Root exudate metabolomes change under drought and show limited capacity for recovery

Root exudates comprise a large variety of compounds released by plants into the rhizosphere, including low-molecular-weight primary metabolites (particularly saccharides, amino acids and organic acids) and secondary metabolites (phenolics, flavonoids and terpenoids). Changes in exudate composition could have impacts on the plant itself, on other plants, on soil properties (e.g. amount of soil organic matter), and on soil organisms. The effects of drought on the composition of root exudates, however, have been rarely studied. We used an ecometabolomics approach to identify the compounds in the exudates of Quercus ilex (holm oak) under an experimental drought gradient and subsequent recovery. Increasing drought stress strongly affected the composition of the exudate metabolome. Plant exudates under drought consisted mainly of secondary metabolites (71% of total metabolites) associated with plant responses to drought stress, whereas the metabolite composition under recovery shifted towards a dominance of primary metabolites (81% of total metabolites). These results strongly suggested that roots exude the most abundant root metabolites. The exudates were changed irreversibly by the lack of water under extreme drought conditions, and the plants could not recover.

The importance of rhizosphere processes is becoming increasingly recognised, and encompasses what happens in the zone of soil that is directly affected by roots and associated soil microorganisms. Such processes can include the acquisition of nutrients, growth inhibition of surrounding competitor plants (i.e. allelopathy), sending chemical signals to attract symbiotic partners (chemotaxis) such as between rhizobia and legumes, or in the promotion of the beneficial microbial colonization on root surfaces 1,2 . Several recent studies suggest a pivotal role of root exudates in rhizosphere function and thus in plant-microbe-soil relationships [3][4][5] . Root exudates comprise a large variety of compounds released by plants into the rhizosphere 1,6,7 , including low-molecular-weight primary metabolites (e.g. saccharides, amino acids, and organic acids), secondary metabolites (e.g. phenolics, flavonoids and terpenoids) 8,9 , and inorganic molecules (e.g. carbon dioxide and water) 3 . Root exudation is an important source of organic carbon in the soil and can account for up to 2-11% of total photosynthetic production [10][11][12] . Exudates can thus provide soil microorganisms, which are often carbon-limited, with an important energy source 3 . This leads to a process called "soil priming", whereby the microbial community becomes more active and may liberate nutrients important for plants through its role in nutrient cycling 13 . Root exudates such as oxalic acid may also directly release organic compounds from organo-mineral aggregates, leading to greater microbial access to these compounds and increased net soil carbon loss 4 . Compounds in exudates also play important roles in various biological processes, including the mobilisation and acquisition of nutrients 14 , attracting and repelling certain microbial species [15][16][17] and inhibiting the growth of competing plant species 18 .
The quantity and composition of root exudates is species-specific, can vary throughout the life of a plant 19,20 , and is affected by abiotic factors. For example, the availability of nutrients in the soil can affect exudation. Some plants in phosphorus (P) poor soils can exude higher amounts of organic acids and phosphatase that help to mobilise recalcitrant P, e.g. Lupinus albus 21,22 , Medicago sativa 23 , and Brassica napus 24 .
Increasing droughts are predicted for many regions worldwide 25,26 , including the Mediterranean Basin, the focus area of this study. Knowledge of the impacts of water stress on plants, soils and their interactions, including root exudation, is consequently vital. The majority of studies on the effects of drought on trees have focused on Effect of recovery on exudate composition. The metabolomes collected after recovery from the various drought durations (except the plants with the longest drought duration of 18 days) shifted towards higher concentrations of several amino acids (aspartate, asparagine, arginine, and valine) and pyruvate but not towards the same metabolomic profile of the controls (Fig. 2). Overall, the metabolite composition under recovery was dominated by primary metabolites (81% of total metabolites). The composition of the exudates thus clearly differed between the drought and recovery pots (P < 0.01), even when plants had experienced low levels of drought. The exudates of the recovery plants were characterised by higher valine and arginine concentrations than the exudates of drought (Tables 2 and 3). This recovery response in exudate composition, with clear differences from pre-drought metabolomes, was in contrast to the recovery of the leaves, shown by F v /F m measurements, which returned to pre-drought levels except for in the two highest drought treatments (Fig. 1).

