Linking pathway gene expressions to the growth inhibition response from the National Cancer Institute's anticancer screen and drug mechanism of action

Novel strategies are proposed to quantitatively analyze and relate biological pathways to drug responses using gene expression and small-molecule growth inhibition data (GI(50)) derived from the National Cancer Institute's 60 cancer cells (NCI(60)). We have annotated groups of drug GI(50) responses with pathways defined by the Kyoto Encyclopedia of Genes and Genomes (KEGG) and BioCarta, and functional categories defined by Gene Ontology (GO), through correlations between pathway gene expression patterns and drug GI(50) profiles. Drug-gene-pathway relationships may then be utilized to find drug targets or target-specific drugs. Significantly correlated pathways and the gene products involved represent interesting targets for further exploration, whereas drugs that are significantly correlated with only certain pathways are more likely to be target specific. Separate pathway clustering finds that pathways engaged in the same biological process tend to have similar drug correlation patterns. The biological and statistical significances of our method are established by comparison to known small-molecule inhibitor-gene target relationships reported in the literature and by standard randomization procedures. The results of our pathway, gene expression and drug-induced growth inhibition associations, can serve as a basis for proposing testable hypotheses about potential anticancer drugs, their targets, and mechanisms of action.

Screening the molecular surface of human anticoagulant protein C: a search for interaction sites

Protein C (PC), a 62 kDa multi-modular zymogen, is activated to an anticoagulant serine protease (activated PC or APC) by thrombin bound to thrombomodulin on the surface of endothelial cells. PC/APC interacts with many proteins and the characterisation of these interactions is not trivial. However, molecular modelling methods help to study these complex biological processes and provide basis for rational experimental design and interpretation of the results. PC/APC consists of a Gla domain followed by two EGF modules and a serine protease domain. In this report, we present two structural models for full-length APC and two equivalent models for full-length PC, based on the X-ray structures of Gla-domainless APC and of known serine protease zymogens. The overall elongated shape of the models is further cross-validated using size exclusion chromatography which allows evaluation of the Stokes radius (rs for PC = 33.15 A; rs for APC = 34.19 A), frictional ratio and axial ratio. We then propose potential binding sites at the surface of PC/APC using surface hydrophobicity as a determinant of the preferred sites of intermolecular recognition. Most of the predicted binding sites are consistent with previously reported experimental data, while some clusters highlight new regions that should be involved in protein-protein interactions.

Iterative sequence/secondary structure search for protein homologs: comparison with amino acid sequence alignments and application to fold recognition in genome databases

MOTIVATION

Sequence alignment techniques have been developed into extremely powerful tools for identifying the folding families and function of proteins in newly sequenced genomes. For a sufficiently low sequence identity it is necessary to incorporate additional structural information to positively detect homologous proteins. We have carried out an extensive analysis of the effectiveness of incorporating secondary structure information directly into the alignments for fold recognition and identification of distant protein homologs. A secondary structure similarity matrix based on a database of three-dimensionally aligned proteins was first constructed. An iterative application of dynamic programming was used which incorporates linear combinations of amino acid and secondary structure sequence similarity scores. Initially, only primary sequence information is used. Subsequently contributions from secondary structure are phased in and new homologous proteins are positively identified if their scores are consistent with the predetermined error rate.

RESULTS

We used the SCOP40 database, where only PDB sequences that have 40% homology or less are included, to calibrate homology detection by the combined amino acid and secondary structure sequence alignments. Combining predicted secondary structure with sequence information results in a 8-15% increase in homology detection within SCOP40 relative to the pairwise alignments using only amino acid sequence data at an error rate of 0.01 errors per query; a 35% increase is observed when the actual secondary structure sequences are used. Incorporating predicted secondary structure information in the analysis of six small genomes yields an improvement in the homology detection of approximately 20% over SSEARCH pairwise alignments, but no improvement in the total number of homologs detected over PSI-BLAST, at an error rate of 0.01 errors per query. However, because the pairwise alignments based on combinations of amino acid and secondary structure similarity are different from those produced by PSI-BLAST and the error rates can be calibrated, it is possible to combine the results of both searches. An additional 25% relative improvement in the number of genes identified at an error rate of 0.01 is observed when the data is pooled in this way. Similarly for the SCOP40 dataset, PSI-BLAST detected 15% of all possible homologs, whereas the pooled results increased the total number of homologs detected to 19%. These results are compared with recent reports of homology detection using sequence profiling methods.

