Pioneering Metabolomic Studies on Diaporthe eres Species Complex from Fruit Trees in the South-Eastern Poland

Fungi from the genus Diaporthe have been reported as plant pathogens, endophytes, and saprophytes on a wide range of host plants worldwide. Their precise identification is problematic since many Diaporthe species can colonize a single host plant, whereas the same Diaporthe species can inhabit many hosts. Recently, Diaporthe has been proven to be a rich source of bioactive secondary metabolites. In our initial study, 40 Diaporthe isolates were analyzed for their metabolite production. A total of 153 compounds were identified based on their spectroscopic properties—Ultraviolet-visible and mass spectrometry. From these, 43 fungal metabolites were recognized as potential chemotaxonomic markers, mostly belonging to the drimane sesquiterpenoid-phthalide hybrid class. This group included mainly phytotoxic compounds such as cyclopaldic acid, altiloxin A, B, and their derivatives. To the best of our knowledge, this is the first report on the metabolomic studies on Diaporthe eres species complex from fruit trees in the South-Eastern Poland. The results from our study may provide the basis for the future research on the isolation of identified metabolites and on their bioactive potential for agricultural applications as biopesticides or biofertilizers.

Molecules 2023, 28, 1175 4 of 31 Table 2. Annotation of specific metabolites in the studied Diaporthe isolates using UHPLC-qTOF-MS/MS in the negative (NI) and positive ionization (PI) modes.          The UHPLC-HRESIMS analysis of extracts from Diaporthe isolates led to the annotation of several compounds from the polyketide group-pyranones such as, dihydrohydroxyphomopsolide B isomers I-III (17,23,32), dihydrophomopsolide A (49), dihydrohydroxyphomopsolide A (26), and furanones such as phomopsolidone B (37) and dihydrohydroxyphomopsolidone B isomers I and II (15, 16) (Figure 1). These compounds were tentatively identified based on the high-resolution mass of the precursor ions and the fragments generated via common fragmentation pathways in positive ionization mode. Namely, the loss of one or two water molecules (−18 Da or 36 Da), followed by the loss of a tiglic acid (2-methylbut-2-enoic acid) residue (-C 5 H 8 O 2 ), giving intense fragment ions with m/z 179 or 197 for compounds 17, 23, and 32 ion at m/z 177 for compounds 26 and 49, and m/z 181 for compounds 15, 16, and 37, was observed ( Table 2). Phomopsolides are common secondary metabolites derived from Diaporthe [25]. They were initially isolated from Phomopsis oblonga, a fungus that provided some protection against elm bark beetle infestations [37]. They have been proved for their antibacterial activity against Staphylococcus aureus [38]. Moreover, phomopsolide A/C (62), from the endophytic Diaporthe sp. AC1 from Artemisia argyi, was proved to inhibit the growth of Fusarium graminearum, F. moniliforme, Botrytis cinerea, and Verticillium dahliae, indicating that the compound may have a broad spectrum of antifungal activity [29]. The UHPLC-HRESIMS analysis of extracts from Diaporthe isolates led to the annotation of several compounds from the polyketide group-pyranones such as, dihydrohydroxyphomopsolide B isomers I-III (17,23,32), dihydrophomopsolide A (49), dihydrohydroxyphomopsolide A (26), and furanones such as phomopsolidone B (37) and dihydrohydroxyphomopsolidone B isomers I and II (15, 16) (Figure 1). These compounds were tentatively identified based on the high-resolution mass of the precursor ions and the fragments generated via common fragmentation pathways in positive ionization mode. Namely, the loss of one or two water molecules (−18 Da or 36 Da), followed by the loss of a tiglic acid (2-methylbut-2-enoic acid) residue (-C5H8O2), giving intense fragment ions with m/z 179 or 197 for compounds 17, 23, and 32 ion at m/z 177 for compounds 26 and 49, and m/z 181 for compounds 15, 16, and 37, was observed ( Table 2). Phomopsolides are common secondary metabolites derived from Diaporthe [25]. They were initially isolated from Phomopsis oblonga, a fungus that provided some protection against elm bark beetle infestations [37]. They have been proved for their antibacterial activity against Staphylococcus aureus [38]. Moreover, phomopsolide A/C (62), from the endophytic Diaporthe sp. AC1 from Artemisia argyi, was proved to inhibit the growth of Fusarium graminearum, F. moniliforme, Botrytis cinerea, and Verticillium dahliae, indicating that the compound may have a broad spectrum of antifungal activity [29].

