2.1 Plant Growth Conditions

Soybean seeds (Glycine max L. Merr. var. Sumatra) were surface sterilized in 1% NaClO (v/v) and 0.01% SDS (w/v) for 40 min, washed with deionized water, and then placed in 0.01 N HCl for 10 min (Labhilili et al. 1995). Seeds were germinated and sown in 1 L pots containing a mixture of vermiculite: perlite (2:5, v/v). Seedlings were inoculated with 1 mL of a hup⁻ Bradyrhizobium diazoefficiens (previously named japonicum) strain UPM792 on days 0 and 3 after sowing. Plants were grown for 1 month under controlled conditions (24°C/18°C day/night temperature, 60%/70% day/night relative humidity, 16 h photoperiod and 450 μmol m⁻² s⁻¹ light intensity) and watered to field capacity twice a week with a N-free nutrient solution (Rigaud and Puppo 1975).

2.2 Experimental Design and Stable Isotope Labeling Method

A total of 72 soybean plants were grown for 4 weeks. The plants were separated into three groups. Two groups were subjected to a gradual water deficit by stopping irrigation, while the third group maintained a full water supply. Based on leaf water potential values (Ψ_leaf), we identified three categories: control, mild drought, and severe drought conditions. To facilitate harvesting at various times throughout the day, each category was further divided into three subgroups comprising eight plants each, with samples taken 2, 5, and 8 h after the start of the photoperiod. This corresponded to samples at t = 0, t = 3, and t = 6 h following sucrose application. Because sample collection is a destructive procedure, a separate set of plants was used for each time point.

To determine the effect of drought on ¹³C allocation, half of the plants were labeled with stable isotope-labeled sucrose, [U-¹³C]-sucrose (Sercon Ltd.), while the other half were treated with unlabeled sucrose to determine natural ¹³C abundance. Upon reaching the two levels of drought stress, both control and stressed plants were labeled (time = 0 h), corresponding to 2 h after the beginning of the photoperiod, following established protocols (Fondy and Geiger 1977). Briefly, the adaxial surface of the youngest fully expanded leaf was gently abraded with a fine-grained sandpaper. Next, 25 μL of a solution containing 20 mM [U-¹³C]-sucrose buffered in 5 mM potassium phosphate buffer (pH 6.5) was applied to the abraded area. Uptake of the sucrose solution was allowed to continue for 3 or 6 h. The following tissues were collected at these time points: labeled leaf, sink leaf (approximately 26% of full lamina length), other source leaves, stems, roots, and nodules. The samples were weighed, and one set of aliquots was immediately frozen in liquid nitrogen and stored at −80°C for analytical determination. The other set was dried at 70°C for 72 h, ground to a fine powder, and weighed into tin capsules for elemental analysis-isotope ratio mass spectrometry (EA-IRMS) and dry weight (DW) measurements. The same procedure was performed for the non-labeled plants.

2.3 Plant Physiological Characterization

To establish the effect of water stress on plants, non-destructive parameters including stomatal conductance, net photosynthesis, Ψ_leaf, and apparent nitrogenase activity (ANA) were measured. These parameters were measured 2 h after the beginning of the photoperiod in the youngest fully expanded leaf.

Stomatal conductance and net photosynthesis were measured in a portable open system mode (model LCpro+; ADC BioScientific Ltd.) using an ADC PLC-7504 leaf chamber. Ψ_leaf was measured using a pressure chamber (Scholander et al. 1965). Nitrogen fixation was measured as ANA (Witty and Minchin 1998) following the H₂ evolution of intact plants in an open flow-through system under N₂/O₂ (79%/21%) using an electrochemical H₂ sensor (Qubit Systems). The H₂ sensor was calibrated with high-purity gases using a gas mixer at the same rate as that of the sampling system (500 mL min⁻¹).

