A_2A adenosine receptor functional states characterized by ¹⁹F-NMR

Significance

G protein-coupled receptor (GPCR) signaling from the extracellular orthosteric drug binding site across the cell membrane to the intracellular contact sites with G proteins and arrestins is enabled by inherent structural plasticity, which can be observed by NMR spectroscopy. Here, we use ¹⁹F-NMR to characterize ensembles of different, simultaneously populated conformations of the human A_2A adenosine receptor in function-related equilibria. The NMR data support an activation model involving both induced fit and reequilibration of conformational ensembles, and experiments with a constitutively active mutant of A_2A adenosine receptor (A_2AAR) established correlations between NMR parameters and GPCR basal activity. The present quantitative measurements with observation of judiciously engineered ¹⁹F-NMR probes complement a previous qualitative overall view of A_2AAR dynamics from high-resolution NMR.

Abstract

The human proteome contains 826 G protein-coupled receptors (GPCR), which control a wide array of key physiological functions, making them important drug targets. GPCR functions are based on allosteric coupling from the extracellular orthosteric drug binding site across the cell membrane to intracellular binding sites for partners such as G proteins and arrestins. This signaling process is related to dynamic equilibria in conformational ensembles that can be observed by NMR in solution. A previous high-resolution NMR study of the A_2A adenosine receptor (A_2AAR) resulted in a qualitative characterization of a network of such local polymorphisms. Here, we used ¹⁹F-NMR experiments with probes at the A_2AAR intracellular surface, which provides the high sensitivity needed for a refined description of different receptor activation states by ensembles of simultaneously populated conformers and the rates of exchange among them. We observed two agonist-stabilized substates that are not measurably populated in apo-A_2AAR and one inactive substate that is not seen in complexes with agonists, suggesting that A_2AAR activation includes both induced fit and conformational selection mechanisms. Comparison of A_2AAR and a constitutively active mutant established relations between the ¹⁹F-NMR spectra and signaling activity, which enabled direct assessment of the difference in basal activity between the native protein and its variant.

G protein-coupled receptors (GPCRs) are cell surface sensory proteins that recognize a vast array of extracellular stimuli and form signaling complexes with intracellular partner proteins that drive physiological responses. Approximately 35% of all FDA-approved drugs target human GPCRs, and development of new GPCR drugs shows no signs of slowing down (1). Drug development is supported by both 3D structures of GPCRs and GPCR complexes obtained by X-ray crystallography and cryo-EM and by knowledge about their dynamic structural plasticity gained using other biophysical methods such as NMR spectroscopy. NMR in solution possesses the unique ability to observe multiple, simultaneously populated GPCR conformers, establish relations between their populations and the efficacies of bound drugs, and thus provide direct insight into dynamic processes that underlie physiological GPCR signaling.

In this study, we apply ¹⁹F-NMR to refine our understanding of signaling mechanisms in the human A_2A adenosine receptor (A_2AAR). A_2AAR functions have been studied in the central nervous system as a key modulator of neurotransmitters (2), in the cardiovascular system as regulating vasodilation (3), and in the immune system as a T-cell surface immune checkpoint protein (4). Correspondingly, A_2AAR has been targeted for various diseases, including Parkinson’s disease (5) and depression (6), and new drugs are under development for cancer therapies (7, 8). To gain a deeper understanding of A_2AAR functionality, X-ray crystal structures have been reported for its complexes with antagonists (9–12) and agonists (13, 14) and for a ternary complex with an agonist and a “mini G protein” (15). The extensive data on structure and function now provide a foundation for other biophysical techniques to provide complementary insights into the signaling mechanisms. Here we used NMR spectroscopy in solution.

In earlier high-resolution NMR studies of stable-isotope–labeled A_2AAR, NMR observations afforded a global view of structural plasticity throughout the 3D structure, and changes in conformational equilibria could be related to variable drug efficacies and to inactivation of an allosteric switch (16, 17). To characterize these structure ensembles in greater detail and to quantify rates of exchange among the different conformers, we now used ¹⁹F-NMR with single extrinsic probes judiciously placed in positions at the intracellular tips of the transmembrane helices (TM) VI, VII, and VIII. The increased sensitivity of this ¹⁹F-NMR approach relative to 2D heteronuclear correlation experiments permitted quantitative measurements of drug efficacy-related changes in the populations of multiple conformers and the exchange rates among them. The results obtained will be placed in context with data collected using different ¹⁹F-NMR probes positioned differently in the A_2AAR structure (18, 19) and by ¹³C-methyl NMR of selectively labeled isoleucine residues on a deuterated background (20).

