Identification and analysis of functionally important amino acids in human purinergic 12 receptor using a Saccharomyces cerevisiae expression system


J. Klovins, Latvian Biomedical Research and Study Centre, Ratsupites Str. 1, LV-1067 Riga, Latvia
Fax: +371 744 2407
Tel: +371 780 8003
V. Ignatovica, Latvian Biomedical Research and Study Centre, Ratsupites Str. 1, LV-1067 Riga, Latvia
Fax: +371 744 2407
Tel: +371 780 8003


The purinergic 12 receptor (P2Y12) is a major drug target for anticoagulant therapies, but little is known about the regions involved in ligand binding and activation of this receptor. We generated four randomized P2Y12 libraries and investigated their ligand binding characteristics. P2Y12 was expressed in a Saccharomyces cerevisiae model system. Four libraries were generated with randomized amino acids at positions 181, 256, 265 and 280. Mutant variants were screened for functional activity in yeast using the natural P2Y12 ligand ADP. Activation results were investigated using quantitative structure–activity relationship (QSAR) models and ligand–receptor docking. We screened four positions in P2Y12 for functional activity by substitution with amino acids with diverse physiochemical properties. This analysis revealed that positions E181, R256 and R265 alter the functional activity of P2Y12 in a specific manner. QSAR models for E181 and R256 mutant libraries strongly supported the experimental data. All substitutions of amino acid K280 were completely inactive, highlighting the crucial role of this residue in P2Y12 function. Ligand–receptor docking revealed that K280 is likely to be a key element in the ligand-binding pocket of P2Y12. The results of this study demonstrate that positions 181, 256, 265 and 280 of P2Y12 are important for the functional integrity of the receptor. Moreover, K280 appears to be a crucial feature of the P2Y12 ligand-binding pocket. These results are important for rational design of novel antiplatelet agents.


extracellular loop


G-protein-coupled receptors


purinergic receptors


purinergic 1 receptor


purinergic 12 receptor


quantitative structure-activity relationship




Purinergic receptors have two structurally different subgroups, the P2X receptors that are ligand-gated ion channels, and the P2Y G protein-coupled receptors (GPCRs) [1–3]. The human genome has eight genes, coding for P2Y–P2Y1, P2Y2, P2Y4, P2Y6 and P2Y11, which all functionally couple to the Gq subunit of G proteins to activate the phospholipase Cβ transduction pathway. However, P2Y12, P2Y13 and P2Y14 mediate their signals via Gi and inhibit adenylate cyclase. Natural agonists of these receptors are ADP, ATP, UDP, UTP and UDP-glucose [4]. P2Y12 is activated only by ADP [5]. Several unnamed receptors have considerable sequence homology with the P2Y receptors, but have yet not been shown to have nucleotide affinity [4,6–8].

P2Y12 is expressed on various human cell types including platelets, where it plays a critical role in platelet aggregation [5,8,9]. Stimulation of P2Y12 by the natural agonist ADP leads to αIIbβ3 integrin receptor activation, which binds fibrinogen, resulting in the formation of a blood clot [10–12]. P2Y12 is a target receptor for drugs such as ticlopidine and clopidogrel, which are frequently used prophylactically to protect against thrombosis after coronary interventions such as stenting or angioplasty [13,14].

Although P2Y12 is a major drug target, little is known about its functional domains. Few publications [15,16] have investigated P2Y12 although many functional studies have been carried out on other P2Y receptors [17–20]. The most studied is probably P2Y1, which is closely related to P2Y12 and binds the same natural agonist, ADP [5,21–23]. Overall, transmembrane (TM) 3, TM6 and TM7 domains are the most influential in signal transduction in the P2Ys. Strong evidence shows that a phosphate group in the ligand attracts polar amino acids (R, K) in these TM regions [19,20,24]. Mutagenesis studies of P2Y1 revealed that K280 and R286 are important in ligand recognition, and the corresponding regions of P2Y12, R256 and R265 influence receptor activity [15,16]. Other mutagenesis studies suggested that the R256 position is involved in ligand recognition [15] by P2Y12, but mutation of this position causes a minor decrease in receptor activity after activation with agonist [16]. Another residue described as important for the functional activity of P2Y12 is K280 [15] in TM7. This residue is conserved in all P2Ys [24] (Fig. 1).

