Evaluating waterlogging stress response and recovery in barley (Hordeum vulgare L.): an image-based phenotyping approach

Abstract

Waterlogging is expected to become a more prominent yield restricting stress for barley as rainfall frequency is increasing in many regions due to climate change. The duration of waterlogging events in the field is highly variable throughout the season, and this variation is also observed in experimental waterlogging studies. Such variety of protocols make intricate physiological responses challenging to assess and quantify. To assess barley waterlogging tolerance in controlled conditions, we present an optimal duration and setup of simulated waterlogging stress using image-based phenotyping. Six protocols durations, 5, 10, and 14 days of stress with and without seven days of recovery, were tested. To quantify the physiological effects of waterlogging on growth and greenness, we used top down and side view RGB (Red-Green-Blue) images. These images were taken daily throughout each of the protocols using the PSI PlantScreen™ imaging platform. Two genotypes of two-row spring barley, grown in glasshouse conditions, were subjected to each of the six protocols, with stress being imposed at the three-leaf stage. Shoot biomass and root imaging data were analysed to determine the optimal stress protocol duration, as well as to quantify the growth and morphometric changes of barley in response to waterlogging stress. Our time-series results show a significant growth reduction and alteration of greenness, allowing us to determine an optimal protocol duration of 14 days of stress and seven days of recovery for controlled conditions. Moreover, to confirm the reproducibility of this protocol, we conducted the same experiment in a different facility equipped with RGB and chlorophyll fluorescence imaging sensors. Our results demonstrate that the selected protocol enables the assessment of genotypic differences, which allow us to further determine tolerance responses in a glasshouse environment. Altogether, this work presents a new and reproducible image-based protocol to assess early stage waterlogging tolerance, empowering a precise quantification of waterlogging stress relevant markers such as greenness, Fv/Fm and growth rates.

Introduction

Climate change is set to drastically alter the environmental conditions of most regions, and water availability is a key factor that will be disturbed. Climate models predict that many areas will face water shortages and droughts while other areas will see increased precipitation and an increased frequency of flooding [1, 2]. Waterlogging causes major crop yield reductions with the magnitude of reduction varying by crop [3, 4], management practices [5], genetic diversity, [6] and stress conditions [7, 8]. This rise in stress conditions comes at a time when demand is growing due to increasing population sizes. Further, this problem is compounded by the necessity to limit agricultural land use and agricultural inputs [9, 10]. In many areas, crops are expected to be exposed to more extreme and frequent abiotic stress over the coming decades and so improving the ability of crops to maintain yield under harsher environmental conditions is a critical step to promote food security [11].

Waterlogging describes the flooding of the root zone with an excess of water while the above ground biomass remains above the water level [12]. Waterlogging in the field creates intricate heterogeneous soil environments resulting in hypoxic stress that greatly reduces yield [13,14,15]. The ability of plants to respond to waterlogging stress is altered by compounding factors such as the growth stage at which the stress occurs [7, 16] temperature [17], and duration of stress [18, 19]. Screening for waterlogging tolerance presents many challenges with several methods and approaches being utilised by researchers [8]. This task has proven to be a slow and difficult undertaking. In controlled environments, such as glasshouses, there is very limited uniformity amongst experimental waterlogging setups across multiple species [8]. A meta-analysis by Tian et al. (2021) explored the differences across experimental waterlogging regimes highlighting the range of waterlogging durations used by researchers across different species [4]. Because waterlogging stress primarily impacts the root system, and as such, root adaptations are key [20,21,22,23], several studies have focused on root phenotyping to screen for waterlogging tolerance. In particular, aerenchyma formation is a hallmark of waterlogging adaptation as pores form within roots allowing improved gas diffusion. Root porosity can be used as a proxy measurement of aerenchyma development in smaller scale experiments.

Barley (Hordeum Vulgare L.) is the fourth most abundant crop across the globe and is grown across a wide range of environments in both high intensity and low input production. Robust germplasm collections and a wide environmental range positions barley as a good candidate for the development of more climate resilient crops [24, 25]. However, despite this large geographical range and climate-resilient status, barley is considered to be more sensitive to waterlogging stress than other cereal crops [3, 7].

When studying barley responses to waterlogging, researchers have employed varying durations of waterlogging stress as well as including or excluding recovery periods [26, 27], limiting comparison between studies and their results’ interpretation. For example, de San Celedonio et al. (2014) observed that barley yield penalties were higher when waterlogging was imposed at Leaf seven appearance on the main stem to anthesis, whereas Masoni et al. (2016) noted that waterlogging for up to 16 days did not affect barley yields [7, 18]. Moreover, reproducibility of waterlogging tolerance screening carried out in the field has proven difficult as cultivar tolerance scores in a given year often show little correlation to the scores of subsequent years [28]. In glasshouse conditions, where yield reduction cannot be used to determine waterlogging tolerance, visual scoring has been used as a proxy to assess waterlogging tolerance Other screening approaches include leaf damage, chlorosis, and/or survival, which has been widely used to rank barley varietal tolerance [29,30,31]. Other studies have focused on aerenchyma formation, which can be imaged and scored directly via microscopy, though this is a destructive and low throughput process [28, 32]. More recently, the use of high-throughput phenotyping with non-destructive imaging has increased the resolution of data collected to assess abiotic stresses [33,34,35]. Nevertheless, to our knowledge, a high throughput image-based phenotyping approach has never been used to study barley under waterlogging stress. The use of image-based phenotyping has the potential to provide high-resolution early screening of waterlogging tolerance in a short time-period and give a detailed account of the stress response over time [36].

