Selected species of BeSO, all known to naturally occur in Swiss soils (Campos-Herrera et al., 2015a; Jaffuel et al., 2016; Schlaeppi et al., 2016; Imperiali et al., 2017), were applied depending on the objective and design of each field experiment (Figure S1). The BeSO that were used included two species of the genus Pseudomonas, three AMF species and four EPN species and they were applied as inoculants either individually or in various combinations in the different experiments (Tables 1, 2).

To monitor the bacteria following field application, the bacterial inoculants, i.e., P. protegens strain CHA0 (Stutz et al., 1986) and P. chlororaphis strain PCL1391 (Chin-A-Woeng et al., 1998) were tagged with a spontaneous resistance to rifampicin following previously described protocols (Natsch et al., 1994). Briefly, spontaneous rifampicin-resistant derivatives were obtained following plating concentrated cell suspensions of each parental strain on King's medium B agar (KMB) (King et al., 1954) supplemented with 100 μg/ml of rifampicin and incubated for 3 days. A CHA0-Rif derivative and a PCL1391-Rif derivative (Table 1), which stably maintained rifampicin resistance and displayed wild-type growth and antifungal and insecticidal activities, were selected. For the preparation of the bacterial field inocula, the selected rifampicin resistant strains were grown overnight at 25°C in lysogeny broth (LB) (Bertani, 1951) containing 100 μg/ml of rifampicin. Aliquots of 200 μl of each culture were spread on KMB plates without antibiotics. After incubation at 27°C for 16 h, bacterial cells were harvested and washed in sterile distilled water. The optical density at 600 nm (OD₆₀₀) of the bacterial cell suspensions was adjusted to 0.15 corresponding to a cell density of 8 × 10⁷ CFU ml⁻¹. These bacterial stock suspensions were maintained on ice until final dilution and use on the field sites.

The AMF strains Claroideoglomas claroideum SAF12, Funneliformis mosseae SAF11 and R. irregulare SAF22 were selected from the Swiss Collection of Arbuscular Mycorrhizal Fungi (SAF) at Agroscope (Reckenholz, Zurich, www.agroscope.ch/saf; Table 1). The inocula were prepared as described by Schlaeppi et al. (2016). Briefly, AMF were propagated over 6 months in the greenhouse in autoclaved sand:soil (85:15%; v/v) as substrate and using Plantago lanceolata as host. The final inoculum contained pieces of plant roots mixed with the substrate containing AMF hyphae and spores (SAF12, Propagation 0510, P. lanceolata roots were colonized by 28% and 763 spores could be washed from 25 g substrate; SAF#11, Propagation 0711, 27% and 29 spores; SAF#22, Propagation 0813; 97% and 475 spores). In addition, a “mock” inoculum consisting of Plantago roots and substrate free of AMF propagules was prepared following the same protocol and this mock treatment was termed “AMF control” (Quality inspection of the mock inoculum, Propagation 0711, roots were not colonized by AMF and no spores could be washed from 25 g of substrate). The COMBINATION experiment also comprised a treatment where nothing was applied, named “control”. AMF inocula as well as the mock-inoculum were mixed in separate plastic bags and stored at room temperature until use. In addition, the commercial AMF inoculum R. irregulare TOP (INOQ GmbH, Schnega, Germany, www.inoq.de) was used as obtained from Otto Hauenstein Samen AG (Rafz, Switzerland, www.hauenstein.ch). For the second trial (PERFORMANCE-2), we utilized the lower dosage of the commercial inoculum based on the COMBINATION experiment results. For the PERFORMANCE-2 trial we utilized autoclaved commercial inoculum as AMF control treatment.

For the EPN, infective juveniles (IJs) of four species were prepared in adjusted suspensions. Heterorhabiditis species were obtained from a commercial source (Andermatt Biocontrol, Grossdietwil, Switzerland, www.andermattbiocontrol.com), whereas Steinernema species were propagated from field collected populations under laboratory conditions following protocols described by Campos-Herrera et al. (2015a) (Table 1). All nematodes were received or harvested within 2 weeks prior to field application. The day before application, the EPN inoculant suspensions were prepared in sterile water. To this end, IJs were counted and their density was adjusted to deliver the required field concentration per experimental unit (Table 2) by using separate containers. Containers were kept at 5°C overnight and transported in coolers to the field. In addition, laboratory infections of Galleria mellonella larvae by the inoculant EPN at field concentrations were used to verify their infectivity for each experiment (Jaffuel et al., 2017).

