Metabolite Profiling and Bioassay-Guided Fractionation of Zataria multiflora Boiss. Hydroethanolic Leaf Extracts for Identification of Broad-Spectrum Pre and Postharvest Antifungal Agents

Hydroethanolic leaf extracts of 14 Iranian Zataria multiflora Boiss. populations were screened for their antifungal activity against five plant pathogenic fungi and metabolically profiled using a non-targeted workflow based on UHPLC/ESI-QTOFMS. Detailed tandem mass-spectrometric analyses of one of the most active hydroethanolic leaf extracts led to the annotation of 68 non-volatile semi-polar secondary metabolites, including 33 flavonoids, 9 hydroxycinnamic acid derivatives, 14 terpenoids, and 12 other metabolites. Rank correlation analyses using the abundances of the annotated metabolites in crude leaf extracts and their antifungal activity revealed four O-methylated flavones, two flavanones, two dihydroflavonols, five thymohydroquinone glycoconjugates, and five putative phenolic diterpenoids as putative antifungal metabolites. After bioassay-guided fractionation, a number of mono-, di- and tri-O-methylated flavones, as well as three of unidentified phenolic diterpenoids, were found in the most active subfractions. These metabolites are promising candidates for the development of new natural fungicides for the protection of agro-food crops.


Introduction
With more than 3000 species, the Nepetoideae is the largest subfamily in the Lamiaceae family [1]. It comprises numerous genera rich in essential oils and includes important culinary and medicinal herbs, such as thyme, oregano, rosemary, sage, savory, basil, mint, and lemon balm. One of the smallest genera in the subfamily Nepetoideae is Zataria with only one described species, Zataria multiflora Boiss. Z. multiflora is a thyme-like, perennial aromatic shrub native to southwest Asia (Afghanistan, Iran, Oman, Pakistan). Similar to thyme, the aerial parts of Z. multiflora are used as a spice and in traditional folk remedies for their antiseptic, analgesic, carminative, anthelmintic, and antidiarrheal properties [2]. Its essential oil (EO) consists mainly of oxygenated monoterpenes and monoterpene hydrocarbons with carvacrol, thymol, linalool, γ-terpinene, and p-cymene as major components [3]. Due to the antibacterial and antifungal properties of the phenolic monoterpenes thymol and carvacrol, the EO of Z. multiflora is used in pharmaceutical preparations for the treatment of fungal infections of the skin, coughs, bronchitis, and digestive disorders [2,4].
In addition to volatile monoterpenes, Nepetoideae species produce a plethora of nonvolatile secondary metabolites with a broad spectrum of bioactivities [1]. Non-volatile compounds include phenolic diterpenoids (e.g., carnosol, carnosic acid), caffeic acid derivatives (e.g., rosmarinic acid, salvianolic acids), and various flavonoids, such as multiple hydroxylated or methoxylated flavones and flavanones. Despite its regional importance as an aromatic and medicinal plant, the fraction of non-volatile secondary metabolites from the leaves of Z. multiflora is less well explored. So far, only a limited number of metabolites have been identified or putatively annotated, including the thymol/carvacrol derivatives zataroside A and B, zatatriol, multiflotriol, multiflorol, the caffeic acid derivatives rosmarinic acid, lithospermic acid A and B, as well as luteolin, luteolin O-glucuronide, vicenin-2, and naringenin as flavonoids [5][6][7].
Fungal diseases are the main cause of severe pre-and post-harvest losses in agriculture and horticulture [8]. Among the broad spectrum of fungal pathogens, the genera Fusarium, Botrytis, Alternaria, and Colletotrichum have a high potential to infect and destroy various crop plant species in pre-and post-harvest stages, as well as contaminate them with mycotoxins that can harm human and animal health [9]. For example, the grey mold fungus, Botrytis cinerea, causes enormous damage to more than 200 crop species, including small fruit crops and vegetables [10]. Synthetic fungicides have been the silver bullet in the fight against these pathogens for decades. Due to their increasing resistance to existing synthetic fungicides, their environmental toxicity, and stricter regulation, the demand for alternative strategies to control fungal diseases in food crops has increased. One alternative to synthetic fungicides is the use of natural substances that either have antifungal properties or stimulate the plants' natural defense mechanisms. Medicinal and aromatic plants in particular are a rich source of biologically active natural products that can be used not only to combat human diseases and preserve food, but also to develop novel bio-based pesticides for agricultural purposes [11,12].
In recent studies, we investigated the composition and antifungal activities of EOs from leaves of Z. multiflora collected from 14 different natural habitats in Iran [3,13]. The observed chemical diversity of EOs and their different antifungal activities prompted us to further characterize the fraction of non-volatile semi-polar secondary metabolites in this sample set with respect to their composition and antifungal activity. To this end, hydroethanolic leaf extracts of Z. multiflora populations were screened for their antifungal potential against five plant pathogenic fungi and classified based on their metabolite profiles obtained by ultra-high-performance liquid chromatography coupled with electrospray ionization quadrupole time-of-flight mass spectrometry (UHPLC/ESI-QTOFMS). After in-depth characterization of the most active leaf extract using tandem mass spectrometric techniques, correlational approaches and bioassay-guided fraction were applied to identify antifungal metabolites as candidate bio-based pesticides.

