Purinergic GPCR transmembrane residues involved in ligand recognition and dimerization

Abstract

We compare the GPCR-ligand interactions and highlight important residues for recognition in purinergic receptors—from both X-ray crystallographic and cryo-EM structures. These include A₁ and A_2A adenosine receptors, and P2Y₁ and P2Y₁₂ receptors that respond to ADP and other nucleotides. These receptors are important drug discovery targets for immune, metabolic and nervous system disorders. In most cases, orthosteric ligands are represented, except for one allosteric P2Y₁ antagonist. This review catalogs the residues and regions that engage in contacts with ligands or with other GPCR protomers in dimeric forms. Residues that are in proximity to bound ligands within purinergic GPCR families are correlated. There is extensive conservation of recognition motifs between adenosine receptors, but the P2Y₁ and P2Y₁₂ receptors are each structurally distinct in their ligand recognition. Identifying common interaction features for ligand recognition within a receptor class that has multiple structures available can aid in the drug discovery process.

1. Introduction

GPCR structural biology has accelerated the discovery of new drug molecules, for example, enabling the de novo design of A_2A adenosine receptor (AR) antagonist AZD4635 (Fig. 1), which is in clinical trials for cancer (Borodovsky et al., 2020; Congreve, de Graaf, Swain, & Tate, 2020). X-ray crystallography has been used to determine most of these structures (Qu, Wang, & Wu, 2020; Wacker, Stevens, & Roth, 2017). In addition, cryogenic electron microscopy (cryo-EM) is increasingly used and has the advantage of more readily including protein partners such as the associated G proteins and detecting multiple conformations of the same protein (Safdari, Pandey, Shukla, & Dutta, 2018). GPCRs contain seven transmembrane helical domains (TMs), but curiously not all 7TM proteins are GPCRs (Vasiliauskaite-Brooks, Healey, & Granier, 2019). In most of the known structures of GPCR-ligand complexes, the recognition of small molecules involves amino acid residues in the TM region facing the centroid within the heptahelical bundle. The canonical, orthosteric ligand binding site of Class A (rhodopsin-like) GPCRs has largely conserved features (Ballesteros and Weinstein, 1995; van Rhee & Jacobson, 1996; Venkatakrishnan et al., 2013). This conserved structure-function aids in predicting the modes of interaction of competitive small molecule ligands, both for cases in which a protein structure is lacking (Jung et al., 2020) and when a known receptor structure is applied to predicting the interactions of different ligand classes (Rodriguez et al., 2016). The correspondence of key TM residues between receptor subtypes and their shared functionality in ligand binding and activation has been analyzed (Lebon et al., 2011; Wacker et al., 2017; Warne, Edwards, Dore, Leslie, & Tate, 2019; Xu et al., 2011) for Class A GPCRs. There is structural information available, as well, for non-class A GPCRs, in which the degree of homology with Class A is low, for example, a recently determined Class D Ste2 receptor dimer (Velazhahan et al., 2021). Key water molecules, H-bonding networks and molecular (micro)switches for activation within the TM region that are shared among various Class A GPCRs have been defined (Filipek, 2019; Venkatakrishnan et al., 2019; White et al., 2018; Zhou et al., 2019). GPCRmd (http://gpcrmd.org/) is an open source online platform for molecular dynamics (MD) simulations and visualization of dynamic ligand and interhelical interactions, conserved structural motifs, water molecules, H-bonding and other interactions in GPCRs (Rodriguez-Espigares et al., 2020).

FIG. 1.

Structures of pharmacological probes for ARs (A) and P2YRs (B) shown in Table 1, including ligands present in reported structures. The structures of adenosine, ADP and ATP are not shown. Theophylline and caffeine (shown under A_2AAR) are nonselective antagonists.

Allosteric small molecule modulators of GPCRs have been shown to bind in nearly all regions of the general structure (Wold, Chen, Cunningham, & Zhou, 2019). Nevertheless, many of the allosteric binding regions that are distal or distinct from the canonical binding site still may involve TM residues. This review does not cover the GPCR intracellular loops (ILs), which interact with the C-terminus of the alpha subunit of the G protein or with β-arrestin (Carpenter & Tate, 2017; Ciancetta, Rubio, Lieberman, & Jacobson, 2019), but it emphasizes the role of TM residues. The extracellular loops (ELs) that have their own effects on GPCR ligand binding and activation (Cao et al., 2018; Marullo et al., 2020), and data for some EL residues are included in this review. EL2, in particular is typically in direct contact with the orthosteric ligands of Class A GPCRs.

GPCR dimerization is also now recognized as being important biologically (García-Recio et al., 2020; Pin, Kniazeff, Prezeau, Liu, & Rondard, 2019). Class C GPCRs form obligatory functional dimers, but there is increasing evidence for functional dimerization and its pharmacological consequences within Class A receptors, as well. It appears that the monomers and dimers may exist in an equilibrium. Thus, we discuss information about the TM regions that are engaged in putative contact during dimerization with other GPCR protomers.

2. AR and P2YR ligands and structures

In this review, we focus on residues of purinergic GPCRs, i.e., ARs and PY receptors (P2YRs), that are involved in ligand and protein recognition. These represent 12 Class A GPCRs (4 ARs and 8 P2YRs), which couple to various G proteins: G_i/_o, G_s or G_q (Table 1). Representative members of these receptor families have been determined structurally by X-ray or cryo-EM methods (Table 2): specifically, A₁AR, A_2AAR, P2Y₁R and P2Y₁₂R (all human). In particular, the A_2AAR has served as a model system for structural probing in many studies. An abundance of ligands has been discovered for these receptors, which have important drug discovery potential (Jacobson et al., 2021). These receptors are important drug discovery targets for immune, inflammatory, cardiovascular, metabolic and nervous system disorders. Some of these orthosteric ligands have been tested in clinical trials or are being developed for mainly chronic, but also some acute conditions, such as stroke (Pedata et al., 2014). Thus, it is valuable for future drug discovery efforts to analyze the commonality between ARs and within the two subfamilies of the P2YRs (Attah et al., 2020; Ciancetta & Jacobson, 2018; Gutierrez-de-Teran, Sallander, & Sotelo, 2017; Jespers et al., 2017; Jung et al., 2020).

Table 1.

