The role of the tryptophan-rich allosteric network and sodium egress in GPCR activation

Significance

While hundreds of class A GPCRs retain a sodium binding site in their inactive states, we have yet to explain why this is so crucial in receptor function. By monitoring both TM6 and a toggle switch adjacent to the sodium pocket, we can observe both toggle switch dynamics and the formation of a precoupled state which becomes stabilized as sodium egresses. This is likely a critical precursor to the onset of G protein coupling. Remarkably, tryptophan residues for the most part play a crucial role in GPCR activation, as the majority lie along computationally derived allosteric pathways. This allostery is explored in detail for A_2AR using rigidity-theory and microsecond timescale molecular dynamic simulations.

Abstract

The human adenosine A_2A receptor (A_2AR) is a prototypical member of the class A family of G Protein–Coupled Receptors (GPCRs), which are engaged by over one-third of FDA-approved drugs. Here, we used ¹⁹F NMR to simultaneously evaluate functional states from the perspective of a CF₃-tag on transmembrane helix-6 (TM6) and all 5-fluorotryptophan reporters. Whereas spectra from the TM6 tag reveal a dynamic conformational ensemble, the tryptophan spectra exhibit more discrete ligand-dependent states. These ligand-dependent signatures provide insights into microswitches including a universal toggle switch which interfaces with a sodium binding pocket. Over 560 GPCRs share this toggle switch — sodium pocket cluster, implying a fundamental role in activation. Computational rigidity-theory reveals tryptophan reporters reside along prominent allosteric activation pathways, underscoring their roles in mechanical signal propagation and corroborating NMR observations. Among them, W246^6.48, associated with the toggle switch, plays a critical role in regulation of allosteric networks spanning the orthosteric pocket and extending via tryptophan-rich pathways through the receptor to the A_2AR-Gβ interface and the nucleotide pocket in G_sα. While higher sodium concentrations (~100 mM) predictably stabilize the inactive conformation of the receptor, lower concentrations (below 40 mM) greatly enhanced the presence of the activation ensemble and in particular, the precoupled state, suggesting that the release of sodium from the conserved pocket—either through basal conditions for the apo receptor, or upon binding of the agonist—enables sampling of a precoupled state, ultimately needed to initiate activation and coupling.

G Protein–Coupled Receptors (GPCRs) represent a superfamily of over 800 7-transmembrane receptors in humans, responsible for sensory, hormonal, and neuronal signaling, cell homeostasis, and immunity. The adenosine A_2A receptor (A_2AR) is a prototypical class A (rhodopsin-like) GPCR, found in the nervous system, brain, platelets, immune cells, lungs, heart, and vasculature (1). The primary function of A_2AR is to engage the stimulatory G protein, G_s, upon activation by its endogenous agonist, adenosine. G protein coupling is a complex dynamic process, wherein the alpha subunit of the G protein, G_sα, dislodges bound GDP through a series of conformational changes involving separation of the RAS and alpha helical domains, enabling exchange of GDP for GTP. GTP binding then initiates dissociation of G_sα from the receptor and the βγ subunits of the G protein (Gβγ) enabling GTP-bound G_sα to activate its downstream signaling partner, adenylyl cyclase (2–4). This entire cascade of dynamic events is a consequence of an exquisite and adaptive allosteric network encompassing the activated receptor and G protein.

The promiscuous coupling potential that most GPCRs exhibit suggests that multiple allosteric activation pathways likely coexist within a specific receptor and are influenced by membrane composition, conformational modifications, and other adjuvants specific to a given cell (5–10). Using NMR, it is often possible to validate computationally derived allosteric networks by examining residue-specific chemical shift or linewidth perturbations along potential transmission pathways as a function of ligand. One such study made use of cysteine mutagenesis and the installation of CF₃ tags on two sites associated with TM6 and TM7 of human A_2AR, which were known to undergo significant conformational rearrangement upon activation: V229C^6.31 and A289C^7.54 (11). Here, 6.31 and 7.54 refer to the location of specific residues on TM6 and TM7, relative to the most conserved TM6 and TM7 residues (12). In this and an earlier study (13), A_2AR was reconstituted into POPC/POPG lipid nanodiscs and ¹⁹F NMR spectra were recorded as a function of ligand efficacy (i.e. inverse agonist, apo, partial agonist, full agonist) and G protein (±GDP). The resulting spectra revealed an ensemble of resonances whose function could be discerned based on response to ligand and G protein. In the case of the TM6 label, two distinct inactive resonances were resolved which were differentiated by the presence or absence of an ionic lock between R^3.50 and E^6.30. In addition, three activation states (A₁, A₂, A₃) were discerned, including specific activation states (A₁ and A₂) which became dominant with agonist and nucleotide-free G protein, and a preactivated state (A₃) which became pronounced for the apo receptor, complexed with GDP-bound G protein. A₁ was observed to be dominant using full agonist while a separate A₂ resonance dominated with partial agonist. From this, the authors proposed unique conformations as part of separate allosteric networks for partial and full agonism. Finally, utilizing either TM6 or TM7 CF₃ reporters, it was also shown that these same A₁ and A₂ states were preferred in the presence of a cognate G protein (G_s) and noncognate G protein (G_o) respectively, again suggesting unique allosteric activation networks (11). In the current study, we shift the focus to tryptophan residues since they are located throughout the receptor, and in particular, near several key microswitches (14). Specifically, since there are 6 tryptophan residues in the A_2AR construct used in this study, located on or near allosteric hubs, ¹⁹F NMR can conveniently register signatures of activation from reporters located throughout the receptor. At the same time, by engineering a mutation on TM6 (V229C) situated near the A_2AR-G protein interface, and labeling this site with a CF₃ reporter, we can observe the resulting activation ensemble (15). Together, these reporters allow us to examine the interplay between specific microswitches and the TM6 activation ensemble.

