The Plant Host Induces Antibiotic Production To Select the Most-Beneficial Colonizers

When grown on a solid biofilm-inducing medium, B. sutbilis biofilms form symmetrical, circular colonies. To determine the effect of a competing S. plymuthica colony on the development of a B. subtilis biofilm, the two species were inoculated next to each other on a solid biofilm medium. After 2 days, the B. subtilis colony reached S. plymuthica, forming a thick wrinkle around its edge and penetrating toward its center. By the third day, B. subtilis biofilm completely engulfed the S. plymuthica colony, covering it with a thin, unstructured film and enclosing it within the circular wrinkle (Fig. 1A). The location and shape of the B. subtilis center of the biofilm did not change during the interaction with S. plymuthica. However, the biofilm colony advanced asymmetrically toward the S. plymuthica colony, breaking from its usual circular shape (Fig. 1A).

We next examined the mechanisms that could mediate this asymmetric expansion toward a competitor. Many bacterial cells are capable of directional swimming using their flagella (28). The genetic basis of flagellar swimming in B. subtilis is well known. B. subtilis has peritrichous flagella (29) that propel the cell forward (30, 31). Each flagellum is composed of a basal body, a hook, and a filament composed of many flagellin monomers (32). The flagellin is encoded by the hag gene, which is transcribed by a σ^D RNA polymerase encoded by sigD (33). In addition, flagellar rotation can also be controlled by chemotaxis (28). To assess whether motility and chemotaxis are responsible for the asymmetrical spreading of the B. subtilis biofilm, two motility mutants, the ΔsigD and Δhag strains, and two chemotaxis mutants, the ΔcheA and ΔcheY strains (28, 31, 34), were tested. Again, the parental B. subtilis and the mutant strains were inoculated next to S. plymuthica at various distances and allowed to grow for 3 days. The asymmetry of each colony was quantified as the ratio between the vertical radius extending toward the S. plymuthica colony and the opposite vertical radius. Wild-type B. subtilis and all four mutants exhibited similar asymmetrical growth toward the S. plymuthica colony, at inoculation distances of <1.5 cm (Fig. 1B). Those results suggest that neither flagellar motility nor chemotaxis is required for the ability of B. subtilis to expand specifically in the direction of a competitor.

We next decided to examine which component of the S. plymuthica colony could promote this behavior. We repeated the experiment, but this time, the asymmetrical expansions toward live S. plymuthica cells, heat-killed S. plymuthica, or a supernatant of S. plymuthica spotted next to the B. subtilis colony were compared. The supernatant of S. plymuthica culture had no effect on the expansion of B. subtilis, further supporting the conclusion that the movement of B. subtilis biofilm was not initiated by signaling molecules secreted by the competitor and did not involve chemotaxis (Fig. 1C). On the other hand, S. plymuthica cells killed by heat promoted asymmetrical expansion, similarly to live S. plymuthica cells (Fig. 1C). This raised the possibility that this collective movement could be aided by organic polymers associated with S. plymuthica cells. We therefore directly tested the possibility of B. subtilis movement being aided by the exopolysaccharide (EPS) of the competitor colony. EPS from an S. plymuthica colony was isolated and spotted next to B. subtilis. Indeed, B. subtilis was able to expand toward isolated EPS but not toward phosphate-buffered saline (PBS) (Fig. 1D). These results suggest that asymmetrical expansion might be achieved by a sliding motion and that EPS can be shared between different species to facilitate this motion.

It was previously shown that B. subtilis is capable of using its own matrix components, primarily EPS encoded by the eps operon, to slide toward new territories (35, 36). We therefore assessed the contribution of matrix components to asymmetric expansion. We tested the ability of the Δeps and ΔtasA mutants (lacking the protein component of the extracellular matrix) to asymmetrically expand toward its competitor. The Δeps mutant could not specifically expand toward S. plymuthica at any of the tested distances. The ΔtasA mutant was capable of asymmetric expansion, but only at shorter distances (Fig. 1E).

Taken together, these results demonstrate that B. subtilis biofilms are capable of directional movement toward their competitors. Interestingly, this collective migration occurs on top of 1.5% agar, which is a concentration 2-fold higher than the maximal concentration permitting sliding in B. subtilis monocultures (37), consistent with similar observations that were recently made for additional interspecies interactions between B. subtilis and Streptomyces spp. (38, 39). This movement does not require flagellar motility or chemotaxis. Instead, the cells in the expanding biofilm colony are likely sliding on the extracellular matrix produced by their own colony and, here, also by their competitor.

