Phytoalexins of the crucifer Barbarea vulgaris Structural profile and correlation with glucosinolate turnover

,


Introduction
Phytoalexins are antimicrobial metabolites elicited by (i.e.biosynthesized de novo in response to) microbial attack of plants and certain abiotic stresses (VanEtten et al., 1994;Pedras et al., 2011).Phytoalexins are of considerable interest as they exhibit diverse chemical structures and bioactivities that are finding use in plant breeding for disease resistance (Pedras et al., 2011;Ahuja et al., 2012;Møldrup et al., 2013;Ahmed and Kovinich, 2021;Kariya et al., 2023).Both UV-radiation and heavy metal salts, e.g., CuCl 2 , are abiotic stresses reported to elicit phytoalexin biosynthesis and routine screening for phytoalexins often uses abiotic stresses (Pedras et al., 2011).One reason for experimental use of abiotic rather than biotic elicitation is that many fungal pathogens metabolize and detoxify phytoalexins, generating additional metabolites that are not phytoalexins (Pedras and Abdoli, 2017).
The GSLs are sulfur-containing metabolites in plants of the Brassicales order, capable of transformation to isothiocyanates (Halkier, 2016;Barco and Clay, 2019;Blažević et al., 2020).Isothiocyanates formed by hydrolysis of preexisting GSLs are by definition not phytoalexins (VanEtten et al., 1994).The GSLs are generally constitutive metabolites although levels are subject to complex regulation (e.g.Jeschke et al., 2019;Malhotra et al., 2022).De novo biosynthesis of Brassicales phytoalexins from GSLs via isothiocyanates has been demonstrated in several cases by labelling studies (e.g.Pedras et al., 2011;Pedras and Yaya, 2013;Pedras and To, 2018) but is not universal (Schuhegger et al., 2007;Møldrup et al., 2013;Mucha et al., 2019).In one case, the phytoalexin brassinin (P1), genes and enzymes involved in biosynthesis via isothiocyanate were discovered in the crop Brassica rapa L. (Fig. 1).The biosynthesis showed a remarkable overall similarity to GSL biosynthesis concerning the kinds of reactions and enzymes involved (Klein andSattely, 2015, 2017).Many phytoalexins in crucifers contain an indole moiety and several are derived from indol-3-ylmethylGSL (IM) (Pedras and Yaya, 2013).They can be termed "indole phytoalexins" in analogy with the term "indole GSLs" commonly used for IM and derivatives.Three indole phytoalexins are discussed in this paper and numbered P1, P2 and P3 (Fig. 1).Brassinin (P1) is a representative example of an indole phytoalexin and a biosynthetic precursor of several derivatives, including cyclobrassinin (P2) (Fig. 1) (Pedras, 2014).The CYP71CR2 enzyme from B. rapa was suggested to catalyze the cyclization of P1 to P2 (Klein and Sattely, 2017).In Nasturtium officinale W.T. Aiton (watercress), a cyclobrassinin isomer with the sulfur atom of the thiocarbonyl group linked to C-4 of the indole moiety was named cyclonasturlexin (P3) (Pedras and To, 2016) (Fig. 1).Administration of isotopically labelled precursors to N. officinale demonstrated that P1 is a precursor of P3, but the enzymology of the conversion is yet unknown (Pedras and To, 2016).
The finding of GSLs and isothiocyanates to be also precursors in de novo biosynthesis of phytoalexins attracts attention to the control of GSL hydrolysis and fates of isothiocyanates.Dedicated thioglucosidase enzymes known as myrosinases are responsible for hydrolysis of the Sglucoside bond in GSLs, which generally happens upon tissue disruption but may also happen in intact tissue (Pastorczyk and Bednarek, 2016;Bhat and Vyas, 2019;Sugiyama et al., 2021;Lv et al., 2022).The term myrosinase (EC 3.2.1.147)is defined functionally and several subclades of glycosidase enzymes are myrosinases (Nakano et al., 2017;Blažević et al., 2020;Yamada et al., 2020;Lv et al., 2022).Enzymes from the BABG subclade of glycosidase enzymes were suggested to be responsible for conversion of IM to the corresponding isothiocyanate and downstream phytoalexins (Klein and Sattely, 2017) (Fig. 1).
Biosynthesis of phenyl-containing phytoalexins was investigated in B. verna and N. officinale using tailor-made labelled precursors (Pedras and To, 2018).This systematic screen of potential precursors of nasturlexin D (P6) (Pedras and To, 2018) ruled out several possibilities but showed two GSLs and their corresponding isothiocyanates to be potential precursors of P6: phenethylGSL (PE) and 3-hydroxyphenethylGSL (3hPE) as well as the corresponding isothiocyanates (Fig. 2).Hence, also in this case, phytoalexin biosynthesis would seem to involve GSL activation.Several related phytoalexins could also be converted to P6, and it could not be concluded at what step the characteristic meta-hydroxy group was introduced (Fig. 2) (Pedras and To, 2018).Glucobarbarins were suggested not to be precursors (Pedras and To, 2018).
The initial publication of phenyl-containing phytoalexins in B. vulgaris and B. verna (Pedras et al., 2015) did not include any data on potential indole phytoalexins in the genus.For this reason, it was relevant to investigate the profile of both indole phytoalexins and phenyl-containing phytoalexins, preferably in genotypes well suited for future genetic investigations.Two genotypes of B. vulgaris, known as the G-and P-types, are genetically well characterized (Kuzina et al., 2011;Liu et al., 2016Liu et al., , 2019a;;Byrne et al., 2017).As the ecological biochemistry and GSL profiles of the two genotypes contrast with each other (Liu et al., 2016(Liu et al., , 2019b;;Erthmann et al., 2018;Günther et al., 2022), it was of interest to compare their phytoalexin profiles and biosynthesis.
The GSLs in B. vulgaris are all potential substrates of myrosinase activated during phytoalexin elicitation, and some are potential direct precursors of phytoalexins.They are all structurally related to either indole phytoalexins or phenyl-containing phytoalexins, with some structural variation among the G-and P-types.Within the phenylcontaining GSLs, the pattern of substitution strongly affects the kind of product formed after myrosinase-catalyzed hydrolysis (Fig. 3).In leaves of both types of B. vulgaris, PE is mainly converted to an epimeric pair of 2-hydroxy derivatives, known by their trivial names as glucobarbarin (BAR) and epiglucobarbarin (EBAR), but low levels of remaining PE are also present in leaves (Agerbirk et al., 2015).The G-type is dominated by BAR while the P-type is dominated by EBAR.Accumulation of phenolic phenyl-containing GSLs is only known for the P-type (Agerbirk and Olsen, 2015).The main hydroxyphenyl derivative is usually 4-hydroxyepiglucobarbarin (4hEBAR) accompanied by low levels of 3-hydroxyepiglucobarbarin (3hEBAR) and 4-hydroxyphenethylGSL (4hPE).From a rare genotype collected in the Caucasus with lowered extent of 2-hydroxylation, m-substituted 3hPE has also been reported once (Agerbirk et al., 2015).Considering the scarcity of 3hPE in previous literature, the observation of synthetic 3hPE being an efficient precursor of the m-substituted phytoalexin nasturlexin D (P6) (Pedras and To, 2018) was unexpected.
As a basis for future elucidation of phytoalexin biosynthesis, we aimed to identify a plant organ exhibiting phytoalexin accumulation (and possibly biosynthesis) and to characterize the metabolite levels during phytoalexin elicitation in the two genotypes of Barbarea vulgaris.For the elicitation process, we decided to study specifically elicitation with a spray of a 10 mM CuCl 2 aqueous solution, as previously done (Pedras et al., 2015;Pedras andTo, 2016, 2018), although 2 mM CuCl 2 had also been used (Pedras and To, 2015).We followed levels of potential GSL precursors of phytoalexins over time in an attempt to identify the actual precursors, and we followed levels of general GSL products as a measure of general GSL activation during phytoalexin elicitation.As a control for any observed correlations with GSLs or products, it was relevant to include metabolites not expected to be biochemically related to phytoalexins.Previous authors had tested flavonoid levels in metabolic studies of phytoalexin elicitation in Brassica rapa L. (Pedras et al., 2008), so we carried out a similar experiment with B. vulgaris.The same previous authors had likewise observed drastically increased levels of a 4-substituted indole GSL in treated plants; we also tested for that relationship in B. vulgaris.
Specifically, we asked the following questions: (1) Are indole phytoalexins and phenyl-containing phytoalexins elicited in roots or leaves  by a spray of 10 mM aq.CuCl 2 in the two types of B. vulgaris, what is the time-course of elicitation, and do the genotypes differ in response?(2) Are potential precursor GSLs present at the site of phytoalexin accumulation, and are levels affected?(3) Is relevant decoration of GSLs or their products affected by phytoalexin elicitation, in particular 4-substitution of indole GSLs or m-hydroxylation of phenyl-containing GSLs? (4) Are levels of general GSL hydrolysis products affected by phytoalexin elicitation?As a control experiment, we also investigated whether flavonoid levels were affected during elicitation of phytoalexins, as they were not considered biochemically related to phytoalexins.

