UBC Faculty Research and Publications

Poplar trees reconfigure the transcriptome and metabolome in response to drought in a genotype- and time-of-day-dependent… Hamanishi, Erin T; Barchet, Genoa L; Dauwe, Rebecca; Mansfield, Shawn D; Campbell, Malcolm M Apr 21, 2015

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

Item Metadata


52383-12864_2015_Article_1535.pdf [ 1.28MB ]
JSON: 52383-1.0223861.json
JSON-LD: 52383-1.0223861-ld.json
RDF/XML (Pretty): 52383-1.0223861-rdf.xml
RDF/JSON: 52383-1.0223861-rdf.json
Turtle: 52383-1.0223861-turtle.txt
N-Triples: 52383-1.0223861-rdf-ntriples.txt
Original Record: 52383-1.0223861-source.json
Full Text

Full Text

RESEARCH ARTICLEPoplar trees reconfigure thmetabolome in responsea,6,Drought is one of the most significant environmental widespread die-off of Populus tremuloides [2].Hamanishi et al. BMC Genomics  (2015) 16:329 DOI 10.1186/s12864-015-1535-zat the physiological level, strategies to contend withreduced water availability can vary from isohydric toToronto, 25 Willcocks St., Toronto, ON M5S 3B2, CanadaFull list of author information is available at the end of the articlestresses that can impinge on the growth and productivityof forests. Recently, more severe and frequent droughtevents have been associated with increased global forest-dieback [1]. In North America, severe drought andTrees of the genus Populus are often characterized bytheir high productivity [3,4]. The rapid growth ratesattributed to poplars are often associated with significantwater requirements. Thus, the growth, productivityand survival of poplars is often dependent on wateravailability [5-7].In response to water limitation, plants may exhibitadaptation at morphological, physiological and biochem-ical level to contend with the abiotic stress. For instance,* Correspondence: malcolm.campbell@utoronto.ca1Faculty of Forestry, University of Toronto, 33 Willcocks St., Toronto, ON M5S3B2, Canada2Centre for the Analysis of Genome Evolution and Function, University ofBackgroundAbstractBackground: Drought has a major impact on tree growth and survival. Understanding tree responses to this stresscan have important application in both conservation of forest health, and in production forestry. Trees of the genusPopulus provide an excellent opportunity to explore the mechanistic underpinnings of forest tree drought responses,given the growing molecular resources that are available for this taxon. Here, foliar tissue of six water-deficit stressedP. balsamifera genotypes was analysed for variation in the metabolome in response to drought and time of day byusing an untargeted metabolite profiling technique, gas chromatography/mass-spectrometry (GC/MS).Results: Significant variation in the metabolome was observed in response the imposition of water-deficit stress.Notably, organic acid intermediates such as succinic and malic acid had lower concentrations in leaves exposed todrought, whereas galactinol and raffinose were found in increased concentrations. A number of metabolites withsignificant difference in accumulation under water-deficit conditions exhibited intraspecific variation in metaboliteaccumulation. Large magnitude fold-change accumulation was observed in three of the six genotypes. In order tounderstand the interaction between the transcriptome and metabolome, an integrated analysis of the drought-responsivetranscriptome and the metabolome was performed. One P. balsamifera genotype, AP-1006, demonstrated a lack ofcongruence between the magnitude of the drought transcriptome response and the magnitude of the metabolomeresponse. More specifically, metabolite profiles in AP-1006 demonstrated the smallest changes in response towater-deficit conditions.Conclusions: Pathway analysis of the transcriptome and metabolome revealed specific genotypic responses withrespect to primary sugar accumulation, citric acid metabolism, and raffinose family oligosaccharide biosynthesis. Theintraspecific variation in the molecular strategies that underpin the responses to drought among genotypes may havean important role in the maintenance of forest health and productivity.Keywords: Balsam poplar, Drought, Metabolome, Transcriptome, Trees, Forestshigher summer temperatures have been linked with thegenotype- and time-of-dErin T Hamanishi1,2, Genoa LH Barchet3, Rebecca Dauwe3© 2015 Hamanishi et al.; licensee BioMed CenCommons Attribution License (http://creativecreproduction in any medium, provided the orDedication waiver (http://creativecommons.orunless otherwise stated.Open Accesse transcriptome andto drought in ay-dependent mannerShawn D Mansfield3 and Malcolm M Campbell1,2,4,5*tral. This is an Open Access article distributed under the terms of the Creativeommons.org/licenses/by/4.0), which permits unrestricted use, distribution, andiginal work is properly credited. The Creative Commons Public Domaing/publicdomain/zero/1.0/) applies to the data made available in this article,Hamanishi et al. BMC Genomics  (2015) 16:329 Page 2 of 16ansiohydric [8]. Poplar trees generally respond by closingtheir stomata during periods of water limitation to reducewater loss, thus limiting the photosynthetic capacity of thetrees [9]; however, this response to water limitation isoften variable among poplar trees.Among three closely related poplar genotypes,Larcheveque et al. [10] found that the response todrought varied at the physiological level. Specifically,two hybrid Populus balsamifera and one native P. bal-samifera genotypes had variable growth rates andwater use efficiencies under drought conditions [10].Previous studies have identified significant variation atthe molecular level among poplar trees that may underpinvariation at the morphological and physiological level.Large-scale microarray experiments studying water-deficitstress have identified many transcripts with known rolesin stress tolerance in Populus [11-15]. For example, amongsix genotypes of P. balsamifera, Hamanishi et al. [14]observed significant variation in patterns of transcript ac-cumulation. The variation in the drought transcriptomesamong the six P. balsamifera genotypes was correlatedwith their ability to maintain growth following waterlimitation [14], highlighting the complexity in the droughtresponse among poplars.The great intra- and interspecific variation seen amongpoplar species is also reflected at the biochemical level.In trees, metabolites involved in osmotic adjustment,protection and stabilization of cellular structure andredox regulation are often involved in drought responses[16]. For example, the amino acids proline (Pro), valine(Val) and isoleucine (Ile), carbohydrates such as sucrose,raffinose family oligosaccharides (RFO) and sorbitol,polyols, and organic acids have been shown to vary inabundance in response to drought [11]. Elevated levelsof sucrose were observed in leaf tissue of water-stressedPopulus tomentosa [17]; whereas a combination of glu-cose, fructose, and sucrose accumulated in Populushybrids in response to drought [18]. Some of thesecompounds are thought to function as osmolytes, main-taining cell turgor and stabilisation of cellular proteins[19]. Similarly, raffinose and the RFO accumulate in re-sponse to water-stress, and are hypothesised to be osmo-protectants, with the capacity for membrane and enzymestability [11,20], along with a putative role as hydroxylradical scavengers.Proline accumulation has long been associated withstress tolerance in plants, and is likely one of the mostwidely distributed osmolytes among plants and animals[17,21]. Similar to carbohydrates, proline is hypothesisedto aid in the osmotic adjustment in response to drought;however, proline is also hypothesised to have roles inreactive oxygen species (ROS) scavenging and mem-brane stability. Proline has been shown to accumulatein severely water-stressed mature Populus nigra leaves[18,22]; whereas no significant increase in proline accu-mulation was observed in field-grown, drought-treatedPopulus hybrids [6].Organic acids have also been implicated in the biochem-ical response to drought. For example, malic acid in-creased in abundance under mild periods of water stress[6,19,23]. Unlike carbohydrate and amino acid accumula-tion, malic acid accumulation may be a function of thestomatal system in plants rather than being osmoticallyactive [24].As the response to drought stress is not simply theproduct of the drought-responsive transcriptome, com-plexity in the whole-plant response to drought is theresult of the interactions between genes, transcripts,proteins, metabolites, and the environment. The modelplant genus Populus provides an opportunity to explorethe relationship between the drought transcriptome andthe drought metabolome. In keeping with this, the rela-tionship between the transcriptome and metabolome forspecific metabolic pathways in Populus has also beencharacterised in response to salt stress, revealing the im-portance of control mechanisms for osmotic adjustment[19,25].In order to test hypotheses related to intra-specificvariation in drought responses in Populus, the