Development of a bioassay-guided genome mining approach for antifungal natural product discovery from pseudomonads Authors: George Lund, Susan Mosquito, David M Withall, John Caulfield, David Hughes, Ian M Clark, Jason Rudd, Tim H Mauchline Published: 2026 ABSTRACT Zymoseptoria tritici causes Septoria Leaf Blotch disease of wheat and has evolved to overcome most chemical and genetic control methods. As such, new tools are required for future disease control. We identified Pseudomonas isolates that antagonise Z. tritici through the production of secreted secondary metabolites, using a novel in vitro antagonism assay. Quantitative assessment identified variation in sensitivity of Z. tritici isolates to antagonism by Pseudomonas isolates. Genome assemblies of 3 strongly antagonistic Pseudomonas isolates contain a predicted Biosynthetic Gene Cluster (BGC) with high similarity to a reference BGC encoding the biosynthesis of known antifungal compound 2,4-diacetylphloroglucinol (2,4-DAPG). Mutagenesis of the core biosynthetic gene phlD resulted in loss of 2,4-DAPG production in Pseudomonas isolate Roth82, and loss of Z. tritici inhibition in vitro. These results demonstrate that our in vitro antagonism assay can be used to identify, quantify, and mechanistically characterise bacterial antagonism of Z. tritici through the production of secondary metabolites. This is the first study to putatively identify that resistance to bacterial antagonism exists as a quantitative trait within natural Z. tritici populations. Our approach of in vitro phenotype-guided genome mining can also prioritise candidate BGCs for discovery of antifungal natural products effective against Z. tritici.Supplementary InformationThe online version contains supplementary material available at 10.1038/s41598-026-48020-5. Introduction Wheat is the UK’s most widely grown crop, but yield is diminished due to diseases caused by pathogenic fungi. The most important disease of wheat in the UK is Septoria tritici blotch, caused by the fungus Zymoseptoria tritici, resulting in up to 50% yield losses1. Wheat has no durable natural resistance and the fungus has become insensitive to all classes of existing commercial fungicides2. Therefore, new ways to protect wheat are urgently needed. One underexplored approach is to exploit natural antifungal chemistries encoded within microbial genomes. The soil microbiome is diverse and largely untapped for the discovery of new molecules that can suppress plant pathogens. To date, most antimicrobial compounds have been derived from soilborne actinomycetes3. Pseudomonas bacteria are an important component of the soil microbiome with exemplar species behaving as human pathogens4(e.g. P. aeruginosa), plant pathogens5(e.g. P. syringae), as well as plant growth promoting bacteria6, and biocontrol agents7,8(e.g. P. fluorescens); yet the diverse secondary metabolite repertoire of this genus, which has potential for the discovery of new high value fungicides, is relatively underexplored experimentally9. Previous studies investigating in vitro antagonism of Z. tritici by Pseudomonas aeruginosa isolate LEC1, found the siderophore pyocyanin10 and 1-hydroxyphenazine11,12 to partially contribute towards antagonism. Pseudomonas fluorescens PMF2 was found to antagonise Z. tritici through the production of 2,4-diacetylphloroglucinol (2,4-DAPG) and 2 additional unidentified antifungal compounds (Levy et al.,13. Tropolone, produced by Pseudomonas lindbergii, was found to inhibit the growth of Z. tritici in vitro14,15. Pseudomonas isolates have also previously been shown to antagonise Z. tritici through the production of volatile compounds, with Pseudomonas putida BK8661 found to antagonise Z. tritici through the production of hydrogen cyanide (HCN) in vitro16. In vitro studies investigating the antagonistic potential of bacterial isolates against fungal plant pathogens, have mostly consisted of single genotypes of fungal plant pathogens16,10,13,12,14. However, Z. tritici has high levels of genetic diversity, with previous studies identifying over 600 different genotypes of the fungus in a single field over a single season17, and individual leaf lesions were found to harbour 2–5 different genotypes18. Quantification of in vitro antagonism of fungal pathogens by bacterial isolates is difficult, owing to the filamentous hyphal growth of many fungal pathogens, and therefore measurements have commonly taken place through qualitative assessments19,20. Quantitative assessment of in vitro Z. tritici antagonism could facilitate the identification of Pseudomonas isolates capable of antagonising Z. tritici in vitro in a broader context. Such assessments could facilitate sensitivity comparisons of genetically diverse Z. tritici isolates to antagonistic Pseudomonas isolates. Pseudomonads are generally amenable to culture and this property facilitates the acquisition of their genomes, which in turn assists in the identification of the genetic components responsible for the biosynthesis of antifungal or fungistatic secondary metabolites21. This study introduces a novel qualitative and quantitative blastospore-based assay for high throughput assessment of pseudomonad growth suppression of multiple Z. tritici genotypes. The radial zones of clearing produced by antagonistic Pseudomonas isolates, confronted with Z. tritici blastospores as a confluent lawn, offer the potential to both qualitatively and quantitatively assess antagonism of Z. tritici by Pseudomonas isolates. This assay in combination with genomics, microbial mutagenesis and analytical chemistry enables the prioritisation of predicted secondary metabolite Biosynthetic Gene Clusters (BGCs) from antagonistic isolates for further characterisation, and comparison with BGCs previously described in literature from the MIBiG 3.0 repository22. We report