Phenylphenalenones protect banana plants from infection by Mycosphaerella fijiensis and are deactivated by metabolic conversion

Plant material

In vitro plants of Musa acuminata var. ‘Williams’ (AAA, Cavendish subgroup) were provided by Universidad Católica de Oriente – Rio Negro (Antioquía, Colombia), and in vitro plants of M. acuminata var. ‘KTR’ (AAA, Ibota subgroup) were originally provided by the International Transit Centre (ITC), Bioversity International, Katholieke Universiteit Leuven, Belgium (Leuven, Belgium). Plants were propagated and cultured individually in medium B5Z according to the standard procedures (Banerjee & Delanghe 1985) and maintained at 26 °C under 12 h photoperiod conditions until they reached a height of 10–12 cm (4–5 weeks). Plants were further transferred to 10 cm diameter plant pots containing sterile soil of a mixture of Klasmann‐Ton‐Substrat (KTS), vermiculite 2–3 mm, sand 0.7–1.2 mm and perlite (in a ratio 2:1:1:2) with fertilizer Oscomote Exact 19‐9‐12 (333 mg L⁻¹ soil), and each plant received a dose of Ferty 3 (0.1%; Planta Düngemittel GmbH, Regenstauf, Germany) every week. The plants were maintained in an incubation chamber, Percival Model AR‐95HILX (CLF Plant Climatics, Emersacker, Germany), at a temperature of 26/23 °C (day/night), 70% humidity and 12 h photoperiod for 1 month. Then the plants were transferred to 4 L plant pots under the same soil conditions and kept for 3 months before the infection experiments began.

Fungal organisms

Mycosphaerella fijiensis strains voucher E22 and voucher Ca10_13 were provided by Dr Gert Kema from the Plant Research International (Wageningen, the Netherlands) and maintained in solid potato dextrose agar medium (PDA, emended with 100 μg mL⁻¹ ampicillin), at 25 °C under dark conditions for 15 d. They were further cryo‐preserved according with the International Network for the Improvement of Banana and Plantain (INIBAP) technical guidelines (Carlier et al. 2002) until the inoculum for the infection protocol and bioassays was prepared.

Analytical equipment

One‐dimensional (1D) and two‐dimensional (2D) NMR experiments were carried out using a Bruker Avance 500 NMR spectrometer (Bruker BioSpin, Karlsruhe, Germany), operating frequency 500.13 MHz for ¹H and 125.75 MHz for ¹³C. A triple resonance inverse (TCI) cryoprobe (5 mm) was used to measure spectra at 300 K. Tetramethyl silane (TMS) (0.05%) was used as an internal standard for referencing ¹H and ¹³C NMR spectra unless otherwise indicated. One‐dimensional Nuclear Overhauser Effect Spectroscopy (NOESY) experiments were performed using a Bruker Avance III HD 700 NMR spectrometer (Bruker BioSpin, Karlsruhe, Germany), operating frequency 700 MHz for ¹H. A TCI H‐C/N‐D 1.7 mm microcryoprobe was used to measure spectra at 298 K. NMR spectrometers were controlled with bruker topspin 2.1 and 3.2 software (Bruker BioSpin, Karlsruhe, Germany) for the 500 and 700 MHz instruments, respectively. Electrospray ionization mass spectra (ESIMS) and LC–ESIMS were recorded on a Bruker Esquire 3000 ion trap mass spectrometer (Bruker Daltonics, Bremen, Germany). High‐resolution ESIMS was recorded on ultra high performance liquid chromatography (UHPLC) system of the Ultimate 3000 series RSLC (Dionex, Sunnyvale, CA, USA) connected to an LTQ‐Orbitrap XL mass spectrometer (Thermo Fisher Scientific, Bremen, Germany). Analytical and semipreparative HPLC was performed on an Agilent series HP1100 (binary pump G1312A, autosampler G1367A, diode array detector G1315A; Agilent Technologies, Waldbronn, Germany). The ultraviolet (UV) spectra were recorded by the DAD during analytical HPLC. The semipreparative HPLC system was coupled with the fraction collector CHF 122SB ADVANTEC (Dublin, CA, USA).

Metabolomics analysis

Analysis of inducible phenylphenalenones in Musa during pathogen attack

Antifungal bioassay

A microtiter well method developed by Peláez et al. (2006) was used with some modifications (Hidalgo et al. 2009). Briefly, M. fijiensis strains (vouchers E22 and Ca10_13) grown for 15–17 d at 26 °C ± 1 in PDA (Fluka Analytical, Steinheim, Germany) were used for preparing the fungal inoculum. Mycelium of the culture was scraped with a smear loop and sterile water, and the dense suspension obtained was then fragmented in a vortex mixer with glass beads of 4 ± 0.3 mm diameter (Carl Roth GmbH, Karlsruhe, Germany) followed by filtration through four layers of sterile cheesecloth to obtain uniform myceliar fragments. The concentration of the inoculum was counted under a light microscope using a Neubauer haemocytometer (Marienfeld, Lauda‐Königshofen, Germany) and adjusted to 2 × 10⁵ mycelial fragments mL⁻¹ with sterile water. Sterile microtiter plates 96 well‐F (Sarstedt AG & Co, Nümbrecht, Germany) were used for the biological incubation. Each well was filled with 50 μL of Sabouraud broth (Fluka Analytical), 50 μL of the fungal inoculum (described in the preceding text) and 50 μL of the corresponding compound to be assessed. Concentrations ranging from 0.001 to 0.1 mg mL⁻¹ of each compound were prepared in aqueous dimethyl sulfoxide (12% DMSO, BioReagent for molecular biology, ≥99.9%; Sigma‐Aldrich, Steinheim, Germany) and passed through a filter (0.1 μm, hydrophilic polyethersulfone membrane sterile; PALL Life Sciences, Ann Arbor, MI, USA) before being administered to the incubation system. The culture medium (50 μL; Sabouraud), 50 μL of inoculum and 50 μL of aqueous DMSO (12%, sterile) were used as a reference for growth control. Propiconazole (Fluka Analytical, 98.4%) was used as a positive control. The same procedure was applied for the negative control except that the inoculum solution was replaced by sterile water. The microplates were incubated under darkness or a photoperiod for 8 d (logarithmic growth phase of the fungi) at 25 °C in an incubation chamber. For the light‐controlled experiments, the plates were placed at a distance of 1 m and irradiated by two cool white light tubes (Sylvania Luxline Plus T8 36W/840, Havells Sylvania Großschirma, Germany) during 12 h d⁻¹. Both experiments (darkness and photoperiod) were carried out simultaneously. The absorbance for the optical density (OD_{595 nm}) was recorded by a spectrophotometer (TECAN Infinite M200 built‐in with multiple reads per well; TECAN, Crailsheim, Germany) at time 0 and 8 d after incubation in order to measure the mycelia growth. The experimental design consisted of three biological replicates (including three technical replicates). The half maximal inhibitory concentration (IC₅₀) value was calculated by plotting the mycelial growth (data previously normalized to the reference control) against the logarithm of each concentration of the compound assessed.

