Here, we show differences in the molecular, morphophysiological and biochemical responses of two contrasting soybean genotypes, MUNASQA (drought tolerant) and TJ2049 (drought susceptible), subjected to mild water deficit treatments in V3 and R5 phenological stages.
1. Molecular insights
Transcriptional changes were assessed in MUNASQA and TJ2049 after 72 h of exposure to mild water deficit. From 38.658 transcripts analysed, drought-stressed MUNASQA and TJ2049 plants exhibited 2952 and 1126 transcripts with significant changes (P < 0.05) in their expression levels (Suppl. Data 1), respectively. After an FDR=0.1, 399 and 15 transcripts were assigned as MUNASQA and TJ2049 DEGs (Suppl. Data 2). The transcript loss detected in TJ2049 might be explained by the larger variation among replicates observed in the water deficit samples and measured as SD/mean ratio for each gene (Suppl. Table 1, Fig. S2). However, in a previous exploratory 454 sequencing experiment, TJ2049 showed significantly fewer DEGs than MUNASQA (data not shown).
Large-scale transcriptional reprogramming has long been recognised as the first response to drought, initiating stress mitigation pathways and metabolic changes14. Moreover, the quickness to sense and respond to stresses could be an essential factor to differentiate tolerant and susceptible genotypes. Here, after a mild drought, far from normal field levels, the genotypes exhibited dramatic changes on their transcriptional profiles and no observable phenotypic alteration. In the heat map (Fig.1A), DEGs expression profiles were classified in 8 clusters (Suppl. Data 3), and almost 50% of MUNASQA DEGs showed repression under stress conditions, a difference reinforced by the Venn diagram (Fig.1C).
Generally, in response to drought, plants initially trigger transcriptional control and hormone signalling, leading to metabolic adjustment for coping with low water availability. Cellular mechanisms such as water/ion uptake and transport, redox homeostasis, ROS scavenging, osmoregulation and membrane protection are accompanied by physiological responses such as stomata regulation, root development and protection of the photosynthesis machinery15. Here, MUNASQA up-regulated several DEGs involved in these physiological responses (Suppl. Table 2). DEGs from "chlorophyll-binding" and "antioxidant activity" (MF category), "thylakoid", "thylakoid membranes" and "chloroplast" (CC category), and "photosynthesis", "response to water" and "response to desiccation" (BP category) (Fig.1B), were identified. Overall, these results indicate a tightly regulation of stress-response, growth and water use mechanisms in drought-stressed MUNASQAs.
The expression of ten randomly selected DEGs by qRT-PCR (Fig. 1D) agreed with the MACE profile data for MUNASQA and TJ2049, showing a high similarity between both methodologies. Moreover, genes encoding detoxifying proteins like SOD, CAT and APX were differentially regulated in both genotypes (Suppl. Table 3). Enzymes like these are stress-response indicators, frequently regulated at both transcriptional and post-transcriptional levels under drought16. Here, we detected two up-regulated SOD genes under stress conditions in MUNASQA and, contrary to TJ2049, three down-regulated genes for APX and CAT. In accordance, SOD activity measurements in R5 corroborated these differences in transcriptional activity between the two genotypes (Suppl. Table 4).
2. Morphophysiological and biochemical phenotyping
MACE assay, by itself, does not fully explain the extent of MUNASQA and TJ2049 responses to the applied stress. Thus, extensive morphophysiological and biochemical measurements were performed to minimise the gap between the transcriptional regulation and phenotypical alterations observed. Here, 22 markers associated with stress-response, growth and water-use were evaluated in both genotypes exposed to mild water deficit in V3 and R5 stages.
