The human (h) P2X(3) receptor and its mutants deficient in one out of four N-glycosylation sites were expressed in HEK293 cells. Concentration-response curves were generated by whole-cell recordings of alpha,beta-methylene ATP (alpha,beta- meATP)induced currents. A gradual change of external pH from the alkaline 8.0 to the acidic 5.0 successively decreased the maximum current amplitude (E-max) without affecting the EC50 value. The replacement of Asn-139 and -170 by Asp (N139D, N170D) abolished the pH sensitivity of the wild-type (WT) hP2X(3) receptor. In the case of N194D, the E-max was again the highest at the alkaline pH value with no change from 7.4 to 6.5, whereas in the case of N290D, there was an inverse pH sensitivity, with an increase of E-max in the acidic range. However, this effect appeared to be due to enhanced protonation by the insertion of Asp into the receptor, because replacement of Asn by the neutral Thr resulted in a comparable potency of alpha,beta- meATP at any of the pH values investigated. In accordance with the reported finding that His-206 is involved in the modulation of WT P2X(3) receptors by protons, we showed that the normal change of E-max by an acidic, but not alkaline pH was abolished after substitution of this His by Ala. However, the double mutant H206A + N290D did not react to acidification or alkalinization with any change in Emax. In conclusion, only fully N-glycosylated P2X3 receptors recognize external pH with a modified sensitivity towards alpha,beta-meATP.

This review deals with three problems: the selection of suitable chemical descriptors from a pool of variables by a simultaneous one-regression/one-observation leaving-out resampling, the comparison of the results with a generalized-regression artificial-neural network, using an unconstrained genetic algorithm (GRNN), and the prediction of protein subdomains as potential molecular drug targets. As an example, the human P2X7 (h-P2X7) receptor subunit and a series of novel 4,5-diarylimidazoline inhibitors [Merriman et al., Bioorg. Med. Chem. Lett., 15, 435 (2005)] is used. GRNN ignores relevant and add noisy descriptors although the goodness-of-fit criterion is large. Therefore, GRNN is considered as supplementary tool which cannot replace the traditional QSAR methodology. Simultaneous one-regression/one-observation leaving-out resampling shows that the h-P2X7 inhibitory activity of 4,5-diarylimidazolines depends on electronic, steric and hydrogen-bonding properties of the substituents. Diagnostic statistic examines the validity of the results. The inhibitors are probably bound to sites that are located mainly in the subdomains 344-347 and 370-375 of h-P2X7.

Wild-type human (h) P2X3 receptors expressed in HEK293 cells responded to the prototypic agonist α,β-methylene ATP (α,β-meATP) with rapidly desensitizing inward currents and an increase in the intracellular Ca2+ concentration. In contrast to electrophysiological recordings, Ca2+ microfluorimetry showed a lower maximum of the concentration-response curve of α,β-meATP in the transiently than in the permanently transfected HEK293 cells. However, the concentrations causing 50% of the maximum possible effect (EC50 values) were identical, when measured with either method. In order to determine the role of certain conserved, positively charged amino acids in the nucleotide binding domains (NBD-1-4) of hP2X3 receptors for agonist binding, the lysine-63, -65, -176 and -299 as well as the arginine-281 and -295 residues were substituted by the neutral amino acid alanine. We observed no effect of α,β-meATP at the K63A, K176A, R295A, and K299A mutants, and a marked decrease of agonist potency at the K65A and R281A mutants. The P2X3 receptor antagonist 2',3'-O-trinitrophenyl-ATP (TNP-ATP) blocked the effect of α,β-meATP at the wild-type hP2X3 receptor with lower affinity than at the mutant K65A, indicating an interference of this mutation with the docking of the antagonist with its binding sites. The use of confocal fluorescence microscopy in conjunction with an antibody raised against the extracellular loop of the hP2X3 receptor documented the expression of all mutants in the plasma membrane of HEK293 cells. Eventually, we modelled the possible agonist and antagonist binding sites NBD-1-4 of the hP2X3 subunit by using structural bioinformatics. This model is in complete agreement with the available data and integrates results from mutagenesis studies with geometry optimization of the tertiary structure predictions of the receptor.

