The study was conducted at the Israeli Center of Research Excellence (iCORE) (http://departments.agri.huji.ac.il/plantscience/icore.Phpon) at the Robert H. Smith Faculty of Agriculture, Food and Environment, The Hebrew University of Jerusalem, Rehovot, Israel. iCORE is a functional physiological phenotyping facility, consisting of the Plantarray 3.0 system (Plant-DiTech, Israel) in a semi-temperature-controlled greenhouse (Dalal et al., 2020; Pandey et al., 2021). The system consists of highly sensitive gravimetric lysimeters, soil and atmospheric probes, data acquisition units and a precise irrigation controller. The system enables continuous assessment of plant–water relations and soil and atmosphere parameters throughout the plant’s entire growth season.

Plants were grown in pots, with each pot connected to its own irrigation controller and data-acquisition unit. Data were recorded automatically every 3 min and saved in a server. Atmospheric variables—air temperature, relative humidity (RH), VPD, and photosynthetically active radiation (PAR)—were recorded in the Plantarray system.

Tef genotypes were selected from our tef diversity panel (TDP-300) based on a previous study (Alemu et al., 2024). Genotypes with similar, medium phenology (heading time), but diverse productivity were selected based on their performances in a replicated field experiment conducted in 2021 ( Supplementary Table S1 ).

A two-way factorial experiment was conducted with six tef genotypes x two treatments (control and drought) and six replicates, altogether 6 x 2 x 6 = 72 pots (experimental units). A random block design was used to ensure uniform exposure to the environment and minimize variations. Load-cell units were calibrated under constant load weights (1 kg and 5 kg) using the Plantarray software. Plastic drainage containers were positioned on the Plantarray lysimeters before transferring the pots into the system, to ensure the accuracy of lysimeter weighing and minimize noise in the system. The initial total pot weight, comprising all components, was incorporated into the Plantarray system before the experiment was initiated. This included weight of the pot, drainage container, pot soil, water in the drainage container, stick and trilling rope, and seedlings’ initial fresh weight. The later was determined through destructive harvest on the day of transplanting (28 days after sowing), averaging the weights of nine seedlings per genotype.

Seedling trays and growing pots were sterilized and thoroughly washed before being filled with the growing medium. Seedling trays (10 ml cone) were filled with growing medium Matza Gan (Shaham, Givat-Ada, Israel). Tef seeds were sown on 22 May 2022 in seedling trays (~4 seeds cone^-1), watered manually once a day, and maintained in the iCORE greenhouse for 28 days. Thinning was conducted after 1 week to 1 seedling cone^-1.

Plastic pots (4 l) were filled with silica sand 20/30 mesh (Negev Industrial Minerals Ltd., Israel) growing medium and washed with water prior to transplanting to eliminate pore spaces. To minimize evaporation, the soil surface was covered with a white polyethylene-vinyl acetate with three equally spaced planting holes. Transplanting was used to establish a single plant in each planting hole. Three seedlings were transplanted per pot (12 pots per genotype) and each pot was placed in a plastic drainage container at its designated position on a Plantarray lysimeter and connected to four outlet drippers to ensure uniform soil moisture ( Figure 1A ). Pre-experimental observation of tef daily transpiration (DT) was used as a reference to determine the initial daily irrigation amount. Irrigation was applied four times during the night (2000 h–0300 h) to minimize “noise” during data measurements. The pot soil’s volumetric water content (Cθ) was about 1200 ml, with 80 ml of water remaining in the drainage containers, providing extra water to the control plants beyond the pot soil ‘s capacity.

