Mining Actinomycetes' Metabolomes and Genomes for Anti‐Phytophthora infestans Compounds

ABSTRACT

Actinomycetes are well‐known for producing a diverse array of specialised metabolites with various bioactivities; yet, identifying metabolites with targeted activity against specific pathogens remains challenging. In this study, we employed a comparative metabolomic and genomic approach on 63 actinomycete strains differing in their ability to inhibit or alter the mycelial growth of Phytophthora infestans , the causal agent of potato late blight. This comparative approach efficiently pinpointed approximately 1000 mass spectrometry features linked to active extracts, out of 16,500 detected features. Our analysis putatively identified over 75 compounds with potential activity against P. infestans , including borrelidin, actinomycin D, antimycin A, macbecin I, myriocin and ikarugamycin. Our study shows that leveraging multi‐omics analysis of phylogenetically related strains with differential activity is a promising strategy which, combined with a relatively high throughput metabolite extraction method, advanced mass spectrometry and cutting‐edge tools for bacterial metabolite annotation and prediction, allowed a straightforward selection of interesting candidate compounds for the biological control of an important plant pathogen such as P. infestans . The methodology outlined here offers broader applicability for identifying bioactive compounds underlying any phenotype of interest, provided this phenotype varies in phylogenetically closely related strains.

Using a comparative metabolomic and genomic approach on 63 actinomycetes, this study identified 75 compounds associated with the strains' ability to inhibit the late blight‐causing agent Phytophthora infestans . This study highlighted the promise of applying multi‐omics techniques on closely related strains differing in biological activity to pinpoint the responsible metabolites.

1. Introduction

Actinomycetes are filamentous bacteria renowned for producing structurally diverse metabolites with remarkable biological activities (Barka et al. 2016; van Bergeijk et al. 2020; Alam et al. 2022). Despite extensive studies on these bacteria, they are still considered vast reservoirs of natural antibiotics (Takahashi and Nakashima 2018). Within this group, the genus Streptomyces, in particular, remains the primary source of clinically important antibiotics and other bioactive compounds (Barka et al. 2016; van Bergeijk et al. 2020). However, so‐called ‘rare actinomycetes’, which encompass all genera except Streptomyces, also possess significant potential in specialised metabolism (Parra et al. 2023) and have yielded major therapeutic compounds, such as erythromycin (Mcguire et al. 1952) and gentamicin (Weinstein et al. 1963).

The diversity of actinomycete specialised metabolites arises primarily from large biosynthetic gene clusters (BGCs) encoding modular enzymatic systems such as nonribosomal peptide synthetases (NRPS) and polyketide synthases (PKS) (Belknap et al. 2020; Kumar et al. 2025). Moreover, many actinomycete metabolites are synthesised by hybrid NRPS/PKS systems, which integrate both peptide and polyketide modules, leading to structurally complex natural products such as macrolactams and lipopeptides (Komaki et al. 2018). While these systems have been extensively explored for drug discovery, their potential in plant protection remains underexplored. Still, some compounds produced by actinomycetes have been reported to inhibit devastating plant pathogens, causing significant economic losses, for example oligomycins and actinomycin D, which inhibit Botrytis cinerea (Xiao et al. 2021; Yong et al. 2022; Louviot et al. 2024), and valinomycin, which suppresses several fungal pathogens, including Phytophthora capsici and Penicillium verrucosum (Huang et al. 2021).

Among oomycetes threatening plant health and food safety, Phytophthora infestans , the causative agent of late blight in potato and tomato, stands out as a particularly destructive pathogen that leads to significant economic losses to agriculture (Haverkort et al. 2009; Whisson et al. 2016; Wang and Long 2023). P. infestans is considered a highly threatening pathogen, capable of rapidly evolving new genotypes through sexual recombination, a trait that could contribute to severe epidemic outbreaks on a global scale (Fry et al. 2015). Thus, it is not surprising that new genotypes of the pathogen that are resistant to fungicides such as fluazinam and mandipropamid have arisen (Schepers et al. 2018; Abuley et al. 2023). Moreover, the recent withdrawal of heavily used fungicides like chlorothalonil in 2020 and mancozeb in 2022 underscores the threats they pose to human health and the environment (Ben Naim and Cohen 2023; Wang and Long 2023). Therefore, there is an urgent call for sustainable approaches in late blight management to minimise crop losses and mitigate the environmental impact of fungicide applications (Majeed et al. 2017). So far, late blight disease has been managed through the application of synthetic fungicides and copper‐based products, which are not sustainable solutions for food production, or for human and environmental health (Majeed et al. 2017; De Vrieze et al. 2018; Wang and Long 2023). Efforts have been made to exploit various bacterial genera for strains and compounds that serve as environmentally friendly P. infestans inhibitors, including Pseudomonas spp. (Tran et al. 2007; De Vrieze et al. 2015, 2020; Hunziker et al. 2015; Morrison et al. 2017; Anand et al. 2020; Biessy et al. 2021; Léger et al. 2021), Bacillus spp. (Sorokan et al. 2020; Wang et al. 2020; Alfiky et al. 2022), Lysobacter spp. (Puopolo et al. 2014; Lazazzara et al. 2017; Vlassi et al. 2020) and Myxococcus spp. (Wu et al. 2021). However, only a few actinomycetes‐derived compounds have been reported to inhibit P. infestans , such as fungichromin (Huang et al. 2007), piericidin A (Fu et al. 2022) and finally, borrelidin, which we recently identified using a comparative metabolomics and genomics study on as few as three closely related Streptomyces strains differing in anti‐Phytophthora activity (Gillon et al. 2023). This study provided the proof of concept that leveraging the strain's phylogenetic relatedness despite differing antagonistic capabilities can be an effective shortcut towards the identification of the metabolites underlying the phenotype of interest (Gillon et al. 2023).

Building upon the success of this proof of concept, in the present work, we applied such comparative metabolomic and genomic approaches to profile a collection of over 60 actinomycete strains previously characterised as having differing activities on the mycelial growth of P. infestans . Not only did these strains differ in the extent of the growth inhibition they caused, but they also induced specific disturbances in mycelial structure and development, which indicated different modes of action (Abdelrahman et al. 2022). Such diversified modes of action are a key element to ensure long‐lasting efficacy of chemical compounds used as crop‐protecting agents, as they would hinder the pathogen's resistance development. Therefore, we grouped the strains within our collection according to their differential effects on P. infestans and identified the metabolites and biosynthetic gene clusters commonly present in the strains displaying the respective effect but absent from the others, with the aim to discover new anti‐Phytophthora metabolites with differing modes of action for later implementation into sustainable crop protection strategies against late blight.

The 63 selected strains included 18 highly active strains that inhibited the mycelial growth of P. infestans by more than 80% in an in vitro dual assay, as previously described (Abdelrahman et al. 2022). Additionally, they included 21 active strains exhibiting 79%–50% growth inhibition, 14 moderately active strains with 49%–20% growth inhibition, and a group of relatively inactive strains exhibiting less than 19% growth inhibition, serving as the negative control for our comparative study. Regardless of their activity, this set of strains included five strains inducing P. infestans altered morphology 1, five inducing morphology 2, seven inducing morphology 3, and finally two strains inducing the mixed of morphology 1 and morphology 3 categorised in morphology 4 (Abdelrahman et al. 2022).