Discussion
The effect of drought on the total metabolites contained within root exudates has not yet been extensively investigated, but, could be important in a context of increasing aridity in several world areas such as the Mediterranean Basin 25,26 . This metabolomic analysis of root exudates during drought and recovery indicated that drought stress has a strong effect on the composition of Q. ilex exudates. Most of the metabolites that increased in concentration under drought and decreased in concentration under recovery (Table 1) were consistent with those previously reported 56,57 . Plant exudates under all drought levels except the highest (18 d) shifted towards a higher expression of secondary compounds (including antioxidants) and soluble sugars, plus compounds that act as osmolytes, thus compounds associated with the response of plants to drought stress. The metabolite composition of the root exudates under recovery shifted towards a composition dominated mainly by some amino acids and with decreases in the majority of saccharides and secondary compounds.
Effect of drought on exudate composition. The compounds that were most activated (those with the highest concentration) by the drought treatment were abscisic acid (ABA), choline, terpenoids, proline, aspartate, leucine, acacetin and maleic acid. The composition of root exudates has rarely been studied; a recent review indicated that only nine studies have reported metabolomic analyses of exudates 42 . The compounds reported in these studies predominantly included primary metabolites such as saccharides, organic acids, and amino acids but also various secondary metabolites such as phenolic compounds and phytohormones 42 similar to our study.
The increase in concentration of ABA was one of the most significant increases during the drought treatment, compatible with its characterisation as a universal plant stress hormone. An increase in the concentration of ABA in the roots is a very consistent physiological change in plants under drought 37,58,59 , although the predominant site of synthesis within the plant (roots or leaves) is not clear [60][61][62] . ABA reduces growth under drought conditions 29,59 and leads to stomatal closure 58,60 . ABA in roots also mediates an enhancement of the biosynthesis of osmolytes, such as the amino acid proline, protective proteins 63,64 , and organic acids such as malic, lactic, aspartic, and fumaric acids that are also associated with osmolytic function and thus water retention 65 . Our results strongly suggest that root exudate metabolite composition reflects the composition of the rest of the plant, with plants producing exudates using the most abundant metabolites under the current environmental conditions 6,45,66,67 . Therefore, as high levels of ABA within roots are predicted during drought, then high levels of ABA in the exudates should also be expected. Relating to other organic acids, the concentration of maleic acid is also increased under drought, as also reported in root exudates of maize plants under drought conditions 40 .
The concentrations of the amino acids aspartate and leucine were the second and third most increased concentrations in the drought-stressed plants. A typical effect of drought is the accumulation of solutes, such as amino acids and saccharides, within plant leaves and roots, which has been reported in many studies [68][69][70] . This accumulation, sometimes termed osmotic adjustment, may help to maintain cell turgor and reduce water loss from cells 68 . A net benefit to the plant that increases fitness in water-stressed environments has not been firmly established 71,72 , but evidence suggests that solute accumulation may help to maintain root development to access deeper reserves of water 72 . A recent review also reported a positive correlation between osmotic accumulation and plant productivity 73 . Previous studies of the effect of drought stress on the foliar metabolome of various species, including Q. ilex (the species used in our experiment) 53 , Triticum aestivum (wheat) 74 , and two common grass species 37,38 , have found changes in foliar metabolite composition indicating enhanced osmoprotection. Also, a recent paper investigating the effects of drought on the metabolome of barley (Hordeum vulgare) leaves and roots reported increases in osmoprotectants and antioxidants 75 . Our finding of increased concentrations of these amino acids indicates that the accumulation of solutes that has been shown in leaves and roots also occurs in root exudates, perhaps indicating the general osmotic status of the roots.
