Introduction

The natural selection has resulted in numerous examples of genetic resistance in plants to above-ground pathogens, but curiously enough not to many below-ground pathogens. Cook et al. (1995) postulated that plants have developed an entirely different strategy to defend themselves against soil-borne pathogens. This strategy involves the ability of plants to selectively stimulate and support populations of soil and rhizosphere microorganisms that are antagonistic to their pathogens. Natural disease-suppressive soils are the best examples in which the activities of specific soil and rhizosphere microorganisms keep susceptible plants almost free of infection in spite of ample exposure to virulent inoculum of soil-borne pathogens. Natural disease-suppressive soils occur worldwide and provide an enormous resource for the discovery of novel beneficial microorganisms and traits (Weller et al., 2002). For most of the disease-suppressive soils, however, the consortia of microorganisms and mechanisms involved in pathogen control have not yet been identified.

The disease-suppressive soils have been classified as ‘long-standing suppressive soils’ and ‘induced suppressive soils’ according to the longevity of disease suppression (Hornby, 1979; Weller et al., 2002). Long-standing suppressiveness is a biological condition naturally associated with the soil, its origin is not known and it appears to survive in the absence of plants. In contrast, induced suppressiveness is initiated and sustained by crop monoculture in the presence of the pathogen (Weller et al., 2002). Soils suppressive to take-all disease of wheat and barley, caused by the fungal pathogen Gaeumannomyces graminis var. tritici, are referred to as take-all decline soils and are well-known examples of induced suppressiveness. The Fusarium-wilt suppressive soils from Châteaurenard (France) and Salinas Valley (CA, USA) are among the best examples of long-standing suppressive soils.

Fusarium wilts are caused by plant pathogenic Fusarium oxysporum leading to significant yield losses in many crops worldwide. Fusarium-wilt suppressive soils substantially limit the disease incidence or severity, but are not suppressive to diseases caused by non-vascular Fusarium species or other soil-borne pathogens (Alabouvette, 1986). The microbiological nature of the disease suppression in Fusarium-wilt suppressive soils was unequivocally shown by a range of approaches, including treatments with moist heat, methylbromide or γ-irradiation leading to loss of soil suppressiveness, and by experiments in which suppressiveness was transferred to wilt-conducive soils by mixing small amounts (0.1–10% wt/wt) of the suppressive soil into the conducive soil (Scher and Baker, 1980; Alabouvette, 1986; Weller et al., 2002). Fusarium-wilt suppressive soils maintain their activity when brought into the greenhouse or laboratory, which greatly facilitates the identification of the microorganisms, mechanisms and genes involved in disease suppressiveness.

Among the bacterial and fungal genera proposed to contribute to Fusarium wilt suppressiveness are Alcaligenes (Yuen and Schroth, 1986), Trichoderma (Sivan and Chet, 1989), Actinomycetes (Amir and Amir, 1989), Pseudomonas (Kloepper et al., 1980; Scher and Baker, 1982; Lemanceau and Alabouvette, 1993) and non-pathogenic F. oxysporum (Alabouvette, 1986; Larkin et al., 1996). Detailed studies on Fusarium-wilt suppressive soils from Châteaurenard (France) and Salinas Valley showed that the suppressiveness was attributed, in particular, to the activity of non-pathogenic F. oxysporum and fluorescent pseudomonads (Scher and Baker, 1982; Alabouvette, 1990; Lemanceau and Alabouvette, 1991; Duijff et al., 1999). For the non-pathogenic F. oxysporum, competition for carbon and induced systemic resistance were the main modes of action proposed so far (Steinberg et al., 2007). For the fluorescent pseudomonads, siderophore-mediated competition for iron was shown to be a major mechanism in the control of Fusarium wilts (Scher and Baker, 1982; Lemanceau et al., 1992). Particularly interesting from these studies was that the association between non-pathogenic F. oxysporum and fluorescent pseudomonads provided enhanced disease suppression (Lemanceau et al., 1993). Work by Duijff et al. (1999) supported and extended the earlier observations that disease suppression by the non-pathogenic F. oxysporum is related to reductions in both population density and metabolic activity of the pathogen on the root surface, and that competition for iron contributes to the suppression by Pseudomonas and enhances the biological activity of the non-pathogenic F. oxysporum.