transcrip-tomes and metabolomes of six genotypes of P. balsami-fera were examined. Shared versus genotype-specific P.balsamifera drought transcriptomes were identified [14]and superimposed onto metabolome variation. This ap-proach identified important pathways in the drought re-sponse, and highlighted genotypic-specific responses thatprovide insight into different mechanisms of acclimationto water-limiting conditions.MethodsPlant material and experimental designPopulus balsamifera ramets were grown in a climatecontrolled growth chamber at the University of Torontousing conditions as described by Hamanishi et al. [14].Un-rooted cuttings of six P. balsamifera genotypes(AP-947, AP-1005, AP-1006, AP-2278, AP-2298 andAP-2300; Alberta Pacific, Boyle, Alberta) were propagatedand grown under well watered conditions for 9 weeks, atwhich point, water-deficit stress was imposed on half thetrees by withholding water, while temperature, light, andrelative humidity were held constant.Foliar tissue was harvested for metabolite and tran-scriptome analysis 15 days after the onset of the waterwithdrawal. For the transcriptome analysis, the first fullyexpanded [leaf plastochron index (LPI = 7)], mature leafwas collected from each tree; three leaves were pooledto create a single replicate. Triplicate replicates werecollected for each genotype and treatment combinationat pre-dawn (PD; 1 hour before the light period) andHamanishi et al. BMC Genomics  (2015) 16:329 Page 3 of 16mid-day (MD; middle of the light period). Leaves wereimmediately flash frozen in liquid nitrogen, and thenground to a fine powder in preparation for RNA isolation,as described by Hamanishi et al. [24]. For the metaboliteanalysis, a single mature, fully expanded leaf was collectedfrom each tree (n = 10 per genotype per treatment at MDand PD) and immediately flash frozen. Harvested foliartissue was weighed to determine fresh weight (FW), subse-quently freeze-dried, and weighed again to determine dryweight (DW).Non-targeted metabolic profiling by gas chromatography/mass spectrometryMetabolite extraction was performed using a methanol/chloroform-based extraction protocol as described byRobinson et al. [19,25]. Four to 10 biological replicateswere sampled per genotype, treatment, time of day(Additional file 1: Table S1). Approximately 0.5 mL ofsample was extracted in 1300 μL 97% methanol with theinternal standard ortho-anisic acid (0.62 mg mL−1) for15 minutes at 70°C prior to centrifugation at 17,000 gfor 10 minutes. The supernatant was transferred to anew 1.5-mL tube. 130 μL chloroform and 270 μL distilled,deionized water was added and the tube was gently shakenprior to centrifugation at 17,000 g for 5 minutes. A 400 μLaliquot of the upper polar phase was transferred to a new1.5 mL tube and dried overnight at 30°C in a Vacufuge(Eppendorf).Samples were then derivatised for gas chromatography/mass spectrometry (GC/MS) analysis by resuspension in50 μL methoxyamine hydrochloride solution (20 mg mL−1in pyridine) and incubated at 37°C for 2 hours. 10 μLof n-alkane standard and 70 μL of N-methyl-N-tri-methylsilytriflouroacetamide (MSTFA) was added, andincubated at 37°C for 30 minutes with constant agita-tion. Samples were then filtered through filter paperand allowed to rest at room temperature until GC/MSanalysis.GC/MS analysis was conducted on a ThermoFinniganTrace GC-PolarisQ ion trap MS, fitted with an AS2000auto-sampler and a split-injector (Thermo Electron Co.,Waltham, MA, USA). The GC was equipped with aRestek Rtx-5MS column (fused silica, 30 m, 0.25 mmID, stationary phase: 5% diphenyl, 95% dimethyl polysi-loxane). The GC conditions were set with an inlettemperature of 250°C, helium carrier gas at a constantflow rate of 1 mL min−1, injector split ratio 10:1, restingoven temperature at 70°C and a GC/MS transfer linetemperature of 300°C. After a sample injection of 1 μL,the oven temperature was held at 70°C for 2 minutesprior to ramping to 325°C at a rate of 8°C min−1. Thetemperature was held at 325°C for 6 minutes beforecooling to the initial resting oven temperature, prior tothe next run.For MS analysis in the positive electron ionisationmode an ionization potential of 70 eV was used and theforeline was evacuated to 40 mTorr with helium gasflow in to the chamber set at a rate of 0.3 mL min−1 andthe source temperature was held at 230°C. Detector sig-nal was recorded from 3.35-35.5 minutes after the injec-tion, and, with a total scan time of 0.58 s, ions werescanned across the range of 50–650 mass units.Metabolome: data processing and statisticsThe raw metabolite data generated by GC/MS for eachmetabolite was normalised through comparison to in-ternal standards and normalised to freeze-dried DW foreach tissue sample. Raw-data was processed using XCMSas described by Krasensky and Jonak [14]. Descriptive sta-tistics were calculated using R 2.14.1 [26]. For subsequentanalyses, the metabolite data were log10 transformed. Thedataset comprised 87 metabolites, 181 samples (n = 4-10per genotype per treatment per time of day).Metabolic profiles for all samples were subjected tohierarchical cluster analysis (HCA) using Pearson correl-ation coefficient [27,28] to search for metabolic similar-ities and differences among samples and metabolites.The uncertainty associated with HCA was assessedgenerating a consensus dendrogram on 1,000 bootstrapreplicates using the R package pvclust [29]. Over-representation of a given metabolite class within acluster was determined using Fisher’s exact test in R[26]. Statistical significance was calculated using athree-way analysis of variance (ANOVA). The p valueswere corrected for multiple hypothesis testing usingthe false discovery rate (FDR) procedure of Benjaminiand Hochberg [30]. A p value of < 0.05 was consideredstatistically significant.RNA isolation and analysisRNA isolation and microarray analysis was performed as de-scribed by Hamanishi et al. [14]; all samples were uploadedto Gene Expression Omnibus (http://www.ncbi.nlm.nih.gov/geo/); accession number GSE21171. For the purposes ofsubsequent analyses, the global drought transcriptome wasconsidered to include all transcripts significant for atreatment-main effect (p < 0.05) with no log2 (fold-change)cutoff. Weighted co-expression network analysis (WGCNA)was performed using the R statistical package WGCNA witha power of 7 [31]. Functional annotations were assignedbased on the most recent version probe-set annotationsfrom Affymetrix (NetAffx build 32). Networks generatedwith WGCNA were plotted using Cytoscape [32]. Analysisof gene ontology (GO) term enrichment was calculatedby comparing the number of annotations within the listof query transcripts to all annotated transcripts on thePoplar Affymetrix Genome Array. Statistical signifi-cance was calculated using Fisher’s exact test in R [26],and applying the Benjamini-Hochberg correction toadjust for FDR. Overrepresentation of GO Slim termswas confirmed and plotted using AgriGO [33]. Mo-lecular pathways relevant to the drought transcrip-tome/metabolome were previously characterised inKyoto Encyclopedia of Genes and Genomes (KEGG;[34-36]).Results and discussionPopulus balsamifera genotypes were subjected to waterwithdrawal to induce a drought responseTo investigate the impact of drought-like conditions onconditions after 15 days of water withdrawal (treatmentmain effect; ANOVA, p < 0.05); however, a significant de-cline only occurred in genotypes AP-1005, AP-1006, andAP-2298 (Welch’s two-sample t-test, p < 0.05; Figure 1).Reduced photosynthetic rates observed in the chamber-grown seedlings are likely attributable to the lower lightintensity in the growth chamber as compared to ambientlevels in field-grown seedlings or trees.Variation in populus balsamifera metabolite profiles wasevidentTo differentiate between genotypic (G), treatment (T),and time-of-day (D) effects, metabolic profiles of P.balsamifera were analysed using gas chromatography/mass spectrometry (GC/MS). Trend analysis was re-stricted to 87 metabolites that were identified acrossall samples (n = 4-10 per genotype per treatment pertime of day), which represented both known and unknownmetabolites (Additional file 1: Table S1; Table S2). A largeHamanishi et al. BMC Genomics  (2015) 16:329 Page 4 of 16the abundance of Populus balsamifera metabolites, sixgenotypes (AP-947, AP-1005, AP-1006, AP-2278, AP-2298and AP-2300) were exposed to a prolonged period of waterwithdrawal. All plants were grown under the same con-trolled growth conditions for 9 weeks, after which half ofthe plants continued to receive water (well watered) andthe other half received no water (water deficit). This diver-gence in treatment continued for 15 days, at which pointfoliar tissue for metabolic and transcriptome analysis wascollected at PD and MD, and aboveground biomass andrelative water content (RWC) was recorded.Under conditions of water deficit, significant declinesin aboveground biomass (Table 1) and RWC were ob-served in most genotypes. In well watered conditions,AP-2278 had significantly lower aboveground biomass,and AP-2300 had