Here we aim to identify an optimal waterlogging protocol to compare stress responses of spring barley genotypes using high-throughput phenotyping. To determine the ideal protocol, we used an image-based phenotyping approach combined with destructive harvest data. We also imaged the root architecture at harvest to compare the tested protocols. The combination of this range of phenotyping methods and tools were then used to determine the efficacy of early vegetative stage screening, and to determine an optimal waterlogging protocol in terms of physiological responses, resource intensity, and reproducibility. To examine the reproducibility, the selected protocol was validated in a separate facility using a fully automated phenotyping platform. All in all, we present a defined waterlogging protocol in spring barley at seedling stage with a clear physiological response and demonstrated reproducibility across environments.

Materials and methods

Plant material

Four accessions of 2-row spring barley were selected for protocol optimization and validation. Two established commercial malting cultivars, namely RGT Planet and Concerto were used, as well as two heritage accessions, namely Glasnevin no.5 (bred in Ireland and released in 1946) and Golden Promise (bred in the UK through gamma-ray treatment of the cultivar Maythorpe and released in 1968). Glasnevin no. 5 seeds were obtained from the Department of Agriculture Food and the Marine (Backweston Laboratory, Co. Kildare Ireland) whereas RGT Planet, Concerto, and Golden Promise were sourced from the James Hutton Institute (Scotland, UK). Barley accessions were chosen for the following reasons: Glasnevin No.5, representing a broader heritage collection; Golden Promise, the leading old cultivar in Europe; RGT Planet, the top cultivated elite cultivar in Europe in the 2020s; and Concerto the top cultivated elite cultivar in Europe in the 2010s. Each of the accessions used had shown good germination rates. Experiment 1 took place at University College Dublin Rosemount Environmental Station glasshouses (Co. Dublin, Ireland) whereas Experiment 2 took place at the plant phenotyping facilities of the University of Picardie Jules Verne (CRRBM, Amiens, France).

Experiment 1-university college Dublin

Growth conditions

Seeds of RGT Planet and Glasnevin no.5 were germinated on filter paper in Petri dishes before being transferred into two litre pots (KPG, 17 cm x 12.8 cm Kreuwel Plastics, Netherlands). Pots were filled to the fill line with 1.36 kg of a mixture of nutrient topsoil (Westland Horticulture, UK) and sand in a 7:3 by volume. Plants were grown in glasshouse conditions in two batches; October/November 2020 and February/March 2021, with a total of six optimisation protocols tested (Fig.1). Day night temperatures were 18 °C/18°C for protocols 3,4,5 and 6 with an average relative humidity of 68% and 19 °C/17°C for protocol 1 and 2 with an average relative humidity of 54%. Five hours of supplemental light was provided to reach ~ 14 H/10H day/night cycle.

Fig. 1

Schematic summary of the six experimental protocol durations performed in Experiment 1. For each protocol (1-6), five samples for each genotype (RGT Planet and Glasnevin no.5) were exposed to the displayed regimes of waterlogging (orange) and subsequent recovery (green). An identical number of control samples were matched for size and maintained under well-water conditions using split-plot experimental design. Both groups were imaged daily before destructive harvest.

Stress imposition

Six protocols consisting of varying stress and recovery durations were assessed (1). The duration of the stress protocols was based on the meta-analysis by Tian et al., (2021), where waterlogging duration experimental protocols ranged from 3 to 15 days [4]. The specific number of stress days was selected to accommodate the imaging logistics. Waterlogging stress was imposed at the 3rd leaf stage (Zadocks stage 13) by placing the pots within a second pot with drainage holes sealed while control pots were placed in a second pot with unsealed drainage [37]. Waterlogging was determined as partial submergence with the water level above 1 cm above the soil surface, corresponding to approximately 120% of field capacity, whereas control pots were well-watered at approximately 60% of field capacity. The levels of water were maintained throughout the experiment depending on protocols. To drain the pots for testing recovery, the outside pot was removed from both control and waterlogged pots on the final day of stress, corresponding to each tested protocol.

Samples were separated by each tested optimisation protocol with each block containing five biological replicates per treatment group (i.e., RGT Planet Control, RGT Planet Waterlogged, Glasnevin no.5 Control, Glasnevin no.5 Waterlogged) giving a total of 20 samples per protocol. Samples were randomized using a split-plot design, and again randomized after imaging each day.

Image-based phenotyping

Using a split-plot design, sample pairs (control and treatment) were paired by size and imaged in a single tray one day prior to stress imposition and imaged daily until harvest using the PlantScreen™ System (Photon System Instruments, Drásov, spol. s r.o., Czech Republic). Red-Green-Blue (RGB) images were obtained with one top-down camera (Fig. 2A) and one-side facing (Fig. 2B) (GigE PSI RGB – 12.36 Megapixels Camera with 1.1” CMOS Sensor). Projected shoot area (PSA) was calculated using the sum of segmented pixels from side and top view images, and plants were imaged at a height of 500 mm above the tray. Image segmentation was obtained using the PlantScreen™ Analyzer software (PSI, Drásov, spol. s r.o., Czech Republic). Samples were staked using plant holders and blue non-slip material was cut to shape and placed above the soil to aid image segmentation.

Fig. 2

Imaging setup for Experiment 1 and Experiment 2 to assess spring barley responses to waterlogging. Top (A) and Side view (B) of tray configuration for Experiment 1. RGT Planet imaged at four days after stress imposition using the PlantScreen™ System (Photon Systems Instruments, Drásov, spol. s r.o., Czech Republic) at University College Dublin. Green line delineates the segmentation area assigned to each sample. Support stakes were used to reduce overlap of sample into adjacent area. (C) Top and (D,(E) two side-views (D = 0°, E = 90°) for Experiment 2 in the PlantScreen™ System (Photon Systems Instruments, Drásov, spol. s r.o., Czech Republic) at University of Picardie Jules Verne tray configuration. Green line delineates the segmentation area assigned to each image taken in the RGB chamber. Example image of control sample on day 14 of protocol. Plants were returned to a fixed position that was initially assigned for each tray inside the growth chamber, after each round of watering and image acquisition

Root imaging

At the end of each protocol in Experiment 1, roots were washed by hand with the length and dry mass of the roots being recorded after imaging. Roots were scanned using an EPSON flatbed scanner V800/V850 (Epson, Suwa, Japan) and root parameters, including length, volume, diameter, and surface area were obtained using RhizoVision Explorer software [38].