In contrast to other disease-suppressive soils, the role of antibiosis in Fusarium-wilt suppressive soils has not been studied to date, in spite of the fact that several antibiotic compounds produced by antagonistic bacteria have shown growth-inhibitory activity against pathogenic F. oxysporum. For example, phenazine antibiotics were shown to play a key role in the biocontrol activity of several Pseudomonas species against F. oxysporum on diverse crops (Anjaiah et al., 1998; Chin-A-Woeng et al., 1998, 2000). Also the antibiotic 2,4-diacetylphloroglucinol (DAPG) has substantial activity against pathogenic F. oxysporum (Schouten et al., 2004), and populations of DAPG-producing pseudomonads were highly enriched in a soil naturally suppressive to Fusarium wilt of peas (Landa et al., 2002). Following similar molecular-based approaches used to unravel the role of antibiosis in take-all decline soils (Raaijmakers et al., 1997; Raaijmakers and Weller, 1998; McSpadden-Gardener et al., 2000; Weller et al., 2002; Souza et al., 2003a) and in soils suppressive to black root rot of tobacco (Wang et al., 2001; Ramette et al., 2003, 2006; Frapolli et al., 2007), the objectives of this study were to determine the frequency, diversity and biocontrol activity of DAPG (phlD+)- and phenazine (phzC+)-producing pseudomonads in the Fusarium-wilt suppressive soil from Châteaurenard (France) in comparison with the conducive soil of Carquefou (France). The results showed that densities of phlD+ populations were not significantly different between the two soils but that their genotypic diversity differed significantly. However, the suppressive ability of isolates representative of the phlD+ diversity found in the two soils was similar. In contrast, phzC+ populations were only detected in the Fusarium-wilt suppressive soil and isolates representative of their diversity significantly improved the biocontrol activity achieved by non-pathogenic F. oxysporum. These results indicate, for the first time, that phenazine-producing pseudomonads are enriched in the Fusarium-wilt suppressive soil from Châteaurenard and, when reintroduced in association with non-pathogenic F. oxysporum, significantly enhance the soil suppressiveness to Fusarium wilt. Bioassays with a phenazine-deficient mutant showed that phenazine production is involved in this augmentation of disease suppression. Collectively, these results suggest a role of redox-active phenazine antibiotics in the natural soil suppressiveness to Fusarium wilts. A model is presented depicting the versatility of the mechanisms and the complex interplay between consortia of antagonistic microorganisms in Fusarium-wilt suppressive soil.

Materials and methods

Microorganisms and soils used

Characteristics of the Pseudomonas reference strains are given in Table 1. All Pseudomonas strains were cultivated on KMB agar (King et al., 1954) at 25 °C. The flax pathogen Fusarium oxysporum f. sp. lini Foln3 and the non-pathogenic strain F. oxysporum Fo47 were obtained from the MIAE Culture Collection (INRA, Dijon, France), and grown on Difco potato dextrose agar (PDA; Difco Laboratories, Detroit, MI, USA) at 24 °C. Soil naturally suppressive to Fusarium wilt was obtained from Châteaurenard, and the conducive soil from Carquefou. The main characteristics of these soils have been described earlier (Lemanceau et al., 1988). Briefly, the soils from Châteaurenard and Carquefou are calcic silt-clay and sandy-loam soils, respectively. The calcium carbonate content and pH of the Châteaurenard soil (37.4% CaCO3, pH 7.9) are much higher than those of the Carquefou soil (8.4% CaCO3, pH 5.3). These two characteristics account for the very low concentration of extractable iron of the Châteaurenard soil (18.8 p.p.m.), 15 times lower than that of the Carquefou soil (279.9 p.p.m.), even though both soils share a comparable concentration of total iron, that is, 11% and 9% for the Châteaurenard and Carquefou soils, respectively. Biocontrol assays were performed in a soil collected from a fallow field in Dijon (France) for which the properties were described in Latour et al. (1996). The phenazine-producing strain P. fluorescens 2-79 and its phenazine-deficient mutant 2-79Z were kindly provided by D Mavrodi and L Thomashow (Washington state University, Pullman, WA, USA) (Khan et al., 2005). DAPG-deficient mutant 17-8D was obtained by plasposon mutagenesis of strain Phl60rif, and its inability to produce DAPG was confirmed by RP-HPLC analysis (data not shown). The deficiency in phenazine production by mutant 2-79Z was previously demonstrated by Khan et al. (2005).

Table 1 Reference strains of 2,4-diacetylphloroglucinol (DAPG) and phenazine-producing Pseudomonas spp. used in this study
Full size table

