Introduction: The color variations of ornamental flowers are often generated by ion-beam and gamma irradiation mutagenesis. However, mutation rates differ significantly even among cultivars of the same species, resulting in high cost and intensive labor for flower color breeding.
Objectives: We aimed to establish a metabolome-based strategy to identify biomarkers and select promising parental lines with high mutation rates using Chrysanthemum as the case study.
Methods: The mutation rates associated with flower color were measured in 10 chrysanthemum cultivars with pink, yellow, or white flowers after soft X-ray irradiation at the floret-formation stage. The metabolic profiles of the petals of these cultivars were clarified by widely targeted metabolomics and targeted carotenoid analysis using liquid chromatography-tandem quadrupole mass spectrometry. Metabolome and carotenoid data were subjected to an un-supervised principal component analysis (PCA) and a supervised logistic regression with least absolute shrinkage and selection operator (LASSO).
Results: The PCA of the metabolic profile data separated chrysanthemum cultivars according to flower color rather than mutation rates. By contrast, logistic regression with LASSO generated a discrimination model to separate cultivars into two groups with high or low mutation rates, and selected 11 metabolites associated with mutation rates that can be biomarkers candidates for selecting parental lines for mutagenesis.
Conclusion: This metabolome-based strategy to identify metabolite markers for mutation rates associated with flower color might be applied to other ornamental flowers to accelerate mutation breeding for generating new cultivars with a wider range of flower colors.
Keywords: Flower color; Machine learning; Mutation breeding; Mutation rate prediction; Targeted carotenoid analysis; Widely targeted metabolomics.