Silencing an N-Acyltransferase-Like Involved in Lignin Biosynthesis in Nicotiana attenuata Dramatically Alters Herbivory-Induced Phenolamide Metabolism

In a transcriptomic screen of Manduca sexta-induced N-acyltransferases in leaves of Nicotiana attenuata, we identified an N-acyltransferase gene sharing a high similarity with the tobacco lignin-biosynthetic hydroxycinnamoyl-CoA:shikimate/quinate hydroxycinnamoyl transferase (HCT) gene whose expression is controlled by MYB8, a transcription factor that regulates the production of phenylpropanoid polyamine conjugates (phenolamides, PAs). To evaluate the involvement of this HCT-like gene in lignin production as well as the resulting crosstalk with PA metabolism during insect herbivory, we transiently silenced (by VIGs) the expression of this gene and performed non-targeted (UHPLC-ESI/TOF-MS) metabolomics analyses. In agreement with a conserved function of N. attenuata HCT-like in lignin biogenesis, HCT-silenced plants developed weak, soft stems with greatly reduced lignin contents. Metabolic profiling demonstrated large shifts (up to 12% deregulation in total extracted ions in insect-attacked leaves) due to a large diversion of activated coumaric acid units into the production of developmentally and herbivory-induced coumaroyl-containing PAs (N′,N′′-dicoumaroylspermidine, N′,N′′-coumaroylputrescine, etc) and to minor increases in the most abundant free phenolics (chlorogenic and cryptochlorogenic acids), all without altering the production of well characterized herbivory-responsive caffeoyl- and feruloyl-based putrescine and spermidine PAs. These data are consistent with a strong metabolic tension, exacerbated during herbivory, over the allocation of coumaroyl-CoA units among lignin and unusual coumaroyl-containing PAs, and rule out a role for HCT-LIKE in tuning the herbivory-induced accumulation of other PAs. Additionally, these results are consistent with a role for lignification as an induced anti-herbivore defense.


Introduction
Plants as sessile organisms are exposed during their development to variable stress conditions which have shaped, from a macro-evolutionary point of view, the emergence of highly adaptive tolerance and resistance traits. The feeding damage of herbivores has provided a major evolutionary selective pressure that has sculpted all aspects of plant metabolism including the composition and size of their pools and the regulatory networks that determine fluxes [1][2][3]. When chewing leaves, insects elicit a burst of the plant hormone, jasmonic acid (JA); this signaling molecule mediates rapid changes in secondary metabolic pathways [4,5]. Native tobacco plants, Nicotiana attenuata, accumulate, as a consequence of JA signaling, large amounts of nicotine, a neurotoxic compound and a spectrum of phenolic amide conjugates (phenolamides, PAs), derived from the phenylpropanoid pathway, to defend against specialist and generalist chewing herbivores [6,7].
As is frequently observed in transcriptomic analyses, herbivoreinduced changes in a plant's metabolome result from specific reorganizations of metabolic pathways. However, how attacked cells rechannel metabolic fluxes towards the production of a specific spectrum of metabolites is frequently unknown. The underlying mechanisms by which plants perceive insect herbivory are known to be controlled by elicitors present in the oral secretions (OS) of feeding larvae [8]. As an essential part of the signal transduction mechanisms, plants employ specific transcription factors (TFs) to synchronize the expression of relevant gene networks. For example, genes controlling PA biogenesis in N. attenuata are tightly regulated by MYB8, an herbivory-inducible TF of the MYB family [6]. Figure 1 summarizes major steps in the formation of PA and connections with the lignin pathway. MYB8 activates the core phenylpropanoid genes and specific Nacyltransferases that redirect the flux of phenylpropanoids towards PA production during insect herbivory. Using microarray analysis and metabolic profiling of MYB8-silenced plants (irMYB8), we recently identified several N-acyltranferase enzymes that conjugate activated phenolic acids (p-coumaroyl-, caffeoyl-, feruloyl) with polyamines (putrescine and spermidine) [9]: AT1 is a hydroxycinnamoyl-CoA: putrescine acyltransferase responsible for caffeoyl-and coumaroyputrescine accumulation during insect herbivory. Another gene (DH29), specific for spermidine conjugation, mediates the initial acylation step in the formation of caffeoyl-, coumaroyl and feruloyl-containing di-acylated spermidine structures. Although this enzyme was not able to perform the second acylation towards diacylated spermidine biosynthesis, another acyltransferase gene, called CV86, proposed to act on mono-acylated spermidines was isolated and partially characterized [9].
Among the herbivory-regulated genes screened in the aforementioned study, we identified a candidate N-acyltransferase which shared a high sequence similarity with previously characterized hydroxycinnamoyl-CoA:shikimate/quinate hydroxycinnamoyl transferase (HCT) genes from N. tabacum (Acc. No. CAD47830) and N. benthamiana (Acc. No. CAD88491) ( Figure S1). HCT enzymes have previously been shown to catalyze developmentally-regulated lignin biosynthesis in tobacco [10,11]. Interestingly, the herbivory-induced expression of this HCT-like gene is dependent on and is temporally synchronized by MYB8 transcriptional activity. This suggests that MYB8 controls, via an HCT-like conserved activity, certain aspects of lignin production in herbivore-attacked leaf tissues. The possibility of cross-talk between HCT activity and the PA pathway (Figure 1), especially when the PA pathway is maximally induced during insect herbivory [9] has not been rigorously examined.
To dissect the metabolic processes controlled by HCT-like activity and its potential competition with PA metabolism during insect herbivory, we silenced the expression of this gene using a virus-induced gene silencing (VIGs) procedure optimized for this plant species. The VIGs approach allows genes to be silenced in developed seedlings with normal morphologies, thus avoiding the confounding effects of developmental alterations on metabolism that are commonly observed when essential genes are stably silenced [12,13]. While HCT-like apparently functions in planta as an HCT enzyme required for lignin deposition, we also uncovered a strong rerouting of the flux of activated coumaric acid units towards the production of developmentally and herbivory-induced coumaroyl-containing PAs, and to a lesser degree, towards phenolic accumulation.

