Scientific Publications - Work Done by Microbiology Reader Bioscreen C

Environmental Microbiology (2000) 2(3), 274-284

Identification of conserved  traits in fluorescent  pseudomonads with antifungal activity

Richard J. Ellis,^t Tracey M. Timms-Wilson and Mark J. Bailey*

Molecular Microbial Ecology, NERC Institute of Virology and Environmental Microbiology, Mansfield Road, Oxford OX1 3SR, UK.

Summary

A collection of 29 fluorescent pseudomonads, some with known biological control activity against a range of phytopathogenic fungi, were characterized phenotypically and genotypically by comparing carbon source utilization patterns, suppression of Pythium ultimum both in plants and in vitro and the potential to produce known secondary metabolites. Fatty acid profiling and restriction fragment length polymorph­ism (RFLP) analysis of the ribosomal DNA operon (ribotyping) were used to determine the diversity of isolates. A small group of genetically related Pseudo­monas spp. with similar properties was identified; each isolate produced a diffusible bioactive product in vitro and was active against Pythium ultimum in planta. However, other isolates that were able to sup­press damping off disease but did not inhibit hyphal extension in vitro clustered outside this group. Pheno­typic analyses revealed that the accumulation of C17:0 cyclopropane fatty acid (17CFA) and the pro­duction of hydrogen cyanide correlated significantly with biological control activity and with the antagon­ism of fungal development. The potential of 17CFA as a marker for the selection of fluorescent pseudo­monads with biocontrol agent (BCA) potential was demonstrated by the isolation of a novel active strain. This was selected after the screening of 13 clonal groups of fluorescent pseudomonads identified from 500 isolates from the phytosphere of sugar beet. Levels of 17CFA synthesis possibly reflect the effi­cacy of the rpoS allele in particular strains.

Introduction

Many fluorescent Pseudomonas spp. are capable of reducing the incidence of plant diseases caused by soil­borne fungi. Few studies have examined the taxonomic and functional diversity of these bacteria. However, it is apparent that a large variety of antifungal metabolites are pro­duced by these organisms. Hydrogen cyanide (HCN) (Voi­sard et al., 1989), phenazines (phz) (Thomashow and Weller, 1988), phloroglucinols (phi) (Fenton et al., 1992; Keel et al., 1992), pyoluteorin (pit) (Kraus and Loper, 1995) and pyrrolnitrin (prn) (Pfender et al., 1993) are pro­duced by many pseudomonads, and attempts have been made to quantify the importance of each in disease sup­pression (Osburn et al., 1989; Maurhofer et al., 1994). However, production of characterized bioactive metabo­lites does not account for all of the observed antifungal activity (Kraus and Loper, 1992).

The suppressiveness of soils directly correlates with the density of populations carrying the phi biosynthetic locus (Raaijmakers et al., 1997), which has been detected in phenotypically and genotypically distinct pseudomonad isolates (Keel et al., 1996). However, expression was dependent on the genetic background of the strain into which the phi locus was inserted (Fenton et aL,1992), indi­cating that a variety of other factors influence biocontrol efficacy for a given pseudomonad. The high capacity of fluorescent pseudomonads to compete for seed exudates has been implicated in their rapid establishment on roots (Gamliel and Katan, 1992) and may thus lead to the exclu­sion of deleterious fungi. Pseudomonas putida N1 R reduces the amount of Pythium-stimulatory seed exudates avail­able to the fungal spores (Paulitz, 1991). The efficient assimilation of iron from soil has also been implicated in biocontrol efficacy (Leong, 1986; Becker and Cook, 1988; Loper and Buyer, 1991) and is attributed to the production of siderophores (Leong, 1986). However, these features alone do not fully explain the known disease suppression attributes of many fluorescent pseudomonads (Hamdan et al., 1991; Paulitz, 1991).

