+ Disruption of the Na ion binding site as a mechanism for positive allosteric modulation of the mu- receptor

Kathryn E. Livingston and John R. Traynor1

Department of Pharmacology, University of Michigan, Ann Arbor, MI 48109

Edited by Leslie Lars Iversen, University of Oxford, Oxford, United Kingdom, and approved November 13, 2014 (received for review August 6, 2014) Positive allosteric modulation of the mu- (MOPr), of in vitro and in vivo literature describing allostery at GPCRs the site of action of all clinically used , represents a poten- (9–11). In contrast, apart from our initial description of mu-PAMs, tial approach for the management of pain. We recently reported very little is known about allostery at MOPr. on positive allosteric modulators of MOPr (mu-PAMs), a class A G Allosteric modulators exhibit probe dependence, meaning they protein coupled receptor (GPCR). This study was designed to show disparate effects depending on the agonist bound to the examine the mechanism of allostery by comparing the degree to orthosteric site (12). A striking example of this is LY2033298, which opioid ligand structure governs modulation. To do this we a PAM of the muscarinic acetylcholine receptors M2 and M4. examined the interaction of the mu-PAM, BMS-986122, with LY2033298 increases the affinity of the agonist oxotremorine, a chemically diverse range of MOPr orthosteric ligands. Generally, while having no effect on the binding of the agonists pilocarpine for full agonists BMS-986122 enhanced the binding affinity and and McN-A-343 (13). Currently, it is not known if all opioid potency to activate G protein with no alteration in the maximal agonists are equally sensitive to the PAM effect of BMS-986122, effect. In contrast, lower efficacy agonists including nor the mechanism underlying the allosteric modulation. Our were insensitive to alterations in binding affinity and showed little initial characterization showed that BMS-986122 causes a shift 2 4 to no change in potency to stimulate G protein. Instead, there was in the potency of the agonist DAMGO ([D-Ala , N-MePhe , an increase in maximal G protein stimulation. Antagonists were Gly-ol]-), but increases the maximal stimulation of G unresponsive to the modulatory effects of BMS-986122. Sodium is protein by morphine (1). Opioid ligands are extremely diverse, a known endogenous allosteric modulator of MOPr and alters ranging from the 31-amino acid endogenous peptide β-endorphin to orthosteric agonist affinity and efficacy. The sensitivity of an small alkaloids like morphine. Therefore, this study sought to an- orthosteric ligand to BMS-986122 was strongly correlated with its swer two questions: (i) does BMS-986122 show probe dependence sensitivity to NaCl. In addition, BMS-986122 decreased the ability for the orthosteric ligand? And (ii) if probe dependence is seen, of NaCl to modulate agonist binding in an allosteric fashion. what is the mechanistic basis for this? Overall, BMS-986122 displayed marked probe dependence that To address these questions we examined the effect of BMS-