2. Experimental Procedures

2.1. Full Genome Sequencing and antiSMASH Analysis

The genomic DNA of 59 out of the 63 selected actinomycete strains was extracted following the procedure outlined by Gillon et al. (2023). Four strains (B50, B53, B93 and B131) were excluded from further analysis due to poor DNA quality or contamination. Subsequently, the genomic DNA of the 59 strains underwent processing by the Next Generation Sequencing (NGS) Platform at the University of Bern according to the protocols and kits of Pacific Biosciences library preparation. The libraries were multiplexed on 1 x SMRTcell 25M on a Revio. After sequencing HiFi long reads, the demultiplexing was performed with lima, the HiFi reads were quality controlled with FastQC and assembled using version 13.0.0.207600 of the SMRTlink pb_microbial_analysis pipeline. The assembly's polishing, circularization and origin rotation were provided by the SMRTlink pipeline if required. The genomes were annotated with Prokka (v.1.14.5) (Seemann 2014) and antiSMASH (v.7.1) (Blin et al. 2023).

The sequence data have been submitted to the ENA database under projects PRJEB59914 and PRJEB67372. The accession numbers are in Table S1.

2.2. Solid–Liquid Extraction of Actinomycetes Metabolites by Sonication

The metabolites from each actinomycete strain were extracted from solid V8 plates using ethyl acetate. V8 medium was prepared by adding 1 g of CaCO₃ (BioXtra ≥ 99%; Sigma‐Aldrich), 15 g of agar (Carl Roth), and 100 mL of V8 100% spicy hot vegetable juice (Campbell Soup Company) to 1 L of distilled water (Miller 1955). A 100 μL of each strain's spore suspension (OD₆₀₀ = 1) was evenly spread on a V8 plate and incubated at 28°C for 1 week. A non‐inoculated V8 medium control plate was also incubated for the same duration. To capture biological variability, metabolite extraction involved pooling half a plate from two plates, resulting in one plate per strain or control. The two halves of the plates were cut into ∼0.5 cm² pieces for extraction. These agar pieces were placed into two Pyrex 30 mL glass tubes, to which 15 mL of ethyl acetate (Thommen‐Furler AG) was added. The mixture underwent sonication for 30 min, and this step was repeated twice. The extracts from each sample were combined and filtered through filter paper (Folded filters 4 V; Cytiva) in a pre‐weighed glass tube. Subsequently, the solution was evaporated using a nitrogen evaporator (Pierce Reacti‐Therm III), and the tube was refilled with the respective extract, repeating the evaporation until the total volume of the obtained extract (~60 mL) was evaporated. The resulting dried crude extract was weighed and then dissolved in 99.9% methanol (Thommen‐Furler AG, Switzerland) to obtain a final 5 μg/μL concentration. Extracts were stored at −20°C until further use or analysis.

2.3. Assessing the Activity of the Extracted Metabolites on P. infestans Growth

The activity of the extracted metabolites against P. infestans was determined using the disk diffusion method following the procedure outlined by Gillon et al. (2023). Organic extracts, dissolved in methanol at a concentration of 5 μg/μL for each actinomycete strain, were applied to 6 mm diameter Whatman paper disks (Whatman AA disks; Cytiva) resulting in a final amount of 200 μg on each disk. Methanol‐spotted and dried disks were used as controls. These disks were placed on V8‐medium plates, positioned at a distance of 2 cm from the plate border with a 5 mm diameter P. infestans plug positioned at a distance of 2.5 cm from the border of the plate.

2.4. Liquid Chromatography–Mass Spectrometry (LC–MS) Analysis and Data Processing

2.4.1. LC–MS Parameters

Chromatographic separation was conducted using a Vanquish Flex UPLC system (Thermo Scientific, Bremen, Germany) interfaced to a Q‐Exactive Plus mass spectrometer (Thermo Scientific, Bremen, Germany), using a heated electrospray ionisation (HESI‐II) source. Instrument control was facilitated through Thermo Scientific Xcalibur 3.1 software. The LC conditions were as follows: column, Waters BEH C18 100 × 2.1 mm, 1.7 μm; mobile phase, (A) water with 0.1% formic acid; (B) acetonitrile with 0.1% formic acid; flow rate, 600 μL/min; injection volume, 2 μL; gradient, linear gradient of 2%–100% B over 10 min and isocratic at 100% B for 2 min. The optimised HESI‐II parameters were as follows: source voltage, 3.5 kV (pos); sheath gas flow rate (N₂), 55 units; auxiliary gas flow rate, 15 units; spare gas flow rate, 3.0; capillary temperature, 350°C, S‐Lens RF Level, 50. The mass analyser was calibrated using a mixture of caffeine, methionine–arginine–phenylalanine–alanine–acetate (MRFA), sodium dodecyl sulphate, sodium taurocholate and Ultramark 1621 in an acetonitrile/methanol/water solution containing 1% formic acid by direct injection. The data‐dependent MS/MS events were performed on the three most intense ions detected in full scan MS (Top3 experiment). The MS/MS isolation window width was 1 Da, and the stepped normalised collision energy (NCE) was set to 15, 30 and 45 units. In data‐dependent MS/MS experiments, full scans were acquired at a resolution of 35,000 Full Width at Half Maximum (FWHM) (at m/z 200) and MS/MS scans at 17,500 FWHM both with an automatically determined maximum injection time. After being acquired in an MS/MS scan, parent ions were placed in a dynamic exclusion list for 3.0 s.

2.4.2. Data‐Processing and Molecular Networking

MS data were converted from .RAW (Thermo) standard data format to .mzML format using MSConvert software from the ProteoWizard package (Chambers et al. 2012). MZmine software suite v.2.53 (Pluskal et al. 2010) was used for data treatment with the following parameter adjustments: the centroid mass detector was used for mass detection with the noise level set to 1.0E5 for MS level set to 1, and to 0 for MS level set to 2. The ADAP chromatogram builder was used and set to a minimum group size of scans of 5, a minimum group intensity threshold of 1.0E5, a minimum highest intensity of 1.0E5 and an m/z tolerance of 12 ppm (Myers et al. 2017). For chromatogram deconvolution, the wavelets algorithm (ADAP) was applied. The intensity window S/N was used as the S/N estimator with a signal‐to‐noise ratio set at 15, a minimum feature height at 1.0E5, a coefficient area threshold at 80, a peak duration range from 0.02 to 1.00 min and the retention time (RT) wavelet range from 0.02 to 0.05 min. Corresponding MS2 were paired with the following parameters: 0.025 (Da) and 0.15 (min). Isotopes were detected using the isotopes peaks grouper with an m/z tolerance of 8 ppm, an RT tolerance of 0.08 min (absolute), the maximum charge set at 4, and the lowest m/z was used as the representative isotope. The resulting deisotoped files were filtered using the feature list rows filter method to obtain one minimum peak in raw and two minimum peaks in isotope pattern, keeping only peaks with MS2 and resetting the peak number ID. Peak alignment was conducted using the join aligner method (m/z tolerance at 12 ppm), an absolute RT tolerance at 0.08 min, a weight for m/z at 30, and a weight for RT at 30. The aligned feature list was exported using the export to GNPS FBMN module.

