Full Repacks And Surface Design

Solvation is one of the hardest effects to quantify in protein design and remains an important hurdle for accurate predictions, especially for cavity and boundary positions. A hydrophobic protein core can be accurately modeled as a continuous, low dielectric medium, making electrostatic effects easy to calculate. Calculating these effects for solvent-exposed residues or buried polar residues, however, becomes complex and is one area where design force fields must make important choices between accuracy and speed.

The simplest observation to be made about the difference between a protein's core and surface is amino-acid composition. Generally, protein cores are hydrophobic in nature to aid in folding and stability, while surface and boundary positions are evenly distributed between polar and nonpolar amino acids to ensure solubility and prevent aggregation. Deviations from this trend are inspired by functional necessities such as oligomer formation or the burial of an active site.

Binary patterning is a method that exploits this observation by restricting the choice of amino acids based on the location of the position. This places little responsibility on the force field to discriminate between favorable and unfavorable solvent effects, as only favorable residues are allowed at solvent exposed positions. One illustration of the power of binary patterning was the design of a library of four-helix bundles by randomly allowing polar residues on the surface and nonpolar residues on the interior of the bundle (Hecht et al. 2004). Short (74-residue) bundles designed with this naive strategy contained a majority of stable, soluble proteins with alpha-helical secondary structure. NMR spectroscopy revealed that most of the proteins had fluctuating cores and more closely resembled molten globule intermediates than well-folded proteins.

Simply extending the bundles (102-residue) caused a majority of the tested proteins to have well-resolved NMR spectra indicative of well-ordered side chains (Hecht et al. 2004). In this case, binary patterning was sufficient to specify a simple protein fold, without high-resolution structural data to guide residue selection. The dependence on length suggests that a given tertiary structure has a minimum size in order to be stabilized. In the case of protein design, patterning can be combined with existing force fields to improve the stability and solubility properties of complex folds.

The first fully automated and rational design of an entire protein, a zinc finger domain, used this technique (Dahiyat and Mayo 1997). Only polar residues were allowed at surface positions, while only hydrophobic residues were allowed in the core. In the boundary positions between surface and core, however, both sets of residues were allowed. Though the potential included an atomic solvation model, it tended to select hydrophobic residues, given the choice between the two, necessitating the patterning approach to ensure solubility. The surface and boundary positions were additionally designed with a hydrogen-bonding potential to maximize specific contacts that would specify the fold.

The resulting protein shared only 21% identity with the wild type domain, the majority of which were buried, consistent with the steric restrictions of the core (Dahiyat and Mayo 1997). Thermal denaturation showed a cooperative unfolding transition that was completely reversible, as expected for a small hydrophobic domain. Structural analysis by solution NMR revealed that the designed protein's backbone packed in the expected topology, with a backbone deviation of 1.98A.

Since the design of the zinc finger, design potentials have attempted to elaborate on simple patterning by developing their solvation models to accurately distinguish between surface and core positions. As a rigorous test of whether a modern design potential could accurately design full-length proteins without the use of patterning, Dantas and coworkers investigated nine globular proteins of varying secondary structure composition (Dantas et al. 2003). An implicit solvent model was used to discriminate between positions and favor hydrophobic placement in the core. The redesigned sequences were 35% similar to wild type, on average, and 50% similar when only core residues were considered.

When expressed and characterized, six of the nine designed proteins appeared monomeric and folded. Furthermore, their stabilities (as measured by chemical denaturation) were comparable to, or, in a few cases, better than the wild type proteins (Dantas et al. 2003). The improvement in stability is not surprising, as the designed proteins are free to abolish function in favor of stability, which is a luxury the wild type protein is not afforded. Still, it is encouraging that the success of Mayo's zinc finger design can be extended to multiple proteins with widely varying structures.

0 0

Post a comment