Aliphatic glucosinolates function in the chemical defense of Capparales. The cytochrome P450 83A1 monooxygenase (CYP83A1) catalyzes the initial conversion of methionine-derived aldoximes to thiohydroximates in the biosynthesis of glucosinolates, and thus cyp83a1 mutants have reduced levels of aliphatic glucosinolates.
Loss of CYP83A1 function leads to dramatically reduced parasitic growth of the biotrophic powdery mildew fungus Erysiphe cruciferarum on Arabidopsis thaliana. The cyp83a1 mutants support less well the germination and appressorium formation of E. cruciferarum on the leaf surface and post-penetration conidiophore formation by the fungus. By contrast, a myb28-1 myb29-1 double mutant, which totally lacks aliphatic glucosinolates, shows a wild-type level of susceptibility to E. cruciferarum.
The cyp83a1 mutants also lack very-long-chain aldehydes on their leaf surface. Such aldehydes support appressorium formation by E. cruciferarum in vitro. In addition, when chemically complemented with the C26 aldehyde n-hexacosanal, cyp83a1 mutants can again support appressorium formation. The mutants further accumulate 5-methylthiopentanaldoxime, the potentially toxic substrate of CYP83A1.
Loss of powdery mildew susceptibility by cyp83a1 may be explained by a reduced supply of the fungus with inductive signals from the host and an accumulation of potentially fungitoxic metabolites.
The compatibility of plants and microbial pathogens leads to disease and symptom development. However, compatibility is rare in nature, whereas resistance or incompatibility is common. This is explained by several aspects of host–parasite interactions. Compatibility requires the adaptation of a pathogenic microbe to its host (Morrissey & Osbourn, 1999; Thordal-Christensen, 2003). An adapted pathogen can recognize specific host-derived signals that trigger pathogen differentiation and expression of virulence, or both. Therefore, pathogens must adapt to the chemical composition of their hosts. This leads to the evolution of an enzymatic toolbox, which the pathogen uses to overcome structural and chemical barriers of its hosts or to metabolize host-derived substrates. In addition, plants possess robust innate immunity that involves defense responses triggered after the recognition of pathogen-derived elicitors (microbe-associated molecular patterns or pathogen effectors) or host-derived elicitors (damage-associated molecular patterns) (Boller & Felix, 2009; Maekawa et al., 2011). Consequently, pathogens require host-specific effector molecules that match host targets for the suppression of immunity and reprogramming of the host for the demands of the pathogen. Loss of susceptibility may therefore result from altered host immunity (gain of resistance functions) or, in a stricter sense, from changes in host components that are required by the microbe for pathogenesis, but do not directly operate in the regulation of defense (de Almeida Engler et al., 2005; Hückelhoven, 2005; Pavan et al., 2010; Hückelhoven et al., 2013).
In contrast with wild-type Arabidopsis plants, loss-of-function mutants of the cytochrome P450 monooxygenase CYP83A1 gene are barely susceptible to the biotrophic ascomycete Erysiphe cruciferarum (Weis et al., 2013), which causes powdery mildew on many Brassicaceae (Adam et al., 1999). The cytochrome P450 gene family in Arabidopsis comprises 244 genes (and 28 pseudogenes) and constitutes one of the largest gene families in plants. P450 enzymes function as monooxygenases in the biosynthesis of diverse metabolites, including pigments, phytohormones and lignin, or defense compounds, such as flavonoids, alkaloids or glucosinolates (reviewed in Bak et al., 2011). The synthesis of aliphatic glucosinolates is divided into three stages: the chain elongation of the amino acid, formation of the glucosinolate core structure and, finally, secondary modifications (Wittstock & Halkier, 2002). The two CYP83 proteins, CYP83A1 and CYP83B1, phylogenetically belong to the CYP71 clade (Hansen et al., 2001; Bak et al., 2011). They non-redundantly function in the core structure synthesis of glucosinolates by catalyzing the initial conversion of aldoximes to thiohydroximates (Bak & Feyereisen, 2001; Naur et al., 2003). CYP83A1 has higher substrate specificity for methionine-derived aldoximes in the synthesis of aliphatic glucosinolates, whereas CYP83B1 preferentially converts tryptophan-derived aldoximes in the synthesis of indole-glucosinolates (Bak & Feyereisen, 2001; Naur et al., 2003).
