Research Could Save Years of Breeding New Miscanthus Hybrids

As climate change becomes increasingly difficult to ignore, scientists are working to diversify and improve alternatives to fossil fuel-based energy. Renewable bioenergy crops, such as perennial Miscanthus grass, hold promise for cellulosic ethanol production and other uses, but current hybrids are limited by environmental conditions and susceptibility to pests and diseases.

Breeders have been working to develop new Miscanthus hybrids for years, but the sterility of the clonal culture, complex genome, and long maturity make conventional breeding difficult. In a new study, researchers at the University of Illinois exploit the crop’s vast genomic potential in an effort to speed up the breeding process and maximize its most desirable traits.

“The method we are using, genomic selection, can cut the time it takes to breed a new hybrid by at least half,” says Marcus Olatoye, lead author of the study and a postdoctoral researcher at the Illinois Department of Crop Sciences. “That is the general objective.”

In conventional breeding, a typical approach is for researchers to grow individuals from a diverse set of populations and select those with the best traits for mating. But, for Miscanthus, those traits don’t appear until the plants are 2-3 years old. Even after the plants of this first generation mate, it takes another 2-3 years for the offspring to reveal whether the desired traits were faithfully transmitted.

In genomic selection, scientists take genetic samples from seeds or seedlings in a target population. This is the group of plants that would normally have to grow to maturity before making experimental crosses. Meanwhile, the researchers compile genetic and phenotypic data from related populations, known as reference or training sets, into a statistical model. Genetic cross-referencing data from the target population with model data enables researchers to predict the phenotypic outcome of hypothetical crossings within the target population.

This allows breeders to get down to business, chasing only the most promising crossings with more field trials.

“Ideally, this process could allow breeders to make selections based on predicted phenotypic values ​​before plants are planted,” says Alex Lipka, associate professor of biometrics in the Department of Crop Sciences and co-author of the study. “Specifically, we want to make selections to optimize winter hardiness, biomass, disease tolerance, and flowering time in Miscanthus, all of which limit crop yield in various regions of North America.”

Although not a simple process at best, genomic selection in Miscanthus is much more challenging than in other crops. The hybrid of interest, Miscanthus × giganteus, is the product of two separate species, Miscanthus sinensis and Miscanthus sacchariflorus, each of which has different numbers of chromosomes and contains a great deal of variation within and between natural populations.

“As far as we know, no one has tried to train genomic selection models of two separate species before. We decided to go completely insane here,” says Lipka. “Unfortunately, we found that the two parent species do not do a good job of predicting biofuel traits in Miscanthus × giganteus.”

The problem was twofold. First, the statistical model simply revealed too much genetic variation between parental subpopulations to capture the impact of genes that control biofuel traits. This meant that the parental populations chosen for the reference set were too diverse to reliably predict traits in the Miscanthus × giganteus hybrid. And second, the genes that control a particular trait, such as those related to biofuel potential, appeared to be different in the two parent species.

In other words, the genomes that contribute to Miscanthus × giganteus are very complex, which explains why the statistical approach had difficulty predicting traits in the offspring of the two parents.

Still, the research team kept trying. In a simulation study, Olatoye created 50 families of Miscanthus × giganteus, each derived from randomly selected parents of both species. It selectively marked each parent’s genetic contributions above and below, and these contributions formed the genetic basis for the simulated phenotypes. The intention of the study was to provide a better insight into which individuals and populations might be most valuable for crossings in real life.

“The results suggest that the best strategy for using diversity in parents is to fit the genomic selection models within each parent species separately, and then add the predicted Miscanthus × giganteus values ​​of the two models separately,” says Olatoye.

Although the researchers have more work to do, the simulation study showed that genomic selection can work for Miscanthus × giganteus. The next step is to further refine which populations are used to train the statistical model and evaluate crossings in the field.

The article, “Training for Population Optimization for Genomic Selection in Miscanthus”, is published in G3: Genes, Genomes, Genetics.