Association and dissociation of autophagy, apoptosis and necrosis by systematic chemical study.

To address the question of whether established or experimental anticancer chemotherapeutics can exert their cytotoxic effects by autophagy, we performed a high-content screen on a set of cytotoxic agents. We simultaneously determined parameters of autophagy, apoptosis and necrosis on cells exposed to -1400 compounds. Many agents induced a 'pure' autophagic, apoptotic or necrotic phenotype, whereas less than 100 simultaneously induced autophagy, apoptosis and necrosis. A systematic analysis of the autophagic flux induced by the most potent 80 inducers of GFP-LC3 puncta among the NCI panel agents showed that 59 among them truly induced autophagy. The remaining 21 compounds were potent inducers of apoptosis or necrosis, yet failed to stimulate an autophagic flux, which were characterized as microtubule inhibitors. Knockdown of ATG7 was efficient in preventing GFP-LC3 puncta, yet failed to attenuate cell death by the agents that induce GFP-LC3 puncta. Thus there is not a single compound that would induce cell death by autophagy in our screening, underscoring the idea that cell death is rarely, if ever, executed by autophagy in human cells.

Physical interaction of apoptosis-inducing factor with DNA and RNA.

Apoptosis-inducing factor (AIF) is a mitochondrial flavoprotein, which upon apoptosis induction translocates to the nucleus where it interacts with DNA by virtue of positive charges clustered on the AIF surface. Here we show that the AIF interactome, as determined by mass spectroscopy, contains a large panel of ribonucleoproteins, which apparently bind to AIF through the RNA moiety. However, AIF is devoid of any detectable RNAse activity both in vitro and in vivo. Recombinant AIF can directly bind to DNA as well as to RNA. This binding can be visualized by electron microscopy, revealing that AIF can condense DNA, showing a preferential binding to single-stranded over double-stranded DNA. AIF also binds and aggregates single-stranded and structured RNA in vitro. Single-stranded poly A, poly G and poly C, as well double-stranded A/T and G/C RNA competed with DNA for AIF binding with a similar efficiency, thus corroborating a computer-calculated molecular model in which the binding site within AIF is the same for distinct nucleic acid species, without a clear sequence specificity. Among the preferred electron donors and acceptors of AIF, nicotine adenine dinucleotide phosphate (NADP) was particularly efficient in enhancing the generation of higher-order AIF/DNA and AIF/RNA complexes. Altogether, these data support a model in which a direct interaction of AIF contributes to the compaction of nucleic acids within apoptotic cells.

Molecular modelling probes: docking and scoring.

A general introduction to molecular modelling techniques in the area of protein-ligand interactions is given. Methods covered range from binding-site analysis to statistical treatment of sets of ligands. The main focus of this paper is on docking and scoring. After an outline of the main concepts, two specific application examples are given.

Prediction of the structure of human Janus kinase 2 (JAK2) comprising the two carboxy-terminal domains reveals a mechanism for autoregulation.

The structure of human Janus kinase 2 (JAK2) comprising the two C-terminal domains (JH1 and JH2) was predicted by application of homology modelling techniques. JH1 and JH2 represent the tyrosine kinase and tyrosine kinase-like domains, respectively, and are crucial for function and regulation of the protein. A comparison between the structures of the two domains is made and structural differences are highlighted. Prediction of the relative orientation of JH1 and JH2 was aided by a newly developed method for the detection of correlated amino acid mutations. Analysis of the interactions between the two domains led to a model for the regulatory effect of JH2 on JH1. The predictions are consistent with available experimental data on JAK2 or related proteins and provide an explanation for inhibition of JH1 tyrosine kinase activity by the adjacent JH2 domain.

Comparison of the 3D models of four different human IL-7 isoforms with human and murine IL-7.

The three-dimensional (3D) models of several alternatively spliced isoforms (ISO1 through ISO4) of human interleukin-7 (hIL-7) are presented. They are based on sequences of mRNA recently discovered in follicular dendritic cells (FDC) and primary cultures of endothelial cells or smooth muscle cells. The structures were docked to a previous model of the human IL-7 receptor, containing the IL-7 specific (IL-7R) and common gamma (gamma(c)) chain. Two different models of murine IL-7 (mIL-7) were generated as well and docked to this receptor. For an evaluation of the structures and the possible biological role of the isoforms, the models were analysed in detail and a series of enthalpy calculations was carried out. Compared with hIL-7, ISO1 appears to bind equally well to hIL-7R, but even better to the gamma(c) chain. This suggests an agonist role of ISO1, which has already been shown experimentally. The prediction that ISO2 exhibits reduced affinity to hIL-7R is supported by experiments where it had been shown to be inactive in a human test system. However, ISO2 as well as ISO3 could represent antagonists for hIL-7. Remarkably, mIL-7 appears to bind significantly less well to hIL-7R, which is in line with experimental observations that it is not active in the human system. The sequences of the isoforms support the helix assignment made for the previous hIL-7 model.