Hot extraction allowed UHPLC-QToF MS detection and quantification of phytoalexins, glucosinolates and their products, and flavonoids
Attempting to include both phytoalexins, GSLs and GSL products in a single extraction, we tested extraction in boiling MeOH-H 2 O (7:3), the preferred extraction solvent for GSLs and in vivo GSL products in B. vulgaris.In initial optimization of extraction, room temperature versus hot MeOH-H 2 O (7:3) solvent were compared; hot solvent systematically gave higher apparent phytoalexin levels at our conditions (165-325% of room temp.control) so we adopted hot extraction.Half of the extract was set aside for GSL analysis (Section 4.6.).For the other half of the extract, the solvent was evaporated and the residue dissolved in a solvent with lower MeOH content, MeOH-H 2 O (3:7), to facilitate subsequent chromatography.
We developed a UHPLC-QToF MS method allowing separation of a wide range of flavonoids, GSL hydrolysis products and phytoalexins in one run (Table 1, Fig. 4).By a combination of m/z value and retention time, it was possible to distinguish all the tested phytoalexins and most GSL hydrolysis products, however, one pair of isomeric GSL hydrolysis products co-eluted (Table 1).The first minutes of the samplechromatograms contained numerous unidentified peaks (in particular an intense peak at 2-2.2 min from m/z 267 and 249), peaks of known GSLs, peaks of known and putative flavonoids, and a peak of raphanusamic acid.At a retention time of 4 min or more, the phytoalexins and GSL hydrolysis products which are the focus of this investigation eluted as sharp, well separated peaks (Fig. 4B).a Precise monoisotopic m/z values for extracted ion chromatograms were calculated from sum formulae; isomers are indicated in bold.The [M+H] + ion was the basis of detection in all cases except for the C 9 H 7 N fragment of brassinin (P1), in which case the proton adduct of the fragment was the basis of detection.
b Retention time backed up by authentic reference except for the putative sulfoxides (P7, P8), in which case the t R is for the suggested candidate peaks.c Slope of the calibration curve in the unit "integration units/nM".Compound names in parentheses indicates that the slope for the indicated compound was used due to lack of an available standard curve for the analyte in question, assuming similar slopes within compound classes.d The lower value is either the limit of detection or the lowest level with detection tested or (in parentheses) an estimated lower limit of detection from dilution series of pure compounds and crude extracts using an assumed slope.The higher range is generally the highest level tested with confirmed linearity, or in case of brassinin (P1), nasturlexin A (P4) and rutin the confirmed upper limit of linearity.For raphanusamic acid, the additional upper limit (25) denotes the upper range with technically acceptable linearity (R 2 = 0.98) but visual signs of detector saturation (Fig. S2I).For brassinin (P1), extensive fragmentation to a resonance stabilized indol-3-ylmethylium fragment ion with formula [M+H] + = C 9 H 8 N + was observed in a concentration-dependent manner.Approximate linearity (R 2 = 0.93) in a modest range was observed, while extended approximate linearity (R 2 = 0.99) was observed for the C 9 H 7 N fragment.Screening for minor amounts of P1 used both extracted ion chromatograms for [M+H] + and the fragment for improved sensitivity.For nasturlexin A (P4), the investigated standard curve to 10 μM was highly curved, likely due to increasing formation of a complex mixture of adducts; approximate linearity suitable for analysis of minor amounts was observed in the indicated range.f In this range of peak areas, relevant for F1 in G-type samples, the calibration curve was reasonably well described (R 2 = 0.9896) by peak area = 309.0x + 873,069, where x = [Rutin] in nM.A dilution series of F1 showed a similar slightly curved shape (Fig. S3G), suggesting that F1 levels could be safely estimated with the rutin standard curve in the range found in samples although the absolute concentration remained unknown.
The phytoalexin peaks were generally the only prominent peaks observed at their m/z values (±0.005Da) (Fig. 4); two were isomers (P2 and P3) but were baseline separated (Table 1).Some phytoalexins did not fragment during the QToF MS analysis, while others exhibited characteristic, logical fragments verified by MS2 (Fig. S1).The peak for barbarin had the same elemental composition as a prominent fragment of BAR/EBAR but was otherwise the only major peak at this m/z value (Fig. 4, Table 1).The peak for resedine was also the only major peak for this m/z value, but in this case, numerous minor peaks at other retention times were also observed.Logical MS2 spectra of barbarin and resedine supported identifications (Figs.S2A-B).
It is well established that the GSL products barbarin and resedine occur as pairs of enantiomers (Fig. 3) reflecting the stereochemistry of their precursors, BAR and EBAR.Whether kjaerin is likewise a pair of enantiomers was uncertain, as the precursor in this case is mainly one stereoisomer (4hEBAR).Using a chiral column, we detected a pair of stereoisomers in 1:1 ratio of natural kjaerin isolated from myrosinasecatalyzed hydrolysis of 4hEBAR (Fig. S2C), suggesting a racemization during the formation of kjaerin as previously suggested from lack of optical activity (Agerbirk and Olsen, 2015).However, at the achiral analysis conditions generally used in this paper, each enantiomeric pair naturally appeared as a single peak.Hence, they are referred to in singularis in the rest of the text: barbarin, resedine, kjaerin.At the general UHPLC conditions of this paper, kjaerin and the isomer m-hydroxybarbarin co-eluted (Figs.S2D-E), reflecting a gradient optimized for the less polar phytoalexins.However, MS2 spectra differed (Figs.S2F-G), and we have previously reported conditions for their separation (Agerbirk and Olsen, 2015).
Standard curves showed linearity well above the level of phytoalexins and GSL hydrolysis products detected in leaf samples (Table 1).In contrast, GSLs exhibited broad, overlapping "peaks" early in these chromatograms (ca.2.2-3.2min), probably due to their much higher levels and hydrophilicity.Therefore, GSLs were determined after desulfation of GSLs in the remaining half of the crude extract (Section 4.6.).While phytoalexin standard curves were in most cases linear in the ranges tested, saturation was observed for raphanusamic acid (Figs.S2H-I).Raphanusamic acid peak areas from root samples were much higher than the upper linear range of the standard curve, so raphanusamic acid was not quantified from root samples.The phytoalexins P1 and P4 (not prominent in the samples analyzed) showed linearity in a restricted range for the [M+H] + ion, probably due to a high tendency for fragmentation (Fig. S1).
Identified Barbarea flavonoids in this text are numbered as in a previous publication (Dalby-Brown et al., 2011), supplemented by "F" for flavonoid, e.g.F1 for the main flavonoid discussed (Table 1).The four tested flavonoids were previously known from the G-type of B. vulgaris only, and the presence of all four were confirmed by extracted ion chromatograms using stringent m/z values (±0.01 Da) (Figs.S3A-B).Furthermore, MS2 spectra of flavonoids agreed with known glycoside moieties and supported identification (Fig. S3C).F1, F2 and F3 were the major peaks of the corresponding m/z value in their extracted ion chromatograms, while F4 was a minor peak (Figs.S3D-E).The tetraglycosylated F1 and F2 were absent or near the limit of detection in P-type samples (Fig. S3D).In G-type plants, the major peak in extracted ion chromatograms targeting F4 was provisionally named 'isoF4a' and eluted 0.2 min earlier than F4; several other apparent isomers were observed (Figs.S3B-E).The diglycosylated F3 or a coeluting isomer was detected in the P-type, while putative F4 was absent or at low levels in P-type samples (Fig. S3E).
Due to the small amounts of F1-F4 available, concentrations of the flavonoid standards could not be determined gravimetrically.Thus we estimated concentrations using a standard curve of rutin, a commercially available flavonoid (Fig. S3F).Slight detector saturation was observed in standard curves of the flavonoid rutin (Fig. S3F), but the standard curve exhibited an extended approximately linear part (R 2 = 0.99) useful in our range of peak areas (Table 1, Fig. S3F).Peak areas of the main investigated flavonoid F1 were also higher than the strictly linear portion of the rutin standard curve but were within an approximately linear part.As a standard curve of F1 itself showed similar shape as the rutin curve (Fig. S3G), the pseudo linear part of the rutin curve was used for F1 (Table 1).

