β1AR NMR constructs are competent for coupling to G proteins.
In this work we reversed the E130W3.41 mutation39 in the β1AR-W construct20, producing the less thermostabilised construct β1AR-E (Supplementary Fig. 1), to facilitate examination of the receptor active state and to evaluate ternary complex formation with G proteins by NMR spectroscopy.
We assessed the functional integrity and pharmacological profiles of the β1AR-E and thermostabilised β1AR-W receptor constructs in cellular and biophysical assays examining ligand binding and agonist-dependent G protein coupling, validating the suitability of β1AR-E for use in studying receptor activation. Studies examining β1AR activation commonly use the therapeutic full agonist isoprenaline in place of the native ligand adrenaline17,41–43. Consequently, we measured isoprenaline binding affinities of the β1AR constructs in HEK293T cells expressing receptors N-terminally tagged with Nluc44 (Supplementary Fig. 2a). NanoBRET signal loss by competition of β1AR antagonist (S)-propanolol-red45 (KD - β1AR-W: 2.81 ± 0.56 nM; β1AR-E: 3.09 ± 0.30 nM, Supplementary Fig. 2b) with isoprenaline at varying concentrations demonstrated low micromolar Ki binding with no significant difference (p = 0.65) in affinities between β1AR-W and β1AR-E (Fig. 1a), indicating ligand binding was little affected by this thermostabilising mutation.
We next assessed G protein coupling using the TRUPATH BRET2 biosensor assay46 (Supplementary Fig. 2c) expressed in HEK293T cells, allowing examination of signalling via the canonical coupling partner Gs. Comparison of isoprenaline-stimulated concentration response curves showed that both β1AR receptor constructs could activate Gs (Fig. 1b). As expected, for a thermostablised construct that favours the inactive states, the β1AR-W construct displayed reduced (p = 0.0098) isoprenaline potency (β1AR-W (pEC50: 7.23 ± 0.06) compared to the less thermostablised construct β1AR-E (pEC50: 7.77 ± 0.10). Therefore, the TRUPATH and ligand binding assays confirmed that reversal of the thermostabilising mutation increased receptor activity without impacting ligand affinity. Also, although less efficacious, the thermostabilised β1AR-W construct still retained functional G protein coupling.
We further characterised ternary complex formation in vitro using a bio-layer interferometry (BLI) assay. This technique facilitated determination of affinity (KD) and kinetics of G protein binding (kon, koff) to β1AR, under similar solution conditions used for NMR spectroscopy. Receptor purified in detergent was immobilised to streptavidin-coated biosensors via an N-terminal biotinylated avi-tag on the receptor (Fig. 1c) to ensure the receptor’s cytoplasmic binding surface remained accessible to bulky binding partners. In anticipation of NMR experiments, we examined ternary complex formation of isoprenaline-bound receptor with the engineered mini-Gs protein47,48 (Supplementary Fig. 2d). The lower molecular weight of mini-G proteins due to removal of the α-helical domain makes them an attractive means to study ternary G protein complexes in solution by NMR, resulting in smaller signal linewidths and improved sensitivity. Binding of mini-Gs to both β1AR-E and β1AR-W was observed (Fig. 1d) with the binding curves agreeing with a 1:1 interaction stoichiometry. Furthermore, it was observed that the β1AR-E construct demonstrated higher mini-Gs affinity (consistent with the results of the TRUPATH assay) primarily due to increased kon rate relative to β1AR-W, confirming the increase in receptor activity following reversal of the thermostabilising mutation. Minimal binding was observed in the absence of ligand (Supplementary Fig. 2e), indicating functional complex formation was agonist specific and basal receptor activity was low, consistent with the literature49. As anticipated, the application of apyrase (Supplementary Fig. 2f) had little effect upon mini-Gs binding, since the mini-G proteins contained a mutation that made them insensitive to nucleotide binding when coupled to GPCRs47. Agonist-dependent binding to both constructs was also shown for the active state-stabilising nanobody, Nb6B9. Again, the more active construct β1AR-E showed increased binding affinity compared to β1AR-W (Supplementary Fig. 2g, Supplementary Table 1). Together, these data illustrated the functionality of receptor constructs both in vivo and purified in detergent and confirmed that the reversal of E130W3.41 to wildtype E1303.41 resulted in a receptor with greater capacity for activation.
