Oligomerization of the Human Adenosine A2A Receptor is Driven by the Intrinsically Disordered C-terminus

G protein-coupled receptors (GPCRs) have long been shown to exist as oligomers with functional properties distinct from those of the monomeric counterparts, but the driving factors of GPCR oligomerization remain relatively unexplored. In this study, we focus on the human adenosine A2A receptor (A2AR), a model GPCR that forms oligomers both in vitro and in vivo. Combining experimental and computational approaches, we discover that the intrinsically disordered C-terminus of A2AR drives the homo-oligomerization of the receptor. The formation of A2AR oligomers declines progressively and systematically with the shortening of the C-terminus. Multiple interaction sites and types are responsible for A2AR oligomerization, including disulfide linkages, hydrogen bonds, electrostatic interactions, and hydrophobic interactions. These interactions are enhanced by depletion interactions along the C-terminus, forming a tunable network of bonds that allow A2AR oligomers to adopt multiple interfaces. This study uncovers the disordered C-terminus as a prominent driving factor for the oligomerization of a GPCR, offering important guidance for structure-function studies of A2AR and other GPCRs.


INTRODUCTION 40
G protein-coupled receptors (GPCRs) have long been studied as monomeric units, but 41 accumulating evidence demonstrates that these receptors can also form homo-and hetero-42 oligomers with far-reaching functional implications. The properties emerging from these 43 oligomers can be distinct from those of the monomeric protomers in ligand binding (1)(2)(3)(4)

), G protein 44
We performed SEC analysis on a mixture of ligand-active A2AR purified from a custom 126 synthesized antagonist affinity column (Fig. S1A). Distinct oligomeric species were separated and 127 eluted in the following order: high-molecular-weight (HMW) oligomer, dimer, and monomer ( Fig.  128   1 and Fig. S1B). The population of each oligomeric species was quantified as the integral of each 129 Gaussian from a multiple-Gaussian curve fit of the SEC signal. The reported standard errors were 130 calculated from the variance of the fit that do not correspond to experimental errors (see Table S1  131 and Fig. S2 for SEC data corresponding to all A2AR variants in this study). As this study sought to 132 identify the factors that promote A2AR oligomerization, the populations with oligomeric interfaces 133 (i.e., dimer and HMW oligomer) were compared with those without such interfaces (i.e., 134 monomer). Hence, the populations of the HMW oligomer and dimer were expressed relative to the 135 monomer population in arbitrary units as monomer-equivalent concentration ratios, henceforth 136 referred to as population levels ( Fig. 1). 137

C-Terminal Amino Acid Residue C394 Contributes to A2AR Oligomerization 144
To investigate whether the C-terminus of A2AR is involved in receptor oligomerization, we first 145 examined the role of residue C394, as a previous study demonstrated that the mutation C394S 146 dramatically reduced A2AR oligomer levels(63). The C394S mutation was replicated in our 147 experiments, alongside other amino acid substitutions, namely alanine, leucine, methionine or 148 valine, generating five A2AR-C394X variants. The HMW oligomer and dimer levels of A2AR wild-149 type (WT) were compared with those of the A2AR-C394X variants. We found that the dimer level 150 of A2AR-WT was significantly higher than that of the A2AR-C394X variants (WT: 1.14; C394X: 151 0.24-0.57; Fig. 2A). A similar result, though less pronounced, was observed when the HMW 152 oligomer and dimer levels were considered together (WT: 1.34; C394X: 0.59-1.21; Fig. 2A). This 153 suggests that residue C394 plays a role in A2AR oligomerization and more so in A2AR dimers. 154

161
To test whether residue C394 stabilizes A2AR dimerization by forming disulfide linkages, we 162 incubated SEC-separated A2AR dimer with 5 mM of the reducing agent TCEP, followed by SDS-163 PAGE and Western Blotting. The population of each species was determined as the area under the 164 densitometric trace. The dimer level was then expressed as monomer-equivalent concentration 165 ratios in a manner similar to that of the SEC experiment described above. Upon incubation with 166 TCEP, the dimer level of the sample decreased from 1.14 to 0.51 (Fig. 2B). This indicates that 167 disulfide bond formation via residue C394 is one possible mechanism for A2AR dimerization. 168 However, a significant population of A2AR dimer remained resistant to TCEP and C394X 169 mutations (Fig. 2), suggesting that disulfide linkages are not the only driving factor of A2AR 170 oligomer formation. This finding agrees with a previous study showing that residue C394 in A2AR 171 dimer is still available for nitroxide spin labeling,(63) suggesting that additional interfacial sites 172 help drive A2AR dimer/oligomerization. 173

