Heterotrimeric G proteins transduce information to intracellular partners by modulating GTP binding and hydrolysis3. Through their interaction with G-protein-coupled receptors (GPCRs) and effectors, G proteins provide the transducer function that is necessary for the conveyance of extracellular information4,5. Heterotrimeric G proteins consist of an α subunit bound to a β and γ subunit dimer; they remain a trimer while the α subunit is bound to GDP6. Receptors provide the transmembrane conduit for a signal between the extracellular agonist and the intracellular G-protein transducer7. Specifically, GPCRs undergo a conformational change that acts to catalyse a reaction between the receptor and the Gα protein8,9,10. This interaction shifts the affinity for Gα binding to GDP to conditions that favour GDP release and GTP binding3. Thus, the receptor acts as a GEF and this reaction is considered to be primarily unidirectional11 (Fig. 1a). However, there have been observations that the GTP loading function of the receptor is reversible—that is, the receptor may facilitate the release of GTP from Gα. Early examples of this reversible interaction used nonhydrolysable forms of GTP such as GTPγS3,12, wherein the dissociation of radiolabelled GTPγS could be observed upon agonist binding to the receptor. One such study examined the kinetics of the release of 35S-GTPγS in cells expressing the mu opioid receptor (MOR) and found that the rate of release of nucleotide was increased as a function of a single saturating concentration of agonist and that partial and full agonists maintained their rank order efficacy in both exchange reactions13 (35S-GTPγS binding and 35S-GTPγS release).
Fig. 1: GPCRs induce both GTP binding and GTP release from Gα proteins and the process is agonist-mediated. a, Schematic of the proposed model, showing the conventional pathway of GDP-to-GTP exchange (left) and the expanded model to allow for both GTP and GDP release (right) as detailed in the linked Article14. A, agonist; R, receptor (asterisks indicate different active states); K a , affinity constant; G protein; G apo , unbound G protein (blue); G GDP , GDP-bound G protein (green); G GTP , GTP-bound G protein (red); α 1 and α 2 , active state affinities. b, DAMGO-stimulated binding and release in CHO-MOR cells presented as raw data (in disintegrations per minute (dpm)) and the normalization to baseline and maximum response. c–g, Normalized binding and release with indicated agonists in CHO-K1 cells expressing MOR (c; n = 3 binding, 3 release), KOR (d; n =3 binding, 5 release), 5-HT 1A R (e; n = 8 binding, 7 release), M 2 R (f; n = 3 binding, 3 release) and SST 2 R (g; n = 4 binding, 4 release). The raw data are presented in Extended Data Fig. 1. MPE, maximum possible effect. b–g, Data are mean ± s.e.m. and potency is presented as mean with 95% confidence interval. h, Comparison of potency in 35S-GTPγS binding versus release by unpaired, two-tailed t-test for each receptor comparing the individual potency (pEC 50 , where EC 50 is half-maximal effective concentration) values measured per experiment. Data are mean with 95% confidence interval. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001; NS, not significant (P > 0.05). Source Data Full size image
Here we investigate the significance of the release mechanism as a function of agonist concentration and how it can influence drug responsiveness in vivo. In a linked Article14, we present extensive pharmacological and biochemical characterizations of the release reaction, which we summarize in the three-state coupling model (Fig. 1a). In the study, we show that the GTP-release function of the receptor adheres to the pharmacological principles that pertain to the GTP-binding function of the receptor. In summary, the release function is dependent on agonist concentration and can be reversed by antagonists, and competitive interactions are preserved between orthosteric agonists and antagonists. Moreover, we demonstrate that the effect is due to activation of the receptor population and not merely a function of receptor occupancy. We provide experimental evidence and a functional state model that establishes that the efficacy and potency of an agonist to promote GTP release can differ from its efficacy and potency to induce GTP binding. Therefore, an agonist may demonstrate selectivity for affecting the equilibrium of the functional active state of the G protein and can show a preference for one state over the other. We also provide evidence that an agonist may have a different rank order potency and efficacy for the two states of the exchange function of a GPCR.
In the present study we show that agonists can induce both GTP binding and GTP release from Gα in a concentration-dependent manner, and that this can be observed for several different GPCRs (Fig. 1). GTP binding is assessed using a conventional method that entails incubating isolated cell membranes in the presence of 35S-GTPγS and increasing concentrations of agonists15,16. To observe GTP release, we use a ‘pulse-chase’ paradigm, which entails first loading the membrane preparations with 35S-GTPγS. Since many GPCRs are negatively regulated by sodium ions, removal of sodium allows for the constitutive activation of all sensitive receptors and the subsequent loading of 35S-GTPγS binding to G proteins. After the pulse, the chase entails dilution of the membranes and inclusion of an excess of unlabelled (cold) GTPγS in the presence of sodium (see Methods).
The two reactions are compared in Fig. 1b using membranes prepared from cells overexpressing mouse MOR. The data are presented as radioactivity counts for both the binding and the release assay; to facilitate comparison of the potencies, the data are also normalized to the baseline (0%) and the highest concentration used in each response (100%), and the curve is inverted for the release function. For the MOR, the potency of DAMGO ([D-Ala 2 , N-MePhe 4 , Gly-ol]-enkephalin), an enkephalin analogue, is conserved for both assays and the same is true for met-enkephalin (Fig. 1c). The exchange effect can also be observed for dynorphin A (1–17) at the kappa opioid receptor (KOR), serotonin (5-HT) at the serotonin 1A receptor (5-HT 1 A R), carbachol at the muscarinic 2 receptor (M 2 R) and somatostatin-14 (SST-14) at the somatostatin 2 receptor (SST 2 R) (Fig. 1d–g). Notably, agonist potencies at MOR and M 2 R are conserved in the two states, whereas dynorphin, serotonin and somatostatin are significantly more potent at their cognate receptors for promoting the release of GTP (Fig. 1h; individual curves are shown in Extended Data Fig. 1).
