Soil samples were collected in 10 farmer’s fields across Switzerland (Figure 1) in May 2013. The main characteristics of the 10 field soils are listed in Table 1. Field sites had a history of multi-year cereal-oriented crop rotation and were chosen to represent predominant Swiss agricultural soil types and climate conditions. All fields were cropped with winter wheat in the year of sampling. For sampling, soil cores of 15–20 cm depth were extracted with disinfected soil recovery augers between the rows of wheat plants at 20 random locations in each field in order to obtain a representative sample. Then, extracted soil samples were sieved (mesh size, 10 by 10 mm) in order to remove stones, plant residues or other larger material, pooled and thoroughly mixed. For each site, approximately 120 kg of sieved soil were collected and stored in barrels for 3 months at 15°C before the start of the experiments in order to equilibrate the soils, i.e., to minimize effects of different environmental conditions (e.g., temperature, soil moisture) prevailing at the different sampling sites at the time of sampling. The storage temperature was chosen because it can be considered as the average temperature in Switzerland during the growing season of wheat from April to September (according to long-term monthly temperature averages recorded by the Swiss Federal Office MeteoSwiss¹). Soil parameter analyses were carried out by the Labor für Boden- und Umweltanalytik (Eric Schweizer AG, Steffisburg, Switzerland) following standard protocols used in Swiss agriculture (Agroscope, 2006). Concentrations of soluble, readily plant-available macronutrients were determined following H₂O extraction, for which soil samples were suspended in distilled water at a ratio of 1:10 (g mL^-1). Reserve macronutrients and micronutrients were extracted with ammonium acetate EDTA, for which soil samples dried at 65°C were suspended at a ratio of 1:10 (g mL^-1) in a solution consisting of acetic acid (0.5 mol L^-1), ammonium acetate and EDTA (0.02 mol L^-1) adjusted to a pH of 4.65. Soil suspensions were then vigorously shaken for 1 h and filtered prior to analysis by mass spectrometry. The soil parameter analyses were carried out (i) directly after sampling and (ii) after the end of the experiments. Except for nitrate, the soil parameters did not change significantly between the two analyses (Table 1).

In order to test the natural resistance of the 10 Swiss agricultural soils toward soilborne pathogens, pot experiments were carried out in the greenhouse with cucumber and Pu and wheat and Gt. The Pu inoculum was prepared by inoculating 25 g of autoclaved millet seeds moistened with 10 mL of sterile water with three plugs of a culture of Pu strain ETH-2 (isolated from a Swiss agricultural soil) grown on Oxoid malt agar (Thermo Fisher Scientific, Reinach, Switzerland) for 7 days. Pu millet seed cultures were grown for 7 days at 18°C and then cut into small fragments for soil inoculation. The Gt inoculum was prepared by inoculating 100 g of autoclaved oat seeds without spelts and moistened with 100 mL of sterile, distilled water with 20 plugs of cultures of Gt strain I-17 (Lebreton et al., 2007) grown on Oxoid potato dextrose agar for 14 days. Gt oat cultures were grown for 4 weeks at 24°C in the dark and then dried in a sterile cabinet on sterile filter paper for 3 days. Seeds of cucumber (Cucumis sativus cv. Chinese Snake) and spring wheat (Triticum aestivum cv. Rubli) were surface-sterilized for 30 min in 1.5% (v/v) NaOCl, rinsed with sterile distilled water and pre-germinated on sterile moist filter paper for 2 days in the dark at 24°C. For the plant experiments, 250-mL plastic pots of were filled with a 4:1 mixture (wt/wt) of the respective field soil and quartz sand (grain size of 0.5–2.2 mm diameter). Pathogen inoculum at different concentrations was thoroughly mixed into soil. Pu inoculum was added at 0.125, 0.25, 0.5, or 1.0 g inoculum per pot, whereas Gt inoculum was added at 0.2, 0.6, 2.0, or 6.0 g per pot. Pots in control treatments contained field soil without pathogen addition. Three seedlings of cucumber or wheat, respectively, were then planted per pot. For each treatment, six replicate pots were prepared. Plants were grown in the greenhouse at 70% relative humidity with light (210 μmol m^-2 sec^-1) for 16 h at 22°C (cucumber) or 18°C (wheat), followed by an 8-h dark period at 18°C (cucumber) or 15°C (wheat). Plants were watered routinely to keep the soil at constant moisture. The position of the pots was changed at random every other day to avoid position effects. After incubation for 10 days (cucumber) or 21 days (wheat), total shoot fresh weights per pot were assessed.

