In my personal experience bacteria are just easier to work with. Selections are easy and its possible to make loads of protein (100+ mg/litre shaking culture), especially if fermentors are used. If it works in bacteria, its the way to go. Also, the available strains of bacteria with different traits is quite diverse nowadays to overcome different problems. There's also a lot of different media and working habits to control expression, protein stability and solubility etc.

However, membrane proteins, glycosylated proteins, proteins with a lot of disulfides and massive single-domain proteins can prove to be very challenging to work with in bacteria. And of course, in science its not always the goal to make as much protein as possible but to make what is relevant. There's no point in producing misfolded or unfolded protein for biochemical or structural studies if that is not the (near-)physiologically relevant form. Yeasts, insects and other eukaryotic systems come into the picture here, but these are much more laborious. S. cerevisiae, for instance, can be very handy with membrane proteins, but selection is difficult. One can use synthetic media (which are very expensive and the cell growth is not very fast) to select but to be able to produce adequate amounts of protein one usually ends up using complex media (like YPD) and then loses selection. This is a major problem with other than integrative plasmids, as plasmid stability in yeast is quite often low. Insect cells (Sf9, Sf21 etc) require already much more culture upkeep and maintenance although the baculovirus system has worked quite well for my colleagues. Mycoplasma is a problem, as for all eukaryotes. Also, the glycosylation performed by insect cells is not always the same as in mammalian systems, for example.

In vitro translation then... well it does translate but there are no PTM's unless you supply the reaction with the machinery capable of these. Also, since it's a cell-free system, there is practically nothing present that degrades your protein. I guess it's quite difficult to produce bucketloads of protein with cell-free systems.

Pichia pastors for cysteine rich proteins, almost as easy to work with as E. coli and I know that I do not get aberrant disulfide bonds. Otherwise I use E. coli.

E. coli wins as a matter of time, practicality, and money. There are many more expression and selection systems in E. coli than yeast, its harder to break yeast than bacteria, harder to avoid yeast proteases, and the doubling time in E. coli is at least 4 times faster. Under optimal growth conditions the doubling time for E. coli is ~20 minutes versus Saccharomyces cerevisiae, which is ~ 1.3–2 hours at 30 °C and Schizosaccharomyces pombe, which is 2-4 hours. Unless it’s necessary, there is no reason to avoid using E. coli.

It seems clear that E. coli is the first choice for all the reasons stated above. I am interested in how often it is necessary to try another system and what the reasons for that might be. Does anyone ever go straight into using a eukaryotic host without trying E. coli first? In the eukaryotic membrane protein field, Pichia pastoris is very popular for example.

Structural biologists are often best at making large amounts of high quality proteins. Structural biologists focused upon mammalian protein complexes will often start with E. coli and baculovirus expression. The advantages of insect cells over E. coli can include better solubility, post-translational modifications, and higher yields at the expenses of more effort and cost. Yeast is sometimes an advantage and I have used yeast expression when I had a good yeast system and E. coli was not working as well. I have also tried Pichia and it is not proven to be worth the trouble for my systems, probably a lab using Pichia regularly can address its value for them. With the help of a colleague who was good at working with yeast I have gotten my best yields and active human enzymes from yeast. Working with my postdocs yeast is generally less successful.

This has been a new field for me during the early times of my postdoc and E. coli has own over my choice. I guess there are some toxic proteins and some mammalian ones too that sometimes need to be codon optimized for expression in bacterial host. But despite that, bacteria are the easiest, most convenient and economically efficient host for this platform. Even so now since there are vectors easily available with different modifications for in vivo disulphide linkages and proper protein folding to be achieved inside the bacterial host. Further, the secretory systems of these host cells can be easily modulated and controlled to have the recombinant protein in different compartments as and when required.

Plants represent a very promising platform to produce many recombinant proteins and it is easily scalable, free of human pathogens and it has very transcription and translation machinery complexes

Thanks for all the input so far. Mattanovich has stated that approximately 20% of biopharmaceutical proteins are produced in yeasts (the vast majority in S. cerevisiae), 30% in E. coli and 50% in mammalian cell-lines and hybridomas. However, E. coli is by far the most commonly used host in research labs (some articles suggest 60-80% of proteins are made in E. coli). How do these two observations stack up?

I have just published this perspective article that may be of interest to those of you who have contributed answers to my question: http://journal.frontiersin.org/Journal/10.3389/fmicb.2014.00085/abstract

The company for which I have started working is aiming to express complex proteins in soybeans, which have the potential advantages of high protein levels in their seeds (~40%), the ability to do many necessary post-translational modifications consistently, ease of isolation (only a few dozen proteins in seeds), and ability to store at ambient temperature for years.

The use of plants also provides opportunities for edible vaccine generation! The cholera toxin B subunit, immunoglobulins, α-interferon, VP1 protein from foot-and-mouth disease virus and glycoprotein S from transmissible gastroenteritis virus all have been expressed in transgenic plants or by means of plant viruses; for example see Nature Reviews Genetics. 4:794-805 (2003).

Lately I'm more familiar to S. cerevisiae to produce recombinant proteins for vaccine generation. It really interest me its metabolism, easy to use and most of all I think it's because is already recognized as safe by the FDA. I've read some of your papers about DoE and I'm currently doing the same to overexpress some viral proteins!

Thanks for your input, Miguel, and good luck with your experiments.

After numerous attempts to express soluble protein in E. coli I switched to a P. pastoris production system and yielded great results, achieving excellent and highly pure yields. During this process, I attempted expression in various E. coli strains (, and various expression methods (such as glucose supplementation at the point of induction and induction growth at various temperatures), but the protein expressed was constantly in an insoluble form. I was able to gain soluble protein upon extraction and purification using urea, but of course this leads to necessary refolding steps (and then you're relying on the protein refolding to the correct configuration and not returning to an insoluble state upon removal of urea). There were also greater contaminating E. coli proteins in my purified protein samples, whereas I gain great purity using a Pichia expression system.

It is, however, worth pointing out that I was trying to express a small protein containing 3 disulphide bonds (thus it could be considered disulphide rich).

My reason for trying an E. coli expression system initially was essentially because it was a system my lab were familiar with, the system is cheap, it has selectivity, and it is reasonably quick to attempt various optimisation methods, etc. That said, I now find the Pichia system to be a dream to work with, with excellent expression yields achieved. Although there is no antibiotic selectivity in this method (thus a high level of sterility is required), the use of the PichiaPink strain allows for selectivity of transformed colonies, making selection of successfully transformed colonies much more efficient than it would otherwise be. Interestingly, most of my lab now express their proteins in Pichia pastoris, simply for the high expression and very pure yields gained.

I feel system preference can largely depend on the protein being expressed, as there are proteins that can efficiently be expressed in bacterial systems, thus making them potentially cheaper and quicker. Some proteins, on the other hand, require a switch to a different expression system in order to gain soluble protein. I would say its always worth starting with an E. coli system first (again, as this is quick and simple), and then switch to another system as required (unless a lab already has resources in place for using a different expression system, or if it is already known that the protein in question expresses better in another system).

Hope that helps and good luck to anyone working in the realms of protein expression.

Thanks, David.

You may be interested in my latest article in Current Opinion in Structural Biology.

http://ac.els-cdn.com/S0959440X1500055X/1-s2.0-S0959440X1500055X-main.pdf?_tid=b2898c00-4c40-11e5-889c-00000aacb361&acdnat=1440627826_7c033b497f5d156ff68d686c57b127eb

Those of you who are still interested in this string, may find our recent article in Methods helpful.

http://www.sciencedirect.com/science/article/pii/S1046202315301134