Engineering Bacterial Cells Combing Bottom-Up� and Top-Down Approaches: E. coli Threonine Production Model

Introduction. Progress in synthetic biology over the last 70 years has enabled scientists to move from using microbial strains to make specialty chemicals to using organisms with fully synthetic genomes and re-coded translational machinery. Now, we are ready to engineer artificial organisms to study fundamental biological problems and pursue multiple industrial applications.

Design. General strain engineering strategy for industrial applications:

1. Complete bottom-up design of multiple biological samples representing combinations of varying genetic constructs, host modifications, expression levels and growth conditions to produce different amounts of a target product.
2. Generate mid- and high-throughput datasets (metabolites and RNA-SEQ data) using these samples and machine learning (ML) models.
3. Construct and test subsequent samples to reproduce ML-generated expression signatures corresponding to the predicted high-production yield.

Results.

1. Sequenced ten industrial “low-level” threonine-producing coli strains and mapped their genomic changes. Generated eight differentially expressed RNA-Seq samples and characterized the relative RNA abundance of target genes.
2. Constructed deletion of ThrABC and asd genes from three hosts in combinations with deletions of another ten genes likely to negatively affect threonine yields.
3. Constructed five redesigned artificial operons coding for a “core” threonine production and carrying genes from coliand C. glutamicum and placed them under tac promoter. Separately cloned additional set of genes for functions likely to enhance threonine production.
4. Developed controlled fermentation media to grow first hundred combinations of the above components of the system and tested the combinations for threonine production and production of all other amino acids and several other metabolites.

Conclusions. We are currently testing additional combinations of the constructed strains, their regulation levels, hosts, and fermentation conditions. In some of the cases, even in the first round of testing, threonine yield exceed yields observed in the original producing strains. We are also starting RNA-Seq data generation and ML model construction.