The following genetic backgrounds were used for the non-targeted metabolomic analysis: P. patens (Hedw.) Bruch & Schimp: Wild type ‘Gransden’, Ppaos1 ko 21 (in the following referred to as Ppaos1), Ppaos2 ko 71 (in the following referred to as Ppaos2). These P. patens KO-mutants were generated as described by Scholz et al., 2012 and by the methods described by Frank et al., 2005.

Moss was grown on agar plates on Knop medium as described (Reski and Abel, 1985), containing for 1 L: 1000 mg Ca(NO₃) x 4 H₂O, 250 mg KCl, 250 mg KH₂PO₄, 250 mg MgSO₄ x 7 H₂O, 12.5 mg FeSO₄ x 7 H₂O (pH 5.8) and 15 g/L agar was added to the medium prior to autoclaving. Before inoculation, the inoculation moss culture was transferred to fresh Knop medium and afterwards homogenized by using an Ultra Turrax T25 basic (IKA, Stauffen, Germany) at a frequency of 19,000 min^-1. Homogenized moss was then dropped on agar plates. Plates were inoculated with drops of about 4 weeks old homogenized liquid cultures, setting about 9 drops per plate. Plates grew at long day conditions (16 h/25°C day to 8 h/18°C night cycle) with light exposure of about 70% (55 μmol/m²) under normal air atmosphere. Moss was harvested after about 60 days of growth.

Moss wounding was performed by scraping off about 100 mg fresh gametophores from agar plate cultures and transferring them into a clean 1.5 mL reaction tube. The moss was then squashed with a plastic pestle. Subsequently, the sample was incubated under light for 30 min. The wounded sample was then frozen in liquid nitrogen. Control plants were not wounded, but incubated in 1 mL sterile tap water for 30 min under light. Three parallel cultures of each genotype and two extractions of each culture were analyzed from two independent experiments.

This procedure was modified according to a method described earlier (Matyash et al., 2008). Samples (about 100 g fresh weight in 1 mL sterile tap water) were transferred to glass vials and the reaction tube in which the sample was stored was washed with 1.5 mL methanol. Afterwards, 5 mL methyl-tert-butyl ether (MTBE) were added, shaken vigorously and the sealed vial was shaken for 1 h at 4°C in the dark. Then, 400 μL LC-MS grade water were added to the sample, shaken vigorously and incubated for 10 min at room temperature under a fume hood. The sample was then centrifuged for 15 min at 450 x g and the supernatant was transferred to a new glass vial. To the remaining lower phase, 1.4 mL methanol:water, (3:2, v/v) and 2.6 mL MTBE were given. After vigorously shaking the sample, the vial was incubated at room temperature for 10 min under a fume hood and afterwards centrifuged for 15 min at 450 x g. The upper phase was then combined with the first upper phase in the new vial.

The combined non-polar upper phase was dried under streaming nitrogen and the remaining sample resolved in 1 mL methanol, then transferred to a new 1.5 mL reaction tube and dried again under streaming nitrogen. Remaining sample was resolved in 30 μL methanol, 30 μL acetonitrile, and 75 μL water. After shaking vigorously, the vial was centrifuged for 10 min at 450 x g and the supernatant was transferred to a new glass vial (storage at -20°C). The polar lower phase of the extraction described in the previous paragraph was dried under streaming nitrogen in 2 mL reaction tubes and resolved in 5 μL methanol, 10 μL acetonitrile, and 120 μL water. Afterwards, the vial was shaken vigorously and centrifuged for 10 min at 450 x g at room temperature. The supernatant was then transferred to new glass vials and stored at -20°C.

