Principles of Plant Genetics and Breeding

BREEDING SOYBEAN 523
easily selected. High heritability, in the narrow sense (Φ2 A /Φ2 P ), means that a high proportion of the total observed phenotypic variability
(Φ2 P ) is accounted for by what is called additive genetic variance (Φ2 A ). Additive genetic variance is the variance due to the
average effects of alleles (Bernardo 2002, p. 91). It measures the variation in the effects that are transmitted from one generation to
the next. It does not imply that the alleles act in a purely additive manner. Rather, segregating loci with dominance, partial dominance,
and overdominance gene actions, as well as additive gene action, can all contribute to Φ2 A (Bernardo 2002, p. 92).
Epistatic gene action can also contribute to Φ2 A , as well as to Φ2 D (dominance genetic variance) and Φ2 I (epistatic genetic variance)
(Bernardo 2002, p. 96). Φ2 D is the variance of dominance deviations and indicates dominance gene action, which describes the
genotypic interaction of alleles within a locus. Φ2 D is not considered useful for making progress from selection because intralocus
interactions are not passed on to progeny. Rather, meiosis determines that only one allele from each diploid intralocus interaction
will be passed on to progeny through gametes. Hence, an individual’s dominance genotypic interactions are not passed on to its
progeny and that is why Φ2 D is not considered useful for making progress from selection. When Φ2 D = 0, dominance gene action is
absent and the intralocus variance is comprised solely of Φ2 A , indicating purely additive gene action within the locus, but not necessarily
between loci (Bernardo 2002, p. 92). Some components of Φ2 I , such as Φ2 AA (additive by additive epistatic genetic variance),
are useful in selection (Hanson 1963, p. 133), because such interlocus interactions can be passed on to progeny. A
broad-sense estimate of heritability (Φ2 G /Φ2 P ) includes the variance components of all types of gene action (Φ2 G ), some of which
(dominance, etc.) are not passed on from parent to offspring. Exceptions to the above, where dominance gene action can be useful
in selection, are when selecting among asexually propagated clones or among single-cross hybrids (Bernardo 2002, p. 109).
If tolerance were controlled by one or a few major genes, with only minimal environmental influence, then heritability would
be expected to be high and selection for charcoal rot tolerance would be effective in the F2 generation among single plants or in
the F3 generation among F2:3 progenies. If tolerance was controlled by multiple genes, but the environment had only a minor
influence, then heritability might still be high and selection in early generations still effective. However, if heritability for tolerance
to charcoal rot was low, then selection might only be effective among advanced generation breeding lines grown in replicated
experiments.
Creating segregating populations of multiple generations (F2 and backcrosses to each parent) from two parents can serve
multiple purposes in inheritance studies. Warner (1952) proposed utilizing the above generations (P1 , P2 , F1 , F2 , BC1P1 , BC1P2 )
to estimate a narrow-sense heritability (Φ2 A /Φ2 P ), but suggested the need to test at least several hundred F2 , BC1P1 , and BC1P2 individuals
in order to reduce potential sampling error. Reinert and Eason (2000) provide an example of estimating a narrow-sense
heritability in a self-pollinated species using the above generations. Bernardo (2002, pp. 110, 146) listed three disadvantages of
estimating Φ2 A and Φ2 D using the above generations:
1 As single F 2 and BC 1 plants cannot be replicated, the lack of replication across environments causes these estimates of
genetic variance to be confounded with the variance for genotype by environment.
2 Any linkage disequilibrium in the non-random-mated F 2 and BC 1 populations will cause the relationship between genotypic
values at two loci to be confounded with Φ 2 A and Φ2 D .
3 Individual plant measurements of quantitative traits are prone to large, non-genetic effects.
Another option for utilizing the above generations could be to determine the number of genes and their modes of action
for simply inherited (one or two genes) traits. If all generations (P 1 , P 2 , F 1 , F 2 , BC 1 P 1 , BC 1 P 2 ) are assayed together for tolerance to
charcoal rot, segregation ratios of the F 2 , BC 1 P 1 , and BC 1 P 2 , along with the assay values of P 1 , P 2 , and F 1 individuals, can determine
if inheritance is simple and if dominance gene action is evident. The segregating F 2 generation can provide one estimate for
an inheritance model, while the two backcross generations can provide a second and confirming estimate. Velez et al. (1998) and
Singh and Westermann (2002) provide examples in dry bean (Phaseolus vulgaris L.) of utilizing the above generations in this way
to determine the qualitative inheritance of resistance to bean disorders. However, when inheritance is affected by many genes
and influenced greatly by environment, the above estimates may be inadequate.
An additional purpose for creating and utilizing the above generations could be to conduct an analysis of generation means.
Generation mean analysis provides information on the relative importance of additive and dominance effects in populations created
from two inbreds. It involves measuring the means of different generations (P 1 , P 2 , F 1 , F 2 , BC 1 P 1 , BC 1 P 2 , etc.) derived from
two inbreds and interpreting the means in terms of the different genetic effects (Bernardo 2002, p. 144). Because the actual means
of single loci are unobservable, generation means estimate the pooled genetic effects across loci. Generation mean analysis is
most useful when the two parents differ greatly in favorable alleles; that is when one parent has most, if not all, of the favorable
alleles and the other parent has few, if any, favorable alleles. The pooled estimates of effects are summed across all loci for which
P 1 and P 2 differ. Generation mean analysis has been commonly used to study disease resistance, where one parent has high resistance
and the other has high susceptibility (Bernardo 2002, p. 145). Useful examples of generation mean analysis have been provided
by Reinert and Eason (2000) in snap bean (P. vulgaris L.), Mansur et al. (1993) in soybean, and Campbell and White (1995)
and Campbell et al. (1997) in maize (Zea mays L.).
Generation mean analysis can also be used to estimate effects due to epistasis, environment, genotype × environment interactions,
and linkage (Mather & Jinks 1971, pp. 83–119). However, the experiments can become much more complex and are
unnecessary if the simpler additive-dominance model accounts for the variability present.