Metabolic engineering

A microbial cell is already a chemical plant: thousands of enzyme-catalysed reactions, coupled through shared pools of metabolites, turning substrate carbon into biomass. Metabolic engineering redirects that traffic. The plant metaphor carries because the underlying logic is the same — conservation of mass, capacity limits, control loops — which is why a production strain can be designed on paper at all, and why the paper is never quite right.

The cell as a network, not a list of genes

Genetics lists parts; metabolism connects them. An internal metabolite at steady state obeys one rule: whatever flows in must flow out. Every reaction the cell runs is a term in a mass balance, and all the balances must close simultaneously. Flux balance analysis formalises this: given the network’s stoichiometry and a substrate uptake rate, it solves for fluxes that satisfy every balance while maximising a chosen objective, usually growth. The output is an upper bound — the maximum carbon that can reach the product without violating mass conservation. The bound earns its keep by killing impossible designs cheaply and revealing which knockouts free up flux. But it assumes steady state, ignores enzyme kinetics, concentrations and regulation, and knows only what it is told. The optimum it reports exists in a cell that obeys linear algebra, not in a fermenter.

Balancing a pathway

A heterologous pathway is a series pipeline. One enzyme makes an intermediate and the next consumes it, so the pipeline moves at the speed of its slowest stage — and if an upstream stage outruns a downstream one, the intermediate accumulates. Many intermediates are worse than useless: aldehydes and CoA thioesters react with proteins, organic acids cross the membrane and unload the proton gradient, and anything water-soluble leaks out of the cell. Balancing is therefore about ratios, not strength: promoters and ribosome-binding sites are tuned so that each stage keeps up with the previous one, and the same physics explains why “overexpress everything” fails.

The stage that most often sets the titre is the last one. Enzymes evolve specificity for their natural substrates; the final acceptor in an artificial route is frequently a molecule the enzyme has never seen, so its kinetics and stability are mediocre. The product itself can poison the step through toxicity, or through a near-equilibrium reaction that only proceeds if the product is kept low. Hence the recurring shape of published fixes for a stalled pathway: screen homologues of the terminal reductase.

Regulation fights back

The host is not inert material but an optimised survivor that spends regulation defending its own flux map. Feedback inhibition closes entry-point enzymes as soon as the product pool fills. Global control reallocates ribosomes, nitrogen and reducing power away from anything that does not serve growth — and the inserted pathway competes for exactly those resources. The modern repertoire works with this rather than against it: sensors that switch on production only after the biomass phase, growth coupling that ties product formation to growth itself, deletion of degradation routes. The honest summary is that the stoichiometric model supplies the map, and the build-test-learn loop is how the territory corrects it. The limiting case of that argument is a reaction vessel with no regulation left to fight — cell-free biosynthesis.