ORCID

http://orcid.org/0000-0003-4022-2460

Date of Award

Winter 12-15-2021

Author's School

McKelvey School of Engineering

Author's Department

Biomedical Engineering

Degree Name

Doctor of Philosophy (PhD)

Degree Type

Dissertation

Abstract

The human gut microbiome is a compositionally and functionally diverse community of microorganisms that profoundly influences the health of the host. Deep characterization of gut microbiomes via high-throughput sequencing has identified associations between the gut microbiome and disease states, and spurred development of engineered microbial therapeutics. Successful translation of these research efforts to the clinic will involve i) consideration of how engineered microbes behave and adapt in the gut, towards the implementation of biosafety mechanisms, and ii) identification and validation of microbial biomarkers of disease. This dissertation describes both investigative (Chapter 2) and engineering (Chapters 3 and 4) approaches to improving the safety and efficacy of microbial therapies, using the commensal Escherichia coli Nissle 1917 (EcN) as a model chassis. This work further investigates the gut microbiome as an early predictor of preclinical Alzheimer Disease (Chapter 5), with corresponding identification of candidate microbial biomarkers. Unlike traditional therapeutics, engineered microbes are subject to selection potentially at the cost of their intended function. In Chapter 2, gnotobiotic and conventional mouse models of colonization were used to identify the major selective forces acting on EcN in the dysbiotic mammalian gut. Functional metagenomic selections and isolate whole-genome sequencing identified access to preferred carbon sources as the main selective pressure on EcN in low-diversity guts. In functional metagenomic selections, EcN populations encoding heterologous glycosyl hydrolases enabling consumption of dietary polysaccharides were strongly enriched. In the absence of this repertoire of heterologous functions, wild-type EcN mutated to better consume host mucins. In addition, EcN was observed to be a reservoir of aminoglycoside resistance-conferring mutations subsequent to a single antibiotic exposure. These findings carry clinical implications for the administration of engineered probiotics in low-diversity, dysbiotic guts. As such, biocontainment of engineered functions and microbes in situ is an important design consideration, with the goal of mitigating potential for pathology and environmental contamination. In Chapter 3, a transcript-barcoding approach enabling measurement of activity of many synthetic constructs in parallel was developed and validated in mice. The relative activities of 30 pooled EcN strains, each harboring a distinct synthetic construct, were ranked in multiple gut sites; this work informed the development of EcN strains engineered for the treatment of phenylketonuria, the efficacy of which were validated in a murine model of the disease (Chapter 2). This transcript barcoding method will facilitate the design and testing of bio-sensing synthetic constructs in vivo, towards the restriction of engineered functions to target sites of interest. In Chapter 4, a CRISPR-based approach to microbial biocontainment was developed and tested in mice. The kill switch design enabled selective removal of EcN from the guts of mice upon chemical induction; inclusion of temperature-responsive element further enabled kill switch induction upon excretion from the gut. Due to the stringent selective pressure imposed by kill switches, such designs are prone to mutational inactivation. Genetic stability was achieved using parallel approaches of engineering and environmental control, the first by inclusion of functional redundancies, a plasmid retention system, and knockouts of SOS response genes, and the latter by the innovative leveraging of inter-strain exclusion behaviors observed in the gut. Specifically, co-administration of kill switch-encoding EcN with a control EcN strain, in tandem with kill switch induction, enabled virtually complete eradication of the kill switch population from the guts of mice. This approach is attractive when the goal is to remove an engineered subpopulation of engineered microbes, and not a probiotic species itself, from the gut. In Chapter 5, the gut microbiome was instead considered as source of candidate microbial biomarkers of preclinical Alzheimer Disease (AD). Identification of microbiome correlates at the preclinical stage has the potential for clinical significance in that canonical biomarkers of preclinical AD require access to PET-CT scanners, or invasive CSF lumbar puncture, whereas stool is a cheaply acquired analyte. Gut microbiome features were correlated with amyloid biomarkers, but not markers of tau or neurodegeneration, highlighting the potential utility of gut microbiome correlates in early screening as abnormal amyloid levels are considered the most antecedent in disease progression. Further, addition of microbiome features to machine learning classifiers for preclinical status improved sensitivity, predominantly when metagenome-assembled genomes were included as features, rather than taxa identified through clade-specific marker sequences. This suggests taxonomic associations with AD status may be strain-specific, and encourages the pursuit of strain-level resolution in microbiome-wide association studies. As we work to translate observations about the gut microbiome to actionable clinical interventions, we must consider the interactions of engineered microbial therapeutics and recipient microbiomes, as well as the resolution necessary to validate gut microbiome-derived biomarkers of disease. This thesis work represents efforts to that end, and will ultimately inform the design of novel gut-directed therapies.

Language

English (en)

Chair

Gautam Dantas

Committee Members

Matthew Ciorba

Comments

Update embargo

Available for download on Monday, December 12, 2022

Share

COinS