The β1AR-E solution active state is similar to the static β1AR-Gs ternary structure but has differences in the cytoplasmic binding interface.
We hypothesised that the more active β1AR-E construct would access the active state with a much higher occupancy than in our previous work with β1AR-W19,20. Therefore, we used the β1AR-E construct to characterise the conformational properties of the active state in solution by NMR spectroscopy. We labelled β1AR-E with 13C methyl methionine, facilitating examination of the NMR response to full agonist on the natively occurring methionine positions M15334.57 (IL2), M1784.62 (EL2), M2235.54 (TM5), M2836.28 (TM6), and M2966.41 (TM6) (Fig. 2a) assigned in previous work using β1AR-W20. We assessed conformational changes of the receptor upon isoprenaline addition through 2D 1H-13C HMQC methionine NMR shift correlation experiments and used peak changes in both dimensions compared to the apo state spectrum to inspect differences in conformations and dynamics (Supplementary Fig. 3). Overlaying β1AR-E apo- and isoprenaline-bound spectra revealed significant differences in peak positions that were indicative of conformational changes across the whole receptor upon agonist binding (Fig. 2b). The affected residues changed their peak positions in directions indicative of overall receptor activation as indicated by our previous spectral assignments20. The isoprenaline-bound spectrum varied substantially from the corresponding spectrum of β1AR-W (Supplementary Fig. 4a), which occupied a pre-active state, and looked similar to the previously recorded ternary state spectrum of β1AR-W bound to isoprenaline and Nb6B920 (Supplementary Fig. 4b). Therefore, the full agonist bound β1AR-E occupied a global conformation considerably closer to the solution Nb6B9 ternary complex than β1AR-W, which we concluded was representative of the solution active state.
In the receptor core, M2235.54 and M2966.41 showed marked shifts which indicated significant conformational rearrangements around the nearby highly conserved P5.50-I3.40-F6.44 motif50. The latter has been discussed to play a crucial role in signal transduction from the binding pocket and stabilization of the active state51. A comparison of the structures of inactive β1AR with the active full agonist and G protein bound β1AR ternary complex showed conserved aromatic residues F2996.44 and Y2275.58 adopt orientations which would result in a strong shielding effect on M2235.54 and deshielding effect on M2966.41, respectively (Supplementary Fig. 4c-d). These changes were consistent with the 1H shift perturbations we observed. Further, the large upfield 13C shift for M2966.41 indicated a more gauche orientation of the χ3 angle in the agonist-bound state, in agreement with the ternary complex structure. Therefore, changes to both M2235.54 and M2966.41 signals strongly indicated that in the solution isoprenaline-bound state, the core of the receptor approached the conformation observed in the ternary complex structure. Additionally, the upfield shift in 13C for M2235.54 indicated a greater degree of trans-gauche conformational exchange, and line broadening of both M2235.54 and M2966.41 suggested that the adopted state of β1AR-E remained dynamic on the µs-to-ms timescale and sampled different local environments. These observations implied an increase in the dynamics of the PIF motif region upon agonist binding, and a substantially more dynamic receptor core relative to the β1AR-W pre-active state.
Effects of agonist binding also permeated to the extracellular and intracellular surfaces. M1784.62 at the top of TM4 resolved into a single peak in the isoprenaline-bound β1AR-E spectrum relative to two peaks in the apo and β1AR-W pre-active states (Fig. 2b and Supplementary Fig. 4a), indicating reduced conformational variability as the receptor orthosteric binding pocket formed favourable interactions with the agonist. Also, the IL2 probe M15334.57, which is distant from the ligand binding pocket (~ 26 Å), showed a more substantial upfield 1H shift upon addition of isoprenaline (Fig. 2b) relative to β1AR-W (Supplementary Fig. 4a), indicating greater activation20 and allosteric conformational changes to either IL2 or the cytoplasmic tip of TM3 as the effect of agonist binding propagated to the intracellular region.