was based upon the efficacy of the orthosteric ligand and can be 986122 on the MOPr properties of a wide range of opioid ligands PHARMACOLOGY explained using the Monod–Wyman–Changeux two-state model + from endogenous peptides to small molecules (Fig. S2). The of allostery. Furthermore, disruption of the Na ion binding site results reveal that the PAM effects of BMS-986122 are de- may represent a common mechanism for allosteric modulation of pendent on the efficacy of the orthosteric ligand and not on the class A GPCRs. structure per se. We find a strong correlation between the pos- + itive action of BMS-986122 and the negative action of Na ions endogenous opioids | | GTPgammaS | ligand binding | receptor states Significance he mu-opioid receptor (MOPr) is the site of action of all Tclinically used opioid drugs. MOPr is a class A G protein- Morphine and related compounds are the gold standard for the coupled receptor (GPCR) that activates heterotrimeric Gi/o management of pain. Such drugs bind to the orthosteric site on proteins. Clinically used opioid agonists bind to the orthosteric the mu-opioid receptor (MOPr), a G protein-coupled receptor site on MOPr and although they are efficacious at causing pain (GPCR). We have proposed that targeting an allosteric site on relief, have a number of unwanted side effects resulting from MOPr could result in improved pain management and have direct MOPr activation. We have recently discovered and pre- reported positive allosteric modulators (PAMs) of MOPr. High- + sented a preliminary characterization of positive allosteric modu- resolution X-ray structures have identified a Na binding site + lators of MOPr (mu-PAMs) and are currently pursuing the idea on multiple GPCRs and have shown how bound Na stabilizes that mu-PAMs could be a viable way to manage pain (1, 2). The inactive receptor states. Here we demonstrate that PAM ac- ligand BMS-986122 (Fig. S1) represents the most active mu-PAM tivity at MOPr allosterically disrupts Na+ binding, thereby + currently identified. It was discovered in a high-throughput screen forming an active-state receptor. The Na binding site is highly for its ability to enhance the recruitment of β-arrestin to MOPr conserved across class A GPCRs so this may represent a funda- by the agonist -1. Although having little agonist mental mechanism of allosteric modulation. activity on its own, this modulator has the ability to enhance the affinity, potency, and/or maximal response of MOPr agonists. In A preliminary report of some of the findings presented in this paper was given at the 45th annual meeting of the International Research Conference, July 13–18, the same systems, BMS-986122 has no activity when the delta 2014, Montreal, Canada. opioid receptor (DOPr) is expressed, indicating the importance Author contributions: K.E.L. and J.R.T. designed research; K.E.L. performed research; of MOPr for BMS-986122 activity. K.E.L. and J.R.T. analyzed data; and K.E.L. and J.R.T. wrote the paper. The study of allosteric modulation of GPCRs has recently gained The authors declare no conflict of interest. momentum (3) and represents a relatively unexplored avenue for This article is a PNAS Direct Submission. drug development (4, 5). Allosteric modulators have been dis- 1To whom correspondence should be addressed. Email: [email protected]. covered for several GPCRs including the muscarinic, cannabinoid, This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. and metabotropic glutamate receptors (6–8) with a growing body 1073/pnas.1415013111/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1415013111 PNAS | December 23, 2014 | vol. 111 | no. 51 | 18369–18374 Downloaded by guest on September 25, 2021 to inhibit agonist binding. Moreover, we also show that BMS- Table 1. Binding affinity of MOPr ligands in the presence or + 986122 allosterically inhibits the ability of Na ions to reduce absence of BMS-986122 agonist binding. The PAM effect of BMS-986122 can conse- + Ligand Ki(Veh), nM Ki(BMS), nM Ki(Veh)/Ki(BMS) quently be explained by an inhibition of the ability of Na ions to stabilize the inactive state of the receptor, thereby allowing the Peptides receptor to shift to an active conformation. Thus, the mechanism β-Endorphin 194 ± 13 47 ± 8*** 4.1 ± ± of positive allosteric modulation can be simply explained by the Endomorphin-1 104 32 17 8* 6.1 ± ± two-state Monod–Wyman–Changeux model of allosterism (14). Leu-Enk 664 67 100 15** 6.6 Met-Enk 423 ± 133 63 ± 16* 6.7 Results Small molecules ± ± We first investigated the effects of a maximally effective con- 0.5 0.1 0.4 0.2 1.3 ± ± centration (10 μM; ref. 1) of BMS-986122 on the MOPr activity 2.4 0.3 1.5 0.8 1.6 ± ± of a range of endogenous opioid peptides (Fig. S2A). Using cell 222 34 89 25* 2.5 215 ± 54 14 ± 2* 15.4 membrane homogenates prepared from C6 glioma cells stably (RS)- 1076 ± 85 100 ± 4*** 10.7 expressing MOPr (C6MOPr) (15), we performed radioligand ± ± 3 (R)-Methadone 382 10 36 10*** 10.6 competition binding assays using H- (DPN, an ± ± γ (S)-Methadone 6,358 2,065 896 35* 7.1 ) in the presence of GTP S and NaCl to gen- Morphine 163 ± 18 143 ± 42 1.1 erate an inactive receptor state known to predominate in native 2.5 ± 0.7 3.4 ± 1.0 0.7 membranes (16, 17). As shown in Fig. 1A and Table 1, BMS- 986122 caused an approximate sixfold enhancement in the Affinity (Ki values) were determined by competitive displacement of 3 affinity of both methionine-enkephalin (Met-Enk) and leucine- H-DPN binding from C6MOPr cell membranes in a buffer containing 10 μM γ enkephalin (Leu-Enk). A similar increase in affinity was seen for GTP S and 100 mM NaCl as described in Methods, in the absence or presence of 10 μM BMS-986122. *P < 0.05, **P < 0.01, ***P < 0.001 compared with control the smaller putative endogenous peptide endomorphin-1 (18). In (vehicle) data by Student t test. Data shown are means ± SEM of three to five independent experiments each in duplicate.

addition to enhancing its affinity to bind MOPr, BMS-986122 caused a leftward shift in the concentration-response for Met- Enk to activate G protein, with no alteration in the maximal response (Emax), as measured by GTPγ35S binding in membrane homogenates (Fig. 1C and Table 2); an effect also seen with Leu-Enk and endomorphin-1 (Table 2). The endogenous opioid β-endorphin, a much larger 31-amino acid peptide, was also modulated by BMS-986122 with rightward shifts in both the af- finity (fourfold; Table 1) and potency (sixfold; Table 2) to acti- vate G protein. Because we had previously seen that BMS-986122 increased the maximal G protein activation by morphine (1), we de- termined the modulation of opioid affinity, potency, and maxi- mal agonist effect of this small molecule MOPr agonist. There was no shift in the affinity of morphine to bind MOPr in the presence of BMS-986122 (Fig. 1B and Table 1). Even at 30μM BMS-986122, a concentration approaching the limits of solubil- ity, there was still no enhancement of morphine affinity [Fig. S3; Ki (veh) = 163 ± 18 nM, Ki (BMS-986122; 30 μM) = 250 ± 110 nM)]. In contrast to this lack of effect on affinity, BMS-986122 did alter the ability of morphine to activate G protein. There was a small 2.9-fold decrease in the potency (EC50), but the most striking effect was a significant increase in the degree of maximal activation. In the presence of BMS-986122, morphine was able to activate G protein to nearly the same extent as the full agonist DAMGO (Fig. 1D and Table 2). Moreover, the rate at which DAMGO activated G protein was unchanged in the presence of Fig. 1. Comparison of the effect of BMS-986122 on Met-Enk and morphine. μ 3 10 M BMS-986122 whereas the rate of morphine-activated G H-DPN binding to MOPr was determined in membranes from C6MOPr cells E in the presence of increasing concentrations of Met-Enk (A) or morphine (B) protein was enhanced (Fig. 1 ). in the presence of vehicle (●)or10μM BMS-986122 (○). The ability of in- We hypothesized that the disparate effects on orthosteric li- creasing concentrations of Met-Enk (C) or morphine (D) to elicit GTPγ35S gand binding seen with BMS-986122 might be explained by the binding in C6MOPr cell membranes was measured in the absence (●)or structure of the ligand: peptides versus small molecules. All of presence (○)of10μM BMS-986122 as described in Methods. Data are pre- the endogenous ligands tested are large, flexible peptides, whereas sented as % stimulation of a maximal concentration (10 μM) of the full morphine is a small, rigid molecule. To address this possibility, we agonist standard DAMGO. (E) Rate of G protein activation by DAMGO measured the effect of BMS-986122 on additional small molecule (dotted line, circles) or morphine (solid line, squares) was measured in the MOPr agonists that are structurally distinct from morphine, namely presence of vehicle (closed symbols) or 10 μM BMS-986122 (open symbols). γ35 ± buprenorphine, fentanyl, methadone, and loperamide as well as the The rate of DAMGO-stimulated GTP S was unchanged (vehicle: 21.5 1.0 B cpm/min; BMS-986122: 20.8 ± 0.7 cpm/min), whereas the rate of morphine- antagonist naloxone (Tables 1 and 2 and Fig. S2 ). Buprenorphine stimulated GTPγ35S was enhanced from 11.5 ± 0.7 cpm/min (vehicle) to behaved like morphine, showing no increase in binding affinity for 35 18.1 ± 0.5 cpm/min in the presence of BMS-986122. All plotted points are MOPr or potency in the GTPγ S assay, but a marked concen- means ± SEM of three to five independent experiments, each in duplicate. tration-dependent enhancement in maximal effect (Fig. S4). With

