True-to-type Carrizo citrange (Poncirus trifoliata L. Raf. x Citrus sinensis L. Osb.) and C. mandarin (Citrus reshni Hort. Ex Tan.) plants were purchased from an authorized commercial nursery (Beniplant S.L., Penyíscola, Spain). One-year-old seedlings of the two citrus genotypes were individually placed in plastic pots filled with perlite and watered three times a week with 0.5 L of a half-strength Hoagland solution. Plants were grown in the greenhouse under natural photoperiod and day and night temperature averaging 25.0 ± 3.0°C and 18.0 ± 3.0°C, respectively, for 1 month to allow acclimation. Subsequently, plants were maintained in growth chambers under a 16-h photoperiod (light fluency of 300 μmol m⁻² s⁻¹ PPF), 25 °C and a relative ambient moisture of 80% for 2 weeks before the beginning of the experiments. Temperature and relative moisture were recorded regularly with a portable USB datalogger (OM-EL-WIN-USB, Omega, New Jersey, USA).

A 24-h experiment was performed in which severe drought, imposed by transplanting plants to dry perlite, was applied alone or in combination with high temperatures (40°C). Prior to imposition of drought, heat stress was applied for 7 days to a group of well-watered Carrizo and Cleopatra plants whereas another group was maintained at 25°C. Thereby, we established four experimental groups for each genotype: Well-watered plants grown at 25°C (CT) and 40°C (HS) and plants subjected to drought at 25 °C (WS) and at 40°C (WS+HS). Fully expanded leaves with an intermediate position in the canopy were harvested 24 h after the imposition of water stress by performing a gentle cut at the petiole with a scalpel. Harvesting was performed during day time and illuminated leaves were immediately submerged in liquid N₂ to deter all metabolic activity. Frozen leaves were crushed and ground to fine powder and stored at −80°C for subsequent metabolomic analyses.

For primary metabolite analysis by GC-MS, 50 mg of fresh leaf tissue was extracted in 300 μL of pure methanol (LC-MS grade, Panreac, Barcelona, Spain) spiked to 0.2 mg ml⁻¹ with ribitol as internal standard (IS), following Roessner et al. (2001). Extractions were performed by ultra-sonication for 10 min at room temperature. After centrifugation, supernatants were recovered and mixed with 200 μL of chloroform and 400 μL of water and centrifuged at 10000 rpm and 4°C for 10 min to allow phase separation. The upper water layer was recovered and evaporated to dryness using a SpeedVac (Jouan, Saint HerblainCedex, France). Dry residues were re-dissolved with 50 μL of 20 mg mL⁻¹ methoxamine in pyridine followed by incubation at 30°C in a water bath for 90 min. Later, 70 μL of methylsilyltrifluoroacetamide (Macherey-Nagel, Germany) were added and subsequently incubated at 37°C for 30 min. Finally, the solution was mixed with 10 μL of a commercial mixture of fatty acid methyl esters (C8-C24 FAME mix, Sigma-Aldrich, Madrid, Spain) as retention index (RI) markers.

Derivatized extracts were independently injected onto a GC-MS system composed of a gas chromatograph equipped with an autosampler and a quadrupole mass analyzer (GCMS-QP2010 Ultra, Shimadzu Corporation, Japan) equipped with an electron impact ion source. The GC separation was performed using a fused silica BPX5 capillary column with a length of 30 m × 0.25 mm ID and a film thickness of 0.25 μm (SGE Analytical Science, Victoria, Australia). The oven program was set as follows: 80°C (2 min); 10°C min⁻¹ to 325°C (3.5 min) for a total runtime of 30 min. Injections of 1 μL of sample extracts were performed in split mode (1:200) at a temperature of 230°C. Helium (99.999 %; Praxair, Valencia, Spain) was used as the carrier gas at a constant flow rate of 2 mL min⁻¹. The interface and source temperatures were set to 325°C and 230°C, respectively. Scan rate was set at 5 scan s⁻¹ within 50 to 700 amu mass range. After acquisition, chromatogram files were converted to NetCDF for further processing.