Stress-response markers
An efficient antioxidant activity is crucial for withstanding low cellular water content, and its importance in plant drought tolerance has been extensively reported17. Here, substantial differences were found in MUNASQA and TJ2049 SOD, APX, POX, and CAT enzymes regulation over time, under water deficit and independently of the phenological stage (Table 2). MUNASQA reached maximum activity for all enzymes after 4 d of water deficit, while the highest activity for TJ2049 occurred to 8 d after the stress onset, except for CAT. Noticeably, under non-stressed conditions, all enzymatic activities, excluding CAT, showed higher activity in the tolerant genotype than in the susceptible one, strengthening the hypothesis that TJ2049, compared to MUNASQA, presents delayed stress perception and response mechanisms. In fact, and according to Carvalho et al. (2019)16, a successful response to water deficit may depend not only on which enzymes are activated but also on the activation timing. Moreover, the general gene expression profile regarding these enzymes regulation was highly consistent with the biochemical results.
Regarding PRO, one of the most common osmoprotectants in plants18, MUNASQA and TJ2049 accumulated the osmolyte in response to water deficit over time and in both V3 and R5 stages (Table 2). Several studies have demonstrated a direct correlation between high osmoprotectant accumulation and drought tolerance in many crops19. Here, the tolerant genotype MUNASQA exhibited a higher and more rapid accumulation of PRO after 4 d of water deficit at both phenological stages. In agreement with our results, a recent study in soybean reported higher PRO accumulation in the drought-tolerant genotype A5009 RG, compared with the drought-susceptible ADM5004820.
When analysing MDA production, an indicator of lipid peroxidation and stress severity21, a higher and more rapid accumulation was detected in TJ2049 plants in response to water deficit, compared to MUNASQA (Table 2).
Drought also affects leaf pigments content21. Changes in photosynthetic pigments can alter various light-harvesting processes, while the accumulation of photoprotective compounds plays an essential role in preventing photo-oxidative damage22. Here, MUNASQA and TJ2049 showed alterations in pigment content under water deficit (Table 2). The CHL was significantly reduced over time due to stress in both genotypes, but this reduction was significantly lower in the tolerant one. Similar results were found in drought-tolerant maise that showed lower CHL reductions under stress than susceptible genotypes23. Regarding CAR levels, TJ2049 and MUNASQA showed an increased synthesis in response to drought, although the last one exhibited greater and faster accumulation.
According to our results, MUNASQA is more efficient at removing reactive oxygen species (ROS) under drought stress conditions, an essential mechanism for stress tolerance24.
Growth and yield markers
The ability to produce high seed yield or biomass under limited water access is considered the optimal indicator of drought tolerance in crops25–27. We evaluated the effect of mild water deficit in MUNASQA and TJ2049 growth and yield by monitoring changes in various markers related to biomass and seed production, including LAI, LAR and NAR, RGR, CGR, relative yield and yield-DSI (Table 3).
Total leaf surface area and LAI are strongly associated with canopy interception, evapotranspiration and photosynthesis28. Here, independently of the phenological stage or water availability, MUNASQA plants exhibited a higher LAI compared to TJ2049, indicating a larger assimilatory capacity and, as a consequence, photosynthetic potential. In vegetative stages, a higher LAI denotes a more rapid canopy development, favouring greater and faster soil coverage and thus less water loss from direct evaporation. In general, drought suppresses leaf initiation and growth and consequently affects LAI29. Therefore, a decrease in LAI is expected in plants exposed to water scarcity. As expected, MUNASQA and TJ2049 plants showed a reduction in LAI in response to drought, more noticeable after 8 d of stress and in the R5 stage.
Drought also alters the LAR: the leaf area development in relation to the total biomass produced30. Here, water deficit only affected LAR during the vegetative stage, displaying a significant reduction in both genotypes exposed to stress. The highest ratio between plant leaf area and biomass is reached at the beginning of the plant life cycle31, which explains the highest LAR values at the first sampling day during the vegetative stage.
In MUNASQA and TJ2049, the relationship between leaf area expansion and the biomass produced over time (NAR) was also reduced due to water deficit at both phenological stages. Changes in photosynthetic efficiency were more significant at V3, where both genotypes exhibited greater NAR values. Noticeably, well-watered MUNASQA plants showed a significantly higher NAR than TJ2049 ones, while, in response to stress, MUNASQA’s NAR increased in contrast to TJ2049 values, which were reduced. Although not as accentuated, a similar pattern was observed in plants exposed to water deficit in R5. Here, LAI, LAR and NAR results indicated that, in response to drought, MUNASQA plants regulated photosynthates allocation to leaves and the maintenance of photosynthetic efficiency more efficiently than TJ2049 ones.
