Root traits are fundamental for the resilience of plants under stress. Image-based phenotyping can provide relevant datasets for the underlying root traits. However, root phenotyping is still hampered by methodological constrains, in particular extraction of root traits from images taken under semi-natural conditions. In this study, we thus propose a strategy for analyzing root images from rhizoboxes. Utilizing three Vicia faba genotypes and two soil moisture conditions, we applied software tools featuring distinctive types of root descriptors. We determined their accuracy of root length measurement, inference from surface-visible root axes towards total root length, inter-relation among root architectural descriptors and their relevance for plant transpiration. Our results show that different image analysis tools provide similar root length estimates in spite of specific segmentation approaches. Several root architectural descriptors are also inter-comparable. Using a structural equation model, we identified relevant phenotyping root traits for root size and branching driving plant transpiration. We conclude that rhizobox systems are a promising approach for root phenotyping. Future developments in image analysis should overcome needs for manual post-processing (e.g. gap closure) to automatize root architecture measurements improving the throughput and thus the range of rhizobox phenotyping applicability for breeding.
There are high expectations that plant breeding for improved root systems will substantially advance adaptation of crops to resource‐limited environments and climate change. Image‐based phenotyping technologies provide novel opportunities to characterize root systems and overcome traditional throughput limitation in sampling and analysis. This chapter reveals interfaces between root systems modelling and phenotyping and demonstrates how modelling can contribute to maximizing the usability of phenotyping data for breeding purposes. After discussing different viewpoints and classification approaches for root systems as multivariate plant organs, the model‐based analysis of inter‐trait relations to reveal essential measurement traits for phenotyping is demonstrated. A major challenge for application of root phenotyping data in practical breeding is restrictions due to experimental duration (phenology) and growth environment (artificial substrates). Root models allow upscaling from seedling data towards yield‐relevant phenological stages and estimate performance under field conditions. Hence, modelling improves inference from phenotyping platforms towards conditions in a breeding nursery. Defining relevant target root traits for selection requires an accurate characterization of the predominant environmental constraints and quantitative understanding of the root trait–environment interactions influencing stress resistance. In silico modelling experiments provide root ideotypes with distinct target traits to overcome the predominant environmental limitations. This is discussed for mobile (water and nitrate) and immobile (phosphorus) nutrients as well as mechanical constraints. Novel measurement methods to improve environment characterization (envirotyping) (i) are of high relevance for defining/selecting promising root traits during phenotyping and (ii) critically influence the capability of models to reliably predict the expected performance of improved cultivars. Deeper integration of phenotyping and modelling will strongly contribute to overcoming current platform specificity and enhance phenotyping data usability in the context of plant breeding. This requires common formats between phenotyping outputs and model parameters to optimize joint applications and guide future developments in novel software modules and scaling capabilities.
Plant phenotyping to date typically comprises morphological and physiological profiling in a high-throughput manner. A powerful method that allows for subcellular characterization of organelle stoichiometric/functional characteristics is still missing. Organelle abundance and crosstalk in cell dynamics and signaling plays an important role for understanding crop growth and stress adaptations. However, microscopy cannot be considered a high-throughput technology. The aim of the present study was to develop an approach that enables the estimation of organelle functional stoichiometry and to determine differential subcellular dynamics within and across cultivars in a high-throughput manner. A combination of subcellular non-aqueous fractionation and liquid chromatography mass spectrometry was applied to assign membrane-marker proteins to cell compartmental abundances and functions of Pisum sativum leaves. Based on specific subcellular affiliation, proteotypic marker peptides of the chloroplast, mitochondria and vacuole membranes were selected and synthesized as heavy isotope labeled standards. The rapid and unbiased Mass Western approach for accurate stoichiometry and targeted absolute protein quantification allowed for a proportional organelle abundances measure linked to their functional properties. A 3D Confocal Laser Scanning Microscopy approach was developed to evaluate the Mass Western. Two P. sativum cultivars of varying morphology and physiology were compared. The Mass Western assay enabled a cultivar specific discrimination of the chloroplast to mitochondria to vacuole relations.