From spring 2014 to summer 2015, three field experiments were conducted in wheat plots and the applications of beneficial soil organisms were adapted for each experiment. All the experiments were carried out with the commercial spring wheat variety “Rubli” in the experimental plots, whereas the commercial triticale variety “Trado” was seeded in the buffer zones. Fields were bordered by strips of non-managed grassland. The three experiments were named as follows: COMBINATION (2014), PERFORMANCE-1 (2014), and PERFORMANCE-2 (2015) (Table 2). The selection of the applied organisms and combinations of treatments were adapted on results of the preceding trial. The first experiment (COMBINATION) was set up to test various species of each group of beneficials and first combinatory treatments. The second experiment was designed to evaluate wheat yield effects after combining bacteria and EPN (PERFORMANCE-1). In this experiment, the AMF treatment was not included due to limitations in scaling the production of inoculum for the large plot size. Finally, the PERFORMANCE-2 experiment consisted of the full bacteria-AMF-EPN combinations during the subsequent season (Table 2).

All the experiments were conducted in neighboring experimental field sites located near Prangins, Switzerland (see Table 2 for coordinates). The sites belong to Agroscope, research center of Changins, (Nyon, Switzerland) and have documented crop and management sequences for the last 30 years. The field sites chosen for the experiments had no overlapping areas to avoid cross-contaminations with inoculants. None of the experiments had irrigation systems. The soil type was sandy loam for the COMBINATION and PERFORMANCE-1 trials and loam for the PERFORMANCE-2 trial (Table 2). General agronomic preparations for all the experiments included tillage (15 cm deep) and harrowing about 4 days before seeding. The seeding machine “Hege Seedmatic” (Hege Maschinen, Eging am See, Germany) allowed the customized seeding for each plot size and arrangement (Table 2; Figure S1) and was modified to keep the seed furrows open after placing the seeds. After seeding, the plots were marked for the corresponding treatments (see Figure S1 for the exact field design of each of the three experiments) and inoculated on the same day with the beneficial soil organisms. In combination treatments, the application followed the order bacteria, EPN, and AMF.

Bacteria were applied as a cell suspension to the seed furrows (plant rows) using treatment-specific watering cans. Final cell suspensions were prepared directly on the field from bacterial stock suspensions (OD₆₀₀ 0.15; i.e., 8 × 10⁷ CFU ml⁻¹) by adjusting with water to obtain the required volumes (400 ml per meter of row) and bacterial cells (8 × 10⁸ CFU per meter of row) (Table 2). Similarly, EPN were applied in variable volumes depending on plot size using treatment-specific watering cans. They were applied to entire plots (not just the rows), and in all the cases, the final concentration was 0.5 Mio. IJs/m² (equivalent to 50 IJs/cm², Grewal and Peters, 2005). Finally, AMF inocula were applied manually employing 250-ml glass beakers. The material was applied directly over the seeds in the furrows, thereby gently mixing seeds and inoculum with a small hoe. AMF control plots received the same quantity of AMF-free substrate. Control plots were treated with the same volumes of BeSO-free water. Immediately after treatment application, the seeds were covered with soil using hoes to close the seed furrows. All equipment used for inoculant application was thoroughly cleaned and disinfested between manipulations using 70% ethanol.

Weed control included the application of the herbicides Azur (Omya AG, Switzerland) against monocots 2 weeks after seeding and Apell (Syngenta AGRO SA) against dicots shortly before earing (BBCH 45-50). When necessary, some persistent weeds (Galium spp., Setaria spp.) were controlled manually. No fungicides nor nematicides were applied during any of the experiments. The insecticide Karate Zeon (Lambda-Cyhalotryne, Syngenta Agro GmbH) was applied in the PERFORMANCE-2 experiment against cereal mining dipterae such as the frit fly and hessian fly conducted in 2015, but not in the 2014 COMBINATION and PERFORMANCE-1 experiments. Plots were fertilized once by supplementing nitrogen in liquid (Lonza-sol N liquid, Basel, Switzerland) at 62 kg ha⁻¹ of to reach 155 units N and potassium (K₂O) at 30.6 kg ha⁻¹. The PERFORMANCE-2 trial was covered with a black hail net during the first 2–3 weeks to protect the seeds and young plants from cold conditions and predation by birds and small mammals.

