Drought stress at the reproductive stage causes severe damage to productivity of wheat. However, little is known about the metabolites associated with drought tolerance. The objectives of this study were to elucidate changes in metabolite levels in wheat under drought, and to identify potential metabo lites associated with drought stress through untargeted metabolomic profiling using a liquid chromatography-high resolution mass spectrometry (LC-HRMS)-based technique called Isotopic Ratio Outlier Analysis. Metabolomic analysis was performed on flag leaves of drought-stressed and control (well-watered) plants after 18 days of post-anthesis drought stress at three-hour intervals over a 24-hour period. Out of 723 peaks detected in leaves, 221 were identified as known metabolites. Sixty known metabolites were identified as important metabolites by 3 different methods, PLS-DA, RF and SAM. The most pronounced accumulation due to drought stress was demonstrated by tryptophan, proline, pipecolate and linamarin, whereas the most pronounced decrease was demonstrated by serine, trehalose, N-acetyl-glutamic acid, DIBOA-glucoside etc. Three different patterns of metabolite accumulation were observed over 24-hour period. The increased accumulated metabolites remained higher during all 8 time points in drought stressed leaves. On the contrary, metabolites that showed decreased level remained significantly lower during all or the most time points. However, the levels of some decreased metabolites were lower during the day, but higher during night in drought stressed leaves. Both univariate and multivariate analyses predicted that N-acetyl-glutamic acid, proline, pipecolate, linamarin, tryptophan, and DIBOA-glucoside could be potential metabolite biomarkers, and their levels could serve as indicators of drought tolerance in wheat.
Wheat is one of the most widely produced and consumed cereal grains worldwide and global wheat demand is expected to rise in excess of 880 million metric tons by 2050 [
Crop yield has shown significant correlation with metabolites under drought [
Metabolic profiling allows for comprehensive analyses of a range of metabolites that have great value in both phenotyping and plant diagnostics [
The soft winter wheat variety ‘SS8641’ (GA-881130/2*GA-881582) is widely grown in the southeastern USA, developed by the University of Georgia Wheat Breeding Program, and was subjected to post-anthesis drought stress and LC-HRMS based untargeted metabolic profiling in this study. The SS8641 is a mid-maturing, high yielding wheat variety and has high test weight with good straw strength. It showed resistance to Hessian fly biotypes B and E, and possesses powdery mildew genes Pm1 and rust resistance genes Lr37/Yr17/Sr38. Since the wheat flag leaf is a vital source of energy assimilates during grain filling [
Initially, seeds of SS8641 were exposed to vernalization at 4˚C for 6 weeks to induce flowering. After vernalization, three germinated seeds were planted in each pot and kept in an environmentally controlled greenhouse located at the campus of the University of Florida in Gainesville. The greenhouse was maintained at day/night temperature of 20/15˚C ± 0.5˚C (16 h daylight and 8 h night time) with a relative humidity of 50% ± 2%. Plants were manually irrigated at three day intervals and maintained at 100% field capacity (FC) until exposed to drought stress. Plants were also fertilized with two splits of 12 g of Scott’s Osmocote (20-4-8 NPK fertilizer) during the experimental period: one at 7 days and another split at 45 days after transplanting. Once plants reached the flowering stage, pots were randomly assigned to two different watering regimes with three replications as follows: control treatment (well-watered, 100% FC) and drought treatment (water-deficit, 25% FC). To determine FC, the weights of pot plus dried soil weight were recorded as average 2.24 kg, then pots were irrigated to 100% FC and average weight increased to 3.44 kg (moisture content was assessed as 1.20 kg/pot). Pots in the control treatment were maintained at 100% FC, while drought treatments were maintained up to 25% of the FC. After 18 days of the drought treatment, plants were characterized for photosystem and chlorophyll damages, and flag leaves were collected for metabolic profiling.