Mass chromatographic features were extracted with xcms (Smith et al., 2006) and subsequently processed with TargetSearch software (Cuadros-Inostroza et al., 2009) using Golm Metabolite Database (available from http://www.mpimp-golm.mpg.de/). Briefly, this software calculates RI for all compounds in chromatograms based on tabulated RI values for the constituents of the FAME mix and then matches RI and mass fragments of compounds in samples with those found in databases. Identified metabolites were cross-referenced in chromatograms using GCMS-solution software (Shimadzu Corp.). Peak areas of identified metabolites were normalized to the IS (ribitol) area before statistical analyses.

Extraction was performed essentially as previously described in Zandalinas et al. (2012) with slight modifications. Briefly, 500 μL of 70% methanol supplemented with biochanin A at 1 mg L⁻¹ (IS) was added to 0.1 g of frozen leaf powder. After 10 min of sonication, extracts were heated at 80°C in a water bath for 15 min and allowed to cool down at room temperature. Then, samples were centrifuged at 10000 g for 10 min at 4°C and supernatants filtered through 0.22 μm PTFE syringe filters (Whatman International Inc., Kent, United Kingdom), prior to UPLC/ESI-QTOF-MS analysis.

Chromatographic separations were performed on an Acquity SDS system (Waters Corp. Ltd., Milford, MA) interfaced to a QTOF Premier from Micromass Ltd. through an ESI source. A reversed-phase column was used as follows: 100 mm × 2.1 mm i.d., 1.8 μm, ProntoSil, Bischoff, Leonberg, Germany). Sample aliquots (10 μL) were injected onto the UPLC system using the partial loop-filling option. Separations were carried out using a flow rate of 300 μL min⁻¹ for 30 min as follows: 0−2 min, isocratic 95% A [water:formic acid, 99.9:0.1 (v/v)] and 5% B [acetonitrile: Formic acid, 99.9:0.1 (v/v)]; 2−17 min, gradient 5−95% B; 17−20 min, return to initial conditions; 20−25 min, re-equilibration period. During analyses, the column temperature was maintained at 40°C and samples were maintained at 10°C to slow down degradation. Samples were analyzed in both negative and positive ionization modes. Two functions were set in the instrument: In function 1, data were acquired in profile mode from 50 to 1000 Da using a scan time of 0.2 s and a collision energy of 2 eV; in function 2, the scan range was the same, but a collision ramp between 4 and 65 eV was set. During all measurements, the electrospray capillary voltage was set to 4 kV and the cone voltage was set to 25 V. The source temperature was maintained at 120°C and the desolvation gas temperature was set at 350°C. Argon was used as the collision gas and nitrogen was used as the nebulizer as well as desolvation gas set at 60 and 800 L h⁻¹, respectively. Exact mass measurements were provided by monitoring the reference compound lockmass leucine-enkephalin.

Data were processed using Masslynx v.4.1. Raw data files were converted to netCDF format using the application databridge from Masslynx and processed using the xcms package as previously described (Smith et al., 2006; Zandalinas et al., 2012). Chromatographic peak detection was performed using the matchedFilter algorithm, applying the following parameter settings: snr = 3, fwhm = 15 s, step = 0.01 D, mzdiff = 0.1 D, and profmethod = bin. Retention time correction was achieved in three iterations applying the parameter settings minfrac = 1, bw = 30 s, mzwid = 0.05 D, span = 1, and missing = extra = 1 for the first iteration; minfrac = 1, bw = 10 s, mzwid = 0.05 D, span = 0.6, and missing = extra = 0 for the second iteration; and minfrac = 1, bw = 5 s, mzwid = 0.05 D, span = 0.5, and missing = extra = 0 for the third iteration. After final peak grouping (minfrac = 1, bw = 5 s) and filling in of missing features using the fillPeaks routine of the xcms package, a data matrix consisting on feature × sample was obtained.