G protein coupling of ARs and P2YRs and their ligands used as pharmacological probes (Jacobson et al., 2020; Jacobson, IJzerman, & Müller, 2021).

Receptor family	Subtype (agonist)		G protein	Synthetic agonists; antagonists (Ki in nM at human receptor, unless noted)
ARs	A1		Gi, Go	Cl-ENBA (0.51), CPA (2.3), MRS7469 (2.14), SPA (7.92); DPCPX (3.9), PSB-36 (0.7)
	A2A		Gs, Golf	CGS21680 (27), UK432097 (4), PSB-0777 (360); SCH442416 (4.1)
	A2B		Gs, Gq	Bay60–6583 (114); MRS1754 (1.97), PSB-603 (0.55)
	A3		Gi	IB-MECA (1.8), Cl-IB-MECA (1.4), MRS5980 (0.70), MRS5698 (3.5); MRS1523 (19), MRS1191 (31)
P2YRs	P2Y1-like	P2Y1 (ADP)	Gq	MRS2365 (0.40); 2-MeS-ADP (2510)a; MRS2179 (331), MRS2500 (0.79), BPTU (6.0)
		P2Y2 (ATP, UTP)	Gq, Gi	2-thio-UTP (35), PBS-1114 (135); AR-C118925XX (58)
		P2Y4 (UTP)	Gq, Gi	MRS4062 (23); PSB-16133 (234), PSB-1699 (407)
		P2Y6 (UDP)	Gq	MRS4383 (493)b, PSB-0474 (71), 5-OMe-UDP (79); MRS2578 (372)b, TIM-38 (4300)b
		P2Y11 (ATP)	Gq, Gs	NF546 (537); NF157 (45)c
	P2Y12-like	P2Y12 (ADP)	Gi	2-MeS-ADP (5)a; cangrelor (0.40), ticagrelor (13), AZD1283 (31.6), PSB-0739 (0.16)
		P2Y13 (ADP)	Gi	2-MeS-ADP (19)a; MRS2211 (1620)
		P2Y14 (UDPG, UDP)	Gi	MRS2905 (0.92), MRS2690 (49); PPTN (0.44), MRS4625 (27.6)b

Adenosine is the principal native agonist for the AR subtypes, while the more diverse endogenous P2YR ligands are indicated above. P2YR affinities are found in Jacobson et al. (2020). Structures of AR ligands are shown in Fig. 1.

^aActivates P2Y₁, P2Y₁₂ and P2Y₁₃ receptors.

^bIC₅₀ value.

^cAntagonist of P2Y₁ and P2Y₁₁ receptors.

Table 2.

Reported AR and P2YR experimental structures (in the Protein Data Bank (PDB)) and their characteristics.

Receptor	Activation state	Ligand	PDB ID (References)
ARs
A1AR	Inactive	PSB36	5N2S (Cheng et al., 2017)
		DU172 (covalent)	5UEN (Glukhova et al., 2017)
	Active (Gi)	Adenosinea	6D9H (Draper-Joyce et al., 2018)
A2AAR	Inactive	ZM241384	3EML (Jaakola et al., 2008), 3PWH (Doré et al., 2011), 3VG9 & 3VGA (Hino et al., 2012), 5UVI (Martin-Garcia et al., 2017), 5JTB (Melnikov et al., 2017), 6AQF (Eddy et al., 2018), 6MH8 (Martin-Garcia et al., 2019), 6S0L & 6S0Q (Nass et al., 2020)
		ZM241384, Na+ allosteric mod	4EIY (Liu et al., 2012), 5IU4 (Segala et al., 2016), 5K2A & 5K2B & 5K2C & 5K2D (Batyuk et al., 2016), 5NLX & 5NM2 & 5NM4 (Weinert et al., 2017), 5VRA (Broecker et al., 2018), 5OLG (Rucktooa et al., 2018), 6JZH (Shimazu et al., 2019), 6LPJ & 6LPK & 6LPL (Ihara et al., 2020), 6PS7 (Ishchenko et al., 2019), 6WQA (Lee et al., 2020)
		12b, Na+ allosteric mod	5IUA (Segala et al., 2016)
		12x, Na+ allosteric mod	5IUB (Segala etal., 2016)
		12c, Na+ allosteric mod	5IU7 (Segala et al., 2016)
		12f, Na+ allosteric mod	5IU8 (Segala et al., 2016)
		XAC	3REY (Doré et al., 2011)
		Caffeine	3RFM (Doré et al., 2011)
		Caffeine, Na+ allosteric mod	5MZP (Cheng et al., 2017)
		Theophylline, Na+ allosteric mod	5MZJ (Cheng et al., 2017)
		PSB36, Na+ allosteric mod	5N2R (Cheng et al., 2017)
		T4G	3UZA (Congreve et al., 2012)
		T4E	3UZC (Congreve et al., 2012)
		T4E, Na+ allosteric mod	5OLZ & 5OM1 & 5OM4 (Rucktooa et al., 2018)
		AZD4635, Na+ allosteric mod	6GT3 (Borodovsky et al., 2020)
		Tozadenant, Na+ allosteric mod	5OLO (Rucktooa et al., 2018)
		Cmpd-1	5UIG (Sun et al., 2017)
		LUAA47070, Na+ allosteric mod	5OLV (Rucktooa et al., 2018)
		Vipadenant, Na+ allosteric mod	5OLH (Rucktooa et al., 2018)
		Chromone 4d, Na+ allosteric mod	6ZDR (Jespers et al., 2020)
		Chromone 5d, Na+ allosteric mod	6ZDR (Jespers et al., 2020)
	Intermediate	UK432097	3QAK (Xu et al., 2011), 5WF5 & 5WF6 (White et al., 2018)
		Adenosine	2YDO (Lebon et al., 2011)
		NECA	2YDV (Lebon et al., 2011)
		CGS21680	4UG2 & 4UHR (Lebon, Edwards, Leslie, & Tate, 2015)
	Active (bound to mini-Gs)	NECA	5G53 (Carpenter, Nehmé, Warne, Leslie, & Tate, 2016)
		NECAa	6GDG (García-Nafría, Lee, Bai, Carpenter, &Tate, 2018)
P2YRs
P2Y1R	Inactive	MRS2500	4XNV (Zhang et al., 2015)
		BPTU	4XNW (Zhang et al., 2015)
P2Y12R	Inactive	AZD1283	4NTJ (Zhang et al., 2014)
	Intermediate	2-MeSADP	4PXZ (Zhang et al., 2014)
		2-MeSATP	4PY0 (Zhang, Zhang, Gao, Zhang, et al., 2014)

The main technique used to solve the structures was by X-ray crystallography, except where noted.