the highest aboveground biomass rela-tive to all other genotypes (Table 1). Stomatal conduct-ance significantly decreased in all genotypes, with thegreatest decline observed in AP-1006, whereas genotypeAP-2278 had the smallest reduction in stomatal con-ductance in response to the imposition of water-deficitcondition. No correlation between aboveground biomassand decline in stomatal conductance was observed amongthe six genotypes. More specifically, larger plants (e.g.,AP-2300) did not show a greater reduction in stomatalconductance under water-deficit conditions. Net photo-synthetic rate decreased in response to water-deficitTable 1 Aboveground biomass for well watered andwater-deficit treated Populus balsamifera genotypes(n = 6–12 for each genotype, treatment)Genotype Aboveground drybiomass, wellwatered (g)Aboveground drybiomass waterdeficit (g)t-stat p-valueAP-947 3.12 2.17 3.04 0.008 *AP-1005 4.23 2.89 2.62 0.007 *AP-1006 3.01 2.18 2.24 0.042 *AP-2278 2.12 1.85 0.74 0.472AP-2298 2.46 2.26 0.40 0.699AP-2300 5.27 3.62 3.52 0.004 **p value < 0.05.degree of variation in metabolite abundance profiles amongsamples was observed, as indicated by the dendrogram.Notably, both genotype and treatment appeared to play animportant role in the segregation of samples (Additionalfile 2: Figure S1). The metabolite profiles from water-deficit samples of AP-1005 and AP-2278 appeared mostdifferent from the other metabolomes. Specifically, themetabolite profile for AP-1005 was separated by treat-ment rather than time of day. Additionally, samples ofAP-947 and AP-2300 clustered in a genotype-wise fash-ion, regardless of time of day or treatment.Although the metabolomes were highly variable amongsamples, further investigation of the relationship amongFigure 1 Box-plot representing net photosynthetic rate (μmol CO2m−2 s−1) for genotype AP-947, AP-1005, AP-1006, AP-2278, AP-2298and AP-2300. Well watered samples (filled); water-deficit-treated sam-ples (empty) (n = 3 per treatment per genotype). The midline of thebox represents the median value for photosynthesis, the upper andlower bounds of the box represent the interquartile range, and thewhiskers extend to the most extreme values that are not outliers.metabolites revealed 13 significant clusters of metabolitesthat had a high degree of similarity in their abundanceprofiles across all samples (Additional file 3: Figure S2), asdetermined by HCA. Unique clustering of these metabo-lites may be indicative of a different mechanism that gov-erns their regulation. For example, three of the 13 clustershad significant over-representation of a given metaboliteclass (Fisher’s exact test; padj < 0.05). Specifically, cluster IIwas predominantly carbohydrates (padj = 0.00366), clusterIX was all organic acids (padj = 0.00245), and cluster XIIwas primarily amino acids (padj = 0.000251).time of day. ANOVA analysis, taking into account intra-levels in response to drought (ANOVA, padj-value < 0.05)Group Metabolite Log2 (fold-change)Amino acid Aspartic acid −0.68L-Isoleucine 3.32L-Threonine −0.39NI (Amino acid; 2) −0.26Carbohydrate Fructose (2) −0.16Glycerol −0.58Melibiose 0.3NI (5C sugar; 2) −1.18NI (6C sugar; 1) −0.16Raffinose 1.13Sucrose 0.28Organic acid Benzoic acid 1.13Citric acid 1.07Fumaric acid −1.3Glycolic acid −0.39Malic acid −0.12Malonic acid −1.58Quinic acid −0.37Shikimic acid −0.55Succinic acid −0.97Threonic acid −0.25Threonic acid 1,4-lactone −0.57Phenolic Catechol 0.45Quercitin −0.24Salicin 0.83Salicyl_alcohol 1.56Sugar alcohol Galactinol 0.52Myo-inositol 0.15Not identified NI (2) −0.14NI (3) −0.43NI (4) 0.36NI (5) 0.69NI (7) 0.15Ni (9) 0.24NI (10) 1.36NI (11) 0.53NI (15) 0.97NI (18) 0.39NI (20) 0.55NI (25) 0.53Hamanishi et al. BMC Genomics  (2015) 16:329 Page 5 of 16A three-way factorial (ANOVA) identified metabolitesthat had significantly different abundance in response todrought treatment (T main effect), genotype (G maineffect), time of day (D main effect), as well as any inter-action between the three experimental factors (Table 2,Additional file 1: Table S3). Similar to the HCA resultsamong metabolites; significant variation in the metabolicprofiles was attributable to genotype. A large proportionof metabolites had differential abundance among geno-types (n = 79; p < 0.05; Table 3). Of the 79 metabolites withsignificant variation among genotypes, no interaction withany factor was found for 38 metabolites.ANOVA analysis also revealed a small subset (n = 11)of metabolites that varied significantly in abundance inresponse to time of day (Figure 2). However, a largernumber of metabolites (n = 15) had abundance that var-ied significantly in response to water-deficit treatment ina time-of-day dependent fashion (TxD interaction;Table 2; Additional file 4: Figure S3). Notably, prolinehad a significantly higher abundance at PD relative toMD (Figure 2A). Conversely, sucrose had higher abun-dance at the MD time point (Figure 2B). In plants, su-crose concentrations fluctuate diurnally, with increasedabundance during light conditions [37-40].A populus balsamifera drought metabolome was identifiableWater withdrawal induced significant changes in me-tabolite abundance. Four to 10 biological replicateswere analysed per treatment, per genotype and perTable 2 Number of metabolites with significant maineffects or interactions (n = 87 metabolites)Number ofmetabolitesPercent (%) ofmetabolitesGenotype (G) 79 91.95%Treatment (T) 40 45.98%Time of day (D) 11 12.64%Genotype:treatment (GxT) 41 47.13%Genotype:time of day (GxD) 6 5.75%Treatment:time of day (TxD) 15 17.24%Genotype:treatment:time of day (GxTxD) 0 0.00%padj-value cutoff = 0.05 (Benjamini-Hochberg).Table 3 Metabolites with significantly different abundancereplicate variation (residual error, Additional file 1:Table S3), identified 40 metabolites with differentHamanishi et al. BMC Genomics  (2015) 16:329 Page 6 of levels in response to drought. Twenty-onemetabolites increased in abundance and 19 decreased inabundance (p < 0.05; Table 3; Figure 3A). No generalclass of metabolites responded to drought. For example,the amino acid (AA) class had variable response todrought. The contribution of amino acids in Populusclones is thought to be small relative to the effect ofcarbohydrates and other osmolytes [41]. However, iso-leucine had the largest fold increase in abundance inresponse to drought of any metabolite assessed, and wasthe only branched chain amino acid (BCAA) to be ana-lysed, whereas aspartic acid and threonine decreased inabundance in the drought-treated samples. Increasedaccumulation of BCAAs has been observed in otherorganisms including Arabidopsis [42] and various wheatcultivars [43]. Although increased accumulation of BCAAsMD PDMD PD5. 2 Time of Day main effect observed for (A) proline and (B)sucrose between mid-day (MD; light grey) and pre-dawn (PD;dark grey).has frequently been observed in response to abiotic stress,little is known about their role in stress tolerance; however,accumulated BCAAs may serve as a substrate for thesynthesis of other stress-induced proteins and may act assignalling molecules in response to drought stress [44].Two organic acids, representative of TCA cycle inter-mediates, succinic and malic acid, had a general declinein abundance; whereas raffinose and galactinol weresome of the most highly accumulated metabolites in re-sponse to water-deficit conditions (Table 3). Although ageneral decline was observed in abundance of malic acid,patterns of accumulation in response to drought inPopulus are often varied; both increased and decreasedlevels of accumulation in response to drought have beenobserved [41,45]. Malic acid is a very abundant organicacid in plants, and its role is likely not restricted to thecitric acid cycle [46]. Sugars have previously been shownto increase in abundance in response to water-stress,having an important role in the osmotic adjustment[47,48]. Raffinose and galactinol have been hypothesisedto be osmoprotectants in drought-stress conditions, andhave frequently been implicated in the drought responsein plants [17,20].Metabolites commonly associated with drought or stressin plants constituted the core drought metabolome (i.e., Tmain effect). For example, substantial accumulation of raf-finose and galactinol, important stress related carbohy-drates, occurred in drought treated trees. Notably, of the40 metabolites that were significant for T main effect, only15 did not show any significant interactions (i.e., TxG orTxD; Figure 3B). Carbohydrates, a sugar alcohol, andsome unknown metabolites had increased abundance inwater-deficit conditions, whereas decreased abundancewas exhibited by a variety of metabolites representative ofdifferent metabolite classes (Figure 3C). As indicated bythe large proportion of metabolites significant for TxGor TxD interactions, the accumulation of metaboliteswas not simply due to the imposition water-deficitstress, rather, metabolite accumulation was a complexresponse shaped by genotype and time of day. Thevariation in metabolite accumulation across genotypesand at different time-points