Development of a novel semi-hydroponic phenotyping system for studying root architecture

Additional keywords: bin system, branching, Lupinus angustifolius, narrow-leafed lupin, root length, root morphology, root system architecture, root traits, RSA.

Quantifying the architecture of root systems is essential because crop productivity is almost always influenced by the availability and accessibility of water and nutrients that are heterogeneously distributed in the soil (Miller et al. 2003; Cichy et al. 2009; Gregory et al. 2009). Root architecture directly impacts the capture of such soil resources and is, therefore, fundamental to crop productivity (Lynch 1995; Casson and Lindsey 2003; Malamy 2005; Watt et al. 2006; Dunbabin 2007; Fita et al. 2008; Gregory et al. 2009). Changes in root system architecture and exploitation of soil water influence the accumulation of crop biomass (Hammer et al. 2009).

Progress in root measurement methodology has enhanced our ability to visualise, quantify and conceptualise root architecture and its relationship to plant productivity (Lynch 1995). Numerous approaches have been developed for the study of root architecture with the support of advanced optical recording techniques. However, reliable approaches for phenotyping large numbers of plants are still being developed. Several non soil-filled methods for evaluating roots are available, including hydroponic, aeroponic and agar-plate systems (Gregory et al. 2009). The soil-filled root observation chambers (Manschadi et al. 2008) or rhizotrons (Wiese et al. 2005), and growth pouches (Bonser et al. 1996; Liao et al. 2004) are often used in studying roots. More recently, an improved pouch system was developed by Hund et al. (2009) for rapid measurement of maize (Zea mays L.) root systems during the first days of lateral root growth. In this system, roots grew on the surface of blotting paper facilitating the two-dimensional observation of root growth over time. However, phenotyping large sets of genotypes beyond very early growth stages remains problematic, particularly for mapping studies of quantitative trait loci (QTL).

The main objective of this work was to design and construct an inexpensive, space-saving, high-throughput phenotyping system for quantifying root architecture for large numbers of plants. This system was evaluated through a screening trial with 20 wild genotypes of narrow-leafed lupin (Lupinus angustifolius L.). This paper describes the construction of the bin system, along with a summary of the data that was collected in the screening experiment using this bin system.

Fig. 1 shows photographs of the operational system. Each unit of the system consisted of a plastic mobile bin, support framework, 20 growth units, and a controlled watering system (Figs 2, 3). A commonly-used mobile rubbish bin (240 L, top 751× 580 mm, height 1080 mm; Silverlock Packaging, Perth, WA, Australia) was chosen. The size of this bin allows for studying deep root systems (up to 1.2 m) and many genotypes within a limited space (40 plants in ~0.5 m²). Also, the bin is mobile (on wheels), which helps in moving and repositioning during the course of the experiment.

Fig. 1.  Operational semi-hydroponic bin system designed, constructed and tested for two-dimensional, non-destructive measurements of root system architecture: (a) a view of inside the bin system showing 10 growth units and pumping system (additional 10 units to be placed into the front part of the bin), and (b) side view of growth units fitted inside the bin with Lupinus angustifolius plants grown in the system showing the operation of the irrigation system.

Fig. 2.  Schematic representation of the semi-hydroponic bin system: (a) Clear flat acrylic panel; (b) supporting material – calico cloth; (c) Lupinus angustifolius plant; (d) 240 L bin; (e) extension column of cloth; (f) water or nutrient solution; (g) pump and irrigation system. See Fig. 3 for ‘support framework’.

Fig. 3.  The support framework of the semi-hydroponic bin system: (a) front view, (b) side view. The rigid PVC tubing was secured to the middle bar of the support frame and the flexible tubes with drippers branched from the rigid tubing onto each dividing bar.

The stainless steel supporting framework consisted of a top frame (400 × 600 mm) connected perpendicularly with two U-shaped support bars (420 mm deep) (Fig. 3). Within the top rectangular frame there were 20 areas divided equally by nine parallel rods and one central dividing rod for inserting the growth units. The bottom wires of the support bars were bent in a zig-zag to hold the growth units and separate them evenly.

