Metabolite modification in oxidative stress responses: a case study of two defense hormones ‡

ABSTRACT


Introduction
Localized or generalized shifts in cellular redox state linked to increased production or availability of reactive oxygen species (ROS) underpin many responses to environmental challenge and also play a key role in coordinating plant developmental processes [1][2][3].The biological impact of such oxidative stress is strongly linked to changes in gene expression that can be observed at the transcriptomic levels [4].
Such changes in gene expression also feed through to other levels, producing altered metabolite signatures [5].Indeed, it is noteworthy that a significant proportion of genes that are strongly induced by oxidative stress encode enzymes involved in metabolite modification, many of which are expressed at relatively low levels in the absence of stress [6].As well as reductases and dehydrogenases involved in maintaining cell redox status in the face of increased ROS production, such induced genes may encode enzymes that perform metabolite repair or that remove reactive organic molecules produced secondarily in response to ROS [7,8].Additionally, conjugases and related enzymes are known to be important in controlling the signaling strength of stress-induced compounds with hormonal properties, such as salicylic acid (SA) and jasmonic acid (JA) [9][10][11][12][13][14][15][16].
Glutathione S-transferases (GSTs), cytochromes P450 (CYPs), and UDP-glycosyl transferases (UGTs) are three of the major gene families encoding enzymes that can alter the biological activity of organic compounds by introducing new functional groups.GSTs act notably to conjugate electrophilic metabolites to glutathione, CYPs generally catalyze oxidation reactions, and UGTs introduce sugar groups (such as xylose or glucose) onto hydroxyl groups that pre-exist or, in many cases, have been introduced by the prior action of CYPs or other enzymes.At least some members of all three enzyme groups are involved in detoxification of herbicides and other potentially deleterious organic molecules of exogenous origin [17,18], but specific members also act in many reactions involved in endogenous secondary metabolism [10,13,19].Certain GSTs were among the first genes found to be induced in early studies of H2O2 signaling [20], and are useful as markers of H2O2 signaling strength [4][5][6]21].Similarly, specific UGTs are known to be associated with oxidative stress conditions [6,22], and some of them have been shown to play important roles in controlling phytohormone signaling in response to stress [23].
However, much progress remains to be made in identifying the functions of many of the genes in these families in oxidative stress conditions.
Our aim in this work was, first, to exploit three cat2 transcriptomic datasets to generate a robust account of how the expression of major metabolite-modifying gene families responds to enhanced intracellular H2O2.We then sought to investigate in detail how gene expression associated with SA and JA synthesis and metabolism are impacted by oxidative stress.To explore the potential functional impact of the changes in gene expression, we used parallel metabolite profiling to determine how oxidative stress affects compounds such as SA and JA, as well as their derivatives.Given the potential biological activity of dihydroxybenzoic acids (DHBA) [30,31], we included these compounds alongside SA (2-hydroxybenzoic acid) in the metabolite analysis.By comparison of transcript and metabolite data, we identify key features of the oxidative stress signatures linked to signalling via benzoic acids and jasmonates.

Plant material
Arabidopsis Col-0 and the cat2 mutant were grown in 16h days at an irradiance of 200 µmol.m - s -1 and 22°C/20°C (day/night).The cat2 mutant is a T-DNA insertion line with about 20% wild-type leaf catalase activity.It was initially characterized by Queval et al. [6] and subsequently used by various groups as a system in which to study the impact of oxidative stress resulting from endogenous H2O2 production by a physiologically relevant pathway (photorespiration) [24].Leaf samples were taken after 21 days of growth in the middle of the photoperiod for transcript and metabolite analysis.All the samples were immediately frozen in liquid nitrogen and stored at -80°C until analysis.

Transcript measurements by quantitative RT-PCR
Total RNA was isolated in triplicates from leaves using Spectrum™ Plant Total RNA Kit (SIGMA, Germany), and samples were treated with DNase during RNA purification.RNA concentrations were determined using Nanodrop and 1 µg of RNA was used to perform reverse transcription using the qScript cDNA Synthesis Kit (Quantabio).Selected genes were chosen for qPCR analysis according to [32].Two or three independent experiments were performed, each with three biological replicates and each with at least 12 plants per genotype.qPCR was performed in technical triplicates and data were normalized relative to two reference genes, ACTIN2 and UBIQUITIN7.For each transcript, differences between cat2 and Col-0 were assessed for statistical significance by t-test of the three biological replicates in each experiment.

