically interacts with 7TM MOR starting at the endoplasmic reticulum, enhancing MOR expression at the plasma membrane through a chaperone like function (Zhang, Xu & Pan, 2020). Heteromerization of 6TM with β2AR has also been reported by Diatchenko and co-workers as a mechanism for synergistic analgesia when the 6TM and β2-adrenoreceptor drugs are co-administered (Samoshkin et al., 2015).

It remains to be seen if the analgesic actions of IBNtxA and other 6TM dependent drugs can be linked to either ORL1 or other GPCRs in vivo. Selectivity of drug action at 6TM targets also remains a challenge. IBNtxA is a very nonselective drug with high affinity and activity at MOR, KOR and DOR (Grinnell et al., 2021; Islam et al., 2020; Majumdar et al., 2011a). Structure-activity relationships (SAR) on the IBNtxA template did lead to partial selectivity for 6TM over DOR, and showed that the phenyl ring at the 6-position of the epoxymorphinan moiety is required for 6TM affinity (Majumdar et al., 2012). SAR studies have been hindered by the lack of a functional assay due to the unknown identity of the partner protein or proteins with which the 6TM variant MOR1G physically associates. Characterization of IBNtxA in ORL1 and 6TM/ORL1 double KO mice would aid in revealing the composition of the site labelled by IBNtxA. Identification of a partner receptor will permit development of an assay and in turn lead to selective 6TM drugs which retain the behavioral profile seen with the parent drug IBNtxA while also elucidating 6TM opioid function in vivo.

Inhibition of Heat Shock Protein 90

One recent target that has been uncovered for novel pain therapy is the central chaperone protein Heat shock protein 90 (Hsp90). Hsp90 is a highly conserved and highly expressed protein in all cells, where it occupies a central role in cell biology. Hsp90 carries out numerous roles, including acting as a chaperone to assist in protein folding and stabilization, regulating target protein location within the cell, forming specific protein complexes, and regulating the specificity and onset of signal transduction. Despite this central importance, Hsp90 has barely been studied in the context of neuroscience, pain, or opioids.

Some studies have suggested that Hsp90 inhibition is anti-inflammatory, suggesting the use of this protein for managing inflammatory pain. Non-selective Hsp90 inhibitors were shown to prevent the onset and duration of neuropathic pain via this anti-inflammatory activity (Lewis et al., 2010). While this suggests potential direct use in the management of inflammatory-related pain modalities, these studies do not show an impact on opioid anti-nociception or signaling. A few studies did show that Hsp90 was upregulated in the brain in response to chronic morphine, and that non-selective Hsp90 inhibition reduced the severity of opioid withdrawal (Abul-Husn et al., 2011). A supportive study found that Hsp90 inhibition reduced cellular markers of dependence and withdrawal in an in vitro model (Koshimizu et al., 2010). These studies are suggestive, but do not establish the role of Hsp90 in regulating opioid signaling and whether this target can be used to enhance opioid therapy.

To this end, we’ve engaged in a long-term investigation into the role of Hsp90 in regulating opioid anti-nociception and signal transduction in the central nervous system. Our studies to date have shown a regional difference in Hsp90 activity. We found that in the brain, Hsp90 promoted ERK MAPK activation and thus promoted opioid anti-nociception; thus inhibition of Hsp90 in the brain blocked ERK MAPK and opioid anti-nociception (Lei et al., 2017). We found this effect consistent across many acute and chronic pain states, including post-surgical paw incision, HIV peripheral neuropathy, and chemotherapy-induced peripheral neuropathy (Lei et al., 2017; Stine et al., 2020). In contrast, we found that in the spinal cord, Hsp90 blocked ERK MAPK activation and thus blocked opioid anti-nociception, with Hsp90 inhibition in the spinal cord thus promoting opioid anti-nociception (Duron et al., 2020). This enhancement of opioid anti-nociception in the spinal cord suggested we could use this activity to enhance opioid therapy. We found this hypothesis was correct, in that spinal Hsp90 inhibition consistently improved morphine potency in acute and chronic pain models by 2–4 fold, while improving tolerance, rescuing established tolerance, and not changing the potency of constipation and reward (unpublished data in review).

