Genomic selection and quantitative trait loci mapping for Fusarium head blight resistance in wheat (Triticum aestivum L.)

Fusarium head blight;genomic selection;quantitative trait loci;disease resistance

Pais De Arruda, Marcio

关键词: Fusarium head blight;    genomic selection;    quantitative trait loci;    disease resistance;   

Others  :  https://www.ideals.illinois.edu/bitstream/handle/2142/78412/PAISDEARRUDA-DISSERTATION-2015.pdf?sequence=1&isAllowed=y

美国|英语

来源: The Illinois Digital Environment for Access to Learning and Scholarship

【 摘 要 】

Fusarium head blight (FHB) is a destructive disease of wheat (Triticum aestivum L.) occurring in most growing areas. The disease is primarily caused by Fusarium graminearum Schwabe [telemorph: Gibberella zeae Schw. (Petch)], in North America, and the majority of current wheat cultivars are susceptible to it. Significant economic losses are associated with FHB since it results in yield reduction, poor kernel quality and grain contamination by mycotoxins, such as deoxynivalenol. Resistance to FHB has been identified in the wheat gene pool, but breeding for it remains a challenge for several reasons, including the complex genetic control of resistance and poor adaptability of the traditional sources. In this context, molecular markers could contribute to the identification of genomic regions associated with FHB resistance. In addition, markers could be used to calculate breeding values for wheat lines, for traits related to resistance disease. In the first study of this dissertation, genomic selection (GS) models were compared for predicting traits associated with resistance to FHB resistance, using 273 breeding lines in use at the University of Illinois’ soft red winter wheat breeding program. Genotyping-by-sequencing (GBS) was used to identify 5,054 single nucleotide polymorphisms (SNPs) which were then treated as predictor variables in GS analysis. Different parameters affecting the prediction accuracy of the genomic estimated breeding values (GEBVs) were tested, including: i) five genotypic imputation methods (random forest imputation – RFI, expectation maximization imputation – EMI, k-nearest neighbor imputation – KNNI, singular value decomposition imputation – SVDIand the mean imputation – MNI); ii) three statistical models (ridge regression best linear unbiased predictor – RR-BLUP, least absolute shrinkage and operator selector – LASSO, and elastic net); iii) marker density (p = 500, 1500, 3000, and 4500 SNPs); and iv) training population size (nTP = 96, 144, 192, and 218). No discernable differences in prediction accuracy were observed among imputation methods. For five of six traits, RR-BLUP outperformed other statistical models (LASSO and elastic-net), and a significant reduction in prediction accuracy was observed when marker number decreased to 3000 or 1500 SNPs, depending on the trait. Lastly, prediction accuracies decreased significantly when the sample size of the training set was less than 192. The second study consisted in a genome-wide association study (GWAS) performed on the same panel used in the first study. A total of 19.992 SNPs were obtained with GBS and ten significant SNP-trait associations were detected for multiple parameters associated with FHB resistance on chromosomes 4A, 6A, 7A, 1D, 4D, 7D, and multiple SNPs were associated with Fhb-1 on chromosome 3B. Fhb-1 is a major effect QTL identified in China, and it is very popular among wheat breeders worldwide. The genomic region on chromosome 6A appears to be new, as no other study reported QTL for that region. In addition, combination of favorable alleles of these SNPs resulted in lower levels of disease. The third study compared marker-assisted selection (MAS) with GS using different sets of genotypic data, including the QTL identified in the second study and Fhb-1. GS greatly outperformed MAS, with cross-validated prediction accuracy varying from 0.24 to 0.74 and from 0.59 to 0.98 for MAS and GS, respectively. Treating QTL as fixed effects in GS models resulted in higher prediction accuracy when compared with a GS model with only random effects. For the same selection intensity, GS resulted in higher selection differentials than MAS for all traits. This study indicates that GS is a more appropriate strategy than MAS for FHB resistance. The last study of this dissertation was concerned with a linkage mapping study using a population of 233 recombinant inbred lines obtained from IL97-181 (resistant) and Clark (susceptible). Neither parent possesses the traditional Asian sources of resistance to FHB in their pedigree. A total of 2275 single nucleotide polymorphisms (SNPs) were detected using genotyping-by-sequencing (GBS) and a genetic map was built covering all 21 wheat chromosomes. Inclusive composite interval mapping (ICIM) analysis identified four genomic regions associated with multiple FHB parameters, across all environments. Four QTL were detected for FHB resistance under field conditions on chromosomes 1B, 2D, 6D, and 7B. Two QTL were associated with type I resistance (6D and 7B), and two were associated with type II resistance (1B and 2D). The percentage of the phenotypic variation explained by these QTL varied between 6.7 and 12.5%. For QTL on sub-genome B, intervals smaller than 2 cM were obtained. The results show that elite germplasm can contribute to FHB resistance.