The phytoalexin response to CuCl 2 of both G-type and P-type Barbarea vulgaris was dominated by nasturlexin D and cyclonasturlexin
Leaves of rosette plants of B. vulgaris are of the compound type: the petiole-midvein axis carries a number of small lateral lobes as well as a large terminal lobe.Aiming at investigating a homogeneous plant part, we sampled and extracted terminal lobes of rosette leaves for phytoalexin analyses.All leaf samples and analyses in this paper concern rosette leaf terminal lobes, which for simplicity will be termed "leaves" unless specificity is required in the context.
In leaves of B. vulgaris plants treated with a spray of 10 mM aq.CuCl 2 , three phytoalexins were conclusively detected in both genotypes (G and P), the indole phytoalexins cyclobrassinin (P2) and cyclonasturlexin (P3) and the phenyl-containing phytoalexin nasturlexin D (P6).These three are referred to as the three major phytoalexins in the rest of the paper.In all three cases, we detected substantial intensity extracted ion peaks of proton adducts at stringent conditions (±0.005 Da).We also observed the expected m/z values, corresponding to [M+H] + and any fragments exhibited by authentic standards, as major signals in the detected peaks of several samples (Fig. S4A).Cyclobrassinin (P2) was consistently detected at lower levels than P3 and P6.Hence, cyclonasturlexin (P3) and nasturlexin D (P6) are referred to as the two dominating phytoalexins.In many samples, cyclobrassinin (P2) was not even detected despite the presence of the two dominating phytoalexins (Fig. 5).
After preliminary experiments for optimizing extraction, three elicitation experiments were carried out, referred to as experiments 1, 2 and 3, respectively (Section 4.3; Table 2).In all three, the phytoalexins and various other metabolites listed in Table 1 were quantified using a single UHPLC-QToF MS analysis.In experiments 2 and 3, this analysis was supplemented with quantification of GSLs, requiring a separate analytical approach (Section 4.6.).After comparison of leaf positions in experiment 1 (Fig. S5A), we concluded that the youngest fully or nearly fully expanded leaves of a rosette plant were representative and generally high in phytoalexins after CuCl 2 treatment; all subsequent leaf analyses considered young rosette leaves.Statistical evaluation was carried out separately for each experiment.The sum of major phytoalexins in treated G-type leaves, averaged across the values at 2 and 3 days after treatment, was on average 7.0 nmol/g fresh wt.(SD 4.6, N = 6) in experiment 2 and 8.4 nmol/g fresh wt.(SD 5.4, N = 10) in experiment 3.For the P-type (experiment 2), the sum averaged across the values at 1, 2 and 3 days after treatment was 2.6 nmol/g fresh wt.
We manually searched all chromatograms for five other phytoalexins available as standards (P1, P4, P5, P8, P10) (Table 1), but in no case conclusively detected any of them.However, trace signals were observed in some cases (Fig. S4B).In experiments 1 and 2 we concluded that none of these five phytoalexins were detectable above mere trace levels.This was likewise concluded in experiment 3.However, in experiment 3 we manually integrated all trace peaks for inclusion in statistical tests; very low levels of brassinin (P1), nasturlexin C (P5) and tridentatol C (P10) in the data reflect such trace peaks but not conclusive analytical identification in any sample.Likewise, minute levels of an isomer of P6 were included in the data for statistical reasons.From reasonable t R and m/z it was tentatively identified as the known nasturlexin C sulfoxide P7 (Pedras et al., 2015) (found: m/z 224.0196 ± 0.0023 (mean ± SD, N = 6), calculated for C 10 H 10 NOS 2 + : 224.0198).However, P7 was not available as authentic standard.Finally, a low intensity peak at 4.7 min tentatively identified as the known nasturlexin D sulfoxide (P8) (Pedras et al., 2015) was detected in many samples.The m/z value matched the expected mass of the proton adduct [M+H] + of the previously reported P8 (found: m/z 240.0138 ± 0.0019 (mean ± SD, N = 10), calculated for C 10 H 10 NO 2 S 2 + : 240.0147).An apparent isomer at 3.6 min was less frequently observed, usually with lower peak area.
The two dominating phytoalexins P3 and P6 were generally present in leaves of CuCl 2 -treated plants 1-3 days after treatment: in both experiment 2 and 3, each sample of a treated (G-type) plant contained both of these two dominant phytoalexins.However, in experiment 1, where only sub-samples of dried leaves were sampled, P3 was entirely absent from seven of 16 samples of treated plants (44%), while P6 was consistently detected in all samples of treated plants.We interpreted this observation as evidence of anatomical heterogeneity of distribution of the major indole phytoalexin P3 in individual leaves, and for this reason, each extracted sample consisted of one entire leaf terminal lobe in the subsequent experiments 2 and 3.
We concluded that the indole phytoalexins cyclobrassinin (P2) and cyclonasturlexin (P3) (Fig. 1) and the phenyl-containing phytoalexin nasturlexin D (P6) (Fig. 2) were the major phytoalexins in leaves elicited by foliar CuCl 2 treatment in both the G-type and P-type of B. vulgaris, with typical combined average levels between 1 and 10 nmol/g fresh wt.

Modest phytoalexin time-course differences between G-and P-type Barbarea vulgaris
The time course of the phytoalexin response in both plant genotypes was considered in experiment 2. In the control plants, phytoalexins were not detected except a trace in a single G-type plant on day 3 (Fig. 5, Fig. S6).In both plant genotypes, the phytoalexin accumulation of treated plants appeared similar in the first two days.In the G-type (treated compared to controls), there was a significant effect of the treatment on phytoalexin levels (Fig. 6A-C), which did not decline on day 3.In the P-type (treated compared to controls), the period with maximum phytoalexin levels tended to be confined to day 1 or 2, and hence be briefer than in the G-type (Fig. 6A-C).Statistical analysis of the P-type phytoalexin levels (ANOVA) showed a significant effect (cyclobrassinin, P2), a tendency (cyclonasturlexin, P3) or no significant effect (nasturlexin D, P6) of the treatment.
We then considered whether there was a difference in the phytoalexin response among the plant genotypes.Comparison of the overall phytoalexin response to CuCl 2 treatment of the G-type and the P-type by ANOVA showed no statistically significant difference of the responses among the two plant types for any of the three major phytoalexins (Fig. 5, statistical details not shown).Due to the tendency for a briefer response in the P-type (Fig. 6A-C), this genotype was not used in subsequent experiments.However, the data did allow investigation of correlation patterns among individual metabolites.
In individual leaves, levels of the two indole phytoalexins (P2 and P3) appeared to be quite well correlated (R 2 = 0.66) in the P-type (Fig. 6D).However, in the G-type, levels of the two indole phytoalexins were not correlated (R 2 = 0.12), reflecting the extended duration of the P3 response.In contrast, levels of the dominant indole phytoalexin P3 were reasonably well correlated with levels of P6 in the G-type (R 2 = 0.78), while any correlation in the P-type relied on a single pair of observations (Fig. 6E).This pattern of variable degrees of correlation between phytoalexin levels supported the tendency for two contrasting time courses of accumulation: some phytoalexins tended to peak early (day 1-2, all phytoalexins in P-type, cyclobrassinin in G-type), while the two dominant phytoalexins tended to peak later in the G-type (day 2-3) (Fig. 6A-C).The lack of correlation of early P2 with later P3 and P6 in the G-type was confirmed by statistical testing in experiment 3 (vide infra).In both plant genotypes, low levels of putative nasturlexin D sulfoxide (P8) were identified in many samples.Levels of the putative sulfoxide P8 appeared to be well correlated with P6 levels as expected for a proposed downstream metabolite of P6 (R 2 = 0.81 in both types) (Fig. 6F).
From the comparison of the two plant genotypes in experiment 2, we concluded that the same three major phytoalexins were formed in both plant types, with no statistically significant difference in the response.As the combined data suggested that the response to CuCl 2 treatment was more prolonged in the G-type than in the P-type, we concentrated on the G-type in experiment 3.In experiment 3, with focus on the G-type, higher number of replicates and consideration of both leaves and roots, the effect of the CuCl 2 treatment was confirmed for all three major phytoalexins in leaves (Table 3).However, to our surprise, low levels of phytoalexins were detected also in control plants (Fig. S7), suggesting that phytoalexin biosynthesis in all plants was already elicited (by an uncontrolled environmental factor e.g.pathogen exposure) to a low extent at the time of the CuCl 2 or mock treatments.No visible signs of pathogen attack or other stress were noticed during the experiment.The phytoalexin profile from this background-elicitation was dominated by the phenylcontaining phytoalexin P6; levels tended to decline over the three experimental days (Table 3).
Traces of the suggested precursor of P6, nasturlexin C (P5) and its sulfoxide P7 were also detected in the leaves, but not the suggested earlier precursor nasturlexin A (P4) (Table 3).Very low levels of the parent indole phytoalexin brassinin (P1) and the phenyl-containing phytoalexin tridentatol C (P10) were also tentatively detected.Analysis of variance (ANOVA) suggested that levels of these trace phytoalexins depended on the treatment, with highest levels in treated plants (Table 3).There was no statistically significant effect of treatment for the putative nasturlexin D sulfoxide (P8); however, linear regression analysis of levels of P8 as a function of nasturlexin D (P6) showed a statistically highly significant positive slope, both when considering treated plants or control plants (Table 4).This result was expected for an obvious precursor-product relation and agreed with correlations observed in experiment 2 (Fig. 6F).
In order to test the tentative observations from experiment 2 of different time courses of appearance of the modest level indole phytoalexin P2 versus the dominant indole phytoalexin P3 and phenylcontaining P6, correlation analysis was supplemented by linear regression analysis using data for experiment 3 (Table 4).Levels of the indole phytoalexin P2 were not correlated with any of the two dominant phytoalexins, reflecting the different timing of their elicitation, with P2 peaking early (day 1) and the two major phytoalexins P3 and P6 continuously accumulating over three days (Fig. S7).This result agreed with tendencies and correlation analysis for the G-type in experiment 2. In contrast, levels of the two major phytoalexins P3 and P6 were correlated (Table 4), as also inferred for the G-type in experiment 2 and in agreement with similar appearance of averaged time course plots (Fig. S7).
Due to known high levels of the expected precursor GSLs PE and IM in roots, we had anticipated that phytoalexin biosynthesis might take place in roots as a systemic response to CuCl 2 treatment of leaves (Koprivova et al., 2023).However, only trace values of nasturlexin D (P6) were detected in roots, with no statistical association to CuCl 2 treatment.Among the phytoalexins searched for (Table 1), no other phytoalexin was detected in roots of G-type plants (Table 3).