Screening of Antifungal Activities of Hydroethanolic Leaf Extracts
Hydroethanolic leaf extracts of Z. multiflora from 14 geographically and environmentally diverse natural habitats in Iran were assayed for their inhibitory effect on the mycelial growth of five phytopathogenic fungi (Table 1). In general, the leaf extracts showed low to moderate, and in some cases even high antifungal activities (inhibition rate > 60%). For all five fungal species tested, antifungal activities were significantly different between Z. multiflora populations (ANOVA, 2.54 × 10 −15 ≤ p adj ≤ 1.8 × 10 −7 ). Leaf extracts from Siriz and Haneshk populations had by far the lowest activity against all tested fungi. In contrast, leaf extracts from Daarbast, Hongooyeh, Jandaq, and Konar Siah populations were the most active. Especially in assays with A. dauci and C. lindemuthianum, the leaf extracts from Jandaq and Daarbast populations caused high mean growth inhibition rates of 64.9% and 65.7%, respectively. Table 1. Antifungal activities of hydroethanolic leaf extracts from plants of 14 Z. multiflora populations against five phytopathogenic fungi. Mean growth inhibition rates and standard deviations determined from three leaf samples per population are shown. Mean values in a column that do not have a common letter are significantly different (Fisher's least significant difference test, p adj ≤ 0.05).

Metabolite Profiling of Hydroethanolic Leaf Extracts
To characterize the metabolite composition, hydroethanolic leaf extracts used for antifungal bioassays (14 populations × 3 samples) were diluted and analyzed by reversedphased UHPLC/DAD/ESI-QTOFMS in positive and negative ion mode. After feature detection and alignment using the XCMS algorithm, two data sets comprising 12,783 and 10,518 molecular features were obtained from the raw data acquired in positive and negative ion mode, respectively. Both data sets were subjected to unsupervised random forest (RF) classification. For data from positive ion mode, visualization of the proximity matrix by multidimensional scaling (Figure 1) revealed three main clusters. Cluster I comprised samples from the Siriz and Haneshk populations, which had the lowest antifungal activities. Samples of the Fasa, Taft, Arsenjan, and Kemeshk populations were located in cluster II, while cluster III included samples from the Darab, Ashkezar, Hongooyeh, and Gezeh populations. Samples from three of the four populations with the highest antifungal activities, including Jandaq, Konar Siah, and Daarbast, and samples from the Gachooyeh population were scattered in both cluster II and III. A corresponding analysis using data from negative ion mode (Supplementary Figure S1) generally confirmed the distribution of populations among the three clusters.
To estimate the variation in metabolite levels between populations, abundances of the detected molecular features were analyzed by ANOVA with population as a factor. Significant differences in abundance between populations were found for 56.7% (7248/12,783) of the molecular features detected in positive and for 61.5% (6479/10,518) of those detected in negative ion mode (p adj ≤ 0.01).

Annotation of Semi-Polar Secondary Metabolites
Due to the high metabolic diversity of the sample set, the in-depth characterization of semi-polar secondary metabolites was limited to a single population that exhibited high antifungal activity and was scattered in clusters II and III. For this purpose, a pooled sample of all leaf extracts from the Konar Siah population was prepared and subjected to accurate mass tandem mass spectral analysis. Based on library searches and manual interpretation of the collision-induced dissociation (CID) mass spectra obtained, a total of 68 metabolites were putatively annotated (Table 2), including 33 flavonoids (1-32), 9 hydroxycinnamic acid (HCA) derivatives (33-41), 14 terpenoids (42-55) and 12 other metabolites (56-67). Of these, 17 were clearly identified using commercially available reference compounds. Representative chromatograms with peak annotation are given in Figure 2, and molecular structures are shown in Supplementary Figure S2. Full chromatographic and mass spectral data and comments on compound identification and fragmentation schemes are given in Supplementary Table S1.

