Allele frequency shifts as an indication of marker-trait associations
D. Z. Skinner and D. E. Obert
USDA-ARS and Agronomy Dept. Kansas State University, Manhattan, KS 66506
There are several reasons why co-segregation studies may fail to reveal marker-trait associations in tetraploid alfalfa (Medicago sativa L). Several methods for associating markers with traits have been described, each with the stated starting point of a cross of two inbred lines. The development of inbred lines in alfalfa is impossible due to inbreeding depression resulting in virtual sterility after only one or two generations of selfing. The autotetraploid nature of alfalfa vastly complicates the evaluation of cosegregation, due in large part to the diploid nature of the gamete. Reassortment of pairs of chromosomes into the gametes can, under many circumstances, yield apparent recombination between the marker and the gene conditioning expression of the trait (progeny plants expressing the marker or the trait, but not both, from parents expressing both the trait and marker), when in fact, no physical recombination has occurred. Genes occurring in triplex or quadriplex yield no gametes without at least one copy of the gene, hence, only simplex or duplex plants yield segregating progeny. However, gene dosage effects may preclude scoring simplex or duplex plants as "positive" for the trait. Finally, many studies of segregation in alfalfa have shown marked segregation distortion, presumably due to lethal gene combinations.
Instead of relying on segregating populations, it may be possible to use shifts in allele frequencies between populations selected for a trait, and the base, unselected population. Typically, alfalfa cultivars are developed from an initial population largely lacking a particular trait; rare plants expressing the trait are selected and polycrossed, selection is then carried out on that population and the selected plants again are intercrossed. This recurrent selection may be carried out for several generations until the level of expression (both in frequency of plants expressing the trait and the level at which they express the trait) is acceptable. The frequencies of alleles not under selection should be virtually identical in both the base population and the selected population. However, the frequency of alleles that affect expression of the trait must occur with significantly greater frequency in the selected population. Generating large numbers of markers from the selected and unselected populations will detect markers physically associated with genes determining the trait. Unfortunately, the power of statistical tests to discriminate frequency differences diminishes with increasing numbers of markers tested. Resampling analysis techniques can be applied to adjust the significance levels used, and marker-trait associations can thus be defined and tested.