Interactions of Bunias orientalis plant chemotypes and fungal pathogens with different host specificity in vivo and in vitro

Within several plant species, a high variation in the composition of particular defence metabolites can be found, forming distinct chemotypes. Such chemotypes show different effects on specialist and generalist plant enemies, whereby studies examining interactions with pathogens are underrepresented. We aimed to determine factors mediating the interaction of two chemotypes of Bunias orientalis (Brassicaceae) with two plant pathogenic fungal species of different host range, Alternaria brassicae (narrow host range = specialist) and Botrytis cinerea (broad host-range = generalist) using a combination of controlled bioassays. We found that the specialist, but not the generalist, was sensitive to differences between plant chemotypes in vivo and in vitro. The specialist fungus was more virulent (measured as leaf water loss) on one chemotype in vivo without differing in biomass produced during infection, while extracts from the same chemotype caused strong growth inhibition in that species in vitro. Furthermore, fractions of extracts from B. orientalis had divergent in vitro effects on the specialist versus the generalist, supporting presumed adaptations to certain compound classes. This study underlines the necessity to combine various experimental approaches to elucidate the complex interplay between plants and different pathogens.

The Brassicaceae Bunias orientalis L. harbours two distinct genetic lineages, which represent two distinct chemotypes, that differ in their glucosinolate profiles and other polar metabolites 23 , as well as in susceptibility to plant enemies [23][24][25] . Plants from one chemotype, collected in the Caucasus region where the species potentially originates 26 , showed lower abundance of various herbivores and pathogens in a field common garden and were of lower host quality for a generalist herbivore in a laboratory experiment than plants of another chemotype, which was collected in the species' invasive range in northern parts of Eurasia 23,25 . However, it is unclear whether differences in pathogen infection spot abundance ultimately represent the impact of infection on the plants and whether chemotype susceptibility of B. orientalis differs in dependence of the degree of host specificity of the pathogens.
Necrotrophic, asexual fungal plant pathogens are ubiquitous in air and soil. If fungal conidia successfully germinate and infect plant tissues, the fungal mycotoxins help to digest and consume these plant parts, ultimately leading to chlorosis, necrosis and wilting of leaves and/or plant death 27 . These necrotrophs can easily be cultivated in the laboratory and thus used for both in vivo and in vitro bioassays. Alternaria brassicae (Berk.) Sacc. (Pleosporaceae) occurs on a wide range of plant species almost exclusively within the Brassicaceae family (~ 640 records on ~ 90 species), including B. orientalis 28,29 . Botrytis cinerea Pers. (Sclerotiniaceae) has an extraordinary high host spectrum (> 3,000 records on > 200 species) including few members of the Brassicaceae (~ 90 records on ~ 20 species but no reports yet on B. orientalis 29,30 ), and may thus moderately tolerate various classes of defence metabolites. These well-studied species could naturally come into contact with B. orientalis, and are thus suitable models to investigate interactions of chemotype traits with fungal attackers with potentially different adaptations to the host chemistry under controlled conditions. Due to their different host range, A. brassicae and B. cinerea are in the following referred to as 'specialist' and 'generalist' , respectively.
In the present study, we used laboratory bioassays to characterise the performance of the fungal pathogen species A. brassicae and B. cinerea in response to plants from two B. orientalis populations from Turkey and Germany, belonging to the distinct chemotypes ('Turkish' versus 'German' chemotype). In an in vivo bioassay pre-damaged leaves were inoculated with pathogens and the degree of plant damage caused by fungal infection was estimated as leaf wilting by measuring leaf water loss. Fungal growth in vivo was measured by quantifying ergosterol as biomarker for fungal biomass in the leaves. To investigate the relevance of plant metabolites in the plant-fungus interaction, fungal growth was measured in vitro in culture broth amended with extracts from the two plant chemotypes using photometer assays. To characterise the activity of groups of metabolites influencing fungal growth, plant extracts were fractioned by polarity and fractions used for in vitro growth assays. We expected divergent effects of plant chemistry on the specialist versus the generalist fungal species. Overall, only a combination of bioassays may allow to pinpoint factors mediating B. orientalis interactions with pathogens.