Using fluorotryptophan reporters, we focus on two of the most conserved motifs in A_2AR (and indeed, in the vast majority of class A GPCRs) and their role in activation — namely the sodium pocket and adjacent toggle switch (W246^6.48) alongside W143, which is situated in the so-called selectivity filter of ECL2 at the interface with TM4 (16). While crystal structures show W143 faces the lipid milieu, the ¹⁹F NMR chemical shift of W143 varies greatly with ligand, implying a strong allosteric connection to the nearby orthosteric pocket. In the case of the sodium pocket, it is well known that sodium binding helps stabilize an inactive pose of the receptor while agonist binding displaces the sodium ion, resulting in a reorientation of the aromatic ring associated with the toggle switch (17). For many receptors that undergo basal signaling, it is also likely that their activation is accompanied by sodium egress. Here, we wish to better understand why the sodium switch is so highly conserved, how sodium egress enables activation, and if there are specific ensuing activation intermediates which subsequently direct the receptor toward G protein coupling. ¹⁹F NMR allows us to directly interrogate specific receptor activation states and potential intermediates in addition to equilibria between the toggle switch inactive and active states, as a function of ligand, G protein, and sodium. The results described below provide insights into the sequence of activation events associated with the toggle switch and other tryptophan reporters, in addition to global activation states associated with TM6, resulting from sodium egress.

Advances in rigidity theory-based modeling have enabled quantitative mapping of allosteric networks in GPCRs and protein complexes (11, 13, 18–21), offering an atomistic-level view of mechanical signal propagation regarding the above phenomena. This computational framework complements ¹⁹F NMR by predicting which residues serve as critical nodes in long-range allosteric pathways. In this study, we integrate Rigidity Transmission Allostery (RTA) (22) analysis with Molecular Dynamics (MD) simulations, Monte Carlo simulations, and ¹⁹F NMR to examine how local tryptophan dynamics reflect the broader allosteric communication landscape of A_2AR.

Results

Allosteric Activation Networks in the GPCR-G Protein Complex Encompass Many Tryptophan Residues.

Microswitches are a key facet of allosteric activation networks, which connect to the agonist binding pocket and extend to the G protein interface (17, 23–25). As shown in SI Appendix, Fig. S1, many of these microswitches include a sodium binding pocket and PIF, CWxP, D(E)RY, and NPxxY motifs, which are observed in A_2AR and most class A receptors, encompassing nearly 85% of all GPCRs (6, 26–28). Furthermore, the extracellular loop 2 (ECL2), which is responsible for admitting and stabilizing the agonist, adopts grossly different structures in the inactive and active states, as do the PIF, CWxP, DRY, and NPxxY motifs. Critically, agonist binding in the G protein–bound complex also triggers a rapid rearrangement of a buried network of hydrogen-bonded water molecules and an associated sodium ion in the inactive state. Release of sodium and a hydrogen bonded water then initiates toggle switch activation and the resulting cooperative allosteric response, beginning with release of hydrophobic contacts between 3.43, 6.41, and 6.40, and rearrangement of the NPxxY motif, resulting in rearrangement of TM5, TM6, and TM7 at the cytosolic interface, to accommodate and engage the G protein (6, 26, 28–33).

Fig. 1A depicts a 5 μs MD-derived activation state of A_2AR, stabilized by agonist and GDP-bound G_s heterotrimer (13). Here, the allosteric pathway between the agonist binding pocket and nucleotide pocket is identified through rigidity-based allostery theory, revealing a prominent role for six tryptophan residues in A_2AR, and additional tryptophan residues in Gα and Gβ, which tend to lie along this pathway, as shown. Note that SI Appendix, Fig. S1 reveals the location of these 6 fluorotryptophan reporters and the TM6 V229C CF₃ reporter, relative to the position of well-known microswitches and activation motifs (PIF, NPxxY, DRY, and sodium pocket). Two of the 6 tryptophan reporters are highly conserved within the class A GPCR family and the adenosine receptor subfamily. Specifically, W143 residing in ECL2 (extracellular loop-2) is locked to the lipid headgroup in the inactive pose while establishing a more pronounced hydrogen bond to the asparagine side chain N175^5.36 in the agonist-bound state, as shown in Fig. 1B. Similarly, the tryptophan toggle switch (W246^6.48) is situated adjacent to the orthosteric pocket and undergoes a reorientation and displacement upon activation and binding by agonist, resulting in a specific local interaction network. W143 and W246 each have unique allosteric roles; W246 plays a strong role in allosteric network coupling, transmitting information from the orthosteric pocket to the G protein interface and into Gα via Gβ. W143 is in turn strongly coupled to W246 and its ligand response likely directly modulates the allosteric activation response initiated at the toggle switch. The importance of these tryptophan residues can also be investigated by exploring W143Y and W246Y mutations via ¹⁹F NMR, MD, and Monte Carlo simulations, as discussed below.

Fig. 1.

A. Allosteric Activation Network in A_2AR. (A) Depiction of the allosteric network revealed through Rigidity Transmission Allostery (RTA) analysis associated with agonist-bound A_2AR, complexed with GDP-bound G_s heterotrimer, highlighting the location of all tryptophan residues. The depicted allosteric pathway (from NECA) connects the orthosteric binding pocket to the transmembrane domain and extends through the G protein subunits to the nucleotide-binding site. Tryptophan residues are shown in stick representation, while NECA (Top) and GDP (Bottom) are shown as gray spheres. Here, allosteric intensity is measured by changes in degrees of freedom across residues experienced upon rigidification of the agonist NECA. Allosteric intensities are visualized using a rainbow color gradient, from red (Highest transmission) to blue (Lowest transmission). (B) Location of the orientation and interaction network of W143 and W246 in the inactive inverse-agonist (ZM) stabilized state (green), and an activation state (blue), stabilized by both agonist and nucleotide-free heterotrimer, G_s.

