I don't know exactly how you're doing your assay but my guess would be the variability is mostly because it's being done over an extraordinarily long expression period rather than the initial clone you pick.

I agree with the first answer that your assay period seems way too long. It's not clear if you still have the cells in your medium while you measure the protease activity and if your protein is secreted to the medium. Do you have a tag on your protein or a specific antibody to see if there is any expression in the first place? If you can't find a substrate that can be used for a faster assay then the enzyme activity is probably not very high or the protein is mostly inactive.

How do you know it's toxic? Did you try any of the strategies recommended for toxic proteins except for using the C41 strain? e.g. https://www.embl.de/pepcore/pepcore_services/protein_expression/ecoli/decreasing_protein_toxicity/index.html

Hüseyin Besir I have tried mostly all of those strategies and they never seem to make a difference. I speculate it's toxic since it's supposedly a digestive protease. The only reason I incubate for so long is to try to expose the enzyme to the substrate for as long as possible, though I'm sure activity stops way before then. I guess my main concern is why there is so much variability in expression from clone to clone, and what I can do to minimize this variability so I can actually get a general idea of how well the enzyme performs against the substrate. For example, one clone may show significant degradation by 96 hours, but another shows nothing.

I recommend showing that your clones express the protein in detectable amounts in the first place. The fact that it's toxic can have the following consequences which may explain the observed variability:

- The timing of your induction could be slightly different, so at different growth stages the expression level might be different, hence the variable expression levels.

- Another effect of toxic genes is the observation that your clones contain random mutations that could make them less active or inactive, so only those cells would survive that do not express the active enzyme. I have seen this once with a kinase domain that was so toxic that even cloning was difficult and gave only a few colonies containing mutated genes resulting in inactive kinase.

Hüseyin Besir I have been using a toxic cell line (BL21-C41) that seems to work okay, and I always retransform fresh as this seems to be an issue for others who are trying to do protein expression.

This particular C41 cell line contains the pLysS plasmid which contains the chloramphenicol resistance gene, so that's an added selection factor. Typically I haven't added chloramphenicol to the cultures because I'm not sure how much of a difference it will make once expression is already underway, but perhaps it will help maintain the "correct" clones vs. incorrect ones, as I agree selection would favor cells that don't produce the protein.

I have confirmed via proteolytic assays that my cells do produce a protein that does what it's supposed to (it's a protease specific to a particular substrate). So overall I'm at a loss.

It would still be more convincing to see how much of the protein is actually produced. Is it a protease that needs disulfide bonds so it should be produced as secreted protein (many proteases are not easily expressed in the cytoplasm)? If you have to incubate a substrate for several days to get it proteolysed, I would say your enzyme is not more than just detectable by its (variable) activity but not enough to see a specific band on SDS-PAGE of even purify in reasonable amount. That means that E.coli might not be the best expression system.

Really, if you have problems with measuring activity, just grow the cells to high density, induce for short period of time, collect cells and medium, concentrate and then measure the activity (max. overnight).

The right way to grow cells for expression (not that I would have done that ever) is to grow various clones, pick the best one, grow it in large volume and the freeze aliquotes (without induction). Next time just get out one aliquot and you will always have the same clone grown to the same density (originally). That should minimize some variability.