Automated phenotyping of root system architecture and shoot growth

The Plant Phenotyping Platform for Plant and Micro-organism Interactions (4PMI) is hosted by the UMR Agroécologie (INRA Dijon, France, https://www6.dijon.inra.fr/umragroecologie/Plateformes/Serres-PPHD). 4PMI is an automated phenotyping platform based on conveyors (LemnaTec, Würselen, Germany). It is composed of four different greenhouses where environmental conditions can be varied independently (temperature, light, hygrometry, individual plant watering regime). For each greenhouse, conveying lanes (in total 60 lanes for 4PMI) carrying 26 carts each (in total 1560 pots), are used to transport plants either towards two watering units or to the imaging units. Watering units consist of two weighing terminals (ST-Ex, Bizerba, Balingen, Germany) and high-precision pump-watering stations (520Du, Watson Marlow, Wilmington, MA, USA). The visible imaging unit designed to acquire non-invasive shoot images of either pots or RhizoTubes is composed of a 3D image acquisition cabin with top and side cameras (Basler piA2400-17gm/gc with a motorized lens Pentax C-Mount 12.5–75 mm C6Z1218M3 2/3″ 6× Megapixel, Basler AG, Ahrensburg Germany) and illumination (HE 28W/865, OSRAM, Augsburg, Germany). The top camera can take zenithal images while the side camera mounted at an angle of 90° to the vertical axis of the plant allows acquiring shoot images at different angles of rotation, whose number and amplitude depends upon the plant type. Circulation of plants via conveyors, image acquisition and watering is regulated by a control personal computer using the LemnaLauncher software bundle (LemnaTec, GmbH, Würselen, Germany).

RhizoTube description

RhizoTubes and RhizoCab were designed in close collaboration with Inoviaflow (Dole, France). RhizoTubes are cylindrical rhizotrons where plants grow and can be phenotyped dynamically. They are 18 cm in diameter and 50 cm high, and weigh approximately 12 kg. Because our visible imaging unit is about 1.6 m, this allows working with plant shoots 1.1 m high, which is sufficient for the time span of our plant root observations on the various species we work on. RhizoTubes are composed (Fig. 1) of concentric tubes, which delimit the outside to the inside of RhizoTube the root growing zone from the substrate zone and lastly the center where nutrient solution is supplied. The root growing zone lies between an inner permeable membrane (mesh size of 18 µm) and the external outer transparent polymethylmethacrylate tube (Fig. 1), separating the plant root from the soil. This membrane has been designed to be permeable to nutrients, water, plant rhizodeposits and microorganisms but it does not allow roots to pass through. As such the full root system is confined in two dimensions (Fig. 1c) and can be photographed using the Rhizocab camera through the outer transparent tube. A central inner tube defines the substrate thickness (about 2.5 cm, around 2.5 L of substrate) (Fig. 1). From one up to six plants can be sown simultaneously in a RhizoTube. Nutrient solution is supplied at the top of the RhizoTube and flows to the surrounding substrate zone.

Conveyors (Fig. 2a) transport either pots or RhizoTubes to the high-throughput sequential phenotyping cabins to phenotype shoots at various wavelengths. Each of the growing units (RhizoTube or pot) is uniquely identified by a Radio Frequency IDentification (RFID) chip in its base. As for pots, when needed, RhizoTubes are conveyed either to the watering stations where they can receive the target amount of nutrient solution, with a possibility of two different ones. Each RhizoTube is drained to evacuate excess nutrient solution through a hole in its base. The solution is automatically released from RhizoTubes when they pass through a drainage station. An opaque shell to shade roots from light covers the transparent outer tube of the RhizoTubes. This shell avoids algal growth in the soil using two half-shells of aluminium, which also provides insulation. This opaque shell is automatically and mechanically opened and smoothly lifted by about 5 mm above the RhizoTube basis by an arm equipped with a plier when RhizoTubes enter the RhizoCab. Opening the shell by about 2 cm allows the light to illuminate only the zone where roots can be imaged by the RhizoCab camera. Lifting the shell permits the RhizoTube to rotate on the RhizoCab turntable.