¹⁹F NMR Spectra (of TM6 and F-Trp) Simultaneously Reveal a Complex Conformational Ensemble and Details of Microswitch Activation States.

Fig. 2A shows the ¹⁹F NMR spectra of A_2AR (1-317, V229C) reconstituted in LMNG/CHS micelles, as a function of orthosteric ligand and upon addition of an engineered G protein mimic (mini-G_s) which stabilizes the nucleotide-free activation state. Spectral deconvolutions and assignments of the CF₃ resonances associated with V229C on TM6 (appearing between −60.5 and −62 ppm) are consistent with prior published work, and consist of a dynamically averaged inactive state resonance, S_1-2, in slow exchange with three activation states, A₁, A₂, and A₃ (10, 13, 15). In lipid nanodiscs, the two inactive state resonances (S₁ and S₂) were delineated and proposed to be distinguished by a salt bridge (also known as the ionic lock) between R102^3.50 of the highly conserved E(D)RY motif on TM3 and E228^6.30 on TM6 (13). In micelles, these two inactive poses appear to be in fast exchange and are represented by a single average state, designated S_1-2 (13). Nonetheless, the activation ensemble is very similar to that observed in nanodiscs, where a preactivated state, A₃, associated with the receptor GDP-bound G protein complex, can be differentiated from two activation states, A₁ and A₂, which are stabilized by the respective addition of full or partial agonist, while A₁ is further stabilized by addition of a nucleotide-free G protein or its mini-G_s surrogate (13). In MNG/CHS micelles, removal of mini-G_s results in an activation state (A_1-2) that is likely the result of averaging and consequently coalescence of the A₁ and A₂ states, as seen in the ¹⁹F NMR spectra of A_2AR with partial- and full-agonist alone.

Fig. 2.

¹⁹F NMR spectra of TM6 (A) and all fluorotryptophan reporters (B) as a function of ligand. Dashed lines designate the TM6 principal resonances associated with the inactive (S_1-2) and active state poses (A₁, A₂, A₃) for V229C. ZM, LUF, NECA and mini-G_s designate inverse-, partial-, and full-agonists in addition to a miniengineered G protein, respectively. Residual error spectra in the resulting deconvolutions are shown in red and gray for the TM6 and fluorotryptophan spectra, respectively.

The fluorotryptophan resonances, spanning −119 to −127 ppm in the above spectra, are shown in Fig. 2B. Guided by observed peak maxima or inflection points in the spectra, the ¹⁹F NMR spectra were deconvolved into six principal resonances of equal area (designated in red) corresponding to the 6 tryptophan reporters. Generally, additional resonances whose intensities amounted to significantly less than that expected for a single tryptophan reporter, were attributed to minor (excited) states, designated in mauve in the deconvolutions.

Subtle W/Y Mutations in ECL2 and the Toggle Switch Result in Long Range Allosteric Perturbations.

To identify the inactive and active state resonances associated with the toggle switch and ECL2, two mutants (W143Y and W246Y) were separately prepared, and the corresponding ¹⁹F NMR spectra were recorded in the presence of inverse agonist (ZM) and agonist (NECA) as shown in Fig. 3A. A comparison of W246Y and WT ¹⁹F NMR spectra in the presence of ZM allows for the unambiguous assignment of the toggle switch as the most downfield resonance, where the assigned W246 resonance in the inactive state, missing from the W246Y spectrum, is represented in green in the deconvolution of the WT spectrum. In the presence of NECA, the assignment of the toggle switch is less straightforward since there are nonnegligible chemical shift perturbations across the spectrum. However, after deconvolution, the outer resonances between −124.5 and −126.5 ppm now integrate to 3/5 the total fluorotryptophan spectral area versus 4/6 of the total spectral area. Thus, we can tentatively assign the active state resonance of the toggle switch as the most upfield resonance. The inactive and active state resonances of W143 in ECL2 can be similarly assigned by comparing the ¹⁹F NMR spectra of WT ZM- and NECA-bound A_2AR with those of W143Y, where the assigned resonances in the inactive and active states are represented in blue. In both cases, this mutation appears to result in enhanced dynamics of all F-Trp resonances, as seen by apparent coalescence of the fluorotryptophan resonances in the W143Y spectra (Fig. 3A). Nevertheless, after accounting for missing area in the spectra, the inactive and active W143 signatures are tentatively assigned to −123.3 ppm and −122.25 ppm respectively.

Fig. 3.

Effects of W246Y and W143Y mutations. (A) ¹⁹F NMR spectra of 5F-Trp-enriched A_2AR WT, W246Y (toggle switch), and W143Y, saturated with ZM or NECA. Resonances in green represent those of the toggle switch while those in blue are tentative assignments of the ECL2 reporter, based on missing spectral area. (B) Identical ¹⁹F NMR spectra from above in the region of the CF₃ reporter at V229C^6.31 (TM6). W143Y results in the absence of any detectable inactive resonance in the ZM-bound A_2AR. Similarly, W246Y suppresses the inactive fraction in ZM-bound A_2AR, while the spectrum of NECA-bound A_2AR exhibits greater exchange broadening over that of WT A_2AR.

The ¹⁹F NMR spectra of the W246Y and W143Y mutants can also be evaluated from the perspective of the TM6 reporter (V229C), as shown in Fig. 3B. In the case of the toggle switch mutation (W246Y) a noticeable destabilization of the inactive pose is observed in the presence of ZM, relative to the wild-type (WT) species, with a substantial shift toward the activation ensemble, accompanied by line broadening. Interestingly, in the presence of saturating concentrations of NECA however, the activation ensemble is not stabilized (relative to the WT species) and is instead characterized by additional line broadening consistent with intermediate timescale exchange broadening. The TM6 spectrum from W143Y reveals no detectable inactive signature in the presence of ZM. Addition of agonist results in an apparent enhancement of the A_2AR activation states, although the spectrum is a significant departure from that of the NECA-bound WT species. These results reflect the underlying long-range allosteric effects of W143 (ECL2) and W246 (toggle switch).