Figure 1.

 Structure of P2Y12. Residues in grey were explored in this study.

The literature is controversial about the influence of extracellular loop (EL) 2 in ligand binding. This loop has been suggested to be involved in both nucleobase recognition [18] and directing the agonist into the binding pocket [23]. Several studies observed an impact of the acidic amino acid residues in EL2 [20,23,25], corresponding to E181 of P2Y12, but this position in P2Y12 has not been studied using mutational analysis.

In this study, we carried out an extensive characterization of four residues in P2Y12: E181, R256, R265 and K280. We generated randomized libraries of these positions in P2Y12 and found that all were important for receptor activation.


P2Y12 expression in yeast

The P2RY12 gene was cloned and expressed in the yeast Saccharomyces cerevisiae using strain MMY23. This strain contains a modified yeast pheromone-signalling pathway that enables cell growth after receptor activation, and a β-galactosidase reporter system that provides a quantitative colour reaction that is dependent on receptor activation level. We used ADP, the natural agonist for P2Y12, to generate a ligand–receptor affinity curve. The EC50 value for activation of P2Y12 with ADP was 279 ± 17 nm, comparable with data in the literature [26], and the system was considered appropriate for further experiments.

Isolation of random P2Y12 variants

Four recombinant constructs randomized at amino acid positions 181, 256, 265 and 280 were generated to obtain randomized libraries of P2Y12. All vector constructs were sequenced to confirm library heterogeneity. Randomized libraries were transformed into yeast, which were grown on agar plates, followed by growth of selected clones in liquid medium for functional analysis. We selected and sequenced 50 colonies for positions 181, 256, and 265; and 100 colonies for position 280. Identified receptor variants are shown in Table 1. From the 50 colonies from the E181 library, 24 contained more than one amino acid variant-containing plasmid at the randomized position as indicated by several nucleotide peaks at the randomized positions in sequence chromatograms. Multiple codons were also detected for 28 clones from the R256 library, 31 clones from R265 and 19 clones from K280. In total, we isolated plasmids representing 15 different amino acid variants at position 181, 10 at position 256, 10 at position 265, and 15 at position 280.

