PlantEcol DOI10.1007/s11258-015-0500-x Transcriptomics of plant responses to apical damage reveals no negative correlation between tolerance and defense . Daniel R. Scholes Ken N. Paige Received:30April2015/Accepted:8July2015 (cid:2)SpringerScience+BusinessMediaDordrecht2015 Abstract While one may expect the loss of plant totalgeneexpressionviaRNA-sequencingofundam- tissue by animal herbivores to be universally detri- aged and mechanically damaged plants of three mental to a plant’s fitness, a wide range of tolerance Arabidopsis thaliana genotypes that differ in their responsesexists,includingundercompensation(lower compensatory performances: Columbia-4, Landsberg fitnesswhendamaged),equalcompensation,andeven erecta, and a recombinant inbred line (RIL) from a overcompensation(increased fitnesswhendamaged). Columbia-4 9 Landsberg erecta cross. Among the Theory predicts that these responses could be con- many genetic pathways that responded to clipping, strained by the investment into defensive chemicals Columbia-4significantlyup-regulatedgenesinvolved and structures produced for improving resistance in insecondary defense chemistryand equallycompen- damaged tissues, and thus tolerance and induced sated for fitness while Landsberg erecta and the RIL defense could be considered alternative strategies to both undercompensated and significantly down-regu- the selection pressure imposed by herbivory. To lated secondary defense pathways. The genotypes’ determine which genetic pathways underlie differ- different compensatory performances are thus posi- ences in compensatory (i.e., growth and fitness) tively correlated with their differential investments responses to damage by tissue loss, and to test their intosecondarymetabolismfollowingtissueloss.This relation with pathways involved in defense, we studyidentifiesdifferentialpost-damagegeneregula- performed a controlled greenhouse study to measure tionofgrowth,developmentalsignaling,andenviron- mentalresponsepathwaysinA.thaliana,andprovides the first transcriptomic evidence counter to the CommunicatedbySantiagoC.Gonzalez-Martinez. presumed tradeoff between tolerance and defense in plant-herbivoreinteractions. Electronicsupplementarymaterial Theonlineversionof thisarticle(doi:10.1007/s11258-015-0500-x)containssupple- mentarymaterial,whichisavailabletoauthorizedusers. Keywords Herbivory(cid:2)Resistance(cid:2)Compensation(cid:2) Chemistry(cid:2)Glucosinolate(cid:2)RNA-seq D.R.Scholes(&) DepartmentofBiology,UniversityofIndianapolis,1400 E.HannaAve.,Indianapolis,IN46227,USA e-mail:[email protected] Introduction D.R.Scholes(cid:2)K.N.Paige A broad, consequential issue in the ecology and SchoolofIntegrativeBiology,UniversityofIllinoisat evolution of plant-animal interactions regards the Urbana–Champaign,515MorrillHall,505S.Goodwin Ave.,Urbana,IL61801,USA ways in which plants respond to herbivore-induced 123 PlantEcol damage.Plantsexhibitawidevarietyofresponsesto toleranceordefenseduetothenecessityofallocating damage by animal herbivores, including defense limited resources (Coley et al. 1985; Mauricio et al. strategies that aim to prevent damage, and tolerance 1997;WiseandAbrahamson2005,2007).Forexam- strategiesthataimtopreventthelossofreproductive ple, fitness following apical damage of Solidago outputuponsustainingdamage(RosenthalandKota- altissima decreased nearly tenfold when plants were nen1994;Stoweetal.2000;Agrawal2011;notethat nutrient-stressed versus fertilized, a result consistent while we use these terms in accordance with their with the limiting resources model of tolerance (Wise historicalusage,‘‘resistance’’and‘‘defense’’areoften andAbrahamson2008).Theproductionofsecondary usedconverselytotheirpresentationhere(Straussand metabolites may also be resource-limited, where the Agrawal1999)).Defensivetraitsoftenentailprevent- constitutivelevelandinducibilityofsuchdefensesare ingdamageviastructuralimpedimentsorbydecreas- negativelyimpactedbylownutrientavailability(e.g., ingthepalatabilityoftheplanttissue,andmayinclude Sampedro et al. 2011). There is also theoretical and the production of thorns, trichomes, lignin, latex, empirical evidence of a negative genetic correlation numerous classes of secondary metabolites, or the between tolerance and defense strategies (van der sequestrationofsilicon,asexamples(Levin1973;Ma Meijden et al. 1988; Fineblum and Rausher 1995; 2004;Mitho¨ferandBoland2012).Thesetraitsmaybe Stowe 1998). Specifically, van der Meijden producedconstitutively, often asdefensive structures et al. (1988) argued that because defended plants and baseline levels of secondary compounds, or they were likely to experience a lower frequency and may be induced upon the detection of a potential intensity of herbivore attack, selection for tolerance herbivore or after sustaining damage (Kaplan et al. ability would be reduced relative to less defended 2008). Defense traits may also be general or specific