AVAILABILITY

Secondary structure alignment homepage at http://lutece.rutgers.edu/ssas

CONTACT

anders@rutchem.rutgers.edu; ronlevy@lutece.rutgers.edu

SUPPLEMENTARY INFORMATION

Genome sequence/structure alignment results at http://lutece.rutgers.edu/ss_fold_predictions.

Simplified amino acid alphabets for protein fold recognition and implications for folding

Protein design experiments have shown that the use of specific subsets of amino acids can produce foldable proteins. This prompts the question of whether there is a minimal amino acid alphabet which could be used to fold all proteins. In this work we make an analogy between sequence patterns which produce foldable sequences and those which make it possible to detect structural homologs by aligning sequences, and use it to suggest the possible size of such a reduced alphabet. We estimate that reduced alphabets containing 10-12 letters can be used to design foldable sequences for a large number of protein families. This estimate is based on the observation that there is little loss of the information necessary to pick out structural homologs in a clustered protein sequence database when a suitable reduction of the amino acid alphabet from 20 to 10 letters is made, but that this information is rapidly degraded when further reductions in the alphabet are made.

Cooperative folding units of escherichia coli tryptophan repressor

A previously published computational procedure was used to identify cooperative folding units within tryptophan repressor. The theoretical results predict the existence of distinct stable substructures in the protein chain for the monomer and the dimer. The predictions were compared with experimental data on structure and folding of the repressor and its proteolytic fragments and show excellent agreement for the dimeric form of the protein. The results suggest that the monomer, the structure of which is currently unknown, is likely to have a structure different from the one it has within the context of the highly intertwined dimer. Application of this method to the repressor monomer represents an extension of the computations into the realm of evaluating hypothetical structures such as those produced by threading.

Synthesis and biological properties of novel pyridinioalkanoyl thiolesters (PATE) as anti-HIV-1 agents that target the viral nucleocapsid protein zinc fingers

Nucleocapsid p7 protein (NCp7) zinc finger domains of the human immunodeficiency virus type 1 (HIV-1) are being developed as antiviral targets due to their key roles in viral replication and their mutationally nonpermissive nature. On the basis of our experience with symmetrical disulfide benzamides (DIBAs; Rice et al. Science 1995, 270, 1194-1197), we synthesized and evaluated variants of these dimers, including sets of 4,4'- and 3,3'-disubstituted diphenyl sulfones and their monomeric benzisothiazolone derivatives (BITA). BITAs generally exhibited diminished antiviral potency when compared to their disulfide precursors. Novel, monomeric structures were created by linking haloalkanoyl groups to the benzamide ring through -NH-C(=O)- (amide) or -S-C(=O)- (thiolester) bridges. Amide-linked compounds generally lacked antiviral activity, while haloalkanoyl thiolesters and non-halogen-bearing analogues frequently exhibited acceptable antiviral potency, thus establishing thiolester benzamides per se as a new anti-HIV chemotype. Pyridinioalkanoyl thiolesters (PATEs) exhibited superior anti-HIV-1 activity with minimal cellular toxicity and appreciable water solubility. PATEs were shown to preferentially target the NCp7 Zn finger when tested against other molecular targets, thus identifying thiolester benzamides, and PATEs in particular, as novel NCp7 Zn finger inhibitors for in vivo studies.