18370 | www.pnas.org/cgi/doi/10.1073/pnas.1415013111 Livingston and Traynor Downloaded by guest on September 25, 2021 Table 2. Stimulation of [35S]GTPγS binding by MOPr ligands in the absence or presence of BMS-986122 Vehicle BMS-986122 (10 μM)

† † Ligand EC50, nM Max, % EC50, nM Max, % EC50(veh)/EC50(BMS)

Peptides β-endorphin 167 ± 21 90 ± 628± 5** 88 ± 66.0 Endomorphin-1 70 ± 494± 821± 3* 93 ± 53.3 Leu-Enk 116 ± 21 106 ± 427± 9** 98 ± 64.3 Met-Enk 169 ± 16 103 ± 249± 7** 107 ± 13.5 Small molecules Buprenorphine 0.5 ± 0.2 21 ± 30.5± 0.2 52 ± 1*** 1.0 DAMGO 104 ± 39 100 19 ± 7 100 5.5 Etorphine 0.28 ± 0.04 99 ± 2 0.17 ± 0.05 102 ± 51.6 Fentanyl 459 ± 173 61 ± 3 156 ± 24 85 ± 1*** 2.9 Loperamide 37 ± 186± 23.9± 0.6*** 91 ± 19.5 (RS)-Methadone 542 ± 149 85 ± 147± 6* 85 ± 311.5 (R)-Methadone 273 ± 12 95 ± 431± 0.6*** 87 ± 38.8 (S)-Methadone 7,737 ± 854 70 ± 4 1,275 ± 216** 85 ± 3* 6.1 Morphine 292 ± 59 61 ± 2 100 ± 11* 85 ± 1*** 2.9 Naloxone dns dns

The [35S]GTPγS assay was performed as described in Methods.*P < 0.05, **P < 0.01, ***P < 0.001 compared with control (vehicle) data by Student t test. Data shown are means ± SEM of three to five independent experi- ments each in duplicate. d.n.s., did not stimulate. †Maximal values relative to the stimulation observed with 10 μM DAMGO.

fentanyl, there was a small (2.5-fold) increase in its affinity for (21–23), by stabilizing an inactive state of the receptor (24–26). + MOPr, together with a 2.9-fold shift in potency and an increase in Agonists vary in their response to Na ions such that sensitivity + the maximal level of ligand-stimulated GTPγ35S binding. Con- to Na ions generally correlates with the degree of intrinsic ac- versely, for the small molecule (RS)-methadone, there was a 10.9- tivity, with a continuum from antagonists that are insensitive to + fold shift in the MOPr binding affinity (Fig. 2A and Table 1) and Na ions to full agonists that are the most sensitive. Because this a large (11.5-fold) shift in the potency of (RS)-methadone to ac- finding matches the responsiveness to BMS-986122, we hypoth- tivate G protein, with no increase in the maximal effect (Fig. 2B esized that there would be a correlation between the sensitivity γ35

