Quantification of spatial metal accumulation patterns in Noccaea caerulescens by X-ray fluorescence image processing for genetic studies

Abstract

Abstract Background Hyperaccumulation of trace elements is a rare trait among plants which is being investigated to advance our understanding of the regulation of metal accumulation and applications in phytotechnologies. Noccaea caerulescens (Brassicaceae) is an intensively studied hyperaccumulator model plant capable of attaining extremely high tissue concentrations of zinc, and nickel with substantial genetic variation at the population-level. X-ray Fluorescence microscopy (µXRF) is a sensitive high-resolution technique to obtain information of the spatial distribution of the plant metallome in hydrated samples We used laboratory-based µXRF to characterize a collection of 86 genetically diverse Noccaea caerulescens accessions from across Europe. We developed an image-processing method to segment different plant substructures in the µXRF images. We introduced the concentration quotient (CQ) to quantify spatial patterns of metal accumulation and linked that to genetic variation. Results Image processing resulted in automated segmentation of µXRF plant images into petiole, leaf margin, leaf interveinal and leaf vasculature substructures. The harmonic means of recall and precision (F1 score) were 0.79, 0.80, 0.67, and 0.68, respectively. Spatial metal accumulation as determined by CQ is highly heritable in Noccaea caerulescens for all substructures, with broad sense heritabilities (H 2 ) ranging from 76–92% correlates only weakly with other heritable traits. Insertion of noise into the image segmentation algorithm barely decreases heritability scores of CQ for the segmented substructures, illustrating the robustness of the trait and the quantification method. Very low heritability was found for CQ if randomly generated substructures were compared, validating the approach. Conclusions A strategy for segmenting µXRF images of Noccaea caerulescens is proposed and the concentration quotient is developed to provide a quantitative measure of metal accumulation pattern, which can be used to determine genetic variation for such pattern. The metric is robust to segmentation error and provides reliable H 2 estimates. This strategy provides an avenue for quantifying XRF data for analysis of the genetics of metal distribution patterns in plants and the subsequent discovery of new genes that regulate metal homeostasis and sequestration in plants.

Plant phenotyping relevance

µXRF画像から植物器官を自動分割し、金属蓄積パターンを定量する画像処理法と指標を開発・検証しており、植物フェノタイピング手法が研究の中心である。

Code and data availability

The authors' image-processing/heritability analysis code for the µXRF plant phenotyping is explicitly deposited in a public GitHub repository. The phenotype datasets themselves are only available on request, and GeoPIXE is a generic third-party tool, not a paper-specific asset.