Disulfide-rich miniproteins are a powerful yet underutilized protein family for reagent, diagnostic and therapeutic applications. They are hyperstable like small molecules, yet are large enough to bind specifically to protein targets with high affinity and thus can be readily used to inhibit protein-protein interactions. Disulfide-rich miniproteins occur naturally, but the natural molecules are challenging to engineer. There is a pressing need for new technologies which enable disulfide-rich miniproteins to be engineered to bind to arbitrary protein targets, as is routine for other protein scaffolds like antibodies. Here, we propose a computational de novo design strategy using Rosetta to build a synthetic miniprotein library that will be generally useful for screening protein affinity reagents. The typical approach to creating protein libraries is to generate a large amount of random sequence diversity in a localized area of a single protein structure. Most of these sequences will be unstable or unable to bind any target (e.g. too polar or too nonpolar). Rather than use random sequences, we propose to explicitly design each member of a synthetic disulfide-rich miniprotein library. Our design library will contain 106 different miniproteins that display the widest possible variety of binding surfaces. The genes encoding this library will be synthesized using oligo pools. To perform computational de novo design at this scale, we extended the SEWING algorithm in Rosetta to generate hundreds of thousands of unique miniprotein structures and sequences. We developed novel filters to assess design model quality and to quantify and compare protein structures and surfaces. As a test case of this approach, we will screen our disulfide-rich miniprotein library via yeast display for binders to Clostridium difficile enterotoxins TcdA and TcdB. C. difficile is the leading cause of healthcare-related infections in the USA, and these two proteins mediate its pathogenicity. At present, the only available treatments for C. difficile enteric infection that target these toxin proteins are antibodies, which must be injected and have poor efficacy. A hyperstable miniprotein binder to the same neutralizing epitope could be administered orally. Therefore, this work presents a new frontier in de novo miniprotein engineering and library design. It will also result in a source of novel, therapeutically interesting molecules for the treatment of C. difficile infection.
Miniproteins are a powerful yet underutilized class of therapeutic molecules. Here, we propose to design a set of one million hyperstable miniature proteins all with unique structures that can be used to develop agents that bind to other proteins. Then, we will use this collection of miniproteins to engineer candidate therapeutic molecules that can bind and neutralize toxins produced by the bacteria Clostridium difficile, which is the leading cause of healthcare-related infections in the USA.