Secondary metabolite production facilitates establishment of rhizobacteria by reducing both protozoan predation and the competitive effects of indigenous bacteria


  • A. Jousset,

    Corresponding author
    1. Darmstadt University of Technology, Institute for Zoology, Schnittspahnstr. 3, D-64287 Darmstadt, Germany; and
      *Correspondence author. E-mail:
    Search for more papers by this author
  • S. Scheu,

    1. Darmstadt University of Technology, Institute for Zoology, Schnittspahnstr. 3, D-64287 Darmstadt, Germany; and
    Search for more papers by this author
    • These authors have contributed equally to this work.

  • M. Bonkowski

    1. Darmstadt University of Technology, Institute for Zoology, Schnittspahnstr. 3, D-64287 Darmstadt, Germany; and
    2. University of Cologne, Institute for Zoology, Terrestrial Ecology and Rhizosphere Research, Weyertal 119, D-50931 Köln, Germany
    Search for more papers by this author
    • These authors have contributed equally to this work.

*Correspondence author. E-mail:


  • 1Rhizosphere bacteria live in close contact to plant roots feeding on root exudates and rhizodeposits. By producing toxic exoproducts rhizobacteria may inhibit plant pathogens thereby functioning as biocontrol agents and increasing plant fitness. However, the evolutionary basis why rhizobacteria protect plants is little understood. To persist toxin production needs to improve the competitiveness of the bacteria themselves.
  • 2We investigated the importance of secondary metabolite production for the establishment of the model soil biocontrol bacterium Pseudomonas fluorescens CHA0 in the rhizosphere of rice. We compared the performance of this toxin-producing strain and its isogenic gacS deficient mutant defective in secondary metabolite production. The bacteria were added to the rhizosphere of rice, where they had to compete with the indigenous flora for resources and to resist predation by the protist Acanthamoeba castellanii.
  • 3Secondary metabolite production strongly enhanced the establishment of the inoculated bacteria by improving competitive strength and predator resistance. The fitness gain due to attenuation of predation exceeded that due to competition by a factor of 2–3, confirming the importance of grazing resistance for rhizosphere bacteria.
  • 4Biocontrol properties of Plant Growth Promoting Bacteria such as P. fluorescens therefore gain a new dimension. Toxicity primary plays a role in the interaction with competitors and especially predators, and not in the protection of the host plant. Thus, establishment and efficiency of biocontrol bacteria may be improved by fostering predator defence via toxin production.


Plants roots are colonized by specific bacterial communities living on root exudates and rhizodeposits. Some rhizobacteria produce exometabolites, which are toxic to many organisms, including plant soil-borne pathogens. Toxin production makes them interesting as biocontrol organisms, and they are increasingly seen as a green alternative to agrochemicals (Weller 2007). However, the evolutionary basis of the plant protective activity is still puzzling, since the effect appears to lack reciprocity. Toxin production is costly for the bacteria and a reward in form of more root exudates would be too unspecific to select for this trait (Denison et al. 2003). Consequently, we hypothesize that there exist direct feedbacks of toxins produced by bacteria improving bacterial fitness. In particular, toxin-producing bacteria may benefit from reduced competition for resources by indigenous microflora and from increased resistance against predation. Indeed, bacteria lacking toxin production are impaired in their ability to colonize the rhizosphere of plants (Natsch et al. 1994), suggesting that there is an intense selective pressure favouring toxic bacteria. Further, there is evidence that bacterial toxins indeed impair predator pressure (Jousset et al. 2006). However, both reduced competition by competitors and exposure to predators lack experimental proof from rhizosphere systems resembling those in the field.

