Adenosine-5'-(N-ethylcarboxamide)

Functional activity was determined as described earlier [34 (link)]. Briefly, cells stably expressing both the hA_2A or hA_2B AR and the plasmid pGloSensor-22F coding for the biosensor were cultured. This biosensor encodes for a genetically modified form of firefly luciferase into which a cAMP-binding protein moiety was inserted. The desiderate cell number was harvested and incubated for 2 h at r.t., with 3% v/v GloSensor cAMP reagent stock solution, 10% FBS, and 87% CO₂ independent medium. Cells were dispensed in the wells of a 384-well plate and the reference agonist NECA or the under-study compounds were added at different concentrations. Since the compounds were unable to stimulate cAMP production, they were studied as antagonists. The antagonist profile was evaluated by assessing the ability to counteract an NECA-induced increase in cAMP accumulation. Responses were expressed as a percentage of the maximal relative luminescence units (RLU). Concentration–response curves were fitted by a nonlinear regression using Prism software. The antagonist profile of the compounds was expressed as the IC₅₀, which is the concentration of antagonist that produces a 50% inhibition of the agonist effect.

Three different A_2AAR conformations were used as starting configurations: the complex with the antagonist ZM241385 (PDB ID: 3EML)^{53 (link)}, the ternary complex with the full agonist NECA and an engineered mini-G protein (PDB ID: 5G53)^{33 (link)}, and a conformation of the same ternary complex with the mini-G protein deleted. For the latter simulation, production simulations were run with the backbone heavy atoms softly restrained to stay close to the fully active state by harmonic restraints with a 0.5 kJ/mol/Ų spring constant. The T4-lysozyme was removed from the inactive structure and initial coordinates for residues 209-219 were obtained from the structure of a thermostabilized A_2AAR variant mutant (PDB: 3PWH)^{54 (link)} for which those residues are resolved. The N- and C-termini were capped with the default chemistry in the CHARMM force-field, a primary amine at the N-terminus and a carboxylate at the C-terminus.
All simulations were performed with Gromacs 2020.4. Each system was prepared individually for production simulation using the CHARMMGUI membrane builder tools^{55 (link)}. The protein was embedded in a mixture of POPS and POPC (Supplementary Table 3), solvated with TIP3P water^{56 (link)}, neutralized with Na+ ions if needed, and NaCl added to bring the ionic strength to 150 mM. Lipids and protein were modeled with the CHARMM36 force field^{57 (link),58 (link)}; ligands were modeled with the CHARMM general force field⁵⁹ . Final system sizes were at least 15 nm in each dimension, yielding lipid-to-protein ratios of ~750:1.
Each system was prepared individually for production simulation through a series of 6 minimization and heating steps: (i) steepest descent to minimize the initial configuration; (ii) 125,000 steps of leapfrog dynamics with a 1 fsec timestep and velocities reassigned every 500 steps; (iii) 125,000 steps of leapfrog dynamics with a 1 fsec timestep, pressure controlled by the Parinello-Rahman barostat^{60 (link)} and velocities reassigned every 500 steps, then a total of 750,000 steps of leapfrog dynamics with a 2 fsec timestep and hydrogen positions constrained by LINCS^{61 (link)}, pressure controlled by the Parinello-Rahman barostat, and velocities reassigned every 500 steps. During equilibration, double bonds were restrained in the cis configuration to prevent isomerization; these restraints are gradually reduced during the final three stages of the equilibration protocol. Production simulations (NPT ensemble) were integrated with leapfrog using the Parinello-Rahman barostat to control pressure (time constant 5 psec; compressibility 4.5e⁻⁵ bar⁻¹; coupled anisotropically to allow independent fluctuation of the in-plane and normal directions) and temperature controlled using Nose-Hoover^{62 (link)} (time constant 1 psec) at a temperature of 25 °C. Hydrogens were constrained with LINCS (expansion order 4), a 2 fsec timestep was used, short range electrostatics were computed directly within 1.2 nm, and long-range electrostatics were computed every timestep using particle mesh Ewald^{63 (link)} with a grid spacing of 1 Å and cubic interpolation. Long range dispersion was smoothly truncated over 10-12 nm using a force-switch cutoff scheme. Residue-residue distances as reported in Fig. 5 were measured between the closest sidechain nitrogen atoms in H230^6.32 and R291^7.56.

Competition binding assays with nanodiscs containing A_2AAR were recorded as previously described^{28 (link)}. Ligand binding was measured with 0.125-0.25 µg nanodiscs containing A_2AAR per sample incubated in buffer containing 25 mM HEPES pH 7.0, 75 mM NaCl, [³H]ZM241385 (American radiolabeled chemicals, SKU: ART 0884-50 µCi) and increasing amounts of ZM241385 or NECA for 60 min at 25 °C. The binding reaction was terminated by filtration with a Microbeta filtermat-96 cell harvester (PerkinElmer). Radioactivity was counted using a MicroBeta2 microplate counter (PerkinElmer). ZM241385 and NECA binding affinities (K_D or K_I) were determined using competition binding experiments. Specific binding of A_2AAR and A_2AAR variants were determined as the difference in binding obtained in the absence and presence of 10 µM ZM241385. Radioligand experiments were conducted in triplicate and IC₅₀ values determined using a nonlinear, least-square regression analysis (Prism 8; GraphPad Software, Inc.). The IC₅₀ values were converted to K_I values using the Cheng−Prusoff equation^{51 (link)}. Error bars for each sample were calculated as the standard error of mean (s.e.m) for n = 3 independent experiments.