Acquiring genotypic data of plants through DNA sequencing has become an inexpensive, high-throughput procedure: The larger current problem is obtaining high-quality phenotypic data for a large number of genotypes. The focus in phenotyping corn has been on the aboveground part, arguably due to ease of access and the nondestructive nature of the process. The root structure has had considerably less attention, and even less in a high-throughput sense. It is, however, imperative that the properties of the plant organ that is actually in contact with soil be investigated in relation to agronomic performance.
Corn roots are intricate structures with abundant detail hidden behind outer root branches. To acquire high-quality images from objects like these, proper lighting is essential. The objective was to obtain evenly diffused lighting, reducing the shading effect that outer root branches may have on inner branches. For this purpose we constructed a "soft box" imaging system (Figure 1).
The imaging system consists of a box with inner dimensions of 61 cm square by 122 cm tall, made from nonreflective white furniture panels with two "shelves." The top shelf contains dual diffusing cloths made from a low-cost white sheet material (not drawn in the figure). Two monochrome cameras (Unibrain Fire-i 701b) with a maximum resolution of 1280*960 pixels were mounted, one between the diffusing cloths in the top shelf and another in a side panel, to obtain top and lateral views. The bottom shelf served as a platform with a spike on which the root was pinned upside down after a hole was punched in the stalk. Underneath the spike was mounted a stepper motor that rotated the root to obtain four lateral images. The light source was a standard photography incandescent bulb of 250 watts. The soft box was fitted with a hinged door for easy frontal access.
Before the operator placed a root in the imaging box, a control program written in MatLab was used to acquire two background images for each root, one from above and one from the side. The operator placed the root on the spike, closed the door, and acquired one top and one lateral image. The machine then automatically rotated the root three times through 90 degrees to obtain three more lateral root images. The results of the imaging procedure are shown in Figure 2.
The top left image shows a root as observed from the bottom. Beside it on the right, roots are shown from different genotypes, depicting significant variation in architecture. The bottom left image shows the side view of the same root as the one above it (also shown in Figure 3). Who would have imagined such beauty in an underground plant organ?
The scientific part of the study is to link quantifiable properties of roots with their genetic code. To quantify the complexity of the roots, a mathematical "trick" was used: We expressed the root complexity in the fractal dimension, first defined by Benoit Mandelbrot in 1983. For two-dimensional images, this is a number between 1 and 2, but it can be a broken number like 1.83. By collecting these fractal dimensions for many roots, we generated a QTL map (shown at bottom right of Figure 2) that allows researchers to link quantifiable information from roots to the genetic composition of the corn plant. For instance, the ellipses in the map show which regions of the genome among 10 chromosomes are responsible for complexity. This information allows breeders to select those genotypes that will yield a desired root complexity as expressed in the fractal dimension.
The next step is to figure out if root complexity matters. Intuitively a complex root is desirable, since it can find resources under limiting conditions. However, we have not shown whether a good root gives more yield or whether it allows the plant to cope with drought and nitrogen deficiency. The method shown here gives researchers a tool to improve their breeding programs by partially automating tedious and laborious procedures.
Agricultural and Biological Engineering