At the grass roots: non-destructive root phenotyping using X-ray computed tomography

Abstract

Plant roots are crucial for accessing water and nutrients and directly impact crop yield. However, because they grow underground in the soil, their phenotyping is very difficult. To allow root visualization, artificial growth systems can be used, in which the plants are grown in a transparent substrate like agar, on filter paper, or in a hydroponic system (Li et al., 2022). The downside of these systems is that they only allow phenotyping at early growth stages, and traits observed in artificial systems often do not align with those in soil due to the complex nature of soil, which includes factors like soil compaction, uneven distribution of water and nutrients, and microbiota (Watt et al., 2013). Therefore, analyzing root traits under field conditions is crucial for identifying stable quantitative trait loci (QTLs) for agricultural applications.

Soil coring is commonly used for root phenotyping in the field. For that, a cylinder is inserted into the ground to extract a monolith of soil and the roots within it (Figure 1a) (Böhm, 1979). However, this method requires washing the soil off the roots and is therefore labor-intensive. This is particularly challenging in clay soils, such as those in rice paddy fields, where washing the roots is difficult. Additionally, when rice roots are removed from the soil, their structure collapses, making it impossible to determine their 3D structure.

Shota Teramoto and Yusaku Uga, the authors of the highlighted publication, developed an X-ray computed tomography (CT)-based method to analyze root structure from monolith samples without disturbing the soil core (Teramoto & Uga, 2024). Uga's group at the National Agriculture & Food Research Organization in Ibaraki, Japan, focuses on designing and developing climate-resilient crops by optimizing root system architecture to enhance adaptation to environmental stresses. One of their ongoing research projects involves developing advanced root phenotyping techniques. Within this project, they developed a method to visualize rice root systems grown in pots using X-ray CT. Teramoto, a Senior Researcher in the group, believed that applying this technology to rice root systems collected from paddy fields could solve the challenges of the phenotyping method.

For phenotyping, monolith samples were isolated from rice paddy fields, scanned using X-ray CT and analyzed using a workflow named RSApaddy3D (Figure 1b,c). The authors developed a special 2D kernel filter to isolate root-shaped fragments from the CT scans, optimized to detect small, dot-like fragments within its ring. Roots in soil appear as 3D tubular fragments creating a cross-section with a circular segment of the root in at least one of the x-, y-, or z-axis planar slices. This segment can be detected with the 2D filter. If the ring diameter is larger than the root diameter, using this filter along all three axes and combining the results isolates only the tubular fragments, that is, the roots. From this, root length and diameter can be extracted.

The authors compared measurements from the same samples using WinRHIZO, a software for analyzing root diameter and length from roots isolated from soil. They divided the monolith block into 5 cm depth segments and washed the roots in each segment. Samples from two different rice cultivars were measured. Crown root length and diameter, as well as total crown root length, showed a correlation between RSApaddy3D and WinRHIZO, except in the uppermost layer. This discrepancy occurred because RSApaddy3D could not segment roots that were touching each other with the ring kernel filter, which is common in the uppermost layer due to high root density. Additionally, the authors developed a vectorization-based filtering workflow to separate roots of different plants within one soil monolith. This allowed for the calculation of average root diameter, total root length, and average root growth angle for individual plants.

To test the application of their method, the authors conducted a genome-wide association study (GWAS) using 133 Japanese rice cultivars. They collected the root samples in soil monoliths and isolated images of the root system architecture using RSApaddy3D, converted them to vectors, and calculated the average root diameter, total root length, and growth angle. They detected a peak for root diameter on chromosome 4, indicating a single nucleotide polymorphism associated with the trait, in the same region as a previously reported quantitative trait locus (QTL) for crown root number, qNCR1 (Teramoto et al., 2022). A QTL for total root length was identified on chromosome 2 (qNCR2), and a significant peak for root growth angle was found on chromosome 6 (qRGA1).

To confirm the significance of these QTLs, the authors divided the 133 cultivars into different haplotypes based on their genotypes and compared their root traits. For qNCR1, they defined three haplotypes, with lines carrying haplotype 3 having 10% thicker roots. Haplotype 3 was present in the rice cultivar Koshihikari and its related cultivars in Japan but was rare worldwide. Therefore, the authors suggest that qNCR1 could be useful for increasing root diameter in cultivars globally, which could improve the penetration of compact soils (Materechera et al., 1992). For qNCR2, the cultivars were divided into two haplotypes, with haplotype 2 having 10% longer roots. Regarding qRGA1, the cultivars were also divided into two haplotypes, with haplotype 2 exhibiting a significantly steeper root growth angle by 10°. Cultivars with haplotype 2, developed in Hokkaido, a northern region of Japan, may be better adapted to the local climate, characterized by lower temperatures and shorter day lengths. Root systems with a steeper growth angle can better access water and nutrients that are present in deep soil layers, providing an advantage under specific soil conditions (Uga et al., 2013).

Compared to other methods, the scanning technique is efficient and not time-consuming. Analyzing approximately 450 monoliths required five person-days for sampling and five person-days for X-ray CT scanning. Each soil monolith takes 2.5 mins to scan, followed by 5–10 mins of fully automated image processing. The soil monoliths can be stored at 10°C for over a year when placed in plastic bags to prevent drying. Some root traits, such as lateral root length and diameter, cannot yet be extracted by the RSApaddy3D workflow. However, because the image data can be stored, it should be possible to re-analyze the data once new image analysis methods are developed.

By utilizing X-ray CT scanning and RSApaddy3D, researchers can now directly identify QTLs for root system traits in rice cultivated in paddies, which will contribute to rice breeding for improved root system architecture.