Notice: Wiley Online Library will be unavailable on Saturday 30th July 2016 from 08:00-11:00 BST / 03:00-06:00 EST / 15:00-18:00 SGT for essential maintenance. Apologies for the inconvenience.
Optimizing nodulation in legumes is a target for crop improvement, and the spatial control of nodulation is just beginning to be unravelled. However, there is currently no method for standard phenotyping of nodulation patterns. Here we present a method and software for the quantitative analysis of nodulation phenotypes.
Roots of nodulated peas (Pisum sativum), wild-type and two mutants, were photographed. Data from the photographs were extracted using custom image and data analysis software.
The software makes it possible to extract each nodule's position along primary and lateral roots, and to represent the nodulated root system in a standardized way independent of the way roots are arranged in the soil. A wide variety of nodulation and root variables are calculated, and average spatial nodulation patterns can be computed from multiple samples.
Standardized spatial analysis of nodulation patterns opens the way for comparative analyses among genotypes of a single legume species, as here in pea. This approach could also be used to compare nodulation patterns among crops, among plants grown under different environmental conditions, or among plants exposed to different pharmacological treatments. The proposed method should therefore prove useful for studies on nodule organogenesis and nodule physiology and for optimizing nodulation in crops.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
The importance of legumes in agriculture cannot be overstated (Shtark et al., 2010). Legumes benefit soils and ecosystems, play an important role in human and animal diets, and are therefore at the centre of sustainable agriculture (Vance, 2001; Tikhonovich & Provorov, 2011). A key aspect of legumes’ physiology is their enhanced nitrogen fixation through a symbiotic interaction with bacteria belonging to the family Rhizobiaceae. The establishment of this symbiosis starts with the production of flavonoids by the roots which act as chemo-attractants for the rhizobia, and ends with the formation of characteristic nitrogen-fixing nodules on plant roots (Ferguson et al., 2010). The number and spatial distribution of nodules throughout the root system are likely to affect root growth and root system architecture, as both the availability (Osmont et al., 2007; Desnos, 2008) and distribution (Linkohr et al., 2002; Walch-Liu et al., 2006) of nitrogen influence lateral root growth. Crop improvement programmes in recent years have therefore recognized the necessity to optimize the root system architecture of legumes (McPhee, 2005; Gonzalez-Rizzo et al., 2009; Schultz et al., 2010) and the efficiency of nutrient uptake via symbiosis with nitrogen-fixing bacteria (Shtark et al., 2010).
The location of nodules within single roots and across the root system is not random. At the level of a single root, Bhuvaneswari et al. (1980) identified a zone where lateral roots are most susceptible to rhizobial infection and where most of the nodules later form. This most susceptible zone has been described as a small restricted region just above the root tip where root hairs have begun differentiating but have not yet matured. At this location, root hair walls are plastic enough to allow deformation and penetration of the rhizobia within a hollow tube, known as an infection thread, that the bacteria use to progress across the root cortex. Usually, bacterial infection triggers the differentiation of nodules opposite xylem poles (Bond, 1948). At the level of an entire root system, not all roots bear nodules. Most often, at least in pea (Pisum sativum) plants, nodules are formed high on the root system in a cluster located close to the crown of the plant (Sagan & Gresshoff, 1996). Two explanations have been provided for this occurrence. The first explanation is that inoculants are usually applied onto the seeds or the substrate, and, as the rhizobia have a low ability to migrate in the soil (Wadisirisuk et al., 1989), they will infect only the roots near the crown of the seedlings (Cardoso et al., 2009). Autoregulation of nodulation offers the second explanation for the tight cluster of nodules close to the plant crown (Ferguson et al., 2010). In this hypothetical systemic regulatory mechanism, the first nodules to be formed on the root system would send an ascending signal to the shoot. In the leaves, the decryption of this signal by a CLAVATA-like leucine-rich kinase would result in the production of an inhibitory signal descending to the root, which would prevent new nodules from developing on the younger roots now developing deeper in the substrate (Ferguson et al., 2010). Another level of control affecting nodule distribution occurs at the genotype level, as seen in soybean (Glycine max), where different cultivars are known to exhibit different nodulation profiles; the nodule position on the root system appears to be affected only slightly by the growth conditions to which the plants are subjected (Burias et al., 1990).
