Autogene Selections

Abstract

The evolution of proteins is more difficult than the evolution of nucleic acids both in principle and in practice. While nucleic acid sequence space has a dimensionality of 4ⁿ, where n is the size of the nucleic acid pool (i.e., G, C, A, and T), protein sequence space has a dimensionality of 20ⁿ. Similarly, while nucleic acids can frequently be directly selected for function from a random sequence population, the corresponding methods for the directed evolution of proteins are generally not as robust, in part because of the larger sequence spaces that must be explored, and in part because protein selection requires a translation step that in turn often requires cellular transformation, an inherently inefficient procedure that limits library size. In addition, the requirement for expression of the protein library in a host places limits on the numbers and types of selections that can be performed. Selecting individual colonies on plates is not well-suited to truly high-throughput methods and generally limits library sizes to on the order of 10⁵. Moreover, the complexity of cellular metabolism provides an almost limitless source of potential artifacts to confound the selection of a given phenotype. For example, attempts to evolve an antibiotic resistance element can be thwarted by the evolution of chromosomal resistance elements or by the evolution of plasmid copy number or promoter strength rather than protein efficiency (1,2). While there are frequently work-arounds for many of the artifacts that might be encountered, they nonetheless ultimately limit the phenotypes that can be selected.