Chlorophyll fluorescence (the ratio of variable, Fv, to maximum fluorescence, Fm) and SPAD chlorophyll content were used as indirect methods to assess thylakoid membrane [
Plasma membrane damage (PMD) was measured using the method described by Ristic and Cass (1993) [
Plants were harvested after physiological maturity and oven-dried at 60˚C for five days. After drying, total biomass was measured per plant. The spikes were hand threshed and the chaff was cleaned off. Grain weight was measured and the harvest index was calculated by dividing grain weight/total biomass. Grain yield per spike was calculated by dividing total grain weight by spike number. Grain number per spike was calculated by dividing total grain number by the number of spikes for that plant. Two hundred grains were randomly selected and weighed, and converted to 1000-grain weight.
Flag leaves were collected in 3-hour increments (06:00 AM; 09:00 AM; 12:00 noon; 15:00 PM; 18:00 PM, 21:00 PM, 24:00 PM and 03:00 AM) over a 24 hour time period. Collected flag leaf from each pot was considered a biological replication. A triplicate samples were collected at each sampling time point as our experiment was maintained in a well environmentally controlled greenhouse and experimental treatments showed high accuracy within replication. Sampled leaf tissues were frozen in liquid nitrogen immediately after collection and then stored at −80˚C until processing. Leaf tissue samples were lyophilized and ground using a tissuelyser (24 samples per treatment for a total of 48 samples). For metabolomic analysis, 5 mg of experimental material was weighed and added to 5 mg of freshly ground wheat internal standard (IS). The IS was an isotopically labeled wheat leaf that had been grown in an atmosphere of 13C labeled carbon dioxide resulting in a uniform and universal labeling of approximately 97% (IROA Technologies) [
The reconstituted samples were analyzed for untargeted metabolites on a Liquid Chromatography High Resolution Mass Spectrometery (LC-HRMS) platform. Untargeted metabolomics profiling was performed on a Thermo Q-Exactive Oribtrap mass spectrometer with Dionex UHPLC and autosampler. All samples were analyzed in positive and negative modes with heated electrospray ionization with a mass resolution of 35,000 at m/z 200. Chromatographic separation was achieved on an ACE 18-pfp 100 × 2.1 mm, 2 µm column with mobile phase A as 0.1% formic acid in water and mobile phase B as acetonitrile. The flow rate was 350 µL/min and column temperature was 25˚C. Total run time per sample was 21 minutes. Quality assurance and quality control (QA/QC) guidelines were followed during untargeted profiling of assays with the addition of stable-isotopic internal standards to evaluate reproducibility, injection standards, and the repeated analysis of a large pooled plasma sample. Injection reproducibility was typically less than 10% even without a ratio to an internal standard. The native Thermo “.raw” output files were converted to .mzXML files using ProteoWizard (Version 2).
One of the greatest challenges of most metabolic profiling experiments is the ability to differentiate peaks of biological origin from artifact peaks, and to accurately identify and quantitate the peaks of interest. Since we used an IROA-labeled plant material as our internal standard this study followed the IROA “Phenotypic” global labeling and bioinformatics protocols in which the Internal Standard (IS) is labeled at 95% 13C. Therefore, all biological compounds are paired natural abundance (NA) and IS, and each pair carries distinct molecular signatures. Molecules can be distinguished from each sample set, as they have differing masses [
For IROA, control and drought treated samples were analyzed as a single composite sample by LC-MS. Algorithms pair identified biological peaks, and unlabeled NA artifacts were identified and discarded. All biological compounds had two paired peaks; the peak from the 12C-media is mirrored by a second peak from the 13C-media. The distance between the monoisotopic peaks readily identified the number of carbons in the compound. The corresponding M+1 and M−1 peaks (and M+2 and M−2 etc. peaks) which are a mass difference of 1.00335 amu (mass difference between a 12C and 13C isotope), gave the IROA peaks a characteristic U-shape “smile” pattern. Accurate mass together with the knowledge of the number of carbons in a molecule greatly facilitated metabolite identification. The ClusterFinder program was used to identify, align and quantitate the IROA peaks in the .mzXML files. The files were scanned for IROA peaks down to an intensity level of 1 million in both positive and negative modes with an assumed maximum of 10 ppm mass error. Settings for the experimental samples in the nontargeted analysis assumed a natural abundance (1.1%) isotopic balance while the IS was assumed to have a 97% isotopic balance. The resulting list of IROA peaks was manually curated to define 394 compounds that were seen in both the IS and in the experimental samples. Once the physical attributes for the 494 compounds were identified, a targeted analysis for all of these compounds was imposed on every sample. This resulted in a non-sparse dataset, i.e. there was a value for every compound for every sample and the dataset was exported for analysis. The metabolites were annotated by searching against an in-house metabolite database, Mass Spectrometry Metabolite Library of Standards (MSMLS) (http://iroa.com/page/Mass%20Spectrometry%20Metabolite%20Library%20of%20Standards).
