Comprehensive Analysis of the Alternaria Mycobolome Using Mass Spectrometry Based Metabolomics

Scope: Alternaria fungi are widely distributed plant pathogens infecting grains and vegetables and causing major harvest losses in the ﬁeld and during postharvest storage. Besides, consumers are endangered by the formation of toxic secondary metabolites. Some of these secondary metabolites are chemically characterized as mycotoxins, but the majority of the Alternaria mycobolome still remains unknown. Methods and results: Fourier-transform ion cyclotron resonance mass spectrometry (FTICR-MS) and LC-MS/MS are combined for the non-targeted and targeted analysis of the metabolome of three A. alternata isolates and one A. solani isolate. Due to the ultra-high resolution of FTICR-MS, unique molecular formulae are assigned to measured m/z signals. The molecular formulae are matched to entries of the databases Antibase and Kyoto Encyclopedia of Genes and Genomes. The non-targeted analysis of the fungal extracts reveals variations in the secondary metabolite proﬁle of A. alternata and A. solani . Diﬀerences in the biosynthesis of dibenzo- 𝜶 -pyrones, perylene quinones, tentoxin, and tenuazonic acid of the A. alternata and A. solani isolates are determined applying targeted LC-MS/MS. Conclusion: FTICR-MS analyses reveal clear diﬀerences in the metabolic proﬁle of the A. solani and the A. alternata isolates.


Introduction
Fungi of the genus Alternaria are widely distributed on seeds, plants, animals, in the soil, and in the atmosphere. [1] Growing in various regions and during different seasons, Alternaria species exhibit endophytic, saprophytic, or pathogenic growth. [1,2] The fungi decompose natural as well as artificial substrates [3] and cause plant diseases such as black rot of tomato, olive, and carrots and black and grey rot of citrus fruits. [1,2] Besides, cereals such as wheat, barley, oats, and sorghum are frequently infected by Alternaria fungi [4] resulting in losses of agricultural products in the field and during postharvest storage. [5] Common Alternaria species are Alternaria (A.) alternata, A. tenuissima, A. arborescence, A. radicina, A. brassicae, A. brassicicola, and A. infectoria. [6] In addition to losses of agricultural goods, Alternaria fungi endanger consumers by the production of mycotoxins and secondary metabolites with partly unknown toxicological potential. [7] The mycotoxins accumulate in agricultural products leading to a decline of the food quality. [8] Over 70 different secondary metabolites are described in the literature and more than 30 exhibit toxicological potential. [9] Common Alternaria mycotoxins belong to five different structural classes, namely 1) tetramic acid derivatives (tenuazonic acid [TA]), 2) dibenzo--pyrones (alternariol [AOH], alternariol monomethyl ether [AME], and altenuene), 3) perylene derivatives (altertoxin I [ATX I], altertoxin II [ATX II], alterperylenol [ALTP], and stemphyltoxin III [STTX III]), [2] 4) miscellaneous structures (tentoxin [TEN]), and 5) A. alternata f. sp. lycopersici toxins (AAL-toxins). [9] In previous studies, the benzopyrones exhibited genotoxic, cytotoxic, and mutagenic effects in vitro [2,5] and caused DNA damages in human colon carcinoma cells. [10] Additionally, AOH and AME were described to act as topoisomerase poison and to inhibit the catalytic activity of topoisomerase II . [5] While the acute toxicity of AOH and AME is rather low, [4] TA is acutely toxic to mice, chicken, and dogs. [2] No mutagenicity was observed for the tetramic acid derivative in bacterial systems, [5,11] whereas the altertoxins were mutagenic in the Salmonella Ames Test [11,12] and strongly genotoxic in mammalian and human cells. [13,14] Contrarily to the other mycotoxins, TEN is www.advancedsciencenews.com www.mnf-journal.com characterized as phytotoxin and inhibits the cyclic photophosphorylation in chloroplasts. [15] As Alternaria mycotoxins are neither legislatively controlled nor routinely analyzed, [16] they are considered as "emerging mycotoxins." [17] Besides, the modifications of AOH and AME with sulfates and glucosides attract increasing attention. [18] These so-called "modified mycotoxins" are either produced by the fungi themselves or the mycotoxins are metabolized by plants for detoxification purposes. [19][20][21] After oral consumption, the modification is potentially hydrolyzed during digestion, which releases the aglycon. [22][23][24] In a recently developed LC-MS/MS method, AOH-9-glucoside and AME-3-sulfate were detected in naturally contaminated foods demonstrating the necessity to routinely analyze also the chemically modified mycotoxins in agricultural products. [17] Furthermore, the modified forms should be included in future risk evaluations [19] and also added to mass spectrometric databases.