To quantify the abundance of DAPG and PHZ producing bacteria in soil, we developed quantitative real-time polymerase chain reaction (qPCR) assays targeting phlD and phzF genes. These genes encode, respectively, a polyketide synthase involved in the synthesis of phloroglucinols from malonyl-CoA (Bangera and Thomashow, 1996; Achkar et al., 2005) and an isomerase involved in the synthesis of phenazine-1-carboxylic acid (Mavrodi et al., 1998; Blankenfeldt et al., 2004). Alignments were created with publicly available phlD and phzF sequences from GenBank² and conserved regions were chosen for the design of primers and probes (Table 2), which was carried out with the Primer 3 Plus software (Untergasser et al., 2007). The parameters were amplicon length between 100 and 200 bp, melting temperature (TM) between 50 and 70°C, TM of probe 5°C higher than TM of primers, and the default setting of the program for self-complementarity and 3′-end stability. Partial sequences of phlD (GenBank accession CP003190.1| :6563260-6563937) of strain P. protegens CHA0 (Jousset et al., 2014) and phzF (locus tag, PFLU3_RS28075) of P. synxantha 2–79 (Nesemann et al., 2015) were used for primer design. The specificity of the primers was tested in silico with Primer-Blast (Ye et al., 2012) and in vitro with genomic DNA from 28 DAPG-producing strains and 38 PHZ-producing strains of the P. fluorescens group and nine additional PHZ-producing strains (Supplementary Table S1). Results of these tests revealed that our qPCR assays amplify phlD and phzF genes exclusively from DAPG and PHZ producing species of the P. fluorescens lineage. The PRN biosynthetic genes were quantified on wheat roots by the qPCR method of Garbeva et al. (2004). That assay targets a gene for the class IA oxygenase PrnD that is involved in the final step of PRN biosynthesis (Kirner et al., 1998). In contrast to our phlD and phzF primers, the primers of Garbeva et al. (2004) have broader specificity and, in addition to Pseudomonas, amplify prnD from PRN-producing strains of Burkholderia and Serratia.

The efficiency of phlD and phzF primers at low gene copy numbers was evaluated using in vitro standard curves prepared by serially diluting genomic DNA of P. protegens CHA0 and P. synxantha 2–79. The genomic DNA was prepared by growing both strains in lysogeny broth (LB) (Bertani, 1951) overnight at 24°C on a rotary shaker at 180 rpm and extracting DNA with the Wizard Genomic DNA Purification Kit (Promega AG, Dübendorf, Switzerland). The concentration of purified DNA was quantified by fluorimetry with Qbit (Thermo Fisher Scientific). We also generated an in vivo standard curve for each qPCR assay to quantify the corresponding target genes on wheat roots. To this end, aliquots of 1 g of 21-days-old roots of spring wheat cv. Rubli grown in autoclaved soil were inoculated with decreasing concentrations of a mixture of bacterial cells belonging to different strains carrying the respective target gene. Strains used for in vivo standard curves are listed in Supplementary Table S1. Bacterial cells were harvested from overnight cultures in LB, washed and suspended in sterile 0.9% NaCl solution. Cell suspensions from each strain were set to the same optical density at 600 nm (OD₆₀₀) and then mixed together at equal proportions. The mixed suspensions were adjusted to an OD₆₀₀ of 0.125, corresponding to approximately 10⁸ CFU mL^-1, serially diluted and inoculated at 10¹, 10², 10³, 10⁴, 10⁵, 10⁶, 10⁷ and 10⁸ CFU g^-1 roots for the preparation of the standard curve. For each concentration and for the control without bacteria, three replicates were performed. The inoculated root samples used for standard curves were processed with the same method as the samples from pot experiments with the different soils (see following chapter). In vivo standard curves were prepared as described above for the phlD, phzF and prnD qPCR, using strains listed in Supplementary Table S1. Since all in vivo standard curves were prepared with bacterial cells recovered from wheat roots, the CT values can be directly converted to numbers of bacteria harboring phlD, phzF, or prnD per g root. Our qPCR data also directly reflect the abundance of the antimicrobial biosynthesis genes because phlD, phzF, or prnD are present in single copy in genomes of the P. fluorescens group (Flury et al., 2016). A survey of published bacterial genomes revealed that phzF and prnD are also found as a single copy in other bacterial species such as Burkholderia (phzF and prnD), Pectobacterium (phzF) or Serratia (prnD).