Data tables with metabolite peaks (mz/rt) at 8 time points under both drought and control conditions were formatted as comma separated values (.csv) files and uploaded to the MetaboAnalyst 3.0 server (http://www.metaboanalyst.ca) [
Univariate analysis (t-test and one way ANOVA) was applied to calculate the statistical significance and fold change of the metabolites between two group means (drought over control). As the multivariate methods take all the variables into consideration, we applied multivariate methods for comprehensive data analysis e.g. supervised methods- Partial Least Squares Discriminant Analysis (PLS-DA), Random Forest (RF) classification, and unsupervised method-Hierarchical clustering with heatmap. The supervised method, PLS-DA was used to maximize the difference of metabolic profiles between control and drought groups to enable the detection of metabolites exist in the biological samples. A heat map was generated based on the Pearson distance measure and the Ward clustering algorithm, showing top 60 metabolites selected by PLS-DA VIP (variable importance in projection) score using a significance level of P ≤ 0.05, and post-hoc analysis of Fisher’s LSD. The samples were arranged according to their sampling time points in both control and drought groups. The important metabolites were identified by using 3 different methods separately: SAM (Significant Analysis of Metabolites), PLS-DA and RF.
The pathway analysis was performed using Metabo Analyst for the identified important metabolites using Oryza sativa japonica and Arabidopsis thaliana pathway libraries. The Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway database (http://www.genome.ad.jp/kegg/pathway.html) was also used for the metabolites that were not found in the rice and Arabidopsis pathway libraries. To identify the potential biomarkers associated with drought stress, the Receiver Operating Characteristic (ROC) curve based ‘Biomarker Analysis’ module of the MetaboAnalyst was applied. Both classical univariate ROC curve analysis and multivariate ROC curve exploratory analysis were used to identify the promising biomarkers with high sensitivity and high specificity.
We measured physiological traits to assess the impact of the drought stress 18 days after stress initiation at anthesis. The chlorophyll content (SPAD value), maximum quantum efficiency of PSII (Fv/Fm) and electrolyte leakage (%) were determined on flag leaves for both control (well-watered) and drought conditions (
Total shoot biomass at maturity, grain weight/plant, grain weight/spike, 1000-grain weight, grains/spike, and harvest index showed significant reduction under drought stress compared to control condition as expected (
Traits | Control (Mean ± SE) | Drought (Mean ± SE) |
---|---|---|
SPAD chlorophyll content | 60.1 ± 0.8 | 53.4 ± 1.8 |
Chlorophyll a florescence (Fv/Fm) | 0.781 ± 0.003 | 0.746 ± 0.017 |
Electrolyte leakage (%) | 38.4 ± 0.7 | 57.1 ± 0.7 |
Plasma membrane damage (PMD %) | - | 30.4 ± 0.5 |
Total shoot dry biomass at maturity (g/plant) | 16.0 ± 2.1 | 10.4 ± 0.1 |
Grain weight/plant | 7.5 ± 0.6 | 5.1 ± 0.1 |
1000 grain weight (g) | 49.4 ± 0.3 | 41.0 ± 1.5 |
Number of grains/spike | 27.8 ± 1.7 | 24.1 ± 1.5 |
Grain weight/spike (g) | 1.8 ± 0.1 | 1.3 ± 0.1 |
Harvest Index | 44.4 ± 1.9 | 40.0 ± 0.4 |
SE = Standard Error.
