Tresnak, Daniel2022-08-292022-08-292022-04https://hdl.handle.net/11299/241332University of Minnesota Ph.D. dissertation. April 2022. Major: Chemical Engineering. Advisor: Benjamin Hackel. 1 computer file (PDF); xi, 272 pages.Antimicrobial resistance is a critical and worsening global health threat, driven by excessive and erroneous use of broad-spectrum antibiotics, coupled with minimal discovery of new antimicrobial therapies. Recent advances within the biotechnology field, in particular DNA synthesis and sequencing technologies and molecular and cellular biology techniques, has empowered protein engineering efforts. The encoding of proteins via efficiently synthesized DNA enables rapid construction and testing of large protein libraries, while combinatorial amino acid diversity provides nearly limitless protein phenotypes. Given this development, ribosomally synthesized antimicrobial proteins are one compelling solution in the design and discovery of new antimicrobial therapies. Yet, development of antimicrobial proteins as clinical therapies has remained limited, in part due to lack of high throughput methods for evaluating antimicrobial activity and statistical models for informing protein design. This work spans sequence-function mapping of two antimicrobial protein families and the development of efficient strategies for their continued engineering. First, genome-mining approaches are applied to investigate the small antimicrobial protein family of class IIa bacteriocins. A library of class IIa bacteriocin variants was designed and experimentally evaluated for inhibitory activity to six strains of enterococcus. Ridge regression modeling yielded moderate predictive performance and elucidated factors impacting bacterial susceptibility to class IIa bacteriocins. Individual characterization of proteins with inhibitory activity yielded a collection of variants with high potency and stability which are compelling for further studies. The latter half of this work focuses on the design of lysin catalytic domains, which degrade critical bonds in the peptidoglycan layer of targeted bacteria, and high throughput methods for screening lysin activity and stability. Structural information and epistatic models trained on natural sequence diversity were used to design lysin catalytic domain libraries, and experimental evaluation yielded one variant which displayed 1.8-fold improvement in catalytic activity and an 11.5 °C improvement in melting temperature compared to the parental catalytic domain. This enhanced variant was then used as a lead molecule across an array of protein diversification and library design strategies. A high throughput depletion-based assay was engineered for screening lysin catalytic domain activity and coupled with on-yeast protease stability assays to functionally evaluate ~5104 lysin catalytic domain variants. Ridge regression modeling was conducted to elucidate sequence-function relationships, compare protein diversification strategies for informing epistatic models, and predict compelling new designs. This work identified several improved variants, expanded the explored lysin catalytic domain sequence space and demonstrated an efficient approach for lysin engineering. Altogether, the research presented here advances protein engineering strategies broadly, validates the utility of high throughput methods for screening antimicrobial proteins, and empowers their continued development as next-generation antimicrobial therapies.enAntimicrobial proteinClass IIa bacteriocinGenome-miningLysinSequence modelingThesis or Dissertation