For testing these hypotheses under natural conditions we established a model rhizosphere system with plant seedlings in soil containing a natural bacterial community with and without predators. The role of toxins for increasing competitiveness and reducing predation was investigated using the model biocontrol organism Pseudomonas fluorescens CHA0 which is widespread in the rhizosphere of mono- and dicotyledonous plants. It owes its biocontrol ability to diverse secondary metabolites, including cyanhidric acid, DAPG and exoproteases (Haas & Keel 2003) which inhibit fungal pathogens, such as Pythium ultimum (Keel et al. 1992) and Fusarium oxysporum (Zuber et al. 2003). As in many pseudomonads, the production of secondary metabolites in P. fluorescens is controlled by a two-component gacS/gacA receptor system (Heeb & Haas 2001), which is involved in the response to density-sensing signals. Activation of the gac cascade up-regulates the production of secondary metabolites through the production of three small RNAs, rsmX, rsmY and rsmZ. These molecules bind to the post-transcriptional inhibitor rsmA (Valverde et al. 2003), thereby inducing translation of the corresponding mRNAs. Interestingly, spontaneous gacS/gacA deficient mutants occur at high frequency (Martinez-Granero et al. 2005). These mutants are defective in secondary metabolite production and are less competitive in non-sterile environments (Natsch et al. 1994; Chancey et al. 2002), but not when bacterial diversity is low (Schmidt-Eisenlohr, Gast & Baron 2003), suggesting that toxins increase bacterial fitness by modulating biotic interactions.

For testing if secondary metabolites indeed increase the competitive strength of P. fluorescens CHA0 we compared the performance of the wild-type (Wt) P. fluorescens CHA0 producing an arsenal of secondary metabolites with its isogenic gacS deficient mutant defective in secondary metabolite production. To determine the role of toxins for improving the competitiveness against other rhizobacteria, we followed the differential ability of these two strains to establish populations in the rhizosphere of plants in absence of predators. For investigating the role of toxins in attenuating predator pressure we compared the establishment of Wt bacteria in presence and absence of protozoan predators (Acanthamoeba castellanii).


experimental system

Natural rhizosphere systems were established in soil microcosms planted with rice. The microcosms were inoculated with a predator-free soil bacterial assemblage. Acanthamoeba castellanii, a ubiquitous soil bacterivorous amoeba, was added as model predator in a factorial design. Soil from a pasture near Heteren, the Netherlands (84·6% sand, silt 8·2%, clay 6·2%, carbon content 2·1%, C : N ratio 16·7; cf. van der Putten et al. 2000) was sieved (2 mm), autoclaved and washed twice with a threefold volume of tap water on a 100-µm mesh to remove nutrients and toxins released by autoclaving. The washed soil was dried for 72 h at 70 °C and rewetted to water holding capacity with distilled water (200 mL kg−1). In order to keep moisture conditions constant during the experiment, 50 mg kg−1 soil of a water retaining polymer were added (Grain d’Eau, La Celle St-Cloud, France). A total of 50 g wet weight soil was filled in 3 × 20 cm glass tubes (Schott, Mainz, Germany) and autoclaved (121 °C, 30 min).

Each tube was inoculated with a protozoa-free soil bacteria assemblage at a concentration of 108 bacteria per gram soil (see below), and incubated in the dark at room temperature in order to allow the bacterial population to grow and equilibrate. After 5 days 106 amoeba per gram soil (total volume 1 mL) were added to the amoeba treatment. The control samples received 1 mL sterile Neff's modified amoeba saline (NMAS; Page 1988). Five days later one rice seedling was transferred aseptically into each tube, and 12 h later the plants were inoculated with 106 P. fluorescens per gram soil. Plants were grown at a constant temperature of 22 °C and 16 h of light (500 µmol s−1 m−2). The tubes were randomized daily.

organisms and culture conditions

Bacteria for inoculation of the sterilized soil were isolated from the experimental soil as described by Kreuzer et al. (2006) with few modifications. Briefly, 10 g of soil were suspended in 100 mL of NMAS and filtered through a paper filter. The filtrate was successively filtered through 5 and 1·2 µm membranes (Millipore, Schwalbach, Germany) to remove protists. The resulting filtrate was mixed in a 1 : 1 ratio with a diluted nutrient solution (0·8 g L−1 nutrient broth in NMAS) and incubated in 10 mL tissue culture flasks. Cultures were checked after 4 and 6 days with an inverted microscope at 100× magnification for contamination by flagellates. Bacterial cultures were harvested by centrifugation (5 000 g. for 5 min) and washed in NMAS prior to inoculation.