Table 1.   P2Y12 variants identified from mutagenised libraries and EC50 values after ADP stimulation. Underlined, yeast preference codon; wt, wild-type codon; STOP, stop codon; MIX, harvested colonies containing more than one codon variant; NA, no activity observed; NT, not tested. *P < 0.05, **P < 0.01, ***P < 0.001 significance of differences for wt P2Y12 EC50 ± SEM versus mutant P2Y12 EC50 ± SEM (t-test); ns, non-significant (P > 0.05 of differences EC50 ± SEM P2Y12 variant activated with ADP versus EC50 ± SEM P2Y12 variant activated with ADP + 1 μm AR-C66096 according to t-test).
Recombinant P2Y12 variantCodonsNumber of clonesAmino acidEC50 ± SEM (ADP, nm)EC50 ± SEM (ADP + 1 μm AR-C66096, nm)Dose ratio (EC50 AR-C66096 vs. EC50 ADP)
wtGAG3279 ± 17749 ± 1642.68
E181RCGG1Arginine297 ± 705354 ± 150418.03
E181KAAA, AAG2Lysine391 ± 77NT
E181HCAC2Histidine463 ± 140NT
E181DGAC, GAT2Aspartate228 ± 151291 ± 421ns
E181AGCC1Alanine379 ± 372194 ± 4725.79
E181CTGT1Cysteine529 ± 217NT
E181VGTA, GTT2Valine627 ± 225NT
E181LCTA, TTA3Leucine1297 ± 210**NT
E181IATC1Isoleucine1188 ± 248*2197 ± 504ns
E181FTTC1Phenylalanine647 ± 262NT
E181NAAT1Aspargine497 ± 114NT
E181SAGC, TCT2Serine366 ± 21*NT
E181TACC1Threonine733 ± 190NT
E181YTAC, TAT2Tyrosine803 ± 2824006 ± 1262ns
R256HCAC1Histidine13483 ± 1893**NT
R256AGCG1Alanine9679 ± 75***NT
R256GGGA, GGC, GGT9Glycine12102 ± 9428NT
R256IATA4Isoleucine1092 ± 19*5906 ± 12035.41
R256NAAC5Asparagine1704 ± 492*NT
R256QCAA, CAG5Glutamine4636 ± 2810NT
R256STCT2Serine3214 ± 246***NT
R265KAAG1Lysine578 ± 244511 ± 60ns
R265EGAG1Glutamate1701 ± 704NT
R265DGAC1Aspartate532 ± 2451906 ± 1953.58
R265AGCC1Alanine1189 ± 268*661 ± 99ns
R265PCCT3Proline821 ± 228NT
R265LCTA, CTG3Leucine1310 ± 223**NT
R265IATA2Isoleucine3562 ± 798*NT
R265FTTC, TTT2Phenylalanine555 ± 1282350 ± 2314.23
R265STCT1Serine522 ± 141732 ± 103ns
R265TACA3Threonine607 ± 216NT
K280EGAA, GAG8GlutamateNANT
K280IATA, ATC, ATT8IsoleucineNANT
K280YACC, ACG6TyrosineNANT

Functional activation of recombinant P2Y12

ADP was used for all activation experiments with recombinant P2Y12 variants. In cases where different codons coded for the same amino acid, we used the variant containing the codon with the best yeast expression preference [27] (Table 1). All activation experiments were repeated at least three times, and the EC50 values and ligand–receptor affinity curves are displayed in Table 1 and Fig. 2, respectively. Functional activation of wild-type P2Y12 was also carried out with 2-methylthio-adenosine-5′-diphosphate, to compare these results with ADP activation and observe stability of ADP in yeast expression system (Fig. S1).

Figure 2.

 Activation curves of recombinant P2Y12s. Substitutions with (A,D,G) hydrophilic–charged, (B,E,H) hydrophobic, and (C,F,I) hydrophilic–neutral residues.

All substitution variants for the E181-randomized library were functionally active, although the E181L and E181I variants showed almost fivefold lower activity compared with wild-type P2Y12. All recombinant variants at position 256 displayed at least fivefold lower activity. Furthermore, the R256P and R256Y variants were functionally inactive. The obtained variants of the amino acid 265 library generally demonstrated decreased activity levels compared with the wild-type P2Y12, with especially low EC50 values for R265E, R265A, R265L and R265I. Any substitution of the 280 position with exception of the wild-type amino acid (P2Y12 codons AAA and AAG isolated from the library and representing the wild-type amino acid at this position were tested) inactivated the receptor completely (Table 1).

Several mutant P2Y12 variants were functionally characterized by the P2Y12 antagonist AR-C66096, for this activation P2Y12 variants with EC50 values with ADP < 1.5 μm were selected. In addition, only one mutant receptor variant from the hydrophilic basic, hydrophilic acidic, hydrophilic neutral and hydrophobic amino acid groups was tested. For six of 12 different receptor variants tested, AR-C66096 produced significant shifts in the ADP dose-response curves (Table 1 and Fig. S2).