Pyrones
Pyrones represent a class of oxygen-based heterocyclic compounds that naturally occur in two isomeric forms as either 2-pyrone (α-pyrone) or 4-pyrone (γ-pyrone). The number 2/4 is assigned based on the position of the carbonyl group relative to the oxygen atom within the ring system [39]. In our study, Diaporthe spp. isolated from fruit trees produced phomopsinone A (31) and pyrenocine P (3) (Figure 2), which belong to the α-pyrones. Their fragmentation spectra showed mainly water (−18 Da) and/or CO losses (−28 Da). However, characteristic UV maxima at around 280 nm indicated α-pyrone structures (Table 2). Previously, phomopsinone A and pyrenocine J-M have been isolated from the endophytic fungus Phomopsis sp. and have shown antifungal, antibacterial, and antialgal activity [27,28]. Phomopsinone A showed very strong antifungal activity against Botrytis cinerea, Pyricularia oryzae, and Septoria tritici. Pyrenocine J-M had strong antibacterial activity especially against the gram-negative bacterium E. coli, since gram-negative bacteria are usually difficult to inhibit. Similarly, all mentioned compounds showed algicidal activity against Chlorella fusca [27,28]. Pyrones represent a class of oxygen-based heterocyclic compounds that naturally occur in two isomeric forms as either 2-pyrone (α-pyrone) or 4-pyrone (γ-pyrone). The number 2/4 is assigned based on the position of the carbonyl group relative to the oxygen atom within the ring system [39]. In our study, Diaporthe spp. isolated from fruit trees produced phomopsinone A (31) and pyrenocine P (3) (Figure 2), which belong to the αpyrones. Their fragmentation spectra showed mainly water (−18 Da) and/or CO losses (−28 Da). However, characteristic UV maxima at around 280 nm indicated α-pyrone structures (Table 2). Previously, phomopsinone A and pyrenocine J-M have been isolated from the endophytic fungus Phomopsis sp. and have shown antifungal, antibacterial, and antialgal activity [27,28]. Phomopsinone A showed very strong antifungal activity against Botrytis cinerea, Pyricularia oryzae, and Septoria tritici. Pyrenocine J-M had strong antibacterial activity especially against the gram-negative bacterium E. coli, since gramnegative bacteria are usually difficult to inhibit. Similarly, all mentioned compounds showed algicidal activity against Chlorella fusca [27,28]. The studied Diaporthe isolates, apart from the metabolite characteristics for the genus Diaporthe, also produced several new bioactive compounds usually present in the other species of fungi but not in Diaporthe [39]. For example, islandic acid-II (1), originally isolated from Penicillium islandicum, in the literature was reported as showing the complete growth inhibition of Yoshida sarcoma tumor cells [40]. Another compound produced by the tested Diaporthe isolates was scirpyrone K (6) (Figure 2). Its fragmentation pathway was very similar to that of compound 3. Previously, it had been isolated from a marine fungus identified as Phialocephala sp. strain FL30r. This compound exhibited weak radical scavenging activity with no cytotoxic activities reported [41].

Oxylipins
Oxylipins constitute a large family of oxidized fatty acids and their derivatives. Bioactive lipid production is widespread among many organisms including filamentous fungi [42]. In many cases, oxylipins have a role in both organismal development and communication with the host on a cellular basis [43,44]. The literature showed that fungal oxylipins are involved in influencing processes in infected host tissues, presumably by mimicking endogenous signal molecules [45,46]. Fungi have the ability to use the host The studied Diaporthe isolates, apart from the metabolite characteristics for the genus Diaporthe, also produced several new bioactive compounds usually present in the other species of fungi but not in Diaporthe [39]. For example, islandic acid-II (1), originally isolated from Penicillium islandicum, in the literature was reported as showing the complete growth inhibition of Yoshida sarcoma tumor cells [40]. Another compound produced by the tested Diaporthe isolates was scirpyrone K (6) (Figure 2). Its fragmentation pathway was very similar to that of compound 3. Previously, it had been isolated from a marine fungus identified as Phialocephala sp. strain FL30r. This compound exhibited weak radical scavenging activity with no cytotoxic activities reported [41].

Oxylipins
Oxylipins constitute a large family of oxidized fatty acids and their derivatives. Bioactive lipid production is widespread among many organisms including filamentous fungi [42]. In many cases, oxylipins have a role in both organismal development and communication with the host on a cellular basis [43,44]. The literature showed that fungal oxylipins are involved in influencing processes in infected host tissues, presumably by mimicking endogenous signal molecules [45,46]. Fungi have the ability to use the host plant's oxylipin to achieve their own benefits. For example, by increasing the production of toxins, they improve their virulence [45,46], and by increasing sporulation they can accelerate reproduction in the tissues of the host plant [47]. Additional functions of fungal oxypilins have also been reported. They are related to fungal development regulation, metabolism, and host-pathogen interaction [42,48,49]. The synthesis of oxylipins proceeds due to substrates released by phospholipids and acylglycerides such as: oleic, linoleic, linolenic, and arachidonic acids [50,51]. Various reactions occurring in an oxidizing environment, in combination with enzymatic activity, contribute to the formation of various oxylipins from a given fatty acid. [52]. In our study we have tentatively identified thirtyseven oxylipins of predominantly C 18 chain (Table 2); among them, trihydroxyoctadecenoic acid isomers I-VII (56, 69, 71, 74, 81, 83, 85) have been found in the tested Diaporthe isolates. It should be mentioned that the differences in fragmentation patterns between structural isomers were minimal and did not allow us to determine the position of double bonds or hydroxyl groups in the analyzed compounds. Previously, similar metabolites have been produced in the tubers of taro (Colocasia antiquorum) as a defense response to inoculation with black rot fungus (Ceratocystis fimbriata) [53]. They were isolated for the first time from the Chinese truffle Tuber indicum [54]. It has been proven, for example, that (9S,12S,13S)-tri-hydroxyoctadeca-10E-enoic acid had antifungal activities against Magnaporthe grisea causing rice blast disease [55], and (13S)-hydroxy-9,11-octadecadienoic acid had nematocidal properties [56].