Significantly altered metabolites were extracted from datasets following ANOVA analysis at p ≤ 0.05 using sample treatment as factor. To investigate metabolite profiles over samples, genotypes and treatments, peak areas of significantly altered metabolites were represented as a heatmap to facilitate visualization. Both heatmaps and hierarchical cluster analysis (HCA) on heatmaps were performed using Dchip software (http://www.hsph.harvard.edu/cli/complab/dchip/). PLS-DA was performed with Simca-P+ 11.0 (Umetrics AB, Umeá, Sweden). In addition, metabolite levels were subjected to analysis of variance using a two-way ANOVA considering the two citrus genotypes and the four stress treatments followed by Tukey post-hoc test (p < 0.05) when a significant difference was detected (Tables S2, S3).

Non-targeted GC-MS analysis was performed to determine different patterns of polar metabolite accumulation in leaves of citrus plants subjected to WS, HS, and WS+HS. Results revealed different metabolite profiles in both citrus genotypes under individual and combined stress conditions (Figures 1, 2). In Carrizo, stress was the main source of variability (21.4%) allowing the differentiation of a “control” cluster of samples from the rest of treated samples. Within this “stressed” cluster, all three treatments (WS, HS, and WS+HS) could be easily differentiated (Figure 1A). Conversely, the main source of variability (17.1%) of polar metabolites in Cleopatra was the imposition of combined stress conditions (WS+HS) with a minor contribution of WS (11.5%). Interestingly, control samples could not be clearly differentiated from HS samples (Figure 1B). Analysis of variance revealed a total of 94 polar metabolites differentially altered (56 accumulated and 38 decreased, out of 1143 detected) in Carrizo plants of which 14 were accumulated only during WS (Figure 2A). Only 8 compounds were found specifically altered in Carrizo during WS+HS and heat stress alone induced the accumulation of 12 compounds (Figure 2A). Stress imposition significantly altered a total of 20 metabolites whereas the impact of individual stresses depressing metabolite levels was more restricted (Figure 2A).

In Cleopatra, stress imposition significantly altered a total of 78 polar metabolites out of 1113 detected in GC-MS analyses, of which 68 were accumulated and 10 reduced (Figure 2B). Stress combination induced the accumulation of 48 metabolites including the glycolytic end-product pyruvate as well as the TCA intermediate succinate, whereas WS applied alone induced the accumulation of only 1 polar metabolite in Cleopatra. Both HS and WS+HS stresses caused the increase in levels of 15 metabolites (Figure 2B). As mentioned above, stress imposition did not have such a strong impact on depressing polar metabolite levels in Cleopatra compared to Carrizo (levels of 4 metabolites decreased upon stress imposition, 5 compounds shared by HS and WS and only one was significantly altered by HS applied alone) (Figure 2B).

To investigate differences between genotypes in the accumulation of different polar compounds, metabolite levels were represented within their respective pathways.

Carrizo and Cleopatra displayed different accumulation of metabolites related to glycolysis and TCA cycle depending on the stress type. As a general trend, stress imposition induced fructose accumulation in both citrus genotypes, especially in response to HS and WS+HS (Figure 3). In addition, WS induced a significant glucose-6-P accumulation in Cleopatra. On the contrary, in this genotype, fructose-6-P levels were depleted in Cleopatra plants subjected to WS or stress combination, in contrast to the response observed in Carrizo (Figure 3). Both genotypes also showed differences at the level of fructose 1,6-bis-P splitting catalyzed by aldolase: Cleopatra preferentially accumulated glyceraldehyde 3-P in response to WS and HS whereas Carrizo showed higher levels of dihydoxyacetone-P in response to all stress treatments but especially WS. Glycerate 3-P levels increased specifically in response to WS in Carrizo and, to a lesser extent in response to HS, whereas WS+HS reduced its levels. In Cleopatra, the sensitive genotype, individual stress imposition reduced glycerate 3-P levels but stress combination significantly increased them. On the other hand, glycerate 2-P showed increased levels in response to HS in Carrizo and in response to WS and HS in Cleopatra. In addition, stress combination reduced its levels below controls in Carrizo. Moreover, stress imposition also induced higher levels of phosphoenolpyruvate in Carrizo, consistent with increased oxalacetate concentration in this genotype. Nevertheless, stress slightly induced pyruvate accumulation in the two genotypes, although to a higher extent in Cleopatra. Among the TCA cycle intermediates, all stress conditions induced citrate accumulation in Cleopatra but only HS had a similar effect in Carrizo. Similarly, isocitrate accumulated in Carrizo in response to HS. In turn, α-oxoglutarate profiles showed a similar trend to that of citrate: In Carrizo, it responded to high temperatures but stress combination reduced its levels below control values. Contrastingly, Cleopatra exhibited increased levels of α-oxoglutarate in response to WS but even higher levels were found under HS and WS+HS. Levels of succinate also showed contrasting behavior in Carrizo and Cleopatra in response to stress (Figure 3). All stress conditions increased succinate levels in Carrizo especially those involving HS whereas in Cleopatra isolated heat or drought caused decreases in succinate concentration and only stress combination slightly increased its levels. Finally, Cleopatra increased fumarate levels in response to all stress conditions whereas in Carrizo no significant differences were found. Malate levels slightly increased under WS and stress combination in Carrizo but it followed an opposite decreasing trend in Cleopatra under WS.

