The selection of cell populations harbouring stably integrated, exogenously introduced DNA is commonly carried out using a strategy of antibiotic resistance. The mode of selection involves the simultaneous introduction, alongside the gene of interest, of an antibiotic-resistance gene that promotes cell survival when grown in the presence of the target antibiotic. Puromycin, a potent inhibitor of protein synthesis not only in microbes, has been widely used for this purpose for over 30 years in conjunction with the dominant selectable marker, puromycin N-acetyltransferase (pac), an enzyme from Streptomyces alboniger. Where there is a need to bias the derivation of cell lines in favour of those expressing high levels of the protein(s) encoded by the gene(s) of interest e.g. therapeutic antibodies, stringent selection strategies have been developed. These have included expressing the selectable marker off a sub-optimal promoter element in the case of the widely used selectable marker, glutamine synthetase. Given the high activity of pac, we used a structure-based approach to engineering pac to be a less effective enzyme in order to confer a greater stringency in stable cell clone production. We have solved crystal structures of the acetyl-CoA bound and the puromycin/acetyl-CoA complexes. There appear to be no catalytic residues that can participate in the acetyltransferase reaction, so a series of mutations have been designed that interfere with the binding of either substrate. These mutations show various levels of activity relative to the wild-type enzyme and still maintain the ability of select cell populations harbouring the marker.