For some plant species (e.g. wheat), seeds are germinated on moistened filter paper and then installed within the RhizoTube. For others (e.g. pea) seeds can be directly installed within the RhizoTube (Fig. 1c). In the case of grapevine, herbaceous cuttings were prepared as previously described [51, 52] until they developed 6 leaves, and then transplanted in the RhizoTube.

Experiments performed

A series of experiments were conducted to assess (1) how the RhizoTube affects plant cultivation as compared to growth in pots and (2) the quality of images captured by RhizoCab for further phenotyping. Plant species and environmental conditions were chosen so that to (1) represent various research thematics of our unit, (2) assess the suitability of our system for different plant root architectures, and (3) taking into account interactions with microorganisms leading to specialized structures such as nodules (e.g. legume plants) or mycorrhiza (grapevine). Experiment 1 was performed on grapevine (Vitis vinifera L.) to assess survival within RhizoTubes of herbaceous cuttings, the usual means of propagation. Experiments 2 and 3 were conducted to characterize root architecture of nodulated pea roots (Pisum sativum) when grown in RhizoTubes or in pots and the influence of either different genotypes or varying soil nitrogen availability. In experiment 4, the responses to soil mineral N availability of a crop (Brassica napus) and a weed (Vulpia myuros) species were compared between RhizoTubes and pots. In experiment 5 the impact of water deficit on legume plants (pea and Medicago truncatula) was compared in pots and RhizoTubes. As RhizoTubes were developed to provide a powerful tool for investigating plant–microorganism interactions, experiment 6 aimed at establishing the persistence of microorganisms on roots within the RhizoTube rooting medium following their early inoculation within the RhizoTube rooting medium.

Growth and mycorrhization of grapevine plants in RhizoTubes

RhizoTubes are suitable for the growth and development of grapevine plantlets obtained from herbaceous cuttings. We observed a well-balanced development of both shoots and the root system. Furthermore, no symptoms of root alteration were detected. Experiment 1 was therefore performed to check whether these systems are adapted to the production of mycorrhized grapevine plants. The shoot/root dry biomass ratio (0.57) measured in RhizoTubes was similar to the shoot/root dry biomass ratio calculated for pot-grown plantlets, indicating that overall plant development was not modified in RhizoTubes (data not shown). As observed in Fig. 3a, within RhizoTubes the number and diameter of the roots, and overall root architecture (namely extent of branching) was determined. The total mycorrhization level (frequency, F%) was around 60 % and a root age-dependent extent of mycorrhization was observed (Fig. 3a).

Nodulated root architecture in RhizoTube compared to that in pot: the pea example

In experiment 2, pea plants (cv Kayanne) were supplied with a high level of mineral nitrogen. Under these conditions, root nodulation did not occur, neither in pots nor in RhizoTubes (Fig. 3b).

Both total plant dry matter and root dry matter were similar in pots or RhizoTubes 28 days after sowing (Table 2). Root phenotype analysis yielded descriptors such as the length of main root (Table 2), and architectural traits such as the distribution along main root of the density of the first- (Fig. 4a) and second- (Fig. 4c) order lateral roots and of their maximal length (Fig. 4b, d). The length of the main root was 47 % higher in RhizoTubes compared to pots (Table 2). The root distribution profiles along the main root showed that number and maximal length of first order lateral root roughly evenly decreased with depth in pots (Fig. 4a, b). In RhizoTubes, they also decreased from main root base, similarly to the situation observed in pots 25 cm more below from the hypocotyl (Fig. 4a, b). There were more first order lateral roots in pots than in RhizoTubes from the hypocotyl down to 25 cm below. More than 25 cm below the hypocotyl, this was inversed: while both the number of first order lateral roots and their maximum length reached a plateau in RhizoTubes, in pots they decreased down to zero (Fig. 4a, b). Conversely, the number and maximum length of second-order lateral roots were similar throughout the distribution profile for plants growing in pots and in RhizoTubes (Fig. 4c, d).

