Image Analysis and Untargeted Metabolomics Reveal Potential Phytotoxins from Fusarium venenatum Against Major Parasitic Weed Phelipanche ramosa (L.) Pomel

Abstract

Branched broomrape (Phelipanche ramosa (L.) Pomel), an obligate parasitic weed with a wide host range, is known for its devasting effects on many crops worldwide. Soil fungi, notably Fusarium sp., are described as pathogenic to broomrape, while the hypothesis of the phytotoxicity of fusaric acid produced by F. verticillioides for parasitic weeds of the genus Orobanche has been proposed. Using image analysis and untargeted metabolomics, this study investigated fungal metabolites phytotoxic for P. ramosa and produced by the F. venenatum MIAE02836 strain, isolated from symptomatic broomrapes and identified as a promising candidate for broomrape biocontrol. Phytotoxicity tests of crude extracts from the fungus alone or in interaction with broomrape on P. ramosa microcalli and quantification of necrosis by image analysis confirmed the phytotoxic potential of F. venenatum MIAE02836 metabolites towards the early developmental stages of P. ramosa. Data analysis of a non-targeted metabolomics approach revealed numerous metabolites produced by F. venenatum MIAE02836. Four of them, accumulated during interaction with the parasitic plant, are known for their phytotoxic potential: maculosin, cyclo(Leu-Phe), phenylalanyl-D-histidine and anguidine. These results suggest that combining image acquisition of the microcalli screening test and untargeted metabolomic approach is an interesting and relevant method to characterize phytotoxic fungal metabolites.

Keywords: Fusarium venenatum; Phelipanche ramosa; fungal metabolites; image analysis; phytotoxicity; untargeted metabolomics.

Figure 1

Example of the application of the necrosis analysis model on microcalli D6 after treatment with water or with fungal filtrates of modality A. (a,b) Images of microcalli treated with water (negative control) and filtrates of modality A (fungus–broomrape interaction) respectively after standardization step; (c,d) Images of the same wells after application of the classification model. Necrotic microcalli are shown in green, healthy ones are in red, seeds are in purple, and the background is in yellow.

Figure 2

Necrosis development on broomrape microcalli depending on fungal filtrates over treatment time. The percentage of necrosis is the corrected relative ratio after subtraction of the average percentage of necrosis observed for the control at D0, D1, D5, and D6. Boxplots are shown with means (bold stars) and standard errors. Asterisks indicate significance of adjusted p-values (ns: not significant; *: 0.05; **: 0.01; ****: <0.001) after ANOVA on quasi-logistic regression post hoc pairwise comparison test between filtrates (Control: distilled water; modality A: filtrates from fungus–broomrape interaction; modality B: filtrates from fungus alone) for each day (Emmeans test with Holm–Bonferroni correction).

Figure 3

Results of metabolomic data analysis (ESI+ mode) for AMs and NAMs. (a) Proportion of metabolites with relative abundance significantly and non-significantly different based on ANOVA by permutation; (b,c) metabolites with relative abundance significantly different distributed according to their origin of production (F. venenatum MIAE02836 or P. ramosa) for NAMs and AMS respectively. The origin of production was supposed after a simple comparison between the normalized means of the relative abundance of the metabolites produced by the fungus alone (modality B) and by the broomrape alone (modality C); (d) distribution by chemical class of annotated fungal metabolites with relative abundance significantly different.

Figure 4

Accumulation of AMs with significantly different relative abundance by modality (broomrape alone, fungus alone, and interaction) (ESI+ mode). Part of the hierarchical tree showing metabolites with relative abundance significantly different between modalities A (broomrape–fungus interaction), B (fungus alone), and C (broomrape alone) according to the one-factor ANOVA by permutation (1000 permutations, p-value < 0.01) using MeV 4.9.0 software. The basal normalized relative abundance is shown in black. When the gradient tends towards yellow, the metabolite is accumulated to a greater extent. When the gradient tends towards blue, the quantity is lower. Metabolites framed in red correspond to fungal metabolites accumulated during interaction (see Table 2).

Figure 5

Mean relative abundance of fungal metabolites accumulated during the interaction with broomrape. Boxplots are shown with means (bold stars) and standard errors of the relative abundance of the fungal peptides and sesquiterpenoids accumulated during the interaction. Anguidin has not been represented, as it is heavily produced (minimum relative abundance of 329,458 metabolites and 2,442,767 metabolites for the fungus alone and for the interaction, respectively). Modality_A: filtrates from fungus–broomrape interaction; Modality_B: filtrates from fungus alone; Modality_C: filtrates from broomrape alone.

Figure 6

Workflow Overview for BRoomrape’s microcalli Automated Image-based Necrosis-detection (BRAIN). This workflow supports batch processing of images following a standardization step (Step 1) where users can choose to (i) run our pre-trained model on broomrape microcalli images (Step 2; requires same image acquisition parameters), (ii) annotate their own images (Step 3) and train a new model (Step 4), or (iii) annotate their own images and further train the pre-trained model (Alternative Step 3 and Step 4).

Figure 7

Sampling plan for phytotoxicity test. Columns 1–4 correspond to the negative control (microcalli with distilled water); columns 5–8 correspond to the filtrates of modality B (fungus alone); columns 9–12 correspond to the filtrates of modality A (fungus–broomrape interaction).