plants. Conversely, tolerant plants would not neces- withregardtotheherbivoreagainstwhichtheyact— sarily face selective pressure to invest in defense if for example, there is substantial evidence for coevo- theydonotincurfitnesslossesfromherbivory(vander lutionbetweenplantsandtheirspecializedherbivores Meijden et al. 1988). The dynamics of the tolerance/ for defense compound toxicity and detoxification, defense tradeoff have been theorized from varied respectively, inanumber ofsystems(Fraenkel 1959; perspectives(Stamp2003),butultimatelyunderthese EhrlichandRaven1964;Berenbaum1983). circumstances tolerance and defense are considered Tolerancereferstoaplant’sabilitytomaintainfitness redundantresistancestrategiessuchthattheywouldbe despite sustainingdamage. Tolerancemechanismsare mutuallyexclusiveifeithertraitreceivedahighlevel generallytheorizedtoincludethereallocationofstored ofinvestment(vanderMeijdenetal. 1988). resourcestoreplacethetissuesremoved,increasedstem Despitethetheoreticaltradeoffbetweentoleranceand productionand/orbranchingfromthehormonalrelease defense strategies, recent research into the genetic of apical dominance, and increases in photosynthetic underpinnings of tolerance provides evidence that output and growth rates following damage, among toleranceanddefensemaynotbealternativestrategies others(StraussandAgrawal1999).Precisemechanisms after all. For example, the pentose phosphate pathway underlying tolerance are not fully understood and are (PPP)isametabolicpathwaythatprovidesfive-carbon likely variable among populations, species, types of sugarsandNADPHforgeneralizedbiosynthesisinthe damage, and environments (Maschinskiand Whitham Calvin cycle (Kruger and von Schaewen 2003), but it 1989;StraussandAgrawal1999).Toleranceoutcomes alsoprovidestherawmaterialsforsecondarymetabolite range from undercompensation (reduced fitness when production via the shikimate pathway (Maeda and damaged), equal compensation (i.e., complete toler- Dudareva2012).Siddappajietal.(2013)demonstrated ance), and overcompensation (increased fitness when that the A. thaliana genotype Columbia-4 had higher damaged), with compensatory abilities varying across expressionofGLUCOSE-6-PHOSPHATEDEHYDRO- genotypes and populations of a wide range of species GENASE 1 (G6PD1), which encodes the enzyme that (Rosenthal and Kotanen 1994; Strauss and Agrawal catalyzesthefirstandrate-limitingstepofthePPP,when 1999;Stoweetal.2000;Tiffin2000). damaged compared to when undamaged and that it Toleranceanddefensestrategieshavetraditionally overcompensatedforfruitproduction.Landsbergerecta, been considered alternative forms of resistance to incontrast,hadreducedG6PD1expressionandunder- herbivory with plants primarily exhibiting either compensated(Siddappajietal.2013).Theexperimental 123 PlantEcol knockoutandcomplementationofthisgeneconfirmed positively corresponds with the expression of sec- G6PD1’s, and therefore the PPP’s, influence in the ondary defense metabolic pathways in these geno- compensatory performance of these genotypes. This types, providing transcriptomic evidence that influence may arise in part via endoreduplication, the toleranceanddefensemetabolismarenotnecessarily replication of the genome without mitosis such that mutuallyexclusivestrategiesandthattheymayinfact cellular genome copy number increases with each beemployedconcurrentlyfollowingdamage. replication. This process is assumed to play important rolesinplantdevelopment,growth,fitness,andgeneral stress tolerance through its effects on cell size, Materialsandmethods metabolism, and gene expression (Sugimoto-Shirasu andRoberts2003;Leeetal.2009;Breueretal.2014;for A.thalianagenotypes,growth,andexperimental reviewseeScholesandPaige2015).AlongwiththePPP, clipping Columbia-4isknowntoincreaseitsendoreduplication during regrowth following damage while Landsberg Columbia-4(Col-4;TAIRstocknumber:CS933;The erecta has no such increase (Scholes andPaige2011). Arabidopsis Information Resource 2014), Landsberg Further, experimentally increasing the endopolyploidy erecta (Ler-0; TAIR stock number: CS20), and one ofanA.thalianagenotypethattypicallyundercompen- RILfromtheircrossing(TAIRstocknumber:CS1936) sates allows it to completely mitigate the otherwise wereselectedforthisstudybasedontheirpreviously detrimentaleffectsofdamageonitsfitness(Scholesand demonstratedfitnesscompensatoryresponses(Scholes Paige 2014). Since the PPP also produces DNA andPaige2011;Siddappajietal.2013).Allgenotypes nucleotides, the ability of a genotype to increase its wereinbredaminimumofeightgenerationstoachieve cellularploidyduringregrowthislikelyconstrainedby fullhomozygosity(ListerandDean1993). itsabilitytoincreasethePPP(KrugerandvonSchaewen Sixty-fiveindividualsofeachgenotype(Col-4,Ler- 2003).Giventheseknownrolesingeneralizedprimary 0,CS1936)weregrown,witharandomizeddistribution, metabolism, endoreduplication, and the production of inagreenhouseatapproximately21 (cid:3)Candona12 h secondarydefensemetabolites,thePPPmay,therefore, light/darkcycle.Whenboltinginflorescencesreached6 notonlyunderlietheabilitiesoftheseplantstotolerate cm in height, the inflorescences of 35 plants of each apical damage but also provide a mechanism for the genotype were clipped with scissors, leaving 1 cm of concurrentproductionofdefensivecompounds. inflorescencetissueashasbeenperformedinprevious Inthisstudy,weprovideatestofthehypothesisthat studieswiththesegenotypes(ScholesandPaige2011). tolerance and defense strategies to herbivory are not This clipping regimen is comparable to natural apical necessarily in opposition, but rather are positively damage of A. thaliana observed throughout its native correlated via the co-regulation of their genetic range,stimulatesasimilarchangeinarchitecturaltraits pathways. Specifically, using three genotypes of following damage, and approximately coincides with A. thaliana that vary in their degree of tolerance, we A.thaliana’smaximumrateofinflorescenceelongation initially ask whether a correlation exists between the (Scholes et al. unpublished). Thirty of the remaining response in expression of genes in the PPP, the plantsofeachgenotyperemainedunclippedascontrols. endocycle, and secondary metabolite biosynthesis Fiveoftheclippedplantsofeachgenotypewereutilized following damage. We additionally ask whether the for transcriptomicanalysis (see ‘‘Transcriptomicanal- expressionofgenesinthosepathwaysalsocorrelates ysis’’sectionfordetails).Rosette diameteratthetime withmeasuresoffitnesscompensationandendopoly- theinflorescencereached6cminheightwasrecordedas ploidyfollowingdamage.Toaddressthesequestions, ameasureofplantsizeatthisstage. we performed quantitative transcriptome sequencing of A. thaliana Columbia-4, Landsberg erecta, and a Phenotypicanalysis Columbia-4 9 Landsberg erecta recombinant inbred line(RIL)todeterminewhetherthegeneticexpression Measurementofendopolyploidy oftoleranceanddefensecorrelatewiththeircompen- satory performances following damage. Our results Attheinductionofsenescence,allinflorescencetissue demonstrate that compensatory performance of 20 plants (10 unclipped, 10 clipped) of each 123 PlantEcol genotype was analyzed for nuclear DNA content via plant’s rosette diameter at an inflorescence height of flow cytometry. Tissue for cytometric analysis was 6 cm(tocontrolfordifferencesinplantsize)andthe chopped with a razor blade, sheared in a nuclear number of days from planting to an inflorescence isolation buffer (sodium citrate, 3-morpholino- heightof6 cm(tocontrolfordifferencesindevelop- propane-1-sulfonic acid, magnesium chloride, Triton mentaltiming)wereusedascovariatestostandardize X-100; Galbraith et al. 1983), filtered for debris between genotype 9 treatment groups. Counts of removal, and stained with the DNA fluorophore silique yield, seed yield, and biomass were also propidiumiodide.Isolatednucleiwerethenmeasured square-root transformed to approximate normality to forDNAcontentviaaBDBiosciences(SanJose,CA, satisfy the assumptions of ANOVA. Linear contrasts USA) FACScanto flow cytometer. Nuclei population were used to test for differences between unclipped gating was performed using De Novo Software FCS andclippedplantsofeachgenotypeforeachmeasure. Express(v.3;LosAngeles,CA,USA)tomeasurethe number of nuclei at each ploidy level (2C, 4C, 8C, Transcriptomicanalysis 16C) for each plant sample. The cycle value, inter- preted as the average number of endocycles that a Samplepreparation,libraryconstruction, nucleus in the sample has undergone and thus an andsequencing overall measure of endoreduplication (Barow and Meister 2003), was calculated by the following Five of the clipped plants of each genotype were equation: randomly selected for transcriptomic analysis, with the clipped inflorescence tissue collected for RNA- Cyclevalue¼ð0(cid:3)n þ1(cid:3)n þ2(cid:3)n 2C 4C 8C sequencing.Uponthegenerationofaxillarystems,the þ3(cid:3)n Þ=ðn þn þn þn Þ 16C 2C 4C 8C 16C firstregrownaxillaryinflorescencetoreachaheightof 6 cm was likewise collected from each plant for where the cycle value is the sum of the number of transcriptomics. Initial and lateral stem tissues col- nucleiateachploidylevelmultipliedbythenumberof lected for transcriptomics (i.e., stems collected at the endocycles required to achieve each corresponding timeofandafterclipping,respectively)werecompa- ploidy level, divided by the total number of nuclei rable in mass, chronological age, and developmental measured. stage,allowingforthecomparisonofgeneexpression beforeversusafterclippingforindividualplantswhile Measurementoffitness controlling for differences between individuals. Tis- sues for transcriptomic analysis were flash-frozen in At the completion of senescence, 40 plants of each liquid nitrogen immediately upon clipping from the genotype(20unclipped,20clipped)wereanalyzedfor plant.TotalRNAwasextractedfromeachsamplevia fitness. Inflorescence