Biomass yield improvement in switchgrass through genomic prediction of flowering time

Authors: TILHOU, NEAL - University Of Wisconsin; CASLER, MICHAEL - Retired ARS Employee

Submitted to: Global Change Biology Bioenergy
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 6/10/2022
Publication Date: 6/15/2022
Citation: Tilhou, N., Casler, M.D. 2022. Biomass yield improvement in switchgrass through genomic prediction of flowering time. Global Change Biology Bioenergy. 14(9):1023-1034. https://doi.org/10.1111/gcbb.12983.
DOI: https://doi.org/10.1111/gcbb.12983

Interpretive Summary: Switchgrass is a potential candidate for conversion of biomass to bioenergy in sustainable and perennial production systems. Development of new switchgrass varieties is focused on identification of late-flowering plants that continue to accumulate biomass until cold temperatures of autumn shut down the growth processes. In addition, genomic prediction, using whole-genome DNA sequencing, is also being used to increase the efficiency of selection and to accelerate the development of new and superior varieties. This research reports the strong association between late flowering time and high biomass accumulation and demonstrates that these individuals can be efficiently identified using DNA sequencing within several highly divergent source populations of switchgrass. These results will be of significant value to switchgrass breeders specifically, but also as a concept that can be applied to other breeding programs of other energy grass species.

Technical Abstract: The seasonal timing of transition from vegetative to reproductive growth has a major impact on biomass accumulation in switchgrass. Late flowering switchgrass cultivars produce greater biomass, a critical trait for sustainable bioenergy production. Genomic prediction (GP) may allow rapid selection of late flowering individuals with reduced time and expense for field evaluations. To evaluate GP, two flowering time traits (heading date and anthesis date) were collected on 1,532 individuals from four breeding groups: Midwest, Gulf, Atlantic, and Hybrid. These individuals were sequenced using genotype-by-sequencing (530,792 SNPs). Predictive ability of single-trait and multi-trait models were evaluated by cross-validation, by prediction of a progeny trial (n=122), and through prediction of yield performance in a parallel experiment (n=52). Predictive ability was not improved by sharing information among breeding groups. Overall, multi-trait models provided an advantage during cross-validation, but a smaller advantage during progeny prediction. Within populations, GP resulted in lower per-cycle progress than previously reported field evaluations (3.1 vs 5 d-1 cycle-1). However, GP cycles are potentially much faster than field evaluations. When directly predicting biomass yield, the Hybrid training population had a predictive ability of 0.54-0.63. This reinforces the strong linkage between biomass yields in swards and flowering time. These results highlight the value of GP for rapid yield gains in switchgrass, particularly in a breeding program designed to share information between biomass yield trials and low-cost flowering time evaluations.