the initial screening of a Pseudomonas library of 534 isolates for suppression of the Z. tritici reference isolate IPO323. Subsequent in vitro testing of 5 antagonistic and 6 non-antagonistic Pseudomonas isolates confronted with 12 genetically diverse Z. tritici isolates, revealed significant differences in resistance to bacterial antagonism. To our knowledge this is the first study to find evidence suggesting sensitivity to bacterial antagonism exists as a quantitative trait within natural Z. tritici populations. We demonstrate the power of a bioassay-mediated genomics approach to define potential antifungal secondary metabolite BGCs involved in bacterial-fungal interactions in vitro. The study also demonstrates the importance of testing genetically diverse individuals from within pathogen populations to ensure consistency of antagonistic responses ahead of further exploitation of bacterial secondary metabolites or biological control agents as future disease control strategies. Results Qualitative screening of Z. tritici IPO323 antagonism A total of 534 Pseudomonas isolates were screened qualitatively for antagonism of Z. tritici IPO323 blastospores, with 52 identified as being antagonistic (Fig. 1, Table S1). Fig. 1A high-throughput qualitative assay to assess the ability of Pseudomonas antagonism of Z. tritici blastospores in vitro. (a) Schematic diagram of the qualitative Z. tritici blastopore antagonism assay. In the bioassay, Pseudomonas isolates are spotted onto a confluent lawn of Z. tritici blastospores on PDA. Created in BioRender. Lund, G. (2024) BioRender.com/p07z107 (b) A photograph of the qualitative Z. tritici blastospore antagonism assay, with 48 Pseudomonas colonies on one assay plate in confrontation with a single fungal isolate. A high-throughput qualitative assay to assess the ability of Pseudomonas antagonism of Z. tritici blastospores in vitro. (a) Schematic diagram of the qualitative Z. tritici blastopore antagonism assay. In the bioassay, Pseudomonas isolates are spotted onto a confluent lawn of Z. tritici blastospores on PDA. Created in BioRender. Lund, G. (2024) BioRender.com/p07z107 (b) A photograph of the qualitative Z. tritici blastospore antagonism assay, with 48 Pseudomonas colonies on one assay plate in confrontation with a single fungal isolate. Pseudomonas isolates demonstrate variable quantitative suppression of a genetically diverse Z. tritici collection Of the 52 Pseudomonas isolates that demonstrated qualitative antagonism to Z. tritici IPO323 blastospores, 5 antagonistic isolates alongside 6 non-antagonistic isolates (Table S2) from the collection were randomly selected and screened quantitatively for antagonism against a collection of 12 genetically diverse Z. tritici isolates collected from across Europe in 2015/1623(Fig. 2). All 5 qualitatively antagonistic Pseudomonas isolates were found to suppress growth of Z. tritici IPO323 as well as the additional 11 genetically diverse Z. tritici isolates in the quantitative antagonism assay (Fig. 2). In addition, all 6 Pseudomonas isolates which had been determined to be qualitatively non-antagonistic towards Z. tritici IPO323, as well as LB Miller Broth medium controls, did not produce an observable zone of clearing when confronted with any of the Z. tritici isolates screened (Figure S7). This indicates that IPO323 confrontation is an appropriate starting test for the identification of Pseudomonas isolates with the potential to antagonise Z. tritici isolates qualitatively in vitro. Fig. 2Significant differences in zones of inhibition of genetically distinct Z. tritici blastospores by Pseudomonas isolates in vitro, using a quantitative antagonism assay. (a) A schematic diagram of the quantitative Z. tritici blastopore antagonism assay. Created in BioRender. Lund, G. (2023) BioRender.com/y78z760 (b) Images of in vitro quantitative antagonism plates exhibiting differences in zones of inhibition produced by Pseudomonas isolates against IPO323 and Zt116 blastospores. (c) Heatmap indicating the average radius of the zone of inhibition produced in each pairwise interaction between antagonistic Pseudomonas and Z. tritici isolates in the quantitative antagonism assay after 12 days (n = 4). (d-h) Zones of inhibition produced by Pseudomonas isolates in confrontation with genetically diverse Z. tritici blastospores in vitro. Mean values of zones of inhibition not sharing any letter are significantly different by Sidak’s test for multiple comparisons at the 5% level of significance for Pseudomonas isolates (d) BFR20 (e) BFB20 (f) Roth82 (g) Roth107 (h) Roth112. Significant differences in zones of inhibition of genetically distinct Z. tritici blastospores by Pseudomonas isolates in vitro, using a quantitative antagonism assay. (a) A schematic diagram of the quantitative Z. tritici blastopore antagonism assay. Created in BioRender. Lund, G. (2023) BioRender.com/y78z760 (b) Images of in vitro quantitative antagonism plates exhibiting differences in zones of inhibition produced by Pseudomonas isolates against IPO323 and Zt116 blastospores. (c) Heatmap indicating the average radius of the zone of inhibition produced in each pairwise interaction between antagonistic Pseudomonas and Z. tritici isolates in the quantitative antagonism assay after 12 days (n = 4). (d-h) Zones of inhibition produced by Pseudomonas isolates in confrontation with genetically diverse Z. tritici blastospores in vitro. Mean values of zones of inhibition not sharing any letter are significantly different by Sidak’s test for multiple comparisons at the 5% level of significance for Pseudomonas isolates (d) BFR20 (e) BFB20 (f) Roth82 (g) Roth107 (h) Roth112. A statistically significant difference in the mean zones of inhibition (millimetres) was detected between genetically diverse Z. tritici isolates confronted with antagonistic