Variable dynamics of indole glucosinolates after phytoalexin elicitation
The dominating indole GSL in all leaf samples was IM.Although IM is the expected biosynthetic precursor of indole phytoalexins (Fig. 1), biosynthesis of indole phytoalexins did not require any significant drain of IM levels, as IM levels were typically 1-2 orders of magnitude higher than levels of indole phytoalexins.However, in experiment 2 a moderate decline of IM levels in treated plants was observed in G-type plants, while IM levels were not affected in P-type plants.In experiment 3, levels of IM in G-type plants were likewise not affected by the treatment (Table 5, Fig. 7A).Accordingly, we concluded that IM levels were not

Table 3
Levels (nmol/g fresh wt.) of phytoalexins in young rosette leaves or roots after CuCl 2 -treatment ("treated") or mock treatment ("control") of G-type Barbarea vulgaris plants (N = 5 in each of six groups; data from experiment 3).Bold font names and data indicate the three major phytoalexins and their levels.) was detected in one leaf sample.The reported trace levels of brassinin (1) were based on manual integration of trace peaks of the m/z 130 fragment ion, a peak for the [M+H] + .ionwas not detected in any sample.
b Inconclusive identification of P5, P7, P8 and P10 due to very low levels close to detection limit.These five metabolites were detected as trace-like peaks tentatively identified by retention time of stringent (±0.005Da) extracted ion chromatograms only; due to low levels, correct fragmentation could not be confirmed.Levels of these five metabolites should hence be regarded as tentative maximum estimates only.c Additional data for roots: Nasturlexin A (P4) traces (typically 0.002 nmol/g fresh wt.) were detected in six root samples with no correlation to treatment.Traces of nasturlexin C (P5) were detected in one control (0.005 nmol/g fresh wt.) and one treated (0.010 nmol/g fresh wt.) root sample.Trace m/z 130 peaks at the t R of brassinin (P1) were detected in some root samples, with no difference between controls and treated plants.No other tested phytoalexins were detected in any roots sample.
P.D. Cárdenas et al. generally affected by the CuCl 2 treatment.
Searching for any yet poorly understood metabolic connection (Pedras et al., 2008) between phytoalexins (such as the 4-substituted P3) and 4-substituted indole GSLs, we compared levels of 4-substituted indole GSLs in treated and control plants.The 4-substituted indole GSLs were 4-hydroxyindol-3-ylmethylGSL (4hIM) and 4-methoxyindol-3-ylmethylGSL (4moIM) (Fig. 1).In experiment 2, comparing both plant genotypes, highly significant effects of considerable magnitude were observed for levels of 4-substituted indole GSLs, but in opposite directions and with different levels in control plants (Table 5, Fig. S8A, Fig. S9).In experiment 3, however, treated and control G-type plants showed comparable levels of 4-substituted indole GSLs (Table 5, Fig. 7B).We concluded that the CuCl 2 treatment resulted in a strong apparent effect on levels of 4-substituted indole GSLs in some but not all experiments, but the timing and direction of the fluctuation was not reproducible.

Depletion of phenethylglucosinolate during phytoalexin elicitation
Two phenyl-containing GSLs had been suggested as precursors in the biosynthesis of the phenyl-containing phytoalexin P6: either the unsubstituted PE and/or the meta-hydroxy derivative 3 hP E. However, the latter was not detectable in these B. vulgaris accessions (Fig. S10), while PE was consistently detected (Fig. S11).
In the P-type, PE levels were significantly affected by number of days after elicitation, reflecting quite low levels of PE in treated plants on days 1 and 2 compared to controls (Table 6, Fig. S12).The treatment effect was trending, but not statistically significant (P = 0.11), perhaps reflecting that levels had returned to control levels on day 3, at which day P6 accumulation had also ceased.Since our aim was to test whether on-going biosynthesis of P6 resulted in decreased levels of PE due to PE being a precursor of P6, testing PE levels within the entire 3-day duration of the data was misleading.Indeed, exclusion of day 3 data resulted in a highly significant treatment effect in ANOVA (Table 6).In the Gtype, significantly lowered levels of PE were observed in both experiment 2 and 3 (Table 6, Fig. 7C, Fig. S12).We concluded that PE levels were reproducibly reduced in treated plants during phytoalexin accumulation.
If phytoalexin biosynthesis involved unspecific GSL hydrolysis (Fig. 3), a general lowering of levels of all GSLs or all phenyl-containing GSLs could be hypothesized.Phenyl containing GSLs dominated the GSL profile of leaves, with total mean levels typically three orders of magnitude higher than levels of phenyl containing phytoalexins (compare Tables 3 and 6).For the P-type, there was no significant difference in total GSLs or total phenyl-containing GSLs in treated versus control plants when considering all three experimental days.However, for this genotype, phytoalexin accumulation was restricted to the first two experimental days, and there was a modest treatment effect when considering the first two days only (Table 6).For the G-type, lowered total GSLs and total phenyl-containing GSLs were found in treated plants compared to controls in experiment 2, but in experiment 3 the trend was different, with increased levels on day 2 and decreased levels on day 3 (Table 6, Fig. 7D).We concluded a tendency for lowered or fluctuating total GSL levels accompanying CuCl 2 treatment.We further investigated the hypothesis based on GSL product levels (Section 2.6.).
Other hydroxyphenyl-containing GSLs were searched for as potential indications of hydroxylation of PE or derivatives in the meta position.The data agreed with previous investigations (Agerbirk et al., 2015), showing low levels of 4hEBAR and 3hEBAR in the P-type and absence or very low levels in the G-type (Fig. S12).No relation to CuCl 2 treatment

Table 4
Correlation and linear regression analysis of levels of selected phytoalexins in young rosette leaves of Barbarea vulgaris the G-type plants, using either CuCl 2 treated plants (T) or mock-treated control plants (C).Mock-treated plants contained low levels of nasturlexin D (P6) apparently due to background elicitation not caused by the experimental treatment.Data from experiment 3. was found.During inspections of desulfo GSL chromatograms for trace peaks, an unidentified isomer of (desulfo) 3hPE, BAR and EBAR was consistently detected in chromatograms (Fig. S10).From comparison of retention times, this isomer was distinct from all six known isomeric (desulfo) GSLs (BAR, EBAR, 3hPE, 4hPE and two methoxybenzylGSL isomers) that were all available as authentic desulfo GSL standards (Agerbirk et al., 2022).If this isomer represents a yet unknown GSL, it could potentially be "ortho-hydroxy PE" or several other isomers.We considered whether it could possibly be related to an alternative pathway to P6 (Fig. 2).The unknown isomer was tentatively named "isoBAR" and showed no relation to the CuCl 2 -treatment (Fig. S13).We concluded that the hypothetic P6-precursor 3hPE was not detected and that no detected hydroxyphenyl-containing GSLs showed any relation to CuCl 2 -treatment.