Significant rearrangement of TM6 is associated with class A GPCR ternary complex formation52. Therefore, we assessed conformational changes in this key region using the introduced probe L289M6.34 which occupies a less solvent-accessible location between M2836.28 and M2966.41 (Fig. 2a). The equivalent mutation was previously introduced into β2AR with only modest effects on ligand affinity10 and the non-disruptive nature of the L289M6.34 mutation was supported by the similarity of the native methionine signal positions between L289M6.34 and L2896.34 spectra, in the apo and isoprenaline-bound conditions (Supplementary Fig. 4e). We observed a pronounced 1H downfield shift of L289M6.34 upon isoprenaline addition, indicating a major conformational rearrangement of TM6 upon agonist binding (Fig. 2d). This peak change agreed with movement of L289 towards Y2315.62 in the β1AR-Gs ternary structure (PDB: 7JJO)41 and the paramagnetic relaxation enhancement (PRE)-based β2AR isoprenaline-bound structure (PDB: 6KR8)25 where full agonist binding induces rotation of the lower half of TM6 (Supplementary Fig. 4f). Therefore, L289M6.34 indicated TM6 underwent a significant conformational change in solution upon isoprenaline addition. In contrast, no such movement was observed in the static inactive cyanopindolol-bound or isoprenaline-bound receptor structures (PDB: 2YCY and 2Y03)43,53.
Corroborating information was obtained for the cytoplasmic tip of TM6 through 1D 19F NMR in combination with single site 19F-labelling at the previously studied A282C6.27 position19 (Supplementary Fig. 5a). The isoprenaline-bound spectrum showed a peak which could be deconvoluted into two components (Supplementary Fig. 5b). The major of the two signals was shifted downfield, indicating a greater solvent accessibility than the minor peak based on a correlation between solvent accessibility and chemical shift observed for β1AR-W19 and confirmed using Gd3+ solvent PRE experiments (Supplementary Fig. 5c). These changes in solvent accessibility were consistent with helical rotation of TM6 to move A2826.27 from behind TM5 into the cytoplasm (Supplementary Fig. 5d), which we hypothesised represented the active state observed with L289M6.34. Conversely, the minor, less accessible peak indicated a conformation in which A2826.27 was still positioned behind TM5, likely representing the pre-active state. Relaxation dispersion (CPMG) experiments did not show exchange on the µs-to-ms timescale between the two states (Supplementary Fig. 5e) and although slower timescales were not probed, slow ms-to-s exchange would be consistent with large helical movements, agreeing with L289M6.34 observations.
Additional information on the nearby NPxxY motif in TM7 and neighbouring IL4 was obtained using 19F-TET labelled C3447.54 (Supplementary Fig. 5a). This position has been examined previously in β2AR with the E122W3.41 mutation54, and in our previous work on β1AR-W which revealed the isoprenaline-bound TETC3447.54 as a single µs-to-ms exchange broadened signal19. The corresponding probe in β1AR-E resulted in an even broader peak (Fig. 2e) that could be deconvoluted into two components (Supplementary Fig. 5f). The more intense TETC3447.54 upfield component was unique to β1AR-E, whereas the weaker component matched the β1AR-W construct peak19. Therefore, we concluded that full agonist-bound β1AR-E predominantly adopted a new conformation in the TM7/ICL4 region but also populated to a lesser extent the pre-active state. The observation of two conformations was similar to that made for extracellular signals of β1AR using 1H-15N NMR18. The more severe signal broadening in the 2D 1H-13C NMR spectra likely prevented the observation of the minor pre-active state signals. Notably, the populations of the states was similar to those for A282C6.27,BTFA, indicating an allosteric link between TM6 and TM7 in the pre-active/active state equilibrium. Furthermore, the upfield location for the TETC3447.54 major peak indicated a more hydrophobic environment relative to the pre-active state19. Occlusion of Y7.53 in a hydrophobic layer was observed in agonist-bound MD simulations of the β2AR, adenosine A2A (A2AR), and κ-opioid receptors55,56, and similarly found in the β2AR PRE structure (PDB: 6KR8)25. Our data support a similar β1AR-E active state conformation with the probe and adjacent Y3437.53 in the hydrophobic layer, suggesting that the cytoplasmic binding cavity remains occluded in the full agonist bound state. We therefore concluded that whilst the β1AR-E active state conformation was similar to ternary complex static structures around the PIF motif TM6 underwent a characteristic rotation of its helix. In the absence of a cytoplasmic coupling partner, however, displacement of TM6 from the receptor core had not occurred, thus retaining the hydrophobic layer around the NPxxY motif. In contrast, the minor pre-active state peak showed smaller rotation of TM6 and reduced TM7 rotation into the hydrophobic layer as we previously demonstrated19.
β1AR-E ternary complex formation with mini-Gs progresses by conformational selection but with further induced fit from the active state.
