Extracellular adenine and uracil nucleotides (e.g., ATP, ADP, UTP, UDP and sugar nucleotides) are signaling molecules involved in several patho physiological phenomena, from short-term signaling (neurotransmission, mechanosensory transduction, secretion and vasodilatation) to long-term functions (proliferation, differentiation, survival and death, development and post-injury repair) . Conversely, cysteinyl-leukotrienes (cysteinyl-LTs) are inflammatory lipid mediators derived from arachidonic acid through the 5-lypoxigenase (5-LO) pathway, and are implicated in bronchial asthma, stroke and cardiovascular diseases . Despite the fact that nucleotides and cysteinyl-LTs originate from totally independent metabolic pathways, several data suggest important functional interactions between two families of signaling molecules and their receptors. To date, eight distinct nucleotide G-protein-coupled receptors (GPCRs), the P2Y receptors have been identified (P2Y1;2;4;6;11;12;13;14) and classified in two distinct phylogenetic subgroups: the first subgroup includes the P2Y1;2;4;6;11 subtypes, whereas P2Y12, P2Y13 and P2Y14 belong to the second subgroup . Only two cysteinyl-LTs responding GPCRs (the CysLT1 and CysLT2 receptors) are instead currently recognized. However, certain reported actions of cysteinyl-LTs are not readily explained by interaction with either CysLT1 or CysLT2, raising the possibility of the existence of additional CysLT receptors [4–7]. There exists a functional cross-talk between the P2Y and CysLT receptor families, since both nucleotides and cysteinyl-LTs massively accumulate at sites of inflammation and both types of receptors are co-expressed in the same peripheral inflammatory cells. This evidence shows a cross-regulated response typical of the chemoattractant systems . Along this line, in rat brain microglial cells, both nucleotides and cysteinyl-LTs, that are co-released as a consequence of the activation of P2Y1 and CysLT receptors, contribute to neuroinflammation and neurodegeneration . Nucleotides can also regulate, via heterologous desensitization, CysLT1 receptor activity  and, in parallel, the CysLT1 receptor antagonists pranlukast and montelukast can functionally influence P2Y receptor signaling pathways in human monocyte/macrophage-like cells . In addition, P2Y12 was found to be promiscuously activated by both nucleotides and LTE4 , further underlying the close relationship between the two families. Both P2Y and CysLT receptors share the typical seven-transmembrane spanning topology of GPCRs. Besides their heterogeneity in function and tissue distribution, P2Y and CysLT receptors share a phylogenetic relationship, given that both families, together with GPR17 and other related receptors, belong to the so called "purine receptor cluster" of GPCRs . This cluster also includes several "orphan" receptors responding to yet-unidentified endogenous ligands. Among these, the orphan receptor GPR17 appeared to us as a possible common ancestral progenitor that originated the two above receptor families. On this basis, we recently cloned the human, rat and mouse GPR17 and demonstrated that they all respond to both nucleotides and cysteinyl-LTs [13, 14].
Thus, GPR17 is a hybrid receptor linking the P2Y and the CysLT receptor families. Besides endogenous ligands, synthetic compounds typical of the two above receptor families are also active at GPR17. Specifically, it has been shown that GPR17 can be activated in vitro by uracil nucleotides (UDP and UDP-sugars) and by cysteinyl-LTs (LTC4, LTD4 and LTE4). GPR17 activation can be contrasted by treatment with two well known P2Y antagonists, MRS2179 (2'-deoxy-N6-methyladenosine 3',5'-biphosphate) and cangrelor (N(6)-(2-methyl-thioethyl)-2-(3,3,3-trifluoropropylthio)-β, γ-dichloromethylene-ATP), and also by the already marketed CysLT receptor antagonists pranlukast (N-[4-oxo-2-(2H-tetrazol-5-yl) chromen-7-yl]-4-(4-phenylbutoxy) benzamide) and montelukast (2-[1-[[(1R)-1-[3-[2-(7-chloroquinolin-2-yl) ethenyl] phenyl]-3-[2-(2-hydroxypropan-2-yl) phenyl] propyl] sulfanylmethyl]cyclopropyl] acetic acid). Furthermore, in a model of focal rodent brain ischemia, its in vivo early knock down with either pharmacological or specific antisense strategies, reduces the progression of cerebral ischemic damage, highlighting GPR17 as novel therapeutic target for ischemia . Since at present this disease still remains without a specific pharmacological treatment, molecules active as GPR17 inhibitors may represent a new class of promising anti-ischemic agents. On the other hand, more recent data has shown that GPR17 indeed has a dual and spatiotemporal-dependent role in the development and post-injury repair of damage in the brain and in spinal cord. While at very early times after injury, GPR17 seems to mediate cell death, at later stages, GPR17 may even participate to repair mechanisms [14, 15]. Thus, GPR17 may be proposed as a "sensor" of damage that is activated by the specific signaling molecules (uracil nucleotides and cysteinyl-LTs) that are released at high levels in the lesioned area, and as a new target for amyelinating post-injury responses. These data further highlight the attractivity of this receptor as a new target for drug discovery.
