Collimonas pratensis Ter91 (β-Proteobacteria), Paenibacillus sp. P4 (Bacilli) and Pedobacter sp. V48 (Sphingobacteria) have been isolated isolates from the rhizopshere of Marram grass in sandy dune soils in The Netherlands (De Boer et al., 1998, 2004); Serratia plymuthica PRI-2C strain (Y-Proteobacteria) was isolated from maize rhizosphere, The Netherlands (Garbeva et al., 2004) Pseudomonas fluorescens Pf0–1 was isolated from an agricultural soil in Massachusetts, USA (Compeau et al., 1988). All strains were pre-cultured from frozen glycerol stocks on 1/10 strength Tryptone Soya Broth agar (CMO129, Oxoid).

The bioassay was performed as described in the Figure S1. The top area of the glass Petri dish contained 12 ml water-agar medium (20 g L⁻¹ of Agar, 5 g L⁻¹ of NaCl, 1 g L⁻¹ of KH₂PO₄ and 0.1 g L⁻¹ (NH₄)₂SO₄; pH 6.5_.). This carbon-limited medium was used to represent the situation in the bulk soil where bacterial growth is limited by availability of easily degradable carbon compounds.The water-agar medium was inoculated with Pseudomonas fluorescens Pf0–1 of which 5.0 × 10⁶ cells were spread over the water-agar surface. The bottom area of the glass Petri dish contained 45 g of sterile washed sea sand (Honeywell Specialty Chemicals Seelze GmbH, Germany) supplemented with 4.5 ml artificial root exudates and bacterial inoculum (3.0 × 10⁶/gr sand) from monocultures of Collimonas pratensis Ter 91; Paenibacillus sp. P4; Pedobacter sp. V48; Serratia plymuthica PRI-2C or a mixture of these soil bacteria. As control treatment P. fluorescens Pf0–1 was exposed only to sand with artificial root exudates without bacterial inoculum. The artificial root exudates (ARE) stock solution contained 18.4 mM glucose; 18.4 mM fructose; 9.2 mM saccharose; 4.6 mM citric acid; 9.2 mM lactic acid; 6.9 mM succinic acid; 18.4 mM L-serine; 11 mM L-glutamic acid and 18.4 mM L-alanine (C/N 10.4). To each plate 4.5 ml of ARE working solution consisting of 1.5 ml of stock solution mixed with 3 ml of 10 mM phosphate buffer (pH 6.5) was added as described in Baudoin et al. (2003). The plates were incubated at 20°C while packed in aluminium foil. After 3 days of incubation bacterial numbers in the top and bottom compartments were determined. P. fluorescens Pf0–1 cells that had developed on the top water-agar area were scraped and suspended in 3 ml 10 mM phosphate buffer (pH 6.5). One hundred and fifty micro liter of this bacterial suspension was used for OD measurements and plating of serial dilutions on 1/10 TSBA medium (the remaining 2.85 ml were used for RNA extraction, see below). For enumeration of bacteria growing in the bottom area 1 g of sand was taken from each plate and transferred into a 20 ml Greiner tube. Ten milli liter of 10 mM phosphate buffer (pH 6.5) were added and the tubes were shaken on a rotary shaker at 350 rpm for 30 min at 20°C. Subsequently, serial dilutions were plated in triplicate on 1/10 TSBA. All plates were incubated at 20°C and bacterial colonies were counted after 48 h.

For total RNA extraction all suspensions retrieved from agar (see above) were diluted in sterile phosphate buffer to the same optical density (OD; 600 nm) to obtain equal amounts of cells for RNA extraction. The cell suspensions were centrifuged at 16,000 × g for 3 min. RNA was extracted from the cell pellets with the Artrum Total RNA Mini Kit (BIO-RAD cat# 732-6820) according to the manufacturer's recommendations. The extracted total RNA was treated with the TURBO DNA-free Kit to remove DNA (Ambion cat#1907).

Transcriptomic analyses were performed using high-density, multiplex (12x72K) microarrays designed and produced by Roche NimbleGen (Cat# A7241-00-01). Arrays consisted of 60-mer probes covering 5735 genes, 6 probes per gene, 2 replicates. cDNA synthesis, labeling of cDNA with Cy3 dye and hybridization were performed by the Micro Array Department (MAD), University of Amsterdam, The Netherlands (www.microarray.nl).

Each treatment and control were performed in triplicates. The lists of differential expressed genes were extracted by comparison of each interaction with the control. The Robust Microarray Analysis (RMA)-normalized gene expression data were analyzed with the Array Star 2 software for microarray analysis (DNASTAR, Madison, Wisconsin, USA). Analysis was performed, with application of false discovery rate (FDR; Benjamini Hochberg) and multiple testing corrections.

