I keep getting pulled into “which host” conversations and they keep landing in the same place, so I’m writing it down. This is for small-molecule and protein production, not for making whole cells as the product.
The default answer in industry is E. coli — it doubles every 20 minutes, the genetics are 40 years deeper than yeast’s, and the fermentation engineers have already solved your problem at scale. But the default is wrong often enough that it’s worth running through the five questions every time. I’ve personally championed E. coli for a target that needed disulfide bonds and watched the team spend a year fighting inclusion bodies. I’ve also championed yeast for something that didn’t need any of yeast’s eukaryotic machinery and watched titer cap at a fraction of where it would’ve been in E. coli. Both directions are easy to get wrong.
1. Does your product need glycosylation or other post-translational modifications? If yes, E. coli is out. It can’t N-glycosylate (in production-grade strains), can’t do most mammalian PTMs, and even engineered glyco-E. coli is not where you want to live for a real program. Yeast at least gets you a baseline of N- and O-glycosylation, though the glycan structures are not human-like without further engineering (humanized strains exist; they are slower and more expensive than wild yeast).
2. Does your product have disulfide bonds? Both can do it, but yeast’s secretory pathway handles disulfides natively, and E. coli requires either periplasmic targeting (limited yields, finicky) or engineered cytoplasm strains (SHuffle, Origami) that grow slowly. If you have more than two or three disulfide bonds, yeast is usually worth the slower growth.
3. How big is the protein, and is it secreted? Yeast secretes proteins. E. coli mostly does not — secretion into the medium is possible but a research project of its own. If you want your product in the supernatant rather than ground out of a cell pellet, that biases yeast. The flip side: yeast caps out somewhere around 50-100 g/L secreted titer for the very best programs; E. coli intracellular expression can hit higher absolute titers for small, simple proteins.
4. Are you making a small molecule that requires a long heterologous pathway? Here it gets interesting. E. coli gives you faster turnaround for pathway iteration — DBTL cycles in days rather than weeks. Yeast gives you better compartmentalization (peroxisomes, mitochondria, the secretory pathway as a chemical reactor), tolerance to nasty intermediates, and access to cytochrome P450s, which barely work in E. coli. Terpenoids, complex alkaloids, and most natural products end up in yeast for these reasons. Aromatics, polyketides, and amino-acid-derived products often stay in E. coli.
5. What does your scale-up partner already run? This is the question new founders skip. If your CDMO has 50 E. coli fermenters and 2 yeast tanks, and your project has any flexibility in host choice, the CDMO’s installed base matters more than you’d guess. The decision isn’t only “which organism grows my molecule best.” It’s “which organism does the supply chain for my molecule already exist for.”
The honest summary: if your product is a non-glycosylated protein under 50 kDa with few disulfides, default to E. coli. If it’s glycosylated, secreted, or sits in a long oxidative pathway, default to yeast. Everything in the middle is where the actual project work lives — and where I keep losing the bet I think I’m winning.