Pseudomonas isolates; BFR20 (F(11, 36) = 11.48, p < 0.001), BFB20 (F(11, 36) = 10.97, p < 0.001), Roth82 (F(11, 36) = 4.455, p < 0.001), Roth107 (F(11, 36) = 5.567, p < 0.001), and Roth112 (F(11, 36) = 8.11, p < 0.001). Significant differences in the mean zones of inhibition produced by antagonistic Pseudomonas were determined using Sidak’s multiple comparison post hoc test (Fig. 2d-h, Tables S4-S13). Genome sequencing, taxonomic estimation, and predictive bioinformatics of secondary metabolism of environmental Pseudomonas isolates Genome assemblies were generated by MicrobesNG for the 11 Pseudomonas isolates examined in the quantitative Z. tritici antagonism assay. All genome assemblies were found to have BUSCO scores > 99%, when analysed using the bacterial odb10 dataset (containing 124 single copy orthologous genes) to check for bacterial contamination within the genome assembly (Figure S1). Additionally, BUSCO analysis of the generated genome assemblies using the Pseudomonadales odb10 dataset (containing 782 single copy orthologous genes) also returned BUSCO scores > 99% for all assemblies. BUSCO analyses indicated high levels of genome assembly completeness and low levels of genome assembly contamination (Figure S2). To predict putative secondary metabolite biosynthetic gene clusters from the Pseudomonas genome assemblies, antiSMASH 7.0.024 was used. A total of 131 BGCs were predicted from the genome assemblies of the 11 Pseudomonas isolates (Figure S3). To compare predicted BGCs between Pseudomonas isolates, and compare to reference BGCs from the MIBiG 3.0 database22, sequence similarity networks of predicted secondary metabolite BGCs from both antagonistic and non-antagonistic Pseudomonas isolates were generated using BiG-SCAPE25, including reference BGCs from MIBiG 3.0. A total of 46 GCF networks were formed of two or more BGCs, with 3 GCFs containing a reference BGC from MIBiG 3.0 (GCF_00413, GCF_02527, GCF_02551, Fig. 3a), indicating similarity of predicted BGCs from Pseudomonas genome assemblies within this study, at the 0.3 threshold in BiG-SCAPE. A total of 13 GCF networks were found to contain predicted BGCs from solely Z. tritici blastospore antagonistic Pseudomonas isolates (Fig. 3a). Of these, GCF_02551 contained reference BGCs from MIBiG 3.0, BGC0000280 and BGC0000281, which have been shown experimentally to encode the biosynthesis of 2,4-diacetylphloroglucinol26, a known antifungal compound effective at inhibiting Z. tritici in vitro(Levy et al.,13. Predicted BGCs within GCF_02551 were aligned using clinker, with high levels of amino acid similarity, and synteny identified between the reference BGCs and those predicted from the generated Pseudomonas genome assemblies (Fig. 3b). Fig. 3Sequence similarity networks generated using BiG-SCAPE reveal Gene Cluster Families associated with the Z. tritici blastospore antagonism phenotype, including reference BGCs with known antifungal activity from the MIBiG repository. (a) Sequence similarity networks indicated predicted BGCs with high sequence similarity to MIBiG reference BGCs for pyoverdine (BGC0000413, GCF_02527), 2,4-DAPG (BGC0000280 & BGC0000281, GCF_02551), and pseudomonine (BGC0000410, GCF_00413), within the Pseudomonas genome assemblies. (b) Amino acid alignment of BGCs forming GCF_02551 with 2,4-DAPG biosynthesis reference BGCs from MIBiG (BGC0000280 & BGC0000281) indicates high levels of amino acid similarity and gene synteny between predicted BGCs. Sequence similarity networks generated using BiG-SCAPE reveal Gene Cluster Families associated with the Z. tritici blastospore antagonism phenotype, including reference BGCs with known antifungal activity from the MIBiG repository. (a) Sequence similarity networks indicated predicted BGCs with high sequence similarity to MIBiG reference BGCs for pyoverdine (BGC0000413, GCF_02527), 2,4-DAPG (BGC0000280 & BGC0000281, GCF_02551), and pseudomonine (BGC0000410, GCF_00413), within the Pseudomonas genome assemblies. (b) Amino acid alignment of BGCs forming GCF_02551 with 2,4-DAPG biosynthesis reference BGCs from MIBiG (BGC0000280 & BGC0000281) indicates high levels of amino acid similarity and gene synteny between predicted BGCs. To estimate the taxonomy of the generated Pseudomonas genome assemblies, autoMLST27 was used. A phylogenetic tree was constructed using 100 common genes (Table S3) from the 11 Pseudomonas genome assemblies, with the closest 50 genome assemblies from RefSeq included in the phylogenetic tree construction. The closest RefSeq type strains to the genome assemblies generated in this study were cross referenced with previous literature to identify clades associated with described subgroups of the P. fluorescens species complex28,29(Fig. 4). Fig. 4Phylogenetic distribution of Pseudomonas isolates with predicted secondary metabolite BGCs with significant similarity to known BGCs in relation to Z. tritici blastopore antagonism phenotypes. Phylogenetic analysis with known Pseudomonas strains, using autoMLST, putatively identifies 11 Pseudomonas strains used within this study within the P. fluorescens, P. corrugata, P. koreensis, P. jessenii, and P. mandelii subgroups of the P. fluorescens species complex. The presence and absence of predicted BGCs that form sequence similarity networks with experimentally characterised BGCs from the MIBiG database are indicated, alongside qualitative Z. tritici blastospore antagonism phenotype against Z. tritici strain IPO323. Type strain Pseudomonas genomes identified according to LPSN (accessed 24th January 2025) are denoted by a superscript ‘T’. Phylogenetic distribution of Pseudomonas isolates with predicted secondary metabolite BGCs with significant similarity to known BGCs in relation to Z. tritici blastopore antagonism phenotypes. Phylogenetic analysis with known Pseudomonas strains, using