Chromones
Chromones are naturally occurring phenolic derivatives of chromone (1,4-benzopyrone or 4H-chromen-4-one) and are isomers of coumarin. They are produced abundantly by many genera of plants, being a part of a normal healthy diet and by fungi. This class of compounds is mainly associated with antioxidant, antimicrobial, anticancer, and antiinflammatory activities [57]. In our study, Diaporthe spp. produced phomochromone A (45) (Figure 3), which can exhibit an antifungal, antibacterial, and algicidal activities, which is supported by the literature. For example, two new chromones, phomochromone A and B, have been isolated from the endophytic fungus Phomopsis spp. from Cistus monspeliensis which showed good antifungal, antibacterial, and algicidal properties towards Septoria tritici, Microbotryum violaceum, Botrytis cinerea, E. coli, Bacillus megaterium, and Chlorella fusca [58]. Amycolachromone E (52) (Figure 3) and the series of other chromone derivatives were isolated from the deep-sea marine actinomycete Amycolatopsis sp. [59].
plant's oxylipin to achieve their own benefits. For example, by increasing the production of toxins, they improve their virulence [45,46], and by increasing sporulation they can accelerate reproduction in the tissues of the host plant [47]. Additional functions of fungal oxypilins have also been reported. They are related to fungal development regulation, metabolism, and host-pathogen interaction [42,48,49]. The synthesis of oxylipins proceeds due to substrates released by phospholipids and acylglycerides such as: oleic, linoleic, linolenic, and arachidonic acids [50,51]. Various reactions occurring in an oxidizing environment, in combination with enzymatic activity, contribute to the formation of various oxylipins from a given fatty acid. [52]. In our study we have tentatively identified thirty-seven oxylipins of predominantly C18 chain (Table 2); among them, trihydroxyoctadecenoic acid isomers I-VII (56, 69, 71, 74, 81, 83, 85) have been found in the tested Diaporthe isolates. It should be mentioned that the differences in fragmentation patterns between structural isomers were minimal and did not allow us to determine the position of double bonds or hydroxyl groups in the analyzed compounds. Previously, similar metabolites have been produced in the tubers of taro (Colocasia antiquorum) as a defense response to inoculation with black rot fungus (Ceratocystis fimbriata) [53]. They were isolated for the first time from the Chinese truffle Tuber indicum [54]. It has been proven, for example, that (9S,12S,13S)-tri-hydroxyoctadeca-10E-enoic acid had antifungal activities against Magnaporthe grisea causing rice blast disease [55], and (13S)-hydroxy-9,11-octadecadienoic acid had nematocidal properties [56].

Chromones
Chromones are naturally occurring phenolic derivatives of chromone (1,4benzopyrone or 4H-chromen-4-one) and are isomers of coumarin. They are produced abundantly by many genera of plants, being a part of a normal healthy diet and by fungi. This class of compounds is mainly associated with antioxidant, antimicrobial, anticancer, and anti-inflammatory activities [57]. In our study, Diaporthe spp. produced phomochromone A (45) (Figure 3), which can exhibit an antifungal, antibacterial, and algicidal activities, which is supported by the literature. For example, two new chromones, phomochromone A and B, have been isolated from the endophytic fungus Phomopsis spp. from Cistus monspeliensis which showed good antifungal, antibacterial, and algicidal properties towards Septoria tritici, Microbotryum violaceum, Botrytis cinerea, E. coli, Bacillus megaterium, and Chlorella fusca [58]. Amycolachromone E (52) (Figure 3) and the series of other chromone derivatives were isolated from the deep-sea marine actinomycete Amycolatopsis sp. [59].

Sesquiterpenoids
Drimane-type sesquiterpenoids are a large group of compounds that have been found in plants and fungi, exhibiting various biological activities [60,61]. During the research conducted by Zang et al. [62] and Chen et al. [63], a variety of new drimane-type metabolites, including diaporols B-I (104), Q, and R, have been isolated from the mangrove endophytic Diaporthe sp. [62,63]. Furthermore, two drimane-type

Phthalides
Phthalides are natural substances used in traditional medicine in Asia, Europe, and North America, which can be found both in plants and fungi [65][66][67][68][69]. In our study Diaporthe spp. produced the convolvulanic acid A isomers I-II (9,13), which was previously reported from Phomopsis convolvulus, a host-specific pathogen of field bindweed (Convolvulus arvensis) [66]. This metabolite showed phytotoxic activity against C. arvensis, proving that it could be used as an herbicide to control this weed effectively [66].

Hybrid Compounds
HRESIMS analysis revealed a molecular formula of C27H33ClO10 ([M − H] − at m/z 551.1692) for compound 135, suggesting close structural analogy to pestalotiopene A [68]. The structural similarity of both compounds was further corroborated by detecting the same mass fragment at m/z 301.1206 with the characteristic chlorine isotope splitting, corresponding to the altiloxin B part of pestalotiopene A. Previously, drimane sesquiterpene-cyclopaldic acids hybrids, pestalotiopens A and B, were isolated from the mangrove-derived fungus Pestalotiopsis sp. obtained from leaves of the Chinese mangrove Rhizophora mucronate [68]. Pestalotiopen A (135), an altiloxin B-O-methylcyclopaldic acid hybrid, showed moderate antibacterial activity against Enterococcus faecalis [68]. Cyclopaldic acid was also produced by Seiridium cupressi, the pathogen of a canker disease of cypress, showing phytotoxic and antifungal activity [67], and by Coccomyces strobi isolated from needles of Pinus strobus, showing moderate growth inhibition of Microbotryum violaceum (=Ustilago violacea) and weak antibiotic activity against Bacillus subtilis, with no inhibition observed against E. coli at the highest tested concentration [69]. In the search for natural products as an alternative to synthetic pesticides, cyclopaldic acid has been reported to possess insecticidal [70], fungicidal [71], as well as herbicidal [72] activities. Recently, Samperna et al. [73], during the investigation of the effects of cyclopaldic acid in Arabidopsis thaliana plants and protoplasts, showed that this metabolite induced leaf chlorosis, ion leakage, membrane-lipid peroxidation, hydrogen peroxide production, and inhibited root proton extrusion in vivo and plasma membrane H + -ATPase activity in vitro. In our study, we report the presence of over twenty-five compounds, ethers of altiloxin A and its derivatives with cyclopolic acid (51, 88, 103, 109, 126, and 127), and ethers of altiloxin B and its derivatives with either (iso)cyclopaldic acid, cyclopolic acid, or salfredins A7/C3 (57, 63, 66, 70, 76, 79, 87, 90, 91, 93, 95, 98, 101, 106, 107, 112-115,   Figure 4. Putative structures of sesquiterpenoids found in the tested Diaporthe isolates.