dry biomass and silique yield an RNeasy Plant Mini Kit (Qiagen; Venlo, Nether- were measured for each plant. Seed yield for each lands). Qualityofextracted RNA wasassessed onan plant was estimated by multiplying the number of Agilent2100Bioanalyzer(AgilentTechnologies,Inc., siliquesbytheaveragenumberofseedspersiliquefor SantaClara,CA,USA;softwarev.1.3)withtheassay the respective genotype 9 treatment group (calcu- class designated as Plant RNA Nano. RNA integrity lated by counting the number of seeds for five valuesaveraged9.0(range7.7–10.0),25S/18SrRNA representativesiliquesfromeachplant). ratios averaged 1.9 (range 1.3–2.7), and adequate quality was confirmed upon visual inspection of Statisticalanalysisofphenotypicmeasures electrophoretic images. cDNA libraries were con- structed by the High-Throughput Sequencing and Analysesforphenotypicmeasureswereconductedas GenotypingUnit,RoyJ.CarverBiotechnologyCenter general linear models with SAS (v.9.3; Cary, NC, of the University of Illinois at Urbana–Champaign USA).Genotypeandtreatment(clipping)weretested (Urbana, IL, USA), using a TruSeq Stranded Total as fixed effects via ANOVA for silique yield, seed RNA Sample Prep Kit (Illumina; San Diego, CA, yield,inflorescencedrybiomass,andcyclevalues.For USA). All libraries were quantified by qPCR and siliqueyield, seedyield,andbiomassmeasures,each sequenced on three lanes (one for each genotype, 123 PlantEcol comprising ten multiplexed samples per lane), dis- the false discovery rate (FDR) at 5% in edgeR based tributed randomly within a single flow cell, for 101 on the total number of genes analyzed for the cycleseachonaHiSeq2000sequencerusingaTruSeq respectivegenotype,generatingFDR-correctedpval- SBS Sequencing Kit (v.2) and CASAVA software ues(i.e.,qvalues;BenjaminiandHochberg1995). (v.1.8.2;Illumina;SanDiego,CA,USA).Onaverage, 20,717,090readsweregeneratedpersample. Genome-wideGOtermenrichment,GOterm redundancyreduction,andtargetedGOtermanalysis Readmapping,countnormalization, andidentificationofdifferentiallyexpressedgenes Gene ontology (GO) analysis was performed to determine pathway-level patterns of gene expression Reads were trimmed to remove adapter sequences, enrichment among the genes deemed significantly sequences with low quality scores (limit = 0.05, differentiallyexpressed(i.e.,q\0.05)relativetothe Phred scores [20; Ewing and Green 1998), and entiresetofanalyzedgenes.GOtermsaredescriptors sequenceswithmultipleambiguousbases(maximum of genes based on the cellular component, molecular two ambiguous nucleotides per read) using CLC function, and biological process for which they are Genomics Workbench (v.6.5; CLC bio; Aarhus, associated,andareoftendeterminedbyexperimental Denmark). After these quality control measures, an demonstration or by sequence similarity to genes of averageof19,303,545readspersampleweremapped known function. GO terms may therefore be used to to the TAIR10 Arabidopsis thaliana annotated refer- inferthefunctionsofgenesofinterest(e.g.,differen- ence genome of 28,642 genes (The Arabidopsis tially expressed genes) or for targeted analysis of InformationResource2014). geneswithspecificfunctions(e.g.,todeterminewhich Aftermapping,geneswithfewerthanfiveuniquely genes within a pathway of interest are differentially mappedreadspertreatmentonaveragewereremoved expressed, and/or if the pathway as a whole is from the analysis given the increased measurement differentiallyregulated).Geneidentitiesandqvalues error typically associated with low read counts wereimportedintoGOrilla(Edenetal.2009),which (Robinson and Smyth 2007). An average of approx- analyzesauser-supplied‘‘target’’genelistforsignif- imately21,707genesremainedpersample.Statistical icant GO term enrichment relative to a user-supplied analysis for differential gene expression was per- ‘‘background’’ gene list. The significantly up-regu- formed using the edgeR (Robinson et al. 2010) latedgenesandthesignificantlydown-regulatedgenes package in R (i386 of v.3.0.3; R Core Team 2014). were analyzed separately for each genotype. GO The numbers of trimmed reads mapped uniquely to analyses were performed with Arabidopsis thaliana each gene were normalized via the edgeR function selected as the experimental organism, with two calcNormFactorstocontrolfordifferencesinsequenc- unrankedlistsofgenes(target:eitherup-regulatedor ing depth among samples (Robinson and Oshlack down-regulated genes; background: all genes ana- 2010). Normalized read counts of each gene were lyzed for the respective genotype), for the biological compared between treatments (before clipping, after process ontology, and with a p value threshold for clipping) via a generalized linear model (GLM; significant GO term enrichment of 10-3. Briefly, for McCarthy et al. 2012), with samples coming from each GO term for which A. thaliana genes are the same individual paired to control for individual associated (2340 GO terms at the time