C-Terminus Truncation Systematically Reduces A2AR Oligomerization 174
To determine which interfacial sites in the C-terminus other than C394 drive A2AR 175 dimer/oligomerization, we carried out systematic truncations at eight sites along the C-terminus 176 (A316, V334, G344, G349, P354, N359, Q372, and P395), generating eight A2AR-ΔC variants 177 (Fig. 3A). The A2AR-A316ΔC variant corresponds to the removal of the entire disordered C-178 terminal region as previously performed in all published structural studies(24, 27, 52-59). Using 179 the SEC analysis described earlier ( Fig. 1) we evaluated the HMW oligomer and dimer levels of 180 the A2AR-ΔC variants relative to that of the A2AR full-length-wild-type (FL-WT) control. Both the 181 dimer and the total oligomer levels of A2AR decreased progressively with the shortening of the C-182 terminus, with almost no oligomerization detected upon complete truncation of the C-terminus at 183 site A316 (Fig. 3B). This result shows that the C-terminus drives A2AR oligomerization, with 184 multiple potential interaction sites positioned along much of its length. 185 Interestingly, there occurred a dramatic decrease in the dimer level between the N359 and P354 186 truncation sites, from a value of 0.81 to 0.19, respectively (Fig. 3B). A similar result, though less 187 pronounced, was observed on the total oligomer level, with a decrease from 1.09 to 0.62 for the 188 N359 and P354 truncation sites, respectively (Fig. 3B) (Fig. 3C). We then compared the HMW oligomer and dimer levels of the resulting 206 variants with controls (same A2AR variants but without the ERR:AAA mutations). We found that 207 the ERR:AAA mutations had varied effects on the dimer level: decreasing for A2AR-FL-WT (ctrl: 208 0.49; ERR:AAA: 0.29) but increasing for A2AR-N359ΔC (ctrl: 0.33; ERR:AAA: 0.48) (Fig. 3C). Given that the structure of A2AR dimers or oligomers are unknown, we next used molecular 220 dynamics (MD) simulations to seek molecular-level insights into the role of the C-terminus in 221 driving A2AR dimerization and to determine the specific interaction types and sites involved in this 222 process. First, to explore A2AR dimeric interface, we performed coarse-grained (CG) MD 223 simulations, which can access the length and time scales relevant to membrane protein 224 oligomerization, albeit at the expense of atomic-level details. We carried out a series of CGMD 225 simulations on five A2AR-ΔC variants designed to mirror the experiments by systematic truncation 226 at five sites along the C-terminus (A316, V334, P354, N359, and C394). Our results revealed that 227 A2AR dimers were formed with multiple interfaces, all involving the C-terminus ( Fig. 4A and S3A). 228 The vast majority of A2AR dimers were symmetric, with the C-termini of the protomers directly 229 interacting with each other. A smaller fraction of the dimers had asymmetric orientations, with the 230 C-terminus of one protomer interacting with other parts of the other protomer, such as ICL2 (the 231 second intracellular loop), ICL3, and ECL2 (the second extracellular loop) (Fig. 4A). hydrogen bonds involving C-terminal residues progressively declined, respectively reaching 5.4 247 and 6.0 for A2AR-A316ΔC (in which the disordered region of the C-terminus is removed) (Fig. 4B  248 and 4C). This result is consistent with the experimental result, which demonstrated a progressive 249 decrease of A2AR oligomerization with the shortening of the C-terminus (Fig. 3B). Interestingly, 250 upon systematic truncation of the C-terminal segment 335-394, we observed in segment 291-334 251 a steady decrease in the average number of electrostatic contacts, from 10.4 to 7.4 (Fig. 4B). This 252 trend was even more pronounced with hydrogen bonding contacts involving segment 291-334 253 decreasing drastically from 21.0 to 7.0 as segment 335-394 was gradually removed (Fig. 4C). 254 This observation, namely that truncation of a C-terminal segment reduces inter-A2AR contacts 255 elsewhere along the C-terminus, indicates that a cooperative mechanism of dimerization exists, in 256 which an extended C-terminus of A2AR stabilizes inter-A2AR interactions near the heptahelical 257 bundles of the dimeric complex. Besides the intermolecular interactions, we also identified a 258 network of intramolecular salt bridges involving residues on the C-termini, including cluster 259 355 ERR 357 (Fig. 7A). These results demonstrate that A2AR dimers can be formed via multiple 260 interfaces predominantly in symmetric orientations, facilitated a cooperative network of 261 electrostatic interactions and hydrogen bonds along much of its C-terminus. 262   Table 1). The HMW oligomer and dimer levels of the four A2AR variants were 282 determined and plotted as a function of ionic strengths. 283 The low ionic strength of 0.15 M should not affect hydrogen bonds or electrostatic interactions, if 284 present. We found that the dimer and total oligomer levels of all four variants were near zero ( Upon closer examination, we recognize that at the very high ionic strength of 0.95 M, the increase 295 in the dimer and total oligomer levels was robust for A2AR-FL-WT, but less pronounced for A2AR-296 FL-ERR:AAA (Fig. 5). Furthermore, this high ionic strength even had an opposite effect on A2AR-297 N359ΔC, with both its dimer and total oligomer levels abolished (Fig. 5). These results indicate 298 that the charged cluster 355 ERR 357 and the C-terminal segment after residue N359 are required for 299 depletion interactions to promote A2AR oligomerization to the full extent. 300 Taken together, we demonstrated that A2AR oligomerization is more robust when the C-terminus 301 is fully present and the ionic strength is higher, suggesting that depletion interactions via the C-302 terminus are a strong driving factor of A2AR oligomerization. The question then arises whether 303 such depletion interactions are the result of the C-termini directly interacting with one another, 304 necessitating an experiment that investigates the behavior of A2AR C-terminus sans the 305 transmembrane domains.

