A Morphometric Discriminant Analysis of Agouti Genotypes in C57BL/6 House Mice

Abstract

Leamy, L. and S. S. Sustarsic (Department of Biology, California State University, Long Beach, Long Beach, California 90840) 1978. A morphometric discriminant analysis of agouti genotypes in C57BL16 house mice. Syst. Zool. 27:49-60.-Multivariate discriminant and canonical analysis techniques were applied to suites of six (A) and eight (B) morphometric characters, respectively, in two separate inbred populations of C57BL/6 house mice. Each population consisted of six classes of male mice which differed only in the particular allele present at the agouti locus. The six genotypes (aa, ata, atat, Aa, Avya A, a) were previously found to exert significant pleiotropic effects on most of the characters, the dominant alleles in general causing a decrease from the standard aa genotype. Among the A characters, the width of the first upper molar was the most powerful genotype discriminator, whereas interorbital width was most important among the B characters. Contributions of the characters to the canonical variates generally reflected known correlation patterns as seen from the results of principal component analyses. Discrimination plots and phenograms of the six genotypes consistently showed the atat and AMa genotypes as distinctive. The percentage of correct classification in the discriminant analysis classification matrices was similar (55% and 60%) for both sets of characters. The implications of the results for topics of interest to systematists (pleiotropy, subspeciation, natural selection) are discussed. [C57BL/6 house mice; multivariate discriminant analysis; canonical analysis; genotypes; pleiotropy.] In recent years both classical and numerical taxonomists have made increasing use of discriminant and canonical analysis in the assessment of morphometric divergence among natural populations. These analyses have been helpful primarily in detecting shape and size differences between taxa especially at the species level, and in a number of instances, have led to new classifications. Of course in virtually all of these studies the natural assumption has been made that observed morphological differences resulted from parallel genetic change. Unfortunately, however, there have been surprisingly few genetically based multivariate studies designed to examine this assumption (Blackith and Reyment, 1971), due in part to early difficulties in adequately assessing genetic variability. With the fairly recent development and refinement of techniques which estimate frequencies of allozymes and/or isozymes these difficulties have been largely overcome, but whether such systems repre1 Division of Biological Sciences, University of Kansas, Lawrence, Kansas 66045. sent reliable indicators of relevant genetic variability remains controversial (Lewontin, 1974). Other studies have focused on morphometric differentiation among chromosomal races, and although these studies are few in number (Atchley, 1974; Atchley and Cheney, 1974; Davis and Baker, 1974), they have generally shown a positive correspondence between phenetic and cytogenetic divergence. Another potentially very promising approach for exploring the genetic-phenetic relationship involves the multivariate morphometric comparison of populations differing by single alleles at a given genetic locus. Such an approach obviously necessitates the use of major (Mendelian) genes for which individual effects are detectable, the assumption being that such genes exert pleiotropic effects on the quantitative characters typically of interest to the systematist. Naturally this can best be performed in laboratory (rather than wild) populations where standard genetic backgrounds may be created for the unambiguous interpretation of the effects of the genes.