Quantitative structure–activity relationship

We studied the physiochemical characteristics of E181, R256 and R265 using quantitative structure–activity relationship (QSAR) analysis. For amino acid position 181, the best QSAR model (R2 = 0.77, Q2 = 0.55) (Fig. 3A) was obtained using five z-scale descriptors from Sandberg et al. [28] and adding a charge description to the model. The most important descriptor in this model proved to be the z1-scale, representing amino acid hydrophobicity/hydrophilicity (Fig. 3B). The highly positive regression coefficient indicated that hydrophobic amino acids in this position resulted in lower receptor activity. The R256 position was better characterized using the Gottfries et al. [29] descriptors system (R2 = 0.80, Q2 = 0.60) (Fig. 3C), which revealed a requirement for an amino acid with a smaller rigidity (i.e. without Tyr, Pro or His rings) and higher flexibility (i.e. longer side chains) at this site (Fig. 3D). None of the descriptor combinations gave a highly predictive model for the R265 position based on our data. The best model (R2 = 0.52, Q2 = 0.20) was obtained using five z-scales (Fig. 3E,F). Positive regression coefficients for the z1- and z2-scales indicated a preference for an amino acid that was hydrophilic and large at this position. However, analysis of the cross-validation results showed that the model could not find a reason for the threefold difference in activity for receptors containing Asp or Glu, or receptors containing Leu and Ile at this position.

Figure 3.

 QSAR models for positions 181, 256 and 265 of P2Y12. Correlation of predicted versus measured activities in models for positions (A) 181, (C) 256 and (E) 265. Orthogonal partial least-square regression coefficients of the used descriptors in models for positions (B) 181, (D) 256 and (F) 265.

Homology modelling and docking of P2Y12

To investigate the possible role of the tested residues in formation of the ADP-binding pocket of P2Y12, we generated a homology model of P2Y12 and performed docking studies using ADP as a ligand. The docking model with the lowest free energy is presented in Fig. 4. In our model, the nucleobase group of ADP docked in the cavity between the TM domains. The residues in extracellular loops, E181 and R265, did not appear to be involved in formation of the ADP-binding pocket, whereas amino acids R256 and K280 were located in close proximity of the docked ADP (Fig. 5). Residue K280 was located only 1.5 Å from the ADP phosphate group, suggesting involvement in binding pocket formation.

Figure 4.

 P2Y12-ADP docking model. ADP docked between TM helices, close to the extracellular side of the receptor cavity.

Figure 5.

 Detailed view of P2Y12-ADP binding pocket from two viewpoints. (A,B) Four studied amino acids (coloured by atom type) and ADP (amber).


This is one of the first studies to describe mutational analysis of P2Y12 using a yeast system [30] and to discuss advantages of the expression of purinergic receptors in S. cerevisiae compared with mammalian cells. We revealed the specific importance of amino acid positions 181, 256, 265 and 280 for functional activation of the receptor and suggested binding pocket characteristics for the natural agonist ADP. We also propose yeast to be a highly sensitive and efficient screening system for P2Y12. EC50 values obtained for both ADP (279 ± 17 nm) and the chemically stable ligand 2-methylthio-adenosine-5′-diphosphate (2.66 ± 0.42 nm) are similar to those reported for P2Y12 activation in the mammalian cell systems [31] and in the yeast-based assays [26], indicating that unwanted ADP hydrolysis is not a major issue that may influence the results obtained in our study. This system may be useful for avoiding problems characteristic of mammalian cell cultures. Most of the relevant mutational studies in past have been carried out on other purinergic receptors, and our initial selection of residues was based on knowledge from other P2Y receptors. Recent studies [15,16] have reported mutations of human P2Y12 at positions 256, 265, and 280, allowing us to compare the results obtained in these studies.