It is therefore highly relevant to establish which features of these fungal antagonistic strains distinguish them from other fluorescent pseudomonads. It is probable that they are all specialized for survival in soil and may be capable of the regulated production of antifungal compounds in the soil-root environment, but do they share other common phenotypes? Some important insight may be gained by

determining the ecological function of secondary metabolite production in bacteria and the environmental pressures that have led to the evolution of these pathways. It may be that production of these metabolites has evolved as a mechan­ism for survival under intense competition (Mazzola et al., 1992). Concomitantly, it has been suggested that those bacteria that demonstrate non-specific plant growth pro­motion may have evolved highly efficient signalling sys­tems with the host plants, which lead to specific plant­microbe interactions (Pierson and Pierson, 1996).

At present, the majority of potential biological control agents have been selected after the screening of large numbers of isolates for their ability to control disease in planta (Campbell, 1989). Seedling assays remain one of the most reliable methods for this, as they ensure that any antifungal activity observed is an in situ phenomenon. However, they are expensive and time-consuming to run, and the results are often variable. The in vitro identification of traits indicative of biocontrol activity would permit screening of larger numbers of isolates at a greatly reduced cost. Thus, the aim of this study was to compare phenotypic and genotypic characteristics of fluorescent pseudomonads with proven biocontrol potential against phytopathogenic fungi and relate the results to the ability to suppress Pythium ultimum infection in pea seeds.

Results

Comparative analysis of antifungal activity

The efficacy of each strain for the control of Pythium ulti­mum infection of pea seeds is shown (Table 2) and is con­trasted with the in vitro inhibition of hyphal extension on agar plates. Three distinct levels of antagonism were observed on plates. First, the bacterial colony was com­pletely overgrown; secondly, Pythium grew to the edge of the bacterial colony but not over it; thirdly, a distinct zone of inhibition was seen around the edge of the bacter­ial colony where Pythium growth was not observed. No correlation between in vitro antagonism and in situdisease suppression was observed. For example, F113 did not inhibit Pythium ultimum growth on plates, but effectively suppressed damping off [control index (CI) = 0.87], whereas PH6 generated a zone of inhibition on agar with a CI of only 0.33 (Table 2).

Fatty acid profiling

Fatty acid profiling of these 29 pseudomonads revealed a considerable degree of divergence within the collection (Fig. 1). By comparison with the commercially available database (MIDI, Newark, DE, USA), all but one of the strains were named within the rRNA homology group I pseudomonads (Table 1). The exception to this was PGSB 8456, which did not match with any entries in the database. An orange pigment, produced in abundance by this strain, was co-extracted with the fatty acid methyl esters (FAMES) and consequently interfered with the gas chromatographic analysis. No correlation was detected between the similarity of total FAME profiles of strains and their ability to suppress disease (Fig. 1). However, with the exception of strain 76/10, isolates with elevated levels of C17:0 cyclopropane fatty acid were effective in the sup­pression of damping off disease in peas. On this basis, the proportion of C17:0 cyclopropane fatty acid in each of the strains was identified as a variable for further correlation and stepwise multiple regression analysis as described below (Table 3).

Ribosomal DNA restriction fragment length polymorphism (RFLP) analysis

Ribotyping, as described here, demonstrated that there was a large degree of genetic variation in this collection of pseudomonads and that the majority of the strains were genetically distinct (Fig. 2). PGS12 and PGSB 8456 could not be differentiated by this method. In addi­tion, both strains also had a very distinctive orange pig­mentation when grown on solid media. Two other closely related strains, CHAO and Pf-5, also had similar FAME nrnfiles (Fig. 1)_

Although there was no direct correlation between taxo­nomic relatedness and absolute biocontrol activity, a small cluster of strains was identified that exhibited similar disease-suppressive and in vitro antagonistic activities (Fig. 2). These pseudomonads had a control index greater than 0.7 and produced an antifungal compound that could diffuse through agar, as identified by a zone of clearing

Fig. 1. Dendrogram indicating the relatedness of the biological control strains listed in Table 1 according to the analysis of whole-cell fatty acids. Euclidean distances are calculated on the basis of the relative proportions of individual fatty acids of pairs of strains.

around the colony. These strains were greater than 63.8 ± 6.8% genetically similar, as determined by their RFLP ribotype. Only two strains within this cluster, 2Ps4 and 1335, did not exhibit this combination of antifungal activity. Both showed some activity on plates, although there was no zone of clearing for 2Ps4, but performed poorly in disease suppression (CI <0.5).