Metabolite profiling by LC-HRMS detected a total of 723 peaks from wheat flag leaves. Among the detected peaks, 221 were identified as known metabolites (266 including duplications) and the remaining peaks were unknown metabolites. The identified compounds included amino acids, sugars, organic acids, organic compounds, polyamines, fatty acids, nucleosides/nucleobases and other compounds.
The supervised clustering method, Partial Least Squares-Discriminant Analysis (PLS-DA) was performed for two conditions at 8 different time points. Five PLS-components (PCs) explained 66.5% of the total variation, with the first and second PCs contributing 45.1% and 8.5%, respectively (
One hundred fifty two metabolites were found significantly different (t-test, p < 0.05) between drought and control conditions (Supplemental
Significantly different metabolites were analyzed by hierarchical clustering with heat map in order to visualize the effect of drought stress over the control at eight different time points (
with opposite pattern of metabolite accumulation. The first cluster was represented by metabolites that accumulated to high levels under drought stress, including tryptophan, proline, linamarin and pipcolate. The second cluster included metabolites that decreased ≥1.5 fold under drought compared to control, including serine, 4-Aminobutanoate (GABA), oxidized glutathione, sucrose, DIBOA-glucoside, raffinose, arabinose, N-acetyl-DL-glutamic acid, Pyruvate, glycerate, nicotinate, tribenuron methyl, palmitoleic acid, nandrolone, calystegin B2, slaframine, pyridoxal and others. The clustering of metabolites in two groups clearly indicates the metabolic changes in flag leaves under stress condition.
A comparison of statistical models was carried out to identify the important metabolites associated with drought condition using: SAM, PLS-DA and RF (
Name of metabolites | Compound ID | Molecular Formula | Compound type | SAM (d.value) | RF (Mean Decrease Accuracy) | PLS-DA **(VIP score) |
---|---|---|---|---|---|---|
L-Proline | C00148 | C5H9NO2 | Amino acid | 3.034 | 0.0180 | 1.369 |
L-Tryptophan | C00078 | C11H12N2O2 | Amino acid | 2.745 | 0.0024 | 1.284 |
L-Aspartate | C00049 | C4H7NO4 | Amino acid | −2.017 | 0.0016 | 1.099 |
L-Serine | C00065 | C3H7NO3 | Amino acid | −1.978 | 0.0012 | 1.078 |
L-Threonine | C00188 | C4H9NO3 | Amino acid | −1.965 | 0.0023 | 1.091 |
4-Aminobutanoate (GABA) | C00334 | C4H9NO2 | Amino acid | −2.184 | 0.0045 | 1.151 |
L-Carnitine | C00318 | C7H15NO3 | Amino acid derivative | −2.609 | 0.0051 | 1.336 |
Clithioneine | UNPD 118571 | C13H22N4O5S | Amino acid derivative | −2.037 | 0.0019 | 1.136 |
D-(+)-Trehalose | C01083 | C12H22O11 | Sugar | −2.827 | 0.0053 | 1.376 |
Sucrose | C00089 | C12H22O11 | Sugar | −2.573 | 0.0028 | 1.268 |
D-Threo-L-galacto-octose | 16019992* | C8H16O8 | Sugar | −2.513 | 0.0065 | 1.299 |
D-(-)-Arabinose | C00216 | C5H10O5 | Sugar | −2.188 | 0.0066 | 1.202 |
D-(+)-Raffinose | C00492 | C18H32O16 | Sugar | −2.071 | - | 1.103 |
N-Acetyl-D-Fucosamine | C15480 | C8H15NO5 | Sugar | −1.892 | 0.0027 | 1.053 |
6-(α-D-glucosaminyl)-1D- myo-inositol, (GlcN)1 (Ino)1 | C15658 | C12H23NO10 | Sugar alcohol | −2.775 | 0.0033 | 1.331 |
Bis-D-fructose 2',1:2,1' -dianhydride | C04333 | C12H20O10 | Sugar dianhydride | −2.720 | 0.0064 | 1.313 |
D-xylonate | C00502 | C5H10O6 | Sugar acid | −3.090 | 0.0204 | 1.469 |