Next, we tested the outcome of the above-described interaction between the competing species regarding their viability. We determined the absolute (Fig. 2A and B) and relative (Fig. 2C) CFU of both species before contact with the competitor, at the time of the initial direct contact, and after engulfment. Contact with B. subtilis led to a dramatic reduction in the viability of S. plymuthica (Fig. 2A and C), while the number of viable B. subtilis cells actually increased during the interaction (Fig. 2B and C). Thus, upon reaching its competitor, B. subtilis was able to efficiently kill it. Unlike the ability to slide in the direction of S. plymuthica, the ability to eradicate it was not dependent on extracellular matrix (ECM) production, as all matrix mutants reduced the amount of viable S. plymuthica cells when inoculated at a short distance to compensate for their inability to expand toward it (Fig. 2D; see also Fig. S1 in the supplemental material).

A nonribosomal peptide that contributes to the biocontrol capacity of B. subtilis is surfactin, a surface- and membrane-acting lipopeptide (40, 41). Surfactin mediates both movement by reducing surface tension to promote collective flagellar and sliding motility (42, 43) and toxicity of B. subtilis toward biofilm competitors from related species (44). While the ΔsrfAA mutant had a defect in expansion toward S. plymuthica colonies consistent with its role in promoting sliding motility (data not shown), it was still capable of eradicating the colony cells after the formation of a direct contact (Fig. S2), further suggesting that the ability of B. subtilis to kill its competitor is independent from motility. Taken together, these results demonstrate that asymmetric expansion creates the opportunity for contact but is not essential for the eradication of S. plymuthica by B. subtilis.

B. subtilis has a repertoire of antibacterial molecules it can produce to kill competitors. To determine which of those molecules are involved in the elimination of S. plymuthica in our system, we tested which antibiotics are secreted by B. subtilis under the conditions used in this study. We allowed B. subtilis to form biofilm colonies on solid medium and analyzed the medium to identify molecules secreted by the bacterial cells using liquid chromatography-mass spectrometry analysis (Fig. S3). We detected the presence of bacillaene (45, 46), a polyketide antibiotic synthesized by enzymatic complex encoded in a pks gene cluster and previously implicated in bacterial predator-prey interactions (47) (Fig. 3A, left). We therefore set out to test whether this is the molecule responsible for killing the competing colony.

To determine the role of bacillaene in competitor eradication, we tested the ability of the Δpks mutant, which is unable to produce bacillaene (Fig. 3A, right), to kill S. plymuthica. As expected, the engulfing wild-type B. subtilis colony promoted the death of S. plymuthica cells, as reflected by a drop in the number of viable cells. However, when engulfed by the Δpks mutant, S. plymuthica cells retained stable cell counts (Fig. 3B). Moreover, when the S. plymuthica cells interacting with the wild-type B. subtilis were examined by light and electron microscopy, there was a clear impact on their morphology, and an increased number of lysed cells was observed. No such effect could be detected in cells interacting with the Δpks mutant strain; the cells retained an intact morphology and a thick extracellular matrix (Fig. 3C and D). Next, we extracted the molecules secreted by a B. subtilis colony and examined their effect on planktonic growth of S. plymuthica. Molecules secreted into the solid medium by the wild-type strain, but not the Δpks mutant, inhibited the growth of S. plymuthica (Fig. S4). Moreover, deletion of the pks operon completely eliminated the toxicity of B. subtilis conditioned medium against S. plymuthica (Fig. 3E). Interestingly, the pks operon is highly conserved in bacilli residing exclusively in the rhizosphere but not in B. cereus and B. anthracis, which reside in a broader range of hosts (Fig. S5). Consistent with the hypothesis that polyketide synthesis (PKS) is relevant to the competitiveness of B. subtilis in the soil, the Δpks mutant could not eliminate an additional Gram-negative competitor from a different genus, Pseudomonas chlororaphis (Fig. S6).

