Unique pharmacodynamic properties and low abuse liability of the μ-opioid receptor ligand (S)-methadone

(R,S)-methadone ((R,S)-MTD) is a racemic μ-opioid receptor (MOR) agonist comprised of (R)-MTD and (S)-MTD enantiomers used for the treatment of opioid use disorder (OUD) and pain. (R)-MTD is used as an OUD treatment, has high MOR potency, and is believed to mediate (R,S)-MTD’s therapeutic efficacy. (S)-MTD is in clinical development as an antidepressant and is considered an N-methyl-D-aspartate receptor (NMDAR) antagonist. In opposition to this purported mechanism of action, we found that (S)-MTD does not occupy NMDARs in vivo in rats. Instead, (S)-MTD produced MOR occupancy and induced analgesia with similar efficacy as (R)-MTD. Unlike (R)-MTD, (S)-MTD was not self-administered and failed to increase locomotion or extracellular dopamine levels indicating low abuse liability. Moreover, (S)-MTD antagonized the effects of (R)-MTD in vivo and exhibited unique pharmacodynamic properties, distinct from those of (R)-MTD. Specifically, (S)-MTD acted as a MOR partial agonist with a specific loss of efficacy at the MOR-galanin 1 receptor (Gal1R) heteromer, a key mediator of the dopaminergic effects of opioids. In sum, we report novel and unique pharmacodynamic properties of (S)-MTD that are relevant to its potential mechanism of action and therapeutic use, as well as those of (R,S)-MTD.


Introduction
Opioid medications are potent and e cacious analgesics, but their use can be associated with serious adverse effects such as tolerance, dependence, and respiratory depression. (R,S)-methadone ((R,S)-MTD) is an opioid medication used as an analgesic and a maintenance therapy for opioid use disorder (OUD) 1,2 . (R,S)-MTD is a long-acting µ-opioid receptor (MOR) agonist that is comprised of equal amounts of (R)-MTD and (S)-MTD enantiomers. The therapeutic properties of (R,S)-MTD are believed to be mediated by the pharmacological actions of (R)-MTD 3 , which is also prescribed alone as a maintenance therapy for OUD 4 .
Furthermore, (S)-MTD's a nity for the MOR is ~300 times greater than its a nity for the NMDAR 10,15 . Finally, (S)-MTD is an established MOR agonist, whereas its NMDAR actions involve noncompetitive antagonism 10 .
(R,S)-MTD and its enantiomers are classi ed as Schedule II controlled substances by the United States Drug Enforcement Administration. Nevertheless, (R,S)-MTD produces weaker activation of midbrain dopamine systems and has lower abuse liability when compared to other opioids 16 . The reduced dopaminergic effects of (R,S)-MTD are dependent on its unique, weak interaction with MOR-galanin 1 receptor (Gal 1 R) heteromers speci cally expressed in the ventral tegmental area (VTA). MOR-Gal 1 R in the VTA are known to mediate the activation of the dopaminergic system by opioids 16 , but the effects of (R) and (S) enantiomers of MTD at these heteromers are unknown.
In order to explore the analgesic and abuse liability pro les of (R,S)-MTD and its enantiomers, and to address the gaps in knowledge about these compounds, we performed an in-depth in vitro, in vivo and in silico pharmacological characterization of (R,S)-MTD, (R)-MTD and (S)-MTD. Our ndings provide a mechanistic basis for the differential in vitro and in vivo properties of the enantiomers, which may impact on their clinical utility.

Results (R)-MTD and (S)-MTD preferentially bind and activate MOR
Each enantiomer was tested for its ability to competitively inhibit binding or activity at a panel of 98 receptors and enzymes that are known targets for drugs of abuse and CNS medications. At 10 µM, (R)-MTD inhibited binding at several receptors (Fig. 1a), while at 100 nM, (R)-MTD inhibited binding only at MOR (98%) and SERT (68%). At 10 µM, (S)-MTD inhibited binding at several receptors, while at 100 nM, (S)-MTD inhibited binding only at MOR (79%). We derived each enantiomer's a nity at MOR using inhibition of [ 3 H]DAMGO binding in rat brain tissue. The Ki values obtained were 15.6 ± 0.1 nM for (R,S)-MTD, 7.5 ± 0.1 nM for (R)-MTD and 60.5 ± 0.1 nM for (S)-MTD (Fig. 1b).