of analysis), variation. Model dispersion was estimated for each GOrilla uses the total number of genes in the user- geneseparatelyviathefunctionestimateTagwiseDisp suppliedbackgroundgenelist(N),thetotalnumberof in edgeR (Robinson and Smyth 2007, 2008). Genes genes associated with the GO term (B), the total whose normalized read counts differed significantly number of genes in the user-supplied target gene list before vs. after clipping were identified using GLM (n), and the number of genes in the target gene list likelihood ratio tests via the function glmLRT in associated with the GO term (b) to calculate an edgeR (Lund et al. 2012). Genotypes were analyzed enrichmentscore[definedas(b/n)/(B/N)];Edenetal. independently with no direct comparisons made 2009). GOrilla then uses the hypergeometric method among them. p values resulting from the GLM to calculate a p value from the enrichment score and likelihood ratio tests were then corrected to control produces an FDR-corrected p value (i.e., q value) to 123 PlantEcol adjust for the number of GO terms in the analysis due to their previously reported involvement in (Benjamini and Hochberg 1995; Eden et al. 2009). tolerance (Scholes and Paige 2011, 2014; Siddappaji Genes ineither user-supplied gene list notassociated et al.2013).To assess the effects ofthetreatment on with a GO term were omitted from the analyses specificGOtermsofinterestselectedapriori,listsof (amongthethreegenotypes,13,261genesonaverage A.thalianageneswithdirectandindirectassociations wereanalyzedforGOenrichment).AcrossallsixGO with each a priori GO term were obtained using the enrichment analyses (up- and down-regulated genes AmiGO 2 (http://amigo2.geneontology.org) Grebe analyzed separately for three genotypes), an average search wizard.Foreachgeneassociatedwithagiven of19.7GOtermsweredeemedsignificantlyenriched GO term ofinterest, the p value calculated by edgeR (Supplementarymaterials1,2,3). wascorrectedforafalsediscoveryrateof5 %based GO terms deemed significantly enriched (i.e., on the total number of genes associated with the GO q\0.05) were imported with their q values into termthatremained afterthequalitycontrolmeasures REViGOtoreducetheredundancyoftheGOtermlist describedabove(BenjaminiandHochberg1995).An (Supeketal.2011).REViGOusesqvalues,semantic exact binomial test was then performed in R via the similarity, and GO term parent-child relationships to binom.testfunctiontodetermineifthenumberofup- calculatea‘‘dispensability’’score(rangingfrom0to1 regulatedgenesdifferedsignificantlyfromthenumber with greater values indicating greater dispensability) ofdown-regulatedgenesforeachGOterm. foreachGOtermandproducesashortenedGOterm list based on a user-indicated threshold (Supek et al. 2011). Analyses in REViGO were performed with a Results dispensabilitythreshold(i.e.,Cvalue)of0.5,imported q values for each user-supplied GO term, the Ara- FitnessandnuclearDNAcontent bidopsis thaliana database selected to provide GO term relationships and sizes, and SimRel as the LandsbergerectaandtheCol-4 9 Ler-0RILCS1936 semanticsimilaritymeasure(Supeketal.2011).After bothsignificantly undercompensatedforsiliqueyield redundancy reduction, an average of 5.8 GO terms [t(67) = 2.88, p\0.01, and t(67) = 2.52, p\0.05, remained across the six analyses (Supplementary respectively;Fig. 1].CS1936alsosignificantlyunder- materials1,2,3). compensated for seed yield [t(67) = 3.08, p\0.01), Giventhenatureofourclippingtreatmentandour while Ler-0 nearly undercompensated for seed yield interest in the integration of the PPP, secondary [t(67) = 1.93, p = 0.058; Fig. 1). Columbia-4 defense chemistry, and endoreduplication following equally compensated for both silique and seed yield plantdamage,sevenGOtermswereselectedapriori [t(67) = 0.41, p = 0.682, and t(67) = 0.16, (Table 1).Theseincludeoverallsecondarymetabolite p = 0.874, respectively; Fig. 1]. There were no dif- biosynthesis as well as the biosynthesis of specific ferences between unclipped and clipped plants in classes of secondary compounds of A. thaliana inflorescence dry biomass for any genotype [Col-4: (Table 1; Kliebenstein 2004). We acknowledge that t(67) = 0.30, p = 0.762; Ler-0: t(67) = 0.95, the classes of secondary compounds assessed may p = 0.348; CS1936: t(67) = 1.20, p = 0.234; compriseavarietyofcompoundsub-classes(e.g.,sub- Fig. 1], though the change in their means closely classes of phenylpropanoids include coumarins, fla- approximates the silique and seed compensatory vonoids, lignins, etc., which may or may not be performanceforeachgenotype(Fig. 1). specifically involved in herbivore defense; Solecka Although cycle values of unclipped and clipped 1997);however,limitationsintheavailabilityofsub- plantsdidnotdiffersignificantlyforanygenotype,the class GO terms, the annotation of genes to these GO changes in direction of their means did correspond terms, and ambiguity in the involvement of specific roughlytotheirfitnesscompensatoryresponseswitha compounds in herbivore defense prevent the assess- trendtowarddecreasedendopolyploidyforLer-0and ment of more finely defined compound sub-classes. CS1936 [t(39) = 0.66, p = 0.514, and t(39) = 0.25, The‘‘pentosephosphateshunt’’and‘‘DNAendoredu- p = 0.807,respectively]andatrendtowardincreased plication’’ GO terms were also selected for analysis ploidyforCol-4[t(39) = 1.67,p = 0.103;Fig. 1]. 