Quantitative analysis of the structural requirements for blockade of the N-methyl-D-aspartate receptor at the phencyclidine binding site.

Blockade of the N-methyl-D-aspartate receptor by uncompetitive antagonists has implications for symptomatic and neuroprotective therapy of various neuropsychiatric diseases. Since the three-dimensional (3D) structure of this ion channel is unknown, the structural requirements for uncompetitive inhibition were investigated by application of a three-step strategy: At first, Ki values were measured for a number of structurally diverse compounds at the phencyclidine (PCP) binding site in postmortem human frontal cortex. Second, a pharmacophore model was developed for this set of molecules employing a mathematical method called graph theory. The resulting pharmacophore provided a very good explanation for the ability of structurally diverse compounds to bind to the same binding site. Using the experimental data and the pharmacophore as a basis for the third step, a three-dimensional quantitative structure-activity relationship (3D-QSAR) applying comparative molecular field analysis (CoMFA) was performed. The QSAR proved to be highly consistent and showed good predictiveness for several additional molecules. The results give a deeper insight into the structural requirements for effective NMDA receptor antagonism and offer the opportunity for improved drug design. The study represents the first quantitative 3D-QSAR model for NMDA receptor blockade, and it comprises structurally very different molecules. That the alignment for a highly consistent CoMFA is based on an automated 3D pharmacophore analysis has important methodological implications.

A structural model of the human thrombopoietin receptor complex.

Thrombopoietin (TPO) is a glycoprotein hormone that regulates red blood cell production. Presented here is a modeling study of the extracellular region of the human thrombopoietin receptor complex, in particular the TPO-receptor interface. The models were developed from structural homology to other cytokines and their receptors. Experimental evidence suggests that the receptor is homodimeric and it was modeled accordingly. Key interactions are shown that correlate with previous cytokine receptor complexes, and the pattern of cysteine bonding (Cys7-Cys151 and Cys29-Cys85) agrees with that experimentally determined for thrombopoietin. These models pave the way for possible mutagenesis experimentation and the design of (ant)agonists.

Homology modeling study of the human interleukin-7 receptor complex.

Following a recent model of human interleukin-7 (IL-7), we present here a modeling study of the extracellular part of the human IL-7 receptor complex, including the IL-7 specific (IL-7R) and the common gamma (gamma c) chains. The investigation is based on structural homology to the complex of human growth hormone (hGH) bound to its receptor (hGHR). For domain 1 of IL-7R two different models are presented which differ in the alignment to hGHR in three regions. However, these differences affect binding to IL-7 in only one region, at the interface between loop EF of domain 1 of IL-7R and helix C of IL-7. The disulfide pattern in domain 1 of IL-7R is predicted to deviate from that observed in hGHR in that the C'-E disulfide (hGHR) is replaced by a C-C' cross-link. The prediction for the gamma c chain is compared with two previous studies. The models of the complex provide insight into the binding of IL-7 to its receptor and have implications for the suggestion of mutagenesis experiments and the design of (ant)agonists.

Comparative molecular field analysis of haptens docked to the multispecific antibody IgE(Lb4)

Using comparative molecular field analysis (CoMFA), three-dimensional quantitative structure-activity relationships were developed for 27 haptens which bind to the monoclonal antibody IgE(Lb4). In order to obtain an alignment for these structurally very diverse antigens, the compounds were docked to a previously modeled receptor structure using the automated docking program AUTODOCK (Goodsell, D.S.; Olson, A.J. Proteins: Struct., Funct., Genet. 1990, 8, 195-202). Remarkably, this alignment method yielded highly consistent QSAR models, as indicated by the corresponding cross-validated r2 values (0.809 for a model with carbon as probe atom, 0.773 for a model with hydrogen as probe atom). Conventional alignment failed in providing a basis for self-consistent CoMFAs. Amino acids Tyr H 50, Tyr H 52, and Trp H 95 of the receptor appeared to be of crucial importance for binding of various antigens. These findings are consistent with earlier considerations of aromatic residues being responsible for the multispecificity of certain immunoglobulins.

Prediction of the three-dimensional structure of human interleukin-7 by homology modeling.