Destructive phenotyping

On the final imaging day of each protocol for Experiment 1 at University College Dublin, length, fresh and dry mass for shoots were recorded. Dry mass was recorded after incubation at 70 °C for 48 h in an incubator (Model 1 H-150, Gallenkamp, UK), and weights were recorded using a scale (Nimbus Precision Balances: NBL 2062e, Adam Equipment, UK).

Experiment 2-University of Picardie Jules Verne

Growth conditions

To test the reproducibility of the selected protocol as defined on Experiment 1, protocol 6 was selected and validated in Experiment 2. Seeds of RGT Planet, Glasnevin no.5, Concerto and Golden Promise were used in Experiment 2 to replicate protocol 6. RGT Planet and Glasnevin no.5 seeds were from the same seed batch as in Experiment 1 and germinated on Whatman paper before being transferred into a three litre double-pot design. Drilled pots (RP 3 L, Ippoland, Poland) were inserted in undrilled pots (KR-3 L, 15 cm, Kreuwel Plastics, Netherlands) prior to being filled with 1.22 kg of a nutrient topsoil (NF U 44–551, Botanic, France). Plants were grown in a PlantScreen™ Modular Facility during June/July of 2021 where controlled conditions were assessed by environmental sensors. A day night photoperiod of 16 H/8H and temperatures of 19 °C/15°C with an average relative humidity of 60% were set up. Blue mats, with a central hole in them to allow the shoot to grow unimpeded, were used to cover the soil of each pot (147 mm of diameter, Photon Systems Instruments, Drásov, spol. s r.o., Czech Republic). This was done to improve image processing. Plant holders (for 3 L-pots, 305 mm Height, Photon Systems Instruments, Drásov, spol. s r.o., Czech Republic) were also used to support the leaves of each plant while also ensuring that no conveyor belt related issues occurred while scanning.

Stress imposition

Waterlogging stress was applied at 3rd leaf stage as described in the ‘Experiment 1’ section. Water level was maintained about 1 cm above the soil surface, corresponding to approximately 140% of field capacity, whereas control pots were well-watered at approximately 100% of field capacity. After a duration of 14 days of stress, the outside undrilled pot was removed from both control and waterlogged pots so that the excess of water could be drained. This was done in order to perform the seven-day recovery phase of the protocol and to reach 100% saturation of the soil for all pots. The different watering regimes were monitored daily at 8:30 am by the high-precision irrigation system of the PlantScreen™, which is fully automated at UPJV and allows the control of watering based on the maintenance of a specific reference weight. A total of 80 samples, consisting of ten biological replicates per treatment group (RGT Planet Control, RGT Planet Waterlogged, Glasnevin no.5 Control, Glasnevin no.5 Waterlogged, Concerto Control, Concerto Waterlogged, Golden Promise Control, and Golden Promise Waterlogged), were randomized within the growth area where each waterlogged plant alternated with its corresponding control.

Destructive phenotyping

Shoot dry mass was quantified on the final day of imaging. Dry mass was recorded after samples had been dried in an incubator (ULM500 Memmert) at 70 °C for 48 h and weighed on a scale (Precision Balance SBA52 Scaltec).

Image-based phenotyping

A first round of daily imaging was done during daytime, at 11:00 am, for Red-Green-Blue (RGB) images. Plant images were obtained with two-side facing (RGB1: 0 and 90° - GigE µEye UI-5580SE-C/M − 5 Megapixels QSXGA Camera with 1/2” CMOS Sensor) and one top-down camera (RGB2: GigE µEye UI-5580SE-C/M − 5 Megapixels QSXGA Camera with 1/2” CMOS Sensor), at a height of 500 mm above each tray of a single central position (Fig. 2C to E).

All plants were imaged daily, with imaging starting one day prior to stress imposition, using the PlantScreen™ Modular System (Photon System Instruments, Drásov, spol. s r.o., Czech Republic). After identifying 10 specific hues of barley greenness, using the MorphoAnalysis GUI software (v1.0.9.5, PSI, Drásov, Czech Republic), the greenness of the test plants was assessed as the mean of the RGB1 two-side view images. The relative and absolute values of the 10 hues was analysed for each genotype on each day to track changes of greenness in response to waterlogging treatment. The PSA was calculated as the sum of segmented pixels from two side angles (0 and 90 degrees) and the top view images. The relative growth rate (RGR) (day ^− 1) was calculated using the equation from Ball et al., (2020): [39]

$$\:RGR\:\left(t1,t2\right)=\text{log}\frac{PSA\:t2}{PSA\:t1},$$

where t₁and t₂ are the days after stress imposition defining an interval.

In the second round of daily imaging, chlorophyll Fluorescence images were taken during night-time at 4:00 am to ensure their full dark-adaptation status. This was achieved using the Kinetic Chlorophyll Fluorescence Imaging Unit of the PlantScreen™ Modular System (1.4 Mega-pixel High quality CCD sensor, 35 × 35 cm) and using an Fv/Fm protocol with a F0 duration of 2s followed by a saturating pulse of 800ms set at 20% (~ 1137µE of intensity). Image segmentation was obtained using the PlantScreen™ Analyzer software (v3.3.10.7, PSI, Drásov, Czech Republic). Dark adapted samples were measured to obtain a minimum value (F0) before light saturation allowed the maximum fluorescence to be obtained (Fm). Fv was calculated as the difference between Fm and F0.