At the most severe levels of drought in our study, there was an increase in the exudation of proline, adenine, cinnamic acid, and homoorientin. Proline synthesis has been up-regulated in several plant species under drought 76 and other circumstances such as after wounding 64 , acting by buffering the cellular redox status during drought. Adenine is a precursor of several cytokinins, secondary compounds associated with triggering several anti-drought mechanisms in plants 77,78 . The flavonoids acacetin and homoorientin (O-methylated flavones) were also present in higher concentrations during drought. Flavonoids are secondary metabolites with an antioxidant function in terrestrial plants and accumulate during stressful environmental conditions [79][80][81] . Flavonoids play a direct role in drought tolerance 82 and are associated with symbiosis, signalling, plant development, and plant defences [83][84][85][86] , and the differential accumulation of specialised metabolites in border cells likely serves important defensive and signalling roles 87 . Vanillic acid is an organic acid that was present in higher concentrations in the drought samples in our study, as also reported in exudates of drought-stressed crested wheatgrass (Agropyron cristatum) 41 .
Overall, our results strongly suggest that plants produce exudates using the most abundant metabolites under the current environmental conditions. This agrees with a previous study that found strong links between exudate composition and the composition of both phloem (relating to organic acids and amino acids) and root biomass (relating to saccharides) 45 . Also supporting our theory is the observation that exudation is generally believed to be a mostly passive process 6 and a recent study found that there was a weak positive correlation between the amount of some saccharides (fructose, glucose and sucrose) exuded and the concentration in the root tissue. This suggests that these compounds are released by passive or facilitated diffusion and that the amount of these compounds present in exudates reflects that of the roots 66 . In contrast, there is recent evidence showing that organic acids and amino acids appear to be mostly actively exuded from roots 66,67 meaning that the metabolite composition of these compounds might not necessarily be expected to reflect that of the rest of the plant.
It is also important to note that there can be differences in the drought response of metabolites in different organs of the plant, for example one study with the highly drought tolerant plant Caragana korshinskii showed that the metabolites that were found to increase under drought varied between the leaves, stem, root collar and roots, reflecting the higher drought tolerance of the stem and roots compared with the leaves 88 . Similarly, a study with two common grass species (Holcus lanatus and Alopecurus pratensis) revealed opposite metabolic responses to drought between shoots and roots. While roots remained active, with a high amount of primary metabolites, leaves showed a decrease in primary metabolism and an increase in some secondary metabolites (such as organic acids, terpenes and phenols) 37,38 . This response of the shoot seems to reflect more the drought response we saw in root exudates. Therefore, there is potentially a complicated interplay between the effect of the leaves and the roots on the composition of exudates, with the leaves having a greater impact during drought and root metabolism being more important during recovery and non-drought times.
Effect of recovery on exudate composition. The longest drought had an irreversible effect on the metabolome as after six weeks of re-watering the exudates did not return to the pre-drought composition. This is in contrast to recent work in the same study system that showed that the amount of carbon exuded following   drought was able to recover, even after extreme drought stress 32 . The compounds that showed the largest increase in concentration during recovery from drought included several amino acids such as alanine, arginine, valine and asparagine, one intermediate of saccharides and amino acid metabolism, which was citrate, and two saccharides, fructose and ribose. This is partially consistent with a recovery of energy production capacity and protein synthesis and thus growth. Saccharides and amino acids are typical semi-polar primary metabolites that should be common in growing plant tissues, such as fine roots of young tree saplings, under conditions of no stress. The main precursor metabolite in glycolysis is a fructose. The yield of glycolysis is ATP and NADH with pyruvate as an end product. Citrate is an important intermediary in the tricarboxylic acid cycle and also serves as a substrate for the cytosolic production of acetyl-CoA. CoA is important in numerous metabolic processes, especially in providing carbon substrates for secondary metabolism. The levels of saccharides, pyruvate and citrate (precursors for CoA biosynthesis), including amino acids alanine, isoleucine and valine were higher in the recovery samples (Table 3). Overall, the dominance of compounds involved in primary metabolism, typical of a growing plant, may again suggest that exudates mainly contain the most abundant metabolites in the fine roots at each moment, depending on the current plant metabolic status. Phenylpropanoids are the precursors of flavonoids, isoflavonoids, anthocyanins, and lignin and are synthesised from the primary amino acid phenylalanine. The conversion of phenylalanine to cinnamic acid by phenylalanine ammonia-lyase is an early step in the biosynthesis of these compounds. Phenylalanine was abundant in both the drought and recovery samples but was higher in the recovery samples (Tables 2 and 3).