Plant material and growth conditions
Wild-type Nicotiana attenuata Torr. ex Watson from an inbred line in its 31 st generation were used for all experiments. The seeds were collected from plants grown on private lands (DI ranch in southwestern Utah, USA) and since N. attenuata is not an endangered plant species, no collection permits were necessary. The owner of the private land at the time of the collections gave The MYB8 transcription factor is up-regulated in response to herbivory and activates the transcription of downstream genes in the phenylpropanoid pathway (PAL, C4H, C3H and 4CL, [6,32]). MYB8-activated Nhydroxycinnamoyl acyltransferases lead to the accumulation of direct defense metabolites including N-caffeoylputrescine (AT1) and N9,N99dicaffeoylspermidine (DH29 and CV86) in attacked and systemic leaves of N. attenuata [9]. Simultaneously, MYB8 also activates the lignin-related acyltransferase HCT-LIKE (this report) which putatively promotes local and systemic lignification and mechanical defense. Enzymatic step are shown by solid lines; experimentally determined MYB8 regulatory steps are indicated by dashed lines. doi:10.1371/journal.pone.0062336.g001 permission to conduct the collections. Seeds were treated for 1 h with diluted smoke and 0.1 mM GA 3 solution and placed on sterile Gamborg's B5 agar medium in petri dishes. Seedlings were maintained at 26uC/16 h, 155 mmol m 22 s 21 light: 24uC/8 h dark cycle in the growth cabinet for 10 days and subsequently transplanted to a peat-based substrate in small Teku plastic pots in the glasshouse. Plants were maintained at day/night cycle of 16 (26-28uC)/8 (22-24uC) h under supplemental light from 600 W high-pressure sodium lamps (Philips Sun-T Agro). After 10 d in the glasshouse, the plantlets were transferred to 1 L pots with soil and placed in a growth chamber under the same light conditions but at 20uC. Plants were kept in the growth chamber under constant humidity ,65% and the low temperature regime (20uC) for the rest of the experiment.