To assess the spectral diversity within the profiled dataset, a molecular network (MN) was generated using the Feature‐Based Molecular Networking (FBMN) workflow on the Global Natural Product Social Molecular Networking (GNPS) website (http://gnps.ucsd.edu) using the .mgf spectra file generated at the previous step (Wang et al. 2016). The precursor ion mass tolerance was set to 0.02 Da and a MS/MS fragment ion tolerance of 0.02 Da. The network was created with edges filtered to have a cosine score above 0.7 and more than 6 matched peaks. Additionally, edges between two nodes were retained in the network only if each of the nodes appeared in each other's respective top 10 most similar nodes. Finally, the maximum size of a spectral family was set to 100, and the lowest scoring edges were removed from molecular families until the molecular family size was below this threshold. The spectra in the network were then searched against GNPS spectral libraries. All matches kept between network spectra and library spectra were required to have a score above 0.7 and at least 6 matched peaks.

2.5. Metabolite Annotation

The entire spectral dataset corresponding to the extract collection was uploaded to the MassIVE repository, enabling an automated identification workflow against the GNPS experimental spectral libraries. Additionally, a taxonomically informed metabolite annotation was conducted using a theoretical spectral database of natural products (built from the LOTUS resource) associated with the biological sources of these products (Allard et al. 2016, 2022; Rutz et al. 2019, 2022). This annotation was executed using the met_annot_enhancer scripts version v0.1: https://github.com/mandelbrot‐project/met_annot_enhancer/releases/tag/v0.1. Finally, Sirius (Dührkop et al. 2019) (v.5.5.7) and CANOPUS (Djoumbou Feunang et al. 2016; Dührkop et al. 2021; Kim et al. 2021) annotation workflows were also employed to achieve metabolite annotation and the assignment of chemical classes to MS/MS spectra within the dataset. In this study, we focused our annotation discussion on selected clusters containing the most intense features, demonstrating a summed precursor intensity of ≥ 1E+07, and composed of three or more nodes. Additionally, some features that did not meet the selection criteria were exceptionally included if they corresponded to metabolites whose biosynthetic gene clusters were encoded in the genome of the respective strain.

The converted mass spectrometry data files, the corresponding metadata table, and the metabolite annotation results table, together with a Cytoscape file corresponding to the full molecular network annotated with the experimental and theoretical spectral matches are available under the following MassIVE id: MSV000095751.

2.6. Validating the Activity of Selected Putatively Annotated Compounds

The inhibitory efficacy of different concentrations of the purchased standards for putatively annotated active compounds was assessed against the mycelial growth of P. infestans using a disk diffusion assay. This assay aimed to investigate the contribution of these compounds to the growth inhibition induced by the respective putatively producing strains. The activity of the following commercially available pure standards was assessed: acetaminophen, actinomycin D, antimycin A, valinomycin (Merck), ikarugamycin, macbecin I, myriocin (Cayman), indole‐3‐carboxylic acid, ochromycinone (Brunschwig), piperafizine B, resistomycin (Adipogen) and borrelidin (Apollo Scientific). The standard compounds were dissolved in methanol at a concentration of 1 mg/mL, and different volumes of each solution were added to 6 mm diameter Whatman paper disks to obtain final amounts of 0.5, 1, 5, 10, 50 and 100 μg per disk and the disks were allowed to dry. On each plate, a Whatman paper disk containing pure methanol was included as a control. The experiment included duplicates, and the average inhibition percentage was calculated for each compound concentration. The half‐maximal inhibitory concentration (IC₅₀) of each active compound was obtained using a linear regression equation.

Additionally, the inhibitory effect of these compounds on P. infestans zoospore germination was tested following the method described by De Vrieze et al. (2020). However, in this study, the assay was conducted in 96‐well plates (Corning Incorporated costar) instead of 24‐well plates. Consequently, different volumes of zoospore suspension and treatments were used. Each well contained a total of 140 μL, comprising 40 μL of zoospore suspension (4 × 10⁴ zoospores/mL), varying volumes of each compound (dissolved in methanol at 100 μg/mL) to reach final concentrations of 0.5, 1, 5 and 10 μg/mL, and distilled water to complete the volume. Two controls were included: a methanol control and a water control. Plates were incubated at 18°C for 4 h. Images were captured at 10× magnification using a Cytation5 plate reader (BioTek, United States) at the end of the incubation period. The average germ tube length of zoospores was measured for each treatment relative to the water control. This experiment was performed in five technical replicates across two independent biological replicates.

3. Results and Discussion

3.1. The Actinomycete Collection Reveals Diverse Metabolomes With Differing Activity on P. infestans Growth

Ethyl acetate extracts were obtained from the 63 actinomycete strains whose genome sequencing revealed that they belonged to the Nocardiopsis (4), Saccharothrix (1) and Streptomyces (58) genera (Figure 1, Table S2). These extracts, together with that of the non‐inoculated V8 medium, were tested against the mycelial growth of P. infestans to evaluate their potential for inhibiting the pathogen and, more importantly, to determine if they contained the molecules responsible for the inhibition we formerly observed with the respective strains (Abdelrahman et al. 2022). This revealed differing inhibition capacities, which we categorised into five groups, from very active to inactive extracts (Figure 1A,B, Table S3).

FIGURE 1.

Genomic relatedness and inhibitory effects of actinomycetes on Phytophthora infestans . (A) Phylogenetic tree based on Average Nucleotide Identity analysis (ANI) of 59 actinomycete strains. The ANI was calculated with FastANI version 1.33, while the tree was visualised using iTOL version 6. The extract inhibition activity was represented on the top of each strain by colour square following the efficiency and the strain inhibition data (inner ring) were taken from a previous study (Abdelrahman et al. 2022). (B) Representative pictures of the five categories of the actinomycete extracts' inhibitory activity against the mycelial growth of P. infestans . A dried methanol disk was used as a negative control.

As illustrated in Figure 1, the metabolites extracted from some active strains exhibited less pronounced activity against P. infestans (outer ring) compared to that of their producing strains (inner ring). This observation aligns with our previous finding that the Streptomyces strain B135 demonstrated inhibitory activity against the examined pathogens when grown in co‐cultivation assays but not when its extracted metabolites were applied (Gillon et al. 2023). Obtaining inactive extracts from active strains suggests a potential limitation in our extraction protocol, possibly due to the proteinic nature or insolubility of the active molecules in our extraction solvent. Alternatively, the observed lack of activity may be explained by the fact that the active metabolites might have been induced during the interaction with the pathogen, which was not present when extracting the strains' metabolites.

Such induction of specialised metabolism has been shown in previous studies reporting that co‐cultivation of different actinomycete strains with various microorganisms can modify their specialised metabolism and uncover new biosynthetic pathways and secreted metabolites (Schroeckh et al. 2009; Onaka et al. 2011; Traxler et al. 2013; Wu et al. 2015; Sung et al. 2017).

To profile the metabolomes contained in these extracts, we used ultrahigh‐performance Liquid Chromatography coupled with high‐resolution tandem Mass Spectrometry (LC–MS). The resulting total ion chromatograms revealed numerous peaks that were unique to each specific strain, shared among certain strains within the collection, or common across the entire collection (MassIVE id MSV000095751). Given that the collection contained three different genera and multiple taxonomic clusters within the Streptomyces genus (Figure 1A, Figure S1 and Table S2), differences were anticipated both across genera and species. To better visualise variations in the chemical profiles among the strains, the acquired LC–MS data were utilised to construct a molecular network with the Feature‐Based Molecular Networking (FBMN) workflow (Nothias et al. 2020).