This review deals with the molecular modelling of a subtype of the membrane-embedded purinoreceptor P2X family, which belongs to the large class of membrane-embedded glycoproteins. The P2X family has two transmembrane domains and a core of five extracellularly occurring disulfide bonds. At present, seven different P2X receptor subtypes (P2X1 – X7) have been cloned. The human purinoreceptor P2X3 (h-P2X3) is a putative drug target for the development of inhibitors against chronic inflammatory, neuropathic and mixed-pain conditions. No details on P2X receptor architecture are known at the atomic resolution level by X-ray or NMR analyses. An attempt was made to predict the conformation of h-P2X3 using homology based comparative modelling and threading, but the modelling could not be carried out due to missing template proteins. State-of-the-art ab initio protein structure prediction methods also failed. A novel approach has been applied and exemplified on the h-P2X3 receptor. The coordinates of the secondary structure of h-P2X3 were determined by a profile-based neural network prediction. The conformation was geometry optimised using the quantum chemistry RHF/3-21G minimal basic set and all-atom molecular mechanics AMBER force field. A dielectric constant of ε = 3.5 was used to simulate the lipophilic environment of the membrane-bound protein. The h-P2X3 protein has a number of interacting peptide modules. An example is an extracellularly occurring triad of sterically interacting domains, which consists of a nucleotide binding domain (amino acids in positions 62 – 66), a PKC phosphorylation site (196 – 198) and a N-glycosylation attachment site (194 – 197). The discovery of this peptide module, and of other interacting modules, raises the possibility of exploiting structure-based strategies to design P2X3 inhibitors. Nevertheless, it should be noted that the predicted structures are defined in a probabilistic sense. Only biological and chemical knowledge can determine whether or not these predictions are meaningful. Thus, the results from the computational tools are probabilistic predictions and subject to further experimental verification.

There is evidence that the P2X1 receptor subunit is involved in apoptosis, platelet aggregation, and smooth muscle contraction. The conformation of the membrane-embedded, ligand-gated mouse P2X1 glycoprotein, a monovalent-bivalent cation channel-forming receptor, is predicted. The first step is based on secondary structure prediction. The secondary structure is converted into a three-dimensional geometry. Then, the secondary and tertiary structures are optimized by using the quantum chemistry RHF/3-21G minimal basic set and the all-atom molecular mechanics AMBER96 force field. The fold of the membrane-embedded protein is simulated by a suitable dielectric. The structure is refined using a conjugate gradient minimizer (Fletcher-Reeves modification of the Polak-Ribiere method). Although the mouse P2X1 receptor subunit is more complex (388 amino acids) than the KcsA protein (160 amino acids), the overall folds are similar. The geometry optimized P2X1 receptor subunit is freely available for academic researchers on e-mail request (PDB format).

There is a hope that inhibitors of the purinoreceptor P2X7, a ligand-gated ion channel, would have a beneficial role in the therapy of inflammatory processes. Therefore, the P2X7 channel is a putative drug target but neither X-ray nor NMR analyses have been made. It was attempted to predict the conformations of P2X receptor subunits using homology-based comparative modelling and threading. The modelling could not be carried out because of missing template proteins. The paper reviews a new method to predict the conformation of proteins. The method is exemplified on the human P2X7 (h-P2X7) receptor. The coordinates of the spatial structure of the h-P2X7 protein were determined by a profile-based neural network prediction. The conformation was geometry optimised using the quantum chemistry RHF/3-21G minimal basic set and allatom molecular mechanics AMBER force field. A dielectric constant of ε = 3.5 was used to simulate the lipophilic environment of the membrane-bound glycoprotein. The h-P2X7 protein has a number of interacting peptide modules. The discovery of peptide module raises the possibility of exploiting structure-based strategies to design h-P2X7 inhibitors. Keywords: Sponge metabolites, Proteomics, Drug target, Human P2X7 receptor, Ligand-gated ion channel

A major problem in traditional quantitative structure-activity relationships (QSARs) analysis is to select suitable chemical descriptors from a large pool of variables. Decisions against or in favor of a particular descriptor depends entirely on the result of statistically based hypothesis testing. Uncertain results may be produced in presence of multicollinear descriptors and flagged observations (high-leverage points, outliers, influential data). To satisfy the assumptions for hypothesis testing, diagnostic statistics and subsequent design repair are employed. Here we show an example with nonnucleoside HIV-1 reverse transcriptase inhibitors.