In 2011, the Panel on Contaminants in the Food Chain of the European Food Safety Authority (EFSA) assessed the risk to human health originating from Alternaria mycotoxins in agricultural products. The estimated chronic dietary exposure of AOH and AME exceeded the threshold of toxicological concern (TTC) of 2.5 ng kg −1 body weight per day and, therefore, additional toxicity data are indispensable for further risk evaluations. TA and TEN are non-genotoxic and the estimated chronic dietary exposure did not reach the TTC value of 1500 ng kg −1 body weight per day. [9] However, recent studies, particularly on infant foods, highlighted that these TTC values were exceeded [25] and required the establishment of a legal limit for TA in food products for infants. [26] To perform reliable risk assessments, the metabolic capabilities of Alternaria fungi need to be investigated and various species need to be identified correctly. Traditionally, the systematics of Alternaria fungi was based on morphological characteristics, [3,8] but morphology alone was insufficient due to overlapping traits of closely related species. [2] In addition to 3D sporulation patterns on agar plates, [8,27,28] Alternaria species were also classified based on the sequence variation in the translation elongation factor 1-. [8] The analysis of further genes resulted in distinct Alternaria species clusters. [3] As DNA-based studies did not always match the species-groups defined in morphological surveys, [1] excreted metabolites produced by the fungi were analyzed by HPLC-UV and LC-MS/MS and used for their differentiation. [29][30][31] The species group of A. infectoria was separated from A. arborescence, A. alternata, and A. tenuissima, but the latter were indistinguishable based on the mycotoxin profile obtained by LC-MS/MS. [30,31] As high-resolution mass spectrometers offer the simultaneous detection of hundreds to thousands of metabolites, these instruments are increasingly in use for the differentiation of fungal species. [32,33] In addition to the species segregation, high-resolution mass spectrometers improve the holistic characterization of the fungal mycobolome. The comprehensive detection of low-molecularweight metabolites of an organism is called metabolomics. [34] As the metabolome of organisms varies depending on genotype, cell cycle stage, or environment, [35] metabolomics approaches focus on the detection of a wide range of possibly produced compounds. [34] It has to be kept in mind that sample preparation and the selection of the ionization mode in the electrospray ionization (ESI) source will significantly influence the detected metabolite profile. [34,36] Metabolomics approaches commonly use time-of-flight (TOF), Orbitrap, and Fourier transform ion cyclotron resonance (FT-ICR) mass analyzers. [34] As FTICR-MS combines ultra-high mass resolution and superior mass accuracy, [34,37,38] unique elemental compositions can be assigned clearly to measured m/z signals. Although the allocation of molecular formulae to metabolites is possible, [39] the structural identification of molecular formulae remains the bottleneck in mass spectrometric metabolomics studies. [36] FTICR-MS analysis was utilized to investigate alterations between Chlamydia pneumoniae-infected and non-infected Hep-2 cells [38] and to identify differences in the growth stages of bacteria. [39] Applying FTICR-MS to foods is called "foodomics" aiming to analyze the functionality, the nutritional value, and the safety of agricultural products. [24] Regarding mycotoxins and related food contaminants, foodomics investigations can support the identification of new or fungal metabolites on foods and can provide further data for proper risk assessments. [24] In the present study, various fungal isolates of A. alternata and A. solani were cultivated in a chemically defined liquid medium and the extracts were analyzed by direct infusion FTICR-MS. Due to the determination of hundreds of metabolites, we aim at obtaining a more holistic insight into the fungal mycobolome. Assigned molecular formulae are intended to be allocated to metabolites using the Antibase and Kyoto Encyclopedia of Genes and Genomes (KEGG) databases. Complementing this non-targeted FTICR-MS approach with targeted LC-MS/MS analysis [25] will help to clearly identify Alternaria mycotoxins, as LC-MS/MS offers advantages in selectivity [40] and the possibility to chromatographically separate isomers such as ATX II and ALTP. Additionally, the LC-MS/MS measurements are intended to provide quantitative results on intra-and extracellular mycotoxin contents.

Selection of Fungal Isolates
A. alternata isolates were analyzed by FTICR-MS to obtain a comprehensive insight into the fungal mycobolome. To cover a "general" A. alternata metabolome irrespective of the origin of the fungi, A. alternata was isolated from different sources such as potato leaves, tomato leaves, and moldy tomatoes. Apart from the three small-spored A. alternata isolates 1, 2, and 3, one A. solani isolate was analyzed to reveal interspecies variations. The fungi were cultivated in a chemically defined liquid medium consisting of salts and glucose, facilitating the detection of metabolites solely biosynthesized by Alternaria. As most of the metabolites produced by the fungi are excreted into the media, [29] the cultivation in liquid medium circumvented the extensive extraction of metabolites after growing the isolates on solid medium. In the literature, metabolic profiling is mainly performed by growing the fungi on solid media due to a higher quantity and a higher number of produced metabolites. [29] In 2018, Zwickel et al. performed a study on the metabolic profiles of different Alternaria species grown on rice. [31] After the cultivation, the mycotoxins were laboriously extracted using a mixture of acetonitrile/water/acetic acid. [41]

Molecular Formula Annotations
The network established during the annotation process is shown in Figure 1. The black nodes in the network represent assigned molecular formulae, which are linked via edges representing biochemical reactions. [42] The smaller sections of the network illustrate the biochemical connections between the molecular formulae of AOH and AME, and between the molecular formulae of ATX I, ATX II/ALTP, and STTX III. The elemental compositions of AOH and AME are linked via the accurate mass difference of CH 2 , whereas the accurate mass difference of H 2 interconnects the molecular formulae of the perylene quinones ATX I, ATX II/ALTP, and STTX III. In addition to ATX I, ATX II/ALTP, and STTX III, molecular formulae of other perylene quinones described in the literature were detected in the data. The molecular formulae of ATX III (C 20 H 12 O 6 ), [43] STTX IV (C 20 H 12 O 7 ), [43] alterlosin I/STTX I (C 20 H 14 O 7 ), [43,44] stemphyperylenol (C 20 H 16 O 6 ), [43] and alterlosin II/7-epi-8-hydroxyaltertoxin I/stemphytriol/6-epi-stemphytriol (C 20 H 16 O 7 ) [43,44] were determined and Figure 2 displays their elemental compositions as well as their chemical structures. Applying LC-Orbitrap, Zwickel et al. determined ATX I, ATX II, STTX III, as well as hydrated and dehydrated forms of these mycotoxins corre-sponding to further perylene quinones. [31] Besides, the latter authors detected additional peaks in the ion chromatograms of ATX I, ATX II, and STTX III, sharing the same elemental compositions as the respective mycotoxins within the mass error range (± 5 ppm) of the applied LC-Orbitrap MS instrument.
Apart from the determination of mycotoxins, a comprehensive analysis of the fungal mycobolome also includes the detection of precursors of metabolites. Previous studies on the biosynthesis of perylene quinones of fungi postulated a dinaphthyl intermediate as a precursor. [45,46] The dinaphthyl intermediate is supposed to originate from two tetralone derivatives biosynthesized from one acetate and six malonyl units. [46,47] [48] and perylene quinone (C 20 H 10 O 4 ) [43] were detected in our data and might represent precursors of the respective perylene quinones. Biosynthetic pathways of fungal secondary metabolites are often not clarified yet and reference compounds for the precursors are often not available. Therefore, the identification of the precursors could not be performed in the present study.