Agonist-stimulated [ 35 S]GTPγS autoradiography in rat brain sections was used to examine the ability of (R,S)-MTD and its enantiomers to activate MOR (Fig. 1c). At 100 nM, only (R)-MTD increased [ 35 S]GTPγS recruitment in the caudate putamen (CPu) (171%) and nucleus accumbens (NAc) (151%) (Fig. 1d, Supplementary Fig. 1). Conversely, when evaluating catalepsy score in rats, (S)-MTD behaved as a partial agonist, unable to achieve maximal cataleptic effects, even at 100 mg/kg. The high-dose effect was ~60% of the maximal cataleptic effects of both (R)-MTD and (R,S)-MTD, which were observed at 3 and 10 mg/kg, respectively (Figs. 2b, Supplementary Fig. 1). The cataleptic ED 50 values (%MPE) for (R,S)-MTD, (R)-MTD and (S)-MTD were 2.1, 0.9 and 59.4 mg/kg, respectively, which were two-to three-fold higher than their analgesic ED 50 values. For the three drugs, the maximal cataleptic effect corresponded to the minimal dose required to produce signi cant hypothermia ( Supplementary Fig. 1). (S)-MTD-induced catalepsy did not observably saturate, and toxic higher doses were not employed. Overall, these experiments demonstrate a MOR agonistic pro le of (S)-MTD, with a lower potency and a possibly lower intrinsic e cacy compared to (R)-MTD. Finally, (R)-MTD and (S)-MTD did not differ in their propensity to interact with e ux transporters or in relation to CYP-dependent metabolism ( Supplementary Fig. 2), indicating that the two enantiomers would demonstrate similar metabolic pro les in vivo.
(S)-MTD exhibits lower abuse liability than (R)-MTD and (R,S)-MTD There is evidence that (R,S)-MTD is self-administered in humans 17 and rats 18 . However, the intravenous self-administration (IVSA) of (R)-MTD and (S)-MTD has not been reported. Moreover, depending on the dose administered, (R,S)-MTD can have either rewarding or aversive effects in rats 19 . IVSA, the standard preclinical approach for predicting abuse liability of drugs in humans 20 , was used to evaluate the reinforcing effects of (R,S)-MTD and its enantiomers in rats. First, we performed dose nding experiments to determine the dose of each drug that maintained IVSA. Rats exposed to various doses of (R)-MTD readily self-administered 50 µg/kg/infusion and consumed a maximum of ~2 mg/kg at the highest dose ( Supplementary Fig. 3). Rats exposed to (S)-MTD never acquired IVSA, even at high unit doses, and did not show any evidence of dose response. Nevertheless, when (R)-MTD-trained rats were switched to (S)-MTD, they showed reliable IVSA at 500 µg/kg/infusion (S)-MTD. The switched rats consumed a cumulative dose of ~30 mg/kg at the highest (S)-MTD dose ( Supplementary Fig. 3).
Next, we performed IVSA studies on another cohort of rats trained on either (R)-MTD (50 µg/kg/infusion), (R,S)-MTD (100 µg/kg/infusion), or (S)-MTD (500 µg/kg/infusion) ( Fig. 2c-e). For the rst 10 days of training, rats were on a xed-ratio 1 (FR1) schedule. During this time, rats in all three groups learned to discriminate the active from inactive lever. On the 11 th session, the schedule was increased to FR5 (5 presses for 1 infusion). Whereas rats trained on (R)-and (R,S)-MTD adjusted lever press rates to maintain stable infusion rates, rats trained on (S)-MTD did not. We then performed a dose response assessment of IVSA (Fig. 2f). Rats trained on (R)-or (R,S)-MTD displayed the typical inverted-U shaped dose-response curve, but rats trained on (S)-MTD showed no evidence of dose response. Rats given (R)-MTD showed peak infusion rates at 25 μg/kg, while rats given (R,S)-MTD peaked at 50 μg/kg. Notably, rats trained on (R,S)-MTD had more infusions at the peak unit dose than those on (R)-MTD, and the (R,S)-MTD curve was signi cantly shifted to the right indicating that larger drug amounts were required to reach the same level of reinforcement.