In plant–herbivore interactions, glucosinolates and their hydrolysis products, such as isothiocyanate, nitriles or epithionitriles, function as deterrents against generalist herbivores, but also as attractants for specialized insects. The knowledge about the function of glucosinolates in plant–fungus interactions is more limited (Bednarek & Osbourn, 2009). However, in Arabidopsis, the peroxisome-associated myrosinase PENETRATION 2 (PEN2) hydrolyzes 4-methoxyindol-3-ylmethylglucosinolate to bioactive products involved in resistance to non-adapted powdery mildew fungi (Lipka et al., 2005; Bednarek et al., 2009). Indole-glucosinolates are further important to balance the mutualistic interaction of Arabidopsis with the beneficial root endophyte Piriformospora indica (Jacobs et al., 2011; Nongbri et al., 2012). Aliphatic glucosinolates are important in resistance to lepidopteran larvae, to non-adapted bacterial pathogens and to the necrotrophic fungus Sclerotinia sclerotiorum (Beekwilder et al., 2008; Fan et al., 2011; Stotz et al., 2011). However, little information is available about the role of aliphatic glucosinolates in interactions with haustorium-forming fungi.
Powdery mildew fungi are obligate biotrophic ascomycete pathogens with a high degree of specialization to a limited range of hosts. They form haustoria from appressoria that sense and directly penetrate the host cuticle and cell wall (Green et al., 2002; Hückelhoven & Panstruga, 2011). Erysiphe cruciferarum is a typical powdery mildew fungus adapted to Brassicaceae (Adam et al., 1999). Hence, it can normally cope with glucosinolates as a result of an unknown mechanism.
In this study, we report an as yet uncharacterized function of CYP83A1 in interactions with fungal pathogens. Loss of CYP83A1 changes the metabolic composition of Arabidopsis in a manner that greatly influences the interaction outcome with biotrophic and necrotrophic ascomycetes.
Materials and Methods
Plants, pathogens and inoculation
Arabidopsis thaliana (L.) Heynh ecotype Col-0 was purchased from Lehle Seeds (Round Rock, TX, USA). cyp83a1-1 (Weis et al., 2013), cyp83a1-2 (= ref2-1, reduced epidermal fluorescence2-1; Hemm et al., 2003) and myb28-1 myb29-1 (Sønderby et al., 2007) have been described previously. Arabidopsis seeds were sown in a soil–sand mixture and stratified for 2 d at 4°C before placement in a growth chamber at 22°C, photoperiod of 10 h and 64% relative humidity.
Erysiphe cruciferarum was maintained on Col-0 and on super-susceptible phytoalexin-deficient pad4 mutants (Reuber et al., 1998). Plants and glass slides were inoculated with a density of 3–5 and 10–15 E. cruciferarum conidia mm−2, respectively, as described previously (Weis et al., 2013).
Botrytis cinerea strain B05.10 was cultivated on Gamborg's B5 plates (Duchefa Biochemie BV, Haarlem, the Netherlands) containing 2% glucose. Based on a protocol of Gronover et al. (2001), conidia were washed from 7–10-d-old plates, and adjusted to a final concentration of 2 × 105 conidia ml–1 using Gamborg's B5 medium containing 2% glucose and 10 mM KH2PO4 (pH 6.4). On pre-germination for 1 h at room temperature, 20 μl of the suspension were dropped onto leaves of c. 5-wk-old Arabidopsis plants. Plants were further cultivated covered with a plastic cap under normal growth conditions. Botrytis cinerea symptoms were evaluated 3 d after inoculation (dai) by calculating the affected leaf area.
Acetic ink staining of epiphytically grown structures of E. cruciferarum was performed as described previously (Weis et al., 2013). Wheatgerm agglutinin–tetramethylrhodamine (WGA-TMR; Invitrogen Molecular Probes GmbH, Karlsruhe, Germany) was used to stain fungal structures including haustoria. Discolored leaves were vacuum infiltrated with a staining solution containing 10 μg WGA-TMR and 10 μg BSA per milliliter of PBS, and incubated overnight in the staining solution at 4°C. Samples were observed via fluorescence microscopy (Leica DM 1000, Wetzlar, Germany). Callose staining of WGA-TMR-treated leaves was performed using 0.05% (w/v) methyl blue in 0.067 M potassium phosphate buffer (pH 5.8). Vacuum was applied twice and leaves were stored in the staining solution overnight at 4°C.