The membrane-embedded, ligand-gated P2X glycoprotein receptor is a monovalent-bivalent cation channel that is activated by physiological concentrations of extracellular ATP. A quantitative structure-activity relationship (QSAR) analysis was developed to model the cation permeability of the P2X2 channel and its mutants. As chemical properties, the helix-coil equilibrium constants and the distribution coefficients of the system octanol/water at pH 7.4 were applied and modified (sliding windows) according to Eroshkin et al. (Comput. Appl. Biosci., 1995, 11, 49-44). The results were visualized by a dimeric P2X2 channel construct. The results support the hypothesis that residues which put into the cavity and contribute to hydrogen bonding forces are involved to a control of the transport of hydrated cations through the P2X2 channel. The model may be useful to develop P2X2 receptor antagonists.

A novel method of structure prediction for membrane-bound proteins is reviewed. The approach is based on a sequence-function analysis, secondary structure prediction and subsequent geometry optimization. The prediction of the structure of ligand-gated P2X4 receptor subunit, which is a membrane-embedded cation channelforming protein with extracellularly occuring ATP binding sites, is shown as an example. The potential Nglycosylation sites of the glycoprotein, location of the five disulfide bridges, and phospholipid-dependent protein kinase C (PKC) phosphorylation attachment sites were determined by sequence-function analysis. Subsequently an attempt was made to predict its conformation using homology-based comparative modeling and threading; however, the modeling could not be accomplished. Because of this, secondary structure prediction of the protein was carried out. The input coordinates of the spatial structure were obtained by a profile-based neural network prediction method. The resulting secondary structure was converted into a three-dimensional geometry. The secondary and tertiary structures were optimized by the quantum chemistry RHF/3-21G minimal basic set and all-atom molecular mechanics AMBER96 force field. The predicted shape is similar to the shape of the experimentally obtained monomeric structure of the classical ion channel, the K+ion channel from Streptomyces lividans (KcsA channel), and agrees with the P2X shape proposed by biological experimenters. The geometry optimized structure of the P2X4 receptor is freely available (Protein Data Bank format) from the authors on e-mail request (magp@medizin.unileipzig. de). Keywords: Proteomics, structural bioinformatics, ATP-activated purinergic receptor, P2X receptor, cation channel, KcsA protein

The whole-cell patch-clamp technique was used to record current responses to nucleotides and nucleosides in human embryonic kidney HEK293 cells transfected with the human purinergic P2X3 receptor. When guanosine 5′-O-(3-thiodiphosphate) was included into the pipette solution, UTP at concentrations that did not alter the holding current facilitated the α,β-methylene ATP (α,β-meATP)-induced current. ATP and GTP, but not UDP or uridine, had an effect similar to that of UTP. Compounds known to activate protein kinase C (PKC) acted like the nucleoside triphosphates investigated, whereas various PKC inhibitors invariably reduced the effects of both PKC activators and UTP. The substitution by Ala of Ser/Thr residues situated within PKC consensus sites of the P2X3 receptor ectodomain either abolished (PKC2 and PKC3; T134A, S178A) or did not alter (PKC4 and PKC6; T196A, S269A) the UTP-induced potentiation of the α,β-meATP current. Both the blockade of ecto-protein kinase C activity and the substitution of Thr-134 or Ser-178 by Ala depressed the maximum of the concentration-response curve for α,β-meATP without altering the EC50 values. Molecular simulation of the P2X3 receptor structure indicated no overlap between assumed nucleotide binding domains and the relevant phosphorylation sites PKC2 and PKC3. α,β-meATP-induced currents through native homomeric P2X3 receptors of rat dorsal root ganglia were also facilitated by UTP. In conclusion, it is suggested that low concentrations of endogenous nucleotides in the extracellular space may prime the sensitivity of P2X3 receptors toward the effect of subsequently applied (released) higher agonistic concentrations. The priming effect of nucleotides might be attributable to a phosphorylation of PKC sites at the ectodomain of P2X3 receptors.