Statistical analyses

All data preprocessing and analysis from both Experiment 1 and Experiment 2 was conducted using the R programming language (R Core Team, 2022). To assess the impact of Genotype and Waterlogging Treatment, two-way ANOVA was performed. Data was visualised using R package ggplot2 [40]. Significant changes in colour hue proportion were determined using ANOVA for each hue on each day. Beta regression was used to determine significant differences between relative hue pixel counts.

Results

Selected protocol integrates prolonged stress and recovery phase

In this study, we sought to determine an optimal waterlogging stress protocol in spring barley that enables us to detect genotypic differences using image-based phenotyping. Because the ability of stressed plants to recover normal development is fundamental, we assessed barley responses during the stress phase but also during the following recovery period. In Experiment 1, we assessed two genotypes (Glasnevin no.5 and RGT Planet) under three different durations of waterlogging stress (5, 10, and 14 days) with and without a seven-day recovery period for a total of six protocols. To establish the most effective waterlogging stress protocol, we determined the time-period where biomass had a notable decline when comparing treatment and control groups, but where it was possible to distinguish the biomass responses of the two genotypes.

To evaluate the effect of the stress non-destructively, the PSA was calculated every 24 h as the sum of pixels segmented from top and side view RGB images [41]. Plant material was segmented from the background from a side and top view images using the PSI Morpho analyser software using the colour formula (4*G-3*B-R). The daily average PSA for each treatment and genotype across each of the six protocols is shown in Fig. 3. Plants exposed to waterlogging exhibited slower accumulation of biomass compared to those in control conditions in all six protocols tested. However, we observed that the difference becomes more pronounced as the duration of stress increases and continues to persist during the subsequent recovery period. Uniformity across overlapping durations between protocols was limited by the low number of biological replicates (five per treatment group).

Fig. 3

Average daily pixel count of five replicates of Glasnevin no.5 and RGT Planet Genotypes under waterlogging or control conditions. Orange area represents stress duration, green area denotes recovery period. Pixel count calculated as the sum of segmented pixels from side and top view. (A-F): protocols 1–6. Mean and standard error of each treatment displayed by day

We observed that protocols 1 through 4 displayed no significant changes in the shoot dry mass upon harvest on the final day of each protocol (Fig. 4A-D), with only a longer time of stress imposition, i.e., 14 days (protocol 5 and 6), showing a significant change between control and waterlogged plants (Fig. 4E-F). Although protocol 5 shows significant effect of treatment on the shoot dry mass (p = 0.03) (Fig. 4E), the addition of a recovery period enabled to observe differences in the recovering ability of different genotypes. The longest protocol (Protocol 6), which included 14 days of stress plus seven days of recovery, produced the most significant change in harvested biomass by treatment (p = 0.003) and was selected as the protocol for subsequent experiments (Fig. 4F). A high correlation (r² = 0.96) was observed between the PSA and harvest biomass (Additional file 1: Figure S1). In Experiment 1, we observed that the two-pot system caused an overlap of biomass between the imaged plants at later growth stages. To overcome this challenge, in Experiment 2, we used one pot per tray and increased the number of biological replicates to ten while expanding the number of genotypes to four.

Fig. 4

Boxplot of shoot dry mass (g) of barley assessed in the six tested protocols. Above ground biomass was harvested and oven dried at 70 °C for 48 h. A-F protocols 1–6. Whiskers indicate minimum and maximum values in the first and fourth quartile. The box boundaries indicate the upper (25th percentile) and the lower (75th percentile) quartiles. The bold line in the box represents the median

To examine the effect of the recovery period in the waterlogging response, we analysed the roots of plants in Protocol 5 (14 days of stress) and Protocol 6 (14 days of stress plus seven-day recovery) in Experiment 1. Root samples from both protocols were washed and imaged upon harvest and a range of parameters were obtained from these images. In Protocol 5, roots were classified into diameter groups. We identified significant shifts in the distribution of total root volume associated with diameter size classes between genotype and treatment. To investigate the effect on root size, we distributed the total root volume into six ranges of root diameter (0–0.1 mm, 0.1–0.25 mm, 0.25–0.5 mm, 0.5–0.75 mm, 0.75–1 mm, >1 mm). Waterlogging significantly increased root volume associated with a root diameter of 0.75–1 mm (p = 0.0003); however, no significant changes were observed in any diameter class in protocol 6 after a period of seven days recovery (Additional file 1: Figure S2).

Waterlogging reduces barley biomass and induces photosynthetic and greenness changes

Based on the results of Experiment 1, we selected Protocol 6 (14 days waterlogging followed by seven days recovery) for further analysis as shoot biomass was most reduced in waterlogged samples (shoot dry mass of waterlogged samples was 54% of the control ones) (Fig. 4). To assess the reproducibility of the protocol, we doubled the number of genotypes to four and increased the number of biological replicates to ten. In addition, to facilitate image analysis, the imaging setup was changed to a one pot per tray system. (Fig. 2). The selected protocol of 14 days of stress and seven days of recovery was adapted for high-throughput phenotyping with the key changes between imaging setup (Fig. 2) and media composition as previously detailed. Furthermore, we performed an initial assay using different soil types to evaluate potential changes in response between samples and ensure the reproducibility of the Experiments. Our results demonstrated that plants behaved in the same way in response to waterlogging when grown in 100% nutrient soil in Experiment 2 in comparison to the 70:30 soil to sand ratio used in Experiment 1 (Additional file 1: Figure S3).