Results

Selecting Locations for ¹⁹F-NMR Probes in A_2AAR.

The chemical reaction used here for the introduction of ¹⁹F-groups targets cysteine side chains via a disulfide bond formation. Wild-type A_A2AR contains 14 cysteines, of which 8 are in disulfide bonds and thus protected from the chemical reagent. In the 3D structure of A_A2AR, the remaining six cysteine residues are located in the membrane interior and therefore also inaccessible for chemical reagents, provided that the technique of in-membrane chemical modification (IMCM) (21) is used for the introduction of the NMR probes. Without the need for other amino acid replacements, ¹⁹F-NMR reporter groups could thus be introduced by design into locations near the intracellular surface, where their possible response to ligand-induced conformational rearrangements can provide insights into the mechanisms enabling the physiological signaling by A_A2AR. Cys residues were thus engineered into three such locations and then reacted with 2,2,2-trifluoroethanethiol (TET), yielding C^TET (22).

The selection of the engineered Cys sites was based on comparison of crystal structures of A_2AAR complexes with antagonists and agonists, which showed that the intracellular ends of the transmembrane helices (TM) VI and VII undergo major rearrangements (10–13, 23, 24). We therefore introduced cysteines into sequence positions 225 and 289, which are located near the intracellular tips of TM VI and TM VII, yielding A_2AAR[L225C] and A_2AAR[A289C] (Fig. 1 A and B). In the selection of these two sites, we also considered earlier NMR and crystallographic studies with rhodopsin, the β₂-adrenergic receptor (β₂AR), and A_2AAR, which had identified similar receptor locations as hotspots for activation-related conformational changes (22, 25–30).

Fig. 1.

Location of ¹⁹F-NMR probes in the crystal structure of A_2AAR and NMR response to variable drug efficacy. (A) Side view of a superposition of the antagonist complex A_2AAR–ZM241385 (brown, PDB-ID: 3ELM) with the agonist complex A_2AAR–UK432097 (cyan, PDB-ID: 3QAK); the extracellular membrane surface is at the top. The three sequence positions selected for the ¹⁹F labels are highlighted by spheres and identified. (B) Same as A, view onto the intracellular surface; Roman numerals indicate the TM numbering. (C) The 1D ¹⁹F-NMR spectra of A_2AAR[A289C^TET] complexes with ligands of varying efficacy, as indicated in the Center. On the Right, the NMR spectra shown on the Left are interpreted by Lorentzian deconvolutions with the minimal number of components that provided a good fit, i.e., P₁ to P₄. The chemical shift positions of P₁ to P₄ are indicated by colored broken vertical lines. (D) The 1D ¹⁹F-NMR observation of ligand exchange in A_2AAR[A289C^TET]. The agonist NECA was added at saturating concentration to the apo-A_2AAR and then displaced by the more strongly binding antagonist ZM241385.

As a control, we also investigated A_2AAR[R304C^TET], where the position of the engineered cysteine in helix VIII does not show major differences between the crystal structures of A_2AAR complexes with agonists and antagonists (10, 12–14).

All three variants of A_2AAR used here were shown to retain the pharmacological activity of the parent receptor, using radioligand saturation binding (SI Appendix, Fig. S1).

The ¹⁹F-NMR of A_2AAR Complexes Shows Different Conformational Ensembles in Response to Variable Efficacy of the Bound Drug.