Phenotypic analysis of isolates

BIOLOG carbon source utilization patterns further demon­strated the metabolic diversity of isolates (Fig. 3). Strains used between 33 (PH6) and 60 (1310 and PGSB 5589) of the 95 different carbon sources available, and no correlation was found between related clusters and antifungal activity.

The metabolic potential, substrate utilization patterns for each isolate and the phytopathogen Pythium ultimum were determined and expressed as the nutritional similar­ity index (NSI). The values obtained for isolates ranged between 0.560 (strain PH6) and 0.857 (strain 54/96) (Table 2). The majority of strains (> 85%) exhibited an NSI of greater than 0.7, demonstrating considerable simi­larity in the substrates used by Pythium ultimum and the fluorescent pseudomonads. Considerable differences in the growth rate of isolates in pea seed exudate were observed (Table 2). Mean generation times (MGTs)

varied between 2.227 h (strain UWC1) and 6.596 h (strain Q2-87) with the MGT for most isolates (> 90%) less than 4 h.

Siderophore production, as determined by the colour change of chrome azurol S in an agar medium, was also highly variable (Table 2). Although some strains (C7) appeared not to produce any compounds with affinity for ferric iron, other isolates produced prolific amounts, creat­ing a large zone of orange (up to 180 mm², i.e. UWC1).

The potential of each strain to produce well-characterized antifungal metabolites was determined. Cyanide and phenazine-1-carboxylate production was assessed by screening bacteria for the products in vitro. Genotypic analyses were performed for the presence of genes homologous to 2,4-diacetylphloroglucinol or pyoluteorin biosynthetic loci (Table 2). Neither of these methods were able to determine whether biosynthesis of the com­pounds occurred in situ. Over 62% of the isolates in the collection showed a positive response to any of the tests, despite their observed ability to control damping off in pea seedlings.

Correlation and regression analysis of data

The matrix indicating the correlations between all charac­teristics analysed is shown in Table 3. The correlation

the only variable that could be used rigorously to predict activity was the ability to produce cyanide. However, the strong correlation of C17:0 cyclopropane fatty acid with both CI and HCN production implied that FAME analysis could serve as a highly suitable marker for the primary screening of collections to identify potentially effective Pseudomonas biological control agents.

between cyanide production and the proportion of C17:0 cyclopropane fatty acid was the highest at 0.7524 (P<0.001). Only these two traits showed significant corre­lations with the control index.

Small but significant negative correlations were found between siderophore production and mean generation time in pea seed exudate (PSE) and siderophore produc­tion versus NSI. A small positive correlation existed between NSI and MGT. All comparative data were sub­jected to stepwise multiple regression to test whether a combination of factors could be used to define the ability of an individual strain to protect pea seeds against Pythium ultimum infection (CI). These tests revealed that

Fig. 3. Dendrogram indicating relatedness between the strains listed in Table 1 as determined by BIOLOG metabolic profiling. Euclidean distances are calculated on the basis of the differences between profiles. These data were then clustered using the unweighted pair group method of averages.

Confirmation of the validity of identified markers

To determine the relevance of this observation, represen­tatives of the 13 numerically dominant clonal groups (with identical ribosomal RFLP banding patterns) identified from more than 500 fluorescent pseudomonads isolated from sugar beet phytosphere and soil (Ellis et al., 1999a) were screened for the characteristic presence of a high relative proportion of C17:0 cyclopropane fatty acid. One of the 13 groups (ribotype 11; Ellis et al., 1999a) consisted of isolates with a C17:0 cyclopropane fatty acid content of 15%, whereas the other groups consisted of isolates with a C17:0 cyclopropane fatty acid content of less than 5%. A single isolate from the ribotype 11 clonal group, P. chlororaphis S34/10, was assayed for the ability to sup­press Pythium ultimum infection. The control index of S34/ 10 (CI ˆ 0.83) was equivalent to those strains that had been selected for biocontrol efficacy by more traditional and protracted (seedling assay) methods. Therefore, FAME analysis of strains from a large collection of natural isolates facilitated the direct and successful identification of an effective antifungal biological control agent.

