In this study, we comprehensively screened the molecular pharmacology of 9 clinically relevant opioids at 9 pain-related receptor targets (Table 1). Such a comprehensive screen has not been performed for these compounds to our knowledge, and our findings could potentially explain discrepancies between clinical opioids in their in vitro vs. in vivo τ values and lack of cross-tolerance (see Introduction). We found many expected interactions, including expected binding affinities and potencies at the MOR, and NET/SERT interaction for tapentadol and tramadol. These expected findings, rank order affinities/potencies, and the expected behavior of positive control compounds all validate our experimental approach [3].

This validation lends confidence to our unexpected findings, summarized in Table 6. Some of these findings are likely to be significant at therapeutic doses, such as σ1R binding that is only 2.3 fold over MOR affinity for hydrocodone and approximately equal affinity to MOR for tapentadol. Similarly, hydrocodone displays CB1 partial agonist activity that is only 3.1 fold over MOR potency, and O-desmethyl-tramadol displays DOR partial agonist activity that is equal in potency to the MOR. However even low potency interactions have the potential to be significant at therapeutic doses. We detected a highly novel but low potency transporter activating activity by buprenorphine; using the transporter inhibitor duloxetine, we show results consistent with a transport activation activity by buprenorphine at a moderate ~A₅₀ dose in mice (Fig 4). These results suggest that even our low potency/affinity novel interactions should be explored for potential roles at therapeutic doses in vivo.

We did observe discrepancies in some cases between the binding (Table 4) and functional data (Table 5). While in some cases binding and function matched well—such as oxymorphone at the MOR (11 vs. 23.1 nM)—in others they did not, as multiple compounds at multiple targets showed functional activity without any apparent binding affinity (O-desmethyl-tramadol at DOR, multiple compounds at CB1 and NET, etc.). In a few cases, binding affinity was significantly greater than functional potency, such as buprenorphine at the DOR (34 nM affinity vs. 1684 nM potency). Since compounds like oxymorphone matched previously reported results well, and positive controls behaved as expected, the answer cannot be as simple as one assay being more or less sensitive than the other. Several potential explanations could be behind these discrepancies. For one, compounds could have high or low intrinsic efficacy, which can more or less efficiently translate binding into a functional response [12, 38]. By this explanation, O-desmethyl-tramadol could have high intrinsic efficacy at the DOR, while buprenorphine could have low intrinsic efficacy at the DOR. For another, some compounds could bind allosterically to produce a functional change, while radioligand binding will only measure displacement of orthosteric ligand binding [12]. This could most readily explain compounds that produced functional changes without detectable binding, such as O-desmethyl-tramadol which produced no binding displacement at the DOR up to a 10 μM concentration, but produced functional changes at the modest potency of 361 nM.

Related to these discrepancies, the affinity and potency values measured are in some cases lower than those reported elsewhere in the literature, particularly for hydrocodone (1757 nM affinity, 472 nM potency). However, our rank order of compounds matches well with the literature, along with the performance of our positive control compounds, suggesting the results are valid [3]. Contrary to common perception, reported molecular pharmacology values for opioid (and other) compounds can vary wildly, and are highly dependent on assay conditions, cell line, temperature used, and other variables. For example, in the case of fentanyl, an affinity range of 20,000 fold has been reported throughout the literature [3]. Considering all factors, our observations appear valid, within the range of literature values, and are confirmed by intrinsic factors such as positive control compounds and potency/affinity rank order.

Among our observed novel interactions, some may be significant, and fall into a few categories (Table 6). Hydrocodone, O-desmethyltramadol, oxymorphone, and tapentadol showed previously unreported KOR and/or DOR partial agonist activity. This is not surprising, as other opioids such as morphine have long been known to show KOR and DOR agonist activity [39, 40]. Evidence from MOR KO mice does suggest that morphine mediates anti-nociception fully through the MOR [4]; however, evidence with drugs like oxycodone suggests that the DOR and/or KOR could mediate part of the anti-nociceptive response with these drugs [19]. DOR and KOR activity should thus be explored as contributors to the pharmacological profile of hydrocodone, O-desmethyl-tramadol, oxymorphone, and tapentadol, especially as some of these compounds have very similar MOR and KOR/DOR potencies. KOR/DOR activity could be the mechanism behind lack of cross-tolerance between morphine and oxycodone [8], and could be involved with the above drugs.