Results
Visual infection differences between fungal species. Plants of the two chemotypes from Turkey and Germany were inoculated with either A. brassicae or B. cinerea and incubated for eight days. Infection by both fungal species was observed in all B. orientalis plants from both chemotypes. However, A. brassicae and B. cinerea developed different infection patterns (Fig. 1b, c). Infection by A. brassicae was typically visible as necrotic leaf veins, which were wide-spread over the leaf blade (Fig. 1b), while in leaves infected by B. cinerea a circular necrotic spot developed directly around the inoculation area, which was surrounded by a chlorotic ring (Fig. 1c). www.nature.com/scientificreports/ Leaf water content and water loss upon infection. Prior to inoculation, the water content measured in untreated control leaves was significantly higher in plants of the Turkish chemotype than in plants of the German chemotype (linear model, LM; df = 1, F = 7.826, P = 0.019, n = 6 replicates per plant chemotype; Fig. 2). After 8 days of incubation with the pathogens, the leaf water loss compared to the initial water content per chemotype was significantly affected by an interaction of the plant chemotype and the fungal species (linear mixed-effects model, LMM; df = 1, Chi 2 = 6.296, P = 0.012, n = 15 per plant chemotype and fungal species). Infection of A. brassicae caused a higher water loss in plants of the Turkish chemotype than of the German chemotype, whereas water loss upon infection of B. cinerea did not differ between plants of both chemotypes (Fig. 2).

Chemical composition of leaf extracts and fractions.
To characterise the composition of semi-polar metabolites of B. orientalis plants from different chemotypes, leaf extracts were prepared and analysed. Chemical analysis revealed clear differences in the overall metabolic composition of different B. orientalis chemotypes, inferred from combined leaf material of all plant individuals used per chemotype (n = 30 per chemotype) (Fig. 3a). Glucosinolate composition was determined from six leaf samples per chemotype with leaves pooled from five plants per sample. The composition of twelve glucosinolates, identified and quantified as in Tewes et al. 23 , largely differed between the two chemotypes, although the total glucosinolate concentration was comparable in both chemotypes ( Table 1). Plants of the Turkish population were dominated by aliphatic glucosinolates, whereas in the German chemotype p-hydroxybenzyl glucosinolate was highly abundant ( Table 1). The patterns correspond with findings from a previous study 23 in which the metabolite and glucosinolate composition of those chemotypes were characterised on the plant individual level.
To separate classes of potentially active compounds present in B. orientalis leaves, methanol leaf extracts, in which plant material from both chemotypes was combined, were used. Extracts were fractionated using solid phase extraction with four different methanol:water mixtures as eluents and chemically analysed. The four fractions of the extract differed in their dry mass as well as in their presumed (Fig. 4a) and measured (Supplementary www.nature.com/scientificreports/ Information Fig. S2) chemical composition. Glucosinolates were mainly present in the 10% methanol fraction ( Table 1, Fig. S2a) which represented ~ 11.7% of the extracted leaf dry mass. The 25% methanol fraction (Fig. S2b) represented ~ 1.6% of the leaf dry mass and contained few glucosinolates and putative flavonoids, based on their UV spectra recorded at 360 nm (not shown). The 50% methanol fraction (~ 2.6% of leaf dry mass) partly overlapped with both the 25% and the 100% methanol fractions ( Fig. S2b-d) and contained putatively flavonoids and yet unidentified metabolites. The 100% methanol fraction (~ 5.5% of leaf dry mass) contained mainly metabolites that could not be detected using the underlying analytical method (Fig. S2d).

Fungal growth in vitro.
