Researchers from the University of Washington have used sequence function maps generated using deep sequencing to improve two in silico-designed influenza inhibitors so that they can better fight different virus strains.
The research team has combined computational drug design and sequencing to optimize two neutralizing antibodies — HB80.3 and HB36.4 — that bind to a portion of the hemagglutinin protein on the surface of influenza viruses.
The group claims in a recently published Nature Biotechnology paper that after improvements enabled by sequencing, one of the in silico-designed proteins, HB80.3, "neutralizes H1N1 viruses with a potency that rivals that of several human monoclonal antibodies." Additionally, it was found to be "cross-reactive against all influenza group 1 hemagglutinins, including human H2."
Specifically, the paper discusses how deep sequencing was used to locate variants in the HB80.3 and HB36.4 inhibitors that resulted in stronger binding interactions between the in silico antibodies and viral proteins.
Because of these results, "we anticipate that our approach … will be widely useful in generating high-affinity and high-specificity binders to a broad range of targets for use in therapeutics and diagnostics," David Baker, a UW biochemistry professor and a co-author on the Nature Biotech paper, said in a statement.
Baker heads UW's newly formed Institute for Protein Design where researchers work on engineering novel proteins for specific functions in medicine.
Using computer modeling techniques, the researchers construct proteins that fit into specific nano-sized targets on flu viruses, in this case points on hemagglutinin molecules, in order to prevent them from changing shape and causing infection.
The computational method that was used to design the inhibitors used in this study was discussed in a paper published a little over a year ago in Science.
As described in that paper, the approach docks "disembodied amino acid residues ... against the target surface to identify energetically favorable configurations" and then computes "shape-complementary configurations of scaffold proteins ... that anchor these energetically favorable interactions."
The method relies on the feature-matching algorithm PatchDock and components of UW's Rosetta software suite, namely the RosettaDock and RosettaDesign algorithms.
The approach was used to generate HB36 and HB80 inhibitors and their more potent variants, which are discussed in the Nature Biotech article.
The researchers report in Nature Biotech that the HB80.3 inhibitor successfully prevented the conformational change in the hemagglutinin protein required for the influenza virus to infect cells and seemed "a promising candidate for generating a broad-spectrum antiviral agent against influenza." However, they added, "additional screening failed to isolate higher-affinity variants."
The problem was that at the time the Science paper was published, the lab didn’t have the technology to test more than a few inhibitor designs at a time, which made it difficult to locate which variants in the sequences improved affinity, Aaron Chevalier, a graduate student in Baker's lab and one of the study's co-authors, explained to BioInform.
Before deep sequencing, the status quo "involved randomly making variants and trying to determine which [ones] are better," he said.
The team needed a way to "test all of the different thousand single-point mutants at the same time" so that "we could get a better idea of the areas that we needed to improve," he explained.
Deep sequencing enabled the researchers to "look at all the possible one-mutation variants" at each position in the inhibitor and locate the best mutations much faster, Chevalier said. "It's like taking a really highly detailed picture of what's going on," he added.
The Nature Biotech paper explains that by combining sequence function maps with computational design, the researchers were able to identify which variants within the inhibitors' sequences improved the binding affinity of the antibodies.
The authors generated sequence libraries containing about 1,000 single-point mutant variants for both the HB80.3 and HB36.4 inhibitors.
They then collected sequences from each library that bound to one of two hemagglutinin subtypes under specific conditions and then performed deep sequencing using an Illumina GAII instrument to "determine the frequencies of each point mutant before and after selection for binding," the paper states.
By comparing the candidate versions of the proteins before and after they had been tested for affinity to the hemagglutinin subtypes, "we could certainly see which mutations got enriched ... which means they were better binders," Chevalier explained.
"We looked at the differences between those two [sets of sequences] to try to determine where on the protein we could make improvements ... and where on the protein it was basically already optimized," he said.
So far, the group has found that both HB80.3 and HB36.4 work well in cell models. The next step is to test their efficacy in animal models, Chevalier said. Eventually, both proteins could be used in human therapeutics, he added.
Meanwhile, researchers at UW are already working on using these techniques to create therapeutic proteins for other diseases, such as malaria, Chevalier told BioInform.
"It certainly is an interesting and new way to make potential protein therapeutics and something we are continuing to look at," he said.
According to the Baker, these methods could be "a powerful route to inhibitors or binders for any surface patch on any desired target of interest."
For example, if a new disease pathogen arises, scientists could figure out how it interacts with human cells or other hosts on a molecular level and then use computational protein design to generate a diversity of small proteins that would block the pathogen's interaction surface, he explained.