2.4 Carbon Isotope Analyses

The percentages of C and δ¹³C were determined using EA-IRMS in a DeltaPlus (ThermoQuest) coupled to an elemental analyser NC2500 (CarloErba, Italy). The results of the C isotope ratio analyses were expressed in delta notation. According to Farquhar et al. (1982), the C isotope composition with respect to Vienna Pee Dee Belemnite (VPDB) is δ¹³C (‰) = [(Rs/R_VPDB)−1] × 1000, where R is the ¹³C/¹²C ratio of the sample and standard, respectively. These results were used to obtain ¹³C (mg) transported to each tissue from the labeled source leaf. First, the calculations to obtain the % of ¹³C were performed according to Ariz et al. (2011) but using the R_VPDB standard. Subtracting the %¹³C of the tissues from labeled plants from that of the unlabeled plants and multiplying the %C and biomass of each tissue, the transported ¹³C (mg) was finally obtained as previously described (Parvin et al. 2020).

2.5 Metabolite Extraction and Determination

For the determination of soluble carbohydrates, 0.1 g aliquots of fresh tissue (including sink leaves, source leaves, stems, roots, and nodules) were extracted, as previously described (Marino et al. 2006). Carbohydrates were determined by ion-exchange chromatography in a 940 Professional IC Vario with an amperometric detector (Metrohm AG, Switzerland) equipped with a Metrosep Carb 2 Guard/4.0 + Metrosep Carb 2−150/4.0 columns. The samples were eluted with 300 mM NaOH/1 mM C₂H₃NaO₂ at 30°C.

To determine the organic acid content, 0.1 g aliquots of fresh tissue (including sink leaves, source leaves, stems, roots, and nodules) were extracted, as described by Gálvez et al. (2005). Anions and organic acids were determined by ion-exchange chromatography in a 940 Professional IC Vario with a conductivity detector (Metrohm AG) by gradient separation with a Metrosep A Supp16 150/4.0 column. Samples were eluted with solvent A (deionized water) and solvent B (20 mM Na₂CO₃ + 2.5 mM NaOH) at 55°C according to the manufacturer's instructions.

Lastly, the same extract used for the measurement of soluble carbohydrates was used to determine N compounds. High-performance capillary electrophoresis (P/ACE 5500; Beckman Coulter Instruments) was used to measure the content of ureides and asparagine according to the method described by Sato et al. (1998). A solution of 0.1 M Na₂B₄O₇ · 10H₂O (pH 9.2) in 25 mL L⁻¹ OFM-Anion BT (Waters Corp.) was used as the electrolyte in a fused-silica capillary tube 60 cm in length. The levels of ureides and asparagine were detected at an optical density of 190 nm.

2.6 Statistical Analysis

Data are presented as the mean ± SE of ≥ 3 biological replicates. Significance of the physiological parameters was analyzed by one-way ANOVA. The Shapiro–Wilk test and Levene's test were used to verify the sample's normal distribution and homogeneity of variances, respectively. In the case of no normality or homogeneity, a log transformation was applied.

Comparisons between each treatment and time point were performed using the least significant difference (LSD) or Dunnett's T3 test at p ≤ 0.05. Metabolite analysis and δ¹³C, total ¹³C, and ¹³C distributions were analyzed by two-way ANOVA. Comparisons between each treatment and time were performed using the LSD or Dunnett's T3 test. Differences were considered significant at p ≤ 0.05. The Perseus software (version 1.6.5.0; Tyanova et al. 2016) was used to represent the heatmap and cluster aggregation. Metabolite levels underwent a base-2 logarithmic transformation, followed by standardization using Z-scores. R was used to perform principal component analysis (PCA) using the prcomp() function on the dataset of metabolite content and ¹³C values for each tissue and drought level.

3.1 Water Deficit Effects on the Physiological Status of Soybean Plants

Plants were subjected to water deprivation for 5–7 days, and leaf water potential was monitored to establish two levels of drought stress: mild drought (Ψ_leaf = −0.44 ± 0.03 MPa) and severe drought (Ψ_leaf = −0.85 ± 0.08 MPa). Well-watered plants were taken as the control group (Ψ_leaf of −0.24 ± 0.05 MPa; Table 1). During the experiment, stomatal conductance remained relatively constant in control plants (0.31 ± 0.01 mol m⁻² s⁻¹), while drought stress provoked an 81% and 93.2% decline under mild and severe water deficit, respectively (Table 1). Net photosynthetic rates declined by 59% under mild drought stress conditions, whereas under more severe conditions, a 91.5% decline was observed (Table 1). The effect of water withholding on nitrogen fixation was measured by hydrogen evolution (ANA) in intact soybean plants. Under mild water stress, ANA was more negatively affected than photosynthesis or Ψ_leaf, declining by approximately 71.4% compared with control plants. Under severe water deficit conditions, almost complete inhibition of nitrogen fixation was observed, with an 88.8% decline (Table 1). These results are in line with those of previous studies in which nitrogen fixation was more negatively affected by mild drought stress than other physiological parameters.