	Total plant dry matter (g/plant)	Root dry matter (g/plant)	Main root length (cm)
Pots	2.04 ± 0.34 (A)	0.28 ± 0.04 (A)	33.8 ± 8.7 (A)
RhizoTubes	1.93 ± 0.33 (A)	0.35 ± 0.06 (A)	49.6 ± 0.05 (B)

Total plant (i.e. shoots and roots) dry matter, length of main roots of Kayanne genotype (Pisum sativum) cultivated in pots or RhizoTubes with 10 mMeq soil mineral nitrogen were measured after 28 days after sowing. Letters in parentheses indicate significant differences (P < 0.05) between traits measured in RhizoTube or in pot

As quoted previously for Kayanne, addition of nitrate prevented nodulation in both rooting media (Fig. 3b). In experiment 3, two pea genotypes (cv Caméor and Kayanne) were grown without mineral N and inoculated with symbiotic rhizobial strains. Root images acquired using the RhizoCab showed that nodulation occurred both in pots and RhizoTubes on the main and lateral roots when nutrient solution was depleted of mineral nitrogen (Fig. 3c, d).

A deeper analysis of nodule distribution along the main and lateral roots was conducted on a set of Kayanne plants cultivated for 28 days without soil mineral nitrogen. While the number of nodules counted on lateral roots was very similar between containers (Fig. 5b), their distribution along the main root differed (Fig. 5a). In both RhizoTubes and pots there was an upper main root zone without any nodules, followed downward by a 16–20 cm long root zone bearing 1–3 nodules per cm, and lastly the terminal region without nodules. However, the main nodule-containing root zone was around 6 cm higher in RhizoTubes compared to pots (Fig. 5a).

Mineral N response of contrasted plant species in RhizoTubes and pots

Growing those species in RhizoTubes permitted the acquisition of root system images using the RhizoCab (Fig. 6). However both species developed very thin and entangled roots, which precluded detailed manual RSA measurements. Root biomass analysis demonstrated that the response to soil nitrogen was significantly higher for oilseed rape than for Vulpia but not significantly different between container types. Root dry matter and leaf area responses to soil mineral nitrogen were highest for the nitrophilic species (Table 4). Although the change in leaf area was significantly less in rhizotubes than in pots for both species, the leaf area of oilseed rape was more responsive than that of Vulpia to mineral N both in rhizotubes (P = 0.017) and in pots (P = 0.001).

Water stress responses of two plant species in RhizoTubes and pots

Whatever the container or species, water deficit impacted less root biomass than either shoot or nodule biomass (Table 5). The water deficit effect was however less severe on pea roots than on Medicago roots but the reverse was observed for plant nodules. Shoot and root growth were slightly more depressed in pea grown in RhizoTubes than in pots. The response of nodule biomass to water deprivation was similar for both species whether cultivated in pots or RhizoTubes (Table 5).

Root colonization by P. fluorescens C7R12

In order to check if microbial strains could persist in RhizoTubes, the bacterial strain Pseudomonas fluorescens C7R12 was inoculated after sowing pea and wheat plants. The goals of this assay were to: (1) characterize the bacterial cells concentrations on plant roots after 4 weeks of culture (measure of bacterial colonies on KING B agar plates) and (2) test if the method used to disinfect RhizoTubes (sodium hypochlorite solution) was effective enough to eliminate P. fluorescens cells after harvesting. This last point was performed in order to avoid microbial contamination for the following cultures.

Pseudomonas fluorescens C7R12 quantification on KING B agar plates showed a good persistence of the bacterial strain on roots in RhizoTubes. After 4 weeks of culture, the LOG of the densities of P. fluorescens C7R12 on roots were 6.19 ± 0.26, 5.69 ± 0.47 and 5.77 ± 0.29 CFU/g for respectively wheat, pea and wheat and pea grown together. Moreover, after culturing, when RhizoTubes were disinfected in sodium hypochlorite solution and checked for P. fluorescens C7R12 persistence, no bacterial colonies were detected on agar plates after cultures, indicating that the bacterial strain did not persist in the RhizoTubes after disinfection.

