Limonene epoxide hydrolase from Rhodococcus erythropolis DCL 14 (LEH) is known to be an exceptional epoxide hydrolase (EH) because it has an unusual secondary structure and catalyzes the hydrolysis of epoxides by a rare one-step mechanism in contrast to the usual two-step sequence. From a synthetic organic viewpoint it is unfortunate that LEH shows acceptable stereoselectivity essentially only in the hydrolysis of the natural substrate limonene epoxide, which means that this EH cannot be exploited as a catalyst in asymmetric transformations of other substrates. In the present study, directed evolution using iterative saturation mutagenesis (ISM) has been tested as a means to engineer LEH mutants showing broad substrate scope with high stereoselectivity. By grouping individual residues aligning the binding pocket correctly into randomization sites and performing saturation mutagenesis iteratively using a reduced amino acid alphabet, mutants were obtained which catalyze the desymmetrization of cyclopentene-oxide with stereoselective formation of either the (R,R)- or the (S,S)-diol on an optional basis. The mutants prove to be excellent catalysts for the desymmetrization of other meso-epoxides and for the hydrolytic kinetic resolution of racemic substrates, without performing new mutagenesis experiments. Since less than 5000 tranformants had to be screened for achieving these results, this study contributes to the generalization of ISM as a fast and reliable method for protein engineering. In order to explain some of the stereoselective consequences of the observed mutations, a simple model based on molecular dynamics simulations has been proposed.