Annotation of Semi-Polar Secondary Metabolites
Due to the high metabolic diversity of the sample set, the in-depth characterization of semi-polar secondary metabolites was limited to a single population that exhibited high antifungal activity and was scattered in clusters II and III. For this purpose, a pooled sample of all leaf extracts from the Konar Siah population was prepared and subjected to accurate mass tandem mass spectral analysis. Based on library searches and manual interpretation of the collision-induced dissociation (CID) mass spectra obtained, a total of 68 metabolites were putatively annotated (Table 2), including 33 flavonoids (1-32), 9 hydroxycinnamic acid (HCA) derivatives (33-41), 14 terpenoids (42-55) and 12 other metabolites (56-67). Of these, 17 were clearly identified using commercially available reference compounds. Representative chromatograms with peak annotation are given in Figure 2, and molecular structures are shown in Supplementary Figure S2. Full chromatographic and mass spectral data and comments on compound identification and fragmentation schemes are given in Supplementary Table S1. a Annotation level: 1: metabolite identified using a reference compound, 2: metabolite putatively annotated based on interpretation of mass spectral data, 3: metabolite class putatively annotated based on interpretation of mass spectral data, 4: unknown metabolite. b Hispidulin and chrysoeriol coelute under chromatographic conditions used for metabolite profiling, for chromatographic separation of both metabolites a modified eluent system is required (see Supplementary Figure S3).  Table 2. Further analytical data is detailed in Supplementary Table S1.

Hydroxycinnamic Acid Derivatives
A total of nine metabolites (33-41) were annotated as HCA derivatives. In the extracted wavelength chromatogram at 260-340 nm of the pooled sample of the Konar Siah population ( Figure 2) the highest peak eluting at 5.69 min was assigned to rosmarinic acid (33) using an authentic reference compound. In a retention time range from 5.4 to 5.9 min, rosmarinic acid is accompanied by other prominent metabolites (34-37) that show similar absorption spectra to rosmarinic acid with absorption maxima around 287 and 332 nm. Under negative ion electrospray conditions, metabolites 34-37 formed deprotonated molecular ions at m/z 717.1461 (34), m/z 1075.2156 (35), m/z 1075.2149 (36), and m/z 1433.2789 (37), while for metabolites 35-37 double deprotonated molecular ions were also detectable at m/z 537.1038 (35), m/z 537.1045 (36), and m/z 716.1388 (37). Therefore, metabolites 34-37 probably represent a series of dehydro-oligomers of rosmarinic acid with elemental compositions of C 18+18n H 16+14n O 8+8n (n = 1-3). Deprotonated 34 showed sequential neutral losses upon CID corresponding to dihydroxyphenyllactic acid (C 9 H 10 O 5 , 198.052 amu) and dehydrated caffeic acid (C 9 H 6 O 3 , 162.032 amu), resulting in a relatively stable fragment ion at m/z 357.061 (C 18 H 13 O 8 − ) from which losses of carbon dioxide and water were observed at higher collision energies. The CID mass spectrum of deprotonated 34 is therefore consistent with a dehydrodimer of rosmarinic acid in which two oxidized rosmarinic acid moieties are linked via a C-C bond, probably forming a phenylcoumaran linkage motif as described for salvianolic acid B/lithospermic acid B. As observed for dehydrodimer 34 Based on elemental composition and fragmentation data from literature [15], metabolite 40 was annotated as 2-(dihydroxyphenyl)ethenyl caffeate, which could represent nepetoidin A or B or geometric isomers thereof, depending on the position of the hydroxy groups. , suggesting that the hexosyl moiety is linked to tyrosol via its aliphatic hydroxy group. In addition to the indole derivative 56, tyramine and N-γ-glutamyl tyramine were annotated as further nitrogen-containing secondary metabolites.

Bioassay-Guided Fractionation of a Hydroethanolic Leaf Extract
To validate the results from correlation analyses, a classical bioassay-guided fractionation of a hydroethanolic extract prepared from a pooled leaf sample of the Konar Siah population was performed. As in the first screening experiment, the same five fungal species were used for monitoring the antifungal activity of the prepared fractions. For a first rough fractionation, the total hydroethanolic extract was separated on Strata C18-E solid phase extraction cartridges using a methanol/water step gradient ( Figure 4A)

Discussion
In the present study, hydroethanolic leaf extracts of 14 populations of Z. multiflora were investigated as potential sources of non-volatile semi-polar secondary metabolites with antifungal activity against five important pre-and postharvest fungal pathogens. Due to their sessile lifestyle, land plants in their natural habitats have to cope constantly with a range of abiotic (drought, heat, UV radiation) and biotic (microorganisms, insects) . The antifungal assays of fractions 1 to 5 were performed in 9 cm Petri dishes using test solutions in ethanol/water, 1/1 (v/v) with a concentration of 10 mg/mL. Due to limited substance amounts, fractions 4-1 to 4-7 were assayed in 4 cm Petri dishes at a concentration of 1 mg/mL. Shown inhibition rates are means of three replicates and reflected by a heatmap. Darker green or blue colors correspond to higher inhibition rates. The complete data set and statistical analyses are given in Supplementary  Table S2. a Assay not performed due to limited substance amounts.