VIGs procedure
A virus-induced gene silencing (VIGs) system based on the tobacco rattle virus (TRV) was used as described by (Saedler and Baldwin, 2004). A 167 bp fragment of the acyltransferase2 (HCT-LIKE) gene was PCR-amplified using a primer pair shown in Figure S2. A gel-purified HCT-LIKE fragment was cloned into BamHI-SalI sites of the polylinker in the pTV00 vector to obtain pTV-HCT-LIKE. Correct plasmid clones were transformed by electroporation into the Agrobacterium tumefaciens strain GV3101: pTV-HCT-LIKE for HCT-LIKE gene silencing, pTV00 plasmid without insert (empty vector; EV) as a negative control and pTV-PD, a phytoene desaturase-silencing construct (pTV-PD) as a positive control. Three leaves of 24-25 day-old N. attenuata plants were infiltrated with a 1:1 mixture of A. tumefaciens strains carrying pBINTRA vector and one of the pTV-HCT-LIKE, pTV00 (EV) or pTV-PD constructs. VIGs progression was monitored in a set of inoculated plants with pTV-PD which bleaches leaves due to the depletion of carotenoids. Experiments on VIGs-HCT-LIKE and VIGs-EV plant treatments were initialized after the pTV-PDinoculated leaves developed the distinct bleaching phenotype demonstrating that the gene silencing process was well underway. 200 bp fragments of the gene to be silenced guarantee effective gene silencing. In general, 23 nt of identity between the fragment and the target gene is sufficient for direct post-transcriptional silencing of a gene. To achieve single gene silencing (i.e to avoid post-transcriptional silencing of closely related gene sequences), sequence identity of more than 22 nt with other genes has to be avoided, and this is usually accomplished by selecting sequences in the un-transcribed regions of the gene. This procedure was strictly followed for selecting the DNA sequence used for HCT-like VIGssilencing. To rule out the possibility of off-target silencing, the HCT-like DNA fragment used for VIGs was blasted against a full transcriptome database obtained by 454-sequencing to ensure that this fragment did not have a sequence identity of more than 22 nt with other genes.

Microscopy
Stems of elongated plants were hand-sectioned with a razor blade at the base of the plant (1 cm above the ground) and stained with 0.05% Toluidine blue O to visualize lignified cell walls. After a brief rinsing of the cross-sections in distilled water, the samples were mounted on microscope slides and observed in bright field with a Leica DM 6000B light microscopy system equipped with a CCD camera HV-D20P (Hitachi Kokusai Electric Inc., Tokyo, Japan). Photographs were processed with a LM Image Manager Elemental formulas and relative mass errors (in ppm) were calculated using Smart Formula from the UPLC-TOFMS operating software (see Method section). Candidate formulas were ranked according to both mass deviation and isotope pattern accuracy reflected in the sigma value. MS/MS+ spectra for some of the reported parent ion and the strategy used for compound annotation are summarized in Files S1 and S2 and characteristic ion fragments reported in [9,15]

Growth parameters
VIGs-EV and VIGs-HCT-LIKE plant growth parameters, i.e. stalk length and diameter (1 cm and 20 cm above the base), were determined at the end of the experiment.

Targeted analysis of secondary metabolites by HPLC-PDA
High performance liquid chromatography coupled to photodiode array detection (HPLC-PDA) was used for the quantification of secondary metabolites. Samples (approximately 100 mg) were ground in liquid nitrogen and extracted by adding 1 mL of acidified 40% MeOH, prepared with 0.5% (v/v) acetic acid/ water, to each sample in 2 mL microcentrifuge tubes with metal balls. Samples were homogenized in the ball mill (Genogrinder 2000, SPEX CertiPrep, Metuchen, New Jersey, USA) for 45 sec at 250 strokes per min and centrifuged at 16,000 g, 4uC for  minutes. Supernatants were transferred into 1.5 mL microcentrifuge tubes, re-centrifuged under the same conditions and finally transferred into 2 mL glass vials before analysis on the Agilent-HPLC 1100 series (http://www.chem.agilent.com). One mL of the sample was separated with a Chromolith FastGradient RP 18-e column (5062 mm, monolithic silica with bimodal pore structure, macropores with 1.6 mm diameter, Merck, Darmstadt, Germany) connected to a pre-column (Gemini NX RP18, 2 x 4.6 mm, 3 mm). Mobile phases -0.1% formic acid +0.1% ammonium water, pH 3.5 as solvent (A) and methanol as solvent (B) -were used in a gradient mode with the following conditions: time/concentration (min/%) for B: 0.0/0, 0.5/0, 6.5/80, 9.5/80 and reconditioning for 5 min to 0% B. The flow rate was 0.8 mL/ min and column oven temperature was set to 40uC. Levels of dicaffeoylspermidine (DCS), caffeoylputrescine (CP), chlorogenic acid (CGA), crypto-chlorogenic acid (crypto-CGA) and rutin were determined as described previously [6]. Calibration curves generated by injecting increasing concentrations of authentic standards were used for the calculation of nicotine, CGA and rutin concentrations. Levels of DCS isomers, crypto-CGA and CP were expressed as CGA equivalents, since no authentic standards were available for these compounds.