All experimental field trials presented a Randomized Complete Block design (Figure S1). The data from each group of beneficial organisms were analyzed following standard procedures for their data presentation, transformation, standardization, and normalization whenever necessary. In the case of the EPN activity, data from the G. mellonella baits were expressed as the percentage of larval mortality per plot, averaged by treatment. The activity was determined with respect to the total mortality caused only by nematodes. For the EPN soil food web analysis, all the organisms (EPN, FLN, NF, and bacteria) quantified by using qPCR were expressed per 100 g of dry soil. The parasitism of nematodes by NF was expressed as “infection rate” (IR), which was calculated by dividing the DNA quantity of each species by the total amount of DNA (Campos-Herrera et al., 2012; Duncan et al., 2013). Similarly, to estimate the total FLN and NF, we divided all data within a species by the highest measurement for that species, which allowed the standardization of the units of measurement among species ranging from 0 to 1 (de Rooij-van der Goes et al., 1995).

Unless specified, all significant differences between treatments were assessed by one–way ANOVA, using Tukey's HSD test, considering block as co-variable (V 20.0, IBM SPSS Inc., Chicago, IL). In some cases, t-tests were employed to compare pre- and post-augmentation or control vs. a specific treatment. If necessary, data were transformed to conform the assumptions of normality and equal variances (transformation method is indicated with the respective statistics). The bacterial colonization data were statistically assessed with a non-parametric Kruskal-Wallis test, followed by a post-hoc test (Dunn's test). With the exception of the Pseudomonas root colonization data (presented as log₁₀ of the obtained values ± SEM) all data are presented as mean ± SEM of untransformed values.

In the COMBINATION trial, P. protegens CHA0-Rif and P. chlororaphis PCL1391 reached similar population densities that surpassed the threshold of ~10⁵ CFU per gram of roots, which is the level needed for a plant-beneficial effect (Haas and Défago, 2005) (Figure 1A; Table 4). However, in the PERFORMANCE-1 trial, the population density of the P. chlororaphis strain on wheat roots was significantly lower (P < 0.05) than the density of the P. protegens strain at all three monitoring times (Figure 1B; Table 4). In this trial, in general, for strain CHAO-Rif, alone or in the combinations, we observed a better progression of the population if compared with the strain PCL1391-Rif. If both strains were present in the same treatment, our agar plates almost only reported P. protegens CHA0-Rif colonies (personal observation). Moreover, in clear contrast to CHA0-Rif, PCL1391-Rif never approached the population threshold for plant-beneficial effects, neither when applied alone nor when combined with other BeSO, (Figure 1B; Table 4). In combination with a commercial population of the EPN H. bacteriophora, the density of CHA0-Rif was significant reduced for the June 2014 sample. July 2014, i.e., about 1 month later, CHA0-Rif still maintained its population density in presence of the nematode inoculant while PCL1391-Rif was no longer detectable (Figure 1B; Table 4). In the 2015 PERFORMANCE-2 trial, bacterial numbers approached or surpassed the threshold for plant-beneficial effects at all three sampling times (Figure 1C; Table 4). In general, no significant differences among treatments were observed, but there was a trend of higher bacterial population densities in the April 2015 and May 2015 samplings when bacterial inoculants where combined with the EPN inoculant mixture. In contrast, an opposite trend was observed for June 2015 samples. Finally, as already observed in the PERFORMANCE-1 trial, strain P. protegens CHA0-Rif dominated the colonization, while P. chlororaphis PCL1391 was hardly detected (Figure S2).

In all three field trials, no rifampicin-resistant bacteria were detected in the non-inoculated control treatments, in the buffer zones or in the grassland border zones at the experimental sites (data not shown), hence, the reported CFU data for the augmented bacteria required no baseline correction.

In general, the applied bacteria survived under field conditions until the end of the crop season (Figure 1), although the threshold required to provoke beneficial plant effects (~10⁵ CFE per g root) was not always attained in all the trials or treatments. Inoculation was successful in all, with good traceability of the different bacterial inocula without cross-contamination. Consistently, the P. protegens inoculant showed higher presence on wheat roots, but the effects of the combination with other BeSO were not conclusive and depended on the BeSO species, on the time of exposure to the field conditions and differed between trials.

