Kinetic modeling of terpenoid production in E. coli: insights into subpopulation emergence and process optimization.

Escherichia coli is a promising host for terpenoid production, yet kinetic models tailored for such strains are limited—hindering effective bioprocess optimization and control. To address this, we developed a kinetic model of E. coli engineered to produce viridiflorol, aiming to capture the dynamics of key bioprocess variables under both uninduced and induced conditions. To account for growth and metabolism of cells, a simplified central metabolic pathway of E. coli was considered by including acetyl-CoA as the intracellular metabolite due to its role as a common precursor for viridiflorol, acetate, and the TCA cycle. Initial modeling efforts based solely on metabolism failed to reproduce experimental trends, prompting the inclusion of cellular stress caused by IPTG-induced heterologous pathway activation. However, this adjustment alone could not fully explain the observed trends, leading us to hypothesize the emergence of two distinct subpopulations post-induction: (1) stressed, slow-growing producers and (2) dormant non-producers that eventually outgrow producers. After iterative refinements, the model successfully replicated experimental trends with R2 value for glucose, cells, acetate, dissolved oxygen, and viridiflorol being 0.98, 0.88, 0.86, 0.70, and 0.94 respectively. Finally, model simulations were performed for insights which suggested that lowering IPTG concentration can improve viridiflorol production by delaying the emergence of the non-producer population. Guided by the simulations, an optimal IPTG concentration (9.375 µM) for a batch condition was identified, resulting in a titre of 0.14 g/L of viridiflorol. In contrast, a non-optimal IPTG concentration (150 µM) yielded a titre of only 0.03 g/L. Furthermore, experimental validation of the subpopulation hypothesis showed the coexistence of producer and non-producer populations. Taken together, the model captured the bioprocess variable trends, predicted an optimal IPTG concentration for a batch process, and provided insights into the existence of two subpopulations.