Recently, results obtained in recombinant systems, have been proposed GPR17 as a constitutive ligand-independent negative regulator of the CysLT1 receptor, that modulates CysLT1-mediated functions at the cell membrane . Although this interesting hypothesis will have to be confirmed in vivo, it may be hypothesized that GPR17 may function as both a ligand-dependent and independent receptor depending upon specific patho-physiological conditions. Definitely, to fully understand the therapeutic potential of GPR17, specific ligands that do not interfere with the other P2Y or CysLT receptors are needed.
Along this line, as a first step to the design of selective ligands, we have recently provided a computational study of GPR17, providing a macroscopic view of a three-dimensional (3D) model of GPR17 complexed with three representative purinergic compounds: the endogenous agonist UDP and the synthetic antagonists MRS2179 and cangrelor . To do so, we used a raw homology model of GPR17, based on the X-ray crystallographic 3D structure of bovine rhodopsin (b Rh) 1U19, deposited at RCSB Protein Data Bank http://www.pdb.org, the best high-resolution 3D template for a mammalian GPCRs that was then available . In fact, for many years, the crystalline structure of the inactive receptor form of b Rh has been widely used as a template, even for significantly distant receptors, on the basis of the commonly accepted assumption that, in evolutionary related proteins, the 3D structure is more conserved than the amino acid sequence. Fortunately, within the GPCRs superfamily, one of the essential determinant for GPCRs activity concerns the 7TM architecture that is well conserved between all GPCRs.
Rhodopsin-based homology models of GPCRs and the subsequent structure-based drug discovery approach are widely accepted, since experimental data have indeed confirmed computational predictions for many GPCR models [19–21]. Recently, thanks to protein engineering, the modified structures of two human GPCRs have been solved, providing new templates suitable for homology modeling: the adenosine A2Areceptor (A2AR) bound to the high-affinity antagonist ZM241385 (PDB code 3EML) ; the β2-adrenergic receptor-Fab (β2AR-Fab) (PDB code 2R4R)  and the β2-adrenergic receptor-T4 (β2AR-T4) (PDB code 2RH1) [24, 25], both bound to their inverse agonist carazolol; the mutated β2-adrenergic receptor-(E122W)-T4 (β2AR(E122W)-T4) (PDB code 3D4S) bound to cholesterol and its partial inverse agonist timolol . In addition, the crystal structure of the turkey β1-adrenergic receptor (β1AR, PDB code 2VT4), in complex with the high-affinity antagonist cyanopindolol, has been also solved , raising the issue of how a range of compounds with very different affinity values can bind to such closely related receptor subtypes. Finally, also the squid Rh (PDB code 2Z73)  has been determined. Analysis of the newly published crystal structures of the squid Rh, the human β2AR, the turkey β1AR and the A2AR further confirm that the TM7 core is conserved among the entire GPCR superfamily. Nevertheless, structural differences have been found even within the TM bundle. Comparison between b Rh and β2AR structures shows that the binding site of the ligand carazolol on β2AR is very similar to that of retinal on rhodopsin, despite the fact that carazolol is a diffusible ligand rather than a covalently-bound ligand like retinal. In contrast to the β-adrenergic ligands and retinal, the A2AR antagonist ZM241385, exhibits a significantly different orientation within the TM bundle. Interestingly, the bound A2AR ligand, while interacting with helices, gets in contact also with EL2 and EL3. The publication of such new structures allows a very detailed assessment on the reliability of models based only on ground state the b Rh: this is a unique GPCR, because of its light-induced activation mechanism driven by the cis/trans isomerization of its covalently-bound ligand. However, its structure has been solved with high resolution, in different crystallization environment, in different states and both with different methodologies (NMR and X-ray). In fact, a detailed analysis of the structure differences connected to crystal packing and binding states reveals that, in spite of the close similarity