Each plant growth unit contained a 5 mm thick acrylic panel (260 × 480 mm) covered with black calico cloth (570 × 900 mm). Four 15 mm fold-back butterfly clips were used to attach the cloth to the panel on both vertical sides. The lower part of the cloth over each panel was cut in half to make two columns for root training.

Each bin system was equipped with a watering system consisting of a fountain pump (Pond Max PM1000F, pump flow rate 550 L h^–1) and drippers (Philmac Dripper True Drip, Philmac Pty Ltd, Adelaide, SA, Australia) connected with rigid and flexible PVC tubing. The rigid PVC tubing was secured to the middle bar of the support frame with cable ties, and the flexible tubes branched from the rigid tubing onto each dividing bar (Fig. 3). A timer was attached to the pump system for periodic water supplying.

The growth units were positioned in the chambers against the horizontal rods and sited on the zig-zagged wire frame at the base (20 growth units per bin). The drippers located in the middle of the growth units were adjusted to face the growth unit. The wrapped acrylic panel with cloth formed pouches for seedlings to grow, with moisture maintained by irrigation. The growth unit allowed easy imaging down to the bottom of the plate, and the extensions of the cloth allowed room for a total length of root growth of ~1.2 m.

A list of equipment and materials needed to construct one single bin system is provided in Table 1.

Table 1.  List of equipment and materials for construction of inexpensive semi-hydroponic bin system

A screening experiment used two bin systems in a controlled-temperature glasshouse in June–July (20/15°C day/night temperatures) at The University of Western Australia (31°58′S, 115°49′E). Twenty wild genotypes of narrow-leafed lupin (Lupinus angustifolius L.) were randomly chosen, as representatives, from a subset of wild germplasm containing 111 genotypes to be screened for genetic intrinsic of root traits as part of an on-going research project. The 111 genotypes were previously selected using Diversity Array Technology (DArT) method which can detect DNA variability in hundreds of loci simultaneously from a large germplasm pool (1301 genotypes). Seeds were scarified by scalpel and sown in pots filled with washed river sand in a glasshouse. Seeds were germinated for 4 days and four germinated seeds of each genotype were randomly transplanted into growth units of two bins (two plants per bin). Bins were filled with 30 L solution containing (in μM): K (1220), P (20), S (1802), Ca (600), Mg (200), Cu (0.2), Zn (0.75), Mn (0.75), B (5), Co (0.2), Na (0.06), Mo (0.03), Fe (20) and N (1000). The nutrient solution was supplied to the plant growth units by the automatic irrigation system through a timer controller. The periodic pumping was set as 10 min on–5 min off during the daytime (0600–1800 hours), and 5 min on–10 min off during the night time (1800–0600 hours). The solution was refreshed weekly.

Root growth was monitored and measured weekly. The pattern of root growth on each plate was photographed by removing the cloth, and an early pattern of root growth (up to 3 weeks) was traced onto a clear panel using a marker pen. After 3 weeks, taproots of most genotypes were beyond the lower base of the plate and the length of taproot was measured. Plants were harvested 6 weeks after transplanting. The roots were subsampled by cutting into 20 cm sections along the taproot. Dry mass of shoot and roots was determined.

Root subsamples from harvest were scanned in greyscale at 300 dpi using a desktop scanner (Epson Expression Scan 1680, Epson America, Long Beach, Canada). Scanned images were analysed in WinRHIZO (V.2009 Pro, Regent Instruments, Montreal, QC, Canada) for root morphological and architectural studies. The debris removal filter of WinRHIZO was set to remove objects with an area smaller than 1 cm² and a length : width ratio lower than 10. Diameter classes were set to 0.17 mm (equivalent to two pixels) with 15 equal intervals. The program distinguished the taproot, lateral roots and root hairs in the images and generated root data of several traits, indicating that the image quality was appropriate. Root length data computed in WinRHIZO were compared with measurements by the Tennant’s (1975) method and by manually tracing root images, with all three methods producing consistent results (data not shown).

