Severe drought stress is affecting selected primary metabolites, polyphenols, and volatile metabolites in grapevine leaves (Vitis vinifera cv. Pinot noir)

lable at ScienceDirect

Plant Physiology and Biochemistry 88 (2015) 17e26

Contents lists avai

Plant Physiology and Biochemistry

journal homepage: www.elsevier .com/locate/plaphy

Research article

Severe drought stress is affecting selected primary metabolites,polyphenols, and volatile metabolites in grapevine leaves (Vitisvinifera cv. Pinot noir)

Michaela Griesser a, *, Georg Weingart b, Katharina Schoedl-Hummel a, Nora Neumann b,Manuel Becker c, Kurt Varmuza d, Falk Liebner c, Rainer Schuhmacher c, Astrid Forneck a

a Division of Viticulture and Pomology, Department of Crop Sciences, University of Natural Resources and Life Sciences Vienna (BOKU),Konrad Lorenz Str. 24, 3430 Tulln, Austriab Center for Analytical Chemistry, Department of Agrobiotechnology (IFA-Tulln), University of Natural Resources and Life Sciences Vienna (BOKU),Konrad Lorenz Str. 20, 3430 Tulln, Austriac Department of Chemistry, University of Natural Resources and Life Sciences Vienna (BOKU), Konrad Lorenz Str. 24, 3430 Tulln, Austriad Departement of Statistics and Probability Theory, Vienna University of Technology, Wiedner Hauptstrasse 7, 1040 Vienna, Austria

a r t i c l e i n f o

Article history:Received 20 October 2014Accepted 14 January 2015Available online 15 January 2015

Keywords:Drought stressVitis viniferaHS-SPME-GC-MSMetabolomicsChlorophyll fluorescence

Abbreviations: cv, cultivar; ROS, reactive oxygen spDA, partial least square discriminant analysis; DOT,adapted; LA, light adapted; PCA, principal componefluorescence; CER, leaf CO2 exchange rate.* Corresponding author.

E-mail address: [email protected] (M.

http://dx.doi.org/10.1016/j.plaphy.2015.01.0040981-9428/© 2015 Elsevier Masson SAS. All rights re

a b s t r a c t

Extreme weather conditions with prolonged dry periods and high temperatures as well as heavy rainevents can severely influence grapevine physiology and grape quality. The present study evaluates theeffects of severe drought stress on selected primary metabolites, polyphenols and volatile metabolites ingrapevine leaves. Among the 11 primary metabolites, 13 polyphenols and 95 volatiles which wereanalyzed, a significant discrimination between control and stressed plants of 7 primary metabolites, 11polyphenols and 46 volatile metabolites was observed. As single parameters are usually not specificenough for the discrimination of control and stressed plants, an unsupervised (PCA) and a supervised(PLS-DA) multivariate approach were applied to combine results from different metabolic groups. In afirst step a selection of five metabolites, namely citric acid, glyceric acid, ribose, phenylacetaldehyde and2-methylbutanal were used to establish a calibration model using PLS regression to predict the leaf waterpotential. The model was strong enough to assign a high number of plants correctly with a correlation of0.83. The PLS-DA provides an interesting approach to combine data sets and to provide tools for thespecific evaluation of physiological plant stresses.

© 2015 Elsevier Masson SAS. All rights reserved.

1. Introduction

Changing environmental conditions can have severe effects onyield and quality of all crop plants. Grapevine (Vitis vinifera) is afruit crop plant of high importance, with 7.8 million hectarescultivated in 2011 worldwide and a yearly production of 67.5million tons of berries (http://www.oiv.int/). Traditionally grape-vine is cultivated without irrigation and thus yield and grapequality depend on the vine to cope with periods of drought by

ecies; ABA, abscisic acid; PLS-days of treatment; DA, darknt analysis; CHF, chlorophyll

Griesser).

served.

adapting to rootstock and scion to edaphic conditions (Lovisoloet al., 2010).

Plants have evolved various adaptive responses to minimize theeffects of stress, as for example during drought stress plants closetheir stomata and accumulate compatible solutes to maintain a lowwater potential and avoid dehydration (Skirycz and Inze, 2010).Grapevine regulates the flow of water to the leaf and from the leafto the atmosphere by aquaporins (Vandeleur et al., 2009), vesselanatomy (Lovisolo et al., 2010) and stomatal conductance (Soaret al., 2006). As a consequence, plants reduce their growthrapidly. In grapevine a reduced leaf and shoot growth, measurableas inhibition of internode extension, leaf expansion, elongation oftendrils and a decrease of average diameter of xylem vessels di-ameters (Lovisolo et al., 2002; Schultz and Matthews, 1988) istypically observed. On the other hand root growth is stimulated to acertain but minor extent in soil compartments where water is still

M. Griesser et al. / Plant Physiology and Biochemistry 88 (2015) 17e2618

available (Dry et al., 2000). Moreover, reduced growth is reflectedby the suppression of photosynthesis due to stomata closure andlimitations to CO2 diffusion within leaf tissues (Flexas et al., 2004).When stomatal conductance drops below 50 mmol H2O m�2 s�1

photosynthetic limitations becomes more dependent on non-stomatal processes, especially decreased mesophyll diffusionconductance to CO2 and impaired photochemistry (Flexas et al.,2004; Flexas and Medrano, 2002).

In response to water stress, grapevine cultivars have beenclassified as isohydric or anisohydric (Schultz, 2003). Isohydriccultivars have the ability to keep their midday leaf water potentialconstant and above the cavitation threshold (�1.5 MPa) regardlessof soil water availability or atmospheric water demand (Lovisoloet al., 2010) by the reduction of the stomatal conductance,whereas anisohydric cultures keep their stomata open even underdecreasing leaf water potential (Sade et al., 2012). Grapevines re-ported as isohydric are Vitis labruscana, the rootstock Richter-110(Vitis berlandieri Planch x Vitis rupestris Scheele) and the V. viniferavarieties Grenache, Trincadeira, Petra, Tempranillo and others.Many traditional V. vinifera cultivars, e.g. Chardonnay, CabernetSauvignon, Syrah and Riesling are described as anisohydric(Lovisolo et al., 2010). However this classification is not alwaysconsistent, as for example Pinot Noir is described as anisohydricpre-veraison and isohydric post-veraison (Poni et al., 1993). Aqua-porins have been identified as key genes for water uptake andtransport in plant roots in general and two grapevine genesVvPIP2;1 and VvPIP2;2 have been described in detail under wellwatered and water stress conditions (Vandeleur et al., 2009) andABA is a key signal in root to shoot communication under droughtstress by regulating stomatal conductance (Schachtman andGoodger, 2008).

The effects of drought stress on grape ripening and grape qualitytraits have been investigated on the transcriptome (Deluc et al.,2009), proteome (Grimplet et al., 2009) and metabolome(Grimplet et al., 2009) level. Biotic and abiotic types of stress areinfluencing the primary and secondary metabolism of plants,having impacts on the content of carbohydrates, and secondarymetabolites as e.g. polyphenols and volatile organic compounds(VOCs) (Deluc et al., 2009; Grimplet et al., 2009; Lawo et al., 2011;Schoedl et al., 2013). In case of primary metabolites, an increase inmyo-inositol and sucrose was observed under water-deficit stresswhich may reflect their respective roles as osmoprotectants andprecursors for the formation of raffinose series sugars to enhancewater-deficit stress tolerance (Grimplet et al., 2009; Taji et al.,2002). Additionally a significant increase of proline in grapevineleaves (Doupis et al., 2011) and berries (Deluc et al., 2009) underwater deficit conditions was observed, assuming this amino acid toparticipate in protection against the formation of excessive reactiveoxygen species (ROS). The production of ROS, such as singlet oxy-gen, superoxide, hydrogen peroxide and hydroxyl radicals, is ageneric stress response of plants and these molecules play impor-tant roles in signaling to elicit defense responses (Vickers et al.,2009). Therefore plants synthesize different protective proteins,such as dehydrins, antioxidants and secondary metabolites toprevent damage of ROS to other proteins and cell membrane(Pardo, 2010).

Secondary metabolites have important ecological functionswithin the defense, protection and signaling mechanisms of plants.Applying selectively water deficit increased anthocyanin accumu-lation in grape skins and genes of the corresponding anthocyaninbiosynthesis pathway (Castellarin et al., 2007). In plants thesesubstances have also the ability to scavenge reactive oxygen speciesproduced upon abiotic stresses and it was shown that berry stemsfor example contain higher amounts of flavonoids, flavan-3-ols andflavonols than ripe berries (Doshi et al., 2006). Recently the

composition and contents of selected polyphenols in grapevineleaves according to their age and insertion level was investigated(Schoedl et al., 2012). The substances cis- and trans-resveratrol-3-O-glucoside, (þ)-catechin, quercetin-3-O-glucoside, caftaric acidand quercetin-3-O-glucuronide differed significantly among leafage groups on at least three of four sampling dates (Schoedl et al.,2012). Flavonoids are also produced upon UV-B radiation as anadaptive procedure in plants to reduce UV-B damage (Ibanez et al.,2008) and the correlation between physiological performance andquercetin-3-O-glucoside and kaempferol-3-O-glucoside in leavesof UV-B stressed vines have been observed (Schoedl et al., 2013).