A massive multi‐informative molecular networking and bioactivity‐based molecular networking approach was employed to prioritise the metabolites of interest (Kurita et al. 2015; Olivon et al. 2017; Nothias et al. 2018). This approach aimed to unveil both commonalities and variations in the metabolomic profiles among the strains, potentially contributing to the observed differences in their respective activities. The dataset comprised a total of 94,237 spectra, categorised into 16,508 nodes (MS2 feature, including the medium and extraction blank features). Among these nodes, 10,319 were assembled into 1647 distinct clusters, while the remaining 6189 nodes were designated as singletons. The FBMN is publicly accessible through this link: https://gnps.ucsd.edu/ProteoSAFe/status.jsp?task=b75553771be84f53a7e54e1a4b44d718.

Among the 1647 feature clusters, we focused on those exclusive or highly abundant in extracts that exhibited a specific mycelial growth inhibition or that were produced by strains inducing a specific disturbance in that mycelial growth. Consequently, the FBMN was mapped and analysed based on two distinct criteria: extract activity and strain‐induced growth disturbances.

3.2. Metabolic Clusters Associated With Extract Activity

The applied comparative metabolomic approach is based on the hypothesis that features detected in active samples but not in inactive samples may be potential compounds contributing to the observed activity. Consequently, we mapped the FBMN based on categories of extract activity rather than strain activity. However, one should keep in mind that some compounds, even if present, may not ionise effectively and, therefore, remain undetectable with LC–MS.

Within the entire FBMN, 28 clusters and 2 singletons were identified as features that were exclusively present or highly abundant in one or more of the five extracts categorised as very active, 50 clusters in one or more of the nine active extracts, and 66 clusters in one or more of the 10 moderately active extracts, as depicted in Figure S2. Our attention was directed towards annotating the features associated with the very active and active extract categories, as the final goal of this analysis was to find promising anti‐ P. infestans compounds. The full annotation of these features is detailed in Table S4.

3.2.1. Metabolite Clusters Associated With Highly Active and Active Extracts

Among the 28 clusters associated with the very active extracts from strains B5, B19, B57, B10 and B113 (Figure 2A,B), one cluster was exclusively present in the extract of the Streptomyces strain B5 (cluster 232 in green in Figure 2B). This cluster was annotated as borrelidin and analogues, an annotation which was expected, given the presence of the borrelidin BGC in the genome of B5 and given our earlier validation of the borrelidin annotation by LC–MS analysis of a purchased standard (Gillon et al. 2023). Borrelidin was previously identified as an anti‐oomycete agent and reported to be a strong inhibitor of Phytophthora sojae (Gao et al. 2012), P. capsici (Chen et al. 2016) and P. infestans (Gillon et al. 2023). Eleven clusters and one singleton were associated with the Streptomyces strain B19 (clusters in white in Figure 2B), with 10 of them putatively annotated, as detailed in Table S4: Sheet 1. Streptomyces strain B19 was identified as an Streptomyces albus strain (Table S2), a species known for its potential to produce multiple alkaloids (Hahn et al. 2009; Estévez et al. 2018; Rodríguez Estévez et al. 2020). This aligns with our observation, as 8 out of the 10 annotated unique clusters in the B19 extract were putatively identified as alkaloids. Furthermore, two compounds produced by strain B19 were annotated following an antiSMASH analysis of its genome, revealing the BGC for dudomycin A (detected in cluster 409) and for xantholipin (detected in the singleton feature 5723). Eight clusters were associated with Streptomyces strain B57, depicted in yellow in Figure 2B and annotated in Table S4: Sheet 1. Finally, eight clusters and one singleton were associated with the extracts of the two Streptomyces strains B108 and B113, and were shared between the two strains, except for cluster 229. This cluster, putatively annotated as linear polyketides, was much more abundant in strain B108 than in strain B113. AntiSMASH analysis identified one more compound produced by both strains, 2‐hydroxyethyl clavam (detected in the singleton feature 11,902) (Table S4: Sheet 1).

FIGURE 2.

Inhibitory effects of very active and active actinomycete extracts against Phytophthora infestans and associated metabolite clusters. (A, C) Representative pictures of the inhibitory activity of bacterial strains (A) B5, B19, B57, B108 and B113, and (C) B1, B23, B50, B52, B93, B111, B123, B171 and B182, along with their extracted metabolites, against the mycelial growth of P. infestans . (B, D) Extracted molecular networks displaying (B) 28 and (D) 50 clusters composed of features that were either exclusive to or highly abundant in extracts categorised as very active (B) or active (D). Cluster IDs (component IDs) are indicated above each cluster. Node colours represent corresponding samples, node sizes indicate summed precursor intensity, and pie charts within nodes depict the relative abundance of each feature across different samples. The category QCs in black represent the quality control samples, which consist of a pooled mixture of all the extracts. The category Others in grey includes extracts classified as moderately, minimally active and inactive.

It is worth noting that among these putatively annotated compounds detected in the highly active extracts, only borrelidin had previously been identified as having anti‐ P. infestans properties (Gillon et al. 2023). However, several other compounds, primarily produced by actinomycetes, have been reported to possess diverse biological activities. For instance, antibacterial properties have been attributed to a number of these compounds, including albonoursin (Fukushima et al. 1973; Lautru et al. 2002), actinomycin D (Wang et al. 2017), xantholipin (Wu et al. 2017), cyclophostin (Nguyen et al. 2017), antibiotic YC 17 (also known as 10‐deoxymethymycin) (Kinumaki and Suzuki 1972; Pilli et al. 1998) and Antibiotic A 83586C (Smitka et al. 1988; Nakagawa et al. 1990). Actinomycin D has also been extensively documented for its antitumor activity and is widely used in clinical practice (Koba and Konopa 2005; Wang et al. 2017). In addition, androsterone has been reported to exhibit potential anticancer properties (Chen et al. 2018). Anti‐inflammatory effects have been associated with lansai C (Taechowisan et al. 2010), while oxopropaline G has demonstrated cytocidal activity (Abe et al. 1993). Other notable biological effects include the insecticidal activity of cyclophostin (Kurokawa et al. 1993) and the inhibition of morphological differentiation in actinobacteria by annimycin (Kalan et al. 2013), with one annimycin derivative also exhibiting antimalarial activity (Zhang et al. 2017). Importantly, and directly relevant to our study, actinomycin D has been shown to possess antifungal activity against B. cinerea, Fusarium oxysporum and Verticillium dahliae (Toumatia et al. 2015; Zeng et al. 2019; Yong et al. 2022). Furthermore, some putatively annotated compounds in our dataset have not yet been found to be produced by actinomycetes, but by other organisms. For example, mycestericin C, produced by Mycelia sterilia, is recognised for its potent immunosuppressive activity (Sasaki et al. 1994), while crambescidin 826, isolated from the marine sponge Monanchora sp., was reported to inhibit HIV‐1 envelope‐mediated fusion in vitro (Chang et al. 2003).

Beyond the highly active extracts described above, 50 clusters of features were identified in extracts from the strains B1, B23, B50, B52, B93, B111, B123, B171 and B182, which exhibited significant, although less strong activity against the mycelial growth of P. infestans (Figure 2C,D, Table S4: Sheet 2).