Topological studies of the amino terminal modules of vitamin K-dependent protein S using monoclonal antibody epitope mapping and molecular modeling

Protein S is an important anticoagulant protein acting as cofactor to activated protein C (APC) in the degradation of membrane-bound factors Va and VIIIa. Binding of protein S to the membrane depends on the Gla-domain, whereas sites for APC-interaction are located in the thrombin-sensitive region (TSR) and the first EGF domain. The aims of the present investigation were to localize the sites on protein S which are involved in APC-cofactor function and to elucidate possible orientations of the TSR in relation to the membrane. For these purposes, we determined the epitope for a calcium-dependent monoclonal antibody (HPS67) against the TSR, which inhibits APC cofactor activity even though it does not impede protein S binding to the membrane. HPS67 did not recognize wild-type mouse protein S but gained reactivity against a recombinant mouse protein in which G49 and R52 were mutated to R and Q (found in human protein S), respectively, suggesting these two residues to be part of a surface exposed epitope for HPS67. This information helped in the validation and refinement of the structural model for the Gla-TSR-EGF1-modules of protein S. The X-ray structure of a Fab-fragment mimicking HPS67 was docked onto the protein S model. The observation that HPS67 did not inhibit phospholipid binding of protein S has implications for the possible orientation of protein S on the membrane surface. In the proposed model for membrane-bound protein S, there is no contact between the TSR and the membrane. Rather, the TSR is free to interact with membrane-bound APC.

A cooperative folding unit in HIV-1 protease. Implications for protein stability and occurrence of drug-induced mutations

We investigated the HIV-1 protease molecule for the occurrence of cooperative folding units, i.e. structural units that exhibit a relatively stronger protection against unfolding than do other parts of the molecule. Calculated unfolding penalties are used to delineate folding units. This procedure identifies a folding core in HIV-1 protease, based on an ensemble of denatured states derived from native structures, comprising a spatially close unit of residues 84-91, 74-78 and 22-32, the last of which contains the active site residues D25, T26 and G27. Observed enzyme mutations of HIV-1 protease, either naturally occurring or induced by drug therapy, are found in regions that are not structurally designed to withstand unfolding. These mutations are especially likely to occur in the flap region, a part of the protein which is not essential for the stability of the protein, but does contribute significantly to the stability of protease-drug complexes. A similar avoidance of structurally protected regions in the reverse transcriptase enzyme is also observed.

The C4b-binding protein-protein S interaction is hydrophobic in nature

C4b-binding protein (C4BP) is a major regulatory molecule of the complement system. By forming a non covalent complex with the anticoagulant cofactor protein S (PS), it also plays an important role in blood coagulation. C4BP is composed of one beta-chain and seven alpha-chains that are essentially built from complement control protein (CCP)-modules. Our group has previously reported that the first (N-terminal) CCP module of the beta-chain (betaCCP1) contains the entire binding site for protein S. We now investigate further the binding of protein S to C4BP and show that the complex formation is essentially dependent on hydrophobic forces with minor contribution from electrostatic interactions. This result is in agreement with homology modeling experiments carried out in conjunction with inter-species sequence comparison and theoretical enumeration of potential binding sites. These methods pinpoint a solvent exposed hydrophobic cluster at the surface of the betaCCP1 module that is of crucial importance for the binding process.

Structural investigation of C4b-binding protein by molecular modeling: localization of putative binding sites

C4b-binding protein (C4BP) contributes to the regulation of the classical pathway of the complement system and plays an important role in blood coagulation. The main human C4BP isoform is composed of one beta-chain and seven alpha-chains essentially built from three and eight complement control protein (CCP) modules, respectively, followed by a nonrepeat carboxy-terminal region involved in polymerization of the chains. C4BP is known to interact with heparin, C4b, complement factor I, serum amyloid P component, streptococcal Arp and Sir proteins, and factor VIII/VIIIa via its alpha-chains and with protein S through its beta-chain. The principal aim of the present study was to localize regions of C4BP involved in the interaction with C4b, Arp, and heparin. For this purpose, a computer model of the 8 CCP modules of C4BP alpha-chain was constructed, taking into account data from previous electron microscopy (EM) studies. This structure was investigated in the context of known and/or new experimental data. Analysis of the alpha-chain model, together with monoclonal antibody studies and heparin binding experiments, suggests that a patch of positively charged residues, at the interface between the first and second CCP modules, plays an important role in the interaction between C4BP and C4b/Arp/Sir/heparin. Putative binding sites, secondary-structure prediction for the central core, and an overall reevaluation of the size of the C4BP molecule are also presented. An understanding of these intermolecular interactions should contribute to the rational design of potential therapeutic agents aiming at interfering specifically some of these protein-protein interactions.

