3.1 Allosteric modulation of the opioid receptors via an allosteric binding site (ABS)

The discovery of small molecules, capable of modulating the response associated with orthosteric ligand binding, while targeting spatially distinct binding sites, opened up new avenues in the realm of GPCR signalling modulation. When druggable, these so-called allosteric binding sites (ABS) provide useful alternative approaches to target the GPCRs. The ABS is typically located towards the extracellular domains, the usually less conserved regions, of the receptor.^70-75 It is thus an ideal target when aiming for subtype selectivity within the same GPCR family. A major advantage of allosteric ligands is that they are, in essence, modulators, which implies that most of them do not show any effect on receptor signalling on their own accord but influence that of an orthosteric ligand. This then allows for a fine-tuning of receptor response while maintaining spatial and temporal fidelity of endogenous GPCR agonist activity. Studies have also shown that allosteric ligands can show functional selectivity or signalling bias.^76-78 Allosteric modulators can therefore favorably influence receptor response away from pathways mediating side effects while reinforcing the desired therapeutic effect.

While structural data on the ABS of the opioid receptors are currently not available, MD studies have identified a plausible allosteric binding pocket located towards the extracellular side of the DOR.⁷⁹ The allosteric binding pocket is surrounded by TM1, TM2, TM7 helices, and part of TM6, according to metadynamics simulations results which were also supported by site directed mutagenesis studies of DOR residues deemed important for maintaining the allosteric response. This allosteric binding site is also comparable to that of the human M2 muscarinic receptor previously revealed by X-ray crystallography.⁸⁰

Allosteric ligands can be classified into three principal categories: positive allosteric modulators (PAMs), negative allosteric modulators (NAMs) and silent allosteric modulators (SAMs). PAMs enhance the affinity and/or efficacy of an orthosteric ligand while NAMs show the opposite effect. SAMs or neutral allosteric modulators bind at the ABS without showing any influence on an orthosteric ligand's activity. They are thus useful tools in probing the activity of a PAM or a NAM.

The earliest known NAM at the opioid receptors is the sodium ion. The first evidence of the negative allosteric nature of sodium ions was shown in in vitro assays involving mouse brain extracts. Sodium inhibited the binding of orthosteric agonists to the MOR while exhibiting no effect on orthosteric antagonist binding.⁸¹ It is now known that the sodium ion stabilizes the inactive conformation of opioid receptors or GPCRs.^{62, 82} The sodium ion occupies a binding site which is conserved across many GPCRs. Crystal structures of inactive GPCRs show that the Na⁺ binding site is buried within the 7-TM bundle and involves a coordination with an aspartate residue (D^2.50) and a serine residue (S^3.39) and an extensive water network across TM II, III, VI and VII (Figure 7).⁸³ The same Na⁺ ion binding pocket is not present in the crystal structures of active GPCRs (Figure 7), likely due to movement of the transmembrane domains occurring during receptor activation which impairs the accommodation of sodium and its corresponding water molecules.³⁷

Other putative NAMs of the opioid receptors include the cannabinoid type-1 receptor (CB1) agonist cannabidiol (19, Figure 8) and the potent KOR agonist salvinorin A (20, Figure 8). Cannabidiol (19) is a NAM of both MOR and DOR while Salvinorin A (20) elucidates its NAM activity only at the MOR.^{84, 85} In kinetic binding studies on rat cerebral cortex membrane homogenates, cannabidiol (19) accelerated the dissociation of the agonist ³H-DAMGO from the MOR by a factor of 12 and [³H]-naltrindole from the DOR by a factor of 2.⁸⁴ Cannabidiol (19) also showed a Hill coefficient corresponding to positive homotropic cooperativity, indicating that it effectively binds to an allosteric site.⁸⁴ However, it is important to note that the allosteric properties of cannabidiol (19) were observed at doses much higher than its activity in vivo. In assays of Chinese hamster ovary (CHO) cells expressing the cloned human MOR, the KOR agonist salvinorin A (20, Figure 8) partially inhibited both the binding efficacy and affinity of the MOR agonists [³H]-DAMGO and [³H]-diprenorphine, respectively.⁸⁵ A right shift in their dose–response curves with a concomitant decrease in E_max were observed for both orthosteric ligands in the presence of increasing concentrations of salvinorin A (20), a typical profile for NAMs.⁸⁵ In addition, saturation binding studies involving the same orthosteric ligands, showed that salvinorin A (20) lowered the MOR B_max and increased the K_d in a dose-dependent nonlinear fashion.⁸⁵ This is in sharp contrast with the dose-dependent linearity usually observed with competitive inhibitors. Results of kinetic binding experiments strongly suggest that salvinorin A (20) increased the dissociation rate of [³H]-DAMGO and [³H]-diprenorphine binding to the MOR.⁸⁵ Salvinorin A (20) also behaved as a noncompetitive inhibitor of DAMGO when tested in the [³⁵S] GTP_γS functional assay.⁸⁵

PAMs and SAMs of the opioid receptors have also been identified in a high throughput screen.²⁸ Initial screening using a β-arrestin recruitment assay revealed two potential PAMs selective for the MOR, BMS-986121 (21) and BMS-986122 (22) (Figure 8), respectively. The two compounds did not exhibit significant β-arrestin recruitment on their own but significantly increased the β-arrestin response of the MOR agonist endormorphin-I (EM-1).²⁸ BMS-986121 (21) augmented the β-arrestin response by 20 nM EM-1 to a maximal effect (E_max) of 76% of the response by a maximally effective concentration (1 µM) of EM-1 with an EC₅₀ of 1.0 μM.²⁸ BMS-986122 (22) also produced a similar potentiation of the effect EM-1 to 83% of the maximal observed EM-1 response with an EC₅₀ of 3.0 µM.²⁸ These compounds were further evaluated in two additional functional assays, namely, a cAMP-accumulation assay, and a G-protein activation assay using [³⁵S]GTPγS binding. BMS-986121 (21) and BMS-986122 (22) considerably increased the inhibition of forskolin-induced adenyl cyclase activity produced by an EC₁₀ (30 pM) concentration of EM-1 in CHO-MOR cells with EC₅₀ values of 3.1 μM and 8.9 µM, respectively.²⁸ In G-protein activation [³⁵S]GTPγS-binding studies in membranes from C6 glioma cells stably expressing recombinant MOR, addition of BMS-986122 (22) and BMS-986121 (21) resulted in a leftward shift in the dose–response curve of the MOR agonist DAMGO by factor of 7 and 4, respectively.²⁸ In an attempt to establish some SAR of the BMS chemical series, 2 analogues of BMS-986122 (22), namely, BMS-986123 (23) and BMS-986124 (24) (Figure 8), were identified as SAMs of the MOR. These two compounds were shown to inhibit the PAM activity of 12.5 μM BMS-986122 (22) in a β-arrestin recruitment assay in the presence of 30 nM EM-1.²⁸

In light of the observed allosteric behavior of BMS-986122 (22) (Figure 8), MD studies were carried out to gain more insight into its mechanism of action.⁸⁶ It was found that in the hypothetical ABS of the MOR, BMS-986122 (22) stabilizes the favorable interactions between the receptor and the full MOR agonist R-methadone while inhibiting the binding of the allosteric sodium ion.⁸⁶

Kandasamy et al. investigated the antinociceptive effects of BMS-986122 (22, Figure 8) in mouse models of acute and inflammatory pain to test their hypothesis that an MOR agonist can potentiate the activity of endogenous opioids released during pain.⁸⁷ In vitro, BMS-986122 (22) showed no agonist activity at endogenously expressed MOR; it did not stimulate G protein in mouse brain homogenates, including periaqueductal gray (PAG).*⁸⁷ This is in agreement with the findings made in heterologous cell systems expressing MOR, where the PAM does not activate the heterotrimeric G protein or recruit β-arrestin.²⁸ However, low activity (EC50 ~ 0.1 mM) of BMS-986122 (22) was observed at the more amplified pathway, adenyl cyclase, and in CHO cells overexpressing MOR.^{28, 87} In C57BL/6 mice, the PAM showed a dose-dependent antinociception when centrally administered.⁸⁷ The duration of the antinociceptive response was short-lived but reversed by the orthosteric antagonist β-FNA (8, Figure 1).⁸⁷ While this response could be due to BMS-986122's weak potency and efficacy, the PAM does not bind at the OBS.⁸⁷ BMS-986122 (22) rather mediates its action by potentiating the activity of endogenous opioid peptides which are then hindered by β-FNA (8) binding to the OBS.⁸⁷ This hypothesis was confirmed by Kandasamy et al. by using a low dose of enkephalinase inhibitor, RB-101, which prolongs the half-life of endogenous peptides in vivo, thereby increasing their available concentration.⁸⁷ RB-101 showed minimal efficacy on its own accord but afforded strong antinociception in the presence of BMS-986122 (22).⁸⁷

BMS-986122 (22, Figure 8) promoted swim-stress antinociception which was reversible with naloxone (7, Figure 1), further implying the involvement of endogenous opioid peptides and MOR.⁸⁷ BMS-986122 (22) showed antinociceptive responses in mouse models of inflammatory pain.⁸⁷ These observations suggest that a continuous peptide release is necessary for the effectiveness of BMS-986122 (22) and that endogenous opioid peptide levels in resting states are not sufficiently high to be influenced by allosteric modulators unless there is a nociceptive insult.⁸⁷ This provides a basis for the therapeutic application of MOR PAMs whereby antinociception can be achieved by potentiating the activity of endogenous opioid peptides without the administration of the classical opioid agonists.⁸⁷

More recently identified PAMs of the MOR are MS1 (25) and ignavine (26) (Figure 8).^{29, 88} MS1 (25) was the most active PAM in a β-arrestin recruitment assay in the search of new PAMs based on the chemical scaffold of BMS-986121 (21) and BMS-986122 (22) (Figure 8). In a competition assay, 10 μM MS1 (25) increased the binding affinity of R-methadone (16, Figure 1) by sevenfold and augmented the potency of the latter by over fourfold in a G-protein activation assay.²⁹ MS1 (25) also displayed strong probe dependency; it had no effect on either the binding affinity or potency of other MOR agonists such as DAMGO and EM-1.²⁹ Nonetheless, MS1 (25) could enhance the potency of morphine to that of a full MOR agonist while showing no influence on its binding affinity.²⁹ Ignavine (26, Figure 8), a diterpene alkaloid and component of Aconitum tuber, affects the MOR activity in a unique manner. In vitro functional assays indicated that ignavine (26) exhibited bidirectional activity in a concentration-dependent manner. An increase in MOR activity produced in the presence of 1 μM of DAMGO was observed at low concentrations of ignavine (26) (1 μM) both in a receptor internalization assay and an intracellular cAMP assay.⁸⁸ On the other hand, higher concentrations (>10 μM) of ignavine (26) led to an inhibition of MOR response in the same assays.⁸⁸ Such results were obtained with other MOR agonists, including, morphine and the endogenous agonist EM-1 but not with the MOR antagonist naloxone (7, Figure 1).⁸⁸ Similar observations were also made in an in vivo analgesia study of ignavine (26). The analgesic response of ignavine (26) was probed by the tail-flick and tail-pressure tests in normal mice. A maximum response was shown at a dose of 0.1 mg/kg which was lowered at higher doses of ignavine (26).⁸⁸