In addition, the relative growth rate (RGR) or biomass produced over time was significantly reduced by water deficit in the vegetative stage (Table 3). Interestingly, the reduction in RGR was more pronounced in TJ2049 plants. Regarding CGR, significant differences were observed in the V3 stage after 8 d of drought, while in R5 were detected after 4 d.
These results agree with the differences in yield and yield-DSI exhibited by MUNASQA and TJ2049 after water-deficit treatments in V3 and R5 (Fig. 2). According to these findings, when drought was applied in the V3 stage, TJ2049 showed a distinct but not significant yield penalty and a yield-DSI considerably higher than MUNASQA. Moreover, after a mild water deficit in R5, a highly moisture-sensitive phenological stage, TJ2049 exhibited the largest yield loss and a significantly higher yield-DSI than MUNASQA. Results that also agree with the ones reported by Pardo et al. (2015)13.
Water use markers
Maintaining tissue/cellular water content and/or metabolic activity at low water potentials are physiological strategies to survive drought32. Traits like pubescence, changes in stomatal density, slow wilting, leaf thickening, or tight control of stomatal closure, canopy temperature, RWC, and WUE are essential to determining drought tolerance in plants. Thus, all these water-use parameters were assessed in MUNASQA and TJ2049 plants in response to drought.
Morphological drought-adaptation traits, such as stomatal, trichrome density and leaf thickness33, showed significant differences after prolonged exposure to mild water deficit (21 d) (Table 4). The stomatal and trichrome densities were substantially altered in response to drought in both genotypes. Interestingly, in MUNASQA, stomatal density increased in response to stress on both leaf sides; in contrast, TJ2049 plants exhibited a significant decrease. Regarding pubescence, no significant alterations were observed on the adaxial surface in response to drought; however, on the abaxial side, an increase of trichrome density was quantified in MUNASQA plants, which was higher than in TJ2049. Generally, plants that grow under water deficit develop denser stomata and trichomes34, mainly on the abaxial surface35, which helps prevent excessive water loss through transpiration. Our results suggest that MUNASQA morphoanatomic adaptations favour the genotype ability to cope with water deficit.
When evaluating the genotypes leaf thickness, no significant differences were detected under well-irrigated conditions. However, and unexpectedly, after 21 d of water deficit, MUNASQA leaves were significantly thinner, while TJ2049 ones were thicker. The knowledge about the links between leaf morpho-anatomy and its function under non-stressed/stressed conditions is relatively poor, especially for traits like leaf thickness36 that is strongly related to transpiration37 and reported by some authors as a drought-tolerance trait that maintains turgor pressure and enhances photosynthesis38. Yet, this feature, only apparent in TJ2049 plants under water deficit, could not be associated with adjustments in transpiration, nor with an increase in photosynthesis, or any other drought-tolerant feature.
On the other hand, the regulation of stomatal aperture reinforces the drought-tolerant character of MUNASQA. Under non-stressed conditions, the susceptible TJ2049 showed more opened stomata (⁓22% more than MUNASQA). Moreover, after 3 d of water deficit in V3, this difference was increased to almost ⁓50% of stomatal aperture (Table 4). Although stomata represent a small percentage of the leaf lamina, large amounts of water evaporate through them33. Thus, the lack of stomatal control in TJ2049 might explain the fast-wilting phenotype displayed in response to air desiccation (Fig. 3). Measuring the water loss of detached leaves is a selection method for drought tolerance39. Here, R5 leaves of both genotypes were removed and air-dried. After 48 h, MUNASQA exhibited a greener, healthier and slow-wilting phenotype, previously linked to drought tolerance in studies with soybean cultivars17.