After establishing the similarities and differences between the full agonist-bound solution active conformation of β1AR and the ternary complex static structures, we questioned what further conformational changes were required to reach the solution ternary complex conformation. We therefore characterised receptor conformational changes upon transition from the isoprenaline-bound active state to the ternary state by recording 2D 1H-13C HMQC experiments of the β1AR-E/isoprenaline/mini-Gs ternary complex. None of the NMR probes had direct interactions with proximal binding partner residues, thus changes to chemical shifts reported receptor conformational changes upon mini-Gs binding.
The ternary complex 2D 1H-13C HMQC spectrum showed peak locations were reminiscent of isoprenaline-bound positions (Fig. 2c), indicating the solution ternary state receptor conformation was similar to the solution active state. Therefore, the major conformational changes involved in receptor activation occurred upon agonist binding rather than coupling to mini-Gs, in contrast to our previous studies using thermostabilised β1AR-W with nanobody20. Nevertheless, notable smaller differences for all methionine probes were observed, indicating there were further minor conformational adjustments upon mini-Gs binding. Further, the generally greater peak intensities of the ternary complex spectrum showed that the ternary state was less dynamic than the agonist-bound state (Supplementary Fig. 6a). Together, these observations suggested that whilst the receptor bound to agonist was already primed in a conformation suitable to engage with mini-Gs, further structural adjustments were required to form the ternary complex, resulting in a more rigid receptor.
Positions of peaks corresponding to residues located in the extracellular and core transmembrane regions were only mildly affected by mini-Gs binding. The change for M1784.62 reflected the established allosteric contraction of the ligand binding pocket that accompanies ternary complex formation18,57,58 which results in an increased affinity of the bound agonist59. The minor 1H shift change for M2235.54 indicated only a small adjustment relative to F2996.44 in the PIF motif, confirming the active state was already similar in conformation to the solution ternary state in this conserved region. The M2235.54 13C shift indicated residual dynamics of the side chain via χ3 trans/gauche conformational exchange in agreement with published structures showing M2235.54 with a range of χ3 dihedral angles. However, the peak intensity increased by 2.4-fold (Supplementary Fig. 6a), indicating that the region around M2235.54, including the PIF motif, adopted a more rigid conformation than in the ligand-bound active state of β1AR-E.
M2966.41 also showed a subtle change in proton shift, likely in response to readjustments of the PIF motif and Y2275.58 as it engaged with Y3437.53 on TM7. The substantial 3.6-fold increase in peak intensity (Supplementary Fig. 6a) indicated that the TM5-TM6 interface around M2966.41 was also less dynamic and thus stabilised by mini-Gs binding.
In IL2, closer to the binding pocket, only a minor change in the 1H dimension was observed for M15334.57 which implied small differences in the surrounding IL2 conformation upon coupling to mini-Gs, consistent with structural comparisons (Supplementary Fig. 6b). Conversely, the M15334.57 13C shift moved more appreciably towards a gauche conformation, suggesting reduced dynamics in the wider IL2 and TM3 cytoplasmic region in the presence of binding partner in agreement with overall receptor rigidity increases.
The L289M6.34 and M2836.28 probes, located on the cytoplasmic side of TM6, showed chemical shift changes more similar in magnitude to those observed upon isoprenaline addition to the apo receptor. This confirmed that the major conformational changes accompanying ternary complex formation concentrated at the cytoplasmic binding interface.
A downfield shift in 1H of the L289M6.34 probe (Fig. 2d) suggested movement towards the proximal side chain of Y2315.62 and supported further opening of the cytoplasmic cavity via adjustments in TM6 orientation upon mini-G protein binding. This was consistent with the smaller L2896.34-Y2315.62 distance observed in ternary complex (PDB:7JJO)41 relative to the agonist-bound β2AR PRE structure (PDB: 6KR8)25 (Supplementary Fig. 4f). M2836.28 at the tip of TM6 appeared as a sharp peak indicating fast local dynamics consistent with a solvent exposed environment. The 13C chemical shift remained in the rapid trans/gauche exchange shift range, supporting a high flexibility of this region and in agreement with the ternary structure (PDB: 7JJO)41 in which M2836.28 was unresolved. For the adjacent A282CBTFA,6.27 19F probe, addition of mini-Gs resulted in a single sharp signal that was downfield shifted relative to the ligand-bound states (Supplementary Fig. 6d-e). The smaller linewidth relative to the agonist-bound state indicated faster dynamics which agreed with M2836.28, likely due to increased flexibility and solvent accessibility. Together, M2836.28 and A282CBTFA,6.27 suggested an increase to solvent exposure and flexibility in the TM6 cytoplasmic tip upon binding partner coupling, consistent with additional displacement farther into the cytoplasm.