123 PlantEcol Table1 Geneontologytermsofinterestselectedapriori ID Term #Genes Col-4 Ler-0 CS1936 GO:0044550 Secondarymetabolitebiosyntheticprocess 310 56 11 51 GO:0016114 Terpenoidbiosyntheticprocess 261 6 1 2 GO:0019761 Glucosinolatebiosyntheticprocess 169 19 10 21 GO:0009699 Phenylpropanoidbiosyntheticprocess 135 40 1 30 GO:0052315 Phytoalexinbiosyntheticprocess 11 0 0 2 GO:0006098 Pentosephosphateshunt 182 0 2 2 GO:0042023 DNAendoreduplication 111 0 0 1 GeneontologyIDs,terms,andnumbersofArabidopsisthalianagenesannotatedtoeachGOtermwerecompiledfromtheAmiGO2 geneontologybrowser(http://www.geneontology.org).Duetodatatrimmingforquality,thenumberofgenesannotatedtoeachGO term isnot necessarily equal to thenumber of genes that constituted the family for which thefalse discovery rate was corrected. ValuesintheCol-4,Ler-0,andCS1936columnsarethenumbersofdifferentiallyexpressedgenesforeachrespectivegenotypeafter falsediscoveryrate(FDR)correction(q\0.05).SeeSupplementarymaterials1,2,and3forfullresults terms). Upon clipping, 302 genes were significantly 10 up-regulated in Col-4 while 663 were significantly %) down-regulated. When considering Col-4 up-regu- e ( 0 lated genes,21GOtermsweresignificantly enriched r u (Fig. 2;Supplementarymaterial1).Uponredundancy s a -10 e reduction, the three remaining GO terms were all m n involved in cell wall biosynthesis (Fig. 2; Supple- ge i -20 mentary material 4). The 30 GO terms significantly n ha -30 # * enriched with down-regulated Col-4 genes were C generallyinvolvedinpollentubegrowthandcellwall * * -40 modification(Supplementarymaterial1).Afterredun- Col-4 Ler-0 CS1936 dancy reduction, eight GO terms remained (Fig. 2; Supplementarymaterial4). Fig.1 Percent change in phenotypic measures of clipped Only four GO terms were significantly enriched Columbia-4, Landsberg erecta, and CS1936 plants relative to unclipped controls. Asterisks(*) significant(a=0.05)differ- with Ler-0’s 38 up-regulated genes, with three ences between unclipped and clipped plants for the given remainingafterredundancy reduction (Fig. 2).These genotype and measure; #p=0.0584. Light gray shading termswereallrelatedtotheresponsetovariousstimuli silique yield, black shading seed yield, white shading mass, (e.g.,chitin,nitrogen,fungi,oxygen-containingcom- darkgrayshadingcyclevalue pound; Supplementary materials 2, 4). Although 27 Geneexpression genesweresignificantlydown-regulated,noGOterms were significantly enriched with Ler-0 down-regu- Differentialgeneexpressionandgenome-wideGO latedgenes(Fig. 2;Supplementarymaterial4). termenrichment The Col-4 9 Ler-0 RIL CS1936 exhibited 124 significantlyup-regulatedgenesand148significantly For those genes differentially expressed between down-regulatedgenes.CS1936demonstratedavaried samples before and after clipping, we conducted a transcriptomic response to the clipping treatment geneontology(GO)analysisthatidentifiesbiological relative to its parents, with 25 and 38 GO terms processes (i.e., GO terms) for which the genes are significantly enriched with up- and down-regulated associated. Here we describe broad themes from our genes, respectively, relative to the genomic back- gene regulation and biological process analyses (see ground (Fig. 2; Supplementary material 3). Upon Supplementary materials 1, 2, and 3 for detailed redundancyreductionofup-regulatedGOterms,only analyses and full lists of significant genes and GO fiveremainedandcollectivelywereassociatedwitha 123 PlantEcol a b with no significant changes in the regulation of the othermetabolitegroups(Fig. 3a).TheCol-4 9 Ler-0 RIL CS1936 displayed intermediate secondary metabolite regulation relative to its parents, with significant down-regulation of glucosinolates, signif- icant up-regulation of phenylpropanoids, and no sig- nificant changes in the regulation of biosynthesis of phytoalexins, terpenoids, or secondary metabolites overall(Fig. 3a).Nogenotypehadsignificantchanges in the regulation of the pentose phosphate shunt or Fig.2 The number of gene ontology terms significantly endoreduplicationuponclipping(Fig. 3b). enriched with a up- and b down-regulated genes for Columbia-4,Landsbergerecta,andCS1936followingclipping inthegenome-wideanalysis.Boldvaluesindicatethenumbers Discussion ofsignificantlyenrichedGOtermsidentifiedbyGOrillaintotal; normal-weight values indicate the numbers of significant GO terms following redundancy reduction by REViGO. See Our results yield insight into the transcriptional ‘‘Materialsandmethods’’sectionfordetails.Valuesinregions