This procedure followed a method described earlier (Feussner and Feussner, 2019). For non-targeted analysis of wounded P. patens, non-polar and polar extractions of the samples were analyzed via ultra-performance liquid chromatography (UPLC, ACQUITY UPLCTM System, Waters Corporation, Milford, USA) coupled to a photo diode array detector (UPLC eLambda 800 nm, Waters Corporation, Milford, USA), followed by an orthogonal electrospray ionization time-of-flight mass spectrometer (ESI-TOF-MS, LCT Premier, Waters Corporation, Milford, USA). For analysis of the polar phase, the UPLC was equipped with an ACQUITY UPLC HSS T3 column (1.0 x 100 mm, 1.8 μm particle size, Waters Corporation, Milford, USA) held at 40°C and a flow rate of 0.2 mL/min. The following gradient program was run for the polar phase: 0 – 0.5 min 1% B, 0.5 – 3 min 1% B to 20% B, 3 – 8 min 20% B to 100% B, 8 – 10 min 100% B, 10 – 10.1 min 100% B to 1% B, 10.1 – 14 min 1% B (solvent system A: water:formic acid (100:0.1, v/v); B: acetonitrile:formic acid (100:0.1, v/v)). Analysis of the non-polar phase applied using an ACQUITY UPLC BEH C18 column (1.0 x 100 mm, 1.7 μm particle size, Waters Corporation, Milford, USA) at a temperature of 40°C and a flow rate of 0.2 mL/min. The gradient program for the non-polar phase was as follows: 0 – 0.5 min 40% B, 0.5 – 6 min 40% B to 100% B, 6 – 12 min 100% B, 12 – 12.1 min 100% B to 40% B, 12.1 – 15 min 40% B (solvent systems A and B equal to polar phase). UV/VIS data were collected in the range of 190-800 nm with a 1.2 nm resolution.

The ESI-TOF-MS was run at both positive and negative ionization mode (ESI) for all samples. In positive ESI-mode, masses from m/z = 85.00 – 1200.00 were scanned over a runtime of 10 min using a capillary voltage of 2700 V. In negative ESI mode, masses from m/z = 50.00 – 1200.00 were scanned over a runtime of 13 min using a capillary voltage of 2500 V. In both modes, the cone voltage was maintained at 30 V, the desolvation temperature at 350°C and the source temperature at 80°C. Data acquisition was carried out by using the MassLynx software (MassLynx V4.1 SCN779, Waters Corporation, Milford, USA) in centroid data format. Nitrogen was used as cone and desolvation gas at a flow of 30 and 800 l/h, respectively. The dynamic range enhancement mode was used for data recording. All analyses were calibrated by applying the lock spray reference compound leucine-enkephaline (Sigma-Aldrich, Deisenheim, Germany) at a concentration of 0.5 μg/mL in acetonitrile:water (50:50, v/v) and a flow rate of 20 μl/min.

The raw mass spectrometry data acquired by the UPLC-HRMS analysis were preprocessed using the MarkerLynx Application Manager for MassLynx software (Waters Corporation, Milford, USA). This software tool manages peak picking and peak alignment from the raw data files to generate data matrixes. In this approach, four data matrixes are obtained: two for the metabolites of each extraction phase (polar and non-polar), analyzed in the positive and negative ESI mode. For peak detection, the following parameters were used: Initial retention time 0.30 min, final retention time 10.00 min, low mass m/z 50, high mass m/z 1200, extracted ion chromatogram window 0.03 Da. Apex track peak parameters were set to automatic.

Identification of high-quality metabolite features was conducted via the MarVis toolbox (Kaever et al., 2015). A metabolite feature is characterized by m/z, a retention time (RT) and the corresponding intensity profile received from all analyzed samples. Ranking of data applied in MarVis-Filter via the ANOVA ranking method. This initial ranking was combined with adjustment for multiple testing, the Benjamini-Hochberg algorithm (false discovery rate, FDR). As a threshold, only features with a FDR smaller than 1 x 10^-4 were used for further data analysis. Adduct correction was performed for the four filtered data sets (non-polar phase in positive and negative ESI mode, polar phase in positive and negative ESI mode) with concern for the following adducts: [M+H]⁺, [M+Na]⁺, [M+NH₄]⁺ (for positively charged adducts); [M-H]^-, [M+CH₂O₂-H]^-, [M+CH₂O₂+Na-2H]^- (for negatively charged adducts).

In order obtain a data set of valid metabolite features, two independent wounding experiments were analyzed by non-targeted metabolomics. Metabolite features which appeared in both data sets were used for further data analysis. All in all, 800 metabolite features remained, which were sorted into a 1D-SOM with 5 clusters, of which 4 showed a wound-depended metabolic pattern. A database search within a mass window of 0.005 Da was performed using internal databases, KEGG (Kanehisa, 2002) and AraCyc (Mueller et al., 2003).

