Computational protocol: Phenotypic responses to microbial volatiles render a mold fungus more susceptible to insect damage

Similar protocols

Protocol publication

[…] To test whether the substrate‐dependent effects of yeast on mold morphology and growth are related to qualitative and/or quantitative differences in the composition of volatile metabolites emitted by S. cerevisiae, we cultivated the yeast on both type of media (“no sucrose” vs. “plus sucrose”) and analyzed the volatiles emitted from the yeast colonies. For this, thoroughly cleaned glass vials of 65 ml volume (13.3 cm height, 2.5 cm diameter) were filled with 20 ml of the respective malt extract agar. The vials were covered with aluminum foil and autoclaved. After hardening of the agar, the substrate was inoculated with 50 μl yeast cell suspension. Eight vials per treatment were prepared and incubated for 4 days at standard climatic conditions (25°C, 12‐hr light cycle). Subsequently, we extracted the headspace volatiles by means of solid phase micro‐extraction (HS‐SPME). The fiber type (85 μm CarboxenTM/PDMS StableFlexTM) and a sampling time of 5 min at 20°C were chosen based on previous studies on fungal volatiles (Nilsson, Larsen, Montanarella, & Madsen, ). Identification of volatile compounds by GC‐MS followed established protocols (see Stötefeld, Holighaus, Schütz, & Rohlfs, ), detailed in the “Specifications of the GC‐MS analysis of yeast volatiles.”For the quantification of volatiles, extracted ion chromatograms (EIC) were integrated manually and recalculated to peak areas of total ion chromatograms (TIC) based on the EIC:TIC ratio of the standard compounds. Peak areas were scaled to their mean (z‐score). To describe sucrose‐mediated changes in yeast volatile formation, we created Venn diagrams for depicting presence/absence of volatiles as a function of treatment, subjected the scaled areas to a hierarchical heat‐map clustering analysis, and calculated fold changes in volatile formation using MetaboAnalyst, (Xia, Sinelnikov, Han, & Wishart, ). To further confirm the differences of the volatile profiles, we applied a machine learning algorithm (random forest analysis), to test whether the samples can be correctly assigned to predefined treatment groups (“no sucrose” or “plus sucrose”) (Breiman, ) based on volatile metabolites (peak areas) in a number of iterations. We used the “randomForest” package implemented in R (R Core Team, ), which returns a confusion matrix that contains a summary of the classification procedure and the mean prediction error, the “out‐of‐bag” error. We drew 10,000 bootstrap samples (trees) with three randomly selected variables at each node. The random forest analysis also provides a measure of importance of individual variables for a correct classification, the “mean decrease accuracy” (MDA) value. Variables that contribute strongly to the correct classification receive high positive MDA values. Small MDA values contribute little to it and negative values counteract a correct classification. [...] Five seven‐day‐old colonies of A. nidulans exposed to yeast volatiles and five nonexposed colonies were used to search for chemical differences developed under these conditions. As for the gene expression analysis, the colonies were grown on cellophane placed on culture medium (see above). Lyophilized, ground mycelium was extracted with acetonitrile–water (84/16) of ten times the sample weight. The extracts were cleared by centrifugation, the solvent was removed in vacuum, and the residues were dissolved in methanol–water (1:1) and defatted with cyclohexane.Nontargeted chemical analysis was conducted by HPLC coupled with an ion trap mass spectrometer equipped with an electrospray ionization source (HPLC‐ESI‐MS). An RP column eluted with a water/methanol gradient was used as described by Ratzinger, Riediger, von Tiedemann, and Karlovsky (). The ion trap was operated in positive mode under conditions described by Khorassani et al. (). Raw data were processed with a component detection algorithm (Windig, Phalp, & Payne, ) implemented in the ACD/MS Manager version 12.0 (Advanced Chemistry Development, Toronto, Canada) with an MCQ (mass chromatography quality) threshold of 0.8 and a window width of three scans (see Chatterjee, Kuang, Splivallo, Chatterjee, & Karlovsky, for details). For peak picking, signals with intensities of less than 500,000 counts and/or signal‐to‐noise ratios of less than 100 were discarded. The resulting peak tables were processed using custom Perl scripts for peak alignment and normalization (Laurentin, Ratzinger, & Karlovsky, ).Targeted HPLC‐MS/MS analysis was used for the identification and quantification of sterigmatocystin (m/z 325 to 281 and m/z 325 > 310), austinol (m/z 459 > 441 and m/z 459 > 423), dehydroaustinol (m/z 457 to 439 and m/z 457 to 421), emericellamide C/D (m/z 596 to 525 and m/z 596 to 507), and emericellamide E/F (m/z 624 to 553 and m/z 624 to 464). The ion trap was operated in a positive ionization mode with the following source conditions: needle voltage +5,000 V, shield voltage +600 V, capillary voltage +100 V, drying gas (nitrogen) 15 psi at 350°C, and nebulizing gas (air/nitrogen) 25 psi. Data were acquired with a scan speed of 15,000 Da/s. The identity of target compounds was confirmed by at least two mass transitions. We used the same statistical routines (Venn diagram, clustering, random forest) as for the analysis of yeast volatiles to describe the chemical differences between the unexposed and volatile‐exposed colonies. […]

Pipeline specifications

Software tools MetaboAnalyst, randomforest, CODA
Applications Miscellaneous, MS-based untargeted metabolomics
Organisms Saccharomyces cerevisiae