The irrigation program was designed to apply differential irrigation treatments (control and drought stress) based on each pot’s transpiration. In the control group, each pot was irrigated at a rate of ~130% relative to its transpiration on the previous day. Water application to the drought-treated genotypes consisted of three phases: pretreatment (13 days, identical to control) ( Figure 1B ), differential irrigation (24 days) ( Figure 1C ), and recovery (6 days) ( Figure 1D ). To avoid rapid water depletion and to mimic the development of soil water deficits in the field, water application was gradually reduced by ~60 ml of each pot’s previous day’s transpiration for 11 days (day 14 to 24). Irrigation remained at this level for 10 days (day 25 to 34), and was then further reduced over 3 days (day 35 to 37) to provide 50% of the previous day’s transpiration. Finally, on day 38, at the onset of the recovery period, full irrigation was resumed until the final harvest on 3 August 2022 (43 days after transplanting).

Environmental conditions in the greenhouse throughout the experimental season were on average (min/max): temperature 22.7/31.1°C, RH 45.5/80.5% and VPD 0.62/2.41 kPa, with an average maximum PAR of 1318 μmol m^-2 s^-1 ( Supplementary Figure S1 ). Commercial fertilizers were applied through the drip irrigation system (fertigation), providing 67, 10 and 42 ppm N, P and K, respectively, and micronutrients. No pesticide application was required during the experimental season.

High-throughput phenotyping platform was employed to continuously monitor whole-plant functional physiological traits related to plant–water–soil–atmosphere kinetics. TR, DT, normalized transpiration (E), canopy stomatal conductance (G_sc) and calculated fresh weight (CFW) were either directly recorded or estimated by the Plantarray system. Real-time data inspection and analysis were carried out using Soil-Plant-Atmosphere-Continuum (SPAC) Analytics web-based software (Plant-DiTech, Israel) (Dalal et al., 2017; Halperin et al., 2017; Gosa et al., 2022).

A conventional phenotyping approach was used to assess phenology, physiology and productivity traits. Days from planting to heading were recorded based on visual observations. Leaf osmotic potential (OP), measured as previously described by Alemu et al. (2024), and chlorophyll content (Chl), measured using a SPAD 502Plus chlorophyll meter (Konica Minolta, Japan), were recorded twice: at the end of the differential irrigation treatment and at the end of the recovery/experimental period. Productivity traits, including shoot, root and total dry weights (SDW, RDW and TDW, respectively), were measured following destructive harvest. Shoot and root biomass were separated, roots were thoroughly washed, and both parts were oven-dried (60°C, 74 h) and weighed.

JMP® version 16 Pro statistical package (SAS Institute, Cary, NC, USA) was used for ANOVA to test the effects of genotype, treatment and their interactions, as well as for correlation analyses.

Piecewise curve fitting was conducted using the SPAC Analytics software to estimate the relationships between midday (1200–1400 h) TR and Cθ for each of the drought-treated pots, enabling a calculation of critical drought point (θ_crit) and the slope of the TR reduction at Cθ < θ_crit.

To assess the drought-recovery rate, we divided DT and CFW by their value on the first recovery day to calculate their relative values (RDT and RCFW, respectively). These relative values were correlated vs. recovery day using a linear regression, to determine their slopes (recovery rate). Differences between slopes of the tested genotypes were examined using GraphPad prism, Version 10.0.1 (GraphPad Software, Boston, MA, USA).

Relative total dry weight (RTDW), calculated as TDW under drought relative to TDW under control conditions, was used to estimate the effect of drought on productivity. WUE was calculated for the pretreatment phase based on plant fresh weight as: WUE_fw = ΔCFW_1-13/ΣTR_1-13, where ΔCFW_1-13 is the difference between the calculated CFW on days 1 and 13 and ΣTR_1-13 is the cumulative transpiration for days 1 to 13. For the entire season, dry weight-based WUE was calculated as WUE_dw = TDW₄₃/ΣTR_1-43, where TDW₄₃ is TDW on day 43 and ΣTR_1-43 is the cumulative transpiration for the entire growing period (Leakey et al., 2019).