^aSolved by cryo-EM.

Numerous selective agonist and antagonist ligands (Jacobson et al., 2021) have been reported for the ARs (Fig. 1A) and P2YRs (Fig. 1B), and receptor complexes have been determined structurally for some of these agonists/antagonists, including both subtype selective and nonselective ligands (Table 2). In the case of ARs, three-dimensional structures have been solved for two receptor subtypes with either agonists or antagonists: A₁AR and A_2AAR. Ligand binding in the orthosteric binding site has many parallels between the A₁ and A_2AARs structures and with the theoretical models of A_2B and A₃ARs (Kose et al., 2018; Tosh et al., 2020; Tosh et al., 2020). There are also many ligand-interacting amino acid residues in common between nucleoside agonists and diverse, non-nucleoside antagonists of the ARs. Although most AR agonists are adenosine derivatives, pyridine-3,5-dicarbonitrile agonists and partial agonists have been extensively explored (Dal Ben et al., 2019). Structural information on P2YRs is also available for two receptor subtypes, P2Y₁R and P2Y₁₂R, with only antagonists in the former example, but agonists and antagonists are represented in the latter. The parallelism that can be observed in the case of ARs, is missing in the case of P2YRs, where more variability has been observed among three-dimensional structures.

We have categorized GPCR-ligand interactions, i.e., identified important residues needed to coordinate the ligands—either from X-ray crystallographic or cryo-EM structures—for G_i/o-coupled A₁ and G_s-coupled A_2A receptors (Glukhova et al., 2017; Jaakola et al., 2008; Lebon et al., 2011; Xu et al., 2011), G_q-coupled P2Y₁ and G_i-coupled P2Y₁₂ receptors (Zhang et al., 2015; Zhang, Zhang, Gao, Paoletta, et al., 2014; Zhang, Zhang, Gao, Zhang, et al., 2014). Often, there is site-directed mutagenesis (SDM) data gathered in multiple studies to support these structures. We have investigated the tridimensional structures deposited in the Protein Data Bank (PDB) (rcsb.org; Berman et al., 2000), analyzed the contacts between ligands and receptor residues and collected the results in bar plots (Fig. 2) enabling the highlighting of most interacting residues. An automatic procedure was employed using a Schrödinger (Schrödinger Release 2020–4: Maestro, Schrödinger, LLC, New York, NY, 2020) command line script to detect ligand-receptor interactions on the PDB complexes, previously prepared with the Protein Preparation Wizard of the Schrödinger suite. Contacts defined as “good contacts” were taken into account, described by the distance of two atoms divided by the sum of their van der Waals radii within the range of 1.30 and 0.89. Furthermore, H-bonds, water mediated H-bonds, π–π interactions (face-to-face or edge-to-face), salt bridges and halogen bonds were examined, and these interaction patterns are discussed in the text. Chain A or R of the PDB files and the alternate state with higher occupancy were employed. TM residues are highlighted using the Ballesteros-Weinstein notation (Ballesteros & Weinstein, 1995) in addition to their number in the protein sequence (x·yz, where x is the helix number and yz is the amino acid residue number relative to the most conserved residue in that helix set as 50).

FIG. 2.

The purinergic GPCR structures deposited in the Protein Data Bank (PDB) so far were surveyed to compare the interaction patterns of different ligands bound to ARs and P2YRs. Maestro by Schrödinger (Schrödinger Release 2020–4: Maestro, Schrödinger, LLC, New York, NY, 2020) was employed to prepare (Protein Preparation Wizard) and analyze the PDB structures, and the data were then plotted with matplotlib using python scripts. Contacts defined as “good contacts” in Maestro were taken into account (distance of two atoms divided by the sum of their van der Waals radii within the range of 1.30 and 0.89). Panels (A–D) show residues in contact with the ligand in the complexes deposited in the PDB. The height of the bar plots represents the number of structures where the contact is observed, with a possible maximum height, corresponding to the number of deposited structures, equal to 3, 57, 2, 3 in the case of A₁AR, A_2AAR, P2Y₁R and P2Y₁₂R, respectively (panels A–D). Different colors highlight structures with agonists and antagonists in the case of A₁AR, A_2AAR and P2Y₁₂R (panels A, B and D), while discriminate the orthosteric antagonist MRS2500 from the allosteric BPTU in the case of P2Y₁R (panel C). A focus on the H-bonds between A_2AAR residues and agonists (panel E) or antagonists (panel F) is further provided. Ligand abbreviations: ADN, adenosine; UKA, UK432097; CGS, CGS21680; ZMA, ZM241385.

3. Analysis of small molecule recognition by TM residues of adenosine receptors

3.1. AR structures and models

X-ray and cryo-EM structures have been reported for two AR subtypes, specifically A₁AR and A_2AAR, for a total of 3 structures in the first case and 57 in the second (Table 2). The structures of the two subtypes show high similarity, with major differences confined to the loops. In particular, EL2 has a completely different conformation in the two receptors, and even the same receptor (A_2AAR) with different bound ligands can show a large reorganization of EL2 (Xu et al., 2011). Human ARs have high sequence identity, ranging from a maximum of 46% for the A₁AR-A₃AR pair to a minimum of 30% for the A_2AAR-A₃AR pair (sequence identity data retrieved from GPCRdb; https://gpcrdb.org/similaritymatrix/render; Kooistra et al., 2021). The similarities among the TM structures of the two subtypes, together with their high sequence identity among ARs, makes structural information for A₁AR and A_2AAR useful also to build homology models for A_2BAR and A₃AR.

Experimental structures and models have been used extensively in medicinal chemistry to design new ligands of the same chemical family and to discover novel chemotypes for a given receptor by virtual screening. For example, the discovery of a sub-micromolar A_2AAR antagonist by virtually screening around 40,000 lead-like compounds in the dark chemical matter was accomplished by Carlsson et al. (Ballante et al., 2020). The use of purinergic structure for rational drug design has been recently reviewed (Salmaso & Jacobson, 2020b).