A role for EL2 in formation of the ligand-binding pockets in GPCRs has been proposed [32], and evidence supports the involvement of this region in ligand recognition of P2Ys. The amino acid E181 is located in EL2 of P2Y12 and was selected for analysis based on mutagenesis data on E209 and E186 in the EL2 of P2Y1 and P2Y11 that affect the activity of these receptors [20,25]. We extensively explored position 181 of P2Y12 and our data suggested that hydrophobity at this site (when substituted with Leu or Ile) lowered the functional activity of P2Y12, but hydrophilic recombinant receptor variants (Asp, Arg, Lys substitutions) functioned nearly as well as wild-type P2Y12.

Residue R265 is located in the EL3 of P2Y12. Previous studies havce shown that substitution of this position (corresponding to R287 of P2Y1) with hydrophobic (Ala) or hydrophilic–neutral (Gln) amino acids, reduces the functional activity of the receptor by 1000-fold [25]. However, a P2Y12–R265W recombinant receptor showed only a minor change (1.5-fold lower activity) [16] in mammalian cell system. By contrast, our data suggested that even substitution of Arg at the 265 position with the chemically related amino acid Lys led to a twofold decrease in activity.

Activation results of the recombinant variants of the R256 library were also in slight disagreement with previous reports. Lys, Ala and Asp recombinant receptor variants have decreased cAMP inhibition after 2-methylthio-adenosine-5′-diphosphate activation [15]. Substitutions of this position with Thr and Gln are shown to have very little effect on P2Y12 activity after ADP stimulation (1.2- and 2.3-fold lower activity, respectively) [16], whereas we observed 17-fold lower activity for the R256Q mutant. It should be noted, however, that these data are not quite comparable with our results because different agonists were used.

Our results might also suggest that expression in the yeast system is more specific for P2Y12 than in most commonly used mammalian cells, supporting assumptions by Schonberg et al. [30]. This is somewhat surprising because yeast systems usually have lower potency than mammalian cell cultures [33,34]. However, we observed a P2Y12 activation potential similar to that described previously [26], and believe that expression of P2Y12 in yeast is highly suitable for this type of study. Similar results have been obtained for P2Y1 receptors where agonists and antagonists were found to be equally potent in the yeast cells and transfected 1321N1 human astrocytoma cells that endogenously do not express P2Y receptors [35]. A number of GPCRs are activated by nucleotide/nucleoside ligands, so other mammalian cell lines are likely to have endogenous background expression of these receptors, leading to inadequate interpretation of activation results. In mammalian cells, P2Y12 might be influenced by conditions such as expression of endogenous P2Y12 ligands or even the presence of a special compensation mechanism when P2Y12 functions are distorted. In conclusion, yeast is a more isolated and straightforward system, with no GPCRs expressed that could lead to possible false-positive results like in mammalian cells.

Interestingly, the QSAR models successfully explained our experimental data on positions E181 and R256. The E181 model proved the necessity of hydrophilic residues at this position for the functional activity of the P2Y12 and this could support the assumption that E181 has important role in ligand binding. However, the question of whether it is a significant component of ligand-binding pocket remains. The R256 model revealed that the most influential characteristic of the amino acid at the 256 position is low rigidity and high flexibility (Fig. 3D). Positively charged amino acids like Arg and Lys are widely thought to be common for all P2Ys in the upper parts of TM helices, and they are believed to form binding pockets for negatively charged nucleotides that are especially attractive to phosphate groups [19,20,25]. Our QSAR results indicate that the R256 position of P2Y12 has specific physical properties that contribute to the spatial organization of the binding pocket but not to the direct interaction of this residue with the ligand.

Docking studies of P2Y12 revealed a possible ADP-binding pocket. We believe that our model is better supported than a previously reported model of P2Y12 [36], because we used data from the A2A adenosine receptor X-ray structure [37] as a template for modelling instead of the bovine rhodopsin receptor. The A2A receptor has a higher amino acid homology to P2Y12 than the bovine rhodopsin receptor and binds a similar class of ligands. Furthermore, the A2A template is in the active state, with bound antagonist, and this more closely mimics the conformation of the receptor activated by ligand in vivo.