In order to assess cuticle permeability, Calcofluor white staining of leaves of c. 5-wk-old plants was performed as described by Bessire et al. (2007). The incubation time was set to 30 s. Calcofluor white was detected under a Zeiss Axioimager Z1 (Zeiss, Jena, Germany) fluorescence microscope using a 368-nm excitation filter, a 385-nm dichromatic mirror and a 420-nm suppression filter.
Seven-week-old single rosette leaves were collected and immersed in 6 ml of 80% ethanol in plastic tubes and then gently agitated on a shaker platform in the dark at 25°C. The amount of chlorophyll extracted into the solution was quantified using a UV4-500 UV/VIS spectrophotometer (UNICAM) and calculated from the light absorption at 647 and 664 nm, as described by Lolle et al. (1997). For this purpose, 1-ml aliquots were measured every 20 min after initial immersion over a 2-h period. Each aliquot was poured back into the same tube after measurement. Data are given as micromoles of chlorophyll per square centimeter of leaf surface area.
Cuticular wax analysis
Leaf cuticular waxes were extracted by immersing 30 intact, 7-wk-old Arabidopsis rosette leaves for 1 min in 10 ml of chloroform at room temperature. The solvent was evaporated under a gentle flow of nitrogen. Sample derivatization, GC-MS and gas chromatography-flame ionization detection (GC-FID) analyses were performed as described previously (Hansjakob et al., 2010). Five replicate samples per plant line were analyzed. Statistical analysis was carried out using the Mann–Whitney U-test for each wax component.
Coating of glass slides and plants
Glass slides were coated with 0.5% Formvar® resin, together with the C26 alkane n-hexacosane, to a final concentration of 6.8 × 10−4 mol l−1, or, in addition, with the C26 aldehyde n-hexacosanal or the C30 aldehyde n-triacontanal to a final concentration of 6.8 × 10−5 mol l−1 in the dipping solution, as described previously by Hansjakob et al. (2010). Depending on the molecular weight of the molecule, a concentration of 6.8 × 10−5 mol l−1 of n-hexacosanal or n-triacontanal in the dipping solution corresponds to a glass surface coverage of c. 50–60 ng cm–2, which roughly reflects the amount of very-long-chain aldehydes present per square centimeter of Arabidopsis leaf surface area.
For chemical applications, detached leaves were placed on water agar and sprayed with a chromatography sprayer with solutions of n-hexacosane or n-hexacosanal in chloroform (each 5 mg ml−1). Mock-treated plants were sprayed with pure chloroform. After evaporation of the solvent, leaves were inoculated with fungal conidia at a density of four conidia mm−2. At some points, larger droplets of chloroform caused small necrotic lesions on the leaves. These areas were excluded from microscopic evaluation. At 8 h after inoculation (hai), leaves were fixed and stained for microscopy of fungal infection structures (see Staining methods).
Fifty milligrams of frozen plant material were extracted with 2 ml of methanol. Methanol was removed in a Speed-Vac concentrator and the extract was re-dissolved in 100 μl of methanol for analysis by liquid chromatography/electrospray ionization-multistage mass spectrometry (LC-ESI-MSn) according to Yin et al. (2012). Data analysis was performed using the DataAnalysis 3.1 software (Bruker Daltonics, Bremen, Germany). Untargeted metabolite profiling was performed using XCMS (Smith et al., 2006). Synthetic 5-methylthiopentanaldoxime (5-MTPO) (Dawson et al., 1993) served as an external standard.
Germination and penetration processes of E. cruciferarum are delayed on cyp83a1 mutants
Recently, we have reported a strong loss of susceptibility to the powdery mildew fungus E. cruciferarum for two independent cyp83a1 mutants (Weis et al., 2013). However, the mechanistic basis of this phenotype was unclear. We therefore investigated the early development of E. cruciferarum on the leaf surface of cyp83a1 mutants (Fig. 1a,b). Col-0, cyp83a1-1 and cyp83a1-2 plants were inoculated with conidia of the adapted powdery mildew fungus E. cruciferarum and analyzed at 3 and 8 hai. The developmental stage of the conidia was divided into non-germinated conidia, conidia showing appressorial germ tubes (AGT) without a mature appressorium and the formation of a mature lobed appressorium (APP) (Fig. 1c). We observed a significantly enhanced percentage of non-germinated conidia on cyp83a1-1 and cyp83a1-2 when compared with Col-0 at 3 hai. Consistent with this, the percentage of conidia with developing AGT was strongly reduced on both mutants. Only very few conidia had formed an APP at 3 hai. By 8 hai, the majority of conidia had formed an APP on Col-0 (53%), whereas fewer conidia had developed an APP on cyp83a1-1 (22%) and cyp83a1-2 (37%) (Fig. 1a).