Correlation between native-state hydrogen exchange and cooperative residue fluctuations from a simple model

Recently, we developed a simple analytical model based on local residue packing densities and the distribution of tertiary contacts for describing the conformational fluctuations of proteins in their folded state. This so-called Gaussian network model (GNM) is applied here to the interpretation of experimental hydrogen exchange (HX) behavior of proteins in their native state or under weakly denaturing conditions. Calculations are performed for five proteins: bovine pancreatic trypsin inhibitor, cytochrome c, plastocyanin, staphylococcal nuclease, and ribonuclease H. The results are significant in two respects. First, a good agreement is reached between calculated fluctuations and experimental measurements of HX despite the simplicity of the model and within computational times 2 or 3 orders of magnitude faster than earlier, more complex simulations. Second, the success of a theory, based on the coupled conformational fluctuations of residues near the native state, to satisfactorily describe the native-state HX behavior indicates the significant contribution of local, but cooperative, fluctuations to protein conformational dynamics. The correlation between the HX data and the unfolding kinetics of individual residues further suggests that local conformational susceptibilities as revealed by the GNM approach may have implications relevant to the global dynamics of proteins.

Identification of cooperative folding units in a set of native proteins

Cooperative unfolding penalties are calculated by statistically evaluating an ensemble of denatured states derived from native structures. The ensemble of denatured states is determined by dividing the native protein into short contiguous segments and defining all possible combinations of native, i.e., interacting, and non-native, i.e., non-interacting, segments. We use a novel knowledge-based scoring function, derived from a set of non-homologous proteins in the Protein Data Bank, to describe the interactions among residues. This procedure is used for the structural identification of cooperative folding cores for four globular proteins: bovine pancreatic trypsin inhibitor, horse heart cytochrome c, French bean plastocyanin, and staphylococcal nuclease. The theoretical folding units are shown to correspond to regions that exhibit enhanced stability against denaturation as determined from experimental hydrogen exchange protection factors. Using a sequence similarity score for related sequences, we show that, in addition to residues necessary for enzymatic function, those amino acids comprising structurally important folding cores are also preferentially conserved during evolution. This implies that the identified folding cores may be part of an array of fundamental structural folding units.

Analysis of protein-protein interactions and the effects of amino acid mutations on their energetics. The importance of water molecules in the binding epitope

A modeling analysis has been conducted to assess the determinants of binding strength and specificity for three crystal complexes; the anti-hen egg white lysozyme antibody D1.3 complexed with hen egg white lysozyme (HEL), the D1.3 antibody complexed with the anti-lysozyme antibody E5.2, and barnase complexed with barstar. The strengths of individual binding components within these interfaces are evaluated using a model of binding free energy that is based on pairwise surface preferences. In all cases the energetics of binding are dominated by a relatively small number of interfacial residues that define the binding epitope. A precise geometric arrangement of these residues was not found; they were either localized to one region, or distributed throughout the binding interface. Surprisingly, interfacial crystal water molecules were calculated to contribute around 25% of the total calculated binding strength. Theoretical alanine mutations were completed by atomic deletions of the wild-type complexes. Strong correlations were observed between the calculated changes in binding free energy (deltadeltaG(calculated)) and the experimental values (deltadeltaG(observed)) for all but three of the 30 single residue mutations in the D1.3-HEL, D1.3-E5.2 and barnase-barstar systems and for all of the double mutations in the barnase-barstar system. This analysis finds that the observed differences in binding strength are consistent with a model that accounts for the changes in binding energy from the direct contacts between each member of the complex and indirect changes due to released crystallographic water molecules that are near the mutation site. The observed energy changes for double mutations in the barnase-barstar system is fully accounted for by considering water molecules bound jointly by each member of the complex.