Plants produce a wide spectrum of biogenic volatile organiccompounds (VOCs), which are largely lipophilic and have enoughvapor pressure to be released into the atmosphere (Loreto andSchnitzler, 2010). Typical substance classes of VOCs are alkanes,alkenes, aldehydes, ketones, aromatic compounds and terpenes(Weingart et al., 2012). Among these, terpenes constitute the mostcomplex group of volatile compounds in plants (Dudareva et al.,2004; Loreto and Schnitzler, 2010). The biosynthesis of volatileterpenes seems to be unaffected by mild drought stress conditions,but is significantly reduced when plants are heavily stressed bydrought (Lavoir et al., 2009). Biotic stresses, such as herbivoreattack, are inducing the emission of plant volatiles, especially greenleaf volatiles (GLVs), which are C6 aldehydes, alcohols and estersfrom the lipoxygenase cleavage of fatty acids, and terpenes(Dudareva et al., 2006).

In depth knowledge of the impact of severe drought stress onthe pattern of grapevine-quality related plant constituents is stillscarce. In this study we tested the effects of severe drought stresson the abundance of selected metabolite classes in leaves dodetermine early as well as high responsive substances to theapplied stress. Several metabolites could be identified and werefurther used to generate a model for the prediction of the physio-logical parameter leaf water potential. To our knowledge this is thefirst attempt to combine metabolite and physiological data withthis specific chemometric approach to characterize specifically se-vere drought stress in grapevine leaves with the final aim toestablish a prediction model for further application.

2. Material and methods

2.1. Drought stress treatment

The drought stress experiment was conducted under green-house (research greenhouse, University of Natural Resources andLife Sciences, Vienna, Austria) conditions in June 2010. 3-year oldsingle shoot V. vinifera plants (cultivar Pinot noir 18 Gm grafted onKober 5BB, 51 plants) potted in 3L pots in a sandy loam soil wereused. All plants were well watered (200 mL per day) at the begin-ning of the experiment (04.06.2010; DAY 0; 5 plants) andwater wassupplied to all control plants once every day (250 mL per day),whereas water supply of stressed plants was stopped. Physiologicalmeasurements and sampling of leaves took place on 07.06.2010(DAY 3; 5 control, 5 stressed plants), 10.06.2010 (DAY 6; 5 control, 5stressed plants) and 12.06.2010 (DAY 8; 5 control, 10 stressedplants). Due to very hot weather conditions in June 2010 theexperiment was stopped after 8 days and 12 available controlplants were used to restart the drought treatment with 6 controland 6 stressed plants on 11.06.2010 and all plants were measuredon 15.06.2010 (DAY 5). The mean leaf temperatures at middaywere: 25 �C (04.06.2010; DAY 0), 31.9 �C (07.06.2010; DAY 3),30.8 �C (15.06.2010; DAY 5), 35.8 �C (10.06.2010; DAY 6) and 35.7 �C(12.06.2010; DAY 8). The mean PAR radiation per day (measuredfrom 6:00 am till 7:00 pm) was 144.1 mmol m�2 s�1. Each plant wasused only once for physiological measurements and sampling of

M. Griesser et al. / Plant Physiology and Biochemistry 88 (2015) 17e26 19

leaves.The drought stress of all plants was monitored by measuring

various parameters as compiled in Table 1. On every day of theexperiment (day 0, 3, 5, 6, 8, Table 1) the pot weight and thevolumetric soil moisture content (ThetaProbe ML2x and handhelddata logger Moisture Meter HH2, Delta-T Devices, Cambridge,United Kingdom) was recorded. The water potential (PWSC Model3000, Soilmoisture Equipment Corporation, Santa Barbara, USA)was determined for the 6th leaf (representing the insertion level ofthe shoot from the basis) of every plant and measurement day.Chlorophyll fluorescence and gas exchange parameters of lightadapted leaves were determinedwith the 4th and 5th leaf, whereasdark adaptationwas performed only with the 5th leaf. Immediatelyafter these non-invasive measurements, the 5th leaf was harvested,frozen in liquid nitrogen and further used for the measurement ofpolyphenols, selected primary metabolites and volatiles (VOCs).

2.2. Physiological measurements

Chlorophyll fluorescence and gas exchange parameters of leaveswere measured using the Ciras-2 photosynthesis system (PP Sys-tems, Amesbury, Massachusetts, USA) equipped with a PLC6 leafcuvette and a chlorophyll fluorescence module (CFM) as describedin (Schoedl et al., 2012, 2013). Dark adaption was achieved bywrapping the leaves for at least 30 min in aluminum foil. The pulseamplitude modulation (PAM) based system uses a definedmeasuring area of 250 mm2, cuvette flow was set at 200 mL min�1,reference CO2 concentrationwas set to 381 ppm and reference H2Owas set to 50% of ambient. Two measurement types per lightadaption state were applied. Measurement type ‘FvFm DA’ persistsof measuring a dark adapted (DA) leaf which was exposed to a lightintensity of 6000 mmol m�2 s�1 for 2.5 s, the ‘Fo Prime LA’ mea-surement includes the measurement of a light adapted (LA) leaf.The type ‘phiPS2’ is applied both after the ‘FvFm DA’ and ‘Fo PrimeLA’ measurement, respectively after adaption to ambient lightconditions and fully re-oxidation of PSII reaction centers indicatedas ‘phiPS2 DA-LA’ and ‘phiPS2 LAeLA’.

2.3. Chemical analysis

Per plant and time point, all measurements of chemical pa-rameters were carried out on the leaf of the 5th node position per

Table 1Experimental design of the drought stress treatment and schedule of physiologicalmeasurements and biochemical screenings.

Drought stress experiment

Dates 4e12 June 2010Genotype Pinot noir

(18 Gm grafted on 5BBa, potted in 2007)Treatments and

durationdry/control (drought period: 8 days)

Experimentaldesign

5e11 plants per treatment

measurementdays

0, 3, 5, 6 and 8 DOTb

Treatment control pot weight (g) soil moisture (vol%)Physiological

Screeningscreening of 38 CHF and CERc parameters (4th and 5th leaf)d

and water potential (6th leaf) in dry leaves vs. controlleaves;

BiochemicalScreening

14 sugar compounds (5th leaf),16 polyphenolic compounds (5th leaf),95 volatile metabolites (5th leaf)

a 5BB, Rootstock (V. berlandieri x V. riparia).b DOT, day of treatment.c CHF, chlorophyll fluorescence; CER, gas exchange.d Leaves counted from the trunk to the apex.

plant. Plant tissue was homogenized with a ball mill (RetschMM400, Haan, Germany) under cold conditions to prevent thaw-ing. In total, 45 leaves were analyzed.

Polyphenols were determined as described previously (Schoedlet al., 2011, 2012). In brief, 500 mg of pulverized leaf samples wereextracted using 5 mL of 0.02% hydrochloric acid (v/v) in aqueous80% MeOH (v/v) during ultrasonication, followed by centrifugationfor 5 min at 3750 rpm (GS-6 Centrifuge, Beckman Coulter, Inc., Brea,CA). The supernatant was collected and one re-extraction step wasperformed. Two dilutions of combined extracts were injected intothe analytical system consisting of a QTrap 4000 LC-MS/MS system(Applied Biosystems, Foster City, CA) equipped with a TurboIon-Spray electrospray ionization (ESI) source and a 1100 series HPLCsystem (Agilent, Waldbronn, Germany). Chromatographic separa-tion was performed at 40 �C on a Gemini RP-18 column,100 � 2 mm inner diameter, 3 mm particle size (Phenomenex,Torrance, CA) protected with a guard column (Phenomenex, Tor-rance, CA) packed with the same material. The mobile phase con-sisted of (A) 0.5% formic acid in H2O and (B) 0.5% formic acid inMeOH and the flow rate was set at 0.4 mL min�1. Detection wascarried out by ESI-MS/MS in selected reaction monitoring (SRM)mode in negative polarity and for quantification of polyphenols,external calibration using standards in pure solvent was applied.