The three-dimensional structure of human interleukin (IL)-7 has been predicted based on homology to human IL-2, IL-4, granulocyte-macrophage colony stimulating factor and growth hormone. The model has a topology common to other cytokines and displays a unique disulfide pattern. Knowledge of the tertiary structure of IL-7 has implications for analysis of key binding regions, suggestions for mutagenesis experiments and design of (ant)agonists. In this context, the model is discussed and compared with other cytokine structures.

Prediction of IgE(Lb4)-ligand complex structures by automated docking.

A mouse monoclonal anti-2,4,6-trinitrophenyl IgE (clone Lb4) was screened with a random set of over 2000 compounds, and several ligands were found to bind with affinities comparable to that of the immunizing hapten (KD in the microM range). An automated docking algorithm was used for the prediction of complex structures formed by 2,4-dinitrophenyl (DNP) and non-DNP ligands in the fragment variable region of IgE(Lb4). All ligands were found to dock in an L-shaped cavity of 15 x 16 x 10 A, surrounded by complementary-determining regions L1, L3, H2 and H3. The ligands were found to occupy the same binding site in different orientations. For rigid ligands the most stable orientation could be predicted with high probability, based on the calculated energy of binding and the occurrence frequencies of identical complexes obtained by repeated simulations. The localization of a flexible ligand (cycrimine-R) was more ambiguous, but it still docked in the same site. The results support a model for heteroligating antibody (Ab) binding sites, where different ligands utilize the total set of available contacts in different combinations. It is suggested that although pseudoenergies calculated by the docking algorithm do not correlate with experimentally measured binding energies, the screening-and-docking procedure can be useful for the mapping of Ab and other receptor binding sites ligating small molecules.

Heteroligation of a mouse monoclonal IgE antibody (La2) with small molecules, analysed by computer-aided automated docking.

A mouse monoclonal anti-TNP IgE antibody (IgE-La2) was screened by a competitive-binding ELISA with a random pool of over 2000 small molecules, mostly drugs, drug derivatives and metabolites. Thirteen of these (naproxene, beta-carboxy-chi-naphthol, oxolinic acid, hymecromone, 8-aminoquinoline, beta-naphthylamine, chi-nitrilo-cinnamic acid, 1,5-diaminonaphthaline, prolonium iodide, diaspirin, 3,4,5-trimethoxy-cinnamic acid, cycrimine, hemimellitic acid) were found to bind as strongly, or stronger, to the antibody as the immunizing hapten. We have used a Monte Carlo search technique for simulated docking of the DNP and non-DNP ligands to a model of the Fv region of IgE(La2). The validity of structural predictions made by the AutoDock program were tested on IgG(ANO2), the three-dimensional structure of which had been obtained previously by X-ray crystallography and 2D-NMR. The rms differences between the experimentally determined and auto-docked complexes in the energetically most favored binding modes were 0.31-0.44 A. Evaluation of structures of IgE(La2)-ligand complexes [including 2,4-dinitrophenol (DNP), 16 DNP amino acids, and the 13 non-DNP ligands listed above] obtained by computer-aided automated docking, suggested the existence of two subsites within an approximately 12 x 18 A2 groove extending between the H and L CDRs. Some of the ligands (DNP-Glu, 8-aminoquinoline, prolonium-I, beta-naphthylamine) were found to bind exclusively to subsite 1, others (DNP-Ala, chi-nitrilo-cinnamic acid, hemimellitic acid, beta-carboxy-chi-naphthol) to subsite 2. The majority of DNP amino acids and other ligands (oxolinic acid, 3,4,5-trimethoxy-cinnamic acid, diaspirin, [R]-cycrimine) were found to occupy an overlapping area including subsites 1 and 2, while some of the compounds (DNP-Asn, DNP-Pro, hymecromone, 1,5-naphthylenediamine) were predicted to interact with either of these subsites with comparable probabilities. When all of the docked La2-ligand complexes were taken into account, five tyrosine residues (H33, L32, L91, L92, L96) were found to provide the majority (53.4%) of all observed contact points. Thus, a multitude of interactions with aromatic residues, and a combinatorial type of interaction within the binding region, seem to be the major factors to explain the mechanism of heteroligation by IgE(La2).

Different electrostatic descriptors in comparative molecular field analysis: A comparison of molecular electrostatic and coulomb potentials.