UPLC/ESI-TOF-MS metabolomics measurements
To analyze the breadth of N. attenuata phenolic-based metabolism, 2 mL of the leaf extracts prepared as above were separated using a Dionex RSLC system (Dionex, Sunnyvale, USA) equipped with Dionex Acclaim 2.2 mm 120A 2.16150 mm column, applying either a short separation binary gradient (flow rate 300 mL min 21 ) with the following parameters, 0 to 0.5 min isocratic 80% A (deionized water, 0.1% [v/v] acetonitrile [Baker, HPLC grade] and 0.05% formic acid), 20%B (acetonitrile, 0.05% formic acid); 0.5 to 2 min linear gradient to 40% B; 2 to 6 min isocratic 40% B, 6 to 10 min linear gradient to 80% B or a long separation gradient optimized for phenolamides (flow rate 300 mL min 21 ) 0 to 5 min isocratic 95% A, 5% B; 5 to 20 min linear gradient to 32% B; 20 to 22 min linear gradient to 80% B; isocratic for 6 min. Eluted compounds were detected by a MicroToF mass spectrometer (Bruker Daltonics, Bremen, Germany) equipped with an electrospray ionization source in positive ionization mode as previously described [15]. Typical instrument settings were as follows: capillary voltage 4500 V, capillary exit 130 V, dry gas temperature 200uC and dry gas flow of 8 L min -1 . Ions were detected from m/z 200 to 1400 at a repetition rate of 1Hz. Mass calibration was performed using sodium formate clusters (10 mM solution of NaOH in 50/50% v/v isopropanol/ water containing 0.2% formic acid).

Non-targeted comparative processing of UHPLC/ESI-TOFMS data
Raw data files were converted to netCDF format using the export function of the Data Analysis version 4.0 software (Bruker Daltonics, Bremen, Germany) and processed using the XCMS package (http://metlin.scripps.edu/download/). Peak detection was performed using the 'centWave' method and parameter settings ppm = 20, snthresh = 10, peakwidth = c (5,18). Retention time correction was achieved using the parameter settings minfrac = 1, bw = 10 s, mzwid = 0.1 D, span = 1, missing = extra = 0. After peak grouping and filling in of missing features using the fillPeaks routine of the XCMS package [14], the obtained data matrix was imported into Microsoft Excel for statistical analysis. Ion traces were deconvoluted and putative in-source pseudo-spectra reconstructed with the R package CAMERA (http:// www.bioconductor.org/packages/release/bioc/html/CAMERA. html) with defaults parameters. Isotopic peaks and multiplecharged m/z signals detected by CAMERA were excluded to reduce the redundancy within the data matrix. Consistent mass features, which were at least present -for a single factorial groupin four out of the five biological replicates with Rt .1 min were considered for further analysis. Zero values remaining after applying the 'filling in' function in XCMS were replaced by half of the minimum positive value of the row in the original data.

Phenolamide annotation and identification
Identifications and annotations of N. attenuata phenolic derivatives are based on MS/MS CID-induced fragmentation experiments performed on a MaXis q-TOFMS and in vivo N-containing fragment labeling with 15 KNO 3 for high precision diagnostic fragment elemental composition assignment [15]. Original data are reported in [15] and examples presented as File S1. Structural rearrangements during in source ionization and fragmentation did not allow for the unequivocal assignment of the phenylpropanoid residues to the different N positions of putrescine and spermidine backbones. N-caffeoylputrescine, N-feruloylputrescine and Ncoumaroylputrescine were unambiguously identified based on comparison of obtained mass spectral data and retention times with those of synthetic standards. Numbers in the compound name column of Table 1 refer to the different annotation levels defined by the Metabolomics Standard Initiative [16].

Statistical analysis
For principal component analysis (PCA) of UHPLC/ESI-TOFMS profiling data, we first filtered the data in order to identify and remove m/z signals that are unlikely to be of use when modeling the data. No phenotype information was used during the filtering process and original m/z features that were considered as near-constant throughout the experiment conditions based on their coefficient of variation (mean divided by standard deviation) were removed. We used the Pareto scaling -mean-centered and divided by the square root of the standard deviation of each variable -method for data normalization. The PCA analysis was performed using the 'prcomp' package for R via the MetaboAnalyst interface [17]. The HPLC data and RT-qPCR data were analyzed with PASW statistic 18 (SPSS) software.

