The ability to use proteins in nonnatural environments greatly expands their potential applications in biotechnology. Because nature has not paid much attention to optimizing proteins for in vitro applications under conditions that differ substantially from their natural surroundings, there is generally room for improvement through alterations in the amino acid sequence. The most effective approach to this protein engineering task depends on the level to which the molecular basis for the desired property is understood. Consistently successful "rational" design using site-directed mutagenesis requires a high level of understanding of structure and mechanisms or, alternatively, a particularly simple strategy for obtaining the desired feature. An example of a generally applicable and easy-to-implement protein stabilization strategy is metal ion chelation by specific surface dihistidine sites, which can affect thermal stability as well as the protein's ability to withstand denaturants such as guanidinium chloride. Random mutagenesis, on the other hand, can be effective even when structure or mechanisms are poorly understood, provided one can conveniently screen or select for the property of interest. This approach is illustrated by the sequential accumulation of random mutations that greatly enhance the catalytic activity of a serine protease, subtilisin E, in polar organic solvents. The random mutagenesis approach, which mimics the natural evolutionary refinement process, can be used to "coax" enzymes into tolerating nonnatural environments.