Potato foliar infection with Phytophthora infestans drives strong, cultivar-specific shifts in rhizosphere communities Authors: Vivien Pichon, Mout De Vrieze, Fares Bellameche, Rares Cristea, Floriane L’Haridon, Laurent Falquet, Laure Weisskopf Published: 2026 ABSTRACT AbstractBackgroundPotato is an important crop worldwide, yet its production is severely threatened by Phytophthora infestans, the causal agent of late blight. Alternatives to the current control strategies are needed, as these rely heavily on environmentally harmful treatments. The recruitment of beneficial microbes by plants upon stress (“cry-for-help” mechanism) may represent an opportunity to find new biocontrol agents but this has not yet been reported for potato. The aim of this study was to analyse whether foliar late blight infection induces shifts in the phyllosphere, rhizosphere and soil bacterial communities associated with two potato cultivars of differing sensitivity to late blight. Moreover, we aimed at isolating plant microbiota members to test whether bacteria putatively recruited upon infection would be particularly active in protecting the plant against late blight.ResultsControlled foliar infection triggered substantial, cultivar-specific shifts in rhizosphere communities across two successive generations. Despite the number of differentially abundant ASVs detected being ten times higher in the second generation than in the first one, the same taxonomic groups were involved in the shifts: Burkholderiales, Flavobacteriales, and Bacillales. Furthermore, the communities linked to the susceptible cultivar consistently shifted more strongly upon infection than the communities linked to the resistant cultivar. The obtained ASV sequences were used to identify 163 corresponding isolates through sequence alignment. Their inhibition potential against P. infestans sporangia and zoospores was assessed through biological assays. These revealed the biocontrol potential of genera otherwise not yet known to inhibit phytopathogenic organisms, such as Advenella, Nocardioides and Phyllobacterium strains. Although we found no correlation between the relative abundance shift of the ASVs upon infection and the activity of the corresponding strains, we observed that the overall activity of strains isolated from the resistant cultivar was higher than that of the strains isolated from the susceptible one.ConclusionTaken together, the higher activity of the strains isolated from the resistant cultivar, along with its comparatively modest microbiome shifts upon infection suggest that the investigated resistant cultivar might harbour specific microbiota enriched in strains that are able to inhibit pathogen development and possibly contribute to its higher resistance against P. infestans.Supplementary InformationThe online version contains supplementary material available at 10.1186/s40793-026-00910-x. Background Originally domesticated about 10,000 years ago in the Andes (South America), potato (Solanum tuberosum L.) is the 6th most produced crop, accounting for 4% of the global agricultural production [1]. It is, however, an extremely challenging crop to cultivate due to its sensitivity to many diseases, one of the most devastating being late blight caused by Phytophthora infestans. Potato late blight is a major challenge for farmers because it is very hard to control due to its rapid infectious cycle and genetic plasticity [2]. Various synthetic fungicides are used in conventional agriculture whereas in some countries, copper-based products are used in organic farming [3]. The application of such treatments has heavy sanitary, environmental and energetic costs, while getting less efficient over time due to the emergence of fungicide resistance. Finding new protective strategies against late blight is therefore urgently needed. Microbial Biological Control Agents (mBCAs), which are microorganisms able to prevent the rise of pests or diseases, have a great potential as they could be environmentally benign in addition to being a source of novel active ingredients [4]. Current attempts to identify new mBCAs require long and costly screenings of isolated bacterial collections, which for practical reasons mostly rely on in vitro tests of their antagonistic activities against the pathogen(s) of interest [5–7]. In addition to these tests being time- and cost-intensive, strains selected under these conditions often fail to provide consistent protection in whole plant assays [8], or even in field experiments. This selection of efficient mBCAs could be considerably improved if researchers could be led directly to the taxa that are the most beneficial to the plants. Recent developments in rhizosphere microbial ecology suggest the existence of an exudate-mediated selection of beneficial microbes by the plants themselves when they are facing challenging conditions [9]. This mechanism, called the cry-for-help hypothesis, might give researchers the much-needed lead to finding more efficient mBCAs. Indeed, many plants have been shown to respond to various biotic and abiotic stresses by secreting specific root exudates into the rhizosphere. Some of these exudates have a direct effect on their target pathogenic organisms [10, 11], while others rather seem to recruit specific members of rhizosphere bacterial communities. Red clover for example, when facing iron deficiency, has been shown to secrete phenolic compounds, which recruited rhizosphere bacteria able to produce siderophores [12]. Arabidopsis thaliana was reported to secrete L-malic acid upon infection with Pseudomonas syringae, and this carboxylate was shown to recruit the plant growth promoting rhizobacterium (PGPR) Bacillus subtilis F017 [13]. Cucumber was found to secrete tryptophan when infected with Fusarium oxysporum, which led to the increased colonisation of the PGPR Bacillus amyloliquefaciens SQR9 [14]. When they were infected with the same pathogen, tomatoes secreted a cocktail of chemical compounds, some of which were shown to be antifungal, while others attracted PGPR Bacillus spp [15]. Some of these root exudates were also reported to activate specific genes in bacterial members of rhizosphere communities: for example, phenolic compounds secreted by barley upon infection with Pythium ultimum activated genes encoding the production of antifungal compounds in Pseudomonas fluorescens [16]. Beyond these reported effects of specific root exudates on specific members of plant-associated communities, altered root secretion following a perceived biotic stress (e.g. pathogen attack) has also been observed to lead to community-wide shifts in the relative abundance of different populations. When Panax notoginseng was infected with the foliar pathogen Alternaria panax [17], changes in relative abundance of different members of the rhizosphere communities were observed, with higher relative abundance of beneficial bacteria and fungi at the expense of pathogenic ones. These community shifts were strong enough to lead to improved germination rates of Panax seeds sown in the soil where infected plants had formerly grown, compared to seeds germinating in soils previously harbouring uninfected plants, and even to reduce disease severity of A. panax on these new plantlets. Although the plant’s ability to shape their rhizosphere to their advantage is actively studied, the influence of the plant genotype on this phenomenon has received limited attention so far. In potato, the comparison of rhizosphere microbial communities between 51 cultivars revealed that the differences were correlated to the plant’s productivity, but not to other genetic features [18]. However, significant genotype-driven recruitment differences were observed in tomatoes, where resistant and susceptible cultivars exhibited distinct microbiome shifts upon late blight infection [19], suggesting that the host genetic background might influence microbiota recruitment. Whether resistant cultivars would recruit bacteria more efficiently than sensitive cultivars due to a better microbiota resource utilization, or less efficiently due to lower symptom severity is an open question. Evidence for targeted microbiome modulation and recruitment by the plants could be used to point us directly to plant-beneficial members of the communities such as strains antagonistic to pathogens, thereby reducing the screening time necessary to identify new mBCAs. Since the cry-for-help hypothesis has not yet been investigated in potato, the aim of this study was to test whether late blight infection in potato would also lead to microbiota shifts and to a rhizosphere recruitment of beneficial bacteria. In addition, we explored the influence of plant genotype on such recruitment, with the ultimate goal to identify new and promising mBCAs against this disease. To this end, we challenged potato plants of two cultivars differing in late blight sensitivity with a controlled infection of P. infestans to induce the plant’s reaction, without allowing the infection to proceed further and destroy the plants. We performed this challenge in two consecutive generations of potted plants and analysed the shifts in phyllosphere, rhizosphere and soil bacterial communities occurring as a result of the challenge (infected vs. uninfected), the cultivars (sensitive vs. tolerant) and the generation (first vs. second). In parallel to the amplicon-based community shift studies, we isolated cultivable members of bacterial communities. Those which could be associated to sequences of the amplicon-based community study were selected and assessed for their protective effects against P. infestans using bioassays targeting different stages of the oomycete’s life cycle. Material and methods Exposure of potato plants to P. infestans over multiple generations The experimental setup used for this transient infection of potato plants is illustrated in Fig. 1. Seed potato tubers of susceptible cultivar Bintje and tolerant cultivar Innovator were certified free from Leafroll Virus and Potato Virus Y by the Agroscope [20]. The tubers were planted in 2.2 L pots filled with standard potting soil made of 35% compost, 15% loam, 17% leaf mould, 25% Toresa (wood-based peat replacement) and 8% sand. This substrate was heated at 90 °C for 