Mapping the Allosteric Sites of the A_2A Adenosine Receptor

Abstract

The A_2A adenosine receptor (A_2AAR) is a G protein-coupled receptor that is pharmacologically targeted for the treatment of inflammation, sepsis, cancer, neuro-degeneration, and Parkinson’s disease. Recently, we applied long-timescale molecular dynamics simulations on two ligand-free receptor conformations, starting from the agonist-bound (PDB ID:3QAK) and antagonist-bound (PDB ID:3EML) X-ray structures. This analysis revealed four distinct conformers of the A_2AAR: the active, intermediate 1, intermediate 2, and inactive. In this study, we apply the fragment-based mapping algorithm, FTMap, on these receptor conformations to uncover five non-orthosteric sites on the A_2AAR. Two sites that are identified in the active conformation are located in the intracellular region of the transmembrane helices (TM) 3/TM4 and the G protein-binding site in the intracellular region between TM2/TM3/TM6/TM7. Three sites are identified in the intermediate 1 and intermediate 2 conformations, annexing a site in the lipid interface of TM5/TM6. Five sites are identified in the inactive conformation, comprising of a site in the intracellular region of TM1/TM7, and in the extracellular region of TM3/TM4 of the A_2AAR. We postulate that these sites on the A_2AAR be screened for allosteric modulators for the treatment of inflammatory and neurological diseases.

Graphical Abstract

We uncover five non-orthosteric sites on the A_2A­ adenosine receptor that can be screened for allosteric modulators for the treatment of inflammatory and neurological diseases. Site one is in the intracellular region of the transmembrane helices (TM) 3/TM4, and site two is the G protein-binding site between TM2/TM3/TM6/TM7. Site three is on the lipid interface of TM5/TM6, site four is in the intracellular region of TM1/TM7, and site five is in the extracellular region of TM3/TM4.

Introduction

The adenosine receptors are a class of four G protein-coupled receptors, A₁, A_2A, A_2B, and A₃, whose endogenous ligand is the purine adenosine [1–3]. The A_2A adenosine receptor (A_2AAR) is expressed in leukocytes, such as T cells, monocytes, macrophages, and natural killer cells (NKT) [2–4], and in the brain, specifically the GABAergic neurons in the basal ganglia that are responsible for voluntary movement [4, 5]. The receptor couples to the G_s protein in the periphery and the G_olf protein in the brain [2, 4]. The G protein stimulates the adenylyl cyclase pathway, causing a surge in cAMP, and the activation of the protein kinase A (PKA) signaling cascade [1–3, 5].

In leukocytes, the activation of PKA by the A_2AAR produces anti-inflammatory effects [2, 3]. Furthermore, a knockout of the A_2AAR in mice causes an accumulation of pro-inflammatory cytokines and leads to tissue damage [6]. When the A_2AAR is activated on T cells, the adenylyl cyclase/cAMP/PKA pathway leads to the inhibition of NF-κB, a pro-inflammatory transcription factor [3]. Additionally, the pathway activates anti-inflammatory transcription factors, which leads to the decrease in pro-inflammatory cytokines [2]. Because of its role in the immune response, A_2AAR agonists are being developed and studied for the treatment of sepsis, which is caused by hyper-inflammation and characterized by multiple organ failure, and multiple sclerosis, where leukocytes attack the myelin on neurons and cause lesions in the brain [1, 2, 7]. Moreover, the anti-inflammatory role of the A_2AAR in the immune system is protective to tumors, suggesting that antagonists of the receptor can be developed to treat cancer [8].

Antagonists of the A_2AAR is a treatment option for Parkinson’s disease. Parkinson’s disease is a motor disorder caused by the degradation of the dopamine pathway. In the GABAergic neurons, the activation of the A_2AAR acts counter to the D₂ dopamine signaling pathway, and suppresses dopamine release [9].

Overall, A_2AAR antagonists can treat Parkinson’s disease [9] and cancer [8]. A_2AAR agonists can treat sepsis [2] and inflammation [1, 7]. Allosteric modulators bind to sites spatially different from the orthosteric binding site [10, 11]. Because allosteric regions are not as conserved as the orthosteric ones across receptor subtypes, modulators that bind to non-orthosteric sites can maintain subtype specificity. Allosteric modulators alter the association or disassociation rate of the endogenous ligand [10]. They include positive allosteric modulators (PAM), which increase the effect of the orthosteric ligand and negative allosteric modulators (NAM), which decrease the effect of the orthosteric ligand [10].

The structure of the A_2AAR consists of 7 transmembrane helices (TM1-TM7) connected by three intracellular loops (ICL1-ICL3) and three extracellular loops (ECL1-ECL3) [12–21]. The receptor has been crystallized in complex with agonists, including adenosine [12–15], antagonists [16–21], and coupled to a G protein fragment [15]. The A_2AAR has also been crystallized with interacting lipids [13, 14, 16, 19–21] and sodium ions [20, 21].