Phenylphenalenone metabolism

The antimicrobial activity of phenylphenalenones and derivatives depends on the chemical structure of the compound (Hidalgo et al. 2009; Otálvaro et al. 2007). In order to test whether the survival of M. fijiensis to the incubation with certain phenylphenalenones is in part due to the ability to metabolize or degrade them, an experimental approach was set up including several parameters to be analysed during in vitro interaction of the fungal pathogen and the phenylphenalenone. Thus, seven phenylphenalenone‐type compounds covering a range from weak to strong antifungal activity against M. fijiensis were selected for this analysis. The protocol described in the preceding text for the antifungal bioassay was used with some modifications. The fungal inoculum was prepared and adjusted to a final concentration of 3 × 10⁵ mycelial fragments mL⁻¹ in identical conditions as has been described in the preceding text. Sabouraud broth (600 μL) with 600 μL of the fungal inoculum and 60 μL of each concentration of the compound to be assessed was incubated in a sterile Safe‐Lock Eppendorf tube (2.0 mL; Eppendorf, Hamburg, Germany) containing five sterile ceramic beads (0.7 mm in diameter). Three solutions of each compound (0.21, 0.42 and 0.84 mg mL⁻¹) were prepared in dimethyl sulfoxide (DMSO, BioReagent for molecular biology, ≥99.9%; Sigma‐Aldrich) to achieve a final concentration of 0.01, 0.02 and 0.04 mg mL⁻¹ in the incubation system (maximum concentration of DMSO in the system: 4.7%). Sabouraud broth (600 μL) with 600 μL of the fungal inoculum and 60 μL of DMSO (0.1 μm filtered) was used as control growth. Similarly, 600 μL of Sabouraud broth with 600 μL of sterile water and 60 μL of each compound solution (at each concentration) were prepared in order to ensure a comparable analysis with the treated samples. Ten replicates of control growth, reference compounds and treated samples were used for reproducibility purposes and additional analysis. The samples were maintained at 25 °C, shaken at 150 rpm under dark conditions and incubated for 8 d (logarithmic growth phase). Fungal biomass, soluble protein, ergosterol production and quantification of each phenylphenalenone either in the culture medium as well as in fungal mycelium were analysed as follows.

Biomass quantification

After the microbial incubation, the fungal mycelium was completely removed from each tube and transferred to a new sterile Eppendorf tube. The fungal biomass was washed with 500 μL (×3) phosphate‐buffered saline (sterile PBS, pH 7.4, 10% DMSO) and centrifuged at 20,000 × g for 10 min, and the supernatant was poured off and collected for further analysis. A final rinse was followed with 500 μL of sterile water, and the mycelium was immediately frozen in liquid nitrogen and lyophilized in a freeze drier for 48 h. The biomass was left for 24 h in a desiccator before being weighed on a microbalance Mettler Toledo XP26 (0.001 mg readability; Giessen, Germany).

Soluble protein quantification

The protein quantification was carried out by the Bradford method. Briefly, after 8 d of incubation with the fungus, 120 μL of each culture medium (control growth, negative controls and treated samples) was placed in a sterile microtiter 96‐well plate and mixed with 120 μL of Coomassie reagent (Coomassie Protein Assay Reagent; Thermo Scientific, Rockford, IL, USA) and then allowed to react for 15 min followed by reading the absorbance at 595 nm as indicated by the manufacturer. Three technical replicates of each sample were measured, and the average was used for quantification. The protein concentration was calculated from a calibration curve with bovine gamma globulin as standard.

Ergosterol quantification

The fungal biomass obtained either from control as well as material treated with phenylphenalenones (n = 4) was ground in a paint shaker using stainless steel beads (1.4 mm in diameter; Carl Roth, Karlsruhe, Germany), and 0.7 mg was weighed, transferred to a Safe‐Lock Eppendorf tube (2.0 mL) and subjected to saponification with 1.5 mL of methanolic solution of KOH (0.1 m) followed by heating at 70 °C and shaking at 600 rpm for 10 min (Eppendorf Thermomixer Comfort 1.5 mL; Eppendorf, Hamburg, Germany). The reaction was left at room temperature, and then the tubes were centrifuged at 15,000 × g for 10 min and the supernatant transferred to a 4 mL glass vial. Distilled water (1 mL) was added to the methanolic solution along with 1 mL of diethyl ether for liquid–liquid extraction. The ether phase was poured off, and the extraction with diethyl ether was repeated twice. The organic phases were pooled and evaporated under a stream of nitrogen. The raw extract was resuspended in 60 μL of CDCl₃. ¹H NMR spectra were recorded in a Bruker Avance 700 MHz spectrometer (Bruker Biospin, Karlsruhe, Germany) using a standard 1D NOESY pulse sequence with water suppression, with 512 scans with 32k data points. Other parameters were used as described in the preceding text for metabolomics analysis, and the chemical shifts were referenced to internal TMS (δ 0.0). Ergosterol was quantified using eretic2 software (Bruker Biospin, Karlsruhe, Germany) built‐in in topspin 3.2 (calibration method). Ergosterol (13.35 μm; purity >98%) in CDCl₃ was used as standard for the calibration procedure. The singlet signal at δ 0.63, which integrates for the protons of the methyl group attached to C‐13, was used for quantification. A minimal signal–noise ratio of 80:1 was calculated. A T₁ relaxation time of 349.3 ms was determined for the signal at δ 0.63, and 5 × T₁ was set up as a repetition time. Five standard solutions of ergosterol in the range 0.01–2.0 mm were prepared and evaluated for the accuracy of the molar quantification method. An area of at least five times the full width at half height (FWHH) of the ¹H NMR signal was integrated and considered for the quantification. Absolute error below 2.34% was determined. This value is in agreement with validation methods using NMR spectroscopy (Pauli et al. 2005).

Quantification of metabolized phenylphenalenones

To quantify phenylphenalenones, the culture medium was extracted with ethyl acetate (500 μL × 5) by shaking for 10 min and centrifuged at 10,000 × g for 5 min. The pooled organic phase was evaporated under a stream of nitrogen gas. The extract was resuspended in 50 μL of methanol, and an aliquot of 5 μL of this solution was injected into HPLC. Separation was achieved in a Hibar column RP‐18e (Purospher STAR, 250–4.6 mm, 5 μm particle size) using water (0.1% TFA, solvent A) and methanol (solvent B) as a mobile phase. A linear gradient was used from 0 min (40% B), 20 min (100% B) and 27 min (100% B). The monitor wavelength was set at 254 nm (HPLC method B). The analysis was carried out by triplicate injections of each sample and reference controls. A coefficient correlation r² ≥ 0.9997 was determined by plotting the peak area (average data) versus the concentration of each reference compound. The data were normalized by comparing the peak area of the compounds (average data) of the treated sample against the peak area (average data) of the reference compound. The LOQ calculated for the phenylphenalenones was used as described in the preceding text.