Concentrations of many amino acids were much higher in the recovery than the drought samples (Table 3). Asparagine was the most abundant amino acid in the recovery samples and increased relative to the drought samples. Asparagine is an end-point amino acid that serves as a major compound for transporting and storing nitrogen in plant cells 89 .
Even for plants exposed to only very mild levels of drought, exudation still differed in composition compared to those in the control drought treatment after six weeks of re-watering, demonstrating that even short-term droughts can lead to long-term measurable changes. This revealed a clear difference with the above-ground response as the observed leaf chlorophyll fluorescence, values returned to normal, pre-drought, levels for plants at all drought treatments except for the two most severe ones. Chlorophyll fluorescence has been shown to be sensitive to photosynthetic drought and recovery in the study species 90 , however other studies have revealed that photosynthesis and chlorophyll fluorescence measurements are not always coupled 91,92 . Therefore, it is possible that the aboveground part of the plant was still affected by the drought, despite this not being shown by our measurements of chlorophyll fluorescence. This highlights the importance of investigating the below-ground processes under drought, and root exudates in particular, as responses may be in contrast to what is seen above-ground.
At the highest drought intensity there was no evidence of any recovery of the exudate metabolome, and the re-watered plants had exudate metabolomes that closely resembled those of the corresponding plants under drought stress. Few studies have investigated how root exudates are able to recover after drought (none to our knowledge investigating the effect on the metabolome), but the amount of carbon released during root exudation has some capacity for recovery in the two studies that have measured it under recovery after a previous drought period 32,35 .

Conclusions
This study is the first to document the changes in the metabolome of root exudates subjected to varying drought levels and then subsequent recovery. We have demonstrated that even small changes in water availability can have a measurable and long-lasting impact on the composition of root exudates, and we have highlighted the compounds whose synthesis is activated at different stages of water stress.
Exudates of drought-stressed plants (except for those with the longest drought) tended to be dominated by secondary metabolites, although the precise composition varied depending on the intensity of the drought, and under the most extreme drought higher concentrations of fructose, arginine, and valine were seen. We did not find evidence for recovery of the root exudate metabolome after six weeks of re-watering, and we also demonstrated that the metabolite composition of root exudates during recovery after drought varied with the severity of the drought.
Overall, changes in root exudation composition, both during drought and recovery, are highly dependent on the severity of water stress. There may also be a threshold level of stress for plants beyond which they cannot recover, and that could be determined from the metabolome of the exudates. We propose that plant root exudate composition reflects the composition of the rest of the plant, with plants producing exudates using the most abundant metabolites available at the time, and therefore varies depending on environmental conditions. Exciting new developments in the collection and analysis of root exudates may lead to greater insight into the response of belowground plant tissues to abiotic stress and into the causes underlying species differences in drought tolerance.

Materials and Methods
Description of drought experiment. Plant and soil material. A greenhouse experiment was established in the experimental fields of the Autonomous University of Barcelona in May 2015. The experiment comprised 144 three-year-old Q. ilex saplings, approximately 80 cm in height and total biomass of 50 g (supplied by Forestal Catalana). The sapling roots were carefully washed in water prior to replanting to remove the previous soil so that the soil communities would be representative of the new soil. These plants were then re-potted in 3.5-L pots, with forty-eight in each of three soil treatments: control, drought-stressed and sterilised. All potting substrates contained 45% autoclaved peat, 45% of sand, and 10% of natural soil inoculum. The soil inoculum was collected from Prades Natural Park, which is the site of a long-term drought experiment (established in 1999) that reduces rainfall by approximately 30%. For details of the site see Peñuelas et al. 93  three different soil inocula representing three different soil histories. Firstly, 'control' soil inoculum was collected from the topsoil of control plots of the long-term drought experiment, to represent a natural holm oak forest soil under ambient precipitation. Secondly, 'drought-stressed' soil inoculum was taken from topsoil of the drought plots of the same long-term drought experiment. There was a third 'sterilised' soil inoculum, which used soil from the control plots of the long-term experiment that was later autoclaved to remove the majority of microorganisms in the soil. However, there was no significant difference in metabolomic profiles of our root exudate samples from plants grown in the different soil types (results of the PERMANOVA), so data was pooled into one group for the analyses presented in this paper. This resulted in there being 18 replicates at each drought level, with nine undergoing drought only and nine being kept for recovery. The plants were given adequate water (top-watering once a day to attain a soil moisture content of approximately 20-25%) for five weeks to allow them to adjust to the greenhouse environment. Mean air temperature during the experiment was 26.7 °C (monitored using ELUSB-2 data logger, Lascar Electronics, Wiltshire, UK). Soil temperature was monitored at a fine scale in five pots, across the different soil types (using a Decagon Em50 data logger with 5TM soil probes, Decagon Devices, Pullman, WA, USA), and the mean temperature was 27.0 °C.