4h prior pot filling. Each pot was inoculated 1 and 7 days after planting with 100 mL of a suspension generated by mixing soils collected from copper-free organic potato fields from three different farms in Switzerland (Muensingen 46.884010, 7.559025; Uettligen 46.985864, 7.389619; Niederbipp 47.267207, 7.707649). Bintje and Innovator plants were kept in the greenhouse during 19 and 16 days after emergence respectively and were then transiently infected with P. infestans. To this end, the plants were transferred to a growth chamber maintained at 16 °C with 85% humidity. A zoospore solution was produced from P. infestans Rec01 strain grown on V8 medium for 10 to 14 days. To harvest the zoospores, the P. infestans petri dishes were wrapped in aluminium foil and 15 mL of ice-cold water was poured into them. They were placed at 4 °C for 2h and then incubated at room temperature for 30 min. After incubation, the liquid was pipetted into a 15 mL Falcon tube covered with aluminium foil. The concentration of zoospores was determined using a Thoma chamber and diluted to 50,000 zoospores/mL. The plants were sprayed lightly with distilled water immediately before being sprayed with either the zoospore solution or distilled water for the control plants. The leaves were sprayed until the solution dripped from the leaves. To keep high humidity conditions, the plants were sprayed generously with distilled water every hour for 6h. After three days, the plants were placed in a second growth chamber at 25 °C with 60% humidity for three days to stop the infection, before being transferred back to the greenhouse for 9 days. At least 3 uninfected leaflets from 3 different Phytophthora-exposed mature leaves, 5 mL of rhizospheric soil and 10 mL of bulk soil (respectively referred to as phyllosphere, rhizosphere and soil samples, were collected per pot (5 pot per treatment as biological replicates) 15 days after the infection and kept at −80 °C. The remaining plants (control and infected ones) were removed from the pots 3 days later. Their soil as well as the rhizospheric soil were collected and mixed with standard potting soil (¼ of previous generation soil with ¾ of standard soil). These mixtures were used to grow the second generation of potatoes. Experimental setup for transient infection of potato plants. Susceptible (Bintje) and tolerant (Innovator) cultivars were planted in pots with soil inoculated with microbial communities from copper-free organic potato fields in Switzerland. After growth in a greenhouse, plants were transferred to a high-humidity growth chamber and transiently infected with either Phytophthora infestans zoospores (orange lightning) or treated with distilled water (control). Post-infection, plants were returned to the greenhouse. Successive generations of plants were grown in soil from infected or uninfected plants to examine microbial community shifts. Samples of phyllosphere, rhizosphere, and soil were collected from infected plants 15 days after infection (blue squares) to analyze microbial dynamics. Rhizospheric soil, roots and leaves from the plants of the third generation were used for microorganism isolation. Ti, time of infection The objective of the second generation was to test the hypothesis that the signal from this first generation could be amplified in a second round: 15 days after emergence, plants which grew in the Phytophthora-challenged plant soil were transiently infected as described above (Fig. 1) and the plants which grew in non-challenged plant soil received the same control treatment as in the first generation. Fifteen days after infection, phyllosphere, rhizosphere and soil samples (6 biological replicates) were harvested and the remaining soil and rhizosphere of the plants were collected, mixed with fresh potting soil (¼ of previous generation soil with ¾ of new soil) and used for the planting of a third generation of potatoes. The third generation’s purpose was to provide material from which to isolate bacterial strains from the phyllosphere, roots and rhizosphere of plants growing in the Phytophthora-challenged plant soil. Microbiome sample DNA extraction and sequencing The DNA of microbiome samples was extracted using the NucleoSpin Soil DNA extraction kit (Macherey–Nagel) according to the manufacturer’s instructions. The V5, V6 and V7 hypervariable regions of the 16rRNA were amplified using the 799F (5′-AACMGGATTAGATACCCKG-3′) and 1193R (5′-ACGTCATCCCCACCTTCC-3′) primers [21]. The Platinum SuperFi II DNA polymerase (Invitrogen, Carlsbad, California) was used for the amplification of soil and rhizosphere samples, and the Phusion DNA polymerase (ThermoFisher, Waltham, Massachusetts) for the phyllosphere samples. The ATCC-MSA-3001 (American Type Culture Collection, Manassas, Virginia) mock community standard was used as an internal standard. The bacterial 16S rRNA were separated from the plant mitochondrial 16S by fragment size selection using the PippinHT (Sage Science, Beverly, USA). The library was prepared with the Illumina Nextera XT kit (Illumina, San Diego, USA) at the Bern Next Generation Sequencing Platform, and the sequencing was done using the Illumina MiSeq System in a single MiSeq v3 flow cell (600 cycles). The quantity of amplicon from the phyllosphere samples were adjusted to get 20% of the sequencing depth compared to the rhizosphere and soil samples. A total of 250 samples including 74 phyllosphere, 64 rhizosphere, 104 soil and 8 control and mocks went through the NGS workflow, and 132 of them (44 for each plant compartment) were selected for the current study. The raw reads are available in the European Nucleotide Archive database, accession PRJEB75732. NGS data cleaning and correction with DADA2 The raw reads were cleaned using the fastp 0.20.1 [22] program and their quality was checked using FastQC 0.11.9 [23] and MultiQC 1.8 [24]. The reads were then corrected using the DADA2 package [25] in R. Firstly the reads were trimmed asymmetrically at 261 bp for the forward reads and 200 bp for the reverse reads, to have a long overlap of about 45 bp, with an expected error of 3 bp between the runs. This asymmetrical trim avoided having an overlap in the V5 hypervariable region. The sequences of both runs were merged with a tolerance of up to two mismatches. In the end there were 23,420 different Amplicon Sequence Variants (ASVs). The taxonomy was assigned until the Genus rank based on the Silva database, release 138 [26]. Initial data treatment and alpha diversity analysis The resulting read table was analysed with R [27]. The samples of soil, rhizosphere and phyllosphere of the plants 15 days after being challenged were selected, and the organelle contaminants were removed by filtering out the ASVs assigned to the Cyanobacteria phylum. At the end of this step, there were 22,557 different ASVs. The rarefaction curves were computed with the rarecurve function from the vegan package [28] and plotted with ggplot (Figure S1). The richness–defined as the number of taxa observed–and Sheldon’s evenness of each sample were calculated with vegan. The effect of the plant generation, compartment (phyllosphere, rhizosphere, soil), treatment (infected, control) and cultivar (Bintje, Innovator) on the alpha diversity were assessed by comparing the samples from the soil and the rhizosphere in a four-way ANOVA with the formula: And the samples from the phyllosphere in a Generalised Least Squares (GLS) model with the formula: The ANOVA and GLS were followed by Tukey Honest Significant Differences (HSD) post-hoc tests to compare the richness and evenness between infected and control plants within each single compartment, generation and cultivar. Low-occurrence ASVs removal and beta-diversity analysis The low-occurring ASVs were filtered out by keeping only the sequences with at least 3 reads in at least 4 samples. There were 6472 ASVs remaining after this step, which represents 29% of the total number of ASVs, but 96% of the total reads from the sequencing. The filtered tables were normalised by Total Sum Scaling (TSS) with the package edgeR [29]. The TSS-normalised table was used for ordinations in order to have an overall idea of the dissimilarity between all the samples. It was firstly plotted based on Bray–Curtis dissimilarity matrix as a PCoA with the phyloseq package [30]. A Distance-Based Redundancy Analysis (dbRDA) based on a Bray–Curtis dissimilarity matrix meant to removing the variability linked to the generation on the ordination and to represent more specifically the effect of the compartment was then performed with the vegan package using the formula: Then another dbRDA aimed at removing the variability linked to the compartment and representing more specifically the effect of the generation was performed with the formula: To find which ASVs became more abundant upon infection, the filtered read table was sliced into twelve subtables, each containing the samples of one generation, one compartment and one cultivar. Each subtable was then normalised by Trimmed Mean of the M Component (TMM) before going through a generalised linear model (GLM) fitness test from the edgeR package, using the model: An ASV was considered differentially abundant between infected and uninfected plants if its adjusted p-value (Benjamini–Hochberg method) was below 0.05. Comparing the number of unique differentially abundant (DA) ASVs between each parameter (generation, compartment and cultivar), supported the choice to focus on the community shifts in the rhizosphere of both cultivars. Indicators were calculated for every taxonomic level in the rhizosphere of the second generation. They were based on the TMM-normalised abundance and expressed in percentages (Table 1 see after the conclusion). The proportion of each taxon compared to the whole community was calculated with this formula: Table 1Differential abundance of chosen taxa per cultivarThe cells of the taxon name were shaded according to their phylogenetic level (Kingdom is darkest and Genus is lightest). Descriptive statistics were calculated for every taxon of every taxonomic level present in our study; all were expressed in percentages. The last letter of the column names (B, I) refers to the cultivar. The shading of the cells is proportional to their