A library of variant enzymes was created by combined shuffling of the DNA encoding the human Mu class glutathione transferases GST M1-1 and GST M2-2. The parental GSTs are 84 % sequence identical at the protein level, but their specific activities with the substrates aminochrome and 2-cyano-1,3-dimethyl-1-nitrosoguanidine (cyanoDMNG) differ by more than 100-fold. Aminochrome is of particular interest as an oxidation product of dopamine and of possible significance in the etiology of Parkinson's disease, and cyanoDMNG is a model for genotoxic and potentially carcinogenic nitroso compounds. GST M2-2 has at least two orders of magnitude higher catalytic activity with both of the substrates than any of the other known GSTs, including GST M1-1. The DNA library of variant Mu class GST sequences contained "mosaic" structures composed of alternating segments of both parental sequences. All clones contained the 5'-end of a GST M1-1 clone optimized for high-level expression in Escherichia coli. The remainder of the sequences derived from segments of GST M2-2 and GST M1-1 DNA. All of the clones analyzed contained between two and seven distinct DNA segments. In addition, each clone contained an average of approximately one point mutation. None of the library clones analyzed was identical with either of the two parental structures. Variant GST sequences were expressed in E. coli, and their enzymatic activities with aminochrome, cyanoDMNG, and 1-chloro-2,4-dinitrobenzene (CDNB) were determined in bacterial lysates. Such screening of more than 70 clones demonstrated a continuous range of activities covering at least two orders of magnitude for each of the substrates. For a given clone, the activities with aminochrome and cyanoDMNG, in spite of their different chemistries, were clearly correlated, whereas no strong correlation was found with CDNB. This functional correlation suggests a common structural basis for the enzymatic mechanisms for conjugation of aminochrome and denitrosation of cyanoDMNG. From an evolutionary perspective, the results show that recombination of segments from homologous proteins gives rise to a large proportion of functionally competent proteins with a range of activities. The data support the proposal that natural evolution of protein functions may involve recombination of DNA segments followed by selection for advantageous functional properties of the resulting proteins. Clearly, the same approach can be utilized in the engineering of proteins displaying novel functions by in vitro evolution.
Copyright 1999 Academic Press.