Dynamic single-cell modelling of nutrient signalling in budding yeast

Sammanfattning: To survive in environments with constantly changing nutrient levels the bud- ding yeast Saccharomyces cerevisiae has a sophisticated nutrient sensing system. A central component of this system is the SNF1 signalling pathway, which primarily helps yeast adapt to glucose limited conditions. Albeit we know the overall behaviour of the SNF1 pathway, there are gaps in our understanding of its regulation upon long-term starvation. Moreover, most studies have focused on the population average behaviour of SNF1 signalling, however, the dynam- ics of single cells can differ substantially from the ensemble average. Hence, in this thesis we set out to elucidate the single cell behaviour of SNF1 signalling by combining single-cell experiments with dynamic mechanistic modelling. In Paper I, we investigated why the expression rate of the SUC2 gene decreases upon long-term glucose starvation. Combining mathematical modelling with time-laps microscopy experiments we showed that a feedback acting on the SNF1 complex can explain this decrease. In the developed mathematical model we assumed cell-to-cell variability arising from stochastic chemical reactions (intrinsic noise) to be negligible, primarily, to be able to estimate unknown model parameters using a non-linear mixed-effects approach. However, a model’s predictive power can be impacted negatively, if this assumption is incorrectly made. To avoid this issue we developed PEPSDI (Paper II), a flexible Bayesian inference framework for mixed-effects models where the model dynamics are driven by a stochastic processes, such as a Markov jump process or a stochastic differential equation. Using PEPSDI, we then studied the Mig1 protein and could deduce hexokinase activity as a source of cell-to-cell variability in SNF1 signalling.