value, the darkest colors correspond to the highest value observed commonly between the cultivars. The Rel_abund columns show the abundance of each taxon as a proportion of the total abundance. The Weighted_dens columns show the proportion of the abundance of the DA ASVs of each taxon as a proportion of the total abundance. The More columns represent the proportion of the abundance of the ASVs which become significantly more abundant upon infection as a proportion of the abundance of the taxon. The Less columns are the same as the More columns, but for significantly less abundant ASVs. Taxa shown in this table were kept if their differential abundance characteristics were striking enough to be reported in the main text of this article, or if they were closely related to such a taxon. The full table is available in the supplementary materials (Table S3). Differential abundance of chosen taxa per cultivar The cells of the taxon name were shaded according to their phylogenetic level (Kingdom is darkest and Genus is lightest). Descriptive statistics were calculated for every taxon of every taxonomic level present in our study; all were expressed in percentages. The last letter of the column names (B, I) refers to the cultivar. The shading of the cells is proportional to their value, the darkest colors correspond to the highest value observed commonly between the cultivars. The Rel_abund columns show the abundance of each taxon as a proportion of the total abundance. The Weighted_dens columns show the proportion of the abundance of the DA ASVs of each taxon as a proportion of the total abundance. The More columns represent the proportion of the abundance of the ASVs which become significantly more abundant upon infection as a proportion of the abundance of the taxon. The Less columns are the same as the More columns, but for significantly less abundant ASVs. Taxa shown in this table were kept if their differential abundance characteristics were striking enough to be reported in the main text of this article, or if they were closely related to such a taxon. The full table is available in the supplementary materials (Table S3). The proportion of DA ASVs in each taxon compared to the whole community was calculated with this formula: The proportion of promoted (or inhibited) DA ASVs within each taxon was calculated with this formula: Phylogenetic tree creation and visualisation Annotated phylogenetic trees were created to visually compare the infection-triggered shifts in the rhizosphere communities between the susceptible and resistant cultivars. The ASVs used for the beta-diversity analysis were aligned with MAFFT 7.490-GCC-10.3.0-with-extensions [31] and the global phylogenetic tree was built using FastTree 2.1.11-GCCcore-10.3.0 [32]. Further work on phylogenetic trees was based on only the data and differential abundance tests of the rhizosphere. The ASVs absent from the rhizosphere samples were filtered out using the phyloseq package. Then the ggtree [33] and ggnewscale [34] R packages were used to build 1) a general tree per generation, which summarises the differential abundance analyses per order and per cultivar, and 2) several detailed trees for the second generation only, showing the differential abundance of every single ASV per cultivar. For each general tree, the ASVs were aggregated per order and split per cultivar. All the calculations were based on the TMM-normalised abundance. The abundance was calculated by adding the values of all the ASVs of each order. The proportion of DA ASVs within each order was calculated by dividing the sum of the abundances of the DA ASVs by the sum of all the ASVs per order: The mean normalised, weighted log-fold change of significantly differentially abundant ASVs of each order, was calculated with this formula: The above indicator shows if the order became more or less abundant upon infection, and the scale of this difference compared to the whole rhizosphere community. The detailed trees were drawn to represent every single ASV in the rhizosphere of the second generation, their abundance and logFC. To keep it readable, the ASVs were subsetted on the basis of their phylogeny in order to have fewer than 900 ASVs per tree. Only the informative trees were kept (see Supplementary materials). Isolation and identification of microbial strains Four different growth media supplemented with either nystatin (10 μg/mL) or ampicillin + kanamycin (respectively 250 μg/mL and 25 μg/mL), were used for the isolation of the microbial strains: 10% Tryptic Soy Agar (Sigma-Aldrich, Darmstadt, Germany) supplemented with nystatin; 20% Potato Dextrose Agar (Carl Roth, Karlsruhe, Germany) supplemented with ampicillin and kanamycin; Actinomycete Isolation Agar (Sigma-Aldrich) supplemented with nystatin and Artificial Root Exudate (ARE) [35] supplemented with either nystatin or ampicillin and kanamycin. The isolation was made from three different compartments (leaves, rhizosphere soil and roots) in a 0.1 M potassium phosphate isolation buffer at pH 7 with 0.1% Silwet. The leaflets from the leaves 2 and 3 of the healthiest-looking plants of generation 3 were ground in a mortar with 10 mL isolation buffer, then 200 μL of the ground tissue were spread either undiluted or 10- and 100-fold diluted in Petri dishes containing the different growth media. For isolation of root and rhizosphere soil communities, the plants were uprooted and shaken to remove the non-rhizosphere soil. Roots were collected from multiple locations across each plant’s root system and placed loosely into a 50-mL Falcon tube. The tubes were then filled up to 45 mL with isolation buffer, shaken and mildly sonicated (47 kHz) for 5 min to separate the soil from the roots. The roots were placed into a new Falcon tube, and the remaining content (soil and root wash corresponding to the rhizosphere soil) from different plants of the same cultivar were mixed together and vortexed. The supernatant was diluted from 10−1 to 10−5 and 200 μL were plated on each growth medium. The roots were washed two extra times until they appeared white and were then ground in 10 mL isolation buffer in a mortar. The ground tissues from the plants of a same cultivar were mixed together, diluted from 10−1 to 10−7, and 200 μL were plated on the different growth media. After 5 days, single colonies were visible. Per cultivar and per isolation medium, all phenotypically distinct colonies were picked using a toothpick and transferred to a new plate for isolation and purification. In total, 636 bacterial isolates were collected. Each bacterial isolate was identified by partial 16S rRNA sequencing using the primers 27F (5′-AGAGTTTGATCCTGGCTCAG-3′) and 1492R (5′-CGGTTACCTTGTTACGACTT-3′). The strains were identified by aligning their 16S rRNA sequence on the SILVA, RDP [36] and NCBI [37] databases. The sequences are available in the European Nucleotide Archive, accession PRJEB75732. Identification of strains of interest for biocontrol of P. infestans As “strain of interest”, we considered the strains for which the abundance of the corresponding ASV differed significantly between transiently infected and control plants. The DA ASVs were identified using the results of the GLM described above. All the ASVs which were significantly differentially abundant between infected and uninfected plants in at least one GLM fitness test were listed. To identify strains corresponding to these ASVs, we proceeded as follows: all 16S rRNA sequences from the isolated strains were assembled as a custom blast database with the BLAST + (2.15.0) package for Ubuntu [38] and blasted against the list of DA ASVs previously identified using blastn (2.11.0) [39] allowing 1 mismatch and no deletion. If an ASV perfectly matched the 16S rRNA sequence of at least one strain, then all the strains which aligned with one mismatch were removed. As a result, 425 strains from the collection were aligned to 74 ASVs. Thirty-five strains aligning with 7 ASVs were removed because they belonged to species listed as opportunistic pathogens by the Swiss Federal Office for the Environment [40]. At this stage, there were 390 strains left aligning with 67 ASVs, among which 144 strains perfectly matching with 39 ASVs and 246 strains aligning with 1 mismatch on 28 ASVs. For the biological assays, all the strains matching perfectly any ASV were kept, but only one strain per ASV was kept for the ASVs which aligned with one mismatch, while excluding the filamentous bacteria (e.g. Streptomyces sp.) which would not fit the high throughput screening described below. This resulted in the selection of 163 strains aligning with 57 ASVs, which were screened for their biocontrol potential against P. infestans. Evaluating the biocontrol potential of the strains of interest Preparation of P. infestans zoospore and sporangia suspensions Sporangia suspension P. infestans 208M2 (GFP-tagged) [41] was grown on V8 medium (10% V8 juice supplemented with 1 g/L of CaCO3 and 15 g/L of agar) for 10 to 21 days. Mycelium was then scraped off the agar and put into a 15 mL Falcon tube, to which sterile distilled water was added until reaching 2 mL. The tube was shaken vigorously to detach the sporangia from the hyphae. The solution was then filtered using a nylon mesh. The sporangia concentration was determined using a Thoma chamber and adjusted to 30,000 sporangia/mL. Zoospore suspension P. infestans 208M2 was grown on V8 medium for 10 to 14 days. The Petri dishes were then wrapped in aluminium foil and 15 mL of ice-cold water was poured into them. They were placed at 4 °C for 2h and then incubated at room temperature for 30 min. After incubation, the liquid was pipetted into a 15 mL Falcon tube coated in aluminium foil. The concentration of zoospores was determined using a Thoma chamber and diluted to 30,000 zoospores/mL. High throughput P. infestans in vitro confrontation assays Bacteria preparation The selected bacteria were grouped in 96-well plated according to their growth speed. They were then incubated at 28 °C in 200 μL of filtered liquid V8 medium (autoclaved 10% V8 juice supplemented with 1 g/L of CaCO3, filtered using 0.22 μm sterile syringe filters) during 12h to 48h. The OD600 of each well was measured using a Cytation5 cell imaging reader (Agilent, Santa Clara, USA). Germination assays The bacterial suspensions in filtered V8 or sterile distilled water were diluted with sterile 0.9% NaCl solution to adjust their OD600. One hundred μL of this solution was mixed with 40 μL of 208M2 sporangia or zoospore suspension with the goal of reaching an adjusted OD600 between 0.3 and 0.7. Adjustments were made to adapt to the diversity of strains screened. A replicate with an OD600 below 0.3 was added for each strain to provide comparison points for 34 strains whose growth plateaued below 0.3. In addition, 40 strains formed compact aggregates that moved with the liquid and the position of these aggregates at the moment of OD600 measurement added variability, including a small number of OD600 values above 0.7. The plates were sealed using parafilm and placed in a box with wet paper towel sheets at 18 °C during 24h for sporangia or 4h for zoospores until imaging. Imaging, scoring and data treatment The sporangia and zoospore germination assays were performed at least 3 times per strain. Imaging was performed using a Cytation5 plate reader. For each well, a bright-field image and a GFP fluorescent image (excitation at 395 nm) were taken. A germination score from 0 to 4 was given manually for each well. The scores for sporangia germination are 0 for no germination; 1 if less than 30% of sporangia had a short germ tube; 2 if less than 30% of sporangia had a long germ tube; 3 if 30% to 90% of sporangia had medium-sized germ tubes; 4 if more than 90% of sporangia had long germ tubes. The scores for zoospore germination were 0 for no germination, 1 if less than 30% of zoospores had a short germ tube, 2 if 30% to 80% of the zoospores had a medium-sized germ tube, 3 if more than 90% of zoospores germinated without appressoria, 4 if more than 90% of zoospores germinated with appressoria. The adjusted score was calculated by dividing the germination score of each strain by the average germination score of the distilled water controls of their respective 96-well plate. The adjusted OD600 was calculated by removing the average OD600 of the controls of each 96-well plate from the OD600 of each strain. The adjusted OD600 and adjusted score were used to calculate the germination index using the following calculation: A germination index below 0.25 indicates strong inhibition, an index between 0.25 to 0.5 signifies moderate inhibition, an index between 0.5 to 0.9 suggests a weak inhibition, and indexes between 0.9 and approximately 1.8 indicate no inhibition. Indexes higher than 1.8 imply that germination was stimulated. The score per strain was calculated by averaging the germination indexes of all the replicates. The strains were then grouped by phylogenetic group at every taxonomic level (Phylum, Class, Order, Family, Genus). The groups with fewer than 3 strains were discarded and the distribution of the germination index of the remaining groups were compared with a Kruskal–Wallis test followed by Dunn’s post-hoc test (p-value adjusted with the Benjamini–Hochberg method). The strains were then grouped by cultivar and compartment of isolation and were compared by a two-way ANOVA with Type II Sum of Squares, followed by Tukey’s HSD. The average index for each ASV was determined by averaging the indexes of all associated strains. ASVs were then classified as “MORE” if they showed a significant increase in abundance upon infection in at least one GLM test (see Low-occurrence ASVs removal and beta-diversity analysis section above), and "EQUAL" otherwise. The germination index of the "MORE" and "EQUAL" groups were compared using a Kruskal–Wallis test. A similar comparison was conducted for strains that decreased in abundance. Results The late blight symptoms were more severe in the first generation than in the second one Lesions characteristic of late blight infection developed on the stems and leaves of zoospore-infected plants of both generations. Their expansion was fast in the cold growth chamber and nearly ceased after the plants were transferred to the warmer growth chamber. In the first generation, Bintje plants had fewer but more severe lesions than Innovator plants. Fewer lesions were observed in the second generation in both cultivars and Bintje had a higher number of more severe lesions than Innovator plants. Transient P. infestans infection did not affect the alpha diversity of bacterial communities The richness of bacterial communities was affected by both the compartment and the generation when comparing the soil and the rhizosphere. As expected, the soil communities were significantly richer and displayed higher evenness than the rhizosphere communities (Table S1, Figure S2). We also observed that communities were richer in the second generation than in the first one for both compartments. However, neither the treatment (transient infection vs. control) nor the cultivar affected the richness of rhizosphere bacterial communities. When compared with the soil or the rhizosphere, communities living in the phyllosphere had much lower richness, but relatively high evenness. Richness was not affected by any factor, but evenness was significantly higher in the second generation compared to the first one, and lower in infected plants compared to uninfected ones. Although no global trends could be observed, transient P. infestans infection altered the richness and evenness of subgroups of this experiment in four instances (Figure S2). These four significant changes all occurred in the first generation, and three of them concerned the sensitive cultivar Bintje: in the phyllosphere, which is the site of infection, richness decreased in infected plants compared to control plants, while the reverse tendency was observed in the soil, but not in the rhizosphere. Surprisingly, although richness decreased in leaf communities upon infection, evenness increased, which also occurred in rhizosphere communities. The only impact of transient infection on communities associated with the resistant cultivar Innovator was an increased evenness in the soil communities. Overall, the communities associated with this cultivar reacted less strongly to infection than those associated with the sensitive cultivar (Figure S2). Leaf infection had a strong cultivar-dependent impact on the beta diversity of soil and rhizosphere communities According to the results of the PCoA, all samples grouped into three major clusters (Fig. 2a), one for the phyllosphere samples, and one for each generation for the soil and rhizosphere samples. In phyllosphere samples (Fig. 2b), communities differed slightly between infected and control plants in the first generation, while such differences could not be observed in the second generation. The soil and rhizosphere samples of the first generation (Fig. 2c) clustered according to their compartment, then according to their infection status. As expected, the difference between communities from infected and uninfected plants was greater within the rhizosphere than within the soil communities. The soil and rhizosphere samples of the second generation (Fig. 2d) also clustered according to their compartment and infection status but with a remarkably similar clustering pattern between soil and rhizosphere samples. In contrast, much greater differences were observed in the effect of infection on both cultivars than in the first generation: in both soil and rhizosphere communities, those associated with the sensitive cultivar Bintje reacted much more strongly to infection than those associated with the resistant cultivar Innovator (Fig. 2d, circles vs. diamonds). This stronger impact of infection on communities from the second generation was further highlighted by the distance-based redundancy analysis, which removed the variability explained by the compartment (Figure S3a). In addition, removing the variability explained by the generation (Figure S3b) revealed that shifts in the soil and the rhizosphere communities followed a remarkably similar pattern. In both cases, communities from infected Bintje plants were clearly separated from the other communities. Principal Coordinate Analysis of a all bacterial communities, b phyllosphere communities, c rhizosphere and soil communities from the first generation, and d rhizosphere and soil communities from the second generation The phylum composition of phyllosphere and rhizosphere communities changed between generations In this section the relative abundance of each taxon is described based on a presumption that Amplicon Sequence Variants (ASVs) read number is proportional to the number of living cells across the community. The Proteobacteria were the most abundant phylum in every compartment (Figures S4 and S5), followed by the Actinobacteriota, the Bacteroidota and the Firmicutes. The Proteobacteria were particularly dominant in the phyllosphere of the first generation where they accounted for 87% of the total community (Figures S4 and S5a). This proportion dropped significantly during the second generation where it went down to 68%, mostly due to the proportion of Firmicutes booming from 7 to 21%. This drop in Proteobacteria in the second generation was not observed in the rhizosphere and soil, where they rose from below 50% to almost 60% (Figure S5 b and c). In the soil, the increase in relative abundance of Proteobacteria was compensated mainly by a drop in Actinobacteriota, which fell from 23 to 12%, while the Bacteroidota and Firmicutes remained stable at about 6% and 10%. In the rhizosphere, the increase in Proteobacteria was mostly compensated by a drop in the Firmicutes population from 15 to 5% in the second generation, while the proportion of Actinobacteriota, Bacteroidota and other phyla all dropped by 1 to 2% each. The phylum distribution was much more even in the rhizosphere than in the soil in the first generation, but these compartments became remarkably similar in the second generation. While the infection had a limited effect on the soil phylum composition (Figure S5c), it led to stronger shifts in compartments closely associated with plants. While the phyllosphere phylum composition of Innovator did not change upon infection in the first generation (Figure S5a), the Actinobacteriota population in Bintje collapsed from 18 to 4% of