(R)-MTD and (S)-MTD do not produce MOR desensitization
Decreases in MOR density and desensitization contribute to the development of opioid tolerance 24 . In contrast to other MOR agonists, (R,S)-MTD does not produce tolerance, due to its ability to induce MOR internalization and recycling of re-sensitized MOR 25 . Thus, we examined to what extent repeated exposure to (R)-MTD (2 mg/kg, sc), (R,S)-MTD (4 mg/kg, sc), or (S)-MTD (30 mg/kg, sc) lead to changes in MOR density and G protein activation using [ 3 H]DAMGO and DAMGO-stimulated [ 35 S]GTPγS autoradiography. We found that neither (R,S)-MTD nor its enantiomers produced any effect on MOR density or G protein activity ( Supplementary Fig. 4).

Divergent pharmacodynamic effects of (R)-MTD and (S)-MTD at MOR in the VTA
In view of the apparent lower reinforcing e cacy of (S)-MTD in rats, we next examined effects of the (R,S)-MTD and its enantiomers on locomotor activity in mice. In contrast to rats, which become cataleptic following opioid exposure, mice display dose-dependent increases in locomotion 26-30 . This opioidinduced hyperlocomotion is dependent on dopaminergic activation [31][32][33] , namely the activation of MORs expressed on GABA afferents onto VTAdopamine neurons 34 . Additionally, locomotor activation can distinguish between full and partial MOR agonists, with partial agonists producing graded increases dependent on e cacy 35 . We found that (R,S)-MTD and (R)-MTD increased locomotion, but (S)-MTD did not (Fig. 3a-d). Speci cally, after 60 minutes of habituation, (R,S)-MTD produced a signi cant locomotor activation at 10 mg/kg (sc) but not at 3 mg/kg (sc), and 30 mg/kg (sc) was less effective than 10 mg/kg. An inverted U shape effect was also observed with (R)-MTD, which was more potent and effective at 3 mg/kg (sc). (S)-MTD did not produce any signi cant locomotor-activating effects, even at 100 mg/kg (sc; Fig. 3d). Moreover, when administered 15 min before (R)-MTD, (S)-MTD (10, 30 mg/kg, sc) dosedependently counteracted the locomotor-stimulating effect of (R)-MTD (10 mg/kg, sc; Fig. 3e-f).
In view of the apparent lower reinforcing e cacy of (S)-MTD in rats we next examined whether the enantiomers of (R,S)-MTD could stimulate MOR receptors in the VTA which are involved in opioid reinforcement. In particular, the weak interaction of (R,S)-MTD with MOR-Gal 1 R heteromers in rat VTA is thought to underlie its reduced dopaminergic activation and lower abuse liability 16 . Thus, we studied the effects of (R)-MTD and (S)-MTD perfusion into the rat VTA, using in vivo microdialysis. We recently showed that the intracranial perfusion of (R,S)-MTD in the VTA was less potent and e cacious than other opioids (e.g., morphine, fentanyl and DAMGO) at eliciting somatodendritic dopamine release 16 . This reduced effect was attributed to (R,S)-MTD's weak activation of MOR-Gal 1 R. Here, we show that local perfusion of (R)-MTD into the VTA produced a concentration-dependent increase in extracellular dopamine, with a signi cant increase at 3 μM and a larger increase at 10 μM (Fig. 3m, n), showing similar potency to that previously obtained with morphine 16 . In contrast, (S)-MTD did not induce any signi cant effect on extracellular dopamine levels when perfused up to 100 μM (Fig. 3n). Notably, the 100 µM concentration of (S)-MTD completely counteracted the effect of (R)-MTD on dopamine release (Fig. 3o).