Transcriptome analyses
Transcriptome profiling of cat2 and Col-0 plants were performed in three independent experiments using the growth conditions described above, with three biological replicates in each case.Transcript abundance data were retrieved from one microarray [21] and two RNA sequencing datasets (in-house datasets).For the RNA sequencing experiments, library preparation and sequencing were performed at the VIB Nucleomics Core (Leuven, Belgium), using TruSeq Stranded mRNA Library Preparation Kit (Illumina).Samples were sequenced on IlluminaNextseq 500 (75-bp single-end reads).Reads were aligned to the Arabidopsis genome by STAR (v2.5.2b) [33] using the Araport11 annotation [34].The number of reads per gene was quantified with the featureCounts function as implemented in the Subread package v1.6.2 [35].

Transcriptomic data mining
To generate the list of genes for CYP and GST subfamilies, the keyword research in Thalemine (bar.utoronto.ca/thalemine/)was used.For UGTs, a search in Uniprot (www.uniprot.org)for reviewed "IPR002213" Interpro accession in A. thaliana was performed.The gene annotations were crossvalidated using Phytozome (phytozome-next.jgi.doe.gov;Araport11).To visualize the datasets, dendrograms, heatmaps and PCA plots were generated using the average of Z score normalized data of each dataset.For CYPs, GSTs, and UGTs, only genes for which data are available in the three datasets were considered.To generate the dendrograms, the "ward.D2" method of the R "stats" package was used with a Pearson's correlation coefficient-based distance matrix.
Genes associated with salicylate and jasmonate synthesis and metabolism were selected based on current knowledge of the different pathways (Supplemental Figures S1-S6).The mean expression and the standard error were calculated from the three biological repeats within each experiment.Fold change (cat2/Col-0) was also calculated and statistical significance between cat2 versus Col-0 was determined using the t-test.

Metabolite analysis
Metabolites were measured in cat2 and Col-0 samples obtained in exactly the same conditions as for the transcript analyses, using a protocol described in detail by Lelarge-Trouverie et al. [36].Briefly, 200 mg snap-frozen leaf tissue was ground in liquid nitrogen then extracted into 800 µL of a solution of 80% acetone, 19% water, 1% acetic acid containing internal standards (d4-SA and DHJA) at a concentration of 4 µM each.Following sonication and vigorous shaking, extracts were centrifuged at 12 000 g and 4 °C for 15 mins.Supernatants were set aside and pellets extracted twice using 400 µL of the above extraction solution.For each extract, the three supernatants were pooled and dried using a TurboVap.Dried samples were then resuspended in 100µL water and transferred into glass vials with inserts, filtered, and submitted to analysis by UPLC-MS.Sample solutions were injected onto a reversed phase column (Kinetex C18, 1.7U, 100A, 2.1 × 100 mm, Phenomenex) on an ACQUITY Ultra Performance LC™ System (Waters, Milford, MA, USA) linked to a Bruker MicrOTOF II time of flight mass spectrometer (Bruker Daltonik GmbH, Germany) equipped with an electrospray ionization (ESI) source operating in negative mode.Mobile phases were water (eluent A) and 0.1% acetic acid in acetonitrile (eluent B).The injection volume was 5µL and chromatographic conditions were as described in [36].Eluting compounds were detected from m/z 50 to 800.Data were acquired and processed using the MicrOTOF-Control, DataAnalysis (version 4.0 SP 1) and QuantAnalysis (version 2.0 SP 1) software packages.Most compounds were identified by comparison of retention times with standards and by searching for ions according to the mass spectra of the standards.Identity was confirmed using exact mass and isotopic profile in combination.For compounds for which standards were not available, identification was done according to theoretical mass spectra and confirmed as above.QuantAnalysis 2.0 was applied to integrate extracted ion chromatograms of corresponding quantifier ions.For SA, SA glucosides, and JA, values were adjusted for internal standard responses (d4-SA for SA and other benzoic acids, DHJA for jasmonates and related compounds) and absolute quantification on a fresh weight basis was performed by comparison with curves generated from authentic standards.For other compounds, relative quantification was performed using quantifier ions and expressed on an area/FW basis, using standard curves generated with authentic compounds where these were available.