Modified forms of mycotoxins attract increasing attention in fungal investigations and sulfo-conjugations were frequently detected by different research groups. [19,31,49] During FTICR-MS measurements, the molecular formulae of the sulfo-conjugated forms of the mycotoxins AOH, AME, altenuisol, and altenuene [31] were not detected, which is not surprising as the intensities of the m/z signals of AOH, AME, and altenuisol in the mass spectra were already low. www.advancedsciencenews.com www.mnf-journal.com  [43,44,47] The number of possible structural suggestions and the information on the biological context were extended by matching the molecular formulae to entries of two different databases. For the annotation of secondary metabolites, the elemental compositions were checked against the subset Alternaria of the database Antibase. [50] In total, 86 of the 2883 molecular formulae were assigned to metabolites, which equals 3 % annotation rate. 97 % of the molecular formulae could not be assigned to metabolites using Antibase. Applying only the subsets A. alternata and A. solani of Antibase to the data, mainly solanapyrones and altersolanols were annotated for A. solani, whereas bicyclo-(BCA) and tricycloalternarenes (TCA) were assigned for A. alternata. Detailed information on the detected experimental masses, the theoretical neutral masses, the molecular formulae as well as the annotated metabolites using the subsets A. alternata and A. solani are depicted in Table 1.
Besides the secondary metabolites, molecules of the primary metabolism can be detected by FTICR-MS analysis. The assigned molecular formulae were compared to the entries of the KEGG database and subsequently allocated to species-specific pathways. [69] Comparing the assigned molecular formulae to the entries of the KEGG database resulted in 527 annotations (18 %) and, due to molecules with the same molecular formula, in 1820 metabolite annotations. The database assignments of less than 20 % illustrate the extent of the still unknown metabolism of Alternaria fungi and demonstrate the low coverage of the total Table 1. Detected experimental masses, theoretical neutral masses, annotation errors, annotated molecular formulae, and metabolite candidates of the database match in Antibase for the subsets A. alternata and A. solani. [50] Experimental mass  [51] solanapyrone E, [52] prosolanapyrone III [53] 330  [68] diversity of all existing metabolites. [70] Besides, great attention should be paid to multiple annotations to only one molecular formula also hampering the identification of the metabolites. [38] This problem was addressed by Nielsen et al., as the researchers determined multiple entries of Antibase 2008 exhibiting identical molecular formulae. For example, the elemental composition of C 15 H 22 O 3 resulted in 113 metabolite candidates. [71] Another example of multiple assignments was given by Zwickel et al. [31] ATX II and ALTP share the same molecular formula and, therefore, exhibit identical m/z values in the mass spectrum. In the survey of Zwickel et al., various Alternaria isolates were analyzed by high-resolution mass spectrometry and four chromatographically separated peaks in the ion chromatogram of ATX II were assigned to the same molecular formula. Only ATX II and ALTP were identified, whereas the other two peaks could not be allocated to metabolites. [31] The results of Nielsen et al. and Zwickel et al. demonstrate the difficulty in dealing with multiple assignments of molecular formulae to metabolites. [31,71] To support the identification of metabolites, MS/MS spectra, specific UV-vis spectra, and authentic reference compounds are indispensable. [30,71,72] In our survey, the lack of reference compounds allowed solely the hypothetical identification of the database assignments. The unambiguous identification of the metabolites was only performed for AOH, AME, ATX I, ATX II, ALTP, STTX III, TEN, and TA using targeted LC-MS/MS analysis (see Section 3.2). Additionally, the metabolites were allocated to metabolic pathways listed in the KEGG database. [69] Figure 3 displays various pathways of amino acids, carbohydrates, and lipids related to A. alternata. The black bars represent the number of annotated metabolites belonging to one specific pathway. Contrarily, the shaded bars show the number of molecules, which belong to the respective pathway, but the molecular formulae of which were not detected in the data. The percentage of detected molecular formulae with regard to the number of all metabolites belonging to the pathways shown in Figure 3 was mainly below 50%. This Figure 3. The illustration displays the number of detected molecular formulae (black bars) of various metabolic and biosynthetic pathways in the KEGG database. [69] The shaded bars represent metabolites belonging to the respective pathways, but the molecular formulae were not detected in the FTICR-MS data.
may be due to the restrictions of the applied methodology, for example, a too small or large molecule size (beyond the detection limit of the MS analyzer), the presence of other elements than CHNOSP in the elemental composition, the SPE conditions during sample preparation, the ionization mode during MS measurements, or simply due to low concentrations of the molecules.
After the annotation of m/z signals to molecular formulae, the hydrogen to carbon ratios (H/C) and the oxygen to carbon ratios (O/C) were calculated from the elemental compositions. [38] The H/C ratios of the molecular formulae were plotted against the O/C ratios [73] and each Alternaria isolate was displayed separately on such a van Krevelen diagram. In the van Krevelen plot, different metabolite classes have their specific position based on different elemental compositions (Figure 4). [36] The positions of the metabolite classes of fatty acids, amino acids and peptides, carbohydrates, and polyphenols are displayed in Figure 4a. For all Alternaria isolates, 72% of the molecular formulae corresponded to a CHO composition, followed by 17% of a CHNO composition, and 11% of CHOS and CHNOS compositions. Comparing the plots of Figure 4a-d with each other, the profiles of the mycobolome differ. In the A. solani samples, fatty acids and condensed terpenoids are displayed, whereas in the A. alternata samples, the polyphenols were dominant. All investigated Alternaria fungi share the presence of amino acids and peptides.
As the molecular formulae of the Alternaria mycotoxins AOH, AME, ATX I, ATX II, ALTP, STTX III, and TA, of further perylene quinones, possible precursors and other secondary metabolites were detected in the FTICR-MS data, we assumed a representative coverage of the Alternaria mycobolome under the given conditions. Besides, the suitability of the liquid medium for the cultivation of Alternaria isolates was confirmed and the sample preparation protocol as well as the conditions during the measurements allowed the detection of a wide range of fungal secondary metabolites.

LC-MS/MS Detection of Alternaria Mycotoxins
In addition to the non-targeted analysis of fungal extracts, the identification and quantification of the mycotoxins AOH, AME, ATX I, ATX II, ALTP, STTX III, TEN, and TA were complemented by targeted LC-MS/MS analysis. The mycotoxins were compared to reference compounds and were identified based on retention times and mass transitions. [25] A chromatogram of AOH, AME, ATX I, ATX II, ALTP, STTX III, and TEN is displayed in Figure S1a, Supporting Information. An additional chromatographic run had to be performed for TA due to different polarity (Figure S1b, Supporting Information). As adequate amounts of stock solutions were not available for ATX II and STTX III, quantitative values could not be calculated for these mycotoxins. The sample preparation and LC-MS/MS analysis were not fully validated for the fungal cultures in this study, and therefore, no limits of detection and quantification were calculated. However, to precisely identify the mycotoxin signals in the LC-MS/MS run, a minimum peak area unit of 10 5 was stated. The peak areas of the mycotoxins are displayed as mean values of areas of the five biological replicates (Figure 5). If one or two of the five replicates  [36] Liu et al., [74] and Schmitt-Kopplin et al. [75] showed peak areas below the area cut off, a peak area of 1 × 10 5 was used for the calculation of mean values. Outliers were detected by applying Dixon's Q testing.
