Nature 403, 617–622 (2000).

This laboratory reported the in vitro evolution of an enzyme with phosphoribosyl anthranilate isomerase activity (ivePRAI) from the α/β-barrel scaffold of indole-3-glycerol-phosphate synthase using a combination of rationally designed libraries, DNA shuffling, and selection with Escherichia coli, JA300, a strain that lacks an active PRAI gene. As part of the ongoing project to characterize the structure and properties of ivePRAI, we discovered that the protein expressed from a variety of vectors that contained a synthetic gene corresponding to the sequence of ivePRAI as published is insoluble and does not complement JA300 (R. L. Weinberg, C. M. Blair and A.R.F., unpublished results), as reported. We conclude that the results are unsound.

It appears that the discrepancy in the results is due to a combination of two episodes of cross-contamination. We are now repeating the directed evolution of ivePRAI using modified procedures to test the design strategy that should eliminate the source of errors of contamination.

The first author of this Article (M.M.A.), who was responsible for most of the analysis and design strategy involving loop transfer and most of the experimental work, wishes to be dissociated from this retraction because she believes that the experimental data are fundamentally sound.