Discussion
In the present study, hydroethanolic leaf extracts of 14 populations of Z. multiflora were investigated as potential sources of non-volatile semi-polar secondary metabolites with antifungal activity against five important pre-and postharvest fungal pathogens. Due to their sessile lifestyle, land plants in their natural habitats have to cope constantly with a range of abiotic (drought, heat, UV radiation) and biotic (microorganisms, insects) stress factors. The corresponding adaption processes of plants to these stressors often include the formation of specialized metabolites with a broad range of bioactivities [17]. For example, the biosynthesis of flavonoids in plants is upregulated in response to high solar irradiance and excess metal ions [18]. In addition, antimicrobial phytoalexins are synthesized de novo following pathogen attack. As a result of these adaption processes, habitat-specific chemotypes can be formed, which are extremely valuable resources for bioactivity studies and help to fully exploit the metabolic repertoire of a given plant species. The metabolite profiles of unfractionated crude extracts of different chemotypes can be correlated with the different biological activities to identify possible target metabolites for isolation and bioassay. In comparison to the classical, time-consuming, bioassay-guided fractionation techniques, such correlation approaches have the potential to accelerate the process of bioactive metabolite discovery [19].
An in-depth analysis of the semi-polar metabolite profile of leaves of the Konar Siah population revealed the presence of flavones, flavanones, and dihydroflavonols as major flavonoids in Z. multiflora. Thus, Z. multiflora has a similar spectrum of flavonoids as described for other Nepetoideae plants, such as rosemary, oregano, sage, basil, and thyme [20][21][22]. Apigenin and luteolin, the two main unmethylated flavones, were present mainly as Oand C-linked mono-and diglycosides. In addition, small amounts of two 6-hydroxyluteolin O-glycosides were detected. From a biosynthetic point of view, the co-occurrence of 6-hydroxyapigenin (scutellarein) and 6-hydroxyluteolin would be plausible. However, the scutellarein O-glycosides were below the detection limit in leaves of the Konar Siah population but could become detectable when analyzing leaf material of other populations. As in numerous other species of the Nepetoideae, nine different O-methylated flavones (17)(18)(19)(20)(23)(24)(25)(26) with one to three methyl groups were detected, of which four of the mono-O-methylated compounds (17)(18)(19)(20) could be clearly identified using reference compounds. O-Methylation of flavones alters their physico-chemical properties, such as lipophilicity, and protects hydroxy groups from conjugation reactions, such as glycosylation. The identification of genkwanin (17) (27), eriodictyol (28), and sakuranetin (29) as well as the dihydroflavonols dihydrokaempferol (31) and dihydroquercetin (32) were identified. Despite a targeted search, no glycoconjugates of flavanones and dihydroflavonols as well as other O-methylated derivatives besides sakuranetin were detected in leaf extracts of the Konar Siah population. According to the proposed biosynthetic scheme of (iso)schaftoside [14], 2-hydroxynaringenin 6,8-di-C-hexoside (30) could be a biosynthetic precursor of the highly abundant apigenin 6,8-di-C-hexoside (4).
Correlation analyses of metabolite abundances with antifungal activities indicated putative antifungal activities of four O-methylated flavones (18,20,25,26), two flavanones (27,28), and two dihydroflavonols (31,32). In bioassay-guided fractionation, various O-methylated flavones were present in subfractions with moderate and high antifungal activity. In particular, the yet uncharacterized di-O-methylated and tri-O-methylated flavones (25,26) from subfractions 4-1 and 4-3 showed promising activities against F. culmorum and B. cinerea. Naringenin (27), eriodictyol (28), and dihydrokaempferol (31) were the major constituents of fraction 3, which was not further fractionated during the bioassay-guided fractionation. Possibly, these three flavonoids could be responsible for the moderate antifungal activity of this fraction. The antifungal activities of some of the identified flavonoids are well-documented in the literature. Naringenin and dihydrokaempferol showed moderate antifungal activity against Candida albicans and Cryptococcus neoformans [23]. In maize, the biosynthesis of non-O-methylated and O-methylated flavones, flavanones, and dihydroflavonols is stimulated in the leaves upon infection with Bipolaris maydis [24]. The antifungal activity of selected representatives, such as naringenin, genkwanin, and xilonenin against the plant pathogens Fusarium graminearum, Fusarium verticillioides, and Rhizopus microspores has been demonstrated [24]. Similarly, sakuranetin (7-O-methylnaringein), detected in subfraction 4-4 with moderate activity against Fusarium culmorum, is a major phytoalexin in rice accumulating in blast-infected leaves and was shown to effectively inhibit spore germination of Pyricularia oryezae [25].
The annotated HCA derivatives comprise the caffeic acid esters rosmarinic acid (33), chlorogenic acid (39), and nepetoidin (40), as well as a glycosylated coumaric acid derivative (41). As in many other species of the Nepetoideae, rosmarinic acid was the quantitatively dominant HCA derivative in Z. multiflora leaf extracts and was accompanied by a number of dehydro-oligomers, including a dehydrodimer (34), two dehydrotrimers (35,36), and a dehydrotetramer (37). Rosmarinic acid dehydrodimers, such as salvianolic acid B/lithospermic acid B and salvianolic acid E have already been isolated from Salvia species [26], but have also been described for other Nepetoideae, such as Origanum vulgare [27] and Melissa officinalis L. [28]. Higher rosmarinic acid dehydrooligomers have been found, for example, in Celastrus hindsii [29]. In order to structurally elucidate the linkage motifs of rosmarinic acid in the detected dehydrooligomers 34-37, their isolation and NMR spectroscopic analysis are required. The detection of nepetoidin in the leaves of Z. multiflora supports its role as chemotaxonomic marker to distinguish the Nepetoideae from the other subfamilies of the Lamiaceae [15]. Both nepetoidin A and B showed antifungal activity against Aspergillus niger [15]. However, in our study, nepetoidin (40) was not highlighted by correlation analyses or bioassay-guided fractionation. One reason for this could be the chemical instability of the nepetoidins, which accumulate in the glands of the leaf and easily decompose when the plant material dries [15].