Recently, we applied molecular dynamics (MD) to the ligand–free (apo) form of two conformations of the A_2AAR X-ray structures, agonist-bound (PDB ID: 3QAK) [13] and antagonist-bound (PDB ID: 3EML) [16]. We uncovered the deactivation process of the A_2AAR from the active conformation of 3QAK to the inactive crystal structure 3EML [22]. We identified four distinct conformations of the receptor, the active, intermediate 1, intermediate 2, and inactive, which were characterized particularly by the motions of the Y288^7.53 residue in the NPxxY motif. The number ‘288’ denotes the residue number, ‘7’ denotes the transmembrane helix number, and ‘53’ represents the location of the residue in the transmembrane helix relative to the most conserved residue (denoted 50) [23]. In the active conformation, the χ₁ angle of Y288^7.53 remained in the trans conformation and the side-chain faced the intracellular region of the receptor. In the intermediate 1 conformation, Y288^7.53 interacted with residues in TM2, TM6, and TM7. In the intermediate 2 conformation, the χ₁ angle of Y288^7.53 switched from trans to gauche conformation, causing a break in the interactions between TM2 and TM6. Finally, in the simulation of the inactive A_2AAR, a salt bridge was formed between the D101^3.50 and E228^6.30 residues, or the ‘ionic-lock’ [22].

In this study, we obtain representative conformations of the A_2AAR from the previous MD simulations. We identify non-orthosteric sites on the active, intermediate 1, intermediate 2, and inactive conformers, and 20 A_2AAR crystal structures using FTMap [24]. FTMap is fragment-based mapping algorithm, which screens the surface of the receptor with small probes, such as ethanol and isopropanol, and identifies hot spots based on regions of the receptor where multiple probes bind at low energy. This tool was previously used to identify five non-orthosteric sites on the β₂ adrenergic receptor (β₂AR), four non-orthosteric sites on the β₁ adrenergic receptor (β₁AR) [25], and seven non-orthosteric sites on the M₂ muscarinic receptor [26]. These receptors are all membrane-bound proteins and the sites identified spanned the extracellular, intracellular and lipid interface regions. We classify two non-orthosteric sites on the active conformations, three on the intermediate 1 and intermediate 2 conformations, and five non-orthosteric sites on the inactive conformations of the A_2AAR.

Materials and Methods

Simulations

In a previous study [22], we performed two simulations on the ligand-free A_2AAR starting from the agonist-bound (PDB ID: 3QAK) [13] and the antagonist-bound (PDB ID:3EML) [16] X-ray structures. 3QAK was crystallized in complex with the agonist UK-432097 at a resolution of 2.71 Å [13]. 3EML was crystallized in complex with the antagonist ZM241385 at a resolution of 2.6 Å [16]. The ligands were removed from each crystal structure. In this study, the term “agonist-bound apo” denotes the ligand-free simulation starting from the 3QAK X-ray structure. The term “antagonist-bound apo” denotes the ligand-free simulation starting from the 3EML X-ray structure. The agonist-bound apo and antagonist-bound apo simulations were performed using all-atom MD simulations on the Anton supercomputer [27] for 1.57 μs and 1.75 μs, respectively, as previously described in Caliman, AD, et al [22].

FTMap

FTMap is a fragment-based site mapping algorithm that identifies potential binding sites on a receptor. The program docks 16 molecules, ethanol, isopropanol, isobutanol, acetone, acetaldehyde, dimethyl ether, cyclohexane, ethane, acetonitrile, urea, methylamine, phenol, benzaldehyde, benzene, acetamide, and N,N-dimethylformamide, to the surface of the receptor [24]. To use this algorithm, PDB files are uploaded to the online server, http://FTMap.bu.edu. The output of FTMap is a PDB file with the coordinates of the receptor and the probes found in low-energy hot spots. Hot spots are probe clusters distributed different sites of the receptor. The hot spots are identified and ranked by the FTMap protocol [24, 28]. All hot spots identified on the A_2AAR were analyzed.

Clustering of Receptor Conformations

In the agonist-bound apo simulation, the receptor transitioned from the active to the intermediate 1 conformation at 250 ns. The receptor transitioned from intermediate 1 to intermediate 2 at 1.25 μs. The antagonist-bound apo simulation remained in the inactive conformation for the entire 1.75 μs [22]. The receptor frames were aligned on the backbone atoms of the transmembrane helices, and RMSD-based clustering of the receptor conformations was performed with a 1.5 Å cutoff to identify representative structures of the active, intermediate 1, intermediate 2, and inactive conformations. There were a total of 16 receptor clusters obtained from the agonist-bound apo simulation and 14 receptor clusters from the antagonist-bound apo simulation. Three receptor clusters were found in the active conformer. Nine clusters were found in the intermediate 1 conformer. Four receptor clusters were found in the intermediate 2 conformer, and 14 receptor clusters were found in the inactive conformer [22]. All 30 receptor clusters were submitted to the FTMap server for analysis.

Crystal Structures

Twenty crystal structures were analyzed using the FTMap server. The antagonist-bound crystal structures included are shown in Table 1 and the agonist-bound crystal structures are shown in Table 2. The crystal structures were prepared with Schrödinger’s Protein Preparation Wizard, where missing side chains, residues, and loops were added [29].

Table 1.