MOTIVATION:No details on P2X receptor architecture had been known at the atomic resolution level. Using comparative homology-based molecular modelling and threading, it was attempted to predict the three-dimensional structure of P2X receptors. This prediction could not be carried out, however, because important properties of the P2X family differ considerably from that of the potential template proteins. This paper reviews an alternative approach consisting of three research fields: bioinformatics, structural modelling, and a variety of the results of biological experiments.MODEL:Starting point is the amino acid sequence. Using the sequential data, the first step is a secondary structure prediction. The resulting secondary structure is converted into a three-dimensional geometry. Then, the secondary and tertiary structures are optimized by using the quantum chemistry RHF/3-21G minimal basic set and the all-atom molecular mechanics AMBER96 force field. The fold of the membrane-embedded protein is simulated by a suitable dielectricum. The structure is refined using a conjugate gradient minimizer (Fletcher-Reeves modification of the Polak-Ribiere method). The results of the geometry optimization were checked by a Ramanchandran plot, rotamer analysis, all-atom contact dots, and the C(beta) deviation. As additional tools for the model building, multiple alignment analysis and comparative sequence-function analysis were used. The approach is exemplified on the membrane-embedded, ligand-gated P2X3 receptor subunit, a monovalent-bivalent cation channel-forming glycoprotein that is activated by extracellular adenosine 5'-triphosphate. From these results, a topology of the pore-forming motif of the P2X3 receptor subunit was proposed. It is believed that a fully functional P2X channel requires a precise coupling between (i) two distinct peptide modules, an extracellularly occurring ATP-binding module and a pore module that includes a long transmembrane and short intracellular part, (ii) an interaction surface with membranes, and (iii) hydrogen bonding forces of the residues and hydrated cations. Furthermore, this paper demonstrates the role of quantitative structure-activity relationships (QSARs) in P2X research (calcium ion permeability of the wild-type and after site-directed mutagenesis of the rat P2X2 receptor protein, KN-62 analogs as competitive antagonists of the human P2X7 receptor). EXPERIMENTAL PROOFS: The predictions are experimentally testable and may provide an additional interpretation of experimental observations published in literature. In particular, there is the good agreement of the geometry optimized P2X3 structure with experimentally proposed P2X receptor models obtained by neurophysiological, biochemical, pharmacological, and mutation experiments. Although the rat P2X3 receptor subunit is more complex (397 amino acids) than the KcsA protein (160 amino acids), the overall folds of the peptide backbone atoms are similar.LIMITATIONS:To avoid semantic confusion, it should be noted that "prediction" is defined in a probabilistic sense. Matches to generic rules do not mean "this is true" but rather "this might be true". Only biological and chemical knowledge can determine whether or not these predictions are meaningful. Thus, the results from the computational tools are probabilistic predictions and subject to further experimental verification.AVAILABILITY:The geometry optimized P2X3 receptor subunit is freely available for academic researchers on e-mail request (PDB format).

GABAergic interneurons with high-frequency firing, fast-spiking (FS) cells, form synapses on perisomatic regions of principal cells in the neocortex and hippocampus to control the excitability of cortical networks. Brain-derived neurotrophic factor (BDNF) is essential for the differentiation of multiple interneuron subtypes and the formation of their synaptic contacts. Here, we examined whether BDNF, alone or in conjunction with sustained KCl-induced depolarization, drives functional FS cell differentiation and the formation of inhibitory microcircuits. Homogeneous FS cell cultures were established by target-specific isolation using the voltage-gated potassium channel 3.1 b subunit as the selection marker. Isolated FS cells expressed parvalbumin, were surrounded by perineuronal nets, formed immature inhibitory connections and generated slow action potentials at 12 days in vitro. Brain-derived neurotrophic factor (BDNF) promoted FS cell differentiation by increasing the somatic diameter, dendritic branching and the frequency of action potential firing. In addition, BDNF treatment led to a significant up-regulation of synaptophysin and vesicular GABA transporter expression, components of the synaptic machinery critical for GABA release, which was paralleled by an increase in synaptic strength. Long-term membrane depolarization alone was detrimental to dendritic branching. However, we observed that BDNF and KCl exerted additive effects, as reflected by the significantly accelerated maturation of synaptic contacts and high discharge frequencies, and was required for the formation of reciprocal connections between FS cells. Our results show that BDNF, along with membrane depolarization, is critical for FS cells to establish inhibitory circuitries during corticogenesis.