To estimate the effects of semi-polar metabolites of B. orientalis plants from different chemotypes on fungal growth of both species, leaf extracts were prepared and used for amendment of liquid culture broth in which fungal growth was measured photometrically. Testing the effects of extracts from www.nature.com/scientificreports/ leaf material pooled from all plants within chemotypes on fungal growth, biomass produced by A. brassicae was significantly affected by plant chemotype in high extract concentration (LM; df = 1, F = 70.251, P < 0.001, n = 9 per plant chemotype). Compared to the control, a significant growth inhibition was caused by extracts from the Turkish chemotype, while a growth facilitation was found in response to extracts of the German chemotype plants ( Fig. 3b; for detailed statistical results on fungal growth change in response to all extracts and extract fractions see Supplementary Information Table S1). In contrast, in low extract concentration no fungal biomass differences in response to plant chemotypes were found (LM; df = 1, F = 0.306, P = 0.588, n = 8-10 per plant chemotype), and likewise extracts caused no significant changes in growth compared to the control (Fig. 3b). Biomass produced by B. cinerea did not differ in presence of extracts from plants of different chemotypes in both high (LM; df = 1, F = 0.433, P = 0.519, n = 10 per plant chemotype) and low (LM; df = 1, F = 2.520, P = 0.131, n = 9-10 per plant chemotype) extract concentrations. Compared to the control, all extracts caused significant growth facilitation in B. cinerea, irrespective of plant chemotype and extract concentration (Fig. 3b). However, on average, growth facilitation seemed overall higher in high extract concentration, and slightly lower in extracts from plants of the Turkish than of the German chemotype (Fig. 3b).
To localise classes of B. orientalis metabolites that influence the specialist versus the generalist fungus in particular, four fractions from solid phase extraction of extracts from leaf material from both chemotypes was combined, were used for amendment of liquid culture broth for fungal growth assays. The glucosinolate-containing 10% methanol fraction caused significant growth facilitation in A. brassicae when tested in high concentration, but significant growth inhibition in B. cinerea compared to the control (Fig. 4b). The 25% methanol fraction in high concentration caused on average growth facilitation in both fungal species, which was significant for A. brassicae (Fig. 4b). The 50% methanol fraction caused no significant changes in growth of both fungal species (Fig. 4b). The 100% methanol fraction in high concentration caused significant growth inhibition in A. brassicae but not in B. cinerea (Fig. 4b).

Discussion
In the present study, pre-damaged leaves of both B. orientalis chemotypes were susceptible to both A. brassicae and B. cinerea, indicated by chlorosis, necrosis and water loss of leaves. However, the degree of plant damage, i.e. leaf water loss, was affected by an interaction of plant chemotype and fungal species. When plants were infected with the specialist, plants from the Turkish chemotype had a higher water loss and thus likely suffered more from infection. This suggests that virulence of A. brassicae, but not of B. cinerea, is highly sensitive to intraspecific variation in the host metabolite profile. In addition to chemical traits, morphological leaf traits may influence fungal development and can differ between plant chemotypes, as in this study found for the leaf water content. Thus, various characteristics of both different B. orientalis chemotypes and fungal species with different host specificity may determine the outcome of this in vivo interaction.
Various constitutive plant (defence) metabolites extractable in methanol and tested for their effect on fungal growth in vitro may be involved in divergent chemotype interactions with the specialist fungus, as discussed below. In addition, phytoalexins may have been induced differently in both chemotypes. These rapidly induced metabolites play an important role in anti-pathogen defence [31][32][33] , including defence against both A. brassicae 34,35 and B. cinerea 36,37 .  www.nature.com/scientificreports/ Importantly, not only plants but also pathogens produce a wide range of metabolites, of which some may be responsible for the differences in plant damage caused by the two fungus species on plants of different B. orientalis chemotypes. Mycotoxins damage host plant cells, which is required for fungi to thrive on the plant tissue 27 . In A. brassicae the cyclodepsipeptide 'destruxin B' is a main factor determining disease development 38,39 . Effects of destruxin B have been found to be host-selective 40 , e.g. due to differing efficiency of enzymatic detoxification by the host plant 41 , and may also differ in the impact on distinct chemotypes within B. orientalis. In contrast, in B. cinerea the non-specific sesquiterpene 'botrydial' acts as virulence factor in interaction with numerous host plant species 30,42,43 . The lack of specificity in this mycotoxin fits with the lack of virulence differences between plant chemotypes in the present study. www.nature.com/scientificreports/ Interestingly, the patterns found in plant damage were not mirrored in the fungal biomass produced after infection in vivo, which did not differ between plant chemotypes, irrespective of fungal species. Nevertheless, the high recovery of ergosterol from leaf material infected by A. brassicae and B. cinerea suggests ergosterol as promising biomarker for biomass of these and other eufungal pathogens in infected leaf material. Comparing fungal growth in vitro in presence of raw leaf extract of plants from the two B. orientalis chemotypes, a divergent influence of plant chemotype on different fungal species was revealed, as only the specialist growth was affected. This finding is in line with the high sensitivity towards chemotype differences suggested above for A. brassicae in in vivo assays. Compared to the control A. brassicae showed significant growth inhibition versus growth facilitation by metabolites present in leaf extracts from the Turkish versus the German chemotype. Regarding the invasion history of B. orientalis, such differences in chemical defence may have been shaped by random genetic drift after population founding 23 or by distinct selection pressures the plants experienced in their native (Turkish) versus the introduced (German) range 25 . A reduced pressure by different specialist enemies in novel habitats may have led to changes in the chemical defence composition in plants of the German population 44,45 . However, growth of the broad generalist B. cinerea was not sensitive towards such chemotypic differences, underlining that controlled bioassays should ideally be conducted with several natural enemy species of distinct host specificity.