Antagonist-bound X-ray structures used for site-mapping. The PDB ID, ligand name, and resolution, interesting features, and references are shown.

PDB ID	Ligand	Resolution in Å	Interesting Features	Reference
3EML	ZMA	2.6	Lipid Interactions	[16]
3PWH	ZMA	3.3		[17]
3REY	XAC	3.31		[17]
3RFM	Caffeine	3.6		[17]
3UZA	T4G	3.27		[18]
3UZC	T4E	3.34		[18]
3VG9	ZMA	2.7	Lipid Interactions	[19]
4EIY	ZMA	1.8	Lipid Interactions, Sodium	[20]
5IU4	ZMA	1.72	Lipid Interactions, Sodium	[21]
5IU7	6DY	1.9	Lipid Interactions, Sodium	[21]
5IUA	6DX	2.2	Lipid Interactions, Sodium	[21]
5IUB	6DV	2.1	Lipid Interactions, Sodium	[21]
5IU8	6DZ	2	Lipid Interactions, Sodium	[21]

Table 2.

Agonist-bound X-ray structures used for site-mapping. The PDB ID, ligand name, and resolution, interesting features, and references are shown.

PDB ID	Ligand	Resolution in Å	Interesting Features	Reference
2YDO	Adenosine	3.0		[12]
2YDV	NECA-agonist	2.6		[12]
3QAK	UKA-agonist	2.71	Lipid Interactions	[13]
4UG2 –Subunit A	NGI -agonist	2.6	Lipid Interactions, Two Subunits Crystallized	[14]
4UG2 – Subunit B	NGI -agonist	2.6	Lipid Interactions, Two Subunits Crystallized	[14]
5G53 – Subunit A	NECA + G protein	3.4	Two Subunits Crystallized	[15]
5G53 – Subunit B	NECA + G protein	3.4	Two Subunits Crystallized	[15]

Probe Occupancy Analysis

The total number of probes within 5 Å of the Cα atom of each residue was calculated for every receptor cluster and X-ray structure. Because these probes are small, if any atom on the probe was within 5 Å of the Cα atom a residue, it was included in the probe occupancy. The probe occupancy designates the probability that a probe was found within 5 Å of a residue. The probe occupancy was calculated by:

$${\mathit{Occupancy}}_i=\frac{1}{P}{\sum }_{j=1}^{N_c}(P_{ij}\times W_j),i=1,2,\dots \dots \dots ..N_{\mathit{res}}$$ (1)

where P_j is the number of probes within 5 Å of a residue i in receptor cluster j, N_c is the total number of receptor clusters, W_j is the number of simulation frames that receptor cluster j represents, N_res is the total number of residues, and P is the total number of possible probes that could interact with that residue, or

$$P=16\times T,$$ (2)

where 16 is the total number of probes and T is the total number of simulation frames. When the probe occupancy is calculated for all antagonist or agonist-bound crystal structures, the weight, W_j is 1, and T is the number of antagonist or agonist-bound crystal structures, respectively.

The binding site residues plotted in the probe occupancy figures include I66^2.64, V84^3.32, L85^3.33, T88^3.36, Q89^3.37, I92^3.40, L167^ECL2, F168^ECL2, E169^ECL2, M177^5.38, N181^5.42, W246^6.48, L249^6.51, H250^6.52, N253^6.55, T256^6.58, H264, L267^6.32, M270^7.35, Y271^7.36, I274^7.39, S277^7.42, and H278^7.43 (Fig. S1 and Fig. 1).

Figure 1. Probe Occupancy Per Residue during the Molecular Dynamic Simulations.

Probe occupancies for the agonist-bound apo simulation [A], and the antagonist-bound apo simulation [B] are shown. Binding site residues are shown in red, non-orthosteric residues are shown in blue, and black boxes represent the transmembrane regions. ICL1 is located in between TM1 and TM2. ECL1 is located between TM2 and TM3. ICL2 is located between TM3 and TM4. ECL2 is located between TM4 and TM5. ICL3 is located between TM5 and TM6, and ECL3 is located between TM6 and TM7.

Center-of-Mass

If the distance between the center-of-mass for one hot spot was within 5 Å of the center-of-mass of another hot spot, they were considered a part of one larger site, and identified as a non-orthosteric site.

Figures 2–7, Supp. Figure 3, and Supp. Figures 5–7 were generated with VMD [30].

Figure 2. The Hot Spots on the A_2AAR.

The hot spots for probe-binding in the active (green) [A], intermediate 1 (cyan) [B], intermediate 2 (orange) [C], and inactive (red) [D] conformers are shown. The orthosteric site (OS) is shown in gray, site 1 in purple, site 2 in blue, site 3 yellow, site 4 in orange, and site 5 in cyan.

Figure 7. Hot spot 5: ‘The Extracellular Cleft’.

is located between TM3 (gray), and TM4 (orange) helices. The probes are shown as cyan spheres [A]. The key interacting residues are shown as bonds [B], and the probes that bind to this site are shown as cyan bonds.

