The motivation of this work is to comprehend and overcome challenges in understanding and in design of optimal heterologous metabolic pathways that lead to production of biofuels and other valued biochemicals in microbial hosts. I specifically address the problem that within a designed pathway, introduced enzymes are often inefficient leading to a reduction of metabolic flux and consequently product yield. This enzymatic underperformance can be attributed to either poor catalytic fitness or poor soluble expression in the host. To help develop technologies that remedy these inefficiencies, the field of metabolic engineering was surveyed for current approaches that identify an optimal pathway variant and the limitations thereof. I identified numerous inadequacies in current isogenic and high-throughput pathway screening and optimization methods. Specifically, the amount of time and the number of unique variants tested in current methods is limiting. With the advent of high-throughput deep sequencing technologies, large population-based studies are now feasible which reduce the amount of time and increase the total number of unique variants tested. Therefore, this work set out to utilize this promising new approach to test unique enzyme variants in a pathway. I developed a new deep sequencing approach to study the enzyme levoglucosan kinase (LGK) from L. starkeyi that was introduced into E. coli. LGK converts levoglucosan into glucose-6-phosphate which is then used for microorganism growth. A growth selection was developed such that growth on levoglucosan as a sole carbon source was dependent on an active LGK enzyme, and the change in growth was correlated to the change in enzymatic activity. This method was able to quantify the effect of over 8,000 single point mutations on specific levoglucosan flux. The datasets were able to predict whether a beneficial mutation improved stability or catalytic efficiency. Combining computational modelling with these datasets aided the creation of nine enzyme designs. One enzyme design incorporating 38 mutations was crystallized to learn the structural basis of the beneficial mutations. The best enzyme design had a 15-fold improvement in growth rate and 24-fold improvement in pathway activity. Developing this deep sequencing method illuminated a number of problems and opportunities: 1) growth selections are difficult to design and may not be feasible for enzymes in secondary metabolism, 2) improving the soluble expression of an enzyme is potentially an easy avenue to increase specific flux however, 3) stabilizing mutations often have small trade-offs in catalytic fitness. Therefore, the second project set out to extend the original deep sequencing method to improve soluble expression of enzymes without trading-off catalytic fitness in the absence of a growth selection. Using three solubility screens: yeast surface display, GFP fusion, and Tat export, I screened two enzymes, TEM-1 beta-lactamase and LGK. Deep sequencing was used to quantify the effect of all single point mutations on soluble enzyme production. Classifiers were developed to identify solubility-enhancing mutations from these datasets that maintain wild-type catalytic fitness with an accuracy of 90%. The final project was a small extension of the solubility work where I developed analytical equations for converting the enrichment of a variant to a fitness metric for plate-based screens like the Tat export pathway. Using isogenic and mixed cultures I show that growth rates and survival percentages correlate for plate selections. This will help further deep sequencing-based studies for interpretation of the datasets.
Read
