Molecular modeling of the human P2Y₁₄ receptor: A template for structure-based design of selective agonist ligands

Abstract

The P2Y₁₄ receptor (P2Y₁₄R) is a G_i protein-coupled receptor that is activated by uracil nucleotides UDP and UDP-glucose. The P2Y₁₄R structure has yet to be solved through X-ray crystallography, but the recent agonist-bound crystal structure of the P2Y₁₂R provides a potentially suitable template for its homology modeling for rational structure-based design of selective and high-affinity ligands. In this study, we applied ligand docking and molecular dynamics refinement to a P2Y₁₄R homology model to qualitatively explain structure-activity relationships of previously published synthetic nucleotide analogues and to probe the quality of P2Y₁₄R homology modeling as a template for structure-based design. The P2Y₁₄R model supports the hypothesis of a conserved binding mode of nucleotides in the three P2Y₁₂-like receptors involving functionally conserved residues. We predict phosphate group interactions with R253^6.55, K277^7.35, Y256^6.58 and Q260^6.62 and nucleobase (anti-conformation) π-π stacking with Y102^3.33 and the role of F191^5.42 as a means for selectivity among P2Y₁₂-like receptors. The glucose moiety of UDP-glucose docked in a secondary subpocket at the P2Y₁₄R homology model. Thus, P2Y₁₄R homology modeling may allow detailed prediction of interactions to facilitate the design of high affinity, selective agonists as pharmacological tools to study the P2Y₁₄R.

Graphical abstract

INTRODUCTION

The P2Y receptors (P2YRs) are a family of membrane proteins that belong to the G protein-coupled receptor (GPCR) superfamily. Eight P2YR subtypes have been identified so far in human tissues and are further classified as P2Y₁-like (P2Y_1,2,4,6,11 subtypes) or P2Y₁₂-like (P2Y_12,13,14 subtypes) receptor subfamilies based on structural homology, sequence alignments, and the identification of conserved residues with key functions in each subtype [1,2]. The P2YRs are involved in specific signaling pathways; five P2Y₁-like receptors couple to a G_q protein to modulate a phospholipase C signaling pathway, while three P2Y₁₂-like receptors couple to G_i to inhibit adenylyl cyclase activity. Specifically, the P2Y₁₄R is involved in the mast cell degranulation pathway, which suggests it may be a potential therapeutic target for asthma treatment [3,4]. The P2Y₁₄R is also likely to be involved in the regulation of neuroimmune functions, but the lack of selective and stable agonists for this receptor have made it difficult to evaluate pharmacologically [5,6].

The P2Y₁₂-like receptors, also including P2Y₁₃R and P2Y₁₄R, share significant sequence identity and structural homology. As a result, the P2Y₁₂-like subtypes have been suggested to share a similar binding mode distinct from P2Y₁-like receptors. Uridine 5′-diphosphoglucose (UDP-glucose or UDPG, 1, Table 1) and other UDP-sugars and uridine 5′-diphosphate (UDP, 2) have previously been revealed to be endogenous agonists for the P2Y₁₄R. Yet, the fast degradation of nucleotides has made it difficult to study this receptor [7,8]. Previous structure-activity relationship (SAR) studies of synthetic nucleotide analogues have yielded highly potent agonist compounds by probing the base, ribose, phosphate, and terminal sugar moieties [9-11]. Such studies have revealed the functionally distinct regions of the ligand that contribute to receptor activation. Thio modification at the 2 position of the uracil base was found to significantly increase potency, the 2′ and 3′ hydroxyl groups of the ribose moiety were found to be necessary for receptor activation, and the hexose moiety was shown to be the most structurally permissive region. So far it has been difficult to interpret these SAR findings structurally due to the lack of a P2Y₁₄R crystal structure. Homology modeling serves as a means for predicting ligand-receptor interactions and understanding ligand conformations in the binding pocket of a receptor for which a structure has not yet been solved. However, a limitation of this method is that the quality of a homology model is dependent on the templates that are available.

