Comparison of Selection Based on Yield of Topcross Progenies and of S² Progenies in Maize (^{Zea mays} L.)¹

Abstract

Three methods of breeding for higher grain yield in maize (Zea mays L.) were compared: (a) recurrent selection for combining ability with the parental population (Parental Tester method), (b) recurrent selection for combining ability with an unrelated inbred line (Inbred Tester method), and (c) recurrent selection for yield of S2 progenies per se (S2 Progeny method).After three cycles of selection for grain yield had been completed with each of the three methods, the nine resulting selected populations were evaluated for grain yield (a) as random‐mated populations (synthetics), (b) as selfed populations (bulked S1 lines), and (c) in crosses with 11 unrelated testers. Significant differences among the breeding methods were noted for the second and third cycles of selection, but not for the first cycle. The Parental Tester method produced the highest yielding random‐mated (syn‐3) population; whereas the S2 Progeny method produced the highest yielding selfed population. Average combining ability with unrelated testers for all three methods was increased significantly (5.2%) by the second and third cycles of selection, but there was no significant difference among methods. Although the Inbred Tester method resulted in significant improvement in yield of selfed populations, it was inferior to the other two methods for population improvement in the over‐all evaluation.Selfing one generation resulted in over 70% reduction in yield of the first cycle populations, which indicates that dominance is of considerable importance in the base population. Inbreeding depression was progressively less drastic in the second and third cycle populations, particularly with the S2 Progeny method.

Inbred‐Progeny Selection Is Predicted to Be Inferior to Half‐Sib Selection for Three Maize Populations

“…Beyond the numerical predictions of selection response we have presented, a better understanding of the underlying genetic principles in maize populations that limit the relationship between inbred and outbred performance is needed to determine if the relationship can be manipulated. Previous authors who proposed that inbred‐progeny recurrent selection would effectively improve outbred population performance generally assumed a single locus genetic model with partial or complete dominance (Eberhart, 1970; Comstock, 1964; Horner et al, 1969; Cockerham and Matzinger, 1985). However, overdominance, epistasis, and or linkage disequilibrium could all change these predictions profoundly, as could other possible genetic phenomena.…”

“…I nbred‐progeny recurrent selection was shown to be a superior method to mass selection and half‐sib recurrent selection based on theoretical arguments by numerous authors over an approximately 20‐yr period (Comstock, 1964;Horner et al, 1969; Eberhart, 1970;Choo and Kannenberg, 1979;Wricke, 1976;Wright, 1980;Cockerham and Matzinger, 1985). In contrast to theoretical predictions, inbred‐progeny selection has not been demonstrated to be universally superior based on empirical work (Lamkey, 1992; Coors, 1999) nor has it been widely adopted in industrial maize ( Zea mays L.) breeding programs, where the vast majority of performance evaluations have been based on testcross performance for decades (Hallauer et al, 1988).…”

“…The result was a wave of theoretical studies espousing the virtues of recurrent selection methods to improve testcross performance. Most of these studies focused on the comparison among inbred‐progeny selection, mass selection, and half‐sib methods on the basis of population per se performance as opposed to testcross performance (Comstock, 1964; Horner et al, 1969; Eberhart, 1970, 1972; Wricke, 1976; Choo and Kannenberg, 1979; Wright, 1980; Cockerham and Matzinger, 1985). A tacit assumption was that improvement of population per se performance was equivalent to improvement of testcross performance.…”

“…Early forms of predicted selection response for inbred‐progeny selection such as those given by Comstock (1964), Sprague (1966), and Empig et al (1981) were technically correct, but the theory did not exist to estimate the correct covariance components for the numerator of the predicted selection response. In absence of empirical estimates, several authors relied on single‐locus, two‐allele models with no overdominance and environmental variance generated to match a desired heritability (Comstock, 1964; Horner et al, 1969; Wricke, 1976; Choo and Kannenberg, 1979). Under these simple conditions, inbred‐progeny selection always had the highest expected response to selection (for population per se performance).…”

Testcross Response to Four Cycles of Half‐sib and S2 Recurrent Selection in the BS13 Maize (Zea mays L.) Population

“…M any theoretical comparisons among selection methods have shown that inbred‐progeny recurrent selection methods are expected to be much more effective than mass selection or half‐sib selection with either related or unrelated testers (Comstock, 1964; Horner et al, 1969; Eberhart, 1970; Choo and Kannenberg, 1979; Wricke, 1976; Wright, 1980; Cockerham and Matzinger, 1985). Such theoretical comparisons were generally based on single locus models that did not include overdominance (Comstock, 1964; Horner et al, 1969; Eberhart, 1970; Choo and Kannenberg, 1979; Wricke, 1976). However, addition of epistasis in two cases did not change the general conclusion that inbred‐progeny selection was predicted to be a superior method of population improvement (Wright, 1980; Cockerham and Matzinger, 1985).…”