311
The Isolated A2AR C-Terminus Is Prone to Aggregation 312 To test whether A2AR oligomerization is driven by direct depletion interactions among the C-313 termini of the protomers, we assayed the solubility and assembly properties of the stand-alone 314 A2AR C-terminus-an intrinsically disordered peptide-sans the upstream transmembrane regions. 315 Since depletion interactions can be manifested via the hydrophobic effect(48), we examined 316 whether this effect can cause A2AR C-terminal peptides to associate. 317 It is an active debate(67) whether the hydrophobic effect can be promoted or suppressed by ions 318 with salting-out or salting-in tendency, respectively(68-70). We increased the solvent ionic 319 strength using either sodium (salting-out) or guanidinium (salting-in) ions and assessed the 320 aggregation propensity of the C-terminal peptides using UV-Vis absorption at 450 nm. We first 321 observed the behavior of the C-terminus with increasing salting-out NaCl concentrations. At NaCl 322 concentrations below 1 M, the peptide was dominantly monomeric, despite showing slight 323 aggregation at NaCl concentrations between 250-500 mM (Fig. 6A). At NaCl concentrations 324 above 1 M, A2AR C-terminal peptides strongly associated into insoluble aggregates (Fig. 6A). 325 Consistent with the observations made with the intact receptor (Fig. 5), A2AR C-terminus showed 326 the tendency to progressively precipitate with increasing ionic strengths, suggesting that depletion 327 interactions drive the association and precipitation of the peptides. We next observed the behavior 328 of the C-terminus with increasing concentrations of guanidine hydrochloride (GdnHCl), which 329 contains salting-in cations that do not cause proteins to precipitate and instead facilitate the 330 solubilization of proteins(71, 72). Our results demonstrated that the A2AR C-terminus incubated in 331 4 M GdnHCl showed no aggregation propensity (Fig. 6A), validating our expectation that 332 depletion interactions are not enhanced by salting-out salts. These observations demonstrate that 333 the C-terminal peptide in and of itself can directly interact with other C-terminal peptides to form 334 self-aggregates in the presence of ions, and presumably solutes, that have salting-out effects. 335 Attractive hydrophobic interactions among the hydrophobic residues are further enhanced by water 336 solvating the protein having more favorable interactions with other water molecules, ions or 337 solutes than with the protein, here the truncated C-terminus(73-75). We explored the possible 338 contribution of hydrophobic interactions to the aggregation of the C-terminal peptides using 339 differential scanning fluorimetry (DSF). In particular, we gradually increased the temperature to 340 melt the C-terminal peptides, exposing any previously buried hydrophobic residues (Fig. S4A) 341 which then bound to the SYPRO orange fluorophore, resulting in an increase in fluorescence signal. 342 Our results showed that as the temperature increased, a steady rise in fluorescence was observed 343 (Fig. 6B), indicating that multiple hydrophobic residues were gradually exposed to the SYPRO 344 dye. However, at approximately 65°C, the melt peak signal was abruptly quenched (Fig. 6B), 345 indicating that the hydrophobic residues were no longer exposed to the dye. This observation 346 suggests that, at 65°C, enough hydrophobic residues in the C-terminal peptides were exposed such 347 that they collapsed on one another (thus expelling the bound dye molecules), resulting in 348 aggregation. Clearly, the hydrophobic effect can cause A2AR C-terminal peptides to directly 349 associate. These results demonstrate that A2AR oligomer formation can be driven by depletion 350 interactions among the C-termini of the protomers.