and Table 2). Hill slopes of binding and GTP S assay data for all of a ligand to the mu-PAM and the sensitivity of a ligand to the PHARMACOLOGY + compounds were not significantly different from 1.0. We further presence of Na ions. Competition binding curves for orthosteric characterized the effect of BMS-986122 on (RS)-methadone- ligands were performed in Tris buffer in the absence or presence mediated G protein activation by performing a series of concentra- of NaCl/GTPγS to calculate the ratio of binding affinity (as pKi tion-response curves in the presence of increasing levels of BMS- values) to active and inactive states of MOPr. Under both con- 986122 (Fig. 2C). Analyses of these curves using the allosteric ditions competition binding curves with Hill slopes not signifi- ternary complex model afforded an alpha (α)valueforfunctional cantly different from one were obtained for all compounds (Fig. cooperativity between (RS)-methadone and BMS-986122 of 18.4 S5). pKi values under the two conditions were then compared and an affinity (Kb) of 1.7 μM. This contrasts with our previously with the shift in affinity (Fig. 3A) or potency (Fig. 3B) of the published cooperativity value of 8 for the interaction between orthosteric ligand caused by 10 μM BMS-986122. We observed endomorphin-1 and BMS-986122 recruitment of β-arrestin (1). a strong correlation between an orthosteric ligand’s loss of binding + For loperamide, there was a similar 15-fold shift in affinity and affinity in the presence of Na /GTPγS and its increased affinity or a 9.5-fold shift in the potency (Tables 1 and 2). potency in the presence of BMS-986122. As expected, there was The above experiments used racemic methadone. The (R)- a relationship between the shift in potency and affinity of ortho- isomer has a higher affinity for MOPr and a higher analgesic steric ligands caused by BMS-986122 (Fig. S6). + potency than the (S)-isomer (19, 20). We therefore examined To analyze whether Na ions and BMS-986122 were antago- whether the differential binding of the isomers to the MOPr nistic, we investigated the ability of BMS-986122 to inhibit the + affected the response to BMS-986122. The affinities of the in- effect of Na ions on agonist binding. As expected NaCl showed dividual isomers for MOPr as well as the racemate were all en- a concentration-dependent inhibition of DAMGO binding, de- 35 hanced in the presence of BMS-986122 (Table 1). In the GTPγ S termined as the inhibition of an EC50 concentration of DAMGO assay for (R)-methadone, there was an 8.8-fold shift in potency (10 nM) to displace 3H-DPN (0.2 nM) (Fig. 3C). This gave an in the presence of BMS-986122, with no change in maximal G inhibitory concentration 50 (IC50) value for NaCl of 6 ± 1 mM. protein stimulation. In contrast (S)-methadone, which is a partial Addition of BMS-986122 resulted in a concentration-dependent agonist, responded to the presence of BMS-986122 with a sixfold rightward shift of the IC50 of NaCl to 16 ± 2 mM in the presence shift in potency as well as an enhancement of the maximal of 3 μM BMS-986122 and 36 ± 8 mM in the presence of 10 μM stimulation. There was no effect of BMS-986122 on the affinity BMS-986122. This effect was saturable as the curve failed to shift of the antagonist naloxone for MOPr (Table 1), nor did BMS- any further in the presence of 30 μM BMS-986122 (Fig. 3C). + 986122 impart any agonist activity to naloxone (Table 2). Finally, to confirm a role for Na ions we examined the ability The above findings suggest that it is the degree of agonist of BMS-986122 to alter etorphine binding and activity. Etor- efficacy of the orthosteric ligands rather than their chemical phine (Fig. S2B) is a potent full agonist at MOPr, but is relatively + structure that governs the response to BMS-986122 and so the insensitive to Na /GTP (Ki = 0.18 ± 0.03 nM in Tris and Ki = + observed probe dependence. Sodium ions are known to reduce 2.4 ± 0.3 nM with Na /GTPγS) compared with other full ago- the affinity of agonists to bind to GPCRs, including the MOPr nists (16). We therefore hypothesized that etorphine would be