Addition of mini-Gs to isoprenaline-bound TETC3447.54 resulted in a 19F downfield shift (Fig. 2e) consistent with a transition to a more hydrophilic solvent exposed environment. Therefore, this shift supported breaking of the hydrophobic layer and extension of Y3437.53 into the cytoplasmic cavity. This suggests outward displacement of TM6, and hence opening of the cytoplasmic cavity, that only occurs upon complex formation. Consequently, the TETC3447.54 ternary complex spectrum further supported our proposed agonist-bound active state conformation. Interestingly, the TETC3447.54 signal deconvoluted as two components with minor differences in shifts but differing linewidths (Supplementary Fig. 6f). This observation, coupled with the lack of proximity to either the agonist or mini-Gs binding surfaces, suggested that the receptor region near IL4 sampled two subtly different environments in the mini-Gs complex, and implied IL4 retains conformational dynamics in ternary complexes.
We confirmed further β1AR-E conformational changes were present upon ternary complex formation with the active-state stabilising nanobody Nb6B9. Although generally similar, discrepancies in peak positions and intensities relative to the mini-Gs ternary state were prominent in the cytoplasmic IL2, TM6, and TM7/IL4 regions (Supplementary Fig. 7). This variation implied differences in local receptor conformation and complex dynamics, emphasising the plasticity and conformational adaptability of β1AR to accommodate structurally different coupling partners.
Overall, our NMR data suggested that further conformational changes, including altering of the cytoplasmic binding cavity through TM6 displacement, were necessary to reach the final ternary complex conformational state in solution. We concluded that full agonist ternary complex formation progressed mainly via conformational selection of the active state followed by smaller amounts of induced fit upon mini-Gs binding.
Ternary Gs complex formation via the pre-active state follows an induced fit mechanism.
In our previous studies, isoprenaline and partial agonist-bound β1AR-W occupied the inactive/pre-active state equilibrium19,20. Likewise, based on our 1H-13C and 19F NMR data we found partial agonist (xamoterol, intermediate efficacy; salbutamol, high efficacy)60 bound β1AR-E populated the same inactive/pre-active state equilibrium, with no evidence of active state populations (Fig. 3a,3c-e, Supplementary Fig. 8). This observation suggested partial agonist bound receptor was unable to overcome the energy barrier to occupy the active state. Consequently, we utilised this to investigate the aptitude of the pre-active state to interact with mini-Gs to form ternary complexes and whether intrinsic differences between the pre-active state and the active state would result in conformational differences in the respective ternary complexes with mini-Gs.
Salbutamol and xamoterol had decreased potency in TRUPATH assays relative to isoprenaline, indicating reduced signalling through β1AR-E (Supplementary Fig. 9a-b). Similarly, BLI assays with salbutamol and xamoterol-bound β1AR-E demonstrated significantly (p < 0.01) reduced mini-Gs binding affinities relative to the equivalent isoprenaline assays (Fig. 3b). The increased koff rates indicated partial agonist ternary states were slightly less stable than for those with full agonist (Supplementary Fig. 9c, Supplementary Table 2), agreeing with assays using fluorescent BODIPY-GTPγS binding to complexes in other studies61. However, as with isoprenaline-bound β1AR-W, the lower kon rates relative to isoprenaline-bound β1AR-E were the major contributor to differences in KD (Supplementary Fig. 9c, Supplementary Table 2) and therefore suggested the more occluded pre-active state had a higher energy barrier to ternary complex formation compared to the active state.
We also conducted BLI assays with partial agonist-bound β1AR-W, which showed only minor decreases in mini-Gs binding affinity compared to β1AR-E (Fig. 3b) mainly influenced by the modestly higher koff rates, indicating similar partial agonist-bound receptor pre-active state populations. Following the kinetic investigations of the pre-active state to mini-Gs binding, we characterised the pre-active state induced ternary complex conformation. The 1H-13C NMR spectral fingerprint of isoprenaline-bound β1AR-W in complex with mini-Gs was very similar to the ternary complex with β1AR-E. Although differences in the vicinity of the W130E3.41 mutation that affected the core region residues (M2235.54, M2966.41), the cytoplasmic region (L289M6.34, M2836.28, M15334.57) was very similar (Supplementary Fig. 10a-b), indicating that the final ternary complex conformation of β1AR was mostly independent of the initial engagement of mini-Gs with the active or pre-active state of the receptor.
