I have never encountered this strategy for overproducing recombinant proteins. It may be that it doesn't work because the plasmid concentration does not increase in response to increasing antibiotic concentration, or perhaps increasing the concentration of antibiotic is detrimental to the health and growth of the recombinant strain. However, it is true that, as beta-lactamase destroys ampicillin, ampicillin levels may decrease below to the level requered to maintain selection, leading to the appearance of plasmid-free segregants (obviously non-producers), particularly at high cell densities. It also leads to the appearance of satellite colonies on plates. Ampicillin being too readily destroyed by beta-lactamase, it is preferable to use other beta-lactam antibiotics, such as carbenicillin or ticarcillin.

Dear Maria Victorova , I have never investigated this phenomena in my lab but I know that the selective pressure can play an important role in determining the copy number of plasmid thus the increased concentration of plasmid. There are different mechanisms through which plasmids can regulate their own copy number. One of them could be the antibiotics in the surrounding environment which can drive the variations of plasmid copy number as suggested in the following report: https://uu.diva-portal.org/smash/get/diva2:1333633/FULLTEXT01.pdf

Maria - In theory you are correct but in practice, probably not. The answer will vary depending upon the plasmid used and the antibiotic used. The phenomenon of resistance gene amplification is well documented in many eukaryotic systems, generally increasing the number of copies of integrated gene clusters. Often drug resistant cancer cells arise due to this mechanism.

However with plasmids in E. coli it becomes complicated. Most common plasmids used in molecular are already very high copy number (hundreds per cell) and it is unlikely that you could select for higher copy number than they have, nor would you see much of a difference. However if your plasmid is a low copy number, ie less than 10-20, it might be possible but probably would just be easier to clone into a high copy plasmid you already have.

As Pierre Béguin points out though, ampicillin would not be a suitable choice for selection, in fact I don't think any beta-lactam antibiotics would be useful for the same reasons he describes (sorry Pierre but I don't agree with you here, carbenicillin is more stable but still has the same problems just on a different scale). Note that a typical plasmid confers resistance to at least 1mg per ml of ampicillin (I've not tried higher but probably is also true).

I think it could work with some other antibiotic resistance genes. Most preferable would be efflux pumps where there is likely to be a direct correlation between resistance and copy number.