Frequency of indigenous phlD+ and phzC+ pseudomonads in soil and rhizosphere

Fluorescent pseudomonads harboring the phlD (phlD+) or phzC (phzC+) genes were isolated and enumerated in bulk soil and in the rhizosphere of flax and tomato according to the methods described by Raaijmakers et al. (1997) and Souza et al. (2003a). Briefly, surface-sterilized seeds of flax (Linum usitatissimum L. cv. Opaline) and tomato (Lycopersicon esculentum Mill. cv. H63-5) were sown in the wilt suppressive and conducive soils, and the plants were grown for 4 weeks under controlled conditions (12:12 h light/dark photoperiod under photosynthetic active radiation of 215 μE m−2 s−1, 25:22 °C light/dark thermoperiod). For each of the two soils, five pots (3 l) were used for flax and five for tomato; each pot contained 2 l of soil and five seeds. One series of five pots was kept uncultivated (bulk soil). Rhizosphere and bulk soil samples were obtained as described earlier (Raaijmakers et al., 1997) and dilution plated onto KMB agar supplemented with cycloheximide (100 mg l−1), chloroamphenicol (13 mg l−1) and ampicillin (40 mg l−1) (Simon and Ridge, 1974). After 48 h of incubation at 25 °C, colonies were counted and fluorescent pseudomonads were differentiated from non-fluorescent colonies under UV light (wavelength, 365 nm). The numbers of fluorescent pseudomonads that harbored the phlD gene or the phzC gene were determined by colony hybridization followed by PCR analysis as described earlier (Raaijmakers et al., 1997; Souza et al., 2003a). The phlD and phzC probes were obtained by labeling of a 629-bp phlD fragment obtained from the reference strain P. fluorescens Phl60rif, and of a 522-bp phzC fragment from the reference strain P. fluorescens 2-79. Digoxigenin labeling of the DNA probes was carried out by random priming using the DIG DNA labeling kit (Roche, Meylan, France). Primers B2BF and BPR4 (McSpadden-Gardener et al., 2001) were used to amplify the 629-bp phlD fragment and primers PHZJR1 5′-CAGGGCCG(G/C)(A/G)(C/T)ATTTCTCGGTTCT-3′ and PHZJR2 5′-GCGCGGGTCGCACAGGCTTTTGTA-3′ for the 522-bp phzC fragment. The hybridized probes were immunodetected with anti-digoxigenin-alkaline phosphatase-Fab fragments and visualized with the colorometric substrates nitroblue tetrazolium salt and 5-bromo-4-chloro-3-indolylphosphate, as described by the supplier (Roche). The colonies that were positive in hybridization were purified and subjected to PCR analysis with primers specific for phlD or phzC. For statistical analysis, population densities were log10 transformed and significant differences were assessed by analysis of variance followed by Student–Newman–Keuls tests (SAS Institute Inc., Cary, NC, USA).

Genetic diversity of phlD+ and phzC+ pseudomonads

Restriction-fragment length polymorphism (RFLP) analysis of the 629-bp phlD PCR product amplified from the indigenous isolates and reference strains was performed using the restriction enzymes HaeIII, MspI and TaqI (McSpadden-Gardener et al., 2001). For RFLP analysis of the 522-bp phzC fragments, the restriction enzymes Fnu4HI, NciI, NlaIV, NdeII, DdeI and Cfr13I were used. Restriction patterns were determined by electrophoresis at 80 V in TAE buffer with 3.5% (w/v) Metaphor agarose (Tebu, Le Perray-en-Yvelines, France). A similarity matrix was calculated using the pair wise Jaccard similarity coefficient (Jaccard, 1908). Cluster analysis was performed by the unweighted pair group method using arithmetic averages allowing the delineation of phlD and phzC genotypes. As phlD genotypes were earlier shown to correlate almost perfectly with those defined by BOX-PCR (Mavrodi et al., 2001; McSpadden-Gardener et al., 2001; Landa et al., 2002, 2006, Weller et al., 2002; De La Fuente et al., 2006), BOX-PCR was only performed for phzC+ isolates. The protocol used was described by Rademaker et al. (1997), except for the Taq Polymerase. Four units of Taq DNA polymerase (Q-BIOgene, Illkirsch, France) were used in each 25 μl BOX-PCR. BOX-PCR fragments were separated overnight at 40 V in a 1.5% (w/v) Seakem LE agarose gel (Tebu) in TAE buffer. For each BOX-PCR fingerprint, a binary data matrix with 79 bands being either present (1) or absent (0) was established. A similarity matrix was calculated using the pair wise Pearson product-moment correlation coefficient. Cluster analysis was performed by unweighted pair group method using arithmetic averages. Isolates presenting Pearson product-moment correlation coefficients ⩾0.75 were assigned to a BOX-PCR genotype named with a same letter (A–J) and, within these genotypes, those having a correlation coefficient equal to 1 were assigned a BOX-PCR genotype named with the same number (that is, A1). Next to RFLP and BOX-PCR analyses, phlD and phzC PCR fragments were cloned (pGEMT Easy Vector System II, Promega, Charbonnières, France) and sequenced (Genome Express, Meylan, France). BLAST analyses and CLUSTAL W (Thompson et al., 1994) were applied for multiple sequence alignments. Phylogenies were determined by neighbor-joining method (Saitou and Nei, 1987) using the Kimura ‘2-parameters’ correction (Kimura, 1980) with the pair wise gap removal option. To estimate tree node validity, results of 1000 bootstrapped data sets were determined.