In addition to antibiotics, B. subtilis biofilm cells are also capable of producing extracellular proteases that allow them to “mine” extracellular material, such as dead cells, breaking these macromolecules down into amino acids that can then be imported and used for further growth (48). As the CFU of B. subtilis increased following interaction with S. plymuthica cells (Fig. 2B), we decided to examine the contribution of the main extracellular proteases (AprE, Vpr, and Mpr) to the ability of B. subtilis to overcome its competitor during the interaction. The deletion of aprE, but not of the other proteases, inhibited the eradication of S. plymuthica (Fig. 3F and S7). However, its effect on viability of the competitor was less dramatic than that of pks (Fig. 3F), suggesting an indirect mechanism. One intriguing possibility is that consuming dead competitor cells provides B. subtilis with a growth advantage, and the increased rate of growth further enhances its ability to successfully compete and eradicate its neighbor.

Overall, our results separate two biological properties that contribute to the success of B. subtilis during the interaction with S. plymuthica, EPS-dependent sliding movement that leads to asymmetric expansion toward its competitor, and bacillaene-dependent killing once the two colonies make contact (summarized in Table S1).

Both B. subtilis and S. plymuthica reside in the soil rhizosphere of temperate habitats, where the temperature fluctuates between 15°C and 23°C (49). We therefore decided to test whether pks-dependent killing also occurs at the environmental temperature. Surprisingly, while being highly effective in killing S. plymuthica at 30°C, we found that B. subtilis could not efficiently kill it at 23°C (Fig. 4A). However, in the natural niche, additional species are present, most importantly, the host plant. Therefore, we decided to set up a plant-host model system. Eruca sativa (salad rocket) is a winter annual undomesticated wild flowering plant that grows in nature mostly around the Mediterranean basin (50). It belongs to the Brassicaceae family, which includes various natural plant hosts for both Bacillus and Serratia species (51).

In the presence of an E. sativa plant host at 23°C, the killing of S. plymuthica was increased from 2-fold to 26-fold in a pks-dependent manner (Fig. 4B). Again, the killing was linked to predation and resulted in an increased number of B. subtilis cells, and this result could not be attributed to growth induction of B. subtilis by the host (Fig. 4C). In contrast, while the growth of S. plymuthica was indeed improved when grown together with E. sativa (Fig. 4B), it still could not overcome the pks-dependent killing in three-party interaction experiments.

Even in the absence of the plant, the addition of root exudate was sufficient to induce the killing of S. plymuthica by B. subtilis, suggesting that this induction is mediated by a factor secreted by the plant (Fig. 4D). In order to better understand the molecular mechanism of induction of pks-dependent killing, we tested the effect of the root exudate on the expression of the pks operon. As shown, purified root exudate of E. sativa had no impact on cell growth (Fig. 4E) but had significantly (P < 10⁻⁹) induced pks expression (Fig. 4F). In contrast, we could detect no induction of pks, eps, or tasA expression by direct contact with an S. plymuthica colony during expansion and engulfment (Fig. S8) or by its conditioned medium (data not shown).

The transcription of the pksA-R operon (the operon containing pksA and pskR and the genetic material in between) initiates from a promoter located upstream of pksC and is regulated by two master regulators (52), CodY, which senses the levels of branched amino acids (53, 54), and the transition phase regulator AbrB (55). In turn, AbrB is inhibited by the master regulator of sporulation and biofilm development, Spo0A (56). Previous publications indicated that plant exudates of tomato and Arabidopsis spp. can induce Spo0A by several independent mechanisms (57, 58) and therefore may repress AbrB. We used a luciferase reporter driven by the pksC promoter to measure the induction of pks expression by root exudate in the wild type and in the Δspo0A, ΔabrB, and ΔcodY mutant strains. The activation of the pks promoter by the root exudate required Spo0A and AbrB but not CodY (Fig. 4G). Two target genes of Spo0A (sinI and sdpA) activated by low Spo0A-P were also induced by exudates of E. sativa similarly to the pks operon (Fig. S9). Altogether, these results support the hypothesis that the plant increases the pks-dependent killing of S. plymuthica by B. subtilis by inducing the transcription of the pks operon in an Spo0A-dependent manner.