In rat brain slices containing the VTA, 1 μM (R)-MTD signi cantly increased [ 35 S]GTPγS recruitment (121.6%, P = 0.0003), which was prevented by preincubation with (S)-MTD. As shown in Fig. 3p, 1 (Fig. 1d, e) and signi cantly counteracting the effect of R-MTD in the VTA, but not in the CPu or NAc (Supplementary Fig. 5).  (Supplementary Fig. 6) were analyzed with the 'dimer receptor model' (see Methods). In both cell lines and for the three compounds, a signi cantly better t was obtained for biphasic versus monophasic curves (p < 0.05 in all cases), indicating the preferred dimeric structure of MOR, forming heteromers or not forming heteromers with Gal 1 R, as previously shown 37 . Supplementary Table 1 shows that (R,S)-MTD, (R)-MTD and (S)-MTD bind MOR with two different a nities and negative cooperativity, both in MOR and MOR-Gal 1 R cells.
None of the obtained binding parameters show signi cant differences between MOR and MOR-Gal 1 R cells for any of the ligands, indicating that the MOR-Gal 1 R-dependent changes in the pharmacodynamic properties of (S)-MTD are not related to changes in its a nity for the MOR, but likely to its intrinsic e cacy. As expected, (S)-MTD had 14 times lower a nity than (R)-MTD in both cell types.
BRET experiments were performed to evaluate differences in the intrinsic e cacy of (R,S)-MTD, (R)-MTD and (S)-MTD at the MOR (Fig. 4a). MOR-Rluc and Gi-YFP constructs were transiently co-transfected to HEK-293T cells, and concentration-response curves of (R,S)-MTD, (R)-MTD, and (S)-MTD were analyzed for E max and EC 50 values (Fig. 4b-d). As expected, E max for (S)-MTD was signi cantly lower than for R-MTD (about 30% lower, Fig. 4c), and EC 50 for (S)-MTD was signi cantly higher than for R-MTD (about 10 times; Fig. 4d). Thus, relative to (R)-MTD, (S)-MTD is a partial and less potent MOR agonist.
CODA-RET experiments were then performed to determine whether MOR-Gal 1 R heteromerization might determine the speci c pharmacodynamic pro le of (S)-MTD (see Methods) (Fig. 4g). HEK-293T cells were co-transfected with MOR fused to nRLuc (MOR-nRLuc), Gal 1 R was fused to cRLuc (Gal 1 R-cRLuc) and Gi-YFP ( Fig. 4h-j). In the presence of Gal 1 R, no detectable increase of response (BRET ratio) could be obtained with (S)-MTD, while the dose-response curve of (R,S)-MTD was shifted to the right, with an EC 50 value signi cantly higher than for R-MTD (~10-fold; Fig. 4j). These results, therefore, indicate that (S)-MTD, but not (R)-MTD, changes its pharmacological pro le and loses its e cacy for the MOR when forming heteromers with Gal 1 R. This implies that the changes in the pharmacological pro le of (R,S)-MTD within the MOR-Gal 1 R heteromer, as previously described 16,37 , depend on the modi ed pharmacodynamic properties of (S)-MTD. Consistent with this, increasing concentrations of (S)-MTD progressively counteracted the effect of a minimal concentration with maximal effect of (R)-MTD (10 μM) (Fig. 4k). At the highest concentration of (S)-MTD (1 mM), the effect of (R)-MTD was completely blocked, and CODA-RET measurements were not signi cantly different from basal values (Fig. 4l). As a control, the same design was applied with BRET experiments with the MOR alone. In this case, the highest concentration of (S)-MTD (1 mM) did not counteract the effect of (R)-MTD (10 μM) (Fig. 4e, f), and only decreased its effect to the expected maximal level of e cacy of (S)-MTD. These results, therefore, complement those obtained with in vivo and ex vivo experiments in the VTA (microdialysis and [ 35 S]GTPγS) and with locomotor activation and psychomotor sensitization in mice, and provide strong evidence for their mediation by MOR-Gal 1 R heteromers.