Comparative molecular field analysis (CoMFA) is a three-dimensional quantitative structure-activity relationship (3D-QSAR) method which correlates precalculated fields surrounding a set of molecules with some target property. Among others, the electrostatic fields are commonly used. These are usually generated by calculating the Coulomb potential between a probe and the molecules bearing atom-centered point charges. The present study was performed in order to investigate up to which extent different methods to calculate electrostatic potentials can influence the results of a CoMFA. Therefore, a variety of charge calculation methods was applied to a data set consisting of 37 ligands of the benzodiazepine receptor inverse agonist-antagonist active site. These methods included Gasteiger-Marsili, semiempirical (MNDO, AM1, and PM3), and ab initio (HF/STO-3G, HF/3-21G*, and HF/6-31G*) charges. Semiempirical as well as ab initio electron populations were derived both from the Mulliken population analysis (MPA) or from fitting the charges to the molecular electrostatic potential (ESPFIT charges). In addition, the molecular electrostatic potentials (MEPs) resulting from ab initio calculations were mapped directly onto the CoMFA grid. With regard to the cross-validated r(2) values (r(2) cv ) of the resulting QSAR models, the ESPFIT-derived potentials yielded generally higher r(2) cv values than those resulting from MPA charges. For example, at the HF/3-21G* level the r(2) cv rose from 0.61 (MPA-derived potentials) to 0.76 (ESPFIT fields). The MEPs mapped directly onto the CoMFA grid were not superior to the corresponding ESPFIT-derived potentials. Semiempirical ESPFIT charges appeared to be of similar quality compared with ab initio ESPFIT electron populations in the CoMFAs. When no scaling between the steric and electrostatic descriptor matrices was applied, the electrostatic contributions were influenced to a high degree by the magnitude of the corresponding field values. © 1996 by John Wiley & Sons, Inc.

3D-quantitative structure-activity relationships of human immunodeficiency virus type-1 proteinase inhibitors: comparative molecular field analysis of 2-heterosubstituted statine derivatives-implications for the design of novel inhibitors.

A set of 100 novel 2-heterosubstituted statine derivatives inhibiting human immunodeficiency virus type-1 proteinase has been investigated by comparative molecular field analysis. In order to combine the structural information available from X-ray analyses with a predictive quantitative structure-activity relationship (QSAR) model, docking experiments of a prototype compound into the receptor were performed, and the 'active conformation' was determined. The structure of the receptor was taken from the published X-ray analysis of the proteinase with bound MVT-101, the latter compound exhibiting high structural similarity with the inhibitors investigated. The validity of the resulting QSARs was confirmed in four different ways. (1) The common parameters, namely, the cross-validated r2 values obtained by the leave-one-out (LOO) method (r2cv = 0.572-0.593), and (2) the accurate prediction of a test set of 67 compounds (q2 = 0.552-0.569) indicated a high consistency of the models. (3) Repeated analyses with two randomly selected cross-validation groups were performed and the cross-validated r2 values monitored. The resulting average r2 values were of similar magnitudes compared to those obtained by the LOO method. (4) The coefficient fields were compared with the steric and electrostatic properties of the receptor and showed a high level of compatibility. Further analysis of the results led to the design of a novel class of highly active compounds containing an additional linkage between P1' and P3'. The predicted activities of these inhibitors were also in good agreement with the experimentally determined values.

A new procedure for improving the predictiveness of CoMFA models and its application to a set of dihydrofolate reductase inhibitors.

A new automated procedure to improve the predictive quality of CoMFA models for both training and test sets is described. A model of greater consistency is generated by performing small reorientations of the underlying molecules for which too low activities are calculated. In order to predict activities of test compounds, the most similar molecules in the previously optimized model are identified and used as a basis for the prediction. This method has been applied to two independent sets of dihydrofolate reductase inhibitors (80 compounds each, serving as training sets), resulting in a significant increase of the cross-validated r2 value. For both models, the predictive r2 value for a test set consisting of 70 compounds was improved substantially.

Replacement of steric 6-12 potential-derived interaction energies by atom-based indicator variables in CoMFA leads to models of higher consistency.

The steric descriptors commonly used in CoMFA--Lennard-Jones 6-12 potential-derived interaction energies calculated between a probe atom and the molecules under investigation--have been replaced by variables indicating the presence of an atom of a particular molecule in predefined volume elements (cubes) within the region enclosing the ensemble of superimposed molecules. The resulting 'atom indicator vectors' were used as steric fields in the subsequent PLS analyses, with and without inclusion of electrostatic Coulomb interaction-derived fields. Application of this method to five training sets (80 compounds each) and five test sets (60 compounds each), randomly selected from an ensemble of 256 dihydrofolate reductase inhibitors, leads to models of significantly higher consistency, as indicated by the cross-validated r2 values for the training sets and the predictive r2 values for the test sets.