The preparation and measurement of volatile metabolites fol-lowed the protocol described by (Weingart et al., 2012). 105 ± 5 mgof leaf powder was weighed into 20 mL headspace glass vials. Thesolid phase microextraction e gas chromatography mass spec-trometry (SPME-GC-MS) measurement was performed as follows:For metabolite extraction with a PDMS/CAR/DVB 2 cm fiber thesample was heated to 90 �C for 30 min and then extracted for60 min at 90 �C. The fiber was then desorbed in the GC inlet at250 �C for 2 min. GC oven program started at 40 �C (2 min) andraised with 6 �C/min to 260 �C (hold 5 min). For chromatographicseparation a non-polar DB-5MS columnwas used. The scan range ofthe MSDwas 35e500 m/z at a scan rate of 3 spectra per second. Foridentification of metabolites we applied strict criteria: recordedspectra had to fit with a minimummass spectral match factor of 90to the NIST Wiley 2008 library (McLafferty, 2008) spectra leading toa list of preliminary identified metabolites. For final identificationthe RI had to bewithin amaximumdeviation of 2% of themedian RIreported in NIST Chemistry Webbook (http://webbook.nist.gov/chemistry/) for corresponding column phase and dimensions. Forseveral metabolites the identity was confirmed with authenticstandards. In total about 300 peaks per sample could be detected.Data processing was done with AMDIS software (version 2.65,www.amdis.net, (Stein, 1999)) and further on with Spectconnect(Styczynski et al., 2007) which is able to handle also the peaksremaining unidentified after AMDIS processing. The data process-ing led to mass spectra representing a total of 95 individual volatilemetabolites.

The analyses of selected primary metabolites followed themethod described recently (Becker et al., 2013). All referencecompounds, the internal standard O-methyl-a-D-galactopyrano-side, anhydrous pyridine, ethyl acetate, N,O-bis(trimethylsilyl)trifluoroacetamide (BSTFA), 4-(dimethylamino)pyridine (DMAP)and trimethylchlorosilane (TMCS) were purchased from Sigma-eAldrich (SigmaeAldrich Handels GmbH, Vienna, Austria). Allstandards, chemicals and reagents were of GC-grade and usedwithout further purification. Approximately 200 mg of finelyground leaf material (fresh weight) were extracted by adding 3 mLof a methanol/water solution (50/50, v/v) to the sample. The firstand second extraction was conducted at room temperature for60 min, the third replication at 50 �C for 60 min. The sample wasthen centrifuged at 4300 rpm for 10 min and the supernatant of thethree extraction replications was collected. After removing

M. Griesser et al. / Plant Physiology and Biochemistry 88 (2015) 17e2620

methanol by evaporation the remaining sample was lyophilized.The lyophilized sample was dissolved in 200 mL of pyridine whichcontained 200 mg mL�1 O-methyl-a-D-galactopyranoside as inter-nal standard (IS). After one hour of activation at room temperature,1.5 mg mL�1 of DMAP dissolved in 200 mL of pyridine and 200 mL ofthe derivatization reagent BSTFA containing 10% TMCS were addedand the mixture was heated to 70 �C for 2 h. The derivatizedsamples were kept at �20 �C until analysis. GC/MS analysis wasperformed on an Agilent 6890N gas chromatograph coupled withan Agilent 5973 mass selective detector. Column: HP-5MS(30 m � 0.25 mm� 0.25 mm; J&W Scientific, Folsom, CA, USA);carrier gas: helium; split/splitless injector: 280 �C; column flow:0.9 mLmin�1; purge flow: 32.4 mLmin�1 (0.6 min); oven program:50 �C (2 min), 5 �C min�1, 280 �C (20 min); MS: EI mode, 70 eV,source pressure: 1.13∙10�7 Pa, source temperature: 230 �C. Scanrange was set from 50 to 950 Da, except ISP, which ranged from 35to 950 Da. The derivatized samples were diluted with 600 mL ofethyl acetate and filtered prior to injection. Aliquots of 0.2 mL wereinjected in splitless mode using a 7683B autosampler (AgilentTechnologies, USA). Peak assignment and quantification wasaccomplished using MSD Chemstation E.2.01.1177 (Agilent Tech-nologies, USA). Peaks were assigned by comparing their retentiontimes and mass spectra with those of respective reference com-pounds in combination with the internal standard.

2.4. Data analyses

Differences between all treatments (control, dry 1 (3e5 days),dry 2 (6e8 days)) were tested non-parametrically with Krus-kaleWallis one-way analysis of variance and Mann-Whitney-U testwith significance levels of p< 0.05. The Bonferroni-Holm procedurewas used to correct for multiple testing.

PCA analyses were performed to detect significantly influencingmetabolites. These data analyses were performed using the pro-gram SPSS for Windows version 16.0 (IBM Corporation, New York,USA).

Chemometric analyses were performed using software in theprogramming language R (Version 3.02) (RCoreTeam, 2013). Thedataset obtained for all plant samples was separated into twogroups (class 1: drought stressed plants with n ¼ 25 and class 2:control plants with n ¼ 20), which included samples starting fromday 3 to day 8. All obtained variables (relative concentrations ofprimary metabolites, polyphenols and volatile metabolites) wereanalyzed with Student's t-test (p � 0.05) by comparing the mean ofboth groups to detect significantly differing variables betweencontrol and drought stressed plants. The found subset of potentiallydiscriminating variables was used to build a classification modelusing partial least squares discriminant analysis (PLS-DA) (Barkerand Rayens, 2003; Varmuza and Filzmoser, 2009). To assess thepredictive power and quality of the PLS-DA models the recentlydeveloped strategy “repeated double cross validation” (rdCV) wasperformed (Filzmoser et al., 2009). This method was used to buildregression models based on the t-test selected metabolites orcombinations of those and validated with rdCV to verify if theselected metabolites were able to separate plants correctly intotheir assigned groups. The developed model was then applied topredict dependent physiological parameters, hence indicating theselectedmodel variables as potential biomarkers for drought stress.

3. Results

3.1. Overview of physiological parameters and metabolites

The drought stress treatment was conducted under glasshouseconditions without temperature control for eight days with single

potted plants of the V. vinifera cv. Pinot noir Gm18 grafted on Kober5BB pruned as one shoot plants. The treatment had to be stoppedafter eight days due to a high temperature period in June 2010,therefore the plants were separated in three groups according tothe measured midday leaf water potential and the volumetric soilmoisture content (Vol%) (Fig. 1): control (n ¼ 26), dry 1 (3e5 days,n ¼ 10), dry 2 (6e8 days, n ¼ 15). The midday leaf water potentialcould differentiate significantly between both drought treatments,therefore this parameter was used in further analyses.

Chlorophyll fluorescence (CHF) and gas exchange parameters(CER) of leaves were measured, whereas 15 parameters of two lightadapted leaves (insertion 4 and 5) as well as 19 parameters of onedark adapted leaf (insertion 5) were obtained. The most importantparameters for CHF and CER of light adapted leaves are shown inTable 2 (seeAppendix S1 in Supporting Information for allmeasuredparameters). Furthermore three different groups of metaboliteswere determined: 11 primarymetabolites (PM),13 polyphenols (PP)and 95 volatiles (VOCs). Among the tested 7 primary metabolites(Table 3),11 polyphenols (Table 4) and 17 of the 40 identified volatilemetabolites (Table 5) were found to show significant differentabundances between the treatment variants. Volatile metaboliteswhich could not be identified according to our classification schemeare shown in Appendix S2 in Supporting Information.

3.2. Metabolites highly correlated with the leaf water potential

The midday leaf water potential could clearly differentiate be-tween our treatments. Correlation analyses of the water potentialwith measured metabolites could identify strong treatment effects(data not shown). Among primarymetabolites glyceric acid did showthe strongest positive correlation (r ¼ 0.789) (Fig. 2B) and ribose aswell as citric acid the strongest negative values with correlation co-efficients of�0.723 (Fig. 2C) and�0.686 (Fig. 2A) respectively. In thegroupofvolatiles, phenylacetaldehyde (Fig. 2F, r¼�0.846), 2-methyl-butanal (Fig. 2D, r ¼ �0.713) and benzaldehyde (Fig. 2E, r ¼ �0.655)were negatively correlatedwith the leafwater potential. The contentsof all studied polyphenols were confirmed to increase with droughtstress. The measured polyphenols kaempferol-3-O-glucoside(Fig. 2G) andquercetin-3-O-glucoside (Fig. 2H)were those substanceswith thehighest correlation coefficients of r¼�0.524 and r¼�0.541.Both substances showedasimilarpatternof response,whichseemstobe common for most polyphenols. The formation of the primarymetabolites glyceric acidwas reducedwith increasing drought stress,whereas the metabolite citric acid, ribose, 2-methyl-butanal phenyl-acetaldehyde, benzaldehyde, kaempferol-3-O-glucoside and quer-cetin-3-O-glucoside were increased, especially for severe droughtstress.

