Tom Blundell

Protein structure and function
With the annual Barbara Bowman Lecture in Medical Genetics, Tom Blundell (Cambridge University) presented a comprehensive
overview of structural data from cell signaling complexes. Structural controls on signaling pathways cannot always be defined by the amount of surface area involved between the two proteins. Very large interfaces are often characteristic of very tight interactions, for example between Gb and Gg, which act as a single unit in the cell. However, for sensitive signaling, smaller interface areas and conformational changes are often involved, as there should not be too much of a barrier to overcome when pulling the molecules apart. All of the data presented was obtained in his group, including two recent structures of the phosducin transducin ßg and cyclin D-dependent kinase Cdk6 bound to the cell cycle inhibitor p19INK4d.

Krishna Rajarathnam, a new UTMB faculty member, showed how small differences in the sequences of the CC and CxC chemokine families can alter their specific activities without changing their overall structure. By using chemical synthesis, he incorporated cysteine analogues at points where cysteine cross linkages stabilize the structure of IL-8. Surprisingly, although these mutants greatly reduced the ability of the protein to bind to its receptor, there was almost no change in the overall structure (determined by NMR). He suggests that the basic scaffolding of this chemokine family is very stable, accounting for why it is used so often for so many different types of cell signalling, while only a few variable residues in exactly the right conformation are important for receptor binding specificity. Wlodek Bujalowski (UTMB) presented a variety of techniques to analyze the interaction of the E. coli primary replicative helicase DnaB protein with DNA substrates. One can estimate the distance between the protein hexamer and the DNA by using fluorescein labeling. Fluorescence energy transfer indicates that the ssDNA passes through the inner channel of the DnaB hexamer. A weak DNA binding subsite of the protein could be the entry site for the dsDNA, while the strong binding site for ssDNA is at the replication fork.

Analysis of protein structure and protein design
The next topic was protein design, which also included several talks on the classification and analysis of known structures. Jane Richardson (Duke University Medical Center) presented small probe contact dot analysis and a new scoring system as an additional tool to assess the quality of crystallographic structures. The new method includes explicit hydrogen atoms in the refinement process and uses a "clash score" as a criterion for structure quality. By rolling a sphere with a radius of 0.25 Å radius over all atoms of a protein, areas of bad van der Waals interactions can be identified. The clash score (clashes per 100 atoms) increases with resolution and jumps greatly for structures >2 Å. Clashes occur mostly in areas where it is difficult to read maps and are especially frequent for the side chains of glycine, glutamine and asparagine. In addition to being a good technique for validating structures, clash scores can also help correct protein and nucleic acid structures and improve rotamer libraries for amino acid side chains. Jane noted further that the super high resolution structures are a gold mine for those designing proteins, as they reveal many details of internal packing in proteins that are important for understanding conformational preferences.

The average protein can take a good deal of random mutagenesis before losing its unique native fold, while achieving a stable fold in a designed protein is very difficult. Typically, a fair amount of "negative design" is needed, which means removing residues whose interactions could drive the peptide chain toward an alternate conformation. Many of these interactions can be identified by using the contact dot system. Clash scores should also be useful for docking ligands into a known structure. Best of all, the programs and the new rotamer libraries are available free of charge (see the website kinemage.biochem.duke.edu for details).

George Phillips (Rice University) described methods to put high resolution structures into motion to aid in redesigning proteins. For example, in the crystal structure of myoglobin, there is no way for the oxygen to get into the heme site. His group began by using molecular dynamics simulations to see how the molecule moves. The group combined computer techniques with physical methods, including diffuse scattering, to get more details on the fine structure and the role of individual residues near the haem group. Computer simulations of generating a geminate state by using light to break the Fe-O2 bond can reveal much about other interactions with the ligand. Further, one can check the predictions of the calculation by using IR to follow changes in the CO stretching frequency (that bond to Fe has a partial triple bond character and absorbs in the region of 1900-2000 cm-1, well separated from the spectrum of the rest of the molecule). A linear relationship between the average stretch frequency (drops from 1920 to 1990 cm-1) and the affinity for ligand in a series of mutants, means electrostatics controls the binding affinity. They were able to convert myoglobin to a peroxidase by moving the position of one of the histidines (Phe43His/His64Leu mutant). A different mutant, with a substantially lower reactivity with nitrous oxide that still binds oxygen well, is being tested for use in blood substitutes by Baxter Hemoglobin Therapeutics.

Stephen Mayo (Caltech) presented the inverse problem: find the amino acid sequence that will best fit a predetermined fold. He noted that although one might be able to define an energy function that would be indicative of the native state, it would be impossible to calculate exhaustively the energy of all the possible conformations for even a small polypeptide. They use a simple configuration energy function and the "dead end elimination theorem" (Desmet et al., Nature 1992) to narrow the number of configurations that must be calculated. With this calculation efficient method, they can run a structure calculation in less than 40 min on 8 processors on an R10000 Silicon Graphics station. Physical tests of their scoring potential, using mutants of the core of the GCN4 dimerization surface, showed that the melting T was linearly related to the calculated van der Waals energy. This correlation was improved by adding a hydrophobic surface energy term and another for burial of hydrophobic residues. Using this design principle, they developed a mutant that resisted unfolding at 50° C up to 4M guanidinium, while the wild type protein is completely denatured by <1M. Using their methods, G-CSF mutants (12-16 h of computer time was required to alter 29-34 out of 174 residues) was prepared that had improved stability (the Tm increased from 55 to 70° C) and bound as well as the wild type to leukocyte receptors.

Click for Next Page