In general, stress induced the accumulation of different intermediates of the chloroplastic phase of the glyoxylate/dicarboxylate cycle in citrus. The accumulation of glycolate 2-P from ribulose-1,5-bis-P was impaired in Carrizo under WS+HS conditions. Nevertheless, glycolate concentration was significantly higher in WS and HS in Carrizo and Cleopatra leaves, respectively. Accumulation profile for the Calvin-cycle intermediate, ribulose 5-P, varied among genotypes. Levels were severely reduced in Cleopatra under all stress conditions whereas they slightly increased in Carrizo in response to WS+HS. Sucrose levels significantly increased in Carrizo in response to all stress conditions, showing the strongest accumulation under HS. In contrast, in Cleopatra, HS had a detrimental effect on sucrose build-up, whereas a moderate induction was observed under WS+HS (Figure 4). Among the peroxysomal metabolites, concentration of glycine followed an opposite trend between the two genotypes: While levels of this aminoacid strongly increased in response to WS and WS+HS in Carrizo leaves, in Cleopatra they decreased under the same stress conditions. Although clear increases in glycine levels were induced in Carrizo by the stress conditions, no parallel effect on serine concentration was observed (Figure 4). Nevertheless, glycerate concentration also differed from that of its precursor serine, showing significant increases in response to WS and WS+HS in Cleopatra and a severe depression in Carrizo in response to all stress conditions.

Accumulation of polar metabolites participating in the phenylpropanoid pathway arising from shikimic acid was analyzed, showing different metabolite profiles between genotypes in response to stress: All adverse conditions increased shikimic acid, prephenic acid, and phenylpyruvic acid concentrations in Carrizo whereas levels of these metabolites generally decreased in Cleopatra under the same stress conditions. Nevertheless, phenylalanine concentration increased in Cleopatra in response to WS, HS and especially WS+HS whereas stress treatments had poor impact on this aminoacid in Carrizo (Figure 5). Levels of cinnamic acid, the first phenolic acid product of the reaction catalyzed by the phenylalanine ammonia lyase, were severely reduced after stress imposition, as well as the product of cinnamate hydroxylation, coumaric acid, in leaves of both citrus genotypes. However, cinnamoyl aldehyde accumulated only in Carrizo leaves subjected to WS and into a lesser extent to HS and WS+HS, whereas its levels were severely reduced in Cleopatra. Moreover, caffeic acid levels increased in both genotypes in response to stress following different profiles. In this sense, its levels were higher in Carrizo plants subjected to WS whereas a minor accumulation could be observed upon WS+HS imposition in this genotype. Contrastingly, all stress conditions induced significant increases in the caffeic acid concentration in Cleopatra, although the highest levels were observed in plants subjected to stress combination. Ferulic acid levels did not show any significant alteration in response to stress but, conversely, scopoletin, derived from hydroxylation of feruloyl CoA (Kai et al., 2008), and scopolin, a glycoside of scopoletin, were accumulated in response to all stress conditions in Carrizo and Cleopatra. Moreover, sinapic acid levels also responded positively to WS+HS in Carrizo. However, in Cleopatra, levels of this metabolite increased significantly only in response to HS, whereas stress combination reduced its levels below control values.