The A. solani isolate did not produce any of the targeted Alternaria mycotoxins above the peak area cut off, neither on the first day of sampling nor on the other days of cultivation. In the literature, A. solani is reported to produce AOH and AME. [76,77] AOH was consistently detected in the medium of the A. alternata isolate 2, whereas the A. alternata isolate 1 and the A. solani isolate did not produce AOH at all. AME was not detected in the media of any isolate. As the benzopyrones were neither detected in the extracts of A. solani nor often analyzed in the samples of A. alternata, the excretion of AOH and AME into the liquid medium might be low. The three A. alternata isolates produced various mycotoxins exceeding the minimum peak area (Figure 5a1-d1). The most frequently detected mycotoxins were the perylene quinones ATX I, ATX II, ALTP, and STTX III. These mycotoxins were produced by all A. alternata isolates and were detected at each day of sampling. During the cultivation, the peak area of ATX II and STTX III decreased to less than one tenth for all A. alternata isolates when comparing the peak area of the 4th to the 11th day of cultivation. Contrarily, a decrease of the peak area of ATX I and ALTP could not be observed. ATX II and STTX III structurally share an epoxy group, [43] probably sensitive to chemical degradation during the cultivation process. The A. alternata isolates 1, 2, and 3 produced TEN, which was detected from the 7th day of cultivation on. TA was produced by the isolates 2 and 3, while the isolate 1 did not produce TA at all. Due to the different sensitivity of the analytes in the mass spectrometer, the peak areas of the mycotoxins were transferred into contents via one-point calibration (Figure 5a2-d2). Quantitative results could not be calculated for ATX II and STTX III due to the lack of adequate amounts of stock solutions. The determined contents of AOH were below 5 µg kg −1 for the isolates 2 and 3. ATX I and ALTP were detected in all A. alternata isolates in contents ranging from 28 to 233 µg kg −1 and from 40 to 182 µg kg −1 . TA was produced by the isolates 2 and 3 and the calculated amounts varied from 5400 to 20 000 µg kg −1 . The highest content of TEN of 37 µg kg −1 was produced by the isolate 1. However, it has to be mentioned that the calculated contents were obtained using only one-point calibration and were not quantified by matrix matched calibration or by using isotopically labeled internal standards. Therefore, the given contents should only be classified as semi-quantitative.
The biosynthetic capabilities of diverse Alternaria species has been reported by Andersen et al. [30] and Zwickel et al., [31] who both have been growing their isolates on solid media. Anderson et al. investigated 87 Alternaria isolates and allocated 22 isolates to the A. arborescence, A. infectoria, A. tenuissima, and A. alternata species groups. [30] Similarly, Zwickel et al. [31] performed studies on 93 isolates of A. alternata, A. arborescence, A. tenuissima, and A. infectoria. In the former study, the A. alternata, A. arborescence, and A. tenuissima isolates frequently produced AOH, AME, Figure 5. The peak areas of AOH, AME, ATX I, ATX II, ALTP, STTX III, TEN, and TA are displayed for the 4th day (a1), 7 th day (b1), 9 th day (c1), and 11 th day (d1) of cultivation. The peak areas are shown as mean values and standard deviations of the five replicates. The A. solani isolate is not displayed, as this isolate did not produce any mycotoxin above the peak area limit. a2-d2) Contents of the mycotoxins AOH, AME, ATX I, ALTP, TEN, and TA in the liquid medium of the 4th, 7th, 9th, and 11th day of cultivation calculated via one-point calibration. Note the logarithmic axis of the peak areas and of the mycotoxin contents. and altenuene, whereas TEN and TA were biosynthesized less frequently. In the study of Zwickel et al., 21 isolates belonged to the A. alternata species group and 90% of the isolates produced ATX I, 81% produced STTX III, and 76% ATX II and ALTP. AOH and AME as a group as well as TA were biosynthesized by 81% and 76% of the A. alternata isolates, respectively. Interestingly, five out of the 93 Alternaria isolates did not biosynthesize any of the analyzed mycotoxins. These results are partly different to ours, as the benzopyrones were not detected in the liquid medium of our A. alternata isolate 1 and were rarely determined in the samples of the A. alternata isolate 3. One explanation for this discrepancy could be given by the study of Söderhäll et al., [78] who investigated the mycotoxin production of A. alternata under the exposure of white light. Depending on the growth phase of the fungus, the biosynthesis of the benzopyrones was almost completely inhibited after the light exposure, [78] which may have been also the case for our cultures. The perylene quinones ATX I, ATX II, ALTP, and STTX III were produced by all A. alternata isolates, which is in good agreement with the results of Zwickel et al. [31] Alike in the study of Zwickel et al., [31] TEN was produced by all A. alternata isolates but was detected above the peak area limit only from the second day of sampling on. The results of Zwickel et al. that some of the A. alternata isolates did not produce any Alternaria mycotoxins clearly demonstrate the differences in the metabolic capabilities of isolates belonging to the same species group. [30,31] Due to the differences in the TA production of the three A. alternata isolates in our survey, large numbers of reliable isolates of the same taxon should be used for differentiation investigations based on mycotoxin profiles. [79] The mycotoxins AOH, AME, ATX I, ATX II, ALTP, STTX III, TEN, and TA were also determined in the extracts of fungal cells from the mycelium (Figure 6a). Contrarily to the liquid medium, AOH was detected in the cells of all investigated Alternaria isolates and AME was determined in the cells of A. solani and of the A. alternata isolates 2 and 3. This confirmed the ability of the A. solani isolate to produce these two mycotoxins. The perylene quinones were verified in the cells of all A. alternata isolates. Interestingly, TEN was only determined in the cells of the isolate 2, whereas TA was not detected in any of the Alternaria cells. Figure 6b displays the contents of AOH, AME, ATX I, ALTP, and TEN calculated via one-point calibration. The highest contents were determined for ATX I and ALTP, ranging from 730 to 1900 µg kg −1 and from 400 to 1300 µg kg −1 , respectively. AOH was detected in contents from 11 to 120 µg kg −1 and AME from 0.5 to 2.7 µg kg −1 . Again, the contents have to be considered as semi-quantitative. Although the benzopyrones were detected frequently in the extracts of the disrupted cells, AOH and AME were only rarely analyzed in the liquid medium. As the benzopyrones were also detected in almost all A. alternata cultures in the study of Zwickel et al., [31] this suggests that these compounds are generally formed, but only excreted under certain conditions to the medium. Contrarily, the perylene quinones ATX I, ATX II, ALTP, and STTX III were determined in the extracts of the cells and were excreted into the liquid medium. A different tendency was observed for TA as this mycotoxin was fully excreted into the liquid medium and was not detectable inside the fungal cells. The transport mechanisms to export these metabolites are still largely unknown. It can be hypothesized that extracellular vesicles reported to carry virulence factors [80] may be involved.

