Dissertation

Computational Design of Protein Function Using Modular Backbone Assembly

Lapidoth, Gideon David

2018

Overview

Computational protein design has made substantial progress over the past years, generating novel conformations, catalysts, binders, and oligomeric assemblies. Prevalent methods to design new conformations de novo have relied on so-called “ideal” folds rich in regular secondary structures and almost devoid of loops and destabilizing elements, such as cavities. Molecular function, such as binding and catalysis, however, often demands non-ideal features, including large and irregular loops, and buried polar interaction networks, which to date protein designers have failed to generate in a general reproducible manner. Currently, to design new function, protein designers repurpose scaffolds from naturally occurring proteins to carry out different functions. These designs are reminiscent of the designed folds mentioned above, since they relied on rigid protein scaffolds with high secondary-structure content whereas natural proteins encode functional elements in regions lacking secondary structure. Additionally, modifying these natural scaffold often compromises the protein’s stability. Herein, I describe a combinatorial backbone and sequence design algorithm, which addresses both issues: designing new scaffolds with non-ideal features which can be tailored for a function of choice, while simultaneously optimizing the protein’s stability. The method leverages the large number of sequences and experimentally determined molecular structures of natural proteins to construct novel protein binders and catalysts. To prove the generality of my design approach the algorithm was applied and experimentally tested on two unrelated protein folds and functions: antibody binders against human insulin and bacterial Acyl Carrier Protein (ACP) and TIM-barrel fold catalysts for hydrolysis of lactones and xylan sugar. 2 anti-ACP and 1 anti-insulin designed antibodies bound to their ligands with mid to high nanomolar affinities before directed evolution and demonstrated native like stability despite having over 30 mutations from mammalian antibody germlines. The designed binding modes were validated using site directed mutagenesis and crystallographic analysis of two of the anti-insulin binders revealed atomic accuracy throughout most of the structure. 43 glycoside hydrolase 10 (GH10) xylanases and 34 phosphotriesterase-like lactonases (PLLs) were also generated using this method. Twenty-one GH10 and seven PLL designs were active and four were as active as natural enzymes in these families. The designs exhibited thermostability on par with natural enzymes from thermophiles despite having over 100 mutations from their closest homologue.

• Publisher: ProQuest Dissertations & Theses
• Subject: Amino acids / Antigens / Binding sites / Bioengineering / Bioinformatics / Computational chemistry / Crystal structure / Genes / Insulin / Molecular biology / Mutagenesis / Plasmids
• ISBN: 9798802709863