3.3. Discrimination of drought stressed plants by multivariateapproaches

Principal component analyses were performed separately withprimary metabolites, polyphenols and volatiles (Appendix S3).Metabolites which were significantly influence by both droughtstress treatments were used (Tables 3e5). The primary and volatilemetabolites were able to differentiate at least the severe droughtstressed from control and moderate drought stressed plants, whilethe results from polyphenols were less suitable. None of the threemetabolite groups could discriminate controls form both stresstreatments. Therefore in a next step a supervised chemometricanalyses was conducted. For simplicity, data of both drought stresslevels were combined and not analyzed separately. Primarily wefocused on measured volatiles, as according to the PCA thediscrimination between control and drought stress plants seemedpossible. A selection of 28 volatiles which did significantly differ

Fig. 1. Results obtained for the midday leaf water potential (A) and the volumetric soil moisture content (B) of control and drought stressed plants. Values represent arithmeticmeans and standard deviations from 51 experimental plants. The treatments can be differentiated: control (n ¼ 26), dry 1 (3e5 days of drought stress, n ¼ 10), dry 2 (6e8 days ofdrought stress, n ¼ 15). Different letters (a, b, c) indicate statistical significant differences.

Table 2List of selected CHF/CER parameters measured in leaves of V. vinifera cv. Pinot noir Gm18 plants applying a drought stress treatment. Data shown are arithmetic means withstandard deviation (control plants n ¼ 26; dry 1 n ¼ 10; dry 2 n ¼ 15). Significant differences were tested with Kruskal Wallis Test (p � 0.05; p values below 0.05 indicatesignificant differences between groups and are indicated with bold letters) and Mann Whitney U-Test. Letters indicate statistical difference (aec, highest content value withletter a). Correction for multiple testing was performed according to Bonferroni-Holm. Results of all parameters measured are provided in Appendix S1 in SupportingInformation.

Code Physiological parameter Unit Measured value (mean ± standard deviation) Kruskal Wallis test

Control Dry 3e5 DOT Dry 6e8 DOT

WP2 Water potential leaf Mpa (*�1) �0.94 ± 0.27 a �1.64 ± 0.26 b �2.58 ± 0.54 c <0.000PHY 20 EVAP (Fo Prime LA) transpiration rate mmol m�2 s�1 2.3 ± 1.2 a 0.4 ± 0.2 b 0.3 ± 0.1 b <0.000PHY 22 gs (Fo prime LA) stomatal conductance mmol m�2 s�1 94.2 ± 82.6 a 10.9 ± 4.7 b 6.3 ± 1.3 c <0.000PHY 24 Ci (Fo prime LA) sub-stomatal CO2 concentration ppm 209.1 ± 36.8 c 350.5 ± 77.3 b 418.1 ± 38.7 a <0.000PHY 26 Pn (Fo Prime LA) photosynthesis rate mmol m�2 s�1 5.4 ± 3.5 a �0.1 ± 0.7 b �0.5 ± 0.2 b <0.000PHY 30 ɸPS20 (phiPS2 LA) yield Units 0.219 ± 0.061 a 0.098 ± 0.051 b 0.057 ± 0.017 c <0.000PHY 31 qP0 (phiPS2 LA) photochemical quenching Units 0.597 ± 0.119 a 0.350 ± 0.247 b 0.163 ± 0.077 c <0.000PHY 32 ETR0(phiPS2 LA) electron transport rate mmol m�2 s�1 92 ± 26 a 41 ± 21 b 24 ± 7 c <0.000

M. Griesser et al. / Plant Physiology and Biochemistry 88 (2015) 17e26 21

between control and drought stress treatment (t-test, p� 0.05, datanot shown) was used for discriminant partial-least squaresregression (PLS-DA) in combination with repeated double crossvalidation (rdCV). Applying PLS regression with 28 VOC concen-trations as x-variables (regressors) and the leaf water potential(WP) as the dependent physiological parameter resulted in goodcalibration models for prediction of WP, correlation coefficient of(0.824), and a very good separation of plants according to controland stress treatment (Fig. 3A). In a second step the number ofmetabolites used to establish the PLS-DA model was reduced, ac-cording to the strongest correlation as shown in Fig. 2, to fivebiochemical compounds, namely: glyceric acid, ribose, citric acid,

Table 3List of measured primary metabolites in leaves of V. vinifera cv. Pinot noir Gm18 plants adeviations of control (n¼ 26) plants, dry 1 (3e5 days, n¼ 10) and dry 2 (6e8 days, n¼ 15)Test (p� 0.05; p values below 0.05 indicate significant different between groups and are s(a-c, highest content value with letter a). Correction for multiple testing was performed

Code Primary metabolite Unit Measured values (m

Control

PM 1 Succinic acid mg g�1 DW 37.2 ± 4.6 bPM 2 Glyceric acid mg g�1 DW 420.3 ± 146.9 aPM 3 Malic acid mg g�1 DW 2051.3 ± 363.7 aPM 4 Tartaric acid mg g�1 DW 2267.5 ± 303.6 bPM 5 D(�)-ribose mg g�1 DW 19.7 ± 5.3 bPM 6 Citric acid mg g�1 DW 311.9 ± 112.6 cPM 7 D(�)-fructose mg g�1 DW 1606.1 ± 276.4PM 8 D(þ)-glucose mg g�1 DW 2074.7 ± 241.6PM 9 Palmitic acid mg g�1 DW 43.7 ± 5.2 bPM 10 Myo-inositol mg g�1 DW 287.8 ± 111.3PM 11 S(þ)-sucrose mg g�1 DW 616.4 ± 121.9

phenylacetaldehyde and 2-methylbutanal (Fig. 3B). As shown inFig. 3A and B the established PLS-DAmodels are equally well suitedto predict the leaf water potential of the samples. They showalmostthe same predictive power with r ¼ 0.834 for WP predicted by thefive selected biochemical compounds and r ¼ 0.824 for the modelsbased on the 28 VOCs. This indicates that the five selected com-pounds are indeed affected by drought stress and have hence thepotential power to predict drought stress.

4. Discussion

In this study we aimed to describe the effect of severe drought

pplying a drought stress treatment. Data represent arithmetic means with standarddrought stress treated plants. Significant differences were tested with Kruskal Wallishown as bold letters) andMannWhitney U-Test. Letters indicate statistical differenceaccording to Bonferroni-Holm.

ean ± standard deviation) Kruskal Wallis test

Dry 3e5 DOT Dry 6e8 DOT

40.0 ± 9.7 b 49.0 ± 20.8 ab <0.017178.7 ± 87.9 b 56.9 ± 14.1 c <0.000

2153.0 ± 357.5 a 1190.1 ± 542.0 b <0.0002252.5 ± 424.6 b 3608.9 ± 1995.3 a <0.001

25.0 ± 10.0 b 56.8 ± 21.6 a <0.000477.6 ± 163.0 b 826.8 ± 249.3 a <0.000

1426.5 ± 451.8 1870.1 ± 1020.3 0.4511871.0 ± 310.3 2423.9 ± 1110.4 0.136

43.8 ± 4.7 b 74.6 ± 54.7 ab <0.028441.1 ± 370.5 307.1 ± 155.7 0.655498.1 ± 162.6 811.6 ± 625.7 0.068

Table 4List of determined polyphenols in leaves of V. vinifera cv. Pinot noir Gm18 plants applying a drought stress treatment. Data represent arithmetic meanswith standard deviationsof control (n ¼ 26) plants, dry 1 (3e5 days, n ¼ 10) and dry 2 (6e8 days, n ¼ 15) drought stress treated plants. Significant differences were tested with Kruskal Wallis Test(p � 0.05; p values below 0.05 indicate significant differences between groups and are shown in bold letters) and Mann Whitney U-Test. Letters indicate statistical difference(a-c, highest content value with letter a). Correction for multiple testing was performed according to Bonferroni-Holm.