To standardize the root material for qPCR quantification of DAPG, PHZ and PRN biosynthetic genes, soil samples from the 10 Swiss field sites were planted with spring wheat cv. Rubli in a greenhouse pot experiment. Plastic pots of 8 cm diameter and 30 cm height were part-filled with field soil and three wheat seedlings, prepared as described above, were planted per pot. Six pots per field soil were prepared. Wheat plants were grown for 2.5 months under the conditions described above for the Gt resistance assays. Root samples were collected, rinsed with tap water, incubated overnight in a sterile 0.9% NaCl solution at 3°C, and then vigorously agitated at 350 rpm for 30 min. Roots were separated from the root-wash suspensions and kept for dry weight assessment. Root-wash suspensions were centrifuged at 3500 rpm for 20 min. The supernatant was discarded and aliquots of 0.5 ml of the resulting root-wash pellet were used for DNA extraction. To each sample, 10⁹ copies of the APA9 plasmid from a cassava mosaic virus were added as an internal standard (Von Felten et al., 2010). DNA extraction was performed with the MPBio soil kit (MP Biomedicals, Illkirch, France) following the protocol of the manufacturer. The concentration of extracted DNA was measured with Qbit. qPCR reactions consisted of 10 μL TaqMan Gene Expression Master Mix (Applied Biosystems, Foster City, CA, USA), 2 μL of the respective forward and reverse primer solutions (10 μM), 2 μL of the respective probe solution (2.5 μM), 0.5 μL of bovine serum albumin solution (20 mg mL^-1), and 2 μL of template DNA in a total reaction volume of 20 μL. Primer and probe sequences are indicated in Table 2. Cycling conditions consisted of 2 min at 50°C (to permit uracil-DNA glycosylase activity), an initial denaturation step of 10 min at 95°C, and 40 cycles of 15 s at 95°C, 30 s at annealing temperature (see Table 2) and 30 s at 72°C. In all samples, the added APA9 plasmid was quantified with the primers listed in Table 2, following the method of Von Felten et al. (2010). The results from the APA9 plasmid quantification were used to normalize DNA extraction.

To compare the abundance of phlD and phzF measured with qPCR with the abundance measured using a cultivation-dependent terminal endpoint dilution assay, also called most probable number PCR (MPN-PCR) method (Ramette et al., 2006), a greenhouse experiment was carried out with soil samples from the Cazis and Taenikon field sites and spring wheat cv. Rubli cultivated under the same conditions as described above. After 3 weeks, phzF and phlD qPCR assays were performed on one fraction of each harvested root-wash pellet as described above, while the other fractions were serially diluted into microtiter plate wells (200 μl volume) containing King’s medium B (King et al., 1954) broth amended with 100 mg L^-1 cycloheximide, 13 mg L^-1 chloramphenicol and 40 mg L^-1 ampicillin. The microtiter plates were incubated at 24°C for 3 days and MPN-PCR was performed as described by Ramette et al. (2006), except for adapting annealing temperatures to the primers used in the present study (Table 2).

Pythium ultimum and Gt populations naturally present on roots of wheat were quantified using the qPCR methods developed by Cullen et al. (2007) and Bithell et al. (2012), respectively. The reaction mix was prepared as described above for antimicrobial genes, and cycling conditions were set as described previously (Cullen et al., 2007; Bithell et al., 2012). In vitro standard curves were performed with genomic DNA of Pu isolate ETH-2 (concentration range from 0.1 ng to 200 ag per reaction) and of Gt isolate I-17 (10 ng to 10^-4 ng per reaction). Genomic DNA of the two pathogens was extracted with the DNeasy plant mini kit (Qiagen, Hombrechtikon, Switzerland) from lyophilized mycelia prepared from cultures grown in potato dextrose broth (Difco, Becton, Dickinson and Company, Franklin Lakes, USA) for 7 days at 24°C with agitation at 180 rpm.

Bacterial strains and plasmids used for generation of Pseudomonas reporter strains for monitoring antimicrobial gene expression are listed in Table 3. Pseudomonas and Escherichia coli strains were routinely cultured at 30 and 37°C, respectively, on nutrient agar plates, in LB and in nutrient yeast broth (Stanisich and Holloway, 1972). When appropriate, selective antibiotics were added to the media at the following concentrations: ampicillin, 100 μg mL^-1; chloramphenicol, 50 μg mL^-1; gentamicin, 10 μg mL^-1; kanamycin, 25 μg mL^-1; and tetracycline, 125 μg mL^-1. Genomic DNA from P. protegens strain CHA0 and P. chlororaphis strain PCL1391 was isolated as previously described (Schnider-Keel et al., 2000). Plasmids were extracted and purified using the QIAprep Spin Miniprep kit (Qiagen) or the JETStar Plasmid Purification Midi kit (Genomed, Basel, Switzerland). PCRs were done using the PrimeSTAR HS DNA polymerase kit (Takara Bio Inc., Shiga, Japan) as described elsewhere (Péchy-Tarr et al., 2005). All DNA digestion and ligation reactions were done using standard techniques (Sambrook and Russell, 2001; Schnider-Keel et al., 2001). DNA extractions from agarose gels were carried out with the QIAquick Gel Extraction kit (Qiagen). Transformations of electro-competent cells with plasmid or purified ligation products were performed by electroporation (Schnider-Keel et al., 2000). To amplify genomic DNA or to detect the presence of recombinant DNA in E. coli colonies by screening, 100–200 ng of DNA were amplified using the GoTaq DNA polymerase kit (Promega, Dübendorf, Switzerland). All PCR constructs intended for transformation were verified by sequence analysis. DNA sequencing was carried out by GATC Biotech AG (Konstanz, Germany). Sequences were analyzed using the DNASTAR Lasergene software package version 11.0.