DISCUSSION 359
The key finding of this study is that the C-terminus of A2AR, removed in all previously published 360 structural studies of this receptor, is directly responsible for receptor oligomerization. Using a 361 combination of experimental and computational approaches, we demonstrate that the C-terminus 362 drives A2AR oligomerization via a combination of disulfide linkages, hydrogen bonds, electrostatic 363 interactions, and hydrophobic interactions. This diverse combination of interactions is greatly 364 enhanced by depletion interactions, forming a network of malleable bonds that give rise to the 365 existence of multiple A2AR oligomeric interfaces. 366 The intermolecular disulfide linkages associated with residue C394 play a role in A2AR 367 oligomerization. However, it is unclear which cysteine on the second protomer is linked to this 368 cysteine. A previous study showed that residue C394 in A2AR dimer is available for nitroxide spin 369 labeling(63), suggesting that some of these disulfide bonds may be between residue C394 and 370 another cysteine in the hydrophobic core of A2AR that do not form intramolecular disulfide The electrostatic interactions that stabilize A2AR oligomer formation come from multiple sites 380 along the C-terminus. From a representative snapshot of a A2AR-C394ΔC dimer from our MD 381 simulations (Fig. 7A), we could visualize not only the intermolecular interactions calculated from 382 the CGMD simulations (Fig. 4B), but also intramolecular salt bridges. In particular, the 355 ERR 357 383 cluster of charged residues lies distal from the dimeric interface, yet still forms several salt bridges 384 (Fig. 7A, inset). This observation is supported by our experimental results showing that 385 substituting this charged cluster with alanines reduces the total A2AR oligomer levels (Fig. 3C). 386 However, it is unclear how such salt bridges involving this 355 ERR 357 cluster are enhanced by 387 depletion interactions (Fig. 5), as electrostatic interactions are usually screened out at high ionic 388 strengths. In our MD simulations, we also observed networks of salt bridges along the dimeric 389 interface, for example between K315 of one monomer and D382 and E384 of the other monomer 390 (Fig. 7A, inset). The innate flexibility of the C-terminus could facilitate the formation of such salt 391 bridges, which then acts as a potential scaffold to stabilize A2AR dimers. 392