responseofplants toapicaldamage, usingthe model ofoverlapindicatethenumbersofGOtermsincommonamong species Arabidopsis thaliana. We demonstrate that genotypes. Values are additive for each genotype (e.g., there were 21 GO terms significantly enriched with up-regulated Columbia-4,Landsbergerecta,andtheRILgenerated genesforCol-4intotal,withthreeremainingafterredundancy from a cross between these two accessions differ in reduction). See Supplementary material 4 fortheidentitiesof theirregulationofgenesassociatedwiththeresponse thesesignificantGOterms tostimuli,cellwallbiosynthesis,andavarietyofother widevarietyofprocesses,includingxylanbiosynthe- processes. The most striking results, however, are sis, transition metal ion transport, and cellular differences between the genotypes intheir regulation responsetoironionstarvation(Fig. 2;Supplementary of secondary defensive chemistry that are positively material 4). Of the 38 down-regulated GO terms, 16 correlated with their compensatory performances. In remained after redundancy reduction, with many the context of these three genotypes, our results involved in growth and developmental signaling indicatethattoleranceanddefensearenotnecessarily (e.g., ethylene biosynthesis and phosphorelay signal alternative responses to the removal of apical domi- transduction) and the response to the environment nance since one genotype appeared to employ both (including ‘‘response to wounding,’’ ‘‘response to strategieswhiletheothertwoemployedneither.While mechanical stimulus,’’ ‘‘response to external stimu- thegeneralitiesofourresultsremainunassessedacross lus,’’etc.;Fig. 2;Supplementarymaterial4). populations and species, our methods were sufficient to produce robust, reliable transcriptomic data that DifferentialregulationofGOtermsofaprioriinterest provide the basis for conclusions for these three commonly studied genotypes that could be more GO terms selected a priori generally fall into two thoroughly investigated in future genetic and natural groups:thebiosynthesisofsecondarymetabolites(i.e., fieldstudies.Together,thisstudycontributesthefirst chemicaldefenses;Fig. 3a)andprocessespredictedto transcriptomic data counter to the presumed tradeoff influence compensatory performance (the pentose between tolerance and defense, as well as transcrip- phosphate shunt and endoreduplication; Table 1; tomic characterization of the plant regrowth process Fig. 3b). Overall, Columbia-4 significantly up-regu- broadly,addinganewperspective tothisconsequen- latedsecondarymetabolitebiosynthesiswhileLands- tialissueinplantdamagedynamics. berg erecta significantly down-regulated this process The A. thaliana genotype Columbia-4 experienced (Fig. 3a). Specifically, Columbia-4 significantly up- no significant change in attributes of fitness when regulated the biosynthesis of glucosinolates and damaged relative to when undamaged (i.e., it equally phenylpropanoids with no significant changes in the compensated). Col-4’s complete tolerance of damage regulation of phytoalexins or terpenoids. Landsberg correspondedwiththesignificantup-regulationofgenes erecta significantly down-regulated glucosinolates involvedinsecondarymetabolism,andparticularlythe 123 PlantEcol a Secondary metabolite Terpenoid biosynthetic Glucosinolate Phenylpropanoid biosynthetic process process biosynthetic process biosynthetic process 3.0 14 20 * * * m er 2.5 12 * 30 t O G * n 15 2.0 10 s i ne 8 20 e 1.5 g 10 * f 6 o e 1.0 g a 4 10 t n 5 * e c 0.5 r 2 e P 0 0.0 0 0 C ol-4LerC-0S 1936 C ol-4LerC-0S 1936 C ol-4LerC-0S 1936 C ol-4LerC-0S 1936 Phytoalexin b Pentose phosphate DNA biosynthetic process shunt endoreduplication 30 1.2 1.0 m m O ter 25 O ter 1.0 0.8 G G n 20 n 0.8 s i s i 0.6 e e n n e 15 e 0.6 g g of of 0.4 e 10 e 0.4 g g a a nt nt e e 0.2 c 5 c 0.2 r r e e P P 0 0.0 0.0 C ol-4LerC-0S 1936 C ol-4LerC-0S 1936 C ol-4LerC-0S 1936 Fig.3 Differential gene regulation of a priori gene ontology regulated genes in the GO term differs significantly (exact termsofinterest.Geneontologytermsincludeathebiosynthesis binomialtest,p\0.05;seeSupplementarymaterials1,2,and3 ofsecondarymetabolitesandbthepentosephosphateshuntand for full results). Black shading up-regulated, gray shading endoreduplication. Asterisks percentage of up- and down- down-regulated production of phenylpropanoids and glucosinolates. chemistry responses relative to its parents, including Landsbergerecta,whichsufferedreducedfitnesswhen significant down-regulation of glucosinolates, the pri- damaged relative to when undamaged (i.e., it under- mary anti-herbivore compounds produced by compensated),displayedsignificantdown-regulationof A.thaliana(Kliebenstein2004;D’AuriaandGershen- generalized secondary chemistry and glucosinolate zon2005;TzinandGalili2010). biosynthesis. The Col-4 9 Ler-0 RIL CS1936 also Theory and historical perspective often consider undercompensated and displayed a mix of defensive tolerance and