The most basic assessment of nodulation is a count of the total number of nodules on a plant. Nodules are generally counted manually by excising them from the roots (e.g. Guinel & Sloetjes, 2000); this work is prone to human error and does not provide information about spatial distribution. Only a few studies document the distribution of nodules, and these have been directed towards the primary roots of young plants (Heron & Pueppke, 1984; Pueppke, 1986). Here, we present a method and software for recording and quantifying nodule distribution patterns, as well as basic secondary root architecture in the nodulated zone of the plant, in a standardized way using semi-automated image analysis tools. We demonstrate that the software can be used to quantify phenotypic differences in nodulation patterns among genetically different pea lines, and we propose that similar experiments could be performed to analyse differences among cultivars, various environmental conditions, and different legume species. Such quantitative phenotyping will make it possible to explore the link between nodulation patterning, root system architecture, and ultimately plant yield.
Materials and Methods
Plant growth conditions
Thirteen seeds of Pisum sativum L. cultivar Sparkle (wild-type) and of its mutants R50 (Pssym16) and E151 (Pssym15) were surface-sterilized and left to imbibe overnight as in Guinel & Sloetjes (2000). They were planted individually in cylindrical black Conetainers® (6.4 cm in diameter, 25 cm in height, and 656 ml in volume; Stuewe & Sons Inc. Tangent, OR, USA), which contained a mixture (1 : 1) of sterile vermiculite and Turface® (Plant Products Company Ltd, Brampton, ON, Canada) wetted the night before planting. The pots were kept in a growth-room under 8 h, 18°C : 16 h, 23°C, dark: light (light intensity of 120–150 μmol m−2 s−1). The seedlings were inoculated 3 d after planting with 5 ml of a 5% solution of Rhizobium leguminosarum bv. viciae 128C53K (a generous gift of Dr Stewart Smith, EMD Crop BioScience, Milwaukee, WN, USA), which had been grown to reach a stationary growth phase. From 10 d after planting until harvest, plants were watered daily with an alternating regimen of water and low-nutrient solution (Guinel & Sloetjes, 2000). Plants were harvested 28 d after inoculation, that is, 31 d after planting.
Preparing the root system for analysis
On harvest day, the plants were removed gently from the pots and their roots were washed delicately to remove soil debris without damaging the nodules or lateral roots. Shoots were excised just above the cotyledons. Root systems were disentangled so that the first-order lateral roots fell naturally into their three planes of branching caused by the triarch pattern of the root stele (Pepper et al., 2007). Second-order lateral roots, which do not bear nodules in pea cv Sparkle, were excised under water with a pair of tweezers, so that they did not overlap the first-order lateral roots during imaging. The root excision was performed as close to the first-order lateral root as possible to avoid any short stumps being later counted by the software as nodules.
The resulting root system was then centred on a rectangle of wetted dark felt (Fig. 1) placed on a tray, with the primary root running as straight as possible along the top of the tray. A dissecting pin was used to indicate the primary root tip such that it did not touch any roots. A 1.5 inch-wide piece of Styrofoam™ was labelled with the plant's identification number and placed on the felt as a size reference. A first photograph (Fig. 1) was used to keep a record of the whole plant. Lateral roots were then prepared for the imaging of individual planes. All roots branching from the same xylem pole within the primary root were set to one side of the primary root and laid out with every effort to prevent them from touching one another (Fig. 2a), as this would impede the ability of the software or a person to identify individual roots. Roots of the planes not being imaged were laid on the opposite side of the primary root and covered with a moist paper towel to prevent desiccation (Fig. 2a); the paper towel was placed so that it did not touch the primary root. After imaging of the plane, the roots of the plane in question were excised. This procedure was then repeated for the remaining two planes. Lateral roots below the root bearing the lowest nodule in the data set (e.g. those at the very bottom of the primary root) were not laid out individually, as our focus was on the nodulation zone (i.e. the region of the root system in which nodules occur). We reasoned that, unless the researcher was interested in the root architecture of the entire plant, these roots well below the nodulation zone did not need to be disentangled.