In the fractions of leaf extract pooled from both B. orientalis chemotypes differences in susceptibility between fungal species were found, which can be linked with the fungal host specificity. Most strikingly, the glucosinolate containing 10% methanol fraction caused growth facilitation in the Brassicaceae specialist but growth inhibition in the generalist. Indeed, a rather high tolerance of glucosinolate breakdown products in vitro, and also a positive correlation of infection and glucosinolate content in vivo has been revealed in Alternaria species specialised to Brassicaceae 2,35,47 . The generalist B. cinerea has been found to be sensitive to these compounds 37,46 , although tolerance of some glucosinolate breakdown products has been suggested 47 . Potential flavonoids and/or other metabolites present in the 25% and 50% methanol fractions did not lead to remarkably divergent growth responses in both fungi. In contrast, one or several compounds only soluble in 100% methanol were likely relevant for B. orientalis defence against A. brassicae but not against B. cinerea. Further bioassay-guided fractionation and analytical techniques are needed to isolate and identify these compounds.
The patterns observed in this study suggesting differing susceptibility of the two B. orientalis chemotypes to the two fungal species were not consistent across experimental approaches. The increased virulence of A. brassicae when infecting the Turkish chemotype in vivo contradicts the higher defence potential of metabolites from that chemotype against this specialist fungus in vitro. Similarly, A. thaliana mutants with a novel glucosinolate were more susceptible to Alternaria brassicicola infection than wild-type plants in vivo, although breakdown products of that glucosinolate were growth inhibitory in vitro 48 , underlining that plant-fungus interactions are determined by an interplay of various factors, rather than by the direct toxicity of certain metabolites. Furthermore, the differences in results on fungal growth in vivo and in vitro in the present study may be linked to the late (8 days) and early (3 days) time points of growth measurement, respectively. While plant chemotype differences seemed to play a role for conidia germination and/or growth in early stages of the infection process, produced biomass may be more balanced after a certain time period of successful infection of the plant.
A previous study suggested lower susceptibility of plants from the Turkish than of the German chemotype to unidentified pathogens when grown under field common garden conditions 25 . Contrasting findings in the present in vivo experiments may be partly related to the influence of mechanical leaf barriers that were to some extent reduced by pre-damaging the leaves prior to infection. As mechanical barriers, epicuticular wax layers can largely determine plant resistance against fungal pathogens 49,50 , including A. brassicae 51 and B. cinerea 36 . Moreover, the density and shape of trichomes can be crucial for plant resistance against various plant attackers 36,52,53 . Indeed, trichome patterns highly differ between B. orientalis populations from the Turkish versus the German chemotype 24 . This suggests an important role of mechanical leaf defence traits for susceptibility of B. orientalis chemotypes in natural scenarios.
