Bacteria - Hydrogel Interactions: Mechanistic Insights via Microelastography and Deep Learning

Bacteria-based cancer therapy (BBCT) holds immense promise in addressing the limitations in treatment of solid tumors. Bacterial strains used for BBCT are engineered to express therapeutics, facilitate precise navigation within the tumor microenvironment by enhancing bacteria's motility, chemotaxis (movement toward or away from specific chemicals), or other mechanisms that aid in reaching and infiltrating the tumor tissue effectively, and complementing traditional chemotherapy and immunotherapies while minimizing side effects. Bacterial motility not only influences the ability of bacteria to navigate within the tumor but also plays a pivotal role in optimizing drug delivery, treatment efficacy, and minimizing potential obstacles associated with the complex microenvironment of human tissues. However, the current understanding of bacterial motility remains limited. In this thesis, we use a reductionist approach and study bacteria motile behavior within human tissue phantoms (collagen and agar) and the bacteria-hydrogel interactions. Apart from motility, it is important to analyze the mechanical properties of the hydrogels the bacteria interact with as they play a vital role in overall behavior and physics of bacteria movement. To that extent, there exists a gap in our understanding of the viscoelastic properties of hydrogels. Lastly, systematic and comprehensive investigation of bacteria behavior in hydrogels requires tracking of thousands of individual cells. Thus, there is an unmet need to develop new automated techniques to reduce the labor-intensive manual tracking of bacteria in low-contrast hydrogel environments, with feature sizes comparable to that of bacteria. To address these gaps, this thesis proposes a trident approach towards mechanistic understanding of bacteria motility in time-invariant agar and temporally evolving collagen hydrogels to bridge critical gaps in understanding bacterial motile behavior in these media, non-destructive microelastography-based mechanical characterization of hydrogels with less than 4.7% error compared with rheology, and the development of deep learning-enabled automated bacteria tracking tools with 77% precision.