the total abundance, which led to an increase in the proportion of Firmicutes and Proteobacteria. In the second generation, the phyllosphere population in Bintje were stable, whereas in Innovator, Proteobacteria increased upon infection at the expense of Firmicutes and Actinobacteriota. The infection-induced shifts observed in the rhizosphere were common to both cultivars in the first generation (Figure S5b), with an increase in Proteobacteria and Bacteroidota at the expense of Firmicutes, which collapsed. Later, in the second generation, the phyla composition in Innovator hardly changed upon infection, while in Bintje, Firmicutes and Proteobacteria increased at the expense of the Bacteroidota and Actinobacteriota. The number of differentially abundant (DA)ASVs was much higher in the second generation (2191) than in the first one (192), including 81 DA ASVs in both generations. (Fig. 3a). More DA ASVs were detected within Bintje-associated communities (1475 specific, 1821 total) than within Innovator-associated ones (481 specific, 827 total) (Fig. 3b), with a large majority of them in the soil (1149 specific, 1791 total), less in the rhizosphere (507 specific, 1145 total) and very few in the phyllosphere (4 specific, 8 total) (Fig. 3c). Venn diagrams summarizing the number of differentially abundant ASVs detected in the GLMs (Table S1a). Each diagram shows the same data split across the different factors: a the generation, b the cultivar and c the compartment. Within a diagram, each set represents the result of several GLM grouped together depending on the samples used in it; each GLM was used in exactly one set. Each set is a list of all the ASVs which were differentially abundant in at least one GLM. The numbers in each section of the diagram represent the number and percentage of DA ASVs that were shared between sets or unique to one According to these results, shifts in bacterial communities following transient infection with P. infestans occurred almost exclusively in the soil and rhizosphere, although infection occurred in leaves. The few observed in the phyllosphere were either driven by significant shifts on very abundant ASVs with a strong impact on community structure, or they were non-significant at the ASV level, leaving them absent from the differential abundance analysis, yet visible at the phylum level. Further analyses were therefore focused on the rhizosphere communities, in view of the higher consistency of their shifts across generations. Rhizosphere community shifts upon infection were stronger in the sensitive cultivar and amplified in the second generation The bacterial community of the rhizosphere shifted in a cultivar-dependent way in both quantity, which is the percentage of ASVs that belonged to DA ASVs considering their abundance, and in quality, which is whether an ASV was promoted (became more abundant upon infection) or inhibited (became less abundant upon infection). In the first generation, 12% of the ASVs were inhibited in Bintje, and 0.1% were promoted, while Innovator had only 1.4% promoted and 0.6% inhibited (Table S2). Most of the inhibited ASVs in Bintje were detected in the Firmicutes (12% of the total abundance), with a sharp decline in abundance, while few ASVs shifted significantly in other major phyla. The shifts in Innovator were spread more evenly across the major phyla with 1% in the Proteobacteria, 0.6% in the Firmicutes and 0.4% in the Bacteroidota. In both cultivars, most of the DA ASVs were concentrated in a single order per phylum. Shifts in the Proteobacteria were mostly in the Burkholderiales, the shifts in the Firmicutes were mostly in the Bacillales, and the shifts in the Bacteroidota were mostly in the Flavobacteriales. Considerably more DA ASVs were detected in the second generation than in the first one. The changes were again much more pronounced in Bintje, where 23% of all ASVs were differentially abundant, than in Innovator where it was 12% (Table 1). However, unlike in the first generation, more DA ASVs were promoted (14% of ASVs in Bintje and 7% in Innovator) than inhibited (9% in Bintje and 5% in Innovator). The most quantitatively significant shifts happened in the most abundant taxa. In the Proteobacteria, all the DA ASVs taken together accounted for 10% of all reads in Bintje and only 6% in Innovator (Table 1), while they accounted respectively for 8% and 2% of ASVs in the Bacteroidota. A remarkably small fraction of the Actinobacteriota exhibited significant shifts, with their differentially abundant ASVs representing only 1% of the total reads in both cultivars. In contrast, a larger proportion of Firmicutes showed shifts, with differentially abundant ASVs constituting 2% of the total reads in Bintje cultivar and 1% in Innovator. The differential abundance results were summarised into trees, displaying the abundance, the proportion of DA ASVs and the direction and magnitude of the shifts for each order (Fig. 4). In addition, subtrees showing the abundance and post infection log-fold change for every ASVs of selected taxa per cultivar were created (Figures S6 to S13). Phylogenetic trees representing the distribution of differentially abundant ASVs across the orders and per cultivar in the first (a) and second (b) generation. Each tip represents an order, and their colors represent the phyla they belong to. The tree is surrounded by three annotation circles, the innermost half of which represents the Bintje cultivar (B), and the outermost half represents Innovator (I). The outer annotation circle represents the TMM-normalized abundance of each order. The middle annotation circle represents the proportion of DA ASVs weighted by the abundance of each ASV. The inner annotation circle represents the influence of the infection on the abundance the DA ASVs of each The differences at the order level highlight the cultivar-dependent nature of the microbiome shifts observed. Very few orders displayed DA ASVs in the first generation, and only the Bacillales in Bintje had a high proportion of DA ASVs (dark purple, Fig. 4a). Most of the orders with a low proportion of DA ASVs had near-neutral weighted averaged logFC values (pale red and pale blue squares in Fig. 4a) except for the Burkholderiales and Flavobacteriales, which were strongly promoted in a few ASVs in Innovator (dark red, Fig. 4a), and inhibited in Bintje (blue, Fig. 4a). The second generation had considerably more DA ASVs (Fig. 4b), but most orders which contained DA ASVs had near-neutral weighted averaged logFC. Only two orders had a strong, consistent response between cultivars: Burkholderiales (Proteobacteria), which was the most abundant order of the study, and Bryobacterales (Acidobacteriota). In Innovator, only one additional order, Caulobacterales (Proteobacteria), shifted sharply and became less abundant. In contrast, Bintje exhibited strong shifts in three major orders, such as Rhizobiales (Proteobacteria) and Chitinophagales (Bacteroidota), which increased in abundance, and Flavobacteriales (Bacteroidota), which decreased. Additionally, several minor orders showed significant shifts in Bintje such as Paludibaculum (Acidobacteriota), Peptostreptococcales (Firmicutes), Clostridiales (Firmicutes), R7C24 (Proteobacteria) and unassigned Gammaproteobacteria. The cultivar-dependent nature of the shifts appeared even more clearly in the detailed trees where we saw that differing ASVs were differentially abundant even when trends appeared similar at higher taxonomic level, such as in the Firmicutes in the first generation (Figure S6) or in the Gammaproteobacteria in the second generation (Figures S7 and S8). Bintje and Innovator had no ASV shifting in the same direction in the first generation, and only 4% of their DA ASVs in the second generation. Proteobacteria Like most taxa, the Proteobacteria displayed weaker shifts in the first generation than in the second one. In the first generation, very few ASVs with differential abundance were observed in the Alphaproteobacteria despite their prevalence in the community. In the second generation, however, they displayed clear opposite cultivar-dependent trends. While Alphaproteobacteria in Bintje were overall promoted with 9% of promoted, and 5% of inhibited taxa, they were mostly inhibited in Innovator with 2% of promoted and 5% of inhibited taxa (Table 1). These differences were mainly due to the Rhizobiales and the Caulobacterales, which were strongly promoted in Bintje and did not shift or were inhibited in Innovator. Additionally, in Bintje, 29% of Devosia ASVs were promoted, with 0.1% showing inhibition. In contrast, Innovator displayed a different pattern, where 4% of ASVs were slightly inhibited and only 1% were promoted. A similar cultivar-dependent trend was observed for Rickettsiales: in Bintje, 53% of ASVs were strongly inhibited and only 5% were promoted, while the response of Rickettsiales in Innovator was more balanced with 23% of ASVs promoted and 27% inhibited (Table 1). On the other hand, Gammaproteobacteria displayed differential abundance during the first generation, essentially in the Burkholderiales where respectively 2% and 12% of the order’s reads were more abundant in Bintje and less abundant in Innovator (Table S2). In the second generation, the Burkholderiales shifted similarly in Bintje and in Innovator, with respectively 16% and 14% of the order reads that were promoted, and 3% and 5% that were inhibited (Table 1). Infection affected ASVs across the whole order (Figure S7) with a particularly strong response in the Comamonadaceae and Rhodocyclaceae families, but without particularly dense differential abundance clusters. Differential abundance clusters were found among other taxa of the Gammaproteobacteria despite their lower abundance (Table 1, Figure S8). Although they generally hardly shifted upon infection, a clear and dense cluster was observed in the Xanthomonadales, corresponding to the Ahniella genus, where 52% and 64% of reads were promoted in Bintje and Innovator respectively (Table 1, Figure S8). More DA ASV clusters were observed to be specific to the Bintje cultivar, such as in the Enterobacterales, the R7C24, or in a group of unassigned Gammaproteobacteria (Incertenae sedis) (Figure S8). Bacteroidota Similarly to what