3.1.1. A₁AR structures

Three structures have been reported for A₁AR, one in the agonist-bound active state (PDB ID: 6D9H, bound to adenosine) and two in the antagonist-bound inactive state (PDB IDs: 5N2S and 5UEN, respectively bound to PSB36 and the covalent antagonist DU172), where both antagonists have a xanthine structure (Fig. 3A and C). A total of 25 residues have been found in contact with a ligand in at least one of the three tridimensional structures, and among them 12 residues are common to all the three ligands (Fig. 2A): V87^3.32, L88^3.33, T91^3.36 (mutated into A91 in structure 5N2S), F171^EL2, E172^EL2, M177^5.35, M180^5.38, W247^6.48, L250^6.51, N254^6.55, I274^7.39 and H278^7.43. A contact between adenosine and T277^7.42 distinguishes this agonist from the two antagonists, which instead make additional contacts with L253^6.54, T257^6.58, T270^7.35 (and with I69^2.64, N184^5.42, H251^6.52 and K265^EL3 in the case of PSB36 in structure 5N2S, and with Y12^1.35, A66^2.61, N70^2.65 and Y271^7.36 in the case of DU172 in structure 5UEN).

FIG. 3.

(A) and (C) Experimental binding mode of the agonist adenosine (orange sticks) and of the covalent antagonist DU172 (green) to A₁AR (dark and light gray when bound to agonist and antagonist, in structures 6D9H and 5UEN, respectively). (B) and (D) Experimental binding mode of the agonist adenosine (orange sticks) and of the antagonist ZM241385 (green) to A_2AAR (dark and light gray when bound to agonist and antagonist, in structures 2YDO and 4EIY, respectively). Relevant receptor residues discussed in the text are shown by sticks.

AR-conserved N254^6.55 is engaged in a bidentate H-bond with the exocyclic amine at position 6, N⁶H, and with the position 7 of the adenine scaffold in the case of adenosine (agonist), and with the carbonyl at position 6 and the NH at position 7 of the xanthine scaffold of the antagonists. The aromatic base scaffolds are furthermore involved in a π–π stacking interaction with F171^EL2. Adenosine is involved in additional H-bonds with T91^3.36 through the 5′-hydroxy group and H278^7.43 through the 3′-hydroxy group, and is in contact with T277^7.42, whose equivalent position in A_2AAR, i.e., S277^7.42, interacts through H-bonds with agonists. The antagonists are engaged in fewer H-bonds compared to agonists, with only the additional involvement of Y12^1.35 in the case of the covalently bound xanthine DU172, but the cyclohexyl (DU172) and 3-tricyclo[3.3.1.0^3,7]nonanyl (of xanthine derivative PSB36) substitution at position 8 face the extracellular portion of the receptor making hydrophobic contacts with F171^EL2, E172^EL2, M177^5.35, L250^6.51, L253^6.54, T257^6.58 and T270^7.35 (and K265^EL3 in structure 5N2S). Moreover, the long propyl-carbamoyl-benzene-1-sulfonyl substituent at position 3 of DU172, covalently bound to Y271^7.36, makes further interactions with TM1 (residue Y12^1.35) and TM2 (residues A66^2.61 and N70^2.65). However, the shorter 3-hydroxypropyl group of PSB36 engages TM2 only through I69^2.64. Differently, PSB36 presents a butyl substituent at position 1 which is bulkier than the propyl substituent of DU172, and probes the orthosteric pocket more deeply, entering in contact with N184^5.42 and H251^6.52.

3.1.2. A_2AAR structures

In the case of A_2AAR, 57 experimental structures have been deposited in the PDB so far. An ensemble of 38 residues appears in contact with a ligand in at least one PDB structure, and among them, 10 residues are common to all (i.e., F168^EL2, L249^6.51, N253^6.55, M270^7.35, I274^7.39) or nearly all (~90% or more) (i.e., L85^3.33, E169^EL2, M177^5.38, W246^6.48, H250^6.52) structures. Thus, these 10 residues define the orthosteric binding pocket of A_2AAR regardless of the agonist or antagonist character of the ligand (Fig. 2B). Moreover, seven residues are only in proximity to antagonists (i.e., Y9^1.35, I64^2.62, I80^3.28, A81^3.29, K153^EL2, A265^EL3, P266^7.31). However, they interact in a small fraction of all the PDB structures (less than 10%), so they cannot be considered a fingerprint for this category of compounds. Four residues are only in proximity to agonists (i.e., Q89^3.37, which is mutated into A89 in some structures, I92^3.40, C185^5.46, and V186^5.47). These residues define a pocket between TM3 and TM5 that is the deepest region of the orthosteric site, which typically hosts the ethyl-amido group at position 5′ of most of the agonists, but it is not explored by any of the antagonists.

In 9 PDB structures, the A_2AAR is bound to an agonist, in particular to adenosine (PDB ID: 2YDO), NECA (PDB IDs: 2YDV, 5G53 and 6GDG), CGS21680 (PDB IDs: 4UG2, 4UHR) and UK432097 (PDB IDs: 3QAK, 5WF5, 5WF6). Among these structures, 5G53 and 6GDG are in the active state, with the receptor bound to an engineered G protein, mini-Gs (truncated Gα_s) alone or in a heterotrimeric state, respectively. The agonists share a similar binding mode (Fig. 3B), where the ligands are surrounded by V84^3.32, L85^3.33, T88^3.36, F168^EL2, M177^5.38, W246^6.48, L249^6.51, N253^6.55, M270^7.35, I274^7.39, S277^7.42, H278^7.43 in all the structures, and by E169^EL2, Q89^3.37 (mutated into A89 in some structures), N181^5.42, H250^6.52 in nearly all (~90%) of the structures (Fig. 2B). Particularly, agonists are involved in a π–π stacking interaction between F168^EL2 and the adenine base, in a bidentate H-bond between N253^6.55 and the endocyclic N7 position of the adenine base and the exocyclic N⁶H, which interacts also with E169^EL2 (missing only in structure 6GDG), in a H-bond between S277^7.42 and the 3′-hydroxyl group and between H278^7.43 and the 2′-hydroxyl group (Fig. 2E).