The methods of bioinformatics, molecular modelling, and quantitative structure-activity relationships (QSARs) using regression and artificial neural network (ANN) analyses were applied to develop safer aldoxime antidotes against poisoning by organophosphorus (OP) agents with high, mean, and low aging rates. We start here from a molecular modelling of the mouse AChE at an atomistic level. Aim is to predict qualitatively the structural requirements of an aldoxime that shows an unique reactivating activity against the three classes of OPs. An antidotal action should occur by a three-site mechanism: the aldoxime groups of the first pyridinium ring should point towards the catalytic site, and the second pyridinium ring and its substituents should be anchored at the peripherical and anionic subsites. Based on this model, it is predicted that a suitable substituent is based on an arginine-like moiety. Then, an ANN-based QSAR analysis using a training set of aldoximes with known structure and activities was applied. Its input layer consisted of seven nodes: the group-membership descriptors that parameterize the type of the OP, the logarithms of the distribution coefficients at pH 7.4 and their squared term, the lowest unoccupied molecular orbital (LUMO) energies, the scaled molar refractions of the substituents, and their squared term. It was shown that the qualitative prediction made by molecular modelling can be quantified by an ANN prediction.

Two different quantitative structure-activity relationship (QSAR) techniques were compared using a series of thromboxane A2 receptor antagonists (benzimidazole and imidazol[4,5- b ]pyridine acid derivatives). In both cases, binary molecular substructural indicator variables were used for encoding the substituents of a lead structure. The first technique is based on the conventional Free-Wilson (FW) analysis (ordinary least-squares regression). The second technique is based on an optimized backpropagation neural network where non-additive substituent contributions can be analyzed, too. It is shown that the two approaches may lead to an overfitting.

The effect of [2-substituted-4-(3-pyridyl)-1,3-dioxan-5-yl]alkenoic acids of thromboxane synthase was examined using a multivariate quantitative structure-activity relationship (QSAR) model. As biological parameters, the dual inhibition of thromboxane receptors (inhibition of the in vitro aggregation of human platelets) and of thromboxane synthase were used. As chemical parameters, the inductive constant, van der Waals radius, the STERIMOL parameters L and B 1 of the substituents, and the dihedral angle energy of the geometry-optimized conformations of the antagonists (hybrid Gasteiger molecular-mechanics approach) were chosen. The backpropagation neural network approach has the ability to approximate multivariate structure-activity functions with satisfactory accuracy.

Strong hydrogen-bonding forces between the Thr26 and Thr26' of the protease stabilize the internal cage of the dimeric triad of the aspartyl HIV-1 protease (Asp25Thr26Gly27 and Asp25' Thr26'Gly27', respectively). The interaction of reversible inhibitors of HIV-1 protease is based on (i) strong hydrogen-bonding forces between the main chain (--CONH--) oxygen atoms of Gly27 and/or Gly27' and hydrogen-bond donating moieties of a drug, and (ii) hydrogen bonds between the oxygen of the catalytic Asp25 and/or Asp25' carboxylates and aliphatic hydroxyl groups of a drug. The free entry of natural substrates into the active-site cavity is sterically hindered by inhibitors, so that the catalytic Asp carboxylates cannot interact with natural substrates. Irreversible inhibitors interact with the nucleophilic carboxylate moiety of Asp25 of HIV-1 protease by covalent bonding.

