Expression hosts in outline

Host choice is a statement about disulfide bonds, sugars, and how much protein the experiment needs. Escherichia coli, yeasts such as Pichia or Saccharomyces, insect cells with baculovirus, and mammalian cells such as HEK293 or CHO lines are all routine research hosts. They are not a ranking of vendors, and a faster host is not a better protein. This guide is the outline a laboratory can use before anyone orders a litre of medium. It is research education, not a manufacturing licence.

How you will know the polypeptide was made is in reading a protein gel. Why a bacterial pellet is often the address of that polypeptide is in codon use and insoluble expression. Capture after the culture is in affinity and ion-exchange protein purification.

What the host is physically able to do

A host is a set of compartments and enzymes. Bacteria make a lot of protein quickly and do not build mammalian N-linked glycans. Their cytoplasm is reducing, so disulfide bonds that would form in a eukaryotic secretory pathway often fail to form unless you redirect the chain to the periplasm or use a strain built for oxidative folding. The outer membrane of a Gram-negative bacterium carries lipopolysaccharide. That molecule comes along as endotoxin unless a later process removes it. None of this makes E. coli a poor host. It makes E. coli the right host when the chain can fold without those eukaryotic extras and when the assay can live with a bacterial background, or when you have planned the cleanup.

Yeast grows densely and can secrete. It glycosylates, and the sugars are its own: often high-mannose, immunogenic in some animal experiments, and perfectly acceptable when the sugar is not the question. Yeast does not contribute bacterial endotoxin. It contributes yeast proteins, media components, and its own proteases. Secretion can simplify purification and can also stall if the protein is foreign to the secretory pathway.

Insect cells infected with a baculovirus vector are a common route to intracellular or secreted protein with more of the eukaryotic folding machinery, including some disulfide formation in the secretory path. Their N-glycans are often paucimannose rather than fully complex mammalian structures. Yields can be attractive relative to mammalian culture. The workflow is slower than a bacterial overnight, and the virus is a biological agent your institution has to authorise.

Mammalian cells provide the closest routine match to human-like disulfide pairing and complex glycosylation, especially for proteins that must be secreted and recognised by mammalian receptors. Transient HEK293-style expression and stable CHO-style expression are different operational commitments. Both are slower and usually lower yielding than a happy bacterial fermentation. They are the rational cost in time when the modification is the point of the experiment.

Match the constraint, then the scale

Start from the protein, not from the incubator you already own. Look up the sequence on a public record such as UniProt and mark cysteines that form known disulfides, signal peptides, and glycosylation motifs. If the active protein is an extracellular domain with disulfides and complex glycans, bacterial cytoplasm is a hypothesis you should expect to fail, or to succeed only as an unfolded mass. If the protein is a cytosolic enzyme with no disulfides, E. coli is a fair first host, and insolubility is then a folding and induction problem rather than a missing organelle.

Yield is the second constraint. Structural work and some binding assays consume more mass than a single blot. A mammalian system that produces a small amount of authentic glycoform can still be the correct choice if that small amount answers the question. Scaling a host that makes the wrong molecule produces more of the wrong molecule. Write the amount as a use: enough for a gel and a binding assay, or enough for a crystallisation screen. Leave currency out of the scientific sentence.

Then ask what must be absent. Endotoxin matters when cells or animals will see the protein and the readout is sensitive to it. A clean SDS-PAGE lane does not answer that. DNA contamination, detergents, and the elution buffer matter when the next assay is kinetic. The host decision and the buffer decision travel together.

Branches when the first host misbehaves

If E. coli makes inclusion bodies, you have three research branches, not a verdict that the gene is bad. Lower temperature and milder induction are empirical levers, described with codon myths in the codon article. A solubility tag or a periplasmic route is a construct change. A move to yeast or insect cells is a host change, justified when folding needs eukaryotic machinery rather than a slower ribosome. Refolding from inclusion bodies is a separate process with its own failure rate. Choose it when the ungycosylated chain is acceptable and the laboratory has a refolding assay, not as a slogan.

If yeast secretes a smeared, hyperglycosylated band, decide whether the smear is the biology you wanted. Deglycosylation as an analytical check can show the polypeptide underneath. It does not turn the prep into a mammalian glycoform. If insect or mammalian expression yields almost nothing, confirm the construct, the signal peptide, and the viability of the culture before you condemn the sequence. A toxic protein will look like a failed transfection.

If the gel shows a strong band and the activity is absent, the host may have made polypeptide without the disulfide or the cofactor. Activity, not the band, is the branch condition. Keep the soluble and insoluble fractions on the same gel so a "no expression" conclusion is not just a soluble-fraction conclusion.

Failure modes that are really constraint mismatches

The common failure is scaling a silent culture. No band in the soluble fraction can mean no expression, or it can mean the protein is in the pellet. Run both. The second failure is accepting a glycan you did not characterise. A sharp band from yeast is not proof of a mammalian-type sugar, and a broad band is not proof the protein is degraded. The third is importing a host the biosafety rules have not named. Baculovirus, mammalian lines, and antibiotic-resistant bacteria are institutional decisions. A methods paragraph does not authorise them.

Codon optimisation will not install a glycosylation pathway. A purification tag will not create a disulfide bond. Those tools belong after the host can, in principle, make the molecule you will assay.

Safety is an institutional decision

Live hosts are biological agents. The World Health Organization laboratory biosafety manual is a public frame for how institutions think about containment. It is not a permit for your building. Follow the biosafety level, the disinfection, and the waste rules you were given. This outline is not medical advice, not a vaccine process, and not an approval to express toxins or to infect animals. Chemical inducers and antibiotics are handled under the chemical assessment as well.

Power, heat, and a culture that is also a cold chain

Bacterial expression at reduced temperature needs a shaker that actually holds that temperature. A power cut returns the flask to a hot room and changes the induction you planned. Insect and mammalian incubators are less forgiving of a long outage than a stationary bacterial pellet. Record the outage, and do not pool that batch with an uninterrupted one.

Strains and viral stocks arrive as a cold-chain handoff. A vial that thawed in transit is a different reagent. Quarantine it from the working freezer and tell the sender. Glycerol archives and freezer practice are a separate habit. They matter here because the host you chose is only as good as the clone you thaw.

What the enquiry needs

State the amino-acid sequence or a public accession, the disulfide and glycan constraints, the host you prefer and why, the evidence you want on a gel, and the buffer the assay requires. Mention endotoxin only if the readout needs it. Reagent classes for media additives and buffers are in the reagents and chemicals catalogue. Host and purification scope can be discussed as an enquiry through the custom protein expression and purification reference. Repeat the constraints in the quote request. A host name without the constraint is not a brief.