There are multiple clinically relevant opioid agonists that span a broad range of pharmacological characteristics (including partial agonists and biased agonists); therefore, we used these tool compounds to determine whether the release function and binding function could be dissociated at the MOR. In addition, we tested two new compounds, which were selected on the basis of their scaffold variation from biased MOR agonists introduced by our laboratory (the SR series—for example, SR-17018) and for their characteristics as full agonists that are less potent than morphine in cellular assays. The latter consideration was based on a desire to not introduce more potent opioid agonists to the scientific literature. For each drug, DAMGO was assayed in parallel to serve as a reference, since DAMGO maintains the same potency in both responses and serves to define the maximum efficacy in both assays in this cell line. Not unexpectedly, several agonists perform similarly to DAMGO, preserving the potency in both responses; however, some agonists show a differential preference for potency (Fig. 2a) and/or efficacy (Fig. 2b) for one state over the other (see Extended Data Fig. 2 for curves and Extended Data Table 1 for parameters). Since the two effects were measured in parallel with DAMGO, we also determined the difference in transduction efficiencies (ΔΔlogR; Fig. 2c and Extended Data Table 1) for each agonist in the release assay and the binding assay. This representation permits normalization between responses to directly compare agonist activity17. The two new agonists show significant gains in the release function, having nearly a hundred-fold gain in selectivity for the release active state, as measured by the difference in transduction efficiencies; we have named these compounds muzepan1 and muzepan2 (Fig. 2d), as they are mu opioid receptor-acting compounds containing an ‘azepane’ ring.
Fig. 2: Opioid agonists exhibit differential preferences for GTP binding and release in CHO-MOR cell membranes. a–c, Comparisons of the mean of the individual 35S-GTPγS binding and release: potencies (a), maximum efficacies (E max normalized to DAMGO (100% versus baseline (0%)) (b), and difference in transduction efficiencies (ΔΔlogR, relative to DAMGO) (c).Data are mean with 95% confidence interval. Unpaired t-test was used for comparing binding and release parameters for each compound. Extended Data Fig. 2 shows concentration–response curves and Extended Data Table 1 presents parameters and number of individual replicates (n ≥ 3). d, Chemical structures of muzepan1 and muzepan2 with binding affinities (pK i with s.e.m., n = 6; K i is the inhibition constant) determined from competition binding assays with 3H-naloxone. Source Data Full size image
Several of the agonists that show a state preference have previously been identified as biased agonists that prefer G-protein signalling over β-arrestin2 recruitment (for example, oliceridine18, PZM2119, herkinorin20, buprenorphine21,22,23 and SR-1701824). When tested in the cellular assays expressing the human MOR that were used to evaluate the biased agonism of SR-1701824, both muzepan1 and muzepan2 show no preference between GTPγS binding and β-arrestin2 recruitment (Extended Data Fig. 3a and Extended Data Table 2). Moreover, the exchange selectivity for GTP release over binding is maintained at the human receptor (Extended Data Fig. 3b, Extended Data Table 3). Therefore, whereas many of the compounds that showed selectivity for release over binding also show preference for G-protein binding over β-arrestin2 recruitment, the correlation is not absolute.
To demonstrate the physiological significance of agonist-induced GTP release, the experiment was repeated in mouse spinal cord membranes. In the binding experiment, DAMGO promotes only a 40% stimulation in the native tissue ((1.4 ± 0.01)-fold; P < 0.001, paired t-test versus baseline; Extended Data Fig. 4a). We determined that the sodium-free conditions lead to very high levels of GTPγS binding, making it difficult to see an effect of DAMGO on release. This is not unexpected, as there are relatively low levels of MOR in the system, as reflected by the approximately 40% stimulation in the binding studies (Extended Data Fig. 4a). Therefore, to isolate the MOR-accessible G-protein pool, we used DAMGO in the presence of sodium and 35S-GTPγS in the pulse phase and diluted 100-fold as part of the chase (Methods and ref. 14). We demonstrate that this is feasible in the CHO-mMOR cell line, where the potencies of DAMGO are similar to those in the sodium-free loading conditions, although the potency for the release function is slightly decreased (19 nM binding versus 43 nM release; P < 0.05, t-test; Extended Data Fig. 4b). In mouse spinal cord membranes, when 100 nM DAMGO is included in the pretreatment period, there is an increase of about 10% in 35S-GTPγS loading (P < 0.001; paired t-test), but no change is evident in the spinal cord of MOR-knockout mice (Extended Data Fig. 4c). We therefore took this modest stimulation as representative of MOR-mediated GTPγS loading in the mouse spinal cord.
In mouse spinal cord membranes, DAMGO is more potent in stimulating 35S-GTPγS binding than muzepan1 and muzepan2, whereas all agonists are full agonists (Fig. 3a). In the release paradigm, DAMGO loses potency, whereas muzepan1 and muzepan2 gain potency (Fig. 3b). Notably, the efficacy obtained by muzepan1 and muzepan2 reach the 10% maximal effect anticipated in the pulse loading (Extended Data Fig. 4c). This is in contrast to DAMGO, which does not reach this plateau, suggesting that in spinal cord, the enkephalin-like agonist may be selective against release. No significant effects were observed in spinal cord membranes from MOR-knockout mice (Extended Data Fig. 4d); therefore, the effects are likely to be due to MOR activation.
... continue reading