alpha-Methylene gamma -butyrolactone | 68352* | C5H6O2 | Sugar acid | −2.114 | - | 1.120 |
Spermidine | C00315 | C7H19N3 | Amine | −2.106 | 0.0078 | 1.130 |
Palmitoleic acid | C08362 | C16H30O2 | Fatty acid | −1.949 | 0.0015 | 1.062 |
N-Acetyl-DL-glutamic acid | C00624 | C7H11NO5 | Organic acid | −3.467 | 0.0267 | 1.568 |
Guanidinoproclavaminic acid | C06657 | C9H16N4O4 | Organic acid | −2.892 | 0.0064 | 1.395 |
Gluconic acid | C00257 | C6H12O7 | Organic acid | −3.163 | 0.0092 | 1.468 |
Cerheptaric acid | UNPD64017 | C7H12O8 | Organic acid | −2.576 | 0.0029 | 1.292 |
Pipecolate | C00408 | C6H11NO2 | Organic acid | 2.442 | - | 1.121 |
Lactobionic acid | 7314* | C12H22O12 | Organic acid | −2.228 | 0.0029 | 1.203 |
12-Hydroxyjasmonic acid | C21385 | C12H18O4 | Organic acid | −2.170 | 0.0019 | 1.161 |
---|---|---|---|---|---|---|
D-Saccharic acid | C00818 | C6H10O8 | Organic acid | −2.170 | 0.0025 | 1.119 |
Glutarate | C00489 | C5H8O4 | Organic acid | −2.009 | 0.0051 | 1.076 |
alpha-Aminoadipate | C00956 | C6H11NO4 | Organic acid | −2.004 | 0.0012 | 1.135 |
Phosphonatoenolpyruvate | C00074 | C3H2O6P | Organic acid | −1.930 | - | 1.064 |
Nicotinate | C00253 | C6H5NO2 | Organic acid | −1.896 | 0.0026 | 1.110 |
Glycerate | C00258 | C3H6O4 | Organic acid | −1.892 | - | 1.043 |
Pyruvate | C00022 | C3H4O3 | Organic acid | −1.890 | 0.0019 | 1.083 |
DIBOA-glucoside | C15772 | C14H17NO9 | Organic compound | −3.455 | 0.0126 | 1.529 |
Lilioside A | 101277416* | C11H20O9 | Organic compound | −3.047 | 0.0119 | 1.472 |
Tribufos | 5125* | C12H27OPS3 | Organic compound | −2.784 | 0.0079 | 1.405 |
P-Cymene | C06575 | C10H14 | Organic compound | −2.782 | 0.0105 | 1.385 |
5'-O-beta-D-Glucosylpyridoxine | C03996 | C14H21NO8 | Organic compound (Vit B6) | −2.751 | 0.0055 | 1.391 |
Strophanthobiose | 14236734* | C13H24O9 | Organic compound | −2.706 | 0.0072 | 1.356 |
Tribenuron methyl | C10962 | C15H17N5O6S | Organic compound | −2.703 | 0.0036 | 1.318 |
Eupalitin 3-O-sulfate | 44259774* | C17H14O10S | Organic compound | −2.590 | 0.0077 | 1.354 |
Linamarin | C01594 | C10H17NO6 | Organic compound | 2.494 | 0.0025 | 1.187 |
Pruyanaside B | UNDP 145434 | C33H36O16 | Organic compound (Phenolicglucoside) | −2.430 | 0.0030 | 1.281 |
Oxidized glutathione | C00127 | C20H32N6O12S2 | Organic compound (antioxidant) | −2.419 | 0.0017 | 1.240 |
Slaframine | C06185 | C12H20N2O3 | Organic compound | −2.331 | 0.0044 | 1.224 |
Fructoselysine 6-phosphate | C16489 | C12H25N2O10P | Phosphorylated compound | −2.588 | 0.0036 | 1.274 |
1-(Indol-3-yl)propanol 3-phosphate | C04229 | C11H14NO4P | Phosphorylated compound | −2.318 | 0.0020 | 1.242 |
Nicotinamide | C00153 | C6H6N2O | Organic compound | −2.308 | 0.0040 | 1.219 |
Nandrolone | C07254 | C18H26O2 | Organic compound | −2.194 | 0.0024 | 1.194 |
Glucogallin | C01158 | C13H16O10 | Organic compound | −2.147 | 0.0017 | 1.187 |
Calystegin B2 | C10851 | C7H13NO4 | Organic compound | −2.137 | 0.0027 | 1.154 |
Asperuloside | C09769 | C18H22O11 | Organic compound | −2.062 | 0.0093 | 1.101 |
1,2,3-Trihydroxybenzene | C01108 | C6H6O3 | Organic compound | −2.026 | - | 1.100 |
4-Nitroacetophenone | C02803 | C8H7NO3 | Organic compound | −1.987 | 0.0013 | 1.052 |
Beta-Cymaropyranose | C08234 | C7H14O4 | Organic compound | −1.968 | - | 1.101 |
2-Valeryl-SN-glycero -3-phosphocholine | 24779499* | C13H28NO7P | Others | −2.457 | 0.0020 | 1.305 |
Carbetamide | C11075 | C12H16N2O3 | Others | −2.446 | 0.0015 | 1.264 |