Biocontrol efficacy of phlD+ and phzC+ pseudomonads

The phlD+ and phzC+ Pseudomonas isolates were grown on KMB agar plates for 48 h at 25 °C and cells were harvested and washed once in sterile distilled water. The cell density was determined spectrophotometrically at 600 nm and adjusted to 107 CFU g−1 of fresh soil. Inoculums of the pathogenic F. oxysporum f. sp. lini Foln3 and non-pathogenic F. oxysporum Fo47 were prepared as described earlier (Lemanceau and Alabouvette, 1991). The fungal suspensions were adjusted to 105 conidia g−1 of soil fresh weight for F. oxysporum Fo47, and 103 conidia g−1 of soil fresh weight for F. oxysporum f. sp. lini Foln3. Flax plants were grown in environment-controlled growth chamber (15:9 h light/dark photoperiod under photosynthetic active radiation of 215 μE m−2 s−1, 26:20 °C light/dark thermoperiod). The relative humidity in the growth chamber was maintained at 60% during the night and 80% during the day. Each flax plant was grown individually in a plastic tray cell filled with 50 ml of autoclaved (three successive treatments at 100 °C for 1 h, each one separated by 24 h) Dijon soil placed on top of a draining layer of extruded clay beads. Surface sterilized flax seeds were sown in the filled plastic tray cells and inoculated with conidial and bacterial suspensions. Each treatment consisted of three replicate trays of 16 plants arranged in a randomized complete block experimental design. Plants showing the typical symptoms of Fusarium wilt were recorded twice a week for a period of 6 weeks, and cumulated numbers of diseased plants were used to determine the area under the disease progress curve (AUDPC) according to the methods described by Steinberg et al. (2004). Differences in disease severity between treatments were determined by analysis of variance followed by Fisher's PLSD-test (SAS Institute Inc.).

Results

Frequency of antibiotic-producing pseudomonads in disease suppressive and conducive soils

Population densities of fluorescent pseudomonads harboring the biosynthesis gene phlD for DAPG or phzC for phenazine production were assessed in bulk soils and in the rhizosphere of tomato and flax seedlings grown in the Fusarium wilt suppressive (Châteaurenard) and conducive (Carquefou) soils. phlD+ pseudomonads were not detected in the two bulk soils (detection limit was 3.3 × 101 CFU g−1 soil fresh weight). In contrast, phzC+ pseudomonads were detected in the wilt-suppressive bulk soil at an average density of 3.9 × 103 CFU g−1 soil fresh weight, representing 1.3% of the total population of fluorescent pseudomonads. phzC+ pseudomonads were not detectable in the conducive soil (data not shown). When flax or tomato seedlings were cultivated in these soils, phlD+ pseudomonads were detected in the rhizosphere of both crops at densities ranging from 4.1 × 103 to 4.7 × 105 CFU g−1 root fresh weight (Table 2), representing on an average 3.7–26.1% of the total population of culturable fluorescent pseudomonads. Owing to variations among replicates, no significant differences in the population densities of phlD+ pseudomonads were found between the wilt suppressive and conducive soils, either in flax or tomato rhizospheres. phzC+ pseudomonads were detected in the rhizosphere of flax and tomato grown in the wilt-suppressive soil at average densities of 1.1 × 103 CFU g−1 and 1.3 × 104 CFU g−1 root, respectively, representing on an average 0.6–13.9% of the total rhizosphere population of culturable pseudomonads. In the rhizosphere of flax and tomato plants grown in the wilt-conducive soil, phzC+ pseudomonads were not detected (Table 2).

Table 2 Densities of total, phlD+ and phzC+ populations of culturable fluorescent pseudomonads in the rhizosphere of flax and tomato plants grown in the Fusarium-wilt suppressive soil (Châteaurenard, France) or in the wilt-conducive soil (Carquefou, France)
Full size table

Diversity of phlD+ pseudomonads

For a total of 95 phlD+ isolates from the suppressive and conducive soils, and 13 phlD+ reference strains, phlD-RFLP analysis was performed to assess the genotypic diversity (Figure 1). The 95 indigenous isolates comprised 24 from Châteaurenard tomato rhizosphere (designated ChPhlTR), 18 from Châteaurenard flax rhizosphere (designated ChPhlLR), 27 from Carquefou tomato rhizosphere (designated CaPhlTR) and 26 from Carquefou flax rhizosphere (designated CaPhlLR). Among the indigenous phlD+ isolates, seven distinct phlD-RFLP genotypes could be distinguished: (i) three phlD-RFLP genotypes corresponded to the known genotypes F, M and O (De La Fuente et al., 2006), and (ii) four genotypes, designated α, β, γ and δ, were not described earlier (Figure 1). The 42 phlD+ isolates obtained from the rhizosphere of tomato and flax grown in the Châteaurenard wilt suppressive soil were classified as the M-genotype (34 isolates) or the F-genotype (8 isolates). Most of the 53 phlD+ isolates obtained from the rhizosphere of tomato and flax grown in the Carquefou conducive soil were different from the Châteaurenard isolates and belonged to six phlD-RFLP genotypes, including F (6 isolates), O (4 isolates), α (1 isolate), β (6 isolates), γ (26 isolates) and δ (10 isolates) (Figure 1). Subsequent sequencing of the 629-bp phlD gene fragment of representative isolates of each of the phlD-RFLP groups followed by phylogenetic analysis strongly supported the classification based on phlD-RFLP analysis and confirmed that the phlD+ isolates belonging to the new RFLP groups α, β, γ and δ were indeed different from the other isolates and from the phlD+ reference strains described so far (Supplementary Figure S1). RP-HPLC analysis confirmed that each of the isolates, for which the phlD gene was sequenced, produced DAPG in vitro (data not shown). These results indicate that, in spite of similar rhizosphere population densities in the suppressive and conducive soils (Table 2), most of the phlD+ isolates from the wilt-suppressive soil were genotypically different from those isolated from the conducive soil (Figure 1).