MS/MS experiments were performed via a UHPLC-ESI-QTOF-MS device. For UHPLC, an Agilent 1290 Infinity series UHPLC system (Agilent Technologies, Böblingen, Germany) equipped with an ACQUITY UPLC BEH C18 column (2.1 x 100 mm, 1.7 μm particle size, Waters Corporation, Milford, USA) kept at 40°C was used. Mass detection was performed with an Agilent 6540 UH Accurate-Mass-Q-TOF mass spectrometer (Agilent Technologies, Böblingen, Germany). Same solvents and gradients were used as for the non-targeted LC-MS analysis for a comparability of the two separations. The mass spectrometer was operated using an ESI source with Agilent Dual jet Stream Technology (Agilent Technologies, Böblingen, Germany) in negative ionization mode. Ionization parameters were as follows: gas temperature 300°C, gas flow 8 L/min, nebulizer pressure 35 psi, sheath gas temperature 350°C, sheath gas flow 11 L/min, VCap 3.f kV, nozzle voltage 100 V. As a collision cell, a linear hexapole collision cell with nitrogen as collision gas at various collision energies (20 eV – 40 eV) was used. For data acquisition, the Mass Hunter Workstation Acquisition Software was used (Version: B05.01.), with Mass Hunter Qualitative Analysis software as analysis tool. For structure elucidation, the fragmentation patterns of compounds were compared with MS/MS spectra deposited in the Metlin (Guijas et al., 2018) database, known from the literature (Oliw and Hamberg, 2019) or by interpretation of fragmentation rules.

For identifying metabolites deriving from the fatty acid 20:4 (ARA), a sample set of wounded gametophores of mutant lines for the Δ6-desaturase Ppdes (Girke et al., 1998) and the Δ6-elongase Ppelo were analyzed (Zank et al., 2002). For both genes new KO-lines were generated in the ‘Gransden’ wild-type strain. Therefore, P. patens protoplasts were transformed (Frank et al., 2005) with a KO-construct for the gene Ppdes (Pp3c5_9590) or PpELO (Pp3c17_7080). The KO-constructs for homologous recombination consisted of 5’ and 3’ flanking sequences of the respective genes and included a nptII selection cassette. Genomic DNA from gametophytic tissue was isolated using cetyl trimethylammonium bromide extraction method. PCR was performed to select transgenic lines in which either Ppdes or Ppelo was knocked out using the following primer combinations: Ppdes- 5’UTR: for 5’- CAAGGTGTTTCATCTCATGGTATTAG-3’/rev 5’-CCATAATAATGTGTGAGTAGTTCC-3’; 3’UTR: for 5’-CTGAGACTGTATCTTTGATATTCTTG-3’/rev 5’-ATGGCTACAGAAGAATGGATGC-3’; Ppelo- 5’UTR: for 5’-CAGCCACATTGTGACTATTCCCC-3’/rev 5’- CCATAATAATGTGTGAGTAGTTCC-3’; 3’UTR: for 5’-CTGAGACTGTATCTTTGATATTCTTG-3’/rev 5’-CCATCTGAAGTGCTCAGGCCG. To confirm the KO of either Ppdes or Ppelo fatty acid profiles were analyzed via GC and these analyses revealed the published fatty acid profiles (Girke et al., 1998; Zank et al., 2002). The lines Ppdes28, Ppdes51, Ppelo35 and Ppelo131 were selected for this work and further analyzed.

In order to identify wounding-related oxylipins in P. patens and to differentiate between an AOS-dependent as well as an AOS-independent accumulation, a non-targeted metabolite fingerprinting approach was applied. As genetic backgrounds, P. patens WT, Ppaos1 and Ppaos2 KO-mutants were used. For comparison we followed a previously optimized protocol and each genotype, unwounded control samples as well as samples harvested 0.5 hours past wounding (hpw) were analyzed by UPLC-HRMS (Stumpe et al., 2006; Stumpe et al., 2010; Scholz et al., 2012).