Livingston and Traynor PNAS | December 23, 2014 | vol. 111 | no. 51 | 18371 Downloaded by guest on September 25, 2021 increase in maximal ability to stimulate GTPγ35S binding with little or no change in potency or affinity in the presence of BMS- 986122. Opioid peptides have a message and address sequence (29, 30) and occupy more of the MOPr binding pocket than smaller organic molecules (31) which could explain their different sensitivities to BMS-986122. However, our findings are not simply a matter of chemical structural differences in terms of peptide versus small molecules because (R)-methadone and loperamide were sensitive to BMS-986122 in the same ways as the endogenous peptides. Indeed, the allosteric action of BMS-986122 on these compounds was much greater than for the peptides. Morphine, as well as fentanyl, buprenorphine, and (S)-meth- adone have reduced efficacy compared with the endogenous peptides and (R)-methadone or loperamide. We have previously reported that binding of the opioid antagonist DPN was un- affected by BMS-986122 (1) and have now demonstrated a lack of modulation of the opioid antagonist naloxone. Thus, a ratio- nal explanation for our findings is that the observed probe selectivity is dependent on agonist efficacy. Table 3 lists the intrinsic efficacy of the opioid compounds examined, determined by the method of Ehlert (32) using values for agonist affinity de- Fig. 2. BMS-986122 enhances the affinity and potency of (RS)-methadone. γ 3 termined in the presence or absence of NaCl/GTP S and agonist (A) Displacement of H-DPN in C6MOPr membranes was measured in the potency in the GTPγ35S assay. The compounds with efficacy presence of increasing concentrations of (RS)-methadone in the presence of β vehicle (●)or10μM BMS-986122 (○). (B) The ability of increasing concen- greater than or equal to -endorphin [namely etorphine, Leu-Enk, trations of (RS)-methadone to elicit GTPγ35S binding in C6MOPr cell mem- loperamide, Met-Enk, (RS)- and (R)-methadone, endomorphin branes was measured in the absence (●) or presence (○)of10μM BMS- 1], all behaved similarly in their response to BMS-986122 with an 986122. Experiments were performed as described in Fig 1. (C) Stimulation increase in potency and ligand affinity, but no change in maximal of GTPγ35S binding by (RS)-methadone was performed in the presence of in- response. In contrast, the lower efficacy agonists showed an in- creasing concentrations (0.3–10 μM) of BMS-986122. Data were analyzed using crease in maximal effect with minimal alteration in potency or the allosteric ternary complex model as described in Methods. Data shown are affinity, resulting in an increase in intrinsic efficacy (Table 3). With means ± SEM of three independent experiments, each in duplicate. morphine we also demonstrated an increase in the rate of G protein activation, a property associated with ligand efficacy (33). Finally (S)-methadone, that has an efficacy value between that of less sensitive to BMS-986122. Indeed, BMS-986122 caused no morphine and DAMGO, showed a shift in potency and an in- shift in the affinity of etorphine (Table 1). In GTPγ35S binding crease in maximal effect. assays, there was also no significant shift in potency and no al- In a simple two-state model of GPCR activation, receptors are teration in the level of maximal stimulation (Table 2). hypothesized to exist in conformational states that differ dra- Discussion matically in their affinity for orthosteric agonists. Agonists have higher affinity for the active, G protein bound state and prefer- In this study we show that the mu-PAM BMS-986122 exhibits entially stabilize this conformation, thus propelling agonist- marked probe dependence across a variety of structurally diverse induced activation of G protein and downstream cellular responses agonists acting at the orthosteric site on MOPr. The sensitivity of (34, 35). The inactive receptor state has lower affinity for ortho- + orthosteric ligands to BMS-986122 correlated with the ratio of steric agonists and is stabilized in the presence of Na ions as well agonist affinities for active and inactive states of the receptor + as guanine nucleotides that drive uncoupling of receptor and G defined by the absence or presence of Na ions and guanine protein. The differential affinity of orthosteric agonists used in this nucleotide and was consistent with the hypothesis that probe study for inactive and active MOPr states varied, but showed dependence of BMS-986122 is defined by the efficacy of the a strong correlation with their relative sensitivity to BMS-986122 orthosteric agonist. The effects of BMS-986122 are in line with and with their potency to stimulate GTPγ35S binding. Thus, BMS- the Monod-Wyman-Changeux two-state model of allosterism 986122 appears to shift the equilibrium toward the active receptor (14, 27) involving a single active state, bound to G protein, and state thereby modulating the activity of the orthosteric agonists single inactive receptor state, uncoupled from G protein and + differentially, depending on their efficacy, an action that is opposite stabilized by Na ions. BMS-986122 favors the active state, + + to the effects of Na ions (23). The action of BMS-986122 on opposes the action of Na ions, and therefore positively modu- maximal response of the lower efficacy agonists is also in line with lates the properties of the orthosteric agonist. a two-state model of GPCR function. Using the idea that efficacy is Although the degree of effect varied, BMS-986122 enhanced based upon an agonist’s ability to shift the equilibrium of receptors the affinity and potency of all endogenous opioid peptides tested toward an active state we would expect BMS-986122, by stabilizing and none showed any enhancement of maximal G protein activa- an active receptor state, to enhance the efficacy of partial agonists. tion. We initially hypothesized that the 31-amino acid β-endorphin The appropriateness of the two-state model of allosterism to explain may be a bitopic ligand, capable of binding to both the orthosteric the probe dependence of BMS-986122 at MOPr mirrors the action site as well as the allosteric site (for review, see ref. 28). Thus, we of the M1 muscarinic acetylcholine receptor modulator BQCA predicted that β-endorphin would compete with BMS-986122 as (benzylquinone carboxylic acid), which enhances the potency of well as with 3H-DPN for binding to MOPr. However, BMS-986122 orthosteric ligands or their maximal response depending on the affected β-endorphin in a manner similar to other endogenous efficacy requirements of the signaling assay (27). ligands, suggesting that it is not bitopic, at least for the allosteric If a two-state model does explain the action of BMS-986122 site occupied by BMS-986122. then we should expect to see agonist activity of the modulator In contrast to the peptides, there were differences seen with even in the absence of orthosteric agonist (27). BMS-986122 small molecule MOPr agonists, indicating orthosteric probe de- alone fails to activate G protein to a detectable level as measured + pendence. Morphine, buprenorphine, and fentanyl showed an by the GTPγ35S assay, even with a lowered Na concentration to

18372 | www.pnas.org/cgi/doi/10.1073/pnas.1415013111 Livingston and Traynor Downloaded by guest on September 25, 2021 + Fig. 3. Relationship between BMS-986122 and Na ions. (A) pKi values were obtained for each ligand in the presence and absence of 100 mM NaCl and

10 μM GTPγS with or without 10 μM BMS-986122. The ratio of the calculated pKi values is plotted. (B)pEC50 values were obtained for each ligand in the absence or presence of 10 μM BMS-986122. The ratio of these values is plotted compared with the ratio of pKi values with and without NaCl/GTPγSasinA.(C) Using C6MOPr membranes, DAMGO (10nM) was incubated with 0.2 nM 3H-DPN and increasing concentrations of NaCl with 10μM GTPγS in the presence of vehicle (●), 3 μM BMS-986122 (triangles with dashed line), 10 μM BMS-986122 (○), or 30 μM BMS-986122 (■). Data shown are means ± SEM of 3–10 in- dependent experiments each in duplicate. In A and B:1,β-endorphin; 2, Buprenorphine; 3, DAMGO; 4, Endomorphin-1; 5, Etorphine; 6, Fentanyl; 7, Leu-Enk; 8, Loperamide; 9, Methadone-R; 10, Methadone-RS; 11, Methadone-S; 12, Met-Enk; 13, Morphine; 14, Naloxone.