Quantitative RT-PCR was performed to verify the gene expression detected by microarray analysis. First strand cDNA was synthesized with random hexamer primers from Invitrogen (cat# 48190-011) using SuperScript™ Double-Stranded cDNA Synthesis kits (Invitrogen cat#11917-010). Two μL of cDNA was subjected to real-time PCR using SYBR Green PCR master mix (Applied Biosystem). For each target gene [five differentially expressed genes: catalase; sulfotransferase, methyl-accepting chemotaxis sensory transducer, cytochrome C oxidase, chemotaxis sensory transducer and two non-differentially expressed control housekeeping genes: 16S rRNA and RNA polymerase (rpoB)], forward and reverse primers were designed using Primer Express software (PE Applied Biosystem, Warrington, UK). All primers used for real-time PCR were first tested using conventional PCR with DNA isolated from P. fluorescens Pf0–1. Real-time PCR was performed using a Corbett Research Rotor-Gene 3000 thermal cycler (Westburg, Leusden, The Netherlands) with the following conditions: initial cycle 95°C for 15 min and 40 cycles of: 95°C for 15 s; 56°C for 50 s and 72°C for 50 s. The relative expression of the genes was normalized to that of the house keeping genes.

For the collection of bacterial volatiles that were produced in sand containing artificial root exudates, glass Petri dishes with leads, to which a steel trap containing 150 mg Tenax TA and 150 mg Carbopack B (Markes International Ltd., Llantrisant, UK) could be fixed, were used (Figure S1b). Volatiles were collected during 72 h of incubation at 20°C, traps were removed, capped and stored at 4°C until analysis.

Volatiles were desorbed from the traps using an automated thermodesorption unit (model Unity, Markes International Ltd., Llantrisant, UK) at 200°C for 12 min (He flow 30 ml/min). The trapped volatiles were introduced into the GC-MS (model Trace, ThermoFinnigan, Austin, TX, USA) by heating the cold trap for 3 min to 270°C. Split ratio was set to 1:4, and the column used was a 30 × 0.32 mm ID RTX-5 Silms, film thickness 0.33 μm (Restek, Bellefonte, PA, USA). Temperature program used was as follows: from 40 to 95°C at 3°C/min, then to 165°C at 2°C/min, and finally to 250°C at 15°C/min. The VOCs were detected by the MS operating at 70 eV in EI mode. Mass spectra were acquired in full scan mode (33–300 AMU, 0.4 scan/s). Compounds were identified by their mass spectra using deconvolution software (AMDIS) in combination with NIST 2008 (National Institute of Standards and Technology, USA, http://www.nist.gov) and Wiley 7th edition spectral libraries and by their linear retention indexes (lri).

The lri values were compared with those found in the NIST and the NIOO lri database. Mass spectra and lri values for identification were also checked by analysis of pure compounds.

Several volatiles produced by Collimonas pratensis Ter91 and Serratia plymuthica PRI-2C were commercially available. A number of these compounds, namely methanthiosulfonate (CH₃SO₂SCH₃); S-methyl thioacetate (C₃H₆OS); dimethyldisulfide (CH₃S₂CH₃) and benzonitrile (C₆H₅CN), were tested for their individual effect on P. fluorescens Pf0–1 growth. Each volatile was applied in concentrations ranging from 3, 12, 30 to 60 μmol as a droplet on a filter paper on the bottom of the Petri dish. The effect of these pure compounds on P. fluorescens Pf0–1 growth was determined by CFU enumeration as described above.

For extraction of secondary metabolites the water-agar inoculated with P. fluorescens Pf0–1 was removed carefully from the plate and cut in small (1-cm-diameter) pieces. These pieces were vigorously shaken in 20 mL of 80% (v/v) acetone for 1 h at room temperature. The acetone solution was centrifuged for 10 min at 4000 × g and the acetone was evaporated under air flow. The water fraction was acidified with trifluoroacetic acid [0.1% (v/v)], mixed with 2 volumes of ethylacetate and shaken vigorously for 5 min. After incubation overnight at −20°C the unfrozen (ethylacetate) fraction that contains the active compounds was carefully transferred to a new flask and dried under air flow. The dried extract was dissolved in 150 μl of 50% (v/v) methanol and subjected to reverse phase high pressure liquid chromatography (RP-HPLC) analysis and test for antimicrobial activity.