Molecular mechanism of the MOR-Gal 1 R-dependent pharmacodynamic pro le of (S)-MTD The recently reported structure of MOR in complex with fentanyl 38 can be used as a template to understand the pharmacological differences among the enantiomers of (R,S)-MTD at the molecular level. We rst performed ve replicas of unbiased 1 μs molecular dynamics (MD) simulations of (S)-MTD and (R)-MTD docked into the MOR monomer (see Methods). Root-mean-square deviations (rmsd) of the simulations show that the proposed docking models of (S)-MTD and (R)-MTD remained highly stable ( Supplementary Fig. 7). In these models, the protonated amine of (S)-MTD and (R)-MTD forms the conserved ionic interaction with D149 3.32 , and both phenyl groups adopt a "V" shaped conformation in the orthosteric binding site but, importantly, with signi cant differences (Fig. 4m, n). In (R)-MTD, both phenyl rings point up to form T-shaped aromatic interactions with H299 6.52 and W320 7.35 , whereas the phenyl rings of (S)-MTD point down to interact with W295 6.48 in a "sandwich" mode in which the aromatic Trp ring is between both phenyl rings ( Supplementary Fig. 7 shows a detailed analysis of the binding modes). We suggest that the phenyl ring of (S)-MTD positioned between W295 6.48 and TM 5, absent in (R)-MTD, restricts the necessary movement of W295 6.48 for activation 38,39 , which explains the decreased ability of (S)-MTD to activate MOR.
To understand the inability of (S)-MTD to activate MOR in the presence of Gal 1 R, at the molecular level, we rst needed to computationally model the MOR-Gal 1 R heteromer ( Supplementary Fig. 8). Previously reported bimolecular uorescence complementation (BiFC) and total internal re ection uorescence (TIRF) microscopy experiments, in the presence of synthetic peptides corresponding to different TM domains of MOR and Gal 1 R, revealed that the interface for the MOR-MOR homodimer changed from the TM 5/6 to the TM 4/5 interface in the absence and presence, respectively, of Gal 1 R 37 . Thus, we hypothesized that the MOR-MOR homodimer interacting via the TM 4/5 interface disables (S)-MTD to activate MOR. To test this hypothesis, we performed ve replicas of unbiased 1 μs MD simulations of the MOR-MOR homodimer, constructed via both the TM 5/6 (not interacting with Gal 1 R) and TM 4/5 (interacting with Gal 1 R) interfaces, in complex with Gi (see Methods and Supplementary Fig. 8). These simulations showed that, in contrast to the TM 5/6 interface, TM 5 of the active Gi-bound protomer moved the extracellular part of TM 5 inward in the TM 4/5 interface. Importantly, this movement of TM 5 relocated the position of the key V238 5.42 (Supplementary Fig. 8). Fig. 4o, p summarizes these ndings. In the TM 5/6 interface (Fig. 4o), W295 6.48 is only partially restricted (depicted as exible ellipses) by the phenyl ring of (S)-MTD because the dynamic behavior of the ligand is not fully constrained by the partner protomer (depicted as exible arrows). In contrast, in the TM 4/5 interface (Fig. 4p), the inward movement of V238 5.42 fully constrained (depicted as a single arrow) the phenyl ring of (S)-MTD, maintaining W295 6.48 in the inactive conformation (depicted as a single ellipse).