Results
We exploited one microarray and two RNAseq transcriptomic analyses of cat2 and Col-0 growing in standard controlled conditions (moderate light and temperature).PCA and clustering analysis performed on the entire datasets clearly separated the two genotypes on the basis of their transcriptomes, with all nine cat2 replicates clustering apart from the Col-0 controls (Supplemental Figure S7).

Transcriptomic profiles of major metabolite modification families in response to oxidative stress
Previous transcriptomic analyses of catalase-deficient Arabidopsis lines have shown that genes encoding enzymes involved in metabolite modification can be strongly up-regulated by oxidative stress [4,6].Such genes notably include members of the GST, UGT, and CYP families, some of which have been shown to play important roles in response to stress and/or to be among the best molecular markers of the intensity of oxidative stress [5,22,23,37].To provide a comprehensive overview of the response of these three gene families to oxidative stress linked to enhanced intracellular H2O2, we compared how they were affected by the cat2 mutation in the three transcriptomic experiments.Data were retrieved for annotated CYPs, GSTs, and UGTs.A complete list is shown in Supplemental Table S1.Genes for which transcript data were available in all three experiments (146 CYPs, 52 GSTs, 78 UGTs) were analyzed for their expression patterns.Hierarchical clustering of these genes allowed a clear visual separation between the genotypes in all experiments, with all nine cat2 replicates clustering apart from the nine Col-0 replicates (Figure 1).Some clustering of replicates within each experiment was also observed but this contributed much less to the separation than the genotype, demonstrating that a marked, reproducible effect of oxidative stress can be observed by analysis of the transcripts of these gene families.
The response of the different genes within these families is presented in the heatmap shown in Figure 2 (mean values for each genotype within each experiment).While this analysis revealed a specificity of response, it also confirmed that all three families contain a substantial number of genes that are strongly and consistently induced by intracellular H2O2.Reproducible induction across the three experiments was observed for about one third of the CYPs, more than half of the GSTs, and almost one third of the UGTs.Much smaller numbers of each family showed the opposite response (decreased expression in cat2 in all three experiments).The remaining genes showed some intra-genotype variability between the experiments that made it more difficult to conclude on any effects of oxidative stress.

Expression of genes associated with salicylate and jasmonate synthesis and modification
Previous analyses have shown that salicylic acid and some related genes accumulate strongly in response to oxidative stress in cat2, and that this mutation also interacts with the jasmonate signaling pathway [5,[25][26][27][28][29].As shown in Figure 2, several CYPs and UGTs are known to be involved in SA or JA metabolism.We therefore investigated how these pathways might be regulated and fine-tuned in response to oxidative stress by a targeted expression analysis of genes associated with SA and JA synthesis and modification, allied with metabolomics to assess the possible impact of altered gene expression on key compounds.
First, we checked that the SA and JA signaling pathways were activated by oxidative stress using specific downstream marker transcripts measured by qRT-PCR.This confirmed that alongside the H2O2 marker transcript, GSTU24, SA-dependent pathogenesis-related (PR) genes were markedly induced in cat2, as were the JA marker genes JAZ1 and VSP2 (Supplemental Table S2).These data are consistent with previous findings that both pathways are activated together in cat2 [26][27][28].To elucidate the impact of oxidative stress on the responses of genes involved in synthesis and metabolism of the two phytohormones and related compounds, we mined the same three transcriptomic datasets.This search was informed by available databases and current concepts in Arabidopsis (Supplemental Figures S1, S2, S4, S5) according to which the isochorismate pathway is the main route of stress-induced SA synthesis [38] but a phenylalanine-dependent pathway might also contribute, whereas JA is synthesized from linolenic acid via OPDA [39].Genes were selected based on these concepts and current knowledge of the enzymes involved in modification of salicylates and jasmonates (Supplemental Figures S3 and S6), and applying an additional filter that transcripts must be detected in at least two out of three experiments.This gave a total of 44 annotated genes associated with SA and benzoic acid biosynthesis and modification (Supplemental Table S3) while, for jasmonates and related compounds, we detected 48 genes (Supplemental Table S4).
To explore the impact of oxidative stress on these gene sets, transcripts that were detected in all three transcriptomic experiments were used in a clustering analysis (Figures 3 and 4).This involved 41 of 44 genes assigned to SA and benzoic acid pathways and 43 of 48 genes associated with jasmonates.Figure 3A shows that using both sets of genes resulted in a clear separation of cat2 and Col-0 samples in all three experiments, and this was confirmed in a hierarchical clustering analysis (Figure 3B).A heatmap of individual genes also revealed that many transcripts showed a clear visual genotype-dependent response for both SA-and JA-related pathways (Figure 4).Just over half of SA-related genes and just under half of the JA-associated genes showed stronger expression in cat2 compared to Col-0 across the three experiments (Figure 4).Analysis of cat2/Col-0 differences by t-test in each experiment is documented in Supplemental Tables S3 and S4, which show that selected genes in the pathways were significantly altered in their expression.ICS1, a key enzyme in SA synthesis, was more strongly expressed in cat2 in all three transcriptomic experiments, although with a slightly higher P value than the chosen threshold of 0.05 in the microarray experiment.The clear induction of ICS1 by oxidative stress was confirmed by qRT-PCR in two independent experiments using parallel biological samples, and accompanied in the transcriptomic experiments by marked induction of EDS5 and PBS3 (Supplemental Table S3), which encode a transporter and an enzyme downstream of ICS1 in isochorismate-dependent SA synthesis, respectively [40].Genes involved in converting chorismate to phenylalanine, and the latter to downstream compounds, were also induced in cat2 although again, this effect was clearer for the two RNAseq datasets than the microarray experiment.Alongside these effects, several genes involved in JA synthesis were induced in cat2, including isoforms of lipoxygenase, allene oxide cyclase, and acyl CoA oxidase, although few of these genes were induced in all experiments (Supplemental Table S4).Interestingly, allene oxide synthase was more weakly expressed in cat2, although effects were significant only in two experiments.Globally, pathways of both SA and JA synthesis were induced by intracellular H2O2, with perhaps the strongest and most consistent effect observed for the isochorismate-dependent production of SA.