We observed a larger effect on biomass accumulation in Experiment 2 with varying responses across genotypes when comparing to Experiment 1 (Fig. 5A). Golden Promise showed a noticeable lower biomass than the other genotypes under control conditions with an average shoot dry mass of 6 g compared to RGT Planet (7.5 g), Glasnevin no.5 (7.4 g) and Concerto (8.8 g). Such lower biomass accumulation in Golden Promise was not mirrored in the waterlogged samples, as evidenced by an average reduction of shoot dry mass of 75% compared to its control counterparts. In comparison, RGT Planet exhibited a dry mass reduction of 73%, followed by Concerto at 80%, and Glasnevin no.5 at 81%. To further demonstrate the reproducibility of the imaging protocol, we compared the plant size in both Experiments and observed a coefficient of 0.86 using the Pearson correlation between the PSA values for Experiment 1 and 2, despite the larger plant size observed in Experiment 2. The PSA data in Experiment 2 shows a more apparent reduction in biomass as validated by PSA data in waterlogged samples with PSA falling below 50% of control for each genotype by day 14 of stress (Fig. 5A).

Fig. 5

Waterlogging effects on growth and photosynthetic performance in spring barley plants. A: Average daily pixel count of ten biological replicates of four genotypes (Concerto, Glasnevin no.5, Golden Promise and RGT Planet) under waterlogging or control conditions. Orange area represents stress duration, green area denotes recovery period. Pixel count calculated as the sum of segmented pixels from two sides and one top-view. Mean and standard error of each treatment displayed by day. Images not taken on day 6 and 14, data was extrapolated for these days. B: Average Fv/Fm value of ten replicates of four genotypes (Concerto, Glasnevin no.5, Golden Promise and RGT Planet) under waterlogging or control conditions. Orange area represents stress duration, green area denotes recovery period

A key aspect of the stress response is the ability to restore a regular growth rate once the stress has subsided. The included recovery period of seven days allowed observation of recuperating relative growth rates after stress. We calculated the RGR of samples using the equation according to Ball et al. [39], and observed that the RGR for each group of genotype x treatment peaked between days 10–12, with higher growth rates under control conditions. Concerto control exhibited a RGR of 1.6 between this period whereas their counterpart waterlogged samples recorded an average relative growth rate of 0.79. Such RGR differences were comparable in each genotype, with Glasnevin no.5 exhibiting a RGR of 1.4 in control vs. 0.73 in waterlogged, RGT Planet (1.3 to 0.76), and Golden Promise (0.97 to 0.73). The lowest RGR value for each genotype under waterlogging conditions was observed between day 12 to 14, with waterlogged samples averaging 0.1071 in the final days of the waterlogging treatment. In the recovery phase, we observed a gradual increase in RGR during days 14 to 18 (0.2287) and finally between days 18–21 (0.2672) (Additional file 1: Figure S4).

To investigate if we could detect this apparent stress recovery prior to the effects seen in the biomass, we investigated photosynthetic health using chlorophyll fluorescence imaging. We observed a higher variability in Fv/Fm values at the beginning of the experiment compared to later time-points, which is attributed to smaller segmented images (i.e., smaller plant size) from which the Fv/Fm measurement is calculated. Once the plants increased their size, the Fv/Fm readings become more stable between days. Because of this highly variable data, we excluded the first five days of chlorophyll fluorescence imaging. From day six onwards, we observed that the average Fv/Fm values in control conditions remained relatively stable for each genotype, ranging from 0.845 to 0.855 across the experimental period for Concerto and 0.848 to 0.859 in Glasnevin no.5. Slightly larger variation was observed for Golden Promise; 0.835–0.853 and RGT Planet, 0.838–0.855 (Fig. 5B). Under waterlogged conditions, the average Fv/Fm values was lower during the stress period (day 6 to 14) with each genotype registering their lowest Fv/Fm value during this period. Notably, on day 8, RGT Planet and Golden Promise recorded their lowest Fv/Fm values at 0.782 and 0.830, respectively. Concerto showed its lowest value on day 9 at 0.777, while Glasnevin no.5 experienced its lowest value on day 10 at 0.785. Interestingly, Fv/Fm values in waterlogged samples of Concerto, Glasnevin no.5, and RGT Planet followed a similar trend as the Fv/Fm values began to increase from day 10 while the samples were still waterlogged, whereas Golden Promise had a lower reduction of Fv/Fm values between control and waterlogged condition throughout the experiment (Fig. 5B).

To further assess physiological changes of barley plants exposed to waterlogging, we analysed changes in greenness. The proportion of ten colour hues prominent in the segmented images was analysed across all genotypes and treatments for Experiment 2 (Fig. 6). Using both absolute and relative pixel counts, the selected hues were categorised into three trends over the course of the waterlogging stress: constant, stress-induced reduction and stress-induced increase.

Fig. 6

Changes in greenness hue abundance in response to waterlogging differ by genotype and through time. (A:D) Concerto, Glasnevin no.5, Golden Promise and RGT Planet, respectively, under control conditions. (E:H) Concerto, Glasnevin no.5, Golden Promise and RGT Planet, respectively, under stress conditions. Red line indicates the end of stress and the beginning of the recovery period. RGB (red-green-blue) and hex values for each hue is given in the figure. Hue values represent the proportion of that specific hue of the entire segmented area. Values shown represent the averages calculated for the ten biological replicates per accession and treatment

Following these observable trends, we noticed that hues 5 (Additional file 1: Figure S9) and 6 (Additional file 1: Figure S10) showed a constant trend in pixel counts for both control (Fig. 6A) and waterlogged plants (Fig. 6B). As anticipated, waterlogged plants showed slower growth due to stress, this resulted in lower absolute pixel values than in control samples. After four days, control plants showed significantly higher values of hues 5 and 6 in Concerto, Glasnevin no.5, and RGT Planet than in their waterlogged counterparts until the end of the experiment (day 21). However, Golden Promise was the exception, exhibiting higher growth in hue 6 in waterlogged plants during the initial phase, characterized by a significantly elevated pixel count from day 4 to 11. However, towards the end of the experiment, control plants surpassed this difference, showing a notably higher pixel count from day 18 to 21 (Additional file 1: Figure S5).