MUNASQA water-saving behaviour was also confirmed through parameters such as RWC and WUE, strongly regulated under drought (Table 5). The RWC, a time-specific measurement of the hydric status of a plant, is considered a physiological character recommended for drought-tolerance selection40. In V3 and R5 stages, well-irrigated MUNASQA plants showed higher RWC than TJ2049, even with a smaller canopy. As expected, under water deficit, the RWC values were reduced by 14% in MUNASQA and 20% in TJ2049, indirectly confirming the effectiveness of the stress imposed.
Water use efficiency (WUE), referring to the biomass produced per water unit, has been widely used as a breeding target in many rainfed crops, including soybean. Conservative water-use strategies are associated with high leaf WUE34. In agreement, and contrarily to TJ2049, V3 and R5 MUNASQA plants showed a gradual increase of WUE in response to water deficit that agrees with the tighter regulation of stomatal movements and the reduced water loss observed in the genotype (Table 5). Moreover, considering the discrete NAR reduction and the maintenance of a ~70% RWC under water deficit, we hypothesise that MUNASQA may display a stomatal control based on partial or total/partial closure intervals, therefore reducing transpiration and saving water through a smaller gas exchange (potential photosynthesis) penalty.
As a surrogate trait for stomatal conductance, the CTD, regarding the canopy temperature difference with the surrounding air, is a good indication of plant transpiration rate41. As expected, in response to drought, MUNASQA plants evidenced lower CTD values, a finding that supports the stomatal aperture results and strongly suggest the genotype water-saving behaviour. Plants with higher stomatal conductance transpire more and thus maintain a cooler canopy42. Thus, in TJ2049 stressed-plants, the high and positive CTD confirmed a higher stomatal aperture and transpiration rate that agrees with a water-spender behaviour. Moreover, TJ2049 also presented higher transpiration rates in unstressed conditions. Finding that could be evidence of a natural and predisposing difference between tolerant and susceptible genotypes.
3. Markers selection
Identifying and exploiting phenotyping markers will improve selection strategies for drought tolerance in legumes crops17. However, to successfully implement markers in a breeding program, it is imperative to validate their (i) accuracy, (ii) feasibility and (iii) strength of association with the desired trait. To further understand the markers contribution to drought tolerance and yield stabilization in MUNASQA and TJ2049, a Principal Component Analysis (PCA) was performed for each set of the parameters, previously grouped in (i) “stress response”, (ii) “growth” and iii) “water use” categories. In addition, markers evaluated at V3 and R5 stages were measured by Pearson's correlation to determine their strength of association between the phenological stages (accuracy) and yield stabilization (desired trait) (Tables 2, 3 and 5).
Biochemical parameters such as enzymatic and non-enzymatic ROS scavengers, leaf pigments and MDA have been confirmed as adaptive responses to desiccation stress, frequently used for selecting plant genotypes under drought43. Clear discrimination of MUNASQA and TJ2049 drought responses were observed in PCA results (Fig. 4A). Here, the first two principal components (PC) explained 96.6% of total variation (PC1 = 75.0% and PC2 = 21.6%). Data were clustered by irrigation treatment in PC1, suggesting that these markers are indicators of phenotypic plasticity. The CHL was associated with well-irrigated plants, while all enzymes, MDA, CAR and PRO, were related to drought stress (Fig.4A).
All these markers showed high accuracy between phenological stages due to significant (P < 0.001), strong (r2 over 0.88) and positive correlations (Table 2). However, the correlation with yield showed inconsistent outcomes. PRO, MDA and CAR were significant and positive correlated with yield. Meanwhile, except SOD, all enzymes exhibited significant and positive correlations that were too variable in the strength of association and could not be linked with a specific genotype or treatment. Thus, we considered these “stress-response” markers suitable for discriminating susceptible/resistant responses during early drought-tolerance screenings in soybean. Still, due to the high environmental effect, their use as breeding traits is limited. Interestingly, a report20 found that PRO and CHL were suitable markers for ranking soybean genotypes in response to drought in vegetative stages (5 days after emergence), while MDA could be useful during R5 as a sensitivity trait.