3.2 Drought Stress Reduces Natural C Isotopic Composition in Source and Sink Leaves

In order to set a baseline for natural variation of C isotopic composition in soybean leaves, we analyzed (1) whether sampling at various time points during the day had an effect on δ¹³C values, and (2) whether plants subjected to different water regimes presented significant variations in δ¹³C values. Regarding the first variable, time-dependent variations in δ¹³C values were observed only under control conditions, leading to more negative values in the source leaves and roots at the end of the experiment (Figure 1A,D). No time effects were observed for the drought stress treatments (Figure 1). Regarding the effect of water status, in control plants, source leaves had more negative δ¹³C values than sink leaves, stems, nodules, or root samples, indicating ¹³C enrichment in non-photosynthetic tissues (Figure 1). Water withholding affects the C isotopic composition, resulting in less negative δ¹³C values in leaves and roots (Figure 1). In the source leaves, there was no significant variation in δ¹³C during the 6 h of the experiment under drought stress. In contrast, in control plants, δ¹³C changed towards more negative values, indicating a ¹³C loss in this tissue (Figure 1A). The effect of water stress on δ¹³C was more significant in sink leaves than in the source leaves. Moreover, the δ¹³C of sink leaves under severe stress became less negative over time, unlike the observed trend in severely stressed source leaves (Figure 1A,B). In the root samples, δ¹³C of drought-stressed plants remained constant over the course of the experiment, whereas δ¹³C of control roots declined over time (Figure 1D). In contrast, there was no significant effect of drought on the δ¹³C in stems and nodules (Figure 1C,E).

3.3 Water Deficit Changes C Allocation Among Source and Sink Tissues

Drought stress is a condition that has an impact on plant metabolism and physiology at several levels. To employ labeled sucrose as a proxy to estimate C allocation dynamics, it was necessary to test whether the total amount of ¹³C did not present variations due to water deficit or time of sampling. To do so, we measured the total amount of ¹³C, calculated as the sum of the ¹³C values in each tissue (Figure 2; Table S1). Neither water deficit nor time produced significant changes in the total amount of ¹³C in the plants during the experiment.

We then calculated the rate of transport of labeled sucrose applied to the source leaf across the plant. Figure 3 shows the percentage of ¹³C fixed in each tissue as a proportion of the total amount of ¹³C found in the soybean plants at each time point. In control plants, approximately 16.2% of ¹³C was transported to other tissues at 3 h post labeling, while 22.4% was distributed at 6 h (Figure 3A). During the first 3 h, the major fate of ¹³C transported was the roots (5.5%), followed by the sink leaves (5.4%); the third and fourth main locations were the nodules (3.9%) and stems (1.9%; Figure 3A). After 6 h of labeling, the strongest sink for ¹³C was found to be the sink leaves (15%), while the ¹³C transported to the remaining tissues was relatively lower compared to the previous time point (Figure 3A).

However, in plants subjected to mild water deficit, there was a reduction in the amount of ¹³C transported from the labeled leaves. At time 3 h, only 8.3% of the total ¹³C was transported, which increased to 14.5% after 6 h of labeling (Figure 3B2). At 3 h post labeling, the allocation of ¹³C was concentrated in the root (3.2%), followed by the stems (2.1%) and nodules (2.1%; Figure 3B). Transport to sink leaves was significantly reduced compared with the control conditions (only 0.1% of the transported ¹³C). The same trend was observed 6 h after labeling. Thus, the largest decline in transported ¹³C was observed in sink leaves, dropping from 5.4% to 0.1% of allocated C at 3 h and from 15% to 0.4% at 6 h after labeling (Figure 3A1,B1). In both roots and nodules, the amount of ¹³C transported under mild drought was slightly delayed but not significantly changed during stress (Figure 3A,B).