Figure 1
figure 1

Diversity and phylogeny of phlD+ fluorescent Pseudomonas isolates obtained from the Fusarium-wilt suppressive soil of Châteaurenard (indicated in light gray) and the conducive soil of Carquefou (indicated in dark gray). The pair wise coefficients of similarity were clustered with the unweighted pair group method using arithmetic averages algorithm of NTSYS-pc 2.02. PhlD+ reference strains are presented in italics. The phlD genotype designations A, B, D, F, K, M and O are based on work of De La Fuente et al. (2006), whereas the novel phlD genotypes α, β, γ and δ were identified in this study. Isolates tested in this study for their efficacy to suppress Fusarium wilt of flax are indicated with an asterisk.

Full size image

Diversity of phzC+ pseudomonads

Twenty-nine phzC+ isolates were selected from the Châteaurenard suppressive bulk soil and from the rhizosphere of flax and tomato seedlings grown in this soil, ten from bulk soil (ChPhzS), three from flax rhizosphere (ChPhzLR) and 16 from tomato rhizosphere (ChPhzTR). Diversity was assessed by phzC-RFLP analysis, and by BOX-PCR and sequence analyses. Among the indigenous phzC+ isolates, five distinct phzC-RFLP genotypes (I, II, III, IV and VII) could be distinguished, two of which clustered with the phenazine-producing reference strains P. chlororaphis PCL1391 (genotype II) and P. fluorescens 2-79 (genotype VII) (Figure 2). BOX-PCR analysis revealed 11 distinct genotypes, designated A1, A2, B1, B2, C1, F1–F3, F5, G1 and H1, providing an additional level of discrimination between the isolates within each phzC-RFLP group (Figure 2). For example, the three phzC+ isolates belonging to RFLP group VII could be discriminated in three BOX-PCR genotypes (G1, H1 and I1). Sequence and phylogenetic analyses of the phzC genes from 10 indigenous isolates, five reference strains, and 12 phzC sequences available in genomic databases showed that seven indigenous isolates clustered with the reference strains PGS12 and 30–84 (Supplementary Figure S2). Consistent with the classification based on phzC-RFLP and BOX-PCR analyses, isolates ChPhzTR44 and ChPhzS26 were most similar to reference strain PCL1391, and isolate ChPhzLR107 to reference strain 2-79 (Supplementary Figure S2). RP-HPLC analysis confirmed that each of the isolates for which the phzC gene was sequenced, produced phenazine antibiotics in vitro (data not shown).

Figure 2
figure 2

Diversity and phylogeny of phzC+ fluorescent Pseudomonas isolates obtained from the Fusarium-wilt suppressive soil of Châteaurenard (highlighted in light gray). The pair wise coefficients of similarity were clustered with the unweighted pair group method using arithmetic averages algorithm of NTSYS-pc 2.02. PhzC reference strains are indicated in italics. For the phzC+ Pseudomonas isolates with RFLP-genotypes I to VIII, the corresponding BOX-PCR genotypes (A1 to J1) are indicated in bold in front of the strains name. Isolates tested in this study for their efficacy to suppress Fusarium wilt of flax are indicated with an asterisk.

Full size image

Suppression of Fusarium wilt of flax by phzC+ and phlD+ pseudomonads

Earlier studies on the Châteaurenard Fusarium-wilt suppressive soil showed that disease suppressiveness was ascribed to the joint activity of non-pathogenic F. oxysporum and fluorescent pseudomonads (Lemanceau et al., 2006). Therefore, the efficacy of phzC+ and phlD+ isolates to control Fusarium wilt of flax was tested when inoculated separately (Figures 3a and c) and in combination with the non-pathogenic F. oxysporum strain Fo47 (Figures 3b and d). The bacterial isolates (indicated with an asterisk in Figures 1 and 2) were chosen as representatives of the diversity of the indigenous populations: (i) an isolate of each of the five phzC-RFLP genotypes from the Châteaurenard suppressive soil (genotypes I–IV) plus strain 2-79 (genotype VII), and (ii) two phlD+ isolates of the phlD-RFLP dominant genotype M from the Châteaurenard suppressive soil, one isolate of each of the dominant phlD-RFLP genotypes γ and δ from the conducive soil, and reference strain Phl60rif (genotype D).

Figure 3
figure 3

Protection of flax against Fusarium wilt by phzC+ Pseudomonas isolates obtained from the wilt-suppressive soil (a and b) and by phlD+ Pseudomonas isolates obtained from the wilt-suppressive or conducive soil (c and d). For each of the Pseudomonas isolates tested, the corresponding genotypes as defined in Figures 1 and 2 are indicated between brackets. Plants were grown in soil infested with pathogenic F. oxysporum f.sp. lini Foln3 (103 conidia g−1 soil fresh weight), and inoculated with cell suspensions of the Pseudomonas isolates (107 CFU g−1 soil fresh weight) and non-pathogenic F. oxysporum Fo47 (105 conidia g−1 soil fresh weight) separately or in combination. The area under disease progress curve represents the cumulative disease severity monitored over a period of 6 weeks of plant growth. Error bars represent the s.d. of the mean. Fisher's PLSD-test was used for pair wise comparison of the treatment with the infested control (Foln3): *P⩽0.1; **P⩽0.05; ***P⩽0.01; ****P⩽0.001.