In our model, the amino acids in the EL of P2Y12 (E181, R265) appear to be far from the docked ligand (Fig. 5). Therefore, the involvement of these residues in binding pocket formation seems implausible. However, if EL has a major role in attracting ligand, EL residues may form a ligand meta-binding site on the outside of the receptor that helps to guide the ligand further into the TM cavity to the final binding pocket. This hypothesis has been proposed previously for P2Y1 [23]. The R256 and K280 residues are close to the docked ADP in our model and might therefore act as the main determinants of receptor–ligand interaction in the binding pocket (Fig. 5). This is supported by our data showing that any substitution at position 280 in P2Y12 leads to nonfunctional P2Y12. Even replacement of the wild-type amino acid Lys with the related amino acid Arg abolished the activity of P2Y12; a minor change in the binding pocket is likely to disarrange it completely (Table 1). Indeed, our docking model demonstrated that K280 is in very close proximity to the second phosphate group on ADP (Fig. 5), supporting this interaction. This could ensure the placement of ADP in the binding pocket within the cavity formed by the TM helices.

Positions 256 and 265 in P2Y12 are reported to be associated with a congenital bleeding phenotype [38]. Interestingly, substitutions of both R256Q and R265W are demonstrated to contribute to a decrease in platelet aggregation, but separately they cause only a minor change in P2Y12 activity [16,38]. We did not attempt to generate the double mutant receptor, but substitution of either position led to a significant decrease in activity (from twofold lower to total abolishment of function). We might have observed this effect because of the higher efficiency and specificity of the yeast expression system, as discussed above, but additional explanations might exist. Studies have demonstrated the role of P2Y1 hetero-oligomerization and its impact on ligand-binding properties [39,40]. If we assume that P2Y12, in mammalian cells as well as in vivo, forms dimers with other GPCRs that alter the ligand-binding cavity structure, the double mutant (R256Q and R265W) could decrease ligand binding, whereas a single substitution might not substantially change binding because a more stable structure is formed by hetero-oligomerization. In yeast, however, stabilization by hetero-oligomerization might not take place because of a lack of other endogenously expressed GPCRs.

To estimate the involvement of the analysed amino acids in antagonist binding, we tested the competitive P2Y12 antagonist AR-C66096 on a selected set of mutants. For the majority of substitutions at the E181 and R256 positions, we observed an increase in the dose ratio at 1 μm AR-C66096 (Table 1 and Fig. S2), indicating that none of these amino acids had a strong prevalence for AR-C66096 binding. In the case of R265K, R265A and R265S substitutions, no significant shifts in ADP response curves were observed, indicating that R265 may be more important for antagonist recognition.

In conclusion, we demonstrated that substitutions at positions 181, 256, 265 and 280 alter the functional activity of P2Y12 in a specific manner, and that these positions are likely to be important for structural and functional modelling of the purinergic receptors. These results can be useful for development of novel antiplatelet agents.