Discussion

Despite the observation that the collection of Pseudomo­nas strains shared the common feature of being able to suppress Pythium ultimum infection of pea seeds in planta, isolates represented a genetically and phenotypi­cally diverse group. FAME profiling identified the majority of the strains as members of the rRNA homology group I pseudomonads (Table 1), as described by Palleroni (1984). In some cases, the identification provided by FAME profiling did not correlate with established affilia­tions; this is a result of the inadequacies of the taxonomic libraries for closely related strains rather than variability in the reproducibility of the FAME profiles of isolates. How­ever, analysis of these and other characteristics facilitated assessment of the extent of inter-relatedness and the importance of a previously unreported FAME marker for antifungal activity. Versatility and variation is typical of Pseudomonas, but it was anticipated that isolates from plant-associated habitats would exhibit some traits in com­mon. The limited genetic similarity between the known BCA strains (Fig. 2) was similar to that seen in other studies using similar methods for the assessment of relatedness

Fig. 2. Dendrogram indicating genetic relatedness between the strains listed in Table 1 as determined by RFLP analysis of the rDNA operon on two enzyme digests of genomic DNA. The scale shows percentage similarity as determined by Dice coefficients.

of fluorescent pseudomonad populations (Rainey et al., 1994; Ellis et al., 1999a). This implies that the level of simi­larity between the predominantly BCA strains in this col­lection is no greater than that of any random group of Pseudomonas spp. selected from soil or plant tissue samples. Thus, antifungal activity appears to be a general and genetically widely dispersed function of soil and phy­tosphere pseudomonads and not a trait specific to a specialized, genetically restricted group of pseudomonads.

Cluster analysis based on phenotypic traits (FAME and carbon source utilization (Figs 1 and 3 respectively) demonstrated the phenotypic diversity of the collection. However, the lack of congruence between clusters from phenotypic (Figs 1 and 3) and genotypic (Fig. 2) data was interpreted as a reflection of functional and genetic diversity of plant growth-promoting bacteria (Rainey et al., 1994; Natsch et al., 1997). Neither of the methods applied provides a definitive classification of the strains and, therefore, care should be taken in interpreting data to reveal inter-relationships that are based on a single method of analysis.

Niche overlap indices, as determined by nutritional pro­filing (NSI), have been used to identify strains that are cap­able of co-existence in the same habitat (Janisiewicz, 1996) and have been reversed to identify organisms that are not compatible (Wilson and Lindow, 1994). However, nutritional profiles of fungi and bacteria have not been compared previously for the purpose of biological control. The ability of pseudomonads to use a wide range of car­bon sources has been well documented (Palleroni, 1992; Schroth et al., 1992) and is reflected in these results, but the nutritional versatility of Pythium ultimum was unex­pected. The relatively high similarity between the meta­bolic potentials of the fungi and bacteria may reflect the range of nutrients available in soil and plant-associated habitats. The problem with comparing metabolic profiles of distantly related organisms is that BIOLOG plates pro­vide no information on the relative affinity of each organ­ism for any given substrate and, therefore, the relevance to the ability to use carbon sources at ecological concen­trations is speculative. However, the profiles do identify the potential for metabolism and thus compare the poten­tial for competition, although the substrates found in BIOLOG plates may be of limited environmental relevance. A posi­tive correlation between NSI and MGT values (Table 3) supports the view that nutrient competition plays an impor­tant role for individual isolates in the control of fungal pathogens, although no correlation was drawn between either NSI or MGT and the control index for the group as a whole.

