UBC Faculty Research and Publications

Transcriptome and metabolite profiling reveals that prolonged drought modulates the phenylpropanoid and… Savoi, Stefania; Wong, Darren C J; Arapitsas, Panagiotis; Miculan, Mara; Bucchetti, Barbara; Peterlunger, Enrico; Fait, Aaron; Mattivi, Fulvio; Castellarin, Simone D Mar 21, 2016

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


52383-12870_2016_Article_760.pdf [ 3.47MB ]
JSON: 52383-1.0361788.json
JSON-LD: 52383-1.0361788-ld.json
RDF/XML (Pretty): 52383-1.0361788-rdf.xml
RDF/JSON: 52383-1.0361788-rdf.json
Turtle: 52383-1.0361788-turtle.txt
N-Triples: 52383-1.0361788-rdf-ntriples.txt
Original Record: 52383-1.0361788-source.json
Full Text

Full Text

RESEARCH ARTICLE Open AccessTranscriptome and metabolite profilingreveals that prolonged drought modulatesthe phenylpropanoid and terpenoidpathway in white grapes (Vitis vinifera L.)Stefania Savoi1,2, Darren C. J. Wong3, Panagiotis Arapitsas1, Mara Miculan2,4, Barbara Bucchetti2, Enrico Peterlunger2,Aaron Fait5, Fulvio Mattivi1 and Simone D. Castellarin2,3*AbstractBackground: Secondary metabolism contributes to the adaptation of a plant to its environment. In wine grapes,fruit secondary metabolism largely determines wine quality. Climate change is predicted to exacerbate droughtevents in several viticultural areas, potentially affecting the wine quality. In red grapes, water deficit modulatesflavonoid accumulation, leading to major quantitative and compositional changes in the profile of the anthocyaninpigments; in white grapes, the effect of water deficit on secondary metabolism is still largely unknown.Results: In this study we investigated the impact of water deficit on the secondary metabolism of white grapes usinga large scale metabolite and transcript profiling approach in a season characterized by prolonged drought. Irrigatedgrapevines were compared to non-irrigated grapevines that suffered from water deficit from early stages of berrydevelopment to harvest. A large effect of water deficit on fruit secondary metabolism was observed. Increasedconcentrations of phenylpropanoids, monoterpenes, and tocopherols were detected, while carotenoid and flavonoidaccumulations were differentially modulated by water deficit according to the berry developmental stage. The RNA-sequencing analysis carried out on berries collected at three developmental stages—before, at the onset, and at lateripening—indicated that water deficit affected the expression of 4,889 genes. The Gene Ontology category secondarymetabolic process was overrepresented within up-regulated genes at all the stages of fruit development considered,and within down-regulated genes before ripening. Eighteen phenylpropanoid, 16 flavonoid, 9 carotenoid, and 16terpenoid structural genes were modulated by water deficit, indicating the transcriptional regulation of thesemetabolic pathways in fruit exposed to water deficit. An integrated network and promoter analyses identified atranscriptional regulatory module that encompasses terpenoid genes, transcription factors, and enriched drought-responsive elements in the promoter regions of those genes as part of the grapes response to drought.Conclusion: Our study reveals that grapevine berries respond to drought by modulating several secondary metabolicpathways, and particularly, by stimulating the production of phenylpropanoids, the carotenoid zeaxanthin, and ofvolatile organic compounds such as monoterpenes, with potential effects on grape and wine antioxidant potential,composition, and sensory features.Keywords: Abiotic stress, Grapevine, Network analysis, RNA sequencing, Transcriptomics, Water deficit* Correspondence: simone.castellarin@ubc.ca2Dipartimento di Scienze Agro-alimentari, Ambientali e Animali, University ofUdine, Via delle Scienze 208, 33100 Udine, Italy3Wine Research Centre, The University of British Columbia, 2205 East Mall,Vancouver, BC V6T 1Z4, CanadaFull list of author information is available at the end of the article© 2016 Savoi et al. Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, andreproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link tothe Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver(http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.Savoi et al. BMC Plant Biology  (2016) 16:67 DOI 10.1186/s12870-016-0760-1BackgroundPlant secondary metabolites include more than 200,000compounds that display a large chemical diversity whileaccumulating in specific organs, tissues, and cells [1].They ensure a plant’s survival in the environment byperforming a multitude of functions, such as defendingplant tissues from pathogens or herbivorous attacks,and aiding reproduction by attracting pollinators orseed dispersers [2]. Berry fruits accumulate a variety ofsecondary metabolites such as polyphenols, stilbenoids,carotenoids, and free and bound volatile organic com-pounds (VOCs) [3, 4]. These metabolites affect fruitpigmentation and flavour, and confer to the fruit well-known health benefits. In several fruit crops, the con-centration of these metabolites significantly impacts thequality of the fruit and, indeed, the economic value ofproduction. As part of the adaptation mechanism of aplant to its environment, secondary metabolism is sen-sitive to biotic and abiotic cues [1]. Hence, in agricul-tural settings the effect of climatic constraints on theaccumulation of these metabolites should be taken intoconsideration for developing cultivation strategies thatoptimize fruit composition and crop economic value.Grapes are one of the major fruit crops in the world[5]. Dry and warm Mediterranean climates are consid-ered optimal for wine grape production; in these cli-mates, grapes are often produced without artificialirrigation. However, limited water availability results inreduced vine vigor and fruit growth, significant losses incrop yield, and changes in fruit composition [6]. More-over, climate change is predicted to exacerbate droughtevents in several viticultural areas; and Hannah et al. [7]postulate that these phenomena may reduce the viabilityof viticulture in regions where grapes have been trad-itionally cultivated.Grapevine berry secondary metabolism is under stronggenetic control and varies among cultivars [8, 9]. Hence,the task of understanding the response of this metabol-ism to environmental cues is complicated. Several stud-ies have investigated the impact of drought and deficitirrigation strategies on berry secondary metabolism inred grape cultivars, focusing specifically on the accumu-lation of phenolics. Recently, Hochberg et al. [10]employed large-scale metabolite analyses to investigatethe impact of deficit irrigation on this metabolism inCabernet Sauvignon and Shiraz grapes, and showed cul-tivar specificity in the magnitude of response. In general,it is recognized that moderate and severe water deficitspromote the synthesis and increase the concentration offlavonoids in red grapes, often resulting into better sen-sory attributes of wines [6]. Besides phenolics, manyother secondary metabolites accumulate in the grapeberry. These include carotenoids [11] and free and gly-cosylated VOCs such as C13-norisoprenoids, terpenes,aldehydes, ketones, esters, and alcohols [12]. Deluc et al.[13] adopted a microarray platform to investigate differ-ences in the transcriptome response to water deficit be-tween Cabernet Sauvignon, a red grape variety, andChardonnay, a white grape variety. The study revealedthat genes of several secondary metabolic pathways weremodulated by water deficit and this metabolic responsevaried with the cultivar considered. In Chardonnaygrapes, water deficit increased the level of expression ofone terpene synthase, indicating that terpenes might bepart of the metabolic response to water deficit.The effect of water deficit on secondary metabolismremains largely unexplored in fruits; particularly, verylittle information is available on the effect of this deficiton the concentration of VOCs, key determinants offruit economic value, and in the case of wine grapes, ofthe wine sensory features. Recently, large scale tran-script and metabolite analyses have been adopted to re-veal the metabolic responses of white grapes to clusterexposure to sunlight and to a biotic stress [14, 15]. Inthis case study, we employed a large-scale metaboliteprofiling and RNA-sequencing analyses to evaluate theimpact of water deficit on berry secondary metabolismin white grapes in a year characterized by high temper-atures and low rainfalls (Additional file 1: Table S1) in aNorth Italian viticultural region where irrigation israrely applied to the grapevines. We hypothesize thatwater deficit may activate the terpenoid pathway andthe production of monoterpenes. Two different waterregimes were applied to Tocai Friulano vines and theeffect of water deficit on the transcriptome programand the phenolic, carotenoid, tocopherol, and free VOCaccumulation were investigated at different stages ofberry development. Finally, an integrated network ana-lysis was undertaken to investigate the impact of thewater deficit on metabolite-metabolite and metabolite-transcript interactions in developing grapes.ResultsImpact of irrigation treatments on plant water status,yield, berry growth, berry soluble solids, and titratableacidityTwo irrigation treatments were applied to vines duringthe season. Irrigated vines (defined as C, controls,henceforward) were weekly irrigated in order to keeptheir stem water potential (ΨStem) above −0.8 MPa,whereas vines subjected to deficit irrigation (defined asD, deficit irrigation, henceforward) were not irrigatedfrom fruit set until harvest, unless they displayed signsof extreme water deficit: ΨStem lower than −1.5 MPa andfading of the canopy.Rainfalls during the 2012 season were very limited(Fig. 1a) and mean temperatures peaked just before ver-aison (the onset of fruit ripening), which was recordedSavoi et al. BMC Plant Biology  (2016) 16:67 Page 2 of 1765 days after anthesis (DAA). ΨStem of D vines decreasedfrom early stages of fruit development (Fig. 1b) whileΨStem of C vines generally remained above −0.8 MPa.