3.1. Chlorophyll and relative water content
After 12 DWD and 14 DWD the photosynthetic pigment (chlorophyll content) index was higher in Genetically Modified Plants (GMP; 1Ea15 and 1Ea2939) compared to non-GMP (BR16) (Fig. 1C-D), suggesting a possible adaptive response of the GM genotypes.
Another important parameter analysed related to water deficit was the leaf relative water content (RWC), which showed a significant reduction of 50%, 15%, and 15% in BR16, 1Ea2939, and 1Ea15 genotypes, respectively, after 12 DWD compared to control conditions, and 50%, 25%, and 60% after 14 DWD compared to control conditions (Fig. 1). Decreased RWC is an early response to water deficit and represents variations in osmotic adjustment, as previously observed in the drought-sensitive BR16 genotype (Isabel D Coutinho et al. 2017). When comparing the RWC of sensitive genotype to GMP, the RWC of GMP showed a smaller decrease, mainly after 12 DWD. The visual phenotype of the drought-sensitive genotype compared to the tolerant one was compatible with typical symptoms of water deficiency (Fig. 1a-b).
3.2. Water limitation effects on metabolite profiles of soybean leaves and roots
A total of 31 metabolites were identified in soybean leaves and root extracts, including various carbohydrate metabolites (such as sucrose, glucose, pinitol, and myo-inositol), amino acids (such as leucine, isoleucine, valine, proline, glycine, phenylalanine, tyrosine, alanine, threonine, asparagine, lysine, and GABA), organic acids (such as citrate, malate, lactate, acetate, succinate, glutamate, formate, and fumarate), phenolic compounds (such as cis-coumaroylquinic acid, trans-coumaroylquinic acid, and three kaempferol derivatives), as well as choline and trigonelline.
To provide a comparative interpretation and visualization of metabolic changes among soybean genotypes, the supervised Partial Least Squares-Discriminant Analysis (PLS-DA) method was applied to the 1H Nuclear Magnetic Resonance (NMR) data. The study included soybean leaf and root samples from three genotypes, including one drought-sensitive conventional cultivar (BR16) and two drought-tolerant GMP (1Ea15 and 1Ea2939) soybean genotypes. Metabolic differences between the sensitive and GMP were visualized using two latent variables, which revealed that the metabolic profiling of the BR16 genotype was significantly altered in response to drought stress compared to the GMP (Fig. 2).
Based on Fig. 2A, the PLS-DA score plot of two latent variables clearly showed discrimination between soybean genotypes under water deficit and control conditions. The differences between BR16 and GMP genetically modified (GM) leaves were more distinct in response to water deficit, with samples located in positive LV1 showing high correlation with signals corresponding to amino acids and sugars (as shown in the loadings plot LV1). Interestingly, the LV2 led to clustering based on days of harvest, with genotypes harvested at 12 DWD located at negative LV1, while samples submitted to 14 DWD were clustered at positive LV1. The loadings plot in Fig. S1 revealed the contribution of different variables to the discriminating ability of each PLS-DA model. Both latent variables one and two from PC1 and PC2 indicated clear metabolic variation between well-watered and drought conditions. According to the loadings plot LV2, the main difference between days of stress was due to the oscillation of patterns of signals corresponding to citric acid, sugars, and phenolic compounds.
Similarly, the PLS-DA model from 1H NMR spectra of soybean root polar extracts (Fig. 2B) showed a similar pattern to the soybean leaves model in terms of stress conditions. The discrimination was mainly due to an increase in sucrose levels and a decrease in citric acid and isoflavone levels in response to drought stress (as shown in Fig. S1). Furthermore, the oscillation in isoflavone levels appeared to be an important target in the root's response to water deficit. The discrimination between GM soybean and BR16 cultivars was clear in soybean roots under 14 DWD, with samples located at positive LV1 and negative LV2.
3.3. Accumulation of putative osmolytes
The direct analysis by NMR offers the advantage of detecting a wide range of metabolites in a quantitative and unbiased manner. In this study, the relative content of amino acids, sugars, organic acids, and phenolic compounds was monitored to evaluate the fluctuations in the metabolome of soybean leaves and roots under different stress periods (12 and 14 DWD), and compared with the respective control. We observed in both leaves and roots that the drought stress significantly altered the levels of most of the metabolites, including carbohydrate (Fig. 3), amino acid content in leaves (Fig. 4) and other metabolites (Fig. S1 - S3). The simple effects of water status and genotype are shown in Tables S1 and S2, respectively. Taken together, these results indicated that water limitation led to the accumulation of several metabolites during the stress period.