Results and Discussion

Probe Distribution in X-ray Structures

The overall distribution of probes in a receptor obtained from the FTMap analysis can help us to identify specific residues within the receptor with a propensity for interacting with drug fragments. The probe occupancy for the simulation starting agonist-bound (PDB ID: 3QAK) and simulation starting antagonist-bound (PDB ID: 3EML) X-ray structures were calculated (Fig. S1A and Fig. S1C). Additionally, the probe occupancy for all antagonist-bound and all agonist-bound X-ray structures as listed in Table 1 and Table 2, respectively, were calculated (Fig. S1B and Fig. S1D).

In the simulation starting agonist-bound X-ray structure, probes interact with residues in the TM, ICL1, and ECL2 regions (Fig. S1A). In all agonist-bound X-ray structures, the distribution of the probes to the receptor is identical to the distribution of the simulation starting X-ray structure, but the number of residues in TM1, ICL1, TM3, ECL2, and TM6 within 5 Å of a probe increases (Fig. S1B). In the simulation starting antagonist-bound X-ray structure, probes interact with the TM, ECL1, ECL2, and ICL2 regions of the receptor (Fig. S1C). In the antagonist-bound X-ray structures, there are fewer residues within 5 Å of a probe in TM1, while there are more residues in the ICL1, ECL1, and TM7 regions that are within 5 Å of a probe (Fig. S1D).

Given that there are only slight differences in the probe distribution between the X-ray structures (Fig. S1), and that X-ray structures represent static snapshots of the receptor, an FTMap analysis on MD structures will provide a more complete description of the probe distribution, and reveal transient sites in different conformers of the receptor.

Residues with the Highest Probe Occupancies

In the MD sampled representative structures of the receptor, probes interact primarily with the transmembrane regions, with residues with the highest probe occupancies being found in the TM2, TM3, ECL2, TM5, and TM7 regions in the agonist-bound apo simulation (Fig. 1A), and in the TM2, TM3, TM6, and TM7 regions in the antagonist-bound apo simulation (Fig. 1B). In the X-ray structures, the residues with the highest probe occupancies are A63^2.61, S67^2.65, A97^3.45, L167^ECL2, and F168^ECL2 (Fig. S1 and Table S1). The residues with the highest probe occupancies in both simulations are T88^3.36 and D52^2.50 (Fig. S1, Fig. 1, Table S1, and Table S2). Another residue with high probe occupancies in both simulations is S91^3.39.

Residue T88^3.36 is located in the sodium ion binding site (Fig. S2) [31] and the orthosteric binding site in the X-ray structures of the agonist-bound A_2AAR [12–14]. The probe occupancy surrounding T88^3.36 is higher in the agonist-bound X-ray structures (0.38) compared with the antagonist-bound X-ray structures (0.03) (Table S1). In the agonist-bound apo simulation, the probe occupancy for T88^3.36 is 0.99, and the probe occupancy in the active conformer is 1.0 (Table S1 and Table S2). This is consistent with previous findings that mutating T88^3.36 affects agonist-binding, but not antagonist binding [12, 13]. The probe occupancy in the intermediate 1 conformer is 0.93, 0.87 in the intermediate 2 conformer, and 0.96 in the antagonist-bound apo simulation or the inactive conformer (Table S1 and Table S2). The high probe occupancies found in the intermediate 1, intermediate 2, and inactive conformers compared to the X-ray structures is due to the entrance of a sodium ion into the sodium ion binding pocket during both simulations (Fig. S2).

The probe occupancy of D52^2.50 is 0.86 in the agonist-bound apo simulation and 0.99 in the antagonist-bound apo simulation (Table S1). D52^2.50 is a conserved residue located in the sodium ion-binding site [20, 21, 31]. A sodium ion is not found in either the simulation starting agonist-bound X-ray structure [13] or the simulation starting antagonist-bound X-ray structure [16]. Appropriately, the probe occupancy surrounding D52^2.50 is 0.0 in the simulation starting agonist-bound and antagonist-bound X-ray structures, as well as in most A_2AAR X-ray structures (Table S1). It has been suggested that sodium binding to D^2.50 triggers deactivation of the A_2AAR and other GPCRs [20, 31, 32]. In the agonist-bound apo simulation, a sodium ion enters the site at 405 ns (Fig. S2B), after the receptor has transitioned from the active state to intermediate 1 [22]. Thus, the probe occupancy for D52^2.50 in the active conformer is 0.28 and is 1 in the intermediate 1 conformer (Table S2). In the antagonist-bound apo simulation, a sodium ion enters the site after 34 ns (Fig. S2C).

Residue S91^3.39 interacts with a sodium ion [20, 31], which enters the sodium binding site in both simulations (Fig. S2). S91^3.39 has a probe occupancy of 0.44 in the agonist-bound apo simulation and 0.42 in the antagonist bound apo simulation, compared to 0.0 in the X-ray structures (Table S1). Surprisingly, the probe occupancy is higher in the active conformer (0.85) where a sodium ion does not bind, compared to the intermediate 1 (0.43), intermediate 2 (0.17), and inactive conformers (0.42) (Table S2). This suggests the modulators could bind to some residues of the sodium ion binding site in the active or inactive conformations.