Following these observable trends, we noticed that hues 5 (Additional file 1: Figure S9) and 6 (Additional file 1: Figure S10) showed a constant trend in pixel counts for both control (Fig. 6A) and waterlogged plants (Fig. 6B). As anticipated, waterlogged plants showed slower growth due to stress, this resulted in lower absolute pixel values than in control samples. After four days, control plants showed significantly higher values of hues 5 and 6 in Concerto, Glasnevin no.5, and RGT Planet than in their waterlogged counterparts until the end of the experiment (day 21). However, Golden Promise was the exception, exhibiting higher growth in hue 6 in waterlogged plants during the initial phase, characterized by a significantly elevated pixel count from day 4 to 11. However, towards the end of the experiment, control plants surpassed this difference, showing a notably higher pixel count from day 18 to 21 (Additional file 1: Figure S5).

We found that darker tones present in hues 2, 3 and 4 (Additional file 1: Figure S6, S7 and S8, respectively) showed a gradual increase in control samples as the plants grew, but no change was observed in waterlogged conditions. In waterlogged plants, these hues appeared on day one (similar to control plants), with their proportions remaining the same after recovery before a general increase. After four days, the control plants showed significantly higher values in darker tones than their waterlogged counterparts, which continued until the end of the experiment. While there was a recovery in the proportions of hues 2, 3 and 4 in waterlogged plants, the increment was lower than the values exhibited by the control plants.

We noticed a ‘stress-induced increase’ trend in the lighter tones, namely hues 7, 8, 9, and 10 (Additional file 1: Figure S9, S10, S11, and S12, respectively), as well as in the darkest hue, hue 1 (Additional file 1: Figure S13). All these hues showed an increase in pixel number due to stress, as determined by a higher pixel count in waterlogged samples when compared to control. Upon recovery, the pixel count for these hues decreased dramatically, reaching similar proportions to control plants (Fig. 6B). The timing and duration of peak values remained highly variable between hues and genotypes; however, we observed an overall trend of significant difference in relative pixel counts from the early onset in waterlogging until halfway through the recovery period. We observed that for these hues (1, 7, 8, 9, and 10), there was a large decrease in the pixel count under waterlogged conditions with a statistical significance reappearing at the end of the experiment (Additional file 1: Figure S9, S10, S11, S12 and S13). In addition, we found that hue 1 was significantly different in all genotypes, in relative pixel count, during the stress period (days 2–14). Golden Promise was significantly different in all days in relative proportions for hue 1.

Discussion

Temporal dynamics of Barley response to waterlogging

Waterlogging has been extensively studied in terms of physiological responses and signalling in plants [42, 43]. Furthermore, considerable knowledge of the physiological responses of barley to waterlogging has been aggregated over the years as reviewed by De Castro et al. [44]. Nevertheless, effect size of waterlogging stress is influenced by many factors, including treatment duration, and a plethora of treatment durations have been tested to study of waterlogging responses [4]. While many traits have been employed to identify waterlogging tolerance, to our knowledge, this study is the first one to examine barley responses to waterlogging stress using high-throughput imaging analysis.

To identify the most suitable waterlogging treatment for phenotyping barley plants, we selected six protocol durations, taking into consideration factors such as feasibility, cost, and durations employed in previous studies. We selected protocol six, consisting of 14 days stress and seven days of recovery, due to the significant reduction in shoot biomass and ability to distinguish genotypic responses. Shoot dry mass under waterlogging conditions was 63% and 56% of control samples for RGT Planet and Glasnevin no.5 respectively, which is consistent with other research using similar waterlogging durations [26, 45, 46]. Our findings agree with the results of Leyshon and Sheard 1974 and Ren et al. 2016, reinforcing that the duration of waterlogging stress is positively correlated with the magnitude of growth inhibition [19, 47]. Such results agree with the findings of a global meta-analysis by Tian et al., which concluded that the proportion of yield reduction in several crop species increased with waterlogging duration [4]. In summary, biomass reduction was observed across various spring barley cultivars and in two experimental locations as discussed in the next section.

Although in this work we focused on shoot response, we also sought to examine the effect of the selected protocols on the roots. Roots play a vital role in sensing and responding to waterlogging stress [31, 48, 49]. We analysed a number of root parameters for Protocol 5 (14 days of stress) and 6 (14 days of stress plus seven days recovery) of Experiment 1, using RhizoVision Explorer software [38]. We found that the effects of waterlogging on root architecture was observed in both RGT Planet and Glasnevin no.5 in protocol 5 whereas no significant effect of treatment was detected in protocol 6, which could be attributed to root plasticity and its recovery over the seven days post stress (Additional file 1: Figure S2). However, unlike other stresses such as drought or heat stress, the removal of waterlogging stress in not instantaneous and lasting effects are present in the soil. There is a dynamic transition from waterlogged to regular water content as the water is permitted to drain with lasting effects on nutrient availability and soil structure. The dynamic transition from anaerobic waterlogging soil to re-oxygenation is consistent with the alleviation of waterlogging expected in field conditions, exposing the plants to a change of environments similar to “real-world” waterlogging events. We observed no significant difference in the average root diameter between control and waterlogged samples, supporting de San Celedonio et al., [50] findings, where no change in average root diameter was observed at harvest in barley waterlogged for 20 days at the beginning of tillering. Our results showed no reduction in total root length in either protocol 5 or 6 due to waterlogging, which was contrary to the observations of de San Celedonio et al. (2017). Presumably, this could be due to the time-point for root analysis as the later study harvested barley at maturity after a much longer growing period. In Protocol 5 (no recovery period), we found that waterlogged samples exhibited a larger proportion of total root volume originating from roots with a diameter less than 1 mm (Additional file 1: Figure S4), suggesting the possibility that this could be a result of the rapid development of adventitious roots in response to waterlogging. Adventitious roots are more porous and better adapted to providing internal aeration to the root system, assisting plants’ adaptation to the new challenging conditions [51, 52]. In fact, after the recovery period of protocol 6, this effect was no longer observed, potentially due to thicker adventitious roots reaching a diameter larger than 1 mm (Visser and Voesenek, 2004) [53]. Such results support the idea that root plasticity and its regrowth during recovery is a fast process and that signs of recovery can be observed at root level before becoming apparent in aboveground biomass. However, this once again highlights the difficulty in understanding responses using only a single time point.