Experiments with A_2AAR[A289C^TET] revealed pronounced ¹⁹F-NMR spectral changes between agonist complexes and antagonist complexes. The complexes with antagonists contain two signals with chemical shifts δ ≈ 11.4 ppm (P₃) and δ ≈ 9.5 ppm (P₁), which coincides with the spectrum of apo-A_2AAR[A289C^TET] (Fig. 1C). Complexes with agonists contain a signal at the chemical shift of P₁ and two new signals, P₂ and P₄, at δ ≈ 10.8 and 13.1 ppm, but there is no signal intensity at position P₃ (Fig. 1C). The line shapes of the dominant components in the spectra of antagonist and agonist complexes at 9.5 and 11.4 ppm, respectively, are closely similar and narrower than the other components. There are at most subtle variations among the spectra of complexes with different antagonists or with different agonists (SI Appendix, Table S1), respectively (Fig. 1 and SI Appendix, Fig. S2), so that the ¹⁹F-NMR spectra of A_2AAR[A289C^TET] represent “fingerprints” for the corresponding functional states. The two different patterns of ¹⁹F-NMR signals seen in Fig. 1C are thus characteristic of conformational ensembles associated, respectively, with the functionally inactive receptor and its active-like state present when an agonist is bound to the extracellular orthosteric site in the absence of interactions with intracellular partner proteins. This assignment is further supported by the ¹⁹F-NMR–detected ligand competition experiments of Fig. 1D. Starting with the two-signal spectrum of apo-A_2AAR[A289C^TET], addition of the agonist NECA resulted in the appearance of the characteristic spectral features for agonist complexes. Subsequent addition of an excess of the high-affinity antagonist ZM241385 yielded a spectrum characteristic of the antagonist-bound receptor. We conclude that we observed transitions from an inactive conformational ensemble of apo-A_2AAR[A289C^TET] to an active-like ensemble of the complex with NECA and then from this active-like state to an inactive ensemble of the complex with ZM241385. It is intriguing that the complex of A_2AAR with the partial agonist LUF 5834 shows the signal of the inactive state, rather than that for the active-like state (Fig. 1 and SI Appendix, Fig. S2).

Observation of the ¹⁹F-NMR signals originating from C^TET in sequence position 225 provided similar results to those obtained from observation of C^TET in position 289. Despite the much smaller dispersion of the chemical shifts (see also the next section), the presence of a new signal for complexes of A_2AAR[L225C^TET] with agonists, compared with complexes with antagonists, is readily apparent (SI Appendix, Fig. S3). The signals obtained from the reporter group in position 225 are again identical among different antagonist complexes and the apo-form of the receptor, and different agonist complexes also show closely similar ¹⁹F-NMR signals (SI Appendix, Fig. S3). Deconvolution of the signals showed a good fit with the assumption of two and three components for the inactive and active-like states (SI Appendix, Fig. S3). While the relative intensities and line widths assigned to the two components of the inactive state signal coincide qualitatively with those observed at C^TET289, the intensity distributions in the signals for the active-like states is clearly different, indicating different responses to drug efficacy at the tips of TMVI and TMVII.

For C^TET in sequence position 304, no differences were seen between the ¹⁹F-NMR signals of the apo-form, complexes with antagonists, and complexes with agonists (SI Appendix, Fig. S4). This absence of a response to different efficacies of the bound ligand is in line with the absence of differences in the spatial arrangement of helix VIII between crystal structures of A_2AAR complexes with agonists or antagonists (Fig. 1 A and B).

In the continuation of this work, we focus on drug complexes with A_2AAR[A289C^TET], making use of the high spectral resolution of the ¹⁹F-NMR spectra recorded with this probe location.

The ¹⁹F-NMR in Solution and Crystal Structures of A_2AAR.

