(626d) Improving Glycosylation Yield in Escherichia Coli Using An Adapted Genome-Scale Metabolic Model and Metaheuristic Optimization Techniques
Protein glycosylation, the post-translational covalent attachment of oligosaccharide groups to specific amino acid residues, is required of many therapeutic proteins. This modification affects various protein properties including pharmacokinetic activity and immunogenicity. Currently, eukaryotes possessing native glycosylation machinery serve as the preferred production host of therapeutic glycoproteins. The discovery of bacterial glycosylation in the pathogen Campylobacter jejuni and the subsequent transfer of these pathways into E. coli has spurred interest in producing non-native glycans in a more genetically tractable host. Recently, human-like glycan production was achieved in E. coli with the recombinant expression of a synthetic glycosylation pathway. Producing glycoprotein from prokaryotic hosts continues to suffer from several limitations including insufficient yield. Glycosylation in E. coli partly depends upon the available pool of nucleotide sugars making up the desired oligosaccharide. Removal of competing metabolic pathways and bottlenecks that diminish this pool may improve glycosylation efficiency. The design of strains displaying improved glycoprotein production may be aided greatly by the incorporation of a metabolic network modeling strategy. Genome-scale metabolic reconstructions of industrial microbes are commonly used to identify genetic perturbations that will produce a desired biochemical production phenotype. In this study, we modify the existing genome-scale E. coli model iAF1260 (1260 open reading frames) to include a variety of glycosylation pathways, including those for C. jejuni and human-like glycans. Added reactions include the biochemical transformations associated with glycan biosynthesis and flipping into the periplasm, as well as the expression and glycan conjugation of a target protein. The adapted network consists of 3601 reactions, 1290 open reading frames, and 2233 species segregated into cytosolic, periplasmic, and extracellular compartments. Optimal gene knockout designs were identified by solving a bilevel optimization problem which sought to couple glycan production to a cellular growth objective. For a given set of gene knockouts, flux balance analysis was used to calculate the flux distribution that optimized the cellular objective of maximizing growth rate. A genetic algorithm was applied to generate strains and select those that maximized the engineering objective, a function of glycan production. This approach has the advantage of generating strains with variable number of knockouts and has the ability to accommodate non-linear engineering objectives. We discuss characteristics of strains that couple growth to glycan production and explore metabolic engineering strategies that may improve glycosylation efficiency in a recombinant prokaryotic host.