Full size image

The results show that none of the five phzC+ strains was effective in disease control when applied alone (Figure 3a). However, when applied in combination with non-pathogenic F. oxysporum strain Fo47, they all improved the suppression achieved by Fo47 (Figure 3b). Regardless of their origins, four of the five phlD+ strains significantly reduced Fusarium wilt of flax when applied alone (Figure 3c) and, when co-inoculated with Fo47, all enhanced the level of disease control (Figure 3d). To further resolve the role of DAPG and phenazine in disease control and in the synergy with non-pathogenic F. oxysporum strain Fo47, biosynthesis mutants deficient in DAPG (mutant 17-8D) or phenazine (mutant 2-79Z) production were tested and their biocontrol efficacy compared with their respective wild-type strains. The level of protection achieved by the two wild-type strains did not differ from that of their respective antibiotic-deficient mutants (Figures 4a and c). In the presence of Fo47, however, differential effects were observed between the DAPG and phenazine producers. DAPG-producing strain Phl60rif and its DAPG-deficient mutant 17-8D both enhanced as efficiently the disease suppression by Fo47 (Figure 4d). In contrast, phenazine-producing strain 2-79 was more effective in combination with Fo47 than its phenazine-deficient mutant 2-79Z (Figure 4b). Collectively, these results suggest that phenazines, but not DAPG, contribute to the enhanced disease suppression achieved in combination with non-pathogenic F. oxysporum.

Figure 4
figure 4

Protection of flax against Fusarium wilt by phenazine-producing strain P. fluorescens 2-79 and by 2,4-diacetylphloroglucinol (DAPG)-producing strain P. fluorescens Phl60rif. Each of the strains was tested alone or in combination with non-pathogenic F. oxysporum Fo47. For strain 2-79, the biocontrol efficacy was compared with that of its phenazine-deficient mutant 2-79Z (a and b); for strain Phl60rif, a comparison was made with its DAPG-deficient mutant 17-8D (c and d). Plants were grown in soil infested with pathogenic F. oxysporum f.sp. lini Foln3 (103 conidia g−1 soil fresh weight), and inoculated with cell suspensions of the Pseudomonas isolates (107 CFU g−1 soil fresh weight) and non-pathogenic F. oxysporum Fo47 (105 conidia g−1 soil fresh weight) separately or in combination. The area under disease progress curve represents the cumulative disease severity monitored over a period of 6 weeks of plant growth. Error bars represent the s.d. of the mean. Fisher's PLSD-test was used for pair wise comparison of the treatment with the infested control (Foln3): *P⩽0.1; ****P⩽0.001. The results of the statistical pair wise comparison between the effects of the wild-type strains and their antibiotic-deficient mutant are indicated with probability values.

Full size image

Discussion

To unravel the role of antibiosis in the natural soil suppressiveness to Fusarium wilt, this study used strategies that were earlier applied for take-all decline soils (Raaijmakers et al., 1997; Raaijmakers and Weller, 1998; McSpadden-Gardener et al., 2000; Weller et al., 2002; Souza et al., 2003a) and for soils suppressive to black root rot of tobacco (Wang et al., 2001; Ramette et al., 2003, 2006; Frapolli et al., 2007). This strategy encompasses a comparative analysis of the densities, diversity and activity of antibiotic-producing pseudomonads in two soils chosen for their opposite level of suppressiveness to Fusarium wilt: one being suppressive (Châteaurenard) and the other being conducive (Carquefou). These two soils were the same as those used earlier to show the role of carbon and iron competition in soil suppressiveness to Fusarium wilts (Lemanceau et al., 1988). At this stage, the two structurally different antibiotics DAPG and phenazines were considered because of their established role in the activity of various Pseudomonas strains against pathogenic F. oxysporum (Georgakopoulos et al., 1994; Anjaiah et al., 1998; Chin-A-Woeng et al., 1998; Schouten et al., 2004) and, for DAPG-producing pseudomonads, because of their prevalence in a soil naturally suppressive to Fusarium wilt of peas (Landa et al., 2002).

