Rapid acquisition of ionomic and morphological data from plant seeds through fast X-ray fluorescence microscopy and computer vision.

Abstract

Plant seeds are one of the most important food sources for humans. As a result, seed morphology and the concentrations of essential and toxic elements in seeds have important implications not only for seed yield and quality, but also for human health. To identify natural variation in the accumulation of various elements in seeds and in seed morphology, high-throughput phenotyping methods are needed. Here, we employed X-ray fluorescence microscopy (μ-XRF) as a method for rapid and high-throughput phenotyping of seed libraries and developed a computer vision-based algorithmic workflow to automatically the extraction of elemental and morphological data from single seeds. This workflow enables rapid segmentation of individual seeds from a genome-wide association study (GWAS) panel with 1163 A. thaliana accessions, and facilitates the extraction of elemental and morphological traits at the individual seed level from the μ-XRF image. A total of 7 and 10 loci, respectively associated with the morphology and elemental concentration of A. thaliana seeds, were identified. The high-throughput and nondestructive method for automated phenotyping of plant seed libraries developed in this study provides a tool for investigating natural genetic variation controlling the seed mineral accumulation and seed morphogenesis.

Code and data availability

The authors explicitly state that the u-XRF source code and algorithm (the computer vision workflow used for seed segmentation and trait extraction from μ-XRF images) are distributed under the MIT License and publicly available at their GitHub repository, making it a paper-specific, public, actionable code asset.