Molecular modeling of protein-ligand interactions: Detailed simulations of a biotin-streptavidin complex

The complex formed between biotin and streptavidin is the strongest noncovalent complex known in nature, and serves as a paradigm for ultra-high affinity ligand binding. We have used high-resolution x-ray crystallography, site-directed mutagenesis, calorimetry, and computer simulation methods to determine how streptavidin forms such an extraordinarily tight complex with biotin. This multi-disciplinary approach has yielded some exciting and unexpected results. The basic experimental and computational strategies used in this project should be of general utility for analysis of other high-affinity protein-ligand and protein-protein complexes. Knowledge gained from study of this system should also be useful in design of high-affinity ligands for protein targets in the future.

Molecular Modeling of Protein Structure and Function 

Molecular modeling tools can be used to supplement information about protein structure and function obtained from experimental techniques. In cases where it is impractical or even impossible to obtain direct structural information from experimental studies, molecular modeling tools can often be used to obtain useful information about protein structure and function. I will describe general modeling strategies that may be used to explore protein structure and structure-function relationships for several different categories of problems: 1) cases where there is little direct experimental data available, such as G protein-coupled receptors 2) cases where there is extensive experimental data, but no direct structural information, such as bacterial chemotaxis receptors, and 3) cases where there is experimental structural information for important protein family members, but no direct information for a specific protein of interest (E.g., class II Major Histocompatibility Complex proteins). 

Structural genomics of Thermotoga maritima - data mining and target prioritization using protein-protein interaction

Structural and functional genomics approaches require an industrialization of traditional methods in the determination of the three dimensional structures of biologicalmacromolecules. This requires the development, implementation and operation of high-throughput methodology for a structure determination pipeline addressing both, shortening the time from target identification to structure solution and number of targets processed in parallel while maintaining high data integrity quality of analysis. The Joint Center for Structural Genomics and its partners have implemented a complete high-throughput system that enables target processing from selection to sample preparation to structure determination. The completion of the development phase of the high-throughput subsystems and its full integration into a structure determination pipeline has enabled the initial processing of the Thermotoga maritima genome. Both, a description of the technology and early results from the proteome analysis of T. maritima will be discussed. Data gathering throughout the process allows for detailed analysis of experimental approaches and optimization of individual process steps.

Analysis of Protein sequences and Identification of Protein-Protein Interaction Sites

Proteins are biological “foot soldiers” in living cells, and protein-protein interactions are crucial for the existence of life itself.  Despite their elaborate structures, proteins interact with their complementary partners through a rather small but specific portion of their surface.  In the post-genomic era, sequence analysis is one of the crucial and daunting tasks.  Identification of these interaction sites and understanding the “communication” between protein partners in a specific physiologic system is one of the final goals in structural biology, proteomics and chemical biology.  Recently, we systematically examined the flanking segments of over 1600 protein-protein interaction sites. This survey indicated that proline residues are commonly found in these flanking segments and the probability of occurrence in flanking segments is 2.5 times greater than elsewhere in the molecule.  Based on these observations, we proposed a structural role for proline residues in protecting the conformation and integrity of the interaction site by blocking the “invasion” of neighboring secondary structures.  They also help in presenting the sites to their complementary proteins.  As a corollary, we have developed methods (a) to design potent bioactive peptides and (b) to identify protein-protein interaction sites directly from the amino acid sequence.  These studies provide strong experimental evidence for the structural role of proline residues in the flanking segments of protein-protein interaction sites. 

The Three Finger Toxin Toolkit

Many venom components are invaluable in molecular, biochemical and biomedical research due to their specificity and potency.  Facilitating this is the tremendous natural variation in venom components between species or even within a species itself. This is because the variation in snake venom components apparently results from frequent gene duplication of toxin-encoding genes, which is sometimes followed by functional and structural diversification and accelerated rates of sequence evolution. The three finger toxins family of snake venom peptides is particularly a good source that shows significant functional diversity through seemingly rapid rate of mutation.   In this study, we carried out a phylogenetic analysis of 276 sequences from this family.  The results revealed a diversity in the toxins far greater than has been previously realised.  A substantial number of the toxins did not fall within any of the toxin clades with characterised activity, and further lacked the functional motifs of those groups. We identified twenty such "orphan groups".  The phylogenetic pattern revealed in the case of the three finger toxins is consistent with that expected under the birth-and-death model of gene evolution. The 'three finger toxin toolkit' constructed by this study will be useful in providing a better picture of the diversity of investigational ligands available within this important class.