Materials and methods

Cloning of P2RY12 and construction of mutant libraries

The P2RY12 gene was cloned from anonymous human DNA, using primers P2RY12-Fw (CCAAAAGCTTATGCAAGCCGTCGACAACCT) and P2RY12-Rs (TA CTCGAGTTACATTGGAGTCTCTTCATTT) (Metabion, Martinsried, Munich, Germany) containing restriction sites HindIII and XhoI (underlined) and were cloned directly into the yeast shuttle vector p426TEF with the URA gene for auxotrophic selection [41]. Consistency of the cloned sequence to the P2RY12 gene (ID: 64805 in GenBank) was confirmed by sequencing. Mutant P2RY12 libraries were constructed by PCR using a QuikChange II XL Site-Directed Mutagenesis Kit (Stratagene, Santa Clara, Canada) and oligonucleotides E181rand-Fw (CATGCCAGACTAGACCGAANNNTGATTTAAGGAAAGAGCA), E181rand-Rs (TGCTCTTTCCTTAAATCANNNTTCGGTCTAGTCTGGCATG), K280rand-Fw (TAACCACAGAGTGCTCTCNNNCACATAGAACAGAGTATT), K280rand-Rs (AATACTCTGTTCTATGTGNNNGAGAGCACTCTGTGGTTA), R256rand-Fw (TCAGGGTGTAAGGAATNNNGGCAAAATGGAAAGGAAC), R256rand-Rs (GTTCCTTTCCATTTTGCCNNNATTCCTTACACCCTGA), R265rand-Fw (CAGTGCAGTCAAAGACATCNNNGGTTTGGCTCAGGGTGT) and R265rand-Rs (ACACCCTGAGCCAAACCNNNGATGTCTTTGACTGCACTG) (Metabion) with randomized trinucleotides (in bold) corresponding to amino acids E181, K280, R256, R265 of P2Y12R and the p426-TEF–P2RY12 construct serving as DNA template. To confirm the heterogeneity of the randomized products, we used direct sequencing with primers P2RY12-Fw and P2RY12-Rs with 3130xl Genetic Analyser (Applied Biosystems, Carlsbad, CA, USA).

Yeast transformation

Yeast cell maintenance, transformation and functional activation of recombinant receptors were as described previously [42]. Briefly, yeast S. cerevisiae strain MMY23 with genotype: MATa his3 leu2 trp1 ura3 can1 gpa1Δ::Gs far1Δ::ura3 sst2Δ::ura3 Fus1::FUS1-HIS3 LEU::FUS1-lacZ ste2Δ::G418R, was kindly provided by SJ Dowell (GlaxoSmithKline Medicines Research Centre, Hertfordshire, UK) [43]. This strain expresses the Gi subunit, and has a receptor activation system fused to a specific signalling pathway with reporter genes for β-galactosidase production and histidine synthesis.

Yeast cells were grown overnight in 4 mL YPD medium (20 g·L−1 bactotryptone, 10 g·L−1 yeast extract; Difco Laboratories, Franklin Lakes, NJ, USA). Transformations with vectors containing the randomized libraries were carried out in 100 mm lithium acetate (Sigma, Deisenhofen, Germany) with 0.2 mg·mL−1 salmon sperm DNA (Sigma, St. Louis, MO, USA), with heat shock for 40 min and growth on SC agar (1.92 g·L−1 yeast synthetic drop-out medium supplement without uracil, 1.7 g·L−1 yeast nitrogen base, 5 g·L−1 ammonium sulfate, 2% glucose, 16.67 g·L−1 agar; Sigma).

Functional activation of recombinant receptors

Colonies from selective agar plates were harvested in 2 mL liquid SC medium lacking uracil and grown overnight. Functional activation was carried out in 96-well plates (Sarstedt, Nümbrecht, Germany) in 200 μL SC medium without histidine and uracil, containing 2 mm 3-amino-1,2,4-triazole (Sigma), a competitive inhibitor of imidazoleglycerol-phosphate dehydratase for control of background synthesis of histidine, 10× BU salts (70 g·L−1 Na2HPO4•7H2O, 30 g·L−1 NaH2PO4 pH 7 adjusted with 2 m NaOH; Reahim, Samara, Russia), 0.1 mg·mL−1chlorophenol red-β-d-galactopyranoside (Roche, Mannheim, Germany) and ligand ADP (Sigma) from 10 nm to 30 μm. Cells were diluted with activation medium to 250 cells·μL−1. For antagonist experiments 1 μm AR-C66096 (Tocris, Bristol, UK) was added. Negative controls for each clone were performed in activation medium without ligand to determine the background activity of the receptor variants. Wt P2Y12 was also activated with 2-methylthio-adenosine-5′-diphosphate (Sigma), to compare with ADP produced response. Cells were grown at room temperature for 24 h or longer. D595 was measured on a Victor3 V reader (Perkin–Elmer, Wellesley, MA, USA).