The tray holding the root system was placed in landscape orientation under a camera secured to a camera stand (Kaiser Fototechnik-5360 ‘reprokid’; BJ Photo Labs Ltd, Waterloo, ON, Canada) and the root plane to be imaged was gently patted with a paper towel to remove moisture. The two cotyledons were positioned close to the top right-hand corner of the felt rectangle with the long axis of the straightened primary root running parallel, and with the first-order lateral roots perpendicular, to the long side of the felt. Two reflector lamps were placed symmetrically on the outside of the stand. Each plane was imaged with a camera exposure time of 1/10, 1/8 and 1/6 s so that the image that showed lateral roots and nodules most clearly could be selected later and used for image analysis. The photograph was taken in such a manner that the lateral roots far below the nodulation zone were not considered within the data set. Once the image was recorded on the computer and before it was processed for analysis, the photographs were cropped so that the only background in the image was the black felt (Fig. 1).
In a preliminary experiment using eight plants of the wild-type line, grown in the same conditions as described above (in the section ‘Plant growth conditions’), manual measurements were taken to later evaluate the reliability of the software measurements. Once a plane had been imaged, the lengths of the primary root and the longest lateral root of each plane were measured and recorded in an Excel file. The lateral roots of that plane were numbered from top to bottom, and nodules were excised from each root to record nodule numbers. The nodules located on the primary root were also counted after all planes had been imaged.
Image analysis programs
To extract root architecture and nodule distribution data from the photographs of the three planes, custom image analysis programs were written in Matlab R2012 (The Mathworks Inc., Natick, MA, USA). The algorithms employ functions from Matlab's Image Processing Toolbox™ add-on (The Mathworks Inc.). A package including programs, user guides for sample preparation and software use, and video demonstrations is available for download at http://hdl.handle.net/10393/30321.
For each plane of each plant analysed, the JPEG image (Fig. 2a) is first converted to a hue, saturation, and value (HSV) image via the Matlab rgb2hsv function, and a binary mask of the root system is extracted through image thresholding (Fig. 2b). A morphological skeleton of the root system is then generated using Matlab's bwmorph function.
To delineate the primary root, the user is prompted to click the base and the tip of the primary root in each image. The primary root is automatically extracted by following the shortest path between those two points on the root system skeleton using Matlab's bwdistgeodesic function. The regions of the image above the primary root are disregarded in all later image analysis steps, so that only the roots and nodules on the plane in question are extracted. Individual lateral roots are first detected automatically by finding objects in the image that have a single contact point to the primary root. The roots are displayed to the user for verification (Fig. 2c).
Automatic detection of an individual lateral root does not work for roots that (1) have more than one apparent contact point to the primary root in the extracted root skeleton (occurring when roots are overlapping or touching each other, or are connected by glare or pale debris on the background) or (2) have no identified contact point to the primary root (occurring when upper parts of the root are too narrow or dark to be detected in the image thresholding step). Thus, an option is included for the user to add roots that were not automatically extracted. This can be done in a semi-automated manner by clicking the attachment point to the primary root and the tip of each root; the software traces out the coordinates by following the shortest path along the root skeleton between the two points using Matlab's bwdistgeodesic function. Sometimes when the tip of the root becomes narrow and transparent, that portion of the root is not extracted by the program (compare Fig. 2a–c for some examples); thus, an option is also included to add points to a root by clicking along its path. Next, the nodules are detected automatically using Matlab's bwmorph and regionprops functions, on the basis of specific morphological and colour characteristics (small pink rounded protrusions along a lateral root) (Fig. 2d). Once a nodule has been identified, the closest root coordinate is used to determine on which root the nodule is and the nodule position along that root. Lastly, software-detected nodules are displayed on the image for verification by the user. For each image, the user clicks three or more points to delineate the nodulated region; the software then zooms in on this region and iteratively plots all automatically recognized nodules on each individual root for the user to verify and edit if needed. More nodules can be added by right-clicking on them and incorrect nodules can be deleted by left-clicking on them. An illustration of the end-product of extracted data is displayed in Fig. 2(e) (to be compared with Fig. 2d).