Profiling of SA-and JA-related metabolites in response to oxidative stress
To assess whether these changes in gene expression are associated with an impact on metabolite accumulation, we used a recently developed LC-MS method to measure SA, JA, and derived compounds [36].About twenty compounds could be detected and quantified on an absolute or relative basis (Supplemental Table S5).The structural diversity of the different compounds is summarized in Figure 5.
First, we quantified data for free SA and the two glucosylated forms (SAG, SGE) in Col-0 and cat2 by this method.Three independent experiments revealed an increase in all three forms in cat2.Quantitively, SAG was consistently the predominant form in cat2 (more than 5-fold free SA levels) while SGE was the least abundant form in both backgrounds (Figure 6).The schemes above the panels indicate enzymes known to be competent for SA glucosylation, together with a summary of their transcriptomic responses to oxidative stress (for quantitative transcript data, see Supplemental Table S3).Analysis of other hydroxybenzoic acids revealed that the SA isomers, mHBA and pHBA, were also significantly increased in cat2 (Figure 7A).Several modified forms were detected in which SA was hydroxylated and, in some cases, also glycosylated.Among these, the most marked effect of oxidative stress was observed for 2,3DHBA, 2,5DHBA-2G, and 2,3DHBA-3X (Figure 7A).Based on the transcriptomic analysis (Supplemental Table S3) and available knowledge, Figure 7B illustrates the expression pattern of candidate genes involved in the production of these compounds in response to oxidative stress.Both SA hydroxylases were strongly and consistently induced (Figure 7B).Of the nine UGTs reported to be able to glycosylate at least one of the DHBAs, the most marked induction was observed for UGTs of the 73 subclass (UGT73B3, B4, B5).The strong induction of UGT73B3 and UGT73B5, as well as S3H and S5H, was confirmed in two independent qRT-PCR analyses.
LC-MS analysis revealed similar levels of free JA and the active conjugate, JA-Ile, in Col-0 and cat2 (Figure 8A).However, several modified forms of JA were markedly higher in cat2, as was OPDA.Two hydroxylated forms of JA, 11-OH-JA and 12-OH-JA, were detected in leaf extracts: the only one that was significantly increased in cat2 was the latter (Figure 8A).In contrast to JA-Ile itself, the modified forms of the conjugate (12-OH-JA-Ile and 12-COOH-JA-Ile) were increased by oxidative stress.The most marked fold increases triggered by H2O2 were observed for JAG, the glucosylated form of free JA, and two compounds derived from 12-OH-JA: these were 12-SJA and 12-O-Glc-JA, a sulfated and a glucosylated derivative, respectively (Figure 8A and B). Figure 8B attempts to integrate the transcriptome data (Supplemental Table S4) for potentially significant enzymes involved in the production of accumulated JA metabolites in cat2.This comparison revealed strong but sometimes specific induction of oxidases, sulfotransferases, hydrolases, conjugases, and CYPs (Figure 8B).