In plants under severe drought stress, the export of ¹³C was further limited and only 2.9% of the C was exported at 3 h, reaching a maximum of 8.2% of ¹³C transported at 6 h after labeling (Figure 3C2). As in mild stress, the amount of ¹³C transported to sink leaves was close to zero, while the sink strength in other tissues was maintained (Figure 3C). However, in nodules, there was a significant reduction in the amount of ¹³C allocated at 6 h post-labeling compared to that in control plants (Figure 3A6,C6).

Taken together, these results show that under drought stress, roots and, to a lesser extent, nodules maintain their relative sink capacity, whereas there is a reduction in the amount of C allocated to young, growing leaves.

3.4 The Concentration and Distribution of Plant Metabolites Were Affected by Both Drought Conditions and Type of Tissue

To obtain an overview of the main metabolic changes occurring in soybeans subjected to a progressive water deficit, we analyzed the levels of major C and N metabolic compounds in different plant tissues during the sampling period. Due to the complexity of the dataset, we ran a PCA to identify the variables most strongly associated with the drought treatment and different tissues (Figure 4A). PCA differentiated the data into four groups: source leaves, sink leaves, and nodules clustered largely independently, whereas stem and root samples were grouped together. PCA showed that drought treatment predominantly influenced the distribution of metabolites and ¹³C allocation in different tissues, with sampling time having a lesser impact. The first PCA component accounted for 29.2% of the variance and was responsible for the differentiation between the sink leaves and root/stem groups. PCA1 was influenced most by starch and raffinose, strongly correlated with the stem/root group and polyols and organic acids on the side of the sink leaves. PCA2, which accounted for 20.8% of the variance, separated the source leaves from nodules. In this separation, the distribution of hexoses during drought was correlated with the source leaves, whereas trehalose, ureides, and ¹³C allocation were associated with nodule samples.

To visualize and establish a hierarchy of the main metabolic changes occurring across the plant under different drought conditions, we plotted the normalized (log₂) levels of the metabolites in a heatmap (Figure 4B). The levels of starch showed a unique profile (red cluster), showing relatively high but stable levels (not affected by water deficit) in source leaves, stems, and, to a lesser extent, nodules, while the levels in roots and particularly sink leaves were relatively lower. The second cluster (orange) included mono- and disaccharides such as sucrose, glucose, and fructose, as well as the main N compounds exported in soybean, allantoin, and allantoic acid. Drought caused a progressive accumulation of these compounds in practically all the tissues tested, with the exception of allantoic acid in the source leaves, whose levels were reduced under water deficit. Interestingly, raffinose was detected mostly in the stems, whereas trehalose was found only in the nodule tissue. Metabolites in the yellow cluster included organic acids, such as oxalate, acetate, and lactate, all of which showed a drought-induced decrease, particularly in nodules. Lastly, the green cluster grouped metabolites whose distribution depended to a greater extent on the plant tissue tested, such as organic acids and polyols. This representation also showed that the effect of drought stress was greater than the effect of harvesting time on the distribution of the detected metabolites. After this outline, to expand the analysis, the most relevant changes for each tissue are summarized below. The values corresponding to each metabolite are provided as Tables S2–S6.

Drought caused a general increase in carbohydrates detected in the source leaves, with a higher accumulation in hexose content (Table S2). In the case of sucrose, its content remained stable under mild drought but increased under more severe stress. Regarding nitrogen compounds, drought caused a decrease in ureide concentrations, and allantoin was not detected under severe drought (Table S2). Instead, asparagine concentration increased with time and drought stress, reaching values of 8.4 ± 2.0 and 13.0 ± 3.2 μmol g⁻¹ DW at 6 h under mild and severe stress, respectively. The organic acid content decreased, with malonate being the most negatively affected, with a decrease of approximately 85% and 70% under mild and severe stress at time 0 h, respectively (Table S2). Under control conditions, the malate content increased during the experiment with 33.31 ± 3.4 μmol g⁻¹ DW at 6 h, but its concentration decreased due to drought. Regarding polyols, water deficit increased the myo-inositol content (Table S2).