These results suggest a translational path to target Hsp90 to improve opioid therapy. Spinal cord inhibition improves the therapeutic index of morphine, enabling a dose-reduction strategy. This means that a lower dose of opioid could be given to patients, which due to enhanced anti-nociceptive potency, would retain analgesic efficacy. However, since side effects are either improved or not changed, the lower opioid dose would result in fewer side effects, especially tolerance. While high efficacy agonists like fentanyl produce less tolerance, they do not produce less reward and similar side effects, suggesting the potential superiority of a dose-reduction approach. One roadblock to this approach however was our finding that non-selective inhibitors given systemically always result in a brain-like phenotype, with blocked opioid anti-nociception (Duron et al., 2020). One way around this roadblock could be to target specific Hsp90 isoforms only active in the spinal cord. In our initial investigations, we found that the Hsp90α isoform and the p23 and Cdc37 co-chaperones were responsible for regulating opioid anti-nociception in the brain (Lei, Duron, Stine, Mishra, Blagg & Streicher, 2019). In contrast, our in-progress studies suggest that additional isoforms and co-chaperones regulate opioid anti-nociception in the spinal cord (Hsp90β, Grp94, Aha1). By using systemic isoform-selective inhibitors targeting these spinal-only isoforms, we were able to boost anti-nociception and reduce tolerance while giving drug by a translationally relevant route (unpublished data).

Together these findings do suggest a way forward to target Hsp90 to improve opioid therapy. General Hsp90 inhibitors can be given to block inflammation, preventing and easing inflammatory and neuropathic pain. More specifically, isoform-selective Hsp90 inhibitors can be given to selectively boost the anti-nociceptive efficacy of opioids to manage multiple pain types, while enabling opioid dose-reduction and reduced side effects. While current studies do suggest a benefit in managing opioid reward as above, this topic must be explored in further detail.

Future Directions

As reviewed above, many promising pre-clinical approaches are in development which may result in the next generation of efficacious but non-addictive painkillers. However, from past experience in the field, it’s clear there is no one “silver bullet” approach. Any of these approaches may fail to bridge the “valley of death” between preclinical development and clinical testing and approval. Many such approaches have failed in the past, even with very promising animal model data. General translational challenges include species differences between rodents/dogs/primates typically used during development and human patients, and business challenges like raising capital for development and clinical testing. Beyond the molecular pharmacodynamic approaches reviewed here, there is a robust debate in the field over translational methods like the animal models used to assess pain and side effects. Future exploration of the basic behavioral science of pain may result in more translational animal models which will improve our ability to predict ligand performance in humans. We must explore every avenue in full and expand our basic science knowledge at all levels of pain and analgesia if we are to be successful in creating that next generation of analgesic drugs. Past experience has shown there are no easy fixes, we must fully understand this system if we are to solve the twin crises of chronic pain and opioid abuse and overdose.

Abbreviations:

Aha1

Activator of 90 kDa heat shock protein ATPase homolog 1

CPA

conditioned place aversion

CPP

conditioned place preference

cryoEM

Cryogenic electron microscopy

DAMGO

[D-Ala², N-MePhe⁴, Gly-ol]-enkephalin

DOR

delta opioid receptor

ERK

extracellular signal-regulated kinase

FDA

food and drug administration

7TM

full length seven transmembrane

GPCR

G protein coupled receptors

GABA

gamma aminobutyric acid

GPR139

G grotein coupled receptor 139

Grp94

glucose-regulated protein 94

Gα_I/O

guanine nucleotide-binding protein alpha family member i/o isoform

Hsp90

heat shock protein 90

Hsp90β

heat shock protein 90 β isoform

KDOR

kappa-delta opioid receptor heterodimer

KOR

kappa opioid receptor

MAPK

mitogen-activated protein kinase

MOR

mu opioid receptor

MDOR

mu-delta opioid receptor heterodimer

N/OFQ

nociceptin/orphanin FQ

NAMs

negative allosteric modulators

NHP

nonhuman primates

NMR

nuclear magnetic resonance

NOP

nociceptin opioid peptide receptor

OPRM1

opioid receptor mu 1

ORL1

opioid-receptor-like 1

PAMs

positive allosteric modulators

RTP4

receptor transporter protein 4

SAMs

silent allosteric modulators

TM7

seventh transmembrane domain of GPCRs

6TM

truncated six transmembrane domain variants