Table 1.

In vitro pharmacological data for UDPG (1), UDP (2) and their respective analogues in the stimulation of phospholipase Cβ by activation of the recombinant hP2Y₁₄R expressed in COS-7 cells and transiently transfected with Gα_qi. X = H, Y = O and Z = O, unless noted.

Compound	Structure	Substitutiona	EC50 at hP2Y14R (nM)b
1	UDP-[1]glucose	-	400±90 [21]
2	UDP	R= H	160±40 [21]
3	2-thio-UDP	R= H; Y= S	1.92±0.69 [21]
4	2-Thio-UDP-[1″]glucose	Y= S	11±5 [21]
5	5-Iodo-UDP-[1″]glucose	X= I	NE [19]
6	5-Azido-UDP-[1″]glucose	X= N3	NE [19]
7	5-Amino-UDP-[1″]glucose	X= NH2	NE [19]
8	2′-deoxy-2′-azido-UDP-[1″]glucose	X= H; Z= N3	NE [19]
9	2′-deoxy-2′-amino-UDP-[1″]glucose	X= H; Z=NH2	NE [19]
10	UDP-0-cyclohexyl		5160±1830 [21]
11	UDP-2″-F[1″]2″-deoxyglucose		2500±900 [20]
12	UDP-3″-F[1″]3″-deoxyglucose		361±94 [20]
13	UDP-4″-F[1″]4″-deoxyglucose		567±156 [20]
14	UDP-6″-F[1″]6″-deoxyglucose		905±429 [20]
15	(S)mc-UDP-[1″]glucose		NE [21]
16	(N)mc-UDP-[1″]glucose		NE [21]

^aSee Figure S2 (Supporting information) for structures of methanocarba analogues 15 and 16.

^bNE, no effect at 10 μM.

Structures of the P2Y₁₂R have recently been determined using X-ray crystallography, and they may serve as the best templates yet for modeling other P2Y₁₂-like receptors [12,13]. P2YRs have been previously modeled using class A GPCRs as templates: Rhodopsin and β₂-adrenergic and CXCR4 chemokine receptors [14-24]. However, the P2YRs belong to a different branch (δ) of class A GPCRs than the aforementioned receptor templates. PAR1 is the only other δ-branch GPCR determined structurally, but even this receptor has some unique distinguishing features, such as the placement of proline residues [25].

Therefore, a homology model based on an agonist-bound structure of P2Y₁₂R (complex with 2-methylthioadenosine 5′-diphosphate, 2MeSADP, 17) may serve as an appropriate template to test the hypothesis that another member of the P2Y₁₂-like subfamily shares a conserved binding mode. We recently constructed a P2Y₁₄R homology model based on the agonist-bound P2Y₁₂R structure to explore nonnucleotide antagonist binding, but this was not applied to the analysis of nucleotide binding [26]. Molecular modeling studies may help to explain the patterns observed in previous SAR studies of P2Y₁₄R agonists while providing additional insight into the binding pocket regions that can be exploited for the design of selective agonist ligands for the P2Y₁₄R [27].

In this study we have constructed several human (h) P2Y₁₄R homology models based on the agonist-bound P2Y₁₂R structure, and we have docked previously published synthetic nucleotide agonists to structurally rationalize SAR findings, to understand the nature of the P2Y₁₄R binding pocket and to probe ligand-receptor interactions. In fact, a GPCR homology model is typically validated first using known ligand SAR prior to its application to the discovery of novel ligands [9-11].

Yet, a consequence of ligand docking to a homology model is the limitation of a perceived static model, which does not represent the dynamic nature of protein structures. To overcome this limitation of ligand docking, molecular dynamics is traditionally applied as a way to refine GPCR homology models through simulation in a solvated phospholipid bilayer [28-32]. However, molecular dynamics is at present limited to short timescales due to the significant amount of computing power that is required to solve Newton’s second law for each atom in an explicit solvent protein-membrane system. Regardless, molecular dynamics permits the exploration of a rugged energy landscape at even short timescales, which can extract useful information when applied to the refinement of ligand poses already docked in an orthosteric binding pocket. Thus, we applied molecular dynamics to the refinement of 2 docked in the orthosteric binding pocket of the P2Y₁₄R.