RhizoTubes are suitable for plant growth; root architectural traits of plants growing in RhizoTubes or pots are similar

RhizoTubes allowed similar plant growth to that observed in pots, for all the plant species studied. Thus, RhizoTubes are suitable tools for studying the root systems of various species. Coupled with a high-definition camera, movable in 2D, we had high-resolution access to the root traits such as root and nodule diameters, emerging nodules and hyphae.

In our study, pea plants allocated similar amounts of their biomass to roots in RhizoTubes and in pots (Table 2). However, pea main root length was longer in RhizoTubes as previously reported for growth in hydroponics (or growth pouches) versus pots [22]. Presumably these results from the lower mechanical resistance in RhizoTubes compared to pots where roots encounter substrate particles. The depth of the RhizoTube (50 cm) is attained after around 1.5 month’s growth, depending on species and growing conditions. As such, root phenotyping during the first month of growth generally allows undisturbed root growth and architecture with easy visualization of individual roots. Later on, roots generally start to entangle and root architecture starts to be constrained by reaching the bottom of the RhizoTube. As such, the deeper environment resulting from growth in RhizoTube might explain the higher length of main root that was observed (Table 2). Similarly a shallower and plateauing distribution profile of laterals on main roots and then fewer, smaller, lateral roots were observed at the bottom of RhizoTubes (Fig. 4). In accordance with the above hypothesis, the number of secondary lateral roots in pots displayed a sharp peak at around 20 cm from the main root base, a distance which corresponds exactly to the pot depth (Fig. 4c, d).

RhizoTubes show similar phenotypic variations between genotypes than those observed in pots

In absence of mineral nitrogen, legumes carry out endosymbiotic nitrogen fixation with rhizobial bacteria, within a specialized structure known as the nodule, which increases the complexity of root system analyses. Nodule biomass and number determine plant nitrogen status [56, 57] and constitute key phenotypic traits of nodulated legume roots.

Similar nodule distribution was observed for pea plants cultivated in RhizoTubes or pots, especially for nodules borne by laterals, which are by far the more numerous. Comparing genotypes, a similar plant biomass and similar biomass partitioning between shoots and roots were observed for Cameor and Kayanne genotypes grown either in pots or in RhizoTubes (Table 3).

RhizoTubes allowed us to rank both oilseed rape and Vulpia species as a function of the response of their root biomasses to soil nitrogen with a ranking similar to that in pots. This demonstrates that RhizoTubes allow not only to compare plant genotype/species in similar environmental conditions but also to compare their ranking in their responses to environmental factors. Although the response of plant leaf area to soil mineral N observed in RhizoTubes was lower than that observed in pots, the ranking of oilseed rape and Vulpia plant species appeared similar. RhizoTubes were suitable tools for highlighting the more nitrophilic response of oilseed rape as compared to Vulpia already demonstrated by changes in leaf area [58].

RhizoTubes are suitable for studying water stress effects on plants, which exhibit similar root responses to those grown in pots

RhizoTubes allowed us to study shoot, root and nodule growth responses to drought in the same ranges as those observed in pots. This applied both to plants growing in a large volume of soil in pots and in the 2D environmental conditions of RhizoTubes without direct contact of roots with soil. Nodule biomass allocation was repressed for both pea and Medicago under water stress conditions as compared to well-watered conditions (Table 4). Whereas drought impact on biomass accumulation was similar for Medicago in RhizoTubes as compared to pots, pea shoot and root biomasses were more reduced when grown in RhizoTubes. This may arise from the difficulty of applying similar water deficit regimes in pots and RhizoTubes. Although we tried to optimize irrigation for plants in RhizoTubes (through sequential addition of nutrient solutions) and in pots (nutrient solution supplies being gravimetricaly controlled), possibly water supplies of plants in RhizoTubes and in pots did not match perfectly so that to lead to a similar level of water stress sensed by plants. This highlights the need to accurately adjust nutrient solution supplies in frequency and amount among containers to get similar plant responses to drought. Despite the tuning required, RhizoTubes are well suited for drought stress response studies on legume plants.