Strains B111 and B182 are notable for producing melanoid pigments, and they share a close genetic relationship with other strains (B170, B175, B178, B180 and B181), which were observed to exhibit similar metabolite profiles as B111 and B182, but with much lower intensities. Despite these similarities, the extracts from these related strains did not demonstrate the same level of activity against the mycelial growth of P. infestans . However, when co‐cultured with the pathogen, they demonstrated high activity (Abdelrahman et al. 2022). This suggests that the metabolite clusters identified in strains B111 and B182 may also contribute to the activity observed in their closely related strains, but these compounds might not have been extracted in sufficient concentrations in the latter strains to classify their extracts as active.

Interestingly, no single cluster was unique to the extract of S. werraensis strain B23, despite its observed activity, which might be due to the fact that the overall metabolome of this strain was very similar to other strains (B13, B28, B32 and B60), whose extracts were categorised as moderately active or minimally active as indicated in the PCA generated by Qiime 2 view (Bolyen et al. 2019) in Figure S3. These strains are closely related, and they belong to the same species, according to their full genome homology (Figure 1A, Figure S1 and Table S2). This probably suggests, again, that the compounds responsible for the observed activity of B23's extract were not produced or extracted in sufficient concentration in strains B13, B28, B32 and B60 to classify their extracts as active.

In the category of active extracts, none of the putatively annotated compounds, except surfactins, have been previously reported to demonstrate anti‐ P. infestans activity. Surfactins from various Bacillus spp. have been noted for their effectiveness against P. infestans (Wang et al. 2020; Alfiky et al. 2022), along with other beneficial properties such as antitumor, antibacterial, antifungal, antiviral, anti‐mycoplasma and insecticidal activities (Meena and Kanwar 2015; Chen et al. 2022; Zhen et al. 2023). In this study, Nocardiopsis dassonvillei strain B123 was found to produce different surfactin analogues that might account for the activity of this strain. Among the putatively annotated metabolites in this category, several compounds exhibit notable antimicrobial properties. Christolane B has demonstrated mild antibacterial activity against S. albus (Gómez et al. 2012), while streptoverticillinone has shown inhibitory effects against the pathogenic oomycete Peronophythora litchii (Feng et al. 2007). Collinomycin (also known as α‐rubromycin) displays both antimicrobial and anticancer activities (Lin et al. 2022), and neoenactin M2 has been reported to demonstrate antifungal activity (Roy et al. 1987; Mazu et al. 2016). Additionally, several well‐characterised antibiotics were identified in this group, each known for their broad and potent bioactivities. Valinomycin, a potassium‐selective ionophore, has been reported to exhibit antifungal, antiviral, insecticidal and antitumor effects (Daniele and Holiant 1976; Huang et al. 2021). Ikarugamycin, a polycyclic tetramate macrolactam, is recognised for its anticancer, antibacterial and antiprotozoal properties (Minamidate et al. 2021; Saeed et al. 2021; Zhang et al. 2023). Macbecin, belonging to the ansamycin class of antibiotics, has been reported to have antitumor and antimicrobial properties (Muroi et al. 1980; Tanida et al. 1980).

However, despite the various activities reported for the compounds putatively detected in the very active and active extracts, their potential efficacy against P. infestans remains unknown, and further investigation is necessary to validate their identification and to understand their contribution to the observed activity of the extracts and their producing strains. To address this point, we later validated some of the acquired annotations by comparing our putative compounds to purchased standards. Additionally, we evaluated the activity of these against both mycelial growth and spore development of P. infestans (see below).

3.3. Metabolic Clusters Associated With Disturbances in P. infestans Mycelial Growth

In addition to analysing clusters associated with active extracts, we examined the dataset concerning the altered morphologies of P. infestans induced by specific strains in our collection. This mapping disregarded the activity levels exhibited by the strains causing morphological disturbances and by their corresponding extracts. The specific aim of this analysis was to identify metabolites inhibiting the growth of P. infestans through distinct modes of action, regardless of the extent of their growth inhibition.

Four distinct P. infestans mycelium morphologies were induced by certain groups of strains as reported previously (Abdelrahman et al. 2022). Morphology one involved the growth of P. infestans under the agar without any visible sporangia formation under microscopic examination; morphology two was characterised by the production of fewer and thinner hyphae by the oomycete, resulting in a faint growth appearance; morphology three exhibited clusters of thick‐walled hyphae, leading to a bushy or ‘frozen’ appearance; and finally, morphology four displayed a mixed appearance, combining features of morphologies one (growth under the agar) and three (‘frozen’ morphology). Representative pictures of the four distinct mycelial morphologies induced by the strains are shown in Figure S4.

Our focus was on clusters comprising features common among the extracts from each group of strains causing the respective morphology, as well as unique clusters specific to each strain within those groups. However, extra caution in interpreting the following results is required, as, on one hand, closely related strains mostly induce the same morphology, making it challenging to distinguish between compounds associated with their core metabolome and those responsible for their activity. On the other hand, most extracts from these strains did not exhibit the same activity as their respective strains. As discussed above for the extract activity, this could be due to these compounds being produced only in the presence of P. infestans or to an unsuccessful extraction method. Nevertheless, some extracts did induce the same morphology as their respective strains, albeit to a lesser extent, such as the extract from strains B123 (Figure 3A) and the extracts from strains B50, B52 and B93 (Figure 3C) inducing P. infestans morphology one and morphology two, respectively.

FIGURE 3.

Inhibitory activity of actinomycete strains and extracts altering Phytophthora infestans mycelial morphology and associated metabolite clusters. (A, C, E, G) Representative images of the inhibitory activity of bacterial strains inducing different P. infestans morphological alterations when preincubated and co‐inoculated with the pathogen, along with the activity of their extracted metabolites against P. infestans mycelial growth. Strains include (A) B56, B66, B67, B114 and B123 (morphology one), (C) B50, B51, B52, B53 and B93 (morphology two), (E) B25, B37, B38, B99 and B131 (morphology three; B23 shown in Figure 2C), and (G) B129 and B148 (morphology four, a mix of morphologies one and three). (B, D, F, H) Extracted molecular networks displaying (B) 13, (D) 30, (F) 37 and (H) 24 clusters composed of features that were either exclusive to or highly abundant in extracts from strains inducing the corresponding morphological changes. Cluster IDs (component IDs) are indicated above each cluster. Node colours represent corresponding samples, node sizes indicate summed precursor intensity, and pie charts within nodes depict the relative abundance of each feature across different samples. The category QCs in black represent the quality control samples, which consist of a pooled mixture of all the extracts. The category Others in grey includes extracts from strains that did not induce abnormal morphology of P. infestans .

Within the entire FBMN, 13 clusters comprised features that were exclusive to or highly abundant in the five extracts from strains inducing P. infestans morphology one (Figure S5A); 30 clusters were associated with morphology two (Figure S5B); 37 with morphology three (Figure S5C); and 24 clusters with morphology four, including clusters containing features shared between extracts from morphology one, three and four (Figure S5D). A full annotation of the most intense features (≥ 1E+07) included in these clusters is detailed in Table S5.

3.3.1. Putatively Annotated Compounds Detected in Extracts From Strains Causing Different Morphological Alterations in P. infestans Mycelial Growth

Nearly all the features assembled in the 13 clusters identified in the extracts from strains inducing morphology one (under‐agar growth) were found in those from N. dassonvillei strains B67, B114 and B123 (Figure 3A,B). Notably, extracts from strains B56 and B66, morphology one inducers, shared features with those from strains B129 and B148, which induced morphology four (a mix of morphology one and three), as illustrated in Figure 3H.