In 1H-13C HMQC spectra of partial agonist bound β1AR-E ternary complexes with mini-Gs (Fig. 3f-g) residual ligand-bound peaks were still present consistent with the lower mini-Gs binding affinities. However, the methyl probe positions corresponding to the ternary complexes were reminiscent of the equivalent peaks of the complex with isoprenaline (Fig. 3h). These data indicated a similar global receptor conformation in solution for mini-Gs ternary complexes irrespective of coupling to receptor bound to full agonist or partial agonist. Notably, the chemical shifts of probes on the cytoplasmic region of TM6, L289M6.34 and M2836.28, were comparable between pre-active and active state induced ternary complexes (Supplementary Fig. 10c-d). Corroboratory 19F NMR experiments with A282CBTFA,6.27 also showed minimal variation between the mini-Gs complexes with isoprenaline-bound β1AR-E and isoprenaline-bound β1AR-W, or with either salbutamol- or xamoterol-bound β1AR-E (Supplementary Fig. 11a). Therefore, we concluded the TM6 conformation and degree of outward displacement was conserved between the ternary complexes, likely reflecting stabilisation by the binding partner.
However, other NMR probes did display minor differences. M1784.62 showed adjustments to the ligand binding pocket whilst M15334.57 displayed changes to 1H shifts reporting on small differences to the IL2 conformation (Fig. 3f-h, Supplementary Fig. 10c-e). These changes were consistent with our previous work which showed these residues to be a sensitive measure of receptor activation and ligand efficacy20.
Slight changes to chemical shifts for M2235.54 and M2966.41 in the receptor transmembrane core (Supplementary Fig. 10e-f) indicated partial agonists influenced the conformational exchange of the PIF motif and TM5-TM6 interactions.
Relative to their equivalent ligand-bound 19F NMR spectra, the salbutamol- or xamoterol-bound β1AR-E and isoprenaline-bound β1AR-W ternary complexes showed downfield shifted peaks for TETC3447.54 (Fig. 3c) confirming the anticipated breaking of the hydrophobic layer in the NPxxY and IL4 region seen during ternary complex formation. In agreement with the different orientations of Y3437.53 as stipulated for the ternary complex formed via the β1AR-E active state, the signals deconvoluted into the same two peak components (Supplementary Fig. 11b-d). However, the complexes formed via the pre-active state preferably populated the more upfield shifted signal component, in contrast to the complexes formed via the active state. Consequently, we hypothesised ternary formation via the active state resulted in greater population of a fully extended Y3437.53 conformation, maintaining the water-mediated hydrogen bond3 between Y5.58 and Y7.53. The greater stability of the hydrogen bond agreed with the higher stability of isoprenaline-bound β1AR-E shown via BLI (Supplementary Fig. 9c). Conversely, complex formed via the pre-active state shifted equilibria towards a conformation with reduced extension of Y3437.53 without the hydrogen bond, thus reducing complex stability. Variation of Y5.58 was also consistent with the subtle changes observed for M2235.54 and M2966.41.
Together, our observations suggested ternary complex formation via the pre-active state required a greater induced fit contribution than via the active state but resulted in similar conformations, though partial agonists still imparted minor differences in the receptor core and binding interface periphery.
Kinetics of ternary complex formation contribute to G protein selectivity of β1AR.
Having characterised the primary coupling of Gs to β1AR, we endeavoured to study the kinetics and dynamics of secondary coupling of non-canonical binding partners using the methodologies we established for Gs. We assessed functional activation of Gq and the Gi/o members Gi1 and Go1 by β1AR using the aforementioned BRET2-based TRUPATH assay46. Concentration-response curves were generated which inferred functional coupling to both β1AR-W and β1AR-E for Gi1, Go1 and Gq, with reduced potencies for the β1AR-W (p = 0.0041, p = 0.029, p = 0.097 respectively) (Fig. 4b). The changes in potencies scaled similarly for each binding partner between the two receptor constructs, which implied receptor thermostabilisation had little effect on G protein subfamily selectivity. Only a minor preference towards Gs activation was observed, similar to observations for β2AR 46.