To understand the effects of mutations at the two most conserved tryptophan sites, 10 μs MD simulations were conducted on W143Y and W246Y A_2AR in lipid bilayers consisting of 4:1 POPC:cholesterol, beginning with previously equilibrated simulations of inverse agonist (ZM) and agonist (NECA)-stabilized A_2AR. As shown in Figs. 1B and 4, W143 is situated on ECL2 at the membrane–water interface and is observed to preferentially hydrogen bond with a resident POPC molecule in the inactive state. In the active state, the indole nitrogen of Trp 143 establishes a hydrogen bond with N175^5.36, thereby anchoring the extracellular terminus of TM5 to TM4. Substitution of W143 with Tyr perturbs this cluster, resulting in a slight disruption of the TM4–TM5 interface. These somewhat subtle extracellular disruptions have significant long-range allosteric effects as evidenced by coalescence of most tryptophan resonances in the active state (Fig. 3A) and the loss of the inactive TM6 signature in the inactive state (Fig. 3B). The toggle switch, W246, can also be seen to form a tight Van der Waals cluster with inward facing residues in the inactive state, as shown in Fig. 1B. Note in Fig. 4A, that two specific ionic lock distances (i.e. from a double or a single bond as shown) are generally observed in the MD trajectories of A_2AR in the presence of ZM. Upon activation, the agonist directly coordinates W246, resulting in a reorientation of the indole ring, and a slight downward displacement of W246, juxtaposing it between H250^6.52 and T388^3.36 side chains. The W246Y mutation results in a significant clash with TM7 and a gross rearrangement of TM6, as shown in Figs. 4 A and B. By measuring the distance between E228{$N^{\zeta }$} ^6.30 to the R102{$C^{\delta }$} ^3.50 we could evaluate the ionic-lock propensity of the W246Y and W143Y mutants from MD simulations. Interestingly, the NECA-bound W246Y mutant exhibited a closed conformation, exhibiting an alternative ionic lock configuration as shown in Fig. 4A. Similarly, the ZM-bound W143Y mutant exhibited a more open (active state) conformation, in keeping with the ¹⁹F NMR results. While the TM6 activation signature is still achieved with agonist binding, as shown in Fig. 3B, the predicted packing disruption explains the observed line broadening that accompanies the W143Y NMR spectrum of V229C^6.31 in the presence of NECA.

Fig. 4.

Ionic lock distance distributions for active and inactive states and MD-derived structures. (A) MD-derived interatomic distances associated with an ionic lock between E3.50 and R6.30 shown for ZM and NECA bound to wt A_2AR, W143Y, and W246Y mutants. (B) Overlay of MD-derived structures of ZM-bound (inactive) and NECA-bound (active) A_2AR, with (yellow) and without (green) specific mutations, W246Y and W143Y. Structures were equilibrated through 5 μs MD simulations.

While MD simulations provide atomistic details regarding local structure, dynamics, and activation trajectories, it is extremely challenging to bridge inactive and active states in a single trajectory and therefore assess the effects of specific mutations at allosteric hotspots on activation processes. Using the known inactive and active states of A_2AR, it is possible to evaluate activation (and deactivation) trajectories by rigidity theory-based Monte Carlo simulations (34). In this case, random moves in configurational space (i.e. dihedral and backbone torsion angles, rigid body motions, and concerted rotational moves) are performed, whereupon the energy of the resulting state is evaluated and subsequent jumps from a given state are evaluated based on the resulting energy change, relative to either the starting or end state. Fig. 5A depicts the Monte Carlo-derived excursions of TM6 beginning with high-resolution structures of A_2AR in the presence of inverse agonist (ZM), partial agonist (LUF), full agonist (NECA), and NECA plus either mini-G_s or G_s heterotrimer. Note that the NECA-saturated sample reveals the broadest distribution of distances, which no doubt reflects the complex activation ensemble, encompassing a preactivated state and distinct activation states, adopted by the receptor in the absence of G protein. The maxima of all distributions also denote a gradation of outward extension of TM6 which increases from inactive to fully active.

Fig. 5.

Monte Carlo-based distributions of interresidue distances and dihedral angles for WT A_2AR and W143Y and W246Y mutants. (A) TM6 excursions beginning with known structures in which A_2AR was stabilized with inverse agonist (ZM, pdb 3PWH), partial agonist (LUF, pdb 8RLN), agonist (NECA, pdb 2YDV), NECA + mini-G_s, pdb 5G53, or NECA + Gs heterotrimer (11). (B) W143 distance excursions (Left) and dihedral angle distributions, p(χ₁), along the respective Monte Carlo trajectories as a function of drug. (C) W246 distance excursions (Left) and dihedral angle distributions, p(χ₁), along the respective Monte Carlo trajectories as a function of drug. (D) Residue 143 (Left) and 246 (Right) distance excursions for inverse agonist and agonist saturated A_2AR assuming WT, W143Y, or W246Y mutations.

Fig. 5B depicts W143 distance and χ₁ dihedral angle excursions from the Monte Carlo analysis of A_2AR as a function of drug, mini-G_s, and G_s heterotrimer. In this case, distance excursions of W143 and corresponding distance distributions are very similar, while χ₁ distributions are notably different, particularly for NECA plus mini-G_s. In contrast, distance excursions can clearly be discerned for the toggle switch, W246, as shown in Fig. 5C, where a gradation of excursions is seen under greater activating conditions. Also of note, the dihedral angle distributions associated with χ₁ are considerably broader over the Monte Carlo trajectories and are noticeably distinct for different ligands, mini-G_s, and G_s. Note that both ZM- and LUF-saturated A_2AR exhibit dihedral angle excursions that overlap partly with that of the NECA χ₁ distribution, revealing a potential local activation pathway.

