Experimental conditions and experimental design. Soybean cultivars NS7209 IPRO (Nidera Seeds, São Paulo, Brazil), NS7011 IPRO (Nidera Seeds, São Paulo, Brazil), Desafio 8473 RSF (Brasmax Seeds, Passo Fundo, Brazil) and 7739 M IPRO (Monsoy Seeds, São Paulo, Brazil) were grown in growth chambers (Instalafrio, Pinhais, PR, Brazil) with controlled conditions of relative humidity (~65%), irradiance (~650 µmol m−2 s−1) and temperature (25/20°C, day/night). The four soybean cultivars were selected due to their characteristics of high productivity, and wide cultivation in Brazil.
Plants were grown in polyethylene pots containing 8 Kg of substrate prepared from Red Latosol (LVdf) soil and sand (2:1). The subtrate used had the following composition: pH CaCl2 – 5.6; P – 0.7 mg dm−3; K – 13.0 mg dm−3; Ca – 1.54 cmolc dm−3; Mg – 0.22 cmolc dm−3; Al – 0.05 cmolc dm−3; H+ Al – 1.3 cmolc dm−3; S – 3.5 mg dm−3; B – 0.8 mg dm−3; Cu – 1.0 mg dm−3; Fe – 37.8 mg dm−3; Mn – 13.2 mg dm−3; Zn – 0.1 mg dm−3; Na – 6.0 mg dm−3; SB – 58%; CTC – 3.1 cmolc dm−3; organic matter – 109%. Based on these characteristics, liming was performed using dolomitic limestone, increasing the base saturation to 60%. The plants were fertilized with 0.2 g dm−3 of mono-ammonium phosphate (MAP), 0.16 g dm−3 of potassium chloride (KCl), 0.18 g dm−3 of potassium sulphate (K2SO4), 0.2 g dm−3 of urea (CH4N2O) and 0.026 g dm−3 of zinc sulfate (ZnSO4), according to the recommendation for Cerrado soils in which soybean are cultivated 86. Two plants were grown per pot, representing one experimental unit.
Treatments consisted of a combination of two water replacements (100 and 40% of soil holding water capacity, HWC) and two temperatures (25/20°C or 40/25°C, day/night) imposed at the V3 development stage. The control of soil moisture in the pots was done using the gravimetric method, by replacing the water lost by evapotranspiration in a daily basis. The high temperature (HT) was imposed when the plants reached the water deficit by gradually increasing from 25 ºC at 10 h until reaching 40 ºC ± 0.5 ºC at 11:30 h, maintained for five hours. After this period, the temperature gradually decreased until it returned to 25 ºC at 18:30 h, repeating this cycle in the following day. Thermal treatments were imposed for an 8-day period, during which time physiological, biochemical, metabolic and morphoanatomical evaluations were made.
The experimental design was arranged in randomized blocks, in a factorial design with two water replacements (100% and 40% HWC) and two temperatures (25 ºC and 40 ºC), with five replicates.
Water relations and leaf temperature. Leaf water potential (Ѱw(am)) was measured using a Scholander pressure chamber (Model 3005-1412, Soilmoisture Equipment Corp., Goleta, CA, USA). Osmotic leaf (Ѱs(leaf)) and root (Ψs(root)) potentials were determined using a vapor pressure osmometer (VAPRO 5600, ELITech, Puteaux, France) and calculated using the Van't Hoff equation, expressed as MPa. Leaf temperature was measured using a digital infrared thermometer (Model TI-920, Instrutherm Ltda, São Paulo, SP, Brazil), approximately 15 cm from the leaf limb. The measurements were performed at 09:00 h, with the chamber at 25°C.
Gas exchange and chlorophyll a fluorescence analysis. Gas exchange was measured in fully expanded leaves to determine the photosynthetic rate (A, µmol CO2 m− 2 s− 1), stomatal conductance (gs, mol H2O m− 2 s− 1), transpiration rate (E, mmol H2O m− 2 s− 1), and the ratio of internal and external CO2 concentration (Ci/Ca). We further estimate intrinsic water-use efficiency (A/gs) and instantaneous carboxylation efficiency (A/Ci,). The measurements were performed using an infrared gas analyzer (IRGA, LI-6400xrt, Licor®, Nebraska, USA), under constant photosynthetically active radiation (PAR, 1000 µmol m− 2 s− 1) at the environmental CO2 concentration (~430 µmol mol− 1), temperature (~25 ºC), and relative humidity (~65%). Chlorophyll a fluorescence was measured using the IRGA coupled to a leaf chamber fluorometer (6400xt, Licor®, Nebraska, USA). The minimal chlorophyll fluorescence (F0) and maximum quantum yield of photosystem II (PSII) (Fv/Fm) was measured after 30 minutes of dark adaptation. In light-adapted leaves, the apparent electron transport rate (ETR), the fraction of opened PSII reaction centers (qL), the effective quantum yield of PSII (YII), and the yield of non-photochemical quenching (YNPQ) were obtained.
Determination of antioxidant enzyme activities. To determine the activities of superoxide dismutase (SOD), catalase (CAT), ascorbate peroxidase (APX), and total peroxidase (POX), fresh leaf tissue was homogenized in potassium phosphate buffer solution (pH 6.8). SOD activity was determined by measuring the ability to photochemically reduce p-nitrotetrazolium blue (NBT), at 560 nm in a spectrophotometer (Evolution 60S, Thermo Fisher Scientific Inc., MA, USA), according to (Del Longo et al., 1993), considering that one SOD unit was defined as the amount of enzyme required to inhibit NBT photoreduction by 50%. CAT activity was assayed according to the method described byHavir and McHale 88 and calculated as the rate of hydrogen peroxide (H2O2) decomposition at 240 nm for 3 min at 25 ºC, using a molar extinction coefficient of 36 M−1 cm−1. APX activity was determined according toNakano and Asada 89 and was measured as the rate of ascorbate oxidation at 290 nm within 1 min at 25°C, using a molar extinction coefficient of 0.0028 M−1 cm−1. POX activity was measured following the method described by Kar and Mishra 90. Purpurogallin production was determined by increasing the absorbance of reaction at 420 nm for 1 min at 25°C, using the extinction coefficient of 2.47 mM−1 cm−1 91. Enzyme activity was expressed based on protein, the concentration of which was determined according to the Bradford method.