For the COMBINATION trial we mainly used the locally well-adapted AMF R. irregulare (Schlaeppi et al., 2016) (Table 1). We confirmed successful wheat root inoculation for both R. irregulare strains that we tested, as well as our custom strain SAF22 and the commercial inoculum INOQ Top (Figure 2A; Table 4). The higher dosage of the inoculum INOQ Top (80 g per row) corresponded approximately to the amount of SAF22 inoculum and it appeared that both R. irregulare strains colonized the wheat roots to a similar extent. The reduced dosage of the commercial inoculum (16 g per row) was reflected in lower levels of root colonization and only showed a minor tendency of augmentation. We also traced the inoculation of the F. mosseae strain SAF11 and the C. claroideum strain SAF12 using specific qPCR primers (Thonar et al., 2012). We did not detect these AMF species at the field site (data not shown) confirming the findings of a previous AMF community profiling (Schlaeppi et al., 2016). Hence, we concluded that these strains failed to establish at the tested field site in the wheat roots. In summary, R. irregulare could be augmented in wheat AMF communities using the strains SAF22 or INOQ Top, while this was not successful for F. mosseae SAF11 and C. claroideum SAF12.

Plots in which soil beneficial organisms showed low AMF colonization levels were not different from control (nothing applied) and mock (application of carrier substrate without AM fungus) plots (Figure 2B; Table 4). These measured abundances of R. irregulare correspond to the native strain in the field and indicated that the application of the carrier substrate on its own did not affect the root colonization by the AM fungus. Although, the level of root colonization by R. irregulare showed a slight tendency to increase in the combination treatments of the AM fungus with bacteria, nematodes or both, the AM fungus was not augmented to the same extent as in single application. These first insights on combining soil beneficial organisms suggest possible negative effects on the AMF inoculum if combined with bacteria or nematodes.

In the PERFORMANCE-2 experiment, the commercial R. irregulare strain INOQ Top was inoculated using a lower dosage level to larger plots compared to the previous experiments (Table 2). Again, there was a tendency of increased colonization of the wheat roots in the combined treatments, however, high inter-plot variation precluded statistic support for this effect (Figure 2C; Table 4). It remains to be validated whether the colonization by this R. irregulare strain is particularly facilitated if applied in combination with the Pseudomonas bacteria.

In summary, R. irregulare successfully colonized wheat roots if inoculated alone, and in the combination experiments, we found varying augmentation efficiencies for R. irregulare if combined with pseudomonads, EPN or both, indicating that interactions with these beneficial soil organisms are context dependent.

In all plots, very low numbers of background populations were detected as also found by Campos-Herrera et al. (2015a) and Jaffuel et al. (2017) in the same area. Five to seven species naturally occurred at the experimental field sites, and these species included the taxa that we augmented. Prior to inoculations (baseline; Table 3), there were no differences among treatments for any measured variable (EPN, free-living nematodes and nematophagous fungi; data not shown) in any of the three field trials. The evaluation of EPN soil food web members (free-living nematodes and nematophagous fungi) only revealed natural temporal fluctuations between baseline (pre-inoculation) and post EPN augmentation (data not shown), whereas their presence was not significantly affected by the EPN augmentation (alone or in combination) (Figure S3; Table 4). The nematophagous fungi and free-living nematodes species were in agreement with those already described by Campos-Herrera et al. (2015a,b) and Jaffuel et al. (2017). Finally, the ectophoretic bacterium P. nematophilus was not detected in any of plots (control or augmented).

In the COMBINATION trial, the EPN species H. megidis and S. carpocapsae were recovered in only 25% of the plots, 4 months after augmentation. In contrast, the species S. feltiae and H. bacteriphora, of which the latter was also combined with other BeSO, were detected in 100% of the plots, at the end of the season. The augmentation with S. feltiae was the only treatment with a significant increase in total EPN numbers compared with the native populations (Figure 3A; Table 4). The remarkable persistence of S. feltiae, which was the only species detected in the soil in their corresponding plots, was in agreement with the nematode activity measured in the laboratory as % mortality of G. mellonella producing progeny. This was the only treatment with significantly higher activity than the control in the whole trial (Figure 4A; Table 4).