Fig. 5D reveals the effects of mutations at either W143 (ECL2) or W246 (toggle switch) on the Monte Carlo generated distance distributions of residues 143 and 246, for both the inverse agonist (ZM) and agonist (NECA) saturated states. In the left panel, W143Y is shown to extend the distance excursions at this site, relative to W143, while W246Y exerts relatively subtle effects on the distance excursions of W143 in the ZM-bound state. In the NECA-bound state, changes to the distance excursions at residue 143 resulting from either W143Y or W246Y are also quite subtle. In contrast, while W143Y exerts a very subtle effect on the positional distribution of W246 for ZM-bound A_2AR, there is a distinct shift in the residue 246 distribution resulting from the toggle switch mutation, W246Y, in the direction consistent with activation of W246. Perturbations to the distance distribution of residue 246 for NECA-bound A_2AR, resulting from W143Y or W246Y are again more subtle. Taken together, these Monte Carlo distributions corroborate the observation that W246Y and W143Y both give rise to active-like signatures of the toggle switch for ZM-bound A_2AR.

Sodium Allostery.

Class A receptors are known to be stabilized in the inactive pose by physiological concentrations of sodium (30). While this negative allostery has been previously investigated by many labs (17, 29–32, 35), we were specifically interested in the effects of sodium egress on activation. Sodium egress from the conserved pocket has been shown to result from agonist binding and is suggested to involve the eventual protonation of D2.50 and a concerted reorganization of water networks, resulting in the intracellular egress of sodium (35). It is also possible that sodium egress and the consequent reorganization of water networks, contribute to allosteric activation propensity and lower the activation energy barrier, allowing sampling of activation states, even in the absence of agonist. The specific role of sodium egress on activation can be partly tested by retaining an antagonist, while lowering the sodium concentration and examining the propensity for sampling of activation states using ¹⁹F NMR of TM6 and the fluorotryptophan reporters.

Sodium Depletion Activates the Toggle Switch and Biases the TM6 Activation Ensemble and in Particular, the Precoupled State.

Fig. 6 reveals the effects of sodium depletion on the ¹⁹F NMR spectra associated with TM6 and all Trp reporters of A_2AR, in the presence of saturating concentrations of antagonist. Since low sodium concentrations tend to destabilize the receptor when reconstituted in detergent micelles, we found that receptor stability could be recovered in the presence of 100 mM ethyl ammonium chloride. At the same time, the dimensions of the conserved sodium pocket in A_2AR are such that ethyl ammonium cation cannot bind. Note that a drop in sodium concentrations from 100 mM to 40 mM results in a shift in the TM6 state distribution toward the activation ensemble resonances, coinciding with pronounced line broadening of the activation ensemble states and an enhancement of a resonance associated with a preactivated state. In earlier work by Huang et al, the preactivated state, A₃, was attributed to binding GDP-bound G protein and is likely an activation intermediate that precedes coupling (13). The fluorotryptophan resonances also reflect chemical shift perturbations toward the active state, while the toggle switch inactive signature is diminished, likely due to reorientations that allow sampling of activation states. An additional drop in the sodium concentration to nearly 0 mM almost completely diminishes the inactive toggle switch state, while the TM6 resonances associated with the active state ensemble become better resolved and are remarkably similar to the NECA (agonist saturated) spectrum, with the exception of a disproportionate fraction of the preactivated state. The tryptophan resonances reveal a coincident loss of the toggle switch inactive signature, while the remaining spectrum is somewhat intermediate between the inactive state and that of the tryptophan NECA (agonist saturated) spectrum as shown in the overlay. One prominent exception is the resonance associated with W143 in the active state, which is not unexpected since ECL2 is known to rearrange upon agonist binding. As discussed below, W143 is also allosterically connected to W246 and contributes to overall activation and coupling.

Fig. 6.

Effects of sodium on A_2AR activation states. ¹⁹F NMR spectra of ZM-saturated A_2AR, showing resonances arising from TM6 (A) and those from fluorotryptophan (B) as a function of diminishing concentration of sodium. For comparison, the TM6 and fluorotryptophan NMR spectra of A_2AR in the presence of 100 mM sodium and saturating concentrations of agonist, NECA, are overlaid in yellow.

Discussion

Prior ¹⁹F NMR studies of A_2AR in micelles and HDL nanodiscs, using the identical TM6 reporter, V229C^6.31 have focused on the delineation of functional states to better understand activation mechanisms and more recently, coupling promiscuity (10, 13, 15). In addition to two inactive states, differentiated by an ionic lock, the activation ensemble was described by three unique conformers: a preactivated state (A₃) associated with TM6 engagement with GDP-bound G_s and two activation states (A₁ and A₂) which were observed to be most prominent when the receptor, in complex with nucleotide-free G protein, was saturated with full (NECA) or partial (LUF) agonists, respectively. A₁ and A₂ were later attributed to a full and partial opening of TM6 that also enabled optimal engagement with G_s and a noncognate G protein (G_o) to A_2AR (10). TM7 signatures also revealed a similar ensemble of activation states, which included a distinct outward facing activation intermediate and two additional activation states, A₁^TM7, and A₂^TM7. Here, we consider the ¹⁹F NMR signatures of the tryptophan sidechains to evaluate their correspondence to the activation ensemble observed for TM6 and TM7.

