| Summary: | Thesis advisor: Tim van Opijnen === Thesis advisor: Welkin Johnson === Pathogenic bacteria can experience various stress factors during an infection including antibiotics and the host immune system. Whether a pathogen will establish an infection largely depends on its survival-success while enduring these stress factors. We reasoned that the ability to predict whether a pathogen will survive under and/or adapt to a stressful condition will provide great diagnostic and prognostic value. However, it is unknown what information is needed to enable such predictions. We hypothesized that under a stressful condition, a bacterium triggers responses that indicate how the stress is experienced in the genome, thereby correctly identifying a stress response holds the key to enabling such predictions. Bacterial stress responses have long been studied by determining how small groups of individual genes or pathways respond to certain environmental triggers. However, the conservation of these genes and the manner in which they respond to a stress can vary widely across species. Thus, this thesis sought to achieve a genome-wide and systems-level understanding of a bacterial stress response with the goal to identify signatures that enable predictions of survival and adaptation outcomes in a pathogen- and stress-independent manner. Here, we first set up a multi-omics framework that maps out a stress response on a genome-wide level using the human respiratory pathogen Streptococcus pneumoniae as a model organism. Under an environmental stress, gene fitness changes are determined by transposon insertion sequencing (Tn-Seq) which represents the phenotypic response. Differential expression is profiled by RNA-Seq which represents as the transcriptional response. Much to our surprise, the phenotypic response and transcriptional response are separated on different genes, meaning that differentially expressed genes are poor indicators of genes that contribute to the fitness of the bacterium. By devising and performing topological network analysis, we show that phenotypic and transcriptional responses are coordinated under evolutionary familiar stress, such as nutrient depletion and host infection, in both Gram-positive and -negative pathogens. However, such coordination is lost under the relatively unfamiliar stress of antibiotic treatment. We reasoned that this could mean that a generalizable stress response signature might exist that indicates the level to which a bacterium is adapted to a stress. By extending stress response profiling to 9 antibiotics and 3 nutrient depletion conditions, we found that such a signature indeed exists and can be captured by the level of transcriptomic disruption, defined by us as transcriptomic entropy. Centered on entropy, we constructed predictive models that perform with high accuracy for both survival outcomes and antibiotic sensitivity across 7 species. To further develop these models with the goal to eventually enable predictions on disease progression, we developed a dual RNA-Seq technique that maps out the transcriptomic responses of both S. pneumoniae and its murine host during lung infection. Preliminary data show that a high entropy is observed in the pathogen’s transcriptome during clearance (a failed infection) compared to a successful/severe infection, while the host transcriptome exhibits a pro-inflammatory and active immune response under the severe infection. Lastly, we characterized evolutionary trajectories that lead to long-term survival success of S. pneumoniae, for instance this means that the bacterium successfully adapts to the presence of an antibiotic and becomes resistant or can grow successfully in the absence of a formerly critical nutrient. These trajectories show that adaptive mutations tend to occur in genes closely related to the adapted stress. Additionally, independent of the stress, adaptation triggers rewiring of transcriptional responses resulting in a change in entropy from high to low. Most importantly, we demonstrate that by combining multi-omics profiles with additional genomic data including gene conservation and expression plasticity, and feeding this into machine learning models, that adaptive evolution can become (at least partially) predictable. Additionally, the genetic diversity in bacterial genomes across different strains and species can indeed influence a bacterium’s adaptation trajectory. In conclusion, this thesis presents a substantial collection of multi-omics stress response profiles of S. pneumoniae and other pathogenic bacteria under various environmental and clinically-relevant stresses. By demonstrating the feasibility of predictions on bacterial survival and adaptive outcomes, this thesis paves the way towards future improvements on infectious disease prognostics and forecasting the emergence of antibiotic resistance. === Thesis (PhD) — Boston College, 2020. === Submitted to: Boston College. Graduate School of Arts and Sciences. === Discipline: Biology.
|
|---|