4-Hydroxy-2-butynal | C02648 | C4H4O2 | Others | −2.384 | 0.0018 | 1.203 |
Pyridoxal | C00250 | C8H9NO3 | Others | −2.295 | 0.0031 | 1.207 |
**Only variance for component 1 has been shown.
the delta value of 1.6, FDR of 0.001 and with less than one (0.5) false positive. Similarly the most important metabolites were also identified by PLS-DA method based on the VIP score using five-component model. Random Forest classification ranked the important metabolites in order of decreasing prediction accuracy (Mean Decrease Accuracy) using 5000 trees (permutation) with an overall (OOB, out-of-bag) error of 0.0417. Overall the results were quite similar across all three methods. The top most important 60 metabolites which were identified by at least 2 different methods are shown in
Over the course of 24 hours 89 metabolites exhibited significant fluctuations (ANOVA, P ≤ 0.05, Fisher’s LSD) across the 8 sampled time-points between the two treatments (Supplemental
To better elucidate the biological functions of identified metabolites, a pathway analysis was performed using Oryzasativa and Arabidopsis thaliana as the pathway libraries. As expected, these metabolites were involved in number of different pathways (
Applying a ROC (Receiver Operating Characteristic)-curve based approach of biomarker analysis [
Compound | Pathway involved |
---|---|
L-Proline | Arginine and proline metabolism, Aminoacyl-tRNA biosynthesis |
L-Tryptophan | Glycine, serine and threonine metabolism, Aminoacyl-tRNA biosynthesis, Phenylalanine, tyrosine and tryptophan biosynthesis (Shikimate pathway), Glucosinolate biosynthesis, Tryptophan metabolism, |
4-Aminobutanoate (GABA) | Arginine and proline metabolism, Alanine, aspartate and glutamate metabolism, Butanoate metabolism |
L-Serine | Glycine, serine and threonine metabolism, Aminoacyl-tRNA biosynthesis, Cyanoamino acid metabolism, Cysteine and methionine metabolism, Methane metabolism, Sulfur metabolism, Sphingolipid metabolism |
Threonine | Aminoacyl-tRNA biosynthesis, Valine, leucine and isoleucine biosynthesis |
Sucrose | Galactose metabolism, Starch and sucrose metabolism |
Raffinose | Galactose metabolism |
N-Acetyl-DL-glutamic acid | Arginine and proline metabolism |
Aspartate | Glycine, serine and threonine metabolism, Arginine and proline metabolism, Carbon fixation in photosynthetic organisms, Alanine, aspartate and glutamate metabolism, Aminoacyl-tRNA biosynthesis, Cyanoamino acid metabolism, beta-Alanine metabolism, Cysteine and methionine metabolism, Lysine biosynthesis, Nicotinate and nicotinamide metabolism |
Pyruvate | Glycine, serine and threonine metabolism, Carbon fixation in photosynthetic organisms, Alanine, aspartate and glutamate metabolism, Cysteine and methionine metabolism, Citrate cycle (TCA cycle), Butanoate metabolism, Pyruvate metabolism, C5-Branched dibasic acid metabolism, Glycolysis or Gluconeogenesis, Valine, leucine and isoleucine biosynthesis, Pantothenate and CoA biosynthesis, Terpenoid backbone biosynthesis |
Phosphoenolpyruvic acid | Carbon fixation in photosynthetic organisms, Citrate cycle (TCA cycle), Pyruvate metabolism, Phenylalanine, tyrosine and tryptophan biosynthesis, Glycolysis or Gluconeogenesis |
Nicotinate | Nicotinate and nicotinamide metabolism |
Glyceric acid | Glycerolipid metabolism |
Oxidized glutathione | Glutathione metabolism |