To quantify phenylphenalenones in the mycelium, fungal biomass was collected and lyophilized as described in the preceding text. Methanol (1 mL) was added to the fungal mycelium (0.7 mg), shaken at 650 rpm for 15 min at room temperature and centrifuged at 15,000 × g during 10 min at 4 °C. The extraction was repeated twice, and the organic extracts were pooled and evaporated under stream of nitrogen gas. The extract was resuspended in 50 μL of methanol, and an aliquot of 5 μL was injected into the HPLC system using method B. Triplicate analysis was conducted, and the data were normalized to the reference controls as was described in the preceding text.

Analysis of phenylphenalenone metabolites

To investigate the products derived from the metabolism of phenylphenalenones by the fungus M. fijiensis, an up‐scaled incubation system with anigorufone (9) as reference compound was implemented. Thus, the fungal inoculum (3 × 10⁵ mycelial fragments mL⁻¹) of the M. fijiensis strain Ca10_13 was prepared as described in the preceding text. A volume of 5 mL of inoculum was added to 50 mL of Sabouraud medium (sterile, incubation in Erlenmeyer flask volume 100 mL), and 2.38 mL of a solution of 1.5 m of anigorufone (prepared in 20% DMSO and passed through 0.1 μm sterile filter) was added. A parallel experiment without any treatment was used to control fungal growth (2.38 mL of a solution 20% of sterile DMSO was added). The incubation was carried out at 25 °C, shaken at 150 rpm in a rotatory shaker under darkness for 12 d. Afterwards, the fungal biomass together with the culture medium was transferred to sterile Falcon tubes (50 mL) for centrifugation at 18,000 × g during 15 min. The supernatant was poured off to a new sterile Falcon tube (50 mL), and the mycelium was washed twice with 30 mL of PBS (pH 7.4) and rinsed with 30 mL sterile water (10% DMSO). A centrifugation step was used in each wash, and the supernatants were transferred to new sterile Falcon tube (volume 50 mL). Both fungal mycelium and the supernatants were frozen in liquid nitrogen, lyophilized for 72 h and then kept in a desiccator until further analysis. The lyophilized supernatants were resuspended in 5 mL of methanol and injected into LC–MS system using a LiChrospher 100 column RP‐18 (250–4 mm, 5 μm particle size) with water (0.1% formic acid, solvent A) and methanol (0.1% formic acid, solvent B) as a mobile phase. A linear gradient was used from 0 min (40% B), 20 min (100% B) and 30 min (100% B). The fungal biomass [1.3 g dry weight (DW)] contained in a Falcon tube was ground using stainless steel beads (1.4 mm in diameter) and shaken in a paint shaker for 5 min. Methanol (10 mL × 3) was used for extraction by shaking for 10 min followed by centrifugation. The methanolic fraction was passed through a sterile membrane filter (0.45 μm) and evaporated under a stream of nitrogen gas. The crude extract was resuspended in methanol (5 mg mL⁻¹), and 5 μL was injected into LC–MS system using the parameters described in the preceding text.

The methanolic extracts either from mycelium or from culture medium incubated with phenylphenalenones 5, 7 and 9–12 (Table 1) were analysed by UHPLC–ESIMS. Phenylphenalenones were separated using an Acclaim C18 column (150 × 2.1 mm, 2.2 μm; Dionex, Sunnyvale, CA, USA) with a flow rate of 300 μL min⁻¹ in a binary solvent system of water (solvent A) and acetonitrile (solvent B) (hypergrade for LC–MS; Merck, Darmstadt, Germany), both containing 0.1% (v/v) formic acid (eluent additive for LC–MS; Sigma‐Aldrich, Steinheim, Germany). Separation was accomplished using a linear gradient from 0 min (0% B), 15 min (100% B) and 20 min (100% B) and an equilibration time at 0% B for 5 min. ESI source parameters were set to 4 kV for spray voltage and 35 V for transfer capillary voltage at a capillary temperature of 275 °C. The samples were measured in positive and negative ion mode in the mass range of m/z 100 to 2000 using 30 000 m Δm⁻¹ resolving power in the Orbitrap mass analyser. Data were interpreted using xcalibur software (Thermo Fisher Scientific, Waltham, MA, USA).

Statistical analysis

Unless otherwise stated, statistical tests were carried out with sigma plot 12.0 (SYSTAT Software Inc., San Jose, CA, USA) using analysis of variance (anova). Leven's and Shapiro–Wilk tests were applied to determine error variance and normality of the data, respectively. Holm–Sidak post hoc test was used for pairwise or multiple comparison. Datasets that did not fulfil the assumptions for anova were natural log, decimal log, root square or rank transformed before analysis. A non‐parametric Kruskal–Wallis one‐way anova on ranks was applied for variables that did not meet the normality assumption.

Differential responses of the interaction of Musa–M. fijiensis detected by principal component analysis

In order to explore whether the susceptible Musa variety ‘Williams’ and the resistant variety ‘KTR’ respond differently to infection with the phytopathogenic ascomycete fungus M. fijiensis, in vitro plants of the two varieties were treated with conidia of the strain E22 as described in Section on Materials and Methods. The progress of the BLSD development was monitored by inspecting the leaf surface of the two Musa varieties using a binocular loupe. Initially, both the susceptible variety ‘Williams’ and the resistant variety ‘KTR’ developed similar symptoms at the site of treatment (Table S1). When BLSD symptoms of the treated plants, according to the classification by Fouré (1986), reached a dimension similar to stages 2 and 3, the necrotic zones (sample A) and healthy zones (B) of treated half of the leaves were probed (Fig. 2c). Simultaneously, further samples were taken from the non‐treated half of each treated leaf (D), non‐treated leaves of the treated plants (E) and untreated plants (control, C). When ¹H NMR spectra of crude extracts of necrotic tissue (A) were compared with spectra of non‐necrotic tissue (D) from the same treated leaf and the control plant (C) (Fig. 1), signals of aromatic metabolites were more pronounced in the extracts of leaf areas where the fungal infection had taken place than in extracts of samples taken from ‘healthy zones’. This was observed both for ‘Williams’ and ‘KTR’. ¹H NMR spectra of samples B and E (not shown in Fig. 1) closely resembled those of (C).

Figure 1

¹H NMR spectra (500 MHz, MeOH‐d₄) of crude extracts of Musa leaf tissues from varieties ‘Williams’ (a) and ‘KTR’ (b). Extracts were prepared from necrotic areas of the infected half (Fig. 2, A) and the non‐infected half (D) of a leaf treated with M. fijiensis strain E22, and a leaf of an untreated control plant (C) was taken 35 d after inoculation with fungal conidia. The extensions show the ¹H NMR signals of induced phenolics in the necrotic leaf areas (A). Spectral intensities were scaled with the signal of trimethylsilyl propanoic acid (TSP), which was added as an internal standard.