Azodicarbonamide inhibits HIV-1 replication by targeting the nucleocapsid protein

Nucleocapsid p7 (NCp7) proteins of human immunodeficiency virus type 1 (HIV-1) contain two zinc binding domains of the sequence Cys-(X)2-Cys-(X)4-His-(X)4-Cys (CCHC). The spacing pattern and metal-chelating residues (3 Cys, 1 His) of these nucleocapside CCHC zinc fingers are highly conserved among retroviruses. These CCHC domains are required during both the early and late phases of retroviral replication, making them attractive targets for antiviral agents. toward that end, we have identified a number of antiviral chemotypes that electrophilically attack the sulfur atoms of the zinc-coordinating cysteine residues of the domains. Such nucleocapside inhibitors were directly virucidal by preventing the initiation of reverse transcription and blocked formation of infectious virus from cells through modification of CCHC domains within Gag precursors. Herein we report that azodicarbonamide (ADA) represents a new compound that inhibits HIV-1 and a broad range of retroviruses by targeting the the nucleocapsid CCHC domains. Vandevelde et al. also recently disclosed that ADA inhibits HIV-1 infection via an unidentified mechanism and that ADA was introduced into Phase I/II clinical trials in Europe for advanced AIDS. These studies distinguish ADA as the first known nucleocapsid inhibitor to progress to human trials and provide a lead compound for drug optimization.

Atomic-scale analysis of the solvation thermodynamics of hydrophobic hydration

Molecular dynamics simulations are used to model the transfer thermodynamics of krypton from the gas phase into water. Extra long, nanosecond simulations are required to reduce the statistical uncertainty of the calculated "solvation" enthalpy to an acceptable level. Thermodynamic integration is used to calculate the "solvation" free energy, which together with the enthalpy is used to calculate the "solvation" entropy. A comparison series of simulations are conducted using a single Lennard-Jones sphere model of water to identify the contribution of hydrogen bonding to the thermodynamic quantities. In contrast to the classical "iceberg" model of hydrophobic hydration, the favorable enthalpy change for the transfer process at room temperature is found to be due primarily to the strong van der Waals interaction between the solute and solvent. Although some stabilization of hydrogen bonding does occur in the solvation shell, this is overshadowed by a destabilization due to packing constraints. Similarly, whereas some of the unfavorable change in entropy is attributed to the reduced rotational motion of the solvation shell waters, the major component is due to a decrease in the number of positional arrangements associated with the translational motions.

Evaluation of selected chemotypes in coupled cellular and molecular target-based screens identifies novel HIV-1 zinc finger inhibitors

Conservation of the Cys-Xaa2-Cys-Xaa4-His-Xaa4-Cys retroviral zinc finger sequences and their absolute requirement in both the early and late phases of retroviral replication make these chemically reactive structures prime antiviral targets. We recently reported that select 2,2'-dithiobisbenzamides (DIBAs) chemically modify the zinc finger Cys residues, resulting in release of zinc from the fingers and inhibition of HIV replication. In the current study we surveyed 21 categories of disulfide-based compounds from the chemical repository of the National Cancer Institute for their capacity to act as retroviral zinc finger inhibitors. Aromatic disulfides that exerted anti-HIV activity tended to cluster in the substituted aminobenzene, benzoate, and benzenesulfonamide disulfide subclasses. Only one thiuram derivative exerted moderate anti-HIV activity, while a number of nonaromatic thiosulfones and miscellaneous disulfide congeners were moderately antiviral. Two compounds (NSC 20625 and NSC 4493) demonstrated anti-cultures. The two compounds chemically modified the p7NC zinc fingers in two separate in vitro assays, and interatomic surface molecular modeling docked the compounds efficiently but differentially into the zinc finger domains. The combined efforts of rational drug selection, cell-based screening, and molecular target-based screening led to the identification of zinc finger inhibitors that can now be optimized by medicinal chemistry for the development of biopharmaceutically useful anti-HIV agents.