Correlation analyses indicated positive associations between the abundance of THQ glycosides 43-47 and antifungal activity. THQ is biosynthesized from geranyl diphosphate, which is cyclized to γ-terpinene by a terpene synthase and subsequently oxidized in position 3 or 6 by a cytochrom P450 monooxygenase and a short-chain dehydrogenase/reductase to thymol or carvacrol. The latter are again hydroxylated to THQ by a cytochrom P450 monooxygenase [30]. With regard to the biosynthesis of THQ, it is not surprising that leaves of the populations Siriz and Haneshk, which have a linalool EO chemotype with comparably low levels of carvacrol and thymol [3], also have the lowest levels of THQ (42) and its glycoconjugates 43-47 ( Figure 3). Furthermore, rank correlation analyses using metabolite abundances in essential oils and hydroethanolic leaf extracts from all 14 populations revealed positive associations between carvacrol and THQ glycosides 43-47, and between thymol and THQ (42) ( Supplementary Table S3), supporting their biochemical relationship. Due to their relatively high polarity, THQ glycosides 43-47 were present in fractions 1 and 2 after the first fractionation step, which showed only low antifungal activities in the majority of cases. Despite the positive correlation between abundance and antifungal activity, it is therefore unlikely that THQ glycosides are potent antifungal agents. THQ (42) itself was detected in the most active fraction 4. However, it could not be recovered after the second fractionation step by semi-preparative HPLC and was therefore not tested for its antifungal activity. The reason for this could be the limited chemical stability of THQ. It has been reported that THQ is spontaneously oxidized to thymoquinone when left to stand in hexane at room temperature [30].
Among the eight structurally unidentified diterpenoids, the abundances of diterpenoids #1-#3 (48-50) and diterpenoids #7-#8 (54-55) correlated significantly positively with antifungal activity. Moreover, diterpenoids #1-#3 (48-50) were found in the most active subfractions 4-6 and 4-7. Diterpenoid #1 (48) had the same elemental composition as the phenolic diterpenoid carnosol, but did not co-elute with an authentic carnosol standard. In the constructed debiased sparse partial correlation network, the annotated diterpenoids, with the exception of 51, formed a module in one of the subnetworks (Supplementary Figure S4), supporting their biosynthetic relationship. Unfortunately, the molecular structures of 48-55 could not be elucidated based on the obtained CID mass spectra. However, the antifungal diterpenoids #1-#3 showed absorption maxima in the range of 273-281 nm and upon CID of the pseudomolecular ions neutral losses of propene, propyl radicals, or propane, which could be associated with aromatic isopropyl groups. Therefore, 48-50 can be assumed to be phenolic diterpenoids with structural similarity to abietane-type phenolic diterpenoids known from other species of the Nepetoideae, such as carnosol and carnosic acid. Interestingly, carnosol and carnosic isolated from Salvia fruticosa showed antifungal activity against Aspergillus tubingensis, Botrytis cinerea, and Penicillium digitatum [31]. The antifungal diterpenoids #1-#3 (48-50) are therefore interesting targets for further investigation, including isolation, structure elucidation by NMR, and determination of the spectrum of activity against various plant pathogenic fungi.
The bioassay-guided fractionation revealed other potential antifungal metabolites that were not highlighted by correlation analyses. Although hispidulin (19-1) and chrysoeriol  showed negative correlations between abundance and antifungal activity, they were present together with apigenin (1) in subfraction 4-2, which showed good antifungal activity against F. culmorum. Similarly, genkwanin (17) and sakuranetin (29) were detectable in the moderately active subfractions 4-7 and 4-4, respectively, but showed no significant correlations between abundance and antifungal activity. Previous studies have demonstrated the antifungal activities of hispidulin against B. cinerea [31], of chrysoeriol against Fusarium graminearum and Pythium graminicola [32], and of genkwanin against Fusarium verticillioides and Rhizopus microsporus [24]. It is therefore likely that these compounds also contribute to the antifungal activity of the hydroethanolic leaf extracts of Z. multiflora. In general, bioactive phenolic and flavonoid compounds can prevent fungal growth by inhibiting cell wall formation, cell division, and RNA and protein synthesis [33]. However, the bioactivity of a crude plant extract is a complex trait determined by the specific activities and absolute concentrations of a number of bioactive metabolites. It can additionally be influenced by synergistic effects that cannot be resolved even by bioassay-guided fractionation. Often bioactive metabolites are biosynthetically related to each other, so their accumulation in plant tissue is co-regulated. Based on a diverse set of crude plant extracts with contrasting antifungal activities, correlation analyses can reveal such relationships and accelerate the discovery of antifungal metabolites. However, correlation approaches are always only a first step in identifying bioactive metabolites and cannot replace fractionation approaches and bioassays in providing causative evidence for the bioactivity of a fraction of a crude plant extract or a purified metabolite.
In conclusion, correlation analyses and bioassay-guided fractionation revealed numerous mono-, di-, and tri-O-methylated flavones (17,19,20,25,26), as well as the yet unidentified phenolic diterpenoids 48-50 as promising candidates for metabolites with broad antifungal activity against plant pathogenic fungi. Further studies are now required to determine the molecular structures of the yet unidentified candidate antifungal metabolites and to accurately quantify and compare their antifungal activity and spectrum of activity using pure substances.