Similar to mini-Gs, minimalised forms of other G proteins have been established48. We measured binding affinities for these proteins by BLI. The Gi1 equivalent, mini-Gi1, had poor thermal stability which prevented reliable quantitative analysis of binding kinetics, although low affinity binding was observed (data not shown).
Since Go1 showed similar activation to Gi1 in the TRUPATH assays, and these proteins have 73% identity (Supplementary Fig. 12a), the engineered mini-Go1 was substituted for mini-Gi1. Additionally, the chimeric mini-Gs/i (a chimera of mini-Gs with the receptor-interacting residues of the α5 helix replaced by Gi1 residues (Supplementary Fig. 12b) and shown to have Gi1-like characteristics48), and equivalent mini-Gs/q chimera were also characterised. The two chimeras were established on the observation that the majority of residues interacting with the receptor were in α5 and thus were proposed as more stable mimetics of Gi1 and Gq, respectively48.
Assessment of complex formation with isoprenaline-bound β1AR-E showed non-Gs mini-G mimetics had significantly (p < 0.05) reduced affinities, between 5–30 fold, relative to the primary binding partner mimetic mini-Gs (Fig. 3c, Supplementary Table 3). These reductions in affinities were predominantly the result of slower kon rates. Notably, the similarity in koff rates between the mini-G proteins implied that the complexes were similarly stable once formed. Accordingly, the higher selectivity of β1AR for Gs was reflected in its enhanced ability to associate with this binding partner, rather than the stability of the final ternary complex conformation.
The affinity for mini-Go1 (1.34 µM) differed 28.5-fold from mini-Gs (47 nM), with its kon rate reduced by 50-fold. Mini-Gs/i showed a 23-fold reduction in affinity (1.09 µM), with a 40-fold lower on-rate, indicating the α5 helix residues contributed significantly (p < 0.01) to the differences in kon rates and affinities. This agreed with the observation that 80% of the contacts between the two proteins in the ternary complex involved this helix27. The KD for mini-Gs/q was lower (0.24 µM), more similar to the mini-Gs complex though still ~ 5-fold greater. This further supported alteration of amino acids in the α5 helix as sufficient to modulate coupling to receptor, given the two chimeras share the same Gs Ras domain.
Equivalent assays with the β1AR-W construct showed that the order of selectivity reflected in the affinities and kon rates of the different mini-G proteins for β1AR-E were maintained (Fig. 4c), in agreement with when the TRUPATH assay was used. However, binding was less tight compared to β1AR-E, which again was primarily the result of reductions in kon rates, consistent with the more occluded pre-active state populated by β1AR-W.
Based on the kinetics of complex formation we hypothesised that higher rates of association for the Gs protein facilitated selectivity by outcompeting association of other G protein family members.
Ternary complexes with different G protein families show global conformational similarities whilst retaining local plasticity.
We assessed whether the observed differences in the kinetics of ternary complex formation were reflected in the solution structures through investigation of β1AR in complex with the corresponding mini-G proteins. Similar to our investigations with mini-Gs, we recorded 1H-13C HMQC experiments with isoprenaline-bound β1AR-E in the presence of mini-Go1 (Fig. 5a) and mini-Gi1. Although mini-Gi1 experiments were limited to shorter experiments even at reduced temperature, due to the reduced mini-G protein stability, the data quality was sufficient to confirm the overall spectral features. We found a high overall similarity of the spectra between the mini-Gi1 and mini-Go1 complexes (Fig. 5b) and given the similar behaviour of Gi1 and Go1 in functional data (Fig. 4), we therefore used the more stable mini-Go1 as a mimetic of the Gi/o family.
Given the lower binding affinities for G proteins associated with secondary coupling, we differentiated between signal broadening due to mini-G binding kinetics from broadening due to conformational exchange in the ternary complex, which is independent of mini-G concentration, by obtaining ternary complex spectra with 2 and 8 equivalents of mini-G protein. At the two mini-G concentrations we found no indication of binding kinetic-induced changes in the spectra (data not shown) and no residual uncomplexed active state receptor peaks were resolvable, suggesting β1AR-E was close to saturation with mini-G in all complexes.