Discussion
Our aims in this study were, first, to provide an inventory of the oxidative stress response in major gene families encoding metabolite modification enzymes and, second, to examine the roles of genes in modifying two important defense phytohormones that are affected by this stress.We chose GSTs because of the strong response of this family to redox perturbation, as well as UGTs and CYPs, since these enzymes catalyse oxidative and conjugation reactions, some of which can be important in defense.The first analysis revealed specific and consistent alteration in the expression of a large segment of each gene family in cat2 that is sufficient to distinguish between oxidatively stressed and unstressed plants (Figures 1 and 2).Most of these distinguishing genes were induced while few were repressed, exemplifying the crucial role that enhancing metabolite modification capacity may play in the oxidative stress response [4,6,23].Similarly, it is evident that plants undergoing oxidative stress produced by intracellular H2O2 can be identified by a targeted transcriptomic analysis of genes selected for their roles in either SA or JA synthesis and modification (Figures 3 and 4).Given the key roles of SA and JA in stress responses, we used a combined metabolomics and transcriptomics approach to focus on assessing the metabolic fate of these phytohormones specifically during oxidative stress.

Isochorismate and phenylalanine pathways of metabolite production
SA accumulation and downstream responses triggered by the cat2 mutation require isochorismate synthase 1 (ICS1).This has been shown by abolition of SA-related gene expression and lesion phenotypes in a cat2 sid2 mutant, in which the sid2 (SA induction-deficient 2) mutation in ICS1 genetically blocks SA synthesis through this pathway [25,29].This is consistent with the marked transcriptional up-regulation of the isochorismate pathway in cat2 reported here in several datasets.Although ICS1 is required for marked accumulation of SA in response to oxidative stress, we cannot exclude that the phenylalanine pathway also makes some contribution alongside the isochorismate pathway when the latter is activated.Indeed, the present datasets also show up-regulation of the phenylalanine synthesis pathway and downstream enzymes involved in production of phenylpropanoids, consistent with accumulation of Phe and trans-cinnamate in cat2 observed through GC-MS analyses [5,27].Given the rich array of compounds produced via the phenylalanine pathway, induction of this pathway may be related to production of defensive coumarins such as scopoletin [22] and SA isomers (mHBA and pHBA), which were observed to accumulate alongside SA in this study (Figure 7A).Several reactions can produce mHBA and pHBA in various organisms but their metabolic origin in plants remains to be clearly established.Formation of pHBA from coumaroyl CoA may be involved in the biosynthesis of quinones, when it can be formed from coumaroyl CoA or kaempferol [41,42].

SA glucosyltransferases in response to oxidative stress: evidence for differential contributions
Glucosylation is an important process regulating SA activity and turnover, and four UGTs have been identified that may convert SA to SAG or SGE.Available data suggest that UGT74F1 is a major producer of SAG while SGE can be produced by UGT74F2 but not UGT74F1 [10,19,43].Other enzymes that may be involved include UGT75B1, which has a low activity for production of SGE but does not produce SAG [19] and UGT76B1, which may contribute to the production of SAG, among other compounds [44].
Accumulation of the two SA-glucosides in response to oxidative stress was accompanied by differential induction of these four candidates at the transcript level (Figure 6), with UGT76B1 in particular being markedly up-regulated (Supplemental Table S3).Based on these observations, the data perhaps imply that the main players producing SA-glucosides in response to oxidative stress-driven accumulation of SA may be UGT76B1 (rather than or addition to UGT74F1) for SAG, and UGT74F2 for SGE.However, a contribution of UGT75B1 to SGE accumulation cannot be discounted, since this gene was 5 to 12 times more highly expressed in cat2, although induction was only significant in one of the two datasets in which transcripts could be quantified (Supplemental Table S3).