could be exploited to fur-ther investigate the unique responses of P. balsamiferagenotypes.The drought metabolome varied among P. balsamiferagenotypesWhile a large proportion of metabolites had significantresponse to water-deficit treatment, many of these variedin a genotype- (G) or a time-of-day- (D) dependentmanner (Figure 3B, Additional files 3 and 4: Figure S2and Figure S3). The abundance of 41 metabolites wassignificantly impacted by TxG interaction (Table 2;Additional file 5: Figure S4). Certain metabolites hadHamanishi et al. BMC Genomics  (2015) 16:329 Page 7 of 16Aopposite patterns of accumulation in response to drought(i.e., higher abundance in one genotype and lower abun-dance in another genotype). Of note, glucose had elevatedabundance levels in AP-947 and AP-1006, but decreasedabundance levels in the remaining four genotypes in re-sponse to water-deficit conditions. Similarly, galactinolwas significant for a G x T interaction (p = 0.0259); thehighest level of galactinol accumulation was observedin drought-treated samples of genotype AP-947 andAP-2278. Other metabolites that had a significant TxGinteraction demonstrated consistent directionality ofresponse to water-deficit stress among the six geno-types. For example, glycolic and threonic acids, twometabolites belonging to cluster IX (Additional file 3:Figure S2), decreased in abundance in response toB CFigure 3 Metabolite accumulation levels for treatment main effect and tresignificant for treatment main effect across all genotypes at two different tdemonstrating the number of metabolites that are significant for treatmenmetabolite abundance for metabolites that are significant for treatment mawater-deficit conditions in all genotypes, with substantialreductions observed in genotype AP-1005 and AP-2278.Moreover, half of the metabolites that had significant dif-ferences in abundance between treatments (T main effect)also varied in response to genotype (n = 20; Figure 3B)confirming the importance of genotype in defining thedrought response observed among samples.Ten drought-responsive metabolites had significantdifferences in abundance for a TxD interaction, indicativeof the variation in metabolite level observed between pre-dawn and mid day. Raffinose abundance was significantfor a TxD interaction, having ~2-fold increase in accumu-lation in response to water-deficit at MD (p = 0.0122), butno significant change in abundance at PD (Additionalfile 4: Figure S3).atment x genotype interaction. (A) Hierarchal clustering of metabolitesime-points [pre-dawn (PD) and mid-day (MD)]. (B) Venn diagramt main effect or a 2-way interaction. (C) Mean log2 (fold-change) ofin effect only.A notable feature of the P. balsamifera drought metab-olome was the magnitude of variation observed betweensamples. On average, peak signal intensity (non-trans-formed data) varied ~3000-fold between minimum andmaximum peak intensity for any given metabolite. Simi-larly, the magnitude of variation in metabolite accumula-tion between water-deficit and well watered samplesvaried considerably. Among the metabolites whose accu-mulation had a significant T main effect, the fold-changevariation ranged from ~3 fold decrease in malonic acid ac-cumulation to ~10 fold increase in isoleucine accumula-tion. Overall variation in the drought metabolome wasexamined by Pearson correlation comparison of the log2(fold-change) of the water-deficit metabolome of thesix P. balsamifera genotypes. This analysis revealedwhich genotypes had metabolome responses that weremore equivalent to others (Figure 4; Additional file 1:Table S4). Genotypes AP-1005 and AP-2278 had themost similar drought metabolomes (r = 0.845; p < 0.05),whereas genotypes AP-2300 (r < 0.550) and AP-2298(r < 0.606) were most divergent when compared to allother genotypes (Figure 4; Additional file 1: Table S4).The magnitude of drought-induced changes in metaboliteabundance among the six P. balsamifera genotypes had ahigh degree of variation (Additional file 6: Figure S5A). Thelargest absolute magnitude change in drought responsivemetabolites occurred in AP-1005 (mean = 0.361, standarddeviation = 0.340) and AP-2278 (mean = 0.327; standarddeviation = 0.224), whereas the smallest magnitude changewas observed in genotype AP-1006 (mean = 0.184; stand-ard deviation = 0.223).There were correlations between drought-responsivemetabolites and specific components of transcriptomeremodellingTo assess relationships between drought-responsive me-tabolites and transcripts, the metabolomes and transcrip-tomes of P. balsamifera were compared. These analysesmade use of previously-reported drought-responsive tran-scriptome data for P. balsamifera [14]. Quantitatively,there was a high level of congruence between the metabo-lome and the transcriptome, where larger magnitudechanges in the transcriptome corresponded with largermagnitude changes in the metabolome, with the not-able exception for genotype AP-1006 (Additional file 6:Figure S5). Specifically, genotype AP-1006 and AP-2278had significantly larger magnitude change in the droughttranscriptome relative to all other genotypes (Bonferroni’sColor KeyAP-1006f P.nifiHamanishi et al. BMC Genomics  (2015) 16:329 Page 8 of 160.4 0.6 0.8 1AP-2298AP-2300Pearson correlation coefficient, Figure 4 Variation in the drought metabolome among six genotypes oheatmap. Differential abundance [log2 (fold-change)] for metabolites sigPCC value was calculated for each pair-wise comparison among genotypesrepresented on both the x- and y-axis in the same order.AP-2298AP-2300AP-1006AP-947AP-1005AP-2278AP-947AP-1005AP-2278balsamifera represented by a Pearson correlation coefficient (PCC)cant for treatment main effect (ANOVA, p < 0.05) are represented. The, and is represented by the colour in the given cell. All genotypes areHamanishi et al. BMC Genomics  (2015) 16:329 Page 9 of 16p < 0.001; Additional file 6: Figure S5B); whereas, the abso-lute magnitude change observed in the metabolome forAP-1006 and AP-2278 was among the smallest and lar-gest, respectively. This suggests that coordination of thetranscriptome and metabolome is variable among geno-types, and that the overall magnitude change in metaboliteabundance does not necessarily reflect the magnitudeof transcriptome variation resulting from water-deficittreatment.A correlation matrix of all pair-wise comparisonsamong drought responsive metabolites and transcriptsrevealed 747 transcripts that were significantly corre-lated with at least one metabolite (Pearson correlationcoefficient, |r| > 0.60, p < 0.05), based on the similarity ofabundance profiles across all samples (Additional file 7:Figure S6). Correlation patterns between metabolitesand transcripts were similar among the organic acidswith the exception of citric, benzoic and shikimic acid. Asignificant proportion of organic acids share similarpatterns of abundance across samples; however, citric,benzoic and shikimic acid do not. Similarly, three aminoacids (aspartic acid, threonine and an unidentified aminoacid) had similar correlation patterns; whereas, the cor-relation pattern for isoleucine was distinct. Unlike theother three amino acids, isoleucine increased signifi-cantly in abundance in response to water-deficit with amore pronounced increase at the mid-day time point.These results suggest that the regulatory control of themetabolites with similar patterns of expression may beshared; whereas, the metabolites with distinct correlationpatterns are likely influenced by distinct molecularmechanisms.Among the transcripts significantly correlated with atleast one metabolite, enrichment for GO terms amongtranscripts was determined. For transcripts with in-creased transcript abundance in response to droughtand correlated with at least one metabolite (n = 404), foursignificant enriched GO biological process terms wereidentified: ‘proline metabolic process’ (GO:0006560), ‘ar-ginine metabolic process’ (GO:0006525), ‘galactose meta-bolic process’ (GO: 0006012) and ‘serine family aminoacid metabolic process’ (GO:0009069; Additional file 1:Table S6). A total of 13 significant GO terms were identi-fied. Among transcripts that had decreased transcriptabundance in response to drought and were correlatedwith at least one metabolite (n = 343), 15 significantlyenriched GO terms were identified. For GO terms as-sociated with biological process, ‘serine family aminoacid metabolic process’ (GO:0009069), ‘tyrosine meta-bolic process’ (GO:0006570) and ‘aromatic amino acidfamily metabolic process’ (GO:0009072) were significantlyenriched.Functional annotation of the correlated transcripts andmetabolites revealed pathways that were perturbed bywater withdrawal (Additional file 7: Figure S6). A func-tional class related to starch and sucrose metabolism(pop00500) was overrepresented among the transcriptsthat are correlated with two identified 5C sugars andglucose (Additional file 7: Figure S6). Photosynthesis-related categories (pop00195 and pop00196) were highlyassociated with malic acid, raffinose and galactinol(Additional file 7: Figure S6).In spinach, raffinose