The ¹H NMR spectra of all extracts are not much different in the regions of primary metabolites such as carbohydrates and amino acids (0.5–5.0 ppm, Fig. 1). However, the aromatic regions (6.0–9.0 ppm, Fig. 1) of the spectra exhibited not only largely enhanced signals obtained from the necrotic zones of infected leaves (A) but, compared with the spectra from extracts of non‐treated tissues, also a significantly increased number of signals. Thus, qualitative and quantitative gains in metabolic activity were observed in the infected zones of the plant and could be interpreted as a response to the pathogen attack. Therefore, it was thought that identification and quantification of the major metabolites produced while the BLSD ran its course would provide insight into the role and turnover of inducible natural products in this plant.

An NMR‐based metabolomic analysis was carried out to gain information about metabolic changes occurring in Musa plants under attack by the pathogen. The full ¹H NMR spectra were used for understanding whether differences other than in the aromatic compounds exist between infected (A), non‐infected (B, D and E) and control (C) Musa plant samples (Fig. 2c). According to the score plots of PCA (Fig. 2a,b), the samples from BLSD‐infected necrotic leaf zones (A, red) were clearly discriminated from those of the control group (C, green) by the first principal component (Fig. 2a,b), mainly owing to aromatic compounds. Interestingly, the other non‐infected plant sections (asymptomatic areas B, non‐treated leaf areas D and E) obtained from plants treated with conidia of M. fijiensis E22 (Fig. 2c) appeared also separated in the PCA (blue, orange and gold) from the group of control Musa plants (green).

Figure 2

Score plots of PC1 versus PC2 of eight‐component principal component analysis (PCA) model of the region 0.5–8.7 ppm of the ¹H NMR spectra (500 MHz) of samples A to E (0.005 ppm buckets). Regions 3.25–3.32 ppm and 4.8–5.0 ppm were excluded; data were normalized to total intensity of the spectrum. (a) ‘Williams’ variety (Pareto scaling was applied with r² = 0.903 of PCA model). (b) ‘Khai Thong Ruang’ (‘KTR’) variety (unit variance scaling was applied with r² = 0.854 of PCA model). (c) Sketch of the sampling procedure: A, black leaf streak disease (BLSD)‐infected necrotic leaf zones; B, healthy zones from the treated half of a leaf; C, control Musa plant; D, non‐treated half of a treated leaf; and E, non‐treated leaf from a treated plant.

With the intention of detecting the signals in the spectra that contribute to the differences between samples, 1D and 2D NMR analysis was carried out along with comparisons with reference compounds and previously reported data (Franzyk et al. 2004; Huang et al. 2004; Kashiwada et al. 1988; Kim et al. 2010; Liang et al. 2006; Lu & Foo 2003; Najbjerg et al. 2011; Ramsay et al. 2014; Shimomura et al. 1987; Strack et al. 1989). This allowed the identification of the metabolites summarized in Table S2 (for chemical structures, see Fig. S1). Taking the information from the PCA, including consideration of the metabolites identified, the corresponding loading plots were constructed (Fig. 3). For both Musa varieties, the red‐orange color in the scale of the loading plots demonstrated that the phenolic metabolites, which are up‐regulated in infected tissue, permit the infected tissue to be discriminated from the non‐infected plant sections (Fig. 3). Interestingly, the level of the 1‐O‐((E)‐p‐coumaroyl)‐β‐d‐glucose was lower in infected tissue than in healthy tissue, perhaps because of consumption in specialized metabolism, which was triggered to produce phenylphenalenones (Table S2) such as irenolone and hydroxyanigorufone, already reported as phytoalexins of Musa plants (Luis et al. 1993, 1994, 1995). Further, among the identified metabolites, dopamine, a well‐known catecholamine in banana fruits (Kanazawa & Sakakibara 2000), and 2‐(3,4‐dihydroxyphenyl)ethyl β‐d‐glucopyranoside (common name dopaol β‐d‐glucoside), which is reported here for a first time as a metabolite from Musa plants, allowed the treated tissue to be discriminated from the untreated Musa tissue (for relevant resonances in the ¹H NMR spectrum, see the region between 6.5 and 6.9 ppm in Fig. S2a,b).

Figure 3

Loading plots for principal component analysis (PCA) model shown in Fig. 2a,b displaying the regions of the metabolites with strong influence (in red and orange) and the metabolites with less influence (light blue and dark blue) during the group discrimination observed in the PCA analysis. (a) ‘Williams’ and (b) ‘Khai Thong Ruang’ (‘KTR’) Musa varieties.

A detailed analysis of ¹H NMR spectra of all extracts of ‘Williams’ and ‘KTR’ showed that dopamine was present in infected leaves and in control plants. While control samples contained only traces of dopaol β‐d‐glucoside (Fig. S2c), its level almost equaled that of dopamine in the necrotic tissue of treated plants (Fig. S2a). In non‐infected zones from (B), (D) and (E) samples of both varieties, the level of the glucoside clearly exceeded the level of dopamine, which was detectable only in very low levels (Fig. S2b). This finding confirms the significance of dopaol β‐d‐glucoside in the differences observed by the PCA analysis.

According to the accumulation of dopaol β‐d‐glucoside in the different leaf areas of Musa plants treated with conidia of M. fijiensis but not in the control plant, this compound could be considered a phytoalexin; phytoalexins are produced in those areas of a plant where the pathogen has not been yet spread but which might already have received electrophysiological or chemical signals. Therefore, the antifungal property of dopaol β‐d‐glucoside was assayed against M. fijiensis in an in vitro experiment. However, antimicrobial inhibition was not observed, and the weak activity was comparable with that obtained for dopamine in the same assay (Fig. S3), ruling out the function as a phytoalexin. Thus, the biological function of this compound in the Musa–M. fijiensis pathosystem remains to be investigated, including its potential role in signalling.

In addition, the roots of infected and control plants were evaluated in both Musa varieties in order to check if specialized metabolites were observed in the roots as they were in the leaves. However, no differences between control and treated plants were found (data not shown), supporting the hypothesis that after Musa leaves are treated with the conidia of M. fijiensis, chemical defences are activated mainly in the above‐ground plant parts, especially at the infection site.