autoMLST, putatively identifies 11 Pseudomonas strains used within this study within the P. fluorescens, P. corrugata, P. koreensis, P. jessenii, and P. mandelii subgroups of the P. fluorescens species complex. The presence and absence of predicted BGCs that form sequence similarity networks with experimentally characterised BGCs from the MIBiG database are indicated, alongside qualitative Z. tritici blastospore antagonism phenotype against Z. tritici strain IPO323. Type strain Pseudomonas genomes identified according to LPSN (accessed 24th January 2025) are denoted by a superscript ‘T’. Phylogenetic analysis indicated that the 11 Pseudomonas strains within this study span 4 distinct subgroups of the P. fluorescens species complex, as defined by Hesse et al.,28. Isolates predicted within the P. corrugata subgroup, Roth82, Roth107 and Roth112, were all found to exhibit antagonistic phenotypes when confronted with Z. tritici blastopores in vitro. Both Z. tritici blastospore antagonistic (BF-R-20) and non-antagonistic (A-B-20, Roth105) Pseudomonas isolates were found within the P. mandelii subgroup. Isolates Roth17 and Roth40, within the P. fluorescens subgroup, and Roth79 and Roth97 within the P. jessenii subgroup were all found to exhibit a non-antagonistic phenotype when confronted with Z. tritici blastospores in vitro. The resulting phylogenetic tree was annotated with the GCF network data generated using BiG-SCAPE, to visualise the distribution of predicted BGCs with a high similarity to MIBiG reference BGCs (Fig. 4), and all predicted BGCs forming GCF networks (Figure S5). All 3 isolates classified within the P. corrugata subgroup (Roth82, Roth107 and Roth112) were found to contain predicted BGCs that formed a sequence similarity network with MIBiG reference BGCs encoding the production of 2,4-DAPG. The genome assemblies of Z. tritici non-antagonistic P. fluorescens subgroup isolates Roth17 and Roth40 were found to contain predicted BGCs with significant similarity to the MIBiG reference BGC encoding the biosynthesis of pseudomonine. Predicted BGCs with a significant similarity to MIBiG reference BGC BGC0000413, encoding the production of pyoverdine, were found in genome assemblies of the non-antagonistic P. fluorescens subgroup (Roth17 & Roth40), P. jessenii subgroup (Roth97), and P. mandelii subgroup (Roth105), as well as the antagonistic P. mandelii subgroup (BF-R-20) isolate. Gene disruption of BGC predicted to encode the biosynthesis of 2,4-DAPG results in a loss of visible Zymoseptoria tritici antagonism, and a reduction in 2,4-DAPG detected in agar extracts of confrontation zones Roth82 was selected for proof-of-concept functional validation because it was readily transformable under our conditions and carried a predicted 2,4-DAPG biosynthetic gene cluster with high similarity to reference clusters. To examine the involvement of known antifungal secondary metabolite 2,4-DAPG in the observed Z. tritici blastospore antagonism by isolate Roth82 (P. corrugata subgroup), a single phlD insertional mutant was constructed by insertional disruption using the pKNOCK-Km suicide vector as a proof of concept, in which the phlD gene was disrupted in Roth82, resulting in Roth82 phlD::pKNOCK-Km (Figure S6). Zymoseptoria tritici blastospore confrontation assays were undertaken with strain IPO323, with methanol agar extracts from the zone surrounding Roth82 and Roth82 phlD::pKNOCK-Km prepared after 12 days of confrontation. No visible zone of inhibition was present for Roth82 phlD::pKNOCK-Km (Fig. 5b), whereas a zone of inhibition was seen for wild-type Roth82 (Fig. 5a). These observed phenotypes correlated with LC-MS analysis of the agar extracts, with wild-type Roth82 extract showing a peak at m/z 209.0455 ± 5ppm [M-H]− with a retention time of 26.65 min (Fig. 5c), corresponding to the authentic chemical standard for 2,4-DAPG (Sigma, Fig. 5c). For the Roth82 phlD::pKNOCK-Km mutant confronted with Z. tritici blastospores, no peak corresponding to the retention time and m/z 209.0455 [M-H]− of the authentic chemical standard of 2,4-DAPG was detected (Fig. 5c). Fig. 5Disruption of phlD in isolate Roth82 (P. corrugata subgroup) results in a loss of Z. tritici blastospore antagonism in vitro, and reduction in the production of 2,4-DAPG. Zymoseptoria tritici IPO323 blastospore antagonism assay with wild-type Roth82 exhibiting a zone of inhibition (a) and Roth82 phlD::pKNOCK-Km showing no zone of inhibition (b). (c) Extracted ion chromatogram at m/z = 209.0455 with a mass tolerance of ± 5ppm, corresponding to [M-H]− (molecular formula C10H9O5) of 2,4-DAPG (C10H10O5). Liquid Chromatography-Mass Spectrometry analysis was performed on agar extracts from the area surrounding the colony of wild-type Roth82 in confrontation with Z. tritici IPO323 blastospores (black), Roth82 phlD::pKNOCK-Km in confrontation with Z. tritici IPO323 blastospores (turquoise), and the authentic chemical standard for 2,4-DAPG (Sigma, orange). Disruption of phlD in isolate Roth82 (P. corrugata subgroup) results in a loss of Z. tritici blastospore antagonism in vitro, and reduction in the production of 2,4-DAPG. Zymoseptoria tritici IPO323 blastospore antagonism assay with wild-type Roth82 exhibiting a zone of inhibition (a) and Roth82 phlD::pKNOCK-Km showing no zone of inhibition (b). (c) Extracted ion chromatogram at m/z = 209.0455 with a mass tolerance of ± 5ppm, corresponding to [M-H]− (molecular formula C10H9O5) of 2,4-DAPG (C10H10O5). Liquid Chromatography-Mass Spectrometry analysis was performed on agar extracts from the area surrounding the colony of wild-type Roth82 in confrontation with Z. tritici IPO323 blastospores (black), Roth82 phlD::pKNOCK-Km in confrontation with Z. tritici IPO323 blastospores (turquoise), and the authentic chemical standard for 2,4-DAPG (Sigma, orange). Discussion Our approach