Our models have suggested a similar nucleotide binding mode in the P2Y₁₄R as in the P2Y₁₂R, thus supporting a conserved binding mode hypothesis for P2Y₁₂-like receptors involving functionally conserved residues in the binding pocket. Furthermore, an analysis of the binding pocket topology within the P2Y₁₄R model provides a qualitative explanation for previously observed SAR patterns. As a result, the P2Y₁₄R model may assist in the structure-based design of selective agonist compounds for studying the pharmacology of the P2Y₁₄R.

RESULTS

We built a homology model of the P2Y₁₄R based on the 2MeSADP-bound P2Y₁₂R crystal structure using the alignment shown in Figure S1 (Supporting information) showing an overall sequence identity of 45%. The C_α root-mean-square deviation (RMSD) between the P2Y₁₄R model and the P2Y₁₂R crystal structure was 0.8998 Å. A binding site comparison between the P2Y₁₄R homology model and the P2Y₁₂R template structure is shown in Figure 1. As in the P2Y₁₂R, the binding pocket was observed to have a bifurcated shape; with a main pocket delimited by TMs 3, 4, 5, 6 and 7 (pocket 1) and a smaller pocket delimited by TMs 1, 2, 3 and 7 (pocket 2). To refine the model and especially the orientation of the binding site side chains we performed induced fit docking of 1 and 2 to the original P2Y₁₄R model. The docking pose of 2 in the P2Y₁₄R model was similar to the conformation of 2MeSADP in the P2Y₁₂R and occupies the pocket delimited by TMs 3, 4, 5, 6 and 7 (Figure 2). The binding pocket where the nucleobase binds (left-half of each binding pocket in Figures 1 and 2) is noticeably larger in the P2Y₁₂R compared to the P2Y₁₄R. This observation was consistent with the preference of P2Y₁₄Rs for binding uracil over adenine nucleotides. The 2MeSADP binding mode at the P2Y₁₂R structure is stabilized by interactions with key conserved residues in the P2Y₁₂-like receptors. 2MeSADP is stabilized in the P2Y₁₂R crystal structure by π-π stacking with Y105^3.33 (numbers in superscript refer to the Ballesteros-Weinstein notation [33]); hydrophobic interactions with Y106^3.34, L155^4.56, S156^4.57, and V190^5.39; bidentate interaction with N191^5.40; H-bonds between the ribose 2′- and 3′-hydroxyl groups and K179^EL2, H187^5.36 and the backbone carbonyl of C97^3.25; and polar contacts between the phosphate groups and R256^6.55, K280^7.35, Y259^6.58 and Q263^6.62. Conserved residues between P2Y₁₂R and P2Y₁₄R were found to be involved in stabilizing the docking poses of 1 and 2 in the P2Y₁₄R model (Table S1, Supporting information). In fact, the uridine moiety of 1 occupied the same region as in the docking pose of 2, while the glucose moiety was accommodated in the second half of the pocket between TMs 1, 2, 3 and 7. A 2D representation of the interactions formed by 1 at the P2Y₁₄R is provided in Figure 3A. For comparison, the SAR of nucleotides at the P2Y₁₄R is summarized (Figure 3B). Key interactions predicted from the docking pose of 1 docked in the orthosteric binding pocket are shown in Figure 4. The models obtained after induced fit docking of 1 and 2 were then used to perform docking of various nucleotide analogues previously tested in functional assays at the P2Y₁₄R to structurally explain the main SAR findings. Structures and functional assay data of the nucleotide analogues analyzed in this study are listed in Table 1. In the following subsections, the analysis of the binding of each ligand moiety and the effect of substitutions are discussed.

Figure 1.