The lack of correlation between in vitro inhibition of fungi and the ability to suppress disease caused by those fungi in planta has been documented previously (Reddy et al., 1993). This may be attributed to the differential expression of genes in situ and in vitro. As the regulation and expres­sion of bacterial genes will be different under different conditions, such as artificial media compared with environ­mental growth conditions, it is axiomatic that all potential biocontrol agents must be screened for their in planta activity. As observed above, the potential to produce anti­fungal metabolites is of relevance, but competitiveness and ecological fitness are also significant traits. Forexample, hydrogen cyanide production correlated with disease sup­pression (Table 2) and, in previous investigations, HCN has been identified as an important factor in biocontrol activity (Voisard et al., 1989), in which the distribution of cyanide production was estimated to be - 50% in some collections of pseudomonads (Bakker and Schippers, 1987; Nielsen et al., 1998). Of the 29 isolates examined in detail here, nine produced HCN under laboratory con­ditions. This was a greater proportion than was identified to produce or have the ability to produce any other sec­ondary metabolite, but the genetic diversity of this collec­tion taken from many sources was greater than those used in previous comparative studies. As with other antifungal secondary metabolites, HCN production is regulated under the control of stationary-phase sigma factors (RpoS) and the global activator (GacA) (Reimmann et al., 1997). The ecological function of secondary metabolites is unclear, but they appear to enhance survival. As cyanide increases exudation of nutrients from plant tissue (Astrom, 1991), strains that produce cyanide may be more adept at niche exploitation with increased competitive fitness in the sper­mosphere and phytosphere.

During the course of these comparisons, the novel cor­relation between C17:0 cyclopropane fatty acid (CFA) and antifungal activity was observed. There is no evidence that indicates that the fatty acid has antifungal activity, merely that it serves as a useful phenotypic marker. As with HCN, and other antifungal metabolites, the formation of C17:0 CFA occurs primarily in the stationary phase of the growth cycle. In Escherichia coli, CFA synthase is pri­marily under the control of rpoS (Wang and Cronan, 1994; Grogan and Cronan, 1997; Eichel et al., 1999); this may also be the case for Pseudomonas species. Thus, the level of CFA in any given strain may indicate the overall efficiency of the production of metabolites controlled by stationary-phase regulators, such as the sigma factor U^s. The importance of the complex signalling cascade, including RpoS, has been demonstrated for secondary metabolites in Pseudomonas aeruginosa (Latifi et al., 1996). It has been shown that different alleles of rpoS exist in E. coli (Ferreria et al., 1999) and that they affect the downstream expression of RpoS-controlled genes (Wang and Cronan, 1994). Thus, an efficient production of CFA in a pseudomonad strain, assuming that it is RpoS controlled, may indicate that the strain possesses one of the more efficient rpoS alleles. Therefore, other RpoS-mediated genes may also be similarly highly expressed, including those coding for secondary metabo­lites with antifungal properties.

The utility of C17:0 CFA as a marker for the effective suppression of damping off disease in pea was demon­strated by the selection of Pseudomonas chloroaphis S34/10 from a collection of sugar beet isolates (Ellis et al., 1999a). C17:0 CFA is common to many fluorescent pseudomonads (Vancanneyt et al., 1996), and all but one strain in this collection had detectable quantities of C17:0 CFA (Table 2). Therefore, screening collections or isolates for elevated synthesis of C17:0 CFA could be of general relevance for the primary screening of biocontrol agents, particularly as the assessment is rapid and does not depend on any prior knowledge of secondary metabolite synthesis.

Although alternative sigma factors are known to affect virulence in P. aeruginosa (Suh et al., 1999) and fungal disease suppression in P. fluorescens (Sarniguet et al., 1995), it is possible that each of the secondary metabolites identified in these and other strains serves only to indicate other unresolved functions in these pseudomonads. This may explain why the products so far identified do not com­pletely account for the observed biological activity. Further work is required to elucidate the importance of the rpoS genotype and related stationary-phase regula­tors in the expression of bioactive secondary metabolites and the associated diversity in CFA production. A full understanding of the regulation of these metabolites, which are produced as bacterial cells responding to envir­onmental signals that induce the general stress response, will significantly aid the search for effective biocontrol agents.