Pryce et al. screened 21 derivatives of MS1 (25, Figure 8) for PAM activity in MOR/Gα_i/o-mediated cAMP inhibition and β-arrestin recruitment assays, in the presence of increasing concentrations of MOR agonists endomorphin-1, DAMGO, or l-methadone.⁸⁹ They identified compound 27 (Figure 8) which showed the most significant allosteric improvement of G_i/o activity when comparing EC₅₀ values at 0 and 3 µM concentrations of allosteric modulators in the cAMP assay with all three MOR agonists.⁸⁹ Of note, MS1 (25) and 27 did not exhibit a significant PAM effect on the endogenous endomorphin-1 but more pronounced effects on exogenous MOR agonists.⁸⁹ In mice models of peripheral inflammation, Pryce et al. observed significant antinociceptive effects in the presence of MS1 (25) and 27 at opioid doses lower than those used to inhibit thermal hyperalgesia.⁸⁹ In a model of postoperative pain whose induced thermal hypersensitivity can be treated with 3 mg/kg of morphine, MS1 (25) and 27 induced antinociception at lowered doses of opioids (2 mg/kg).⁸⁹ Both MOR PAMs exhibited probe dependence which differed between in vivo and in vitro experiments. 27 enhanced the antinociceptive response to morphine and oxycodone in both female and male mice in different pain models which is also reflective of the results obtained in the BRET assays.⁸⁹ However, contrary to the in vitro data, 27 did not potentiate the antinociceptive effects of methadone in the hot place assay.⁸⁹ In vitro data suggest a more significant probe dependence for MS1 (25) than 27.⁸⁹ Both 27 and MS1 (25) showed a rather balanced effect on the improvement of potency of the agonists l-methadone, oxycodone and morphine at the G_αoA or β-arrestin 2 pathways in BRET assays.⁸⁹ These results contradict the earlier observations that MS1 (25) behaved as a β-arresting-biased MOR PAM.²⁹ However, the bias quantification methods employed by Pryce and colleagues are reported to be more accurate. 27 and MS1 (25) also showed promising side effect profiles as they did not worsen the development of somatic withdrawal symptoms or respiratory depression and did not favor analgesic tolerance.²⁹

A structure–activity relationship study using a β-arrestin recruitment assay led to the first reported series of DOR PAMs, from which BMS-986187 (28, Figure 8) was the most potent PAM with a 100-fold selectivity for the DOR over the MOR.³⁰ When evaluated in four different assays, namely, β-arrestin recruitment assays, forskolin-stimulated cAMP accumulation assays, [³⁵S]GTP_γS binding assay and ERK1/2 phosphorylation assays primarily using CHO cells, BMS-986187 (28) potentiated the activity of the DOR orthosteric agonists leu-enkephalin (Leu-Enk), SNC80 and TAN67 and completely inhibited the propagation of cAMP when administered on its own (at concentrations higher than 3 µM).³⁰ The latter observation suggests that BMS-986187 (28) might be an effective DOR-selective PAM-agonist.

3.2 Intramolecular allosteric crosstalk within opioid receptor dimers

In the late 90 s and early 2000s, the idea that GPCRs exist as functional oligomers in the form of dimers and were the “actual targets” of neurotransmitters and small molecules binding at the OBS started to gain in popularity.^{90, 91} The DOR was the first opioid receptor for which dimerization was described, and the existence of the DOR dimers was later confirmed by time-resolved fluorescence resonance energy transfer (FRET) and bioluminescence resonance energy transfer (BRET).^{31, 32} The first structural evidence of opioid receptor dimerization was reported by Granier and Kobilka.³⁹ They observed that MOR receptor molecules associated into pairs along the crystallographic two-fold axis through two distinct interfaces, one mediated by TM1, TM2 and helix 8 and another more extensive interface consisting of TM5 and TM6.³⁹ More recently, Zhuang et al. also observed that the MOR-G_i complexes with the opioid agonists, morphine, fentanyl (14) and PZM21 (4) existed as anti-parallel dimers.⁵⁴ However whether these dimeric complexes are merely artifacts of crystal structures remain unclear. Today, GPCR dimers and oligomers are well-characterized with a vast library of ligands, including peptides and small molecules, specifically designed to target these macromolecular agglomerations.^92-97

Dimers can be further classified as homomers where the monomers are identical or heteromers where the monomers are of different subtypes or of different receptor classes. Dimerization naturally provides another level of allosteric modulation of these receptors, whereby a monomer, when bound by an agonist or an antagonist, can influence the binding properties of the partner monomer within a dimer.⁹⁸ The additional level of three-dimensional molecular organization brought by dimerization confers different functional and pharmacological properties to the dimers in comparison to their monomeric counterparts.^{56, 65, 99} Over the years, dimers have established themselves as targets of choice when probing for subtype or functional selectivity unattainable by solely aiming for the OBS of monomers.

Opioid receptors have been reported to form both homomers and heteromers in which crosstalk is observed between 2 different receptor subtypes or between an opioid receptor subtype and another GPCR receptor.^{65, 100, 101} Homomers represent interesting targets when aiming for an increase in affinity and/or potency through positive allosteric cooperativity.¹¹ Heteromers of the opioid receptors in various subtype combinations are described in the literature, e.g., MOR/DOR, MOR/KOR, or DOR/KOR, each showing distinct functional properties.^{65, 100} Additionally, opioid receptors have also been identified in dimeric complexes with other class A GPCRs such as the cannabinoid receptor (CB1), the neurokinin receptor (NK1), and the dopaminergic receptors (D₂R-like).¹⁰⁰ Such heteromers provide new avenues for multi-target approaches, particular useful when the aim is to avoid certain side effects.

However, the existence of heterodimers in vivo is still a subject of debate. The presence of opioid heterodimers has primarily been shown in vitro using co-immunoprecipitation, BRET and FRET assays with a few studies aiming to establish their existence and functional relevance in vivo (see reviews by Ugur et al.⁶⁹ and Zhang et al.⁶⁸). Heterologous systems usually make use of overexpression which does not exclude the possibility that the observed heterodimers could be mere artifacts. Research in the field of heterodimers of opioid receptors are still at the early stages. This further reinforces the need of heterodimer-selective ligands to investigate the translation potential of heterodimers as therapeutic targets.

3.3 Bivalent and bitopic ligands of the opioid receptors

Opioid receptors were the first ever GPCRs to be probed by bivalent ligands in 1982 by Philip Portoghese, even before the concept of GPCR oligomerization was well-accepted. At the time, Portoghese postulated that the sum effect of pharmacophores combined over a linker would exceed the efficacy of the individual units.¹⁰² His pioneering work provided the first evidence of opioid receptor dimers and has since paved the way for the design and synthesis of bivalent ligands of other GPCRs.¹⁰²

Structurally, bivalent ligands consist of two primary pharmacophore units, connected over a linker, each aiming for the OBS of the monomers within a dimeric complex (Figure 9A). They may either be homobivalent ligands where the phamacophore units are identical or heterobivalent when the pharmacophore units are different. Bivalent ligands distinguish themselves from bifunctional ligands consisting of a single pharmacophore capable of interacting with multiple targets. One of the earliest examples of a bifunctional ligand of the opioid receptors is nalorphine (5), which is now known to be a MOR antagonist and KOR agonist.^{103, 104} Nalorphine (5) is able to exhibit antinociceptive effects when acting at the KOR but also can reverse those of morphine at the MOR (for a review of bifunctional ligands of opioid receptors see Rehrauer and Cunningham³⁵).^{103, 104}

While Portoghese initially designed bivalent ligands with the aim to target the OBS of two receptors located in close proximity to each other, he did not preclude the possibility of the second pharmacophore subunit of these ligands binding to an accessory binding site within the same receptor. Ligands with the latter binding characteristics are now the so-called bitopic ligands. From a structural design perspective, bitopic ligands are a logical extension from bivalent ligands, in that they also consist of two pharmacophore units, linked over a spacer, one usually fitting within the conserved OBS and the other, the less-conserved ABS (Figure 9B).

In recent years, rationally designed bitopic ligands have been used to probe class A GPCRS like the dopaminergic receptors,^105-107 the sphingosine-I-phosphate GPCRs^{108, 109} and the adrenergic receptor^{110, 111} providing useful insight into the intrinsic signalling mechanism of this class of receptors. In our research group, endeavors into the field of bitopic ligands have primarily been focused on the dualsteric targeting of the muscarinic acetylcholine receptors (mAChRs). The latter were “ideal” targets for rational bitopic ligand design considering their well-detailed allosteric pharmacology in literature and their significant physiological roles. We have designed and synthesized bitopic ligands showing a range of activities at mAChRs (from agonists to partial agonists^112-114 to antagonists¹¹⁵) with improved affinities and receptor-subtype or functional selectivities.¹¹⁶

Bivalent and bitopic ligands deliver where traditional orthosteric ligands based solely on one pharmacophore have failed. The extensive work by Portoghese has shown how subtype selectivity can effectively be achieved using a bivalent targeting approach. Bitopic ligands with their concomitant occupation of both the OBS and less-conserved ABS also tend to show more subtype selectivity. Both bivalent and bitopic ligands may display improved affinity and/or efficacy. Bivalent and bitopic ligands, owing to their highly specific binding interactions, also show functional selectivity, positive or negative allosteric modulation or cooperativity.

3.4 Understanding bivalent and bitopic binding modes

When the two pharmacophores are linked through a spacer of appropriate length, bivalent ligands are expected to exhibit potencies higher than the combined effect of their individual monovalent ligands. This synergy occurs because of the thermodynamic advantage associated with sequential binding. Univalent binding of the first pharmacophore places the second pharmacophore, by virtue of being tethered, in close proximity to the neighboring unoccupied OBS (Figure 10B).²⁶ Thus, the binding of the second pharmacophore resulting in a bivalent binding mode (Figure 10C) should be favored over the second univalent binding of a second ligand (Figure 10D).²⁶ In situations where the monomers within a dimer are negatively allosterically coupled, synergy may not be observed as binding enhancement might then be hindered.²⁶

In his early studies of homobivalent ligands, Portoghese observed potencies and affinities higher than double those of the corresponding monovalent ligands, suggesting a bivalent binding mode.^{117, 118} Studies of more recent bivalent ligands of other GPCRs, including the serotonine,¹¹⁹ muscarinic,¹²⁰ histamine¹²¹ and dopamine¹²² receptors indicate that homobivalent ligands generally show an increase in binding affinities compared to their monovalent pharmacophores. With heterobivalent ligands of mixed agonist/antagonist pharmacophores, an enhancement of binding affinity is not usually observed. Such a combination of pharmacophores usually requires careful investigation in multiple assays to obtain more information about their pharmacology. Another approach is to investigate the bivalent ligands in heterologous cell systems co-expressing the target GPCRs. If they efficiently bridge a dimer, bivalent ligands may show a higher selectivity for the co-expressed receptors than the singly expressed receptors.^123-125

Careful analysis of Hill slopes of the binding curves of bivalent ligands in radioligand binding assays can confirm their bivalent binding modes. Kuhhorn et al. observed that the Hill slopes of the competitive binding curves of the monovalent antagonists are usually close to 1.¹²⁶ Binding of bivalent antagonists showed an increased Hill slope of 2, suggesting a bivalent binding mode.¹²⁶ Bivalent agonists displayed steeper binding curves in comparison to their monovalent pharmacophores.^{95, 127, 128} The displacement of the monovalent agonists by radiolabelled antagonists results in shallow curves of Hill slopes of 0.5 to 0.7.^{95, 127, 128} On the other hand, the Hill slopes of bivalent agonists significantly increased to 1.3 to 1.4.^{95, 127, 128} Another feature of bivalent ligands is that they show biphasic competition binding curves with two affinity values, K_high and a K_low, corresponding to the bivalent and monovalent binding modes, respectively.^129-131