Dihydrobenzoic acids and their glycosides in the oxidative stress response
Alongside hydroxybenzoic acids and conjugates (SA, SAG, SGE, mHBA, pHBA), several dihydroxybenzoic acids (DHBAs) and their glycosides accumulated in response to intracellular H2O2 (Figure 7A).Literature data suggest that the most markedly accumulated DHBAs are formed by hydroxylation of SA through the action of two Fe-dependent monooxygenases.The first, SA 3-hydroxylase (S3H), is active against SA [45] and was induced in cat2 in all experiments (albeit just above the significance threshold in one RNAseq dataset).S3H induction correlated with marked accumulation of 2,3-DHBA (>150-fold higher in cat2) and its xylose glycoside (2,3DHBA-3X: >20-fold higher in cat2).The second hydroxylase (S5H) can convert SA to 2,5-DHBA [46].Although technical limitations prevented the detection of free 2,5-DHBA in this study, the accumulation of several conjugates of this molecule may have required SA hydroxylation at the 5' position by S5H, consistent with the marked induction of the corresponding gene in cat2.
Our analysis detected a total of four DHBA-glycosides that accumulated in response to oxidative stress (two xylose derivatives and two glucose derivatives).UGTs have been described that can produce three of these from the two free forms.In the case of 2,5DHBA-5G, which only accumulated moderately in cat2, several UGTs that may be involved were induced while two were repressed.It is difficult to draw firm conclusions from these patterns.First, several of the induced UGTs are also implicated in the production of compounds that we were not able to detect in this study.For example, the free form of 2,4-DHBA was somewhat accumulated in cat2, and several induced UGTs are competent to glycosylate this compound, but we were not able to detect any such derivatives.Second, the most consistently induced UGTs that are potentially involved in producing 2,5DHBA-5G were UGT73B3 and UGT73B5, which have been previously implicated in responses to both biotic and oxidative stress [22].While UGT73B3 has been shown to be competent for 2,5DHBA-5G production, evidence is as yet lacking that this is the case for UGT73B5.Additionally, UGT76D1 was induced in cat2 in two out of three transcriptomic experiments.This gene has been shown to be induced by SA and to be involved in lesion formation and PR gene expression [31].Moreover, the recombinant UGT76D1 can catalyze formation of glucosides and xylosides of both 2,3DHBA and 2,5DHBA [31].Interestingly, both 2,5DHBA-5G and 2,5DHBA-5X showed a similar modest accumulation in cat2 (about 2-fold), possibly reflecting the involvement of UGT76D1, although 2,3DHBA-3X accumulated much more strongly (Figure 7A).
In terms of fold change, 2,5DHBA-2G was the most strongly accumulated glycoside in cat2, and indeed has been described as a major product of SA degradation [10].Based on what is known about their substrate affinities [19], none of the UGTs that are induced in cat2 are good candidates for producing 2,5DHBA-2G by glucosylation of 2,5DHBA.Rather, our gene expression data are consistent with a previous suggestion that this compound is generated by hydroxylation of SAG [10].In this case, 2,5DHBA-2G would be formed from free SA by first, glucosylation and second, hydroxylation.Indeed, the marked and consistent up-regulation of S5H is consistent with a role for this enzyme in hydroxylating both free SA to 2,5DHBA and SAG to 2,5DHBA-2G in cat2.