Additional microscopic analyses were carried out at 2 dai. In this case, only germinated conidia were analyzed with respect to APP formation without penetration (APP), establishment of a first haustorium (HAU) with formation of elongated hyphae (EH) or the subsequent formation of secondary haustoria (sec HAU) (Fig. 1d). On the wild-type, c. 14% of the conidia had reached a stage at which the appressorium (APP) had formed without successful haustorium formation. However, the majority of fungi (71%) had established a haustorium and elongated hyphae (HAU + EH) by this time; 15% of the fungi had already established a secondary haustorium in the host cells. By contrast, the number of conidia with an APP only was significantly enhanced on cyp83a1-1 and cyp83a1-2 at 45% and 40%, respectively. Consequently, the number of fungi with haustoria and elongated hyphae was significantly reduced to 53% on cyp83a1-1 and 59% on cyp83a1-2. The percentage of conidia which had established secondary haustoria was also significantly reduced on both cyp83a1 mutants. On cyp83a1-1 and cyp83a1-2, only 1% of the germinated E. cruciferarum conidia had established secondary haustoria (Fig. 1b).
Finally, we evaluated local and whole-cell callose deposition as typical plant responses to powdery mildew fungi (Micali et al., 2008; Fig. 1e). No significant differences in local callose deposition were detected in the two cyp83a1 mutants when compared with the wild-type at 2 dai. In Col-0, cyp83a1-1 and cyp83a1-2, callose deposition was visible in c. 90% of the attacked or penetrated epidermal cells (Fig. 1f). Whole-cell callose accumulation, indicative of cell death (Jacobs et al., 2003), was rare at 2 dai. The percentage of interaction sites with cell death was slightly enhanced on cyp83a1-1 and cyp83a1-2 compared with the wild-type (Fig. 1f) but this could not explain the strong resistance phenotype of the mutants.
In summary, the data indicate that the germination and penetration of E. cruciferarum are substantially delayed on both cyp83a1 mutants when compared with the wild-type Col-0. However, the mutant's cellular defense of the fungus appears to be only marginally altered.
Aliphatic glucosinolates are not required for the development of E. cruciferarum
The T-DNA knock-out mutant cyp83a1-1 and the loss-of-function point mutant cyp83a1-2 (= ref2-1) have a strongly reduced content of aliphatic glucosinolates when compared with the wild-type Col-0 (Hemm et al., 2003; Weis et al., 2013). Because chemical defense might alternatively serve as a signal for fungal development when overcome by adapted parasites, we tested whether E. cruciferarum normally infects the myb28-1 myb29-1 double mutant, which lacks aliphatic glucosinolates (Sønderby et al., 2007). Five-week-old Col-0, cyp83a1-1, cyp83a1-2 and myb28-1 myb29-1 plants were inoculated with conidia of E. cruciferarum. As shown previously, both cyp83a1 mutants were only slightly susceptible to powdery mildew when compared with Col-0 (Fig. 2a; Weis et al., 2013). Microscopy confirmed this observation by the evaluation of the number of conidiophores per fungal colony at 5 dai (Fig. 2b). However, the myb28-1 myb29-1 double mutant showed wild-type-like E. cruciferarum colonization, both macroscopically and microscopically (Fig. 2). Hence, aliphatic glucosinolates do not appear to be required for a compatible interaction of Arabidopsis with E. cruciferarum.