When bitopic/dualsteric binding occurs (Figure 11A), the orthosteric moiety directs the binding affinity. In the presence of a radioligand, competition for the OBS will be observed and the bitopic ligand will show pronounced inhibitive action on radioligand binding.^132-134 The bitopic ligand can however adopt a solely allosteric binding mode (Figure 11C). In that case, the binding of a radiolabelled orthosteric ligand might be favored through the formation of a ternary complex (Figure 11D).¹³² Provided the ternary complex is predominant in vitro, a concentration-dependent increase in radioligand binding may be observed.¹³² Bock et al. used the bitopic ligand iper-6-naph to illustrate how different ligand binding modes can control agonist efficacy at the muscarinic receptors.¹³⁵ They were able to show that iper-6-naph can cause a concentration-dependent increase of the radioligand [³H] NMS in displacement experiments, direct evidence of an allosteric binding mode of iper-6-naph.¹³⁵

The bitopic ligand can also adopt a multiple binding mode, in a “flip-flop” manner between two or more receptor topographies, a purely allosteric and a purely orthosteric binding (Figure 11E).^{133, 134} This usually occurs where the pharmacophores possess low binding affinities or when the linker lengths are not optimal to afford bridging of the OBS and ABS. This “flip-flop” binding mode is experimentally indistinguishable from a bitopic binding mode.¹³⁴ Mutagenesis studies can be used to validate the binding mode of the bitopic ligand.^{133, 134} Mutations of key residues important for binding within the OBS and ABS generally result in a decrease in the binding affinity of the bitopic ligand.¹³³ In their study of bitopic ligands of the M₂ muscarinic receptors, Antony et al. generated M₂ muscarinic receptors with loss-of-function mutations in either the OBS or the ABS. They observed a loss in binding affinity for the bitopic ligand at the mutants.¹³² Moreover, mutations in the OBS led to a decreased inhibition of radioligand binding, suggesting that the bitopic ligand exhibited primarily allosteric behavior, enhancing the formation of the ternary complex.¹³² On the other hand, mutations in the ABS favored orthosteric behavior of the bitopic ligand, resulting in predominantly in the inhibition of radioligand binding.¹³²

Molecular validation of the bitopic binding mode is not always possible, especially when the exact topography of the ABS is not known. In such cases, the effect on the binding affinity of the bitopic ligand in response to coaddition of an excess of the allosteric pharmacophore can be investigated.^{134, 136} When the bitopic ligand simultaneously occupies both the OBS and ABS, addition of the allosteric pharmacophore will have no pronounced effect on its binding affinity. On the contrary, if the ideal occupation of both binding sites is not afforded by the bitopic ligand, the excess allosteric pharmacophore might have an effect on the binding affinity.

3.5 Bivalent ligands of the opioid receptors

3.5.1 Homobivalent ligands of the opioid receptors

Philip Portoghese's extensive work on the bivalent ligands of the opioid receptors, culminating in the discovery of a KOR- and DOR-selective ligands, norbinaltorphimine (31, Figure 13) and naltrindole (9, Figure 1), respectively, is of remarkable significance. In an attempt to further knowledge of the effects mediated by each opioid receptor subtype in the late 80 s, Portoghese designed and synthesized bivalent ligands based on both agonist (oxymorphine) and antagonist (naltrexone, 6) pharmacophores (Figure 12).¹¹⁷ His rationale behind subtype-selective bivalent ligands derived from the assumption that spatial arrangement of proximate binding sites might be unique for each opioid receptor subtype based on their heterogeneity. Accordingly, Portoghese also hypothesized that the combined effects of bridging the gap between neighboring binding sites would be twice or more than that of a monovalent ligand.

Amongst his noteworthy findings at the time was the observation that homobivalent ligands (of linker lengths 0-4) based on the oxymorphine scaffold (compounds 29a-d, Figure 12) possessed agonist potencies and affinities more than twice of those of the monovalent agonists when tested in guinea pig ileum longitudinal muscle (GPI) and mouse vas deferens (MVD) preparations.¹¹⁸ Similar observations were made for bivalent antagonist ligands based on the naltrexone scaffold (compounds 30a-e, Figure 12) at both the MOR and KOR which strongly suggested concomitant engagement of two vicinal binding sites. Of note was the effect of linker length at the MOR and KOR. While peak potency was observed for the agonists when the spacer consisted of glycyl units (n = 2) at the MOR (Table 1, 29c) maximum binding selectivity was obtained for the antagonist with the shortest spacer (n = 0) at the KOR (Table 2, 30a).¹¹⁸ This was evidence of the differential molecular organization within different receptor subtypes. To further investigate whether the bivalent antagonist ligands bind two vicinal recognition sites or the OBS and an accessory binding pocket, Portoghese synthesized the corresponding meso-isomer of the most potent bivalent antagonist at the MOR (30c) by linking the active enantiomer (-) to the inactive enantiomer (+). Such an isomeric combination resulted in a 30-fold loss in antagonist potency (Table 2), supporting the hypothesis that the bivalent occupation of two vicinal binding sites occurred.¹¹⁸

Table 1. Opioid Agonist Potencies in GPI, (R = −CH₃, C6 configuration = α).¹¹⁸

Compound no.	n	Morphinel C50, nM	Potency ratioa	Relative potencyb
29a	0	30.2 ± 8.7	4.6 ± 0.8	2.4
29b	1	22.0 ± 5.0	4.0 ± 0.6	2.1
29c	2	3.9 ± 0.6	36.0 ± 6.6	18.9
39d	3	25.1 ± 5.2	5.5 ± 0.7	2.9
Monomer		66.8 ± 11.7	1.9 ± 0.3	1.0

• ^a Potency ratio is the ratio of morphine (MOR agonist) IC50 to compound IC50 on the same tissue
• ^b Relative potency ratio is the bivalent potency ratio divided by monovalent potency ratio.

Table 2. Opioid antagonist potencies in GPI, (R = −CH₂CH(CH₂)₂, C6 configuration = β).¹¹⁸

Compound no.	n	Morphine IC50 ratioa	Ethylketazocine (EK) IC50 ratioa	Selectivity ratiob
30a	0	27.6 ± 6.5	30.4 ± 9.6	1.1
30b	1	57.4 ± 18.9	22.3 ± 4.7	0.4
30c	2	181.0 ± 34.3	17.0 ± 2.4	0.2
30d	3	61.4 ± 23.4	8.1 ± 2.4	0.1
30e	4	40.7 ± 23.4	7.2 ± 1.2	0.2
Monomer		66.8 ± 11.7	1.2 ± 0.1	0.2
Meso-isomer of 30c	2	7.8 ± 1.7	-	-

• ^a IC₅₀ ratio is the IC₅₀ of morphine or EK (KOR antagonist) in the presence of the antagonist (10 nM) divided by the control IC₅₀ (same preparation without the antagonist).
• ^b EK IC₅₀ ratio divided by morphine IC₅₀ ratio.

Based on the observation that the bivalent ligand with the shortest spacer was the most potent KOR antagonist, Portoghese went on to synthesize norbinaltorphimine (Nor-BNI) (31, Figure 13) which consists of two naltrexone pharmacophore subunits joined via ring fusion through a rigid pyrrole heterocycle.¹³⁷ Nor-BNI (31) was the first reported selective antagonist of the KOR with no significant activity at the MOR and DOR. Nor-BNI (31) was later identified to be a monovalent ligand with parts of structure of the other naltrexone pharmacophore subunit binding a second subsite within the KOR which confers the subtype selectivity.¹³⁸ This was the initial basis of the “message-address” concept later applied to aim for opioid receptor subtype selectivity. Further application of this concept led to the design and synthesis of naltrindole (9, Figure 1), a highly selective and potent non-peptide DOR antagonist.^{139, 140}

In the early 2000s, the research group of Neumeyer used a similar approach as Portoghese to design and synthesize potent mixed KOR and MOR ligands based on new morphinan pharmacophores, cyclorphan (32) and its N-cyclobutylmethyl analogue butorphan (33) (Figure 14).¹⁴¹ Both ligands have shown high affinity for the two receptor subtypes, being full KOR agonists and partial MOR agonists.^{142, 143}

Dimeric ligands of cyclorphan (32) and butorphan (33) were synthesized with the introduction of varying length ether spacers between the two pharmacophores at the phenolic position 3 (Figure 14).¹⁴¹ Further modifications such as the incorporation of ester moieties or a conformationally restricted fumaryl group within the linker led to the identification of 3 potent ligands, namely MCL-139 (34), MCL-144 (35) and MCL-145 (36) (Figure 15).¹⁴¹ These dimeric ligands showed an increased selectivity for the MOR and KOR over the DOR (more than 90-fold), with MCL-139 (34) and MCL-144 (35) behaving as full KOR agonists and partial MOR agonists, respectively (Table 3).¹⁴¹ Further work by Neumeyer on this series of compounds included the introduction of methyl groups α to the ester moiety to improve hydrolytic stability; however, this also led to a subsequent decrease in binding affinities at both the MOR and KOR.¹⁴⁵ An alternative approach to improve hydrolytic stability of the bivalent ligands was to introduce ether linkers with hydroxyl groups strategically placed beta to the ether moiety.¹⁴⁶ This led to the discovery of two new bivalent ligands (38 and 39, Figure 15) with sub-nanomolar binding affinity at the MOR and KOR, respectively, and significant selectivity for the two receptor subtypes over the DOR.¹⁴⁶

Table 3. Potencies (EC₅₀) and binding affinities (K_i) of the homobivalent ligands of the MCL series.

Compound	% Emax (KOR)	EC50 in nM (KOR)	Ki in nM (KOR)	% Emax (MOR)	EC50 in nM (MOR)	Ki in nM (MOR)	Pharmacological properties
MCL-139 (34)141	70 ± 5.7	4.5 ± 1.7	0.076 ± 0.002	50 ± 4.4	7.4 ± 1.8	0.16 ± 0.01	KOR full agonist MOR partial agonist
MCL-144 (35)141	60 ± 0.68	0.85 ± 0.053	0.049 ± 0.001	50 ± 6.6	1.3 ± 0.15	0.090 ± 0.004	KOR full agonist MOR partial agonist
MCL-145 (36)141	90 ± 6.0	2.3 ± 0.34	0.078 ± 0.01	50 ± 2.5	2.9 ± 0.50	0.20 ± 0.032	KOR and MOR partial agonist
MCL-692 (37)144	83 ± 0.95	2.4 ± 0.57	0.37 ± 0.040	42 ± 2.1	1.8 ± 0.36	0.26 ± 0.037	KOR and MOR partial agonist
MCL-715 (40)144	130 ± 3.3	5.5 ± 0.31	0.37 ± 0.024	25 ± 1.8	130 ± 42	8.8 ± 0.66	KOR full agonist
MCL-714 (41)144	130 ± 12	7.3 ± 0.81	0.27 ± 0.0023	38 ± 3.7	22 ± 5.3	2.5 ± 0.19	KOR full agonist

In a similar approach, homobivalent ligands were also derived from 14-methoxybutorphanol and ATPM pharmacophores.¹⁴⁴ MCL-715 (40) and MCL-714 (41) (Figure 15) were found to be full agonists of the KOR, with a 24 and ninefold selectivity for the KOR over the MOR (Table 3).¹⁴⁴ Both ligands also showed pronounced selectivity for the KOR over the DOR, 460- and 140-fold, respectively.¹⁴⁴ The homobivalent ligand of 14-methoxybutorphanol, MCL-692 (37) (Figure 15) showed partial agonist activity at both the KOR and MOR and maintained selectivity for these 2 receptor subtypes over the DOR (Table 3).¹⁴⁴

3.6 MOR/DOR

The MOR and DOR subtypes both mediate analgesic effects, albeit to different extent. While DOR agonists possess fewer side effects, their analgesic properties are weaker than those of MOR agonists.¹⁵⁷ Studies have shown that Coadministration of small amounts of a DOR antagonist, naltrindole (9, Figure 1), with a MOR agonist, leads to a slower development of tolerance.¹⁵⁸ These observations, taken together with evidence of MOR and DOR oligomerisation,¹⁰⁰ have provided the basis for the design and synthesis of bivalent ligands targeting MOR/DOR dimers with enhanced antinociceptive properties but possibly fewer side effects.