Several thousand features from data matrices of the polar and the non-polar extraction phase measured both in positive and negative ionization mode were used for data mining. 800 features with a false discovery rate (FDR) < 1 x 10^-4) occurred reproducible in two independent wounding experiments ( Supplementary Table 1 ). They were used for generating a principle component analysis (PCA) ( Figure 1A ) to investigate sample-based clustering. It shows that the unwounded moss samples clustered together, implying that the different genotypes (WT, Ppaos1 and Ppaos2) did not have a significant effect on the metabolome of untreated P. patens. However, wounded samples are clearly separated from unwounded samples along the principle component (PC) 1 and are more widespread along the PC2, implying that the KO-lines do react differently to wounding stress. Remarkably, wild type and Ppaos2 lines seems to react more similarly on metabolic level to the wounding stress than Ppaos1, indicated by a clear separation of Ppaos1 from the other two lines.

In order to analyze the data on a global metabolite level and to identify compounds of interest, the 800 features were used for clustering by means of one-dimensional self-organizing map (1D-SOMs) using the MarVis-Cluster application of the MarVis toolbox (Kaever et al., 2009). Features with similar intensity profiles cluster together and are represented by five prototypes ( Figure 1B ). The heat map colors represent average intensity values and the width of each cluster is proportional to the number of features assigned to this cluster. The prototypes 1, 3, 4 & 5 represent metabolite features, which are either depleted (prototype 1, 182 features) or enriched (prototype 3 - 5, 505 features) upon wounding in all genotypes. Prototype 5 (marked by a red box, 107 features) summarizes features that represent compounds accumulating in all genotypes except in Ppaos1 after wounding. As already shown by PCA ( Figure 1A ), Ppaos2 and WT moss are indistinguishable on an overall metabolite level, whereas Ppaos1 shows a unique chemotype after wounding.

In order to identify the metabolites behind the features of interest and to reconstruct biosynthetic pathways, the accurate mass information of the features was used for a metabolite set enrichment analysis (Kaever et al., 2014). The pathway analysis identified 20 metabolites (represented by 46 features) from prototypes 3 - 5, which are connected to four variants of the oxylipin-pathway. These pathways start from different fatty acids: roughanic acid (RA, 16:3n-3), α-LeA, linoleic acid (LA, 18:2n-6) or arachidonic acid (ARA, 20:4n-6), respectively ( Figure 2 ). The identities of 11 of the 20 metabolites were unequivocally verified by high resolution-MS/MS analyses and/or by co-elution with identical standards ( Supplementary Table 2 ).

The first oxylipin formed from α-LeA is 13-hydroperoxy octadecatrienoic acid (13-HPOT). It was enriched in wounded samples of the WT and mutant lines ( Figure 2A ). Other AOS-independent metabolite patterns were observed for those metabolites deriving directly from 13-HPOT via LOX and hydroperoxide lyase (HPL) activities, namely 13-keto octadecatrienoic acid (13-KOT), traumatin and traumatic acid, as well as 13-hydroxy octadecatrienoic acid (13-HOT), which is spontaneously formed from 13-HPOT. 13-HOT accumulated in Ppaos1 and Ppaos2 even stronger than in WT samples. In contrast to that, OPDA and two variants of α-LeA-derived ketols, which are formed in direct consequence of PpAOS1 activity, were absent in the Ppaos1 mutant, but still present in Ppaos2 and WT samples. 12,13-epoxy octadecatrienoic acid (12,13-EOT), the direct reaction product of AOS1 and the precursor of OPDA, is very unstable and could therefore not be detected during measurements.

Furthermore, several metabolites of the LA-, ARA-, and the RA-derived LOX-pathway ( Figures 2B–D ) accumulated after wounding. The LA-derived LOX-product 13-hydroperoxy octadecadienoic acid (13-HPOD) as well as its follow-up products 13-keto octadecadienoic acid (13-KOD), 13-hydroxy octadecadienoic acid (13-HOD), 12,13-epoxy 9-hydroxy octadecadienoic acid (12,13-E-9-HOD) and 9,12,13-trihydroxy octadecaenoic acid (9,12,13-TriHOME) accumulated in all wounded genotypes. No compounds deriving from AOS1 or AOS2 catalysis were detected in the LA-pathway. ARA was the only LOX-substrate that was identified in this experiment. It showed a wound dependent accumulation in all genetic backgrounds. 12-hydroperoxy eicosatetraenoic acid (12-HPETE), the only ARA-derived oxylipin detected here, showed a similar accumulation pattern as the α-LeA-, LA-, and RA-derived hydroperoxides. The assumed ARA-derived PpAOS1 product 11-oxo prostatrienoic acid (11-OPTA) was not detectable in this experiment. Oxylipins deriving from RA were detected as 11-hydroperoxy hexadecatrienoic acid (11-HPHT) and 11-hydroxy hexadecatrienoic acid (11-HHT). With dnOPDA and a RA-derived ketol, two wounding related compounds with an AOS-dependent accumulation pattern were identified in the RA-derived pathway.

