Abstract:
Pyramiding multiple desirable genes is an important method for the development of improved breeding materials and/or new cultivars. When the number of genes to be pyramided is many, or the genes are tightly linked in repulsion, it is practically impossible to recover the desirable recombinants in a single generation using a realistic population size, and repeated selection at several generations is required. The availability of markers tightly linked to the desirable genes makes it possible to conduct effective individual selection at early generations. This reduces the number of lines tested in the later generations and increases the desirable genotype frequency in the selected progeny. Computer simulation was used to develop such a marker-based pedigree selection strategy for the development of a barley line that contains 6 desired genes from 3 parental breeding lines (HS078 (H): 221222; PI366444 (P): 212222; Sloop Vic. (S): 122111; with 1 and 2 representing desirable and undesirable alleles, respectively), using the top cross H/P//S. The 6 genes targetted contribute to photoperiod sensitivity, Russian wheat aphid resistance, leaf rust resistance, boron tolerance, earliness per se, and cereal cyst nematode resistance. Under the assumption that perfect markers were available for all the 6 genes, a TC1 population of 300 plants was required to obtain 3 or more lines of the best genotype ?211222/122111', in which 3 loci were fixed for the desirable alleles, while the remaining 3 were kept as heterozygous. When single seed descent was used from the TC2 generation until complete homozygosity, the probability of obtaining lines of the desirable genotype (fixed for the desirable alleles at all 6 loci) was low due to the tight repulsion linkage between some of the genes. About 4000 individuals would be required to ensure with 99% probability the recovery of at least 1 line with the desirable genotype. The total number of lines that would need to be genotyped would be at least 5000. When the pedigree method was used in all test-cross generations, many schemes resulted in more lines of the fixed desirable genotype by genotyping fewer lines. The various options were compared using the genetic simulation software module QuLine, based on the QU-GENE simulation platform. The optimum scheme in terms of high success rate and relatively low genotyping costs consisted of the following steps: (1) in TC1 genotyping of 300 individuals allows for 3 or more individuals with the genotype ?211222/122111' to be identified; (2) in the TC2 individuals that are fixed for 3 loci and segregating for the remaining 3, loci can be selected from among 500 TC2 plants; (3) in the TC3, 50 or more individuals per TC3 line are genotyped for the 3 segregating loci, and individuals fixed for 5 loci and segregating for the 6th locus can be detected (genotyping is only needed for the segregating loci); (4) 25 individuals per TC4 line are genotyped for the single remaining segregating locus and several individuals of the desirable genotype (111111/111111) are finally selected. The desirable line is then obtained by collecting selfed seed from the selected TC4 plants. Using this scheme, on average, 320 desired TC5 lines were obtained by genotyping fewer than 2000 lines. When markers were tightly linked to the target genes but not diagnostic (perfect), not only was more genotyping required, but also appropriate phenotyping at the end of the marker selection process was necessary to confirm the presence of all the target genes. Under the assumption that recombination between marker and target gene was 5%, the best selection scheme identified, on average, 30 fixed desirable lines by genotyping 8000 lines and phenotyping 700 TC5 lines. If double haploid lines were produced from the F1 generation between H and P, and marker and phenotypic screening were conducted, followed by crossing of the individual with the target 2 loci in desired homozygous allelic status with parent S, the total amount of genotyping and phenotyping could be halved. This study showed that genetic simulation allows for numerous strategies to be compared using real data, and to develop an optimal crossing and selection strategy to combine desired alleles in the most effective and efficient way. This approach could likewise be used in other marker-assisted breeding programs.