One-way ANOVAs were performed on root traits using PASW Statistics 18 (SPSS Inc., Chicago, IL, USA). Non-significant differences were found on most of root traits between bins (95% confidence, paired-samples t-test) and, therefore, four plants were treated as four replicates in the analysis. The multivariate standard errors of skewness and kurtosis were 0.29 and 0.58, respectively, when all parameters were included in the analysis, indicating no significant departure from multivariate normality in the dataset. General correlations between parameters were examined with Pearson’s correlation coefficients. Correlations were considered statistically significant and are indicated at *, P ≤ 0.05 or **, P < 0.01. Selected traits were subjected to principal components analysis (PCA; eigenvalue >1, Tabachnik and Fidell 1996) and k-means clustering analysis to compare traits with similar functions among genotypes.

Growth of both shoots and roots was vigorous in the system tested. The morphology of lupin root systems was recorded photographically by removing cover cloth. Genotypic variation in root growth dynamics was apparent. The root systems of all genotypes tested primarily consisted of a taproot and first-order branches in the first 3 weeks (Fig. 4) and second-order branches were also observed for some genotypes in the later growth stage (Fig. 5a).

Fig. 4.  Example images showing root morphology and development of a wild Lupinus angustifolius genotype grown in our constructed bin system. Images were taken at (a) 2, (b) 4 and (c) 6 weeks after planting (scale bar = 5 cm).

Fig. 5.  Example images of Lupinus angustifolius root showing (a) taproot, first order branches, second-order branches and root hairs on both order branches (scale bar = 5 cm), and (b) a close show of dense root hairs on first-order branches of a different genotype (scale bar = 1 cm).

By the end of week 6, there was genotypic variation in the density of first-order branches, and in the abundance of second-order branches. It is notable that several genotypes produced occasional fast-growing geotropic first-order branches that resembled taproots to some extent. A few genotypes also had dense root hairs usually on the first-order branches, creating an increased surface area (Fig. 5a, b).

Phenotyping of 20 wild genotypes of narrow-leafed lupin produced root systems that were readily imaged and evaluated. Variations in some root traits between the tested genotypes were large (Table 2). Greatest variations were observed in the number of links, specific root length, total root length and root volume (coefficient variation, CV = 0.87, 0.84, 0.79 and 0.75, respectively). Root area, link length, root tissue density, branch number and density, link area and branch intensity also differed significantly between genotypes (CV = 0.67, 0.66, 0.64, 0.63, 0.58, 0.58 and 0.57, respectively). Mean root diameter showed moderate variation (CV = 0.46). A large proportion of root length occurred in diameter classes of 0.34 to 1.19 mm with 67% in relative diameter-class length (Fig. 6). A diameter class of 0.68–0.85 mm had the longest root length (20% of the total length).

Table 2.  Descriptive statistics for root traits of 20 genotypes of narrow-leafed lupin 6 weeks after planting
Branch density, the number of branches per unit taproot length; branch intensity, the number of branches per unit total root length; root tissue density, root mass per unit volume; specific root length, root length per unit mass; a link, a root segment that is between two apexes or an apex and a tip. Data of four plants for each genotype were included in analysis

Fig. 6.  Mean root length in diameter class length (DCL, with s.e.) and relative diameter class length (rDCL) of 20 Lupinus angustifolius genotypes tested 6 weeks after planting. Data are mean across the 20 genotypes (n = 4).

Pearson’s correlation coefficient analysis revealed strong correlations among morphological (e.g. root length, specific root length, root area, average diameter) and architectural traits (branch density and branch intensity) (Table 3). In particular, the total root length was highly correlated with root area (r = 0.94), branch number (0.90), root volume (0.85) and branch density (0.84) (each at the 0.01 level of significance). Specific root length increased as numbers of branches, branch density and branch intensity increased (r = 0.77, 0.70 and 0.50, respectively, P < 0.01). Greater branch density was strongly correlated with increasing branch intensity and root diameter (r = 0.83 and 0.61, respectively, P < 0.01), and decreasing root tissue density (r = −0.61, P < 0.01).