RhizoTubes are suitable systems for characterizing plant–microorganism interactions

Legume plants grown with ample nitrogen supply in the rooting medium do not establish nodulation with rhizobia [14, 57, 59], as consistently observed here both in pots and RhizoTubes. However in absence of soil mineral nitrogen, as demonstrated above, RhizoTubes allow legume-rhizobia symbiosis to be established, as in pots. This is made possible as rhizobia cross the membrane, which in turn also demonstrates that early plant–bacteria communication operates successfully via rhizodeposit and nod factor exchanges through the RhizoTube membrane between plant and bacteria. The very slight shift of nodule distribution profile towards the lower part of the main root in RhizoTubes might be explained either by mechanical constraints resulting from the lower root volume and higher substrate resistance of soil in pots than in the RhizoTube.

The production of mycorrhiza-colonized grapevine plants was possible in RhizoTubes with a total mycorrhization level and a root age-dependency similar to that encountered in pots. These observations are in accordance with previous results obtained for grapevines grown in pots, with the same inoculum, showing the reliability of the RhizoTube for mycorrhiza studies (unpublished data). Interestingly, the visualization of the mycelium was possible at 3600 ppi (Fig. 3), illustrating the high resolution of imaging, even if image-based quantification was not possible (as the mycelium is in part hidden by roots).

In microbial ecology, assessing the impact of microbial strains, populations or communities on root development remains a challenge. The strategy usually applied consists of inoculating plants with different microbial strains and then collecting, analyzing and comparing root architectures at different developmental stages [60, 61]. However, these experiments are not easy to manage in term of numbers of replicates for each treatment and sampling date and are usually time consuming. In order to determine whether free-living microorganisms could persist in RhizoTubes, the bacterial strain P. fluorescens C7R12 was inoculated on pea and wheat roots. The bacterial strain persisted in the root systems of plants in RhizoTubes, with densities of P. fluorescens C7R12 comparable to that observed in other systems [60]. Moreover, the strain was completely removed after sodium hypochlorite disinfection, allowing the subsequent re-use of RhizoTubes for plant–microbe studies. Taken together, these results showed that RhizoTubes are suitable for inoculating both symbiotic (rhizobia and arbuscular mycorrhizal fungi) and free-living microorganisms on plant roots. RhizoTubes can therefore be used to measure kinetically and non-invasively the interactions between plant roots and microorganisms.

RhizoTube and RhizoCab pros and cons

Growing plants in RhizoTubes did not change the extent of the plant’s response to the environmental treatments tested here. The suite of tools presented here is powerful not only for measuring phenotypic traits but also for ranking genotypes for a given response to environmental factors.

The study of RSA traits such as lateral root number and length is complicated by the inaccessibility of the soil matrix. RhizoTubes allow visual access to the entire root system. Another advantage is that the cylindrical configuration yields a surface area of about 0.5 m², which for flat rhizotrons would be problematic during movement of plants on conveyors to imaging cabins.

Our system is also highly flexible as it allows working with a range of substrates, because plant roots never enter in direct contact with the substrate. However thanks to the RhizoTube membrane, the substrate can still have its role of hosting microbes which can receive their share of rhizodeposits while at the same time plants receive their nutrient solution and establish symbioses with introduced bacteria or fungi. RhizoTubes are however not compatible with all substrates. The substrates used in RhizoTubes should be dense enough to exert a uniform pressure on the membrane so that roots grow tightly between the membrane and the transparent outer tube. This allows creating a mechanical constraint on roots, which can be modulated when substrate is filled in the RhizoTube. Substrates such as peat do not allow sufficient pressure on membrane, due to their density and cannot be used in RhizoTubes. In order to avoid substrate particles masking roots and so perturbing image analysis, the RhizoTube substrate has not to cross the membrane, which has so been designed with a low porosity. However, the membrane porosity cannot be lowered too much, as it would be detrimental to water and nutrient exchanges with the plant roots. As such, filling RhizoTubes with natural soil only is not recommended. However, in case of studies concerning between plant and soil microbial community interactions, natural soil hosting these microbial communities can be mixed with sand. Substrates such as sand, perlite/sand, sable/billes d’argile, atapulgite/billes d’argile mixtures have been successfully used. These allow water and nutrient solution supply management, for a given water retention capacity. In addition there is no need for tedious root separation from the substrate during harvest and even the thinnest roots are recovered. The membrane allows having soil-free root material, a disadvantage when plants are cultivated in pots. Lastly, while inevitably variable amounts of root parts, depending on both species and their age, get lost during the root harvest and washing process in pots, RhizoTubes in contrast allows the recovery of the entire roots giving unbiased plant root biomass comparison during the various experiments.

