Computational Design of Hyperstable, De Novo Miniproteins Targeting PD-1

Thesis (Ph.D.)--University of Washington, 2017-12Computational protein design has recently advanced to a new era with the de novo design of stable proteins targeting native protein ligands. In this dissertation, I will present the first de novo protein binder with an all beta interface targeting the T cell receptor, programmed cell death protein 1 (PD-1). Expressed on activated T cells, PD-1 inhibits T cell function and proliferation to prevent an excessive immune response. Tumor cells often take advantage of this pathway by over-expressing one of the ligands of PD-1, PD-L1 or PD-L2, to evade immune destruction. Additionally, impairment of the PD-1 pathway through a variety of mechanisms can lead to autoimmunity. Using a combination of computational design and experimental approaches, we have developed a de novo miniprotein that specifically binds PD-1 at the ligand interface. This protein binds murine PD-1 at a Kd of approximately 1 µM on yeast. The apo crystal structure shows that the binder folds as designed with a backbone RMSD of 1.3 Å to the design model. The 4.5 kDa protein proved to be very stable by chemical denaturation in GuHCl likely due to its three disulfide bonds. Over the years, I have identified several other binders using the canonical method of yeast surface display that ultimately had to be abandoned because of their inability to be produced as soluble proteins. I hypothesize this results from the use of the highly expressed native yeast protein Aga2p for display of the protein-of-interest (POI) on the cell surface. Here, I present a new method that replaces the Aga2p fusion with a minimal tag for covalent capture of the secreted protein and takes advantage of the natural yeast quality control pathways to discriminate between misfolded and well-folded proteins. This novel secretion capture system is a powerful tool for the screening, optimization, and production of well-expressed, functional proteins. Improved high-throughput methods for screening and optimizing stable, functional proteins enable generation of de novo binders that are more readily amenable to a variety of applications. The small, hyperstable PD-1 binding domain presented here has potential use in a variety of cancer and autoimmune therapy platforms