Phenolic acids are usually converted into the respective aldehydes via conjugation with coenzyme A and subsequently reduced to alcohols, constituting the building blocks for lignins (Boerjan et al., 2003). Sinapoyl aldehyde exhibited a general increment in response to individual stresses in Cleopatra whereas it was reduced below control levels in Carrizo in response to HS. Coniferyl alcohol (derived from coniferyl aldehyde) accumulated in Carrizo in response to WS and WS+HS and reduced in response to HS. In Cleopatra, levels of this metabolite decreased in response to WS. Coumarin, which is synthesized from cinnamic acid and performs different cell protective roles, was significantly induced by HS in Carrizo and to a lesser extent by WS+HS. Moreover, caffeic acid diverted into the production of chlorogenic acid (Humphreys and Chapple, 2002), showing significantly increased levels in Cleopatra in response to WS and HS. However, levels of this metabolite slightly decreased in Carrizo upon stress imposition (Figure 5).

To determine the accumulation of semi-polar metabolites in leaves of citrus plants subjected to WS, HS and their combination, we performed a LC-MS analysis of compounds extracted from leaves of both citrus plants subjected to the different stresses. Xcms processing of LC-MS metabolite profiles rendered a total of 7195 and 8453 peaks in Carrizo and Cleopatra, respectively, of which 1001 and 2730 had significant variations in area values in the different treatments, as revealed by ANOVA analysis. Partial Least Squares coupled to Discriminant Analysis (PLS-DA) showed a clear impact of heat stress on secondary metabolite composition in Carrizo (26.3% of explained variability) whereas water stress had a lower contribution (18.8%) (Figure 6A). In Cleopatra, however, both heat and water stress had an influence on secondary metabolite composition (21.1%) although the stress combination had a stronger contribution (26.2%) (Figure 6B).

The specific profile for each stress treatment as well as the metabolite overlapping among treatments was studied and results are depicted as Venn diagrams in Figure 7. Carrizo leaves displayed a clear response to HS applied individually or in combination with WS. In this sense, levels of 87 and 66 metabolites increased and decreased, respectively, during HS and WS+HS, likely as a result of the incidence of HS (Figure 7A). Identification of metabolites is shown in Table 1. Among them, we found significantly altered flavonoids such as isorhamnetin hexoside desoxyhexoside, limonoids (nomilin deoxyhexoside), or the phospholipid derivative lysophosphatidyl choline (Table 1 and Figures 8–10 and Figure S1). Furthermore, we also found that levels of several compounds decreased in response to HS and WS+HS in Carrizo leaves (Figures 8, 9 and Figure S1) including flavonoids like kaempferol derivatives or apigenin deoxyhexose hexose. On the other hand, as indicated also by PLS-DA (Figure 6B), we observed a significant response of Cleopatra to WS, finding 14 secondary metabolites whose levels were significantly altered during WS applied individually, including flavonoids (quercetin hexoside deoxyhexoside and hesperetin hexoside deoxyhexoside), limonoids (nomilin deoxyhexoside), or lipid derivatives such as linolenic acid hexoside hexoside (Figures 8, 9 and Figure S1). Additionally, 49 compounds were accumulated in response to all stress treatments (Figure 7B), including those depicted in Figure 8, of which 43 were specifically accumulated in response to HS and WS+HS in Cleopatra leaves. Moreover, levels of 6 secondary metabolites were altered in Cleopatra plants in response to WS and WS+HS and only 3 compounds were found to be significantly accumulated in response to WS+HS. Moreover, a total of 115 metabolites specifically decreased in Cleopatra leaves subjected to HS and WS+HS. Only levels of one compound decreased in response to all stress treatments and 2 metabolites also decreased in Cleopatra plants in response to WS and WS+HS (Figure 7B).