Differentiation between Alternaria Species
Apart from the comprehensive description of the Alternaria mycobolome, similarities and differences between samples can be detected by FTICR-MS and subsequent Principal Component Analysis (PCA). Before performing the PCA, the matrix was filtered by keeping only m/z signals that occurred in at least four out of five biological replicates ensuring the biological importance of the remaining signals. The PCA was performed on the whole data set and each day of sampling is displayed as an individual PCA plot (Figure 7). In the scores plot of the 4th day of cultivation (Figure 7a), the replicates of A. solani and the controls revealed distinct clusters displaying differences in the second component. For the A. alternata isolate 1, four of the five replicates clustered together while the fifth replicate was determined in the cluster of the A. solani isolate. The replicates of the A. alternata isolates 2 and 3 showed high variation and did not form distinct clusters. In the PCA model, 26.9% of the total variance are explained in the PC1 and 8.6% are explained in the PC2. On the 7th day of cultivation (Figure 7b), clear clusters are formed by the replicates of the isolate 1, the A. solani isolate, and the controls. The clusters of the isolates 2 and 3 are partly overlapping and, therefore, the two isolates cannot be separated clearly. A similar cluster formation to the 7th day of cultivation is obtained on the 9th day of cultivation (Figure 7c). However, the replicates of the A. solani isolate and the controls are overlapping. At the 11th day of cultivation (Figure 7d), the A. solani replicates and the controls form distinct clusters, which are clearly separated in the second component. The isolate 1 forms a distinct cluster and the replicates of the isolates 2 and 3 are overlapping. Again, the highest variation was determined between the five replicates of the isolate 3. Due to the formation of clusters in the PCA, differences in the mycobolome of the different Alternaria isolates are obvious. The distinction of the A. alternata and A. solani isolates in the PCA was based on more than 3000 m/z values, whereas the differentiation by LC-MS/MS was performed using eight mycotoxins. The FTICR-MS measurements confirmed the LC-MS/MS results according to which the A. solani isolate differs from the A. alternata isolates in the mycotoxin production. Besides, the A. alternata isolate 1 was proven to vary from the other two A. alternata isolates.
In the literature, different approaches were applied to differentiate between various Alternaria species. Zwickel et al. compared the metabolic capabilities of various Alternaria isolates belonging to A. alternata, A. arborescence, A. tenuissima, and A. infectoria. [31] Based on the metabolic profile and overall low mycotoxin production, the A. infectoria isolates were segregated from A. alternata, A. arborescence, and A. tenuissima. Contrarily, the analysis of various mycotoxins by HPLC-MS/MS could not separate the isolates of A. alternata, A. arborescence, and A. tenuissima. [31] Andersen et al. obtained identical results when analyzing the mycotoxin profiles of A. infectoria, A. alternata, A. arborescence, and A. tenuissima. [30] Besides the analysis of mycotoxin profiles, Andersen et al. performed a metabolic differentiation of A. alternata, A. gaisen, A. limoniasperae, A. longipes, A. tangelonis, and A. turkisafria based on direct infusion MS. In the mass spectra, 100 to 400 ions were detected, respectively, and the clustering of the isolates resulted in a separation of four of the six Alternaria species. A. gaisen, A. turkisafria, A. tangelonis, and A. alternata clustered in four separated clusters, whereas one isolate of A. limoniasperae and A. longipes clustered apart from the other isolates of the related species-group, respectively. [32] In our survey, we analyzed three different A. alternata isolates by FTICR-MS and the A. alternata isolate 1 clustered apart from the other two isolates in the PCA plots. Accordingly, Andersen et al. did not obtain distinct clusters for the A. limoniasperae and A. longipes isolates as one isolate clustered apart from the other isolates belonging to the same species-group. [32]

Determination of Discriminating Metabolites
To identify metabolites, which are responsible for the grouping in the PCA plots, volcano plots were created. Only m/z values that were assigned to molecular formulae were considered to be relevant metabolite candidates. [81] The volcano plots were created by plotting the -log 10 p-value against the log 2 fold change of the MS signal intensities of the A. solani and the A. alternata replicates. [82] A volcano plot was created for the A. solani and the A. alternata isolate 1 at the 11th day of cultivation (Figure 8a). The horizontal line in the plot represents the significance value of 0.01. The higher the y value of one molecular formula is, the more significant is the difference. Interesting molecular formulae are located on the upper left part and upper right part of the plot. [82] To allocate the discriminating molecular formulae to metabolites, all discriminating elemental compositions were checked against Antibase. [50] The assignments are displayed as blue triangles for discriminating molecular formulae of A. solani and as green hashes for A. alternata. If one allocated molecular formula was discriminating in the A. solani-A. alternata isolate 1 comparison, as well as in the A. solani-A. alternata isolate 2

Figure 8. The volcano plot (a) displays the discriminant masses between
A. solani and the A. alternata isolate 1 at the 11th day of cultivation. The volcano plot (b) depicts the discriminant masses between the A. alternata isolate 1 and 2 at the 11th day of cultivation. The triangles and hashes represent molecular formulae that were assigned to metabolites by the Antibase database. [50] Due to the lack of reference compounds, the identification of these metabolites could only be performed for STTX III, APML, and TA (displayed as squares) using targeted LC-MS/MS analysis. Details on the extraction and purification, on the structure elucidation, on the mass spectrometric fragmentation pattern, and UV-vis absorption spectrum of APML are displayed in Supporting Information. and A. solani-A. alternata isolate 3 comparison ( Figure S2, Supporting Information) its m/z value, the theoretical neutral mass, the error of annotation, the molecular formula as well as the metabolite assignment were listed in Table 2. For A. solani, four molecular formulae were significantly different in all comparisons and were allocated to the metabolites altechromone