Cθ in the control group remained at 26–27% (apparently soil field capacity) throughout the entire experiment ( Figure 2A ). In the drought-treated group, Cθ was similar to the control during pretreatment phase (days 1–13), decreased gradually from day 14 to 24, stabilized at 11–13% on days 25–34, further decreased on days 35–37 reflecting the extreme drought applied, and then increased back to field capacity at the onset of the recovery phase.

During pretreatment phase, DT exhibited a gradual increase with highly significant variation (p <0.01) between genotypes ( Figure 2B , Supplementary Table S2 ). Then, during differential irrigation, it continued to increase in the control group until about day 25 when it leveled off ( Figure 2B ). At the same time, the drought-treated group exhibited a gradual reduction in DT, even when Cθ was rather stable (days 25–34). Significant variations were observed in DT between treatments (all days) and genotypes (most days) with no genotype by environment interactions ( Supplementary Table S2 ). Under control conditions, genotypes RTC-273b and RTC-359 consistently showed the highest and lowest DT, respectively, whereas under the drought treatment, RTC-290b and RTC-273b usually presented the highest and lowest DT, respectively. However, under extreme drought, toward the end of the differential treatments, only minor variation was observed between genotypes.

Following 24 days of differential water application, all plants were subjected to full irrigation ( Figure 2A ), thus allowing the drought-treated plants to rapidly absorb water and increase DT ( Figure 2B , Supplementary Table S2 ). During the 6-day recovery period, all genotypes in the drought group exhibited a rapid increase in DT, while genotypes in the control group displayed either stable or reduced DT as compared to the differential irrigation phase. Highly significant variations were observed in DT between treatments ( Supplementary Table S2 ). Consistent variations were observed between the genotypes; RTC-290b showed the highest DT under both treatments, whereas the lowest values under control conditions were recorded for RTC-273b and RTC-275, and under drought conditions for RTC-359.

CFW of the tef genotypes was recorded across the entire experiment, using non-destructive measurements ( Figure 2C ). During the pretreatment, there was a substantial increase in CFW with significant differences between genotypes observed during most days ( Figure 2D , Supplementary Table S3 ). During the differential irrigation, the drought-treated group exhibited slower biomass accumulation compared to the control group. Variations in CFW between treatments became highly significant at the middle of the differential irrigation phase, while differences between genotypes and genotype by environment interactions became consistently significant at the last 4 days ( Supplementary Table S3 ). At the end of this phase, significant variation was noted in CFW within the control group, whereas within the drought-treated group, no significant variation was evident ( Figure 2E ). During the recovery phase, CFW was significantly affected by genotypes, treatments and their interactions ( Supplementary Table S3 ). Furthermore, at the end of the recovery period, significant variation was observed in CFW between genotypes in the control group, but not between the drought-treated genotypes ( Figure 2F ). RTC-290b consistently demonstrated the best performance in terms of CFW under both control (significant) and drought (non-significant) treatments throughout all periods, while RTC-359 and RTC-273b exhibited the lowest performance under control (significant) and drought (non-significant) treatments, respectively ( Figures 2D–F ).

The diurnal patterns of G_sc and E, for selected days of each experimental phase, are presented in Figure 3 alongside PAR and VPD. Tef genotypes exhibited variations in G_sc and E in response to the diurnal pattern of VPD and PAR, with a major effect of Cθ. Both G_sc and E displayed, in most cases, a rapid increase during the morning (0500–1000 h), followed by stable values throughout midday (1000–1600 h) and a decrease in the late afternoon (1600–1900 h), following the patterns of PAR. A clear deviation from this trend was observed in the drought-treated plants which displayed their maximum G_sc and E at about 1100 h and a continuous decrease thereafter until the evening.

