Computational biologists designed and produced two novel proteins that strongly bind to a crucial flu protein that enables the virus to enter cells. The new creations, built with the help of more than 200,000 personal computers around the world, may one day serve as effective antiviral therapies, according to a study published today (May 12) in Science.
To design proteins to interact with a desired target, such as a pathogen's protein, researchers can scan extensive libraries of protein structures in search of a few that roughly complement the target molecule, then tweak those structures slightly to produce a tighter fit. Alternatively, they can introduce the pathogen to an animal to coerce its immune system to respond to the target, and then select from the antibodies that are generated.
While the former approach grants researchers control over where and how the designed proteins will bind to the target, they may not bind as strongly to the target. The latter, more "natural" approach, on the other hand, may yield antibodies that have a high affinity for the target molecule, but researchers have little control over the dynamics of binding.
But with quickly-evolving target molecules such as the influenza virus's surface protein, hemagglutinin, which has a large area that is constantly mutating and changing to evade antibody binding, even antibodies that bind well are often rendered obsolete in time.
To tackle this challenge, computational biologist David Baker of the University of Washington and his colleagues decided to focus on a region of hemagglutinin that tends to be quite stable and is conserved among many influenza strains. Antibodies that bind to this region have been shown to prevent the virus from fusing its membrane with a host cell's and cause infection.
To target this region of the protein, the researchers had to work on the problem in reverse, first searching for "nooks and crannies" in that region where a protein would be able to take hold, Baker explained, and then identifying strings of amino acids that could fit in those spaces and act as hooks.
Once they created an entire library of these hooks, they searched proteins with known structures for those that would roughly fit the conformation of hemagglutinin and serve as the main protein bodies to hold the hooks.
The researchers then modified the orientation and sequence of these scaffold proteins to hold the hooks in positions so that they could interact with hemagglutinin. For this critical, time-consuming step, the researchers reached out to the public for help in solving and optimizing the 3D structures of the proteins. Around 250,000 volunteers downloaded free software developed by Baker's lab called Rosetta@home, which allowed their personal computers at home to contribute computing power for the complex calculations.
"The design approach was really extraordinary," said Tanja Kortemme, a computational biologist at University of California, San Francisco, who did not participate in the study. "They turned the problem around by first finding the amino acid side chains that formed the interactions that they wanted, and then finding a backbone that could display those side chains."
All in all, the researchers came up with around 80 novel proteins. When expressed on the membranes of yeast, however, only two were able to bind to hemagglutinin, and the binding strength had to be further improved by slightly tinkering with the amino acid sequences.
"The success rate is still very low," Baker said. But comparing a crystal structure of one of the two designs that bound to hemagglutinin with the initial computational model from which the protein was designed, he found they were essentially superimposable, an extremely rare accomplishment in de novo protein design. Thus, although the model still needs improvement, it was able to successfully predict an interaction between two proteins.
The fact that the researchers produced two very different designs that worked for the same target is also cause for great optimism, added Karanicolas, who did his postdoc in Baker's lab at the University of Washington. "The real strength of this method is that it allows control of the design in the very early stages."
Fleishman, S.J., et al., "Computational design of proteins targeting the conserved stem region of influenza hemagglutinin," Science, 332:816-21, 2011