We also found that hydrocodone, hydromorphone, and morphine have CB1 partial agonist activity, while buprenorphine has CB1 inverse agonist activity, all unreported; hydrocodone furthermore has a CB1 potency only 3.1 fold lower than MOR, suggesting potential activity at therapeutic doses. CB1 is an extremely widespread and highly expressed modulatory system in the brain, that among other roles, promotes anti-nociception by itself and in synergy with the opioid system [26, 41]. We hypothesize that if these drugs promote CB1-mediated anti-nociception, this effect could synergize with the opioid system and promote a dose-reduction effect to decrease opioid and CB1-mediated side effects, as has been shown with separate opioid and CB agonists [42]. In contrast, if buprenorphine acts as a CB1 inverse agonist, this could cause a pro-nociceptive effect, limiting potential anti-nociception (as we observed for buprenorphine transporter activity in Fig 4). These effects could again explain some discrepancies between opioids in vitro and in vivo, and could suggest drug discovery approaches to refine these CB1 vs. MOR activities to enhance anti-nociception.

Another novel finding is that hydrocodone and tapentadol bind to the σ1R with affinities quite close to that for the MOR (2.3 fold for hydrocodone, same for tapentadol), suggesting activity at therapeutic doses. The σ1R is an intracellular chaperone-like protein mainly expressed in the endoplasmic reticulum and strongly concentrated in the brain; it may act to selectively amplify signal transduction, and σ1R antagonists are being explored as novel analgesics [27, 43, 44]. Hydrocodone and tapentadol binding to the σ1R could thus promote anti-nociception. However one drawback to our approach is that the binding assay does not reveal the functional impact of these drugs on the σ1R. Further research will thus be required to determine how hydrocodone and tapentadol modulates σ1R function, and what is the resulting impact on anti-nociception.

Lastly, we found that buprenorphine has a monoamine transporter activating activity, a highly novel finding. Although low potency, our in vivo experiments suggest that this activity could be relevant at moderate doses (Fig 4). In the entire clinical pharmacopeia, we could find no other drugs that activate monoamine transport. In the entire literature, we could find only a single report of 2 natural product compounds that causes monoamine transporter activation [45]. The literature and our data (Fig 4) suggests that this effect is pro-nociceptive. However, as we see the highest efficacy at the DAT, this effect could act to reduce dopamine release in the brain, which we would expect to reduce reward and drug abuse liability [46]. This effect could thus explain the efficacy of buprenorphine as a treatment for drug addiction, and could be further enhanced to improve efficacy for this indication [47].

These findings should also be placed into a broader context of novel drug discovery and development for the opioid receptors. New findings in regards to the molecular pharmacology of ligand:receptor interaction and downstream signaling pathways suggest additional complexity to the system that can be utilized to improve opioid therapy, in greater detail than the simple molecular pharmacology screening performed in this study (we review some of these novel approaches in [48]). For instance, receptors can interact with each other in homodimer or heterodimer pairs, or higher order oligomeric complexes, that evoke unique signaling when compared to the monomers that can be taken advantage of to improve opioid therapy [49]. We recently created a bivalent antagonist specific for the mu-delta opioid receptor heterodimer that enhances opioid anti-nociception while decreasing side effects like withdrawal, suggesting that heterodimers could be promising future drug discovery targets [32]. Looking to downstream signaling, seminal work from the lab of Laura Bohn identified the signaling regulator βarrestin2 as a molecule which decreased morphine anti-nociception while enhancing side effects [50–52]. This finding led to the development of ligands biased against βarrestin2 recruitment, which showed decreased side effects in animal models, and some of which are in clinical trials [53–55]. These novel approaches show that the simple ligand:receptor interactions profiled in this manuscript are only part of the story, and that the clinical opioids profiled here among other novel opioid drugs should be examined in a broader context of receptor interaction which could include pathway bias, preference for heterodimer interaction, or similar. A full understanding of these factors will likely be necessary for the successful creation of new opioid drugs without the side effects of current ligands.