The relationships between mid-day TR and Cθ under the drought treatment, revealed significant variation between genotypes in their responses to water availability. Two genotypes, RTC-273b and RTC-275, exhibited the highest TR (>1 g min^-1) under high to moderate Cθ, while RTC-359 and RTC-47a exhibited the lowest values (~0.74 g min^-1) ( Figure 4A ). Tef genotypes reflected significant variation in θ_crit (p = 0.001), with RTC-275 exhibited the highest level (0.20 cm³/cm³), significantly different from two genotypes, whereas RTC-273b exhibited an intermediate level (0.18 cm³/cm³), not differing from the highest or lowest genotypes (0.162 cm³/cm³) ( Figure 4B ). The high-transpiring genotypes (RTC-275, RTC-273b) exhibited high values of θ_crit, whereas the medium-transpiring (RTC-290b) and low-transpiring (RTC-47a, RTC-359, RTC-364) genotypes exhibited lower θ_crit. As Cθ was further reduced below θ_crit, TR declined rapidly at rates (slopes) which differed significantly (p < 0.0002) between genotypes. RTC-273b and RTC-275 exhibited the fastest reduction rates, whereas RTC-364, RTC-359 and RTC-47a showed slowest rates, and RTC-290b reflected an intermediate reduction rate that did not differ from any of the other genotypes.

Following the resumption of full irrigation, all genotypes exhibited remarkable recovery from drought stress, as reflected by increasing DT and CFW ( Figure 2 , Supplementary Tables S2 , S3 ). The slopes of RDT and RCFW during the recovery phase displayed significant differences (p = 0.003 and p < 0.0001, respectively) between genotypes ( Figure 5 ). RTC-290b exhibited the greatest RDT slope (recovery rate), significantly different from all other genotypes ( Figure 5A ), whereas for RCFW, RTC-290b, RTC-364 and RTC-47 exhibited significantly greater slopes than the other genotypes ( Figure 5B ).

Tef phenology (days from planting to heading) was significantly affected by water availability and genotype ( Table 1 ). Drought stress imposed, on average, a ~6-day delay in heading time; however, genotypic ranking across treatments was similar. Heading time under both treatments was earliest in genotypes RTC- 273b and RTC- 275, followed by RTC-364 and RTC-47a, and ending with RTC-290b and RTC-359.

The impact of drought stress was also evident at the end of the differential irrigation period as a 0.31 MPa lower OP under drought stress compared to the control ( Table 1 ). While under control conditions, no significant differences were found between genotypes, the drought-treated group reflected significant variation, with three genotypes (RTC-290b, RTC-359 and RTC-364) showing the lowest values. After recovery for 6 days, all genotypes showed increased OP, to a level similar to the control, with RTC-359 showing the lowest values under both treatments. Drought stress induced an increased Chl content, however responses of various genotypes were inconsistent ( Table 1 ).

ANOVA revealed significant differences between treatments and genotypes for most productivity variables, and non-significant treatment-by-genotype interactions ( Table 2 ). Drought stress led to a substantial reduction in TDW, SDW and RDW, by 59%, 62% and 44%, respectively (average across genotypes) ( Table 2 , Supplementary Figure S2 ), and an increase of 50% in the RDW-to-SDW ratio, relative to control conditions. Notably, RTC-290b displayed the highest TDW and SDW, while RTC-47a exhibited the lowest values under both control (not significant) and drought treatments. On the other hand, no significant differences were detected between genotypes in RDW and RDW-to-SDW ratio under either treatment. In addition, the analysis of RTDW revealed non-significant differences between genotypes, with RTC-290b and RTC-275 exhibiting the highest values.

Significant differences were found between genotypes for the pretreatment WUE_fw, whereas for WUE_dw, differences between both genotypes and treatments were significant ( Table 2 ). A rather small, albeit significant reduction in WUE_dw was recorded under drought relative to control conditions. RTC-290b exhibited the highest WUE in all cases under both treatments, the lowest WUE values were noted for RTC-47a under control conditions, and for RTC-273b and RTC-47a under drought, whereas all other genotypes exhibited intermediate levels.