Identification and quantification of induced defence metabolites in Musa after infection with M. fijiensis

Quantification of phenylphenalenone‐type compounds based on the progress of the disease

In order to study the hypothetical correlation between the virulence of M. fijiensis to Musa plants and the formation of phenylphenalenones, in vitro plants of ‘Williams’ and ‘KTR’ were treated in independent experiments with the fungal strains, E22 and Ca10_13, after which the symptoms were evaluated and the leaf tissues quantitatively analysed for the presence of phenylphenalenones. The two M. fijiensis strains were chosen according to their different tolerance (Ca10_13) or susceptibility (E22) to the fungicide propiconazole, and their conidia were used as infective propagules for exploring differential responses of susceptible and resistant Musa varieties. The virulence of the two strains was evaluated by visually inspecting the number and area of the lesions on the surface of the Musa leaves at different time points after infection (Fig. 4). At early stages of infection, the resistant variety ‘KTR’ exhibited some symptoms characterized by small red specks (~1 mm in diameter) on the lower surface of the leaf in response to each M. fijiensis strain, E22 and Ca10_13; these symptoms were stronger compared with symptoms observed in the susceptible ‘Williams’ variety (Table S1). This was evident by symptom‐free leaves in ‘Williams’ at 8 dpi. The further development of the BLSD symptoms is clearly different for the infection of the two Musa varieties treated with the strain E22 or Ca10_13. While the number and area of the lesions constantly increased in ‘Williams’, these symptoms remained almost unaltered on the leaves of ‘KTR’ after infection with E22 but increased after infection with Ca10_13, reaching severe damage after 50 dpi (Fig. 4, Table S1).

Figure 4

Number and area of black leaf streak disease (BLSD) symptoms determined for Musa ‘Williams’ (a,c) and ‘Khai Thong Ruang’ (‘KTR’) (b,d) varieties during the interaction with M. fijiensis strains E22 and Ca10_13. Letters a–d in the box plots indicate significant differences among the treatments. One‐way analysis of variance (anova), P < 0.001 = ***, P < 0.05 = *.

After evaluating the BLSD symptoms, an HPLC–DAD approach was applied to quantify the inducible phenylphenalenone‐type compounds produced by the two Musa varieties ‘Williams’ and ‘KTR’ during their response to the pathogen. The concentration of phenylphenalenones was determined either in an early stage (25 dpi) or in a late stage (50 dpi) of the disease. The results for the ‘KTR’ variety (Fig. 5a) showed irenolone (4) as the major metabolite, with a concentration exceeding 25 nmol mg⁻¹ DW. Metabolites 2, 4, 7, 9, 11 and 12 were produced in a concentration of >10 nmol mg⁻¹ DW. All other compounds were produced in a concentration range between 1.5 and 9.5 nmol mg⁻¹ DW.

Figure 5

Quantification of the phenylphenalenones [nmol phenylphenalenone (PP) mg⁻¹ dry weight (DW)] identified in ‘Khai Thong Ruang’ (‘KTR’) (a) and ‘Williams’ (b) varieties after infection with M. fijiensis strain E22 at 25 and 50 d post inoculation (dpi) and with M. fijiensis strain Ca10_13 at 25 and 50 dpi. Letters a–d indicate significant differences among treatments within the specified time (dpi) and compound [three‐way analysis of variance (anova), Holm–Sidak post hoc test: P < 0.05; n.s: not significant], n.d: not detectable. For chemical structures of compounds 1–15, see Table 1.

A significant increase in the concentration of compounds 6, 10 and 13–15 (P < 0.001) during the observed period of time from 25 to 50 dpi was found when the ‘KTR’ plants were treated with M. fijiensis strain E22. In contrast, compounds 5 (hydroxyanigorufone) and 9 (anigorufone) were significantly reduced in concentration between 25 and 50 dpi (P < 0.001) (Fig. 5a). When the plants were treated with M. fijiensis strain Ca10_13, mainly compounds 4 and 14 but also 8, 10 and 15 were significantly induced at 50 dpi, while metabolite 9 again was found to be significantly reduced (P < 0.001).

The Musa ‘KTR’ variety reacted differently to the infection by each fungal strain. Overall, a significant increase in the production of compounds 2, 3, 4, 12 and 14 were observed during the response against the strain Ca10_13, while compounds 6, 7, 9 and 11 were more induced by the infection with the strain E22 when compared with Ca10_13 (P < 0.001).

Like ‘KTR’, the variety ‘Williams’ also exhibited one dominant metabolite in addition to a number of medium‐concentrated and minor compounds. However, unlike ‘KTR’, which mainly produced irenolone (4), its positional isomer, hydroxyanigorufone (5) was the major compound in ‘Williams’ (25–52 nmol mg^–1 DW) (Fig. 5b). Only two further compounds, 4 and 13, also exceeded the concentration of 10 nmol mg⁻¹ DW, while compounds 1–3, 8, 9, 12 and 15 remained below that level and were produced in the range of 1.3–8.5 nmol mg⁻¹ DW.

According to the statistical analysis, a significant increase in the time‐dependent induction of metabolites 2 and 3 were observed at 50 dpi, independently of the fungal strain used for the infection. Compounds 12 and 13 followed an induction pattern similar to that of 2 and 3 when the plant was treated with M. fijiensis strains E22 and Ca10_13, respectively. In contrast, the concentration of compounds 4, 5 and 9 was dramatically reduced during the time BLSD developed, when the strain Ca10_13 was used (P < 0.001). A remarkable drop of the concentration of hydroxyanigorufone (5) from ~52 to ~27 nmol mg⁻¹ DW was observed between 25 and 50 dpi. Similar results were found for compounds 5 and 9 after infection with strain E22 (Fig. 5b). The de novo biosynthesis of metabolites 1, 8 and 15 remained almost unaffected both by the time point evaluated and by the fungal strain.

An overall overview of the total content of phenylphenalenones (and structural analogues) by the two Musa varieties was obtained by adding up the amount of each metabolite per milligram of plant dry material. Thus, both after inoculation with the strain E22 and the strain Ca10_13, the susceptible variety ‘Williams’ produced lower levels of phenylphenalenones compared with the resistant variety ‘KTR’ (Fig. S4). For treatment with E22, ~68 nmol mg⁻¹ DW was found in ‘Williams’, and ~120 nmol mg⁻¹ DW was found in ‘KTR’. No significant statistical differences in the total phenylphenalenone content were detected between plants at 25 dpi showing early BLSD symptoms and plants at 50 dpi showing stronger symptoms. However, large differences in total phenylphenalenone concentration between the early (25 dpi) and the late stage of the disease (50 dpi) were observed for both Musa varieties when infected with the M. fijiensis strain Ca10_13. ‘Williams’ and ‘KTR’ responded differently as the disease progressed. While the phenylphenalenone content in ‘Williams’ dropped significantly from ~100 nmol mg⁻¹ DW at 25 dpi to ~70 nmol mg⁻¹ DW at 50 dpi, the phenylphenalenone content in ‘KTR’ increased from ~120 to ~140 nmol mg⁻¹ DW (Fig. S4).