defense to be alternative strategies to 123 PlantEcol damage based on the necessity of allocating limited non-defended tissues are preferred by their animal resourcesandapresumednegativegeneticcorrelation herbivores(Coley1983;Berenbaumetal.1986;Steele between the strategies. Across the literature, there is et al. 1993; Newman et al. 1996; Shroff et al. 2008; no conclusive evidence for the generality of this Stowe 2013). This trait prevents damage to valuable tradeoff, however, and in fact there is growing structures,andisthusaformof‘‘defense,’’yetitalso evidencesupportingtheco-occurrenceofthesestrate- allows plants of these species to tolerate damage of gies. For example, classical resource availability other structures since there is no evident cost to theory predicts that limited resources would promote reproduction(Stowe2013). theinvestmentofaplantintoanti-herbivoredefenses Becausethecorrelationbetweencompensationand given the plant’s likely impaired growth rate, and the regulation of secondary chemistry is present at subsequentlyitsimpairedabilitytorapidlyregrowand both extremes in this study (i.e., we observed a maintainfitnessfollowingdamage(Coleyetal.1985). genotype that invested in both, as well as genotypes However, mathematical models suggest that plant thatinvestedinneither,atleastintermsoftheprimary tolerancecouldactuallybemaximizedwhenresources anti-herbivore compounds; Kliebenstein 2004; are limiting since growth rates are reduced for D’Auria and Gershenzon 2005; Tzin and Galili undamaged plants as well, allowing damaged plants 2010), a genetic pathway that influences both could with multiple stems to match undamaged plants in beunderlyingtheresistancestrategiesoftheseplants. total biomass and fitness production (Hillbert et al. Ourpriorstudieshave indicated thattheinduction of 1981;AlwardandJoern1993).Clearlytheorydoesnot the pentose phosphate pathway (PPP) and endoredu- provideadequategeneralizationsfortheemployment plicationfollowingapicaldamageisimportantforthe of tolerance versus defense strategies, and empirical compensatoryperformanceofA.thalianaplants,with evidence often supports mixed-strategy hypotheses. consistent differences in the responses of these Specifically, numerous studies have discovered that pathways between genotypes that differ in their genotypesofmanyspeciesgenerallydisplayamixof compensation(ScholesandPaige2011,2014;Scholes resistance strategies and that these strategies are thus et al. 2013; Siddappaji et al. 2013). Although we did notaproductofdiversifying selection (i.e.,agenetic not observe a concurrent regulatory response of the tradeoff between strategies; Leimu and Koricheva PPP or endoreduplication, this does not necessarily 2006;Nu´n˜ez-Farfa´netal.2007;CarmonaandFornini indicate that these processes are not integrated with 2013; Turley et al. 2013). For example, studies of secondarymetabolism,aswetheorize.Becausewedid tolerance (measures of fitness) versus defense (tri- not observe overcompensation in Columbia-4, nor chome and glucosinolate production) in A. thaliana significant changes in ploidy (though trends were in revealed that natural populations exhibited mixed the expected directions) as in our previous studies resistance strategies due to selection for the mainte- (Scholes and Paige 2011, 2014; Scholes et al. 2013; nance of both traits, and thus no negative genetic Siddappaji et al. 2013), a primary constraint may be correlation exists for these plants (Mauricio et al. variation in the year-to-year environment to which 1997). In fact, mixed resistance strategies may be correlations in pathway regulation might have been expected when considering the concurrent selection weak or absent. Further, previous studies have pressures applied by a diverse herbivore community observed a delay in the induction of the PPP and and/or a genetically diverse plant population, as are endoreduplicationrangingfromafewtoseveraldays commonly present in natural systems (Fornoni et al. after clipping (Scholes and Paige 2011; Siddappaji 2004;Agrawal2011;CarmonaandFornini2013).The et al. 2013), yet these pathways clearly influence growing number of similar examples has even fitness compensation as measured at senescence. spawned a revision of the semantics surrounding the While we used all available data to design our discussion of plant resistance strategies, with Stowe experimental protocol to maximize our detection of (2013)expressingthatthedistinctionbetween defen- differential gene regulation, it is also possible that siveandtolerancetraitsisnotusefulinpracticedueto effectsofclippingongeneregulationcouldhavebeen their often interdependent phenotypic expression and missed if the timing of sampling did not correspond evolution.Forexample,manyspeciesallocatedefen- with differential expression. There thus remains a sivechemicalstovaluabletissues,likeseeds,suchthat theoretical framework by which the PPP supports 123
Description: