4 | CONCLUSION
Oleochemical-producing microbes with arbitrary, user-defined product profiles remain an elusive goal of metabolic engineering (Marella et al., 2018; Sharma and Yazdani, 2021; Yan and Pfleger, 2019). The activities and substrate specificities of critical enzymes are difficult to predict (or change), and their integration into new pathways typically requires multiple rounds of iteration; titer and product profile, in particular, tend to be nonintuitively coupled (Greenhalgh et al., 2021; Grisewood et al., 2017; Sarria et al., 2018). In this study, we developed a set of kinetic models that facilitates the design and analysis of oleochemical pathways. These models provide good fits to experimental data and can capture the effects of previously reported pathway modifications. The final models enable mechanistic studies of unexpected phenomena (e.g., the inhibition of FAEE biosynthesis by TesA overexpression) and facilitate large, system-wide analyses that are experimentally intractable. Our sensitivity analysis, for example, allowed us to probe the influence of isolated shifts in protein concentration within a broad set of pathways. Findings indicate that product profiles are most sensitive to concentrations of FabB, FabF, and acyl-ACP thioesterase (when present) and indicate that coordinated shifts in these concentrations can adjust the product profiles of pathways with promiscuous enzymes downstream of the FAS. Oleochemical-specific enzymes, in turn, primarily influence total production, an indication that their overexpression can improve titer without altering final product profiles. These observations provide guidance for optimizing enzyme expression levels in oleochemical pathways.
The incorporation of different oleochemical pathways into our FAS model allowed us to refine general rules for characterizing pathway enzymes. Results indicate that substrate-specific Michaelis-Menten parameters are sufficient to capture the activities of a broad set of enzymes. These parameters rely on steady-state assumptions that are reasonable during oleochemical production in stationary phase. Importantly, our consistent parameterization of enzymes between pathways suggests that our models can capture their activities in different biochemical contexts. Enzyme-specific kinetic parameters are most accurately measured within vitro kinetic assays (e.g., the CAR-based alcohol pathway), but our results indicate that in vivo product profiles that differ in the identity or expression level of individual enzymes can also support reasonable estimates (e.g., alternative thioesterases). Finally, protein expression levels, which can vary between strains (Monk et al., 2016), are important for accurate and consistent kinetic models. The regular use of proteomics to measure protein concentrations in engineered strains could improve model accuracy and help explain strain-to-strain variability in oleochemical pathways.
The paper concludes by developing a gui to facilitate the rapid implementation of our models. This gui is a useful tool for exploring the influence of various pathway modifications on final product profiles and could guide experimental studies (e.g., promoter selection or configuration). It complements the complete model files, which enable detailed analyses of oleochemical production (e.g., plots of various metabolites or analyses of tradeoffs between objectives) with minor changes in code or short optimization routines. Together, the gui and its core models provide a versatile kinetic framework for studying oleochemical biosynthesis.