To link the NMR data with the available A_2AAR crystal structures, we performed aromatic ring current shift (δ_R) calculations (SI Appendix, Table S2) (31, 32)_. Since only small ring current shifts of <0.1 ppm were calculated for the C^TET groups at positions 225 and 304, which is due to the absence of nearby aromatic residues and consistent with the small chemical shift dispersion observed at these two sites (SI Appendix, Figs. S3 and S4), the following discussion is focused on A_2AAR [A289C^TET]. For the A_2AAR–ZM214385 complex, a small downfield shift of δ_R ≈ 0.2 ppm was calculated, and for A_2AAR–NECA there was a large upfield shift of δ_R ≈ −1.4 ppm. The difference of about 1.6 ppm is due to conformational rearrangements that bring the aromatic rings of Phe286, Phe295, and Phe299 into proximity of residue 289 in the structure of the agonist complex. These calculated shifts are in close agreement with the experimental chemical shift difference of 1.9 ppm between signals P₁ and P₃ (Fig. 1C), whereas all other combinations of pairs of possible corresponding peaks in the two spectra do not fit the crystal data. The implication is that the most highly populated states P₃ and P₁ in the complexes with ZM241385 and NECA, respectively, coincide with the corresponding crystal structures, whereas the conformers giving rise to the NMR signal P₁ in the antagonist complex and those represented by P₂ and P₄ in the agonist complex were not seen in the crystals. Additional ring current calculations using the presumed fully active state in the crystal structure of the ternary complex of A_2AAR, NECA, and a “mini G protein” (15) yielded an upfield shift of δ_R ≈ −1.2 ppm for Ala289, which is close to that for the active-like A_2AAR–NECA state. In summary, the ring current calculations yielded assignments for the dominantly populated conformations of the inactive (P₃) and active-like (P₁) states and support that the crystal structures do not represent the full repertoire of A_2AAR conformations present in solution at ambient temperature.

Rate Processes in Function-Related Conformational Ensembles of A_2AAR.

The presence of individually resolved NMR signals establishes an upper limit for the exchange rates (k_ex) between the A_2AAR conformations represented by these signals, i.e., k_ex ≲10⁻³ s⁻¹. Here, we explore the dynamic processes in these conformational ensembles further with saturation transfer difference experiments (STD) (33, 34) and 2D exchange spectroscopy (EXSY) (35).

For the A_2AAR[A289C]–ZM241385 complex, 1D ¹⁹F-STD experiments were performed with preirradiation on P_1, monitoring changes in the intensity of P₃, and collecting reference data with preirradiation at 13.3 ppm. The saturation times were 0.05–1.0 s. The measurements were then performed also with preirradiation at P₃ and detection on P₁ (Fig. 2 A and B). Analysis of the STD data using the Bloch–McConnell formalism (36, 37) established that k_ex (3→1) = 2.8 s⁻¹ and k_ex (1→3) = 4.0 s⁻¹ (Fig. 2C).

Fig. 2.

Conformational exchange in the A_2AAR complex with the antagonist ZM241385 by ¹⁹F-NMR saturation transfer. (A, Left) The 1D ¹⁹F-NMR spectrum. The Lorentzian deconvolution introduced in Fig. 1 is indicated. A red arrow indicates the carrier position for the preirradiation, and a black arrow indicates the position for the reference measurement. The observed peak is indicated by “detection.” (A, Right) Plots of the normalized intensity of the observed peak, P₃, versus the saturation pulse length after irradiation on P₁ (cyan) and at the control (black). (B) Same as A, with inverted direction. (C) Survey of the conformational exchange rates in the A_2AAR[A289C^TET]–ZM241385 complex.

For agonist-bound A_2AAR[A289C^TET], STD experiments were performed with preirradiation of all three signals (Fig. 3 A–D, red arrows), and reference data were recorded with preirradiation at the chemical shifts indicated by the black arrows in the figure. Analysis of the STD data resulted in the following exchange rates: k_ex (4→1) = 4.5 s⁻¹, k_ex (1→4) = 0.5 s⁻¹, k_ex (4→2) = 4.1 s⁻¹, k_ex (2→4) = 1.8 s⁻¹, k_ex (2→1) = 23.7 s⁻¹, and k_ex (1→2) = 8.0 s⁻¹ (Fig. 3E). It is remarkable that with the exception of k_ex (2→1) and k_ex (1→2), slow exchange was measured for all pairs of conformers.

Fig. 3.

Conformational exchange in the A_2AAR complex with the full agonist NECA by ¹⁹F-NMR saturation transfer. (A–D) Four individual measurements linking P₁, P₂, and P₄, same presentation as in Fig. 2 A and B. (E) Survey of the conformational exchange rates in the A_2AAR[A289C^TET]–NECA complex. Please note in C and D that the saturation and control saturation are not symmetrical to the detection position, and this arrangement was chosen to prevent falsification of the data by direct irradiation of the tail of the observed signal.

