Construction of Enzymes with Synthetic Allosteric Regulation to Control Metabolic Pathways of Escherichia coli

In metabolic engineering strains are created that overproduce a certain product. For that, the production pathway is often released from any transcriptional, translational and post-translational regulation, resulting in a high abundance of enzymes in the production pathway and enzyme variants with feedback-resistance. However, complete dysregulation has several disadvantages: the pathway cannot respond to internal and external perturbations and metabolism of the host is overloaded, resulting in a lowered cellular fitness and reduced growth rates. To circumvent this problem, it is desirable to implement new layers of regulation in the metabolic network and in particular in the overproduction pathway. So far, such regulation has usually been implemented by controlling enzyme abundance. Two-phase processes for instance are dividing a bioprocess in two phases, a growth phase in which a sufficient amount of biomass is accumulated, and a production phase. In this second phase, the expression of enzymes needed for overproduction is induced, often in combination with the introduction of metabolic bottlenecks in competing pathways. However, the regulation of enzyme abundances does not allow fast response at the second or minute time-scale. Especially in large-scale bioreactors fast response is important, because of fluctuating availabilities of nutrients and oxygen caused by insufficient mixing which leads to the formation of microenvironments and dead-zones. Cells in which the overproduction pathway is either dysregulated or regulated only by implemented control of enzyme abundance are not able to adjust their metabolic networks according to fast changing microenvironments, leading to stressed and therefore unproductive strains which might negatively affect the stability and durability of bioprocesses. This highlights the need for faster acting dynamic control of metabolic pathways, for example through enzymes with synthetic allosteric regulation. However, the creation and usage of such enzymes is very challenging. A major goal of this work was to create such enzymes with synthetic allosteric regulation and to test their ability to control fluxes through their pathway. Synthetic allosteric enzymes are ‘metabolic valves’ that implement bottlenecks in the reaction they are catalyzing and we sought to characterize functioning of these valves in vivo. Therefore, our first goal was to examine functioning of these valves, resulting metabolic bottlenecks and their impact on the general fitness (Chapter 3). For that, we analyzed growth and metabolic profiles of wildtype isolate and laboratory strains and could show that in laboratory strains a previously reported bottleneck caused by low pyrE gene expression causes insufficient fluxes through the pyrimidine biosynthesis pathway and subsequently lowered growth rates. In addition to that, we used CRISPR interference to introduce artificial bottlenecks in 30 reactions in different parts of the metabolic network. In 16 of the resulting 30 strains we were able to detect elevated substrate or lowered product concentration, indicating a metabolic bottleneck. However, only 6 of these 16 strains also had a reduced growth rate, underlining that the impact of metabolic bottlenecks on the growth rate is generally dependent on the reaction and strength of the bottleneck. In the second part, we then evaluated two methods to create synthetic allosteric enzymes, both of which are based on the concept of directed evolution: Split Proteins (Protein Fragment Complementation, Chapter 4) and Domain Insertion (Chapter 5). With the Split Protein approach we were able to couple two fragments of a split dihydrofolate reductase (DHFR) to the conditionally interacting proteins FRAP and FKBP12, resulting in a rapamycin-dependent metabolic enzyme that can be used to control the folate biosynthesis pathway and consequently the growth rate. With the Domain Insertion approach, we created enzyme-regulatory domain chimera consisting of 2-Isopropylmalate synthase (LeuA) and murine DHFR as enzymes and the maltose binding protein MBP as regulatory domain. We isolated functional proteins, but could so far not identify a variant that is sensitive to the effector. We optimized the protocol to an extent that libraries of thousands of strain variants expressing potentially switching enzymes can be generated and the screening for strains of interest became the limiting factor. In a third part of this work we therefore evaluated the fluorescent single cell growth rate reporter TIMER for its utilization in E. coli and especially to enrich slow growing cells out of large genetic variant strain libraries in a high-throughput manner using fluorescence-activated cell sorting (Chapter 6). The herewith developed enrichment method is planned to be applied in the future to strain libraries created with the Domain Insertion library approach