utilised a novel high throughput phenotyping screening methodology to identify environmental Pseudomonas isolates with an ability to suppress growth of the wheat pathogen Z. tritici in vitro. We combined an in vitro phenotyping method with genome mining of antagonistic Pseudomonas isolates to identify putative antifungal secondary metabolite biosynthetic gene clusters and associated gene cluster families. This led to the identification of 52 isolates from a collection of 534 with an ability to suppress Z. tritici IPO323 blastospores in vitro. Genome sequencing and assembly of a subset of this collection, comprising 5 antagonistic and 6 non-antagonistic strains was conducted, followed by the prediction of 131 BGCs that formed 46 discrete GCF networks. In addition to demonstrating high throughput qualitative screening, our results also indicate that the blastospore antagonism assay can be used to quantitatively assess antagonism of Z. tritici growth by Pseudomonas isolates in vitro, albeit with lower throughput due to increased spacing of bacterial cultures on test plates and the requirement to adopt digital image analysis to quantify inhibition zones. We believe that this assay can be expanded to assess the ability of other microbial species to suppress Z. tritici growth, though we recommend that this is not done in a disparate manner (e.g., wholesale screening of bacterial culture collections), but rather tailored toward specific bacterial genera to reduce the likelihood of potential quality control issues when simultaneously screening bacteria with inherently different growth kinetics. Assessment of fungal antagonism by bacterial isolates is commonly conducted qualitatively, owing to the difficulty in quantifying antagonism of hyphal networks growing across entire plates in vitro. A high throughput semi-solid medium based screen has been described, using spectrophotometer measurements (OD600) and turbidimetric measurements to quantify growth of the pathogenic fungus Aspergillus flavus when exposed to chemical libraries, to screen for antagonism of the fungus through a reduction in growth rate compared to control30. However, the application of this technique is limited for our purposes, due to the overlapping spectra commonly used to quantify the growth of bacteria and Z. tritici, hampering the discrimination of bacterial isolates and Z. tritici growth. We report the first study to find significant differences in the response of genetically diverse isolates of Z. tritici to bacterial antagonists. Our ability to identify subtle differences in the in vitro growth profiles of Z. tritici in a quantitative manner, when confronted with a given antagonistic pseudomonad, indicates that natural variation in sensitivity or resistance to bacterial antagonists and associated secondary metabolites exists amongst Z. tritici isolates obtained from across Europe. We found that significant differences in the variability of inhibition between the original Z. tritici reference isolate IPO323 (isolated in 1984) and the more recently isolated strains were observed in response to confrontation with some pseudomonads, such as BFR20. This suggests that selection for broad spectrum antifungal insensitivity in Z. tritici across this period, may be correlated with differences in response to in vitro bacterial antagonism and further studies to identify mechanistic understanding of this phenomenon are required. A number of quantitative traits have been identified in Z. tritici populations, such as temperature tolerance31, and fungicide resistance32. The results presented within this study highlight the possibility that sensitivity to bacterial antagonism may also exist as a quantitative trait within natural Z. tritici populations. The patterns of sensitivity of Z. tritici isolates to antagonism by Pseudomonas isolates show that no Z. tritici isolate tested was consistently the most sensitive or resistant towards all antagonistic Pseudomonas isolates screened. Our antagonism phenotype paired bioinformatic analyses suggest that this could indicate differing underlying mechanisms of antagonism by the Pseudomonas isolates and potentially differing genetic targets for antagonism in Z. tritici blastospores, or differing levels of tolerance to antifungal metabolites in the Z. tritici test set. Differences in Z. tritici genotype-specific sensitivity to 2,4-DAPG could reflect variation in tolerance mechanisms such as metabolic detoxification as previously proposed for tolerant Fusarium oxysporum isolates33, although this was not investigated within this study. In future work it will be interesting to generate mapping populations of sexual crosses between Z. tritici isolates resistant and/or sensitive to particular bacterial antagonists, to identify possible ‘modes-of-action’ of bacterial antagonists. This could be achieved through the screening of the mapping population in the blastospore confrontation assay, and the use of genome-wide association study (GWAS) approaches to identify genomic regions in Z. tritici associated with heightened sensitivity or resistance to antagonism. A limitation of our described approach is that the observed zones of clearing produced by antagonistic Pseudomonas isolates are potentially constrained by the diffusion of secreted compounds involved in antagonism through the agar medium. This could limit the ability of this screen to identify subtle changes in sensitivity of Z. tritici isolates to classes of secreted high-molecular weight compounds such as lipopeptides34–36 in a high throughput manner. However, lower throughput microscopic image analysis to assess fungal blastospore density in response to confrontation will facilitate the determination of infraspecific suppressive activity of pseudomonads. A further limitation is that in vitro antagonism does not necessarily