In the PERFORMANCE-1 trial, we only augmented certain plots with H. bacteriophora. This EPN was detected in about 50% of the plots when applied alone or in combination with the bacterial inoculant P. chlororaphis and in about 75% of the plots when applied with the P. protegens in different combinations. No significant difference in EPN populations (qPCR measurements) and their activity (% larval mortality) was observed between plots where the EPN were applied alone or in combination with bacterial inoculants. Nor were they different from control plots (Figures 3B, 4B; Table 4). In both trials, COMBINATION and PERFORMANCE-1, there was a slight trend to detect more H. bacteriophora in the combined treatment with AMF and/or bacterial inoculants (N2+B1+F1 and N2+BM, respectively) (Figures 3A,B). The same trend was also observed for nematode activity (Figures 4A,B). In the PERFORMANCE-1 field trials, Steinernema affine (Table S1) was the dominant native EPN species in the soil of the experimental plots as determined by qPCR (Figure 3B).

In the PERFORMANCE-2 trial, the augmented EPN species (a mix of S. feltiae, H. megidis, and H. bacteriophora) could be detected in 100% of the plots inoculated with the three EPN, alone or in combination with the Pseudomonas inoculants, in 91% of the plots where they were combined with AMF and in only 44.4% of the plots when combined with both bacterial and AMF inoculants. Again, the species S. affine was dominant among the native taxa as displayed in the proportional chart, although, contrary to the PERFORMANCE-1 trial, native species were largely displaced in all the treatments where EPN were applied (Figure 3C). All plots with EPN application showed significantly higher total numbers of EPN than the control plots (Figure 3C; Table 4). All the augmented EPN species were detected in each of the plots, but S. feltiae and H. bacteriophora dominated. The nematode activity was low and did not significantly vary among treatments (Figure 4C; Table 4). The progeny from the activity tests belonged mainly to H. bacteriophora (62.5%), followed by S. feltiae (34.5%), in all the cases we found mixed EPN-free-living nematodes emergence as observed in previous studies in Swiss soils (Jaffuel et al., 2016, 2017).

In general, inoculated EPN persisted during the crop season and remained active until the time for wheat harvest, but with limited pest suppressive potential as measured with a Galleria larvae infection assay. We observed that EPN application increased the total numbers of EPN only in specific treatments, displacing at least partially the native populations (Figure 3). No long-term effect was observed with respect to soil organisms that can be expected to be modulated by EPN augmentation, such as nematophagous fungi and free-living nematodes. The combined application of EPN with other BeSO indicated compatibility with respect to their persistence, prevalence, and activity, when compared with the single EPN application, but some differences depending on EPN species and co-inoculant identity were observed. As for the bacterial and AMF inoculants, the success of EPN inoculants appeared to be context dependent.

The 2014 trials (COMBINATION and PERFORMANCE-1) were intentionally not subjected to standard pesticide treatments and suffered from heavy attack by frit flies (Oscinella frit). For the small scale COMBINATION trial, insect damage was very patchy and therefore not agronomically representative and not included in the plant performance analyses. The larger plot sizes in the PERFORMANCE-1 trial permitted analysis of agronomically relevant plant density and seed yield data (Table 4). The % of plot surface covered with plants was significantly higher in augmentation plots than in the control treatment when the two bacterial inoculants, P. protegens and P. chlororaphis, were applied individually (treatments B1 and B2, respectively) or as a mixture with and without the EPN (treatments BM and B1+B2+N2, respectively) (Figure 5A). Seed yield per plot followed a similar pattern, but only the combined treatment with both bacterial strains and the EPN showed significantly higher values than the control (Figure 5C). AMF effects could not be examined in the PERFORMANCE-1 trial due to limited inoculum production. Nevertheless, the neighboring COMBINATION experiment indicated that seedling survival after frit fly attack tended to be higher in plots inoculated with R. irregulare (Figure S4). In the 2015 PERFORMANCE-2 trial, plots were subjected to pesticide treatment, no pest damage was observed and all plant traits were considered in the analysis. However, none of these measures, including plant density and seed yield per plot (Figures 5B,D) nor the other plant performance traits (Figure S5; Tables 3, 4) differed significantly from the control treatment.

In summary, when wheat was exposed to biotic stress (i.e., a heavy insect pest attack in 2014) a significant positive effect of the application of BeSO, notably Pseudomonas bacteria, on performance of the crop was observed. The presence of the EPN was only beneficial when combined with both bacteria together. In absence of a biotic stress conditions, as in the PERFORMANCE-2 trial in 2015, there was no measurable plant-beneficial effect of the presence of BeSO, highlighting the context dependence of their protective effect on the crops.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.