To complement the pathway reconstruction, transcriptome data from detached leaflets of P. patens (representing mild wounding stress) were analyzed (Hiss et al., 2014). Since experimental conditions were different, changes for the transcript levels of PpLOX1-7, PpAOS1 and 2 as well as PpAOC1 and 2, 1 and 2 hpw are only depicted as arrows showing a general trend in Figure 2A . Six of the seven LOXs present in P. patens were upregulated upon wounding. PpLOX5 was downregulated. Transcript levels of PpAOS1 rose upon wounding, while the transcription rate of PpAOS2 did not change. This is in agreement with previous findings that PpAOS1, but not PpAOS2, is involved in the wounding response of P. patens (Scholz et al., 2012). Additionally, the OPDA forming enzymes PpAOC1 and PpAOC2 were upregulated upon wounding, too. Together these independent transcriptome data support and complement the metabolite data from the shown oxylipin pathways.

Previous reports showed that the oxylipin synthesis up to OPDA and dnOPDA are present in P. patens while the occurrence of jasmonates is still under discussion. We wondered, if P. patens forms other yet unknown metabolites, which would possibly derive from OPDA or dnOPDA. These compounds could possibly fulfill similar functions as ligands of COI1, as it was described for the liverwort M. polymorpha, where dnOPDA has the same function as JA-Ile in flowering plants (Monte et al., 2018).

In our non-targeted metabolome data only 46 out of 505 features with a wound-accumulating pattern could be assigned to the LOX-pathways ( Figure 1B , prototype 3-5 and Figure 2 ). Of those markers that showed an AOS1-dependent pattern, 8 out of 107 features could be assigned to 5 AOS1-dependent metabolites, like OPDA, dnOPDA and ketols ( Figure 1B , prototype 5). The low coverage of metabolome data by searching public databases, like KEGG and MetaCyc, is a major bottleneck of the non-targeted approach, especially for the metabolome of not extensively studied organisms. This bottleneck can be addressed by implementing in-house databases into the framework of metabolite set enrichment analysis in MarVis-Pathway. We used the strategy of in silico database extension to increase the number of hypothetical oxylipin derivatives. Therefore, the core structures OPDA as well as dnOPDA were attached in silico to moieties like hydroxy, carboxy, sulfate, hexosyl, and amino acid groups, as well as combinations of those modifications. By that strategy, a database of 708 hypothetical OPDA and dnOPDA derivatives was created. Beside of OPDA and dnOPDA, 20 features representing five compounds could be identified in the data set of the non-targeted metabolome analysis by searching this in silico extended oxylipin database. These wound- and AOS-dependent compounds contain 18 or 16 carbon atoms (like OPDA or dnOPDA) and either an equal or higher number of oxygen and hydrogen atoms as OPDA or dnOPDA ( Figure 3A ). OPDA- or dnOPDA-like compounds linked to an amino acid or a sulfate residue were not detected.

In order to make further assumptions of the molecular structure of these compounds, LC-HRMS/MS analyses were applied to generate accurate mass fragmentation patterns. First, the identity of OPDA was confirmed. It showed three main fragments in negative ionization mode, which were assigned to loss of water [M-H₂O-H]^- 273.186, loss of the carboxy group [M-CO₂-H]^- 247.207, as well as the analytical OPDA fragment of m/z 165.125 for [M-C₇H₁₀O₂ -H]^- ( Figure 3B ). These fragments were in agreement with an earlier report (Oliw and Hamberg, 2019). In a similar way the identity of dnOPDA was confirmed ( Figure 3B ). Two other wound- and AOS-dependent compounds identified in the data set are isobaric to OPDA (292.2028 Da, C₁₈H₂₈O₃) and to dnOPDA (266.1725 Da, C₁₆H₂₄O₃), but show earlier RTs in comparison to OPDA and dnOPDA ( Figure 3A ). The fragmentation pattern of these two compounds showed a remarkable similarity to -OPDA and dnOPDA, including the three fragments m/z 273.184, m/z 247.207 and m/z 165.129 of 12-OPDA, and m/z 219.175, m/z 191.144, m/z 165.13 and m/z 59.014 of dnOPDA, respectively ( Figure 3B ). In accordance with a report of Oliw and Hamberg, these two compounds were therefore assigned as iso-OPDA (12‐oxo‐9‐(15Z)‐phytodienoic acid) and iso-dnOPDA, respectively (Oliw and Hamberg, 2019).