Antifungal bioassay of the major phenylphenalenones

The major phenylphenalenone‐type compounds identified in both Musa plants were assayed for their antifungal properties against the two M. fijiensis strains E22 and Ca10_13 used in this study. The IC₅₀ values were generated using the method reported by Peláez et al. (2006) with modifications as described in the Section on Materials and Methods and using propiconazole as a positive control. The results showed that both fungal strains were sensitive not only to the commercial fungicide propiconazole but, though to a different extent, also to most of the phenylphenalenone‐type compounds assessed. Among compounds 3–5, 7, 9–11, 14 and 15 tested, only hydroxyanigorootin (14) and 2‐phenyl‐1,8‐naphthalic anhydride (3) did not display significant inhibitory activity against both M. fijiensis strains. Most of the IC₅₀ values were found to be in the range between ~0.05 and ~0.15 mm (Fig. 6). Isoanigorufone (11), a 4‐phenylphenalenone found in the resistant variety ‘KTR’ but not in susceptible ‘Williams’, displayed antimycotic activity in IC_50‐E22 = 0.032 ± 0.002 mm under light conditions. This value represented the best antifungal bioactivity among all the compounds tested. The antimicrobial activity of irenolone (4), hydroxyanigorufone (5) and 4′‐methoxyirenolone (12) against both M. fijiensis strains also increased significantly when the experiments were carried out under light‐controlled conditions (P < 0.05). The antifungal activity of anigorufone (9) and methoxyanigorufone (7) was also significantly light‐dependent but only in response to M. fijiensis Ca10_13 (P < 0.05) (Fig. 6). The results also showed that M. fijiensis Ca10_13 was not only tolerant to propiconazole but also to the phenylphenalenones assessed in this study. This became especially evident by comparing the growth of the strain E22 treated with anigorufone (9), irenolone (4) (under darkness only, P < 0.05) and methoxyanigorufone (7) (under darkness and photoperiod treatments, P < 0.001) with the growth of the strain Ca10_13 under the same conditions (Fig. 6).

Figure 6

Half maximal inhibitory concentration (IC₅₀) of the major inducible metabolites found in both Musa varieties during infection with M. fijiensis. Letters a–c indicate significant differences among treatments within the specified compound [two‐way analysis of variance (anova), Holm–Sidak post hoc test: P < 0.001; n.s: not significant]. For chemical structures of compounds 3–5, 7, 9–12, 14 and 15, see Table 1.

Metabolism of phenylphenalenones by the fungus

Quantification of phenylphenalenones in M. fijiensis cultures

The question of whether phenylphenalenone‐type compounds are taken up by the mycelial tissue or persist in the medium during incubation with the fungus was assessed by HPLC. Fungal cultures of strain Ca10_13 were incubated for 8 d with three different concentrations (0.01, 0.02 and 0.04 mg mL⁻¹) of the corresponding phenylphenalenones, as described in the preceding text. Each phenylphenalenone was then quantitatively determined in the medium as well as in the mycelium. The results showed a very different distribution of the phenylphenalenones between the medium, the mycelium and a non‐recovered portion of the compounds (Fig. 7). A considerable portion of most compounds persisted in the medium. For compounds 9–12, this portion was between 27 and 45% at the lowest concentration tested and increased up to approximately 70 to 80% at the highest concentration. Simultaneously, a certain percentage of compounds 9–12 accumulated in the mycelium biomass and showed a tendency to decline at higher doses of the applied compounds. According to the data shown in Fig. 7, the portions of compounds 9–12 detected in the medium and in the fungal mycelium do not sum up to 100%: an essential portion was missing, exceeding 40% in some cases. Again, the non‐recovered portion tends to decline with increasing doses of the applied compounds. However, generalizing with regard to the data is problematic, and an individualized consideration of the distribution pattern of most compounds seems to be reasonable, at least in the case of hydroxyanigorufone (5) and compounds 3 and 7. Remarkably, compound 5 was no longer detectable after incubating a dose of 0.01 mg mL⁻¹ with M. fijiensis strain Ca10_13, and even more than 70% of the highest applied dose of 0.04 mg mL⁻¹ remained unrecovered. Unlike all other phenylphenalenones applied, methoxyanigorufone (7) and 2‐phenyl‐1,8‐naphthalic anhydride (3) were taken up by the mycelium only to a relatively small (~10% for 7) or even negligible extent (3); that is, they remained mostly unchanged in the medium. For compound 3, this finding coincided with its inactivity on the biomass production of M. fijiensis strain Ca10_13 as shown in Fig. S6.

Figure 7

Quantification of phenylphenalenones after incubation of M. fijiensis strain Ca10_13 with different doses (0.01, 0.02 and 0.04 mg mL⁻¹) of the specified compound. Data were normalized as a percentage based on the peak area generated by high‐performance liquid chromatography (HPLC)–diode array detector (DAD) analysis of the negative controls of each compound. Phenylphenalenone recovery from culture medium (dark‐grey bars), mycelium (light‐grey bars) and the unrecovered portions of phenylphenalenones were determined as the difference between the applied dose and the concentration determined in the medium and the mycelium after 8 d of incubation (white bars). (a) Isoanigorufone (11); (b) anigorufone (9); (c) hydroxyanigorufone (5); (d) 4′‐O‐methylirenolone (12); (e) 4′‐O‐methylanigorufone (10); (f) methoxyanigorufone (7); and (g) 2‐phenyl‐1,8‐naphthalic anhydride (3). Different letters or asterisks indicate significant differences among treatments (concentrations) of the compound analysed [one‐way analysis of variance (anova), Holm–Sidak post hoc test: P < 0.05; n.s: not significant].

Monitoring Musa's metabolic responses during M. fijiensis attack

¹H NMR‐based metabolomics revealed similar biochemical profiles for both Musa varieties, ‘Williams’ (susceptible) and ‘KTR’ (resistant), in response to their interaction with the fungus M. fijiensis. Indeed, infection by this pathogen not only triggered the primary metabolism but also strongly affected the biosynthesis of specialized metabolites, especially aromatic compounds. These effects of infection were inferred from evaluating the group discrimination of each Musa plant part after the PCA (Fig. 2) and the corresponding loading plots (Fig. 3). While the levels of carbohydrates, represented mainly by glucose and sucrose, rose in the infected plant tissue, the levels of most amino acids (such as l‐alanine and l‐threonine) declined. l‐Methionine was an exception and tended to be more up‐regulated in the infected Musa plants than in the non‐treated plants. These findings confirm previous reports of plant–pathogen interactions that suggest that the translocation of sugars from the roots or other plant organs to the infected tissue may help counteract the metabolic imbalances caused by microbial invaders (Heil & Bostock 2002).

