Artificially evolved variants of proteins with roles in photosynthesis may be selected most conveniently by using a photosynthetic organism, such as a cyanobacterium, whose growth depends on the function of the target protein. However, the limited transformation efficiency of even the most transformable cyanobacteria wastes much of the diversity of mutant libraries of genes produced in vitro, impairing the coverage of sequence space. This highlights the advantages of an in vivo approach for generating diversity in the selection organism itself. We constructed two different hypermutator strains of Synechococcus sp. strain PCC 7942 by insertionally inactivating or nutritionally repressing the DNA mismatch repair gene, mutS. Inactivation of mutS greatly increases the mutation rate of the cyanobacterium's genes, leading to an up-to-300-fold increase in the frequency of resistance to the antibiotics rifampin and spectinomycin. In order to control the rate of mutation and to limit cellular damage resulting from prolonged hypermutation, we placed the uninterrupted mutS gene in the cyanobacterial chromosome under the transcriptional control of the cyanobacterial nirA promoter, which is repressed in the presence of NH 4 ؉ as an N source and derepressed in its absence. By removing or adding this substrate, hypermutation was activated or repressed as required. As expected, hypermutation caused by repression in PnirA-mutS transformants led to an accumulation of spectinomycin resistance mutations during growth.Artificial evolution (sometimes called directed or laboratory evolution) has, in recent years, been applied to a wide range of individual genes, successfully changing properties of encoded proteins ranging from thermotolerance, to catalytic specificity, to total activity (reviewed in references 3, 21, 29, 35, and 39). Using artificial evolution, whole organisms also can be evolved rapidly to exhibit new and complex traits (25,40).Techniques for artificial evolution typically entail a mutagenesis step, in which sequence diversity is generated in particular genes, and a step in which phenotypic expression of variant genes confers the desired characteristic that is selected. These techniques range from entirely in vitro procedures (for example, in vitro co-compartmentalization of an evolved gene and its protein product in aqueous droplets emulsified in oil [10]) to fully in vivo procedures. One of the major factors that limit the efficacy of artificial evolution is the number of permutations of a particular sequence that can be screened for the desired functional characteristic (sequence-space coverage). In many artificial evolution protocols, mutagenesis is performed in vitro and selection is performed in vivo. When the selection must be done in a photosynthetic organism, such as a cyanobacterium, to select variants of proteins with roles in photosynthesis, for instance, transformation efficiencies are typically low (8), and this wastes much of the generated sequence diversity. This loss can be avoided by generating the ...