to the b Rh general architecture, mutual rearrangement of the helices involved in the activation mechanism are observed. Recently, also the crystal structure of the native retinal-free GPCR bovine opsin (b Ops) has been solved : the breakage of the so-called ionic lock restraints the helical pack in the resting structure and allows a rotation along the axes of the helical bundle . In our previous study, we utilized the rhodopsin-based model of GPR17, with or without ligands, and embedded in fully hydrated phospholipid bilayer. This model was then refined by means of docking combined with molecular mechanics (MM) and molecular dynamics techniques (MD). Our MD simulations on the rhodopsin-based model of GPR17 suggested that the primary nucleotide binding pocket in GPR17 is contained in an accessible crevice enclosed between transmembrane (TM) helices (mainly TM3, TM5, TM6 and TM7) and extracellular loop (EL) 2, in general agreement with the binding site proposed for small molecules to other class A rhodopsin-like GPCRs and for nucleotides to already known P2Y receptors [31–33]. Based on our computational data, we also hypothesized that at the extracellular interface of the receptor, the N-terminus (Nt) region and EL2 and EL3 form accessory binding surfaces that could address ligands to the deeper main binding pocket. We finally proposed that the driving force for binding of nucleotides to GPR17 was the electrostatic interaction between the phosphate groups of incoming ligand and the basic arginine residue at position 6.55 (See Ballesteros and Weinstein's numbering system for residues index ) that was a recurrent target for all the nucleotidic ligands docked in our GPR17 model. This residue belongs to the conserved motif H-X-X-R/K typical of all the related P2Y and CysLT receptors; this motif is commonly believed to be a key extracellular recognition for nucleotides since 1995, when the first hypothesis on nucleotides binding mode on P2Y1 was formulated [35–40]. The overall configuration of the identified binding pocket shares common features with the ones described for the P2Y receptors [33, 41], albeit showing some interesting differences.
For the P2Y receptors, it's today commonly accepted that the driving force attracting nucleotides is provided by a triplet of conserved positively charged residues, buried in the TM bundle of the receptors: these are believed to interact with the negative charges of the phosphate groups of nucleotides [35–37]. For the P2Y1-subgroup, residues R3.29 (TM3), R/K6.55 (TM6) and R7.39 (TM7) have been proposed to be critical for nucleotide recognition. Between the three residues, only 6.55 is conserved as a basic one among all the P2Y receptor family members, and belongs to H-X-X-R/K motif cited above, whereas R3.29 and R7.39 are only typical of the P2Y1-subgroup. The last residue belongs to the Y-Q/K-X-X-R motif in TM7 and is shared by P2Y1, P2Y2, P2Y4, P2Y6 and P2Y11[3, 42]. In the P2Y12-subgroup, it was proposed that two lysines, one located in EL2 (immediately before the cysteine residue involved in the formation of the conserved disulphide bridge), and the other one located in TM7 at position 7.35 (belonging to the K-E-X-X-L motif conserved among P2Y12, P2Y13, and P2Y14), can account for cationic coordination of the phosphate moiety instead of residues R3.29 and R7.39. Furthermore, the residue close to the conserved cysteine in EL2 appears to be involved in interactions with the phosphates also for the the P2Y1-like receptors [33, 41]. More recently, a variant of this model has been proposed for P2Y14, where two of the basic residues (6.55 and 7.35) are instead assumed to bind the hexose moiety of sugar-nucleotides . Interestingly, multiple sequence alignment of GPR17 with P2Y family members showed that GPR17 lacks the basic triplet and the residue 6.55 binding the phosphates is the only one conserved in the putative pocket , despite the many positively charged amino acids typical of this peculiar receptor. We also found that the binding pocket appears to be shared by both nucleotide agonists and antagonists, even if the modality of binding differs in some details, highlighting a heterogeneity in the binding pocket recently arisen also within the P2Y receptor family.