Levels of glucobarbarin hydrolysis products depended on CuCl 2treatment and were correlated with phytoalexin levels
For an independent assessment of GSL hydrolysis, levels of GSL products were investigated using linear models.We initially used data from experiment 3 due to the higher number of replicates in that experiment, and subsequently compared with data from experiment 2. Representative chromatograms are illustrated (Figs.S8B-D).
A precursor-product relationship of barbarin and resedine in intact leaves had previously been suggested (Fig. 3), and this suggestion was supported by linear regression analysis of resedine levels as a function of barbarin levels in CuCl 2 -treated leaves (Fig. 8A).The slope was statistically different from zero (P < 0.001), suggesting that levels of resedine indeed were related to levels of barbarin.In control plants and roots, however, no relationship was distinguishable (Fig. 8A, Fig. S14).Data from experiments 1 and 2 agreed with a linear relationship of barbarin and resedine levels in treated leaves (Figs.S15A-B, Figs.S16A-B).
The same approach was used for testing for a statistical relationship of barbarin levels and total levels of glucobarbarins (BAR and EBAR, the known GSL precursors of barbarin).If levels of barbarin in intact plants reflected a steady turnover of glucobarbarins, we expected barbarin levels to be correlated with glucobarbarin levels.However, this was not the case, neither in leaves (Fig. 8B) nor in roots (Fig. S14 B).In experiment 2, similar complex patterns were seen, although a statistically significant (P < 0.05) relationship was seen for treated G-type plants (Figs.S15C-D).Experiment 1 could not be included because GSLs were not determined.
Our alternative hypothesis was that barbarin levels were linearly related to phytoalexin levels.That relationship would be expected if phytoalexin biosynthesis is accompanied not only by hydrolysis of precursor GSLs but also hydrolysis of other GSLs.Data from experiment 3 agreed with this hypothesis, as the slope of the regression line of barbarin levels as a function of phytoalexin levels was greater than zero (Fig. 8C).As the glucobarbarins are the dominating GSLs but are considered unlikely phytoalexin precursors (Pedras and To, 2018), this result suggested general GSL hydrolysis (i.e., without structural specificity) accompanying phytoalexin biosynthesis.The linear relationship of barbarin versus total phytoalexins was less robust when tested in experiments 1 and 2; the slope was significantly (P < 0.01) higher than zero in one experiment, while the other three only supported the hypothesis by trend (P-values 0.13, 0.07, and 0.06, respectively) and showed similar slopes (Fig. S16B; Figs.S17A-B).We concluded that the combined data agreed with barbarin levels generally related to phytoalexin levels.
If indeed barbarin levels were related to total phytoalexin levels because general GSL hydrolysis accompanied GSL hydrolysis for phytoalexin biosynthesis, levels of resedine should similarly be related to total phytoalexin levels.Indeed, there was a statistically highly significant linear relationship (P < 0.001) between levels of resedine and combined phytoalexins (Fig. 8D).The relationship of resedine versus total phytoalexins was relatively robust when tested in experiments 1 and 2; slopes were in the same range and in three of four plots a positive slope was statistically significant (Fig. S16C, Figs.S17C-E).
Considering the apparent statistical relation of barbarin and resedine to phytoalexin accumulation, we considered mean levels of these metabolites in the three days following the phytoalexin elicitation (Fig. 9A+B).Barbarin showed a modest (P < 0.05) dependency of the CuCl 2 -treatment (Fig. 9A), while resedine showed a statistically highly  significant (P < 0.001) relationship with much elevated mean levels in treated plants on days 2 and 3 (Fig. 9B).
The accumulation of barbarin and resedine in parallel to phytoalexin elicitation provided a means of estimating the unspecific hydrolysis of glucobarbarins accompanying phytoalexin biosynthesis.The slope with confidence intervals of the regression line of barbarin levels as a function of major phytoalexin levels would suggest hydrolysis of between five and 25 times as much glucobarbarins as the molar amount of precursor GSL channeled into phytoalexin biosynthesis (Fig. 8C).This was a minimum estimate as any barbarin further transformed to resedine was

Table 6
Levels of phenethylglucosinolate (PE) as well as the sum of all phenyl-containing glucosinolates (GSLs) and all GSLs in young rosette leaves after CuCl 2 -treatment ("treated") or mock treatment ("control") of G-type and P-type Barbarea vulgaris plants on the day before Day 1. Units for the GSL levels are indicated separately for each entry.not included.Linear regression analysis of the sum levels of barbarin and estimated resedine as a function of the sum levels of major phytoalexins suggested a somewhat higher turnover of 14-39 times as much glucobarbarins as precursor GSLs channeled into phytoalexin biosynthesis (Fig. S18).However, this estimate is less certain because resedine levels were estimated using a standard curve for barbarin (Table 1).Overall, we concluded that phytoalexin accumulation was accompanied by marked accumulation of the downstream GSL metabolite resedine and frequently the GSL hydrolysis product barbarin, indicating appreciable general turnover of GSLs accompanying elicitation of phytoalexins.
In addition to the barbarin-resedine pair from hydrolysis of glucobarbarins, we considered a more general type of GSL metabolite, raphanusamic acid.Raphanusamic acid can be formed from in vivo detoxification and turnover of any isothiocyanate; which in B. vulgaris could include detoxification of phenethyl isothiocyanate (Fig. 3) and perhaps indol-3-ylmethyl isothiocyanates.The observed levels were of the same order of magnitude as the drop of PE levels during phytoalexin accumulation, and showed statistically significant effects of treatment and time after elicitation, peaking already at day 1 (Fig. 9C).Due to the peak at day 1, a linear relationsship between raphanusamic acid and phytoalexins was not expected.Actually, plots of raphanusamic acid levels as a function of phytoalexin levels had a peculiar hyperbolic character with raphanusamic acid-phytoalexin pairs strongly associated with the x-and y-axes (Fig. S19).That is, high raphanusamic acid levels coincided with less than maximum phytoalexin levels.The plot reflected the early rise and immediate decline of this general isothiocyanate detoxification product after phytoalexin elicitation, in contrast to the steady accumulation of phytoalexins, barbarin and resedine.
Enzymes responsible for the m-hydroxylation in P6 or p-hydroxylation in P9 and P10 could hypothetically form kjaerin or m-hydroxybarbarin.For this reason, we investigated all samples for kjaerin and mhydroxybarbarin.Occasional traces were detected, but neither were conclusively detected in any sample.However, since the limits of detection were rather high relative to phytoalexin levels, relevant levels might be below the limit of detection.We concluded that major levels of kjaerin and m-hydroxybarbarin were absent.
Finally, in an attempt to reveal any general metabolic disturbance after CuCl 2 elicitation of phytoalexins, we considered flavonoid levels (Pedras et al., 2008).Numerous flavonoid-like peaks were detected in G-type and P-type plants, in agreement with previous reports (Figs.S3D-E).We selected the dominant of the four identified flavonoids, tetraglycosylated flavonoid F1, as a representative of general accumulated metabolites in G-type B. vulgaris.Analysis of variance revealed no statistically significant differences between F1 levels in treated versus control plants (Fig. S20).