Finally, we wondered whether the plant host has an underlying preference toward B. subtilis. Both B. subtilis and S. plymuthica are considered biocontrol species (40, 59). Therefore, we explored their relative contributions to the activation of the plant’s immune system. Both bacterial species were inoculated on the root, either alone or in competition, and the bacteria were allowed to form root-associated communities. Next, the pathogen Pseudomonas syringae was injected directly into the leaves of precolonized E. sativa plants. After a week, the area of leaf necrosis was measured (Fig. 5A). As no direct contact occurred between the root-associated biofilms and P. syringae, enhanced resistance is the readout of the immune response of the plant hosts. Biofilms containing B. subtilis alone or together with S. plymuthica, but not S. plymuthica alone, provided significantly enhanced systemic resistance (Fig. 5B). When inoculated alone, the Δpks mutant could provide protection similar to that of the wild-type B. subtilis, further demonstrating that pks is not required for this aspect of interaction. However, when inoculated together with S. plymuthica, the Δpks mutant offered no protection to the plant (Fig. 5C). This was due to its inability to compete with S. plymuthica, as demonstrated by the CFU count (Fig. 5D). This result is a strong indication that the pks operon is fundamental to B. subtilis-plant interactions in competitive environments. Furthermore, these results could indicate a mechanism by which plants shape the rhizosphere communities by specifically regulating antibiotic production.

All experiments were performed with the natural B. subtilis isolate NCIB 3610 and its derivatives. The strains used in this study are listed in Table 1 and were referenced previously (45, 46, 57, 79–85). A natural isolate of S. plymuthica was received from the lab of Leonid Chernin, and a natural isolate of P. chlororaphis was received from the lab of Roberto Kolter. 16S rRNA gene analysis was used to confirm the species of the S. plymuthica and P. chlororaphis strains. Assays were carried out with the biofilm-inducing medium MSgg, which contains 5 mM potassium phosphate, 100 mM morpholinepropanesulfonic acid (MOPS) (pH 7), 2 mM MgCl₂, 50 μM MnCl₂, 1 μM ZnCl₂, 2 μM thiamine, 0.5% glycerol, 0.5% glutamate, 50 μg/ml tryptophan, 50 μg/ml phenylalanine, 50 μg/ml threonine, 700 μM CaCl₂, and 50 μM (for growth assays) or 125 μM (for biofilm assays) FeCl₃ (79, 86). Solid MSgg medium was obtained by adding Bacto agar (Difco) to a final concentration of 1.5%. When necessary, selective medium was prepared with LB agar or LB broth, supplemented with 100 μg/ml ampicillin (Amp; AG Scientific), 10 μg/ml kanamycin (Kan; AG Scientific), 10 μg/ml chloramphenicol (Cam; Amresco), 10 μg/ml tetracycline (Tet; Amresco), 100 μg/ml spectinomycin (Spec; Tivan Biotech), and 1 μg/ml erythromycin (Erm; Amresco) plus 25 μg/ml lincomycin (Linc; Sigma-Aldrich). To create the transcriptional luc reporter, the promoter of pksC was amplified from genomic DNA (gDNA) of B. subtilis NCIB 3610 with the following primers: primer 5′ P_pksC-lux P1, GTCCTAGTAAGGTCGACAGGAGGACTCTCTGCAAATCGCCCGGCCATTCGATAAAGG, and primer 3′ P_pksC-lux P2, GTATGTAAGCAAAAAGTTTCCAAATTTCATTCTCTCAAAGCCACCCTTCCGATTAGT, and further integrated into pBS3Clux by restriction-free cloning. After confirmation by sequencing, pBS3Clux was cut and integrated into the bacterial genome of B. subtilis NCIB 3610 at the neutral sacA locus.

The deletion of codY was generated using the long-flanking homology (LFH) PCR mutagenesis protocol of Wach (87), replacing an endogenous locus with a kanamycin resistance gene from pIKec14. The primers used were as follows: P1, TCGATATGGATGAAGTCGGCCAGGAA; P2, CAATTCGCCCTATAGTGAGTCGTCCGCAGCTTGCAGCATGGAGTTAATA; P3, CCAGCTTTTGTTCCCTTTAGTGAGTCAGGCTTATATCAAGGCGAGAAATGTAGTT; and P4, TTCTGTAAGGCACCCACTCTCCATTC.

The product was first introduced by transformation into strain B. subtilis PY79 and the deletion further integrated into the NCIB 3610 wild type or mutant by transformation. Transformation of B. subtilis PY79 and NCIB 3610 by natural competence with linearized plasmid or genomic DNA was done as described previously (88) and as performed previously to assess natural competence in NCIB 3610 (89).