Results and Discussion
HCT-LIKE is a MYB8-controlled and herbivory-induced Nacyltransferase gene A search of public databases containing tobacco expressed sequence tags (ESTs) yielded a large number of candidate Nacyltransferase genes classified into the BAHD protein superfamily. In an earlier study [9], we analyzed the expression patterns of these genes using a microarray system designed for N. attenuata (data-set publicly available at GEO identifier GPL13527) and discovered that several of these genes shared temporal and tissuespecific and patterns of expression with that of MYB8, a phenolamide (PA)-specific transcription factor (TF). The candidate N-acyltransferase gene was not only co-regulated with MYB8 but it also showed a strongly down-regulated pattern of expression in elicited irMYB8 leaves (Figure 2), consistent with its inclusion in the MYB8-regulated gene network presented in Figure 1. Interestingly, the deduced protein sequence of this N-acyltransferase showed high similarity to the previously characterized hydroxycinnamoyl-CoA:shikimate/quinate hydroxycinnamoyl transferase (HCT) enzymes from N. tabacum and N. benthamiana ( Figure S1) which participates in lignin biosynthesis [10,11]. Hence the homology-predicted function of this N-acytransferase contrasted with its expected function in PA biosynthesis based on its MYB8dependent regulation.
To understand the influence of HCT-LIKE on PA and lignin biogenesis in N. attenuata, we transiently silenced the expression of this HCT-LIKE gene via virus-induced gene silencing (VIGS). The VIGs procedure has been extensively used to identify the function of key N. attenuata metabolic genes, including a recent study on PA biosynthetic genes [9]. This transient gene-silencing procedure has the distinct advantage over constitutive gene silencing procedures of avoiding severe developmental alterations that compromise the conclusions about the metabolic function of particular genes. As expected, transcripts levels of HCT-LIKE were reduced by approximately 90% in leaves of N. attenuata VIGs-HCT-LIKE plants, both in unwounded leaves as well as in those on which Manduca sexta larval attack was simulated by applying oral secretions to mechanical wounds (W+OS) ( Figure S2).