accumulation reduced electronand cyclic photophosphorylation in photosynthesis [49],and it has been hypothesised that raffinose and otherRFOs play an important role in the protection of cellularmetabolism, especially photosynthesis in chloroplasts inArabidopsis [17]. Evidence herein suggests there may bea functional relationship in P. balsamifera between raf-finose accumulation and transcripts associated withphotosynthesis. An association between photosyntheticmetabolic processes and RFO accumulation may high-light unique relationships that can be garnered fromtranscriptome-metabolome relationships in Populus.Energy metabolism and secondary metabolite biosynthesisvaried in a genotypic-dependent manner in response todroughtGalactinol accumulation varied in response to water-deficit stress in genotype AP-1006 [log2 (fold-change) = −0.4526]; whereas galactinol accumulatedconsistently in the other genotypes. Raffinose accumula-tion was significant in water-deficit-treated plants, withthe exception of trees of the genotype AP-2300. Therewas drought-responsive variation in transcript accumula-tion of genes hypothesised to be involved in the galactosemetabolism pathway. All genotypes showed increasedabundance of transcripts corresponding to galactinolsynthase (EC:, raffinose synthase (EC: stachyose synthase (EC:; Additional file 8:Figure S7). Galactinol synthase transcript accumulationvaried in magnitude in response to water-deficit condi-tions among the six genotypes, with the largest increase intranscript accumulation observed in genotypes AP-2278and AP-1006.Elevated levels of RFOs in Arabidopsis plants increaseddrought tolerance, highlighting the importance of theseoligosaccharides in the response to osmotic-stress [20].Increased accumulation of raffinose has been observedin desiccation tolerant seeds [50], chloroplasts of frost-hardy Brassica oleracea leaves [49], and in Populus leavesexposed to osmotic stress [16,25]. Increased transcriptabundance of galactinol synthase and raffinose synthasehas been observed in response to drought in Arabidopsis[20,51] and Populus [14,25].Mounting evidence suggests that the role of raffinoseand other RFOs is consistent across species; however,the magnitude of change is variable, as was observedHamanishi et al. BMC Genomics  (2015) 16:329 Page 10 of 16among the six P. balsamifera genotypes reported herein.Similarly, in four Populus hybrids, variable raffinose andgalactinol content was shown under drought [16]. Thissuggests the existence of genotypic specific metaboliteprofiles related to these oligosaccharides, and that thelevel of accumulation may influence the overall droughtresponse. Moreover, the data suggest that AP-1006 maynot accumulate elevated levels of galactinol in responseto drought; faster metabolism turnover or flux throughthis pathway may be of lower importance.Unique relationships were also observed in the citratecycle (TCA) pathway (KEGG, pop00020). TCA cycle in-termediates. For example, succinic and malic acid, showsignificant variation with respect T and TxG (Additionalfile 1: Table S3). The metabolic rate of the TCA cycle isknown to be influenced by drought [52]. The magnitudechange between well watered and water-deficit treatedsamples for transcripts associated with the TCA cyclevaried among genotypes. Citrate synthase (EC: increased transcript accumulation in water-deficit-treated samples of AP-947, AP-1006, AP-2278 and AP-2300; however, decreased transcript accumulation wasobserved in the other genotypes (Additional file 9:Figure S8). Similarly, malate dehydrogenase (EC: <1 log2 (fold-change) in response to drought in AP-1006 and AP-2298, whereas >1 log2 (fold-change) increasewas observed in AP-947 and AP-2278. In Arabidopsis,malate dehydrogenase demonstrated increased transcriptaccumulation in response to drought, cold or high-salinitystress [19]; however, the variation in the genotypic re-sponse in P. balsamifera highlights the complexity in thisresponse.Similar to other genotypes, variations among genes in-volved in the TCA metabolic pathway were observed ingenotype AP-1006 (Figure 5A). Pair-wise comparisonswithin the TCA cycle for select transcripts and metabo-lites found weak relationships among transcripts, andmalic and citric acid accumulation profiles for AP-1006(Figure 5B); however, succinic acid and malate de-hydrogenase (EC: were significantly negativelycorrelated (r = −0.67, P = 0.0204) in genotype AP-1006(Figure 5B). Pathway analysis highlights the influenceof genotype on the drought-induced modifications tothe TCA cycle in AP-1006, and, more broadly in P.balsamifera.Comparative pathway analysis among genotypes hasproved useful in Populus. In two different genotypes ofPopulus with varying salt-tolerance, pathway analysisrevealed different mechanisms of tolerance between thetwo genotypes. Janz et al. [25] found that the salt-tolerantPopulus eupharatica demonstrated moderate transcrip-tome changes in response to stress when compared to asalt-sensitive Populus hybrid. However, stress tolerancein P. eupharatica was not dependent on transcriptomemodification under conditions of stress; instead, itwas linked to greater energy requirements for cellularmetabolism [25]. In P. balsamifera there are varyingdegrees of transcriptional remodelling in response todrought among genotypes; however, further analysis is re-quired to understand the subtleties in these differences.Network analysis illuminated the nature ofgenotype-specific responses to droughtTo identify genotype-specific transcriptome alterations,a network was created including all genes that weredeemed significantly differentially expressed in a T-maineffect manner for each genotype using WGCNA [31].Weighted Pearson correlation matrices were calculatedand used to determine topological overlap (TO) amonggenes. The TO calculated in WGCNA measured connect-ivity of a gene within a network relative to its neighbours.HCA based on the TO scores for all genes in the droughttranscriptome grouped genes with equivalent transcriptabundance profiles across all samples.Overall, 10 network modules with equivalent tran-script abundance patterns were identified. Many networkmodules were similar across genotypes. For example, asignificant proportion of the network modules fromAP-1006 were preserved in the other five genotypes(Table 4). AP-1006 was chosen as a reference becausethe transcriptome of AP-1006 had the highest magni-tude change in response to drought relative to theother five genotypes. Not surprisingly, all of the mod-ules were highly correlated with treatment; whereasonly three were significantly correlated with time ofday (M2_1006, M5_1006 and M8_1006; Table 5). Not-ably, M3_1006 (black) was shared between AP-1006and AP-2278 with 62% overlap with respect to genemembership (Table 4). Among those transcripts be-longing to M3_1006, there was an overrepresentationof transcripts involved in ‘intracellular signalling cas-cade (GO: 0007242)’. M5_1006 (brown) demonstrateda high degree of overlap among genotypes, with the ex-ception of AP-2298. Functional characterization ofM5_1006 revealed that the module was made up ofgenes that are often associated with drought responsesin plants, and included an overrepresentation of GO termssuch as: ‘response to abiotic stimulus (GO:0009628)’, ‘cel-lular catabolic process (GO: 0044248)’ and ‘response towater deprivation (GO: 0009414)’. The high degree ofoverlap between modules identified for AP-1006 and theother genotypes validated the presence of a highly con-served drought transcriptome in P. balsamifera.Although there was a high degree of network modulepreservation among genotypes, organisation withinmodules varied among genotypes. When visualisingthe top (n = 1000) network connections of each geno-type, and labelling the nodes according to their moduleHamanishi et al. BMC Genomics  (2015) 16:329 Page 11 of 16Malic AcidCitric Acid Succinic Acid. within the drought transcriptome, twogeneral observations could be made (Figure 6). First,transcript connectivity varied among genotypes. In certaingenotypes, there was a higher degree of topological over-lap between individual genes (nodes), as indicated by thecolour of the edges. Nodes connected with a higher TOare indicated with a red/purple colour, whereas lower TOis indicated with a blue colour. For example, modulesfound in AP-1005 demonstrate higher TO indicatingstronger connectivity among nodes and modules asAcetyl CoSuccinyl-CoASuccinateFumarateMalatePyruvateOxaloa4.