JA inactivation during oxidative stress
Although oxidative stress in cat2 did not lead to accumulation of either JA itself or the active form, JA-Ile (Figure 8A), our gene expression and metabolite data are consistent with some activation of JA biosynthesis.In addition to up-regulation of JA marker genes (Supplemental Table S2), genes implicated in JA synthesis that were induced in at least two experiments included LOX4, AOC3, two transporters involved in OPDA movement from the chloroplast to the peroxisome, OPR3, MFP2 and KAT2, as well as JAT2, encoding an ABC transporter that may be involved in exporting JA from the peroxisome [47][48][49][50][51][52][53][54].Of the six Arabidopsis genes encoding acyl-CoA oxidases, ACX1 and ACX5 are the two most closely implicated in the JA synthesis pathways [55].Several isoforms showed significant induction in cat2, including ACX1, ACX2, and ACX3 in all three transcriptomic experiments (Table S4).Interestingly, ACX2 and ACX3 have been described to interact with CAT2 [56].
Our metabolite and transcript profiling data suggest that any stimulation of JA synthesis was accompanied by activation of JA-modifying pathways, reflected in accumulation of specific oxidized, sulfated, and glucosylated forms (Figure 8A).As shown in the scheme in Figure 8B, oxidative stress triggered accumulation of metabolites that are formed from both JA and JA-Ile, suggesting activation of enzymes that can produce the latter from the former.JAR1 is a well characterized aminotransferase competent for this reaction [9].The level of JAR1 expression has recently been shown to affect drought resistance [57] and a double cat2 jar1 mutant presented an enhanced lesion phenotype compared to cat2 [29].However, no consistent up-regulation of this gene was observed in our experiments.By contrast, transcripts for another recently described JA-Ile synthetase, DFL2 [58], were consistently more abundant in cat2, albeit with a modest fold change (Supplemental Table S4).Interestingly, three hydrolases able to convert JA-Ile back to JA were up-regulated in oxidative stress conditions (Figure 8B), although it should be noted that all the corresponding proteins have additional metabolic functions [59].
Accumulation of a hydroxylated form of JA (12-OH-JA) in cat2 was correlated with induction of three JOX, ascorbate-and iron-dependent oxygenases that are competent for this reaction [14,60].Formation of this compound may already allow a partial inactivation of JA signaling [11] and further modification by sulfation of the hydroxylated form to 12-SJA or glucosylation to 12-O-Glc-JA can occur.Strong induction of a specific gene encoding a sulfotransferase (ST2A) alongside accumulation of 12-SJA is consistent with a role for this isoform during oxidative stress (Figure 8).Indeed, of the two Arabidopsis sulfotransferases (ST2A and ST2B), only ST2A is competent for the sulfotransferase reaction [61] and this gene was the only one of the two that was responsive at the transcript level to oxidative stress in our study.
Although the glucosylated form of 12-OH-JA also accumulated strongly in cat2, this effect was not accompanied by up-regulation of the three UGTs that have been described to convert 12-OH-JA to 12-O-Glc-JA [62].Similarly, strong accumulation of JAG was also observed (Figure 8A), but we were not able to identify an induced gene that may be responsible for glucosylating JA.It is not clear how these glucosides are formed in oxidative stress conditions, but it is possible that the UGT(s) involved are among the cluster that is strongly up-regulated in cat2 (Figure 2C).
Like JA, JA-Ile can also be hydroxylated at the C12 position, a reaction catalyzed by CYPs rather than JOXs [63,64], with further oxidation converting 12-OH-JA-Ile to 12-COOH-JA-Ile (Figure 8B).Approximately four-fold accumulation of both these JA-Ile derivatives was observed in cat2, an effect accompanied by induction of a specific CYP (Figure 8B).This isoform, CYP94C1, is at the centre of a group of CYPs that show consistent and marked induction in response to intracellular H2O2 (Figure 2A).The role of JA-Ile oxidation is not clearly established: while some studies suggest that this reaction attenuates jasmonate signaling [13], others have concluded that 12-OH-JA-Ile may function through COI1 in a similar way to JA-Ile [65].Nevertheless, our transcript and metabolite data for the JA pathway are consistent with induction of a network of genes to limit accumulation of any JA-Ile that results from the response to oxidative stress.This network appears to involve enzymes that either hydrolyze JA-Ile or that metabolize JA and JA-Ile through oxidizing and conjugating reactions that act on the terminal carbon of the JA moiety.

Figure 1 .
Figure 1.Experiment and genotype distribution for genes associated with CYPs (A), GSTs (B) and UGTs Figure 1.Experiment and genotype distribution for genes associated with CYPs (A), GSTs (B) and UGTs (C).The dendrogram is based on normalized values (Z score) of gene expression.Hierarchical clustering was done using Ward linkage and Pearson distance.Col-0 and cat2 genotypes are coloured in green and purple, respectively.The experiments are indicated by color shade from dark, intermediate to light corresponding respectively to Microarray, RNAseq1 and RNAseq2.Samples are labelled as Genotype_Replicate_Experiment.

Figure 2 .
Figure 2. Expression in Col-0 and cat2 of genes belonging to the CYP (A), GST (B) and UGT (C) families.The heatmaps represent the mean of Z score normalized gene expression using three biological repeats.Hierarchical clustering on genes and conditions was done using Ward linkage and Pearson distance.Conditions are labelled as Genotype_Experiment.Genes are labelled as AGI_SYMBOL.Genes labelled with a circle (•) and a star (*) are associated with salicylic acid and jasmonic acid, respectively.