Plant Material, Fungal Species and Chemicals
A total of 123 Z. multiflora plants were collected at flowering stage in 14 natural habitats across five provinces from the center to the south of Iran, including their major growing areas in the provinces of Isfahan, Kerman, Yazd, Fars, and Hormozgan (Supplementary Table  S4). Six to eleven individual shrubs were sampled from each habitat. A voucher specimen Formic acid (≥98%, for LC-MS) was purchased from Sigma-Aldrich. Ultrapure water (resistivity ≥ 18.2 MΩ cm) was obtained from an Arium 611 water purification system (Sartorius). Sources of reference compounds used for metabolite identification are listed in Supplementary Table S5.

Preparation of Hydroethanolic Leaf Extracts for Bioassays and Metabolite Analyses
Dried leaf material (approx. 3-4 g) from each of the 123 samples was ground to a fine powder (5 min at 30 s −1 ) using a mixer mill (Retsch MM2) and a steel ball (∅ 8 mm). To reduce the number of samples for bioassays and metabolite analyses, equal aliquots of homogenized leaf material of 2-4 individual shrubs from the same habitat were combined and thoroughly mixed, resulting in a total of 42 pooled leaf samples (14 habitats × 3 pooled samples). Homogenized leaf material (2.50 g) was weighed into a 15-mL polypropylene centrifuge tube. After addition of 10 mL ethanol/water, 1/1 (v/v), the mixture was vortex-mixed (1 min), sonicated (15 min, 20-25 • C), and shaken (30 min, 2000 min −1 , room temperature). The supernatant was transferred into a 25-mL volumetric flask after centrifugation (10 min, 4696× g, 20 • C). The remaining residue was extracted once again with 10 mL ethanol/water, 1/1 (v/v), as described above. Both supernatants were combined, and their volume was adjusted to 25 mL using ethanol/water, 1/1 (v/v). The resulting stock extract (100 mg dry leaf material per mL extract) was stored at 6 • C in the fridge until further use in antifungal assays on 9 cm Petri dishes. For LC/MS-based metabolite profiling, an aliquot of the stock extract was diluted 1/50 with ethanol/water, 1/1 (v/v). For preparation of a quality control (QC) sample, equal aliquots of each of the 42 diluted stock extracts were pooled.

Antifungal Assays
Antifungal assays were performed according to the agar well diffusion method with slight modifications [34,35]. Potato dextrose agar (approx. 20 mL) was poured into Petri dishes (∅ 9 cm). After solidification, an agar plug (∅ 6 mm) was removed using a sterile cork borer. The resulting hole was filled with 100 µL hydroethanolic leaf extract (see Section 4.2), 100 µL dissolved extract fraction (see Section 4.4), or 100 µL ethanol/water, 1/1 (v/v) as control. Afterwards, an agar plug (∅ 6 mm) containing actively growing mycelia was excised from a fungal preculture and placed onto the agar at a distance of 4 cm from the punched hole. Petri dishes were incubated for 4-7 days at 20 • C in darkness and radial growth was monitored every 24 h. Antifungal activity was evaluated by measuring the growth of the mycelia in the direction of the punched hole [34]. The inhibition rate (IR) was calculated using the formula, IR (%) = [(C − T)/C] × 100, where C is the average mycelial growth of the control and T the mycelial growth of the treatment. All assays were performed in triplicate.
In case of limited substance amounts, antifungal assays were performed in smaller Petri dishes (∅ 4 cm) filled with 8 mL of potato dextrose agar. The diameter of the agar hole and the inoculum was reduced to 4 mm, and the assay was run with 30 µL test solution.