The conformational exchange rates from STD measurements were qualitatively confirmed with the use of 2D exchange spectroscopy (EXSY). In the 2D [¹⁹F,¹⁹F]-EXSY spectrum of the A_2AAR[A289C^TET] complex with the agonist NECA, exchange crosspeaks between P₂ and P₁ were observed, but there were no crosspeaks between P₄ and either P₁ or P₂ (Fig. 4). There were no crosspeaks between the two NMR signals of the antagonist-bound receptor (Fig. 1). The EXSY data thus confirm the coexistence of widely different exchange rates among the conformational substates of the different activation levels of A_2AAR (Figs. 2C and 3E).

Fig. 4.

Conformational exchange in the A_2AAR complex with the full agonist NECA observed by 2D exchange spectroscopy. A contour plot is shown of a 2D [¹⁹F, ¹⁹F]-EXSY spectrum collected at 280 K with a mixing time of 100 ms. The diagonal peak positions of P₁, P₂, and P₄ are labeled, and a dashed box indicates crosspeaks observed between P₁ and P₂.

The ¹⁹F-NMR–Based Comparison of A_2AAR with a Constitutively Active Variant.

A_2AAR[S91A], which contains a single amino acid replacement near the allosteric switch at D^2.50 in the transmembrane-spanning region of the receptor, has been demonstrated to exhibit significant basal activity and increased signaling at full activation in HEK293T cells stimulated with the full agonist CGS21680 (38). The ¹⁹F-NMR spectra of A_2AAR[S91A, A289C^TET] show qualitatively similar signals and exchange rates as A_2AAR (SI Appendix, Fig. S5), confirming the structural integrity of the modified protein. Nonetheless, the functional differences relative to A_2AAR are paralleled by important differences between the NMR spectra. The apo-form of the variant protein shows increased intensity for the signal P₁, compared with apo-A_2AAR[A289C^TET], and there is a weak peak at the position P₄, which had no signal intensity in apo-A_2AAR. Upon addition of the agonist NECA, spectra of A_2AAR[S91A, A289C^TET] showed increased intensities for the signals P₁ and P₄, compared with A_2AAR[A289C^TET], where the increased intensity of P₄ was particularly striking. In competition binding experiments, addition of an excess amount of the high-affinity antagonist ZM241385 to the NECA-bound A_2AAR[S91A, A289C^TET] yielded a spectrum with similar signal positions as for the apo-protein, but different relative signal intensities (Fig. 5B). Overall, in contrast to A_2AAR, the spectra of the apo-form and the antagonist-bound complex of the constitutively active mutant protein contain admixtures of ¹⁹F-NMR signals that are characteristic of the active-like state. For both the inactive state and active-like state, the exchange rates among the different conformers represented by the individual ¹⁹F-NMR signals are closely similar to those measured for the wild-type protein (Figs. 2 and 3 and SI Appendix, Fig. S5).

Fig. 5.

Dynamic processes in the ¹⁹F-labeled constitutively active mutant A_2AAR[S91A, A289C^TET] observed by 1D ¹⁹F-NMR. (A) Same ligand exchange experiment as in Fig. 1D. (B) Lorentzian deconvolutions of the spectra shown in A.

Discussion

The present quantitative description of A_2AAR structural ensembles by ¹⁹F-NMR probes complements a recent qualitative many-parameter overview of A_2AAR structural plasticity by high-resolution NMR, which revealed that there are structural manifolds linked to function (16, 17). The ¹⁹F-NMR spectra observed for probes in TM VI and TM VII are qualitatively similar, but only the spectra TM VII are sufficiently well resolved to enable quantitative measurements, due to the location of the –CF₃ group near aromatic residues that reflect the efficacy of bound drugs.

The STD and 2D EXSY (Figs. 2–4) showed that exchange between the two conformers of the inactive state and with the component P4 in the active-like state is slow. The corresponding high energy barrier indicates that differences between these substates likely involve major structural rearrangements of the polypeptide backbone, while the active-like state contains also two substates, P₂ and P₁ (Figs. 3 and 4), with a low energy barrier for interconversion. The available data indicate that all substates of the active-like conformation are separated from the inactive state by high energy barriers.

Overall, two conclusions can be drawn from the observations. First, considering the high energy barriers separating at least one component in each activation state from the other structures in the ensemble, it could be that only one of the conformers is linked to activation. Second, the difference in resolution between the probes on TM VI and TM VII (Fig. 1 and SI Appendix, Fig. S3) illustrates how the NMR data can be optimized by placement of probes near indigenous aromatic amino acid residues, or possibly by introduction of extrinsic aromatic residues near the NMR probe (39).