Few differences can be observed among the agonists: T88^3.36 makes a H-bond with all the ligands’ 5′-amido substituent, while adenosine’s 5′-hydroxyl group is involved in a water mediated H-bond with N181^5.42 and H250^6.52 (PDB ID: 2YDO), with H250^6.52 interacting instead with the carbonyl of the 5′-amido substituent of NECA, CGS21680 and UK432097 in some of the reported structures. E169^EL2, involved in a salt bridge with H264^EL3, interacts through a H-bond with the unsubstituted N⁶H moiety of adenosine, NECA (not in 6GDG) and CGS21680. However, in the case of UK432097, the E169^EL2–H264^EL3 salt bridge is broken, and E169^EL2 makes a bidentate H-bond with the ureidic group of the substituent at position 2, interacting also with Y271^7.36. The ethylamide substituent at position 5′ distinguishes the other agonists from adenosine, and enters the binding pocket more deeply, making contacts with Q89^3.37 (mutated into A89 in some structures), I92^3.40, C185^5.46, and V186^5.47 in different structures. As compared to adenosine and NECA, agonists CGS21680 and UK432097 bear a long substituent at position 2, which fills the orthosteric binding pocket entering in contact with residues of TM2 (A63^2.61, I66^2.64 and S67^2.65), EL2 (L167^EL2), EL3 (H264^EL3) and TM7 (L267^7.32 and Y271^7.36) in some of the PDB structures. In addition, UK432097’s bulky N⁶-2, 2-diphenylethyl substituent also engages the tip of TM6 (I252^6.54 and T256^6.58). Several conserved water molecules are noted to surround the AR agonists. For example, the stabilizing contribution of the conserved water bridging the adenine nitrogen at position 3 and the 2′-hydroxyl has been explored using a combination of free energy perturbation (FEP) and chemical synthesis (Matricon et al., 2021). The penalty for driving this water out of the binding site is a dramatic reduction of affinity at all AR subtypes.

Forty-eight structures of antagonist-bound A_2AARs are present in the PDB, with [1,2,4]triazolo[2,3-a][1,3,5]triazine derivative ZM241385 in 27 cases (PDB IDs: 3EML, 3PWH, 3VG9, 3VGA, 4EIY, 5IU4, 5JTB, 5K2A, 5K2B, 5K2C, 5K2D, 5NLX, 5NM2, 5NM4, 5OLG, 5UVI, 5VRA, 6AQF, 6JZH, 6LPJ, 6LPK, 6LPL, 6MH8, 6PS7, 6S0L, 6S0Q, 6WQA, a ZM241385-related compound in 4 cases (PDB IDs: 5IU7, 5IU8, 5IUA, 5IUB), a xanthine derivative in 5 cases (PDB IDs: 3REY, 3RFM, 5MZJ, 5MZP, 5N2R), a 1,2,4-triazin-3-amine derivative, with an aryl substituent at position 5 and 6, in 6 cases (PDB IDs: 3UZA, 3UZC, 5OLZ, 5OM1, 5OM4, 6GT3), vipadenant (PDB ID: 5OLH), tozadenant (PDB ID: 5OLO), LUAA47070 (PDB ID: 5OLV), a triazole-carboximidamide, Compound-1 (PDB ID: 5UIG), chromone 14 (4d) (PDB ID: 6ZDR), chromone 5d (PDB ID: 6ZDV) in one case each. Nearly all the antagonists are in contact (in at least 90% of the PDB structures) with the pattern of 10 residues cited at the beginning of the paragraph, comprising: L85^3.33, F168^EL2, E169^EL2, M177^5.38, W246^6.48, L249^6.51, H250^6.52, N253^6.55, M270^7.35 and I274^7.39 (Fig. 2B). N253^6.55 is involved in a direct H-bond with the ligand in all the structures, and a π–π stacking interaction with F168^EL2 is often observed as well (Fig. 3D). A π–π edge-to-face interaction with H250^6.52 is made by several ligands (ZM241385 and ZM241385-related compounds, 1,2,4-triazin-3-amine derivatives, vipadenant and LUAA47070), and a H-bond with E169^EL2 is encountered in most complexes with ZM241385, ZM241385-related compounds and 1,2,4-triazin-3-amine derivatives, and in the case of vipadenant. However, some ligand-specific interactions are observed for these antagonists.

In the case of ZM241385 and ZM241385-related compounds, the [1,2,4]triazolo [1,5-a][1,3,5]triazine aromatic scaffold participates in the π–π face-to-face interaction with F168^EL2, and N1 (or sometimes the furan oxygen atom) and the 7-amino group interacts with N253^6.55 through bidentate H-bonding (even if with a deviation from ideal geometry in the case of structures 3VG9 and 3VGA) (Fig. 2F). In addition, a π–π edge-to-face interaction is observed between the furan ring and H250^6.52, and a H-bond is formed between the 7-amino group and E169^EL2 (missing only in structures 3PWH, 3VG9, 3VGA). A water molecule bridges a H-bond between Y271^7.36 and the 5-amino group in the 75% of the PDB structures, and often co-crystalized water molecules interact also with nitrogen atoms at positions 3 and 4 of the ligand. Moreover, in the case of ZM241385-related compounds bearing a piperidine or piperazine moiety in the 5-substituent, a salt bridge with E169^EL2 can be observed.

X-ray structures for A_2AAR in complex with four different xanthine antagonists have been reported so far, in particular XAC (PDB ID: 3REY), caffeine (PDB IDs: 3RFM, 5MZP), theophylline (PDB ID: 5MZJ), PSB36 (PDB ID: 5N2R). Caffeine and theophylline are naturally-occurring prototypical AR antagonists, and antagonist XAC was introduced in 1985 as a high affinity, functionalized congener for coupling to carrier or reporter moieties (Jacobson, 2009). The extended position 8 ethylamino chain of XAC was later confirmed in an A_2AAR structure to be reaching toward the extracellular region, thus explaining its ability to be acylated without preventing receptor binding. Xanthine bases are involved in the π–π face-to-face interaction with F168^EL2, while the H-bond interaction with N253^6.55 is fulfilled by the carbonyl at position 6 (maintained in both the alternate states of caffeine observed in structure 5MZP), accompanied by the free NH at position 7 in theophylline and PSB36. In order to maintain a H-bond between N253^6.55 and the carbonyl of XAC in structure 3REY, N253^6.55 is flipped as compared to its common conformation. Moreover, a water molecule mediates an interaction between carbonyl at position 2 of theophylline, caffeine and PSB36 with A81^3.29 in structures 5MZJ, 5MZP and 5N2R, respectively.