Faba bean (Vicia faba L.) is an important source of protein but breeding for increased yield stability and stress tolerance is hampered by the scarcity of phenotyping information. Because comparisons of cultivars adapted to different agro-climatic zones improve our understanding of stress tolerance mechanisms, the root architecture and morphology of 16 pan-European faba bean cultivars were studied at maturity. Different machine learning (ML) approaches were tested in their usefulness to analyse trait variations between cultivars. A supervised, i.e. hypothesis-driven, ML approach revealed that cultivars from Portugal feature greater and coarser but less frequent lateral roots at the top of the taproot, potentially enhancing water uptake from deeper soil horizons. Unsupervised clustering revealed that trait differences between Northern and Southern cultivars are not predominant but that two cultivar groups, independently from major and minor types, differ largely in overall root system size. Methodological guidelines on how to use powerful machine learning methods such as random forest models for enhancing the phenotypical exploration of plants are given.
There is a need for flexible and affordable plant phenotyping solutions for basic research and plant breeding. We demonstrate our open source plant imaging and processing solution ('PhenoBox'/'PhenoPipe') and provide construction plans, source code and documentation to rebuild the system. Use of the PhenoBox is exemplified by studying infection of the model grass Brachypodium distachyon by the head smut fungus Ustilago bromivora, comparing phenotypic responses of maize to infection with a solopathogenic Ustilago maydis (corn smut) strain and effector deletion strains, and studying salt stress response in Nicotiana benthamiana. In U. bromivora-infected grass, phenotypic differences between infected and uninfected plants were detectable weeks before qualitative head smut symptoms. Based on this, we could predict the infection outcome for individual plants with high accuracy. Using a PhenoPipe module for calculation of multi-dimensional distances from phenotyping data, we observe a time after infection-dependent impact of U. maydis effector deletion strains on phenotypic response in maize. The PhenoBox/PhenoPipe system is able to detect established salt stress responses in N. benthamiana. We have developed an affordable, automated, open source imaging and data processing solution that can be adapted to various phenotyping applications in plant biology and beyond.
Spectral imaging makes use of different wavelength to infer on plant properties and processes. In the context of plant phenotyping, spectral imaging mostly uses multispectral sensors with defined broad- band wavelength in then VIS (400-700 nm), NIR (700-1100 nm) and SWIR (1100-2500 nm) regions. Hyperspectral imaging on the contrary captures the entire spectrum with up to several hundred narrow-band channels. It is expected that the comprehensive spectral signature obtained from hyperspectral imaging can provide deeper insights into plant properties of potential use for structural-functional phenotyping. On the other hand the resulting spatial-spectral datasets are substantially larger compared to multi-spectral images, targeting defined wavelength to obtain spectral indices (e.g. NDVI), and require adequate methods to extract information from the data cloud. Here we present the application of hyperspectral imaging to the root zone of plants grown in soil filled rhizoboxes. Essential steps in processing the hyperspectral datasets to obtain structural and functional information on the root system are demonstrated and implications for phenotyping application are discussed. Plants of Triticum durum are grown in soil (silty loam topsoil, 2 mm sieve-size) filled rhizoboxes (30 x 1000 x 1 cm) at optimum moisture (field capacity) for imaging via a transparent mineral glass side. Spectral images are taken via a spectral scanner (1000-1700 nm, 222 bands, spatial resolution 0.1 mm). Different pre-processing, dimensionality reduction and segmentation algorithms for separating root foreground and soil background pixels are discussed. Chemometric analysis of the segmented root images is exemplified for spectral distinction of root regions. Results demonstrate that pre-processing of spectral images is the most important step for classification of root vs. soil pixels. Thereby the heterogeneity of the root axes as well as the soil background (water content, surface morphology) can be significantly reduced. As an example, polynomial de-trending with subsequent scatter correction via standard normal variate increases the Bhattacharyya distance between pixel histograms of root vs. soil from 0.47 for raw data to 2.76 for pre-processed data, with maximum distinction at a wavelength of 1462 nm. Based on pre-processed images and identification of most distinctive wavelength, segmentation (e.g. via fuzzy clustering) provides accurate binary images of the root system that can be further analysed with different chemometic approaches. This is exemplified by identifying central vs. boarder regions on the root axes showing different spectral signature. It is hypothesized that these spectral feature represents the distinction between parts belonging to the central cylinder with water conducting xylem (lower reflectance at water sensitive bands) vs. the cortex region. First results demonstrate that hyperspectral imaging can provide novel insights into the root zone via distinctive spectral characteristics of different domains. Due to the heterogeneous biophysical and biochemical nature of the root zone, a key requirement for successful application of hyperspectral imaging to plant phenotyping is the use of efficient image processing tools in order to extract features of interest capturing root structure and functionality.