happened in the Proteobacteria, the Bacteroidota displayed fewer DA ASVs in Bintje than in Innovator (respectively 1% and 2% of the group abundance) in the first generation, and most of the DA ASVs in Bintje were inhibited while most of the DA ASVs in Innovator were promoted. The differential abundance was detected almost exclusively in the Flavobacteriales (Table S2). By contrast, much more DA ASVs were observed in the second generation, with a stronger response to infection in the sensitive cultivar (16% and 32% of respectively promoted and inhibited reads in Bintje) than in the tolerant one (7% and 10% of respectively promoted and inhibited reads in Innovator). Interestingly, each order of the phylum had a different reaction to the infection (Figure S9): Like in the first generation, the Flavobacteriales was the order with the strongest inhibition in Bintje (Fig. 4) with 38% inhibited reads (Table 1). They were composed of three families with very contrasted shifts upon infection: The Flavobacteriaceae were strongly inhibited in Bintje and reacted weakly in Innovator, while the NS9 marine group and the Weeksellaceae were strongly promoted in both cultivars. The Chitinophagales observed were mainly Saprospiraceae and showed a strong response in Bintje, with 43% of ASVs promoted and 20% inhibited, while their response was less pronounced in Innovator, where 13% of ASVs were promoted and only 1% inhibited (Table 1). In contrast, the Sphingobacteriales were balanced between promoted and inhibited ASVs mainly in Bintje (Table 1, Figure S10). This apparent balance originates from the two most important families of this order having opposite reaction, i.e. the KD3-93 group was inhibited in both cultivars while the AKYH767 group was promoted in both groups, with an entire branch promoted specifically in Bintje. Lastly the Cytophagales showed hardly any feedback to the infection. Firmicutes The Firmicutes had by far the strongest DA ratio of the first generation as their population collapsed in Bintje, with 78% of inhibited reads (Table S2), mainly coming from a few clusters of very abundant ASVs among the Bacillales (Figure S6). On the other side, only 3.7% of the reads were inhibited in Innovator, despite a strongly reduced abundance (Figure S5). The decrease in the Innovator population was carried by many non-significantly differentially abundant ASVs. In contrast to the first generation, the Firmicutes were mostly promoted in infected plants in the second generation, with stronger changes in Bintje (Figure S11, Table 1). While the Bacillales reacted less prominently in Bintje than in Innovator, with respectively 13% and 21% of promoted reads, the Clostridia were very strongly promoted in Bintje (67%) and showed an opposite tendency in Innovator, where no ASV was promoted and 4% were inhibited (Figure S11, Table 1). Actinobacteriota The Actinobacteriota were remarkably conserved in both generations. In the first generation, there was no differential abundance in Bintje and 3.7% of the Gaiellales order belonged to promoted ASVs in Innovator (Table S2). In the second generation, the phylogenetic tree showed a few scattered DA ASVs, with an unusual majority of inhibited ASVs in Bintje. One cluster was still observable among the Micrococcales (Figure S12). Here too, different subgroups showed different, cultivar-dependent responses to infection, e.g. the Micrococcaceae were inhibited in both cultivars and the Demequinaceae were promoted in Bintje only (Table 1). Minor phyla No notable shift was observed upon infection in the less abundant phyla of the first generation (Table S2), but several taxa showed a strong reaction in the second generation (Figure S13, Table 1). The Acidobacteriota reacted significantly: both of its major orders, Bryobacterales and Paludibaculum, were highly promoted upon infection in both cultivars (Figure S13). Within the Myxococcota, the order blfdi19 had two peculiarities: first, it was located far away from the other Myxococcota on the phylogenetic trees (Myxococcales, Figs. 4 and S13), and second, it was the only group which responded with strong, balancing shifts in Bintje (42% promoted, 48% inhibited) but with a high promotion in Innovator (57% promoted and no inhibition). The Dependentiae only contained ASVs of the Babeliales order. These reacted very strongly in both cultivars, with 46% and 70% of promoted reads in Bintje and Innovator respectively, and 15% and 8% of inhibited reads in the respective cultivars (Figure S13, Table 1). In summary, infection triggered substantial, cultivar-specific shifts in the rhizosphere microbial communities of both generations, although they differed in nature. Despite the number of DA ASVs detected being ten times higher in the second generation than in the first one, both generations had many common points. Firstly, Bintje shifted more strongly than Innovator in both generations, with most of the shifts being concentrated in a few orders: Burkholderiales, Flavobacteriales, and Bacillales, while the Actinobacteriota stayed remarkably stable. Each phylum comprised numerous taxa with distinct responses, emphasising the importance of detailed analysis for this type of study. Beyond characterising community shifts following infection, the sequences of DA ASVs were used to identify corresponding bacteria among the strains that were isolated from the third generation of potatoes. Anti-phytophthora activities of strains isolated from leaf-infected plants From the retrieved isolates, the partial 16S rRNA gene sequence of 163 strains aligned with 57 different ASVs with a maximum of 1 mismatch (Table S4). These strains comprised 90 Proteobacteria, 48 Firmicutes, and 25 Actinobacteriota. Among these, 14 strains originated from leaves, 74 from roots, and 75 from rhizosphere soil. 88 strains were isolated from Innovator and 75 from Bintje. The ASVs linked to these strains frequently exhibited differential abundance upon infection, with 21 strains becoming more abundant in at least one test, 8 becoming less abundant in at least one test, and 32 remaining unchanged across all tests. The differential abundance of these ASVs varied across compartments: none became differentially abundant in the phyllosphere, 17 did so in the rhizosphere, and 18 in the non-rhizosphere soil. Nine of the selected ASVs exhibited differential abundance in the first generation, while 21 did so in the second generation. Additionally, 21 of these ASVs became differentially abundant in Bintje, vs. 9 did so in Innovator. These strains were tested for their potential inhibition of the germination of the two types of spores produced by P. infestans, sporangia and zoospores. The ability to inhibit spore germination depended on the cultivar the strains were isolated from and on their phylogenetic identity Overall, strains were much more efficient in inhibiting sporangia than zoospores, with average germination indexes of 0.46 for sporangia, i.e. a moderate inhibition compared to 0.8 for zoospores, i.e. a weak inhibition. The most efficient genera against sporangia germination mostly belonged to Proteobacteria, e.g. Phyllobacterium, Advenella and Pseudomonas, with respective median indexes of 0.18, 0.21 and 0.31 (Fig. 5a). Nocardioides from the Actinobacteriota was also extremely efficient (0.19), as well as the 14 Bacillus strains associated with the ASVs 77 and 68 (0.22 and 0.26 respectively). In contrast, for zoospore germination, only strains belonging to the Pseudomonas genus were able to efficiently inhibit this Phytophthora developmental stage, with an index of 0.27 (Fig. 5b). Average final germination indexes per strain for each genus for a sporangia germination and b zoospore germination. A lower index corresponds to a higher germination inhibition, the red line corresponds to an index of 0.9, below which germination is considered inhibited. The significance letters can be compared within each sub-figure. Interestingly, there was an effect of the cultivar of isolation on the strains’ inhibition efficiency of both zoospore and sporangia germination (Figure S14): strains isolated from Innovator exhibited slightly higher inhibitory activity against both zoospore (Dunn’s test p = 0.037) and sporangia (p = 0.044) germination compared to those isolated from Bintje. Additionally, the compartment of isolation significantly influenced zoospore germination inhibition, whereby the strains isolated from the roots were more effective in inhibiting zoospore germination than those isolated from the rhizosphere soil (p = 0.011). However, we found no overall correlation between the strains’ activity and the differential abundance of the corresponding ASVs in infected vs. uninfected plants, and hence no evidence of a specific plant recruitment mechanism which would have led to enrichment of active strains upon leaf infection (Figure S15). Furthermore, the taxa that most efficiently inhibited spore germination were not associated with ASVs whose abundance strongly differed upon infection in the rhizosphere, except for the Bacillales in the first generation. Discussion In this study, we analysed the bacterial community shifts upon late blight infection in two different potato cultivars, Bintje (sensitive) and Innovator (resistant), over two consecutive generations. The overall composition of the microbial communities was similar between the two cultivars, with a few groups slightly more abundant in Bintje (Enterobacterales, Peptostreptococcales, Chitinophagales, Burkholderiales and undefined Gammaproteobacteria). This increased abundance was balanced with a higher proportion of Streptomycetales in Innovator. These minor differences between the rhizospheres of both cultivars were expected, since plant genotype is considered a secondary driver of the rhizosphere composition behind soil composition, as shown by the comparison of ten common bean cultivars [42]. Another study on seven cultivars of sorghum came to the same conclusion but found one cultivar with a significantly different rhizosphere composition than the others and linked this divergence to differences in the strigolactone root secretion [43]. In potato, the comparison of the rhizosphere communities of 51 potato cultivars in one soil revealed that bacterial networks