+ increase apparent efficacy (Fig. S7) (36). Downstream of G pro- increase in agonist affinity in the absence of Na ions (1) sug- teins, MOPr orthosteric agonists inhibit adenylate cyclase (AC), gesting it can stabilize an active receptor conformation. Second, a response that is more sensitive to lower efficacy compounds, the same degree of shift in the inhibitory effect of NaCl on due to increased amplification (37). At high concentrations, DAMGO binding is seen with both 10 μM and 30 μM BMS- BMS-986122 does inhibit AC (1), thus confirming the appro- 986122, suggesting saturation is reached as expected for allo- priateness of the two-state model. In addition, we would predict sterism, rather than the surmountable parallel shifts expected if + BMS-986122 to have activity at even more amplified downstream the antagonism were competitive (40). Third, the Na -H2O signaling pathways (27). cluster binding pocket is conserved between MOPr and DOPr An apparent anomaly to the hypothesis that probe dependence (24) and indeed across all Class A GPCRs (26), yet the PAM is based on orthosteric ligand efficacy and can be explained by a activity of BMS-986122 is selective for MOPr over DOPr (1). two-state model is that etorphine shows no cooperativity with In conclusion, this study further confirms the use of the Monod– + – BMS-986122, and is relatively insensitive to Na ions, yet this Wyman Changeux two-state model of allosterism as the simplest PHARMACOLOGY compound is a highly efficacious MOPr agonist (Table 3). Bio- mechanism to explain PAM activity at GPCRs (27). Furthermore, physical studies with the β2-adrenergic receptor have shown that the results directly relate the action of a small molecule PAM to + + agonists destabilize the receptor, but the receptor exists in a variety interference with Na binding at a GPCR. Disruption of the Na of conformationally heterogeneous states that are not fully stabi- binding pocket during receptor activation may be a general mech- lized unless G protein is bound (38). Thus, it is feasible that etor- anism for allosteric modulation across many class A GPCRs that phine promotes a state that has very high affinity for G protein, but + have a conserved Na -H2O cluster binding pocket. For example, the enriches this population to a lesser extent than other agonists. CB1 receptor PAM ORG27569 and the M1 receptor PAM BQCA Further support for the two-state model comes from the re- + both enhance the high affinity state of their respective receptors lationship between BMS-986122 and Na ions. High-resolution X-ray structures of several class A GPCRs (25), including the + + DOPr (24), have identified the Na site as a Na -H2O cluster in Table 3. Efficacy of MOPr agonists a cavity in the middle of the 7-transmembrane (TM) helices. This Efficacy cavity in the DOPr is formed by side chains of 16 residues dis- tributed across TM domains 2, 3, 6 and 7 and is fully conserved in Ligand Vehicle BMS-986122 the MOPr (24). Comparison of inactive and active GPCR structures and molecular dynamics simulations performed with Etorphine 4.7 5.0 the adenosine A2A receptor show that agonist binding causes Leu-Enk 3.6 2.3 molecular rearrangements that are not compatible with concur- Loperamide 2.9 2.1 + rent Na binding (25). In particular, the addition of agonist DAMGO 2.1 2.0 + dramatically reduces the size of the binding pocket for the Na - Met-Enk 1.8 1.2 + (RS)-Methadone 1.3 1.3 H2O cluster. BMS-986122 decreased the potency of Na ions to inhibit binding of the agonist DAMGO and there was a strong Endomorphin-1 1.2 0.8 correlation between the opposite effects of the BMS-986122 and (R)-Methadone 1.1 0.9 + Na /GTP on opioid ligands. Thus, the binding of both BMS- β-endorphin 1.0 1.2 + 986122 and Na to MOPr is incompatible. We conclude that (S)-Methadone 0.6 0.7 BMS-986122 favors the active receptor conformation with dis- Fentanyl 0.5 0.7 + Morphine 0.5 1.0 ruption of the Na -H2O cluster binding pocket and this explains + its PAM activity. The interaction between Na and BMS-986122 Buprenorphine 0.2 0.5 could be through direct competition, for example as with the Naloxone 0 0 diuretic amiloride (39), or indirectly through an allosteric in- Ligands are listed in order of decreasing efficacy (vehicle). Efficacy values teraction. However, the evidence favors an allosteric mechanism were calculated from the ligand affinity and [35S]GTPγS data in Tables 1 and for several reasons. First, BMS-986122 does cause a small 2 according to the method of Ehlert (32) as described in Methods.

Livingston and Traynor PNAS | December 23, 2014 | vol. 111 | no. 51 | 18373 Downloaded by guest on September 25, 2021 3 (27, 41). This may, however, not be true for all agonists, in particular, H-DPN Assays. Assays were performed using C6MOPr membranes as de- + 3 those whose affinity is less sensitive to Na ions such as etorphine, scribed (15). Competitive displacement of 0.2 nM H-DPN was measured + or receptors such as the β1 adrenergic receptor where Na does not using increasing concentrations of orthosteric ligand in the presence of ve- appear to be involved in the transition from inactive to active states hicle (1% DMSO) or BMS-986122. + (42). Nonetheless interference with the stability of the Na -H O 35 2 GTPγ S Assays. Assays were performed using C6MOPr membranes as de- binding pocket of GPCRs may be a generally applicable mechanism scribed (15). Orthosteric and allosteric ligands were included where appro- that provides a basis for the discovery of novel modulators and priate. Rate experiments (Fig. 1E) are described in SI Methods. the identification of potential endogenous modulators (for re- view, see ref. 43). Data Analysis. Data were analyzed using GraphPad Prism version 6 (GraphPad) to