Carbohydrates are also involved in the biosynthesis of a broad spectrum of specialized metabolites and can mediate the gene expression of pathogenesis‐related proteins (Aliferis et al. 2014; Herbers et al. 1996; Herbers et al. 2000; Scharte et al. 2005). Gene expression demands a substantial pool of amino acids for the synthesis of proteins commonly associated with the defence against pathogens. For example, chitinases and glucanases have been reported to play a role in the interaction between Musa and M. fijiensis and in other plant pathosystems (Harish et al. 2009; Torres et al. 2012). The production of defence‐related proteins could explain the reduced content of amino acids found in the infected plant tissue of both Musa varieties. Surprisingly, l‐methionine, which was the only amino acid positively regulated during the infection, may be associated with enzymatic reactions that make up the methionine S‐methyltransferases involved in the O‐methylation of some phenylpropanoids and phenylphenalenone‐type compounds, as has been reported by Otálvaro et al. (2010).

An enhanced number of signals in the ¹H NMR region from δ 6.0 to 9.0 were the most relevant features observed in the loading plots (Fig. 3) of both Musa varieties. Phenolic metabolites were more strongly up‐regulated in the infected local tissue of the resistant variety ‘KTR’ than in the infected local tissue of the susceptible variety ‘Williams’, thus providing the first hints that the two varieties respond differently to the infection with the pathogen.

Dopamine, an animal neurotransmitter also occurring in plants (Kulma & Szopa 2007; Li et al. 2015), is a catecholamine identified previously in banana fruits (Kanazawa & Sakakibara 2000). In our experiments, it was found in Musa leaves. We observed that it declined in the non‐treated areas of infected plants (Fig. 2, samples B and D). Simultaneously, a positive correlation was observed for dopaol‐β‐d‐glucoside, identified for the first time in healthy Musa plant tissue. Dopamine was thought to serve as a substrate in the biosynthesis of its glucoside, which then played a role as a phytoalexin. However, this assumption was ruled out because dopaol‐β‐d‐glucoside (and dopamine as well) did not show any antifungal properties against M. fijiensis (Fig. S3). Thus, the role of those metabolites in Musa plant defence remains to be studied.

A group of phenylpropanoid conjugates was identified for the first time as constituents of Musa plants (Table S2, entries 14–17). 1‐O‐((E)‐p‐Coumaroyl)‐β‐d‐glucose may function as a depot compound, which, after being hydrolyzed to p‐coumaric acid by a putative cell wall‐associated β‐glucosidase (Chong et al. 1999; Chong et al. 2002), could be used as a precursor for the biosynthesis of phenylphenalenone‐type compounds (Hölscher & Schneider 1995; Schmitt et al. 2000). This phenomenon could explain the reduced levels of such phenylpropanoids in the infected tissue. Moreover, phenylpropanoids are not only involved in the biosynthesis of phytoalexins of the phenylphenalenone‐type but also play a role in lignification processes in order to counteract the invading pathogen (Liang et al. 2006).

Irenolone (4) and hydroxyanigorufone (5), the two major phenylphenalenones identified from the metabolomic analysis (Table S2), were induced only in the infected plant tissue; along with other phenolic compounds, these major phenylphenalenones constituted the chemical basis for the Musa response to M. fijiensis. Unfortunately, because of the overlapping ¹H NMR signals, metabolic changes of the other phenolics could be not observed in the loading plots (Fig. 3) but, rather, required a conventional phytochemical approach.

Incompatible Musa–M. fijiensis interaction depends on the pathogen virulence

Phytochemical analysis of the infected Musa tissue led to the identification of 15 phenylphenalenone‐type compounds and structural analogues from the resistant variety ‘KTR’. A smaller number (10) of such metabolites appeared in the defence response of the susceptible variety ‘Williams’. With the exception of 2‐hydroxy‐9‐(4′‐methoxyphenyl)‐1H‐phenalen‐1‐one (10), all other compounds had been previously reported in different varieties of Musa, and their occurrence has been correlated with the resistance phenotype exhibited by some Musa varieties against pathogens and herbivores (Hölscher et al. 2014; Otálvaro et al. 2007). Nonetheless, there has been little quantitative analysis of phenylphenalenones as BLSD progresses, and resistant traits have not been studied with isolates of M. fijiensis that differ in their virulence. Therefore, two strains of M. fijiensis, E22 and Ca10_13, were used as inoculum. According to the method proposed by Cañas‐Gutierrez et al. (2009), the strain E22 was classified as ‘sensitive’ and Ca10_13 as ‘tolerant’ to the fungicide propiconazole. It is important to note that sensitivity to the fungicide propiconazole does not necessarily correlate with a high degree of virulence of the pathogen. However, fungicide tolerance could help the microorganism overcome the plant's chemical defence and/or synthetic fungicides.

For the variety ‘KTR’, an early response was observed by the appearance of small specks on the lower surface of the infected leaf 7–9 dpi with the fungus, regardless of which M. fijiensis strain was used; no symptoms were detectable in the variety ‘Williams’ up to 15–17 dpi (Table S1). This difference between varieties is not surprising because an early recognition of virulence effectors from the pathogen triggers the defence mechanism in Musa (as they do in other plant–pathogen systems), conferring resistance on plants that are attacked (Hoss et al. 2000; Torres et al. 2012).

The resistance conferred on Musa ‘KTR’ in its interaction with M. fijiensis strain E22 was confirmed by the slow progress of BLSD determined at the two points of time. In contrast, M. fijiensis Ca10_13 not only exhibited a higher degree of virulence than E22 but also was able to overcome the resistance of ‘KTR’ and cause severe foliar damages in the infected leaf (Fig. 4, Table S1). According to previous reports, the breakdown of the resistance in Musa caused by M. fijiensis has also been observed in other Musa varieties such as the variety ‘Yangambi Km5’, which is classified as highly resistant to this particular pathogen (Fullerton & Olsen 1995; Mouliom‐Pefoura 1999).

For the variety ‘Williams’, the necrotic symptoms of BLSD were larger than those observed for ‘KTR’ during the progress of the fungal infection; this difference was expected for a susceptible variety. Nonetheless, leaf damage was increased when M. fijiensis Ca10_13 was used as a pathogen, confirming the strong virulence of this pathogen.

Priming defence mechanisms in plants at early stages of infection, after the recognition of the effectors signals produced by a pathogen, are crucial in the incompatible plant–pathogen interactions (Conrath et al. 2006; Frost et al. 2008). As demonstrated for some resistant Musa varieties, the production of pathogenesis‐related proteins, lignification processes, hypersensitivity response and accumulation of phytoalexins must work in a concerted action to enable plants to block pathogenesis and herbivory (Harelimana et al. 1997; Hölscher et al. 2014; Otálvaro et al.2002a, 2002b; Quiñones et al. 2000; Torres et al. 2012).