Molecular-based detection revealed that DAPG- and phenazine-producing pseudomonads were present in the rhizosphere of flax and tomato plants grown in the wilt-suppressive soil. DAPG-producing pseudomonads were found at similar densities in the conducive soil, whereas phenazine-producing pseudomonads were only detected in the suppressive soil. To our knowledge, this is the first report establishing differences in population densities of phenazine-producing pseudomonads between disease suppressive and conducive soils. Furthermore, they also show the coexistence of DAPG- and phenazine-producing Pseudomonas in a naturally suppressive soil. Although phenazine-producing strain P. fluorescens 2-79 was originally isolated from a take-all suppressive soil and was shown to efficiently suppress take-all disease through the production of phenazine-1-carboxylic acid (Thomashow and Weller, 1988), the role of phenazine-producing Pseudomonas populations in the natural soil suppressiveness to take-all and other soil-borne diseases could not be confirmed (Raaijmakers et al., 1997; Mavrodi et al., 2006). The higher density of phenazine-producing Pseudomonas in the suppressive soil of Châteaurenard compared with the conducive soil of Carquefou may be related to differences in the physico-chemical properties of these two soils. Indeed, owing to the high pH (7.9) and CaCO3 content of the Châteaurenard soil, the concentration of extractable iron was shown to be 15 times lower than that in the Carquefou soil (Lemanceau et al., 1988). As phenazines are redox-active antibiotics (Dietrich et al., 2008) and may contribute to iron mobilization in soils (Hernandez et al., 2004; Price-Whelan et al., 2006; Wang and Newman, 2008), the ability of bacteria to produce these metabolites may give them a competitive advantage under the iron-limiting conditions prevailing in the Châteaurenard suppressive soil.

At the genotypic level, the phzC+ isolates were distributed in five groups, four of them being closely related to reference strains known to produce different phenazine derivatives, including phenazine-1-carboxylic acid, 2-OH-phenazine-1-carboxylic acid and phenazine-1-carboxamide (reviewed by Mavrodi et al., 2006). The phlD genotypes found in the two soils differed substantially, with genotype M (81% of the isolates) being dominant in the suppressive soil. Interestingly, the two genotypes M and F identified in the suppressive soil of Châteaurenard were described earlier as being dominant in European take-all decline soils (Souza et al., 2003a). Phylogenetic analysis further revealed that the four phlD genotypes found in the conducive soil, designated α, β, γ and δ, were not described so far (De La Fuente et al., 2006). Collectively, these results correspond well with the results obtained by Ramette et al. (2006) for the soils suppressive or conducive to black root rot of tobacco; in their work, DAPG-producing Pseudomonas spp. were also detected at similar densities in the conducive and suppressive soils and were genotypically diverse. However, in their study the different genotypes also differed in their biocontrol efficacy, whereas in this study the phlD+ isolates from the wilt-suppressive soil were equally effective as the isolates from the conducive soil. None of the genotypically different phzC+ isolates from the Fusarium-wilt suppressive soil were able to control Fusarium wilt of flax when inoculated separately; this includes isolate ChPhzS26 which belongs to the same phzC-RFLP genotype (II) as P. chlororaphis PCL1391, a strain that effectively suppressed crown and root rot of tomato caused by F. oxysporum f. sp. radicis lycopersici (Chin-A-Woeng et al., 1998).

In contrast with suppressiveness to take-all, which has been ascribed mainly to fluorescent pseudomonads (Weller et al., 2002), soil suppressiveness to Fusarium wilt has been related to both non-pathogenic F. oxysporum (Rouxel et al., 1979) and fluorescent pseudomonads (Scher and Baker, 1980). Densities of indigenous populations of non-pathogenic F. oxysporum were shown earlier to be much higher in Fusarium wilt suppressive than in conducive soils (Alabouvette, 1986). Additional suppression by the combination of non-pathogenic F. oxysporum strain Fo47 and specific isolates of fluorescent pseudomonads was reported earlier (Lemanceau and Alabouvette, 1991; Lemanceau et al., 1992, Duijff et al., 1999) and proposed to account for the efficacy and stability of the natural suppressiveness (Lemanceau et al., 2006). Data shown here confirm that most of the isolates tested improved the protection achieved by Fo47. On the basis of the experiments with mutants affected in antibiotic biosynthesis, DAPG biosynthesis does not seem to be involved in this increased protection. In contrast, this additional level of disease suppression was shown to be related to phenazine biosynthesis. Interestingly, the additional suppression provided by the bacterial and fungal combination was ascribed earlier to an interplay between carbon and iron competition achieved by these two groups of antagonistic microorganisms (Lemanceau et al., 1988, 1993): pyoverdine-mediated iron competition achieved by the fluorescent pseudomonads was shown to reduce the efficacy of carbon metabolism of the pathogenic F. oxysporum making it more susceptible to carbon competition with non-pathogenic F. oxysporum (Lemanceau et al., 1993). Considering that phenazines are redox-active antibiotics (Dietrich et al., 2008) and may contribute to iron mobilization in soils (Hernandez et al., 2004; Price-Whelan et al., 2006), they also could play a role in iron competition, thereby making pathogenic F. oxysporum more susceptible to carbon competition with non-pathogenic F. oxysporum. Alternatively, phenazines also could act against pathogenic F. oxysporum as redox-active antibiotics leading to the accumulation of toxic oxygen radicals (Mavrodi et al., 2006). Whether one or both modes of action of phenazines operate in Fusarium wilt suppressiveness remains to be explored.