Experimental procedures

Bacterial strains: storage and culture conditions

The fluorescent pseudomonad strains and other bacteria and fungal strains used are given in Tables 1 and 4 respectively. Pseudomonas isolates were cultured on Pseudomonas agar base supplemented with 10 mg 1^-' cetrimide, 10 mg 1^-' fucidin and 50mgl^-' cephaloridine (PSA-CFC; Unipath) at 28°C. Inocula for all assays were produced by growing isolates in LB at 28°C with shaking at 180 r.p.m. for 18 h. Cells were then pelleted by centrifugation of the cultures at 5000g for 10 min followed by washing twice in sterile double-distilled water (SDDW) before resuspending in the original culture volume of SDDW.

Propagation and maintenance of fungi

Pythium ulfimum was stored in the dark at 15°C as infested soil prepared by drying and sieving spent soil from infected seed trials (Murray, 1994). The pathogen was isolated from infested soil by plating diluted soil suspension onto potato dextrose agar (PDA; Unipath) supplemented with 320 mg 1^-' aureomycin (Cyanamid; PDAA; Thompson et at, 1993a) or tap water agar [TWA; 1.2% agar no. 3 (Unipath) in tap water]. Plates were incubated at 20°C for 18 h. Pythium was identified by its rapid growth rate (Stanghellini and Hancock, 1970).

Oospore suspensions were produced as described pre­viously (Ellis et at, 1999b). Oospores were finally resus­pended in SDDW and stored at 4°C.

In planta quantification of disease suppression

An assay for studying the factors involved in the biocontrol of Pythium was adopted (Ellis et at, 1999b). Bacteria were added by applying a suspension in water, prepared as described above, directly to the soil. Twenty-five peas were pressed into the surface of the soil, and the assays were incubated in the dark at 21°C to optimize the development of disease. Triplicate plates were set up for all treatments, including uninoculated and mock-inoculated controls, and scored for disease after 7 days. The control index (CI) was calculated as described previously (Ellis et at, 1999b) and is essentially the relative proportion of infected seeds com­pared with the mock-inoculated controls.

In vitro screening of fungal inhibition

Each bacterial strain was grown on PSA-CFC overnight at 28°C. Purified oospores (100 µ1; ^-1 x 10⁸ propagules ml^-') were spread on the surface of fresh PSA-CFC (low-iron)

plates and allowed to dry. A single colony of each strain was then picked from fresh overnight plates and spotted onto the Pythium-inoculated plates. After incubation at 28°C for 30 h, the extent of inhibition of mycelial development was assessed. Bacterial colonies overgrown by hyphae scored zero, hyphae at the edge of the colony -1 and a distinct zone of clearing around the colony -2.

Fatty acid profiling

Phenotypic differences were assessed by fatty acid methyl ester (FAME) analysis. Isolates were streaked on tryptic soy broth (Difco) with 1.2% (w/v) agar (TSBA) in triplicate and incubated at 28°C for 24 h. Cells (50 mg wet weight) were har­vested into glass test tubes and the FAMEs extracted as described previously (Thompson et at, 1993b).

Ribosomal RFLP analysis

Genomic DNA was isolated from bacteria by an adaptation of the CTAB method described previously (Ellis et at, 1999a). Approximately 1 µg of DNA was digested with either of the restriction enzymes Kpnl or EcoRl (10 units; Boehringer Man­nheim) at 37°C overnight. DNA fragments were separated by field inversion gel electrophoresis (FIGE; Bio-Rad). Electro­phoresis and transfer of DNA to membranes was carried out as described previously (Ellis et at, 1999a). Membranes were probed with the 4.5kb BamHI fragment from pAC10 carrying both the 16S and the 23S ribosomal RNA genes (Housiaux et at, 1988). The fragment was purified from agar­ose, and 50 ng was labelled with the non-radioactive ECL system as described by the manufacturers (Amersham). Hybridization and visualization of the genes were carried out according to t