Data analysis programs
To evaluate the spatial distribution of nodules, each nodule's position (x, y) is recorded with respect to the root system. We define the x position as the distance along the lateral root from the primary root to the nodule, and the y position as the distance along the primary root from its base to the insertion point of the lateral root on which the nodule occurs (Fig. 2f). Those distances are independent of the roots’ position in the soil, and make it possible to relate nodulation patterns to root system architecture. In the same manner, we also record the position and length of each lateral root along the y-axis. This makes it possible to represent the nodulated root system as a diagram where the primary root is oriented vertically (i.e. along the y-axis) and the extended lateral roots are displayed horizontally to the right of the primary root. Nodules are then placed on the diagram at their proper (x, y) position (Fig. 3a). We refer to the resulting virtual depiction of the nodulated root system as a root system equivalent (RSE). The RSE can be used to extract estimates of spatial coverage of nodules and roots within the nodulation zone by calculating the two-dimensional (2D) area of nodule and root convex hulls on the RSE plot for an individual plant (Fig. 3a). The RSEs from all plants of a data set can be overlaid to give an overall representation of nodule and root distribution for the data set of interest (Fig. 3b).
The data extracted are used to compute various nodule and root-related measurements which are displayed graphically and/or exported numerically to Excel files (Microsoft Corp., Redmond, WA, USA). Three Excel files are generated to record data per plant (Table 1), per root (Table 2), or per nodule (Table 3), for later analysis using a statistical software package (statistical analyses for the Results section were performed in ibm spss statistics v.20; IBM Corp., Armonk, NY, USA).
Table 1. Variables extracted by the software at the whole-plant level
List of recorded variables for each plant (one row in the whole-plant Excel file will give variable values for one plant)
NZD, nodulation zone defined for the data set; that is, zone including roots above the lowest nodule of the data set; NZP, nodulation zone defined for the plant; that is, zone including roots above the lowest nodule on the plant; RSE, root system equivalent.
‘Lateral roots’ throughout the table means ‘first-order lateral roots’ as higher order lateral roots, which did not bear nodules, were excised from the plants and therefore not included in the analyses.
Plant identification number
Total number of nodules
Number of nodules on the primary root
Number of nodules on lateral roots
Distance from base (top) of primary root to first nodule along the root
Distance from base of lateral root to last nodule along the root
Minimum x position of nodules on the RSE (i.e. minimum distance from base of lateral root to first nodule along the lateral root)
Maximum x position of nodule on the RSE
Vertical nodulation range for the RSE
Lateral nodulation range for the RSE
Length of primary root
Total length of lateral roots within the NZP; number of lateral roots within the NZP
Total length of lateral roots within the NZD; number of lateral rootsa within the NZD
Number of nodulated roots
Length from base of primary root to lowest nodule of the data set (this measure is the same for all samples and can be used for calculations of further variables of interest)
Total length of lateral nodulated roots
Presence of nodules on the primary root (coded as 0 for no nodules on the primary root, and 1 for nodules on the primary root)
Area of the 2D hull of RSE nodules
Area of the 2D hull of roots within the NZP
Area of the 2D hull within the NZD
Table 2. Variables extracted by the software at the root level
List of recorded variables for each roota (one row in the corresponding root data Excel file will give values for one root)
Data on primary roots were included in the root data file; primary roots are given plane identification number 0. In that case, entries relevant specifically to lateral roots are left empty.
NZD, nodulation zone defined for the data set; that is, zone including roots above the lowest nodule of the data set; NZP, nodulation zone defined for the plant; that is, zone including roots above the lowest nodule on the plant.