Strains of P. fluorescens CHA0 and its isogenic gacS deficient mutant CHA19 tagged with gfp were used (Jousset et al. 2006). The strains were routinely kept on nutrient agar (blood agar base 40 g L−1, yeast extract 5 g L−1). Prior to inoculation a single colony was picked and incubated overnight in NYB medium (nutrient broth 25 g L−1, yeast extract 5 g L−1) at 28 °C and agitation of 300 r.p.m. Cultures were washed twice in phosphate buffer saline (PBS) and resuspended in NMAS. Concentration of bacteria was determined on the base of the OD600 and checked under an epifluorescence microscope as described below.

Acanthamoeba castellanii was isolated from a woodland soil (Bonkowski & Brandt 2002) and kept axenically on PYG medium (peptone 20 g L−1, yeast extract 5 g L−1, glucose 10 g L−1). Prior to inoculation 10 mL of a stationary phase culture were collected and washed twice by centrifugation (100 g, 10 min). The pellet was resuspended in NMAS, and the cell concentration was determined with a Neubauer counting chamber.

Rice seeds (Oryza sativa cv. Zhonghua11) were dehusked by grinding lightly with a pestle in a mortar, and surface sterilized by soaking in 96% ethanol for 1 min, and for 30 min in a solution containing 30 g L−1 NaCl, 13 g L−1 NaClO, 1 g L−1 Na2CO3 and 1·5 g L−1 NaOH (Hurek et al. 1994). Seeds were separately pre-germinated in the dark at 28 °C, in a 96-well microtiter plate containing 100 µL NMAS per well. After 6 days germinated seeds were checked for sterility with an inverted microscope (Nikon Diaphot, 100× magnification). Only sterile seeds were used for the experiment.


On days 4, 8 and 16 after the set up, the microcosms were destructively sampled.

The roots were gently removed by shaking off the soil and rhizosphere bacteria were subsequently extracted as described in Normander, Hendriksen & Nybroe (1999). Briefly, the roots were gently rinsed with sterile PBS to remove soil aggregates and placed in 8 mL PBS buffer. Samples were vortexed for 1 min, bath sonicated on ice for 2 × 30 s and again vortexed for 1 min. Samples (1 mL) were then fixed in 3% formaldehyde, filtered on a 0·22-µm TGTT membrane (Millipore), and stained with 2·5 µg mL−1 DAPI; gfp tagged cells and total bacteria were enumerated with a Zeiss Axioscop 2 epifluorescence microscope at 400× magnification. The filter settings for DAPI and gfp counts were as follows: excitation filter 365 and 470 nm, dichroic filter 395 and 493 nm and emission filter 397 and 505–530 nm, respectively. Images were taken in triplicate using a mounted Canon Powershot G5 digital camera. Bacteria were automatically counted using the ImageJ software ( In order to check the accuracy of the automated counts some samples were counted by eye.

Soil microbial respiration, microbial biomass and the respiratory quotient (qO2, i.e. the ratio of respiration to biomass) were determined using a substrate-induced respiration method (SIR; Anderson & Domsch 1978; Beck et al. 1997) with an automated respirometer as described in Scheu (1992).