Fractionation of Hydroethanolic Leaf Extracts
Equal aliquots of air-dried and homogenized leaf material from all samples collected in the Konar Siah habitat were pooled. The resulting sample (10 g) was extracted twice with 100 mL ethanol/water, 1/1 (v/v), following the method described in Section 4.2. Both extracts were combined and evaporated to dryness in vacuo to give 2.95 g crude extract. An aliquot of the crude extract (1.50 g) was solubilized in 150 mL of methanol/water, 2/8 (v/v) and subjected to solid-phase extraction (SPE) using six Strata C18-E Giga Tubes (10 g/60 mL, Phenomenex, Torrance, CA, USA). SPE cartridges were conditioned with 60 mL of methanol and equilibrated with 60 mL of methanol/water, 2/8 (v/v). Afterwards, 25 mL of the solubilized leaf extract and 25 mL of methanol/water, 2/8 (v/v), were successively applied on the cartridge and the eluate was collected (fraction 1). The cartridge was then eluted with 50 mL methanol/water, 4/6 (v/v), 50 mL methanol/water, 6/4 (v/v), 50 mL methanol/water, 8/2 (v/v), and 50 mL methanol to give fractions 2 to 5, respectively. Fractions of individual SPE cartridges were combined and evaporated to dryness in vacuo yielding 868 mg of fraction 1, 402 mg of fraction 2, 114 mg of fraction 3, 48 mg of fraction 4, and 60 mg of fraction 5. Aliquots of the crude extract and of fractions 1 to 5 were dissolved in ethanol/water, 1/1 (v/v) at a concentration of 10 mg/mL and subjected to antifungal assays on 9 cm Petri dishes.
For further fractionation, 30 mg of fraction 4 was dissolved in 1.3 mL methanol/water, 8/2 (v/v) and subjected to semi-preparative HPLC, which was performed on an 1100 Series HPLC system (Agilent Technologies, St. Clara, CA, USA) equipped with a ReproSil XR 120 C18 column (250 × 8 mm, 4 µm particle size, Dr. Maisch, Ammerbuch, Germany). Water and methanol were both acidified with 0.1% (v/v) formic acid and used as eluents A and B, respectively. The following binary gradient program at a flow rate of 3 mL min −1 was applied: 0-1 min, isocratic 1% B; 1-1.5 min, linear from 1% to 35% B; 1.5-17 min, linear from 35% to 75% B, 17-17.5 min, linear from 75% to 100% B; 17.5-20 min, isocratic 100% B. The column temperature was maintained at 40 • C. Eluting compounds were monitored with a diode array detector at 280 nm and 340 nm. A total of 24 separations were carried out using an injection volume of 50 µL. Seven fractions were collected in the retention time range between 12.5 and 18.5 min. After removal of volatiles compounds in vacuo, 0.2 mg of fraction 4-1, 1.3 mg of fraction 4-2, 0.6 mg of fraction 4-3, 0.9 mg of fraction 4-4, 1.4 mg of fraction 4-5, 2.4 mg of fraction 4-6, and 1.8 mg of fraction 4-7 were obtained. Fractions 4-1 to 4-7 were dissolved in ethanol/water, 1/1 (v/v) at a concentration of 1 mg/mL and subjected to antifungal assays on 4 cm Petri dishes and after further dilution to LC/MS analysis.