Spermidine | Arginine and proline metabolism, beta-Alanine metabolism, Glutathione metabolism |
Palmitoleic acid | Fatty acid biosynthesis |
Pyridoxal | Vitamin B6 metabolism |
Linamarin | Cyanoamino acid metabolism; biosynthesis of secondary metabolites |
Pipecolate | Lysine degradation; biosynthesis of secondary metabolites |
Malonate | Pyrimidine metabolism; beta-Alanine metabolism |
Alpha-Aminoadipate | Lysine biosynthesis; lysine degradation; biosynthesis of secondary metabolites |
Piperidine | Tropane, piperidine and pyridine alkaloid biosynthesis; protein digestion and absorption |
namely, N-acetyl-glutamic acid, L-proline, L-tryptophan, linamarin, pipecolate, and malonate (
had high AUC value of 0.82 and high sensitivity (0.8) and specificity (0.7) in univariate analysis.
Drought stress can severely affect physiological and morphological mechanisms that could potentially affect performance, functionality and ultimately survival of plants. To maintain growth and productivity, plants must adapt to stress conditions by exercising specific tolerance mechanisms. The alternation of different attributes related to photosynthesis is a good indicator of stress tolerance. Damage to photosynthetic capacity due to water stress during grain filling can limit supply of assimilates to growing grain and thus reduces yield potential. Chlorophyll fluorescence, SPAD chlorophyll content, and electrolyte leakage or plasma membrane damage are very sensitive and quick responding variable for monitoring water stress. In the present study, drought stressed flag leaves exhibited a reduction of chlorophyll fluorescence, SPAD chlorophyll content and higher electrolyte leakage, which indicate significant damage to photosystem II and the plasma membrane due to drought stress at grain filling stage of wheat [
In addition to physiological and morphological traits, levels of metabolite accumulation during stress at a particular growth stage can provide a more specific and accurate indication of stress tolerance. In the present study, the non-targeted metabolite profiling revealed significant changes in 152 metabolites in wheat flag leaves under drought stress condition compared to control. Of these differentially accumulated metabolites, 60 metabolites were identified as the most important metabolites by three different methods: PLS-DA, RF and SAM, which included an array of different amino acids, sugars, organic acids or compounds, polyamines, fatty acids and their derivatives. Amino acids including tryptophan, proline, phenylalanine, tyrosine and isoleucine accumulated up to 7.1 fold in drought treated flag leaves compared to control, although the levels of some amino acids (e.g. serine, glutathione, glutamine, GABA, threonine) decreased due to drought stress. The accumulation of amino acids under drought stress has been reported previously in several studies [
Tryptophan acts as an osmolyte in ion transport regulation and in modulating stomatal opening, and maintaining water balance between air and plant [
Proline level was increased in flag leaves under drought conditions and the accumulation of proline has been correlated with stress tolerance in a wide range of plants by different scientists [
Sugars and their derivatives were significantly reduced under drought stress. Lower accumulation of glucose and sucrose is potentially due to reduced photosynthetic capacity, as demonstrated by lower chlorophyll fluorescence and SPAD chlorophyll content, of the drought stressed leaves at the grain filling stage [