Docking enzyme-inhibitor complexes using a preference-based free-energy surface

We present a docking scheme that utilizes both a surface complementarity screen as well as an energetic criterion based on surface area burial. Twenty rigid enzyme/inhibitor complexes with known coordinate sets are arbitrarily separated and reassembled to an average all-atom rms (root mean square) deviation of 1.0 A from the native complexes. Docking is accomplished by a hierarchical search of geometrically compatible triplets of surface normals on each molecule. A pruned tree of possible bound configurations is built up using successive consideration of larger and larger triplets. The best scoring configurations are then passed through a free-energy screen where the lowest energy member is selected as the predicted native state. The free energy approximation is derived from observations of surface burial by atom pairs across the interface of known enzyme/inhibitor complexes. The occurrence of specific atom-atom surface burial, for a set of complexes with well-defined secondary structure both in the bound and unbound states, is parameterized to mimic the free energy of binding. The docking procedure guides the inhibitor into its native state using orientation and distance-dependent functions that reproduce the ideal model of free energies with an average rms deviation of 0.9 kcal/mol. For all systems studied, this docking procedure identifies a single, unique minimum energy configuration that is highly compatible with the native state.

On the origins of the hydrophobic effect: observations from simulations of n-dodecane in model solvents

The importance of the small size of a water molecule as contributing to the hydrophobic effect is examined from simulations of n-dodecane in different solvents. The earlier observations of the origin of hydrophobicity, derived from cavity formations by Pratt and Pohorille (1992, Proc. Natl. Acad. Sci. USA. 89:2995-2999) and Madan and Lee (1994, Biophys. Chem, 51:279-289), are shown to be largely consistent for a hydrocarbon-induced water pocket. In effect, the small size of a water molecule limits the probability (and hence free energy) of finding an appropriate void in the fluid that will accommodate a solute. In this work a simulated collapse of an n-dodecane molecule in H2O, CCl4, and a water-like Lennard-Jones solvent indicates that the induced entropy and enthalpy changes are qualitatively similar for hydrogen-bonded and Lennard-Jones water solvents. These results suggest that a large part of the hydrophobic response of solutes in aqueous solutions is due to the small size of the solvent. Important quantitative differences between the studied water solvents indicate that the hydrogen-bonded properties for water are still needed to determine the overall hydrophobic response.

A preference-based free-energy parameterization of enzyme-inhibitor binding. Applications to HIV-1-protease inhibitor design

The interface between protein receptor-ligand complexes has been studied with respect to their binary interatomic interactions. Crystal structure data have been used to catalogue surfaces buried by atoms from each member of a bound complex and determine a statistical preference for pairs of amino-acid atoms. A simple free energy model of the receptor-ligand system is constructed from these atom-atom preferences and used to assess the energetic importance of interfacial interactions. The free energy approximation of binding strength in this model has a reliability of about +/- 1.5 kcal/mol, despite limited knowledge of the unbound states. The main utility of such a scheme lies in the identification of important stabilizing atomic interactions across the receptor-ligand interface. Thus, apart from an overall hydrophobic attraction (Young L, Jernigan RL, Covell DG, 1994, Protein Sci 3:717-729), a rich variety of specific interactions is observed. An analysis of 10 HIV-1 protease inhibitor complexes is presented that reveals a common binding motif comprised of energetically important contacts with a rather limited set of atoms. Design improvements to existing HIV-1 protease inhibitors are explored based on a detailed analysis of this binding motif.

A simplified amino acid potential for use in structure predictions of proteins

A simplified description and a corresponding force field for polypeptides is introduced. Each amino acid residue is reduced to one interaction site, representing the backbone, and one or two side chain sites depending on its size and complexity. Site-site interactions are parameterized after a hydrophobicity criterium. The treatment of backbone sites is in addition designed to reproduce typical polypeptide hydrogen bonding patterns, as well as yielding conformations in accord with the allowed phi and psi angles through an effective angle potential. There are no explicit charges in the model. The derived energy functions, which are based on thermodynamic data and sterical consideration of allowed backbone conformations, correspond to the introduction of an effective potential. The model is tested on two small proteins, avian pancreatic polypeptide and a parathyroid hormone-related protein, by simulating folding from an initially extended state using Monte Carlo methods. The reduced amino acid description is able to satisfactorily reproduce the experimentally determined native structures.