The results showed significant interactions between stress conditions and genotypes for several metabolites, particularly in the sensitive genotype BR 16, as shown in Fig. 3 and Fig. 4. Genotype effects were also observed in both leaves and roots. The sensitive genotype BR 16 exhibited higher accumulation of several compounds compared to the GMP (Table S2). Specifically, a remarkable accumulation of amino acids was observed in both leaves and roots of BR 16 cultivar after 12 DWD (Fig. 4 and Fig. S2). In contrast, the tolerant plants showed a smaller accumulation of amino acids and sugars, with a decrease in alanine levels in GMP in response to water limitation (Fig. 3 and Fig. 4). Additionally, differences were noted in the pattern of glucose and fructose accumulation, with glucose levels being higher in the 1Ea2939 genotype compared to BR 16 and 1Ea15 after 14 DWD (Fig. 3). Notably, sucrose levels decreased in GMP after 12 DWD, while the opposite trend was observed in the sensitive cultivar (Fig. 3). However, after 14 DWD, sucrose accumulation increased in all genotypes.
In the leaves, the effect of stress on genetically modified soybean was lower compared to the sensitive cultivar, with significant variations in levels of amino acids, glucose, and sucrose observed mainly after 14 DWD (Fig. 3 and Fig. 4). Organic acids also showed differences among genotypes, with slight metabolic fluctuations in acetate and fumarate levels, and notable accumulations of citrate, glutamate, succinate, and lactate (Fig S3 and Fig. S4). Malic and succinic acid increased in BR 16 and 1Ea15 genotypes but decreased in the 1Ea2939 genotype du.ring water limitation. Citric acid content was generally lower in response to water deficiency, while lactate levels increased in all genotypes. Organic sugars such as pinitol and myo-inositol increased in all genotypes under drought conditions, along with choline, trigonelline, and 2-hydroxyisobutyrate. These findings highlight the significant impact of water deficit on the metabolite profiles of soybean leaves and roots, and the differential responses of genotypes, including genetically modified plants, to water deficit conditions.
In the roots, the levels of glucose and sucrose were significantly altered in plants subjected to water deficit conditions. In general, the levels of amino acids and sugars increased after 12 and 14 DWD, except for alanine in 1Ea2939 after 12 DWD, as previously observed in the leaves' metabolome. Regarding organic acids, quantitative changes were observed in the tricarboxylic acid cycle in response to drought stress. The levels of citric and succinic acid were significantly reduced in the roots after 12 DWD. However, a decrease in citrate levels was observed in both roots and leaves after 14 DWD. The abundance of myo-inositol was altered by drought stress, with opposite patterns observed in leaves and roots (Fig. S3 and Fig. S4). Decreased levels were observed in roots, while increased levels were observed in leaves. Choline levels increased in both leaves and roots, but in roots but tended to decrease in the 1Ea15 genotype (Fig. S3 and Fig. S4).
The relative concentrations of isoflavones, hydroxycinnamic acids, and kaempferol derivatives were determined based on specific areas of the corresponding signals in the 1H spectra. The concentration of Daidzin decreased in the roots in response to water deficiency in all genotypes after 12 DWD (Fig. S5). Interestingly, the levels of Daidzein decreased in the 1Ea2939 genotype, but increased in the 1Ea15 genotype (Fig. S5). Hydroxycinnamic acid and kaempferol glycosides were only detected in soybean leaves. The changes in the profile of hydroxycinnamic acids were quite similar among genotypes, but the adjustments in kaempferol levels were different. Drought sensitive genotypes showed higher concentrations of kaempferol glucosides in control conditions compared to GMP, but these levels significantly decreased in response to water limitation. On the other hand, in GMP, the abundance of kaempferol increased in response to water deficit (Fig. S6).
3.4. Polysaccharide and lipid profiles by SSNMR spectroscopy
SSNMR experiments were conducted using the quantitative MultiCP pulse sequence to identify the relationship between the spectra profiles of lyophilized soybean leaves and roots in terms of cellulose and starch variability. Figure 5 provide an overview of the 13C signal shape dependence from these polysaccharides recorded at different ratios, the main variability assigned to real samples spectra, while the overall tendency after DWD conditions for each genotype is shown in Fig. 6.