The MOR/DOR heterodimer is perhaps the best described heterodimer of the opioid receptors. Some of the approaches to confirm their presence and function in vivo include the use of antibodies that selectively recognize the endogenous heterodimers in the brain,^{159, 160} TAT peptides that disrupt the heterodimer¹⁶¹ and double fluorescent knock-in mice¹⁶² to investigate physical proximity of the MOR and the DOR. These studies suggest that the MOR/DOR heterodimer could behave as an anti-opioid negative feedback loop, triggered by chronic opioid exposure to mitigate the opioid system. These approaches, however, are indirect and thus provide no direct evidence of the physiological existence and role of the MOR/DOR heterodimer in vivo, therefore generating some controversy. Other studies suggest no functional MOR/DOR interaction in vivo.¹⁶³ Moreover, the discovery of small molecules, namely, CYM51010 (42),¹⁶⁴ NNTA (43)¹⁶⁵ and, more recently, MP135 (44)¹⁵⁶ (Figure 16) which selectively engage the MOR/DOR heterodimer over isolated MOR or DOR monomers challenged the notion that the concomitant binding of two pharmacophores at each opioid monomer are needed to probe heterodimers. Seemingly, the binding of a small molecule at one monomer could favor the activation of the other.

3.6.1 Peptidic heterobivalent ligands

Bivalent ligands of the MOR/DOR were synthesized by linking the DOR antagonists TICP-NH₂[Ψ] and TIPP-OH to the highly potent MOR agonist [Dmt¹]-DALDA via a tail-to-tail connection through a short spacer or the Lys side chain.¹⁴⁷ Despite showing agonist and antagonist properties at the MOR and DOR, the binding affinities (Table 4) of TICP[Ψ]-DALDA (45, Figure 17) were considerably lower than that of its constituent monovalent pharmacophores (DALDA, K_iMOR = 0.143 nM and TICP[Ψ] = 0.259 mM).¹⁴⁷ Such a loss in affinity was attributed to the steric hindrance resulting from the introduction of the linker.¹⁴⁷

In another study, the bivalent pseudopeptide Tyr-Pro-Phe-Phe-NH-CH₂-CH₂-NH-Tic-Dmt (46, Figure 17) was synthesized by the tail-to-tail connection of the DOR antagonist Dmt-Tic and the MOR agonist endormorphin-2 (EM-2) via an ethylenediamine linker.¹⁴⁸ The bivalent ligand showed comparable MOR and DOR affinities (Table 4) to that of their monovalent pharmacophores (EM-2, K_iMOR = 0.69 nM, K_iDOR = 1.45 nM; Dmt-Tic, K_iDOR = 1.22 nM), with a mixed MOR agonist/DOR antagonist pharmacological profile.¹⁴⁸

Gao et al. were able to improve the binding affinity of (EM-2), a MOR-selective agonist, to the DOR subtype.¹⁴⁹ They synthesized “balanced agonists” which show affinity for a broad spectrum of opioid receptors, by linking two EM-2 analogue pharmacophore subunits in a tail-to-tail approach via diamine linkers of various lengths. Interestingly, two of the bivalent peptides, 47 and 48 (Figure 17), with the deleted Phe⁴ residue, showed improved binding affinity at both the MOR and DOR (Table 4) compared to their monomer, [Tyr-Pro-Phe-NH-] (K_iMOR = 140 nM and K_iDOR = >10000 nM).¹⁴⁹ They were thus able to demonstrate that the affinity and selectivity profiles of a monovalent ligand can be altered by linking it to another pharmacophore.

More recently, Olson et al. designed a new series of bivalent ligands by linking the moderate/low affinity DOR (H-Tyr-Tic-OH, D) and the MOR (H-Tyr-Pro-Phe-DINal-NH₂-, M) antagonists to develop MOR/DOR selective antagonists which could also be used as pharmacological tool compounds.¹⁵⁰ The linker chain lengths employed ranged from 15 to 41 atoms resulting in the synthesis of six different ligands. One of the bivalent ligands, D24M (49) (Figure 17), emerged as a selective MOR/DOR antagonist (~ 90-fold when compared to the DOR) with sub-nanomolar affinity and potency (Table 4).¹⁵⁰ D24M (49) also strongly reduced naloxone-mediated behavior in models of acute and chronic morphine dependence.¹⁵⁰ These findings, combined with the observed MOR/DOR selectivity of D24M (49), support the implication of the MOR/DOR receptor in the modulation of analgesia, tolerance, and dependence.

3.6.2 Morphinan heterobivalent ligands

The compounds of the MDAN series were amongst the first bivalent ligands, derived from a morphinan-scaffold, targeting the MOR/DOR dimer. These bivalent ligands were designed and synthesized by Daniels et al. to determine if both receptor subtypes mediate their effects in a dimeric state.¹⁶⁶ In this regard, the MOR agonist, oxymorphone, was linked to the DOR antagonist, naltrindole (9, Figure 1), by means of a variable length spacer. Chronic intracerebroventricular (i.c.v.) administration in the tail-flick test in mice revealed that the length of the linker could influence the development of tolerance. Compounds with shorter linkers (e.g., MDAN-16, 50, Figure 18) showed tolerance similar to that of morphine while those of longer linkers (e.g., MDAN-21, 51, Figure 18) showed no significant tolerance.¹⁶⁶ In additional immunocytochemical trafficking and receptor activation studies, MDAN-21 (51) induced significantly less receptor MOR/DOR internalization than MDAN-16 (50), suggesting that these two ligands bind differently (bivalent vs. monovalent binding modes).¹⁶⁶ These findings also confirmed the role of the MOR/DOR heteromer in mediating the underlying mechanisms of tolerance development.

In a similar approach, Harvey et al. synthesized two bivalent ligands, L2 (52) and L4 (53) (Figure 18), by tethering high affinity MOR compounds to low affinity DOR compounds.¹⁶⁷ The idea behind such a combination of pharmacophores was to develop ligands with tuned affinity for the MOR/DOR heteromer over the DOR subtype that could ultimately be used as selective pharmacological tool compounds to study the MOR/DOR heteromer. L2 (52) was designed from the MOR agonist oxymorphone and the DOR antagonist ENTI (Figure 18) and L4 (53), from the MOR antagonist naltrexone and the DOR agonist DM-SNC80 (Figure 18). L4 (53) showed high affinity at the MOR (pK_iMOR > 9 nM), and improved affinity (~10-fold) for the MOR/DOR over the DOR.¹⁶⁷ Harvey et al. were thus able to prove that it is possible to fine-tune affinity and specificity of bivalent ligands for a particular receptor heteromer.

3.6.3 Mixed peptidic/non-peptidic heterobivalent ligands

Several bivalent ligands were also designed from structurally different pharmacophore subunits, i.e., a peptide and a morphinan scaffold. The first bivalent ligand synthesized with such a combination of pharmacophores was MCL-450 (54, Figure 19) consisting of the DOR-selective peptide Dmt-Tic and the MOR/KOR mixed agonist butorphan.¹⁵¹ Esterification of the phenol moiety allowed the introduction of a two-methylene spacer which was then coupled via amide formation to Dmt-Tic. MCL-450 (54) was shown to be a DOR antagonist with high affinity and moderate potency at the DOR (Table 4).¹⁵¹ The antagonistic properties of MCL-450 (54) was attributed in part to the Dmt-Tic pharmacophore. On the other hand, a slight loss in both affinity and potency, ranging from 3- to 6-fold, were observed at both the MOR and KOR (Table 4) when compared to the monovalent ligand butorphan (K_iMOR = 0.23 nM and K_iKOR = 0.079 nM).¹⁵¹

Lee et al. later studied bivalent ligands derived from enkephalin-like pharmacophores and parts of the fentanyl scaffold, namely, 1-phenethyl-piperidin-4-yl-phenyl-amine, 1-phenethyl-piperidin-4-ylamine and N-phenyl-N-piperidin-4-yl-propionamide.¹⁵² Their aim was to identify ligands with mixed MOR and DOR activities with analgesic properties but devoid of side effects. They reported a total of eight bivalent ligands showing broad biological activities at both the MOR and DOR. In general, improved binding affinities for the DOR was obtained by tethering enkephalin-like structures to parts of the fentanyl moieties.¹⁵² Interestingly, the selectivity profiles for the MOR or the DOR varied depending on the fentanyl moiety present at the C-terminal. For example, 55 (Figure 19), with the 1-phenethyl-piperidin-4-ylamine at the C-terminal displayed MOR-selectivity (DOR/MOR = 4.2) (Table 4) in the GTPγS binding assay while 56 (Figure 19), bearing the N-phenyl-N-piperidin-4-yl-propionamide at the same position, showed selective DOR binding affinity (DOR/MOR = 0.03) (Table 4).¹⁵² Further modification to 56 included the substitution of the Tyr residue, in the enkephalin analogues, for Dmt which is known to confer higher DOR affinity, resulting in 57 (Figure 19) which showed significantly improved binding affinities (Table 4) at both the MOR (60-fold increase) and DOR (twofold increase).¹⁵² 57 also showed increased potency (Table 4) in the MVD (13-fold) and in the GPI (24-fold) assays, compared to 56.¹⁵²

Additional work by Lee et al. involved structural modification of the bivalent ligands to improve lipophilicity.¹⁵³ A series of SAR studies were done on the enkephalin analogue, 57 (Figure 19), while keeping the Dmt¹ residue and the propionamide moiety constant. Substituting the d-Ala² with d-Nle² and para-halogenation of Phe⁴ led to improved opioid agonist activity and efficacy at both the MOR and DOR without significant alteration to their binding affinities.¹⁵³ This could be due to the higher bioavailability of the ligands as a result of their increased lipophilicity.¹⁵³ Truncating the Gly³ residue resulted in higher MOR/DOR selectivity and para-halogenation of Phe⁴ afforded MOR sub-nanomolar activity and affinity.¹⁵³ 58 (Figure 19), with tuned MOR and DOR affinities (Table 4), possessed potent antihyperalgesic and antiallodynic properties in in vivo assays.¹⁵³

In a further study, bivalent ligands targeting the MOR and DOR subtypes were designed and synthesized by tethering enkephalin analogues to the 4-anilidopiperidine derivatives.¹⁵⁴ The peptide pharmacophore subunit, an enkephalin moiety with either a N-terminal Tyr- or Dmt- were linked, with or without a spacer, to the non-peptide pharmacophore subunit consisting of a 5-amino substituted (tetrahydronaphthalene-2yl) methyl with 4-anilidopiperidine derivatives.¹⁵⁴ Introduction of a short β-alanine spacer between the enkephalin analogue and the 4-anilidopiperidine moiety and substitution of the N-terminal Tyr- for Dmt- yielded the bivalent ligand 59 (Figure 19) with high affinity for the MOR and DOR (Table 4) but no activity at the KOR.¹⁵⁴ However, 59 also showed short duration of action (< 15 min) and moderate activity in vivo.¹⁵⁴

3.7 KOR/DOR

Spinally administered KOR and DOR agonists can potentiate antinociception¹⁶⁸ and the KOR and DOR have been reported to be co-localized on spinal neurons.¹⁶⁹ Moreover, the observation of KOR/DOR dimerization and oligomerisation in cultured cells⁶⁵ and porcine ileum has prompted the design and synthesis of bivalent ligands specifically targeting the KOR/DOR heterodimer.¹⁷⁰