In the case of the 1,2,4-triazin-3-amine A_2AAR antagonists such as T4E (PDB ID: 3UZC, 5OLZ, 5OM1, 5OM4), T4G (PDB ID: 3UZA) and AZD4635 (PDB ID: 6GT3), the triazine moiety participates in the π–π interaction with F168^EL2. This interaction is face-to-face in most cases, apart from structure 3UZA where it is closer to an edge-to-face interaction, and in structure 3UZC where the conformation of F168^EL2 does not permit an ideal π–π interaction. The exocyclic 3-amino group and the endocyclic N4 are involved in the typical bidentate H-bond with N253^6.55 (conserved across all ARs), and the exocyclic 3-amino group also interacts with E169^EL2 in all structures except 3UZA and 3UZC, where the salt bridge between E169^EL2 and H264^EL3 is not formed (Fig. 2F). The 5-phenyl group makes a π–π edge-to-face interaction with H250^6.52, while the 6-aryl group, consisting in a phenol in the case of compound T4E (ODB IDs: 3UZC, 5OLZ, 5OM1, 5OM4), interacts though its p-hydroxyl group with H278^7.43, and with A59^2.57 through a water mediated H-bond (not present in structure 3UZC).

Triazolo[4,5-d]pyrimidin-5-amine derivative vipadenant (PDB ID: 5OLH) interacts through its [1,2,3]triazolo[4,5-d]pyrimidine structure with F168^EL2 and through its 5-amine and N⁶H with N253^2.55, making the typical π–π stacking interaction and bidentate H-bond. The furan ring attached to the position 7 makes a π–π edge-to-face interaction with H250^6.52 and an H-bond with A81^3.29.

In the case of tozadenant (PDB ID: 5OLO), the typical π–π interaction with F168^EL2 is granted by the benzothiazole structure, which contributes through its nitrogen atom to the bidentate hydrogen with N253^6.55, together with the exocyclic nitrogen attached to position 2 of the scaffold. T256^6.58 is engaged in an additional H-bond with the hydroxyl group at the position 4 of the piperidine moiety.

Thiazole derivative LUAA47070 (PDB ID: 5OLV) lacks the π–π stacking with F168^EL2, even if its benzamide structure assumes a position not far from enabling that interaction. The amide nitrogen makes the H-bond with N253^6.55, accompanied by the nitrogen of the attached thiazole group, which is also engaged in the π–π edge-to-face interaction with H250^6.52.

In the case of structure 5UIG (Sun et al., 2017), the interactions of antagonist Compound-1 with F168^EL2 and N253^6.55 are present with the 5-amino-1,2,3-triazole scaffold; while in the case of chromones 4d and 5d in structures 6ZDR and 6ZDV, the π–π face-to-face interaction with F168^EL2 is accomplished by the chromone group, while N253^6.55 makes a single H-bond with the nitrogen of the thiazole ring (Jespers et al., 2020).

3.1.3. Examples of AR models

The overall sequence identity between A₃AR and A₁AR is 46%, and it goes up to 56% focusing on TM regions; while the overall sequence identity between A₃AR and A_2AAR is 30%, with 47% value for the TM regions (sequence identity data retrieved from GPCRdb; https://gpcrdb.org/similaritymatrix/render; Kooistra et al., 2021). The high sequence similarity, especially of the TM region, among ARs makes A₁AR and A_2AAR good templates for homology modeling of the other AR subtypes. For instance, a hybrid model of A₃AR has been recently built using an A_2AAR structure as template for the TM helices and an A₁AR structure for TM2 (Tosh, Salmaso, Rao, Bitant, et al., 2020; Tosh, Salmaso, Rao, Campbell, et al., 2020). The upper part of TM2 is observed to move outward from the TM bundle in the case of antagonist-bound A₁AR, and this is in agreement with an established hypothesis for A₃AR, where the outwardly displaced TM2 would allow binding of agonists with bulky substituents at position 2. A similar model was previously built using the TM2 of opsin, outward displaced as well (Tosh et al., 2012). An A₃AR homology model enabled understanding of agonism by NECA-like amides vs 5′-CH₂OH derivatives (Dal Ben et al., 2014). These models enable docking of agonists in a conformation analogous to that experimentally assumed by agonists at A₁AR and A_2AAR binding sites and described in the previous paragraphs. In particular, the H-bonds of hydroxyl groups 2′ and 3′, and of the hydroxyl or amidic moiety at 5′, the π–π interaction of the aromatic base, the bidentate H-bonds of the adenine 6-amino group and N7 might occur in the A₃AR with residues at equivalent positions to the other ARs, in particular with H272^7.43, S271^7.42, T94^3.36, F168^EL2 and N250^6.55. An A_2BAR homology model based on a ZM241385-A_2AAR complex (PDB ID: 3EML) was reported (Kose et al., 2018), which features a close similarity to other AR subtypes. The proposed binding mode of a fluorescent A_2BAR antagonist derived from a 3-propylxanthine bore a close resemblance to the xanthine pose in the XAC-A_2AAR complex (PDB ID: 3REY).

4. Analysis of small molecule recognition by TM residues of P2Y receptors

4.1. P2YR structures and models

There are two subfamilies of P2YRs, based on both sequence identity and on G protein coupling. The eight P2Y receptors have much more sequence diversity than the four AR subtypes, with sequence identity values below 20% between human P2Y₁-like and P2Y₁₂-like P2YRs. The X-ray structures of two P2Y receptors (Table 2) have been reported (P2Y₁ and P2Y₁₂), representative of each of the two subfamilies (Zhang et al., 2015; Zhang, Zhang, Gao, Paoletta, et al., 2014; Zhang, Zhang, Gao, Zhang, et al., 2014). These two P2YR subtypes are much less similar to each other than the intrafamily similarity of ARs, with a sequence identity of 19% (sequence identity data retrieved from GPCRdb; https://gpcrdb.org/similaritymatrix/render; Kooistra et al., 2021). Ligand binding in the orthosteric binding sites have identified common residues in the recognition of nucleotide agonists, which are completely different between P2Y₁ and P2Y₁₂Rs. However, several features are in common between the P2Y₁ and P2Y₁₂Rs. There is a predominance of positively charged Lys and Arg residues in the EL regions, which coordinate the negatively charged phosphate moieties of the nucleotide ligands. Also, an unusual disulfide bridge is formed between the N-terminus and EL3 that is common to all P2YRs. This disulfide bridge was first discovered by SDM and led to early modeling suggesting that the ELs are important in the path taken by nucleotide ligands (Moro, Hoffmann, & Jacobson, 1999). Modeling of other P2YR subtypes based on the X-ray structures has been reported (Attah et al., 2020; Jung et al., 2020; Toti et al., 2017).