Taken together, the examined interactions of B. orientalis chemotypes and two fungal pathogen species revealed to be highly complex, potentially depending on the leaf structure and the specific impacts of various plant and fungal metabolites. We could highlight the pronounced influence of fungal host specificity on the fungal response to plant chemotype differences. Overall, our study underlines that the outcomes of different types of field studies and laboratory bioassays strongly depend on the various factors in-or excluded within the experiment. The combination of different types of bioassays helps to avoid misinterpretations and to gain insights in factors influencing individual plant-fungus interactions. www.nature.com/scientificreports/ In vivo assay of plant damage and fungal biomass production. Five days prior to inoculation, one leaf from the youngest fully developed leaf pair was chosen for later inoculation and the other leaf from the pair was harvested, weighed, frozen in liquid nitrogen and stored at − 80 °C. Before weighing, leaves of five plants were pooled into one sample (material of n = 30 plants in n = 6 samples per chemotype). These samples were lyophilised to calculate the water content in untreated leaves and to prepare samples of leaf material for chemical analysis and in vitro bioassays (see below). For each fungal species 30 plants (n = 15 per chemotype) were inoculated between 1 and 5 pm. Per fungal species, five culture plates were used and per plate 42 discs (6.7 mm diameter) with growing mycelium with conidia were taken. Additionally, six replicates of five pooled discs per fungal species were weighed and frozen at − 80 °C as negative control samples for determination of ergosterol content. Prior to inoculation the target leaf was slid in the 25 mm hole of the lid from a plastic cup and the opening sealed with cotton wool (Fig. 1a). The leaves were mechanically damaged five times using a tool of four combined needles (0.4 mm diameter, arranged in a square of 3 × 3 mm). Subsequently, 20 µL of a growth medium, containing 26.5 g L −1 potato dextrose broth (Carl Roth) and 0.03% [w/v] Triton X-100 (Sigma-Aldrich, Steinheim, Germany) in millipore water, was spread on the damaged area and five discs from the fungal cultures were gently pressed upside down on that area. After inoculation, plastic cups were placed on the lids (total volume: 800 mL) and fixed to a wooden stick with wire (Fig. 1a). Eight days after inoculation, the tips of the infected leaves were cut below areas in which fungal infection was visible. All samples were weighed, frozen in liquid nitrogen and stored at − 80 °C. After lyophilisation, samples were weighed again to calculate the water content. The water loss of infected leaves was calculated by relating the water content of each leaf to the water content of the respective pooled untreated control leaves of the same plant population (i.e. one control per five samples).

Methods
Ergosterol is a mycosterine specific for fungal cell membranes and not present in plants, and thus a suitable biomarker for fungal biomass 54,55 . For ergosterol extraction, the lyophilised infected leaf samples and negative controls were pulverised and samples extracted threefold in methanol. The combined supernatants were concentrated and analysed using high performance liquid chromatography coupled with a diode array detector (1260 and 1290 Series, Agilent, Santa Clara, CA, USA). The UV-absorption of ergosterol was quantified at 282 nm and related to a commercial ergosterol reference standard solved in 100% methanol. For details on the extraction and instrument settings see Supplementary Information Methods S2.
In vitro assay of fungal growth. Photometric measurement of mycelium produced in amended culture broth is a robust, small-scale in vitro method to investigate the effects of chemicals on fungal growth 56 . For bioassays, leaf material harvested from the experimental plants prior to infection was used and pooled for each of the two chemotypes. Six technical replicates per chemotype of 30 mg leaf material were extracted three times in 0.6 mL methanol, dried at 35 °C and the dry mass of the extract was determined. For extract fractionation dried extracts from leaf material, combined from both chemotypes, were suspended in 10% methanol in millipore water and applied to solid phase extraction columns (Chromabond C8, 1 mL reservoir, 100 mg bed weight, Macherey-Nagel, Düren, Germany). Columns were eluted with different methanol:water mixtures of decreasing polarity, resulting in 10%, 25%, 50% and 100% methanol fractions of the leaf extract, which were dried and weighed. For all extracts and extract fractions blank samples were prepared, which were treated like the test samples, but contained no leaf material. To determine the chemical composition of extracts and extract fractions, further samples were analysed using ultra high performance liquid chromatography coupled to a diode array detector (Dionex UltiMate 3000, Thermo Fisher, San José, CA, USA) and a quadrupole time of flight mass spectrometer (compact, Bruker Daltonics, Bremen, Germany). Therefore, extracts were prepared from healthy leaf material harvested to calculate the water content as described above (material of n = 30 plants pooled in n = 6 samples per chemotype). In addition, individual extracts of leaf material in the same composition as used in the bioassays, i.e., leaf material pooled from all plants per chemotype, and fractions of extracts from leaf material pooled from all plants including both chemotypes, were prepared. Samples were extracted in 90% methanol and chemical analysis was performed according to Schrieber et al. (2019) 57 , but in negative electrospray ionisation mode and with further modifications of instrument settings and in processing raw data from chromatograms (see Supplementary Information Methods S3). UV absorption was recorded at 360 nm. Glucosinolates were identified using their mass spectra and quantified according to   23 (for details see Supplementary Information Methods S3).