As mentioned earlier, clusters 117 and 1103 from strain B123 were tentatively identified as surfactins from Streptomyces according to GNPS (Tables S4 and S5). However, these surfactins were absent in the extracts of B67 and B114, which exhibited lower activity compared to the extract from strain B123, despite belonging to the same species and inducing the same morphology. This suggests a potential role of surfactins in the activity of these strains, given their absence in other N. dassonvillei strains that lacked such activity, together with the reported biological activity of surfactins on P. infestans (Wang et al. 2020; Alfiky et al. 2022). Apart from the 13 clusters in Figure 3B, extracts from strains B67 and B114 shared certain features in cluster 53 along with features exclusive to the highly active extract from S. albus strain B19 (the grey features in cluster 53 in Figure 2B). These features were putatively annotated as albonoursin. Notably, the albonoursin BGC was detected in the genomes of the three N. dassonvillei strains B67, B114 and B123 (highlighted in blue in Table S6).

When exploring clusters potentially underlying the activity of strains inducing morphology two (faint mycelial growth), we detected 30 (Figure 3D), of which the third comprised features annotated as Streptomyces angucyclines, such as different analogues of landomycinone, rabelomycin, tetrangomycin, ochromycinone, panglimycin and eMycin. However, other chemical classes were also detected in the extracts of these strains (Table S5: Sheet 2).

Despite the general lack of activity observed in most extracts from strains inducing morphology three (bushy/frozen hyphae), except for B5 (Figures 2A and 3G) and B23 (Figure 2C), we proceeded to analyse their associated metabolites. Our motivation was to investigate the potential presence of bioactive compounds in these extracts, considering the biological activities of their producing strains, and the possibility that these active compounds might have been extracted at insufficient concentrations to demonstrate observable antagonistic activity against P. infestans . A total of 37 clusters were associated with these strains, which were predominantly shared among the closely related strains B25, B37, B38, B99 and B131 (Table S2). Notably, there was no overlap between the clusters found in the extracts from this group of strains and those from B5 and B23, although they caused the same morphological alterations (Figure 3E,F, Table S5: Sheet 3).

Borrelidin was detected within this group of clusters as the unique compound of the highly active extract from Streptomyces sp. strain B5. Previous observations indicated that borrelidin induces P. infestans morphology three, suggesting that other strains causing morphology three likely produce distinct compounds, apart from borrelidin, to induce such a phenotype. Among the clusters unique to strains B25, B37, B38, B99 and B131, and absent in B5 and B23, eight clusters contained features annotated as phenazine alkaloids, and phenazines such as streptophenzine (L, K and G), phenazostatin B and saphenic acid methyl ester were detected in the extracts of these strains. This finding is not surprising given that these strains are characterised by an orange colour, a common trait of many phenazines. Furthermore, mining the genome of these strains with antiSMASH revealed two streptophenazines BGCs in strain B25 and one in strains B38 and B99 (highlighted in orange in Table S6). We anticipate that these phenazines likely contributed to the observed activity of these strains, as phenazines from Pseudomonas spp. have been previously reported to inhibit the growth of P. infestans (Morrison et al. 2017; Biessy et al. 2021; Léger et al. 2021).

Twenty‐two clusters and two singletons were identified to be associated with strains inducing morphology four, a mixed phenotype between morphologies one and three induced by Streptomyces strains B129 and B148 (Figure 3H, Table S4: Sheet 4). Notably, the metabolome of these two strains, especially strain B129, closely resembled that of strains B56 and B66, which induced morphology one. Moreover, the full genome analysis revealed that B56, B66 and B129 belong to the same Streptomyces species (Table S2).

Two compounds potentially responsible for causing morphology four were annotated through mining the genomes of strains B56, B66 (morphology one inducers), B129 and B148 (morphology four inducers) using antiSMASH analysis. Their BGCs were found to have a relatively high gene match percentage (highlighted in green in Table S6). The first compound, naphthyridinomycin, was identified in the genomes of both strains and detected in their extracts as well, appearing in clusters 618 and 936 along with the singleton feature 8975. The second compound, resistomycin, was unique to strain B148 and detected in the singleton feature 14,747. In addition to naphthyridinomycin and resistomycin, several compounds were annotated within the extracted network of strains inducing P. infestans morphology four, including antimycin (A20/A6a/A2/A4), phenatic acid A, actinoquinoline B and monacyclinone A. These compounds represent promising candidates for anti‐Phytophthora activity, although none have yet been reported to act specifically against oomycetes. However, antimycin A has been reported to exhibit significant antifungal activity, particularly against Rhizoctonia solani and Magnaporthe oryzae (Paul et al. 2022), while phenatic acid A has been shown to enhance the antifungal efficacy of miconazole against Candida albicans (Fukuda et al. 2005).

In summary, our metabolomic analysis unveiled several compounds potentially active against P. infestans . Employing two mapping strategies within our molecular network (extract activity and induced morphologies in P. infestans ), we identified compounds present in bioactive extracts and absent in inactive ones. These include borrelidin, albonoursin, oxopropaline G, lansai A /C, actinomycin D, annimycin, mycestericin C, antibiotic YC 17, antibiotic A 83586C, louisianin D, collinomycin, neoenactin M2, surfactins A /B /C /D, valinomycin, ikarugamycin and macbecin.

Moreover, our analysis suggests that surfactins might be responsible for P. infestans altered morphology one, along with compounds such as piperafizine B and marinopyrone D. Angucyclines might contribute to the induction of morphology two, alongside compounds like neoenactin M2 and idamycin. Phenazines could be linked to morphology three, while compounds like naphthyridinomycin, antimycin, phenatic acid A, actinoquinoline B and resistomycin, could be responsible for the induction of P. infestans altered morphology four. The chemical structures of selected putatively annotated compounds are shown in Figure S6.

Furthermore, our analysis uncovered additional metabolites with annotated chemical classes and structures, although lacking common names. These annotations warrant further investigation and validation processes, as they may harbour other, potentially new, bioactive molecules that could not be discussed in our result interpretation.

Importantly, the identification of these compounds was achieved through the combination of a relatively high‐throughput and relatively low‐cost metabolite extraction procedure, using as little as 60 mL ethyl acetate per strain, and of advanced community‐developed computational annotation workflows. This approach stands in contrast to traditional methods requiring extensive extraction of bacteria over longer periods, larger volumes of solvents and subsequent chromatography separation and analytical chemistry tools.

3.4. Validating the Annotation of Some Putatively Detected Compounds

Since the metabolomic analysis performed yields putative annotations derived from computational workflows and existing database knowledge, additional experimental validation is required to confirm the identification of each compound. One straightforward approach for validation is comparison with commercially available standards. Accordingly, we assessed the annotation of 11 commercially available pure standards. Detection and chemical data for these compounds are provided in Table S7. All purchased compound standards were purified from streptomycetes except for myriocin, which is known to be produced by Streptomyces and other organisms but is commercially available from the fungus M. sterilia.

The annotation of eight out of the 11 compounds was confirmed through LC–MS analysis of the pure standards, compared to the putative compound peaks in the extracts of their respective producing strains (blue circles in Figure S7). These compounds were actinomycin D, antimycin A, ikarugamycin, indole‐3‐carboxylic acid, ochromycinone, piperafizine B, resistomycin and valinomycin. Macbecin I and myriocin (orange circles in Figure S7) were detected with different adducts and retention times, suggesting that these strains might have produced different isomers of these standard compounds. This assumption, however, requires further investigation. The antimycin A standard consists of four components: antimycin A1, A2, A3 and A4, which were all detected in the strain extract (Figure S7).