As previous results have suggested that cyp83a1 mutants are affected in their ability to contain cell death after treatment with a fungal toxin (Weis et al., 2013), we tested whether CYP83A1 had an impact on the interaction of Arabidopsis with a necrotrophic pathogen. Therefore, we inoculated Col-0, cyp83a1-1, cyp83a1-2 and myb28-1 myb29-1 plants with the necrotrophic fungus B. cinerea. Interestingly, a strongly pronounced symptom development was observed on the myb28-1 myb29-1 mutant (Supporting Information Fig. S1). Both cyp83a1 mutants also showed significantly increased susceptibility to necrotization by B. cinerea when compared with the wild-type Col-0 at 3 dai. However, they were less susceptible to B. cinerea than the myb28-1 myb29-1 mutant (Fig. S1). This suggested a function of aliphatic glucosinolate metabolism in basal resistance to B. cinerea.
cyp83a1 mutants exhibit an altered leaf cuticular wax composition
During the course of the microscopic analysis of the powdery mildew phenotype, we observed a cuticle/wax phenotype on both cyp83a1 mutants. On cyp83a1 mutants, acetic ink not only visualized epiphytically grown fungal structures, but also haustoria and chloroplasts, which were not reached by the dye on wild-type samples (not shown). To confirm increased cuticle permeability, Calcofluor white staining was conducted, as described previously by Bessire et al. (2007). Fluorescence of Calcofluor white occurs after binding to cellulose, and thus indicates penetration of the dye through the cuticle of plants. We observed a strong laminar Calcofluor white fluorescence of cyp83a1-1 and cyp83a1-2 leaves, but not of Col-0, where only marginal fluorescence was detected (Fig. 3a). Leaves of myb28-1 myb29-1 showed wild-type-like Calcofluor fluorescence (Fig. S2). A chlorophyll leaching assay showed faster chlorophyll release from leaves of cyp83a1-1 and cyp83a1-2 than from leaves of Col-0 (Fig. 3b). Together, these results suggested an altered cuticle/wax composition of the cyp83a1 mutants.
To further investigate the cuticular wax composition of cyp83a1 mutants, the surface of Arabidopsis leaves was extracted with chloroform and subjected to GC-FID/GC-MS analysis. This revealed a significantly reduced amount of total wax in cyp83a1-1 and cyp83a1-2 compared with Col-0 (Fig. S3a). Moreover, there was an altered composition of the major single substance classes of the Arabidopsis leaf cuticular wax: alkanoic acids (fatty acids), n-alkanes, primary alkanols, branched alkanols, n-aldehydes and sterols. The data indicate significant differences in the content of branched alkanols and aldehydes in the wax layer of Col-0 compared with both cyp83a1 mutants (Fig. S3). The percentage of branched alkanols was slightly enhanced in cyp83a1-1 and cyp83a1-2, whereas the amount of aldehydes was dramatically reduced in both mutants when compared with the wild-type Col-0. Indeed, no aldehydes were detectable in the leaf cuticular wax of cyp83a1-1 or cyp83a1-2 (Fig. 4).
Furthermore, we measured significant differences with respect to the chain length of the components. For instance, an obvious shift from C32–C36 fatty acids to C24–C28 fatty acids in both cyp83a1 mutants was detected (Fig. S3c). The amount of n-alkanols, in particular C28, C29 and C30n-alkanols, was significantly enhanced in cyp83a1-1 and cyp83a1-2 when compared with Col-0 (Fig. S3d). No obvious differences in n-alkane composition were observed (Fig. S3e). In summary, significant differences were observed in the amount and composition of the leaf cuticular wax in cyp83a1-1 and cyp83a1-2 when compared with the wild-type. Most remarkable, very-long-chain aldehydes were absent from the wax layer of both cyp83a1 mutants.