The antimicrobial compounds dissolved in 50% methanol were tested for activity against the fungi Rhizoctonia solani AG2.2IIIB and Fusarium oxysporum, and the bacteria Bacillus sp. V102 and Collimonas pratensis Ter91 (as described in Garbeva et al., 2011a,b).

All experiments were performed in triplicate with three independent replicates for each treatment and controls. ArrayStar 2 (DNASTAR, Madison, WI) was used for statistical analysis of differentially expressed genes applying Student's t-test with Benjamini-Hochberg false discovery rate correction. The statistical analyses of fungal biomass, bacterial enumeration, antagonistic tests and qRT-PCR were carried out with XLStat 2010 (Addinsoft, New York, USA) using a Student's t-Test. Data were considered to be statistically different at p ≤ 0.05.

After 3 days of incubation, the four bacterial strains that were grown in sand containing artificial root exudates had reached similar cell densities (number of CFUs) (Figure 1). All four strains and also the mixture of strains produced volatiles in the sand microcosms (see next section), but the effect of these volatiles on the growth of P. fluorescens was different (Figure 2A). Volatiles produced by C. pratensis and S. plymuthica stimulated the growth of P. fluorescens, whereas volatiles emitted by Paenibacillus sp., Pedobacter sp. and the mix of all 4 bacteria did not affect P. fluorescens growth.

GC-MS analysis revealed that besides commonly known bacterial VOCs such as dimethylsulfide, 2-pentanone, 4-heptanone, 2-heptanol, and 2-undecanone, each bacterial species produced a different blend of volatiles in sand supplied with artificial root exudates (Table 1). The highest numbers of unique volatile compounds were emitted by C. pratensis and S. plymuthica. Several of these volatiles (like S-methyl thioacetate, methyl thiocyanate, dimethyl disulfide, benzonitrile) were produced by both bacteria. Paenibacillus sp. and Pedobacter sp. produced less different volatile compounds and this was also the case for the mixture of four bacterial species. Interestingly, the volatiles produced by the bacterial mix included compounds that were not detected in the spectrum of volatiles produced by the different bacterial monocultures (like 1-tetradecanol, isopropyl dodecanoate, branched alcane, and unknowns).

Microarray-based analyses did reveal strong differences in expression of P. fluorescens gene when exposed to volatiles emitted by the different bacterial species (Figure 2B). Only a small set of 22 genes was differentially expressed by volatiles of all bacteria, including the mixture. These genes were mainly involved in amino acid transport and—metabolism, energy production and conversion, signal transduction mechanisms, inorganic ion transport and—metabolism, secretion and cell motility (Table 2). In addition, all exposures to bacterial volatiles resulted in increased expression of a gene encoding catalase, an enzyme involved in the protection of cells against damage by reactive oxygen species. The RT-PCR analysis of 5 selected differentially expressed genes confirmed the microarray data (Figure S2). The highest number of differentially expressed genes in P. fluorescens was obtained when exposed to volatiles produced by C. pratensis, Paenibacillus sp. and S. plymuthica (Tables S1–S3) whereas volatiles emitted by Pedobacter sp. and the bacterial mix affected the expression of much less genes (Tables S4, S5). There was high similarity in the effect of volatiles of C. pratensis and S. plymuthica on gene expression (116 common differentially expressed genes) of P. fluorescens which corresponds to the high similarity in the composition of volatiles produced by these two bacteria (Tables S1, S3).

Four volatiles produced by C. pratensis and S. plymuthica namely methanthiosulfonate; S-methyl thioacetate; dimethyldisulfide and benzonitrile, were tested individually for their effect on P. fluorescens. Benzonitrile and dimethyldisulfide stimulated the growth of P. fluorescens growth when applied in concentrations above 3μmol (Figure S3). Methanthiosulfonate and S-methyl thioacetate did not reveal any effect on the Pf0–1 growth.

To test if bacterial volatiles triggered P. fluorescens Pf0–1 secondary metabolites production, ethylacetate extracts were made from water-agar on which P. fluorescens had been grown while exposed to volatiles emitted by C. pratensis. Comparison of the HPLC profiles and activities of these extracts with those extracts obtained from controls revealed that P. fluorescens Pf0–1 produced many more secondary metabolites (11 vs. 7) with higher intensity when exposed to volatiles produced by C. pratensis (Figure S4). Furthermore, these extracts showed antimicrobial activity against a Gram-positive bacterium (Bacillus sp.) and the plant pathogenic fungus (Fusarium oxysporum) but did not affect growth of C. pratensis.

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.