Bacillaene was extracted from the biofilms of strain NCIB 3610. A single B. subtilis wild-type (WT) colony, isolated on a solid LB plate, was inoculated into 3 ml of LB broth and grown overnight at room temperature, and 2 μl of the culture was plated onto solid MSgg medium. The plate was incubated for 3 days at 30°C to allow proper biofilm development. A ∼44-mm² MSgg agar segment was cut out of the plate (together with a piece of the biofilm) and crushed manually, and 700 μl of isopropanol was added to it. The isopropanol-containing crushed MSgg mixture was then incubated at 50°C for 1 h, with occasional vortexing, followed by centrifugation (14,000 rpm for 5 min) to precipitate MSgg agar flakes. Supernatant was extracted and evaporated using a SpeedVac. As a control, the same procedure was performed with noninoculated solid-MSgg medium. To verify that the activity of the extract was pks dependent and not due to the extraction of surfactin, the procedure was also carried out with the Δpks mutant. To analyze the activity of the extracted molecules, 6.6 μl dimethyl sulfoxide (DMSO) and 193.4 μl distilled water (dW) were added to the evaporated residues and vortexed to dissolve them (resulting in a molecule concentration of 3.5×). For S. plymuthica growth curves, liquid MSgg was supplemented with 30% of the 3.5× solution, achieving a final concentration of ∼1×. Growth assays were performed as follows: cells were grown at room temperature with agitation for 18 h in a microplate reader (Synergy 2; BioTek), and the optical density at 600 nm (OD₆₀₀) was measured every 15 min.

Isopropanol extracts were analyzed by liquid chromatography-mass spectrometry. High-performance liquid chromatography (HPLC) separations were carried out using a Kinetex hexyl-phenyl column (2.1 by 150 mm; particle size, 2.6 μm; Phenomenex). The mass spectrometer was operated in positive ionization mode, and the ion source parameters were as follows: spray voltage, 3.5 kV; capillary temperature, 300°C; ion-transfer optics parameters optimized using automatic tune option; sheath gas rate (arb), 35; and auxiliary gas rate (arb), 15. Mass spectra were acquired in the range of m/z 150 to 2,000 Da. The liquid chromatography-mass spectrometry (LC-MS) system was controlled, and data were analyzed using the Chromeleon and Xcalibur software (Thermo Fisher Scientific, Inc.). Bacillaene was detected at m/z 581.3585. Dihydrobacillaene was not detected.

To acquire S. plymuthica supernatant and dead cells, an overnight LB culture of S. plymuthica grown at room temperature was inoculated into liquid MSgg medium (dilution of 1:100) and incubated at 30°C for 24 h. Cultures were then centrifuged (14,000 rpm, 10 min), and the supernatant was filtered, evaporated by SpeedVac, and rehydrated with distilled water (dW), achieving a 20× concentration. The remaining cells were washed with dW and resuspended in 2 ml dW. Cells were heat killed at 85°C for 25 min. For the interaction, 3 μl of either the 20× supernatant or the dead cells was inoculated next to a 2-μl B. subtilis LB culture (grown at room temperature overnight) and then incubated at 30°C for 3 days. For EPS extraction, S. plymuthica cells were grown at room temperature, inoculated into liquid MSgg medium (dilution of 1:100), and grown for 3 days without shaking to generate floating biofilms. Growing medium was collected and filtered, and ECM was precipitated with isopropanol as previously described (90). Solvent was evaporated overnight in a chemical hood.

Single B. subtilis or S. plymuthica colonies, isolated on solid LB plates, were inoculated into 3 ml of LB broth (Difco) and grown overnight at room temperature. For interaction analysis, 2 μl of each culture was inoculated onto solid MSgg medium at the desired distances. Plates were incubated at 23°C and 30°C for 5 and 3 days, respectively. Cells were visualized using a SteREO Discovery.V20 microscope equipped with a PlanApo S 0.5× FWD 134-mm objective or a PlanApo S 1.0× FWD 60-mm objective (Zeiss) attached to an Axiocam camera. Data were captured and analyzed using the Zen Pro AxioVision suite software (Zeiss, Oberkochen, Germany) and ImageJ2 (91).