In legumes, features like LAI, smaller leaf area, leaf area maintenance and dry matter partitioning have been used to screen for drought tolerance17. Here, in the “growth” PCA (Fig. 4B), the first two PC explained 93.0% of total variation (PC1 = 73.0% and PC2 = 20.0%). In PC1, data were clustered by genotype, suggesting that these markers explained differences between MUNASQA and TJ2049 growth responses on their intrinsic genetic variability. Moreover, the markers analysed were only associated with MUNASQA, LAR and RGR associated with stressed plants. However, only LAI and LAR showed a weak correlation between phenological stage and yield (Table 3).
Water-saving features like denser leaf pubescence, a higher number of stomata, warmer canopies, RWC maintenance and increased WUE have been associated with drought tolerance in legumes and applied in drought-resistance breeding17. Here, “water-use” markers contributed the most to discriminating drought-tolerant and susceptible responses. In the two PCA performed, one for physiological markers and the other for morphological ones, data were clustered by genotype in PC1; thus, these markers are indicators of genetic variability.
In CTD, RWC and WUE PCA (Fig. 4C), the first two PC explained of total variation 87.7% (PC1 = 54.6% and PC2 = 33.1%). Markers RWC and WUE (associated with MUNASQA) and CTD (associated with TJ2049) showed significant (P < 0.001), strong (r2 over 0.87) and positive correlations between phenological stages (Table 5). Moreover, WUE and CTD were significantly associated with yield, showing positive and negative correlations depending on the water treatments applied (Table 5).
In the PCA made with morphological “water-use” markers, the first two PC explained 96.6% of total variation (PC1 = 73.8% and PC2 = 22.8%) (Fig. 4D). Here, pubescence and stomata abaxial density were strongly related to MUNASQA. Although the data were insufficient to execute good correlation analysis, these morphological markers have been demonstrated as clear indicators of water-saving strategies in legumes34.
A good drought-resistance marker linked to yield stabilization must also be accurate, cost-effective, if possible non-destructive and easily measurable. Hence, after evaluating marker accuracy (amid phenological stages) and assessing which ones better explained the phenotypic variability between genotypes and treatments, a final selection was performed by cost-feasibility (CF) and statistical weight (SW) rankings. Based on the CF values, the degree of complexity and cost to assess each indicator, 9 markers were further selected encompassing the categories 1 (easy and cheap) and 2 (easy and expensive) (Table 6). Subsequently, after SW re-selection, 4 markers remained with “High” SW in both PC. The selected markers were (i) stomatal density on the adaxial and (ii) abaxial leaf surface, (iii) trichrome density on the abaxial side and (iv) CTD. These four traits were categorised as the most efficient phenotyping markers considering their high accuracy, strong association to water-saving strategies, feasible and non-destructive measurement and amenability to high throughput screening using high-resolution imaging.
Taken together these results: (i) confirmed the effectiveness of the phenotyping methodology and (ii) its successful application in early phenological stages, (iii) the usefulness of MUNASQA and TJ2049 as model genotypes for screening potential drought-tolerance markers and, (iv) suggested that pubescence, stomatal density and CTD are informative markers to discriminate soybean adaptations to withstand drought via dehydration avoidance.
Traditional crops phenotyping is labour-intensive, time-consuming and often destructive. Hence, a well-characterized drought-tolerant phenotype could be useful for developing new commercial genotypes or identifying drought-tolerance markers. In this context, our findings confirm that MUNASQA and TJ2049 constitute suitable models for drought-tolerance screening in soybean. The deep-phenotyping results allowed us to ensure the methodology effectiveness and its possible implementation in early vegetative stages like V3, reducing time and money costs. Finally, identifying stomatal densities, abaxial pubescence and CTD as profitable and non-destructive markers, strongly associated with water-saving adaptation strategies, provides us with promising tools for early selection and prediction of drought tolerance in soybean.