399
We also found that depletion interactions can enhance the diversity of interactions that stabilize 400 A2AR oligomer formation ( Fig. 5 and 6). Depletion interactions could be the key factor to the 401 cooperative mechanism by which A2AR oligomerization occurs. As revealed by our MD 402 simulations, an increasing number of contacts are formed along segment 291-334 when the rest 403 of C-terminus is present (Fig. 4B and 4C). As more of the C-terminus is preserved, the greater 404 extent of depletion interactions limits the available dimer arrangements, forcing segment 291-334 405 into an orientation that optimizes intermolecular interactions. 406 Our finding that A2AR forms homo-oligomers via multiple interfaces (Fig. 4A)  Not only did we find multiple A2AR oligomeric interfaces, we also found that these interfaces can 417 be either symmetric or asymmetric. This finding is supported by a growing body of evidence that 418 there exists both symmetric and asymmetric oligomeric interfaces for A2AR(24) and many other 419 GPCRs. Studies using various biochemical and biophysical techniques have shown that 420 heterotetrameric GPCR complexes can be formed by dimers of dimers, including μOR-δOR (97) dimers. Specifically, as more of the C-terminus was preserved, we observed a progressive increase 435 in the helical tilt of TM7 (Fig. 7B). This change in helical tilt occurred for the entire heptahelical 436 bundle, with an increase in tilt for TM1, TM2, TM3, TM5, and TM7, and a decrease in tilt for 437 TM4 and TM6 (Fig. S3). The longer C-terminus in the full-length A2AR permits greater 438 rearrangements in the transmembrane regions, leading to the observed change in helical tilt. This 439 result hints at potential conformational changes of A2AR upon oligomerization, necessitating future 440 investigation on functional consequences. 441 C-terminal truncations prior to crystallization and structural studies may be the main reason for 442 the scarcity of GPCR structures featuring oligomers. In that context, this study offers valuable 443 insights and approaches to tune the oligomerization of A2AR and potentially of other GPCRs using 444 its intrinsically disordered C-terminus. The presence of A2AR oligomeric populations with partial 445 C-terminal truncations means that one can now study its oligomerization with less perturbation 446 from the C-terminus. We also present evidence that the multiple C-terminal interactions that drive 447 A2AR oligomerization can be easily modulated by ionic strength and specific salts (Fig. 5 and 6). 448 Given that ~75% and ~15% of all class-A GPCRs possess a C-terminus of > 50 and > 100 amino 449 acid residues(108), respectively, it will be worthwhile to explore the prospect of tuning GPCR 450 oligomerization not only by shortening the C-terminus but also with simpler approaches such as 451 modulating ionic strength and the surrounding salt environment. 452

CONCLUSION 453
This study emphasizes for the first time the definite impact of the C-terminus on A2AR 454 oligomerization, which can be extended to include the oligomers formed by other GPCRs with a 455 protracted C-terminus. We have shown that the oligomerization of A2AR is strongly driven by 456 depletion interactions along the C-terminus, further modulating and enhancing the multiple 457 interfaces formed via a combination of hydrogen, electrostatic, hydrophobic, and covalent 458 disulfide interactions. The task remains to link A2AR oligomerization to functional roles of the 459 receptor(109). From a structural biology standpoint, visualizing the multiple oligomeric interfaces 460 of A2AR in the presence of the full-length C-terminus is key to investigating whether these 461 interfaces give rise to different oligomer functions. 462

Cloning, Gene Expression, and Protein Purification 464
The multi-integrating pITy plasmid(110), previously used for overexpression of A2AR in 465 Saccharomyces cerevisiae(111), was employed in this study. pITy contains a Gal1-10 promoter 466 for galactose-induced expression, a synthetic pre-pro leader sequence which directs protein 467 trafficking(112, 113), and the yeast alpha terminator. The genes encoding A2AR variants with 10-468 His C-terminal tag were cloned into pITy downstream of the pre-pro leader sequence, using either 469 splice overlapping extension(114) or USER cloning using X7 polymerase(115, 116), with primers 470 provided in Table S3. The plasmids were then transformed into S. cerevisiae strain BJ5464 471 (MATα ura3-52 trp1 leu2∆1 his3∆200 pep4::HIS3 prb1∆1.6R can1 GAL) (provided by the lab of 472 Anne Robinson at Carnegie Mellon University) using the lithium-acetate/PEG method(117). 473 Transformants were selected on YPD G-418 plates (1% yeast extract, 2% peptone, 2% dextrose, 474 2.0 mg/mL G-418). 475 Receptor was expressed and purified following the previously described protocol(118). In brief, 476 from freshly streaked YPD plates (1% yeast extract, 2% peptone, 2% dextrose), single colonies 477 were grown in 5-mL YPD cultures over night at 30ºC. From these 5-mL cultures, 50-mL cultures 478 were grown with a starting OD of 0.5 over night at 30ºC. To induce expression, yeast cells from 479 these 50-mL cultures were centrifuged at 3,000 x g to remove YPD before resuspended in YPG 480 medium (1% yeast, 2% peptone, 2% D-galactose) at a starting OD of 0.5. signal. The HMW oligomer peak in some cases could not be fitted with one curve and thus was 537 fitted with two curves instead. The reported standard errors were calculated from the variance of 538 the fit and did not correspond to experimental errors. The results are detailed in Fig. S2 and Table  539 S1. 540 Western blot was analyzed with Fiji. The Gels analysis plugin was used to define each sample lane, 558

