Genetic evaluation and genomic approaches for drought tolerance in pearl millet (Pennisetum glaucum (L.) Br.) and chilling tolerance in grain sorghum (Sorghum bicolor (L.) Moench)

Abstract

Abiotic stresses such as drought and early-season chilling significantly limit cereal productivity in U.S. and semi-arid production systems. This dissertation employed integrated phenotypic and genomic approaches to improve abiotic stress tolerance in pearl millet and grain sorghum. In pearl millet, a diverse panel of 29 seed (B) and 27 restorer (R) parental lines was evaluated under irrigated and rainfed conditions across two years using a split-plot randomized complete block design. Significant variation among genotypes, moisture regimes, and years indicated substantial genetic diversity for drought-responsive traits. Multivariate analyses, including principal component analysis, genotype plus genotype x environment (GGE) biplot, rank summation index, and multi-trait stability index, identified superior drought-adapted parents based on grain yield per plant, number of seeds per plant, and 1000-grain weight. Seven seed parents and six restorers were selected as promising sources for developing drought-tolerant hybrids. Early planting in the U.S. Midwest can improve sorghum productivity by using residual soil moisture and avoiding late-season heat and drought stress, but exposure to sub-optimal temperatures can reduce seedling establishment and yield. Developing hybrids with early-stage chilling tolerance and stable grain yield is therefore critical. This dissertation evaluated 215 grain sorghum hybrids derived from a partial factorial mating design involving elite parents (41 female, 16 male) from four breeding programs: Kansas State University (KSU), Texas A&M (TAM), USDA-ARS (ARS), and S & W Seed Company (SW). Hybrids were phenotyped under growth chamber (control vs. chilling stress) and field conditions across two locations using early and regular planting dates. Seedling traits: emergence percentage (EP), emergence speed index (EI), and seedling biomass (SB) were used to assess early-stage chilling tolerance in growth chamber and field conditions, while days to flowering (DF), plant height (PH), and grain yield (GY) were measured in the field. Mixed model analysis revealed significant effects of planting window and hybrid, with early planting generally reducing seedling performance and GY. Hybrid performance rankings varied across planting windows and environments. Variance partitioning showed that both additive (GCA) and non-additive (SCA) genetic effects contributed to seedling traits, while GY particularly under early more strongly influenced by SCA and interaction effects. A two-step selection pipeline combining multi-trait stability index and GGE biplot analysis identified chilling-tolerant, high-yielding, and stable hybrids. Consistently selected hybrids included ARS82A/KSU69R and KSU52A/KSU64R, with additional promising hybrids such as KSU50A/KSU61R, S-W11A/KSU72R, and TAM33A/KSU62R. The prevalence of cross-program hybrids among top selections highlights the value of germplasm sharing for improving chilling tolerance and yield stability. To further improve selection efficiency, genomic prediction models were developed to predict hybrid performance for seedling chilling tolerance and GY under early planting conditions. Genomic prediction models achieved moderate to high prediction accuracy for both seedling and agronomic traits, with EI showing the highest predictability among seedling traits. Population structure analysis indicated moderate differentiation between female and male parental lines, consistent with the heterotic structure typical of hybrid sorghum breeding programs. Growth chamber seedling traits were positively associated with field seedling traits, indicating the possibility of using growth chamber data to predict the field seedling traits. Incorporating growth chamber phenotypes as additional environments improved prediction accuracy for EP and SB, while including as covariate did not improve prediction accuracy. Including field EP as a covariate substantially improved grain yield prediction accuracy, highlighting the importance of field evaluated seedling traits in improving genomic prediction of GY for early planting.