Heterosis, or hybrid vigor, is one of the most important tools in plant breeding and has previously been demonstrated for plant freezing tolerance. Freezing tolerance is an important trait because it can limit the geographical distribution of plants and their agricultural yield. Plants from temperate climates increase in freezing tolerance during exposure to low, non-freezing temperatures in a process termed ‘cold acclimation’. Metabolite profiling has indicated a major reprogramming of plant metabolism in the cold, but it has remained unclear in previous studies which of these changes are related to freezing tolerance. In the present study, we have used metabolic profiling to discover combinations of metabolites that predict freezing tolerance and its heterosis in Arabidopsis thaliana. We identified compatible solutes and, in particular, the pathway leading to raffinose as crucial statistical predictors for freezing tolerance and its heterosis, while some TCA cycle intermediates contribute only to predicting the heterotic phenotype. This indicates coordinate links between heterosis and metabolic pathways, suggesting that a limited number of regulatory genes may determine the extent of heterosis in this complex trait. In addition, several unidentified metabolites strongly contributed to the prediction of both freezing tolerance and its heterosis and we present an exemplary analysis of one of these, identifying it as a hexose conjugate.

BackgroundHybrids represent a cornerstone in the success story of breeding programs. The fundamental principle underlying this success is the phenomenon of hybrid vigour, or heterosis. It describes an advantage of the offspring as compared to the two parental lines with respect to parameters such as growth and resistance against abiotic or biotic stress. Dominance, overdominance or epistasis based models are commonly used explanations.Conclusion/SignificanceThe heterosis level is clearly a function of the combination of the parents used for offspring production. This results in a major challenge for plant breeders, as usually several thousand combinations of parents have to be tested for identifying the best combinations. Thus, any approach to reliably predict heterosis levels based on properties of the parental lines would be highly beneficial for plant breeding.Methodology/Principal FindingsRecently, genetic data have been used to predict heterosis. Here we show that a combination of parental genetic and metabolic markers, identified via feature selection and minimum-description-length based regression methods, significantly improves the prediction of biomass heterosis in resulting offspring. These findings will help furthering our understanding of the molecular basis of heterosis, revealing, for instance, the presence of nonlinear genotype-phenotype relationships. In addition, we describe a possible approach for accelerated selection in plant breeding.

A recombinant inbred line (RIL) population, derived from two Arabidopsis thaliana accessions, and the corresponding testcrosses with these two original accessions were used for the development and validation of machine learning models to predict the biomass of hybrids. Genetic and metabolic information of the RILs served as predictors. Feature selection reduced the number of variables (genetic and metabolic markers) in the models by more than 80% without impairing the predictive power. Thus, potential biomarkers have been revealed. Metabolites were shown to bear information on inherited macroscopic phenotypes. This proof of concept could be interesting for breeders. The example population exhibits substantial mid-parent biomass heterosis. The results of feature selection could therefore be used to shed light on the origin of heterosis. In this respect, mainly dominance effects were detected.