Leaf morphoanatomical characterization. Samples (~3 cm2) from the middle region of the last fully expanded leaf were collected and fixed in Karnovsky solution. After 24 h, the material was pre-washed in phosphate buffer and dehydrated in a gradual ethyl alcohol series, pre-infiltered and infiltered in historesin (Leica, Germany), according to the manufacturer’s recommendation. Samples were then transversely sectioned to 5-µm thickness in a rotating microtome (1508R model, Logen Scientific, China). Sections were stained with toluidine blue (0.05% in 0.1 M phosphate buffer, pH 6.8) and anatomical observations of adaxial and abaxial epidermis, palisade and spongy parenchyma and mesophyll were made using images photographed with an Olympus microscope (BX61, Olympus, Tokyo, Japan) coupled to a DP-72 camera using the light-field option. The micromorphometry measurements were obtained from the previous images using ImageJ software (Image Processing and Analysis in Java, v.147, USA). Ten observations per replicate were measured for each structure evaluated.
Growth analysis. Growth parameters such as plant height (PH, cm), stem diameter (SD, mm), and leaf area (LA) were determined. Shoot (leaves and stem) and roots were separated and dried at 65 ºC for 72 h to obtain the shoot dry matter (SDM, g) and root dry matter (RDM, g). The ration RDM/SDM were also calculated.
Metabolic extraction and analysis. Leaf samples were collected and immediately immersed in liquid nitrogen, and later homogenized and lyophilized. For metabolites extraction, the samples were microwave dried to prevent metabolic turnover92 and then approximately 40 mg of dry leaf samples was extracted in methanol/chloroform/water (12:5:3, v/v/v) at 75 ºC for 30 min. The water fraction of the extraction mixture consisted of a 0.1% solution of internal standard. Samples were centrifuged and the supernatants were collected and mixed to chloroform and Milli-Q water to facilitate phase separation. The water–methanol soluble fractions were collected and stored at -20 ºC for further analysis.
Analysis of soluble carbohydrates, sugars and organic acids were performed using gas chromatography (Agilent 7890A) coupled to triple quadrupole mass spectrometer (7000 Agilent Technologies Inc, Santa Clara, CA, USA). A HP5 column (0.25-mm internal diameter, 30-m long, 0.25-µm film thickness) was used for chromatographic separation. Extracts were dried down in a SpeedVac and resuspended in 400ul of anhydrous pyridine. Samples were then derivatised using a 1:10 mixture of N,O-Bis(trimethylsilyl)trifluoroacetamide (BSTFA): trimethylcholorsilane (TMCS). Samples were incubated for 35 min at 75 ºC and were analysed using GC-MS within 24 h. A 20:1 split injection was made at 300 ºC initial oven temperature program of 60 ºC for 2 min, ramping to 220 ºC at 10 ºC min−1 (hold for 5 min) then ramping at 10 ºC min−1 to 300 ºC (hold for 5 min). Peak integration was made using Agilent MassHunter software (Agilent). A mixed standard was made from a stock solution containing 500 µg mL−1 of each analyte. Appropriate aliquots were taken to make standard concentrations between 0.5 and 50 µg mL−1. The results were expressed based on dry weight.
Analysis of amino acids was performed using the underivatized extracts on a 1290 Infinity liquid chromatography (LC)-MS system (6520 QTOF, Agilent Technologies Inc, Santa Clara, CA, USA). A 3.5 µL sample was injected into a Zorbax SB-C18 column (2.1 x 150 mm, 3.5 µm) and separation was achieved by gradient elution with water and methanol. The QTOF was tuned to operate at low-mass range (<1700 AMU). Data acquisition was performed in scan mode (60–1700 m/z) and ionization performed in positive ion mode. LC-MS results were identified based on their retention times relative to standards as well as their formula mass. Peaks were integrated and their relative quantities were calculated using MassHunter software (Agilent®). A mixed standard was made from a stock solution containing 500 µg mL− 1 of each analyte. The solutions were kept frozen at -20 ºC. Appropriate aliquots are taken to make resulting standard concentrations between 0.1 to 20 µg mL− 1. The results were expressed based on dry weight.
Metabolic network analysis. Metabolomics and physiological parameters data were used to create correlation-based networks, in which the nodes are the metabolites and the links are the strength of debiased sparse partial correlation (DSPC) coefficient (r) 93. The correlation-based networks were created by using Metscape on CYTOSCAPE94,95 and limiting the significance of correlation between nodes to P < 0.05 or by restricting r values to -0.5 < r > 0.5 96. The parameters network density, network heterogeneity and network centralization were obtained using the java plugin NetworkAnalyzer on CYTOSCAPE software 97.
Statistical analysis. The data were subjected to factorial analysis of variance and the means were compared using the Tukey test (p < 0.05), using Analysis System Program Variance (SISVAR, version 5.4). The metabolomics data were analysed by multivariate analysis such as partial least square-discriminant analysis (PLS-DA) and biomarker analysis based on receiver operating characteristic (ROC) curves using the MetaboAnalyst platform 98. These metabolic analyses were normalized by using Log and Auto-scaling transformations on MetaboAnalyst.