Amoebae were enumerated using a modified most probable number method (Darbyshire et al. 1974). Five grams of soil were gently shaken for 15 min in 20 mL NMAS, and 100 µL aliquots (four replicates per sample) were successively diluted threefold in a suspension of 107 Escherichia coli per millilitre in NMAS in 96-well microtiter plates and incubated in the dark at 15 °C. Wells were checked for presence of amoebae after 3, 7 and 14 days under an inverted microscope (Nikon Diaphot, 100× magnification).

statistical analyses

The experiment followed a 2 × 2 factorial set up investigating the effects of the inoculated bacterial strains (Wt and gacS mutant) and predators (presence and absence of A. castellanii), with seven replicates per treatment. The results were analysed by anova using the glm procedure and type III SS. The factors investigated the effect of the inoculated strain (‘Strain’), and of the presence and absence of predators (‘Amoeba’). Two contrasts were set up to further analyse the role of secondary metabolite production for the establishment of P. fluorescence strains: (i) The importance of toxicity in the interspecific competition with other rhizobacteria was evaluated by comparing the performance of the Wt and gacS strain in absence of predators (‘Competition’); (ii) the importance of toxicity against predators was inspected by comparing the performance of the Wt strain in treatments with and without predators (‘Predation’). The relative importance of secondary metabolite production in these two interactions was estimated by comparing the proportion of the total variance explained by these two contrasts. The expected negative effect of the toxic Wt strain on the density of amoebae was evaluated by a one tailed Student's t-test comparing the density of amoebae in the microcosms inoculated with the two P. fluorescens strains. Prior to analyses the variables were inspected for homogeneity of variance, total bacterial densities were log-transformed and percent values were arcsin square root transformed. Statistical analyses were carried out using sas 9·1 (SAS Institute, Cary, NC).



At the end of the experiment, the abundance of amoebae (active and encysted) was two times higher in treatments containing the gacS strain, reaching 4·2 106 ind. g−1 soil, compared to 2·1 106 ind. g−1 soil in treatments with the Wt strain (t-test, P = 0·006), suggesting that Wt bacteria negatively affected predator growth.

total bacteria

Total bacterial density (DAPI counts per gram root) increased from 5·0 108 ind. g−1 roots on day 4 to 6·9 108 ind. g−1 roots on day 8, and then decreased to 4·3 108 ind. g−1 roots on day 16, suggesting an active growth phase until day 8. Amoebae significantly reduced bacterial density by 37% on day 16 (Table 1, Fig. 1). Interestingly, this effect was less pronounced in presence of the Wt strain (significant Amoeba × Strain interaction; Table 1) suggesting that the presence of toxic bacteria protected to some extent the whole bacterial community from grazing (Fig. 1). Similar to total microbial numbers, amoebae reduced soil microbial respiration from 1·003 to 0·802 µL O2 h−1 g−1 soil (anova, F1,28 = 5·7, P = 0·02) with the reduction tending to be more pronounced in the gacS treatment (–28%) than in the Wt treatment (–10%; anova, F1,28 = 3·6, P = 0·07). Microbial biomass (260·6 ± 55·0 µg C g−1 soil) did not significantly differ between treatments.

Table 1. anova table of F- and P-values on the effect of Strain (Wt or gacS) and Amoebae (with and without) on the density of total rhizobacteria and on the relative density of Pseudomonas fluorescens (as percentages of total bacteria) 4, 8 and 16 days after inoculation. In addition to effects of main factors and their interaction, contrasts have been calculated to evaluate the importance of competition and predation on the relative density of Wt or gacS strains of P. fluorescens. Significant effects (P < 0·05) are highlighted in bold
Total bacterial densityd.f.Day 4Day 8Day 16
  Amoebae11·50·2390·50·48816·1< 0·001
  Strain × Amoebae11·30·2692·30·1445·20·031
Relative density of P. fluorescens
  Strain12·10·16638·2< 0·00143·9< 0·001
  Amoebae10·30·60214·2< 0·00113·00·001
  Strain × Amoebae10·60·4336·210·0199·8< 0·001
  Predation10·80·38117·4< 0·00122·6< 0·001
Figure 1.