Plant identification number
Plane identification number (arbitrary value between 1 and 3 for each plane; value of 0 for the primary root)
Root identification number
Root position along the primary root (if the root is not the primary root)
Number of nodules
Lateral range of nodules (if the root is not the primary root)
Distance from base of lateral root to first nodule along the root
Distance from base of lateral root to last nodule along the root
In NZP (value of 1 if the root is within the NZP; value of 0 if the root is below the plant nodulation zone)
In NZD (value of 1 if the root is within the NZD; value of 0 if the root is below the nodulation zone of the data set)
Table 3. Variables extracted by the software at the nodule level
List of recorded variables for each nodule (one row in the corresponding nodule data Excel file will give values for one nodule)
Plant identification number
Plane identification number
Root identification number
Vertical distance from the base of the primary root
Lateral distance from the primary root along the lateral root (= 0 if the nodule is on the primary root)
In order to obtain estimates of the spatial variation in nodulation for a given group of plants, we compute average spatial maps (Rolland-Lagan et al., 2009) of nodulation patterns. These maps are computed by overlaying a grid onto the RSE of a plant to sample nodulation or root variables within each square of the grid. Areal nodule density is computed locally as the number of nodules within each 1 cm × 1 cm square (nodules per cm2). Root density is calculated locally as the cumulative length of all roots passing through that square (e.g. an RSE having two roots passing through the grid square, one crossing the whole square and the other one stopping in the middle of the square, would have a root length of 1.5 cm within the grid square, giving a root density of 1.5 cm of root per cm2). The nodule-to-root density within each grid square is computed as the number of nodules per cm of root within that square (this is equal to the ratio of areal nodule density to root density). We then compute and display average values across plants for each grid square, as well as associated standard errors, to obtain average maps and associated standard error maps for these variables (as in Remmler & Rolland-Lagan, 2012).
We developed a methodology and software to quantify nodulation patterns on pea root systems. Pea possesses a triarch vascular pattern which causes first-order lateral roots to naturally fall into one of three planes (Fig. 1). We imaged one plane of lateral roots at a time (Fig. 2a) and we used custom software capable of extracting the coordinates of all roots and nodules from the image (Figs. 2b–f). For all three planes of a plant, the software records each nodule's spatial coordinates as its position along the lateral root bearing it (x position) and the position of that lateral root along the primary root (y position); we refer to this 2D representation of the root system as an RSE (Fig. 3a) (see the 'Materials and Methods' section for more details). The method was applied first to the quantification of nodules in the nodulation zone of the wild-type Sparkle and then, as proof-of-principle, to that of the low-nodulation mutants R50 and E151, first introduced by Kneen et al. (1994). In total for the three genotypes (39 plants), we identified 4049 nodules (4034 of those being on lateral roots), and 2727 roots, or 1942 roots if we only consider the roots within the nodulation zone of each plant.
Nodule counts and distribution with respect to the root system of the wild-type Sparkle
With the software, we are able to extract a wealth of variables (Tables 1-3). These variables provide simple measurements of nodulation (e.g. mean number of nodules per plant or per root) and root architecture (e.g. mean lateral root length), and measurements combining variables linked to both nodule and lateral roots. For instance, the plant lateral nodulation range is defined as the horizontal distance between the nodule closest to and the nodule farthest from the primary root, all lateral roots being considered. The lateral root nodulation range is defined as the same distance but measured on a single lateral root. The relative lateral root nodulation range is obtained by dividing the lateral root nodulation range by the lateral root length, and the linear nodule density is determined by dividing the number of nodules on the lateral root by the lateral root length.
Average root and nodulation measures per plant are given in Table 4. Nodules occurred on the lateral roots within the first 4.5 cm down the primary root, and on average were spread over c. 5 cm in the lateral direction (Table 4). The wild-type bore c. 180 nodules which were borne by 27 roots. There were just a few nonnodulated lateral roots in the nodulation zone of the plant (NZP). Considering the cumulative length of all the root system within the NZP, the wild-type formed a ratio of one nodule per 2 cm of lateral root (Table 4). Nodulation measures can also be recorded per lateral root and related to the positioning of the lateral root along the primary root (Fig. 4a–e). The wild-type had all its nodules located on lateral roots restricted to a zone close to the cotyledons (Fig. 4a) and those nodulated lateral roots were the longest of the root system (Fig. 4b). Nodule count per root, lateral root length, lateral root nodulation range (defined as the distance between the nodule closest to the lateral root base and the nodule closest to the lateral root tip), linear nodule density (i.e. the number of nodules on the lateral root divided by the lateral root length) and relative lateral root nodulation range (defined as the lateral nodulation range divided by the lateral root length) were all significantly correlated to the point of attachment of the nodulated lateral roots (based on the calculation of Spearman's correlation coefficients, with P <0.001 for all statistical tests except P =0.01 for lateral root length). The fact that the relative lateral root nodulation range was correlated to the y position of the lateral root means that the reduction in the nodulation observed as y increased cannot be explained only by the shortening of the lateral roots (Fig. 4c–e).