Discussion
In this study of two genetically well-characterized genotypes of the crucifer Barbarea vulgaris, G-type and P-type, we have characterized phytoalexin responses to foliar CuCl 2 treatment (a known elicitor of phytoalexins): mainly the indole phytoalexin P3 (cyclonasturlexin) and the phenyl-containing phytoalexin P6 (nasturlexin D), and lesser amounts of the indole phytoalexin P2 (cyclobrassinin).Two suggested precursor GSLs, the phenyl-containing PE and the indole-containing IM, were present in leaves.A third suggested precursor, 3hPE, a derivative of PE which is extremely rare in nature, was not present above the limit of detection.We have additionally performed a first characterization of the additional metabolic response to CuCl 2 treatment, with focus on interactions with GSL hydrolysis.Four interactions with GSL hydrolysis were particularly evident.First, depletion of PE during the first three days after elicitation, as expected if the phenyl-containing GSL PE is precursor of the biosynthesis of phenyl containing P6.Second, a brief accumulation of raphanusamic acid, a known product of the turnover of isothiocyanates (Pastorczyk and Bednarek, 2016;Jeschke et al., 2019;Sugiyama et al., 2021), on the first day after elicitation.Third, continued accumulation of products of the turnover of a pair of non-precursor GSLs, the glucobarbarins BAR and EBAR, during the first three days after elicitation.Levels of these products of GSL turnover, barbarin and resedine, also correlated with phytoalexin levels.The statistical association with phytoalexin levels and CuCl 2 treatment was particularly strong for the down-stream product resedine.Fourth, drastic fluctuations of 4-substituted indole GSLs were observed to accompany the treatment but in a poorly reproducible way.In contrast, no m-hydroxyphenyl-containing GSLs or GSL products showed increased levels associated with phytoalexin biosynthesis, suggesting that anticipated m-hydroxylation in phytoalexin biosynthesis did not happen to GSLs or their products.
Combining phytoalexin biosynthesis from IM and PE appears to be unusual among crucifers, which usually produce phytoalexins from the parent indole GSL IM (Pedras et al., 2011).From patterns of correlation or lack of it, we tentatively observed that the biosynthesis of the two indole phytoalexins P2 and P3 differed in timing in the G-type, while their timing was more consistent in the P-type.Similarly, we observed a signature of different timing of the phenyl-containing phytoalexin P6 between the two genotypes.Overall, these observations suggest a somewhat different biosynthesis and regulation of each of the three major phytoalexins in our genotypes, although the kinds of phytoalexins were identical.Different time courses, brief or extended, could have biological consequences in interactions with pathogens.In general, the results of decades of study of interactions between GSLs, phytoalexins and detrimental as well as beneficial microbes are accumulating (Plaszkó et al., 2022;Poveda et al., 2020).In contrast, identification of biosynthetic genes of GSL-derived phytoalexins is still in its infancy (Klein and Sattely, 2017).Hence, the focus of the present investigation was to characterize profile, time course and metabolic interactions of cyclonasturlexin (P3) and nasturlexin D (P6) biosynthesis in the genetically most well characterized system containing these promising phytoalexins: B. vulgaris.Based on previously reported sequences of enzymes in the biosynthesis of brassinin (P1) and P2, we searched the G-type genome for homologs and previously identified homologs of the atypical myrosinase candidate BrBABG (Agerbirk et al., 2018).We have now further identified homologs of the dithiocarbamate methyl transferases BrDTC-MT a and b, and the P1 to P2-converting BrCYP71CR 1 and 2 (manuscript in preparation).These identifications suggested to us that B. vulgaris could be a suitable experimental system for investigating the biosynthesis of indole phytoalexin P3 and phenyl containing phytoalexins such as P6.Such investigations would complement the investigations of P1 and P2 biosynthesis in B. rapa (Klein and Sattely, 2017).Although myrosinase activity in Barbarea vulgaris upon tissue disruption is well-established (e.g.Agerbirk and Olsen, 2015;Müller et al., 2018) and sequences of several candidate genes have been reported (Liu et al., 2016), the gene(s) and enzyme(s) have apparently not been named and characterized phylogenetically.For this reason, the general abbreviation MYR is used here (Figs. 1, Fig. 2, Fig. 3).Classifying myrosinases in B. vulgaris and unraveling their roles in general GSL hydrolysis and biosynthesis of two distinct groups of phytoalexins is obviously a future priority.
Identifying the site of phytoalexin biosynthesis is crucial for the planning of biosynthetic investigations.Our results suggest (terminal lobes of) rosette leaves, but not roots, to be a site of phytoalexin biosynthesis after foliar abiotic elicitation, based on detection of phytoalexins in leaves but barely in roots, and detection of a plausible metabolic signature in leaves.However, phytoalexin biosynthesis in general can happen in roots (Koprivova et al., 2023), possibly also in B. vulgaris.Phytoalexin biosynthesis may be part of a hypersensitive response; this would suggest in situ biosynthesis.However, GSL levels are not uniform in leaves; patterns of distribution have been reported (Schroff et al., 2008).Indeed, complex transport processes and non-uniform patterns of gene expression seem to be involved in creating the patterns of GSL distribution (Madsen et al., 2014;Sønderby et al., 2010;Nintemann et al., 2018).Petioles or other basal rosette parts could have a role in phytoalexin biosynthesis in analogy with the prominent part in GSL biosynthesis played by the vascular system and basal leaf parts.These hypotheses could be tested by MS imaging (Lorensen et al., 2023;Maia et al., 2022) and visual reporter genes based on genes presumably involved in phytoalexin biosynthesis, e.g. the atypical myrosinases and methyl transferases.Comparable investigation of camalexin biosynthesis in A. thaliana showed the biosynthesis to be restricted to the necrotic spots due to the eliciting pathogen (Schuhegger et al., 2007).
Although strong fluctuations in 4-substituted indole GSLs were observed after phytoalexin-elicitation in some experiments, the reproducibility was poor.These indole GSLs are biosynthesized from IM (Pfalz et al., 2011).Marked increase in a 4-substituted indole GSL was also reported upon CuCl 2 treatment (Pedras et al., 2008) or pathogen exposure (Aghajanzadeh et al., 2023) of B. rapa.Levels of the parent indole GSL IM were more stable in our investigation, although IM is the biosynthetic precursor of the indole phytoalexins P1-P3.However, due to the considerable levels of IM, even a substantial drain of the IM pool might not be distinguishable in metabolic analyses.
From tracer experiments in a related species, B. verna (Pedras and To, 2018), two potential GSL precursors of phenyl-containing phytoalexins had been suggested, PE and its m-hydroxy derivative 3hPE (Fig. 2).Levels of PE were consistently reduced after CuCl 2 treatment compared to mock treatment.We interpret this result as consistent with nasturlexin D (P6) biosynthesis from PE, happening in the analyzed leaves.The additional potential precursor, 3hPE, was not detectable during the period of P6 accumulation.PE is one of the most widespread GSLs in crucifers (Bell and Wagstaff, 2017;Agerbirk et al., 2021a), including cruciferous crops such as radish (Shang et al., 2022) and oilseed rape (Missinou et al., 2022).The biosynthesis of PE is also well-understood (Liu et al., 2016;Wang et al., 2021).Hence, biotechnological (Poveda et al., 2020) or classical genetic introduction of nasturlexin D (P6) biosynthesis starting from PE would potentially be feasible for several cruciferous crops.
The parallel accumulation of barbarin and resedine is evidence for general GSL hydrolysis coinciding with the elicitation of phytoalexins, but whether this general GSL hydrolysis is mechanistically linked to phytoalexin biosynthesis or rather a mechanistically independent effect of the CuCl 2 treatment cannot be concluded.An independent effect could involve two different myrosinases, one of which could be BABG (Fig. 2), acting independently on PE and glucobarbarins, respectively.The association of resedine with phytoalexin accumulation tended to be stronger than the association of barbarin with phytoalexins, in agreement with an enzyme ("oxazolidinethionase") (Agerbirk et al., 2018) converting barbarin to resedine during phytoalexin biosynthesis.This enzyme is among several recently reported pathways in GSL turnover (Sugiyama et al., 2021;Andernach et al., 2023), and the association of resedine with phytoalexin accumulation suggest that oxazolidinethionase is part of a response to infection and could be studied in CuCl 2 -treated plants.The rapid decline of raphanusamic acid levels would likewise suggest turnover of this early detoxification product of free isothiocyanates (Jeschke et al., 2019;Sugiyama et al., 2021).In contrast, resedine accumulated; whether resedine is eventually turned over is yet unclear.
The observation of accumulating products of phenyl-containing GSLs yet essentially stable levels of this group of GSLs is not surprising given the actual levels observed: even the highest levels of barbarin and resedine combined were less than one μmol/g, while total GSL levels were on average between five and ten μmol/g.Hence, the biochemical conditions of CuCl 2 -treated leaves were far from those of physically disintegrated leaves, in which the entire pool of GSLs is rapidly hydrolysed (Agerbirk and Olsen, 2015).Furthermore, levels of GSLs represent a balance between biosynthesis and turnover, and in another case of leaf stress (herbivory), biosynthesis of B. vulgaris GSLs was indeed induced (Liu et al., 2016).De novo IM biosynthesis was also responsible for an IM-derived response to pathogen challenge A. thaliana (Hunziger et al., 2020).Overall, rosette leaves would represent a suitable plant part for future detailed investigations of the dynamics of GSL biosynthesis and turnover that occurs in parallel to or as part of biosynthesis of nasturlexin D (P6).
Flavonoids have so far been little investigated in G-and P-types of B. vulgaris, although four structures are known from the G-type (Dalby-Brown et al., 2011) and genetic loci responsible for tetraglycosylated flavonoids had been genetically mapped from a G-type/P-type F2 population (Kuzina et al., 2011).Presence of all four previously reported G-type flavonoids were confirmed for the G-type plants investigated here.Since levels of the dominant identified flavonoid F1 in G-type plants were not dependent of the CuCl 2 -treatment according to ANOVA or related to phytoalexin levels in a linear model, we concluded that there were no signs of a disturbance of flavonoid metabolism from the elicitation of phytoalexins using 10 mM CuCl 2 .A similar experiment was reported from experiments with a crop, B. rapa (Pedras et al., 2008).In that case, levels of four flavonoids were unaffected or moderately increased by CuCl 2 treatment.In somewhat contrast to our hypothesis of independence between phytoalexin elicitation and flavonoids, a recent paper reported metabolic cross talk between GSLs and flavonoids (Tian et al., 2023).
In conclusion, we have identified genotypes, conditions and plant parts that should allow future molecular investigations of the biosynthesis of three agronomically promising phytoalexins in B. vulgaris.This biochemical system adds to the potentially useful genes of this popular eco-model plant, stressing the relevance of continued development of a comprehensive toolbox for genetic, chemical and molecular biological investigations of this natural genetic resource (Nielsen et al., 2010;Van Molken et al., 2014a, 2014b;Christensen et al., 2019;Hauser et al., 2021).Indeed, the insect deterrent saponins of B. vulgaris are unique among crucifers, and the structurally remarkable indole phytoalexin cyclonasturlexin (P3) and phenyl-containing nasturlexin D (P6) are confined to a narrow group of cruciferous species.In general, the remarkable chemical profiles of B. vulgaris and N. officinale confirms the general value of screening wild species and non-model systems (e.g.Pedras et al., 2011;Pedras and Alavi, 2020;Zhou and Jander, 2021) as a supplement to the investigation of the usual model systems for insect and disease resistance mechanisms.