−2 −1 0 1 2Fold Ratio [Log2(Fold-change)]Color K eyEC:1EC:1.1EC:4EC:2EC:1.1 EC:6EC:4BFigure 5 Pathway analysis related to the citric cycle (TCA cycle). (A) Correlpathway pop00020 ‘Citrate cycle (TCA cycle)’ for genotype AP-1006. Colorsand blue represents negative correlation values. (B) Map displays selectedscript or metabolite abundance between water-deficit and well watered trein water-deficit-treated samples and blue indicates lower abundance in wa1.1.1.37, malate dehydrogenase; EC:, isocitrate dehydrogenase (NADEC:, fumarate hydratase, EC:, aconitate hydratase, EC:, su−1 −0.5 0 0.5 1Pearson Correlation CoefficientColor Keycompared to genotype AP-2278. Interconnectednessamong genes in AP-1005 are more tightly correlated ascompared to other genotypes (Figure 6). Second, theimportance of any given module varied among genotypes.For example, the nodes of top network connections ingenotype AP-1005 were from module M4_1006 andM6_1006, whereas genotype AP-947 had nodes thatbelonged to many other modules. More specifically, genesthat played a more central “hub” role in the drought tran-scriptome networks varied among genotypes.CitrateAcis-AconitateIsocitrate2-ketoglutaratecetatecolysis IV1.1.1.412. balsamifera, genotype AP-1006ation among selected transcripts and metabolites from the KEGGrepresent Pearson correlation value. Red indicates positive correlationsteps from citrate cycle pathway. Colours indicate fold-change in tran-ated samples for genotype AP-1006; red indicates higher abundanceter-deficit-treated samples. Enzymes are given as EC numbers. EC+); EC:, succinate dehydrogenase; EC:, citrate synthase;ccinate-CoA ligase, beta subunit.Table 4 Module membership in the drought transcriptome neamong the other genotypesColour Module identifiedin AP-1006Number of modulemembers (genes)APyellow M1_1006 192 -blue M2_1006 355 49black M3_1006 93 -turquoise M4_1006 399 40brown M5_1006 199 48green M6_1006 188 -pink M7_1006 72 -74Hamanishi et al. BMC Genomics  (2015) 16:329 Page 12 of 16AP-1006 had a genotype-specific transcriptome responseto droughtDue to the uniqueness of the transcriptome of AP-1006,genes central to the network modules in this genotypewere interrogated. The largest magnitude of change wasobserved in the drought transcriptome for AP-1006.Transcript abundance of hub genes in AP-1006 revealedsamples clustered according to treatment; however, tran-script abundance profiles were more similar betweenwell watered samples, regardless of time of day. Theabsolute magnitude change in abundance of transcriptscentral to the network modules in AP-1006 was signifi-cantly higher than the magnitude change of thetranscripts in the other genotypes [absolute log2 (fold-change) AP-1006 = 2.36]. Many hub transcripts hadsignificant changes in transcript abundance in response todrought in AP-1006. There are 195 hub transcripts (TONetwork Ratio > 0.5) that have decreased abundance inresponse to drought in AP-1006; whereas there are 104hub transcripts that had increased abundance in responseto drought.Enrichment of GO terms within the set of centralnetwork hub transcripts from genotype AP-1006 re-vealed the components of the genotype-specific droughtred M8_1006 111transcriptome. For example, transcripts implicated in theTable 5 Module-treatment or -time of day relationshipsof the P. balsamifera (AP-1006) drought transcriptomeModule AP-1006 Treatment (T) Time of day (D)M1_1006 −0.833* −0.484M2_1006 −0.662* 0.709*M3_1006 −0.77* 0.432M4_1006 −0.952* 0.164M5_1006 0.759* −0.638*M6_1006 0.983* −0.038M7_1006 0.928* 0.342M8_1006 0.699* 0.708*Columns 2 and 3 represent the correlation between the mean expression ofthe module and the experimental factor (T or D). *p value < 0.05.response to stress and stimuli were enriched. Of the hubtranscripts with significant declines in abundance, genesimplicated in carbohydrate metabolism were enriched, in-cluding those with GO terms for: sucrose (GO:0005985),starch (GO: 0005982) and disaccharide (GO: 0005984)metabolic processes (Additional file 10: Figure S9A). Con-versely, core hub transcripts with increased accumulationin response to drought in AP1006 were enriched forbiological processes, including response to stimulus(GO:0050896) and stress (GO: 0006950) as well astransport (GO:0006910) and regulation of cellular pro-cesses (GO:0050794; Additional file 10: Figure S9B).However, it should be noted that there was a large pro-portion of transcripts that had unknown function. Thetranscripts that played a central role in the network or-ganisation of the drought transcriptome in AP-1006were likely important regulators of the drought response,and the analysis of transcript co-expression relationshipsmay help with functional annotation; albeit, not with im-mediate interpretation.There were strong correlates between specific transcript-metabolite pairs in response to drought in AP-1006Strong correlations between transcript and metabolitetwork of AP-1006 and preservation of drought modules-947 AP-1005 AP-2278 AP-2298 AP-230068 121 - -387 242 - -- 58 - -309 140 44 244283 183 - 38140 149 97 51201 81 - -128 53 - -abundance in response to drought in AP-1006 were ob-served in metabolic pathways, including: ‘plant hormonesignal transduction’, ‘arginine and proline metabolism’,and ‘glycolysis/gluconeogenesis’ (Additional file 11:Figure S10). As previously noted, one of the largestmagnitude change in transcript abundance was observedin AP-1006 (Figure S5). Transcripts, including thoseencoding genes homologues to Arabidopsis thalianaRAC-like 2 protein (ARAC2) and IRREGULAR XYLEM 9(IRX9) had significantly larger fold-change decrease intranscript abundance in response to water-deficit condi-tions as compared to other genotypes. Conversely, severaltranscripts annotated as universal stress proteins, or thoseinvolved in hormone signalling had significantly highertranscript abundance in AP-1006 in response to water-deficit stress. Correlation network analysis revealed coreHamanishi et al. BMC Genomics  (2015) 16:329 Page 13 of 16transcripts that might have played a role in the under-lying mechanisms regulating metabolite accumulationin AP-1006. Transcripts most strongly correlated withmetabolite levels were identified. Although no particularclass of metabolites or transcripts appeared specific toAP-1006, a large number of transcripts highly correlatedwith succinic acid, raffinose and galactinol accumulation(Additional file 1: Table S7). For example, strong positivecorrelations were observed between raffinose, galactinoland a photosystem II reaction center PsbP family protein(r = 0.871 and 0.835, respectively; Additional file 1:Table S7). Strong correlations between drought respon-sive metabolites and transcripts reveal pathways thatmay be of importance in the drought tolerance mechanismsin a genotype.Figure 6 Transcript correlation networks obtained from WGCNA for (A) AP2300. The top 1000 interactions for each genotype are represented. Nodesto other transcripts. Each node is colored according to the modules defineConclusionThe metabolomics response to drought in Populusbalsamifera in these experiments was complex, andthe variation within the metabolome was highlightedby variation among genotypes and between time-of-dayresponses. Although common drought-responsive metab-olites could be identified across all six P. balsamiferagenotypes, a significant proportion of metabolites variedin a genotype or time-of-day dependent manner. Thecomplexity of the genotype-metabolite relationship wasnotable, and likely attributable to the function of manygenes, the environment and their interaction. Integratingtranscriptome and metabolome data identified significantmetabolite-gene correlation, whereby biologically mean-ingful correlations were derived. Metabolite-transcript-947, (B) AP-1005, (C) AP-1006, (D) AP-2278, (E) AP-2298 and (F) AP-in the graphs represent individual transcripts that connect via edgesd in Table 4.1006. Figures generated using AgriGO (http://bioinfo.cau.edu.cn/Hamanishi et al. BMC Genomics  (2015) 16:329 Page 14 of 16relationships from the same and different pathways wereidentified, and may be useful for future elucidation of im-portant drought response mechanisms. Integration of thetranscriptome and metabolome data at individual pathwaylevels revealed variation in metabolite flux and transcriptaccumulation among genotypes in energy and galactosemetabolism.The impacts of environmental stress on forest healthand productivity are becoming of increasing concern.The results presented herein demonstrate that futureexperiments aimed at understanding the complexities ofthe responses of forest trees to environmental stimulimust take into consideration the intraspecific variationin these responses. Although common drought responsesamong genotypes of P. balsamifera could be identified,significant intraspecific variation was observed. The intra-specific variation in the molecular strategies that underpinthe responses to drought among genotypes may have animportant role in the maintenance of forest health andproductivity, particularly amidst future challenges imposedby reduced forest integrity and fluctuating environmentalconditions.Availability of supporting dataThe microarray data set supporting the results of thisarticle is available in the Gene Expression Omnibus(GEO) database of the National Center for BiotechnologyInformation of the USA as series GSE21171 (http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE21171).The metabolite data set supporting the results of this art-icle is included within the article, and can be found in theSupplemental Information (Additional file 1: Table S1).Additional filesAdditional file 