A. P2Y₁₄R model binding site surface comparison with the B. Agonist-bound P2Y₁₂R crystal structure (right) binding site surface. The gray areas represent the surface area in the receptor pocket predicted to be available for ligand binding. As expected, the nucleobase binding pocket delimited by TM3, TM4 and TM5 is larger in P2Y₁₂R compared to P2Y₁₄R (TM1=Blue, TM2=Cyan, TM3= Cyan/Green, TM4=Green, TM5=Yellow, TM6=Orange, TM7=Red).

Figure 2.

A. Docking pose of 2 at the P2Y₁₄R homology model. B. Crystal pose of 2MeSADP at the P2Y₁₂R (TM1=Blue, TM2=Cyan, TM3= Cyan/Green, TM4=Green, TM5=Yellow, TM6=Orange, TM7=Red).

Figure 3.

A. A 2D representation of the P2Y₁₄R residues predicted to interact with 1.The dotted lines indicate hydrogen and ionic bonding interactions, the solid line represents π-π stacking, and the curved lines indicate steric hindrance observed with nucleotide modification at a particular position. The base and glucose moieties each dock to one-half of the bifurcated binding pocket. B. Summary of structure-activity relationships of 1 in the P2Y₁₄R [20].

Figure 4.

Key interactions in the P2Y₁₄R homology model with 1 docked in the binding pocket.

Nucleobase

Docking results at the P2Y₁₄R model showed that the binding conformation of the nucleobase of 2 around the glycosidic bond could dock in either a syn or anti conformation. If the base adopted a syn conformation, more H-bonding interactions were observed between the uracil base and residues lining the binding pocket. However, in an anti conformation less steric hindrance was observed, and 2 is in its natural energetically favorable torsional conformation around the glycosidic bond. In both conformations, the pyrimidine base is stabilized by π-π stacking with Y102^3.33. A syn glycosyl conformation of the nucleobase provided for optimized H-bonding interactions in the receptor binding pocket. In fact, the carbonyl at the 2 position and the nitrogen at the 3 position would be involved in bidentate interactions with N188^5.39. However, optimized hydrogen-bonding interactions in the orthosteric binding pocket may not be functionally relevant considering that the 2MeSADP-bound P2Y₁₂R crystal structure suggests a preference for an anti glycosyl conformation in the orthosteric binding pocket.

In order to resolve the glycosyl conformation of the nucleotide compounds docked in the orthosteric binding pocket of the P2Y₁₄R, we applied explicit molecular dynamics refinement to the docking pose of 2. When 2 was docked in a syn glycosyl conformation, we observed a base-flipping event during the first 500 fs of the equilibration and rearrangement of N188^5.39 (Movie 1, Supporting information). Then, the docking pose remained stable in the binding pocket in an anti glycosyl conformation during a 10 ns production run. Thus, molecular dynamics refinement suggested an anti glycosyl conformation was more energetically favorable for 2 docked in the P2Y₁₄R model binding pocket.

Uracil nucleobase analogues modified at the 2 and 5 positions (compounds 3-7) were docked to the P2Y₁₄R model. Analogues substituted at the 5 position are previously reported to be inactive at the P2Y₁₄R. A possible explanation for the lack of activity of uracil derivatives substituted at the 5 position is steric hindrance with F191^5.42, which reduced the size of the orthosteric binding pocket in the P2Y₁₄R model compared to the P2Y₁₂R crystal structure. Modification at the 2 position, as in 3, was previously associated with a high potency in activating the P2Y₁₄R. The P2Y₁₄R homology model predicted that there is enough space in the pocket to accommodate the bulkier thio group at the 2 position of the nucleobase and a better fit with the cavity can be an explanation for the increase in potency of 3 compared to 2 (Graphical abstract).