Very-long-chain aldehydes promote appressoria formation by E. cruciferarum
The cuticular wax composition is a crucial factor for the germination of fungal conidia and subsequent penetration of host barriers (Podila et al., 1993; Kolattukudy et al., 1995; Tsuba et al., 2002; Zabka et al., 2008; Hansjakob et al., 2010, 2012; Inada & Savory, 2011). Very-long-chain aldehydes have been shown to function as signals to initiate germination and appressorium differentiation of the powdery mildew fungus Blumeria graminis f.sp. hordei on its host barley (Tsuba et al., 2002; Zabka et al., 2008; Hansjakob et al., 2010, 2012). As cyp83a1 mutants lacked very-long-chain aldehydes, we tested whether such aldehydes can trigger germination and differentiation of E. cruciferarum. Therefore, we coated glass slides with Formvar® resin and the C26 alkane n-hexacosane, or C26 aldehyde n-hexacosanal or C30 aldehyde n-triacontanal, and inoculated them with conidia of E. cruciferarum. We microscopically analyzed fungal development by 9 hai. The developmental stage of the conidia was divided into non-germinated conidia, conidia showing an immature appressorial germ tube (AGT) or a mature appressorium (APP) (Fig. 5b). On n-hexacosane-coated slides, the majority of conidia of E. cruciferarum (77%) did not germinate; 17% of the conidia formed an appressorial germ tube (AGT) and only 6% established a mature appressorium (APP). By contrast, on n-hexacosanal- and n-triacontanal-coated glass slides, the percentage of non-germinated conidia was significantly reduced to 44% and 54%, respectively. The percentage of conidia which formed only an AGT was also reduced to 9% on n-hexacosanal-coated slides, and significantly reduced to 7% on n-triacontanal-coated slides. Simultaneously, the amount of conidia showing mature APP formation significantly increased to 47% and 40% on glass slides coated with the C26 and C30 aldehydes, respectively (Fig. 5a). To further test whether the lack of very-long-chain aldehydes might be responsible for the delay in fungal development on cyp83a1 mutants, we performed a chemical complementation experiment. Therefore, we sprayed detached leaves of the wild-type and cyp83a1-2 with the aldehyde n-hexacosanal dissolved in chloroform, the corresponding alkane n-hexacosane or chloroform alone (mock treatment). At 8 hai, the fungus had developed mature APP from c. 70% of the conidia in the wild-type control, but only from c. 35% of conidia on cyp83a1-2. The alkane n-hexacosane did not promote germination or APP formation on cyp83a1-2. By contrast, n-hexacosanal complemented the rate of germination and APP formation on cyp83a1-2 to the level of the wild-type control, but did not further promote fungal development on the wild-type (Fig. 6). In summary, very-long-chain aldehydes promote germination and differentiation processes of E. cruciferarum conidia in vitro and on plants.
Accumulation of unidentified metabolites and 5-MPTO in cyp83a1 mutants
To further chemically phenotype the mutants, we conducted analytical LC-MS of methanol extracts from leaves. LC-MS-based metabolomic data were analyzed using XCMS (Smith et al., 2006) to identify differences in the metabolic profiles of Col-0, cyp83a1-1, cyp83a1-2 and myb28-1 myb29-1. According to this, mutations in CYP83A1 had a severe and statistically significant effect on the content of many soluble metabolites, as described previously for phenylpropanoids for cyp83a1-2/ref2-1 (Hemm et al., 2003). Most of the differentially accumulating metabolites were specifically enriched or depleted in cyp83a1 mutants (Fig. S4).
As CYP83A1 catalyzes the initial conversion of methionine-derived aldoximes to S-alkyl-thiohydroximate, we wondered whether aldoximes might accumulate in cyp83a1 mutants, as suggested previously by Hemm et al. (2003). As oximes are toxic to fungi (Møller, 2010), an accumulation of methionine-derived aldoximes might further contribute to the resistance of the cyp83a1 mutants to E. cruciferarum. Hence, we conducted analytical LC-MS measurements using 5-MTPO as external standard. 5-MTPO is the precursor of the major aliphatic glucosinolates, for example, 4-methylsulfinylbutylglucosinolate, in Arabidopsis Col-0 leaves (Haughn et al., 1991; Hansen et al., 2001). Methanol extracts of cyp83a1-1 and cyp83a1-2 leaves showed a peak of the ion trace at m/z 148, identical in retention time and fragmentation pattern to the standard 5-MTPO (Figs 7a, S5). Another even more pronounced peak was detected in both independent cyp83a1 mutants with the same m/z and fragmentation pattern as 5-MTPO, but a different retention time. This peak probably represents an isomer of 5-MTPO (Figs 7b, S5). In samples of either Col-0 or myb28-1 myb29-1, peaks of the 5-MTPO and 5-MTPO isomer were negligible (Fig. 7).
In summary, we detected strong metabolic perturbation and a specific accumulation of 5-MTPO, the substrate of CYP83A1, in cyp83a1 mutants when compared with Col-0.