Table 4. Examples of measures extracted at the whole-plant level for Pisum sativum Sparkle, R50 and E151
n =13 samples
n =13 samples
n =13 samples
Mean ± SE
Mean ± SE
Mean ± SE
The results of the statistical analysis (Mann–Whitney U-test) performed on these data are shown in Supporting Information Table S1.
NZP, nodulation zone of the plant; RSE, root system equivalent.
Primary and first-order lateral roots.
The nodule-to-root ratio is the total number of nodules divided by the total length of roots (primary and first-order lateral) within the nodulation zone.
The nodule-to-nodulated roots ratio is the total number of nodules divided by the total length of roots (primary and first-order lateral) that are nodulated.
The vertical range is the vertical distance between the nodule of the data set having its y coordinate closest to the base of the primary root and the nodule of the data set having its y coordinate furthest from the base of the primary root.
The lateral range is the lateral distance between the nodule of the data set having its x coordinate closest to the primary root and the nodule of the data set having its x coordinate furthest from the primary root.
Calculated within the plant nodulation zone, defined vertically from the top of the primary root to the point of attachment of the first-order lateral root bearing the lowest nodule for the plant.
Average mapping of nodulation patterns for the wild-type Sparkle
RSEs from all plants are used to compute average spatial maps of nodulation patterns and their associated standard error maps (see the 'Materials and Methods' section). Those maps illustrate spatial variations in local measures of average areal nodule density, average root density and average nodule-to-root density (Fig. 5; see also Supporting Information Fig. S1 for standard error maps). Wild-type nodules were restricted to a small area within the root system, that is, within 0–5 cm along the primary root axis, and this despite the presence of lateral roots below that area. The highest mean areal nodule density occurred c. 1–2 cm below the base of the primary root and 1–2 cm away from the primary root (Fig. 5a). The areal nodule density declined in clear gradients moving down and away from the most densely nodulated areas; this resulted in a roughly triangular shape for the nodulated zone (Fig. 5a). The root density map showed a clear gradient too, with the highest mean root density c. 1 cm below the base of the primary root and decreasing from its base to its tip (Fig. 5b). The map covered grossly a triangular area with the longer lateral roots higher on the primary roots than the shorter lateral roots (Fig. 5b). The spatial patterns in areal nodule density reflect to some extent the spatial patterns in root density: areas with high nodule counts are also areas with high root densities. To determine whether differences in nodule spatial patterns are independent of root architecture, we can display a nodule-to-root density map (Fig. 5c). This map illustrates that the number of nodules per cm of root also varied spatially, showing again a roughly triangular pattern (Fig. 5c).
Nodulation pattern of the two mutants compared with that of the wild-type
Although the wild-type and the mutants exhibited primary roots of similar lengths, their nodulation patterns were dissimilar, the mutants having a reduced number of nodules and yet a larger vertical nodulation range (Table 4). One of the most obvious differences between the two mutants and the wild-type was the presence of an additional nodulated zone. R50 and E151 exhibited a second wave of nodules which were borne by lateral roots located low on the primary root (Fig. 4f–j and Fig. 4k–o, respectively). E151 had a stronger nodulation phenotype than R50; in particular, whereas the wild-type and R50 had a ratio of one nodule per 2 and 3 cm of lateral root, respectively, E151 exhibited only one nodule per 16 cm (Table 4, Fig. 4i,n). The E151 lateral nodulation range was also much smaller than those of the two other pea lines, with nodules covering a much shorter distance on the lateral root than those of both the wild-type and R50 (compare Fig. 4h and 4m to 4c). Another striking difference was seen in the number of lateral roots borne by their primary roots (Table 4), although a comparison is difficult because some roots far below the zone of nodulation were missed in the wild-type. Both mutants, but E151 more than R50, possessed a higher number of lateral roots, which were generally much shorter than those of the wild-type (Fig. 4). As in the wild-type, R50 nodule counts per root, lateral root length, lateral nodulation range, linear nodule density and relative nodulation range were all significantly correlated to the y position of the nodulated lateral roots (based on the calculation of Spearman's correlation coefficients, with P <0.001 except P =0.012 for linear nodule density in R50). In E151, the root length and the linear nodule density were not significantly correlated to the y position of the nodulated lateral roots (P =0.45 and P =0.096, respectively), but all other correlations were significant at the 0.05 level.