Higher Probe Occupancies Observed in MD Simulations of the A_2AAR

Residues C128^4.49, F93^3.41, and V283^7.48 have higher probe occupancies in the agonist-bound X-ray structures compared to the antagonist- bound X-ray structures (Fig. S1B, Fig. S1D, and Table S1). The TM5, ECL2, and residue S281^7.46 have higher probe occupancies in the agonist-bound apo simulation structures compared to the antagonist-bound apo simulation structures (Fig. 1A and Table S1). TM5 is more flexible in the agonist-bound apo simulation than in the agonist-bound apo simulation [22]. Additionally, during receptor deactivation, residue Y197^5.58 rotates the side chain towards the lipid interface, while in the antagonist-bound apo simulation, Y197^5.58 faces the helical bundle [22]. Residue S281^7.46 is also a part of the sodium ion binding pocket [20, 31] (Fig. S2). S281^7.46 has a probe occupancy of 0.57 in the agonist-bound apo simulation and 0.09 in the antagonist-bound apo simulation, compared to 0.0 in the starting X-ray structures (Table S1). The probe occupancy is 0.63 in the active and intermediate 1 conformers, and 0.18 in the intermediate 2 conformer (Table S2).

TM2, TM3, A81^3.29, and D101^3.49 have higher probe occupancies in the antagonist-bound X-ray structures than in the agonist-bound X-ray structures (Fig. S1D and Fig. S1B). TM4 and TM6 (including residue W246^6.48), and residue L85^3.33 have higher probe occupancies in the antagonist-bound apo simulation, compared to the agonist-bound simulation. W^6.48 is located in the conserved CWxP site, a rotameric trigger for GPCR activation [33], and residue in the sodium ion binding site (Fig. S2A) [31]. W246^6.48 is also a key residue identified in the intrinsic water pathway in A_2AAR [34]. During the antagonist-bound apo simulation, intracellular water molecules enter into the receptor and interact with W246^6.48 (Fig. S4B). W246^6.48 has a probe occupancy of 0.44 in the antagonist-bound apo simulation and 0.11 in the agonist-bound apo simulation (Table S1). The probe occupancy is 0.01 in the active conformer, 0.1 in the intermediate 1 conformer, and 0.14 in the intermediate 2 conformer (Table S2).

The probe occupancy surrounding L85^3.33 is 0.31 in the agonist-bound apo simulation (0.54 in the active, 0.15 in the intermediate 1, and 0.63 in the intermediate 2) and 0.74 in the antagonist-bound apo simulation. L85^3.33 is a residue in the orthosteric binding site in the agonist-bound and antagonist-bound X-ray structures [13, 16]. However, the probe occupancy in the simulation starting agonist-bound X-ray structure is much higher (0.34) than in the simulation starting antagonist-bound (0.0) X-ray structure (Table S1). The probe occupancy for residue L85^3.33 is 0.43 in the all agonist-bound and 0.03 in all antagonist-bound X-ray structures (Table S1).

Hot Spots of Probe Molecules Identified in the A_2AAR during Deactivation

Different conformers of the A_2AAR were identified from the long time-scale molecular dynamics simulations: the active, intermediate 1, intermediate 2, and inactive [22]. FTMap analysis revealed sites and relevant key residues on each of these conformers (Table 3 and Fig. 2). The inactive conformer has more non-orthosteric sites compared with the active conformer (Fig. 2), consistent with findings on the simulation starting agonist and antagonist-bound X-ray structures (Fig. S3).

Table 3.

Non-orthosteric sites in the A_2AAR conformers are listed. The site number, location, transmembrane helices and residues are shown.

Site Number	Location	Regions	Residues
1	Intracellular Crevice	TM3/TM4/TM5	I3.40, F3.41, L3.44, A3.45, D3.49 (TM3) I4.45, I4.48, C4.49 (TM4) Y112 (ICL1) C5.46, P5.50 (TM5)
2	G Protein-Coupling Site	TM2/TM3/TM6/TM7	N39 (ICL1) T2.39, N2.40 (TM2) D3.49, R3.50 (TM3) H6.32, S6.36, F6.44 (TM6) Y7.53 (TM7) I292 (C-Term)
3	The Lipid Interface	TM5/TM6	P5.50, M5.54 (TM5) V6.41, F6.44, W6.48 (TM6)
4	C-Terminus Cleft	TM1/TM7	L1.45, G1.49, L1.52 (TM1) V7.47, P7.50, F7.51 (TM7)
5	Extracellular Cleft	TM3/TM4	C3.30, F3.31, V3.34 (TM3) L4.58, G4.57, F4.54 (TM4)

Each conformer and the X-ray structures have an orthosteric site (Fig. 2 and Fig. S3). Residues in the orthosteric site include I66^2.64, V84^3.32, L85^3.33, T88^3.36, Q89^3.37, I92^3.40, L167^ECL2, F168^ECL2, E169^ECL2, M177^5.38, N181^5.42, W246^6.48, L249^6.51, H250^6.52, N253^6.55, T256^6.58, H264^ECL3, L267^7.32, M270^7.35, Y271^7.36, I274^7.39, S277^7.42, and H278^7.43. This site is smaller in the inactive conformer and the simulation starting antagonist-bound X-ray structure than in the active conformer and simulation starting agonist-bound X-ray structure (Fig. 2A, Fig. 2D, and Fig. S3).