Unfortunately, no definitive 3D model of any CysLT receptors complexed with their ligands has been proposed yet; so, a convincing hypothesis on the basis of recognition is currently unavailable, despite several anti-leukotriene agents are already available in the market and others have already successfully started their track in drug development trials. In this respect, the CysLT1 receptor antagonist zafirlukast (Accolate) was the first CysLT receptor antagonist to be marketed in the USA; montelukast (Singulair) has been introduced to market since 1998 in the treatment of asthma and allergic rhinitis ; pranlukast is still waiting for a global extension of its commercialization and it is currently available only in Japan [7, 45, 46].
In the present paper, to get more insight into the role of residues suggested to be crucial for the recognition mechanism by our previous computational data, the basic residue R6.55 of our GPR17 wild-type (WT) receptor model has been mutated to isoleucine, giving a mutant (R255I) receptor model of GPR17. The effects of this mutation on recognition nucleotides have been studied in silico, by simulating the "unbinding processes" of two docked ligands (the endogenous purinergic agonist the UDP and the leukotriene receptor antagonist pranlukast) from both the wild-type (WT) and the mutant (R255I) receptor model of GPR17 using steered MD (SMD) simulations. The comparison between the two simulations clearly shows that the energy required to force the unbinding of UDP from the WT receptor model was significantly higher than the work spent for the unbinding of the ligand from the R255I receptor. These data suggest that the same target residue (R255) could play a different role in either the recognition of distinct classes of ligands or in the modulation of receptor's activity when activated by ligands. Although the in silico hypothesis presented here still has to be confirmed experimentally, it represents an interesting starting point for in vitro validation. For example, according to our computational hypothesis, the actual involvement of the residue R6.55 in recognition of nucleotide phosphates has been also confirmed by experimental data recently produced by our group. Using a frontal affinity chromatographic-based method coupled to a mass spectrometry detection (FAC-MS), we evaluated the elution time of UDP and other nucleotide-derivative ligands on two chromatographic columns where cell membrane expressing both the native and the mutated form of GPR17 were entrapped on the surface of the stationary phase. For the natural agonist UDP, we found that the retention time on the WT receptor-containing column was higher than for the mutate receptor-containing column, suggesting that the lack of R255 may reduce the affinity for this ligand (unpublished data). In the present paper, we report the results obtained by applying the same computational approach to simulate the forced unbinding of the leukotriene receptor antagonist pranlukast, in order to investigate if the mutation affects the binding of pranlukast and if the putative target R255 is shared by the two molecules. At present, the study of the unbinding processes at atomic scale is available with the use of atomic force microscopy (AFM) , where external forces are applied to molecules to probe their mechanical resistance. The virtual mimics of such experiments are provided by steered MD (SMD) and constant force MD (CFMD), that mimic the so-called force-ramp and the force-clamp methods used in AFM, respectively. In the force-ramp method the mechanical resistance of biomolecules is measured applying a time-dependent force , while in the force-clamp methods a constant force is used . With the use of such external forces, the MD path becomes irreversible and gives access to processes involving non-covalent bonds that cannot be achieved in the same time scale with the conventional MD simulations . Based on SMD/CFMD, several reliable predictions of binding/unbinding [51–58] and folding/unfolding [59–65] processes have been obtained for various biological complexes. In unbinding experiments, the analysis of the interactions of dissociating ligands and the evolution of applied forces and ligand positions provide qualitative information about the irreversible work spent in the unbinding process: in this way, insights in structural features of receptor-ligand complexes and possible binding pathways are gained. However, our propose here was not to use SMD to define the exact ligand unbinding pathway/mechanisms, an issue that would require a more accurate analysis, but, as already mentioned before, to elucidate the role of R255 as possible target for GPR17 ligands. In fact, the simulations of the unbinding of ligands, such as UDP and pranlukast, from GPR17 receptor models presented here can also provide some attractive hypothesis on the unknown recognition mechanism and could thus be helpful to the planning of experimental mutagenesis studies and ligand affinity measurements.