The lower accumulation of sugars was further accompanied by reduced levels of organic acids or compounds including pyruvate, phosphoenolpruvate, α-ketoglutaric acid which are involved in Krebs cycle in leaves of stressed plants. Drought stress in our study didn’t trigger higher accumulation of organic acids or compounds except pipecolate and linamarin. A decreased level of organic acids or compounds was reported in wheat [
Pipecolate is reported to have a defensive action on proteins and nucleic acid structures by maintaining a stable osmotic status in plants under variable soil water and salt stress [
The level of different metabolites changed in wheat flag leaves at 8 different time points across a diel cycle and they changed differently under drought and control conditions (
The pathway analysis by using MetaboAnalyst was able to link with 33 metabolic and biosynthesis pathways of the KEGG pathway database and Rice Annotation Project database (
We also conducted the analysis to find potential metabolite biomarkers associated with drought tolerance in wheat for the screening of drought tolerant genetic resources of wheat. Obata et al. (2015) opined that promising metabolic traits were stronger in explaining variability in maize grain yield than classical agronomic yield components [
This study is the first study in wheat which used a non-targeted LC-HRMS Isotopic Ratio Outlier Analysis (IROA) Global Metabolomics method to identify accumulated metabolites in flag leaves under drought and control conditions at 8 time points during 24 hours day-night period. The study reports increased or decreased level of numerous metabolites simultaneously and consistently at different time points. Some of metabolites received considerable research attention under drought stress, but many of these did not. Metabolites like tryptophan, proline, pipecolate, linamarin, N-acetyl-glutamic acid, DIBOA-glucoside accrued to a greater level after drought stress, which was likely the indication of acclimation in responses to the drought stress. Elevated tryptophan and proline are common in other crops and are not surprising, but increased levels of pipcolate, linamarin, and decreased levels of N-acetyl-glutamic acid and DIBOA-glucoside in response to drought stress have not been widely reported. These metabolites may serve as metabolite biomarkers for screening drought stress tolerance in wheat germplasms to develop climate resilient wheat. The finding of these metabolites could potentially be the effect of genetic variation, as we used soft wheat which is different than other wheat genotypes used in the study before, including in Bowne et al. (2012) [
We are grateful to Southeast Center for Integrated Metabolomics (SECIM), University of Florida for all valuable assistance for metabolic profiling and data cleaning for further analysis.
This work was supported by the Dean’s Research Initiative, IFAS, University of Florida.
The authors declare that they have no conflict of interests.
Rahman, M.A., Akond, M., Babar, M.A., Beecher, C., Erickson, J., Thomson, K., De Jong, F.A, and Mason, R.E. (2017) LC-HRMS Based Non-Targeted Metabolomic Profiling of Wheat (Triticum aestivum L.) under Post-Anthesis Drought Stress. American Journal of Plant Sciences, 8, 3024-3061. https://doi.org/10.4236/ajps.2017.812205