3.5. Assessing the Activity of Selected Compounds Against Two P. infestans Developmental Stages

In addition to annotation validation, we aimed to assess the contribution of each annotated compound to its strain's activity against P. infestans , as well as to determine the concentration required to induce the observed inhibition. Therefore, we evaluated the activity of the 11 purchased compounds against two developmental stages of P. infestans . Borrelidin activity against P. infestans mycelial growth was evaluated in a previous study (Gillon et al. 2023).

The activity of the tested compounds on P. infestans mycelial growth was assessed using a disk diffusion method with varying concentrations. Antimycin A was the most active compound against P. infestans with a half maximal inhibitory concentration (IC₅₀) value of 64.7 μg, followed by actinomycin D (IC₅₀ = 81.9 μg), macbecin I (IC₅₀ = 105.4 μg), myriocin (IC₅₀ = 518.7 μg) and a very slight activity was exhibited by ikarugamycin (Figure 4). In contrast, the other six compounds were inactive at the tested concentrations, as shown in Figure S8. The inhibitory activity of antimycin A, actinomycin D, macbecin I and myriocin increased with higher concentrations, indicating a dose‐dependent effect, consistent with previous observations for borrelidin, although borrelidin exhibited significantly stronger activity against P. infestans mycelial growth with an IC₅₀ value of only 0.99 μg.

FIGURE 4.

Representative plates showing the growth inhibition of Phytophthora infestans in the presence of different amounts of the pure compounds actinomycin D, antimycin A, macbecin I, myriocin and ikarugamycin. The left disks are methanol blanks, and the right disks are the different concentrations tested. The experiment was performed once with two technical replicates.

Beyond mycelial growth inhibition, antimycin A, ikarugamycin and borrelidin compounds exhibited remarkable activity against zoospore germination, an epidemiologically relevant developmental stage of P. infestans (Tani and Judelson 2006; Kasteel et al. 2023). These three compounds inhibited the zoospore germination with IC₅₀ values below the tested concentration range (< 0.5 μg/mL), as shown in Figure 5 and Table S7. The remaining compounds showed varying degrees of inhibition on germ tube length, with IC₅₀ values ranging from 5.17 μg/mL (myriocin) to 14.44 μg/mL (acetaminophen) (Figure 5, Table S7).

FIGURE 5.

Inhibition activity of varying concentrations of standard compounds on Phytophthora infestans zoospore germination. (A) Representative pictures showing the inhibition of P. infestans zoospore germination in the presence of different concentrations of pure compounds. (B) Box plot depicting the percentage of P. infestans germ tube length relative to the water control for each standard compound at four different concentrations. Different letters indicate significant differences between treatments, as determined by a two‐way analysis of variance followed by Tukey's multiple comparisons test (p < 0.05).

It is worth noting that some compounds demonstrated differing levels of activity across the two developmental stages, suggesting distinct modes of action. However, these differences may also result from the distinct experimental setups, solid media versus liquid media, which affect compound diffusion rates. For example, ikarugamycin exhibited weak inhibition of mycelial growth, with only slight inhibition near the filter paper disk (Figure 4), but strongly inhibited zoospore germination in the 96‐well plate assay (Figure 5).

Several of the validated compounds have previously been reported to exhibit activity against various fungal pathogens. For instance, actinomycin D has demonstrated antifungal effects against B. cinerea, F. oxysporum and V. dahliae (Toumatia et al. 2015; Zeng et al. 2019; Yong et al. 2022), antimycin A against R. solani and M. oryzae, and ikarugamycin derivatives against Aspergillus fumigatus and C. albicans (Lacret et al. 2015). Similarly, myriocin has been reported to inhibit Fusarium graminearum (Shao et al. 2021), resistomycin against Valsa mali, Magnaporthe grisea (rice blast) and Pyricularia grisea (Zhang et al. 2013), and valinomycin against B. cinerea (Park et al. 2008). However, to the best of our knowledge, this study is the first to demonstrate the potent activity of these compounds against P. infestans , highlighting their potential significance for the control of late blight disease.

Beyond validating the potent anti‐ P. infestans activity of these compounds, whose activity could be validated using pure standards (in bold in Table 1), this study also highlights a significant set of 75 candidate molecules with potential applicability for managing late blight disease caused by the oomycete P. infestans . These candidate compounds, along with their associated phenotypes, are summarised in Table 1. Although some of the validated compounds exhibited varying levels of activity against mycelial growth (six active compounds) and zoospore germination (11 significantly different from the control), this represents a relatively small number compared to the overall set of candidate molecules. Therefore, the activity of non‐purchasable compound standards should be further validated following their purification or synthesis. Notably, several unvalidated compounds exhibit strong potential for anti‐ P. infestans activity based on their consistent detection across multiple highly active extracts or P. infestans morphology alteration categories. For example, 2‐hydroxyethyl clavam was uniquely detected in both the genome and metabolome of the two most active Streptomyces strains, B108 and B113. Albonoursin was identified in both the genome and metabolome of the highly active S. albus strain B19, as well as in two N. dassonvillei strains, B67 and B114 (morphology one). Additionally, naphthyridinomycin was detected in both the genome and metabolome of strains associated with morphologies one and four. Further investigations are warranted to confirm the structural annotation and evaluate the bioactivity of these compounds through both in vitro and in planta assays.

TABLE 1.

Candidate compounds associated with mycelial growth inhibition/altered morphology.

Compound	Associated with
	BGC	VA	A	M1	M2	M3	M4
11‐Deoxylandomycinone a					x
2‐Hydroxyethyl clavam a	x	x
8‐Cyanooctanoic acid			x
Acetaminophen a		x
Acrovestone			x
Actinomycin D a	x	x
Actinoquinoline B a				x			x
Albonoursin a	x	x		x
Androsterone a		x
Annimycin a	x	x
Antibiotic A 83586C a		x
Antibiotic X 14873A a						x
Antibiotic YC 17 a		x
Antimycin A a	x	x		x		x	x
Borrelidin a	x	x				x
Boseongazepine A a		x
Christolane B a			x
Clifednamide A			x
Colistin heptapeptide						x
Collinomycin a	x		x
Combamide C a			x
Crambescidin 826		x
Cyanophycin a			x
Cyclophostin a		x
Dehydrovomifoliol				x
Desoxyrabelomycin a					x
Destruxin A4/A5				x
Dudomycin A a	x	x
eMycin E/D a					x
Idamycin (Idarubicin) a					x
Ikarugamycin a	x		x
Indole‐3‐acetaldehyde a						x
Indole‐3‐carboxylic acid a						x
Istamycin X0 a					x
Landomycinone a					x
Lansai A/C a		x
Louisianin D a			x
Macbecin I a			x
Marinopyrone D a				x
Monacyclinone A a				x			x
Mycestericin C		x
Myriocin a				x			x
Naphthyridinomycin a	x			x			x
Neoenactin M2 a			x		x
N‐furfurylideneisobutylamine			x
Ochracenomicin C a					x
Ochromycinone a					x
Oxopropaline G a		x
Panglimycin A/C/E a					x
Phenatic acid A a				x			x
Phenazostatin B a						x
Phenylalanyl alanine a						x
Piperafizine B a				x
Rabelomycin a					x
Resistomycin a	x						x
Rhodopeptin C4 a						x
Rubiginone B1 a					x
Salinazinone A a					x
Saphenic acid methyl ester a						x
Schizopeptin 791						x
Semiplenamide D			x
Streptazone a			x
Streptophenzines L/K/G a	x					x
Streptopyridin B a			x
Streptoverticillin a			x
Streptoverticillinone a			x
Sulforaphane nitrile				x
Surfactins A/B/C/D a			x	x
Tariquidar				x
Tetrangomycin a					x
Thiomarinol H				x
Thiopheneethanamine				x
Valinomycin a	x		x
Xantholipin a	x	x

Note: Compounds in bold were tested as commercial standards.