Phthalides
Phthalides are natural substances used in traditional medicine in Asia, Europe, and North America, which can be found both in plants and fungi [65][66][67][68][69]. In our study Diaporthe spp. produced the convolvulanic acid A isomers I-II (9,13), which was previously reported from Phomopsis convolvulus, a host-specific pathogen of field bindweed (Convolvulus arvensis) [66]. This metabolite showed phytotoxic activity against C. arvensis, proving that it could be used as an herbicide to control this weed effectively [66].

Hybrid Compounds
HRESIMS analysis revealed a molecular formula of C 27 H 33 ClO 10 ([M − H] − at m/z 551.1692) for compound 135, suggesting close structural analogy to pestalotiopene A [68]. The structural similarity of both compounds was further corroborated by detecting the same mass fragment at m/z 301.1206 with the characteristic chlorine isotope splitting, corresponding to the altiloxin B part of pestalotiopene A. Previously, drimane sesquiterpenecyclopaldic acids hybrids, pestalotiopens A and B, were isolated from the mangrovederived fungus Pestalotiopsis sp. obtained from leaves of the Chinese mangrove Rhizophora mucronate [68]. Pestalotiopen A (135), an altiloxin B-O-methylcyclopaldic acid hybrid, showed moderate antibacterial activity against Enterococcus faecalis [68]. Cyclopaldic acid was also produced by Seiridium cupressi, the pathogen of a canker disease of cypress, showing phytotoxic and antifungal activity [67], and by Coccomyces strobi isolated from needles of Pinus strobus, showing moderate growth inhibition of Microbotryum violaceum (=Ustilago violacea) and weak antibiotic activity against Bacillus subtilis, with no inhibition observed against E. coli at the highest tested concentration [69]. In the search for natural products as an alternative to synthetic pesticides, cyclopaldic acid has been reported to possess insecticidal [70], fungicidal [71], as well as herbicidal [72] activities. Recently, Samperna et al. [73], during the investigation of the effects of cyclopaldic acid in Arabidopsis thaliana plants and protoplasts, showed that this metabolite induced leaf chlorosis, ion leakage, membrane-lipid peroxidation, hydrogen peroxide production, and inhibited root proton extrusion in vivo and plasma membrane H + -ATPase activity in vitro. In our study, we report the presence of over twenty-five compounds, ethers of altiloxin A and its derivatives with cyclopolic acid (51, 88, 103, 109, 126, and 127), and ethers of altiloxin B and its derivatives with either (iso)cyclopaldic acid, cyclopolic acid, or salfredins A7/C3 (57, 63, 66, 70, 76, 79,  87, 90, 91, 93, 95, 98, 101, 106, 107, 112-115, 117, 122, and 135) (Figure 5). The identity of these compounds was tentatively established by the similarity of fragmentation spectra to those of compound 135 (Table 2). 117, 122, and 135) ( Figure 5). The identity of these compounds was tentatively established by the similarity of fragmentation spectra to those of compound 135 (Table 2).

Metabolite-Based Chemotaxonomy
As a preliminary step in multivariate statistical analysis, PCA analysis provided an unsupervised overview of LC-MS fingerprints obtained in both ionization modes (NI and PI). Both NI and PI PCA score plots revealed a close clustering of the QC samples ( Figure  6A), indicating that the separation, observed between fungal isolates into two distinct chemotypes was mainly due to biological reasons.

