3D Structures of Engineered Proteins
A variety of techniques can produce structural information about the engineered protein. Among the techniques available, only protein X-ray crystallography and NMR spectroscopy have the routine ability to determine directly the experimental 3-D arrangement of atoms comprising the protein at atomic resolution. A computer can be used for protein modeling, when insufficient quantities of the engi-neered protein are available, or if X-ray crystal-lography and NMR are not amenable. While a detailed discussion of each of these techniques is well beyond the scope of this chapter, a very brief introduction of each is valuable to introduce the concepts. The references cited will provide the reader with useful resources to further explore these techni-ques in detail.
Protein X-ray crystallography is a tremendously powerful technique (Stubbs II, 2007). Following formation of appropriate crystals of the protein, an X-ray diffraction pattern is obtained for the crystal. An electron density map is derived from the diffraction data, which subsequently provides the atom positions in the protein. While the X-ray structure obtained is a structure averaged over all of the mutant protein molecules found in the crystal (various subtle conformational differences among chemically identi-cal molecules of the engineered protein), the techni-que provides a view of the spatial arrangement of the protein’s atoms. Molecular interactions (i.e., ligand-protein binding) and the mechanism of catalytic reactions can be studied at the molecular level. New and exciting techniques have been developed to validate protein 3-D structure; important for modern methods of rational drug design (Joosten, et al., 2007; Laskowski and Swaminathan, 2007).
Protein X-ray crystallography is severely limited by the availability of appropriate crystals for analysis. While sufficient quantities used to be a problem, biotechnology techniques have addressed this diffi-culty (Bauer and Schnapp, 2007). Other limitations include the inability to get most hydrogen positional information and the fact that the X-ray structure represents a crystal structure, which is not necessarily equal to a solution structure (Schulz, 2007). An exciting advance that may overcome the serious handicap of the necessity for growing quality protein crystals to solve a 3D structure is the use of high-resolution single molecule diffraction images (Miao, et al., 2001). This new methodology utilizes a powerful X-ray free electron laser (X-FEL) and an algorithm that can solve protein structures from the X-FEL-produced diffrac-tion patterns from single biomolecules rather than multiple molecules held in a specific crystal lattice.
Should the need to obtain quality protein crystals be eliminated, there would be an explosive increase in the number of high-resolution 3D protein structures avail-able for use in rational, structure-based drug design (Hogg and Hilgenfeld, 2007).
The concurrent development of NMR techniques and molecular biology has led to an increased study of the 3-D structure and dynamics of proteins in solution (Clore and Gronenborn, 1998; Carlomagno, et al., 2007). Like X-ray crystallography, NMR spectroscopy generates information directly about the proximity of atoms and about the lifetimes of the through-space interactions of those atoms. Significant advantages of the protein NMR approach over X-ray crystallogra-phy include its ability to study proteins in solution, to obtain structural information about dynamic (flex-ible) portions of the molecule and to look specifically at hydrogen atoms. X-ray crystallography, unlike NMR, provides spatial information about all non-hydrogen atoms. NMR is now widely used as a drug discovery tool, to understand both the 3-D structure of receptors and drug-receptor interactions (Klages and Kessler, 2007).
Protein modeling (in the broader sense, molecular modeling) is a collection of computer techniques, including computer graphics, computational chemis-try, statistical methods and database management, applied to the description, analysis, and prediction of protein structures and protein properties (Charifson, 1997; Murcko et al., 1999). Protein folding (including prediction of 3-D structures), dynamics simulations, protein function and protein-molecule (ligand, DNA, protein, etc.) interactions are some of the problems that are being studied currently by protein modeling. Protein modelers often study products from protein engineering. The 3-D protein structures used in modeling are frequently derived from X-ray crystal-lography and NMR analyses. When structures are not available from these structural techniques, either de novo methods or homology modeling approaches must be used (Charifson, 1997; Laskowski and Swaminathan, 2007). De novo methods involve the prediction of secondary protein structure from an analysis of the amino acid sequence. Homology modeling uses the known structures of homologous or similar proteins as 3-D templates on which one constructs the framework of the protein being studied. Validation of the computer model resulting from either of these methods with experimental observations is necessary. The limited success rate for homology modeling approaches to accurately predict a protein’s 3-D structure (ascertained bycomparing a protein’s homology modeling predicted structure with its experimentally determined X-ray crystallographic structure) has limited their broader use (Baker and Sali, 2001).