On maps, the two nodulation zones of the mutants are obvious (Fig. 5d,g), and the differences between the two zones are easily seen. Whereas the upper zones were more or less similar to the triangular shape observed in the wild-type, albeit with fewer nodules overall (Fig. 5a,d,g), the lower ones were irregularly shaped and barely populated. The area with the highest areal nodule density in the upper nodulation zone was shifted in both mutants. In R50 it was extended towards the tips of both the lateral and the primary roots (Fig. 5d), whereas in E151 it was just shifted towards the lateral root tip (Fig. 5g). The root density map of R50 was similar to that of the wild-type in that the highest root density was found 2–3 cm down the primary root and 3–4 cm away from it (Fig. 5e). However, root density was noticeably lower in the first cm of the primary root. E151 had yet a different root density map (Fig. 5h) with an even greater reduction in root density within the first cm of the primary root, but a higher root density lower on the root system. This was a result of a multitude of short lateral roots from the base of the primary root downwards. In both mutants, as in the wild-type, the nodule-to-root density maps (Fig. 5f,i) show that the spatial variations in areal nodule density were not solely a result of spatial patterns in root density (i.e. areas with low nodule numbers did not simply result from low root density).
Nodulation is a potential target for legume crop improvement (McPhee, 2005; Gonzalez-Rizzo et al., 2009), yet measurements of quantitative traits relating to nodulation remain limited. Studies of plant nodulation generally rely on manually counting nodules on the root system (Guinel & Sloetjes, 2000), which is prone to human error and does not preserve spatial information. Here we propose a new methodology and software to quantify nodulation patterns in a standardized way in relation to the root system, using RSEs.
RSEs as a framework for quantifying nodulation
Because RSEs are based on the overlay of all lateral roots from all planes of each plant, the measurements are proxy measurements. We are aware that the RSEs do not have a direct biological meaning as they are not based on the position of the roots and nodules within the soil; however, they can be useful in the interpretation of plant signalling mechanisms because they provide information on distances and positions with respect to plant architecture. Characterizing nodule positions relative to the root system using RSEs allows the combination of data from multiple samples and the spatial analysis of nodules and roots within a common framework, permitting statistical comparisons between groups of plants. Such spatial analysis would be very difficult without an RSE representation, because the real distributions of nodules and roots within the soil are lost as soon as the plant is harvested for sample preparation.
We demonstrated here the relevance of the methodology by comparing three pea lines, wild-type, R50 and E151, based on their RSEs. R50 and E151 have both been screened as low-nodulation mutants 21 d after inoculation (Kneen et al., 1994), while our study was on plants harvested 28 d after inoculation. Here, the presence of a second zone of nodulation suggests that although these mutants are low nodulators they retain the ability to form nodules where the lowest lateral roots develop. Our study establishes clearly that R50 and E151 differ in the severity of their phenotype, suggesting that the two mutations affect different stages of nodule development and/or nodulation regulation mechanisms. We found that in the wild-type and R50 the relative nodulation range decreased with the y position, and these findings support the systemic autoregulation of nodulation (Ferguson et al., 2010). However, the absence of such a correlation in E151 suggests that there may be another regulatory mechanism at play. E151 root architecture is also distinct from that of both the wild-type and R50; it is characterized by a multitude of short lateral roots throughout the root nodulation zone. It will thus be a useful mutant in which to study whether the developments of nodules and lateral roots are coordinated (Desbrosses & Stougaard, 2011). In E151, the development of those two lateral organs could be uncoupled, the development of the lateral root could override the development of the nodule or the development of the lateral root could compensate for that of the nodule.
