Â鶹´å

Shiladitya Banerjee, Chair
Oana Carja, Â鶹´å
Fangwei Si, Â鶹´å
Andrew Mugler, PITT

How do single-celled microorganisms adapt and learn to survive in dynamic
environments? Bacteria, as some of the simplest life forms, endure rapid shifts in
nutrient availability or recurring antibiotic stress by allocating cellular resources based
on prior experience. However, the decision-making strategies which facilitate rapid
adaptation, and the mechanisms by which they are executed, are currently not well
understood. In this thesis, we present theoretical frameworks to understand bacterial
adaptation and learning at the single-cell level. Through mechanistic and computational
models, we connect intracellular physiology to cell-level behavior to uncover decision-
making strategies used by cells to alter their physiology to adapt to dynamic stressors
such as nutrient shifts and antibiotic application, identifying fundamental resource
allocation tradeoffs cells must navigate in these regimes. Furthermore, by recasting the
problem of adaptation as physical computation, we demonstrate, for the first time, the
emergence of learning-like behavior by single bacterial cells in fluctuating nutrient
environments and identify a biologically plausible reaction network capable of exhibiting
such learning. Finally, the same properties which underpin microorganisms’ remarkable
ability to adapt to unpredictable environmental stressors also make them hard to
control, especially when attempting to suppress their growth. Thus, to conclude we
present reinforcement learning approaches to learn optimal feedback-guided policies
which exploit cellular adaptive mechanisms through environmental modulation to drive
population extinction– a difficult problem with great clinical significance.