Ribose

The ribose H-bonding interactions included H184^5.35, K176^EL2, and the carbonyl backbone of C94^3.25. It is known that the ribose ring conformation is an important determinant of recognition in the P2YR family [17]. Analogues of 1 containing a conformationally constrained North (N) or South (S) methanocarba (bicyclo[3.1.0]hexane, mc) ring system in place of the ribose moiety (compounds 15 and 16) are reported to be inactive at the P2Y₁₄R [11]. A structural explanation for the loss of activity of these analogues could be steric hindrance of the bicyclic ring system with V99^3.30, i.e. the presence of the fused cyclopropyl ring might interfere with ligand binding. However, when the ribose itself was constrained in either extreme conformation, the nucleotide 1 was able to dock to the homology model. Thus, we docked 1 with the ribose moiety frozen in a 2′-endo or 3′-endo pucker to further probe the selectivity of the model for a given ribose configuration. Figures 5A and 5B show both the frozen (S)- and (N)-conformations of 1 docked to the P2Y₁₄R model. (S)-1 maintained the ability to dock in the binding pocket with the preferred pose, while (N)-1 did not dock with the preferred pose. The P2Y₁₄R model suggests the (S)-ribose pucker may be preferred for receptor activation as a result of the relative positioning of the ribose hydroxyl groups for H-bonding interactions with C94^3.25 and K176^EL2.

Figure 5.

A. UDPG 1 in the ribose 2′-endo conformation (purple carbons) docked to the P2Y₁₄R model. This docking pose with the ribose ring frozen in a (S) pucker was consistent with the proposed binding mode. B. UDPG 1 in the ribose 3′-endo conformation (purple carbons) docked to the P2Y₁₄R model. This docking pose with the ribose ring frozen in a (N) pucker was not consistent with the proposed binding mode.

Modifications of the hydroxyl groups of the ribose ring, such as in 2′-modified compounds 8 and 9, display a loss of in vitro activity. These analogues fit to the P2Y₁₄R binding site with a similar orientation as 1, but were not able to form all of the H-bonding interactions observed in the native nucleotide structure. A possible explanation for the loss of activity in this series is the lack of favorable interactions with the highly conserved H-bond donor K176^EL2.

Phosphates

Ionic and H-bonding interactions in the P2Y₁₄R model were observed between the ligand phosphate groups and the following residues, which are conserved with the P2Y₁₂R: R253^6.55, Y256^6.58, Q260^6.62, and K277^7.35. An additional ionic interaction is formed with R274^7.32 that is not conserved in the P2Y₁₂R.

Glucose

After inspecting the pose for the hexose moiety, we found that multiple conformations were possible for the pyranose ring in the second half of the binding pocket of the P2Y₁₄R model (Figure 1B, pocket 2). Among those conformations, we selected the preferred glucose docking pose involving H-bonding interactions between residues in the P2Y₁₄R model and each hydroxyl group of the glucose ring to maximize the stability of this moiety in the binding pocket (Figure S4, Supporting information). The residues lining the hexose binding pocket included R274^7.32 E278^7.36, K77^2.60, D81^2.64, the backbone carbonyl of I170^EL2, and N90^3.21. The cyclohexyl derivative, compound 10, is a less potent analogue of compound 2. A lack of interactions to stabilize the pyranose ring in the receptor pocket due to the lack of hydroxyl groups may explain the loss of potency for 10 compared to 2.

Compounds 11 - 14, which are analogues of 1 that are fluoro-substituted on the glucose moiety, were docked to the model and their poses are shown in Figure 6. Importantly, the 2′′-fluoro modification was correlated with a loss of favorable interactions with three residues: R274^7.32, E278^7.36 and K77^2.60. Modification at the 3′′, 4′′ and 6′′ positions of the glucose moiety altered the docking conformation in the binding pocket and resulted in a loss of favorable interactions with R274^7.32 and N90^2.64. The in vitro activity pattern of the fluoro-modification series indicates that loss of a hydroxyl group at the 2′′ position has a greater impact on altering receptor activity than modification at the 3′′, 4′′ and 6′′ positions. The loss of H-bonding interactions with R274^7.32, E278^7.36 and K77^2.60 through modification at the 2′′ position suggested these residues may be involved in a H-bonding network that could be important for binding and/or receptor activation.