Conserved Codon Composition of Ribosomal Protein-coding Genes in Escherichia coli, M. tuberculosis and Saccharomyces cerevisiae:  Lessons from Supervised Machine Learning in Functional Genomics

Genomics projects have resulted in a flood of sequence data.  Functional annotation currently relies almost exclusively on inter-species sequence comparison, and is restricted in cases of limited data from related species, and widely divergent sequences with no known homologues. Here, we demonstrate that codon composition - a fusion of codon usage bias and amino acid composition signals - can accurately discriminate, in the absence of sequence homology information, cytoplasmic ribosomal protein genes from all other genes of known function in Saccharomyces cerevisiae, Escherichia coli and Mycobacterium tuberculosis using an implementation of support vector machines, SVM^light. Analysis of these codon composition signals is instructive in determining features that confer individuality to ribosomal protein genes. Each of the sets of positively charged, negatively charged and small hydrophobic residues, as well as codon bias, contribute to their distinctive codon composition profile. The representation of all these signals are sensitively detected, combined and augmented by the SVMs to perform an accurate classification. Of special mention is an obvious outlier - yeast gene RPL22B - highly homologous to RPL22A, but employing very different codon usage, perhaps indicating non-ribosomal function. Finally, we propose that codon composition be used in combination with other attributes in gene/protein classification by supervised machine learning algorithms. 

BioWare - the data warehousing system for molecular biology

Biological databases keep growing exponentially. This growth is reflected both in the rapid growth of existing databases and in proliferation of new databases. A major concern for molecular biologists is the access to and the selection of data relevant to their research from the vast pool of biological information. The recently introduced technologies for Knowledge Discovery from Databases (KDD) and Data Mining enable the extraction of new knowledge (concepts, patterns, or explanations, among others) from the data stored in databases. An infrastructure for KDD typically requires: a) mechanism to facilitate the construction of a subject-specific data warehouse (SSDW) from the diverse data sources, b) integration of tools to enable data mining from the SSDW, c) automated updating of the SSDW, and d) the ability to easily integrate both new data sources and new analysis tools. We have designed BioWare - a system that provides a framework for biological data mining. The BioWare prototype provides three separate subsystems. First, BioWare-Retrieve extracts data from multiple public database sources and integrates them into a singular dataset. Second, BioWare-Prep provides a semi-automated mechanism to biologists for rapid annotation of database entries, including the addition of new data fields. Third, the TEMPLAR subsystem facilitates a rapid creation of a new searchable subject-specific data warehouse integrated with the searh tools. We used the BioWare system to create a database of  snake toxins.

Identification of Functional Residues in Scorpion Toxins: A Bioinformatics Approach 

An important and exciting challenge in the post-genomic era is to understand the functions of newly discovered proteins based on their structures. The main thrust is to find the common structural motifs that contribute to specific functions. Using this premise, we have identified a novel class of weak potassium channel toxins from the venom of the scorpion Heterometrus fulvipes. These toxins, k-hefutoxin1 and k-hefutoxin2, exhibit no homology to any known toxins. NMR studies indicate that k-hefutoxin1adopts a unique three-dimensional fold of two parallel helices linked by two disulfide bridges without any b-sheets. Based on the presence of the functional diad (Y⁵/K¹⁹) at a distance (6.0 ± 1.0 Å) comparable to other potassium channel toxins, we hypothesized its function as a potassium channel toxin. k-Hefutoxin1 not only blocks the voltage-gated K+-channels: Kv1.3 and Kv1.2, but also slows the activation kinetics of Kv1.3 currents - a novel feature of k-hefutoxin1 unlike other scorpion toxins, which are considered solely pore-blockers. Alanine mutants (Y5A, K19A and Y5A/K19A) failed to block the channels indicating the importance of the functional diad.

Bioinformatics approach to structure-function study of scorpion toxins

Scorpion toxins have been used as research tools for characterisation of various ion channels, preparation of vaccines and antitoxins, drugs, insecticides and in phylogenetic studies. However, multiple research groups have focused on isolating, purifying and characterising individual toxins, or small groups of toxins. Consequently, there is an increasing number of characterised scorpion toxins reported in literature and molecular databases. Current scorpion toxin data are scattered across multiple databases, or reported as long lists of aligned sequences in literature reviews. Thus, it is becoming more difficult for researchers to get an overall view of the structure-function relationship and classification of scorpion toxins. SCORPION database contains at least 300 scorpion sequences classified into defined categories based on primary sequence homology. This talk focuses on systematic bioinformatic-based approach in the study of structure-function relationship of scorpion toxins. Sequences in the defined categories were further classified into basic structure-function units so that each unit contains peptides that share same functional and structural properties. By comparing peptides that have similar structures and different functional properties, in the context of their units, putative functional sites and sequence motifs related to specific functions could be identified. These motifs will be used to built prediction tools for annotation of newly identified scorpion toxins.