A, [83] quadrilineatinmethylether, [84] 2,4-dihydroxy-6-acetonylbenzoic acid, [85] and (8R,9S)-9,10-epoxy-8-hydroxy-9-methyldeca-(2E,4Z,6E)-trienoic acid [86] (Table 2). For A. alternata, seven molecular formulae were assigned to metabolites (Table 2), for example, ATX III [12] /STTX III, [87] xanalteric acid I/II, [88] alterlosin I, and alterlosin II. [44] One of the orange marked squares in the volcano plot a (Figure 8) represents the molecular formula of the mycotoxin STTX III, identified by targeted LC-MS/MS analysis. [89] In the LC-MS/MS studies, the perylene quinones ATX I, ATX II, ALTP, and STTX III were only produced by the A. alternata isolates. In the literature, the production of the perylene quinones by A. alternata and, additionally, by A. arborescence and A. tenuissima is reported. Contrarily, the biosynthesis of these mycotoxins by A. solani is not mentioned. [31] Interestingly, the molecular formulae of ATX I (C 20 H 16 O 6 ) and ATX II/ALTP (C 20 H 14 O 6 ) were also detected in the samples of A. solani in our FTICR-MS measurements but were not identified as ATX I, ATX II, and ALTP by LC-MS/MS analysis using reference compounds. As these molecular formulae were also detected in the samples of A. solani, these elemental compositions are not identified as discriminant masses by the volcano plots. Obviously, our A. solani isolate produces different compounds with the same molecular formulae as the perylene quinones, which points to the need of using these complementary methods for differentiating the metabolomes. The second orange marked square in Figure 8a represents the molecular formula (C 23 H 20 O 9 S) of a discriminating metabolite named alterperylenepoxide A-9-mercaptolactate (APML). In this study, APML was extracted from overgrown rice, purified by various stationary and mobile phases, and characterized by 1 H and 13 C NMR analysis as well as 1 H-1 H correlated spectroscopy (COSY), Heteronuclear Single Quantum Coherence (HSQC), and Heteronuclear Multiple Bond Correlation (HMBC) 2D NMR analysis (Tables S2-S4 and Figures S3-S11, Supporting Information).
Due to the lack of reference compounds, the other assigned elemental compositions were only tentatively allocated to metabolites. The remaining molecular formulae that were also significantly different among A. solani and A. alternata could not be assigned by Antibase. [50] This reveals the still unknown metabolites produced by Alternaria fungi and the necessity for additional studies on the comprehensive analysis of the Alternaria mycobolome.
Apart from the differences in the mycobolome of A. solani and A. alternata, variations between different A. alternata isolates can be determined. The volcano plot b in Figure 8 displays the differences in the mycobolome of the A. alternata isolates 1 and 2. The discriminating elemental compositions were checked against the entries of Antibase [50] and five (three) molecular formulae were assigned to metabolites of the isolate 1 (isolate 2) ( Table 3). The orange marked square (Figure 8b) represents the molecular formula of the mycotoxin TA, which was already proven to be discriminant by targeted LC-MS/MS analysis. Although the two isolates belong to the same species group and were cultivated under the same conditions, some metabolites differ in averaged signal intensities or are only produced by one of the two isolates. To confirm and improve the results, more isolates from the species groups of A. alternata and A. solani need to be analyzed by FTICR-MS and also fungal isolates from different origins should be included. As it is not possible to fully characterize the fungal mycobolome using only one type of instrument, [72] different approaches and various analytical techniques are required to receive a more holistic picture of the mycobolome of Alternaria fungi. [93] A more holistic knowledge on the secondary metabolism of the food contaminating fungus Alternaria enables proper risk evaluations on food and feed. After isolating and characterizing known Alternaria mycotoxins as well as new secondary metabolites, their toxicological capabilities should be investigated. Besides, accurate quantitative analytical methods are needed to obtain more data on the occurrence and contents of Alternaria metabolites in agricultural commodities. A sound knowledge on the toxicology as well as on the occurrence of the fungal food contaminants is essential to allow the EFSA to perform proper risk evaluations on food and feed products. www.advancedsciencenews.com www.mnf-journal.com Table 2. Discriminant molecular formulae between A. alternata and A. solani assigned to metabolites by Antibase. [50] Species Experimental mass Theoretical neutral mass

Concluding Remarks
The mycobolome of various Alternaria isolates was analyzed by FTICR-MS and complemented with LC-MS/MS analyses. From the total number of detected m/z signals of the FTICR-MS, 35% could be assigned to unequivocal molecular formulae of potential metabolites. Of these formulae, only 3% could be verified as specific fungal metabolites using the Antibase database, which particularly focuses on fungi. Performing an additional database search against the KEGG database resulted in only 18% of assignments of the 2883 molecular formulae to general cellular metabolites. This result indicates that only about one fifth of the metabolome signals are potentially known. However, this estimation of the current database knowledge becomes even worse, when we consider that only a small number of the metabolites, whose molecular formulae were effectively detected by FTICR-MS, could be assigned by targeted LC-MS/MS using authentic reference compounds as revealed by the missing perylene www.advancedsciencenews.com www.mnf-journal.com quinones in the A. solani extracts in the LC-MS/MS measurements. On the other hand, complementing the FTICR-MS data with targeted LC-MS/MS analyses is necessary, as the molecular formulae of the direct infusion FTICR-MS measurements generally can be assigned to several metabolites, due to lack of retention time information in direct infusion FTICR-MS. Another striking result from the pathway analyses was the missing detection of over 50 % of expected metabolites in both LC-MS/MS and FTICR-MS measurements, which also indicates that our coverage of the metabolome is still very low. The low percentage of database allocations demonstrates the need for further comprehensive investigations of the Alternaria mycobolome. The clear identification of interesting metabolites remains the bottleneck of metabolomics and authentic reference compounds of fungal metabolites are necessarily needed for identification purposes.
