A Novel, Rapid Phenotypic Assay for a Beta-Lactam Antibiotic Susceptibility and an Analysis of its Theoretical Limits

Author: Liaw, Eric Jer-Jiun

Year: 2023

Degree: Dissertation (Ph.D.)

Advisor: Ismagilov, Rustem F.

Committee Members: Murray, Richard M.; Ismagilov, Rustem F.; Newman, Dianne K.; Cai, Long

Option: Biology

DOI: 10.7907/qhvg-7q92

Abstract

Current management of bacterial infections is limited by the slow turnaround time of culture-based antibiotic susceptibility testing (AST). Culture-free phenotypic AST methods, though faster, are limited not only by analytical sensitivity but also by the low number, density, and purity of live pathogens present in clinical specimens before culturing. Separating and concentrating pathogens from clinical specimen matrices and improving the analytic sensitivity of phenotypic measurement technologies remain active areas of research. However, to date, the literature lacks consensus over what is a reasonable goal for the minimum number of pathogens in a clinical specimen needed to accurately perform phenotypic AST.

I describe "bulk filtration AST" and "digital filtration AST," two new filtration-based AST methods that improve an AST method previously published by others and myself. These methods use nucleic acid quantification to assess the activity of antibiotic classes (and only those classes) targeting peptidoglycan turnover, specifically the beta-lactams, which are the most frequently prescribed class of antibiotics. I use filtration AST to quantify the in vitro pharmacodynamics of beta-lactam antibiotics over time scales shorter than two hours, and I simultaneously validate the methods' accuracies on clinical isolates of Enterobacteriaceae. To analyze filtration AST results, either for fitting parameter values or for predicting susceptibility, I derive probabilistic models for the outcomes of each of the two filtration AST methods, then perform Bayesian parameter inference from my data.

I then propose a general mathematical framework for defining the concepts of the phenotypic assay and the ideal phenotypic assay. Within this framework, I calculate the ideal filtration AST performance as a function of the number of cells assayed, my fitted pharmacodynamic parameters, and other variables. Interestingly, the observed performance of my implementation of digital filtration AST is consistent with the implementation's approaching the ideal performance. I hope my demonstration of these new methods and my theoretical framework will help guide future research into rapid phenotypic AST.

Files