The overall appearance of the mini-Go1 ternary complex spectra was similar to the mini-Gs spectrum (Fig. 5a), suggesting comparable overall ternary state conformations and similar interactions between receptor and binding partner. The negligible differences for M1784.62 showed that the extracellular side was unaffected by G protein subtype. Interestingly, the shift and intensity changes of M2235.54 were minor when comparing mini-Gs and mini-Go1 complexes, indicating that the proximal PIF motif showed no significant changes. Similar 1H shifts for M2966.41 corroborated this, suggesting F2996.44 adopted comparable conformations between complexes. Therefore, the similarity of the PIF conformations indicated that the activation level of β1AR was not altered between primary and secondary coupling. Notably, the similar 1H shifts of L289M6.34 on TM6 indicated comparable proximities to Y2315.62 on TM5, thus the outward displacement of TM6 did not differ substantially between the mini-Gs and non-canonical mini-G protein complexes. This agreed with the β1AR cryo-EM structures in complex with heterotrimeric Gs and Gi (PDB: 7JJO and 7S0F, respectively)27,41 and other GPCR structural studies with multiple G proteins35,62 but contrasted with earlier reports that found varying levels of TM6 opening to accommodate Gs or Gi when comparing the structures of complexes with different receptors31–34. The signal intensities of corresponding methionine residues showed little variation between the complexes suggesting that the local and global dynamics of the receptor were similarly affected in the presence of the different G proteins (Supplementary Fig. 12c).
However, there were some small differences in peak positions in probes located across the receptor, which indicated subtle variations in the global receptor conformations. In complex with mini-Gs, the residues M2966.41 (Fig. 5c), L289M6.34 (Fig. 5d), and M2836.28 (Fig. 5a) positioned along the lower cytoplasmic half of TM6 showed 13C chemical shift values that were indicative of increased χ3 trans-gauche exchange of the methionine side chains relative to the mini-Go1 and mini-Gi1 complexes, which adopted slightly more fixed trans or gauche conformers. Consequently, these probes suggested coupling to mini-Gs (primary coupling) marginally reduced interactions between TM5 and TM6 along the helix interface, relative to coupling of non-canonical mini-G proteins (secondary coupling). Although these observations correlated with G protein selectivity, the small magnitude of the changes were incompatible with the large kinetic variation observed between mini-G proteins in our binding assays and therefore likely did not relate to selectivity.
Subtle conformational variation was also observed in IL2 as reported by 1H shift differences for M15334.57 (Fig. 5c) between mini-Gs and mini-Go1 complexes. These shifts were distinct from the isoprenaline-bound position of M153, in agreement with IL2 representing a key interacting region on the receptor. Other recent studies examining Gs and Gi binding to β2AR28,29 and muscarinic acetylcholine receptor type 228 have also noted variation in the IL2 region between different G protein family ternary complexes.
We observed the most substantial differences between binding partners in the TM7 and IL4 region in 19F NMR experiments of TETC3447.54 with mini-Gs and mini-Go1 complexes (Fig. 5e). The mini-Go1 ternary complex peak was shifted substantially more downfield than mini-Gs relative to the ligand-bound peaks, consistent with an increased solvent accessibility19 and suggesting that the NPxxY and IL4 environment was modulated by the presence of the different mini-G proteins. This indicated reduced contacts between the mini-Go1 C-terminus and IL4 region relative to mini-Gs, possibly due to a less productive engagement of the two proteins during complex formation in this region.
We further recorded spectra with chimeric mini-Gs/i and mini-Gs/q proteins to examine how mutations in the α5 helix affected receptor conformation. Again, the general spectral features of the complexes were similar to complexes with mini-Gs proteins (Supplementary Fig. 12d), indicating a similar overall structure of the complexes irrespective of the α5 helix identity. However, small variations were still present relative to mini-Gs protein that extended across the entire receptor (Figs. 5d-e). Evidently, only a few mutations in the α5 helix were sufficient to affect the entire interaction interface, including regions around the binding cavity such as IL2, TM6, and TM7. Probes in the receptor core region and tip of TM6 more clearly showed specific influences of both the Ras domain identity and α5 helix on receptor conformations in ternary complexes (Fig. 5c, Supplementary Fig. 12e).
Together, our NMR data suggested that the global conformations of β1AR in complex with G proteins from different families were highly similar. The small local conformational variations of the receptor in the different complexes seemed indicative of an intrinsic adaptability of β1AR to adjust to different binding partners rather than suggestive of a role in G protein selectivity.