among the cultivars with the highest growth were more developed, but the variations in the rhizosphere composition itself could not be linked to the cultivar’s characteristics [18]. It would have been particularly interesting to characterize the root exudates in our study to determine whether the observed differences were due to variations in the exudation profile between the two genotypes in absence and/or presence of infection. In our dataset, the main factors influencing the structure of bacterial communities were the plant compartment and the generation, followed by infection status and cultivar. Infection did not affect the alpha diversity in a conclusive way, but significant differences were found in the community structure of the first generation followed by major shifts in the second generation. Other reports about community changes in the rhizosphere infected with various Phytophthora species have yielded contrasting results. The diversity of the bacterial rhizosphere of avocado trees did not change if they had symptoms of P. cinnamomi-induced root-rot compared to asymptomatic plants, but several phyla including Actinobacteria and Firmicutes became less abundant while the Bacteroidetes became more abundant [44]. In another study, no significant changes were found in the alpha- and beta-diversity of the rhizosphere microbial communities upon root infection of soybean with P. sojae [45]. The few studies analysing the impact of the cultivar showed contrasting results: infecting tomato with P. infestans led to an increased rhizosphere microbial richness in a susceptible cultivar, but not in the resistant one [19]. In contrast, infecting the roots of common bean with Fusarium oxysporum caused increased richness in both sensitive and resistant cultivars [46]. Although we observed that the phyllosphere communities shifted in a cultivar-dependant way upon infection in both generations, only 8 ASVs were differentially abundant in total in this compartment, which is extremely low, especially compared to what has been found after a foliar infection on citrus, where 109 ASVs (20% of the whole community) were promoted in melanose-infected trees [47], or to the extent of the changes observed in rust-infected crabapple, where 17 bacterial orders were differentially abundant compared to healthy trees [48]. A likely explanation for this difference could be that these studies analysed variations in the infected leaves themselves from severely infected plants, which trigger a dysbiosis resulting from interplays between the pathogen, the plant microbiota and the plant itself [49]. Our study, however, focused on variations in uninfected tissues from infected plants, a methodology aimed at the description of shifts expected to be controlled by the plant. Furthermore, the phyllosphere of our greenhouse-grown plants is expected to harbour much lower bacterial richness and abundance compared to plants grown outside due to the importance of environmental factors in the establishment of the phyllosphere communities [50], as reported in lettuce by one study comparing the phyllosphere of greenhouse- and field-grown plants and showing a reduced number of colony forming units by 10- to 100-fold in greenhouse-grown lettuce compared with field-grown ones [51]. Foliar infection caused strong community shifts in the rhizosphere soil in both generations, while causing the bulk soil communities to shift slightly in the first generation and more substantially in the second. These shifts in both soil compartments suggest the existence of a mechanism altering the underground microbiome based on the plant health status, as was already observed in several other plant species infected by various leaf pathogens such as Arabidopsis thaliana infected with downy mildew [52], soybean affected by brown spot disease [45], tomatoes infected with late blight [19], and tea trees infected with grey blight [40]. A third of the ASVs of the second generation were differentially abundant in the underground compartments between infected and uninfected plants. These results suggest that the plant triggered rhizosphere shifts upon foliar infection with P. infestans. These changes were marginal in the bulk soil of the first generation, but led to the more pronounced shifts in the second generation, probably through an amplification of the signal occurring with the repeated infection, keeping in mind that in potted plants, “bulk soil”, although not physically attached to the roots, might still be close enough to the roots to be influenced by the plant. In a setting similar to ours, larger differences were observed in community structure in the bulk soil than in the rhizosphere of A. thaliana infiltrated over 6 generations with Pseudomonas syringae, compared to plants infiltrated with a sterile solution [54]. Meanwhile, the rhizosphere communities stayed under the direct influence of the plant, which led to important changes. Although both cultivars had relatively similar communities at first, the plant genotype strongly influenced microbial dynamics upon infection, with cultivar-dependent differences in DA ASV counts across both generations. Bintje, the more sensitive cultivar exhibited more than 1700 DA ASVs compared to fewer than 800 in Innovator. The higher responsiveness in Bintje was consistent between generations, especially in the rhizosphere. In the first generation, the DA ASVs in Bintje were almost exclusively inhibited and a large majority of Innovator’s were promoted, while in the second one, most orders had their taxa shifting in the same direction between cultivar, with a majority of promoted ASVs. The studies that compared rhizosphere shifts in resistant and susceptible cultivars showed different reactions but rarely give details on the nature of the observed shifts. P. infestans infection has a stronger impact on the alpha-diversity of the rhizosphere communities of susceptible tomatoes than in the resistant ones [19]. The same study also reported differing beta-diversity shifts between these cultivars, without quantifying them. Another study on common bean infected by the soil-borne pathogen Fusarium oxysporum reported rhizosphere shifts in both resistant and susceptible cultivars, with clear family- and genus-level differences in the communities of these cultivars [46]. Interestingly, the rhizosphere of susceptible rapeseed infected with Plasmodiophora brassicaceae became a lot more interconnected upon infection, whereas resistant cultivars only showed minor changes [55]. It must be emphasized that unlike in our study, the shifts observed in studies using root infections could be caused by a direct action of the pathogen, not only by a root exudate-mediated reaction of the plant. The taxa that reacted the most to infection were spread across several phyla: Firmicutes, Proteobacteria, Bacteroidota and Acidobacteriota. Most of these DA orders and genera are known to host PGPR. In the Firmicutes, the abundance of the Bacillales collapsed upon infection during the first generation in both cultivars, as previously observed in avocado [44], but still increased upon infection. This order is known for the extensively studied Bacillus genus, which contains several plant-beneficial species. Some were shown to promote plant growth by secreting phytohormones in the rhizosphere [56] or by increasing nutrient availability by nitrogen fixation [57, 58] and phosphorus solubilization [59, 60]. Furthermore, rhizospheric Bacillus were reported to help the plant against pathogens, firstly by triggering Induced Systemic Resistance, allowing tomatoes to resist viral [61] and fungal [62] infections, secondly by secreting a wide range of powerful pathogen inhibitors, which were able to protect tomatoes from bacterial canker [63] and wheat against take-all [64]. The Clostridiales and the Peptostreptococcales on the other hand, were much less abundant than the Bacillales but were heavily promoted in Bintje following infection in the second generation. While we know that the Clostridiales can interact with plants, either as sweet potato pathogens [65] or as a nitrogen-fixing endophytes in grasses [66], the Peptostreptococcales were never reported to interact with plants to our knowledge. The Burkholderiales was the most abundant order of the Proteobacteria. They were slightly promoted upon infection in the rhizosphere of Innovator during the first generation and strongly promoted in the rhizosphere of both cultivars during the second generation. Among them, the Comamonadaceae had the highest proportion of DA. It is a large family, which contains many cultivable species, including many plant growth promoting strains as well as phytopathogens. Among them, members of the Variovorax genus were reported to improve barley and fenugreek plant growth in agricultural soil when inoculated with a nitrogen-fixing Sinorhizobium [67]. Members of the Comamonas were reported to be great nitrogen fixers [68], and some Acidovorax strains improved barley seedlings survival and promoted root growth [69], while others are known cucurbit pathogens [70]. The Flavobacteriales were the most prominent and reactive order of the Bacteroidota. They were promoted in the resistant cultivar during the first generation but were strongly inhibited in both cultivars in the second generation. In our study, they were essentially from the Flavobacterium genus, which is abundant in the rhizosphere of several species including cucumber [71], lettuce [72], maize [73] and A. thaliana [21], and in both the rhizosphere and the phyllosphere of wheat [74]. Several isolates could promote plant growth by secreting phytohormones in the rhizosphere [75–77], while a strain helped capsicum plants resist Phytophthora capsicii root infection [78]. Besides, the Saprospiraceae (most of the Chitinophagales reads) were heavily promoted in the rhizosphere of Bintje during the second generation. Although they have been observed in the rhizosphere of maize [79] and soils fertilized with manure [80], they have been detected much more frequently in aquatic environments and bioreactors [81]. Lastly in the Acidobacteriota, infection led to a considerable promotion of Bryobacter genus during the second generation in both cultivars. Several elements