Metabolite-Based Chemotaxonomy
As a preliminary step in multivariate statistical analysis, PCA analysis provided an unsupervised overview of LC-MS fingerprints obtained in both ionization modes (NI and PI). Both NI and PI PCA score plots revealed a close clustering of the QC samples ( Figure 6A), indicating that the separation, observed between fungal isolates into two distinct chemotypes was mainly due to biological reasons. To avoid biased group assignment of the PCA plots, samples were statistically assigned into 2 clusters (chemotypes) based on the k−means clustering algorithm in NI mode, and the groups generated by the k−means clustering algorithm in the negative mode were assigned to the positive mode ( Figure 6B). The clustering of the data was easily visualized in both ionization modes, and confirmed by clusters obtained separately by HCA ( Figure 7A,B). The first five PCs explained 75.1% of the variance in NI and 75.4% in PI modes, and 57.1% of the total variance was projected in the first two PCs in NI, while 56.0% in PI, which suggested the similar quality of data obtained in both ionization modes. Indeed, the PCA score plots showed similar patterns with specific host plants (understood as metadata) grouped together (pear, sweet cherry, and walnut), while the rest were much more dispersed, and there were no clear associations between the To avoid biased group assignment of the PCA plots, samples were statistically assigned into 2 clusters (chemotypes) based on the k-means clustering algorithm in NI mode, and the groups generated by the k-means clustering algorithm in the negative mode were assigned to the positive mode ( Figure 6B). The clustering of the data was easily visualized in both ionization modes, and confirmed by clusters obtained separately by HCA ( Figure 7A,B). The first five PCs explained 75.1% of the variance in NI and 75.4% in PI modes, and 57.1% of the total variance was projected in the first two PCs in NI, while 56.0% in PI, which suggested the similar quality of data obtained in both ionization modes. Indeed, the PCA score plots showed similar patterns with specific host plants (understood as metadata) grouped together (pear, sweet cherry, and walnut), while the rest were much more dispersed, and there were no clear associations between the metadata and the groups in the PCA. We decided to use NI mode for further work due to the lower complexity of LC-MS data (high amount of in-source collision-induced dissociation in PI).
Molecules 2023, 28, 1175 20 of 32 metadata and the groups in the PCA. We decided to use NI mode for further work due to the lower complexity of LC-MS data (high amount of in-source collision-induced dissociation in PI). To validate the k−means/HCA model and to identify the features responsible for the classification, we performed a supervised PLS-DA analysis, and overall, 52.1% of the total variance was displayed on the first two principal component axes of the PLS-DA score To validate the k-means/HCA model and to identify the features responsible for the classification, we performed a supervised PLS-DA analysis, and overall, 52.1% of the total variance was displayed on the first two principal component axes of the PLS-DA score plot ( Figure 8A), with R 2 X = 0.946 and Q 2 = 0.924 calculated from the first three components via a 10-fold cross-validation method, with Q 2 as the measured performance. Since PLS-DA tends to overfit data, the model was validated to understand whether the separation is statistically significant or is due to random noise. This hypothesis was tested using the permutation test-separation distance (B/W), with 100 permutations with observed statistics having a p < 0.01 ( Figure 8C).
Molecules 2023, 28, 1175 21 of 32 plot ( Figure 8A), with R 2 X = 0.946 and Q 2 = 0.924 calculated from the first three components via a 10-fold cross-validation method, with Q 2 as the measured performance. Since PLS-DA tends to overfit data, the model was validated to understand whether the separation is statistically significant or is due to random noise. This hypothesis was tested using the permutation test-separation distance (B/W), with 100 permutations with observed statistics having a p < 0.01 ( Figure 8C).  Table 2) (B); and permutation test results of the PLS-DA model (statistical test: separation distance (B/W)), the number of permutations set at 100 (C).
A p-value below 0.01 in 100 permutations means that not even once (<0.01 × 100) did the permutated data yield a better performance (higher B/W) than the original label, suggesting the significant difference between these two clusters.
Potential variables to separate clusters 1 and 2 in the dendrogram were identified as potential biomarkers using VIP values which estimate the importance of each variable in the projection used in a PLS-DA model. The greater-than-one rule is usually considered for detecting the descriptors with the greatest importance in the projection. However, we decided, due to a large number of significant metabolites (>300), to use VIP scores > 1.8 ( Figure 8B). The peak intensity ratios were also subjected to an unpaired non-parametric test (Wilcoxon rank-sum test, also known as the Mann-Whitney U test) within MetaboAnalyst, and false discovery rates (FDR < 0.05) were calculated to discover if those features are significantly different between cluster 1 and 2. A large fold change (FC > 10) between the two putative chemotypes was also considered a selection criterion, with FC > 100 indicating the presence/absence of the feature in question. As a result, 43 features meeting these conditions (VIP = 2.20-1.81, FDR adj. p-value = 3.23 × 10 −19 -4.71 × 10 −18 , FC = 348-19) were selected (Table 3) for receiver operating characteristic (ROC) analysis in order to assess their potential as chemotaxonomical biomarkers. ROC curves are used to evaluate classification and prediction models in bioinformatics. They are often summarized in a single metric known as area under the curve (AUC), where AUC = 1.0 indicates an excellent classifier and AUC = 0.5 means the classifier has no practical utility [74]. In this regard, we calculated the AUC for each selected candidate biomarker, and the AUC values obtained ranged from 0.972 to 1.000 (Table 3). Furthermore, to consider factors other than genetics, i.e., host plant, year of strain isolation or storage time, a combination of multiple individual markers must be considered into a single multivariate model, providing improved levels of discrimination and confidence. To this end, we applied the PLS-DA model to combine our 43 selected markers to obtain the AUC ( Figure  9A), and predicted the classification probability into each chemotype ( Figure 9B). The  Table 2) (B); and permutation test results of the PLS-DA model (statistical test: separation distance (B/W)), the number of permutations set at 100 (C).
A p-value below 0.01 in 100 permutations means that not even once (<0.01 × 100) did the permutated data yield a better performance (higher B/W) than the original label, suggesting the significant difference between these two clusters.