Effect of inoculation with Pseudomonas fluorescens Wt and gacS strains and the presence of predators (Acanthamoeba castellanii; –A, without; +A, with) on total density of bacteria (DAPI counts; ind. g−1 fresh weight of roots) in the rhizosphere of rice. Black bars: Wild-type (Wt) strain, white bars: gacS strain. Error bars represent ± SE. Horizontal bars show significant effect between treatments (*P < 0·05, **P < 0·01, ***P < 0·001) as described in Table 1.

pseudomonas fluorescens

Root colonization by P. fluorescens (GFP counts per gram root) rapidly increased during the experiment, with the differences between treatments being most pronounced at the end of the experiment. Numbers of gacS mutant bacteria remained low during the whole experiment. Their relative densities decreased slowly from 3·4% of the total rhizobacteria on day 4 to 1·7 and 2·0% on days 8 and 16 (Fig. 2). The Wt strain more successfully colonized the roots. In absence of predators its density significantly exceeded that of the gacS strain by factors of 2·9 and 1·7 at days 8 and 16, respectively (effect of ‘Competition’; Table 1, Fig. 2), suggesting that the toxin-mediated increase in competitive strength against other rhizobacteria was most pronounced on day 8 when the number of total bacteria was at a maximum. The increase in density of the Wt strain was even more pronounced in presence of amoebae. Compared to the gacS strain it was significantly increased by factors of 3·7 and 3·7 at days 8 and 16, respectively (significant effect of ‘Predation’; Table 1, Fig. 2), suggesting that compared to increasing competitiveness secondary metabolites confer an even stronger advantage by attenuating predator pressure. Indeed, the effect of ‘Predation’ were not significant at day 4 (Table 1), but explained 19% and 24% of the total variance in density of P. fluorescens at days 8 and 16, respectively, whereas the effect of ‘Competition’ only explained 11% and 5% (Table 1).

Figure 2.

Relative colonization of the rhizosphere of rice by Pseudomonas fluorescens Wt and gacS strains in absence (–A) or presence (+A) of predators (Acanthamoeba castellanii); data are expressed as percentages of the total number of rhizosphere bacteria in respective treatments (see Fig. 1). Black bars: Wild-type (Wt) strain, white bars: gacS strain. Error bars represent ± SE. Horizontal bars show significant effect between treatments (*P < 0·05, **P < 0·01, ***P < 0·001) as described in Table 1.


Using a semi-natural rhizosphere system, we investigated for the first time the role of bacterial secondary metabolites in modifying the two major structuring forces of food webs, that is, bottom-up (competition for resources) and top-down (predation) control. Conform to our hypothesis and in agreement with past observations in non-sterile soil (Chancey et al. 2002), bacteria lacking secondary metabolite production were less competitive than toxin-producing Wt bacteria. Other traits associated with gac deactivation, such as increased siderophore production (Heeb & Haas 2001), unlikely contributed to reduced competitiveness. This suggests that secondary metabolites of biocontrol bacteria indeed primarily function in improving bacterial fitness via targeting other rhizosphere organisms rather than improving plant growth and pathogen resistance. In addition to affecting competitors, bacterial toxins from P. fluorescens also reduced predator pressure by protists. Remarkably, the gain in fitness (measured as increase in relative density) of toxic bacteria by avoiding predation on themselves and increasing predation on competing bacteria exceeded that caused by improved competitive strength. Further, the advantage due to attenuated predator pressure increased faster than that due to increased competitive strength. Top-down control is an important factor for the establishment of bacterial populations in the rhizosphere (Christensen, Bjornlund & Vestergard 2007). Our results suggest that toxin production is an efficient strategy to alleviate losses from predation and by increasing predation on competitors gaining in competitiveness. In fact, for improving competitiveness this strategy may be more efficient than inhibiting competing bacteria via, for example, investing in bacteriotoxic substances.

This functioning of metabolites in attenuating predator pressure sheds new light on the classical theory of antibiosis, according to which the primary function of bacterial toxins is to damage competing microorganisms (Clardy, Fischbach & Walsh 2006). The limited advantage of Wt strain bacteria in absence of predators in the present and previous experiments (Johansen et al. 2002) and the reduction in predator pressure suggest that in fact secondary metabolites primarily target against predators rather than against competing bacteria and fungi. Rhizobacteria in fact generally appear to suffer heavily from predation, especially by protists and nematodes (Bonkowski 2004), supporting our conclusion that attenuating the impact of predators is vitally important.