SDS-PAGE and Western
and to generate an intensity profile. Peaks were manually selected and integrated with the measure 559 tool to determine the amount of protein present. 560

Coarse-Grained MD Simulations 561
Initial configuration of A2AR was based on the crystal structure of the receptor in the active state 562 (PDB 5G53). All non-receptor components were removed, and missing residues added using 563 MODELLER 9.23(121). Default protonation states of ionizable residues were used. The resulting 564 structure was converted to MARTINI coarse-grained topology using the martinize.py script(122). 565 The ELNeDyn elastic network(123) was used to constrain protein secondary and tertiary structures 566 with a force constant of 500 kJ/mol/nm 2 and a cutoff of 1.5 nm. To optimize loop refinement of 567 the model, a single copy was embedded in a 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine 568 (POPC) bilayer using the insane.py script, solvated with MARTINI polarizable water, neutralized 569 with 0.15 M NaCl, and a short MD (1.5 µs) run to equilibrate the loop regions. Subsequently, two 570 monomers of the equilibrated A2AR were randomly rotated and placed at the center of a 13 nm × 571 13 nm × 11 nm (xyz) box, 3.5 nm apart, with their principal transmembrane axis aligned parallel 572 to the z axis. The proteins were then embedded in a POPC bilayer using the insane.py script. 573 Sodium and chloride ions were added to neutralize the system and obtain a concentration of 0.15 574 M NaCl. Total system size was typically in the range of 34,000 CG particles, with a 280:1 575 lipid:protein ratio. Ten independent copies were generated for each A2AR truncated variant. 576 v2.2 of the MARTINI coarse-grained force field(124) was used for the protein and water, and v2.0 577 was used for POPC. All coarse-grained simulations were carried out in GROMACS 2016(125) in 578 the NPT ensemble (P = 1 atm, T = 310 K). The Bussi velocity rescaling thermostat was used for 579 temperature control with a coupling constant of τt = 1.0 ps(126), while the Parrinello-580 Rahman barostat(127) was used to control the pressure semi-isotropically with a coupling constant 581 of τt = 12.0 ps and compressibility of 3 x 10 -4 bar -1 . Reaction field electrostatics was used with 582 Coulomb cut-off of 1.1 nm. Non-bonded Lennard-Jones interactions were treated with a cut-off of 583 1.1 nm. All simulations were run with a 15 fs timestep, updating neighbor lists every 10 steps. 584 Cubic periodic boundary conditions along the x, y and z axes were used. Each simulation was run 585 for 8 µs. 586

Atomistic MD Simulations 587
Three snapshots of symmetric dimers of A2AR for each respective truncated variant were randomly 588 selected from the CG simulations as starting structures for backmapping. Coarse-grained systems 589 were converted to atomistic resolution using the backward.py script(128). All simulations were 590 run in Gromacs2019 in the NPT ensemble (P = 1 bar, T = 310 K) with all bonds restrained using 591 the LINCS method(129). The Parrinello-Rahman barostat was used to control the pressure semi-592 isotropically with a coupling constant of τt = 1.0 ps and a compressibility of 4.5 x 10 -5 bar -1 , while 593 the Bussi velocity rescaling thermostat was used for temperature control with a coupling constant 594 of τt = 0.1 ps. Proteins, lipids, and solvents were separately coupled to the thermostat. The 595 CHARMM36 and TIP3P force fields(130, 131) were used to model all molecular interactions. 596 Periodic boundary conditions were set in the x, y, and z directions. Particle mesh Ewald (PME) 597 electrostatics was used with a cut-off of 1.0 nm. A 2-fs time step was used for all atomistic runs, 598 and each simulation was run for 50 ns. 599