Loss of disease susceptibility occurs in mutants with constitutive or primed defense responses. Other mutants do not support pathogenesis because they lack components on which the pathogen relies for the accommodation of infection structures (Hückelhoven et al., 2013). Alternatively, mutants can lose susceptibility because of changes in their chemical composition. This may involve a lack of nutrients or cues for correct development of the pathogen, or the accumulation of antibiotic substances. The cyp83a1/ref2 mutants analyzed here may represent an example with altered host chemical composition. Mutants show both a lack of cuticular wax components that are recognized by E. cruciferarum for appressorium differentiation, as well as an accumulation of a potentially fungitoxic aldoxime.
Metabolic changes in cyp83a1 mutants cause chemical incompatibility with E. cruciferarum
Loss-of-function mutations of cyp83a1 cause metabolic perturbations in Arabidopsis (Hemm et al., 2003). In accordance with this, we measured strong differences in the concentration of many methanol-soluble metabolites (see also Fig. S4). Specific phenotypes in the host–parasite interaction support the view that certain metabolic changes may largely explain the altered development of E. cruciferarum on cyp83a1. Erysiphe cruciferarum showed delayed germination and differentiation of appressoria on the leaf surface of cyp83a1 mutants, and developed dramatically fewer conidiophores per microcolony post-penetration. This supported the view that fungal development was hampered on cyp83a1 during both the initiation and maintenance of pathogenesis. It appears possible that distinct metabolic changes are responsible for the failure of E. cruciferarum to develop on cyp83a1 before and after penetration of the host cuticle.
Adapted pathogens may utilize chemical defense compounds of their hosts as cues or nutrients for their own development (Nielsen et al., 2006). We therefore speculated that reduced amounts of aliphatic glucosinolates might cause a loss of susceptibility to the adapted powdery mildew fungus E. cruciferarum. However, the myb28-1 myb29-1 double mutant, which does not contain significant amounts of aliphatic glucosinolates (Sønderby et al., 2007, 2010; Beekwilder et al., 2008), displayed wild-type-like susceptibility to E. cruciferarum. Therefore, reduced amounts of aliphatic glucosinolates per se do not explain why cyp83a1 mutants do not support powdery mildew. Recently, it has also been reported that the myb28 myb29 double mutant shows unaltered susceptibility to Phytophthora capsici (Wang et al., 2013). Interestingly, both the myb28-1 myb29-1 double mutant and cyp83a1 mutants were extremely susceptible to the gray mold fungus B. cinerea, which belongs to the Leotiomycetes class of the Ascomycota such as E. cruciferarum (compare Figs 2 and S1). Therefore, the data suggest that an intact aliphatic glucosinolate metabolism is required for basal resistance to B. cinerea, whereas E. cruciferarum can cope with basal levels of aliphatic glucosinolates in the wild-type.
Aldoximes are highly reactive and potentially cytotoxic (Drumm et al., 1995; Sakurada et al., 2009). It has been suggested that wild-type plants channel precursors and intermediates of glucosinolate metabolism through multienzyme complexes, so-called metabolons, which allow for the controlled release of products. Metabolons may be required to prevent the undesired release of reactive metabolites that can interfere with other metabolic pathways (Møller, 2010). The characterization of ref2 mutants, which are defective in CYP83A1 (Hemm et al., 2003), provided indirect support for the existence of a glucosinolate metabolon. cyp83a1/ref2 mutants show defects in phenylpropanoid metabolism, and it has been suggested that aldoximes, which are processed by CYP83A1 in the wild-type, accumulate in cyp83a1/ref2 mutants and inhibit the enzymes of phenylpropanoid metabolism, such as caffeic acid O-methyltransferase. As a result, sinapoyl malate, as an O-methylated product of phenylpropanoid metabolism, is absent from the ref2 mutants, explaining the lack of epidermal fluorescence (Hemm et al., 2003). Here, we show that cyp83a1/ref2 mutants indeed accumulate the postulated CYP83A1 substrate aldoxime 5-MTPO. Aldoximes are toxic to fungi and are considered as potential antibiotics and biofungicides (Drumm et al., 1995). Therefore, it seems possible that aldoximes contribute to the inhibition of the post-penetration growth of E. cruciferarum on cyp83a1.
