The Language of Gene Interaction

Abstract

The study of variation and heredity, in our ignorance of the causation of those phenomena, must be built of statistical data, as Mendel knew long ago; but, as he also perceived, the ground must be prepared by specific experiment. The phenomena of heredity and variation are specific questions. That is where our exact science will begin. Otherwise we may one day see those huge foundations of “biometry” in ruins. W. Bateson (1902) VERY soon after the rediscovery of Mendel it was realized that the multilocus nature of inheritance could not be understood solely by examining the action of individual genes and then predicting how these genes would behave in concert simply by combining the separate observations. Frequently genes interact with one another, distorting simple Mendelian ratios and sometimes leading to novel phenotypes. Ninety years ago, William Bateson coined the term “epistasis” to describe this sort of interaction (he appears to have first used the term in 1908, although it is most clearly spelled out in Bateson 1909). Epistasis translates directly to “standing upon,” and was meant to describe the situation in which the action of one locus masks the allelic effects at another locus, much in the same way that complete dominance (the way in which most genes appeared to work at the time) involves the masking of one allele by another. The locus being masked is said to be “hypostatic” to the other locus. At the same time that he was investigating all manner of Mendelian segregation ratios, Bateson was involved in a heated battle with the Biometrical school of genetics, exemplified by the views of Karl Pearson and W. F. R. Weldon (Provine 1971). The Biometricians focused on the continuous range of variation among relatives in contrast to the discrete differences studied by the Mendelians. Eighty years ago, in one of the seminal papers in genetics, R. A. Fisher (1918) conclusively ended the debate by showing analytically that Mendelian segregation was compatible with the Biometrician's laws of heredity (something that had been argued qualitatively by a number of others before Fisher). In this same work, he also established several branches of statistics and laid the groundwork for the entire field now known as quantitative genetics. It was quite a paper. The essential problem with which Fisher had to deal was how to map the discrete segregation of alleles onto the continuous range of measured traits. Fisher noted that in conducting this mapping, it was entirely possible (although, as it was later clear, not likely to his mind), that one would not be able to predict the quantitative phenotype of a particular two-locus genotype by simply adding the effects of two loci together. With undoubted allusion to Bateson's unusual segregation ratios, Fisher called these nonadditive interactions “epistacy.” This term rapidly became simply “epistasis,” and two related, but distinct meanings of the same word entered the geneticist's vocabulary (Wade 1992a). It is perhaps telling that this duality in meaning rarely causes much confusion for those using the term. Mendelian and molecular geneticists tend to use epistasis in the strict sense of Bateson, while evolutionary and quantitative geneticists use epistasis to mean just about any form of gene interaction. After briefly tracing the history of the use of “epistasis,” I suggest that there is good reason to start thinking about eliminating this potential source of confusion. Some of the difficulty in choosing the right word to describe gene interactions results from the lingering conflict between the Biometrical and Mendelian views of characterizing gene action. The resolution to this difficulty points the way to the possible final synthesis of these two approaches. epistasis 1. Genetics. An interaction between nonallelic genes, especially an interaction in which one gene suppresses the expression of another. American Heritage Dictionary of the English Language (1992) Although officially given a name some ninety years ago, the phenomenon of gene interaction was discovered a few years earlier. In their study of the genetics of chicken combs, Bateson and Punnett (in Bateson et al. 1905) noted that one type of comb, the single, was rarely produced in their crosses and that its presence was difficult to reconcile with a simple Mendelian genetic system. Making good use of Punnett's recently derived method of creating a matrix, a “Punnett square,” to describe all possible combinations of gametes, they concluded that comb inheritance could be described by Mendelian segregation of two factors, with the single phenotype appearing only in the rare double-recessive homozygotes. Punnett later found similar results in his parallel work on sweet peas (reviewed in Punnett 1920). Weinberg (1910) also noted the possibility of such interactions, calling them komplizierter polyhybridismus (complicated polyhybridisms). It is perhaps not surprising that Weinberg's term did not catch on, and h