Abbreviations: A, active extract; BGC, biosynthetic gene cluster detected in the genome; M, morphology alteration induced by the strains; VA, very active extract.

^a Compounds known to be produced by actinobacteria.

3.6. Sixty‐Eight Biosynthetic Gene Clusters Were Associated With Active Strains

Beyond the compounds detected in our metabolomic analysis, we analysed the genomes of our strains to identify BGCs encoding bioactive compounds that might not have been produced or successfully extracted. Beyond validating our metabolite annotations, analysing the strains' genomes might also reveal other metabolites associated with active strains but not detected with our metabolomic analysis. Leveraging our large number of samples, we aimed to exclude many biosynthetic gene clusters that are also present in inactive strains, facilitating the identification of distinct clusters of metabolites carried in the genomes of active strains. Around 300 BGCs were detected across the sequenced genomes (Table S8: Sheet 1). Of these, 68 BGCs were found in the genomes of active strains, showing more than 50% similarity, as illustrated in Table S8: Sheet 2.

Fourteen of these BGCs corresponded to compounds that were identified in the metabolomes of their respective producing strains, validating the chemical annotations: 2‐hydroxyethyl clavam, 4‐Z‐annimycin, 7‐prenylisatin, actinomycin D, albonoursin, antimycin, borrelidin, collinomycin, dudomycin A, ikarugamycin, naphthyridinomycin, resistomycin, streptophenazine and xantholipin (Table S8: Sheet 3).

Although we focused on BGCs with relatively high similarity, we cannot exclude the possibility that some BGCs with lower similarity also encode their respective compounds. For example, we detected and annotated valinomycin in the metabolome of strain B171, even though the corresponding BGC in the strain's genome had only 17% similarity to the valinomycin BGC in the database.

Based on the findings of this work, the next crucial steps involve validating as many potential active compounds as possible and assessing their activity against different developmental stages of P. infestans grown in vitro and in planta. Validation of annotations could be achieved by purchasing commercially available compounds, synthesising non‐commercially available ones, or isolating and purifying these compounds to elucidate their structure using NMR. Additionally, characterising biosynthetic gene clusters associated with active strains and examining their expression on different media is a possible approach to exploit the potential of these strains. Subsequently, once the bioactive compounds are identified, the next step would be to characterise their potential toxicity on plants and animal models, as ensuring safety in developing biocontrol agents is crucial. Compound activity and cytotoxicity would serve as prioritising criteria for further investigations. Moreover, considering each compound's mode of action, if reported, or investigating it if yet unreported, is essential for prioritising compounds suitable for further application as anti‐ P. infestans treatments, as compounds with specific, targeted activity would be expected to have less impact on non‐target organisms.

Additionally, the compounds identified in this study may hold potential as biocontrol agents beyond P. infestans , warranting further investigation against a broader spectrum of plant pathogens. To support such efforts, the associated metabolomic and genomic datasets generated here can be repurposed to pinpoint compounds exhibiting selective or enhanced activity against other agriculturally relevant pathogens.

4. Conclusion

In conclusion, this study establishes that a multi‐omics approach applied to phylogenetically related bacterial strains exhibiting differential bioactivity is a robust strategy for discovering biologically relevant metabolites. By leveraging the expectation that closely related strains share a common metabolite background, this strategy served as a practical shortcut for pinpointing the most promising active metabolites against our target pathogen. Coupled with a streamlined metabolite extraction protocol, high‐resolution mass spectrometry, and advanced computational tools for metabolite annotation and biosynthetic prediction, this framework enabled the identification of over 75 candidate compounds with potential activity against P. infestans . Importantly, this integrative methodology is broadly applicable for elucidating bioactive metabolite profiles associated with any phenotypic variation among closely related microbial strains.

Author Contributions

Ola Abdelrahman: conceptualization (equal), methodology (equal), data curation (lead), investigation (lead), validation (equal), formal analysis (lead), supervision (supporting), visualization (equal), writing – original draft (lead), writing – review and editing (equal). Quinn Coxon: methodology (supporting), data curation (equal), investigation (equal), visualization (equal), writing – review and editing (supporting). Eliane Abou‐Mansour: conceptualization (supporting), methodology (supporting), supervision (supporting), writing – review and editing (supporting). Floriane L'Haridon: supervision (supporting), project administration (supporting), resources (equal), writing – review and editing (supporting). Laurent Falquet: conceptualization (supporting), methodology (equal), software (lead), data curation (equal), validation (equal), supervision (supporting), resources (equal), writing – review and editing (supporting). Pierre‐Marie Allard: conceptualization (supporting), methodology (equal), software (lead), data curation (equal), investigation (supporting), validation (equal), supervision (equal), visualization (supporting), writing – original draft (supporting), writing – review and editing (equal). Laure Weisskopf: conceptualization (equal), investigation (equal), validation (equal), supervision (lead), funding acquisition (lead), project administration (lead), resources (equal), writing – original draft (supporting), writing – review and editing (lead).

Funding

This work was supported by the Swiss National Science Foundation (grant nos. 207917 and 229296 to Laure Weisskopf, grants nos. 10000895 and 10002786 to Pierre‐Marie Allard), by the research pool of the University of Fribourg (grant no. PO2422), and by swissuniversities (Swiss Open Research Data Grants (CHORD) in Open Science I).

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Figures S1–S8: mbt270269‐sup‐0001‐Figures.docx.

Table S1: mbt270269‐sup‐0002‐TableS1.xlsx.

Table S2: mbt270269‐sup‐0003‐TableS2.xlsx.

Table S3: mbt270269‐sup‐0004‐TableS3.xlsx.

Table S4: mbt270269‐sup‐0005‐TableS4.xlsx.

Table S5: mbt270269‐sup‐0006‐TableS5.xlsx.

Table S6: mbt270269‐sup‐0007‐TableS6.xlsx.

Table S7: mbt270269‐sup‐0008‐TableS7.xlsx.

Table S8: mbt270269‐sup‐0009‐TableS8.xlsx.

Abdelrahman, O. , Coxon Q., Abou‐Mansour E., et al. 2025. “Mining Actinomycetes' Metabolomes and Genomes for Anti‐Phytophthora infestans Compounds.” Microbial Biotechnology 18, no. 12: e70269. 10.1111/1751-7915.70269.

Data Availability Statement

All data required for this study is included in this paper, its Supporting Information and through the provided hyperlinks. The full spectral data set corresponding to the extract collection has been uploaded to the MassIVE repository under the following MassIVE id MSV000095751. The genome sequences of the 59 actinomycete strains can be accessed in the European Nucleotide Archive under the project accession numbers PRJEB59914 and PRJEB67372.