Bhushan et al. investigated a series of KOR/DOR bivalent ligands by connecting the KOR-selective antagonist 5’-GNTI (11, Figure 1), over a linker, to the DOR-selective antagonist NTI (9, Figure 1).¹²³ In intrathecal in vivo studies, the bivalent ligand, KDN-21 (60, Figure 20), was identified as a potent antagonist of the DOR-subtype δ₁-selective agonist, [d-Pen] enkephalin (DPDPE) and the KOR-subtype κ₂-agonist, bremazocine (17, Figure 1).¹²³ Binding studies in human embryonic kidney (HEK) cells with singly expressed KOR and DOR or co-expressed KOR/DOR revealed that KDN-21 (60) has a 200-fold higher affinity for the latter (Table 4) suggesting that KDN-21 (60) effectively binds a KOR/DOR heterodimer.¹²³ Another observation of significance was the different antagonism profiles of KDN-21 (60) in intrathecal (i.t.) and intraventricular (i.c.v.) in vivo studies. KDN-21 (60) did not antagonize intraventricularly administered agonists deltorphin-II (δ₂- selective), U50448 (18, Figure 1, κ₁-selective) or bremazocine (17, κ₂-selective) which suggests differing population of KOR and DOR subtypes in the brain and in the spinal cord.¹²³ These findings led to the conclusion that the bivalent antagonist KDN-21 (60) selectively bound a specific population of δ₁-κ₂ present in the spinal cord but not in the brain.¹²³

The studies by Daniels et al. on a series of bivalent ligands helped to gain more insight into the organization of DOR and KOR which is manifested as the δ₂ and κ₁ phenotypes.¹²⁵ The KDAN series of bivalent ligands was designed by tethering together, via variable length oligoglycyl linkers, the high-affinity DOR-selective antagonist NTI (9, Figure 1) and the high-affinity KOR(κ₁)-selective agonist ICI-199441 (61 and 62, Figure 20). The spacer length was varied from 12 atoms (KDAN-12, 62, Figure 20) to 20 atoms (KDAN-20). In intrathecal (i.t.) studies in mice, the bivalent ligands all showed antinociceptive activity with full agonism properties.¹²⁵ The bivalent ligand, KDAN-18 (61, Figure 20), exhibiting the highest potency (Table 4) in the in vivo studies, was potently antagonized by both norBNI (31, Figure 13) (κ-selective) and naltriben, NTB (10, Figure 1, δ₂-selective) but not by CTOP (µ-selective) which suggests that the MOR did not play a role in the antinociceptive activity mediated by KDAN-18 (61).¹²⁵ It was observed that the length of the linker is of significance as bivalent ligands of shorter linker lengths, e.g., KDAN-12 (62, Figure 20), were not antagonized by NTB (10).¹²⁵ This implied that KDAN-18 (61) likely binds neighboring δ₂ and κ₁ phenotypes with a linker length of 18 atoms facilitating their bridging. Studies on HEK cells which consist of singly expressed or co-expressed DOR and KOR showed that KDAN-18 (61) had a ~ 65-fold higher affinity for the co-expressed cell line in comparison to the mixed cell lines when the δ₂-selective [³H]-deltorphin-2 was used.¹²⁵ This provided further evidence of the binding of KDAN-18 (61) to associated KOR and DOR of the δ₂ and κ₁ phenotypes. Interestingly, the heterobivalent KDN-21 (60, Figure 20) which selectivity binds the δ₁ and κ₂ phenotype heterodimer, did not antagonize KDAN-18 (61), indicating that the latter bridges a different phenotypic heterodimer, namely, δ₂ and κ₁.¹²⁵

3.8 MOR/KOR

Peng et al. studied a series of heterobivalent ligands containing KOR agonist and MOR agonist/antagonist pharmacophore subunits.¹⁷¹ The KOR agonist butorphan (33, Figure 14) and other N-substituted derivatives were linked via ester- or amide- linkers to the MOR mixed agonist/antagonist cyclorphan (32, Figure 14). A 10-carbon ester linker, the stereochemistry of the pharmacophores and the N-substituents were significant parameters contributing to affinity at the MOR/KOR heterodimer.¹⁷¹ The heterobivalent ligands were either full or partial agonists at the KOR and MOR with a few showing mixed agonist/antagonist properties at the MOR.¹²⁴ While they all possessed relatively similar affinities at the MOR or KOR as their monomeric pharmacophores, they provided a good basis for the design and synthesis of selective pharmacological tool compounds for the MOR/KOR.¹⁷¹

The bivalent ligand KMN-21 (63, Figure 20) emerged as a selective MOR/KOR antagonist from a series of ligands designed by tethering the KOR antagonist 5’-GNTI (11, Figure 1) and the MOR-antagonist naltrexamine (β-NTX) via linkers of varying length.¹²⁴ KMN-21 (63), with a 21-atom linker between the two pharmacophores, significantly antagonized both U69508 (KOR selective agonist)- and DAMGO (MOR selective agonist)- induced Ca²⁺ release in cells with co-expressed MOR and KOR.¹²⁴

3.9 MOR/NOP

There is evidence pertaining to the heterodimerisation or the co-localization of the MOR and the nonclassical opioid receptor NOP.^{155, 172, 173} It has been suggested that probing the MOR/NOP heterodimer might provide new avenues for improved analgesia with reduced side effects.^174-177 The possible heterodimeric pairing of the MOR and NOP has been observed in both in vitro and in vivo experiments with nonhuman primates.¹⁷⁸ Ko and Naughton observed that N/OFQ, when combined with a single dose of morphine, potentiated intrathecal morphine-induced antinociception in monkeys.¹⁷⁸ Their findings indicated that mixed MOR/NOP agonists might show potent antinociception with fewer side effects.¹⁷⁸ Moreover, the mixed MOR/NOP peptide agonist, [Dmt¹] N/OFQ (1-13)-NH₂, showed a 30-fold increase in potency when compared to N/OFQ and exhibited robust and long-lasting antinociception in monkeys.¹⁷⁹ More recently, the new opioid analgesic, cebranopadol (64, Figure 21), a mixed MOR/NOP ligand, now in phase III clinical trials,^180-182 has shown favorable safety profile in preclinical trials.¹⁷⁷ Additionally, an in vitro study by Evans et al. demonstrated the co-localization of the MOR and NOP via the use of tagged receptors.¹⁷³ Administration of N/OFQ to the co-expressing cell lines resulted in the co-internalization of both the MOR and NOP indicating that these two receptor subtypes might exist either co-localized or as heterodimers.¹⁷³ Studies in CHO cells showed the co-immunoprecipitation of NOP and MOR and the displacement of [³H]-N/OFQ which does not occur in single cell expression systems.¹⁷² While heterodimerisation of these two receptor subtypes can negatively impact MOR signalling,¹⁸³ targeting the heterodimers may result in an improvement in side effect profiles, previously observed in studies with the 6TM MOR and NOP heterodimer.¹⁸⁴

To study the dimerization of the MOR and NOP and the role of the latter on MOR signalling, Bird et al. designed two bivalent ligands, DeNO (65) and De101 (66) (Figure 21) and evaluated their activity in a human HEK co-expression system (HEK_MOR/NOP).¹⁵⁵ DeNO (65), synthesized from the MOR agonist dermorphin and N/OFQ (Figure 21), displays full agonist activity at both the MOR and NOP.¹⁷⁵ De101 (66) was synthesized by tethering together dermorphin and the high-affinity antagonist analogue of N/OFQ, UFP-101 (Figure 21).¹⁵⁵

DeNO (65) and De101 (66) (Figure 21) were shown to be able to displace 140% and 120% of [³H]-N/OFQ in HEK_MOR/NOP membranes, which is significantly higher than that of their pharmacophore subunits which were unable to produce more than 100% displacement.¹⁵⁵ In GTPγ[³⁵S] and cAMP assays, DeNO (65) showed a decreased potency in the HEK_MOR/NOP co-expression system compared to single cells, HEK_MOR and HEK_NOP (Table 4).¹⁵⁵ De101 (66) showed similar potencies in both HEK_MOR/NOP and single cells expression systems (Table 4). The bivalent ligands were also evaluated for their ERK1/2 activity, which is involved in a number of cellular functions, all mediated by the spatio-temporal activation of the receptor. Typically, activation at the MOR occurs within 2 to 5 min following drug administration with a peak observed from 7.5 to 10 min post administration.¹⁸⁵ DeNO (65) and De101 (66) both showed a significant delay in the activation of the ERK1/2 pathway in the HEK_MOR/NOP cell line.¹⁵⁵ The data observed in these three assays suggest that NOP does impact MOR downstream signalling. However, when evaluated in vivo in rats, DeNO (65) showed similar effects as demorphin and N/OFQ upon spinal administration, with no improvement in potency or efficacy indicating that no synergistic antinociceptive effects are obtained with the simultaneous targeting of the MOR and NOP.¹⁷⁵ In vivo characterization of De101 (66) has not been reported. While Bird et al. were not able to confirm heterodimerisation of the classical MOR and the nonclassical NOP, they were able to demonstrate their co-localization and their results now pave the way for further pharmacological investigation of the possible MOR/NOP heterodimer. It is important to note that like their peptidic parent compounds, the instability of DeNO (65) and De101 (66) might limit their applications in vivo.

3.10 Heterobivalent ligands of opioid and non-opioid heteromers

The dimerization of opioid receptors with other non-opioid receptors of the GPCR family involved in the mechanisms of pain has also been pursued in the search for safer, efficient analgesics. The heterodimerisation of the MOR with other GPCRs such as the cholecystokinin receptor (CCK₂), the adrenergic receptor (A₁AR), the cannabinoid receptor (CB₁), the dopamine receptors (D2-like receptors) and neurokinin-1 receptor (NK1) amongst others is described in the literature.^{130, 186-189} Bivalent ligands addressing these heterodimers are important pharmacological tool compounds in the investigation of their functional roles and their significance in pain therapy. In this section, we aim to summarize some of the efforts in the development of bivalent ligands targeting dimeric associations of an opioid receptor with a non-opioid receptor (Table 5).

The cholecystokinin (CCK) receptors also belong to the class A GPCR family. The two CCK receptors, CCK₁ and CCK₂, share a 50% sequence homology^190-194 with the CCK₁ receptors located primarily in the peripheral alimentary tract, while the CCK₂ receptors are distributed in the CNS.¹⁹⁵ Activation of both CCK receptor subtypes recruits the effector molecule G_q which leads to the stimulation of phospholipase C and intracellular calcium.^190-195 At physiological level, this translates into stimulation of gall bladder contraction, pancreatic exocrine excretion, enteric mobility, gastric secretion, and modulation of neurotransmission.^190-195

CCK₂ and opioid receptors are present in the parts of the CNS involved in the modulation of pain.^{196, 197} Activation of the MOR or the DOR triggers CCK release which then negatively impacts opioid receptor-induced analgesia by interacting at the CCK₂ receptor. The interactive mechanisms of action of these two receptors have also been demonstrated in in vivo studies; CCK₂ receptor knockout mice show increased opioid-mediated algesia and an upregulated endogenous opioid receptor system.^{198, 199} Antagonism of the CCK₂ receptor has also been reported to block morphine tolerance.²⁰⁰ Considering their opposite roles in mediating analgesia, it is tempting to investigate the activity of a MOR/CCK₂ bivalent ligand based on a MOR agonist and a CCK₂ antagonist, assuming that these two receptors can form functional dimers.