Code Metabolite Unit Measured values (mean ± standard deviation) Kruskal Wallis test

Control Dry 3e5 DOT Dry 6e8 DOT

PP 1 4-coumaric acid mg g�1 DW 0.01 ± 0.00 b 0.01 ± 0.00 b 0.02 ± 0.02 a <0.000PP 2 Caffeic acid mg g�1 DW 0.05 ± 0.02 b 0.05 ± 0.01 b 0.10 ± 0.12 a <0.000PP 3 Ferulic acid mg g�1 DW 0.12 ± 0.05 b 0.12 ± 0.10 b 0.21 ± 0.12 a <0.000PP 4 cis-resveratrol-3-O-glucoside mg g�1 DW 3.3 ± 1.9 b 3.0 ± 2.1 b 17.1 ± 15.1 a <0.000PP 5 trans-resveratrol-3-O-glucoside mg g�1 DW 0.9 ± 0.7 b 1.0 ± 1.0 b 4.9 ± 5.3 a <0.001PP 6 (þ)-catechin mg g�1 DW 7.0 ± 7.3 6.6 ± 7.0 19.3 ± 27.9 0.191PP 7 (�)-epicatechin mg g�1 DW 0.5 ± 0.4 b 0.7 ± 0.5 b 3.3 ± 5.9 ab <0.003PP 8 Caftaric acid mg g�1 DW 114.7 ± 102.8 130.5 ± 107.0 205.0 ± 225.7 0.428PP 9 (�)-epicatechin gallate mg g�1 DW 0.2 ± 0.2 b 0.3 ± 0.2 b 0.9 ± 1.5 ab <0.010PP 10 kaempferol-3-O-glucoside mg g�1 DW 1.1 ± 0.1 b 1.1 ± 0.7 b 6.5 ± 4.7 a <0.000PP 11 cyanidin-3-O-glucoside mg g�1 DW 0.1 ± 0.1 b 0.1 ± 0.0 b 1.2 ± 1.6 a <0.000PP 12 quercetin-3-O-glucoside mg g�1 DW 11.7 ± 3.1 b 11.2 ± 4.0 b 40.3 ± 20.1 a <0.000PP 13 quercetin-3-O glucuronide Area 7654395 ± 2672234 b 8372541 ± 3422453 b 19456045 ± 14106735 a <0.002

Table 5List of determined and identified volatiles in leaves of V. vinifera cv. Pinot noir Gm18 plants applying a drought stress treatment. Values for volatile metabolites representchromatographic peak areas of extracted ion chromatograms. Data shown are arithmetic means with standard deviations of control (n ¼ 26) plants, dry 1 (3e5 days, n ¼ 10)and dry 2 (6e8 days, n ¼ 15) drought stress treated plants. Significant differences were tested with Kruskal Wallis Test (p � 0.05; p values below 0.05 indicate differencesbetween groups and are shown with bold letters) and Mann Whitney U-Test. Letters indicate statistical difference (a-c, highest content value with letter a). Correction formultiple testing was performed according to Bonferroni-Holm. VOCid (volatile compound identified).

Code Metabolite Unit Measured values (mean ± standard deviation) Kruskal Wallis test

Control Dry 3e5 DOT Dry 6e8 DOT

VOCid 1 2-Methylbutanal Area 1358.6 ± 2513.2 b 1717.9 ± 1013.9 b 7932.1 ± 7413.2 a <0.000VOCid 2 1-Penten-3-one Area 4298.1 ± 2313.7 5514.4 ± 2983.0 4313.2 ± 2507.4 0.611VOCid 3 (E)-2-Pentenal Area 8739.3 ± 3265.1 a 10601.3 ± 3942.1 a 5003.2 ± 2498.9 b <0.000VOCid 4 Methylbenzene Area 12053.5 ± 28459.9 b 16106.7 ± 37421.1 b 101634.8 ± 125103.4 a <0.001VOCid 5 (Z)-3-Hexenal Area 34425.3 ± 16747.6 47477.5 ± 23350.9 29348.9 ± 17803.4 0.078VOCid 6 Hexanal Area 85337.5 ± 37622.6 116209.4 ± 58539.4 74635.8 ± 27842.4 0.116VOCid 7 Furfural Area 12349.2 ± 4872.4 b 24206.5 ± 6816.1 a 25401.9 ± 15531.1 a <0.000VOCid 8 (E)-2-Hexenal Area 644620.6 ± 178359.9 724631.6 ± 227611.5 576546 ± 169661.1 0.210VOCid 9 2-Ethylthiopene Area 1184.1 ± 424.4 1307.8 ± 564.9 1191.7 ± 658.4 0.735VOCid10 Heptanal Area 2542.1 ± 1470.2 2349.2 ± 1114.6 2959.4 ± 1979.0 0.780VOCid 11 (E,E)-2,4-Hexadienal Area 76031.8 ± 21730.7 91487.6 ± 25430.7 60226.2 ± 35675.7 0.053VOCid 12 5-Methyl-2-Furancarboxaldehyde Area 2031.7 ± 468.1 3024.3 ± 691.4 2969.6 ± 953.0 m.v.VOCid 13 Benzaldehyde Area 27437.7 ± 7574.5 b 42386.9 ± 11150.1 a 68084.3 ± 35606.0 a <0.000VOCid 14 6-Methyl-5-hepten-2-one Area 2496.6 ± 1091.3 a 3530.6 ± 1794.3 a 6888.6 ± 8142.5 a <0.038VOCid 15 2-Pentylfuran Area 2317.8 ± 910.7 2816.9 ± 1111.8 2810.3 ± 1453.2 0.458VOCid 16 cis-2-(2-Pentenyl)furan Area 6373.4 ± 2900.8 9816 ± 5655.7 7035.5 ± 4325.3 0.113VOCid 17 Octanal Area 12369.4 ± 8042.2 12140.9 ± 6877.6 13811.4 ± 10127.7 0.998VOCid 18 (E,E)-2,4-Heptadienal area 9030.9 ± 2625.9 9056.4 ± 2569.9 10097.0 ± 3814.4 0.561VOCid 19 Phenylacetaldehyde Area 23978.1 ± 17258.9 c 117463.9 ± 51813.7 b 226350.1 ± 112435.9 a <0.000VOCid 20 4-Methylbenzaldehyde Area 1305.8 ± 367.4 b 1841.2 ± 471.0 a 2051.1 ± 1550.1 b <0.016VOCid 21 p-Cymenene Area 838.9 ± 444.3 1176.8 ± 637.8 901.3 ± 246.4 0.040VOCid 22 Nonanal Area 126965.2 ± 72470.4 103377.8 ± 75718.9 158019.1 ± 110413.4 0.378VOCid 23 4-Ethylbenzaldehyde Area 18231.5 ± 6155.4 b 25054.9 ± 8166.6 a 8851.0 ± 6456.6 c <0.000VOCid 24 2,4-Dimethylbenzaldehyde Area 9294.8 ± 2911.2 12468.9 ± 3646.8 10445.5 ± 4999.4 0.054VOCid 25 5-Hydroxymaltol Area 3586.4 ± 2012.8 a 3814.1 ± 1364.1 a 1798.8 ± 3844.7 b <0.000VOCid 26 20-Methylacetophenone Area 2741.2 ± 969.3 3133.8 ± 791.1 2579.2 ± 1385.9 0.192VOCid 27 1,3-Cyclohexadiene-1-carboxaldehyde Area 8263.1 ± 2547.9 b 14005.7 ± 3228.7 a 12301.9 ± 5154.8 a <0.000VOCid 28 Decanal Area 16781.7 ± 10488.9 18908 ± 5876.6 21582.8 ± 8442.3 0.068VOCid 29 beta-Cyclocitral Area 19481.9 ± 8045.7 a 22584.8 ± 6918.2 a 14530.1 ± 8840.8 b <0.007VOCid 30 (E,E)-2,4-Decadienal Area 1726.8 ± 471.3 2133.9 ± 675.5 2360.9 ± 1121.7 0.076VOCid 31 Eugenol Area 4336.5 ± 5088.4 3210 ± 1749.4 5492.9 ± 3743.4 0.054VOCid 32 1,2-Dihydro-1,1,6-trimethylnaphthalene Area 12099.5 ± 4305.6 b 17290.2 ± 4576.5 a 15906.6 ± 8095.9 ab <0.015VOCid 33 Geranyl acetone Area 53111.2 ± 18274.8 b 73444.3 ± 23198.9 b 108890.3 ± 104305.1 ab <0.017VOCid 34 beta-Ionone Area 227659.7 ± 85679.6 a 263706.5 ± 67423.7 a 126568.9 ± 43976.6 b <0.000VOCid 35 5,6,7,7A-Tetrahydro-2(4H)-benzofuranone Area 120165.8 ± 26406.3 a 128810.6 ± 13645.3 a 90920.8 ± 25735.0 b <0.000VOCid 36 Tetradecanal Area 21202.6 ± 12397.6 a 9233.6 ± 4673.3 b 12929.3 ± 8862.5 b <0.002VOCid 37 Benzyl benzoate Area 1300.8 ± 589.9 1407.5 ± 988.1 1045.1 ± 827.9 m.v.VOCid 38 Hexahydrofarnesyl acetone Area 47918.6 ± 20573.9 42156.1 ± 11911.7 49604.9 ± 23003 0.691VOCid 39 Isophytol Area 30617.5 ± 17495.6 23497.9 ± 14393.9 26178.8 ± 10673.9 0.411VOCid 40 Hexadacanoic acid Area 70884.5 ± 28252.2 65060.4 ± 27295.5 61622.4 ± 47172 0.169

M. Griesser et al. / Plant Physiology and Biochemistry 88 (2015) 17e2622

Fig. 2. Selected biochemical primary metabolites, volatiles and polyphenols are shown, which were strongly correlated with the midday leaf water potential (in MPa). Data areshown as median values with minimal and maximal values separated according to the treatments. A. citric acid (r ¼ �0.686), B. glyceric acid (r ¼ 0.789), C. ribose (r ¼ �0.723), D. 2-methyl-butanal (r ¼ �0.713), E. benzaldehyde (r ¼ �0.655), F. phenylacetaldehyde (r ¼ �0.846), G. kaempferol-3-O-glucoside (r ¼ �0.524), H. quercetin-3-O-glucoside (r ¼ �0.541).Control (n ¼ 26), dry 1 (3e5 days, n ¼ 10), dry 2 (6e8 days, n ¼ 15).