point toward a positive influence of this group on plant survival as they were detected in the rhizosphere of plants from environments as inhospitable as the Atacama Desert and Antarctica [82]. Furthermore, their presence was reported to promote duckweed growth [83] and was associated with healthy sesame [84] and Polygonatum kingianum [85] plants. In addition, metagenomic studies have found many plant growth promoting traits in assembled genomes of members of this genus [86]. When we assessed the biological activity of the isolated strains on P. infestans sporangia and zoospore germination, we observed that the genera Advenella, Nocardioides and Phyllobacterium were the best at inhibiting sporangia germination. Advenella contains known Plant Growth Promoting Bacteria (PGPB), capable of alleviating salinity stress [87] and reported to be a necessary element of a simple synthetic community capable of reducing root rot incidence by triggering Induced Systemic Resistance [88]. To our knowledge, phytopathogen inhibition activity has only been reported once in this genus [89]. Nocardioides are usually found in the soil but are also connected to plants, as some species contain endophytic strains [90, 91]. They have received some attention lately and are extensively studied for their bioremediation potential but are not usually linked to the inhibition of phytopathogenic organisms [92]. Several Phyllobacterium species are known for helping plant growth and resilience in several ways. Their presence in the rhizosphere increased resistance to drought stress [93, 94], they increased lateral root growth [95] and solubilized phosphate [94]. Broad-spectrum fungal inhibition was reported once [96], but anti-oomycete activity has never yet been reported to our knowledge. Inhibiting zoospore germination seemed more challenging than inhibiting sporangia germination, as only Pseudomonas strains were able to do so. This genus is very diverse, and several species are extensively studied for their role as plant growth promoters and in plant protection [97]. Interestingly, the strains isolated from the resistant cultivar Innovator were in average more efficient in inhibiting both spores than the strains isolated from the sensitive cultivar Bintje, suggesting that the resistant cultivar’s microbiota might be enriched in protective strains, which might contribute to their host’s higher disease resistance. Despite these findings, we could not link the differential abundance of the ASVs to the direct inhibition of P. infestans spores by the ASV-associated strains. This conclusion is constrained by the limited number of DA ASVs that could be associated to cultivable strains from our collection. Several taxa with strong shifts were not successfully isolated, such as the Flavobacterium and Saprospiraceae among the Chitinophagales, Bryobacter among the Acidobacteria, and obligate anaerobic taxa as Clostridiales and Peptostreptococcales. Furthermore, our method focused on in vitro direct inhibition, which may not represent in planta protection [98]. Additionally, we cannot exclude that protective mechanisms might rely on interactions between multiple organisms, which we would not see in our experimental setup. For example, A. thaliana was protected from downy mildew by a consortium of three rhizosphere-isolated strains that triggered induced systemic resistance [52], and Fusarium wilt suppression in a field resulted from the synergistic action of a siderophore-producing Pseudomonas and a non-pathogenic Fusarium [99, 100]. Deciphering the role of the strains associated to DA ASVs, as single strains or in combinations, will hence require further experiments in ulterior studies in planta. Conclusion Foliar infection with Phytophthora infestans induced shifts in the potato-associated microbiome, which were first detected in the rhizosphere of the first generation and led to a community divergence with much wider differences in the bulk soil and rhizosphere in the second generation, while the phyllosphere remained stable. Plant genotype also played a major role, as shifts were consistently more important in the susceptible cultivar Bintje that in the resistant cultivar Innovator. While the shifts in the first generation in Bintje were almost exclusively from inhibited ASVs in the Bacillales, Innovator’s differentially abundant ASVs were primarily promoted in the Burkholderiales and Flavobacteriales. The shifts observed in the second generation were more similar between cultivars with common non-clustered shifts among the Burkholderiales and the Flavobacteriales, alongside cultivar-specific clustered shifts in Chitinophagales, Clostridiales, Xanthomonadales, Bryobacterales, Paludibaculum and Peptostreptococcales. Biological assays to assess the spore inhibition potential of strains isolated from potato plants revealed the biocontrol potential of strains otherwise not studied for their antagonism towards phytopathogenic organisms, such as Advenella, Nocardioides and Phyllobacterium. Although we did not see any correlation between differential abundance and biological activity, the overall higher activity of strains isolated from the resistant potato cultivar still suggests that the plant microbiota could play a significant role in its host’s resistance to diseases. It would therefore be of interest to test, in future experiments, whether these active strains associated with the resistant cultivar could be used to improve the protection of the sensitive one. Supplementary Information Figure S1 (JPG 91 KB) Figure S2 (JPG 88 KB) Figure S3 (JPG 61 KB) Figure S4 (PDF 6 KB) Figure S5 (PDF 9 KB) Figure S6 (PDF 49 KB) Figure S7 (PDF 104 KB) Figure S8 (PDF 91 KB) Figure S9 (PDF 104 KB) Figure S10 (PDF 21 KB) Figure S11 (PDF 51 KB) Figure S12 (PDF 119 KB) Figure S13 (PDF 80 KB) Figure S14 (JPG 95 KB) Figure S15 (JPG 71 KB) Table S1 (XLSX 12 KB) Table S2 (CSV 49 KB) Table S3 (CSV 57 KB) Table S4 (XLSX 28 KB) Legends of all the figures and tables (DOCX 19 KB) Acknowledgements We deeply thank the Gerbert Rüf Stiftung, which funded this project. Further financial support from the Swiss National Science Foundation (grant 207917 to L.W.) is gratefully acknowledged. We are grateful for the technical assistance of Aurélie Esseiva, Elissa Feghali, Camila Morales, Eva Trutman, Nicolas Rappo and Tom Lüthi. We would also like to thank the Next Generation Sequencing Platform in Bern for their precious support, Tobias Gelencsér (Research Institute of Organic Agriculture, FiBL) for helping us find the soil inoculum, and Natacha Bodenhausen (FiBL) for her valuable inputs on the project. We are also grateful to the team of the Botanical Garden of Fribourg for their continued help in the greenhouse. Abbreviations Amplicon sequence variant Microbial biological control agents Differentially abundant Distance-based redundancy analysis Generalised linear model Generalised least squares model Tuckey’s honest significance differences Optical density at 600 nm Principal coordinate analysis Plant growth promoting rhizobacterium Trimmed mean of the M component Total sum scaling Author contributions Vivien Pichon did the infection experiment, managed the bacterial strain isolation, the data analysis and the redaction of the article. Mout De Vrieze supervised the lab experiments, and directed the confrontation assays with Rares Cristea. Farès Bellameche participated in isolating microbial strains and reviewing the article. Floriane L’Haridon managed the laboratory and provided technical support for the project. Laurent Falquet supervised the microbiome analysis part of the project. Laure Weisskopf acquired the funding, designed the project and supervised it, except for the microbiome analysis part, and contributed to writing and revising the article. All authors read and approved the final manuscript. Funding The main funding for this project was provided by the Gerbert Rüf Stiftung. Further financial support from the Swiss National Science Foundation (grant 207917 to L.W.) is gratefully acknowledged. Data availability The Miseq sequencing data and the 16S rRNA sequences of the isolated strains are available in the European Nucleotide Archive, accession PRJEB75732 at the following URL: https://www.ebi.ac.uk/ena/browser/view/PRJEB75732. The scripts used during the current study are available in the PotatoCryForHelp repository on Github, available at this URL: https://github.com/VivienPichon/PotatoCryForHelp. The intermediate data and figures generated by the scripts are publicly available on Zenodo, at this URL: https://zenodo.org/records/13384779?preview=1&token=eyJhbGciOiJIUzUxMiJ9.eyJpZCI6IjQ3OWMwMzg1LWE1MDAtNGNkMS04N2RhLTgxZWQxYTg0NDY5MCIsImRhdGEiOnt9LCJyYW5kb20iOiJmMzQyYTAyMjZiMmI2YWI2MDM3Zjg3YzI2NDMxZTQxYSJ9.DLnhZkjhpd7Isb4Dl7inUIFAYP6r0w7jKhCLr13Z3DUr4yU6KcFBgdL0uoO8Y3S_WkjaorkOHjzF3f_Dy_pyMg. Declarations Ethics approval and consent to participate Not applicable. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials Figure S1 (JPG 91 KB) Figure S2 (JPG 88 KB) Figure S3 (JPG 61 KB) Figure S4 (PDF 6 KB) Figure S5 (PDF 9 KB) Figure S6 (PDF 49 KB) Figure S7 (PDF 104 KB) Figure S8 (PDF 91 KB) Figure S9 (PDF 104 KB) Figure S10 (PDF 21 KB) Figure S11 (PDF 51 KB) Figure S12 (PDF 119 KB) Figure S13 (PDF 80 KB) Figure S14 (JPG 95 KB) Figure S15 (JPG 71 KB) Table S1 (XLSX 12 KB) Table S2 (CSV 49 KB) Table S3 (CSV 57 KB) Table S4 (XLSX 28 KB) Legends of all the figures and tables (DOCX 19 KB) Data Availability Statement The Miseq sequencing data and the 16S rRNA sequences of the isolated strains are available in the European Nucleotide Archive, accession PRJEB75732 at the following URL: https://www.ebi.ac.uk/ena/browser/view/PRJEB75732. The scripts used during the current study are available in the PotatoCryForHelp repository on Github, available at this URL: https://github.com/VivienPichon/PotatoCryForHelp. The intermediate data and figures generated by the scripts are publicly available on Zenodo, at this URL: https://zenodo.org/records/13384779?preview=1&token=eyJhbGciOiJIUzUxMiJ9.eyJpZCI6IjQ3OWMwMzg1LWE1MDAtNGNkMS04N2RhLTgxZWQxYTg0NDY5MCIsImRhdGEiOnt9LCJyYW5kb20iOiJmMzQyYTAyMjZiMmI2YWI2MDM3Zjg3YzI2NDMxZTQxYSJ9.DLnhZkjhpd7Isb4Dl7inUIFAYP6r0w7jKhCLr13Z3DUr4yU6KcFBgdL0uoO8Y3S_WkjaorkOHjzF3f_Dy_pyMg.