A study by Zheng et al. investigated the possibility of the MOR and CCK₂ forming heterodimers using the BRET technology.¹⁸⁶ Bivalent ligands consisting of the MOR-agonist oxymorphone linked via a variable length spacer, to the CCK₂ antagonist, L-365260, (67a-c, Figure 22) were studied. Their findings point to a heterodimerisation of the MOR and CCK₂ which is non-constitutive but rather induced in the presence of bivalent ligands of appropriate linker lengths (16 to 22 atoms) sufficient to bridge the two receptors.¹⁸⁶ The BRET signal for bivalent ligands with spacer lengths 18 and 22 atoms (67b and 67c, Figure 22), respectively, significantly increased and plateaued at approximately 2, which suggests specific saturable interaction between MOR and CCK₂ receptors.¹⁸⁶ Conversely, the BRET signal of the bivalent ligand with a linker length of 16 atoms (67a, Figure 22), were significantly lower, approximately half that of the signal obtained with the bivalent ligands of linker lengths 18 and 22 (67b and 67c, Figure 22), respectively.¹⁸⁶ Introducing the CCK₂ antagonist, L-365260, attached solely to a 22-atom linker resulted in a decrease in the BRET ratio, indicating the competition between the latter and the antagonist pharmacophore of the bivalent ligands.¹⁸⁶ The CCK₂ antagonist triggered a switch from bivalent binding mode to univalent, which resulted in the disruption of the non-constitutive heterodimers, which was then observed as a reduction of the BRET signal.¹⁸⁶

Based on the evidence that the heterodimerisation of the MOR and CCK₂ receptor is non-constitutive, there seems to be no direct correlation between MOR/CCK₂ receptor dimer formation and the development of tolerance.¹⁸⁶ Zheng et al. determined the i.c.v ED₅₀ of their bivalent ligands and observed no significant tolerance, independent of linker lengths, thereby reinforcing the notion that the underlying mechanism of tolerance is not mediated by a possible MOR/CCK₂ receptor heterodimer.¹⁸⁶

MOR is also co-expressed with neurokinin-1 (NK1) receptors in areas of the CNS involved in pain transmission that also controls opioid dependence and reward.²⁰¹ The NK1 receptor is the target of tachykinin peptides such as substance P which has been reported to play an important role in the mediation of pain signals.¹⁸⁹ Yamamoto et al. designed and synthesized a number of bifunctional ligands based on a peptide scaffold which showed excellent agonist activity for both the MOR and DOR and high antagonist activity at the NK1 receptor both in vitro and in vivo.^202-206 Vardanyan et al. used a similar approach to develop non-peptide bivalent ligands of the MOR/NK1 heterodimer.¹⁸⁹ Bivalent ligands based on the structures of the opioid agonist fentanyl and the NK1 antagonist L732138 were either tethered via a linker or synthesized as an ion pair (68a-c and 69a-c, Figure 22). These ligands were evaluated in competitive radioligand binding assays in recombinant cell lines expressing the human MOR or the human NK1 receptor and in functional assays using guinea pig isolated ileum/longitudinal muscle myenteric plexus (GPI/LMMP) for the MOR and NK1 receptor and mouse vas deferens (MVD) for the DOR.

In radioligand competition binding assays, the bivalent ligands, both covalently bonded and ion paired, showed a significant decrease in affinity for the MOR along with a dramatic loss of analgesic activity when compared to fentanyl.¹⁸⁹ Such a loss of MOR activity was attributed to the introduction of a negatively charged carboxyl group on the fentanyl moiety which impairs MOR recognition.¹⁸⁹ On the other hand, the activity of the bivalent ligands was unaltered at the NK1 receptor; the bivalent ligands showed high affinity in radioligand competition binding assays using [³H] Substance P.¹⁸⁹ In functional assays, all bivalent ligands displayed antagonism of the NK1 receptor with ligand 68b (Figure 22) showing a 10-fold higher antagonist affinity (Table 5) than the other ligands.¹⁸⁹ 68b was also the most potent bivalent ligand at the MOR, making it a suitable lead compound for further development of non-peptide bivalent ligands of the MOR/NK1 heterodimer.¹⁸⁹

There is evidence of a crosstalk between the A1 adenosine receptors (A₁ARs) and the MOR. When activated this results in a decrease in cAMP levels in their target cells.²⁰⁷ Cross tolerance and cross withdrawal interactions between A₁ARs and the MOR have been observed in the PNS suggesting that these receptors are co-localized on the same primary nociceptors and might even function as dimers.²⁰⁷ Targeting both the A₁ARs and the MOR might be of significant therapeutic applications in some disease conditions; activation of A₁ARs and the release of adenosine leads to hypotension during severe sepsis or septic shock, and ultimately tissue hypoperfusion and ischemia.²⁰⁸ Additionally, the MOR antagonist naloxone (7, Figure 1) has been employed against lowered blood pressure during septic or hypovolemic shock, implicating the participation of endogenous opioids in the severity of the latter syndromes.^{209, 210}

Mathew et al. synthesized the heterobivalent ligand 70 (Figure 22), based on the MOR agonist fentanyl and the highly selective A₁AR agonist N₆-cyclopentyl adenosine (CPA) linked via an amide linker at the 5’-position of CPA.¹⁸⁷ 70 showed antagonistic properties at both the MOR and the A₁AR when evaluated in nociceptive tests in mice and in cAMP production assays.¹⁸⁷ 70 was able to reverse the increased latency observed in both intra cerebro ventricular (icv) and intraperitoneal (ip) administration in mice.¹⁸⁷ 70 also activated cAMP production in a dose-dependent manner both in CHO K1 cells over expressing MORs and in CHO Chem 3 cells overexpressing A₁ARs with a maximum stimulation of 29% and 17%, respectively.¹⁸⁷ In comparison, the MOR antagonist naloxone (7, Figure 1) exhibited an increased cAMP production with a maximum stimulation of 50% while the A₁AR antagonist, 8-cyclopentyl-1,3-dipropyl-xanthine (DPCPX, 74, Figure 23) enhanced cAMP production with a maximum stimulation of 45%.¹⁸⁷ While its affinity was sufficient to trigger biological response at reasonable concentrations, 70 showed lower binding affinities (Table 5) than the antagonists, naloxone (7) (K_iMOR = 1.5 nM) and DPCPX (74) (K_iA1AR = 12.9 nM).¹⁸⁷

The MOR and the cannabinoid receptor (CB₁) have been reported to associate as heterodimers in cultured cells^{211, 212} and they are both present in overlapping regions of the CNS.^213-217 Both receptor types share pharmacological properties and are involved in physiological mechanisms mediating analgesia, tolerance, and dependence after prolonged administration.^218-220 It was suggested that both receptors have a significant role in antinociception and that the MOR/CB₁ may be involved in the development of tolerance. Cross tolerance between tetrahydrocannabinol (THC, 75) (Figure 23) and morphine has been observed in studies where antagonism of antinociception was obtained with both naloxone (7, Figure 1) and CB₁ antagonist SR141716 (76, Figure 23).²²¹ Coadministration of morphine and CB₁ antagonist prevented acute and chronic tolerance to morphine.²²² Self-administration studies show that both receptors are involved in reward processes in which the CB₁ antagonist SR141716 (76) and opioid antagonist naloxone (7, Figure 1) decreased self-administration of morphine or THC (75).^{223, 224} In CB₁ knockout mice, dependence and tolerance were decreased for morphine but not for other drugs of abuse.²²⁵ Comparatively, in MOR and DOR knockout mice, the cannabinoid withdrawal syndrome is significantly inhibited.²²⁶

Le Naour et al. studied a series of heterobivalent ligands possibly targeting the MOR/cannabinoid heterodimer by tethering the MOR agonist oxymorphamine to the CB₁ antagonist/inverse agonist SR 141716 (76, Figure 23) via a spacer of variable lengths (ranging from 2 to 20 atoms).¹⁸⁸ Based on previous observations that the MOR/DOR heterobivalent ligand MDAN-21 (51, Figure 18) which effectively bridges the two receptors shows no internalization of co-expressed MOR and DOR in HEK-293 cells in imaging studies,²²⁷ similar experiments were carried out on co-expressed MOR and CB₁ receptors. Interestingly, the MOR/CB₁ heterobivalent ligand 72c (Figure 22) with a linker length of 20 atoms showed no co-internalisation of the MOR and CB₁ receptors with both receptors remaining localized on the plasma membrane.¹⁸⁸ These data strongly imply that 72c can bridge the two receptors.

The heterobivalent ligands were also assessed in in vivo assays using the mouse tail-flick test in which they were evaluated for antinociception and tolerance following intracerebroventricular (i.c.v.) or intrathethal (i.t.) injection. With the exception of the heterobivalent ligand 71 (Figure 22) with the shortest spacer length of two atoms, the bivalent ligands of linker lengths ranging from 6 to 20 atoms (72a-c, Figure 22) exhibited no tolerance.¹⁸⁸ This observation suggests that the decrease or loss of tolerance associated with CB₁ antagonism is not likely to be mediated via a MOR/CB₁ heteromer. Le Naour et al. attributed the low potency and tolerance observed for heterobivalent compound 71 to its shorter spacer which places the two pharmacophores in close proximity to one another, hence compromising CB₁-mediated antagonism and MOR-mediated antinociception.¹⁸⁸ Of significance was the finding that 72c (Figure 22) with a 20-atom spacer is 20-fold more potent than its 19-atom counterpart, 72b (Figure 22) (Table 5) when administered supraspinally.¹⁸⁸ Comparatively, 72c shows only a fourfold higher potency than 72b (Table 5) via the i.t. route.¹⁸⁸ These data were indicative of a higher supraspinal distribution of MOR-CB₁ heterodimers.

The dopaminergic D₂-like receptor subtypes (namely, D₂R and D₄R) and the MOR are expressed in areas of the brain involved in schizophrenia, Parkinson's disease, addiction, and pain.¹³⁰ They are co-localized in many nuclei of the brain which suggests intermolecular receptor interactions, thus making them relevant targets for the treatment of addiction and chronic pain.^{228, 229} In vivo studies support the presence of cross-regulation between the D₂-like receptors and the MOR. Activation of the dopaminergic receptors leads to an intermediate reduction in MOR immunoreactivity and D₂-like receptor/MOR interactions influence morphine-triggered upregulation of certain transcription factors like c-Fos, δFosB, and P-CREB.^230-232

Qian et al. investigated the pharmacological properties of heterobivalent ligands of the D₂R/MOR and the D₄R/MOR heterodimers in mitogen-activated protein kinase (MAPK) and β-arrestin recruitment assays.¹³⁰ The bivalent ligands were designed by linking either a MOR agonist (hydromorphone) or an antagonist (naltrexone, 6) to D₂-like receptor agonist, 5-hydroxy-2-(dipropylamino) tetralin (DPAT, 77, Figure 23) or antagonist, 1,4-disubstituted aromatic piperazines (DAPs, 78, Figure 23) via a variable length PEG spacer. The heterobivalent ligands were evaluated in radioligand binding competition assays involving HEK293T cells with either singly expressed D₂R, D₄R and MOR cells or co-expressed D₂R/MOR and D₄R/MOR cells. The bivalent ligands based on the D₂-like receptor antagonist DAP (78) and the MOR agonist hydromorphone (73a-c, Figure 22) showed a general decrease in binding affinities when compared to DAP (78) when tested in HEK293T cells expressing solely D₄R.¹³⁰ However, the heterobivalent ligand 73a (Figure 22) with a shorter spacer length (18 atoms) showed a twofold higher affinity than heterobivalent ligands 73b and 73c (Figure 22) with longer spacers (Table 5).¹³⁰ Interestingly, when evaluated in HEK293T cells co-expressing both D₄R and MOR, a biphasic competition curve was obtained for 73a with two distinct affinity constants (K_ihigh = 1.2 nM and K_ilow = 207 nM), which implies a bivalent binding mode.¹³⁰ The high-affinity K_i value results from a bivalent receptor-bridging binding mode while the low affinity K_i value represents monovalent binding to only the D₄R.¹³⁰ When similar experiments were performed on HEK293T cells co-expressing D₂R/MOR, a monophasic binding competition curve was obtained for bivalent ligand 73a which could be indicative that the latter is unable to simultaneously bind D₂R and MOR.¹³⁰