M. Griesser et al. / Plant Physiology and Biochemistry 88 (2015) 17e26 23

stress on three different groups of metabolites in grapevine leavesto identify early responsive metabolic pathways and to evaluatetheir potential to discriminate between control and stressed plants.

4.1. Experimental setup and effects on plant physiology

In our experiment a severe drought stress treatment with astomatal conductance (gs) of 10 and 6 mmol m�2 s�1 respectivelyfor dry1 (3e5 DOT) and dry2 (6e8 DOT) was applied. The responseof photosynthesis to drought stress is correlated with stomatalconductance (Escalona et al., 1999) and can be divided in threephases of drought stress: no stress (gs > 150 mmol m�2 s�1),moderate (50 mmol m�2 s�1 <gs < 150 mmol m�2 s�1) and severedrought stress (gs< 50mmolm�2 s�1) (Flexas andMedrano, 2002).

The control plants were watered in the morning, which led to amoderate drought stress according to the measured stomatalconductance of 100 mmol m�1 s�1. This low stomatal conductanceand the reduced photosynthetic rate can also indicate climaticstress induced by high temperatures and high vapor pressuredeficit (VPD), which is determined by air temperature and relativehumidity. It has been shown that the VPD is highly correlated withthe leaf and stemwater potential in vineyards (Williams and Baeza,2007). This effect should be circumvented by controlling environ-mental conditions during the experiment to prevent a midday toafternoon depression of leaf water potential, as the water flow intograpevine leaves under high irradiance and vapor pressure deficit isinsufficient to compensate evapotranspiration (Chaves et al., 2010;Schultz, 2003). A constant watering during daytime or the use of

Fig. 3. Results for modeling the leaf water potential by the concentration of selected volatiles, using the strategy PLS-rdCV. (A) 28 volatiles, correlation coefficient betweenexperimental WP and test set predicted WP is 0.83. (B) Subset with five metabolites (citric acid, glyceric acid, ribose, phenylacetaldehyde and 2-methyl-butanal); correlationcoefficient is 0.75. The variability of the predictions is caused by the repeated random splits into calibration and test set (rdCV).

M. Griesser et al. / Plant Physiology and Biochemistry 88 (2015) 17e2624

bigger pot sizes would have decreased this effect on control plants.We determined the midday leaf water potential (Jmidday) as anindicator of the drought stress of each plant. As leaf samples in ourexperiment were taken at noon, we decided to measure theJmidday as a parameter determined at the time of sampling. Otherstudies have reported that the stem water potential (Jstem) is amore reliable indicator of the grapevine water status (Williams andBaeza, 2007). It has also been reported that the pre-dawn leaf waterpotential (Jpre-dawn) shows a better correlation with themeasured stomatal conductance (gs) as Jmidday (Intrigliolo andCastel, 2006). According to these observations, the measurementof the Jstem should be considered apart from Jmidday in followup experiments. Concerning physiological measurement it wasshown that moderate drought stress represents a transition be-tween predominant stomatal to non-stomatal limitations, whennet photosynthesis and substomatal CO2 concentration (Ci) de-creases and water use efficiency increases to reachmaximum levels(Medrano et al., 2002). Whereas under severe drought stress non-stomatal limitations of photosynthesis are dominant and theselimitations cannot be restored even when using very high CO2concentrations (Flexas et al., 1999). Our results are in agreementwith this observation.

4.2. Primary and secondary metabolites affected by severe droughtstress in grapevine

Two different drought stress treatments were applied: shortstress with 3e5 DOT (dry1) and prolonged stress with 6e8 DOT(dry2). The content of selected primary metabolites, polyphenolsand volatile metabolites were determined in grapevine leaves ofcontrol and stressed plants. Targeted analytical methods were usedand allowed the absolute quantification of the tested primarymetabolites as well as of the polyphenols. For the VOCs an untar-geted metabolomics approach was selected however, whichenabled the detection and (putative) identification of a total of 95metabolites. For these metabolites comparative relative quantifi-cation was based on peak areas of extracted ion chromatogram(EIC) peaks. Substances significantly influenced by our droughtstress treatment were detected among the different groups ofmetabolites analyzed.

The primary metabolites citric acid, palmitic acid, succinic acidand tartaric acid as well as D(�)-ribose, were increased in stressedgrapevine leaves. Glyceric acid and malic acid were decreased.Interestingly, citric acid and glyceric acid were also significantlydifferent between both drought stress levels. In a former study, the

accumulation of catechin, sucrose, alanine and myo-inositol and adecrease of glutamate and tartrate was observed in the pulp ofgrape berries in drought stressed plants of Cabernet Sauvignon(Grimplet et al., 2009). The increase in myo-inositol and sucroseunder water-deficit stress may reflect their respective roles asosmoprotectants and precursors for the formation of raffinose-derived sugars to enhance water-deficit stress tolerance (Delucet al., 2009; Grimplet et al., 2009; Taji et al., 2002). In our caseD(�)-fructose, D(þ)-glucose, D(þ)-sucrose were not significantlyinfluenced in drought stressed grapevine leaves, maybe reflectingadaptation differences in leaves compared to berries. This may alsoreflect differences in the regulation and adaptation to stress situ-ations in source and sink organs of a plant.

Polyphenols are important quality traits of grapes and thereofproduced wines, last but not least due to their ability to act asantioxidants. This mode of action is considered to be beneficial forthe plant itself as well as reactive oxygen species produced uponabiotic stresses can be effectively scavenged (Doshi et al., 2006).According to the results of our study, prolonged drought stressleads to an increase of the formation of the polyphenols cis-resveratrol-3-O-glucoside, trans-resveratrol-3-O-glucoside,(�)-epicatechin, (�)-epicatechin gallate, kaempferol-3-O-gluco-side, cyanidin-3-O-glucoside, quercetin-3-O-glucoside and quer-cetin-3-O-glucuronide. All of these substances were increased uponthe prolonged drought stress treatment (6e8 DOT), whereas thiseffect could not be observed after the short term treatment (3e5DOT). Polyphenols in grapevine leaves can be therefore consideredas indicators for severe and prolonged drought stress. In agreementwith our findings, higher amounts of flavonoids, flavan-3-ols andflavonols formed in response to abiotic stress had also beenobserved in grape berry stems in an earlier study (Doshi et al.,2006). Analyses of polyphenols in leaves are scarce, as the effectsand possibilities to increase polyphenols as quality traits in grapeswere the main focus of previous research (Deluc et al., 2009). Fla-vonoids are also produced upon UV-B radiation as an adaptiveprocedure in plants to reduce UV-B damage (Ibanez et al., 2008).The flavonols quercetin-3-O-glucoside and kaempferol-3-O-gluco-side showed a significant correlation with chlorophyll fluorescenceparameters (Schoedl et al., 2013) indicating their possible applica-tion as UV-B stress indicators in addition to the biomarkers quer-cetin-3-O-glucoside and kaempferol-3-O-glucoside (Schoedl et al.,2013). The flavonols quercetin-3-O-glucoside and kaempferol-3-O-glucoside were also induced in our experiment. A correlationbetween kaempferol-3-O-glucoside and FvFm as determined withUV-B stress (Pearson's correlation coefficient of 0.759) (Schoedl

M. Griesser et al. / Plant Physiology and Biochemistry 88 (2015) 17e26 25

et al., 2013) could not be observed in our experiment (Spearmancorrelation coefficient �0.564). Polyphenols were especiallyincreased in our experiment 6e8 days of drought stress treatment.Unfortunately temperatures and UV-B radiation were not recordedduring the experiment and side effects of elevated UV-B radiationand temperature sums on our data can therefore not be excluded.The elevated contents may reflect the need of the plant to scavengehigh ROS contents caused by high irradiation and low photosyn-thetic activity due to drought stress.