Plant material, cultivation and elicitation
Seeds of the G-type of Barbarea vulgaris were generally of accession B44 (NGB 31789) while seeds of the P-type of the same species were of accession B4 (NGB 23547) (Agerbirk et al., 2021b).Both accessions are available from https://nordic-baltic-genebanks.org, registered under the indicated NGB numbers.However, for experiment 1 ("leaf position experiment"), two other G-type and two other P-type accessions were used as detailed elsewhere (Agerbirk et al., 2022).Both accessions agree with descriptions of ssp.arcuata (Opiz.)Fries in standard floras (e.g.Hegi, 1958), but are currently not named to subspecies level due to their contrasting phylogeny (Lange et al., 2022).
Plants were sown and grown in a standard, peat based artificial soil and cultivated in a heated greenhouse (typically around 20 • C) with supplemental light (16 h L: 8 h D) and standard complete mineral fertilizer applied (Azelis pioner basis NPK 13-2-23, Tanggaard, Bjaeverskov, Denmark).Two weeks before elicitation experiments, individual rosette plantlets (3-4 weeks old, typically in the 4-6 true leaves stadium) were potted in individual pots (9 cm diameter, 7 cm height) and allowed liberal space (8 pots per tray of ca.1500 cm 2 ) to develop optimally.

Elicitation experiments
The general conditions were used for elicitation experiments No. 2 and 3, which took place in the spring (exp.2) and summer (exp.3) 2019 in the Copenhagen University greenhouse facilities.The treatment consisted of spraying with either 10 mM CuCl 2 (aq) or water as control.For spraying, in order to avoid contamination of our general facilities with CuCl 2 , both control and test plants were taken outdoor in mild weather, avoiding extremes of temperature or UV light.The plant to be sprayed was placed in a cardboard box open to one side, and sprayed with liberal amounts of the relevant solution both from below and above leaves using a simple plastic spray bottle obtained in a local supermarket (separate bottles for CuCl 2 (aq) and water).Surplus solution was allowed to drip off.Separate cardboard boxes were used for CuCl 2 (aq) and water to avoid contamination, and physical contact across the treatment groups was avoided.After the elicitation, plants were returned to the greenhouse and placed in a block design for attempted uniform growth conditions within the plant types (G or P).For each plant type, two blocks in trays placed diagonally were used for test (CuCl 2 ) or control plants.For daily sampling in the following three days, plants were taken at random from both diagonally placed blocks.In experiments comparing the two plant types (G and P), a symmetrical design to randomize the conditions experienced by each plant type was not used, but the G-type blocks and P-type blocks were kept at apparently similar conditions along the center of the same green-house shelf and were treated and sampled in parallel on the same days.
Three biological experiments were carried out in 2017-2019.Experiment No. 1 ("leaf position experiment") in June 2017 investigated relative phytoalexin levels in a gradient of leaf positions in both P and Gtype plants.The conditions differed somewhat from the general conditions (plants grown in a growth chamber, only leaves sampled, leaves dried over silica gel and stored frozen with silica gel until extraction).The four youngest fully developed or nearly fully developed rosette leaves were sampled from each rosette plant (Leaf 2-5), and only one sampling time was used; 2 days after elicitation, resulting in 20 P-type and 20 G-type treated leaf samples representing four leaf positions from 5 G-type and 5 P-type plants, and an equal number of control samples.GSL levels and lack of aliphatic GSLs from some of these leaves were reported elsewhere (Agerbirk et al., 2021b).For the present paper, we extracted all treated G-type leaves, leaves of three G-type controls and one P-type control, and the four leaf positions of one treated P-type plant in triplicate, using 10 mg dry leaf tissue for each extraction.
Experiment No. 2 ("P and G leaf elicitation experiment") in May 2019 was a 2 × 2 x 3 factor experiment with 3 biological replicates comparing two plant types (P and G) and two elicitation regimes (10 mM CuCl 2 (aq) or water) over 3 days (1, 2 or 3 after the treatment), resulting in 36 samples analyzed for phytoalexins and GSLs.
Experiment No. 3 ("G leaf versus root elicitation experiment") in August 2019 repeated the G-type half of experiment 2 but both leaves and roots of the G-type plants were sampled and the number of biological replicates set to 5 per treatment, resulting in 2 × 2 x 3 × 5 = 60 samples analyzed for phytoalexins and GSLs.One leaf GSL sample (Day 1 control #5) was dismissed as it exhibited clear signs of contamination with a root sample.

Extraction and preparation of extracts for metabolite analyses
A combined extraction of B. vulgaris leaves for both GSL and phytoalexin analysis were carried out as follows.For sampling leaves, only the terminal lobe of the selected rosette leaf was extracted.In experiments 2 and 3, we used the innermost fully or nearly fully developed rosette leaf.For sampling roots in experiment 3, peat around the entire root system P.D. Cárdenas et al. of the rosette plant was manually removed as well as possible, requiring around 1 min, and remaining peat rinsed of during a brief rinse with tap water followed by gently drying between tissue paper.Immediately before each extraction, the relevant plant part was picked and the fresh wt.(typically 0.2-0.7 g) was measured on an analytical balance in a beaker.Within ca. 1 min, extraction solvent, ca. 5 mL MeOH-H 2 O (7:3), was brought to boiling on a hotplate and poured over the plant material in the beaker, and extraction was performed for 60 s, keeping the solvent boiling using the hotplate.The extract was poured into a volumetric flask (10.00 mL) and a second round of extraction was performed.The combined extracts were adjusted to 10.00 mL by addition of extraction solvent (or in a few cases by evaporation of surplus solvent under a gentle air stream).The combined extract was poured into a beaker with attention to complete mixing, and 4.50 mL was taken twice for determination of phytoalexins and GSLs (Section 4.6.),respectively.For phytoalexin determination, the solvent was evaporated under a gentle air-stream and the residue was dissolved and transferred quantitatively in 3 × 200 μL of MeOH-H 2 O (1:2), centrifuged for 2 min at 20.000 g, filtered through a 0.22 μm filter (Merck Millipore, Cork, IRL) and subjected to UHPLC-analysis (Section 4.5.).