Transpiration under drought stress can differentiate resistant from susceptible genotypes (Bacher et al., 2022). Plants that exhibit high transpiration under optimal water supply, and a moderate decrease under drought, combined with stress resilience, can secure both high and stable productivity across a range of water availabilities (Appiah et al., 2023). During drought stress, plants shift into a survival mode at the expense of their productivity, thus reducing transpiration and carbon fixation as compared to controls (Moshelion, 2020). Tef genotypes exhibited increasing DT during the pretreatment phase, which has rapidly decreased upon exposure to drought stress ( Figure 2 , Supplementary Table S2 ). During the differential irrigation phase, plants exhibited variations in DT between treatments and genotypes, suggesting G_sc-regulated changes in gas exchange. During the recovery phase, the drought-treated plants responded immediately to re-watering by increasing their DT and CFW accumulation ( Figure 2 ). During all three phases and under both treatments, tef genotypes manifested considerable differences in DT, E and G_sc, presumably reflecting genetic variation ( Figures 2 , 3 , Supplementary Table S2 ), which ultimately led to differences in productivity and WUE ( Table 2 ). Compared to the other genotypes, RTC-290b exhibited moderate DT under control and high DT under drought ( Figure 2 ), as well as rapid recovery ( Figure 5 ), indicating its resilience to water stress.

The interactions between VPD, PAR, G_sc and E are crucial to maintaining physiological processes and optimizing plant water use and productivity under different water regimes (Gosa et al., 2022). VPD plays a vital role in water transport and regulation of stomatal conductance (Appiah et al., 2023). Drought-treated plants displayed a distinctly different diurnal pattern in G_sc and E compared to the control group ( Figure 3 ), suggesting different regulation and water-use strategy. Upon resumption of full irrigation, G_sc and E showed similar diurnal patterns under both treatments, with somewhat lower values in the drought-treated group, reflecting rapid recovery from stress.

Osmotic adjustment (i.e., reducing OP) is a common plant response to drought stress, enabling the maintenance of water absorption, turgor pressure and structural integrity (Turner, 2017; Condorelli et al., 2022). Similar to our previous study (Alemu et al., 2024), tef plants exhibited a significant reduction of OP in response to drought ( Table 1 ). It is worth noting that leaves were water-saturated prior to the OP measurement and therefore, the reduction in OP under drought represents active solute accumulation per se, rather than passive water loss. During the recovery period, OP of the drought-treated plants was rapidly restored, showing values comparable to control plants after only 6 days of full irrigation.

Chl is an indicator of plant photosynthetic capacity under various environmental conditions (Fiorentini et al., 2019; Hasanuzzaman et al., 2022; Yang et al., 2023). Higher Chl under drought-stress conditions is associated with higher Chl density per unit leaf area and increased leaf thickness (Hasanuzzaman et al., 2022). Drought-treated tef exhibited increased Chl in previous studies (Girija et al., 2022; Alemu et al., 2024). In the current study, genotypes RTC-47a and RTC-290b exhibited the highest increase in Chl under drought, which was retained after recovery ( Table 1 ). These two genotypes were among the three that exhibited the highest recovery in terms of CFW accumulation ( Figure 5B ), which might be related to their higher Chl.

The critical drought point (θ_crit) is the soil water content below which plants fail to extract water, ultimately leading to reductions in G_sc and TR (Mishra et al., 2022; Appiah et al., 2023; Paul et al., 2023). Based on their stomatal plasticity in response to drought, plants are classified as either water-conserving (isohydric) or non-water-conserving (anisohydric); however, an intermediate strategy is also evident (Yi et al., 2019). Tef genotypes reflected significant variation in θ_crit and the reduction in TR (slope) at Cθ < θ_crit, indicating diversity in their responsiveness to drought ( Figure 4 ). RTC-275 and RTC-273b exhibited isohydric characteristics (high TR, θ_crit and slope), thus prioritizing water conservation over productivity, whereas RTC-359, RTC-364 and RTC-47a exhibited anisohydric characteristics (lower TR, θ_crit and slope), prioritizing productivity over water conservation. An intermediate strategy was displayed by RTC-290b (medium TR, θ_crit and slope), which might have supported its rapid recovery ( Figure 5 ) and high productivity ( Table 2 ).