1: Table S1. Original data matrix for all metabolites(n = 87). Table S2. Summary statistics for all metabolites (n = 87). Meanpeak intensity values represented for all samples, wet samples and drysamples ± standard deviation. Relative abundance between water-deficitand well watered samples represented as log2 (fold-ratio); positive valuesindicate increased abundance in water-deficit conditions, whereas negativevalues indicate decreased abundance in water-deficit conditions relative tocontrol conditions. p values for treatment (T) main effect as calculated perthe factorial ANOVA; metabolites significant for treatment main effectdenoted with an * (FDR < 0.05). Metabolite classes are as follows: AA =AminoAcid; C = Carbohydrate; P = Phenolic, SA = Sugar Alcohol, and NI = NotIdentified. Table S3. ANOVA results: metabolite abundance. Significance:***P < 0.001; **P < 0.01; *P < 0.05; . P < 0.1 Table S4. Pair-wise comparisonsamong genotypes for absolute magnitude log2 (fold-change) variation forthe drouglht metabolome (n = 40). Bolded asterisks indicate significantdifferences according to Bonferroni’s test (P < 0.05). Table S5. Pair-wisecomparisons among genotypes for absolute magnitude log2 (fold-change)variation for the drought transcriptome (n = 763). Bolded asterisks indicatesignificant differences according to Bonferroni’s test (P < 0.05). Table S6. GOterm enrichment among transcripts that are significantly correlated with atleast one metabolite. (A) For transcripts with increased abundance inwater-deficit treated samples and (B) for transcripts with decreasedabundance in water-deficit treated samples. Table S7. Pair-wise M:TSpearman correlation values for genotype AP-1006.Additional file 2: Figure S1. Dendrogram obtained after HCA of themetabolic profiles of six P. balsamifera genotypes. Metabolite profilescollected for six genotypes under well watered and water-deficitconditions at mid-day and pre-dawn time point. Rows representspecific metabolites (n = 87). Columns represent mean intensity of allreplicates for each genotype, treatment and time of day sample.Plotted values are the mean of n = 4–10 replicates for each sample.Metabolite classes: AA = Amino Acid; C = Carbohydrate; P = Phenolic,SA = Sugar Alcohol. NI = Not Identified.Additional file 3: Figure S2. HCA reveals 13 significant clusters (P < 0.05).Significant clusters are numbered (I through XIII) for identification.Hierarchical clustering was done using pvclust (Suzuki and Shimodaira [29]),with a correlation distance measure and a complete agglomerativeclustering method.Additional file 4: Figure S3. Metabolites that are significant for atreatment (T):time-of-day (D) interaction (P <0.05, n = 15).Additional file 5: Figure S4. Metabolites that are significant for atreatment (T):genotype (G) interaction (P <0.05, n = 41).Additional file 6: Figure S5. Box-plot illustrating the interplay ofgenotype and treatment in shaping the drought metabolome andtranscriptome of six P. balsamifera genotypes. The average absolutelog2 (fold-change) between well watered and water-deficit-treatedsamples for all (A) metabolites (n = 40; P < 0.05) and (B) transcripts(n = 1848; p < 0.05) with significant variation in their abundancebetween treatment conditions at the mid-day (MD) timepoint.Additional file 7: Figure S6. Heatmap of drought responsive transcriptand metabolite correlations. Of all the drought responsive transcripts, 747unique transcripts were correlated with at least one metabolite (|r| > 0.6;p < 0.05). The rows in the heatmap represent metabolites, and thecolumns represent transcripts. The columns are clustered based on theirexpression across samples, and the metabolites are grouped according tofunctional categories. Pearson correlation coefficient (r) are representedfor each pair-wise metabolite-transcript comparison, and were calculatedusing R.Additional file 8: Figure S7. Pathway analysis related to the galactosemetabolism. (A) Pathway map displays selected steps from galactosemetabolism pathway. Colours indicate fold-change in transcript ormetabolite abundance between water-deficit and well wateredtreated samples for all six genotypes; red indicates higher abundancein water-deficit-treated samples and blue indicates lower abundancein water-deficit-treated samples. Enzymes are given as EC numbers. EC2.4.1.123, galactinol synthase; EC:, raffinose synthase; EC:,stachyose synthase. (B) Heatmap representing Spearman correlation valuesamong transcripts related to galactose metabolism and raffinose orgalactinol.Additional file 9: Figure S8. Analysis of the citrate cycle (TCA;pop00020) pathway in genotype AP-947, AP-1005, AP-2278, AP-2298 andAP-2300. (a) Correlation among select transcripts and metabolites fromthe KEGG pathway pop00020 ‘Citrate cycle (TCA cycle)’ for genotypeAP-1006. Colors represent Pearson correlation value. Red indicates positivecorrelation and blue represents negative correlation values. (b) Map displaysselected steps from citrate cycle pathway. Colours indicate fold-change intranscript or metabolite abundance between water-deficit and well wateredtreated samples; red indicates higher abundance in water-deficittreated samples and blue indicates lower abundance in water-deficittreated samples. Enzymes are given as EC numbers. EC, malatedehydrogenase; EC:, isocitrate dehydrogenase (NAD+);EC:, succinate dehydrogenase; EC:, citrate synthase;EC:, fumarate hydratase, EC:, aconitate hydratase, EC:, succinate-CoA ligase, beta subunit. Pearson correlation andpathway maps for AP-1006 can be found in Figure 5.Additional file 10: Figure S9. Overrepresentation of GO termsassociated with transcripts that have (A) decreased transcriptabundance in AP-1006 and (B) increased transcript abundance in AP-agriGO). Significant overrepresentation is represented by darkershaded boxes (p < 0.05).Hamanishi et al. BMC Genomics  (2015) 16:329 Page 15 of 16Additional file 11: Figure S10. Heatmap of representative functionalclasses (transcripts) from the correlation data. The averaged Spearmancorrelation value is represented for significant functional class: metabolitecomparisons (coloured squares). Red indicates positive correlation,whereas blue indicates negative correlation. The number of transcriptsrepresented in each functional group is represented in brackets.AbbreviationsARAC2: Arabidopsis thaliana RAC-like 2 protein; BCAA: Branched chain aminoacids; D: Time-of-day; DW: Dry weight; FDR: False discovery rate; FW: Freshweight; G: Genotype; GC/MS: Gas chromatography/mass-spectrometry;GO: Gene ontology; HCA: Hierarchal cluster analysis; IRX9: IRREGULAR XYLEM9; MD: Mid day; MSTFA: N-methyl-N-trimethylsilytriflouroacetamide; PD: Predawn; Pro: Proline; RFO: Raffinose family oligosaccharide; ROS: Reactiveoxygen species; RWC: Relative water content; T: Treatment; TCA: Citrate cycle;TO: Topological overlap; Val: Valine; WGCNA: Weighted co-expressionnetwork analysis.Competing interestsThe authors declare that they have no competing interests.Authors’ contributionsETH conducted this work as part of her PhD. ETH generated physiologicaland transcriptome data, did the bioinformatic analysis and drafted themanuscript. GLHB, RD and SDM provided GC/MS data. MMC concievedexperiments and drafted the manuscript. All authors read and approved thefinal manuscript.AcknowledgementsWe are grateful to Bruce Hall and Andrew Petrie for greenhouse assistance,and Joan Ouellette for technical assistance. Research infrastructure andtechnical support was provided by the Centre for Analysis of GenomeEvolution & Function.Funding statementSDM is a Canada Research Chair. This work was generously supported byfunding from NSERC to MMC, ALP and SDM, as well as the CanadaFoundation for Innovation (CFI), and the University of Toronto to MMC.Author details1Faculty of Forestry, University of Toronto, 33 Willcocks St., Toronto, ON M5S3B2, Canada. 2Centre for the Analysis of Genome Evolution and Function,University of Toronto, 25 Willcocks St., Toronto, ON M5S 3B2, Canada.3Department of Wood Science, University of British Columbia, 4030-2424Main Mall, Vancouver, BC V6T 1Z4, Canada. 4Department of Cell & SystemsBiology, University of Toronto, 25 Willcocks St., Toronto, ON M5S 3B2,Canada. 5Department of Biological Sciences, University of TorontoScarborough, Toronto, ON M1C 1A4, Canada. 6Current address: Départementdes sciences agronomiques et écologiques, Université de Picardie JulesVerne, Amiens 80039 cedex, France.Received: 27 January 2015 Accepted: 13 April 2015References1. Allen CD, Macalady AK, Chenchouni H, Bachelet D, McDowell N, VennetierM, et al. A global overview of drought and heat-induced tree mortalityreveals emerging climate change risks for forests. For Ecol Manage.2010;259:660–84.2. Anderegg WRL, Berry J, Smith DD, Sperry JS, Anderegg LDL, Field CB. Theroles of hydraulic and carbon stress in a widespread climate-induced forestdie-off. Proc Natl Acad Sci U S A. 2012;109:233–7.3. Eckenwalder JE. Systematics and evolution of Populus. In: Biology of Populusand its implications for management and conservation. Ottawa, ON,Canada: NRC Research Press; 1996. p. 7–32.4. Monclus R, Dreyer E, Villar M, Delmotte FM, Delay D, Petit J-M, et al. Impactof drought on productivity and water use efficiency in 29 genotypes ofPopulus deltoides x Populus nigra. New Phytol. 