The development of aerenchyma and adventitious roots with higher porosity are a fundamental aspect of the response of barley to waterlogging and an important yet elusive target for large scale phenotyping [52, 54]. Hence, more research into root porosity and aerenchyma formation is needed to understand root dynamic responses and architectural changes under waterlogging Stress. To further investigate the root plasticity in response to waterlogging and its recovery, root imaging over multiple timepoints is expected to provide insights into the adaptability and recovery of different barley genotypes. Progress in aerenchyma quantification methods indicates that the capability to screen large numbers of samples efficiently and affordably may be possible in the near future [55, 56]. Our root analysis was limited by the destructive and the two-dimensional nature of the imaging method. After harvesting and root washing, valuable information about root system architecture is lost, highlighting the need for more in depth investigations into changes in root architecture through less invasive methods such as clear pots, rhizoboxes, or X-ray Computed Tomography (CT) scanning [57]. Altogether, we demonstrated that the inclusion of a recovery period provides us with the best chance of observing differences in growth rate over the course of the stress and beyond. Including a recovery period after stress has previously shown to highlight phenotypic differences in barley populations subject to water stress [33, 58].

Waterlogging stress relevant markers identified through high-throughput image phenotyping

To confirm the findings of Experiment 1, we replicated the selected protocol in a second facility. We examined protocol 6, comprising 14 days of stress, seven days recovery, with an increased number of genotypes and replicates as well as several experimental improvements informed by Experiment (1) In Experiment 2 we used a fully automated system which improved our analysis by increasing the number of replicates to 10, while also facilitating a switch to one plant per tray. Converting to a single plant per tray allowed for the collection of an additional side image of the samples to be taken at 90 degrees to the first image, thus increasing the robustness of the PSA as a proxy for biomass. The additional side view increased the correlation of the PSA to shoot fresh biomass from an R² value of 0.97 in Experiment 1 to 0.99 in Experiment (2) Automatic watering allowed for a consistent maintenance of the water level. Furthermore, in Experiment 2 we were able to conduct chlorophyll fluorescence imaging during the night, ensuring that samples were fully dark adapted. The use of the PlantScreen™ system allowed for the dynamic monitoring of maximum quantum efficiency of PSII (Fv/Fm value) as well as analysis of the whole plant area, whereas previous studies on waterlogging stress, using traditional fluorometers, focused on a small proportion of leaf area with destructive harvest [45, 59].

The maximum quantum yield of photosystem II (Fv/Fm) has been used as a proxy for performance under stress as unstressed plants typically record a highly consistent value of ~ 0.83 [60]. In fact, the Fv/Fm value has been previously shown to be sensitive to stress in barley, including drought [61] and salinity [62]. Our analysis showed similar results at endpoint to Zeng et al. [46], providing a better insight into stress responses by highlighting a rapid decrease during the first days of stress, and a dynamic recover up to a normal Fv/Fm value during subsequent days. Such a dynamic response of the maximum quantum yield of photosystem II enabled us to present an accurate discrimination of barley recovery capacity to waterlogging stress. Although Fv/Fm was reduced in all genotypes, Golden Promise maintained the smallest reduction of Fv/Fm, throughout the course of the waterlogging. Indeed, in Golden Promise, the Fv/Fm value remained consistent between stressed and control samples compared during the course of the protocol. Once Fv/Fm measurements stabilised after day 5, the average Fv/Fm value in waterlogged samples ranged between 0.829 and 0.85 while control samples ranged from 0.832 to 0.853. RGT Planet, Glasnevin no.5 and Concerto displayed similar responses with heavily reduced quantum yield, with the largest reductions occurring between days 8–12 with an average reduction of 7% across the three genotypes in this period. The effect on quantum yield appears to begin to recover towards the end of the stress period and into the seven days of recovery. Fv/Fm measurements produced a clear distinction between control and waterlogged groups for all genotypes except for Golden Promise in Experiment 2. The observation of different genotypical responses would not have been observed without time-series measurements. The benefits of using such a dynamic analysis of Fv/Fm value has already been shown in barley under water stress [33].

Waterlogging stress has previously been shown to reduce Fv/Fm values in barley and recover once stress is removed [26]. In our study, we observed a similar pattern of reduced Fv/Fm in dark-adapted waterlogged samples when compared to control groups, indicating the photo-inhibiting effects of waterlogging stress. A high level of variability was observed in early Fv/Fm values with measurements fluctuating within treatment groups. This is likely due to the small plant size at the beginning of the experiment. Because Fv/Fm is calculated from top view sensors, we can speculate that the orientation of the leaves will have had a large influence on Fv/Fm value readings. This effect is reduced as the plants grow and expand, thus increasing the area from which the value is measured. To further explore the effect of waterlogging on photosynthetic characteristics, we recommend more detailed chlorophyll fluorescence imaging protocols that capture a wider range of parameters. Moreover, the advantages of combined imaging (chlorophyll fluorescence and spectral signatures such as the normalized difference vegetation index -NDVI) will further improve the detection of early stress symptoms due to damage of photosynthesis-related proteins. Such changes have already been described in a proteomics approach of waterlogging in barley in a low throughput phenotyping study [31].