Non-orthosteric site 1 is located in the intracellular crevice between TM3, TM4, and TM5 (Fig. 2 and Fig. 3). Non-orthosteric site 2 is located in the G protein-coupling site of TM2, TM3, TM6, and TM7 (Fig. 2 and Fig. 4). These sites are found in the active, intermediate 1, intermediate 2, and inactive conformers (Fig. 2), as well as the simulation starting agonist-bound and antagonist-bound X-ray structures (Fig. S3). Non-orthosteric site 3 is located on the lipid interface of TM5 and TM6 (Fig. S3 and Fig. 5). This site is found in the intermediate 1, intermediate 2, and inactive conformers (Fig. 2B, Fig. 2C, and Fig. 2D), and in the simulation starting antagonist-bound X-ray structures (Fig. S3B). Non-orthosteric site 4 is located on the intracellular end of TM1 and TM7, near the C-terminus (Fig. 2D and Fig. 6). It is found in the inactive conformer (Fig. 2D), the simulation starting agonist-bound X-ray, and the simulation starting antagonist-bound X-ray structures (Fig. S3). Non-orthosteric site 5 is located between the extracellular ends of TM3 and TM4 (Fig. 2D and Fig. 7). This site is found in the inactive conformer (Fig. 2D).

Figure 3. Hot spot 1: ‘The Intracellular Crevice’.

is located in the intracellular region between TM3 (gray), TM4 (orange) and TM5 (tan). The probes in the active [A] and the intermediate 1 conformers [C] are shown as purple spheres, with the numbers ‘1.1’, ‘1.2’ and ‘1.3’ denoting different regions of the pocket. The key interacting residues for the active clusters [B] and intermediate 1 conformers [D] are shown as bonds, and the probes that bind to this site are shown as purple bonds.

Figure 4. Hot spot 2: ‘The G Protein-Coupling Site’.

is located between TM2 (green), TM3 (gray), TM6 (purple), and TM7 (red). The probes in the active [A] and inactive [C] conformers are shown as blue spheres. The key interacting residues for active conformers [B] and inactive conformers [D] are shown as bonds, and the probes that bind to this site are shown as blue bonds.

Figure 5. Hot spot 3: ‘The Lipid interface’.

is located between TM5 (tan) and TM6 (purple). The probes are shown as yellow spheres [A]. The key interacting residues are shown as bonds [B], and the probes that bind to this site are shown as yellow bonds.

Figure 6. Hot spot 4: ‘The C-Terminus Cleft’.

is located between TM1 (blue), TM7 (red), and TM8 (red). The probes are shown as orange spheres [A]. The key interacting residues are shown as bonds [B], and the probes that bind to this site are shown as orange bonds.

The key residues in each of these sites are listed in Table 3.

Site 1: The Intracellular Crevice

The “intracellular crevice” is located in between TM3, TM4, and TM5. The three regions in this site are denoted 1.1, 1.2, and 1.3 (Fig. 3), and correspond to the different probe clusters in the receptor. Site 1.1 is positioned closest to the intracellular region and involves residues Y112^ICL2, D101^3.49 and A97^3.45, and I124^4.45 (Fig. 3A and Fig. 3B). D^3.49 is a conserved residue of the DRY motif and Y112^ICL2 is a key residue in the G protein-coupling site [15]. This region of the intracellular crevice is located in the active, intermediate 2, and inactive conformers (Fig. 2A, Fig. 2C, and Fig. 2D), and the simulation starting antagonist-bound X-ray structure (Fig. S3B). The aromatic and hydrophobic probes that interact with this region are cyclohexane and benzene.

Site 1.2 is present in every structure (Fig. 2 and Fig. S3) and encompasses residues I92^3.40, C185^5.46, and P189^5.50 (Fig. 3D). Hydrophobic cyclohexane, benzene, and phenol, and hydrophilic isobutanol interact with this site. Site 1.3 is only located in the intermediate 1 conformer (Fig. 2B), and comprises of residues I92^3.40, C185^5.46, and P189^5.50 (Fig. 3D). Probes in this region include benzaldehyde, benzene, cyclohexane, acetone, and phenol. Site 1.1 and 1.2 are separated by I124^4.45 and A97^3.45, while site 1.2 and 1.3 are separated by F93^3.41 (Fig. 3B and Fig. 3D).

The intracellular crevice site is also found in the β₁AR, β₂AR, denoted sites 3 and 5 in Ref. [25], and the M₂ muscarinic receptor, denoted site 4 in Ref. [26]. In the M₂ receptor, there are two components to this site, 4.1 and 4.2. Site 4.1, located closest to the intracellular region, can be found in the inactive and intermediate 2 conformers. In ref. [26], the conformers of the M₂ receptor include the inactive, intermediate 1, intermediate 2 and the active, where the intermediate 1 is found between inactive and intermediate 2, and the intermediate 2 is found between intermediate 1 and active. Site 4.2 is identified in the intermediate 1 and active states of the M₂ receptor.