Potential variables to separate clusters 1 and 2 in the dendrogram were identified as potential biomarkers using VIP values which estimate the importance of each variable in the projection used in a PLS-DA model. The greater-than-one rule is usually considered for detecting the descriptors with the greatest importance in the projection. However, we decided, due to a large number of significant metabolites (>300), to use VIP scores > 1.8 ( Figure 8B). The peak intensity ratios were also subjected to an unpaired non-parametric test (Wilcoxon rank-sum test, also known as the Mann-Whitney U test) within MetaboAnalyst, and false discovery rates (FDR < 0.05) were calculated to discover if those features are significantly different between cluster 1 and 2. A large fold change (FC > 10) between the two putative chemotypes was also considered a selection criterion, with FC > 100 indicating the presence/absence of the feature in question. As a result, 43 features meeting these conditions (VIP = 2.20-1.81, FDR adj. p-value = 3.23 × 10 −19 -4.71 × 10 −18 , FC = 348-19) were selected (Table 3) for receiver operating characteristic (ROC) analysis in order to assess their potential as chemotaxonomical biomarkers. ROC curves are used to evaluate classification and prediction models in bioinformatics. They are often summarized in a single metric known as area under the curve (AUC), where AUC = 1.0 indicates an excellent classifier and AUC = 0.5 means the classifier has no practical utility [74]. In this regard, we calculated the AUC for each selected candidate biomarker, and the AUC values obtained ranged from 0.972 to 1.000 (Table 3). Furthermore, to consider factors other than genetics, i.e., host plant, year of strain isolation or storage time, a combination of multiple individual markers must be considered into a single multivariate model, providing improved levels of discrimination and confidence. To this end, we applied the PLS-DA model to combine our 43 selected markers to obtain the AUC (Figure 9A), and predicted the classification probability into each chemotype ( Figure 9B). The performance of this model was tested using a balanced Monte-Carlo cross-validation procedure, and as a result the average accuracy based on 100 cross-validations was 0.991.  Hierarchical clustering with a heat map is also shown to easily visualize the concentration variation of the top 100 tentatively identified metabolites (according to t−tests) expressed in the tested Diaporthe isolates (Figure 10). A sharp contrast of their accumulation is observed, while at the same time the samples are clearly grouped by their group membership, determined by HCA and k−means analyses. Hierarchical clustering with a heat map is also shown to easily visualize the concentration variation of the top 100 tentatively identified metabolites (according to t-tests) expressed in the tested Diaporthe isolates (Figure 10). A sharp contrast of their accumulation is observed, while at the same time the samples are clearly grouped by their group membership, determined by HCA and k-means analyses.
The study on the utilization of metabolites as chemotaxonomic markers for species identification refers to the genus Penicillium, Aspergillus, Fusarium, Alternaria, and the Xylariaceae family [75,76]. However, in the case of Diaporthe, this type of research was limited. In the research conducted by Horn et al. [77,78] on endophytic Phomopsis (=Diaporthe) from woody host, three metabolites named phomodiol, phomopsolide B, and phomopsichalasin were indicated as potential chemotaxonomic markers for this fungi. In addition, Abreu et al. [79] showed that the production of secondary metabolites by Phomopsis and related Diaporthales may be species-specific, indicating the value of utilizing the metabolic analysis in taxonomic research on closely related species.
In our research, isolates belonging to the Diaporthe eres species complex isolated from fruit trees produced 153 metabolites from which 43 were recognized as potential chemotaxonomic markers, mostly belonging to the drimane sesquiterpenoid-phthalide hybrid class. This group included mainly phytotoxic compounds such as cyclopaldic acid and altiloxin A, B and their derivatives. It is noteworthy that during our investigation, the phytotoxic compound cyclopaldic acid was produced not only by the pathogenic Diaporthes species but also by the endophytic D. eres isolate 1420S, previously described by Abramczyk et al. [32] and used in the present study. Following the observations of Graniti et al. [67] and McMullin et al. [69], the production of phytotoxic cyclopaldic acid may be related to Diaporthe changing its lifestyle from endophytic to pathogenic, under favorable conditions. Thus, it is possible that endophytic D. eres isolate 1420S [32], is a weak opportunistic pathogen, switching from an endophytic to a pathogenic phase when the host tissue becomes weakened. This issue requires more advanced research in the future.  Table 3 are enclosed in red rectangles.
The study on the utilization of metabolites as chemotaxonomic markers for species identification refers to the genus Penicillium, Aspergillus, Fusarium, Alternaria, and the Xylariaceae family [75,76]. However, in the case of Diaporthe, this type of research was limited. In the research conducted by Horn et al. [77,78] on endophytic Phomopsis (=Diaporthe) from woody host, three metabolites named phomodiol, phomopsolide B, and phomopsichalasin were indicated as potential chemotaxonomic markers for this fungi. In addition, Abreu et al. [79] showed that the production of secondary metabolites by Phomopsis and related Diaporthales may be species-specific, indicating the value of utilizing the metabolic analysis in taxonomic research on closely related species.
In our research, isolates belonging to the Diaporthe eres species complex isolated from fruit trees produced 153 metabolites from which 43 were recognized as potential chemotaxonomic markers, mostly belonging to the drimane sesquiterpenoid-phthalide hybrid class. This group included mainly phytotoxic compounds such as cyclopaldic acid and altiloxin A, B and their derivatives. It is noteworthy that during our investigation, the phytotoxic compound cyclopaldic acid was produced not only by the pathogenic Diaporthes species but also by the endophytic D. eres isolate 1420S, previously described by Abramczyk et al. [32] and used in the present study. Following the observations of Figure 10. Hierarchical clustering with the heat map generated from the top 100 tentatively identified metabolites present in the tested Diaporthe isolates, according to t-tests, using Pearson distance for similarity measure and Ward's linkage algorithm for clustering. Clusters were grouped based on the HCA/k-means analyses shown in Figures 6B and 7. Cell colors indicate relative concentration values as high (dark brown) or low (dark blue), with samples in columns and features (MS DIAL ID in NI) in rows. Features from Table 3 are enclosed in red rectangles.