Analysis of activation data

Data analysis was performed using graphpad prism (GraphPad Software, La Jolla, CA, USA). Data sets were normalized according to highest and lowest values, and transformed using function X = log(X), and EC50 values and standard error of the mean (SEM) were calculated automatically. Significance of differences between wild-type P2Y12 EC50 and mutant P2Y12 EC50 values as well as significance of antagonist induced EC50 shift was estimated by t-test.


QSAR analysis explored possible physiochemical characteristics of the analysed positions required for P2Y12 functioning. EC50 values from functional mutant receptor activation experiments were converted to logarithmic scale. We included inactive recombinant P2Y12 variants in the QSAR models, assigning them an EC50 value of 105 nm to indicate dramatically decreased receptor activity. Mutated amino acids were described by five z-scales, z1 to z5, derived from Sandberg et al. [28] Z-scales were obtained by principal component analysis of 26 different physicochemical properties of amino acids; they were mutually uncorrelated and essentially represented hydrophobibity/hydrophilicity (z1), steric/bulk properties (z2), polarity (z3) and electronic effects (z4 and z5). Amino acids were also characterized by two descriptors proposed by Gottfries et al. [29] (t-Rig, rigidity; t-Flx, flexibility). We added charge as a descriptor (1, positive; −1, negative; 0, not charged). Combinations of these descriptors were explored to obtain a model that best correlated the amino acid properties with the receptor activity from the mutant libraries. Models were created using orthogonal partial least-square regression by simca p-11 software (Umetrics AB, Umea, Sweden). Goodness of fit of the models was estimated by calculating the fraction of the explained activity variation, R2. Predictive ability was estimated by calculating the fraction of predicted activity variation, Q2, according to sevenfold cross-validation.

Homology modelling of P2Y12 and docking of ADP

Homology modelling was performed with molsoft icm-pro 3.5 software (MolSoft LLC, San Diego, CA, USA) using as a template the 2.6 Å resolution X-ray structure of human A2A adenosine receptor (PDB entry 3EML) [37]. Alignment of sequences was adjusted according to GPCR AlignmentBuilder (;jsessionid=6f83fb65e2cfa25a405be0c3acdb), which implements a special alignment algorithm for class A (rhodopsin-like) GPCRs. This algorithm is designed to consider conserved residues in TM regions of rhodopsin-like GPCRs, as described previously [44]. A model of P2Y12 was created by the ICM-Homology modelling algorithm as implemented in MolSoft icm-pro 3.5. After alignment of the P2Y12 sequence to the template structure and replacement of nonidentical residues, simultaneous global energy optimization was carried out and side chain torsion angles predicted. This procedure was based on biased probability Monte Carlo conformational search and optimization [45,46]. The quality of the generated model was verified by a specialized ICM function that predicted possible backbone deviation between the model and the template [47].

The homology model of P2Y12 was used for ADP docking, using the ICM-Docking algorithm in MolSoft icm-pro 3.5. First, a binding pocket was identified and receptor energy maps generated with grid cell size 0.5 Å. Flexible ligand docking was then carried out from multiple conformations and locations. Finally, refinement of a P2Y12 structure was made for all docked ADP conformations to adjust the receptor conformation to the ligand. The docking model with the lowest free energy value was considered the most likely to correspond to in vivo binding.


The study was supported by the Latvian Council of Science Grant LZPSP10.0010.10.04, and the Latvian Research program 4VPP-2010-2/2.1. VI and KM were supported by ESF grant 1DP/ ML was supported by the Swedish Society for Medical Research and Åke Wiberg foundation. HS was supported by the Swedish Research Council. English proofreading and editing was supported by ERAF grant 2DP/ Yeast strain MMY23 was kindly provided by SJ Dowell (GlaxoSmithKline Medicines Research Centre, Hertfordshire, UK) [43].