translate to in planta efficacy, where compound stability and diffusion alongside many biotic and abiotic factors, can strongly influence outcomes, as previous studies have shown37. The validation of candidate metabolites for antifungal activity against Z. tritici in agronomically relevant scenarios represents an important next step outside the scope of the present proof-of-concept study. The 11 Pseudomonas isolates screened for antagonism of genetically diverse Z. tritici isolates were analysed bioinformatically to obtain predictions of secondary metabolite BGCs from genome assemblies. This analysis indicated that known antifungal secondary metabolite BGCs, such as genes for 2,4-DAPG production, were implicated in fungal suppression in vitro. However, we also found that for P. mandelii subgroup isolate BFB20, a mechanism of antagonism through the production of known secreted antifungal secondary metabolites could not be inferred from predicted BGC similarity to MIBiG reference BGCs. Additionally, the genome assembly of P. mandelii isolate BFR20 only contained one predicted BGC found to be significantly similar to the MiBIG reference BGC for Pf-5 pyoverdine (BGC0000413), though BGCs in this GCF were also predicted from non-antagonistic strains Roth17, Roth40, Roth97, and Roth105. The observed antagonism phenotype of these strains, BFB20 and BFR20, could therefore be due to the production of potentially novel secondary metabolites, encoded by BGCs dissimilar to reference BGCs from the MIBiG repository. As such these strains, and predicted BGCs within GCFs deriving solely from antagonistic isolates, namely GCF_00030, GCF_00023, and GCF_00028 (all predicted from BFR20), GCF_02559 (predicted from BFR20 and Roth112), and GCF_00016 (predicted from BFB20), should be prioritised in future work for further characterisation and the potential discovery of new antifungal molecules. These findings are consistent with previous studies that found many predicted secondary metabolite BGCs from Pseudomonas type strains were significantly dissimilar to known reference BGCs from the MiBIG repository, and therefore potentially encode novel secondary metabolites9. Characterisation of predicted BGCs with significant similarity to reference BGCs encoding 2,4-DAPG production in isolates Roth107 and Roth112 (P. corrugata subgroup) will determine whether the observed in vitro antagonism phenotype is solely due to 2,4-DAPG production, or if other predicted BGCs encoding unknown secondary metabolite products contribute to the phenotype. Notably, Roth112 showed consistently stronger antagonism than Roth82 against Z. tritici isolates in vitro (Fig. 2), which may reflect increased 2,4-DAPG contribution and/or additional bioactive metabolites acting alongside 2,4-DAPG. Roth112 is therefore a priority target for future functional characterisation to dissect the basis of this phenotype. However, our results showed that the in vitro antagonism of isolate Roth82 (P. corrugata subgroup) appeared to be solely due to 2,4-DAPG production. Three antagonistic bacteria were bioinformatically predicted to have the potential to biosynthesise the antifungal polyketide 2,4-DAPG. Production of 2,4-DAPG has been described from two Pseudomonas species subgroups, P. corrugata and P. protegens38–42, and more recently in the genus Chromobacterium43. This corroborates with the putative classification of Roth82, and Roth107, and Roth112 as within the P. corrugata subgroup based on taxonomic inference within this study, and the presence of a predicted BGC with high sequence similarity to 2,4-DAPG reference BGCs38,26,41,42. Predicted BGCs with a high sequence similarity to those encoding the biosynthesis of pyoverdines44 and pseudomonine45 were also predicted from the genome assemblies of Pseudomonas isolates within this study but were either found in an equal number of Z. tritici antagonistic and non-antagonistic strains (pyoverdine), or solely from Z. tritici non-antagonistic isolates (pseudomonine). Whilst these BGCs were not taken forward for further characterisation in respect to the Z. tritici antagonism phenotype, it is possible that diversity in the expression profiles of these BGCs exist between the Pseudomonas isolates, or that the siderophore products of these BGCs differ between isolates. For this reason, predicted BGC presence/absence was only used to enable prioritisation of BGCs associated with Z. tritici antagonism phenotypes, rather than provide definitive identifications of similarity. Whilst these predicted BGCs did not appear to be strongly associated with the Z. tritici blastospore antagonism phenotype, they could be implicated in antagonistic interactions with other fungi, or with Z. tritici under different environmental conditions or contexts. For example, differing carbon sources have previously been shown to influence 2,4-DAPG production by Pseudomonas strain F11346. This highlights a strength of a phenotype-guided genome mining for secondary metabolites involved in microbial interactions within simplified experimental contexts. As a proof of concept, we generated a 2,4-DAPG insertional mutant, by disrupting the core gene encoding the biosynthetic enzyme PhlD. This enzyme is responsible for the production of monoacetylphloroglucinol (MAPG), that undergoes subsequent acetylation to produce 2,4-DAPG26. This resulted in no detectable production of 2,4-DAPG in Roth82 phlD::pKNOCK-Km, and a loss of the visible zone of inhibition in the Z. tritici blastospore antagonism assay, confirming the validity of our approach to determine mode of action of known antifungal BGCs. It will be interesting to use this approach to quantify the relative contribution of predicted secondary metabolite BGCs to the Z. tritici antagonism phenotype where multiple antifungal BGCs are predicted in a strain. For bacterial