Site 2: G Protein-Coupling Site

The “G protein-coupling site” is located between the intracellular ends of the TM2, TM3, TM6, and TM7 (Fig. 4). Allosteric modulators designed to bind to this site could prevent or enhance the binding of the G protein, attenuating the signaling cascade. In the active conformer, residues encompassing the G protein-coupling site include H230^6.32, S234^6.36, Y288^7.53, R102^3.50, and I292^C-TERM, and the probes that interact with this site are cyclohexane, N, N-dimethylformamide, and phenol (Fig. 4B). In the inactive conformer, the interacting residues are D101^3.49, R102^3.50, T41^2.39, F242^6.44, I287^7.52, Y288^7.53, N42^2.40, and N39^ICL1 (Fig. 4D).

Probes found in this site include methylamine, acetone, cyclohexane, ethanol, urea, benzene, N, N-dimethylformamide, acetamide, and benzaldehyde. The G protein coupling site is smaller in the active conformer and the simulation starting agonist-bound structure (Fig. 2A, Fig. S3A, and Fig. 4A), than in the intermediate 1, intermediate 2 and inactive conformers, and simulation starting antagonist-bound X-ray structure (Fig. 2B, Fig. 2C, Fig. 2D, Fig. S3B, and Fig. 4C). In the antagonist-bound apo simulation, an internal water channel floods the intracellular region of the receptor (Fig. S4B). In the agonist-bound apo simulation, fewer water molecules enter the receptor (Fig. S4A).

The residues in this site are key to deactivation of the A_2AAR [22]. Residue Y288^7.53 is oriented toward the intracellular region of the receptor in the active conformation. In the intermediate 1, Y288^7.53 forms hydrogen bonds with N42^2.40 and S234^6.36. Next, the side chain of Y288^7.53 switches from the gauche to the trans conformation. In the inactive conformation, a salt bridge is formed between R102^3.50 and E228^6.30. Additionally, F242^6.44 is a residue in the sodium ion binding site [35]. R^3.50 is a key residue in the G protein binding site [15].

This G protein-coupling site is found in the β₁AR, β₂AR, denoted site 4 in ref. [25], and the M₂ receptor, denoted site 7 in ref. [26]. In the M₂ receptor, this site is identified only in the active state.

Site 3: The Lipid Interface

The “lipid interface” site is located on the lipid exposed region of TM5 and TM6 (Fig. 5). It is present in the intermediate 1, intermediate 2, and inactive conformers (Fig. 2B, Fig. 2C, and Fig. 2D), and the simulation starting antagonist-bound X-ray structure of the A_2AAR (Fig. S3B). The residues P189^5.50, W246^6.48, F242^6.44, M193^5.54, and V239^6.41 interact with probes, such as acetone, benzene, ethane, acetaldehyde, acetamide, isobutanol, isopropanol, urea, ethanol, dimethyl ether, and acetonitrile (Fig. 5B).

The conformations of P189^5.50 and F242^6.44 determine whether this site is present in the conformers (Fig. 5 and Fig. S5). In the active conformer, the side chains of key residues in this site, P189^5.50 and F242^6.44, face towards each other. The distance between the Cα atoms of these two residues is smaller than 10 Å (Fig. S5B). When the receptor transitions to the intermediate 1 conformation in the agonist-bound apo simulation, the distance between P189^5.50 and F242^6.44 increases to an average of 12 Å (Fig. S5B). This distances increases to 15 Å in the intermediate 2 conformer (Fig. S5B). In the antagonist-bound apo simulation, the distance between these two residues is consistently above 15 Å (Fig. S5C).

While this site is not present in the β₁AR or β₂AR receptors [25], it is present in the inactive M₂ muscarinic receptor, denoted site 2 in ref. [26]. Allosteric modulators designed to bind to site 3 could lock the protein in the inactive or intermediate conformers, by interacting with residues P189^5.50, F242^6.44, and W246^6.48 (the “toggle switch” for GPCR activation) [33].

Site 4: The C-Terminus Cleft and Site 5: Extracellular Cleft

The “C-terminus cleft” is located in the intracellular end of TM1 and TM7 (Fig. 6). It is present in the inactive conformer (Fig. 2D) and the simulation starting antagonist-bound X-ray structure (Fig. S3B). The residues located in the C-terminus cleft include V282^7.47, F286^7.51, P285^7.50, L26^1.52, G23^1.49, L22^1.48, and L19^1.45. These residues interact with benzene, cyclohexane, phenol, N, N-dimethylformamide, acetone, cyclohexane, isobutanol, and benzaldehyde (Fig. 6B).

This cleft is caused by a tilt in TM1 in the inactive conformer relative to the active conformer (Fig. S6). The side chains of residues V282^7.47 and L19^1.45 move towards each other with a distance of 4.9 Å, while the side chain of residues F286^7.51 and L26^1.52 move away from each other with a distance of 10.0 Å, forming a cleft (Fig. S6).