For quantification of a symmetric spreading of B. subtilis biofilm colony, the radius ratio was determined by dividing a vertical radius extending from the center of the biofilm toward the S. plymuthica colony by the vertical radius extending to the opposite side.

Two-day-old interactions of B. subtilis WT or the Δpks mutant with S. plymuthica were fixed overnight at 4°C with 2% glutaraldehyde, 3% paraformaldehyde, 0.1 M sodium cacodylate (pH 7.4), and 5 mM CaCl₂. Samples were washed twice for 15 min with double-distilled water (ddW) and dehydrated with a series of ethanol washes at 30%, 50%, 70%, 96% and 100%. Samples were then dried overnight on filter paper (Whatman) at room temperature, mounted, and stored in a vacuum. The mounted samples were sputtered with gold-palladium before examination by scanning electron microscopy (SEM) XL30 with field emission gun (FEI).

To analyze the cell numbers for each strain during the interaction between B. subtilis and S. plymuthica, colonies were seeded as described above and incubated for the required time period. As shown in Fig. 1 and 2A to D, the interacting colonies were divided into three sections, as follows: (i) B. subtilis, consisting of the entire B. subtilis biofilm, excluding the thick wrinkle that surrounded S. plymuthica; (ii) the manually separated wrinkle, containing cells from the contact area (also designated interaction area); and (iii) S. plymuthica, consisting of the entire S. plymuthica colony, excluding the cells that were attached to the wrinkle. For other experiments, both interacting colonies were harvested, with no further separation. Colonies of B. subtilis and S. plymuthica grown alone served as controls.

Each sample was harvested, resuspended in 500 μl phosphate-buffered saline (PBS), and mildly sonicated (3× for 5 s at 15%). The samples were then diluted, plated on solid agar, and incubated at 30°C overnight to allow formation of colonies. Colonies formed by the two species are easily distinguishable on LB medium due to differences in size and colony shape. For B. subtilis biofilm mutants, confirmation of the identity of the colonies was performed by dual plating on nonselective LB and LB with the appropriate antibiotics.

To test bacterial interaction over roots, seeds of the annual E. sativa (Brassicaceae), originated from natural population (92), were germinated on Nitsch (93) agar plates at 25°C with an 8-h/16-h day/night photoperiod. One-week-old seedlings were transferred to solid MSgg medium plates, and bacteria were inoculated (as described above) on the next day at a distance of 0.8 cm from the root and between the inoculants. Samples were incubated in a growth chamber (MRC) with 8-h/16-h day/night photoperiods at 23°C and 30°C for 5 and 3 days, respectively. For interspecies interactions population assessment, the overall community was separated from the root, mildly sonicated, and plated to detect single colonies in triplicate.

Root exudate of E. sativa seedlings was extracted from Nitsch agar growth medium 2 weeks from germination. After seedlings were removed, exudate was obtained by extracting Nitsch agar with isopropanol (3:2), followed by vigorous vortexing and incubating at room temperature (RT) for 10 min. Samples were centrifuged in 10,000 rpm for 10 min to separate the aqueous phase. Exudate samples were evaporated in a SpeedVac concentrator (Eppendorf) and rehydrated in dW. Prior to biofilm inoculation, plates were treated with either 50-μl drops of root exudate or Nitsch agar (plant growth medium) extraction used as a control.

pks expression in the presence of plant metabolites was analyzed as follows: the luminescence intensity of cells harboring a P_pksC-lux transcriptional promoter fusion was analyzed. Cells were grown to a mid-logarithmic phase, diluted 1:100 in 150 μl liquid MSgg medium, either with or without root exudate. Cells were grown in a 96-well microplate (Thermo Scientific, Roskilde, Denmark) with agitation at 30°C for 16 h in a microplate reader (Synergy 2; BioTek, Winooski, VT, USA), and the luminescence intensity and the optical density at 600 nm (OD₆₀₀) were measured every 15 min.

All studies were performed in duplicate or triplicate at least three separate and independent times. Data are expressed as average values ± standard deviations of the means. For triple-party interactions, data are expressed as average values ± standard errors. Parametric testing was performed when appropriate after confirming that raw data were normally distributed. A paired Student's t test was used to determine if the set of treated versus the untreated control are different from each other, and P values of less than 0.1 were considered significant.