In order to tag P. protegens CHA0 and P. chlororaphis PCL1391 with green fluorescent protein (GFP), a single copy of a gfp variant gene constitutively expressed from the P_tac promoter, was inserted into the chromosome using the pBK-miniTn7-gfp2 delivery plasmid (Koch et al., 2001) and the Tn7 transposition helper plasmid pUX-BF13 (Bao et al., 1991) as described previously (Péchy-Tarr et al., 2013). For use as reporters to monitor the expression of HCN, DAPG, PRN, and PHZ biosynthetic genes, strain CHA0-gfp was transformed with pME9011 (hcnA-mcherry), pME9012 (phlA-mcherry) or pME11011 (prnA-mcherry), and strain PCL1391-gfp with pME11017 (phzA-mcherry) (Table 3). To construct the prnA-mcherry reporter plasmid pME11011, a 629-bp fragment containing the CHA0 prnA promoter was amplified from pME7116 (Baehler et al., 2005) using primers P5 BamHI new and P6 (Table 3). The obtained fragment was digested with BamHI and EcoRI and ligated into the mcherry-based promoter-probe vector pME9010 (Rochat et al., 2010) opened with the same enzymes. Similarly, for the construction of the phzA-mcherry reporter plasmid pME11017, a 961-bp fragment containing the phzA promoter (Chin-A-Woeng et al., 2001) was amplified from genomic DNA of PCL1391 (Flury et al., 2016) using primers phzAF and phzAR (Table 3). The PCR product was digested with BamHI-SacI and the resulting fragment was first cloned into pUK21 and, from there, into pME9010, both opened with the same restriction enzymes.

Antimicrobial gene expression and colonization levels of GFP-marked P. protegens CHA0 and P. chlororaphis PCL1391 harboring mCherry-based reporter plasmids were monitored on roots of spring wheat grown in the soils sampled at the 10 different Swiss field sites. Untreated seeds of spring wheat cv. Rubli were surface-sterilized for 12 min in 4% NaClO (vol/vol) and then washed with sterile distilled water. Seeds were germinated on soft agar (Agar Agar Serva at 9 g L^-1; Serva, Heidelberg, Germany) for 48 h at room temperature in the dark. The wheat seedlings were then transferred to 200-mL Erlenmeyer flasks (5-cm opening; Simax, Czech Republic) containing 60 g of soil. In each flask, three seedlings were placed into the soil and inoculated with 1 mL of a suspension of washed cells of the respective Pseudomonas reporter strain adjusted within a range from 3.6 10⁷ to 7.8 × 10⁷ cells mL^-1. Washed cells were prepared from LB cultures grown without antibiotic addition under the conditions described above to an OD₆₀₀ of 0.8 to 1.5, depending on the reporter strain used, whereby choosing a growth stage at which no significant expression of the antimicrobial genes occurred. This was done by determining, with a fluorimetry assay (Baehler et al., 2005), the time point at which the relative red fluorescence emitted by the Pseudomonas strains carrying mCherry-based reporter plasmids was not yet significantly different from the red background fluorescence emitted by control strains carrying the empty vector pME9010. Wild-type and GFP-tagged P. protegens CHA0 and P. chlororaphis PCL1391 (with and without pME9010) were included as control treatments for properly setting green and red fluorescence backgrounds for the FACS-based flow cytometry analysis described below. Flasks were sealed with cotton wool plugs and incubated in a growth chamber set to 60% relative humidity for 16 h with light (176 μE m^-2 s^-1) at 25°C, followed by an 8-h dark period at 20°C. After 5 days of incubation, wheat roots from each flask were removed, washed using distilled water to remove loosely adhering soil particles from roots and pooled in 10 mL of autoclaved, ultrapure water contained in a sterile 50-mL Falcon tube. Tubes were agitated for 20 min at 300 rpm in order to remove the majority of bacteria from the roots. The resulting suspensions were filtered using a 5.0-μm sterile syringe single-use filter (Sartorius Stedim Biotech GmbH, Goettingen, Germany), transferred on ice and immediately analyzed by FACS as described below. Dry weights of wheat roots were recorded and the number of GFP-marked Pseudomonas cells present in the root washes were determined by FACS and recorded as cells g^-1 of root.