Silencing HCT-LIKE affects lignin deposition and stem morphology
Silencing HCT-LIKE (VIGs-HCT-LIKE) strongly affected the plant's morphology, producing smaller plants with leaning, weak soft stems compared to VIGS-EV plants ( Figure 3A). While stalk lengths, measured 40 days after VIGs inoculation, was significantly reduced in VIGs-HCT-LIKE compared to VIGs-EV plants ( Figure 3B), the lower and upper stem diameters measured 1 and 20 cm from the base of the rosette were reduced in VIGs-HCT-LIKE, although these differences were not statistically significant ( Figure 3C). In agreement with the high sequence conservation between HCT-LIKE and the HCT genes involved in lignin biosynthesis [10,11], the growth phenotype in VIGs-HCT-LIKE plants suggested that HCT-LIKE functions as an HCT enzyme in N. attenuata.
To further examine alterations in the lignification of VIGs-HCT-LIKE plants, we analyzed the microscopic lignin deposition patterns in these as well as in VIGs-EV control plants. Lignin is a major component of the cell wall that provides mechanical strength and impermeability to vascular tissues. Chemically, lignin Figure 6. Precursor ions of abundant coumaroyl-containing PAs are up-regulated in response to HCT-LIKE silencing. Representative (n = 5 biological replicates) extracted ion chromatograms (EIC) for precursors of VIGs-HCT-LIKE-specific coumaroyl-based mono-acylated putrescines and mono-and di-acylated spermidines (Table 1) is a phenolic heteropolymer derived from hydroxycinnamoyl alcohols, mainly p-coumaryl alcohol, coniferyl alcohol and sinapyl alcohol, through a free radical coupling reaction [18]. Lignin deposition can be rapidly assessed by a polychromatic staining of the cell walls by Toluidine blue O staining [19]. Staining of the hand-cut cross-sections of the stem 1 cm above the base of the plants' rosette with a 0.05% [w/v] solution of Toluidine blue O in 0.1 M phosphate buffer (pH 6.8) revealed strongly decreased lignification which resulted in thin irregular cell walls in the xylem vasculature of HCT-LIKE-silenced plants ( Figure 3D). Additionally, a slightly red-brown color was observed in the xylem of peeled stem segments of VIGs-HCT-like plants ( Figure 3E). The altered xylem morphology determined from the inspection of crosssections of stems was comparable to that observed in HCT-silenced N. benthamiana plants, changes which are known to be due to reduced lignin deposition [10]. In contrast, VIGs-EV plants showed strong and stable stems with thick-walled xylem cells, confirming that the VIGs procedure itself did not interfere with lignin deposition ( Figure 3D).
The generation, using the RNAi technology, of stably HCT-LIKE-silenced N. attenuata plants was also conducted. Not surprisingly, T 0 transformed plants showed severely stunted growth (not shown). This stunted phenotype is consistent with observations made for HCT mutants [10] and has been functionally associated not only with reduced lignification but also with altered auxin transport due to flavonoid over-accumulation [20]. Recent studies have ruled out the role of flavonoids in the stunted phenotype of Arabidopsis HCT mutants and are consistent with the idea that a biochemical pathway downstream of coniferaldehyde is disrupted in HCT mutants [21]. These strong developmental alterations compromise the utility of these transgenic plants for comparative metabolic studies with wild type plants. In VIGs-HCT-LIKE plants, the new lignin produced after HCT-LIKE-silencing (induced at an advanced developmental stage) is added to the wild-type lignin produced before the VIGs procedure which likely explains why transient silencing of HCT-LIKE by VIGs did not result in stunted growth. Hence, comparative metabolomics experiments presented below were performed exclusively with VIGs-EV and VIGs-HCT-like plants.
No visible lignin deprivation-related phenotypes have been observed in plants stably silenced for MYB8 expression [6]. This suggests that HCT-like involvement in lignin biosynthesis is most likely regulated by an independent transcriptional activator during growth and development. Additionally, it remains to be determined whether MYB89s role is direct or via a retro-route involving the flux of MYB8-dependent metabolites into lignin biosynthesis. Gene expression analyses of N. attenuata suggest an important role for lignification in the plant's response to herbivory. For example, transcripts of the lignin biosynthesis-related cinnamyl alcohol dehydrogenase (CAD) gene have been shown to accumulate in herbivory-elicited leaves of N. attenuata plants [22]. Consistent with a protective function of lignin, stem-boring weevil larvae have been shown under natural conditions to take advantage of the softer stems of lignin-deprived N. attenuata plants [23].