The non-targeted analysis of the fungal mycobolome by FTICR-MS unraveled variations in the metabolome of A. solani and A. alternata and enabled the detection of discriminating metabolites. One of these discriminating metabolites was identified as alterperylenepoxide A-9-mercaptolactate by 1 H, 13 C, and 2D NMR analysis after isolating the metabolite. In addition to the interspecies variations, differences within the A. alternata species were determined. Further isolates from the species group of A. alternata and A. solani need to be analyzed by FTICR-MS to confirm the intra-and inter-variation of the mycobolome. Additionally, isolates from different small spored Alternaria species such as A. tenuissima and A. arborescence should be included in future investigations. Combined to additional analytical tools, FTICR-MS is a promising tool for the chemotaxonomic differentiation of fungal isolates. In future mycobolome investigations, FTICR-MS should also be applied to Alternaria fungi-plant interactions or in the field of food contaminants related to Alternaria mycotoxins.
Preparation of Stock Solutions: Stock solutions of Alternaria mycotoxins were prepared in acetonitrile (AOH, AME, TEN) or methanol (ATX I, ALTP, TA) in concentrations ranging from 10 to 100 µg mL −1 . For quantitative measurements, the stock solutions were further diluted. All solu-tions were stored in the dark at −20°C. The absorptions of the solutions were measured by a Genesys, 10S, UV-vis spectrophotometer (Thermo Fisher Scientific, Madison, Wisconsin, USA) and the concentrations were confirmed by applying the published extinction coefficients. [107] However, ATX II and STTX III were only qualitatively included in the method, as the available amounts of these reference compounds were not detectable by UV-vis.
Preparation of Synthetic Nutrient-Poor Agar: Glucose (0.2 g), magnesium sulfate heptahydrate (0.5 g), potassium chloride (0.5 g), potassium dihydrogen phosphate (1 g), potassium nitrate (1 g), and sucrose (0.2 g) were dissolved in 100 mL of water. To adjust the pH to 5.5, 600 µL of sodium hydroxide (1 mol L −1 ) were added. 22 g of agar were solved in 900 mL of water and, after the unification of both solutions, the medium was autoclaved for 20 min at 121°C. [108] Isolation of Alternaria Fungi: A. alternata was isolated from potato leaves (Uelzen, Germany, isolate 1), tomato leaves (Aitrang, Germany, isolate 2), and tomatoes (Aitrang, Germany, isolate 3). The A. solani isolate originated from potato leaves (Kirchheim, Germany). After harvesting, the plant leaves were dried. The surface of the dried leaves and fresh tomato was sterilized using 3% of sodium hypochlorite. Subsequently, small pieces of plant tissue showing typical symptoms of infection were placed on SN agar and cultivated at 22°C and 65% relative humidity for 1 week. Alternately, the isolates were exposed to black light for 12 h and subsequently cultivated in the dark for 12 h. Single spores were isolated from the overgrown agar plates and cultivated on synthetic nutrient-poor agar at 22°C and 65% relative humidity for 2 weeks. Again, 12 h of black light exposure was followed by the cultivation in the dark. The overgrown agar plates with pure isolates were used for further experiments. All fungi were obtained in 2015.
The sterile liquid medium was inoculated with defined spore suspensions. For the preparation of the spore suspensions, 3 mL of detergent solution (0.5% Tween 20) was pipetted on the overgrown agar plates and the mycelium and spores were scratched. The spores were counted using a Thoma chamber and the spore suspensions were diluted to 8.75 × 10 5 spores per milliliter for A. alternata and to 2 × 10 5 spores per milliliter for A. solani. [108] 25 µL of the spore suspensions of the A. alternata isolates and 100 µL of the spore suspension of the A. solani isolate were added to the liquid medium to receive equal amounts of total spores. During FTICR-MS measurements, contaminations originating from chemicals, solvents, plastic, and glass surfaces were detected, and these contaminations were compensated for by the analysis of control samples. The control samples were obtained by adding 25 µL of pure detergent solution to the liquid medium. All samples were prepared in replicates of five. The Erlenmeyer flasks were sealed with septa allowing sterile sampling during the cultivation process after 4, 7, 9, and 11 days using cannulas and syringes. The fungi were cultivated in the dark (26°C, 110 rpm) [109] and the isolates were exposed to artificial daylight for half an hour a day.
The liquid medium was analyzed by FTICR-MS after 4, 7, 9, and 11 days of cultivation. 3 mL of the medium was sterilely taken from the Erlenmeyer flasks and centrifuged (15 000 × g, 10 min) to separate the mycelium. The pH of the supernatant was adjusted to pH 2 using formic acid. To protect the ESI source of the instrument from contamination, salts of the liquid medium were removed by solid phase extraction (Discovery DSC-8, Supelco, Bellefonte, PA, USA). The C8 material was washed with 1 mL of methanol and conditioned with 1 mL of water (pH 2). After the sample loading (2 mL), the column was washed with 5 mL of water (adjusted to pH 2 using formic acid). The elution of the analytes was performed with 1 mL of methanol.
Besides, the mycelia of the fungi after eleven days of cultivation were also analyzed by FTICR-MS and LC-MS/MS. After separating the liquid medium from the mycelium by centrifugation, the latter was thoroughly washed with water to remove remaining liquid medium. 200 mg of the mycelium were weighed into a nucleo spin bead tube (type A, 0.6-0.8 mm, Machery Nagel, Düren, Germany), previously cleaned with 5 mL of water and 5 mL of methanol. After adding 1 mL of ice-cold methanol/water (90/10, v/v), the disruption of the cells was performed at 6800 rpm (4 × 30 s) utilizing a Precellys homogenizer (Bertin instruments, Montigny-le-Bretonneux, France). The homogenizer operated at −10°C using liquid nitrogen. After the disruption, the cell suspension was centrifuged at 21 000 × g for 10 min, the supernatant was dried under nitrogen and the residue was resolved in 2 mL of water (adjusted to pH 2 using formic acid). To ensure comparability of medium and mycelium samples, the aqueous solutions were desalted by solid phase extraction (Discovery DSC-8, Supelco, Bellefonte, PA, USA). The C8 material was washed with 1 mL of methanol and conditioned with 1 mL of water (adjusted to pH 2 using formic acid). After loading the sample onto the column, the latter was washed with 5 mL of water (adjusted to pH 2 using formic acid). The elution of the analytes was performed with 1 mL of methanol.