UHPLC-QToF MS of phytoalexins, glucosinolate hydrolysis products and flavonoids
Phytoalexins, GSL hydrolysis products and flavonoids were separated, identified and quantified by UHPLC-QToF MS, performed on a Dionex UltiMate 3000 Quaternary Rapid Separation UHPLC + focused system (Thermo Fisher Scientific, Germering, Germany).Injection volume was 10 μL.Separation was achieved on a Kinetex 1.7 μm XB-C18 column (100 × 2.1 mm, 1.7 μm, 100 Å, Phenomenex).For elution, 0.05% (v/v) formic acid in H 2 O and acetonitrile [supplied with 0.05% (v/v) formic acid] were employed as mobile phases A and B, respectively.Gradient conditions were as follows: 0.0− 0.5 min 10% B; 0.5− 4.0 min 10− 45% B, 4.0− 10.5 min 45-75% B, 10.5− 11.5 min 75-100% B, 11.5− 12.5 min 100% B, 12.5− 12.6 min 100− 10% B, and 12.6− 15.0 min 10% B. The flow rate of the mobile phase was 300 μL/ min.The column temperature was maintained at 30 • C. UV absorbance chromatograms were additionally recorded at 229, 260, 310, and 345 nm.The UHPLC was coupled to a Compact micrOTOF-Q mass spectrometer (Bruker, Bremen, Germany) equipped with an electrospray ion source (ESI).The ion spray voltage was maintained at +4500 V in positive ion mode.Dry temperature was set to 250 • C, and the dry gas flow was set to 8 L/min.Nitrogen was used as the dry gas, nebulizing gas, and collision gas.The nebulizing gas was set to 2.5 bar and collision energy to 10 eV in positive ion mode.MS spectra were acquired in an m/ z range from 50 to 1000 and MS/MS spectra in a range from 100 to 800.Sampling rate was 2 Hz.Na-formate clusters were used for mass calibration.All files were automatically calibrated by postprocessing.Chiral UHPLC-QToF MS of kjaerin was carried out using different conditions as previously reported (Müller et al., 2018).
Quantification was based on external calibration curves.Received synthetic phytoalexins were stored over silica gel, weighed (ca.0.3-1.8mg) using a 5 decimal analytical scale (MC 210 S, Sartorius, Göttingen, Germany), each dissolved in 250 μl 80% aq.MeOH and diluted to 100 μM stocks in 30% aq.MeOH (assuming 100% purity) used to prepare pure dilutions in the analytical range as well as mixtures, from which factor 2 dilution series were made from 10.0 μM to 9.8 nM (Table 1).
Serially diluted standards for rutin and F1 were made similarly except that a factor 3 dilution curve was used (rutin) or an equidistant curve was made in the relevant concentration range (F1) (Figs.S3F-G).Standards of raphanusamic acid were made in eight concentrations between 125 nM and 100 μM (Figs.S2H-I).Standards of GSL products were as previously reported (Agerbirk and Olsen, 2015;Agerbirk et al., 2018), based on ε 245 = 1.9 • 10 4 M − 1 cm − 1 for barbarin (Kjaer and Gmelin, 1957).For peak recognition and integration, the automatic integration tool in Bruker Compass Data Analysis and the calibration curves (Table 1) were used.For quantification of analytes (in samples as well as unknowns), we created extracted ion chromatograms for the calculated monoisotopic mass of proton adducts using stringent conditions: ± 0.005 Da for metabolites below 500 Da and ± 0.010 Da for metabolites with mass equal to or higher than 500 Da.Initial inspection of chromatogram shapes and peak areas supported the use of these stringency parameters; less stringent extracted ion chromatograms did not generally result in higher peak areas while more stringent parameters affected peak shape and area.Manual integration was used for distinct peaks that escaped recognition by the program.For peaks well established in the plant type, manual integration of apparent traces or noise at the correct retention time was carried out for statistical reasons even if presence of the analyte was uncertain; the corresponding results served as maximum estimates.

Glucosinolate analysis using desulfation and UHPLC-QToF MS analysis
For GSL determination, the crude extract was stored frozen at − 20 • C, then thawed and applied to an enzymatic on-column procedure for GSL desulfation followed by UHPLC-DAD-QToF MS analysis of the resulting desulfoGSLs, both exactly as described elsewhere (Agerbirk et al., 2022).For logistic reasons, the conversion to desulfoGSLs had to be carried out during three separate days; samples from all treatment groups as well as blanks and external standards were distributed symmetrically over the three days.Linearity testing showed linearity of the UV detector up to at least 6 mM of desulfoGSL, which was well above even major GSL peaks in the samples.However, linearity using the MS detector was much more limited, but extending up to ca. 10 μM in samples.This range allowed determination of minor GSLs such as PE and 4moIM from extracted ion chromatogram traces, while all major GSLs had to be determined from the UV peak areas measured at λ = 229 nm.For the peaks quantified from UV peak areas, levels were calculated from comparison with a reference of 0.535 μmol benzylGSL subjected to the desulfation procedure (corresponding to 0.9 mM desulfo benzylGSL in injected samples).For the peaks quantified from extracted ion chromatogram peak areas, levels were calculated from comparison with a reference of 8.0 μM desulfo BAR.From integration of minor peaks, the LOD was estimated as < 0.1 nmol/g fresh wt.

Fig. 2 .
Fig. 2. Two suggested biosynthetic routes to nasturlexin D (P6), a phenylcontaining phytoalexin detected in Barbarea vulgaris.More potential intermediates are known, and more metabolic connections from the left hand side pathway to the m-hydroxy pathway are possible (Pedras and To, 2018).Abbreviations PE, phenethylGSL; 3hPE, 3-hydroxyPE; other abbreviations as in Fig. 1.

Fig. 4 .
Fig. 4. Representative UHPLC-MS trace of a G-type Barbarea vulgaris leaf sample, showing a sample (Day 3, treated) with relatively high levels of phytoalexins (P3, P6) and glucosinolate hydrolysis products (resedine, barbarin).A. MS base peak chromatogram and UV (229 nm) chromatogram.For selected peaks that were distinguishable in these unspecific traces, identities are indicated.B. Extracted ion chromatograms of selected analytes using the calculated exact m/z value for [M+H] + ions ±0.005 Da for metabolites up to 500 Da and ±0.01 Da for higher masses.

Fig. 5 .
Fig. 5. Representative UHPLC-MS data for treated and control G-type plants in experiment 2. In each panel, the upper trace represents the extracted ion chromatogram ( ± 0.005 Da) for [M+H] + of the phenyl containing phytoalexin P6 and any isomer, while the lower trace represents the extracted ion chromatogram for the indole phytoalexins P2, P3 and any isomer.The legends in panels A-D specify day and treatment/control.

Fig. 6 .
Fig. 6.Phytoalexin response in young rosette leaves to CuCl 2 -treatment of leaves in two genotypes (G-type and P-type) of the plant Barbarea vulgaris.N = 3 treated and 3 control plants of each plant type on each day; since control plants were generally devoid of phytoalexins except a trace of nasturlexin in one control G-type on day 3, only levels for treated plants are shown.A-C, levels of individual phytoalexins as a function of time after elicitation; D-F, pairwise correlation plots of phytoalexins or structural groups of phytoalexins, each dot refers to the same leaf.Data in plots A-C were analyzed by two-way ANOVA; statistical significance of effect of treatment (H o : no effect of treatment) is indicated as follows: *, P < 0.05; **, P < 0.01; ***, P < 0.001; n. s., not significant.No significant effects of day or treatment x day effect was found for any of the three phytoalexins.Bars indicate means, whiskers indicate standard deviations.In E, the correlation analysis of the Ptype dataset was repeated after omission of one observation pair, showing that the apparently high correlation was due to this observation pair only and hence with low support.

Fig. 9 .
Fig. 9. Levels of selected glucosinolate hydrolysis products as a function of time in young rosette leaves after CuCl 2 -treatment (treated) or mock treatment (control) of G-type Barbarea vulgaris plants.The hydrolysis products considered are (A) barbarin, (B) resedine, and (C) raphanusamic acid.Bars depict means, whiskers indicate standard deviations.

Table 1
Chromatographic parameters for B. vulgaris metabolites detected at the UHPLC-QToF MS conditions used for phytoalexin determination.

Table 2
Main characteristics of three experiments (Section 4.3.)considering metabolites in leaves after spraying B. vulgaris plants with either an aqueous solution of CuCl 2 ("treated") or pure water as a mock treatment ("control").
a Counted from the youngest visible leaf in the rosette (not yet fully developed) defined as position 1. b Number of treated plants: number of controls.