HCT-LIKE-based lignin disruption results in pathwayspecific deregulations of soluble phenolic derivatives
Major advances have recently been made in our understanding of the metabolic cross-talk within the downstream branches of the phenylpropanoid pathway by analyzing the impact of the silencing of lignin biosynthetic genes [18,20,[24][25][26][27]. In tobacco, HCT catalyzes the formation of p-coumarate esters using shikimate or quinate units as acyl acceptors [11] and silencing leads to the constitutive redirection of this metabolic flux into flavonoid production through the activity of chalcone synthase enzymes [11,20] but its influence on PA accumulation has not been examined. In MYB8-silenced N. attenuata plants, PA disruption strongly influenced the accumulation of multiple activated phenolic residues [9], indicating that the herbivory-induced increase in the production of PA results in a strong metabolic demand on the flux of phenylpropanoids. We therefore initially hypothesized that silencing HCT-LIKE would directly channel activated coumaric acid units towards conjugation with polyamine and that this flux may be modulated by the strength of the herbivory treatment inflicted to the leaves of VIGs plants.
We first performed routine HPLC-UV measurements of Manduca sexta-attacked VIGs-HCT-LIKE and VIGs-EV leaves (4 day feeding; Figure 4). As expected, and in agreement with HCT-LIKE-activated deregulations predicted to specifically influence phenylpropanoid derivatives, the accumulation of nicotine was unchanged in VIGs-HCT-LIKE plants. Consistent with the revealed compensatory response observed in HCT-silenced tobacco plants, chromatographic peaks corresponding to chlorogenic (CGA) and crypto-chlorogenic acids (CGA) were significantly increased by 40 to 50% in VIGs-HCT-LIKE compared to in EV plants. Additionally to HCT, the silencing of other genes downstream in the PAL pathway is known to increase CGA accumulations [10]. The increased CGA levels are consistent with the idea that HCT-LIKE, like HCT in tobacco plants, acts predominantly as a shikimate, rather than a quinate, conjugating enzyme. In contrast with reports made for constitutive levels of different flavonoid glycosides in HCT-silenced Arabidopsis plants [20], levels of rutin, the most abundant metabolite of this compound class in N. attenuata leaves, did not change in response to reduced HCT-LIKE transcript levels. Caffeoylputrescine (CP) and dicaffeoylspermidine (DCS) isomers, the most abundant PAs in N. attenuata leaves [28], were not significantly influenced by HCT-LIKE silencing. However, inspection of the UV-trace obtained from VIGs-HCT-LIKE plants revealed a peak corresponding to an unknown PA derivative and whose intensity was strongly amplified compared to VIGs-EV plants (Figure 4), motivating a higher resolution, MS-based analysis.
Silencing HCT-LIKE diverts activated p-coumaric acids to coumaroyl-based PA production To achieve a higher resolution in the characterization of HCT-LIKE-silenced plants, we performed metabolomics analyses by UHPLC-ESI/TOF-MS of leaves collected from uninduced (Ctrl), M. sexta-attacked and W+OS-treated VIGs plants. In a previous study [9], we demonstrated that direct M. sexta feeding resulted in more pronounced PA accumulations than did the W+OS treatment. A dramatic increase in the levels of mono-acylated spermidine conjugates, the first step in the biogenesis of polyacylated derivatives, was observed compared to the almost undetectable levels of these metabolites after a single W+OS elicitation. Analyzing the metabolic responses elicited by these two treatments allowed us to capture the broadest picture of PA metabolism by taking advantage of the high temporal reproducibility afforded by the W+OS procedure and of the severe metabolic response that results from continuously chewing insect larvae.
Deconvoluted metabolic profiles from both uninduced and elicited leaves of VIGs-HCT-LIKE plants were clearly separated from those from VIGs-EV plants in PCA projection plots ( Figure S3) indicating larger deregulations than those originally described from the HPLC-UV traces. Consistent with the hypothesis that competition between PA and HCT-like-dependent metabolism is modulated by the strength of the herbivory stress inflicted, the proportion of 2-fold regulated ion signals (671 after herbivore attack vs 434 after W+OS treatment) was maximal in VIGs-HCT-LIKE herbivore attacked leaves ( Figure S4). To better understand this stronger metabolic response, we performed a hierarchical clustering analysis for 555 m/z features influenced by HCT-like expression (File S2 for the complete peak matrix) and annotated the in-source pseudo-spectra present in clusters c5, c6, c7 in which the largest changes were detected ( Figure 5). Many of the pseudo-spectra reconstructed using the CAMERA processing program contained the typical signature fragment of coumaroyl moieties ([M+H] + m/z 147.0460.05) released after the cleavage of the ester bond linking them to a core molecule (sugar, polyamine, small acid). We computed the trace for this m/z signal ( Figure 5  ) and their monohydrated versions that strongly increased after direct or simulated herbivory of VIGs-HCT-LIKE leaves. Structural rearrangements during in-source ionization and fragmentation did not allow for the unequivocal assignments of the phenylpropanoid residues to the different N positions of the putrescine and spermidine backbones. Coumaroylshikimate, the product of HCT activity, could not be detected in VIGs-HCT-LIKE as well as in VIGs-EV leaves, probably due to its rapid turnover. Figure 6 shows the computed traces for some of the most highly up-regulated metabolites, confirming that the production of many of these coumaroyl-based metabolites is specifically activated after HCT-LIKE-silencing. As is apparent in Figure 7, HCT-LIKEsilencing resulted in a clear diversion of activated p-coumaric acid units towards the accumulation of PA, but not towards quinate conjugates. This results most likely from the MYB8-mediated prioritization of the phenolic flux towards PA formation during insect feeding in N. attenuata plants. Consistent with this perspective, silencing MYB8 results in the formation of quinate conjugates, which are otherwise absent or in much lower concentrations in WT plants [9]. Interestingly, the fact that isomers of coumaroyl-PA metabolites responded differently to the different herbivore stress treatments (W+OS vs direct feeding), suggests that the strength of the competition between the lignin production and this part of PA metabolism may be compound-or even isomer-specific ( Figure 6). This could be achieved, as discussed in [29], through the independent regulations of substrate-specific and regio-isomer-specific N-acyltransferases involved in PA production. We have shown that AT1 which controls the formation of N-caffeoylputrescine and N-coumaroylputrescine and is among the most strongly elicited genes during insect feeding is not responsible for N-feruloylputrescine production, that is one of the major PAs in N. attenuata [9]. Moreover, silencing CV86, an enzyme proposed to act on mono-acylated spermidine has been shown to have isomer-specific consequences on the production of diacylated spermidines [9].
No clear changes were detected in the intensity of caffeoyl-and feruloyl-based PAs which normally account for the main part of the PA spectrum in herbivory-elicited N. attenuata leaves (Figure S5). We could only detect reductions in the production of the monohydrated forms of N9,N99-dicaffeoylspermidine ( Figure S5, C 25 H 32 N 3 O 6 + ), which suggests that diverted coumaroyl-coA units are not recycled into the phenylpropanoid pathway or metabolized after conjugation for the formation of caffeoyl-/feruloylbased PAs. Indeed, P450 enzymes that can perform the 3-hydroxylation of coumaroyl rings integrated into quinate and spermidine conjugates have been identify in other plants species [30,31]; the importance of such enzymes for PA production remains unknown in N. attenuata. HCT activity has been shown to act both upstream and downstream of the 3-hydroxylation step of the phenylpropanoid pathway [11], in other words to produce both coumaroylshikimate and caffeoylshikimate derivatives from activated p-coumaric acid and caffeic acid units (Figure 1). It was therefore surprising that no diversion, as reflected by increased caffeoyl-based PA levels, was detected after silencing HCT-LIKE. This observation argues for a central role of HCT-LIKE as a coumaroylshikimate/quinate-, rather than a caffeoylshikimate/ quinate-, forming enzyme in N. attenuata leaves. No peaks corresponding to coumaroylquinate could be detected in the leaves of VIGs-HCT-LIKE and VIGs-EV plants to evaluate the importance of HCT-LIKE activity as a coumaroylquinate-forming enzyme.