We originally discovered CYP83A1 in a biochemical screening for interaction partners of BAX INHIBITOR-1 (BI-1) (Weis et al., 2013). BI-1 is an endoplasmic reticulum (ER)-resident protein that controls cell death and susceptibility to biotic and abiotic stress. BI-1 further modulates plant metabolism and is a direct interactor of cytochrome b5 in fatty acid hydroxylation (Ishikawa et al., 2011). We do not yet understand the function of the molecular interaction of BI-1 and CYP83A1. However, it is tempting to speculate that BI-1 could function in the assembly or stabilization of ER-resident metabolons, perhaps as a molecular scaffold. BI-1 could thus support coordinated metabolism under stress conditions and protect the plant from the uncontrolled release of cytotoxic metabolites.
Lack of very-long-chain fatty aldehydes in cyp83a1 mutants explains the developmental defects of E. cruciferarum
The initial observation that acetic ink can penetrate the cuticle/wax layer of fixed leaves of cyp83a1 provoked a more detailed analysis. This suggested that the physicochemical properties of the cuticle of cyp83a1 mutants were altered. The mutants showed a greatly enhanced permeability for soluble substances into and out of leaves (Fig. 3). Accordingly, our analytical data showed an altered amount and composition of cyp83a1 wax when compared with the wild-type. Most strikingly, very-long-chain aldehydes were absent from the chloroform extracts, whereas very-long-chain alcohols were increased and very-long-chain fatty acids were increased or decreased depending on the chain length. The data suggest an as yet unrecognized metabolic phenotype of cyp83a1 mutants in the biosynthesis of very-long-chain fatty substances. It remains to be elucidated whether CYP83A1 itself could be directly involved in fatty acid metabolism, such as other cytochrome p450 monooxygenases (Höfer et al., 2008; Pinot & Beisson, 2011).
Very-long-chain aldehydes were able to promote conidia germination and appressorium differentiation of E. cruciferarum in vitro. This is in accordance with the finding that very-long-chain aldehydes also promote the differentiation of other powdery mildew fungi (Tsuba et al., 2002; Zabka et al., 2008; Hansjakob et al., 2010, 2012). Importantly, the very-long-chain aldehydes were sufficient to chemically complement cyp83a1 mutants in terms of fungal germination and support of appressorium formation. This suggests that a lack of very-long-chain aldehydes on the surface of cyp83a1 mutants fully explains the delayed differentiation of fungal infection structures. Chemical complementation of early fungal development suggests a pivotal function of the very-long-chain aldehydes in chemical compatibility, which appears to be conserved in monocots and dicots. Erysiphe cruciferarum penetrates many times the cuticle and cell wall of wild-type Arabidopsis for the formation of haustoria before it sporulates. It is therefore possible that the lack of sporulation can also be partially explained by a cumulative effect from multiple failed or delayed penetration events. This might be supported by the fact that E. cruciferarum, once it had germinated, was also delayed in the formation of secondary haustoria (Fig. 1b).
In summary, E. cruciferarum is well adapted to the chemical composition of Arabidopsis thaliana. It utilizes very-long-chain aldehydes at the plant surface as chemical cues for the induction and differentiation of infection structures and can cope with aliphatic glucosinolates in the wild-type. Genetic disruption of aliphatic glucosinolate metabolism interfered with this chemical compatibility with this powdery mildew fungus at CYP83A1, but not at the level of pathway regulation by MYB28 and MYB29. We suggest that very-long-chain aldehydes are pivotal for chemical compatibility at the level of fungal penetration. Failure of fungal sporulation might be explained by a cumulating effect of multiple penetration failures and potential fungitoxic metabolites that accumulate in cyp83a1 mutants.
We thank Katharina Beckenbauer, Andrea Knorz and Johanna Hofer for excellent technical support. We are grateful to Barbara Halkier and Meike Burow (University of Copenhagen, Denmark) for providing the myb28-1 myb29-1 mutant, and to Clint Chapple (Purdue University, West Lafayette, IN, USA) for providing the ref2 mutant. The standard 5-methylthiopentanaldoxime was generated in the laboratory of John Pickett (Rothamsted Research, Harpenden, UK) and obtained from Barbara Halkier. The Botrytis cinerea strain B05.10 was kindly provided by Paul Tudzynski (Westfälische-Wilhelms-Universität, Münster, Germany). Financial support for the project came from grants from the German Research Foundation to R.E. and R.H. (DFG EI835/1) and to the laboratories of R.H. and W.S. in the collaborative research centre SFB 924.