UHPLC/DAD/ESI-QTOFMS
LC/MS analyses were performed on an Infinity 1290 series UHPLC system (Agilent Technologies) consisting of a binary pump (G4220A), an autosampler (G4226A, 20 µL loop), an autosampler thermostat (G1330B), and a thermostatted column compartment (G1316C) which was coupled in series with a diode array detector (G4212B) and an iFunnel Q-TOF mass spectrometer (G6550A, Agilent Technologies) via a dual Agilent jet stream electrospray ion source. MassHunter LC/MS Data Acquisition software (Agilent Technologies, version B.06.01) was used for controlling the instrument and data acquisition as well as MassHunter Qualitative and Quantitative Analysis software (Agilent Technologies, version B.07.00) for data evaluation. The mass spectrometer was operated in a low mass range (m/z 1700) and an extended dynamic range (2 GHz) mode. Using these settings, the mass resolution (full width at half maximum) at m/z 922 was approx. 23,000. The instrument was autotuned and calibrated according to manufacturer s recommendations using ESI-L tuning mix (Agilent Technologies). Reference mass correction was used throughout all experiments. For this purpose, a solution of purine (20 µM) and hexakis-(2,2,3,3-tetrafluoropropoxy)phosphazine (20 µM) in acetonitrile/water, 95/5 (v/v) was continuously introduced through the second sprayer of the dual ion source at a flow rate of 20 µL min −1 using an external HPLC pump equipped with a 1:100 splitting device.
Extracts (1 µL) were separated on a Zorbax RRHD Eclipse Plus C18 column (100 mm × 2.1 mm, 1.8 µm particle size, Agilent Technologies) using water and methanol both acidified with 0.1% (v/v) formic acid as eluent A and B, respectively. The following binary gradient program at a flow rate of 400 µL min −1 was applied: 0-10 min, linear from 5% to 95% B; 10-13 min, isocratic, 95% B; 13-15 min, isocratic, 5% B. The column temperature was maintained at 40 • C and the autosampler temperature was kept at 6 • C. Eluting compounds were sequentially detected in a wavelength range of 190-600 nm and in an m/z range of 70-1700, either in positive or negative ion mode. Absorption spectra were acquired using an acquisition rate of 2.5 spectra per second, and centroid mass spectra using an acquisition rate of 3 spectra per second. Instrument settings were as described by Tais et al. [36].
To correct for systematic instrumental drift within metabolite profiling experiments, pooled QC samples were repeatedly analyzed after six analytical samples as well as in the beginning and the end of the experiment. CID mass spectra were acquired in targeted-MS 2 mode using scheduled precursor ion lists and the following parameters: acquisition rate MS, 3 spectra per second; acquisition rate MS/MS, 3 spectra per second; isolation width, narrow (1.3 m/z); collision energy, 10, 20, 30, and 40 V; collision gas, nitrogen. For acquisition of CID mass spectra of in-source fragment ions (pseudo-MS 3 ), funnel exit DC voltage was increased from 50 to 120 V.
For relative quantification of annotated metabolites 1-67, extracted ion chromatograms were generated for respective quantifier ions (Supplementary Table S1, m/z width 0.02) and integrated using MassHunter Quantitative Analysis software. Systematic instrumental drift was corrected individually for each metabolite using repeatedly injected pooled QC samples and a LOWESS/Spline interpolation algorithm (http://prime.psc.riken.jp/ compms/others/main.html#Lowess, accessed on 26 April 2022). The obtained signal-drift corrected peak areas were used for ANOVA, for construction of a heatmap and a debiased sparse partial correlation network, as well as for rank correlation with antifungal activities.

Statistical Analysis
Statistical analyses were conducted using R statistical software (R Foundation for Statistical Computing, Vienna, Austria). To control the false discovery rate in multiple testing, p values were adjusted using the Benjamini-Hochberg procedure implemented in the function p.adjust from the package "stats" (version 4.1.0). The function anova from the same package was used for variance analyses. Metabolite abundances were log 2 -transformed prior to ANOVA. Fisher s least significance difference test was performed with the function LSD.test from the package "agricolae" (version 1. [3][4][5]. Unsupervised random forest classification was performed using the function randomForest from the package "randomForest" (version 4.7-1.1). The obtained proximity matrix was visualized by multidimensional scaling using the function cmdscale from the package "stats". Spearman s rank correlation coefficients and corresponding p values were calculated using the function rcorr from the package "Hmisc" (version 4.7-0). The heatmap was generated in MS Excel 2016 (Microsoft, Redmond, WA, USA) using conditional formatting. For this purpose, signal-drift corrected peak intensities of quantifier ions of metabolites 1-67 were averaged for each population, normalized to the mean of all populations and log 2 -transformed. A debiased sparse partial correlation network was generated from log 2 -transformed instrumentaldrift-corrected peak areas of quantifier ions of metabolites 1-67 using MetaboAnalyst (https://www.metaboanalyst.ca, accessed on 21 June 2022).

Supplementary Materials:
The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/molecules27248903/s1, Figure S1. Unsupervised random forest classification of metabolite profiles in negative ion mode; Figure S2. Molecular structures of identified metabolites; Figure S3. Identification of hispidulin and chrysoeriol; Figure S4. Debiased sparse partial correlation network; Table S1. Chromatographic and tandem mass spectral data of annotated metabolites; Table S2. Antifungal activities of fractions; Table S3. Spearman's correlation coefficients calculated from THQ glycoconjugates abundances and essential oil compounds; Table S4. General geographic and climatic information of sampled natural habitats of Z. multiflora; Table S5. Commercial sources of reference compounds used for metabolite identification.