Conclusion
This study provides new insights into herbivory-induced PA production and metabolic interactions with the lignin-associated metabolic network in which it is embedded [1,32] (Figure 1). This analysis supports recently developed systems-level views on the lignin perturbation response [27] and uncovers a previously undescribed network of coumaroyl-based PAs for which a strong ''metabolic tension'' exists with the lignin pathway. The metabolic reconfigurations associated with silenced HCT expression revealed a dramatic enrichment in coumaroyl-containing phenolamides. The magnitude of this metabolic diversion depended on the nature of the herbivory treatment applied, and different PA isomers were differently affected, results not readily explained by a ''simple'' redirection of coumaroyl-coA but rather from more complex and likely MYB8-prioritized reactions, as, for example, underscored by the lack of evidence for increases in quinate conjugates. Finally, this study highlights the value of combining gene-silencing and metabolomics approaches in unraveling the clearly complex interplay among herbivory and developmentally-controlled metabolic responses. Figure S1 Alignment of N. attenuata HCT-LIKE (HCTlike, NaAT2) deduced protein sequence and highly similar HCT proteins with hydroxycinnamoyl-CoA: shikimate/quinate hydroxycinnamoyltransferase activity from N. tabacum (NtHCT; CAD47830) from N. benthamiana (NbHCT; CAD88491, gene fragment). (TIF) Figure S2 Design of the VIGs-HCT-LIKE construct and HCT-LIKE gene-silencing efficiency. (A) Nicotiana attenuata HCT-LIKE fragment (167 bp) was amplified by PCR using the primer pair shown in grey. The amplified fragment was cloned into the BamHI-SalI sites of the polylinker in the pTV00 vector to obtain pTV-HCT-LIKE. A pTV00 plasmid without insert (empty vector; EV) was used as a negative control in the experiments. To rule out the possibility of off-target silencing, the HCT-like DNA fragment used for VIGs was blasted against a full transcriptome database obtained by 454-sequencing to ensure that this fragment did not have a sequence identity of more than 22 nt with other genes. (B) Silencing efficiency. RNA extraction and cDNA synthesis was followed by qRT-PCR analysis with a primer pair designed outside the VIGS silencing region. Elongation factor (EF)-1a gene from tobacco was used for normalization of transcript levels. Like for many other genes, the responsiveness of HCT-like expression to the W+OS elicitation (Wound; W + Manduca sexta oral secretions; OS) decreases while plants elongate. Therefore, unlike in rosette-stage plants (Figure 2