Fungal Strains and Culture Conditions
We investigated 40 Diaporthe strains isolated during previous studies from different species of fruit trees growing in south-eastern Poland (Table 1) [4,31]. All axenic cultures were deposited at the Fungal Collection of Phytopathology and Mycology Subdepartment, University of Life Sciences in Lublin (Poland). Thirty-nine came from shoots with visible disease symptoms and one from healthy Prunus domestica as endophyte, described previously by Abramczyk et al. [32]. Diaporthe strains were isolated according to the methodology described by Król [80]. Healthy fragments of the tested plants were properly disinfected by rinsing several times, first in a 10% sodium hypochlorite solution, then in sterile distilled water. After drying, the plant fragments were placed on potato dextrose agar (PDA, Difco) and incubated for 5 days at 25 • C, in the dark. When the fungus colonies appeared, pure cultures were prepared according to the methodology described previously [80].

DNA Extraction, Amplification and Sequencing
Strains were incubated on PDA at 25 • C for 7 days before to DNA extraction. The total genomic DNA was extracted using the FastDNA ® SPIN Kit and the FastPrep ® Instrument (Qbiogene, Inc., Carlsbad, CA, USA), according to the manufacturer's protocol. All extracted DNA was stored at −20 • C until use.
The amplification of the fragment of the internal transcribed spacer region (ITS) of the nuclear ribosomal RNA gene, the universal primers ITS1: TCCGTAGGTGAACCTGCGG and ITS4: TCCTCCGCTTATTGATATGC were used [81]. For the amplification of ITS regions, 25 µL of the reaction mixture was prepared, which consisted of the following components: 1 µL of genomic DNA (5 ng/µL), DreamTaq™ Green PCR Master Mix (2×) (Thermo Scientific, Waltham, MA, USA) in a volume of 12.5 µL, primers (10 µM) in a volume of 1 µL each and purified water in a volume of 9.5 µL. The PCR reaction was run under the following conditions: 95 • C for 3 min, followed by 39 cycles of 95 • C for 30 sec, 55 • C for 50 sec, 72 • C for 1 min, and final extension at 72 • C for 10 min. Sequencing of the obtained PCR products was performed in the Genomed S.A. (Warsaw, Poland). The sequence data received were deposited in GenBank ( Table 1). The Bionumerics 7.6 (Applied Maths NV., Sint-Martens-Latem, Belgium) and SEED v.2.1.05 (Institute of Microbiology CAS, Prague, Czech Republic) software was used for bioinformatic analyses.
The obtained sequences were blasted against the NCBIs GenBank nucleotide database to determine the closest related species.

Extraction of Fungal Metabolites
For metabolite extraction, 40 Diaporthe strains were three-point inoculated on 90 mm Petri plates containing PDA, and incubated for 28 days at 23 • C under a 12 h photoperiod, referring to the methodology of Abreu et al. [80], with modifications. Fungal discs (5-mm diameter) were collected in three individual biological repetitions each (n = 3). Each fungal culture (120 total) and three non-inoculated medium samples were freeze-dried (Christ Gamma 1-16 LSC, Martin Christ, Osterode am Harz, Germany) and subsequently ground with a mortar. Dried material (25 mg) was transferred to a 5 mL screw-capped centrifuge tube (Eppendorf, Hamburg, Germany) and added to 2.5 mL of extraction solvent mixture, MeOH/H 2 O 80:20 (v/v). Samples were then thoroughly vortex-mixed for 1 min and ultrasonicated for 20 min under 4 • C. Samples were centrifuged (18,000× g for 20 min under 4 • C), and the supernatants were transferred to separate vials and analyzed using UHPLC-QTOF HRMS. A QC (Quality Control) sample (aliquot of all samples) was also prepared and injected six times before randomized sample injection for column conditioning and at every forty samples to evaluate the performance of the LC-MS method during the detection.
Ionization spray voltages were set to 4.0 kV (for PI) and 3.0 kV (for NI); dry gas flow was 6 l/min; the dry gas temperature was 200 • C; collision cell transfer time was 90 µs; and nebulizer pressure was 0.7 bar. MS1 and MS/MS data (range 80-1800 m/z) were collected using Bruker DataAnalysis 4.3 software in data-dependent acquisition (DDA) mode-after each full MS1 scan, the two most intense ions were fragmented with collision energies of 20 eV for PI and 30 eV for NI.