Green fluorescent protein and mCherry expression levels in Pseudomonas reporter cells in natural soils were quantified with a Becton–Dickinson LSRFortessa flow cytometer. Size and granularity of Pseudomonas cells and particles were determined by measuring the forward scatter (FSC-A) and side scatter (SSC-A) signals, respectively. FSC-A signals were collected with a photodiode detector (set to 350 V), in the range of 483 to 493 nm (488/10 BP filter), with a threshold set to 200. SSC-A signals were detected with a photomultiplier tube (PMT) (detector G, set to 300 V) in the range of 483 to 493 nm (488/10 BP filter). Green fluorescence signals were collected with the PMT detector E (set at the voltage of 676 V), in the range of 515 to 545 nm (530/30 BP filter, 505 LP mirror). Red fluorescence signals were detected using the PMT detector C (set to 700 V), between 600 and 620 nm (610/20 BP filter, 600 LP mirror). For FACS analysis, aliquots of 300 μL of filtered root-washes were placed into Nunc MaxiSorp flat-bottom 96-well plates (Sigma–Aldrich, Buchs, Switzerland), and each sample was mixed three times. The analyzed volume was standardized to 200 μL, allowing the detection of 500–10,000 GFP events, depending on the analyzed soil. Gating of GFP-marked bacteria was done by delimiting on the FSC-A/FITC-A density plot particles with green fluorescence values above the background fluorescence noise (i.e., autofluorescence emitted by root and soil particles, cell fragments or bacterial cells not expressing GFP). Control samples obtained from soils amended with pure water, wild-type P. protegens CHA0 and P. chlororaphis PCL1391, and GFP-tagged strains CHA0-gfp and PCL1391-gfp were used to identify the GFP background fluorescence in soil extraction samples. The red fluorescence emitted by the gated GFP-tagged Pseudomonas cells was then analyzed on the PE-Texas Red-A histogram and on the FSC-A/PE-Texas Red-A density plot, allowing to detect and analyze all GFP-marked cells actively expressing their mCherry-based reporter fusion. Control samples including CHA0-gfp and PCL1391-gfp without mCherry-based vector, or carrying the empty pME9010 vector were used to define the mCherry background fluorescence among the GFP-marked Pseudomonas population. The median of red fluorescence emitted by the Pseudomonas cells was calculated using the BD FACSDiva software version 8.0 (Becton–Dickinson). To determine root colonization levels of Pseudomonas reporter strains, the GFP tag was used to count by FACS the number of reporter cells present in the analyzed 200 μL of filtered root-wash and then to calculate their concentration per gram of dry roots.

Statistical data analysis was carried out with the open source software R version 3.2.3 (R Core Team, 2015). Shoot fresh weights of cucumber and wheat plants obtained from the Pu and Gt infection assays, respectively, were analyzed as a proxy for soil disease resistance. Shoot weights from samples where pathogen inoculum was added were normalized against the shoot weights from non-infested control plants grown in the same soil, to minimize variation due to different nutrient contents in the different soils. Data were checked for normal distribution with the Shapiro–Wilk test and by plotting QQ-Plots. Equality of variance was verified with Bartlett’s test. Analysis of variance was carried out with a non-parametric test (Kruskal–Wallis test, significance level p < 0.05), followed by a post hoc test (kruskalmc, R package ‘Pgirmess’).

The abundance of phlD, phzF, and prnD harboring bacterial cells on wheat roots was calculated with the in vivo standard curves described above. Efficiencies and detection limits of qPCR assays determined by in vivo standard curves are indicated in Supplementary Table S2. Cycle threshold values obtained from the in vivo standard curves and from the samples were normalized for differences in DNA extractions as described by Von Felten et al. (2010). Normalized values were used for further analysis. The average values obtained from the three technical replicates of each qPCR assay were used for statistical analysis, which was performed as described above for shoot weights.

Data on the expression of reporters of antimicrobial genes in the different soils represent the medians of three independent repetitions of the same experiment, with nine replicates per treatment in each experiment. Significant differences between treatments were calculated with a non-parametric Kruskal–Wallis test (significance level p < 0.05), followed by Dunn’s test for post hoc comparisons.