The accumulation profile of different flavonoid-related compounds in leaves of Carrizo and Cleopatra subjected to WS, HS, and WS+HS was also studied. As shown in Figure 9, different patterns of flavonoid accumulation were observed between both citrus genotypes and also among stress treatments. Therefore, levels of flavones such as the different tangeretin isomers were constitutively higher in Cleopatra than in Carrizo (Figure S1) but WS significantly increased their levels in Carrizo. However, in Cleopatra, WS+HS induced the accumulation of these compounds. Flavones, such as apigenin and its derivatives, showed an opposite constitutive trend: While apigenin aglycone levels were much higher in Cleopatra apigenin deoxyhexoside hexoside levels were higher in Carrizo (Figure S1). In turn, apigenin deoxyhexoside hexoside showed a significant accumulation in response to WS in Carrizo whereas it was slightly reduced below control levels in Cleopatra under HS and WS+HS (Figure 8). Meanwhile, apigenin aglycone levels were significantly increased in response to WS+HS in Cleopatra whereas in Carrizo, the levels of this metabolite were reduced with respect to control values during WS. Stress treatments had no impact on the accumulation of the different hesperetin derivatives (Figure 9, Figure S1 and Table S3) in both citrus genotypes. However, hesperetin derivatives were constitutively higher in Cleopatra than in Carrizo (Figure S1). Furthermore, individual and combined stress conditions induced the accumulation of kaempferol derivatives in Cleopatra with almost identical profiles: WS and WS+HS showing the highest levels (Figure 9). In addition, levels of kaempferol hexoside deoxyhexoside #2 were in general much higher in Cleopatra than in Carrizo (Figure S1). Regarding quercetin derivatives, the effect of stress treatments was similar to that of kaempferol derivatives (except for quercetin hexoside hexoside that was not detected in Cleopatra), showing a significant accumulation in response to WS and WS+H in Cleopatra. On the other hand, no impact on the accumulation of isorhamnetin-related compounds was observed in Carrizo leaves in response to stress. Moreover, constitutive levels of isorhamnetin hexoside deoxyhexoside were higher in Cleopatra and those of isorhamnetin methylhexose hexose higher in Carrizo (Figure S1). On the other hand, in Cleopatra, levels of isorhamnetin hexoside desoxyhexoside decreased in response to HS whereas isorhamnetin methylhexose hexose was accumulated under WS in this genotype.

Limonoids are naturally-occurring triterpenoids in rutaceae whose accumulation or depletion in response to individual and combined stress conditions was evaluated (Figure 10). Nomilin was not detected in Cleopatra leaves while in Carrizo the levels of this metabolite were reduced in response to individual WS and HS. Moreover, the accumulation of nomilin deoxyhexoside followed opposite patterns in both citrus genotypes: While Carrizo decreased the concentration of this metabolite under WS and accumulated it under HS and WS+HS, Cleopatra showed significantly increased levels under WS but reduced under HS and WS+HS. Obacunone and limonin levels increased in Cleopatra plants in response to WS and WS+HS whereas in Carrizo, the accumulation of both metabolites was especially marked in response to WS+HS (Figure 10).

Constitutive levels of linolenic acid, linolenic acid dihexoside, and lysophosphatidyl choline were higher in Cleopatra than in Carrizo and, in addition, showed different regulation patterns in the two genotypes (Figure 8 and Figure S1). Lysophosphatidyl choline levels increased mainly in response to WS+HS in both genotypes although to a higher extent in the sensitive genotype Cleopatra. Conversely, levels of linolenic acid dihexoside decreased in response to WS+HS and increased under WS conditions in Cleopatra. In Carrizo, however, HS and WS+HS reduced levels of linolenic acid dihexoside (Figure 8 and Figure S1). Finally, stress had little impact on linolenic acid levels, showing slight reductions in response to WS and WS+HS in Carrizo and in response to WS and HS in Cleopatra.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.