We have shown that greenness analysis (colour hue analysis) provides a fast and observable change in response to waterlogging. It may be concluded that greenness analysis is of interest to highlight plants’ capacity to adapt and recover from waterlogging using specific hues. Of the ten selected hues, we found that three hues (2,3,4) became less prominent in waterlogged samples whereas five hues (1,7,8,9,10) were found to increase in response to stress. Consistent with our findings supported by the Fv/Fm data, the response of the ten hues was very similar in RGT Planet, Concerto, and Glasnevin no.5, but Golden Promise showed a significantly different response (Additional file 1: Figure S9, S10, S11, S12 and S13). This strategy has already been applied by Awlia et al., (2016), facilitating the discrimination of an Arabidopsis population in response to salinity stress [63]. More recently, colour hues have been used to discriminate barley genotypes exposed to drought stress [34]. In our study, a notable difference was observed between protocol 6 of Experiment 1 and Experiment 2, which was a larger size of the samples in Experiment 2, despite waterlogging treatment beginning at the three-leaf stage (Zadocks stage 13) in both experiments [37]. These size differences are likely due to optimised growing conditions within the high-throughput system of Experiment 2 as samples spent less time in transportation and outside of the glasshouse. Harvested shoot dry mass for samples under waterlogged conditions were 475% larger in Experiment 2 while control samples were 678% larger. We speculate that this difference is a result of the less variable growing conditions and the less invasive sample transportation using conveyor belts. Despite this discrepancy in size, a high correlation for the PSA was observed between Experiment 2 and Protocol 6 of Experiment 1 (R² = 0.85). Furthermore, in Experiment 1 the plant size was smaller, but the PSA values were higher, which is a result of the higher resolution RGB sensor used in Experiment 1 (12.36 Megapixels) compared to Experiment 2 (5 Megapixels). The selected protocol (14 days of stress plus seven days recovery) produced the largest differences in biomass between control and stress conditions while also allowing the observation of the recovery of growth rate. We have demonstrated that reproducibility was achieved by exposing the same genotypes to the same duration of stress and recovery in an alternative facility and adjusted for high-throughput phenotyping. This work provides a methodology for the comparison of barley response to waterlogging stress using imaging phenotyping to observe the stress response over time. Projected shoot area, growth rates, maximum quantum yield of photosystem II (Fv/Fm), and colour hue analysis provided a non-destructive assessment of plant health under stress, enabling us to quantify genotypic responses to waterlogging.

Deciphering the dynamic variation of different phenotyping traits

We have identified an optimal waterlogging protocol to compare stress responses of barley, reinforcing previous results in terms of stress indices [13] that have assessed waterlogging responses using classic phenotyping methods such as harvested biomass and chlorophyll content. Image-based phenotyping of waterlogging has provided a higher resolution view of this stress response as PSA allows for a fast and accurate quantification of plant biomass throughout time. Using PSA, we calculated growth rates and observed a resuming of regular growth shortly after the conclusion of the stress period as demonstrated by an average RGR of 0.2672 for all waterlogged samples between days 18–21, which was higher than that of the control samples (0.2371). Resumption of growth rates was previously observed by Sepp, 2021 [64] in barley waterlogged for seven days, where after a 14-day recovery period the RGR of waterlogged samples exceeded that of control by 48%. Hence, we may hypothesize that the speed at which biomass accumulation recovers can indicate a genotype’s ability to maintain production under waterlogging conditions during seasonal periods. In fact, the RGR of barley exposed to waterlogging at later developmental stages has been shown to correlate with yield, with RGR during recovery periods being more representative than during the stress [65]. Additional work will need to be conducted to determine the extent to which recovery of growth rate, after waterlogging and at the tillering stage, translates to lower yield penalties over a growing season. The use of imaging sensors and high-throughput phenotyping has been utilised for better characterisation of the dynamic responses to abiotic stress such as drought [41]; however, to our knowledge, this work is the first using high-throughput phenotyping to screen for waterlogging tolerance in barley. Integrating PSA as a proxy for biomass accumulation with more sensitive parameters such as chlorophyll fluorescence and colour hue analysis has provided us with an in-depth insight into barley stress responses, highlighting genotypic differences that may have been missed by using classic phenotyping methods. Additionally, our work paves the way for researchers and breeders to use high-throughput automation to analyse a large number of samples and genotypes under waterlogging stress, thus enabling an efficient screening of germplasm responses to this stress. The stress duration selected in our optimal protocol allowed for the identification of parameters sensitive to waterlogging stress. We expect that using a larger number of genotypes and sensors (e.g., hyperspectral imaging and x-ray CT scanners), this set of parameters can be improved and expanded to aid and enhance the screening of barley germplasm in response to stress. The ability to effectively screen for waterlogging tolerance at tillering stage under controlled conditions will reduce the time and resources needed for crop improvement in response to rapidly changing climate conditions.

Conclusions

In this work we delve into the intricate responses of spring barley plants to waterlogging stress using high-throughput imaging analysis, a novel approach in this context. Our research investigated various treatment durations and identified protocol 6 (14 days of stress followed by seven days of recovery) as the most suitable for observing significant reductions in biomass and growth rates under waterlogged conditions. We also introduced a range of imaging parameters, including projected shoot area, chlorophyll fluorescence, and colour hue analysis, to provide a comprehensive view of the stress response. The dynamic nature of Fv/Fm values revealed genotypic differences in overcoming waterlogging stress, with Golden Promise displaying distinct characteristics. Screening for waterlogging tolerance under controlled conditions at the tillering stage offers several advantages over field trials, yet further work is required to determine the relationship between early-stage tolerance and agronomic yield penalties. This work underscores the advantages of high-throughput image phenotyping in uncovering subtle genotypic responses that classic methods might miss. The waterlogging and imaging protocol presented here could be applied to a range of species, namely cereal crops, limited only by the physical dimensions of the imaging chamber. Optimal stress duration and growth stage may need to be slightly adjusted depending on species and research hypothesis. By deciphering the dynamic variations of various phenotyping traits, this study advances our understanding of barley’s waterlogging tolerance and provides a protocol duration suitable for identifying genotyping changes in waterlogging stress response.