The simultaneous measurement of fluorotryptophan spectra of A_2AR with the above TM6 reporter provides a comprehensive measure of specific activation states and allosteric response to ligands and sodium. A similar approach was taken by Eddy et al who were able to biosynthetically label the identical adenosine A_2A receptor with ¹⁵N-enriched tryptophan, while separately mutating an additional residue to a tryptophan at TM5 (F201W^5.62), TM6 (K233W^6.35), and TM7 (Y290W^7.55) (36). These authors also identified a Bg12 strain of Pichia that permitted higher growth rates and expression yields in cell culture media containing D₂O, ultimately reaching ¹⁵N-Trp enriched and 70% deuterated A_2AR in LMNG micelles (36). In general, the NH TROSY spectra of ¹⁵N-Trp enriched A_2AR proved to be sufficient to completely resolve and assign all six tryptophan residues in addition to the biosynthetically incorporated ¹⁵N-Trp on TM5, TM6, or TM7 (14, 36). Moreover, the addition of a 21-residue Gα_s C-terminal peptide resulted in subtle chemical shift perturbations that pointed to an activation state of W233^6.35, while additional shift perturbations were observed for W32 and W29, on the cytosolic interface of the receptor. While the resolution displayed by the NH TROSY spectrum was exceptional, a complex ensemble such as that seen by ¹⁹F NMR for a CF₃ reporter on TM6 (V229C^6.31) could not be observed for W233^6.35. Moreover, additional activation states of TM5 and TM7 were not observed by ¹⁵N,¹H TROSY NMR, either because of exchange broadening associated with the active states or degeneracy of the (A₁, A₂, A₃) activation state chemical shifts (36). Thus, it seems that ¹⁹F NMR has the advantage that the ensemble and possible excited states are easier to resolve.

In the current study, the ¹⁹F NMR spectra of both the TM6 CF₃ reporter and the fluorotryptophan reporters reveal exquisite sensitivity to the addition of drug and/or mini-G_s. In particular, the CF₃ reporter at V229C^6.31 faithfully reproduces an ensemble described by two inactive states, in addition to an activation ensemble consisting of a preactivated state (A₃) and two activation states (A₁ and A₂) (13). The remaining fluorotryptophan reporters were deconvoluted to 6 peaks of equivalent area, with the clear presence of additional minor peaks, associated with excited states. Note for example the deconvolved minor peaks, observed in the ZM-saturated fluorotryptophan spectra (Fig. 2) closely resemble the active state spectrum observed for NECA or NECA + mini-G_s.

The toggle switch W246^6.48 adopts a significant downfield chemical shift signature in the inactive pose, likely arising from the unique aromatic ring-shifted environment (14, 37). Upon mutating the toggle switch to tyrosine (W246Y^6.48) and recording ¹⁹F NMR spectra in the presence of saturating concentrations of inverse agonist (ZM) and agonist (NECA), the toggle switch inactive and active resonances were determined to be −120 ppm and −126 ppm, respectively (Fig. 3). In the spectrum associated with partial agonist (Fig. 2), this upfield (toggle switch) resonance is much broader, suggesting greater disorder or exchange dynamics of the toggle switch in the active state. In addition, all of the tryptophan resonances appear to exhibit a degree of coalescence indicative of extensive sidechain dynamics in the partial agonist saturated state. In contrast, the fluorotryptophan spectra associated with agonist-stabilized A_2AR are well resolved and the addition of mini-G_s appears to diminish the minor peak signatures, consistent with a predominant activation state. Spectral deconvolutions also clearly also point to an underlying ensemble and the presence of minor (excited state) peaks. MD simulations (Fig. 1B) and MC simulations (Fig. 5C) also reflect a continuum of environments, distance excursions, and χ1 dihedral side chain distributions for the toggle switch, W246^6.48, as a function of ligand.

Mutations, W143Y and W246Y, both diminished the inactive state populations of the inverse agonist-saturated receptor, as judged by ¹⁹F NMR of V229C^6.31. This was also borne out from histograms associated with MD simulations around the ionic lock (Fig. 4A). In this case, W246Y results in an alternate arrangement of the ionic lock in the presence of NECA, thereby restricting sampling of the activation state, while W143Y perturbs the inactive state in the presence of ZM. Monte Carlo sampling of A_2AR reveal ligand-dependent distance and dihedral angle excursions around W143 in the presence of the toggle switch mutation (W246Y) and distance/dihedral excursions around W246 in the presence of the ECL2 W143Y mutation. Although MC simulations do not capture significant change around W143 (likely because much of this is mediated through lipid interactions), it is clear from Fig. 5D that W143Y and W246Y mutations both limit the distance excursions associated with active and inactive signatures around the toggle switch site (W/Y246), in keeping with ¹⁹F NMR. Note for example that in the presence of agonist, W143Y resulted in significant exchange broadening in the fluorotryptophan resonances and peak shifts in the V229C TM6 spectra that do not result in the formation of the active state, A₁, recognized for G protein coupling (Fig. 3B). In contrast, the V229C^{6.31 19}F NMR spectra of NECA-bound A_2AR, in the presence of the toggle switch mutation W246Y, resemble that of wt-A_2AR, with the exception that there is additional line broadening. The equivalent fluorotryptophan NMR signatures of the toggle switch mutation also reveal additional line broadening. Severe exchange-broadening in our case, might suggest a steric clash regarding requisite activation steps and G protein coupling. Interestingly, earlier studies showed that a W246A mutation grossly compromised agonist binding and activation, based on cAMP assays (38).

Allostery and Tryptophan Dynamics.