give Ki values and EC50 values. The method of Ehlert (32) was used to Methods calculate agonist efficacy as described in SI Methods. Functional cooper- Materials. [3H]Diprenorphine and [35S]GTPγS were from PerkinElmer Life ativity (α) between BMS-986122 and methadone and affinity (Kb) of BMS- Sciences. BMS-986122 [2-(3-Bromo-4-methoxyphenyl)-3-[(4-chlorophenyl) 986122 were obtained using an allosteric ternary complex model (45) as sulfonyl]-thiazolidine] was from Bristol-Myers Squibb. Morphine sulfate, described in SI Methods. leucine- and methionine-enkephalin, β-endorphin, DAMGO, , and endomorphin-1 and all chemicals were from Sigma-Aldrich. Other ACKNOWLEDGMENTS. We thank Drs. Arthur Christopoulos and Meritxell opioids were from the Opioid Basic Research Center at the University of Canals for helpful discussions on allostery and critical reading of the Michigan. Cell culture materials were from Invitrogen. manuscript. We thank Drs. Neil Burford and Andrew Alt for the generous gift of BMS-986122. This work was funded by the National Institutes of Health Grant R01 DA035316 (to J.R.T.) and an Endowment for Develop- Cell Lines and Membrane Preparation. Generation of C6 rat glioma cells ment of Graduate Education (EDGE) award (to K.E.L.) from the Endowment heterologously expressing only the mu-opioid receptor (C6MOPr; Bmax = 2.1 for the Basic Sciences, University of Michigan. K.E.L. was also supported by pmol/mg protein) and membrane preparation was as described (44). the National Institutes of Health Training Grant T32 DA007267.

