Exploiting Genetic Interactions in Metabolic Engineering

By assembling parts with given properties, engineers are able to build tools and devices of increasing complexity. For example, a simple Ikea table is assembled from pieces of wood, screws and bolts with
precisely designed dimensions, diameters and thread angles. The properties of these parts do not change upon assembly, allowing for a predictable outcome: a table.

This simple reductionist principle is also central to synthetic biology, the field interested in engineering living organisms. In synthetic biology, DNA sequences encoding biological parts found in the living world are copied, assembled and reintroduced in other organisms to engineer new biological functions. For example, the enzymes taking part into a metabolic pathway synthesising a plant pharmaceutical ingredient
can be expressed in the yeast Saccharomyces cerevisiae allowing for its synthesis by cultivating the modified yeast in a bioreactor.

The applicability of the reductionist approach underlying traditional engineering disciplines is however limited in synthetic biology. Indeed, the properties of biological parts are often dependent on the context
of the host they are expressed in. For example, all parts expressed in a living host are synthesised from a family of building blocks (e.g. amino acids, lipids, sugars) which availability is not infinite. Cross-talk
between parts might arise simply by competing for the same pool of resources. These interactions limit the predictability of designs and, in practice, many iterations of trial-and-error are often needed to engineer
a synthetic biological mechanism.

In experimental genetics, the unexpected outcome arising from the combination of two genetic modifications is known as a genetic interaction. Genetic interactions can be complex and often elude our understanding, even in well studied model organisms. Yet, in rare instances, they produce a surprisingly strong phenotype hinting that they may be exploited in an engineering context.

The motivation of this thesis is to turn interactions into a source of improvement for a desired biological function. Indeed, interactions between a metabolic pathway and a modification of a host gene may lead to higher product synthesis. In this thesis, I first attempt to predict interactions with machine learning. Failing to do so, I turn to a high throughput screening to identify positive interactions between a metabolic pathway and all non-essential genes in S. cerevisiae. More specifically, we use CRI-SPA, a new high throughput gene delivery method, to deliver a metabolic pathway in all the strains of either the Yeast Knock Out or the Yeast Over-expression Libraries. This systematically tests for the presence of an interaction between a gene (its absence or over-expression) and the metabolic pathway which might improve its yield. We show that this method can identify positive interactions in two case studies improving the synthesis of the plant pigment betaxanthin or that of the platform chemical cis-cis-Muconic acid.

Altogether, this work shows that genetic interactions can be used to improve a desired engineered trait in synthetic biology and advocate for a shift from its original reductionist dogma.