Towards predictive metabolic engineering: kinetic modeling and experimental analysis of a heterologous mevalonate pathway in E. coli
- Author(s): Weaver, Lane Justin
- Advisor(s): Keasling, Jay D
- et al.
Owing to economic, political, and environmental concerns, the nature of finite natural resources will increasingly necessitate a transition to renewable resources over the next century. Biology, enabled by its use of enzymes, is a master craftsman and is thus well suited to utilize renewable resources, converting them into desired products. As such, the field of metabolic engineering has grown up over the past 20 or so years to harness biology's catalytic and synthetic capabilities. In particular, the Keasling lab has successfully exploited the potential for the mevalonate pathway for many applications, most notably the production of Artemisinic acid, an anti-malarial compound.
However, despite many years of engineering efforts, there is still no definitive way to predict how expressing a certain enzyme at a certain level will effect pathway function, which can largely be attributed to the lack of availability of a model including a kinetic representation of pathway flux.
To aid in debugging efforts to increase yield, titer, and productivity of the pathway, an ordinary differential equation (ODE) model was built using constants culled from literature and enzyme concentrations derived from experiment. To account for uncertainty in model parameters, a global sensitivity analysis was performed. The model demonstrates that amorphadiene synthase (and in general terpene synthase) activity is limiting and needs to be addressed to increase pathway flux. Furthermore, the model predicts that in a local regime, the pathway is fairly insensitive to product inhibition, a hypothesis that had previously been posited to be limiting flux.
To experimentally test these specific predictions, three E. coli strains were constructed, the first acting as a base strain, the second encoding a homologous mevalonate kinase from S. aureus that is less susceptible to product inhibition, and a third strain in which the expression of amorphadiene synthase was increased. These strains were profiled by metabolomic and proteomic methods, which helped validate the predictions of the ODE model. Furthermore, they enabled the determination of an in vivo kcat value for amorphadiene synthase, which differed ~3-fold from its in vitro counterpart--- demonstrating that in vitro parameters are not always representative of in vivo conditions, further bolstering the case for ensemble-type kinetic modeling.
Having identified targets for engineering, high-throughput screens were developed in an attempt to leverage directed evolution for improvement of production. While a screen was unable to be developed for amorphadiene synthase, a platform for screening of another rate-limiting enzyme, Acinetobacter wax ester synthase, was demonstrated.
This type of analysis--starting with in silico pathway analysis, followed by in vivo characterization and finally targeted enzyme engineering--will become increasingly important as metabolic engineering is applied to the production of a variety of new targets. With this ability, metabolic engineers will more rapidly be able to move from concept to production of designer molecules in high titers, yields, and productivities.