Experimental design. The pots were subjected to drought by stopping the addition of water. Eight levels of drought were determined by the length of time without water: 0, 2, 4, 7, 9, 11, 16, and 18 days. Each drought level therefore had 18 pots, divided between the three soil treatments. These pots were arranged into six blocks. Root exudate samples were collected (method described below) from half of the pots at the end of each drought period and are hereafter referred to as "drought-stressed plants". The remaining pots entered a six-week recovery phase with optimal watering and are hereafter referred to as "recovery" plants. Mean soil moisture (measured using ML3 Theta Probe connected to a HH2 Moisture Meter from Delta-T Devices, Cambridge, UK) at the start of the experiment was 22.6% and it decreased exponentially to 1.0% at the end of the most extreme drought treatment. Soil moisture recovered quickly to ca. 20% within one week of re-watering (top-watering as required to attain a soil moisture content of approximately 20-25%) and was 24.7% on average during the recovery phase (Fig. 4).
To measure the impact of the drought treatment and recovery on the aboveground part of the plants, chlorophyll fluorescence measurements were taken. For each level of drought, plants were measured 1-3 days before the collection of exudates during both the drought period and the recovery period. Three healthy and mature leaves at the top of each plant were measured at midday using a portable chlorophyll fluorometer PAM-2500 (H. Walz, Effeltrich, Germany). A saturating light pulse was applied to dark-adapted (at least 30 mins) leaves for the determination of minimum (F o ) and maximum (F m ) fluorescence. The maximum photochemical efficiency of photosystem II (PSII) was then calculated according to Genty et al. 94 Collection and preparation of exudates samples. Rhizodeposition was measured at the end of each drought period (for unwatered plants) and then at the end of each six-week recovery period (for re-watered plants) using a direct measuring technique adapted from that of Phillips et al. (2008). Briefly, a root was carefully excavated from the soil, cleaned, placed in moist sand, and then wrapped in aluminium foil. After one day of acclimation, the root was cleaned and placed in a cuvette containing small glass beads (to apply physical pressure to the root, as if it were in soil) and a carbon-free nutrient solution to prevent desiccation (0.5 mm NH 4  , which can recover a variety of metabolites such as polar and semipolar compounds, e.g. saccharides, amino acids, terpenoids, polysaccharides, flavonoids, and many other secondary metabolites. One millilitre of MeOH/H 2 O was added to Eppendorf tubes containing 150 mg of powder from each sample. The tubes were vortexed for 10 min, sonicated for 5 min at room temperature, and then centrifuged at 23 000 g for 5 min. After centrifugation, 0.7 mL of the supernatant from each tube was collected using crystal syringes, filtered through 0.22-µm pore microfilters, and transferred to a labelled set of high-performance liquid chromatography (HPLC) vials. The vials were stored at −80 °C until the LC-MS analysis. This procedure was repeated for two extractions of the same sample. (+H). Recordings from both the diode array detector and the high-resolution mass spectrometer were monitored and saved to verify system operation and evaluate later results.