1. Burford NT, et al. (2013) Discovery of positive allosteric modulators and silent allosteric 23. Selley DE, Cao CC, Liu Q, Childers SR (2000) Effects of sodium on agonist efficacy for modulators of the μ-opioid receptor. Proc Natl Acad Sci USA 110(26):10830–10835. G-protein activation in μ-opioid receptor-transfected CHO cells and rat thalamus. Br J 2. Burford NT, Traynor JR, Alt A (2014) Positive allosteric modulators of the μ-opioid Pharmacol 130(5):987–996. receptor: A novel approach for future pain medications. Br J Pharmacol, 10.1111/ 24. Fenalti G, et al. (2014) Molecular control of δ-opioid receptor signalling. Nature bph.12599. 506(7487):191–196. 3. Luttrell LM, Kenakin TP (2011) Refining Efficacy: Allosterism and Bias in G Protein- 25. Gutiérrez-de-Terán H, et al. (2013) The role of a sodium ion binding site in the allo- Coupled Receptor Signaling, eds Luttrell LM, Ferguson SSG (Humana Press, Totowa, NJ). steric modulation of the A(2A) adenosine G protein-coupled receptor. Structure – 4. Nickols HH, Conn PJ (2014) Development of allosteric modulators of GPCRs for 21(12):2175 2185. treatment of CNS disorders. Neurobiol Dis 61:55–71. 26. Katritch V, et al. (2014) Allosteric sodium in class A GPCR signaling. Trends Biochem Sci – 5. Wootten D, Christopoulos A, Sexton PM (2013) Emerging paradigms in GPCR allo- 39(5):233 244. 27. Canals M, et al. (2012) A Monod-Wyman-Changeux mechanism can explain G protein- stery: Implications for drug discovery. Nat Rev Drug Discov 12(8):630–644. coupled receptor (GPCR) allosteric modulation. J Biol Chem 287(1):650–659. 6. O’Brien JA, et al. (2004) A novel selective allosteric modulator potentiates the activity 28. Lane JR, Sexton PM, Christopoulos A (2013) Bridging the gap: Bitopic ligands of of native metabotropic glutamate receptor subtype 5 in rat forebrain. J Pharmacol G-protein-coupled receptors. Trends Pharmacol Sci 34(1):59–66. Exp Ther 309(2):568–577. 29. Chavkin C, Goldstein A (1981) Specific receptor for the : 7. Price MR, et al. (2005) Allosteric modulation of the cannabinoid CB1 receptor. Mol Structure—activity relationships. Proc Natl Acad Sci USA 78(10):6543–6547. Pharmacol 68(5):1484–1495. 30. Schwyzer R, Kriwaczek VM, Wunderlin R (1981) A method for mapping peptide re- 8. May LT, et al. (2007) Structure-function studies of allosteric agonism at M 2 muscarinic ceptors. Naturwissenschaften 68(2):95–96. – acetylcholine receptors. Mol Pharmacol 72(2):463 476. 31. Manglik A, et al. (2012) Crystal structure of the μ-opioid receptor bound to a mor- 9. Kruse AC, et al. (2013) Activation and allosteric modulation of a muscarinic acetyl- phinan antagonist. Nature 485(7398):321–326. – choline receptor. Nature 504(7478):101 106. 32. Ehlert FJ (1985) The relationship between muscarinic receptor occupancy and ad- 10. Ma L, et al. (2009) Selective activation of the M1 muscarinic acetylcholine receptor enylate cyclase inhibition in the rabbit myocardium. Mol Pharmacol 28(5):410–421. achieved by allosteric potentiation. Proc Natl Acad Sci USA 106(37):15950–15955. 33. Alt A, McFadyen IJ, Fan CD, Woods JH, Traynor JR (2001) Stimulation of guanosine-5′- 11. Byun NE, et al. (2014) Antipsychotic drug-like effects of the selective M4 muscarinic ace- o-(3-[35S]thio)triphosphate binding in digitonin-permeabilized C6 rat glioma cells: tylcholine receptor positive allosteric modulator VU0152100. Neuropsychopharmacology Evidence for an organized association of mu-opioid receptors and G protein. JPharmacol 39(7):1578–1593. Exp Ther 298(1):116–121. 12. Keov P, Sexton PM, Christopoulos A (2011) Allosteric modulation of G protein-cou- 34. Kenakin T (2013) New concepts in pharmacological efficacy at 7TM receptors: IUPHAR pled receptors: A pharmacological perspective. Neuropharmacology 60(1):24–35. review 2. Br J Pharmacol 168(3):554–575. 13. Valant C, Felder CC, Sexton PM, Christopoulos A (2012) Probe dependence in the 35. Chung KY, et al. (2011) Conformational changes in the G protein Gs induced by the allosteric modulation of a G protein-coupled receptor: Implications for detection and β2 adrenergic receptor. Nature 477(7366):611–615. validation of allosteric ligand effects. Mol Pharmacol 81(1):41–52. 36. Szekeres PG, Traynor JR (1997) Delta opioid modulation of the binding of guanosine- 14. Monod J, Wyman J, Changeux JP (1965) On the nature of allosteric transitions: A 5′-O-(3-[35S]thio)triphosphate to NG108-15 cell membranes: Characterization of ag- plausible model. J Mol Biol 12:88–118. onist and inverse agonist effects. J Pharmacol Exp Ther 283(3):1276–1284. 15. Clark MJ, Harrison C, Zhong H, Neubig RR, Traynor JR (2003) Endogenous RGS protein 37. Clark MJ, Linderman JJ, Traynor JR (2008) Endogenous regulators of G protein sig- action modulates mu-opioid signaling through Galphao. Effects on adenylyl cyclase, naling differentially modulate full and partial mu-opioid agonists at adenylyl cyclase – extracellular signal-regulated kinases, and intracellular calcium pathways. J Biol Chem as predicted by a collision coupling model. Mol Pharmacol 73(5):1538 1548. 38. Nygaard R, et al. (2013) The dynamic process of β(2)-adrenergic receptor activation. 278(11):9418–9425. Cell 152(3):532–542. 16. Lee KO, Akil H, Woods JH, Traynor JR (1999) Differential binding properties of or- 39. Gao ZG, Ijzerman AP (2000) Allosteric modulation of A(2A) adenosine receptors by ipavines at cloned mu- and delta-opioid receptors. Eur J Pharmacol 378(3):323–330. amiloride analogues and sodium ions. Biochem Pharmacol 60(5):669–676. 17. Carroll JA, Shaw JS, Wickenden AD (1988) The physiological relevance of low agonist 40. Arunlakshana O, Schild HO (1959) Some quantitative uses of drug antagonists. Br affinity binding at opioid mu-receptors. Br J Pharmacol 94(2):625–631. Pharmacol Chemother 14(1):48–58. 18. Zadina JE, Hackler L, Ge LJ, Kastin AJ (1997) A potent and selective endogenous ag- 41. Ahn KH, Mahmoud MM, Kendall DA (2012) Allosteric modulator ORG27569 induces onist for the mu- receptor. Nature 386(6624):499–502. CB1 cannabinoid receptor high affinity agonist binding state, receptor internalization, and 19. Scott CC, Robbins EB, Chen KK (1948) Pharmacologic comparison of the optical iso- Gi protein-independent ERK1/2 kinase activation. J Biol Chem 287(15):12070–12082. – mers of methadon. J Pharmacol Exp Ther 93(3):282 286. 42. Miller-Gallacher JL, et al. (2014) The 2.1 Å resolution structure of cyanopindolol- + 20. Kristensen K, Christensen CB, Christrup LL (1995) The mu1, mu2, delta, kappa opioid bound β1-adrenoceptor identifies an intramembrane Na ion that stabilises the ligand- receptor binding profiles of methadone stereoisomers and morphine. Life Sci 56(2): free receptor. PLoS ONE 9(3):e92727. PL45–PL50. 43. Christopoulos A (2014) Advances in G protein-coupled receptor allostery: From 21. Pert C, Pasternak G, Snyder S (1973) Opiate agonists and antagonists discriminated by function to structure. Mol Pharmacol 86(5):463–478. receptor binding in brain. Science 182:1359–1361. 44. Emmerson PJ, et al. (1996) Characterization of opioid agonist efficacy in a C6 glioma 22. Simon EJ, Groth J (1975) Kinetics of opiate receptor inactivation by sulfhydryl re- cell line expressing the mu opioid receptor. J Pharmacol Exp Ther 278(3):1121–1127. agents: Evidence for conformational change in presence of sodium ions. Proc Natl 45. Christopoulos A, Kenakin T (2002) G protein-coupled receptor allosterism and Acad Sci USA 72(6):2404–2407. complexing. Pharmacol Rev 54(2):323–374.

18374 | www.pnas.org/cgi/doi/10.1073/pnas.1415013111 Livingston and Traynor Downloaded by guest on September 25, 2021