The prominent chemical shift response of fluorotryptophan reporters to ligand and mini-G_s reflects their role in allosteric transmission. Tryptophan residues play critical roles in allosteric transmission. This is clearly observed in Fig. 1A, where tryptophan residues largely reside within an allosteric network bridging the ligand binding pocket and the nucleotide pocket of Gs. Many key tryptophan residues reinforce this allosteric pathway via Gβ (a β-propeller protein) bridging the receptor and the Gα subunit, where Gβ was suggested to serve as an allosteric chaperone critical to dictating efficacy in the receptor-G protein complex (13). The extent to which these tryptophan residues contribute to allostery can also be appreciated by evaluating the pairwise residue allosteric propensity, W_ij. In this case, W_ij represents the effective change in degrees of freedom associated with tryptophan j upon rigidification of those residues clustered around tryptophan residue i. Fig. 7 depicts the allosteric connectivity between individual tryptophan residue pairs, W_ij, in the A_2AR-G_s heterotrimeric complex, using RTA theory (39). This 2D allosteric profile reveals that the tryptophan residues in the Gβ subunit, play a critical role in conveying allostery between the agonist binding site and the nucleotide pocket of Gα, while many of the tryptophan residues in A_2AR are also critical to allostery (namely, W29^1.55, W32^1.58, W129^4.50, W143 (ECL2), W246^6.48, and W268^7.33). The resulting correlation map shows for example, that the toggle switch W246^6.48 is allosterically connected to both W143 and hence, ECL2, and a chain of Trp residues spanning Gβ and Gα. Note that while W268 seems less allosterically coupled to other tryptophan residues, it is still coupled with W246 and the orthosteric pocket. The strong allosteric coupling between tryptophan residues across A_2AR and G protein subunits is consistent with their dense positioning in mechanically constrained zones which play a critical role in receptor-G protein complexes as cross-domain communication hubs essential for allosteric coupling. Other tryptophan residues also emerge as significant contributors to the allosteric network. W129, for instance, plays an important role in allosteric communication and exhibits strong conformational variability depending on ligand and G protein engagement, which is consistent with their dynamic χ₁ dihedral angle distributions observed in Monte Carlo simulations (SI Appendix, Fig. S2). Specifically, W29, W129, W246, and W268 exhibit broad and pronounced dihedral angle distributions which depend sensitively on ligand.

Fig. 7.

2D allosteric heat map involving tryptophan residues predicted by RTA analysis. Pairwise allosteric correlation, W_ij, between all tryptophan residues in the agonist-bound A_2AR-Gs complex are shown. A significant number of tryptophan residues are involved in allosteric networks in the receptor- G-protein complex.

Steered MD simulations can also be leveraged to examine possible correlations between Trp residues as shown in SI Appendix, Fig. S3. In this case, upon applying a constant torque associated with the χ₁ dihedral angle of either W143 (Left) or W246 (Right), correlated side chain dynamics can then be investigated. Under these conditions, W246 and W129 exhibit correlated dynamics for NECA bound A_2AR upon steering W143, as do W32 and W29, recapitulating what is observed generally in the allosteric heat map, shown in Fig. 7.

Tryptophan enrichment along allosteric pathways likely reflects their unique biophysical versatility. Tryptophan is known to form interactions via pi stacking, intercalation into hydrophobic domains and establishing van der Waals interactions, bridging membrane water interfaces, and enabling hydrogen bonds via the indole nitrogen (40–42). This chemical versatility of interactions likely enables the tryptophan aromatic rings to serve as highly effective switches by adopting multiple stable configurations that respond sensitively to environmental and structural cues.

Collectively, the spatial positioning, network connectivity, and dynamic properties of tryptophan residues make them particularly effective sensors and transducers of allosteric signals within the A_2AR–G protein complex. In terms of rigidity theory, tryptophan residues often anchor into dense clusters constrained by hydrogen bonds and hydrophobic contacts. Their positioning within these mechanically critical regions makes them ideal nodes for propagating allosteric signals through a complex. Consistent with this, tryptophan residues in A_2AR are disproportionately enriched along predicted allosteric pathways (Fig. 1A) and exhibit strong pairwise allosteric coupling (Fig. 7). Tryptophan mobility in the protein (as a function of drug and/or G protein) should reflect allosteric propensity. If tryptophan residues lie along an allosteric pathway, we might also expect their dihedral angle excursions to be correlated.

Allostery can also be viewed from another perspective — namely the contribution to activation as judged by allosteric propensity between any residue and the nucleotide pocket in the A_2AR-Gs complex. Fig. 8 depicts projected allosteric networks revealed through RTA analysis upon rigidifying a cluster centered around either W143 (Fig. 8A) or W246 (Fig. 8B). While the toggle switch is central to allosterically driving activation and G protein coupling, ECL2 makes its own unique contributions to the receptor and in particular, to W246.

Fig. 8.

Allosteric Activation Network in A_2AR, originating from two highly conserved tryptophan residues in ECL2 and the toggle switch. (A) Depiction of the allosteric network revealed through RTA analysis associated with W143 (ECL2) to the nucleotide pocket for the A_2AR- GDP-bound G_s heterotrimer complex. (B) Predicted allosteric pathway from W246 to the nucleotide pocket.

Together, the above results suggest that tryptophan residues act as conformationally sensitive allosteric transducers. Their high allosteric propensity arises not only from their interaction chemistry, but also from their mechanical integration within the receptor–G protein complex. This dual role—structural and dynamic—makes fluorotryptophan an exceptional reporter type for ¹⁹F NMR.

Sodium Allostery.

Sodium is known to enhance inverse agonist binding in the vast majority of pharmacology studies of class A GPCRs. In a seminal study, Katritch and coworkers proposed that sodium egress resulting from agonist binding could contribute to the free energy of activation, by coupling activation to the free energy of transfer of sodium from the extracellular to the cytosolic domain (30). Later, Vickery et al. simulated a sodium egress pathway and suggested that protonation of D2.50 was critical to the eventual egress of sodium and subsequent activation events (35). In the current study, we have made use of TM6 ¹⁹F NMR spectra and fluorotryptophan signatures to evaluate the consequence of sodium egress. In particular, after first stabilizing the system by ethyl ammonium chloride and maintaining an excess of antagonist, ZM, we observe a pronounced enhancement in sampling of activation states accompanied by exchange broadening, once sodium concentrations are reduced to 40 mM. Importantly, the toggle switch, whose indole group is known to coordinate sodium via a hydrogen bonded water molecule in the inactive state, is also clearly destabilized at low sodium concentrations. Remarkably, at 0 mM sodium, the TM6 signature is very similar to that of full agonist bound A_2AR with 100 mM NaCl, with the exception of a disproportionately high fraction of the precoupled state, A₃. Similarly, the fluorotryptophan NMR signature is similar to the NECA-bound equivalent, with the notable exception of W143 which predictably adopts a ZM-like chemical shift. While the current study does not explore activation energetics, it does underline the importance of sodium egress in enabling critical activation events. Notably, sodium egress biases the ensemble toward activation states, and leads to an enhancement in the preactivated state, which is likely a precursor to subsequent G protein coupling steps.

