Non-additive gene effects in populations under different methods of selection.

The genetic parameters of two quantitative traits, 13-day larval weight and pupal weight, in Tribolium populations developed by reciprocal recurrent selection (RRS) and by within-line purebred selection (WLS) were compared each with the other and also with the parameters of the unselected base populations using the genetic model of Carbonell, Nyquist and Bell. The variability for two and three-way crosses of inbred lines derived from "companion" populations (two strains, breeds, or varieties used for a terminal cross or hybrid) was analyzed into genetic effects: autosomal additivity ((*) g), autosomal heterosis ((*) s), sex-linked additivity (L), sex-linked heterosis (LL), general maternal (m), specific maternal or reciprocal (r), additive by additive epistasis (aa), and deviations from the model due, among other causes, to higher order epistasis (dev). One series of crosses involved companion populations with diverse origins. For contrast, a second series of crosses involved companion populations originating from a common heterogenous base population. For the heterotic trait larval weight, (*) g and (*) s effects were equally important and accounted for over 50% of the total variation. The aa epistasis contributed another 20% and was followed in importance by higher order epistasis and general maternal effects. For the more highly heritable trait, pupal weight, (*) g effects were most important with (*) s, aa, and m effects having smaller but significant influences. Sex-linked and reciprocal effects were statistically significant for many crosses, but they were relatively unimportant overall. In general, the unselected base populations showed higher (*) g variation than either RRS or WLS populations with the reverse true for (*) s effects. In agreement with theoretical expectations, RRS was more effective than WLS in exploiting (*) s effects. The aa epistatic effects for larval weight were of major importance in the unselected populations, but RRS and WLS did not differ significantly for exploiting superior aa gene combinations. Companion populations with diverse origins revealed significantly larger variation due to (*) g and (*) s effects in crosses than did populations initiated from a common heterogeneous base.