Analysis of Computational Results 600
All trajectories were post-processed using gromacs tools and in-house scripts. We ran a clustering 601 analysis of all dimer frames from the CG simulations using Daura et. al.'s clustering algorithm(132) 602 implemented in GROMACS, with an RMSD cutoff of 1.5 Å. (An interface was considered dimeric 603 if the minimum center of mass distance between the protomers was less than 5 Å.) This method 604 uses an RMSD cutoff to group all conformations with the largest number of neighbors into a cluster 605 and eliminates these from the pool, then repeats the process until the pool is empty. We focused 606 our analysis on the most populated cluster from each truncated variant. Electrostatic interactions 607 in the dimer were calculated from CG systems with LOOS(133) using a distance cutoff of 5.0 Å. 608

Isolated C-Terminus Purification 621
Escherichia coli BL21 (DE3) cells were transfected with pET28a DNA plasmids containing the 622 desired A2AR sequence with a 6x His tag attached for purification. Cells from glycerol stock were 623 grown in 10 mL luria broth (LB, Sigma Aldrich, L3022) overnight at 37˚C and then used to 624 inoculate 1 L of fresh LB and 10 μg/mL kanamycin (Fisher Scientific, BP906). Growth of cells 625 were performed at 37°C, 200 rpm until optical density at λ = 600 nm reached 0.6-0.8. Expression 626 was induced by incubation with 1 mM isopropyl-β-D-thiogalactoside (Fisher Bioreagents, 627 BP175510) for 3 hrs. 628 Cells were harvested with centrifugation at 5000 rpm for 30 min. Harvested cells were resuspended 629 in 25 mL Tris-HCl, pH = 7.4, 100 mM NaCl, 0.5 mM DTT, 0.1 mM EDTA with 1 Pierce protease 630 inhibitor tablet (Thermo Scientific, A32965), 1 mM PMSF, 2 mg/mL lysozyme, 20 μg/mL DNase 631 (Sigma, DN25) and 10 mM MgCl2, and incubated on ice for 30 min. Samples were then incubated 632 at 30ºC for 20 minutes, then flash frozen and thawed 3 times in LN2. Samples were then centrifuged 633 at 10,000 rpm for 10 min to remove cell debris. 1 mM PMSF was added again and the resulting 634 supernatant was incubated while rotating for at least 4 hrs with Ni-NTA resin. The resin was loaded 635 to a column and washed with 25 mL 20 mM sodium phosphate, pH = 7.0, 1 M NaCl, 20 mM 636 imidazole, 0.5 mM DTT, 100 μM EDTA. Purified protein was eluted with 15 mL of 20 mM sodium 637 phosphate, pH = 7.0, 0.5 mM DTT, 100 mM NaCl, 300 mM imidazole. The protein was 638 concentrated to a volume of 2.5mL and was buffer exchanged into 20 mM ammonium acetate 639 buffer, pH = 7.4, 100 mM NaCl using a GE PD-10 desalting column. Purity of sample was 640 confirmed with SDS-PAGE and western blot. 641

Aggregation Assay to Assess A2AR C-Terminus Assembly 642
Absorbance was measured at 450 nm using a Shimadzu UV-1601 spectrophotometer with 120 µL 643 sample size. Prior to reading, samples were incubated at 40°C for 5 minutes. Samples were 644 vigorously pipetted to homogenize any precipitate before absorbance was measured. Protein 645 concentration was 50 µM in a 20 mM ammonium acetate buffer (pH = 7.4). 646

Differential Scanning Fluorimetry (DSF) 647
DSF was conducted with a Bio-rad CFX90 real-time PCR machine. A starting temperature 20ºC 648 was increased at a rate of 0.5ºC per 30 seconds to a final temperature of 85ºC. All samples 649 contained 40 μL of 40 µM A2AR C-terminus, 9x SYPRO orange (ThermoFisher S6650), 200 mM 650 NaCl, and 20 mM MES. Fluorescence was detected in real-time at 570 nm. All samples were 651 conducted in triplicate. 652

Hydrophobicity and Charge Profile of C-Terminus 653
The hydrophobicity profile reported in Fig. S4 was determined with ProtScale using method 654 described by Kyte & Doolittle(136), window size of 3. 655

FUNDING AND ACKNOWLEDGMENTS 656
This material is based upon work supported by (1)