A_2AAR is the second human GPCR studied by ¹⁹F-NMR in solution, after the β₂-adrenergic receptor (β₂AR), which contained equilibria between one inactive conformational state and one active-like state (26), which were in slow exchange (25) and showed different relative populations at different probe sites. Considering the present data for A_2AAR, we conclude that β₂AR cannot serve as a model for human GPCRs overall, or even for class A GPCRs. It is further intriguing that the relative populations of active-like states in β₂AR could be directly related to biased agonism (26). Since there are no published findings of biased agonists for A_2AAR, ¹⁹F-NMR experiments may guide future efforts to explore A_2AAR biased agonists.

Since the intensity of the NMR signal P₄ of agonist-bound A_2AAR could be related to the activation level, it now provides a means for assessing basal activity (Fig. 6). As no admixtures of active-like states were detected in either apo- or antagonist-bound A_2AAR[A289C^TET] (Fig. 1 C and D), we conclude that the ¹⁹F-NMR data reflect the inherently low basal activity of A_2AAR measured in cells (38). Recent NMR studies using a ¹⁹F probe in a position on helix VI, which provided a limited chemical shift dispersion comparable to the one shown in SI Appendix, Fig. S3, was interpreted to show that the basal activity amounted to 70% of the activity of agonist-bound A_2AAR (18, 19). Clearly, this contrasts with the present NMR studies and with literature data of the pharmalogical activity of A_2AAR (38). There are additional apparent discrepancies in refs. 18 and 19 relative to the present data; in view of the large chemical shift dispersion and the resulting high spectral resolution by our ¹⁹F probe, we are confident in the interpretation of our results.

Fig. 6.

Biological response versus ligand concentration as manifested in the ¹⁹F NMR spectra. A plot is shown of the biological activity (i.e., G protein signaling) versus ligand concentration for A_2AAR and A_2AAR[S91A], as labeled on the right of the individual sigmoidal response curves. Relative biological activity was determined by observation of the intensity of the peak P₄ in ¹⁹F-NMR spectra of A_2AAR[A289C^TET] and A_2AAR[S91A, A289C^TET]. The dashed lines represent the basal signaling level of the two proteins.

From the relations between the ¹⁹F-NMR data and the A_2AAR crystal structures revealed by ring current shift calculations (SI Appendix, Table S2), only one combination of substates between inactive and active-like A_2AAR is consistent with predicted chemical shifts from the crystal structures. This pair of substates is dominantly populated and shows narrower NMR lines than all other substates except for P4 (Fig. 1). The NMR data in solution at ambient temperature thus manifest conformations that are not observed in crystals. This is most apparent for signal P₄, which is shifted downfield by ∼1.5 ppm from the predicted chemical shift for the active-like state. The increased intensity of the NMR signal P₄ for the antagonist complex of a constitutively active mutant establishes a direct relation to A_2AAR activation. Since all available A_2AAR crystal structures thus represent only a fraction of the repertoire of function-relevant conformations, it could even be that the physiological action of A_2AAR is mainly based on a conformer that has not yet been seen in crystal structures. In this context, it is also intriguing that A_2AAR[A289C^TET] with the bound partial agonist LUF5834 yields spectra that are those of antagonist-bound A_2AAR[A289C^TET], which suggests that the local conformation at the intracellular end of TM VII corresponds to an inactive state for this partial agonist.

Materials and Methods

The TET-labeled A_2AAR variants used in this study were expressed and purified as previously described (21). The solutions used for NMR measurements contained 25 to 50 µM protein in 50 mM Hepes at pH 7.5, 150 mM NaCl, 0.05% DDM, 0.01% CHS, and excess ligand. All NMR data were measured on a Bruker AVANCE 600 spectrometer using a QCI ¹H/¹⁹F-¹³C/¹⁵N quadruple resonance cryoprobe with shielded z-gradient coil. Spectra were processed using Bruker TOPSPIN version 3.1, and signals were deconvoluted using the MNOVA software version 10.0.0.2. Additional details are described in SI Appendix.