Data Processing and Metabolite Identification
LC-MS raw data were first converted into the 'Analysis Base File' (ABF) format [82] using Reifycs Abf (Analysis Base File) Converter (https://www.reifycs.com/AbfConverter/ (accessed on 25 May 2021)) and processed with MS-DIAL (RIKEN, version 4.90) [83]. MS1 and MS2 tolerances were set to 0.01 and 0.05 Da, respectively, in centroid mode for each data set (PI and NI). In PI and NI modes, automatic feature detection was performed between 3.0 and 27.0 min for mass range between 80 and 1800 Da. The minimum peak height intensity was set to 2000 for NI and 3000 for PI modes, respectively; linear-weighted moving average as the smoothing method using 5 scans and peak width 5 scans. Peaks were aligned on a QC reference file with an RT tolerance of 0.10 min and a mass tolerance of 0.015 Da and retained in the feature table if they appeared in at least 3 samples. All peaks detected from non-inoculated medium were removed from the generated matrix if their "Sample average/blank average" ratio was lower than 10, thus removing the background and contaminants and preserving the true biological mass signals from LC-MS data.
The kept significant features were exported to the MS-FINDER program (RIKEN, version 3.52) for in silico-based annotation using the hydrogen rearrangement rules (HRR) scoring system [84]. The MS1 and MS2 tolerances were set to 10 and 25 ppm, respectively, and the isotopic ratio tolerance set to 20%. The formulas were filtered to exclusively contain only C, H, O, N, P, S, and Cl atoms. Selected compounds were searched against the builtin database in the MS-FINDER system: NANPDB (Northern African Natural Products Database), KNApSAcK, COCONUT, T3DB (the toxin and toxin target database), and NPA (Natural Products Atlas), and only structures with a score above 5 were retained for thorough analysis. Fungal metabolites were tentatively identified by their high-resolution mass data, MS/MS fragmentation pattern analysis, UV data, and published literature.

Multivariate Statistical Analysis
The aligned data table was LOWESS (locally weighted scatterplot smoothing), normalized using the pooled QC samples and exported from MS-DIAL software to comma-separated value (CSV) format prior to analysis using MetaboAnalyst (version 5.0) [85]. The data were filtered by removing variables showing low repeatability among QC samples (RSD > 20%). Two data matrices were constructed, one in PI mode (120 isolates × 3557 metabolites) and the second in NI mode (120 isolates × 1759 metabolites). The samples were then normalized by the sum to account for the effects of sample dilution (different content of culture medium in the samples), data were log10-transformed to correct for heteroscedasticity and Pareto-scaled to reduce the influence of intense peaks, which transformed the data matrix into a more Gaussian-type distribution [86,87]. First, unsupervised principal component analysis (PCA) was used as an exploratory data analysis to provide an overview of LC-MS fingerprints. Unsupervised groups from the PCA were assigned by k-means clustering analysis and confirmed by hierarchical cluster analysis (HCA) performed to obtain a dendrogram of fungal strains according to metabolite profiling (Pearson distance measure, Ward's clustering algorithm). On the clusters obtained, a partial least squares discriminant analysis (PLS-DA) was conducted using clusters as Y value, and their potential variables were selected based on variable importance in projection (VIP > 1.0) values and false discovery rate (FDR < 0.05) by Wilcoxon rank-sum test.

Conclusions
The results of our study demonstrated a rich diversity of metabolites secreted by the tested Diaporth eres species complex. The characterization of these compounds could be the basis for the future research on their isolation and bioactive potential for agricultural applications as biopesticides or biofertilizers.
Furthermore, the future research should include a larger population of Diaporthe from fruit plants from various areas of Poland. It would be worth determining their metabolic profile, then isolating more important compounds to confirm their structure and bioactive properties. In addition, the optimization of culture media and cultivation conditions for producing richer metabolite profiles are necessary for a more conclusive chemical classification of these fungi.
Although the bioactivity of cyclopaldic acid and altiloxins (the main components of the drimane sesquiterpenoid-phthalide hybrids) identified in the present study as potential biomarkers for species belonging to the Diaporthe eres complex is known, as described above, the genes involved in their biosynthesis have not yet been defined. In general, the eukaryotic genes involved in a single metabolic pathway are scattered throughout the genome, whereas the genes required for a fungus to produce a given secondary metabolite are very frequently clustered, adjacent to one another on the chromosome [88]. Such clusters are found in the majority of filamentous fungi and may range from only a few to more than 20 genes [89]. Thus, identifying a biosynthetic gene cluster for the main compounds reported as biomarkers for species from Diaporthe eres complex, could be the next step to supplement the current research by the results relied on the genetic methods used on a larger Diaporthe population.
Over the last decade, multi-locus DNA sequence data and morphological characterization have been extensively used to identify Diaporthe on a species level [3,7,10,20,21,[90][91][92]. The gene regions most commonly used for this purpose in Diaporthe are the internal transcribed spacer (ITS), together with translation elongation factor-1α (EF-1α), β-tubulin, partial histone H3 (HIS), and calmodulin (CAL) [3,6,20,21,93,94]. However, they are still limited to those species for which the comparative sequence data have been deposited in the public database. Nevertheless, a multi-locus sequencing should always be used for identification of Diaporthe species [6]. In agreement with the study of Abreu et al. [79] and Horn et al. [77], the metabolite profiling may support phenotypic species recognition in Diaporthe. Thus, when studying closely related species in the Diaporthe eres complex, a holistic approach combining morphological characterization, metabolic profile and multi-locus sequencing for species identification is certainly worth considering [79].
Characterizing metabolites biosynthesized by Diaporthe infecting shoots of fruit trees is vital for the phytotoxic properties and chemotaxonomy. It is also essential to better understand the conditions under which the fungi start producing the toxins and switch their lifestyle from endophytic to pathogenic.
Finally, it is hoped that the results from our initial research will enrich the biodiversity of the chemical compounds of species from Diaporthe eres complex and provide a series of new information for this genus.