Table 3.  Pearson’s correlation matrix for root traits of 20 genotypes of narrow-leafed lupin 6 weeks after planting
Correlation is significant at the **, P < 0.01 level (two-tailed). Root topological traits with CV values greater than 0.45 were included in the matrix (see Table 2 for CV values). Genotype mean data (n = 4) were used for Pearson’s correlation analysis

Eleven root traits with large variations (e.g. CV > 0.50, see Table 2 for CV values) were included in the PCA analysis, which showed that 92% of the observed variation of root architecture was captured by four principal components (Tables 4, 5). The first and strongest principal component was associated with genotypes differing primarily in total root length and specific root length representing 31% of the variability. PC2 consisted mostly of root diameter and volume and accounts for 26% of the variability. PC3 and PC4 add another 24% and 11% of the variability, respectively, and were strongly influenced by branch intensity (PC3) and taproot length (PC4).

Table 4.  Results of principal component analysis (PCA) on 11 selected root traits (see Table 5) of 20 genotypes of narrow-leafed lupin 6 weeks after planting – the proportion of variation explained with the first four components presented
Rotation method: varimax with Kaiser Normalization. Components with eigenvalues >1 are considered significant (Tabachnik and Fidell 1996). Genotype mean data (n = 4) were used for PCA

Table 5.  Results of principal component analysis (PCA) on 11 selected root traits (see Table 5) of 20 genotypes of narrow-leafed lupin 6 weeks after planting – rotated loading scores of traits on each component

The tested genotypes varied significantly in the development of taproots and branches. Taproot length ranged from 31 to 111 cm with an average of 73 cm (Table 2). The total root length differed by up to 19-fold among genotypes. Clustering algorithms were used to automatically assign genotypes in terms of variability among root traits. The k-means clustering analysis was performed on the global data of 10 primary traits of each genotype. The 20 genotypes tested were generally separated into two major clusters, with nine and 11 genotypes in each. Branch number, total root length and branch density were the most important contributors to this separation (F = 56, 42 and 37, respectively, P < 0.001; Table 6). In contrast, taproot length and root diameter had little influence on the separation. The value of the final cluster centre indicated genotypes in cluster 1 possessed larger root systems with more root length, branch numbers, and branch density than those in cluster 2.

Table 6.  ANOVA table and final cluster centres of k-means clustering analysis on 10 root traits of 20 genotypes of narrow-leafed lupin 6 weeks after planting
Genotype mean data (n = 4) were used for k-means clustering analysis. Significant values are indicated: **, P < 0.01; *, P < 0.05

The reason we developed an efficient phenotyping system was to examine genetic intrinsic variations in root traits among wild genotypes of narrow-leafed lupin and to provide basis for root models to constructing three-dimensional root architecture for selecting superior root traits for breeding program. The evaluation experiment using wild narrow-leafed lupin genotypes was to test efficiency of the established phenotyping system and the performance of wild lupin in the system. Therefore, this preliminary investigation has provided information for a follow-up large screening experiment with the same species. The rationale for the choice of narrow-leafed lupin genotypes was also because of the availability of a large germplasm pool of wild narrow-leafed lupin from landraces and diverse climatic and geographic locations (Clements and Cowling 1991; JC Clements, pers. comm.).

Care is needed when handling growth units for observation and photography. This is because lupin roots are relatively fragile and root damage may occur. During root observations, roots and cloth remain moist, but exposure time for observation should be minimised to prevent roots from drying out. The irrigation system can be amended, e.g. by adding another opening for each unit if competition for water occurs between neighbouring plants. The volume of water or solution in the bin can be adjusted depending on the frequency of refreshment and duration of the experiment. Watering frequency and water flow can be optimised with a timer controller attached to the pump system.