Figure 6.

The fluoro-atom modification at each position prevented hydrogen bonding interactions with key residues in the hexose binding pocket. A. Compound 11 (purple carbons) lost interactions with three residues: R274^7.32, E278^7.36, and K77^2.60. B. Compound 12 (purplecarbons) lost interactions toR274^7.32. C. Compound 13 (purple carbons) lost interactions withR274^7.32. D. Compound 14 (purple carbons) lost interactions with N90^3.21.

DISCUSSION

The P2Y₁₄R model has provided in silico validation for a previously predicted conserved binding mode in the P2Y₁₂-like receptors. The P2Y₁₄R model was initially refined to enhance the receptor conformation induced upon binding 1 and 2. We were able to induce a binding site conformation, which applied to both 1 and 2, that is consistent with the binding mode observed for 2MeSADP in the P2Y₁₂R (Figure 2)The refined models were then used to dock several derivatives to explain the SAR patterns previously empirically observed for P2Y₁₄R agonists.Interestingly, nucleotide-like antagonists of the P2Y₁₂R are known, while nucleotide antagonists of the P2Y₁₄R have not been reported. Molecular dynamics calculations of 2 docked in the P2Y₁₄R homology model provide in silico evidence of the thermodynamic stability of the selected docking conformation in the binding pocket during a 6 ns equilibration and a 10 ns trajectory (Movie 1, Supporting information).

In the P2Y₁₄R model we observed that the positively charged residues highly conserved in P2Y₁₂-like receptors K277^7.35 and R253^6.55 were involved in ionic interactions with the negatively charged phosphates, consistent with a previous prediction based on sequence analysis of P2Y₁₂-like receptors [34]. K176^EL2 is also highly conserved in P2Y₁₂-like receptors and was involved in stabilizing the ribose moiety by complementing the hydroxyl group at the 2′ position as a H-bond donor. The P2Y₁₄R model suggested that the (S) pucker is the favored ribose conformation for receptor activation. This conformation in the P2Y₁₄R model is also in agreement with the 2MeSADP (S) ribose conformation observed in the crystal structure of the P2Y₁₂R.

The motif K-E-X-X-L in TM7 has been identified through sequence alignments as highly conserved in P2Y₁₂-like receptors, which corresponded to the residues K277^7.35, E278^7.36, F279^7.37, T280^7.38 and L281^7.39 in the P2Y₁₄R. As already mentioned, K277^7.35 stabilized the negatively charged β phosphate. The docking pose of the glucose moiety of 1 placed the hydroxyl group at the 2′′ position near E278^7.36. It is interesting that fluoro modification at the 2′′ position introduced in compound 11 significantly reduced in vitro activity, which we attribute to a loss of favorable interactions with residues in the hexose binding pocket of the P2Y₁₄R model – including the E278^7.36 constituent of the K-E-X-X-L motif. The effect of the missing interaction with E278^7.36 on the activity of compound 11 supported its functional role in mediating receptor activity.

The π-π stacking interaction involving Y105^3.33 observed in the agonist-bound P2Y₁₂R structure was conserved in the P2Y₁₄R model to form the binding pocket for the nucleotide base. Molecular dynamics refinement of compound 2 docked at the P2Y₁₄R model suggests an anti conformation may be favored in order to minimize steric hindrance in the orthosteric binding pocket. An anti glycosyl conformation has also been determined through spectroscopy as the favored conformation in solution for most nucleotides. However, structural modification of the nucleobase through the addition of substituents may alter the preference for a given glycosyl conformation since substitution of the nucleobase may affect the most energetically favorable torsional conformation [26,35].