One of the major difficulties when plants grow deep in pots or in RhizoTubes is overlapping roots, which renders individual root identification tedious in branched or old root systems. Although analyzing root systems of mature plants in our systems remains a bottleneck, a reasonably long experimental duration can be assured for a given species in our system, which avoids (or reduces) the need for untangling roots harvested from soil pots to organize them into a 2D conformation, and precludes root damage.

However, RhizoTubes can be used for an extended period of time beyond 10–20 days depending on the speed of root development which is variable both for a given species according to environmental conditions and between plant species. This limitation mostly arises from the RhizoTube dimensions that will be increased in future development up to 1 m height and 0.35 m diameter, for other platforms having specific needs of either increased duration of plant root observation or devoted to work with species having much larger root systems. Their use implies that the platform’s specifications can handle them, with sufficient phenotyping cabin height (i.e. higher than the 1.6 m of our 4PMI phenotyping cabins) for plant entry, sufficient depth of view for zenithal images and powerfull motors to tackle the higher RhizoTubes weight.

Our equipment allow to dynamically following growth processes (root elongation, increase in nodule biomass, nodulation waves etc.) of a given plant throughout a large span of its growth cycle (an example is given in Fig. 7 for root and nodule growth dynamic). This decreases the inter-plant heterogeneity resulting from destructive observations on different plant samples across time. It also reduces the number of replicates and the overall cost of experiments.

Lastly, RhizoTubes can be moved on conveyors thanks to their specially designed base. This allows their implementation in high throughput platforms using the “plant to sensor” concept, comprising automatic watering and imaging, completing the shoot imaging usually deployed on such platforms.

Image quality and analysis

The resolution of the images (from 42 µm per pixel at 600 ppi down to 7 µm per pixel at 3600 ppi) allows detection of the thinnest roots and nodules (Fig. 2c, e) and to see hyphae (Fig. 3). Image acquisition is very easy and fast in our system but analysing them may be rate limiting. In this study root phenotypic traits were manually determined. However powerful algorithms under development will allow an automatic extraction of a variety of nodule and root morphometric traits from the images obtained with RhizoCab.

Digital imaging used in 4PMI for shoot and root phenotyping allows extracting from individual images non-invasively and throughout a large spanning of plant lifecycle a number of phenotypic traits. Image acquisition, the first step in a high-throughput phenotyping work flow, is easily performed using RhizoTubes and the associated imaging cabin RhizoCabs for root traits while shoot images are acquired sequentially for both plants growing in pots or in RhizoTubes in the other imaging cabins of 4PMI. Every shoot measurement made on a pot, can be ported on a RhizoTube because it can be simply considered as a 50 cm high pot. Subsequently, a binary image of the RGB original image or a single wavelength is used to improve image segmentation. Both plant shoot support frames and RhizoTube membranes are colored in blue. This greatly facilitates image processing and segmentation. Focusing here on root images, while a number of softwares can already be used to extract phenotypic traits (http://www.plant-image-analysis.org/), specialized image analysis software is still necessary for producing an automatic, rapid extraction of root phenotypic traits of interest for the variety of plant root systems that may be encountered in research [41]. Towards this challenge, image analysis algorithms are under development in our group in collaboration with scientists possessing the necessary programming skills [62]. With this dedicated software (Han S., unpublished) numerous morphological parameters will be automatically and dynamically extracted and quantified from root images such as projected area of root and nodules, number, length, diameters and positions of the various roots and nodules on roots.