In sink leaves, both levels of drought and time also caused an increase in carbohydrates, except for starch, whose concentration was almost negligible (Table S3). Under mild drought conditions, glucose and sucrose levels decreased during the experiment, while severe drought changed the trend towards the highest values (Table S3). The ureide content varied with harvesting time and drought level. In contrast, asparagine concentration increased as drought stress increased (Table S3). Regarding organic acids, unlike the trend observed in source leaves, drought increased malate content in sink leaves, presenting values 4.6- and 7.5-fold higher than in source leaves under mild and severe stress at 6 h, respectively. Moreover, the highest pinitol content was found in sink leaves. Drought induced the accumulation of polyols, except sorbitol, the content of which decreased under severe stress. The greatest increase was observed in the myo-inositol concentrations under both drought conditions (Table S3).

In the stem, hexose concentration increased under severe drought conditions (Table S4). Raffinose, which was only found in stems and nodules in this study (Tables S4 and S6), and sucrose accumulated due to drought, and this stress caused their accumulation during the experiment. This trend was also observed for ureide content (Table S4). Asparagine reached values more than 6-fold under both drought conditions, depending on time (Table S4). Within the organic acids group, there was a general decrease in their content under mild or severe stress (Table S4). Regarding polyols, only the myo-inositol content increased under mild or severe stress (Table S4).

In the roots, there was a general increase in the concentration of carbohydrates due to drought, with hexoses showing the largest relative increase (Table S5). Only the starch content remained stable. The concentration of N compounds was almost negligible in the control roots, but there was a large increase under drought (Table S5). A different trend was observed in organic acids, where water deficit increased the content of malonate and oxalate, and the contents of citrate and lactate were reduced (Table S5). Drought increased the concentrations of myo-inositol and pinitol (Table S5).

In general, drought increased the concentration of all detected carbohydrates in nodules, except for raffinose, which was only detected under control conditions (Table S6). The concentration of ureides increased as drought increased, although a decrease in allantoin content was observed under severe stress conditions at 3 and 6 h (Table S6). The asparagine content was higher than that in the other tissues and increased due to the water deficit (Table S6). Regarding organic acids, drought stress caused a general decrease, with the effect on malate content being more significant, which decreased around 50% under both mild and severe stress (Table S6). As for polyols, nodules had the highest content of myo-inositol, although drought did not have any effect on their concentration. In contrast, there was an increase in pinitol content, while sorbitol content decreased (Table S6).

4.1 Soybean Reduces C Allocation to Sink Leaves, Prioritizing Roots and Nodules Under Water Deficit Conditions

Progressive drought stress imposed on nodulated soybean plants caused changes in the long-distance transport of C. In well-watered plants, sink leaves were the dominant sink tissue, receiving 15% of the total translocated ¹³C (Figure 3A). In contrast, under drought stress, translocation towards this tissue was almost completely inhibited (Figure 3B,C), while ¹³C allocation to stems and roots was maintained (Figure 3C). It is well known that during a water deficit period, plants arrest their growth (Schubert et al. 1995; Lemoine et al. 2013), allocating C resources to the root system to ultimately improve the plant capacity to acquire water (Geiger et al. 1996; Palta and Gregory 1997; Arndt and Wanek 2002). This C allocation trend was also observed in the metabolic profiling of roots and nodules, in which carbohydrate accumulation was detected (Figure 4B; Table S5). Similar results were obtained by Purcell et al. (1997), who reported a higher tolerance to drought associated with continued allocation of photosynthates to nodules. However, under severe drought, ¹³C translocation to the nodules decreased (Figure 3C). Similarly, in faba bean, the sink activity of nodules was higher than that of other tissues, although this did not lead to increased ¹³C incorporation under drought conditions (Parvin et al. 2020).

4.2 Drought-Stressed Soybean Nodules Show a Progressive Accumulation of Sucrose While Malate Levels Are Reduced

Metabolite analysis showed increased levels of carbohydrates (sucrose, fructose, and glucose) in the nodules and sink leaves under water deficit conditions (Figure 4B). In the case of nodules, based on ¹³C measurements, the transport of sucrose was maintained despite the water deficit, although to a lesser extent than in the control plants. This observation, together with the fact that malate, the final C product used for bacteroid respiration, accumulates during drought stress in nodules, supports the hypothesis that the regulation of nitrogen fixation by C is associated with a decline in sucrose synthase activity (González et al. 1995; Gálvez et al. 2005). Indeed, the ¹³C content was found to be inversely correlated with hexose and sucrose contents in the PCA analysis (Figure 4A), whereas it was positively correlated with the levels of ureides and lactate, which are relevant metabolites in nodules.