Around 600 volatile metabolites can be found in a single GC-MSchromatogram of grapevine leaves. Applying very stringentassignment criteria, 63 of these metabolites were identified(Weingart et al., 2012). In our experiment 40 volatiles were suc-cessfully assigned according to the same criteria and out of these 17were significantly affected by the applied drought stress treatment.Ten metabolites were increased and seven decreased in their con-tent in drought stressed grapevine leaves. Several aldehydes wereamong the significantly influenced substances which may act asROS scavengers that can be easily oxidized. In this context the al-dehydes can be products of lipoxygenase cleavage of fatty acids as aconsequence of ROS inactivation. It has been observed that thebiosynthesis of volatile terpenes was not affected by mild droughtstress, but is significantly reduced when plants are heavily stressedby drought (Lavoir et al., 2009). Interestingly this was not the casein our experiment where the contents of more than half of theaffected metabolites increased. A higher emission of green leafvolatiles (C6-alcohols, -aldehydes and -esters) was also observed inpotted apples trees subjected to severe water stress (Ebel et al.,1995). In general abiotic stress is reducing stomatal conductanceand influences negatively the photosynthetic reaction pathways,whereas the emission of volatiles, such as isoprene was notchanged (Loreto et al., 1996). Alternative sources of carbon, likerespiration- or starch degradation products, are meanwhile usedfor the synthesis of isoprene and stop their supply once the plantsare rewatered as reported recently (Loreto and Schnitzler, 2010).The biotic stresses, such as herbivore attack, induce the emission ofplant volatiles, especially green leaf volatiles (GLVs), which are C6-aldehydes, alcohols and esters from the lipoxygenase cleavage offatty acids, and terpenes (Dudareva et al., 2006; Niinemets et al.,2013). This has also recently been observed in phylloxera-infestedroot tips of the cultivar Teleki 5C (V. berlandieri � V. riparia)(Lawo et al., 2011). In that work a total of 38 metabolites could beidentified and 12 of these substances were significantly increasedin phylloxera infested nodosities. The elevated compoundsbelonged to terpenoids, phenylpropanoids and C6-compoundsindicating an increase in metabolic activity within the mevalonate,phenylpropanoid and lipoxygenase pathways (Lawo et al., 2011).

4.3. Selection of metabolites as indicators for severe drought stressin grapevine

According to correlations with the midday leaf water potential,five metabolites namely citric acid, glyceric acid, ribose, phenyl-acetaldehyde and 2-methylbutanal were selected to perform adiscriminant partial-least squares regression (PLS-DA). The modelwas validated with repeated double cross validation (rdCV). ThePLS-DAmodel was used to predict the leaf water potential based onthe measured plant metabolites. With the selected five metabolitesthe model could correctly classify 79% of all control plants and 99%of all plants observing drought stress. The correlation coefficientbetween WP predicted by a PLS model and the experimental WPwas high (0.83), but not all drought stressed plants did show a cleardistinction from the control plants. This overlap of values ispossibly due to the observed moderate drought stress of controlplants, or the high climatic demand due to increased VPD or the low

difference in metabolite profiles between the two groups on day 3.Future experiment validation of these metabolites with adaptedexperimental conditions should be conducted to evaluate theirapplicability as possible markers for severe drought stress. Bio-markers for one specific type of environmental stress are scarce anddepend on the degree of plant adaptations to stresses and the ho-mogeneity of the investigated population (Ernst and Peterson,1994). It was already proposed that a stringent physiologicalassessment is necessary to analyze and confirm candidate bio-markers, both temporally and spatially in the plant (Schoedl et al.,2013). Data analysis is also an important point in handling of largedata sets and identification of biomarkers, as visual observation isextremely difficult and time consuming (Hantao et al., 2013). Inbiological experiments where the variation within a group can begreater than the variation between two groups the use of partialleast squares (PLS) is an interesting alternative to principalcomponent analysis (PCA) (Hobro et al., 2010). The use of U-PLS-DA(unfolded-PLS-DA) with orthogonal signal correction (OSC) wassuccessfully employed with volatile metabolites to differentiatebetween Eucalyptus globulus leaves with and without infectionwith the necrotroph fungus Teratrophaeria nubilosa. A predictionefficiency of 100% was obtained (Hantao et al., 2013). In the pre-sented work PLS-DA was used (1) to detect characteristic metabo-lites for either control or drought stressed plants, (2) to findcombinations of metabolites that lead to a discrimination betweendrought stressed and control plants and (3) to detect possible linksbetween physiological parameters and measured metabolites.

5. In conclusion

We could show that the content of different groups of primaryand secondary metabolites is significantly influenced by severedrought stress in grapevine leaves. The content of the majority ofthe metabolites (around 60% of primary metabolites, around 85% ofpolyphenols and about 40% of the detected and identified VOCs)increased upon drought stress treatment. Among these especiallythe primary metabolites citric acid and glyceric acid were stronglyinfluenced by the short as well as the prolonged drought stresstreatment, whereas all polyphenols were only induced upon theprolonged drought stress treatment. Further studies are necessaryto validate our candidates, citric acid, glyceric acid, ribose,kaempferol-3-O-glucoside and quercetin-3-O-glucoside, aspossible indicators for severe drought stress in grapevine. Addi-tionally PLS regression was applied to establish a model based onselected metabolites to predict physiological parameters, in ourcase the leaf water potential. In our point of view this methodseems promising and suitable to combine analytical data into amodel to discriminate specific biological treatments as in our casedrought stress and predicts other measureable parameters.

Authors' contribution

MG, performed the experiment, measured physiologicalparameter, analyzed data and wrote the article; GW, analyzed andidentified volatiles; KSH, analyzed polyphenols; NM, performedPLS-DA analyses; MB, analyzed primarymetabolites; KV, developedrdCV; FL, primary metabolites analyses; RS, volatile analyses, dis-cussing of results, writing of manuscript; AF, concept of experi-ment, discussing results, writing of manuscript.

Acknowledgment

The work was supported by the Austrian Federal Ministry ofAgriculture, Forestry, Environment and Water Management (proj-ect “Physiological Fingerprinting in Viticulture”, project number

M. Griesser et al. / Plant Physiology and Biochemistry 88 (2015) 17e2626

100196).

Appendix A. Supplementary data

Supplementary data related to this article can be found at http://dx.doi.org/10.1016/j.plaphy.2015.01.004.

References

Barker, M., Rayens, W., 2003. Partial least squares for discrimination. J. Chemom. 17,166e173.

Becker, M., Zweckmair, T., Forneck, A., Rosenau, T., Potthast, A., Liebner, F., 2013.Evaluation of different derivatisation approaches for gas chromatographic-massspectrometric analysis of carbohydrates in complex matrices of biological andsynthetic origin. J. Chromatogr. A 1281, 115e126.

Castellarin, S.D., Matthews, M.A., Di Gaspero, G., Gambetta, G.A., 2007. Water def-icits accelerate ripening and induce changes in gene expression regulatingflavonoid biosynthesis in grape berries. Planta 227, 101e112.

Chaves, M.M., Zarrouk, O., Francisco, R., Costa, J.M., Santos, T., Regalado, A.P.,Rodrigues, M.L., Lopes, C.M., 2010. Grapevine under deficit irrigation: hints fromphysiological and molecular data. Ann. Bot. 105, 661e676.

Deluc, L.G., Quilici, D.R., Decendit, A., Grimplet, J., Wheatley, M.D., Schlauch, K.A.,Merillon, J.M., Cushman, J.C., Cramer, G.R., 2009. Water deficit alters differen-tially metabolic pathways affecting important flavor and quality traits in grapeberries of Cabernet Sauvignon and Chardonnay. BMC Genomics 10.

Doshi, P., Adsule, P., Banerjee, K., 2006. Phenolic composition and antioxidant ac-tivity in grapevine parts and berries (Vitis vinifera L.) cv. Kishmish Chornyi(Sharad Seedless) during maturation. Int. J. Food Sci. Technol. 41, 1e9.

Doupis, G., Chartzoulakis, K., Beis, A., Patakas, A., 2011. Allometric and biochemicalresponses of grapevines subjected to drought and enhanced ultraviolet-B ra-diation. Aust. J. Grape Wine Res. 17, 36e42.

Dry, P.R., Loveys, B.R., During, H., 2000. Partial drying of the rootzone of grape. II.Changes in the pattern of root development. Vitis 39, 9e12.

Dudareva, N., Negre, F., Nagegowda, D.A., Orlova, I., 2006. Plant volatiles: recentadvances and future perspectives. Crit. Rev. Plant Sci. 25, 417e440.

Dudareva, N., Pichersky, E., Gershenzon, J., 2004. Biochemistry of plant volatiles.Plant physiol. 135, 1893e1902.

Ebel, R.C., Mattheis, J.P., Buchanan, D.A., 1995. Drought stress of apple-trees altersleaf emissions of volatile compounds. Physiol. Plant. 93, 709e712.

Ernst, W.H., Peterson, P.J., 1994. The role of biomarkers in environmental assess-ment (4). Terrestrial plants. Ecotoxicology 3, 180e192.

Escalona, J.M., Flexas, J., Medrano, H., 1999. Stomatal and non-stomatal limitationsof photosynthesis under water stress in field-grown grapevines. Aust. J. PlantPhysiol. 26, 421e433.

Filzmoser, P., Liebmann, B., Varmuza, K., 2009. Repeated double cross validation.J. Chemom. 23, 160e171.

Flexas, J., Escalona, J.M., Medrano, H., 1999. Water stress induces different levels ofphotosynthesis and electron transport rate regulation in grapevines. Plant CellEnviron. 22, 39e48.