In a MAPK phosphorylation assay using HEK293 cells stably expressing D₄R, all the bivalent ligands were able to activate the MAPK signalling pathway.¹³⁰ The heterobivalent ligand 73a showed high potency (EC₅₀ = 0.12 µM) and relatively high efficacy (E_max = 92%) in comparison to the monovalent ligand DAP (78, Figure 20, EC₅₀ = 0.21 µM and E_max = 75%).¹³⁰ The heterobivalent ligands were also tested for their MOR activity in a β-arrestin 2 recruitment assay. All the bivalent ligands possessed similar EC₅₀ values as their parent monomer hydromorphone with 73a displaying excellent potency (EC₅₀ = 12.73 nM) and efficacy (E_max = 85%).¹³⁰

3.11 Bitopic ligands of the opioid receptors

The discovery of bitopic ligands of the MOR happened much by serendipity. In the search for novel selective antagonists of the MOR, Li et al. introduced heterocyclic substituents at position 6’ of naltrexone (6, Figure 1) via an amide linker.²³³ This led to the discovery of NAQ (79, Figure 24) which showed sub-nanomolar binding affinity (Table 6) to the MOR with over 200-fold selectivity for the MOR over the DOR and 50-fold selectivity over the KOR.²³³ NAQ (79) behaved as a competitive antagonist producing a right shift in DAMGO concentration-³⁵GTP[γS] stimulation curve in CHO cell lines.²³⁹ NAQ (79) also exhibited significantly fewer opioid withdrawal effects compared to commonly used opioid antagonists such as naloxone (7, Figure 1) or naltrexone (6, Figure 1), even at 100-fold higher doses.²³⁹

NAQ (79, Figure 24) was initially designed following the message-address concept in which the naltrexone core would confer the activity and the heterocyclic moiety at position 6’ selectivity for the MOR.²³³ The rationale behind the choice of aromatic substituents was to aim for π-π stacking interactions within a binding subdomain located extracellular at the MOR consisting of primarily aromatic residues, a subdomain absent at the DOR.²³³ The nitrogen atom on the isoquinoline ring was introduced to gain possible hydrogen bond interactions with Y210 (ECL2) or W318^7.35 on the ELs of the MOR, which would then grant selectivity over the KOR.²³³ Subsequent structure–activity relationships on NAQ (79) led to the discovery of other selective ligands of the MOR like NCQ (80)²³⁴ or NAN (82) (Figure 24).²³⁶ Structurally, though initially designed as orthosteric ligands, NAQ (79), NCQ (80) and NAN (82) can also be perceived as bitopic ligands simultaneously targeting the OBS and a less-conserved allosteric binding subdomain. However, there is no structural evidence of this presumed bitopic binding mode till date.

Interestingly, NCQ (80, Figure 24) which is structurally similar to NAQ (79, Figure 24), albeit two additional substitutions at positions 1′ and 4′ of the isoquinoline ring, behaved as a partial MOR agonist in the [³⁵S]GTPγS binding assay.²³⁴ NCQ (80) showed a binding affinity of 0.55 nM for the MOR with a 40-fold selectivity for the MOR over the KOR and a 62-fold selectivity over the DOR.²³⁴ Generally, SAR studies suggest that the presence of 1′- or 4′- substituents on the isoquinoline ring play a role on the binding affinities of the NAQ analogues.²³⁴

Wang et al. performed docking and MD studies to gain more insight into the differing pharmacology of NAQ (79) and NCQ (80) (Figure 24) at a molecular level.²⁴⁰ NAQ (79), being an antagonist, was docked into the crystal structure of the inactive MOR, MOR^inactive while NCQ (80) was docked into both the inactive and active MOR, MOR^active, crytal structures. For both ligands, the epoxymorphinan core formed the stereotypical interactions observed in the OBS, hydrophobic interactions with M151^3.36, W293^6.48, and H297^6.52, and the anchoring ionic interaction of the quarternary amine with D147^3.32 (Figure 25).²⁴⁰ While the “address” portion of NAQ (79) and NCQ (80) occupied the same subdomain of the ABS in the MOR^inactive, in the MOR^active, the 1′,4′-substituted isoquinoline ring of NCQ (80) bound within an alternative subdomain of the ABS (see arrows, Figure 25).²⁴⁰ The exact topography of the MOR ABS is yet to be confirmed; however, the two subdomains of the ABS which accommodated the “address” portions of NAQ (79) and NCQ (80) were identical to those of previously published allosteric sites of other class A GPCRs.²⁴⁰ In MOR^inactive, the binding subdomain was delineated by TMs 6, 7 and ECL3 (ABD1), the isoquinoline moiety of NAQ (79) and NCQ (80) finding interactions with V300^6.55 and W318^7.35, respectively (Figure 25A,B).²⁴⁰ In MOR^active, the isoquinoline ring of NCQ (80) could form direct interactions with Q124^2.60, W133^ECL1, and I144^3.29 within an allosteric binding subdomain formed by TMs 2,3, and ECL1 (ABD2) (Figure 25C).²⁴⁰

Similar interactions were observed in MD studies of the binding poses of NCQ (80) and NAQ (79) (Figure 24) in the MOR^inactive as in docking studies.²⁴⁰ However, following 100 ns of MD simulations, the “address” portion of NCQ (80) shifted from ABD1 to ABD2.²⁴⁰ Conversely, in MOR^active, the isoquinoline moiety of NCQ (80) occupied ABD2 throughout the entire duration of the MD simulation.²⁴⁰ The “address” portion of NCQ (80) evidently preferred ABD2 in both the MOR^active and MOR^inactive.²⁴⁰ When comparing the docking poses of NAQ (79) and NCQ (80) in MOR^inactive, the binding subdomain ABD1 was not spacious enough to accommodate the additional substituents at positions 1′ and 4′ of the isoquinoline ring of NCQ (80).²⁴⁰ Consequently, following 100 ns of MD simulations, the hydrophobic interactions between the “address” portion of NCQ (80) and residues of ABD1 decreased, at the expense of the gain in interactions within ABD2.²⁴⁰ The net effect of the electron withdrawing chloride substituent and the electron donating methoxyl substituent was an increase in the π-electron density of the isoquinoline ring, which further promoted π-π stacking interactions with W133^ECL1 and CH-π interactions with I144^3.29 within ABD2.²⁴⁰

Docking and MD studies indicate that the interactions of the “address” portion of NCQ (80, Figure 24) within ABD1 may bring the aromatic ring of the epoxymorphinan ring closer to W293^6.48 resulting in a conformational change in the side chain of the latter.²⁴⁰ As previously described (see “2. Structure of the opioid receptors and insight into their activation mechanism” section), W293^6.48 is a residue playing a pivotal role in receptor activation; it forms interactions with F289^6.44, one of the residues of the FPI core triad which is a “microswitch” for MOR activation (Figure 3D). Additionally, average structures from 81 to 100 ns of MD simulations showed that due to the movement of the side chain of W293 in the MOR^inactive-NCQ bound system, TM5 moved inward while TM6 displayed an outward movement when compared to the docking poses.²⁴⁰ Thus, TM5 and TM6 in MD simulations of MOR^inactive-NCQ bound system adopted conformations similar to those of the MOR^active-NCQ bound system. In summary, the “address” portion of NAQ (79, Figure 24) binding within ABD1 had no effect on the “message” portion fitting the conventional OBS, similar to a SAM. On the other hand, the “address” portion of NCQ (80, Figure 24), much like a PAM, modulates the function of the epoxymorphinan ring, when bound within ABD2, further stabilizing the active receptor conformation. The docking and MD studies by Wang et al.²⁴⁰ provide new insights to design bitopic ligands with positive allosteric modulation, for example, with the introduction of bulky electron-donating substituents so as to interfere with binding within ABD1 while favouring occupation of the allosteric subdomain ABD2.

NAN (82, Figure 24) is another putative bitopic ligand identified from SAR studies of NAQ (79, Figure 24).²³⁶ With an indole ring attached via an amide linker at position 6’ of naltrexone, NAN (82) behaved as a competitive MOR antagonist in [³⁵S]-GTPγS functional assays.²³⁵ Co-incubation with 5 nM NAN (82) resulted in a parallel right shift in the DAMGO concentration-effect curve in both mMOR-CHO cells and thalamus (K_e= 2.08 nM and 1.55 nM, respectively).²³⁵ Moreover, NAN (82) inhibited DAMGO-stimulated Ca²⁺ flux in a Ca²⁺ flux assay in a concentration-dependent manner.²³⁵ Opioid withdrawal studies in morphine-pelleted mice revealed NAN to produce significantly less withdrawal symptoms than similar doses of naltrexone (6, Figure 1).²³⁵ NAN (82) is also unlikely to produce cellular MOR adaptations that may eventually lead to opioid tolerance.²³⁵

Obeng et al. also performed docking studies and MD simulations to identify the key epitopes contributing to NAN (82, Figure 24) antagonism despite it sharing structural similarities to the agonist INTA (81, Figure 24) which reportedly binds to KOR-DOR and KOR-MOR heterodimers.²³⁶ INTA (81) and NAN (82) both share an identical epoxymorphinan “message” portion with their “address” portion slightly differing in linker and substitution pattern. Similar to the results obtained from the docking studies of NAQ (79) and NCQ (80), it was observed that while the “message” moiety of both INTA (81) and NAN (82) maintained the stereotypical interactions observed with morphinan ligands, their “address” portion occupied different binding subdomains within the ABS.²³⁶ In both docking studies and MD simulations at MOR^active, the indole moiety of INTA (81) found hydrophobic interactions with H54^N-terminus and π-π stacking interactions with W318^7.35.²³⁶ While NAN (82) did not form any interaction with H54^N-terminus in the MOR^inactive, it did find similar interactions as INTA (81) with W318^7.35, although less strong as estimated based on distance between the residue and the indole moiety.²³⁶ However, a 10 ns of MD simulations, the “address” portion of NAN (82) formed new hydrophobic interactions with L232^5.38 and K233^5.39, accommodating an alternative subdomain of the ABS.²³⁶ This switch in the binding poses of the “address” moiety of NAN (82) observed between docking and MD studies could be due to the α-configuration of the indole substituent which seems to favor the inactive conformation of the receptor.²³⁶ Hence, the indole moiety with a high π-electron density seems to stabilize an inactive receptor conformation, imitating the function of a NAM.²³⁶

More recently, Faouzi et al. aimed to target the sodium ion allosteric binding site (Na⁺-ABS) conserved in opioid receptors and class A GPCRs.²³⁷ The bitopic ligands were designed based on the fentanyl (14, Figure 1) scaffold where the benzene ring was replaced with alkyl chain linkers connected to either an amine or a guanidine moiety (Figure 26). The latter positively charged functional groups were introduced to gain interactions within the Na⁺ binding pocket, particularly with the carboxylate side chain of the residue D^2.50 (Figure 7). Fentanyl (14) was first docked into an active state of the MOR so as to obtain an estimation of the distance between the OBS and the Na⁺-ABS. According to the models’ prediction, the amide nitrogen of fentanyl and the carboxylate group of D^2.50 were separated by a distance of 13 Å.²³⁷ Hence, the aliphatic chain linker lengths were varied ranging from 3 to 11 methylene groups (n  = 3, 5, 6, 7, 9, 11) (Figure 26). Faouzi and co-workers predicted that the linker length, n = 7, would be optimal for a direct interaction, via a salt bridge, between the amino functional group (83d) and D^2.50 while a slightly shorter linker (n  = 6) would allow for similar interactions for the guanidine moiety (84c).²³⁷

