These methods generally start with a known, high-resolution
structure of the target protein and then try to optimize the packing
of different amino-acid side chains while keeping the backbone template
(carboxyl and amino groups) fixed to arrive at new low-energy sequence
solutions. The key features of these methods are an efficient search
protocol for sampling the theoretically vast number of sequence
permutations and an energy function designed to model the physical
forces that hold natural proteins together.
The collaborators extended these concepts in their RosettaDesign
method. However, they were faced with the additional challenge of
sampling protein-backbone structural space as well as sequence space,
since their goal was the creation of a novel fold (where no natural
backbone template was available). To this end, they constructed
RosettaDesign to iterate between full-scale optimization of the
sequence for a fixed-backbone conformation and gradient-based optimization
of the backbone coordinates for a fixed sequence. Beginning with
a simple back-of-the-envelope sketch of the target, a novel α/β
fold, and this protocol, they designed Top7, a 93-residue α/β
protein with a topology not observed in the Protein Structure Database
(PDB), i.e., an artificial protein.
By means of a variety of biophysical techniques, the researchers
determined the synthesized Top7 protein to be monomeric, highly
soluble, and extremely stable to chemical and thermal denaturation.
Preliminary NMR analysis also showed that Top7 had a rigid structure
consistent with the target topology. Finally, thanks to the ALS
Howard Hughes Medical Institute Beamline 8.2.1, they solved an x-ray
structure of a single selenomethionyl-substituted variant of Top7
to 2.5-Å resolution with single-wavelength anomalous diffraction
(SAD) data.
This high-resolution crystal structure revealed that the Top7 protein
adopted the designed topology and in fact was strikingly similar
to the design model at atomic resolution (1.17-Å root-mean-square
deviation or RMSD over all backbone atoms). The two models differ
most in the region surrounding the first N-terminal (amino-group
end) hairpin, but even here the all-atom RMSD did not exceed 2.8Å.
In contrast, the C-terminal (carboxyl-group end) halves of the crystal
structure and the designed model are very similar, and core side-chain
atoms are virtually superimposable.
Comparison of the x-ray crystal structure of Top7 (yellow) to
the computationally designed model (green). The side-chains in
the core of the C-terminal portion of the Top7 structure are effectively
superimposable with the model.
The successful design of Top7 has two major implications. First,
it is a strong validation of the understanding and description of
the energetics of proteins and other macromolecules, much of which,
incidentally, has been a consequence of the determination of high-resolution
structures of those macromolecules. Second, it suggests that the
development of protein therapeutics and molecular machines need
not be limited to the structures sampled by the biological evolutionary
process.
Research conducted by G. Dantas, G. Varani, and D. Baker (University
of Washington, Seattle); B. Kuhlman (University of North Carolina,
Chapel Hill); and G.C. Ireton and B.L. Stoddard (Fred Hutchison
Cancer Research Center, Seattle).
Research funding: The National Institutes of Health and the Cancer
Research Fund of the Damon Runyun–Walter Winchell Foundation.
Operation of the ALS is supported by the U.S. Department of Energy,
Office of Basic Energy Sciences.
Publication about this research: B. Kuhlman, G. Dantas, G.C. Ireton,
G. Varani, B.L. Stoddard, and David Baker, “Design of a novel globular
protein fold with atomic-level accuracy,” Science 302,
1364 (2003). |