Figure 3 .
Figure 3. Overview of distribution of gene expression associated with salicylic acid (left) and jasmonic acid (right) synthesis and modification.A, Principal component analysis.B, Dendrogram showing separation of biological replicates.Analyses were based on normalized values (Z score) of gene expression.Hierarchical clustering was done using Ward linkage and Pearson distance.Col-0 and cat2 genotypes are colored in green and purple, respectively.The experiments are indicated by color shade from dark, intermediate to light corresponding respectively to Microarray, RNAseq1 and RNAseq2.Samples are labelled as Genotype_Replicate_Experiment.

Figure 4 .
Figure 4. Heatmaps of individual genes involved in synthesis and modification of salicylic acid and related compounds (A) and jasmonic acid (B) in Col-0 and cat2.Means of Z score-normalized gene expression using three biological repeats are represented.Hierarchical clustering on genes and conditions was done using Ward linkage and Pearson distance.Conditions are labelled as Genotype_Experiment.Genes are labelled as AGI_SYMBOL.

Figure 5 .
Figure 5.Chemical structures of the compounds analyzed by LC-MS.Top, salicylates and related benzoic acids.Bottom, jasmonate-related compounds.Tables include definition of abbreviations and of sidegroups, represented by R1,R2, etc, based on the structures shown.For jasmonates, the structure of the precursor, OPDA, is also shown on the left.

Figure 6 .
Figure 6.Effect of intracellular oxidative stress on free SA and SA-glucosides.Top, free SA.Middle, SAglucoside.Bottom, SA-glucose ester.Black bars, Col-0.White bars, cat2.Samples were extracted and analyzed as described in the Materials and methods.Data are means ± SE of three (experiment 1) or six (experiments 2 and 3) biological repeats.*Significant difference from cat2 at P < 0.05.The schemes above the bottom two figures include names of genes encoding UGTs implicated in the two conversions, and how their transcripts respond to oxidative stress.A red colour with asterisks indicate significant induction of the gene in cat2 compared to Col-0 in *one, **two, or ***three transcriptomic experiments while a blue colour indicates no significant change (P < 0.05).

Figure 7 .
Figure 7. Response of benzoic acid synthesis and modification pathways to intracellular oxidative stress.A, Contents of benzoic acids and conjugates in Col-0 (black bars) and cat2 (white bars).For definition of abbreviations, see Figure 5, top.Absolute values have been adjusted to allow bars to be plotted on the same scale.DHBA values have been multiplied by 5 while 2,5DHBA-5G and 2,5DHBA-2G have been divided by 200 and 2,5DHBA-5X and 2,3DHBA-3X by 1000.Samples were extracted and analyzed as described in Materials and methods.Data are means ± SE of six biological repeats.*Significant difference from cat2 at P < 0.01.Numbers above the bars indicate cat2/Col-0 fold change.B, Scheme showing some of the measured compounds and the response of associated genes.The colour scheme indicates genes that are significantly induced (red) or repressed (green) in cat2 compared to Col-0 (P < 0.05) in *one, **two, or ***three independent experiments, each with three biological replicates.Blue question marks indicate conversions where no gene has yet been ascribed or where gene function has not yet been definitively established.

Figure 8 .
Figure 8. Response of jasmonate synthesis and modification to intracellular oxidative stress.A, Contents of jasmonates and related compounds in Col-0 (black bars) and cat2 (white bars).For definition of abbreviations, see Figure 5, bottom.Absolute values have been adjusted to allow bars to be plotted on the same scale.Areas for 12-OH-JA, 11-OH-JA and 12-COOH-JA-Ile have been multiplied by 2 while areas for OPDA, JAG, 12-O-Glc-JA, 12-OH-JA-Ile and JA-Ile have been multiplied by 10.Samples were extracted and analyzed as described in Materials and methods.Data are means ± SE of six biological repeats.*Significant difference from cat2 at P < 0.01.Numbers above the bars indicate cat2/Col-0 fold change.B, Scheme showing some of the measured compounds and response of associated genes.The colour scheme indicates genes that are significantly induced (red), significantly repressed (green), or unchanged (blue) in cat2 compared to Col-0 (P < 0.05) in *one, **two, or ***three independent experiments, each with three biological replicates.A blue question marks indicates that no gene has yet been definitively ascribed.