ΨStem of D vines reached the seasonal minimum(−1.5 MPa) at 67 DAA. Afterward, three consecutive ir-rigations together with some rainfalls initiated a partialrecovery of ΨStem values in D vines.Irrigation treatments significantly affected vine prod-uctivity and D reduced both cluster weight and yield pervine (Additional file 2: Table S2). Moreover, water deficitseverely reduced berry weight in D during most part ofthe season (Fig. 1c), produced increased soluble solids (agood indicator of sugar concentration) before veraison(41 and 54 DAA) and at harvest (93 DAA) (Fig. 1d), andincreased and decreased the concentration of acidsbefore (41 DAA) and after (68 and 82 DAA) veraison,respectively, but not at harvest (Fig. 1e).Impact of water deficit on secondary metabolites andintegrated networks of metabolitesBerries were sampled for secondary metabolite analyses(Additional file 3: Table S3) six times during the season:three times before ripening (27, 41, and 54 DAA), one atthe beginning of ripening (68 DAA), one at mid-ripening (82 DAA), and one at late ripening (93 DAA)that coincided with the harvest date of the vineyard.Large scale metabolite analysis identified 27 phenolics, 8carotenoids, 2 tocopherols, and 37 VOCs. A principalcomponent analysis over the metabolite profiles of the48 samples analyzed (two treatments x six developmen-tal stages x four biological replicates) was performed(Additional file 4: Figure S1). The analysis indicates thatthe metabolite profile largely varied based on the berrydevelopment, with a sharp distinction between before rip-ening (27, 41, 54 DAA) and ripening stages (68, 82, 93DAA), largely driven by the PC1. The irrigation treatmentalso affected the metabolite profile, with a clear separationof C and D samples at late ripening (93 DAA).Water deficit affected the concentration of 20 out of27 phenolics at one or more developmental stages(Fig. 2a, Additional file 5: Figure S2). Water deficitgenerally increased the concentration of derivatives ofcinnamic and benzoic acids, and modulated the accu-mulation of flavan-3-ols and proanthocyanidins. Theirconcentration was increased and decreased by waterFig. 1 Weather conditions at the experimental site and impact ofirrigation treatments on plant and fruit physiology. a Daily rainfalland average temperature. Progress of b stem water potential(ΨStem), c berry weight, d soluble solid accumulation, and e titratableacidity in fully irrigated (C) and deficit irrigated (D) vines. Dotted linesindicate veraison. Bars represent ± SE. Asterisks indicate significantdifferences between treatments at P < 0.05 (*), P < 0.01 (**), P < 0.001(***) evaluated by one-way ANOVASavoi et al. BMC Plant Biology  (2016) 16:67 Page 3 of 17deficit before (27, 41, and 54 DAA) and after (68, 82,and 93 DAA) veraison, respectively. Limited effects ofwater deficit on stilbenoid accumulation were ob-served. In contrast, D largely affected the accumulationof carotenoid and tocopherols in the berry (Fig. 2b,Additional file 6: Figure S3). The concentration ofmost carotenoids was increased and decreased in Dbefore and after veraison, respectively. Zeaxanthin, α-tocopherol, and γ-tocopherol concentrations were higherin D than in C after veraison. Water deficit also increasedthe concentration of 12 VOCs (Fig. 2c, Additional file 7:Figure S4) at late ripening (93 DAA). At this stage, D pro-moted the accumulation of monoterpenes such as hotrie-nol, linalool, nerol, and α-terpineol.Differences in metabolic network properties could beobserved between C and D (Additional file 8: Table S7A)for the phenolic (Fig. 3a,b) and VOC (Fig. 3c,d) net-works, but not for the carotenoid and tocopherol ones(Additional file 9: Figure S5). Water deficit affected thephenolic and VOC network topology by increasing thenetwork connectedness in comparison with the controls.In general, the majority of both C and D metabolite-metabolite correlations are based on positive interactionsamong nodes, but negative correlations were observedespecially under D, in particular for gallic acid. We ob-served two highly interconnected clusters within theVOC network of D berries; one of these clusters con-tained many of the VOCs that were significantly modu-lated under D.Impact of water deficit on berry transcriptomeTo investigate the molecular changes that take place inthe berry under water deficit, and to relate thesechanges to the observed changes in the berry metabol-ite profile, we compared the transcriptome of C and Dberries at three selected developmental stages, 41 DAA(before ripening), 68 DAA (beginning of ripening), 93DAA (late ripening).After filtering for organelles contamination and qualitytrimming, the average number of unique reads thatmapped the V1 version of the grape genome [16] was25.4 M (Additional file 10: Table S4). Among the 29,971genes of the grapevine genome, 23,603 (78.8 %) wereexpressed at 41 DAA, 22,259 (74.4 %) at 68 DAA, andFig. 2 Effect of water deficit on secondary metabolites during fruit development. Heatmaps represent log2FC(D/C) of the a phenolic, bcarotenoid and tocopherol, and c VOC concentration under water deficit conditions at 27, 41, 54, 68, 82, 93 DAA. Blue and red boxes indicatelower and higher concentration in D, respectively. Asterisks indicate significant differences (P < 0.05) between treatments. Metabolites werehierarchically clustered based on their response to water deficitSavoi et al. BMC Plant Biology  (2016) 16:67 Page 4 of 1722,349 (74.7 %) at 93 DAA. At harvest, the number ofexpressed genes was significantly higher in D (22,655)than in C (22,042).A strong relationship was found between the RNA-seqand qPCRs gene expression values of 15 genes selectedfor validating the transcriptomic dataset (Additional file11:Table S5, Additional file 12: Figure S6). Coefficient ofcorrelation between RNA-seq and qPCR gene expressionranged between 0.792 and 0.999, indicating the reliabilityof the whole transcriptome assays.A principal component analysis over the transcriptomeprofiles of the 18 samples analyzed (two treatments xthree developmental stages x three biological replicates)was performed (Fig. 4a). The first three principal compo-nents explain 52.9, 26.5, and 7.1 % of the variance amongsamples, respectively. Similarities and differences amongFig. 3 Network representation of phenolics and VOCs in C (a, c) and D (b, d) berries during development. Nodes represent ‘metabolites’ and edgesrepresent ‘relationships’ between any two metabolites. Edges colored in ‘red’ and ‘blue’ represent positive and negative correlations (P < 0.001),respectively. Metabolites in bold indicate a significant effect of water deficit on the concentration of that metabolite at one or more developmentalstages. Number of correlating edges were 13, 35, 11, 42 in (a, b, c, and d), respectively. The average node neighborhood was 1.53, 3.89, 1.57, and 3.11in (a, b, c, and d), respectively. The clustering coefficient was 0.08, 0.53, 0.00, and 0.49 in (a, b, c, and d), respectivelyFig. 4 Analysis of the berry transcriptome in fully irrigated (C) and deficit irrigated (D) vines. a Principal component analysis (PCA) of the berrytranscriptome of 18 independent samples collected from C and D vines at 41, 68, and 93 DAA. Circles, triangles and squares represent berries at41, 68, and 93 DAA, respectively. Full and open symbols identify C and D berries, respectively. b Common and unique DE genes at 41, 68, and 93DAA are represented in the Venn diagramSavoi et al. BMC Plant Biology  (2016) 16:67 Page 5 of 17berry transcriptomes were mostly driven by the develop-mental stage when berries were sampled. C and D sampleswere mixed within the group of the samples harvested at41 DAA, but were clearly separated at 68 and 93 DAA,with the majority of the variance explained by the secondprincipal component.The total number of differentially expressed (DE)genes between C and D was 4,889 (Additional file 13:Table S6A, B, C). The number of DE genes changed dur-ing fruit development. D modulated the expression of1,016 genes (316 up-regulated; 700 down-regulated) at41 DAA, 2,448 genes (1,119 up-regulated; 1,329 down-regulated) at 68 DAA, and 2,446 genes (1,142 up-regulated; 1,304 down-regulated) at 93 DAA. Somegenes were differentially regulated in unison among twoor three developmental stages (Fig. 4b, Additional file13: Table S6A, B, C, D).Seventeen plant GO categories (slim biological pro-cesses) were significantly overrepresented among DEgenes (Additional file 13: Table S6E). Before ripening (41DAA), carbohydrate metabolic process, development, andresponse to biotic stress were the three major GeneOntology (GO) categories within up-regulated genes,while response to stress, transport, and response to abioticstress were the major GO categories within down-regulated genes. At the beginning of ripening (68 DAA),response to stress, carbohydrate, and response to abioticstress were the three major GO categories within up-regulated genes, and response to stress, transport, and de-velopment were overrepresented GO categories withindown-regulated genes. At late ripening (93 DAA), re-sponse to stress, development, and response to abioticstress were the three major GO categories within up-regulated genes, and response to stress, transport, andcarbohydrate metabolic process were enriched GO cat-egories within down-regulated genes. The GO categorysecondary metabolic process was overrepresented withinup-regulated genes at all the stages of fruit developmentconsidered, and within down-regulated genes at 41 DAA.Impact of water deficit on phenylpropanoid, flavonoid,carotenoid, and terpenoid pathwayBecause this study focuses on the impact of waterdeficit on secondary metabolism, we did identify theDE genes that belonged to the major secondary meta-bolic pathways in the grapevine berry during develop-ment (Additional file 13: Table S6 F, G, H). Theimpact of water deficit on the expression of thesegenes was expressed as the log2 fold change of thetranscript level in D compared to C. Finally, the geneswere mapped into the related metabolic pathways (Figs. 5,6, 7).Water deficit modulated the expression of many genesthat codify for structural enzymes of the phenylpropanoidand flavonoid pathway (Fig. 5). Most of these genes wereup-regulated under D, particularly at 41 and 93 DAA.Among the DE genes, three genes annotated asphenylalanine ammonia lyases (VviPALs) were up-regulated by D at 41 and 93 DAA. One trans-cinna-mate 4-monooxygenase (VviC4H; VIT_06s0004g08150)was up-regulated by D at 41 and 93 DAA, while anotherVviC4H (VIT_11s0065g00350) was down-regulated at 41and up-regulated at 68 DAA. Four 4-coumarate-CoAligase (Vvi4CL; VIT_02s0025g03660, VIT_02s0109g00250, VIT_11s0052g01090, VIT_16s0039g02040) wereup-regulated by D at different developmental stages.Other two Vvi4CL (VIT_16s0050g00390, VIT_18s0001g00290) were down-regulated at 41 DAA. One p-coumaroyl shikimate 3'-hydroxylase (VviC3H) and onehydroxycinnamoyl-CoA:shikimate/quinate hydroxycin-namoyltransferase (VviHCT) were up-regulated by Dat 93 DAA. Two caffeic acid 3-O-metyltransferase(VviCOMT) were up-regulated by D: one (VIT_02s0025g02920) at 68 and 93 DAA, the other one(VIT_08s0007g04520) only at 68 DAA. Finally, acaffeoyl-CoA 3-O-methyltransferase (VviCCoAMT;VIT_03s0063g00140) was down-regulated at 68 andup-regulated at 93 DAA, while another VviCCoAMT(VIT_07s0031g00350) was up-regulated at all the threestages of development.In parallel, water deficit modulated the expression ofmost structural flavonoid genes; particularly three chal-cone synthases (VviCHSs), two chalcone isomerases(VviCHIs), one flavonoid-3′5′-hydroxylase (VviF3′5′H),two flavanone-3-hydroxylases (VviF3Hs), one dihydrofla-vonol reductase (VviDFR), and two leucoanthocyanidindioxygenases (VviLDOX). All the above genes exceptone VviLDOX (VIT_08s0105g00380) were up-regulatedby D. The flavonol synthase (VviFLS) is a key enzymefor flavonol production. Water deficit significantly pro-moted the expression of one VviFLS (VIT_18s0001g03470) at 68 and 93 DAA while down-regulating the ex-pression of another VviFLS (VIT_18s0001g03430) at 68DAA. The leucoanthocyanidin reductase (VviLAR) andanthocyanidin reductase (VviANR) are key regulators ofthe flavan-3-ol and proanthocyanidin biosynthesis. Vvi-LAR1 was up-regulated by water deficit at 41 DAA,while VviLAR2 was down-regulated in the same condi-tion at 68 DAA and up-regulated at 93 DAA. VviANRwas up-regulated by water deficit at 41 DAA and down-regulated at 68 DAA.Despite the fact that VviMyb14 (VIT_07s0005g03340)and VviMyb15 (VIT_05s0049g01020)—transcription fac-tors that regulate stilbene synthesis in grapevine[17]—were differentially expressed in D at 68 DAA(Additional file 13: Table S6B), transcript levels of the48 annotated VviSTSs [18] were never affected bywater deficit.Savoi et al. BMC Plant Biology  (2016) 16:67 Page 6 of 17The effect of water deficit on the carotenoid pathwaywas analyzed according to the Vitis vinifera carotenoidgenes identified by Young et al. [11]. A phytoene syn-thase gene (VviPSY2) was upregulated under water def-icit but only at 68 DAA (Fig. 6). The same wasobserved for a ζ-carotene desaturase (VviZDS1). On thecontrary, water deficit down-regulated the expressionof a lycopene β-cyclase (VviLBCY), a β-carotene hy-droxylase (VviBCH2), and a carotene hydroxylase (Vvi-LUT5) at 68 DAA, and of a carotenoid isomerase(VviCISO1) at 93 DAA. The expression of a lycopene ε-cyclase (VviLECY1) was down-regulated by D at 41 andup-regulated at 93 DAA.In plants, carotenoids are also the substrate for theproduction of norisoprenoids. Some C13-norisoprenoids,such as β-ionone and β-damascenone, are important de-terminants of the grape and wine aroma [12]. The en-zymes (9,10) (9′,10′) cleavage dioxygenase (CCD4) and(5,6) (5′,6′) (9,10) (9′,10′) cleavage dioxygenase (CCD1)are key enzymes in the norisoprenoid synthesis. In thisstudy, D up-regulated the expression of VviCCD4b at 68DAA and down-regulated the expression of VviCCD4aat 93 DAA.Plant terpenes are synthesized in the plastids throughthe 2C-methyl-D-erythritol-4-phosphate pathway (MEP),and in the cytosol through the mevalonate (MVA)Fig. 5 Modulation of phenylpropanoid and flavonoid pathway under water deficit. Log2FC (D/C) levels of differential gene expression arepresented at 41 (left box), 68 (central box), and 93 (right box) DAA. Blue and red boxes indicate down- or up-regulation of the gene under water deficit,respectively. Bold margins identify significant differences (P < 0.05) between treatments. Symbols identify commonly regulated steps of the pathway.Transcript levels, expressed as normalized counts, in C and D berries at 41, 68, and 93 DAA, are reported in Additional file 13: Table S6 FSavoi et al. BMC Plant Biology  (2016) 16:67 Page 7 of 17pathway. Water deficit modulated the expression ofseveral genes of the two pathways (Fig. 7). Genes regu-lating early steps of the MEP pathway, such as one 1-deoxy-D-xylulose-5-phosphate synthase (VviDXS1) andthe 1-deoxy-D-xylulose-5-phosphate reductoisomerase(VviDXR) were down-regulated by D at 41 DAA, whileanother VviDXS was down-regulated at 68 DAA andup-regulated at 93 DAA. Terpene synthases (VviTPSs)were generally up-regulated under water deficit, par-ticularly at 93 DAA. The terpene synthases gene familywas recently characterized in Vitis vinifera [19]. Waterdeficit modulated the expression of seven terpenesynthases of the TPS-a family (VIT_18s0001g04280,VIT_18s0001g04530, VIT_18s0001g05240, VIT_18s0001g05290, VIT_18s0001g05430, VIT_19s0014g04810,VIT_19s0014g04930), one of the TPS-b family (VIT_12s0134g00030), and one of the TPS-g family (VIT_00s0266g00070).The impact of water deficit on the expression of keygenes of the phenylpropanoid, flavonoid, and terpenoidpathway was then investigated at all the six samplingdates with targeted gene expression analyses (Additionalfile 14: Figure S7). VviPAL2, VviCHS1, VviFLS, andVviANR were up-regulated by water deficit at several de-velopmental stages in parallel with the observed increaseof phenolic concentration under the same conditions(Fig. 2a). Similarly, the expression profile of two VviTPSs(VIT_12s0134g00030 and VIT_19s0014g04930) indicatedthat water deficit stimulated a higher synthesis of ter-penes from 82 DAA.Impact of water deficit on integrated networks ofmetabolites and transcriptsThe increased average node degree, clustering coeffi-cient, and network density between the C and Dmetabolite-metabolite networks prompted us to per-form an association study between metabolites andtranscripts in order to reveal the major transcriptsthat were associated with changes in metabolite net-works (Additional file 8: Table S7B, C, D). Emphasiswas given on biosynthetic genes of the metabolitepathways considered. The number of positive correla-tions between phenolic compounds and phenolic bio-synthetic genes slightly increased under D particularlybecause of an increase in the number of correlationswithin benzoic and cinnamic acid pathway elements(Additional file 8: Table S7C, D). VOC-transcript linkswere also affected by water deficit. Correlations be-tween geraniol, citronellol, and hotrienol levels andterpenoid transcripts were observed in controls only(Additional file 8: Table S7B, D). In contrast, correla-tions between nerol and α-terpineol levels and terpen-oid transcripts were observed only under water deficit(Additional file 8: Table S7C). Water deficit also mod-ulated the correlations between the non-terpenoidVOCs and the fatty acid related transcripts: reducingthem for (EE)-2,4-hexadienal and (E)-2-pentenal, andincreasing them for nonanal, hexanol, and 3-hexenol.The number of carotenoid-transcript correlations wasnot affected by water deficit.The knowledge of the regulation of monoterpene bio-synthesis is lacking. Because of the remarkable effect ofwater deficit on the VOC networks, we furthered ouranalysis into gene-metabolite relationship focusing onripening-related monoterpenes induced by water deficit.These included linalool, nerol, and α-terpineol. Thegene-metabolite network included the top 100 gene cor-relators for each of these monoterpenes (Fig. 8a). Amongthe 222 genes present in the network, 116 genes (52 %)were differentially expressed under water deficit. Therewere 49, 48, and 64 gene-metabolite relationship thatwere specific for α-terpineol, nerol, and linalool,Fig. 6 Modulation of carotenoid pathway under water deficit.Log2FC (D/C) levels of differential gene expression are presentedat 41 (left box), 68 (central box), and 93 (right box) DAA. Blueand red boxes indicate down- or up-regulation of the geneunder water deficit, respectively. Bold margins identify significantdifferences (P < 0.05) between treatments. Symbols identifycommonly regulated steps of the pathway. Transcript levels,expressed as normalized counts, in C and D berries at 41, 68,and 93 DAA, are reported in Additional file 13: Table S6 GSavoi et al. BMC Plant Biology  (2016) 16:67 Page 8 of 17respectively. Inspection of the overall network showedthat a large proportion of these correlated genes were in-volved in terpenoid (18), lipid (10), and hormone (7) me-tabolism, as well as various transport (11) and signaling(13) mechanisms (Additional file 8: Table S7E). Elevengene-metabolite interactions were found for all the threemetabolites and 29 interactions were in common be-tween α-terpineol and nerol. We highlight several poten-tial transcriptional regulators annotated as MYB24(VIT_14s0066g01090), C2H2 Zinc finger (VIT_07s0005g02190), and Constans-like 11 (VIT_19s0014g05120), which significantly correlated with these mono-terpenes. Promoter enrichment analysis of the top 100correlated transcripts for each metabolite furtherrevealed that many of the genes within each networkcontain significantly enriched (P < 0.01) MYB recogni-tion (such as MYBZM, MYBCOREATCYCB1, MYB1AT,MYBPLANT, MYBCORE, MYB2CONSENSUSAT) andvarious drought-responsive (RYREPEATBNNAPA, LTRECOREATCOR15, DRECRTCOREAT, MYCCONSENSUSAT, MYCATRD22) motif elements (Fig. 8b, Additionalfile 8: Table S7F).DiscussionThe prolonged and severe water deficit imposed inthis experiment modulated the accumulation ofFig. 7 Modulation of terpenoid pathway under water deficit. Log2FC (D/C) levels of differential gene expression are presented at 41 (left box), 68(central box), and 93 (right box) DAA. Blue and red boxes indicate down- or up-regulation of the gene under water deficit, respectively. Boldmargins identify significant differences (P < 0.05) between treatments. Transcript levels, expressed as normalized counts, in C and D berriesat 41, 68, and 93 DAA, are reported in Additional file 13: Table S6 HSavoi et al. BMC Plant Biology  (2016) 16:67 Page 9 of 17phenylpropanoids, flavonoids, carotenoids, and severalVOCs in the berry.At present, little information is available on the ef-fect of water deficit on phenolic accumulation inwhite grapes. Our study indicates that the phenylpro-panoid and the flavonoid pathway respond to waterdeficit at the transcript and metabolite level, and de-termine a general increase in phenolic concentrations.In red grape cultivars, water deficit strongly promotesaccumulation of flavonoids, particularly anthocyanin[13, 20]. Anthocyanin biosynthesis is limited in whitegrapes; however, these grapes do accumulate othermajor flavonoids such as flavonols, flavan-3-ols, andproanthocyanidins. Recent studies reported that waterdeficit increases flavonol concentration [9, 13] and re-duces [10] or does not affect proanthocyanidin con-centration in grapes [20]. Water deficit can increasethe concentration in the berry of secondary metabo-lites produced in the skin and in the seed by reducingthe berry volume and increasing relative skin andseed masses [21–23]. This was not the case in thisstudy, since relative skin and seed masses were not af-fected by water deficit (Additional file 15: Figure S8). Ourgene expression analysis indicated that many phenylpro-panoid and flavonoid genes were up-regulated underwater deficit, and the modulation of these pathways in-creased the concentration of derivatives of benzoic andcinnamic acids and of several flavonoids. Interestingly, keystructural genes for the flavonol and flavan-3-ol biosyn-thesis, such as flavonol synthases (VviFLSs) and leu-coanthocyanidin reductases (VviLARs), were up-regulatedat late stages of development, while flavonols, flavan-3-ols,and proanthocyanidin increased in concentration underwater deficit only at early stages of development (exceptprocyanidin B1, which was also higher at harvest). Simi-larly, in Cabernet Sauvignon vines exposed to water def-icit,VviLAR,VviANR, and VviFLS were up-regulated afterthe onset of fruit ripening, but no differences in flavonoland proanthocyanidin concentration were observed [20].Our combined transcript and metabolite data suggest thata competition for precursors between enzymes of the fla-vonoid and phenylpropanoid pathways is occurring, withFig. 8 Predicted gene-metabolite networks related to linalool (1), α-terpineol (2), and nerol (3) in grapevine berries during development. a Genesand metabolites are represented by circle and square nodes respectively. Edges represent associations (P < 0.001) between transcripts andmetabolites. The top 100 correlators for each metabolite are shown. Node borders in red represent DE transcripts. Node colors indicatethe pathway of the transcripts. b Heatmap of cis-regulatory elements enriched (P < 0.01) within the networks in a. Cis-regulatory elementsin bold and underlined are associated with ABA/drought response and MYB binding, respectively. Light and dark red color denotes enrichment scoresbetween 2 (P < 0.01) and 4 (P < 0.0001) respectively. White color represents no significant enrichment. *, **, ***, and **** denotes otherPLACE cis-regulatory motifs sharing similar consensus sequence with the associated motif (Additional file 8: Table S7F)Savoi et al. BMC Plant Biology  (2016) 16:67 Page 10 of 17phenylpropanoid enzymes being more efficient in direct-ing the substrates into the production of benzoic and cin-namic acid derivatives than the flavonoid enzymes insequestrating these precursors for the productions of fla-vonoids, particularly after the onset of fruit ripening whenthe accumulation of flavan-3-ols and proanthocyanidinsdecreases dramatically. To this extent, the increase in con-nectivity (with negative connections) of gallic acid withsome flavan-3-ols and proanthocyanidins observed underwater deficit highlights the role of this benzoic acid deriva-tive in the drought response, particularly at late stages ofdevelopment.Water deficit affected the concentration of carotenoidsand tocopherols in the berry, but the modulation of ca-rotenoid genes was much lower than for the phenylpro-panoid and flavonoid genes. Carotenoids are normallydegraded after the onset of fruit ripening [11], and ourdata indicate that this degradation is increased underwater deficit. However, water deficit increased the con-centration of zeaxanthin—the only carotenoid synthe-sized after the onset of berry ripening (Additional file 6:Figure S3). Zeaxanthin’s role in drought tolerance hasbeen already hypothesized in plants. Nerium oleanderincreased zeaxanthin content in the leaf under waterdeficit [24], and the enhancement of zeaxanthin levels inthe transgenic tobacco lines made plants more tolerantto drought stress [25]. As in our work with berries, pre-vious studies have shown a positive correlation betweentocopherol accumulation and water deficit in photosyn-thesizing tissues [26, 27]. However, we did not observeda consistent upregulation of key genes of the tocopherolpathway under water deficit and, at late stages of ripen-ing (93 DAA), the gene VviHPT (VIT_11s0052g00610)that encodes for one key enzyme of the pathway was ac-tually down-regulated in D berries (Additional file 13:Table S6), indicating that the accumulation of tocoph-erols might be affected by post-translational regulationand/or that other unknown genes might be involved inthe regulation of this pathway. This result agrees withprevious findings in Arabidopsis, where wild-type plantssubjected to water deficit increased the accumulation ofthe same tocopherols in the leaves without a parallelmodulation of tocopherol biosynthetic genes [27].The VOC profiling indicated that in Tocai Friulano,VOCs are primarily accumulated at early stages of devel-opment (Additional file 7: Figure S4), but prolongedwater deficit can stimulate the accumulation of severalVOCs at late stages of development. The network ana-lysis revealed that most of the VOCs modulated by waterdeficit mapped in the same network, but only underwater deficit conditions. This suggests that water deficithas a major effect on the accumulation of these VOCs,regardless the pathway they belong to or the type of ac-cumulation pattern they have under normal conditions.Among these VOCs, four monoterpenes—key aromaticsof several white grapes [4]—were largely increased underwater deficit, in parallel with an up-regulation of keystructural genes of the MEP pathway. In particular, keygenes for monoterpene production in the grapes such asVviDXS and two VviTPSs [19, 28] were up-regulated.The induction of monoterpene production under waterdeficit has been reported in several plants (reviewed in[29]), including the recent studies in grapevine leaves[30, 31], but the information on the effect of drought onmonoterpene biosynthesis in fruits (where monoterpenesimpact the quality and value of production) is lacking[32]. Besides monoterpene synthases, water deficit alsoup-regulated seven sesquiterpene synthases [19]. Weidentified only two sesquiterpenes, α-humulene andtrans-caryophyllene, which were accumulated in theberry only at early stages and were not affected by waterdeficit. However, the molecular data indicate that a moredetailed profiling of the sesquiterpene accumulated inthe berry is necessary to investigate the role of thesecompounds in the response to water deficit. Other keyodorants of grapes and wines are the carotenoid degrad-ation products C13-norisoprenoids that were observed toincrease in red grapes subjected to a limiting irrigationregime (reviewed in [12]). Interestingly, despite thehigher degradation of carotenoids observed underwater deficit, no clear modulation in the concentra-tion of C13-norisoprenoids, such as β-damascenoneand β-ionone, and of β-cyclocitral—a 7,8 cleavageproduct of β-carotene [33]—was observed.Metabolomic studies coupled with network analysiscomparing contrasting genotypes, stress perturbations,and tissues have been useful for understanding themechanism of genotype—environment interactions ofplants [34]. Deficit irrigation increased the metabolitenetwork connectivity for primary metabolites in grape-vine leaf, but the effect was genotype-dependent forphenolic networks [35]. Network-based analysis con-ducted here on secondary metabolism revealed thatwater deficit contributed significantly to restructuringthe underlying network properties of fruit metabolites.The higher network connectivity of secondary metabo-lites observed under water deficit also coincided withthe modulation of several genes of the related biosyn-thetic pathway. It is therefore likely that the observeddifferential network connectivity between irrigationtreatments may be determined by regulation at the tran-script level [36, 37]. In support of this hypothesis, wehave observed strong gene-metabolite correlations withphenolic and VOC pathway genes. Our results alsoshowed that this observation could be extended to otherpathways, such as that for terpenoid biosynthesis. Thehigher number of positive metabolite-transcript correla-tions for phenolics and terpenoids further strengthensSavoi et al. BMC Plant Biology  (2016) 16:67 Page 11 of 17our finding that those metabolic pathways take part ofthe grape response to water deficit, producing secondarymetabolites that potentially enhance grapevine fitnessunder this abiotic stress.In grapes, correlation network analysis has been usedrecently to ascribe functions to candidate genes poten-tially involved in the regulation of color development[38, 39]. Ma et al. [40] recently adopted a metabolite-transcript network approach to identify key regulators ofthe terpenoid pathway in Artemisia annua. Similarly, weaimed to identify new genes potentially involved in theregulation of terpenoid metabolism during developmentand under water deficit. The comprehensive metabolite-transcript network constructed with three key monoter-penes whose synthesis was promoted under water deficitshowed strong positive association of these metaboliteswith terpenoid transcripts. Interestingly, transcriptsrelated to hormone synthesis (salicylic and jasmonicacid) and signaling (auxin and brassinosteroid) werealso highly correlated. Terpene levels were signifi-cantly increased with the application of BTH (a sali-cylic acid analog) and methyljasmonate to berries [41]in grapevine. Additionally, our analysis allowed us toidentify a transcription factor annotated as VviMYB24(VIT_14s0066g01090) as a promising regulatory candi-date for monoterpene and fatty acid biosyntheticpathways in grapevine. Closer inspection of the anno-tated MYB gene showed high homology towards Ara-bidopsis MYB24, MYB21, and MYB57, all of whichare involved in regulating terpenoid biosynthesis [42].Recently MYB24 was found to be strongly up-regulated under solar UV radiation in grape skins, inparallel with the up-regulation of three terpenoidstructural genes [43], suggesting a major role in thegrapevine berry response to abiotic factors.Previous studies have shown that terpenoid metabol-ism responds to light and UV stimuli in berry and leaftissues [43–45], and we discovered that light-responsivemotif elements were significantly enriched throughoutthe monoterpene gene network (Additional file 8: TableS7F). This data also indicates that the effects of waterlimitation on berry terpenes may be indirect, in partowing to changes in the fruit microclimate due to a re-duction in canopy density. Water deficit can reduce can-opy growth determining a higher level of clusterexposure and, by consequence, higher berry temperature[46, 47]. The accumulation of grape volatiles can be in-fluenced by light exposure and temperature. Indeed, ex-posure of berries to sunlight favors accumulation ofnorisoprenoids and monoterpenes, and other free VOCs[15, 45, 48–50]. However, recent studies showed thatwater deficit and ABA treatments significantly increasedthe monoterpene and sesquiterpene concentration ingrapevine leaves [30, 31] even in the absence of UVradiation [30]. Moreover, the enrichment of droughtassociated elements (e.g., MYB and DRE motifs) inthe promoter region of many up-regulated terpenoidgenes observed in our study suggests a major directmodulation of the terpenoid pathway at the transcrip-tional level, possibly via an ABA mediated stimulus.These elements were frequently associated with abi-otic stress responses and particularly to drought andABA regulation [51]. Nevertheless, further tests undermore climate-controlled conditions are necessary toreveal to what extent the impact of water deficit onfruit metabolism is due to the modification of berrymicroclimate.The stage when deficit is applied and the severity ofdeficit certainly impact the response of fruit metabolism[32]. In grapes it is known that these factors stronglyaffect the physiological and metabolic response of theberry to water deficit [6]. In this case study, large effectson fruit metabolism were observed with drought occur-ring from early stages of berry development to harvest; acondition that also determined a lower yield and higherberry sugar concentration. However, further investiga-tions that compare water deficit imposed at differenttimings and at different level of severity should be car-ried out to fully understand the fruit response to thisabiotic stress in white grapes, and how this response isconsistent among seasons.ConclusionOur study sheds new light into the metabolic mecha-nisms of fruit response to drought events. Recently itwas hypothesized that an overproduction of key odor-ants, such as terpenoids and the carotenoids-derivednorisoprenoids, might be part of the adaptation ofwhite grapes to environmental stresses [5]. Our tran-scriptome and metabolite analyses showed that, besidethe flavonoid pathway, phenylpropanoid and terpenoidpathways can take part in the berry’s response towater deficit in non-pigmented berries; suggestingthat an over-production of monoterpenes is part ofthe fruit response to drought. Our network and pro-moter analyses highlighted a transcriptional regulatorymechanism that encompasses terpenoid genes, tran-scription factors, and drought-responsive elementsenriched in the promoter regions of those genes; thismechanism might be the basis of monoterpenes over-production. Overall, these results indicate that waterdeficit conditions can potentially impact the quality ofwhite wines by increasing the accumulation of poten-tial antioxidant and flavor compounds (e.g., deriva-tives of benzoic and cinnamic acids, zeaxanthin, andmonoterpenes) in the grapes. These results are alsopivotal to future studies that evaluate the impact ofdeficit irrigation strategies on wine quality.Savoi et al. BMC Plant Biology  (2016) 16:67 Page 12 of 17MethodsField experiment, physiological measurements, and samplepreparationThe field experiment was conducted in 2012 in a vine-yard at the University of Udine’s (Italy) experimentalfarm (46°01'52.3"N 13°13'30.6"E). The field experimentwas conducted in accordance with local legislation andno specific permission was required for the study.Climatological data were recorded during the experi-ment by an automated weather station located 100 mfrom the experimental site. Monthly mean tempera-tures and amount of rainfalls measured in 2012, as wellas the averages of the 2000–2012 period, are shown inAdditional file 1: Table S1. Seven years old Tocai Friulano(also known as Sauvignon vert and Sauvignonasse in Chileand France, respectively) grapevines grafted onto SO4were planted at 2.5 m x 1.0 m spacing in north–south ori-ented rows, and trained to a cane-pruned ‘Guyot’ system.Two irrigation treatments were established. Control (C)vines were irrigated in order to maintain midday stemwater potential (ψStem) above −0.8 MPa. Deficit irrigated(D) vines were not irrigated unless the ψStem was mea-sured lower than −1.5 MPa. Plant water status was moni-tored weekly by measuring ψStem using a Scholanderpressure chamber [52]. Irrigation was supplied when rain-fall in the preceding week was below 100 % ETc. or ψStemwas measured lower than −1.5 MPa, as discussed above. Asurface drip irrigation system with emitters (0.5 m x2.5 m) set to an 8 L h−1 application rate was used. At themaximum rate, water was supplied at approximately 40 Lper vine per week. Due to a prolonged drought, irrigation(20 L per vine) was applied to D vines at 67, 70, 76 daysafter anthesis (DAA) in order to mitigate the extremewater deficit. Each irrigation treatment was replicated onfour plots of six vines each, arranged in a completely ran-domized design. No effect of the irrigation treatments wasobserved on the number of shoots and clusters per vine.Samplings were carried out at 27, 41, 54, 68, 82, and93 DAA. Three sets of berries were randomly collectedfrom each plot. The first set of 60 berries was used tomeasure berry weight, total soluble solid concentration,and titratable acidity. The second one of 10 berries wasused to measure skin and seed weights and calculate therelative skin and seed masses. The third set of 40 berrieswas used for transcript and metabolite analyses. Berrieswere carefully trimmed off the cluster at the pedicel witha pair of scissors, quickly brought to the laboratory,weighed, and processed for soluble solids and titratableacidity [52] or immediately frozen at −80 °C for tran-script and metabolite analyses. Before metabolite andRNA extractions, pedicel was removed with a scalpeland berries were ground to a fine powder under liquidnitrogen using an analytical mill (IKA®-Werke GMbH &CO). One quality control (QC) sample was prepared bypooling an aliquot of all the samples and was used forQC runs in the metabolite analyses.Grapes were harvested for commercial wine productionat 93 DAA, when titratable acidity reached approximately5 mg/L in both treatments; yield per vine, number of clus-ters per vine, and cluster weight were recorded.Metabolite analysesPhenolic compounds were determined accordingly toVrhovsek et al. [53] with some modifications. Briefly,0.8 mL of chloroform and 1.2 mL of a mix of methanoland water (2:1) were added to one gram of frozen pow-der of ground berries. A 50 μL aliquot of o-coumaricacid solution (2 mg/mL in MeOH) was added as an in-ternal standard. The extraction mixture was shaken for15 min on an orbital shaker (Grant-Bio Rotator PTR-60)and then centrifuged for 10 min at 1000 g. The upperaqueous-methanolic phase was collected. The extractionwas repeated by adding 1.2 mL of methanol and water.The aqueous-methanolic phase was collected and com-bined with the previous one, brought to a final volumeof 5 mL with Milli-Q water, and filtered with a 0.2 μmPTFE filter (Millipore). The chromatographic analysiswas carried out using a Waters Acquity UPLC system(Milford) coupled to a Waters Xevo triple-quadrupolemass spectrometer detector (Milford). Compounds wereidentified based on their reference standard, retentiontime, and qualifier and quantifier ion, and were quanti-fied by their calibration curve.Carotenoids and tocopherols were analyzed accord-ingly to Wehrens et al. [54]. Briefly, the chloroformphase of the extraction solution described above wascollected. Twenty μL of trans-β-apo-8′-carotenal(25 μg/mL) was used as internal standard. Ten μL ofa 0.1 % triethylamine solution was added to preventrearrangement of carotenoids. After extraction, sam-ples were dried with N2, and stored at −80 °C untilanalysis. Dried samples were suspended in 50 μL ofethyl acetate, and transferred to dark vials. The chro-matographic analysis was performed in a 1290 InfinityBinary UPLC (Agilent) equipped with a RP C30 3 μmcolumn coupled to a 20 x 4.6 mm C30 guard column.Spectra components and elution profiles were deter-mined as in Wehrens et al. [54]. Compounds werequantified from linear calibration curves built withstandard solutions.Free (non-glycosylated) VOCs were analyzed accord-ingly to Fedrizzi et al. [55] with some modifications. Onthe day of analysis, four grams of frozen grape powderwere weighed out in a 20 mL SPME dark-glass vial.Three grams of NaCl, 15 mg of citric acid, 15 mg of as-corbic acid, 50 μL of sodium azide, and 7 mL of milliQwater were added to the sample. Fifty μL of a solutioncontaining five internal standards, d10-4-methyl-3-Savoi et al. BMC Plant Biology  (2016) 16:67 Page 13 of 17penten-2-one (1 g/L), d11-ethyl hexanoate (1 g/L), d16-octanal (1 g/L), d8-acetophenone (1 g/L), d7-benzyl alco-hol (1 g/L), was added to each sample. Prior to injection,the sample was pre-incubated at 60 °C in a SMM SingleIncubator (Chromtech) for 10 min stirring at 450 rpm.Next, the sample was incubated in the same conditionsfor 40 min with a DVB-CAR-PDMS 50/30 μm x 2 cm(Supelco) fiber in the headspace for absorption. FreeVOCs were thermally desorbed in splitless mode for4 min at 250 °C. Extractions and injections were carriedout with a CTC Combi-PAL autosampler (Zwingen).The analysis was performed with a Trace GC Ultra gaschromatograph (Thermo Scientific) coupled to a TSQQuantum Tandem mass spectrometer. GC separationwas performed on a 30 m Stabilwax (Restek Corpor-ation) capillary column with an internal diameter of0.25 mm and a film thickness of 0.25 μm with the condi-tions described in Fedrizzi et al. [55]. VOCs were identi-fied by comparing the retention times of individualpeaks with the retention times of their reference stan-dards, and by identifying the mass spectra using theNIST library. The ratio of each VOC area to the d16-oc-tanal internal standard area was considered to reducetechnical variability among extractions and chromato-graphic runs and VOCs quantity were expressed asμg/kg of berry of d16-octanal equivalents.Extractions and injections of the samples were per-formed in a random sequence and QC samples wereinjected at the beginning of the sequence and every sixsample injections.A list of the secondary metabolites analyzed in thisstudy is reported in Additional file 3: Table S3.RNA extraction and RNA sequencing analysisSamples collected at 41, 68, and 93 DAA were se-lected for transcriptome analyses. Three biologicalreplicates per treatment were considered. Total RNAwas extracted with the ‘Spectrum Plant total RNA’ kit(Sigma-Aldrich) from 0.2 g of ground berries. Thequantity and quality of the RNA were determinedwith a Caliper LabChip® GX (Perkin-Elmer).Library preparation was performed using the TruSeqRNA Sample Prep Kit v2.0 according to the manufac-turer’s instructions (Illumina). Libraries were quantifiedusing a 2100 Bioanalyzer (Agilent Technologies). Multi-plexed cDNA libraries were pooled in equimolaramounts, and clonal clusters were generated using Cbot(Illumina). Sequencing was performed with an IlluminaHiSeq 2000 platform (Illumina pipeline 1.8.2) at IGATechnology Services (Udine, Italy).An average of 28.9 M 50-nt single-end reads was gener-ated per sample (Additional file 10: Table S4). Trimmingfor quality and length, and filtering for mitochondria andchloroplast contamination were performed by the ERNEpackage version 1.2 tool ERNE-FILTER [56]. The mini-mum PHRED score accepted for trimming was 20, andreads shorter than 40 bp were discarded. Reads werealigned against the reference grapevine genome PN4002412x [16] using the software TopHat version 2.0.6 [57] withdefault parameters. Aligned reads were counted with ahtseq-count (version 0.6.0) in intersection-non-emptymode for overlap resolution [58]. Vitis vinifera annotationGTF-file (V1) was downloaded from Ensembl Plants web-site. Differentially expressed (DE) genes (false discoveryrate less than 0.05) analysis was performed with the Rpackage DeSeq2 [59]. Functional annotations of geneswere retrieved from Grimplet et al. [60] and VitisCyc [61].Gene ontology analyses were carried out for each sam-pling. Overrepresented genes categories were identifiedwith the BINGO app 3.0.2 of Cytoscape 3.1.1 [62]. Plant-GoSlim categories, referred to biological processes, wereused to run the gene enrichment analysis using a hyper-geometric test with a significance threshold of 0.05 afterBenjamini and Hochberg false discovery rate correction.Quantitative real-time polymerase chain reactionThe validation of RNA-Seq data was carried out on aset of DE genes using the quantitative real-time poly-merase chain reaction (qPCR) technique. RNA was ex-tracted from independent biological samples collectedat the same stage than the ones used for RNA-Seq ana-lysis. The reverse transcription of RNA samples wasperformed with the QuantiTect Reverse TranscriptionKit (Qiagen) and the Quantiscript Reverse Transcript-ase (Qiagen). Specific primers for 15 selected genes weredesigned with Primer3web version 4.0.0 (Additional file 11:Table S5). qPCR reactions, conditions, and calculation ofrelative expression values were carried out as in Falginellaet al. [63]. The annealing temperature was 58 °C for all pri-mer pairs except the VviUbiquitin housekeeping gene pair,which annealed at 56 °C. Correlation analysis based on thePearson Correlation Coefficient (PCC) was carried out be-tween the RNA-seq normalized counts and qPCR relativegene expression (Additional file 11: Table S5, Additionalfile 12: Figure S6).qPCR was also carried out to determine the level of ex-pression of selected structural genes of the phenylpropanoid,flavonoid, and terpenoid pathway at each sampling date.Statistical, network, and promoter analysesA one-way ANOVA was performed using JMP 7 (SAS In-stitute Inc.) to detect significant differences (P < 0.05) be-tween irrigation treatments at each sampling point.Heatmaps representing log2 fold change (log2FC) of me-tabolite concentrations between treatments (D/C) andprincipal component analysis (PCA) on the metaboliteprofiles and on the entire transcriptome dataset were con-structed and performed, respectively, using R software.Savoi et al. BMC Plant Biology  (2016) 16:67 Page 14 of 17The metabolite correlation network was constructed foreach condition (C and D) using all 74 metabolite accumu-lation profiles separately. The PCC was used as an indexof similarity between any two variables (i.e. metabolites).Correlation pairs were deemed statistically significantwhen the |PCC| > 0.8 and P value < 0.001 (2,000 permuta-tions). The Cytoscape software (version 3.1.1) [64] wasused for network visualization and analysis of networkproperties such as the average node degree, clustering co-efficient, and network density. Additionally, the two matri-ces (C and D) of metabolite and transcript datasets weremerged and used for the construction of a globalmetabolite-transcript network focused on structural genesof phenylpropanoid, flavonoid, carotenoid, fatty acid, andterpenoid pathway. Selected networks were constructedfor three monoterpenes (linalool, nerol, and α-terpineol)considering the top 100 correlating genes. All calculationsand permutation tests were performed in R using the‘rsgcc’ package [65].Promoter motif enrichment analysis was conducted asdescribed previously in Ma et al. [66]. A total of 29,839grapevine promoter sequences (1 kb upstream of the 5′UTR) based on the 12x grapevine genome assembly wereretrieved from Gramene v45 (http://www.gramene.org/).Known cis-regulatory motifs of plants were retrieved fromPLACE [67]. Enrichment of motifs was validated using thehypergeometric distribution test. Cis-regulatory motifswere considered significantly enriched if the associated Pvalue was < 0.01 and at least 10 promoters were associatedwith the given motif.Availability of Data and MaterialsAll raw sequence reads have been deposited in NCBI Se-quence Read Archive (http://www.ncbi.nlm.nih.gov/sra).The BioProject and SRA accession are PRJNA313234and SRP070855, respectively.Additional filesAdditional file 1: Table S1. Comparison of the climatological data inthe season of the study (2012) with data of the 2000–2012 period.(DOC 35 kb)Additional file 2: Table S2. Effect of water deficit on crop production.Yield per vine, clusters per vine, and cluster weight of fully irrigated (C,controls) and deficit irrigated (D, water deficit) grapevines reported as theaverage ± the standard error. Numbers in bold indicate significantdifferences between treatments (P < 0.05) identified by one-way ANOVA(n = 4). (DOC 29 kb)Additional file 3: Table S3. List of phenolics, carotenoids, tocopherolsand free VOCs identified in the study. (DOC 88 kb)Additional file 4: Figure S1. Principal component analysis (PCA) of theberry secondary metabolite profile of 48 independent samples collectedfrom C and D vines at 27, 41, 54, 68, 82, and 93 DAA. Full and opensymbols identify C and D berries, respectively. (PNG 27 kb)Additional file 5: Figure S2. Phenolic profiling. Trends of phenolicconcentrations in C and D berries during fruit development. (DOCX 358 kb)Additional file 6: Figure S3. Carotenoid and tocopherol profiling.Trends of carotenoid and tocopherol concentrations in C and Dberries during fruit development. (DOCX 163 kb)Additional file 7: Figure S4. VOC profiling. Trends of VOCconcentrations in C and D berries during fruit development. (DOCX 468 kb)Additional file 8: Table S7. Network analysis metrics. Detailedinformation concerning the various metabolite/transcript annotations,correlation relationships (PCC), and statistical significance (empirical P value)are depicted for the global (C and D) metabolite-metabolite network (A)and metabolite-transcript network in C (B) and D (C). Frequency tables forthe significant metabolite-transcript correlations from (B) and (C) aresummarized in (D). Correlation values (PCC, statistical significance ofPCC, and correlation ranks) for the predicted monoterpene (linalool,nerol, and α-terpineol)-transcript network reported in Fig. 8a (E). Contingencytables containing enriched (P < 0.01) PLACE cis-regulatory elementsand associated information (e.g., motif description, occurrence of motifwithin promoter regions) identified in the combined monoterpene-transcript network presented in Fig. 8a (F). (XLSX 91 kb)Additional file 9: Figure S5. Carotenoid network. Networkrepresentation of carotenoids in C (A) and D (B) berries duringdevelopment. Nodes represent ‘metabolites’ and edges represent‘relationships’ between any two metabolites. Edges colored in red andblue represent significant (P < 0.001) positive and negative correlations,respectively. Metabolites in bold indicate a significant effect (P < 0.05) ofwater deficit on the concentration of that metabolite at one or moredevelopmental stages. Number of correlating edges, average nodeneighborhood, and clustering coefficient were similar between C and Dnetworks. (PNG 63 kb)Additional file 10: Table S4. RNA sequencing analysis metrics.Transcriptome analyses were performed in C and D berries at threeselected berry developmental stages (41 DAA, before ripening; 68 DAA,beginning of ripening; 93 DAA, late ripening) using an Illumina HiSeqplatform. (DOC 54 kb)Additional file 11: Table S5. List of genes assayed for expression byqPCR. For each, the coefficient of correlation between RNA-Seq data andqPCR data, the P value, forward and reverse primers used, and literaturereferences are shown. (DOC 38 kb)Additional file 12: Figure S6. Scatterplot of the correlation betweenthe fold changes (log2(D/C)) in the expression levels of the 15 genesreported in Additional file 11: Table S5 obtained by RNA-seq and qPCRanalyses. (TIF 1009 kb)Additional file 13: Table S6. Differentially expressed genes. Summaryof differentially expressed genes (P < 0.05) and associated information(12x V1 identification number, predicted annotation, fold change values,and adjusted P value) identified between C and D at 41 (A), 68 (B), and93 (C) days after anthesis (DAA). The ‘Venn’ column in A, B, C identifieswhere each gene maps in the Venn diagram of Fig. 4. List of the genesdifferentially expressed at all the developmental stages (D). Plant geneontology (GO) slim biological process categories enriched (P < 0.05)within significantly up- and down-regulated genes at 41, 68, and 93 DAAare represented as the relative (%) contribution of genes ascribed to thatbiological process over the total DE genes at each developmental stage(E). Level of expression of DE genes of the phenylpropanoid and flavon-oid (F), carotenoid (G), and terpenoid (H) pathways at 41, 68, 93 DAA in Cand D berries. Red-blue color scale identifies high (red) and low (blue)levels of expression. (XLSX 1175 kb)Additional file 14: Figure S7. Target gene expression analysis. Impactof water deficit on transcript abundance of selected genes of thephenylpropanoid, flavonoid, and terpenoid pathway. Using qPCR, geneexpression was analyzed at each sampling time. Gene expression levelsanalyzed with RNA-sequencing at 41, 68, and 93 DAA are reported ininset graphs for comparison. Bars represent ± SE. Asterisks indicatesignificant differences between treatments at P < 0.05 (*). (PNG275 kb)Additional file 15: Figure S8. Evolution of the relative skin and seedmasses expressed as % of berry fresh weight (FW) across development inC and D berries. (TIF 1797 kb)Savoi et al. BMC Plant Biology  (2016) 16:67 Page 15 of 17Abbreviations4CL: 4-coumarate-CoA ligase; ANR: anthocyanidin reductase; BCH: β-carotenehydroxylase; C: control; C3H: p-coumaroyl shikimate 3'-hydroxylase;C4H: trans-cinnamate 4-monooxygenase; CCD1: (5,6) (5′,6′) (9,10) (9′,10′)cleavage dioxygenase; CCD4: (9,10) (9′,10′) cleavage dioxygenase;CCoAMT: caffeoyl-CoA 3-O-methyltransferase; CHI: chalcone isomerase;CHS: chalcone synthase; CISO: carotenoid isomerase; COMT: caffeic acid3-O-metyltransferase; D: deficit irrigation; DAA: days after anthesis;DE: differentially expressed; DFR: dihydroflavonol reductase; DVB-CAR-PDMS: divinylbenzene-carboxen-polydimethylsiloxane; DXR: 1-deoxy-D-xylulose-5-phosphate reductoisomerase; DXS: 1-deoxy-D-xylulose-5-phosphate synthase; ETc.: crop evapotranspiration; F3′5′H: flavonoid-3′5′-hydroxylase; F3H: flavanone-3-hydroxylase; FLS: flavonol synthase;FW: fresh weight; GO: gene ontology; HCT: hydroxycinnamoyl-CoA:shikimate/quinate hydroxycinnamoyltransferase; HPLC-DAD: high performance liquidchromatography-diode array detector; HPT: homogentisate phytyl transferase;HS-SPME-GC-MS: headspace-solid phase microextraction-gas chromatography–mass spectrometry; LAR: leucoanthocyanidin reductase; LBCY: lycopeneβ-cyclase; LDOX: leucoanthocyanidin dioxygenases; LECY: lycopeneε-cyclase; LUT: carotene hydroxylase; MEP: 2C-methyl-D-erythritol-4-phosphate; MVA: mevalonate; PAL: phenylalanine ammonia lyase;PCA: principal component analysis; PCC: Pearson correlation coefficient;PSY: phytoene synthase; QC: quality control; qPCR: quantitative PCR;RNA-seq: RNA sequencing; STS: stilbene synthase; TPS: terpene synthase;UFGT: UDP-glucose:flavonoid 3-O-glucosyltransferase; UHPLC-MS/MS: ultra-high performance liquid chromatography-tandem mass spec-trometer; VOCs: volatile organic compounds; Vvi: Vitis vinifera; ZDS: ζ-Carotene desaturase; ΨStem: stem water potential.Competing interestsThe authors declare that they have no competing interests.Authors’ contributionsSS participated in the design of the study, carried out the metaboliteextractions and analyses, RNA extractions, and part of the transcriptomedata analyses, and drafted part of the manuscript; DCJW carried out thenetwork analysis and part of the transcriptome data analysis, and draftedpart of the manuscript; PA contributed to the metabolite extractions andanalyses; MM performed part of the transcriptome analysis; BBperformed the field experiment; EP coordinated the field experiments;AF critically revised the manuscript; FM participated in the design of thestudy, supervised the metabolite analysis, and critically revised themanuscript; SDC conceived the study, coordinated the experiments,supervised the field, transcriptome, and network analyses, interpreted theresults, and drafted part of the manuscript. All authors read andapproved the final manuscript.AcknowledgementThis study was funded by the Italian Ministry of Agricultural and ForestryPolicies (VIGNETO); the Fondazione Edmund Mach (GMPF), the COST ActionFA1106 Quality Fruit, the University of British Columbia (10R18459), andNatural Sciences and Engineering Research Council of Canada (10R23082).We thank Georg Weingart for his contribution in the sample preparation formetabolite extractions and Christopher J. Walkey for critically reading thearticle.Author details1Department of Food Quality and Nutrition, Research and Innovation Centre,Fondazione Edmund Mach, Via E. Mach 1, 38010 San Michele all’Adige, Italy.2Dipartimento di Scienze Agro-alimentari, Ambientali e Animali, University ofUdine, Via delle Scienze 208, 33100 Udine, Italy. 3Wine Research Centre, TheUniversity of British Columbia, 2205 East Mall, Vancouver, BC V6T 1Z4,Canada. 4Istituto di Genomica Applicata, Parco Scientifco e Tecnologico LuigiDanieli, via Jacopo Linussio 51, 33100 Udine, Italy. 5The Jacob BlausteinInstitutes for Desert Research, Ben-Gurion University of the Negev, MidreshetBen-Gurion, Israel.Received: 11 December 2015 Accepted: 15 March 2016References1. Hartmann T. From waste products to ecochemicals: Fifty years research ofplant secondary metabolism. Phytochemistry. 2007;68:2831–46.2. Kliebenstein DJ. Secondary metabolites and plant/environment interactions:A view through Arabidopsis thaliana tinged glasses. Plant Cell Environ. 2004;27:675–84.3. Klee HJ, Giovannoni JJ. Genetics and control of tomato fruit ripening andquality attributes. Annu Rev Genet. 2011;45:41–59.4. Lund ST, Bohlmann J. The molecular basis for wine grape quality–a volatilesubject. Science. 2006;311:804–5.5. Ferrandino A, Lovisolo C. Abiotic stress effects on grapevine (Vitis vinifera L.):Focus on abscisic acid-mediated consequences on secondary metabolismand berry quality. Environ Exp Bot Elsevier BV. 2014;103:138–47.6. Chaves MM, Zarrouk O, Francisco R, Costa JM, Santos T, Regalado AP, et al.Grapevine under deficit irrigation: hints from physiological and moleculardata. Ann Bot. 2010;105:661–76.7. Hannah L, Roehrdanz PR, Ikegami M, Shepard AV, Shaw MR, Tabor G,et al. Climate change, wine, and conservation. Proc Natl Acad Sci. 2013;110:6907–12.8. Mattivi F, Guzzon R, Vrhovsek U, Stefanini M, Velasco R. Metaboliteprofiling of grape: flavonols and anthocyanins. J Agric Food Chem.2006;54:7692–702.9. Degu A, Morcia C, Tumino G, Hochberg U, Toubiana D, Mattivi F, et al.Metabolite profiling elucidates communalities and differences in thepolyphenol biosynthetic pathways of red and white Muscat genotypes.Plant Physiol Biochem Elsevier Masson SAS. 2015;86:24–33.10. Hochberg U, Degu A, Cramer GR, Rachmilevitch S, Fait A. Cultivar specificmetabolic changes in grapevines berry skins in relation to deficit irrigationand hydraulic behavior. Plant Physiol Biochem Elsevier Masson SAS. 2015;88:42–52.11. Young PR, Lashbrooke JG, Alexandersson E, Jacobson D, Moser C, Velasco R,et al. The genes and enzymes of the carotenoid metabolic pathway in Vitisvinifera L. BMC Genomics. 2012;13:243.12. Robinson AL, Boss PK, Solomon PS, Trengove RD, Heymann H, Ebeler SE.Origins of grape and wine aroma. Part 1. Chemical components andviticultural impacts. Am. J Enol Vitic. 2014;65:1–24.13. Deluc LG, Quilici DR, Decendit A, Grimplet J, Wheatley MD, Schlauch KA, etal. Water deficit alters differentially metabolic pathways affecting importantflavor and quality traits in grape berries of Cabernet Sauvignon andChardonnay. BMC Genomics. 2009;10:212.14. Blanco-Ulate B, Amrine KCH, Collins TS, Rivero RM, Vicente AR, Morales-CruzA, et al. Developmental and metabolic plasticity of white-skinned grapeberries in response to Botrytis cinerea during noble rot. Plant Physiol. 2015;169:2422–43.15. Young P, Eyeghe-Bickong HA, du Plessis K, Alexandersson E, Jacobson DA,Coetzee ZA, et al. Grapevine plasticity in response to an alteredmicroclimate: Sauvignon Blanc modulates specific metabolites in responseto increased berry exposure. Plant Physiol. 2015;170(3):1235–54.16. Jaillon O, Aury J-M, Noel B, Policriti A, Clepet C, Casagrande A, et al. Thegrapevine genome sequence suggests ancestral hexaploidization in majorangiosperm phyla. Nature. 2007;449:463–7.17. Höll J, Vannozzi A, Czemmel S, D’Onofrio C, Walker AR, Rausch T, et al. TheR2R3-MYB transcription factors MYB14 and MYB15 regulate stilbenebiosynthesis in Vitis vinifera. Plant Cell. 2013;25:4135–49.18. Vannozzi A, Dry IB, Fasoli M, Zenoni S, Lucchin M. Genome-wide analysis ofthe grapevine stilbene synthase multigenic family: genomic organizationand expression profiles upon biotic and abiotic stresses. BMC Plant Biol.2012;12:130.19. Martin DM, Aubourg S, Schouwey MB, Daviet L, Schalk M, Toub O, et al.Functional annotation, genome organization and phylogeny of thegrapevine (Vitis vinifera) terpene synthase gene family based on genomeassembly, flcdna cloning, and enzyme assays. BMC Plant Biol. 2010;10:226.20. Castellarin SD. Matthews M a, Di Gaspero G, Gambetta GA. Water deficitsaccelerate ripening and induce changes in gene expression regulatingflavonoid biosynthesis in grape berries. Planta. 2007;227:101–12.21. Roby G, Matthews MA. Relative proportions of seed, skin and flesh, in ripeberries from Cabernet Sauvignon grapevines grown in a vineyard eitherwell irrigated or under water deficit. Aust J Grape Wine Res. 2004;10:74–82.22. Roby G, Harbertson JF, Adams DA, Matthews MA. Berry size and vine waterdeficits as factors in winegrape composition: Anthocyanins and tannins.Aust J Grape Wine Res. 2004;10:100–7.Savoi et al. BMC Plant Biology  (2016) 16:67 Page 16 of 1723. Bucchetti B, Matthews MA, Falginella L, Peterlunger E, Castellarin SD. Effectof water deficit on Merlot grape tannins and anthocyanins across fourseasons. Sci Hortic. 2011;128:297–305.24. Demmig B, Winter K, Krüger A, Czygan F-C. Zeaxanthin and the heatdissipation of excess light energy in Nerium Oleander exposed to acombination of high light and water stress. Plant Physiol. 1988;87:17–24.25. Zhao Q, Wang G, Ji J, Jin C, Wu W, Zhao J. Over-expression of Arabidopsisthaliana β-carotene hydroxylase (chyB) gene enhances drought tolerance intransgenic tobacco. J Plant Biochem Biotechnol. 2013;23:190–8.26. Munné-Bosch S, Schwarz K, Alegre L. Enhanced formation of α-Tocopheroland highly oxidized abietane diterpenes in water-stressed rosemary plants.Plant Physiol. 1999;121:1047–52.27. Cela J, Chang C, Munné-Bosch S. Accumulation of γ-rather than α-tocopherolalters ethylene signaling gene expression in the vte4 Mutant of Arabidopsisthaliana. Plant Cell Physiol. 2011;52:1389–400.28. Battilana J, Emanuelli F, Gambino G, Gribaudo I, Gasperi F, Boss PK,et al. Functional effect of grapevine 1-deoxy-D-xylulose 5-phosphatesynthase substitution K284N on Muscat flavour formation. J Exp Bot.2011;62:5497–508.29. Selmar D, Kleinwächter M. Stress enhances the synthesis of secondary plantproducts: The impact of stress-related over-reduction on the accumulationof natural products. Plant Cell Physiol. 2013;54:817–26.30. Alonso R, Berli FJ, Bottini R, Piccoli P. Acclimation mechanisms elicited bysprayed abscisic acid, solar UV-B and water deficit in leaf tissues of field-grown grapevines. Plant Physiol Biochem. 2015;91:56–60.31. Griesser M, Weingart G, Schoedl-Hummel K, Neumann N, Becker M, Varmuza K,et al. Severe drought stress is affecting selected primary metabolites,polyphenols, and volatile metabolites in grapevine leaves (Vitis vinifera cv.Pinot noir). Plant Physiol. Biochem. Elsevier Masson SAS. 2015;88:17–26.32. Ripoll J, Urban L, Staudt M, Lopez-Lauri F, Bidel LPR, Bertin N. Watershortage and quality of fleshy fruits—making the most of the unavoidable.J Exp Bot. 2014;65:4097–117.33. Sánchez AM, Winterhalter P. Carotenoid cleavage products in saffron(Crocus sativus L.). Carotenoid Cleavage Prod.. American Chemical Society;2013; 45–63.34. Stitt M, Sulpice R, Keurentjes J. Metabolic networks: how to identify keycomponents in the regulation of metabolism and growth. Plant Physiol.2010;152:428–44.35. Hochberg U, Degu A, Toubiana D, Gendler T, Nikoloski Z, Rachmilevitch S,et al. Metabolite profiling and network analysis reveal coordinated changesin grapevine water stress response. BMC Plant Biol. 2013;13:184.36. Zamboni A, Di Carli M, Guzzo F, Stocchero M, Zenoni S, Ferrarini A, et al.Identification of putative stage-specific grapevine berry biomarkers andomics data integration into networks. Plant Physiol. 2010;154:1439–59.37. Degu A, Hochberg U, Sikron N, Venturini L, Buson G, Ghan R, et al.Metabolite and transcript profiling of berry skin during fruit developmentelucidates differential regulation between Cabernet Sauvignon and Shirazcultivars at branching points in the polyphenol pathway. BMC Plant Biol.2014;14:188.38. Wong DC, Sweetman C, Drew DP, Ford CM. VTCdb: a gene co-expressiondatabase for the crop species Vitis vinifera (grapevine). BMC Genomics.2013;14:882.39. Costantini L, Malacarne G, Lorenzi S, Troggio M, Mattivi F, Moser C, et al.New candidate genes for the fine regulation of the colour of grapes. J ExpBot. 2015;66:4427–40.40. Ma D-M, Wang Z, Wang L, Alejos-Gonzales F, Sun M-A, Xie D-Y. A Genome-wide Scenario of Terpene Pathways in Self-pollinated Artemisia annua. MolPlant. 2015;8(11):1580–98. Elsevier Ltd.41. Gómez-Plaza E, Mestre-Ortuño L, Ruiz-García Y, Fernández-Fernández JI,López-Roca JM. Effect of benzothiadiazole and methyl jasmonate on thevolatile compound composition of monastrell grapes and wines. Am J EnolVitic. 2012;63(3):394–401.42. Reeves PH, Ellis CM, Ploense SE, Wu M-F, Yadav V, Tholl D, et al. A regulatorynetwork for coordinated flower maturation. PLoS Genet. 2012;8:e1002506.43. Carbonell-Bejerano P, Diago M-P, Martínez-Abaigar J, Martínez-Zapater JM,Tardáguila J, Núñez-Olivera E. Solar ultraviolet radiation is necessary toenhance grapevine fruit ripening transcriptional and phenolic responses.BMC Plant Biol. 2014;14:183.44. Gil M, Pontin M, Berli F, Bottini R, Piccoli P. Metabolism of terpenes in theresponse of grape (Vitis vinifera L.) leaf tissues to UV-B radiation.Phytochemistry. 2012;77:89–98.45. Reynolds AG, Wardle DA. Impact of various canopy manipulationtechniques on growth, yield, fruit composition, and wine quality ofGewürztraminer. Am J Enol Vitic. 1989;40:121–9.46. Castellarin SD, Pfeiffer A, Sivilotti P, Degan M, Peterlunger E, Di Gaspero G.Transcriptional regulation of anthocyanin biosynthesis in ripening fruits ofgrapevine under seasonal water deficit. Plant Cell Environ. 2007;30:1381–99.47. Santos TP, Lopes CM, Rodrigues ML, de Souza CR, Ricardo-da-Silva JM,Maroco JP, et al. Effects of partial root-zone drying irrigation on clustermicroclimate and fruit composition of field-grown Castelão grapevines.VITIS - J Grapevine Res. 2005;44:117.48. Gerdes SM, Winterhalter P, Ebeler SE. Effect of sunlight exposure onnorisoprenoid formation in white Riesling grapes. In Carotenoid-Deriv.Aroma Compounds. P Winterhalter and R. Rouseff (eds) 262–272. ACSSymp. Series 802. 2002.49. Lee S-H, Seo M-J, Riu M, Cotta JP, Block DE, Dokoozlian NK, et al. VineMicroclimate and norisoprenoid concentration in Cabernet sauvignongrapes and wines. Am J Enol Vitic. 2007;58:291–301.50. Skinkis PA, Bordelon BP, Butz EM. Effects of sunlight exposure on berry andwine monoterpenes and sensory characteristics of Traminette. Am J EnolVitic. 2010;61:147–56.51. Shinozaki K, Yamaguchi-Shinozaki K. Gene networks involved in droughtstress response and tolerance. J Exp Bot. 2007;58:221–7.52. Herrera JC, Bucchetti B, Sabbatini P, Comuzzo P, Zulini L, Vecchione A, et al.Effect of water deficit and severe shoot trimming on the composition of Vitisvinifera L. Merlot grapes and wines. Aust J Grape Wine Res. 2015;21:254–65.53. Vrhovsek U, Masuero D, Gasperotti M, Franceschi P, Caputi L, Viola R, et al.A versatile targeted metabolomics method for the rapid quantification ofmultiple classes of phenolics in fruits and beverages. J Agric Food Chem.2012;60:8831–40.54. Wehrens R, Carvalho E, Masuero D, De Juan A, Martens S. High-throughputcarotenoid profiling using multivariate curve resolution. Anal Bioanal Chem.2013;405:5075–86.55. Fedrizzi B, Carlin S, Franceschi P, Vrhovsek U, Wehrens R, Viola R, et al.D-optimal design of an untargeted HS-SPME-GC-TOF metabolite profilingmethod. The Analyst. 2012;137:3725.56. Del Fabbro C, Scalabrin S, Morgante M, Giorgi FM. Evaluation of readtrimming effects on Illumina NGS data analysis. PLoS One. 2013;8:e85024.57. Trapnell C, Roberts A, Goff L, Pertea G, Kim D, Kelley DR, et al. Differentialgene and transcript expression analysis of RNA-seq experiments withTopHat and Cufflinks. Nat Protoc. 2012;7:562–78.58. Anders S, Pyl PT, Huber W. HTSeq—a Python framework to work with high-throughput sequencing data. Bioinformatics. 2015;31:166–9.59. Love MI, Huber W, Anders S. Moderated estimation of fold change anddispersion for RNA-seq data with DESeq2. Genome Biol. 2014;15:550.60. Grimplet J, Hemert JV, Carbonell-Bejerano P, Díaz-Riquelme J, Dickerson J,Fennell A, et al. Comparative analysis of grapevine whole-genome genepredictions, functional annotation, categorization and integration of thepredicted gene sequences. BMC Res Notes. 2012;5:213.61. Naithani S, Raja R, Waddell EN, Elser J, Gouthu S, Deluc LG, et al. VitisCyc: ametabolic pathway knowledgebase for grapevine (Vitis vinifera). Front PlantSci. 2014;5:1–11.62. Maere S, Heymans K, Kuiper M. BiNGO: A Cytoscape plugin to assessoverrepresentation of Gene Ontology categories in Biological Networks.Bioinformatics. 2005;21:3448–9.63. Falginella L, Di Gaspero G, Castellarin SD. Expression of flavonoid genes inthe red grape berry of “Alicante Bouschet” varies with the histologicaldistribution of anthocyanins and their chemical composition. Planta. 2012;236:1037–51.64. Shannon P, Markiel A, Ozier O, Baliga NS, Wang JT, Ramage D, et al.Cytoscape: a software environment for integrated models of biomolecularinteraction networks. Genome Res. 2003;13:2498–504.65. Ma C, Wang X. Application of the gini correlation coefficient to inferregulatory relationships in transcriptome analysis. Plant Physiol. 2012;160:192–203.66. Ma S, Shah S, Bohnert HJ, Snyder M, Dinesh-Kumar SP. Incorporating motifanalysis into gene co-expression networks reveals novel modular expressionpattern and new signaling pathways. PLoS Genet. 2013;9:e1003840.67. Higo K, Ugawa Y, Iwamoto M, Korenaga T. Plant cis-acting regulatory DNAelements (PLACE) database: 1999. Nucleic Acids Res. 1999;27(1):297–300.Savoi et al. BMC Plant Biology  (2016) 16:67 Page 17 of 17


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items