Data used for the heat map showing rankings of pathogen suppression, abundance of antimicrobial genes, and expression of antimicrobial genes by reporter strains in the different soils were normalized using the function ‘scale’ (R package ‘stats’). Correlations between pathogen suppression, gene abundance, gene expression and abiotic soil parameters were inferred with Spearman’s rho rank correlation (significance level p < 0.05). Data were displayed in a heat map with the functions ‘levelplot’ (R package ‘lattice’) or ‘corrplot’ (R package ‘corrplot’).

The general resistance of 10 Swiss agricultural soils (Figure 1 and Table 1) to soilborne pathogens was tested in a greenhouse assay in which increasing quantities of Gt or Pu inoculum were added to soil samples planted with spring wheat or cucumber, respectively. Resistance to both pathogens varied between soils. In the Gt-infested soils at 0.6 g inoculum per pot, shoot fresh weights of wheat plants ranged from 31% in Grangeneuve soil to 107% in Taenikon soil of the weights of control plants grown in not artificially infested soils (Figure 2A). The shoot fresh weights of wheat grown in infested Taenikon soil were significantly higher (1.3–3.5-fold) than those of plants grown in infested soils from Cazis, Eschikon, Grangeneuve, Vouvry, and Witzwil. Similar trends were observed at other inoculum quantities (Supplementary Figure S1). Plants grown in Taenikon soil had the highest shoot fresh weights at all Gt concentrations used except the highest (6 g of inoculum per pot), at which the fungal pathogen heavily affected plant growth in all the soils and reduced shoot weights by 60–90% (Supplementary Figure S1D).

When the soils were artificially infested with Pu inoculum at 0.125 g per pot, median shoot fresh weights of cucumber plants ranged from 0% in Vouvry soil to 106% in Cazis soil of those of control plants grown in non-infested soils (Figure 2B). The shoot fresh weights of cucumber grown in Pu-infested Cazis soil were significantly higher (three–fivefold) than those of plants grown in infested soils from Cadenazzo, Courtedoux, Delley, Taenikon, and Vouvry (Figure 2B) and similar trends were observed for other inoculum densities (Supplementary Figure S2). Plants grown in soil from the Cazis field site had the highest shoot weights at all Pu concentrations, while plants grown in soil from the Vouvry field had the lowest shoot weights at all levels of the pathogenic oomycete except at 0.25 g per pot.

Individual soils did not display equal resistance levels to the two soilborne pathogens. While the soils from the Taenikon and Delley field sites were the most resistant against Gt, they were among the least resistant against Pu (Figure 2 and Supplementary Figures S1, S2). Likewise, the soil from Cazis was the most resistant against Pu, but was only moderately resistant against Gt.

To account for potential effects of resident Gt and Pu populations on the outcome of the soil resistance experiments, a qPCR method targeting the ITS rRNA gene region was used to detect and quantify the two pathogens in the rhizoplane of spring wheat plants grown in the not artificially infested control treatments of all soils. Gt could not be detected in any of the samples. By contrast, Pu was detected on the plant roots in all ten soils, but there was no significant difference in the abundance of the oomycete pathogen among the individual soils (Supplementary Figure S3).

In summary, the 10 agricultural soils strongly varied in their resistance against soilborne pathogens. However, the soil resistance levels observed for the two investigated pathogens in general were different; i.e., some soils displaying high resistance to Pu were highly susceptible to Gt and vice versa, pointing to specificity in the buffering capacity of individual soils toward specific soilborne pathogens.

The abundance of bacterial cells harboring phlD, phzF, and prnD required for the biosynthesis of the antimicrobials DAPG, PHZ, and PRN, respectively, was quantified by qPCR on roots of spring wheat grown in the 10 Swiss agricultural soils. As detailed in section “Materials and Methods,” we assume that phlD⁺ and phzF⁺ cells quantified in our assays correspond to cell numbers of DAPG and PHZ producing pseudomonads, whereas prnD⁺ cells correspond to cell numbers of PRN producing bacteria. Since the investigated genes are present as one copy per bacterial cell, we also refer to the abundance of cells harboring an antimicrobial gene as gene abundance.