Drought recovery refers to the plants’ physiological capacity to resume growing and producing after experiencing severe drought stress; it is considered an indicator of resilience (Fang and Xiong, 2015; Chen et al., 2016; Bongers et al., 2017; Appiah et al., 2023). Tef has the ability to maintain physiological function during drought stress and recover after water resumption, but exhibits variations in degree of recovery (Ginbot and Farrant, 2011; Kamies et al., 2017; Girija et al., 2022). In this study, tef genotypes exhibited significant variation in recovery rate (slopes) for RDT and RCFW ( Figure 5 ) after 24 days of drought stress. RTC-290b displayed a high rate of recovery in both RDT and RCFW, whereas RTC-364 and RTC-47a exhibited high recovery for RCFW. Further studies are required to confirm these results and investigate recovery at later phenological stages.

The impact of water availability varies across different genotypes, reflecting their ability to adapt and respond to diverse water regimes and determining their productivity and WUE (Leakey et al., 2019). Under drought, tef growth and development are reduced, dependent on stress severity, growth stage and genetic makeup (Girija et al., 2022). In the current study, tef genotypes exhibited different performances under contrasting water regimes ( Supplementary Figure S2 , Supplementary Tables S2 , S3 ). Drought-treated genotypes exhibited a slower accumulation of biomass ( Figure 2 ) and a remarkable (~60%) reduction in their final dry weight compared to controls ( Table 2 ). In our previous field experiments, tef productivity was reduced by 20–39% under drought stress (Alemu et al., 2024), similar to other studies (Admas and Belay, 2011; Abraha et al., 2017, 2018). In the field, tef develops a deep root system (Degu et al., 2008), which slows down the development of drought stress relative to growth in pots. The distribution of resources among various organs during drought stress serves as a drought-adaptive strategy (Yan et al., 2023). Drought-treated tef genotypes exhibited higher root-to-shoot mass ratio than the controls ( Table 2 ), presumably as an avoidance mechanism (Zhou et al., 2018), by which they improve water uptake and reduce transpiration.

WUE involves physiological trade-offs and sensitivity to genotype-by-environment-by-management interactions (Leakey et al., 2019). Plants exhibit variations in G_sc and E in response to water availability, which in turn alter biomass accumulation and WUE (Sun, 2023). Being a ratio between photosynthesis and TR (or alternatively, biomass and water use), high WUE can be obtained by either increased productivity or reduced water use. Therefore, in breeding for drought resistance, it is important to combine high WUE with high productivity (Merchuk and Saranga, 2013). In the current study, WUE showed significant differences between treatments and genotypes ( Table 2 ). In this respect, it is highly important that RTC-290b demonstrated the highest WUE and productivity (TDW) consistently across different water availabilities ( Table 2 ).

High-throughput greenhouse-based phenotyping systems, such as the one used in the current study, can potentially accelerate functional plant phenotyping and the development of drought-resilient crop genotypes. However, it is important to confirm that the findings obtained from such systems are relevant to field conditions and determine what truly works under the relevant environments (Khaipho-Burch et al., 2023). Correlation analysis revealed a significant association (r² = 0.72, p = 0.03) ( Figure 6 ) between RTDW values in the current pot experiment and those recorded for the same genotypes in our previous field experiment. Two genotypes (RTC-275 and RTC-290b) exhibited the highest productivity while other genotypes exhibited intermediate or low values across both experiments. It is worth noting that correlation between TDW and SDW under drought in the current study vs. TDW under drought in the field fell somewhat above the common significance threshold (p < = ~0.09). Additional studies are required to confirm these association with a wider set of genotypes and to fine-tune the greenhouse experimental procedures to better mimic plant responses in the field.