Discussion
(R,S)-MTD is a DEA Schedule II controlled medication with known abuse liability that is prescribed for pain management and treatment of OUD. However, the individual contributions of its enantiomers to its abuse liability and clinical e cacy are not well understood. We found that both (R)-MTD and (S)-MTD produced full agonistic effects on analgesia but only (R)-MTD was reliably self-administered. These ndings are in agreement with results from recent studies indicating that (S)-MTD does not lead to reinforcing effects, physical dependence nor withdrawal signs in rats 40 and that it lacks opioid effects, or withdrawal signs and symptoms in humans 5 , suggesting that the abuse liability of (R,S)-MTD is mediated by (R)-MTD and not by (S)-MTD. Indeed, our data indicate that (S)-MTD can attenuate the abuse liability of (R)-MTD under some conditions.
Although some experimental and clinical effects of (S)-MTD have been attributed to its NMDAR antagonism 5,7-9,41 , we demonstrate here that, at pharmacologically signi cant doses, (S)-MTD does not interact with NMDARs in vivo. Instead, (S)-MTD signi cantly occupies MORs at doses that promote the classical behavioral effects of opioids in rats: analgesia, catalepsy, and hypothermia. For example, the effective dose at which (S)-MTD produced analgesia is within the range of doses used to produce antidepressant-like effects in rats 6,7 . For both (R)-MTD and (S)-MTD, the predicted free brain concentrations coincided with their in vitro MOR a nities as well as their capacity to selectively occupy MORs in vivo. In contrast, we failed to detect in vivo NMDAR occupancy at the same doses where MOR occupancy was observed. Therefore, we can conclude that (S)-MTD selectively binds MORs at brain concentrations relevant to its analgesic and antidepressant-like e cacy. Thus, the currently assumed role of NMDAR blockade in the purported antidepressant effects of (S)-MTD should be reassessed and explained in the frame of its MOR agonistic properties.
We demonstrated that (S)-MTD does not promote activation of the dopaminergic system, likely due to the inability of (S)-MTD to activate MOR-Gal 1 R in the VTA, previously shown to mediate the dopaminergic effects of opioids 16 . On the other hand, (R)-MTD promoted a signi cantly stronger activation of the VTA dopaminergic system than the reported effect of (R,S)-MTD 16 . The speci c lack of effect of (S)-MTD was due to its loss of intrinsic e cacy for MOR-Gal 1 R, which also explains its antagonism of (R)-MTDinduced effects in the VTA including dopamine release, [ 35 S]GTPγS recruitment, and locomotor activation.
The signi cant analgesic and cataleptic effects of (S)-MTD indicate dopamine-independent mechanisms, not mediated by MOR-Gal 1 R. As opposed to neuroleptic-induced catalepsy, opioids do not induce catalepsy by inhibiting striatal dopaminergic neurotransmission, but possibly by inhibiting the MORexpressing striatal and pallidal GABAergic neurons that project to the output structures of the basal ganglia 30,42,43 . These two functionally opposite MOR-dependent effects, locomotor activation and catalepsy, are both present but differentially dominate in mice and rats, respectively. In fact, locomotor activation can also be elicited in rats with the intracranial injection of opioids in the VTA 44 , and catalepsy has been reported with relatively high doses of opioids in mice 45  MTD, are consistent with the dopaminergic hypothesis. Nevertheless, at high doses, (S)-MTD was able to substitute for (R)-MTD in rats trained on (R)-MTD, which support the involvement of additional nondopaminergic mechanisms in opioid reinforcement. Importantly, the dose response of (R,S)-MTD IVSA was qualitatively different from that of (R)-MTD, with a signi cantly higher peak and a pronounced shift to the right. This could be explained by (S)-MTD counteracting the effect of (R)-MTD at su ciently high doses of (R,S)-MTD.
One potential adverse effect of (R,S)-MTD use is that it can cause cardiac arrythmia 50 . This has been attributed to high concentrations of the drug and perhaps to the presence of (S)-MTD, which one study reported blocks the Ether-à-go-go-Related Gene 1 (hERG) channel 3.5-fold more potently than (R)-MTD 51 . However, the stereoselective contribution of (R,S)-MTD enantiomers to these effects has been