The addition of an amino group at the 5 position of the nucleobase suggests improved π-π stacking interaction due to the presence of a strong electron donating group. Yet, synthetic analogues modified at the 5 position of the nucleobase are inactive at the P2Y₁₄R. A question that remains unanswered with respect to the SAR of the nucleobase is whether the loss of activity of compounds 5, 6 and 7 is a result of an alteration in the favorable torsional conformation in the binding pocket or a result of steric effects with nearby residues in the nucleobase binding pocket. Ligand docking and molecular dynamics suggest an anti glycosyl conformation is thermodynamically favored for 2 in the orthosteric binding pocket of the P2Y₁₄R model, similar to the torsional conformation of 2MeSADP in the P2Y₁₂R crystal structure. Thus, the binding mode for nucleotides at the P2Y₁₂-like subtype might require an anti glycosyl conformation for proper receptor activation behavior. We hypothesize that F191^5.43 in the P2Y₁₄R may provide a means for selectivity of pyrimidine nucleotides against purine nucleotides. Comparably, the distal residue C194^5.43 in the base binding pocket of P2Y₁₂R may permit binding of adenine nucleotides by expanding the size of the binding pocket to accommodate purine nucleotides. This provides additional rationale for a conserved agonist binding mode in P2Y₁₂-like receptors by complementing the topology of endogenous uridine nucleotides selective for the P2Y₁₄R over the adenine-binding P2Y₁₂R and P2Y₁₃R. Additionally, the high potency of the 2-thio modification of the nucleobase may be explained by a better fit of the region delimited by TM3 and TM4 that helps anchoring compound 3 in its anti conformation in the binding pocket.

We sought to compare the conformation of the P2Y₁₄R model from this study to the previously reported P2Y₁₄R model in which the final receptor conformation was induced by flexible docking of the highly selective P2Y₁₄R antagonist 18 (MRS4174, Figure S2, Supporting information), which contains a fluorophore for receptor detection [13]. The P2Y₁₄R RMSD between the two models is 4.3 Å; this high value reflects mainly the large conformational differences in the flexible loop domains. The conformation of residues in the trans-membrane region of the binding pocket is similar in both models The residues K171^EL2, K77^2.60, and K176^EL2 are involved in stabilizing 18 in the extracellular region of the receptor binding pocket. In our P2Y₁₄R model, K176^EL2 was oriented downward, and we propose that the orientation of this residue may function to stabilize nucleotide compounds in the orthosteric binding pocket of the P2Y₁₄R by inducing a (S) pucker in the ribose moiety. Conversely, K176^EL2 is oriented upward in the antagonist 18-bound P2Y₁₄R model to stabilize the ligand through H-bonding interactions.

In conclusion, the P2Y₁₄R model has provided an explanation for the structure-activity relationships of various previously reported synthetic nucleotide agonists. The results from this study have assigned roles to several residues highly conserved within the subfamily in mediating receptor-ligand interaction networks, which supports the validation of previous experimental work. The P2Y₁₄R model has also predicted a previously unidentified non-conserved residue, F191^5.42, as potentially allowing for structural selectivity of the nucleobase at the P2Y₁₄R. This indicates that the P2Y₁₂R agonist-bound structure is an adequate template for a P2Y₁₄R model. Thus, P2Y₁₄R homology modeling may allow detailed prediction of interactions to facilitate the design of high affinity and selective agonists as pharmacological tools to study the P2Y₁₄R. This study also illustrates the potential of homology modeling, ligand docking and molecular dynamics as tools for understanding and predicting structural-functional relationships in G protein-coupled receptors.

EXPERIMENTAL SECTION

Homology modeling

The 2MeSADP-bound P2Y₁₂R structure at 2.5 Å resolution was obtained from the Protein Data Bank (PDB ID: 4PXZ)[12] and used as a starting template. The BRIL protein fused in intracellular loop 3, 2MeSADP, cholesterols and waters were removed from the P2Y₁₂R structure. Hydrogen atoms and missing side chains in unresolved regions were added, the hydrogen network was optimized, and minimization was performed using the OPLS-2005 forcefield with the Protein Preparation Wizard tool in the Schrödinger suite [36]. The FASTA sequence for P2Y₁₄R was obtained through the Uniprot database and aligned against the P2Y₁₂R sequence using the Prime tool in the Schrödinger suite. The alignment was visually inspected to ensure no gaps were present in predicted alpha-helical regions of the model. Then, the P2Y₁₄R model was built using the Prime homology modeling tool (energy-based method) of the Schrödinger suite.