The “extracellular cleft” is located between the extracellular ends of TM3 and TM4 (Fig. 7). It is present in the inactive conformer (Fig. 2D). The residues located in the extracellular cleft site include F83^3.31, C82^3.30, L137^4.58, G136^4.57, F133^4.54, and V86^3.34 (Fig. 7B). This site interacts with acetone, benzene, dimethyl ether, ethane, and phenol.

The extracellular cleft is only present in the inactive conformer of the A_2AAR (Fig. 2D), but in the M₂ receptor, the extracellular cleft is found in intermediate 2 and active conformers, denoted as site 3.2 in ref [26].

Conclusion

The A_2AAR is a GPCR that plays a key role in the immune and neurological systems [3, 9]. In the immune system, the receptor is anti-inflammatory, and activation of the A_2AAR leads to an inhibition of pro-inflammatory cytokines and a secretion of anti-inflammatory cytokines [2, 3, 6]. In the neurological disorders, such as Parkinson’s Disease, A_2AAR activation acts counter to the D₂ dopamine receptor [9]. Given these roles, A_2AAR agonists are being developed for the treatment of immunological diseases, such as sepsis, and A_2AAR antagonists are being developed for the treatment of PD [5].

Allosteric modulators bind to regions outside of the orthosteric site. Some of these modulators do not have activity on their own, but either increase or decrease the activity of the endogenous ligand [10]. In this study, we applied standard FTMap, a fragment mapping technique, to analyze representative receptor structures obtained from previous agonist-bound apo and antagonist-bound apo MD simulations [22] and the 20 X-ray structures of the A_2AAR to identify allosteric sites on the receptor. FTMap identified non-orthosteric sites on this membrane receptor distributed in the lipid interface, the extracellular and intracellular surfaces. Many ligands do not bind to the lipid interface of the receptor, but there are known ligands that can partition into the lipid bilayer [36].

First, we calculated the probe occupancy per residue. In the MD simulation representative clusters and the X-ray structures, probes mostly interacted with residues in the TM regions. Many residues with high probe occupancies, I92^3.40, S281^7.46, T88^3.36, D52^2.50, L85^3.33, and W246^6.48, are involved in sodium binding. Sodium cannot bind to the active state of the receptor. Previous A_2AAR studies have shown that a sodium ion can bind to when the receptor is in complex with an antagonist, but cannot bind when the receptor is bound to an agonist [31]. Additionally, studies of the M₃ muscarinic receptor reveal that sodium binding locks the receptor in the inactive conformation [32]. The propensity of these probes to bind to residues of the sodium ion binding site suggest that negative allosteric modulators could be developed to target this region of the receptor.

In the agonist-bound apo and antagonist-bound apo MD simulations, four key receptor conformers, active, intermediate 1, intermediate 2, and inactive, were identified during deactivation of the A_2AAR [22]. We identified five non-orthosteric sites on the A_2AAR that can be targeted for designing non-orthosteric modulators. Overall, the inactive conformer and antagonist-bound X-ray structures exhibited more non-orthosteric sites than the active conformer and agonist-bound X-ray structures.

The intracellular crevice, present on all conformers, is located between the intracellular ends of TM3/TM4/TM5. This site is also present in the β₁AR, β₂AR, and M₂ muscarinic receptor. Key interacting residues include D101^3.49 of the conserved DRY motif. The G protein-coupling site, present on all conformers, is located in the intracellular mouth of TM2/TM3/TM6/TM7. This site is also present in the β₁AR, β₂AR, and M₂ muscarinic receptor. Key interacting residues include Y288^7.53 of the NPxxY motif, and R102^3.50 of the conserved DRY motif. This site is larger in the inactive conformer than in the active conformer due the intracellular influx of water molecules. Given the role of two key residues involved in GPCR activation, Y288^7.53 and R102^3.50 of the ionic lock, either PAMS or NAMs could be developed to bind this site.

The lipid interface site, present on the intermediate 1, intermediate 2 and inactive conformers, is located on the lipid interface between TM5/TM6. This site is present in the M₂ muscarinic receptor, and includes key residues including W246^6.48. The C-terminus cleft is located in the intracellular end of TM1/TM7. The extracellular cleft is present on the extracellular region of TM3/TM4. These sites are not found in the active conformer, and thus could be targeted for designing novel NAMs.

Overall, these sites provide an array of available non-orthosteric sites on the active, intermediate 1, intermediate 2, and inactive A_2AAR conformers. These sites, including the sites found on the lipid interface, can be screened for novel PAMs for the treatment of sepsis and novel NAMs for the treatment of PD and cancer.

Supplementary Material

Supp info

Table S1. Probe Occupancies of Key Residues. Probe Occupancies of Key Residues

Table S2. Probe Occupancies of Key Residues by Deactivation Conformation.

Figure S1. Probe Occupancy for A_2AAR X-ray Structures.

Figure S2. Sodium Ion Binding Site.

Figure S3. Receptor Hotspots in Starting Structures.

Figure S4. Internal water channel.

Figure S5. Conformation of P189^5.50 and F242^6.44 in Site 3.

Figure S6. Conformation of TM1 and TM7 in Site 4.