The abundance of the phlD⁺ pseudomonads in all studied soils in general was higher than the abundance of the phzF⁺ and prnD⁺ bacteria, and in some soils reached 10⁷ cells per gram of root dry weight, whereas the abundance of the latter remained below 10⁶ gene copies per gram of root dry weight (Figure 3). For individual genes, pronounced differences in abundance between soils were observed. The biggest differences were found for phlD⁺ cells with approximately 10⁴-fold higher numbers on roots of wheat grown in soil from Taenikon compared to those grown in soil from Vouvry (Figure 3A). The abundance of phlD⁺ pseudomonads was significantly higher in soils from Courtedoux, Delley, Eschikon, Grangeneuve, Taenikon, Utzenstorf, and Witzwil than in soils from Cadenazzo, Cazis, and Vouvry. The abundances of phzF⁺ pseudomonads and prnD⁺ bacteria varied at maximum 100-fold between the different soils. The number of pseudomonads harboring the phzF gene was significantly higher on roots samples from Courtedoux and Delley soils compared to samples extracted from Vouvry soil (Figure 3B). The abundance of prnD⁺ bacteria was significantly higher in soil samples from Cadenazzo, compared to those from Courtedoux, Cazis, and Vouvry (Figure 3C). Taken together, for all three investigated antimicrobial genes pronounced differences in the abundances were found between the individual agricultural soils, indicating that the different soils may sustain to different extents specific populations of pseudomonads producing DAPG, PRN, and/or PHZ.

The relative capacity of the 10 different Swiss agricultural soils to sustain the expression of biosynthetic genes for the antimicrobial compounds DAPG, HCN, PRN, and PHZ on roots was followed using GFP-tagged Pseudomonas strains carrying mCherry-based reporter plasmids inoculated into soil microcosms planted with spring wheat. P. protegens CHA0-gfp carrying plasmid pME9012 (phlA-mcherry), pME9011 (hcnA-mcherry), or pME11011 (prnA-mcherry), or P. chlororaphis PCL1391-gfp carrying pME11017 (phzA-mcherry) served as reporter strains. GFP fluorescence (identifying the tagged reporter strains) and relative mCherry fluorescence intensities (reporting expression levels of respective antimicrobial genes) of cells in root washes extracted from the different soils were recorded with FACS-based flow cytometry. Data presented as total antimicrobial gene expression by all cells of the respective reporter strain per gram dry weight of roots, and they were calculated by multiplying the median gene expression per cell with the total number of reporter cells per gram dry weight of roots (Supplementary Table S3).

The levels of total expression of all investigated antimicrobial genes varied significantly among the 10 field soils, with the highest variations observed for DAPG and HCN biosynthetic genes (Figure 4). The soil from Cadenazzo supported the highest levels of total phlA expression on the roots, which were approximately 16-fold higher than those measured on roots growing in soil from Taenikon, which yielded the lowest expression levels (Figure 4A). The Cadenazzo soil also supported highest levels of total hcnA expression among all 10 soils, and these levels were approximately 20-fold higher than hcnA expression levels recorded in the soil from the Grangeneuve field site, which was the least favorable to the expression of this antimicrobial gene (Figure 4B). Levels of total expression of PRN and PHZ biosynthetic genes appeared to be less variable among the 10 soils. In particular, prnA expression levels were eightfold higher in the soil from Utzenstorf that supported the highest expression compared to the soil from Eschikon that supported the lowest expression (Figure 4C). Levels of overall phzA expression varied only about fivefold between the most contrasting soils from the Cadenazzo and Taenikon field sites, respectively (Figure 4D). Several individual soils sustained total expression levels for all investigated antimicrobial genes to a similar extent. This was evident notably for soils from Cadenazzo and Vouvry in which all four antimicrobial genes attained high total expression levels, while soils from Eschikon, Grangeneuve and Taenikon supported significantly lower overall expression levels of these genes (Figure 4).

The Cadenazzo and Vouvry soils appeared to favor also highest levels of root colonization (1.9 × 10⁷ to 2.4 × 10⁷ cells g^-1 of dry root weight) by P. protegens and P. chlororaphis among the 10 soils tested, while the soils from Eschikon, Grangeneuve, and Taenikon were among those supporting lower levels of root colonization (2.9 × 10⁶ to 9.7 × 10⁶ cells g^-1 of dry root weight) (Figure 5 and Supplementary Table S3). However, whereas the soil from Cadenazzo was indeed also the field soil supporting highest phlA, hcnA, prnA, and phzA expression levels in individual reporter cells, the median single cell expression of these genes was significantly lower in the soil from Vouvry (Supplementary Table S3). For the Vouvry soil, it thus seems that the high total expression in the reporter population on the roots was mainly due to the higher colonization levels attained in this soil. Likewise, the relatively high single cell expression but low colonization levels in the Eschikon and Taenikon soils (Figure 5 and Supplementary Table S3) resulted in the significantly lower overall expression as compared to the Cadenazzo and Vouvry soils (Figure 5). By contrast, in the soil from Grangeneuve, single cell expression levels of antimicrobial genes as well as colonization levels were consistently relatively low (Supplementary Table S3). Similar overall gene expression levels may therefore reflect contrasting levels of single cell gene expression and colonization in individual soils.