The elemental composition of the three remaining unknown compounds (UNC) show 18 carbons but a higher number of hydrogen and oxygen atoms in comparison to OPDA. In detail, UNC328, shows an exact mass of 328.2248 Da and a deduced sum formula of C₁₈H₃₂O₅ and is represented by 3 features ( Figure 3A ). For UNC344 (344.2200 Da, 3 features, Figure 3A ) an elemental composition of C₁₈H₃₂O₆ and for UNC326 (326.2086 Da, 3 features, Figure 3A ) a formula of C₁₈H₃₂O₅ was deduced. All three UNCs eluted earlier than OPDA on reverse phase UPLC, indicating a higher hydrophilicity than 12-OPDA because of additional functional groups, like possibly hydroxy-, oxo- or carboxy-groups.

The wound- and AOS-dependent compound UNC328 was analyzed by MS/MS fragmentation ( Supplementary Figure 1 ). MS/MS analyses in negative ionization mode showed two times a loss of water (m/z 309.209 and m/z 291.198), which could suggest for additional hydroxy-groups in comparison to OPDA. The main fragments of UNC328 show an accurate ion mass of m/z 197.1194 and m/z 185.1194 with deduced elemental compositions of C₁₁H₁₇O₃ and C₁₀H₁₇O₃, respectively and with it a mass shift of one carbon atom. Although it shares two of its fragments (m/z 291.198 and m/z 247.207) with OPDA and iso-OPDA, the structure for the compound UNC328 could not be solved by MS/MS analysis.

The two other putative oxylipins (UNC344 and UNC326, Figure 3A ) were also analyzed via MS/MS fragmentation. However, no suitable fragment pattern were obtained for both compounds because of low signal intensities. All in all, OPDA, iso-OPDA, dnOPDA, iso-dnOPDA, and three unknown oxylipins consisting of 18 carbon atoms and a higher number of oxygen and hydrogen atoms in comparison to OPDA were detected in wounded P. patens samples with an AOS-dependent intensity pattern.

JA-Ile is the active phytohormone in flowering plants formed after wounding. Neither JA and JA-Ile nor their direct precursors, OPC-8, OPC6 and OPC-4 were detected in any of the analyzed samples of this study. However, a compound of the same accurate ion mass as JA ([M-H]^- 209.1173) that accumulates upon wounding was detected in P. patens WT, Ppaos1 and Ppaos2 lines. This unknown compound (UNC210) elutes at a slightly earlier RT than the JA standard and is not diminished in Ppaos1 and Ppaos2 ( Supplementary Figure 2 ). MS/MS-fragmentation experiments of JA and UNC210 in the negative ionization mode show for both compounds a main fragment of m/z 59.01, which represents the carboxylic acid moiety in JA. While m/z 59.01 is the only fragment generated by JA, UNC210 produces further fragments by collision induced fragmentation: m/z 191.106 for [M-H₂O-H]^-, m/z 147.116 for [C₁₁H₁₅]^-, m/z 135.115 for [C₁₀H₁₅]^-, m/z 119.085 for [C₉H₁₁]^-, and m/z 93.069 for [C₇H₉]^-. To investigate, if UNC210 is derived from α-linolenic acid or from different PUFAs we tested P. patens mutant lines of Δ6 desaturase (Ppdes) as well as of Δ6 elongase (Ppelo) for the occurrence of UNC210. Ppdes harbors no γ-18:3, 20:3(n-6), 20:4(n-6) and 20:5 PUFAs (Girke et al., 1998). Interestingly, in Ppdes lines wound induced accumulation of UNC210 is reduced. However in Ppelo mutants, which harbor no C20-fatty acids UNC210 is absent and does not accumulate after wounding (Zank et al., 2002). UNC210 therefore derives from C20 PUFAs and is formed independently from the activity of PpAOS1 or PpAOS2.