One may consider that the colonization level of the indigenous phzC+ pseudomonads in the suppressive soil was lower than that of introduced phenazine-producing strains in the biocontrol experiments, and therefore may not be high enough for effective disease suppression. This complex issue highly depends on the spatial distribution of the bacteria and the level of expression of the antagonistic traits involved. Various studies performed so far show that the relationships between bacterial densities on plant roots and the level of disease control are nonlinear and have an asymptotic nature (Bull et al., 1991; Johnson, 1994; Raaijmakers et al., 1995; Raaijmakers and Weller, 1998). These studies further show that when indigenous or introduced Pseudomonas populations reach a density in the rhizosphere of 103–105 CFU g−1 of root fresh weight, several fungal diseases, including Fusarium wilt, are controlled to some degree. At densities above 105 CFU g−1 of root, there is no further improvement in the level of disease control (Raaijmakers et al., 1995; Raaijmakers and Weller, 1998), which is probably related to the spatial heterogeneity of the indigenous/introduced bacteria and the level of colocalization of the bacteria and the pathogen. These quantitative dose–response relationships were shown for several Pseudomonas strains and mechanisms, including phenazine-mediated antibiosis by strain 2-79 (Bull et al., 1991), DAPG-mediated antibiosis (Raaijmakers and Weller, 1998), siderophore-mediated competition for iron and induced systemic resistance (Raaijmakers et al., 1995). Therefore, the population densities of the indigenous phzC+ pseudomonads reported in our study do reach population densities within the range, where at least some level of disease control can be expected. This is certainly the case for phzC+ Pseudomonas populations in the tomato rhizosphere but to a lesser extent for flax. However, when acting in synergy with other antagonistic microorganisms and mechanisms, these relatively low population densities may contribute to disease control (Olivain et al., 2004).

The observation that DAPG production did not play a significant role in the enhanced disease suppression achieved by strain Phl60rif in combination with Fo47 may have been due to various factors. One important factor in this respect is the choice of the phlD genotype tested, which may substantially affect the level of disease control (Raaijmakers and Weller, 2001; Ramette et al., 2006). Therefore, more phlD genotypes and their respective DAPG-deficient mutants need to be tested to more conclusively assess the contribution of the antibiotic DAPG in the natural soil suppressiveness to Fusarium wilt disease. Furthermore, DAPG biosynthesis may also have been repressed by specific environmental conditions (Duffy and Défago, 1999) or by fusaric acid (Figure 5; Notz et al., 2002; Duffy et al., 2004), a phytotoxin produced by pathogenic F. oxysporum and non-pathogenic F. oxysporum strain Fo47 (Schouten et al., 2004). However, given that also phenazine production can be repressed by fusaric acid (Van Rij et al., 2004, 2005), it seems unlikely that this phytotoxin played a major role in the multitrophic interactions occurring in the biocontrol assays.

Figure 5
figure 5

Schematic model presenting the proposed mechanisms that contribute to the natural soil suppressiveness to Fusarium wilt. The significant higher microbial biomass in the suppressive soil as compared with the conducive soil contributes to a higher level of carbon competition in the suppressive soil. On this background of general competition, the higher density of non-pathogenic F. oxysporum in the suppressive soil further increases the carbon competition. The suppressive soil also differs from the conducive soil by its lower concentration of extractable iron, due to its high pH and CaCO3 content, making pyoverdine-mediated iron competition between the pathogen and the fluorescent pseudomonads stronger in the suppressive than in the conducive soil. Carbon and iron competition act in synergy to suppress the saprophytic growth of pathogenic F. oxysporum, leading to a reduced activity and rate of root infection. The results of this study suggest that also redox-active phenazine antibiotics may play a role in Fusarium wilt suppressiveness and may act in synergy with the carbon competition achieved by non-pathogenic F. oxysporum. In return, fusaric acid produced by F. oxysporum may affect the activity of the fluorescent pseudomonads by interference with the expression of antibiotic biosynthesis genes.

Full size image

In conclusion, the data presented in this study provide, for the first time, evidence that phenazine-producing pseudomonads are enriched in the Fusarium-wilt suppressive soil of Châteaurenard, and when combined with non-pathogenic F. oxysporum, significantly enhance disease suppressiveness either directly through antibiosis and/or indirectly through iron competition (Figure 5). The suppressive soil of Châteaurenard shows a high pH and CaCO3 content, and hosts an abundant microflora. These microbial characteristics lead to strong competition for carbon, whereas the physico-chemical characteristics of the Châteaurenard soil contribute to a strong competition for the poorly available iron. Non-pathogenic F. oxysporum, which are much more abundant in the suppressive than in the conducive soil, successfully compete for carbon sources with pathogenic F. oxysporum with which they share similar trophic requirements. At the same time, fluorescent pseudomonads decrease the iron availability for pathogenic F. oxysporum, which are more susceptible to iron starvation than non-pathogenic F. oxysporum (Figure 5). We postulate that redox-active phenazine antibiotics play a role in the control of pathogenic F. oxysporum by indigenous fluorescent pseudomonads, and when acting in synergy with carbon and iron competition contribute to the longstanding nature of the soil suppressiveness to Fusarium wilt disease.