LC-MS chromatograms were obtained with a Dionex Ultimate 3000 HPLC system (Thermo Fisher Scientific/ Dionex RSLC, Dionex, Waltham, USA) coupled to an LTQ Orbitrap XL high-resolution mass spectrometer (Thermo Fisher Scientific, Waltham, USA) equipped with an HESI II (heated electrospray ionisation) source. Chromatography was performed on a reversed-phase C18 Hypersil gold column (150 × 2.1 mm, 3-µm particle size; Thermo Scientific, Waltham, USA) at 30 °C. The mobile phases consisted of acetonitrile (A) and water containing 0.1% acetic acid (B). Both mobile phases were filtered and degassed for 10 min in an ultrasonic bath prior to use. The elution gradient, at a flow rate of 0.3 mL min −1 , began at 10% A (90% B) and was maintained for 5 min. The content of mobile phase A increased to 90% during the subsequent 15 minutes. This composition was then maintained for 5 min, after which the system was over a period of 5 min gradually equilibrated to the initial proportions (10% A and 90% B). The column was then washed and stabilised for 5 min before the next sample was injected. The injection volume of the samples was 5 µL. HESI was used for MS detection. The Orbitrap mass spectrometer was operated in Fourier Transform Mass Spectrometry mode, and full-scan spectra were acquired for mass range m/z 50-1000 in the positive mode and 65-1000 in the negative mode at high-mass resolution (60 000). The resolution and sensitivity of the Orbitrap were controlled by the injection of a mixed standard after the analysis of each batch (30 samples), and resolution was also checked with the aid of lock masses (phthalates). Blank samples were also analysed during the sequence. The assignment of the metabolites was based on the standards, and the metabolites were identified against our in-house library (>200 compounds) and manually annotated with the retention time (RT) and mass of the assigned metabolites in both positive and negative ionisation modes ( Table 4).  Table S1 for details). The chromatograms were baseline corrected, deconvoluted, aligned, and filtered, and the numerical database was then exported in "csv" format. Metabolites were assigned by comparison with the analyses of the standards (RT and mass) (see Table S1 for details). Assigned variables corresponding to the same molecular compounds were summed. The LC-MS data for the statistical analyses corresponded to the absolute peak area at each RT. The area of a peak is directly proportional to the concentration (i.e. µg mL −1 ) of its corresponding (assigned) metabolite in the sample, so a change in the area of a peak indicates a change in the concentration of its assigned metabolite. Most metabolites were related to important metabolic pathways. The increases and decreases in concentrations of these metabolites showed up and down regulation in these metabolic pathways.
Statistical analyses. The LC-MS data were analysed by univariate and multivariate statistical analyses. We conducted permutational multivariate analyses of variance (PERMANOVAs) (Anderson et al., 2008) using Euclidean distances, with drought/recovery and drought intensity (eight levels of drought) as fixed factors and individuals as random factors. Partial least squares discriminant analyses (PLSDAs) were performed to detect patterns of sample ordination in the metabolomic variables. The PLSDAs were initially constructed from the LC-MS analysis of exudates to enable the identification of clusters, groups, and outliers (Sandasi et al., 2011) (Fig. 2). The PC scores of the cases were subjected to one-way ANOVAs to determine the statistical differences among groups with different levels of the categorical independent variables (drought and recovery). The PERMANOVAs and PLSDAs used the mixOmics package of R (R Development Core Team 2015). The Kolmogorov-Smirnov (KS) test was performed on each variable to test for normality. All assigned and identified metabolites were normally distributed, and any unidentified metabolomic variable that was not normally distributed was removed from the data set. Statistica v8.0 was used for the ANOVAs, post hoc tests, and KS tests. Chlorophyll fluorescence data were analysed using repeated-measures analyses of variance (ANOVAs), with differences considered statistically significant at P < 0.05 (R Development Core Team 2015).

No. Metabolites
Abbreviation RT(min)  Table 4. The 65 metabolites identified in roots exudates of Quercus ilex. Their abbreviations, relative retention times and mass to charge (m/z) ratios used for identification and quantification are also depicted. RT, retention time.