4.1.1. P2Y₁R structures

Two X-ray structures have been deposited in the PDB so far for P2Y₁R. The receptor is co-crystallized with an orthosteric antagonist, MRS2500 (PDB ID: 4XNW), in one structure, and to one allosteric antagonist, BPTU (PDB ID: 4XNV), in the other, and the two ligands bind to two different binding pockets, with no common residues (Fig. 4A and C). MRS2500 is a nucleotide derivative, but its ribose ring has been sterically constrained by a bicyclic ring system (methanocarba) in a receptor-preferred North conformation, unlike native ribose which is freely twisting between North and South conformations. 18 residues in P2Y₁R appear in contact only with the nucleotide antagonist, while 11 residues are in proximity only to the allosteric antagonist, bound on the outer surface in contact with the phospholipid bilayer (Fig. 2C).

FIG. 4.

(A) and (C) Experimental binding mode of the orthosteric antagonist MRS2500 (orange sticks) and of the negative allosteric modulator BPTU (green) to P2Y₁R (dark and light gray when bound to MRS2500 and BPTU, in structure 4XNW and 4XNV, respectively). (B) and (D) Experimental binding mode of the agonist 2MeSATP (orange sticks) and of the antagonist AZD1283 (green) to P2Y₁₂R (dark and light gray when bound to agonist and antagonist, in structures 4PY0 and 4NTJ, respectively). Relevant receptor residues discussed in the text are shown by sticks.

Bicyclic nucleotide derivative MRS2500 binds the receptor close to the extracellular region, at the interface among the N-terminus, extracellular loops and tip of TM6 and TM7. This antagonist is surrounded by C42^N-ter, L44^N-ter and K46^N-ter of the N-terminus, Y110^2.63, R195^EL2, T201^EL2, Y203^EL2, D204^EL2 and T205^EL2 of EL2, N283^6.58, A286^6.61, R287^6.62, Q291^EL3 of EL3, N299^7.28, V302^7.31, Y303^7.32, Y306^7.35, R310^7.39 (Fig. 2C). Among these residues, N283^6.58 interacts though a bidentate H-bond with the endocyclic N7 and exocyclic N⁶H positions of the adenine scaffold, D204^EL2, T205^EL2, Y306^7.35 and R310^7.39 are involved in H-bonds with the 5′-phosphate group, which interacts also with R287^6.62 though a water mediated H-bond, K46^N-ter, R195^EL2, T201^EL2 and Y303^7.32 are involved in H-bonds with the 3′-phosphate group, with K46^N-ter and R195^EL2 contributing also through salt bridges (Fig. 4A). In addition, Y303^7.32 makes an edge-to-face π–π interaction with the adenine aromatic scaffold, and the C42^N-ter carbonyl is involved in a halogen bond (at a ~180 degree angle) with the iodine atom at position 2 of the base. Differently, BPTU binds at the interface between TM1, TM2 and TM3 on the membrane side, out of the TM bundle, in a binding site defined by F62^1.43, F66^1.47, L102^2.55, T103^2.56, P105^2.58, A106^2.59, F119^EL2, M123^3.24, L126^3.27, Q127^3.28, I130^3.31 (Fig. 2C), with the backbone carbonyl of L102^2.55 interacting through a H-bond with the ureidic potion of the ligand, and F62^1.43 and F119^EL2 through an edge-to-face π–π interaction with the phenyl and pyridine substituents of the urea group of BPTU (Fig. 4C).

4.1.2. P2Y₁₂R structures

In the case of the P2Y₁₂R, two agonist-bound (2MeSATP and 2MeSADP, in PDB ID: 4PY0 and 4PXZ, respectively) and one non-nucleotide antagonist bound (AZD1283, PDB ID: 4NTJ) X-ray structures are reported. The three ligands bind the receptor in a similar region, sharing the contacts with residues V102^3.30, Y105^3.33, F106^3.34, Y109^3.37, M152^4.53, L155^4.56, S156^4.57, N159^4.60, H187^5.36, V190^5.39, N191^5.40, C194^5.43, R256^6.55, Y259^6.58, and K280^7.35. Thus, 15 residues in P2Y₁₂R appear in proximity of both agonist and antagonist ligands, while 6 residues are in contact only with the antagonist, and 12 are in contact only with agonists (Fig. 2D). However, it is to be noted that the ELs in the P2Y₁₂R structures are mostly undefined in the antagonist-bound structure but well-defined with nucleotide agonists bound.