2006;169:765–77.5. Tschaplinski TJ, Blake TJ. Water-stress tolerance and late-season organicsolute accumulation in hybrid poplar. Can J Bot. 1989;67:1681–8.6. Tschaplinski TJ, Tuskan GA. Water-stress tolerance of black and easterncottonwood clones and four hybrid progeny. II. Metabolites and inorganicions that constitute osmotic adjustment. Can J For Res. 1994;24:681–7.7. Ceulemans R, Impens I, Steenackers V. Genetic variation in aspects of leafgrowth of Populus clones, using leaf plastochron index. Can J For Res.1988;18:1069–77.8. McDowell N, Pockman WT, Allen CD, Breshears DD, Cobb N, Kolb T, et al.Mechanisms of plant survival and mortality during drought: why do someplants survive while others succumb to drought? New Phytol.2008;178:719–39.9. Blake TJ, Tschaplinski TJ, Eastham A. Stomatal control of water use efficiencyin poplar clones and hybrids. Can J Bot. 1984;1984:1344–51.10. Seki M, Umezawa T, Urano K, Shinozaki K. Regulatory metabolic networks indrought stress responses. Curr Opin Plant Biol. 2007;10:296–302.11. Krasensky J, Jonak C. Drought, salt, and temperature stress-inducedmetabolic rearrangements and regulatory networks. J Ex Bot.2012;63:1593–608.12. Brosche M, Vinocur B, Alatalo E, Lamminmaki A, Teichmann T, Ottow EA,et al. Gene expression and metabolite profiling of Populus euphraticagrowing in the Negev desert. Genome Biol. 2005;6:1–17.13. Wilkins O, Waldron L, Nahal H, Provart NJ, Campbell MM. Genotype andtime of day shape the Populus drought response. Plant J. 2009;60:703–15.14. Hamanishi ET, Raj S, Wilkins O, Thomas BR, Mansfield SD, Plant AL, et al.Intraspecific variation in the Populus balsamifera drought transcriptome.Plant Cell Environ. 2010;33:1742–55.15. Yan D-H, Fenning T, Tang S, Xia X, Yin W. Genome-wide transcriptionalresponse of Populus euphratica to long-term drought stress. Plant Sci.2012;195:24–35.16. Barchet GLH, Dauwe R, Guy R, Schroeder WR, Soolanayakanahally RY,Campbell MM, et al. Investigating the drought-stress response of hybridpoplar genotypes by metabolite profiling. Tree Physiol. 2013;00:1–7.17. Nishizawa A, Yabuta Y, Shigeoka S. Galactinol and raffinose constitute anovel function to protect plants from oxidative damage. Plant Physiol.2008;147:1251–63.18. Kozlowski T, Pallardy S. Acclimation and adaptive responses of woodyplants to environmental stresses. Bot Rev. 2002;68:270–334.19. Seki M, Umezawa T, Urano K. Shinozaki K. Regulatory metabolic networks indrought stress responses. 2007;10:296–302.20. Taji T, Ohsumi C, Iuchi S, Seki M, Kasuga M, Kobayashi M, et al. Importantroles of drought- and cold-inducible genes for galactinol synthase in stresstolerance in Arabidopsis thaliana. Plant J. 2002;29:417–26.21. Bartels D, Sunkar R. Drought and Salt Tolerance in Plants. Crit Rev Plant Sci.2005;24:23–58.22. Cocozza C, Cherubini P, Regier N, Saurer M, Frey JG, Tognetti R. Early effectsof water deficit on two parental clones of Populus nigra grown underdifferent environmental conditions. Funct Plant Biol. 2010;37:244–54.23. Escobar-Gutiérrez AJ, Zipperlin B, Carbonne F, Moing A, Gaudillére JP.Photosynthesis, carbon partitioning and metabolite content during droughtstress in peach seedlings. Aust J Plant Physiol. 1998;25:197–205.24. Wilkinson S, Davies WJ. ABA-based chemical signalling: the co-ordination ofresponses to stress in plants. Plant Cell Environ. 2002;25:195–210.25. Janz D, Behnke K, Schnitzler J-P, Kanawati B, Schmitt-Kopplin P, Polle A.Pathway analysis of the transcriptome and metabolome of salt sensitiveand tolerant poplar species reveals evolutionary adaption of stress tolerancemechanisms. BMC Plant Biol. 2010;10:150.26. R Development Core Team. R: A Language and Environment for StatisticalComputing. Vienna: R Foundation for Statistical Computing; 2009.27. Eisen MB, Spellman PT, Brown PO, Botstein D. Cluster analysis and display ofgenome-wide expression patterns. Proc Natl Acad Sci U S A.1998;95:14863–8.28. Rabbani MA, Maruyama K, Abe H, Khan MA, Katsura K, Ito Y, et al.Monitoring expression profiles of rice genes under cold, drought, andhigh-salinity stresses and abscisic acid application using cDNA microarrayand RNA gel-blot analyses. Plant Physiol. 2003;133:1755–67.29. Suzuki R, Shimodaira H. Pvclust: an R package for assessing the uncertaintyin hierarchical clustering. Bioinformatics. 2006;22:1540–2.30. Benjamini Y, Hochberg Y. Controlling the false discovery rate: A practicaland powerful approach to multiple testing. J R Stat Soc Ser B.1995;57:289–300.31. Langfelder P, Horvath S. WGCNA: an R package for weighted correlationnetwork analysis. BMC Bioinformatics. 2008;9:559.32. Lopes CT, Franz M, Kazi F, Donaldson SL, Morris Q, Bader GD. CytoscapeWeb: an interactive web-based network browser. Bioinformatics.2010;26:2347–8.33. Du Z, Zhou X, Ling Y, Zhang Z, Su Z. agriGO: a GO analysis toolkit for theagricultural community. Nucleic Acids Res. 2010;38 Suppl(web server):W64–70.Submit your next manuscript to BioMed Centraland take full advantage of: • Convenient online submission• Thorough peer review• No space constraints or color figure charges• Immediate publication on acceptance• Inclusion in PubMed, CAS, Scopus and Google Scholar• Research which is freely available for redistributionHamanishi et al. BMC Genomics  (2015) 16:329 Page 16 of 1634. Kanehisa M, Goto S. KEGG: Kyoto encyclopedia of genes and genomes.Nucleic Acids Res. 2000;28:27–30.35. Masoudi-Nejad A, Goto S, Endo TR, Kanehisa M. KEGG bioinformaticsresource for plant genomics research. Methods Mol Biol. 2007;406:437–58.36. Kanehisa M, Goto S, Sato Y, Furumichi M, Tanabe M. KEGG for integrationand interpretation of large-scale molecular data sets. Nucleic Acids Res.2011;40:D109–14.37. Hoffman DE, Jonsson P, Bylesjo M, Trygg J, Antti H, Eriksson ME, et al.Changes in diurnal patterns within the Populus transcriptome andmetabolome in response to photoperiod variation. Plant Cell Environ.2010;33:1298–313.38. Urbanczyk-Wochniak E, Baxter C, Kolbe A, Kopka J, Sweetlove LJ, Fernie AR.Profiling of diurnal patterns of metabolite and transcript abundance inpotato (Solanum tuberosum) leaves. Planta. 2005;221:891–903.39. Geiger DR, Servaites JC. Diurnal regulation of photosynthetic carbonmetabolism in C3 plants. Annu Rev Plant Physiol Plant Mol Biol.1994;45:235–56.40. Zeeman SC, Smith SM, Smith AM. The diurnal metabolism of leaf starch.Biochem J. 2007;401:13–28.41. Tschaplinski TJ, Tuskan GA, Gunderson CA. Water-stress tolerance of blackand eastern cottonwood clones and four hybrid progeny. 1. Growth, waterrelations and gas exchange. Can J For Res. 1994;24:364–71.42. Joshi V, Jander G. Arabidopsis methionine γ-lyase is regulated according toisoleucine biosynthesis needs but plays a subordinate role to threoninedeaminase. Plant Physiol. 2009;151:367–78.43. Bowne JB, Erwin TA, Juttner J, Schnurbusch T, Langridge P, Bacic A, et al.Drought responses of leaf tissues from wheat cultivars of differing droughttolerance at the metabolite level. Mol Plant. 2012;5:418–29.44. Nambara E, Kawaide H, Kamiya Y, Naito S. Characterization of an Arabidopsisthaliana mutant that has a defect in ABA accumulation: ABA-dependentand ABA-independent accumulation of free amino acids during dehydration.Plant Cell Physiol. 1998;39:853–8.45. Koussevitzky S, Suzuki N, Huntington S, Armijo L, Sha W, Cortes D, et al.Ascorbate peroxidase 1 plays a key role in the response of Arabidopsisthaliana to stress combination. JBC. 2008;283:34197–203.46. Maclennan DH, Beevers H, Harley JL. “Compartmentation” of acids in planttissues. Biochem J. 1963;89:316–27.47. Chaves MM, Maroco JP, Pereira JS. Understanding plant responses todrought - from genes to the whole plant. Funct Plant Biol. 2003;30:239–64.48. Regier N, Streb S, Cocozza C, Schaub M, Cherubini P, Zeeman SC, et al.Drought tolerance of two black poplar (Populus nigra L.) clones:contribution of carbohydrates and oxidative stress defence. Plant CellEnviron. 2009;32:1724–36.49. Santarius KA. The protective effect of sugars on chloroplast membranesduring temperature and water stress and its relationship to frost,desiccation and heat resistance. Planta. 1973;113:105–14.50. Castillo EM, De Lumen BO, Reyes PS. Raffinose synthase and galactinolsynthase in developing seeds and leaves of legumes. Aust J Agr Res.1990;38:351–5.51. Taji T, Seki M, Satou M, Sakurai T, Kobayashi M, Ishiyama K, et al. Comparativegenomics in salt tolerance between Arabidopsis and Arabidopsis-relatedhalophyte salt cress using Arabidopsis microarray. Plant Physiol.2004;135:1697–709.52. Rodríguez-Calcerrada J, Atkin OK, Robson TM, Zaragoza-Castells J, Gil L, ArandaI. Thermal acclimation of leaf dark respiration of beech seedlings experiencingsummer drought in high and low light environments. Tree Physiol.2010;30:214–24.Submit your manuscript at www.biomedcentral.com/submit


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