What does electrophoretic variation tell us about protein variation?

Before the introduction of DNA sequencing methods into population genetics by Kreitman (1983), the major source of information about genetic variation among organisms in natural populations came from electrophoretic studies of proteins. The amount of information acquired during 25 years of major electrophoretic activity is staggering and, for purely technical reasons, is likely to remain orders of magnitude greater in number of genes, number of individuals, and number of species examined than can be acquired from nucleotide sequence data. It would be extremely desirable if this mass of data on electrophoretic phenotypes could be translated into data on patterns of amino acid variation, using the information now available for a handful of cases where amino acid variation in natural populations has been definitively ascertained from DNA sequence variation. It is the purpose of this paper to show that while the total amount of amino acid variability can be inferred from electrophoretic data, the different patterns of electrophoretic class frequency distributions do not allow any other inferences, because those patterns can be generated by the simplest null model of the generation of electrophoretic variation from amino acid variation, without the need to make any assumptions about selection or population structure. The charge-state model (or the stepwise mutation model) was originally proposed by Ohta and Kimura (1973) to deal with electrophoretic data. This model assumes that only amino acid changes resulting in significant charge changes are clearly detectable. Hence gel electrophoresis ideally detects changes in net (integer) charge in a protein but not in fractions of charge. Although the model is somewhat simplistic, Brown, Marshall, and Weir (198 1) concluded after a review of the existing evidence that there was general support for the model, which can be considered at one end of the spectrum of models describing electrophoretic variation.