Porting a comment I made on The Panda’s Thumb:
(I)n another paper by Axe and Ann Gauger (Model and Laboratory Demonstrations That Evolutionary Optimization Works Well Only If Preceded by Invention—Selection Itself Is Not Inventive” BIO-Complexity 2015 (2):1–13. doi:10.5048/BIO-C.2015.2), the authors asked if two variants of a canonical TEM beta-lactamase - one inactive due to targeted deletion of the active site, and the other the low-activity, temperature-sensitive variant that was used in Axe’s 2004 JMB paper - could be optimized by repeated rounds of amplification and selection. In Tables 1 and 2 of this report, they state that the lowest concentrations of ampicillin that completely inhibited growth of cells expressing the respective constructs had overlapping ranges (5-10 µg/mL and 10-45 µg/ml for the inactive and low-activity variants, respectively). Moreover, the ranges of highest concentrations of ampicillin that permitted appropriately-transfected E. coli to grow (5-7 and 5-40 µg/mL, respectively) also overlapped. These concentration ranges include the concentration (10 µg/mL) Axe used in his 2004 paper. The ramification of this is that the assay Axe used in the 2004 study cannot accurately distinguish between constructs (plasmids) encoding enzymes with very low activity and plasmids that encode enzymes devoid of any activity. He does not know if variants scored as nonfunctional may actually have low (say, 2 or 5-fold lower than the starting enzyme) activity, or for that matter if variants scored as active are actually inactive. In my opinion, since the purpose of the 2004 study was to enumerate functional variants (even those with much lower activity than the initial enzyme) as well as non-functional variants, the fact that the assay cannot reliably discern between the “starting point” and controls with no activity (let along variants with slightly lower activity) renders the entire study meaningless.