Protozoan predation is known to structure bacterial communities in aquatic and terrestrial habitats including the rhizosphere of plants (Bonkowski 2004; Matz & Kjelleberg 2005; Pernthaler 2005). Consequently, adaptations for reducing predation by microfaunal predators, such as biofilm formation and toxin production, are widespread among bacteria (Matz et al. 2004; Queck et al. 2006), including P. fluorescens, which have been shown previously to harm eukaryotic predators (Jousset et al. 2006).

Results of the present study indicate that repelling predators not only reduces losses from predation, but even allows toxic bacteria to increase in numbers. This suggests that toxic bacteria take benefit of predators, presumably by redirecting them towards neighbouring bacteria. The potential enemy is thus transformed into an ally and toxic strains profit three times. They avoid losses due to predation, the density of competitors is reduced and thereby, through excretion, release nutrients locked up in the cells of competitors are released, making them available for growth of toxin-producing bacteria.

Predators are driving agents of the community composition of rhizosphere bacteria (Rønn et al. 2002). High density of toxic bacteria therefore likely alters the impact of predators on the whole bacterial community. Supporting this assumption, total bacterial densities were less affected by predators in presence of the P. fluorescens Wt strain, and in turn, the density of amoebae was lower. This supports the above suggestion that via changing predator pressure toxin-producing bacteria indirectly affect the structure of the whole bacterial community. As indicated by reduced microbial respiration in presence of the gacS mutant, toxin-mediated attenuation of predator pressure not only affects microbial community structure and increases total bacterial density but also increases microbial community functioning.

The colonization of the P. fluorescens Wt and mutant strain increasingly diverged during the experiment, suggesting that the advantage of producing secondary metabolites increased with time. This indicates that biotic interactions are of minor importance during early phases of rhizosphere colonization, but that in mature communities of older roots biotic interactions predominate, thereby favouring toxin-producing strains. This is in agreement with the preferential colonization of root tips by toxin-deficient mutants of Pseudomonas (Achouak et al. 2004) a niche with reduced competition for resources and exposure to predators (Folman, Postma & Veen 2001).


The results of the present study challenge previous views on the role of secondary metabolite production by plant growth promoting rhizobacteria. Conform to our hypotheses, secondary metabolites of biocontrol rhizobacteria primarily improved bacterial fitness by manipulating biotic interactions with other rhizosphere organisms. Secondary metabolites improved the competitive strength against other rhizobacteria but in particular they attenuated losses due to protozoan predators. Thus, bacterial fitness and biocontrol properties are linked in a causal way providing an evolutionary explanation for the biocontrol activity of rhizosphere bacteria such as P. fluorescens. Since predators may promote the establishment of P. fluorescens strains, knowledge on soil protozoan predators and their interactions with bacterial preys may allow improving management strategies employing plant growth promoting rhizobacteria in arable systems. Biotic interactions between biocontrol bacteria and their predators have been traditionally neglected, the effects on soil protozoa being seen as mere side effect of inoculation (Winding, Binnerup & Pritchard 2004). In contrast to this view, results of the present study suggest that the functioning of biocontrol bacteria can only be understood considering biotic interactions in particular top-down forces. Potentially, the biocontrol activity of bacterial secondary metabolites exerted on plants itself results from side effects of the toxins due to the fact that they evolved for manipulating eukaryotic organisms (predators).


We thank Dr Claudio Valverde (Universidad Nacional de Quilmes, Argentina) for his helpful advices, Dr Xin Ke (Institute of Crop Breeding and Planting, Chinese Academy of Agricultural Sciences, China) for providing the rice seeds and Katja Rosenberg (TU Darmstadt, Germany) for providing axenic Acanthamoeba cultures. This work was partially funded by the fellowship program of the German Federal Foundation for the Environment (DBU).