Ligand docking

Ligands were built using Maestro and prepared using the Ligprep module and OPLS-2005 forcefield in the Schrödinger suite. The sitemap tool was used to identify probable binding site pockets in the receptor. Molecular docking of ligands to the P2Y₁₄R model was performed by means of the Glide package from the Schrödinger suite. In particular, a Glide grid was positioned on the centroid of residues located within 5 Å from the previously identified cavity. The Glide grid was built using an inner box (ligand diameter midpoint box) of 10 Å × 10 Å × 10 Å and an outer box (box within which all the ligand atoms must be contained) that extended 20 Å in each direction from the inner box. Induced fit was used to refine the binding site by separately docking 1 and 2 to provide optimized receptor models for the base and hexose binding pockets. The poses that best fit the coordinates of superimposed 2MeSADP-bound P2Y₁₂R were used for further inspection of ligand-receptor interactions. The Glide package was used for SP (standard precision) docking to the optimized models of all previously reported synthetic nucleotide agonist compounds. The conformational poses were selected based on similarity to the docking pose of 1 and agreement with available experimental data.

Molecular dynamics

P2Y₁₄R models were uploaded to the “Orientations of Proteins in Membranes (OPM)” database and a suggested orientation for each structure was provided based on the 2MeSADP-bound P2Y₁₂R orientation (PDB: 4PXZ) [37]. Each receptor model was then positioned in a 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC) lipid bilayer (70 Å × 70 Å) generated by a grid-based method using the VMD Membrane Plugin tool, and overlapping lipids within 0.6 Å were removed upon combining the protein-membrane system [38,39]. Each protein-membrane system was then solvated with TIP3P water using the Solvate 1.0 VMD Plugin tool and neutralized by 0.154 M Na⁺/Cl⁻ counterions [40].

This study utilized the high-performance computational capabilities of the Biowulf Linux cluster at the National Institutes of Health, Bethesda, Md (http://biowulf.nih.gov). Molecular dynamics simulations with periodic boundary conditions were carried out using Nanoscale Molecular Dynamics (NAMD) software and the CHARMM36 Force Field [41,42]. The ligands were parameterized by analogy using the ParamChem service (0.9.7.1) and implementing the CHARMM General Force Field for organic molecules (2b8) [43-46]. A 3000-step conjugate gradient minimization was initially performed to minimize steric clashes. The protein and ligand atoms were kept fixed during an initial 8 ns equilibration of the lipid and water molecules. Atom constraints were then removed and the entire system was allowed to equilibrate. The UDP-bound P2Y₁₄R structure was allowed to equilibrate for 6 ns (first plateau reached at around 1 ns, Figure S4), and the orthosteric binding pocket was visually inspected to ensure solvation of the binding pocket. An aporeceptor P2Y₁₄R model was allowed to equilibrate for 8 ns. A 10 ns production run was then carried out on the equilibrated UDP-bound P2Y₁₄R and apo-P2Y₁₄R models. The temperature was maintained at 300 K using a Langevin thermostat with a damping constant of 3 ps^-1. The pressure was maintained at 1 atm using a Berendsen barostat. An integration time step of 1 fs was used, while hydrogen-oxygen bond lengths and hydrogen-hydrogen angles in water molecules were constrained using the SHAKE algorithm [47]. VMD 1.9 was used for trajectory visualization and movie making. The PyMOL Molecular Graphics System, Version 1.6.0 Schrödinger, LLC was used for making figures.

Abbreviations

EL

extracellular loop

GPCR

G protein-coupled receptor

2MeSADP

2-methylthioadenosine 5′-diphosphate

TM

transmembrane helical domain

UDP

uridine 5′-diphosphate

UDPG

uridine 5′-diphosphoglucose