isolates where mechanisms of antagonism cannot be bioinformatically predicted, analytical chemistry of isolate extracts can facilitate the identification of molecules implicated in fungal suppression. This, when coupled with other approaches such as the creation and screening of transposon mutant libraries can facilitate the identification of previously unknown antifungal molecules. Similarly, the in vitro Z. tritici blastospore antagonism assay could be adapted to screen fractions and/or compounds of interest and could be used in a bioassay-guided fractionation approach. The efficacy of biological control agents added to environmental settings is often unpredictable47. Our approach of identifying molecules responsible for the suppression of Z. tritici does not rely on phyllosphere competence and/or production of antifungal secondary metabolites in situ on the plant surface. The method described can be used to efficiently and cost-effectively mine bacterial strain collections for novel antifungal secondary metabolites that are effective against Z. tritici in an in vitro setting. Furthermore, the described Pseudomonas-Zymoseptoria interaction system could be further exploited as a potential ‘stepping-stone’ for the identification of broad spectrum antifungal secondary metabolites with activity against other fungal plant pathogens, that are less amenable to high-throughput in vitro screening, as well as medically important fungi such as Candida and Aspergillus spp. Methods Pseudomonas isolate collection The library of 534 Pseudomonas wheat root associated isolates utilised in this study were generated previously at Rothamsted Research in Harpenden, UK20,48 and maintained as glycerol stocks in the Rothamsted microbial culture collection at -80 °C. Zymoseptoria tritici isolate collection Twelve genetically diverse Z. tritici isolates utilised in this study were isolated from across Europe in 2016 as part of a previous study23, with the exception of Z. tritici isolate IPO323, the reference strain first isolated in the Netherlands in 198449. Culturing of Pseudomonads and Z. tritici Zymoseptoria tritici blastospores were cultured by spreading 30 µl of defrosted glycerol stock on potato dextrose agar (PDA, Formedium) plates, before inverting and placing in darkness at 16 °C for 5 days. Blastospores were harvested using a sterile loop after 5 days of growth. Pseudomonas isolates were cultured from glycerol stocks, by inoculating 10 ml of LB Miller Broth (Formedium), with a chip (approximately 5 µl volume) of frozen glycerol stock. The inoculated broth was placed in a shaking incubator at 180 rpm and 25 °C overnight (18 h). Zymoseptoria tritici blastospore antagonism assay To qualitatively assess Z. tritici antagonism, a 20 µl loop of harvested Z. tritici blastospores was vortexed with 5 ml of distilled water, resulting in the formation of a suspension of approximately 1 × 106 blastospores per ml23. Next, 800 µl of spore suspension was added to a 120 mm x 120 mm Petri dish plates containing 50 ml PDA (Formedium). Plates were tilted to ensure a confluent inoculation of the blastospore suspension across the entire plate before excess spore suspension was recovered. Plates were air-dried for 20 min in a laminar flow hood, before 1 µl of 0.6 OD600 of a given Pseudomonas overnight culture was applied. A total of up to 48 Pseudomonas isolates were arrayed per plate in triplicate. Plates were sealed with Parafilm (Amcor) and inverted, before incubation in darkness at 16 °C for 12 days, prior to assessment for fungal growth inhibition. Images were taken after 12 days from a standardised height of 78.5 cm, using a Nikon D80 digital camera with a Sigma 70 mm F2.8 DG Macro lens (settings ISO 400, RAW image quality, large image size). When performing quantitative antagonism measurements, the assay was conducted as described above with increased spacing; 11 Pseudomonas isolates and an overnight LB Miller Broth (Formedium) negative control were included per plate and four replicate plates were prepared in a factorial design. Images of confrontation plates were used to measure fungal growth inhibition as described above. Measurements were taken from the edge of the bacterial colony to the outer boundary of the visible zone of clearing of Z. tritici blastospores, in a radial manner from the centre of the bacterial colony, using FIJI/ImageJ (version 2.1.0/1.53c) to quantify the number of pixels (2253 pixels = 1 mm). For each bacterial colony, three independent radial measurements were taken using a randomly generated measurement angle (1–360°) from the colony centre. Where a randomly selected radial intersected a region in which zones of clearing from neighbouring colonies met or overlapped, such that the boundary could not be unambiguously attributed to a single colony, that direction was treated as non-measurable and a new random angle was generated. Figures 1a and 2a were created using BioRender. Genome sequencing of Pseudomonas isolates Genome sequencing was provided by MicrobesNG (http://www.microbesng.com). Draft genome assemblies for 11 Pseudomonas isolates were generated using 250 bp paired end Illumina MiSeq. Sequences were passed through their bioinformatic pipeline consisting of Kraken50, BWA mem51, before genomes were assembled de novo using SPAdes52 and annotated using Prodigal53. BUSCO (Benchmarking Universal Single-Copy Orthologs) analysis BUSCO analysis of genomes was carried out on a local server using BUSCO version 5.2.254, using the bacterial odb10 reference single copy orthologous genes dataset for bacterial species to assess genome assemblies for bacterial contamination, and the Pseudomonadales odb10 reference dataset to assess genome assembly completeness. Phylogenetic tree construction A phylogenetic tree was constructed using Automated Multi-Locus Species Tree (autoMLST), based on a concatenation of 100 common genes (Table S3) found across the genome assemblies of the 11 Pseudomonas isolates within this study, and the closest 50 reference genome assemblies from RefSeq27. Annotations to the phylogenetic tree were added using ITOL55. Clades within the phylogenetic tree were annotated, to indicate the Pseudomonas subgroups present as defined by28. Type strain Pseudomonas genomes were identified using The List of Prokaryotic names with Standing in Nomenclature (LPSN) (https://lpsn.dsmz.de/) accessed on the 24th January 2025. Secondary metabolite biosynthetic gene cluster prediction and sequence similarity network (SSN) generation Secondary metabolite biosynthetic gene clusters were predicted using antiSMASH 7.0.0 on a local server for 5 antagonistic and 6 non-antagonistic Pseudomonas genome sequences. Sequence similarity networks (SSN) were constructed from the antiSMASH 7.0.0 output from the 11 antagonistic Pseudomonas genome sequences, using BiG-SCAPE version 1.1.525 and Pfam 3556 on a local server. A cut-off value of 0.3 was used, and 1,808 reference sequences from the MIBiG database were included in the construction of Gene Cluster Family (GCF) networks based on sequence similarity, with visualisations of SSNs generated and edited in Cytoscape 3.10.157. Regions containing predicted BGCs that formed a SSN with MIBiG reference BGCs for 2,4-diacetylphloroglucinol (BGC0000280 & BGC0000281) were aligned using clinker version 0.0.2358, before trimming to remove sequence upstream and downstream of the predicted BGCs. Colour schemes of plots were modified using Inkscape version 1.3 (Available from: https://inkscape.org). Construction of phlD insertional mutant To generate Roth82 phlD::pKNOCK-Km an internal fragment (333 bp) of the phlD gene was amplified by PCR using primers (pKNOCKPhlD3_Fw: GTCGGATCCACATCGTGCACCGGTTTCAT and pKNOCKPhlD3_Rv: GTCGGTACCGTAAGACCCGGTATTGGCG), verified by DNA sequencing and cloned as a BamHI-KpnI fragment into corresponding sites of pKNOCK-Km59, resulting in pKNOCKphlD. Electrocompetent Roth82 cells were prepared as described by Choi et al.,60. We added 300 ng of pKNOCKphlD DNA to the prepared electrocompetent cells and transferred to a pre-chilled 2 mm gap electroporation cuvette. The plasmid construct was delivered to electrocompetent Pseudomonas isolate Roth82 by electroporation (2.5 kV, 25 µF and 200 Ω). Immediately following electroporation, 1 mL of LB was added and cells recovered at 25 °C with 200 rpm shaking for 1 h, before plating. Transformants were then selected by plating 100 µL of the recovery culture on LB agar supplemented with kanamycin (50 µg/mL), before incubation at 25 °C for 18 h. Single colonies of putative transformants were purified on LB agar supplemented with kanamycin (50 µg/mL) prior to PCR confirmation. The mutant Roth82 phlD::pKNOCK-Km was confirmed by colony PCR using external primers flanking the insertion region (pKnockPhlD2_Fw: ATGATGCCCTCGCTGAC, pKnockPhlD2_Rv: CCGGGTTCCAAGTCCAGT, pKNOCKcheckFw: GGACAACAAGCCAGGGATGTA and pKNOCKcheckRv: ATGTAAGCCCACTGCAAGCTA), resulting in the expected size shift (wild-type 527 bp; mutant 2098 bp, Figure S6), and by Sanger sequencing at Eurofins (Germany). Chemical extraction of secondary metabolites from Z. tritici antagonism assay A 10 mm diameter section of agar surrounding the Pseudomonas colony edge was excised from wild-type Roth82 and mutant Roth82 phlD::pKNOCK-Km grown with and without Z. tritici confrontation on PDA plates. Chemical extracts were prepared by macerating agar sections in 1 ml HPLC grade MeOH, with resulting solutions centrifuged at > 13,000 rpm for 10 min. Supernatant was collected for analysis by liquid chromatography-mass spectrometry (LC-MS). Chemical analysis of secreted secondary metabolites produced by Pseudomonas isolates UPLC was conducted on agar extracts using an ACQUITY UPLC® system equipped with an ACQUITY BEH C18 (1.7 μm, 2.1 × 50 mm) analytical column (Waters, USA). Extracts were eluted over 38 min from 5% to 95% organic (UPLC grade water + 0.1% formic acid and MeOH + 0.1% formic acid) at a flow rate of 0.21 mL min− 1. The UPLC system was coupled with quadrupole time-of-flight mass spectrometer SYNAPT G2-Si (Waters, USA), with electrospray ion source run in negative and positive ionisation mode and data acquired over 50–1200 m/z range. The lock mass solution consisted of Leucine Enkephalin (200 pg/µL). The lockspray was acquired during UPLC-MS acquisition and corrected for ([M + H]+ m/z 556.2771 & [M-H]− m/z 554.2615). Instrument control and data acquisition was carried out in MassLynx (Waters, USA). Data was converted to mzML format using Waters2mzML V1.2.0 (available at https://github.com/AnP311/Waters2mzML), before Extracted Ion Chromatograms (EICs) were generated using mzmine 4.3.061. Statistical analyses Analysis of Variance (ANOVA) was used to determine significant differences in the size of zones of inhibition. Zone of inhibition data of Z. tritici isolates in confrontation with Pseudomonas isolate BFR20 were log transformed, to fulfil the requirement of analysis of variance (ANOVA) for data to be approximately normally distributed. Zone of inhibition data for all other antagonistic Pseudomonas isolates were found to be approximately normally distributed without transformation. Sidak’s multiple comparison post-hoc test was used to examine pairwise data comparisons between Z. tritici genotypes at the 5% level of significance. All statistical analyses were carried out in RStudio (2023.03.0 Build 386, R version 4.2.3), using packages ggplot2 (3.4.4)62, dplyr (1.1.4)63, ggpubr (0.6.0)64, emmeans (1.10.0)65, multcomp (1.4–25)66 and svglite (2.1.3)67. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 1 Supplementary Material 2 Supplementary Material 2 Supplementary Material 3 Supplementary Material 3