As the leaves and roots are predominantly composed and structured by lignocellulose and starch at varying ratios, prior signal assignments to the 13C NMR spectra of pure and standard compounds seemed advantageous, considering that the signals of the glucopyranose ring, which are present in both cellulose and starch, are commonly overlapped. To achieve the correct assignment of cellulose and starch contributions in the compositions of leaves and roots, a specific weight of pure compounds was mixed in predetermined proportions. The respective 13C NMR spectra of cellulose, starch and their mixtures are shown in Fig. s7. It is worth noting that the anomeric carbon C1 was useful for quantifying the percentage of the compounds, and the deconvoluted values closely matched the weight proportions. Although the other 13C signals also vary with respect to the compound content, the relative distinction is biased, and the quantification through these signals is at least doubtful due to complete signal overlap across a broad range, which is not the case for the C1 signal. An additional morphological dependence, which is evident in all cellulose and starch signals, also plays a role in the correct assignment and contribution in a given mixture of the two polysaccharides. It should be considered that the crystallinity index inside the leaves and roots is unknown. While the C1 carbon signal from cellulose shows no significant variation, the one from starch starch is severely affected by morphological variations, as can be seen in Fig. S8. Fortunately, despite this clear dependence, the deconvolution procedure was able to overcome this issue, as the C1 signal overlap between cellulose and starch is negligible. Therefore, C1 was employed to accurately quantify the levels of cellulose and starch consumption in the leaves and roots of GMP and Br16 samples.
Figure 5 shows 13C NMR spectra normalised to δ 102–105 ppm signal intensity of the lyophilized soybean leaves and roots under water deficit and control. The spectral regions were attributed according to literature (Carvalho et al. 2009; Dick-Pérez et al. 2011; Hatcher 1987; Iulianelli and Tavares 2016; Lemma et al. 2007; Mutungi et al. 2012) (Knicker and Lüdemann 1995) and starch and cellulose standards. Six different chemical shift regions were clearly observed in both spectra (leaf and root). In leaves spectra, signals δ 101–105 ppm were assigned to C-1 of the starch glycosidic unit, the δ 72 ppm signal was assigned to carbons 2, 3 and 5, signal δ 83 ppm for C-4 and δ 62 ppm corresponding to C-6. The signals δ 120–140 ppm were assigned to the aromatic carbons of lignin and proteins. The region between δ 165 ppm and 180 ppm was attributed to the acyl functional group (C = O) of aliphatic carboxylic acids, esters and amide groups of proteins (δ 173 ppm). The main qualitative difference between the 13C spectra of leaves and roots of lyophilized soybean is related to the pattern of C1, C4 and C6 atom signals. Signals from δ 105 ppm to δ 60 ppm were attributed to starch in leaves, while the same region was attributed to cellulose in soybean roots. According to the deconvolution of the C1 carbon signal, soybean leaves clearly show a major content of starch, whereas cellulose plays this whole in roots.
The deconvolution procedure was applied to all samples, and a similar trend was observed regardless of the genotype, as mentioned earlier (Fig. 7). However, the gap between cellulose and starch content appeared to decrease from 12 DWD and 14 DWD) in leaves, rather than in the roots. This overall trend supports the previous finding in Fig. 3, indicating an increase in the levels of sugar molecules, likely resulting from the hydrolysis of cellulose and starch.
1H NMR experiments were conducted using the 1H ECHO-MAS pulse sequence to obtain spectral profiles of compounds with lower mobility, such as lipids (Fig. S9). The signals corresponding to methyl and methylene groups, as well as signals at δ 5.15 ppm and 5.37 ppm, were assigned to 1H in unsaturated linkages (Table S3). No signal was observed at δ 4.2 ppm, indicating the absence of glycerol units, triacylglycerols, and phospholipids (Forato et al. 2000), which suggests the presence of free fatty acids. Additionally, the intense signal at δ 0.97 ppm is likely attributed to methyl groups of the terpene unit. The main difference observed in the 1H NMR spectra of lyophilized soybean leaves between conventional genotypes and PGM under stress and control conditions is the lower intensity of spectral lines relative to the BR16 conventional cultivar in response to water deficit.