From Bottomup Strategies To The Design Of Uniquely Folded Structures

The field of protein design advanced considerably in the second half of the 1990s as researchers (1) tackled the design of proteins with both alpha and beta secondary structures; and (2) began to apply the rapidly developing tools for computational searching and optimization.

In 1996, Imperiali and coworkers reported the design of a 23-residue protein with a ppa architecture similar to that found in DNA-binding zinc finger modules (Struthers et al. 1996). Their goal was to design a novel sequence that would fold into the desired structure without the use of disulfide bridges or metal-binding sites. In contrast, natural zinc fingers require metal to fold. This was accomplished by starting with a native zinc finger backbone as a template and incorporating a type II' p turn (Sibanda and Thornton 1985) and a D-proline. The resulting peptide, BBA1, was soluble and monomeric. The NMR structure showed that the protein had the desired secondary structure and type II' p turn.

Progress in protein design was greatly enhanced in the mid-1990s as researchers took advantage of newly emerging computational approaches. A year after the rational design of BBA1 by Imperiali and coworkers, Dahiyat and Mayo (1997) devised a fully automated computational algorithm to redesign the ppa zinc finger motif. A natural zinc finger module was chosen as the design template and its coordinates were used as the fixed backbone for the sequence selection algorithm. After computationally screening 1027 amino-acid sequences, the designed sequence FSD-1 was selected as the optimal sequence for the desired ppa fold. FSD-1 had low sequence identity to any known protein sequence. The design was experimentally validated by solving the NMR structure of FSD-1, which showed that this novel protein was well-folded and looked very similar to the designed structure. This landmark study demonstrated that a purely computational approach could successfully design a small, well-folded protein containing alpha, beta, and turn structural elements (Dahiyat and Mayo 1997).

Although initial attempts at computational protein design were limited to the design of new side chains onto fixed backbone structures that were "borrowed" from the coordinates of natural proteins, further advances in computational methods enabled new algorithms that allowed flexible backbones (Desjarlais and Handel 1999). While the fixed backbone algorithms were sufficient to redesign side chains onto pre-existing natural structures, the ability to incorporate backbone flexibility is essential for the computational design of truly novel folds for which no backbone template pre-exists.

0 0

Post a comment