Surprisingly, the amino-fentanyl compounds (Figure 26, 83a-f) all showed decreased affinity for the MOR, indicating that the amine group did not engage in any interactions with the polar and negatively charged residues of the Na⁺- ABS.²³⁷ These include the residues N^3.35, N^7.45, S^7.46 and S^3.39 (Figure 7). Further extension of the aliphatic linker chains in an attempt to probe for possible interactions did not result in any notable effects on the binding affinities of these bitopic ligands.²³⁷

On the other hand, the guanidino counterparts showed moderate to high affinity for the MOR with K_i values ranging from 590 nM for 84f (Figure 26, N = 11) to 4.1 nM for 84c (Figure 26, N = 6, C6 guano).²³⁷ 84c thus showed a slightly better affinity for the MOR than its parent compound fentanyl (K_i = 9 nM).²³⁷ The guanidine moiety seemed better than the corresponding amino moiety in terms of finding additional hydrogen bond interactions with the polar cavity usually harboring the Na⁺ ion. The K_i values of the bitopic ligands with the guanidine moiety were up to 1000-fold lower when compared to that of their amino counterparts.²³⁷ Linker lengths considerably affected the binding affinity of the bitopic ligands with 84b (Figure 26, N= 5, C5 guano) and 84c showing the highest affinities (Table 6).²³⁷ 84a (Figure 2, N = 3) displayed a 16-fold lower affinity for the MOR (Table 6) possibly due to the shorter linker unable to reach into the Na⁺- ABS while 84f (Figure 26, N = 11) had the lowest affinity (Table 6) which could result from possible steric clashes with some residues in the Na⁺-ABS imposed by a longer linker.²³⁷

The bitopic ligands were also tested in functional assays, namely, the Tango assay for β-arrestin-2-recruitment and cAMP assays for G_i protein activation. As expected, the bitopic ligands with the amino group only weakly stimulated the G_i pathway, with EC₅₀ values all above 293 nM.²³⁷ The bitopic ligand 84b (Figure 26, N = 5, C5 guano) showing the highest affinity for the MOR was a full agonist in cAMP assays with a subnanomlar potency (Table 6) comparable to that of the full MOR agonist DAMGO (EC₅₀ = 0.9 nM).²³⁷ As observed in the determination of binding affinities, an increase in linker length resulted in a decrease in potency with 84 f (Figure 26, N = 11) showing the lowest potency (Table 6).²³⁷ Interestingly, 84b with the highest potency in cAMP assays, did not show a similar propensity to recruit β-arrestin-2 in the TANGO assays. In these assays, 84b had an EC₅₀ of 4710 nM and an E_max of 54% which is significantly lower compared to the values obtained for the full agonists DAMGO (EC₅₀ = 723 nM and E_max = 100%) and fentanyl (EC₅₀ = 66 nM and E_max = 119%).²³⁷ Therefore, 84b displayed considerable functional bias for the G_i protein activation pathway over β-arrestin-2 recruitment.

Faouzi et al. were also able to obtain cryo-EM structures of the MOR-G_i complex bound to the bitopic ligands 84b (Figure 26, N = 5, C5 guano) and 84c (Figure 26, N = 6, C6 guano) at 3.2 and 3.3 Å, respectively. As predicted, the fentanyl scaffolds of 84b and 84c overlapped with that of the fentanyl model from docking studies in the OBS with the guanidine moiety reaching towards the Na⁺-ABS.²³⁷ In both structures, the stereotypical interaction between the anchoring residue D^3.32 and the piperidine nitrogen atom was observed.²³⁷ The guanidine moiety of 84b was positioned within 4 Å of D^2.50 in the Na⁺-ABS which indicated the formation of a possible interaction between the two, albeit weak considering the distance over 3 Å.²³⁷ However, this interaction could also be mediated by water molecules not resolved at a resolution of 3.2 Å. Faouzi et al. then employed MD to confirm that water molecules could play an important role in mediating the observed interaction between the basic guanidine moeity and acidic D^2.50 residue.²³⁷ Conversely, the guanidine functional group of 84c was positioned slightly closer to D^2.50, within 3 Å, thus allowing direct interaction via a salt bridge.²³⁷

The intrinsic efficacy of bitopic ligands 84a-f (Figure 26) was also investigated in BRET-based assays for arrestin recruitment and G-protein heterotrimer dissociation (TRUPATH) at all G_α subtypes (G_i1, G_i2, G_i3, G_oA, G_oB and G_z) through which the MOR signals. The physiological roles mediated by each of these G_α subtypes are yet to be completely understood. Some studies have tried to investigate the role of the G_α subtypes in mutant mice models. Mice lacking or deficient in G_αi2 have demonstrated inflammatory bowel disease²⁴¹ with pronounced increases in pro-inflammatory T_H1-type cytokines²⁴² and increased basal production of IL-12 in some subsets of APC.²⁴³ The immune abnormalities in mouse deficient G_αi2 precede colitis.²⁴⁴ An increased incidence of colonic carcinomas arising from chronic inflammation was also observed in these mutant mice.²⁴⁵ While mice lacking G_z were viable and did not show any obvious neurological effects, G_z deficient mice displayed CNS defects.^{246, 247}

If the on-target side effects of opioids are also associated with G-protein signalling, individual probing of each of these subtypes might provide new directions for the discovery of opioids deprived of side effects. Moreover, it was recently proposed by Gillis et al. that rather than G-protein over β-arrestin bias, low intrinsic G-protein efficacy results in reduced side effects.¹⁵ Therefore, more understanding into the G-protein signalling mechanism of opioids is highly valuable. Faouzi et al. observed differing efficacy and order of potency for diverse opioids, including fentanyl, DAMGO and morphine. Interestingly, despite their structural similarities, 84b and 84c (Figure 26) showed unique signalling profiles across the G_i, G_o and G_z subtypes. 84b had a wide range of potencies (EC₅₀ values ranging from 350 nM for G_i3 to 8.8 nM for G_z) but displayed a narrow efficacy range (E_max from 80% for G_i1 to 92% for G_oA).²³⁷ In comparison, 84c exhibited a narrow potency range (EC₅₀ from 133 nM for G_i1 to 25 nM for G_z) but wide range of efficacies (E_max from 50% for G_z to 79% for G_i2).²³⁷ Of note, 84c exhibited marginal decreases in G-protein potency and lower intrinsic efficacy compared to typical MOR agonists such as DAMGO, fentanyl, morphine and oxycodone, and showed the lowest efficacy for the G_z subtype.

To investigate how its unique signalling profile compared to other opioid agonists, especially its lower efficacy at the G_z subtype translates in vivo, 84c (Figure 26) was assessed with a standard warm water tail-withdrawal assay in mice. 84c displayed supraspinal MOR-mediated analgesia in various rodent pain models with reduced abuse potential.²³⁷ Following i.c.v administration, 84c had an average antinociceptive dose (ED₅₀) of 18.77 nmol which was slightly higher than that of morphine (ED₅₀ = 6.6 nmol).²³⁷ It was suggested that such a unique in vivo profile for 84c might be due to its lower G_z efficacy and/or at all G_α subtypes.²³⁷ The bitopic ligands synthesized by Faouzi et al. are the first aiming to target the Na⁺- ABS concomitantly with the classical MOR OBS. They successfully synthesized valuable pharmacological tool compounds to study the distinct signalling pathways of the MOR, the β-arrestin pathway or those mediated by the individual G_i, G_o and G_z subtypes.

The KOR is an alternative therapeutic target in the development of nonaddictive pain therapeutics but it also mediates psychotomimesis, sedation and dysphoria which have limited the clinical application of KOR agonists as analgesic drugs.^1-3 Muratspahic et al. used the available crystal structure of KOR bound to the dual KOR/DOR opioid agonist MP1104 and computational methods to design peptide–small molecule conjugates of the KOR.²³⁸ They aimed to exploit the advantages provided by both moieties; the small molecule would bury deeply into the OBS while the peptide would make extensive interactions, aiming for specificity. These peptide–small molecule conjugates therefore follow the design concept of bitopic ligands with pharmacophores simultaneously occupying the OBS and alternative secondary binding sites. The small molecule would provide affinity for the KOR while the peptide, modulate selectivity and efficacy.

Muratspahic et al. designed and synthesized four different cyclic peptide–small molecule conjugates (DNCP-β-NalA) by coupling 5-amino acid residues de novo cyclic peptides to β-NalA.²³⁸ All four DNCP-β-NalA (1-4) conjugates showed nanomolar affinities for the mouse KOR, with DNCP-β-NalA-1 (85, Figure 27) showing the highest binding affinity, K_i = 3.9 nM, an 18-fold increase compared to β-NalA.²³⁸ In cAMP assays, the conjugates behaved as full agonists (E_max ranging from 89% to 101%), a considerable improvement in efficacy in comparison to β-NalA which behaves only as a partial agonist (EC₅₀ of 130 nM and Emax of 61%).²³⁸ DNCP-β-NalA-1 (85) and DNCP-β-NalA-4 were the most potent conjugates with EC₅₀ of 2.0 and 1.0 nM, respectively.²³⁸ Similarly, in the [³⁵S]GTPγS binding assays using CHO cells stably expressing the human KOR, DNCP-β-NalA-1 (85) was a full agonist (Table 6).²³⁸ In a BRET assay, DNCP-β-NalA-1 (85) only partially recruited β-arrestin-2 and β-arrestin-1 at KOR (Table 6).²³⁸ In TRUPATH assays to evaluate G_αi/o coupling, DNCP-β-NalA-1 (85) showed partial agonism at Gαi2, Gαi3, GαoA, GαoB and Gαgastducin subtypes with E_max values ranging from 48% to 77%, whereas it elicited full agonism at Gαi1 and Gαz with E_max values of 81% and 101%, respectively.²³⁸ These results indicate that DNCP-β-NalA-1 (85) shows preferential efficacy bias for G proteins over β-arrestins.

DNCP-β-NalA-1 (85, Figure 27) also exhibits an approximately 80-fold and 330-fold selectivity for KOR over DOR, and for KOR over NOR, respectively but was equipotent at the MOR.²³⁸ However, the conjugate is an antagonist at the MOR but an agonist at the KOR.²³⁸ This mix activity of biased agonism at the KOR and antagonism at the MOR might result in antinociception with a favorable side effect profile of DNCP-β-NalA-1 (85) observed in male mice.²³⁸ DNCP-β-NalA-1 (85) was shown to produce significant and potent antinociception in the formalin test and CFA-induced inflammatory hyperalgesia as well as anti-inflammatory properties, without KOR-mediated liability of motor dysfunction/sedation following subcutaneous (s.c.) administration in male mice.²³⁸

The cryo-EM structure of human KOR bound to DNCP-β-NalA-1 (85, Figure 27) and Gαi1/Gβ1/Gγ2 heterotrimer revealed that residues in ECL2/3 and TM6/7 controlled the functional selectivity of the human KOR.²³⁸ The mutations E209^ECL2A, E297⁶.58A and L309^7.32A resulted in an improvement in β-arrestin efficacy bias and G_i potency of DNCP-β-NalA-1 (85).²³⁸ Additionally, the small molecule part of the conjugate, βNalA, adopted a similar conformation as KOR agonists U50488 and MP1104 in the OBS while the cyclic peptide formed extensive interactions with ECL2/3 and TM6/7 residues.²³⁸ This supports the hypothesis that ECL2 can modulate bias at GPCRs.^49-51