Beneficial antiviral HIV-1 chemotherapy is associated with adverse reactions. To optimize the desired actions and to lower the side effects of nonnucleoside HIV-1 reverse transcriptase (RT) inhibitors (NNRTIs), quantitative structure-activity relationships (QSARs) were studied by using a series of HEPT derivatives of NNRTIs. Hypothesis testing requires that certain assumptions are approximately satisfied in statistically based QSARs, however. A complementary approach is based on artificial neural network analysis. Model building can be made without the manifold assumptions of statistically based QSAR approaches but the problem is that the number of neural weights increase exponentially (danger of overfitting) under certain circumstances. A way to get more reliable results is to reduce the dimensionality of the two subsets (biological and chemical variables). A suitable method is the canonical correlation analysis. The two subsets of canonical variates are used as outputs (biologically derived variates) and inputs (chemically derived variates) of an optimized backpropagation neural network approach. The contribution summarizes the most recent results of this canonical-correlation backpropagation-neural network QSAR approach. It is shown that noncovalent interactions (lipophilic, steric, hydrogen-bonding, and inductive forces of the substituents) are responsible for the antiviral and cytotoxic actions. The outcome of this analysis produces an internally highly self-consistent result (model robustness). The predictive performance is tested. The butterfly-like conformation of the predicted compound is consistent with the butterfly-like model of other NNRTIs. Molecular simulation shows that the complexed drug interacts with the Tyr181 and Tyr188 residues of the RT. The uracil ring of the drug binds directly with Lys101, and the acyclic side chain (with an intact free hydroxyl function) binds with Lys103. The suggested noncovalent interaction forces are equivalent with that found by the QSAR analysis.

Beneficial antiviral HIV-1 chemotherapy is associated with adverse reactions. To optimize the desired actions and to lower the side effects of nonnucleoside HIV-1 reverse transcriptase (RT) inhibitors (NNRTIs), quantitative structure-activity relationships (QSARs) were studied by using a series of HEPT derivatives of NNRTIs. Hypothesis testing requires that certain assumptions are approximately satisfied in statistically based QSARs, however. A complementary approach is based on artificial neural network analysis. Model building can be made without the manifold assumptions of statistically based QSAR approaches but the problem is that the number of neural weights increase exponentially (danger of overfitting) under certain circumstances. A way to get more reliable results is to reduce the dimensionality of the two subsets (biological and chemical variables). A suitable method is the canonical correlation analysis. The two subsets of canonical variates are used as outputs (biologically derived variates) and inputs (chemically derived variates) of an optimized backpropagation neural network approach. The contribution summarizes the most recent results of this canonical-correlation backpropagation-neural network QSAR approach. It is shown that noncovalent interactions (lipophilic, steric, hydrogen-bonding, and inductive forces of the substituents) are responsible for the antiviral and cytotoxic actions. The outcome of this analysis produces an internally highly self-consistent result (model robustness). The predictive performance is tested. The butterfly-like conformation of the predicted compound is consistent with the butterfly-like model of other NNRTIs. Molecular simulation shows that the complexed drug interacts with the Tyr181 and Tyr188 residues of the RT. The uracil ring of the drug binds directly with Lys101, and the acyclic side chain (with an intact free hydroxyl function) binds with Lys103. The suggested noncovalent interaction forces are equivalent with that found by the QSAR analysis.

The effect of substitution on the phenyl ring of orally active beta-lactam inhibitors (3,3'-diethyl-2-azetidinones) of human leucoctyte elastase was examined using quantitative structure-activity relationship (QSAR) models. Tradionallly used QSARs are ordinary least-squares multiple regression analysis (Hansch), nonleast-squares (NLS) and partial-least squares (PLS) regression models. The results were compared with that of backpropagation (BP) and generalized-regression genetic-neural network (GRNN) approaches. It was found that a Hansch analysis can only be used with reservation due to the multicollinearity of the physicochemical descriptors. The problem of multicollinearity using Hansch analysis is circumvented by PLS and NLS regression. BP did not improve the goodness of fit. GRNN has the ability to approximate structure-activity functions with satisfactory accuracy.