Y105^3.33 is involved in a π–π stacking interaction in all the structures, interacting with the adenine base in the case of the agonists and with the pyridine in the case of the antagonists; R256^6.55 and K280^7.35 contribute through H-bonds with the 5′-phosphate in the case of the agonists and with the sulfonyl-carbamoyl substituent in the case of AZD1283. However, ligand binding of the non-nucleotide antagonist AZD1283 is unlike other purinergic receptors and has no similar structure among other GPCRs. The antagonist-bound structure consistently deviates from the agonist-bound P2Y₁₂R structures: the tips of TM5, TM6 and TM7 bent toward the membrane environment, and the disulfide bond between C175^EL2 and C97^3.25, linking EL2 to TM3, is broken. This is unusual for a GPCR structure, as this same disulfide is considered a conserved structural feature in other Class A receptors. AZD1283 in fact spans the TM bundle, going from a cleft defined by TM3, TM4 and TM5, where the ethyl ester engages N159^4.60 in a H-bond and occupies the same hydrophobic pocket that hosts the 2-methylsulfanyl substituent of 2MeSADP and 2MeSATP in the agonist-bound structures, to the interface between TM7 and TM6, where it is surrounded by T283^7.38, K280^7.35, V279^7.34, L276^7.31, Y259^6.58, R256^6.55, A255^6.54, F252^6.51, stabilizing the ligand’s benzylic group through an edge-to-face π–π interaction (Fig. 4D). Differently, the TM bundle of the agonist-bound structures is more compact, and the two agonists share the same binding mode where the adenine base, involved in the π–π stacking interaction with Y105^3.33, directs the exocyclic amino group toward the bottom of the pocket, where it is engaged in a H-bond with N191^5.40, and the ribose ring toward the extracellular region, where it makes H-bonds with H187^5.36 through the 2′-hydroxyl group and K179^EL2 and C97^3.25 through the 3′-hydroxyl group (Fig. 4B). The α-phosphate interacts through H-bonds with Y105^3.33, Q263^6.62, in addition to the aforementioned R256^6.55 and K280^7.35, which have also a salt bridge character. The β-phosphate interacts via H-bonds with Y259^6.58, K280^7.35 and R93^3.21 (also salt bridge in the case of the last two), while the γ-phosphate of 2MeSATP interacts with R93^3.21, C175^EL2, S176^EL2.

4.1.3. Example of P2YR models

As an example of modeling for P2YRs, we report P2Y₁₄R subtype, whose antagonists may have therapeutic interest in treating inflammatory diseases, asthma, acute kidney injury, etc. P2Y₁₄R share with P2Y₁₂R 44% sequence (44% overall sequence identity, 52% sequence identity of TM regions) (sequence identity data retrieved from GPCRdb; https://gpcrdb.org/similaritymatrix/render; Kooistra et al., 2021), so X-ray structures of the second can be used as templates for homology modeling. Specifically, a model was built using the 2-MeSADP-bound P2Y₁₂R structure as a template and then refined using MD (Salmaso & Jacobson, 2020a) simulations after docking of the known biaryl antagonist PPTN (Junker et al., 2016). This model has been used to drive the design and synthesis of PPTN analogs as P2Y₁₄R antagonists. The suggested binding mode of PPTN and related compounds involves a salt bridge between the ligand’s carboxylate and K77^2.60 and K277^7.35, which behave also as H-bonds donor together with Y102^3.33 (Jung et al., 2020; Yu et al., 2018). Cation-π interactions stabilize the aromatic moieties of the ligands, involving R253^6.55 and R274^7.32. Although the pose of P2Y₁₄R antagonists does not resemble the experimental binding mode of P2Y₁₂R ligands, and even though PPTN is selective for P2Y₁₄R subtype, some of the residues involved in binding are equivalent to those enumerated for P2Y₁₂R in the previous paragraph, i.e., Y102^3.33, K277^7.35 and R253^6.55.

5. Receptor domains involved in dimerization

There is still some controversy about the stoichiometric nature of class A GPCRs (Felce et al., 2017). While class C GPCRs are known to dimerize, there is no consensus in the case of class A receptors. Evidence suggests that various ARs and P2YRs can form putative dimers, either homo- or heterodimers, with other receptor protomers (Borroto-Escuela et al., 2020; Hill, May, Kellam, & Woolard, 2014; Hinz et al., 2018; Navarro et al., 2016; Neumann, Müller, & Namasivayam, 2020). The contact regions for this dimerization have been determined experimentally, in limited cases, and through modeling (Al-Shar’i & Al-Balas, 2019). In some cases, the association between protomers is facilitated by loop regions of the receptors, for example, intracellular loops of the A_2AAR (Fernandez-Duenas et al., 2012), but also a critical role of TM residues is proposed (Johnston, Wang, Provasi, & Filizola, 2012; Townsend-Nicholson, Altwaijry, Potterton, Morao, & Heifetz, 2019). A tool for exploring the location of residues involved in GPCR dimerization based on reported structures is available on the web (DIMERBOW, http://lmc.uab.es/dimerbow/; García-Recio et al., 2020).

An early modeling study of potential A₃AR homodimerization pointed to a likely TM4-TM5 interface (Kim & Jacobson, 2006). A₃AR homodimerization was later demonstrated pharmacologically (May, Bridge, Stoddart, Briddon, & Hill, 2011), and the TM4-TM5 interface was confirmed as an interface also for A_2AAR-D₂R heterodimerization, by means of peptide segments from the sequences of A_2AAR’s TM4 and TM5 which interfered with heterodimerization (Borroto-Escuela et al., 2018). The same interface, which was experimentally observed in the X-ray structure of the oligomeric β₁-adrenergic receptor (Huang, Chen, Zhang, & Huang, 2013), was proposed for homodimerization of A₁AR and A_2AAR, while the same work suggested the TM5-TM6 interface for a A₁AR-A_2AAR heterotetramer (Navarro et al., 2016). Moreover, A₁AR crystallized in dimeric form in one of the X-ray structures (PDB ID: 5UEN), resulting in the burial of hydrophobic areas of TM4 and TM5 (Glukhova et al., 2017). The role of TM5 in A_2AAR dimerization was previously anticipated in a study exploring the self-assembly of TM5 in oligomeric structures, which also found that a mutation of M193 (5.54) alters the dimerization capability of the whole length A_2AAR (Thevenin, Lazarova, Roberts, & Robinson, 2005).

6. Summary

We compare the GPCR-ligand interactions and highlight important residues for recognition—from both X-ray crystallographic or cryo-EM structures for purinergic receptors. These include A₁ and A_2AARs and P2Y₁ and P2Y₁₂Rs that respond to nucleotides. Molecular modeling, including MD simulation, has been successful in predicting ligand interactions across most of this family of 12 purinergic GPCRs (Salmaso & Jacobson, 2020a). This review catalogs the residues of the binding pocket, with specific interest in transmembrane helical (TM) regions, that engage in contacts with ligands or with other GPCR protomers in dimeric forms. In most cases, orthosteric ligands are represented, except for one allosteric P2Y₁ antagonist. There is extensive conservation of recognition motifs within the adenosine receptor, but the P2Y₁ and P2Y₁₂ receptors are each structurally distinct in their ligand recognition.