Here, the BLSD symptoms appearing on the leaves of the ‘KTR’ variety after infection with the strain Ca10_13 clearly demonstrated that resistance in this Musa variety could be overcome by a highly virulent pathogen. Analysis of the chemical composition of both Musa varieties during their responses to the two fungal strains showed that phenylphenalenone‐type compounds constituted the major induced metabolites both in ‘Williams’ and in ‘KTR’. The susceptible variety ‘Williams’ produced not only a smaller number but also lower levels of these metabolites (less than 10 nmol mg⁻¹ DW) compared with ‘KTR’. Hydroxyanigorufone (5) appeared to be the exception: it was the most prominent compound synthesized in ‘Williams’ in response to the strain Ca10_13. Nevertheless, as the disease developed, the concentration of hydroxyanigorufone (5) and anigorufone (9) decreased, while that of other compounds (e.g. 1–4) stayed constant or even slightly increased. The reduced levels of compounds 5 and 9 at 50 dpi could be due to an oxidative metabolic reaction and may result in the naphthalic anhydrides 1 and 3, respectively. Reactive oxygen species like hydrogen peroxide, which have been reported to be produced in the infected tissue of Musa plants (Torres et al. 2012), could function as the oxidizing agent. In ‘Williams’, the susceptibility seemed to be associated with a late recognition of the pathogen, which then successfully colonizes the plant tissue. This is consistent with a relatively low total level of phenylphenalenones (Fig. S4), although hydroxyanigorufone (5), which is one of the active phenylphenalenones (Fig. 6), was overproduced in comparison with the other metabolites (Fig. 5b).

In contrast, the ‘KTR’ variety not only rapidly recognized the pathogen but also responded with a burst of phenylphenalenones, providing an enhanced chemical response to both fungal strains. This variety produced a number of phenylphenalenones in concentrations over 10 nmol mg⁻¹ DW; irenolone (4) was the most abundant compound (>25 nmol mg⁻¹ DW) (Fig. 5a). Compound 4 and its congeners seemed to have a strong inhibiting effect on the fungal colonization of plant tissue infected by the strain E22 but not on tissue infected by Ca10_13 (Fig. 5a). The response of the two Musa varieties to the two fungal strains E22 and Ca10_13 produced slight quantitative differences but no qualitative differences in the metabolic patterns (data not shown).

In ‘KTR’, the concentrations of compounds 2–4, 12 and 14 increased significantly in response to Ca10_13 at 25 or 50 dpi but not in the response to the strain E22. Irenolone (4) was the metabolite with the highest response against Ca10_13 at 50 dpi in comparison with all other induced metabolites (Fig. 5a). This finding suggests that, at least from the chemical point of view, ‘KTR’ was able to distinguish between the virulence capacities of the fungal strains.

Pathogen virulence associated with metabolism of phenylphenalenones

The breakdown in the resistance of ‘KTR’ caused by the virulent strain Ca10_13 encouraged us to explore whether some phenylphenalenones (e.g. compounds 3, 9, 11, 12), although up‐regulated in response to M. fijiensis, did not significantly inhibit the growth of this particular fungus in planta. The antifungal bioassay showed that phenylphenalenones were active against both fungal strains (Fig. 6). The fungal pathogen E22 was more sensitive in comparison with Ca10_13 when irenolone (4), anigorufone (9), isoanigorufone (11) and methoxyanigorufone (7) were assessed under dark conditions. Furthermore, these compounds displayed an enhanced antifungal activity when light conditions were applied to the experimental system (Fig. 6). Phenylphenalenones (Flors & Nonell 2006; Lazzaro et al. 2004) and perinaphthenones (Hidalgo et al. 2009) have been reported to be light‐dependent singlet oxygen producers, and these could significantly affect the fungal growth rate. Overall, the hypothesis that M. fijiensis strain Ca10_13 could tolerate higher concentrations of phenylphenalenones than the strain E22 was found to hold only for some compounds and under darkness.

Differences in the fungal growth rate between both M. fijiensis strains became more apparent when the incubation experiments were scaled up to almost 10 times the volume of liquid culture used for the microtitre bioassay. The growth rate of M. fijiensis strain E22 was much slower than that of strain Ca10_13 when phenylphenalenones in the range of 0.01 to 0.04 mg mL⁻¹ were supplied in the incubation medium; hydroxyanigorufone (5) and 2‐phenyl‐1,8‐naphthalic anhydride (3) were the exceptions (Fig. S5). Therefore, the strain Ca10_13 was selected for further experiments. Because some plant pathogens have developed strategies to evade plant defences through a complex detoxification process (Barz & Welle 1992; Jeandet et al. 2010; Pedras & Taylor 1993), we wondered whether the resistance of M. fijiensis to some phenylphenalenones could be attributed to metabolic detoxification.

In fact, the strain Ca10_13 was able not only to metabolize most of the compounds into sulfate‐derivatives (compounds 5 and 9–12; Table 2) but also to further degrade them into undetectable metabolites or even decompose them completely as in the case of hydroxyanigorufone (5) at the lowest concentration assessed. In general, the formation of sulfate conjugates through sulfotransferases constitutes a biochemical reaction, which is involved in such physiological processes as intracellular signalling, extracellular interactions and detoxification of xenobiotics (Chapman et al. 2004; Zhang et al. 1996). The present work is the first report of an inactivation or detoxification of phenylphenalenones by M. fijiensis through sulfate conjugation. In addition, to the best of our knowledge, ours is the first report of a detoxification reaction of any phenylphenalenone by a fungal organism. New strategies to control BLSD through inhibitors that target sulfotransferases in the microorganism may result from this new understanding of the role of sulfate conjugates.

Interestingly, only phenylphenalenones containing a hydroxyl group attached to C‐2 in the molecule were found to be sulfated (for hydroxyanigorufone, a sulfation in the free hydroxyl group at C‐4′ cannot be ruled out). Because conversion to sulfates seems to minimize the toxic effects of most phenylphenalenones, the fungal biomass production remained almost unaffected (Fig. S6). However, methoxyanigorufone (7), a metabolite which is missing a hydroxyl group (instead of the OH, an O‐methyl group is attached to C‐2), enormously affected growth and biomass production (Fig. S6). This supports the hypothesis that an α‐hydroxyketone feature helps the molecule to be recognized by the aryl sulfotransferase. In addition, the chemical pressure generated by compound (7) on the metabolism of the strain Ca10_13 was confirmed by the overproduction of soluble extracellular proteins (Fig. S7) and the inhibited production of ergosterol (Fig. S8). Because no metabolites derived from phenylphenalenones were found in the liquid culture medium during the incubation with the fungus, the overexpression of extracellular proteins observed in response to compounds 7 and 11 could not be associated with the external enzymatic detoxification of these metabolites. Instead, detoxification may suggest that these metabolites increase the physiological stress in the fungus, which then results in the overexpression of fungal proteins involved in the suppression of the plant defences (Dodds & Rathjen 2010; Escobar et al. 2015). Identification of the proteins, which were overexpressed by the interaction with specific phenylphenalenones, would help explain the pathophysiological processes taking place during the interaction between Musa and M. fijiensis. Because of their role in the defence of Musa plants against M. fijiensis, methoxyanigorufone (7) and isoanigorufone (11) are considered the most relevant metabolites for further exploration of the molecular mode of phenylphenalenones.