Plants of wild narrow-leafed lupin genotypes were tap-rooted confirming previous observations (Clements et al. 1993). Each wild lupin genotype displayed a unique root system architecture when cultivated in our phenotyping system. The root system of each genotype retained similar morphological pattern, particularly in the proximal part, in terms of the number of the first order branches and branch angles during the 6 week growth period. We compared weekly root images and data and concluded that the 6 week root systems presented larger diversity of root traits than the early roots, although variation could be seen as early as 2 weeks after planting. Differences in root architecture among the tested genotypes primarily include taproot length, branch number and branch density. For example, some genotypes developed deeper taproot, or substantially high branch density in the upper root. Root architectural differences have significant impact on the efficient acquisition of water and nutrients in soils (e.g. Nibau et al. 2008; Rengel and Damon 2008; Rose et al. 2009), and, therefore, have relationship to plant productivity (Lynch 1995).

During the preparation of this report, we have obtained comparable data from two consequent phenotyping experiments using the same semi-hydroponic system developed in this study. The large screening experiment examined root trait variation among 111 wild genotypes of narrow-leafed lupin, followed by a repeat experiment with 10 selected genotypes undertaken in the same phenotyping system under the similar grown conditions. There was no significant difference between bins (replications) (P ≤ 0.05). Consistent rankings of genotypes in each root trait between these two separate experiments confirmed the value of the approach for screening large numbers of genotypes using our designed semi-hydroponic phenotyping system. These screening experiments demonstrated a reliable and efficient phenotyping platform for the development of lupin roots. Further examinations are required to explore the potential for evaluation of root system architecture of other crop species on the phenotyping system, and to investigate the biological effects of the specific adaptations of selected genotypes with novel root traits.

Numerous rooting parameters can be collected using this bin-based system coupled with a digital photography and image analysis software (e.g. WinRHIZO). The root architectural traits achieved can be divided into two general categories: topological properties (describing the pattern of root branching) and geometric properties (such as the growth angle of root axes) (Manschadi et al. 2008). Both traits are critical in studying root architecture (Lynch and Brown 2001; Kato et al. 2006; Manschadi et al. 2008). According to Fitter (1991), five types of data are required to reconstruct a three-dimensional model of a root: the number of internal and external links (without and with top meristem, respectively); the length of the links; the distribution of branches within the root, i.e. the topology; the branching angles; and the diameter per link. If one is interested in total size, rather than three-dimensional distribution, the branching angles can be left out. For the total length, rather than volume or weight, the diameters can be excluded and only the first three types of information are needed (van Noordwijk et al. 1994). Data collected from our constructed phenotyping system can be used to simulate a three-dimensional model using root modelling software such as ROOTMAP (Diggle 1988; Dunbabin et al. 2002) and SimRoot (Lynch et al. 1997).

To help understand genotypic variation in root architecture and functions, routine data transformation of some traits were performed to normalise the disparity between plants of different sizes. For example, specific root length (calculated as root length divided by root mass) is an important measure of how root biomass can be used for nutrient acquisition, and is widely used in studying root architecture and functions (Ostonen et al. 2007). Root branch density (number of branches per unit taproot length) and branch intensity (number of branches per unit root length) govern exploration of the soil matrix and may also affect nutrient acquisition (Fitter 1991).

Unlike other hydroponic or aeroponic phenotyping systems (reviewed by Gregory et al. 2009), our bin system is semi-hydroponic and partly aeroponic. It is notable that the bin system significantly reduces environmental stresses such as hypoxia and drought effects by optimising water supply with the equipped automatic irrigation system with a controller. This system using nutrient solution has the advantage of being faster and more accurate for analysing a range of root traits, since it eliminates the challenges of soil contamination, and root loss during washing. Because the supply of water and nutrient can be adjusted for each individual bin system, there is a potential to use this system in studying root plasticity in morphological and physiological responses to water and nutrients.

To conclude, our constructed phenotyping system made from inexpensive readily-available materials was efficient in screening root traits in wild narrow-leafed lupin. It can be used to investigate relatively deep root systems and phenotype root architecture in many genotypes in a relatively small space. Furthermore, the range of root architectures observed in our phenotyping platform can be simulated using root modelling software, i.e. ROOTMAP and SimRoot, for rapid and reliable reconstruction of root systems grown in a particular environment.