In summary, the expression of the investigated antimicrobial genes strongly varied between the studied agricultural soils. Some soils seem to favor higher levels of overall antimicrobial gene expression whereas others apparently support the expression of these genes much less. The findings suggest that the expression of different antimicrobial genes may be induced by the same soil factors (or similar combinations thereof).

Results obtained from experiments on pathogen resistance of soils, abundance and expression of antimicrobial genes were displayed in a gradient map (Figure 6). Soils were grouped in three clusters. The first cluster consisted of soils supporting high abundances but low expression levels of antimicrobial genes (i.e., soils from Eschikon and Taenikon field sites). The second cluster consisted of soils supporting high antimicrobial gene expression levels but low abundances of bacteria harboring antimicrobial genes (i.e., soils from Cadenazzo and Vouvry field sites). The third cluster consisted of soils ranging in between the two other clusters. Two soils displayed high resistance against soilborne plant pathogens, i.e., the soil from Taenikon against Gt and the soil from Cazis against Pu. Although the soil from Taenikon was the one supporting the highest abundance of pseudomonads harboring the DAPG biosynthetic gene phlD, resistance to the soilborne pathogens did not generally cluster with high abundances or expression levels of antimicrobial genes (Figure 6).

Correlation analysis revealed that the expression of the four studied antimicrobial genes was positively correlated (Figure 7B). A similar trend could be observed for the abundance of the antimicrobial genes (Figure 7A). Resistance to soilborne pathogens was not significantly correlated to the abundance of antimicrobial genes, although a weak positive correlation could be observed between the abundance of bacteria harboring individual antimicrobial genes and resistance to pathogens (Figure 7A). These results suggest that the expression of the investigated antimicrobial genes is similarly influenced by biotic and abiotic factors prevailing in the respective soils. The same could be true for the abundance of bacteria harboring these antimicrobial genes. Our results further indicate that the abundance of such bacteria probably plays only a limited role in the resistance of the investigated cereal crop-oriented agricultural soils to the soilborne pathogens Pu and Gt.

The physical and chemical properties of the 10 Swiss agricultural soils investigated in this study were analyzed (Figure 1 and Table 1) and correlated with their resistance to pathogens, the abundance of antimicrobial genes and the expression of antimicrobial genes (Figure 8). Macronutrients were extracted from soils with the water or ammonium acetate-EDTA (AA-EDTA) soil extraction procedures routinely used in Swiss agriculture to account for, respectively, soluble, i.e., readily plant-available macronutrients and bound, reserve macronutrients that become available to plants at mid or long term (Agroscope, 2006; Stünzi, 2007). Plant micronutrients were extracted with AA-EDTA.

Organic carbon and clay, and silt and sand inversely influenced antimicrobial gene abundance and expression. Abundance of DAPG biosynthesis genes (recorded for phlD) was significantly positively correlated with clay and significantly negatively correlated with silt, while expression of these genes (recorded for phlA) was significantly negatively correlated with organic carbon and clay and significantly positively correlated with silt (Figure 8). Similar trends were also observed for the pyrrolnitrin and phenazine biosynthetic genes prnA and phzF, respectively. No clear positive or negative correlation was found between pH and antimicrobial gene abundance or expression (Figure 8). Nitrate concentration in soil was positively correlated with the abundance of bacteria harboring the investigated antimicrobial genes; in particular, it was significantly positively correlated with the abundance of phlD⁺ pseudomonads (Figure 8). Antimicrobial gene expression was neither clearly positively nor clearly negatively correlated with nitrate. Of the other macronutrients, potassium inversely influenced abundance and expression of antimicrobial genes. Reserve potassium extracted with AA-EDTA was significantly positively correlated with the abundance of phlD⁺ and phzF⁺ pseudomonads and significantly negatively with phlA and phzA expression (Figure 8). A similar trend was also observed for water-extracted, i.e., readily plant available magnesium. Among the measured micronutrients, a significant effect was only observed for manganese, which was significantly negatively correlated with phzA and hcnA expression (Figure 8).

Resistance to soilborne pathogens was not significantly correlated to any soil factor in the case of Gt, and significantly positively correlated to nitrate and magnesium extracted with water in the case of Pu (Figure 8). The resistance to both pathogens was positively, but not significantly correlated to organic carbon, clay, potassium, manganese, and zinc.

Taken together, these analyses indicate that soil physical and chemical properties have contrasting and subtle effects on the abundance and expression of antimicrobial genes. No pronounced correlations between soil properties and general disease resistance of soils could be observed for the 10 agricultural field sites investigated. Remarkably, all soil factors that were positively correlated with the abundance of phlD⁺ and phzF⁺ pseudomonads were also positively correlated with Pu resistance.