Limitations and Future Directions of This Work.

Fluoroaromatics have long been known to provide excellent chemical shift dispersions in protein NMR studies (43, 44). Monofluorotryptophan, and in particular, 5- and 7-fluorotryptophan are proven reporters for delineation of functional states and the study of dynamics via CPMG relaxation dispersion and T₁-based relaxation experiments (18, 45). In principle, resolution could be further improved through two-dimensional (¹⁹F,¹³C) TROSY NMR, or the use of site-specific labeling of fluorotryptophan via stop codon suppression (46–48). However, two-dimensional NMR experiments would require much greater sample concentrations (or sensitivity) while stop codon suppression technologies would first need to be adapted to yeast or other eukaryotic expression systems to have application to GPCRs.

Like Escherichia coli, Saccharomyces cerevisiae and Pichia pastoris both utilize the shikimate pathway to synthesize all three aromatic amino acids. Therefore, monofluorotryptophan could in principle be incorporated during heterologous expression in yeast, after supplementing the medium with fluorinated tryptophan or fluorinated precursors. This was demonstrated recently by Pan et al who were able to record ¹⁹F NMR spectra of 5-fluorotryptophan enriched neurokinin 1 receptor (NK1R) in LMNG/CHS micelles (49). In this preliminary study, the authors utilized 5-fluoroindole as a precursor, whereas we have used 5-fluoroanthranilate primarily for reasons of cost and solubility. Yeast expression times are usually days rather than hours and this presents new challenges. We observed that incorporation of fluorotryptophan reached only around 10 to 20% if the fluorotryptophan precursor was introduced just prior to induction in yeast minimal media. We hypothesized that over days, unlabeled tryptophan might arise from other proteins that are hydrolyzed through proteasomal pathways, unless the cells are first expanded in the presence of fluorotryptophan or its precursors. Indeed, there is a precedent in the literature for the presence of a persistent tryptophan pool in yeast (50). We therefore resorted to first enriching the BMGY starter cell culture with 25 mg/liter 5-fluoroanthranilate for 12 to 24 h, after which the cells were transferred to BMMY minimal media, enriched with 25 mg/liter 5-fluoroanthranilate, and induced over three days. During induction, the media were replenished with 5-fluoroanthranilate every 24 h. Under these conditions we were able to achieve 60 to 70% enrichment of 5-fluorotryptophan, based on the relative areas of the fluorotryptophan to CF₃ (V229C) signals. An alternative strategy might be to use minimal media under tryptophan auxotrophy, which is typically achieved through the addition of glyphosate which inhibits EPSP synthase (51). However, in yeast, glyphosate is less accessible to the cytoplasm while EPSP synthase is less inhibited. This could in principle be reversed by engineering yeast strains with greater glyphosate sensitivity or by simply resorting to a tryptophan auxotroph. The persistence of a tryptophan pool also suggests that greater enrichment could be achieved by supplying the cells with even more of the fluorinated analog to outcompete the endogenous tryptophan pool.

One of the most significant challenges (in our hands) to recording robust ¹⁹F NMR spectra of fluorotryptophan-enriched A_2AR is one of signal to noise, at least in the case where sample concentrations are less than 100 μM. Relaxation effects through dipolar and chemical shift anisotropy (CSA) mechanisms associated with 5-fluorotryptophan are such that linewidths vary between 100 and 250 Hz (nearly twice that of the CF₃ reporter). One way of recovering sensitivity is to pre-equilibrate the NMR sample in a chamber under ~10 to 15 bar oxygen, thereby reducing ¹⁹F T₁ relaxation times from ~ 1 s to ~ 100 ms. In this case, we were able to effectively record three to four times as many transients in a given period of time, without compromising resolution (52). Another way to recover resolution (and sensitivity) in ¹⁹F NMR experiments is to make use of MAS ¹⁹F NMR as was recently shown by Eddy et al in ¹⁹F NMR studies of A_2AR in liposomes (53). Since the CSA of 5-fluorotryptophan hovers around 50 ppm (54, 55), spinning rates above 40 kHz would be anticipated to contribute to substantial improvements in resolution.

Materials and Methods

Information on plasmid construction, heterologous expression of A_2AR, membrane protein reconstitution and purification, MD simulation methods, allostery analyses, and constrained geometric Monte Carlo simulations is provided in SI Appendix. Briefly, a Pichia Pastoris cell culture, containing a high copy number of the gene encoding A_2AR-V229C, was grown for 1 h at 19 °C, then induced with 0.5% (v/v) methanol every 12 h for a total of 72 h. 1 g/L N-(phosphonomethyl)glycine and 25 mg/L 2-amino-5-fluorobenzoic acid was added every 24 h to achieve enrichment in 5-fluoro-tryptophan. Cell pellets were resuspended and lysed using a microfluidizer. The lysate was centrifuged at 8,000×g for 30 min and supernatant was then collected and centrifuged at 100,000×g for 60 to 90 min. The resulting pellet was resuspended in LMNG/CHS detergent as detailed in the SI Appendix. The solubilized membrane was centrifuged at 10,000×g whereupon the supernatant was incubated with TALON resin for purification, BTFMA labeling, and elution.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

All study data are included in the article and/or SI Appendix.