In legumes, the accumulation of N compounds has been suggested to play a role in the feedback inhibition of nitrogen fixation, including that of ureides (Atkins et al. 1992; Serraj et al. 1999; Vadez et al. 2000), and asparagine (Bacanamwo and Harper 1997). In this study, drought caused the accumulation of ureides and asparagine in both roots and nodules (Figure 4B; Tables S5 and S6). Instead, in the source leaves, the concentration of ureides declined until they were undetectable under severe stress (Table S2). These results suggest that the observed inhibition of nitrogen fixation is likely not related to the accumulation of ureides in leaves, in line with previous studies (King and Purcell 2005; Ladrera et al. 2007). The role of asparagine in the inhibition of nitrogen fixation has also been questioned, as a general increase in single amino acid content observed in both nodules and leaves suggests more complex N signal regulation (Gil-Quintana, Larrainzar, Arrese-Igor, et al. 2013). Interestingly, in sink leaves, ureide content remained stable under severe stress (Figure 4B; Table S3). These results suggest that different urea catabolism pathways act in the various tissues analyzed rather than impairing long-distance transport from nodules to shoots. Gil-Quintana, Larrainzar, Seminario, et al. (2013) showed that urea catabolism is more negatively affected than de novo synthesis under water-deficit conditions. Thus, our results support the co-existence of both malate limitation and N compound accumulation regulatory mechanisms in soybean.

4.3 Drought Stress Alters the Levels of Organic Acids and Polyols in Different Tissues Allowing to Establish a Metabolic Fingerprint

Organic acids play a major role in providing redox equilibrium in plants (for a review, see Igamberdiev and Eprintsev 2016). In our study, drought stress, a type of oxidative stress, provoked differential variation in the levels of organic acids across plants. In contrast to the above-mentioned decline in malate in nodules, both in the sink and source leaves, malate was found to accumulate. Interestingly, malate accumulation in the cytosol or vacuoles of guard cells is related to the regulation of stomatal movement (Hedrich and Marten 1993). In contrast to the well-established role of malate in nodules, malonate is described as a poor C source for bacteroids (Karunakaran et al. 2013), and its function in this symbiotic organ remains unknown (Booth et al. 2021). Li and Copeland (2000) suggested that its metabolism is not related to abiotic stress, but rather a consequential effect related to malate metabolism. The variation in the contents of these two organic acids observed in the current study seems to agree with this suggestion (Figure 4B; Table S6). Trinchant and Rigaud (1996) suggested that oxalate oxidation is important in nodules, preventing the production of ROS near bacteroids, which could explain the decline in oxalate content in drought-stressed nodules (Table S6).

Several studies have highlighted the role of polyols in osmotic regulation under drought conditions (Noiraud et al. 2001; Streeter et al. 2001; Dumschott et al. 2019). Although sorbitol synthesis has also been described in nodules, polyol synthesis occurs mainly in source leaves (Colebatch et al. 2004). Thus, its occurrence in sink tissues indicates its likely translocation (Noiraud et al. 2001; Streeter et al. 2001). In the current study, we observed that drought altered polyol levels differently, depending on the tissue analyzed. For instance, the concentrations of mannitol and sorbitol in sink leaves showed an effect of drought prior to this in source leaves, indicating both lower assimilation in sink leaves and transport impairment under severe stress (Figure 4B; Tables S2 and S3). The PCA results also highlighted the association of polyol levels as part of the drought response of sink leaves. The biosynthesis of pinitol from glucose-6-phosphate includes myo-inositol biosynthesis (Dumschott et al. 2019). In the source leaves and roots, there was an accumulation of myo-inositol, but no effect on pinitol, whereas in nodules, there was no effect on myo-inositol, although the pinitol content increased. This suggests that pinitol synthesis is regulated during drought, or that its translocation to sink tissues increases.