Transferability of the methodology
We envision that this type of approach could be used in different types of study. Developmental analyses could be performed to understand better nodule organogenesis. Younger and older plants could be used as long as developing and senescent nodules, respectively, are distinguishable. Because the method is semi-automatic, the user is always capable in the end of deleting misidentified nodules or adding structures thought to be nodules (see the 'Materials and Methods' section). The measures extracted by the software may also be used to compare nodulation patterns among cultivars, among different pharmacological or bacterial treatments, or among varied environmental conditions. While there are other methods for quantifying nodulation, none have included average spatial maps of nodule distribution and density or relative spatial occupancy of nodules with respect to the root system. Spatial information is essential for evaluating differences in nodule patterning that could reveal genotypes optimal for specific environmental conditions. For example, two cultivars might have the same number of nodules, on average, but may differ in nodulation range, making the cultivars nodulating lower on the root system a preferable choice for cultivation in dry climates or during periods of extended drought (Burias et al., 1990). Detailed spatial information about the root system itself is also critical for the analysis of nodulation by making it possible to determine whether a group of plants that nodulates less or more than a control group is affected in nodulation per se, or whether its lower/higher nodule number may simply be a result of an altered root phenotype (e.g. a mutant with few roots may as a result have few nodules). Because the list of variables extracted is quite exhaustive, other related nodulation measures could also be computed.
Software validation, limitations and future improvements
The proposed software allows highly detailed measurements of nodulation to be quantified in a reasonable period of time. The preparation and photographing of plants takes slightly under 1 h per plant. Once images have been acquired, full processing of a data set of 13 plants takes c. 4 h (an average of 20 min per plant), which is a reasonable time considering the wealth of information that we extract. Using the software, although the root tracing and the length calculations are highly accurate (Notes S1, Fig. S2), an average of 10.0% of the nodules on a plant are missed. This is because any nodules occurring on the face-down sides of the roots during imaging will not be visible in the photograph. Furthermore, roots sometimes overlap each other, obscuring more nodules from the camera view. However, as there is no reason why the nodules of two different treatment groups would have different proportions of nodules obscured in the photographs, this discrepancy does not diminish the ability of this method to detect nodulation differences between groups.
Parameters for the automated nodule and root extraction algorithms were developed for the nodule and root features (colour and size) that were typical for the conditions of the plants used in this study. However, nodule and root features might vary in other mutants, different species, plants of different ages, or plants grown in different conditions, so the automated extraction may not work well for all studies. In those cases, the nodules will simply have to be extracted fully with user input (i.e. by clicking on a nodule or along a root to record its coordinates). Alternatively, the parameters of the automatic nodule detection could be adjusted in the source code by a researcher with programming and image analysis skills to suit other root or nodule colours.
While the software allows for quick and easy extraction of the root and nodule data from the photographs, disentangling and separating the three planes of lateral roots for imaging can be time consuming. Therefore, the method would not be well suited for denser root systems. The method is also only applicable to root systems that have nodules on first-order lateral roots (as is the case in the pea cultivar we studied). However, in such cases where tertiary roots are of interest, we envision that the method could be adapted to first record the attachment points of all first-order lateral roots along the primary root; then one would excise the first-order lateral roots and process them in the same way that we treated the primary root in this study.
Our software allows us to quantify nodulation in terms of root architecture but not in terms of symbiotic efficiency, that is, the effectiveness of the symbiosis at fixing atmospheric nitrogen. However, one could remove the nodules once the plant has been imaged, and keep shoots, roots, and nodules aside to measure their dry weights to then evaluate the symbiotic efficiency of a particular association. For example, return on nodule construction cost – which is an estimation of the amount of nitrogen obtained by the plant in return for the photosynthates it provides (Oono & Denison, 2010) – could be analysed.
With the modifications and exceptions described in the previous section, the method we have presented in this paper will be broadly applicable to a number of future studies. It may be used to compare and evaluate nodulation levels and distributions in other mutants, different cultivars, or different legume species; to evaluate the effects of various physiological treatments on nodule formation; or to determine the optimal timing, temperature or other conditions for ideal rhizobial symbiosis. The method could also be applied to a time-series study, the results of which could be used to develop or refine models of nodule patterning and autoregulation of nodulation.
The research was supported by operating grants to F.C.G. and A-G.R-L. provided by the Natural Sciences and Engineering Council of Canada. We thank Emily Macdonald for the idea of nodulation maps and for helpful discussions and Jocelyn Pender for processing images for the trial data set.