Plant extracts and extract fractions were suspended in 100% methanol in a concentration of 10,000 ppm, based on the extract or fraction dry mass. These extracts, extract fractions or blanks were added to potato dextrose broth in a 1:10 ratio and the solvent was evaporated at 50 °C for 1 h. The amended culture broth was further diluted to 100 ppm or 200 ppm plant extract concentration. For each extract or extract fraction in each concentration, ten bioassay replicates were set-up in 96-well photometer plates, whereby 180 µL of amended culture broth was applied to each test well. Control samples without plant metabolites were prepared likewise from blank samples with 10-20 replicates per extract or fraction and concentration.
Fungal conidia used for bioassays were harvested from culture plates in potato dextrose broth with 0.03% Triton X-100. Depending on conidia size, the conidia suspensions were filtered through two (A. brassicae) or three (B. cinerea) layers of Miracloth tissue (Millipore Corp., Billerica, MA, USA) and diluted to a concentration of approximately 0.5 × 10 5 conidia per mL (A. brassicae) or 5 × 10 5 conidia per mL (B. cinerea). To each test well with amended culture broth 20 µL of conidia suspension were added and the plates were sealed and incubated in the dark while shaking. Optical density of test wells was measured on a multiplate reader at 492 nm at 0 and 72 h after preparation. To calculate relative fungal growth for each replicate the background value at 0 h was subtracted from the 72 h value. The individual values were further related to the mean value of the control replicates to calculate the percentage of relative change in growth. Samples in which aerial mycelium was produced were not suitable for photometrical measurements and thus excluded from the analysis.
Scientific RepoRtS | (2020) 10:10750 | https://doi.org/10.1038/s41598-020-67600-7 www.nature.com/scientificreports/ Statistical analysis. All statistical analyses and figures were done with R (version 3.5.2 58 ). To compare the water content of untreated control leaves between plants of different chemotypes, a LM (i.e. one-factorial ANOVA with F test) was performed. To analyse the influence of plant chemotype and fungal species on the water loss of infected leaves, a LMM was calculated (lme4 package 59 ), using culture plate identity (1-5) nested within fungal species as random factor. The total ergosterol amount in infected leaves was compared between plant chemotypes within each fungal species with LMMs, using the culture plate identity and the day of the extraction procedure (1-3) as random factors. In all LMMs response variables were transformed prior to the analysis, whereby arcus sinus square-root transformation was used for water loss, and log-transformation was used for ergosterol amount. LMMs were fitted with a maximum likelihood approach and P-values revealed based on likelihood ratio tests (Chi 2 tests). Relationships between fungal damage and fungal growth were evaluated by calculating Spearman rank correlations for the untransformed raw data of water loss and ergosterol amount for each fungal species separately.
To investigate whether extracts from plants of different chemotypes differently influence fungal growth in vitro, separate LMs were calculated for different extract concentrations (100 ppm, 200 ppm) within each fungal species. To analyse whether amendment of culture broth with different plant extracts or extract fractions influences fungal growth relative to the controls, the relative inhibition values of replicates within treatments were investigated for being significantly different from zero (i.e. the control mean values) using one-sided sign tests (BSDA package 60 ). The residuals of LMs and LMMs were inspected visually and analysed for normality and homoscedasticity with the Shapiro-Wilk test and the Levene test (car package 61 ), respectively, and did not reveal obvious deviations from these assumptions.