Locating the polyanion-binding residues in the'sushi' domain 7 of human complement factor H

Factor H, a secretory protein comprising 20 short consensus repeat (SCR) or 'sushi' domains of about 60 amino acids each, is a regulator of the complement system, an alternate pathway in immune response. The complement-regulatory functions of factor H are targeted by its binding to polyanoins such as heparin and sialic acid, involving SCRs 7 and 20. The SCR 7 heparin-binding site was shown to be co-localized with the Streptococcus Group A M protein-binding site. We have a combination of sequence analysis of all heparin-binding domains of factor H and its closest homologues, molecular modeling of SCRs 6 and 7, and surface electrostatic potential studies. The residues implicated in heparin/sialic acid binding to SCR 7 have been localized to four regions of sequence space, containing  stretches of basic as well as histidine residues. The heparin-binding site is spatially compact and lies near the interface between SCRs 6 and 7, with residues in the interdomain linker playing an important part. Experimental mutation results of the proposed heparin-binding residues will be presented.

k-Hefutoxin1, a Novel Toxin from the Scorpion Heterometrus fulvipes with Unique Structure and Function: Importance of the Functional Diad in Potassium Channel Selectivity

An important and exciting challenge in the post-genomic era is to understand the functions of newly discovered proteins based on their structures. The main thrust is to find the common structural motifs that contribute to specific functions. Using this premise, we have identified a novel class of weak potassium channel toxins from the venom of the scorpion Heterometrus fulvipes. These toxins, k-hefutoxin1 and k-hefutoxin2, exhibit no homology to any known toxins. NMR studies indicate that k-hefutoxin1adopts a unique three-dimensional fold of two parallel helices linked by two disulfide bridges without any b-sheets. Based on the presence of the functional diad (Y⁵/K¹⁹) at a distance (6.0 ± 1.0 Å) comparable to other potassium channel toxins, we hypothesized its function as a potassium channel toxin. k-Hefutoxin1 not only blocks the voltage-gated K+-channels: Kv1.3 and Kv1.2, but also slows the activation kinetics of Kv1.3 currents - a novel feature of k-hefutoxin1 unlike other scorpion toxins, which are considered solely pore-blockers. Alanine mutants (Y5A, K19A and Y5A/K19A) failed to block the channels indicating the importance of the functional diad.

Strategies for large scale protein-protein interaction studies

Experimental approach of large scale protein:protein interaction studies remains to be challenging. With the advent of high-throughput protein production, genome-wide scale protein:protein interaction becomes feasible. This lecture will cover present strategies of high-throughput protein production and platforms for genome-wide scale protein:protein interaction such as protein microarray and mass spectrometry. Future challenges of this field will also be discussed.

Prediction of Peptide Binding to Families of Related Receptors

The Protein-Protein Interactions Database (PpiDB) has been created to help understand several functional and evolutionary relationships existing in biological knowledge. Protein-protein interactions have been predicted using the "Rosetta stone" approach described by Ed Marcotte (Eisenberg Lab, UCLA). The basic principle of this idea is that certain proteins exist as large multifunctional proteins within one species, while their corresponding functions in another species are carried out by smaller individual proteins. The larger multifunctional protein thus serves as a "Rosetta Stone" for predicting protein-protein interactions within the latter species. We have taken this approach and augmented it by basing the interaction prediction on domain information (Pfam) rather than sequence. This does result in a large amount of false positives but that is where the next set of steps is implemented.

Prediction of Peptide Binding to Families of Related Receptors

Major histocompatibility complex (MHC) molecules bind peptides and present them on cell surface for recognition by T cells of the immune system. MHC molecules are encoded by genes that show significant polymorphism. In humans, there are more than 500 characterised allelic variants of MHC for each MHC class I and class II. We developed a prediction system called MULTIPRED that was trained using virtual sequences that represent peptide-MHC interactions. The virtual sequences were constructed by combining the interaction sites from both peptide (ligand) and the receptor (MHC) inferred from the three-dimensional structure of MHC molecules. We applied the MULTIPRED system to a selection of human MHC class II  molecules HLA-DR, and to the human MHC class I superfamily HLA-A2. MULTIPRED showed high accuracy in predicting HLA-binding peptides. In addition, we have shown that MULTIPRED can accurately predict peptide binding to HLA molecules for which no binding data are available.