Flexas, J., Bota, J., Loreto, F., Cornic, G., Sharkey, T.D., 2004. Diffusive and metaboliclimitations to photosynthesis under drought and salinity in C(3) plants. PlantBiol. (Stuttg) 6, 269e279.

Flexas, J., Medrano, H., 2002. Drought-inhibition of photosynthesis in C-3 plants:stomatal and non-stomatal limitations revisited. Ann. Bot. 89, 183e189.

Grimplet, J., Wheatley, M.D., Jouira, H.B., Deluc, L.G., Cramer, G.R., Cushman, J.C.,2009. Proteomic and selected metabolite analysis of grape berry tissues underwell-watered and water-deficit stress conditions. Proteomics 9, 2503e2528.

Hantao, L.W., Aleme, H.G., Passador, M.M., Furtado, E.L., Ribeiro, F.A., Poppi, R.J.,Augusto, F., 2013. Determination of disease biomarkers in Eucalyptus bycomprehensive two-dimensional gas chromatography and multivariate dataanalysis. J. Chromatogr. A 1279, 86e91.

Hobro, A.J., Kuligowski, J., Doll, M., Lendl, B., 2010. Differentiation of walnut woodspecies and steam treatment using ATR-FTIR and partial least squaresdiscriminant analysis (PLS-DA). Anal. Bioanal. Chem. 398, 2713e2722.

Ibanez, S., Rosa, M., Hilal, M., Gonzalez, J.A., Prado, F.E., 2008. Leaves of Citrusaurantifolia exhibit a different sensibility to solar UV-B radiation according todevelopment stage in relation to photosynthetic pigments and UV-B absorbingcompounds production. J. Photochem. Photobiol. B 90, 163e169.

Intrigliolo, D.S., Castel, J.R., 2006. Vine and soil-based measures of water status in aTempranillo vineyard. Vitis 45, 157e163.

Lavoir, A.V., Staudt, M., Schnitzler, J.P., Landais, D., Massol, F., Rocheteau, A.,Rodriguez, R., Zimmer, I., Rambal, S., 2009. Drought reduced monoterpeneemissions from the evergreen Mediterranean oak Quercus ilex: results from athroughfall displacement experiment. Biogeosciences 6, 1167e1180.

Lawo, N.C., Weingart, G.J., Schuhmacher, R., Forneck, A., 2011. The volatile metab-olome of grapevine roots: first insights into the metabolic response uponphylloxera attack. Plant Physiol. Biochem. 49, 1059e1063.

Loreto, F., Ciccioli, P., Brancaleoni, E., Cecinato, A., Frattoni, M., Sharkey, T.D., 1996.Different sources of reduced carbon contribute to form three classes of terpe-noid emitted by Quercus ilex L leaves. Proc. Natl. Acad. Sci. U. S. A. 93,

9966e9969.Loreto, F., Schnitzler, J.P., 2010. Abiotic stresses and induced BVOCs. Trends Plant Sci.

15, 154e166.Lovisolo, C., Perrone, I., Carra, A., Ferrandino, A., Flexas, J., Medrano, H., Schubert, A.,

2010. Drought-induced changes in development and function of grapevine(Vitis spp.) organs and in their hydraulic and non-hydraulic interactions at thewhole-plant level: a physiological and molecular update. Funct. Plant Biol. 37,98e116.

Lovisolo, C., Schubert, A., Sorce, C., 2002. Are xylem radial development and hy-draulic conductivity in downwardly-growing grapevine shoots influenced byperturbed auxin metabolism? New Phytol. 156, 65e74.

McLafferty, F., 2008. Wiley Registry of Mass Spectral Data with NIST 2008. Wiley-Blackwell.

Medrano, H., Bota, J., Abadia, A., Sampol, B., Escalona, J.M., Flexas, J., 2002. Effects ofdrought on light-energy dissipation mechanisms in high-light-acclimated,field-grown grapevines. Funct. Plant Biol. 29, 1197e1207.

Niinemets, U., Kannaste, A., Copolovici, L., 2013. Quantitative patterns betweenplant volatile emissions induced by biotic stresses and the degree of damage.Front. Plant Sci. 4.

Pardo, J.M., 2010. Biotechnology of water and salinity stress tolerance. Curr. Opin.Biotechnol. 21, 185e196.

Poni, S., Lakso, A.N., Turner, J.R., Melious, R.E., 1993. The effects of pre-veraison andpost-veraison water-stress on growth and physiology of potted Pinot-noirgrapevines at varying crop levels. Vitis 32, 207e214.

RCoreTeam, 2013. R: a Language and Environment for Statistical Computing. RFoundation for Statistical Computing, Vienna, Austria.

Sade, N., Gebremedhin, A., Moshelion, M., 2012. Risk-taking plants: anisohydricbehavior as a stress-resistance trait. Plant Signal. Behav. 7, 767e770.

Schachtman, D.P., Goodger, J.Q., 2008. Chemical root to shoot signaling underdrought. Trends Plant Sci. 13, 281e287.

Schoedl, K., Forneck, A., Sulyok, M., Schuhmacher, R., 2011. Optimization, in-housevalidation, and application of a liquid chromatography-tandem mass spec-trometry (LC-MS/MS)-based method for the quantification of selected poly-phenolic compounds in leaves of grapevine (Vitis vinifera L.). J. Agric. FoodChem. 59, 10787e10794.

Schoedl, K., Schuhmacher, R., Forneck, A., 2012. Studying the polyphenols ofgrapevine leaves according to age and insertion level under controlled condi-tions. Sci. Hortic. 141, 37e41.

Schoedl, K., Schuhmacher, R., Forneck, A., 2013. Correlating physiological parame-ters with biomarkers for UV-B stress indicators in leaves of grapevine cultivarsPinot noir and Riesling. J. Agric. Sci. 151, 189e200.

Schultz, H.R., 2003. Differences in hydraulic architecture account for near-isohydricand anisohydric behaviour of two field-grown Vitis vinifera L. cultivars duringdrought. Plant Cell Environ. 26, 1393e1405.

Schultz, H.R., Matthews, M.A., 1988. Resistance to water transport in shoots of Vitisvinifera L. : relation to growth at low water potential. Plant physiol. 88,718e724.

Skirycz, A., Inze, D., 2010. More from less: plant growth under limited water. Curr.Opin. Biotechnol. 21, 197e203.

Soar, C.J., Speirs, J., Maffei, S.M., Penrose, A.B., McCarthy, M.G., Loveys, B.R., 2006.Grape vine varieties Shiraz and Grenache differ in their stomatal response toVPD: apparent links with ABA physiology and gene expression in leaf tissue.Aust. J. Grape Wine Res. 12, 2e12.

Stein, S.E., 1999. An integrated method for spectrum extraction and compoundidentification from gas chromatography/mass spectrometry data. J. Am. Soc.Mass Spectrom. 10, 770e781.

Styczynski, M.P., Moxley, J.F., Tong, L.V., Walther, J.L., Jensen, K.L.,Stephanopoulos, G.N., 2007. Systematic identification of conserved metabolitesin GC/MS data for metabolomics and biomarker discovery. Anal. Chem. 79,966e973.

Taji, T., Ohsumi, C., Iuchi, S., Seki, M., Kasuga, M., Kobayashi, M., Yamaguchi-Shinozaki, K., Shinozaki, K., 2002. Important roles of drought- and cold-inducible genes for galactinol synthase in stress tolerance in Arabidopsisthaliana. Plant J. Cell Mol. Biol. 29, 417e426.

Vandeleur, R.K., Mayo, G., Shelden, M.C., Gilliham, M., Kaiser, B.N., Tyerman, S.D.,2009. The role of plasma membrane intrinsic protein aquaporins in watertransport through roots: diurnal and drought stress responses reveal differentstrategies between isohydric and anisohydric cultivars of grapevine. PlantPhysiol. 149, 445e460.

Varmuza, K., Filzmoser, P., 2009. Introduction to Multivariate Statistical Analysis inChemometrics. CRC Press (Taylor & Reancis), Boca Raton, FL, USA.

Vickers, C.E., Gershenzon, J., Lerdau, M.T., Loreto, F., 2009. A unified mechanism ofaction for volatile isoprenoids in plant abiotic stress. Nat. Chem. Biol. 5,283e291.

Weingart, G., Kluger, B., Forneck, A., Krska, R., Schuhmacher, R., 2012. Establishmentand application of a metabolomics workflow for identification and profiling ofvolatiles from leaves of Vitis vinifera by HS-SPME-GC-MS. Phytochem. Anal. 23,345e358.

Williams, L.E., Baeza, P., 2007. Relationships among ambient temperature and vaporpressure deficit and leaf and stem water potentials of fully irrigated, field-grown grapevines. Am. J. Enol. Viticult. 58, 173e181.