FTICR-MS Analysis: The acquisition of ultra-high-resolution mass spectra was performed on a Bruker Solarix Fourier transform ion cyclotron resonance mass spectrometer (FTICR-MS) (Bruker Daltonics GmbH, Bremen, Germany) coupled to a 12 Tesla superconducting magnet (Magnex Scientific Inc., Yarnton, GB). The direct infusion of samples was performed with an APOLO II ESI source (Bruker Daltonics GmbH, Bremen, Germany) which operated in the negative ionization mode. The samples were diluted with methanol (1/10; v/v) prior to injection and were introduced into the ESI source at a syringe flow rate of 120 µL h −1 by a Gilson autosampler (Gilson, Inc., Meddleton, WI, USA). The spectra were externally calibrated by using ion clusters of arginine (10 mg L −1 in methanol) and were acquired with a time domain transient of four mega words in size. Measured masses ranged from m/z 150 to m/z 1000. For each sample, 300 scans were accumulated. This was equivalent to 13 min of analysis time. The capillary voltage was set to 3600 V and the spray shield voltage was −500 V. The drying gas flow rate and the drying gas temperature were adjusted to 4.0 L min −1 and 200°C. The ion accumulation time was 0.3 s. A resolving power of 600000 at m/z 300 was achieved. Subsequently, internal calibration was carried out on each mass spectrum by using a calibration list of Alternaria metabolites described in the literature. The calibration list covered an m/z range of 160 to 730 Da. The Data Analysis Version 4.2 (Bruker Daltonics GmbH, Bremen, Germany) was used to process raw spectra. The m/z values with a signal to noise ratio of 7 and a relative intensity threshold of 0.01% were exported as mass lists. The mass lists were de-noised from the well-known Gibbs sidelobes (wiggles) by the use of a special program of denoising [110] and the clean mass lists were subsequently aligned using an in-house written program (peak alignment window width: ± 1 ppm). [39,111] In total 120 spectra were measured.

Molecular Formula Annotation and Database Assignments of FTICR-MS Data:
The m/z signals that occurred in at least two out of five biological replicates were assigned unequivocally to molecular formulae by an inhouse written software tool named NetCalc. The annotation of the molecular formulae is based on a mass difference network consisting of nodes and edges. The nodes represent m/z values (metabolite candidates) and edges constitute biochemical reactions. [42] The biochemical reactions can be expressed as mass differences between substrates and products and are predefined in a mass difference list covering 191 reaction-equivalent mass differences such as oxidation, reduction, hydroxylation, methylation, and the loss of CO 2 . [34,42] As starting points of the network, 41 reference masses (Alternaria metabolites) with exact deprotonated mass and molecular formula were specified. Originating from the references, measured m/z values were assigned to molecular formulae by comparing the mass differences of all signals of a mass spectrum to the mass difference list. [34,42] Signals of isotopes and masses with an unusual mass defect were not included in the network. [42,81] The assignments of the elemental compositions contained only C, H, N, O, S, and P. The molecular formula allocation was performed on 8139 m/z values resulting in 4467 monoisotopic elemental compositions (55%) with 3285 molecular formulae ex-hibiting an annotation error within ± 0.2 ppm (74%). [81,112] During FTICR-MS measurements, adduct ions such as chloride adducts are formed for some metabolites. 402 adducts were removed and 2883 annotated elemental compositions were used for further investigations. The calculated mass difference network was constituted using Gephi 0.9.2 software. [113] To assign the experimental m/z values to metabolite candidates, the annotated molecular formulae were matched to the entries of Antibase. [50] For Alternaria and other genera of fungi, 95-98 % of the metabolites described in the literature are included in the database. [71] Additionally, the molecular formulae were matched to the entries of the KEGG database (http://www.genome.jp/kegg/compound/). [69] The annotated elemental compositions were used to calculate the H/C and O/C ratios. Subsequently, the ratios were displayed in a 2D van Krevelen diagram to visualize the variations within the metabolic profile of the fungal isolates. [36] Statistical evaluations were only performed on m/z values that occurred in at least four out of five biological replicates. PCA, an unsupervised statistical method, was applied for reducing the complexity of the data. After zscore normalization, the variation in the data was displayed as a set of new independent variables called the principal components. It was used for providing an overview of the complex multivariate data and for detecting outliers and relations between samples. [114,115] The PCA was performed using Simca-P 9.0 software (Umetrics, Sweden).
To determine the most discriminative molecular formulae between the A. alternata and the A. solani isolates, volcano plots were created. Volcano plots display the log 2 fold change (ratio of averaged intensities of measured m/z signals) on the x-axis. The y-axis shows the -log 10 p-value of measured intensities to significantly determine discriminating molecular formulae between the A. alternata and the A. solani isolates. [82] LC-MS/MS Analysis of Alternaria Mycotoxins: AOH, AME, TEN, ATX I, ATX II, STTX III, and ALTP were chromatographically separated on a Shimadzu Nexera X2 UHPLC system (Shimadzu, Kyoto, Japan). As stationary phase, a HyperClone BDS-C18 column (150 • 3.2 mm, 3 µm, 130 Å, Phenomenex, Aschaffenburg, Germany) was utilized. Further details on the instrument conditions, the solvent mixtures, and the gradient were published recently. [25] TA had to be analyzed in an additional LC-MS/MS run, due to the more polar character of the molecule. A Gemini-NX C18 column (150 × 4.6 mm, 3 µm, 110 Å, Phenomenex, Aschaffenburg, Germany) served as stationary phase. The solvent mixtures, the gradient, and further details on the chromatographic separation of TA are listed in the literature. [25] The LC was connected to a triple quadrupole mass spectrometer (LCMS-8050, Shimadzu Corporation, Kyoto, Japan). All analytes were detected in the negative ESI mode. Details on the mass spectrometric conditions were published recently [25] and are listed in Table S1, Supporting Information. The LabSolutions software (Shimadzu, Kyoto, Japan) was used for data acquisition and data analysis.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.