INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Certain plant-associated bacteria, particularly fluorescent Pseudomonas spp., have been exploited for suppression of crop diseases, and the importance of these bacteria in agriculture is expected to grow (5). Commercial development of biocontrol entails large-scale production of inoculants. Bacterial inoculants, regardless of their intended use (e.g., agricultural, pharmaceutical, food processing, manufacturing), are typically mass produced in industrial fermentors, and small batches are used to inoculate increasing fermentation volumes, a process referred to as scale-up (33). A stream-lined process (i.e., a cost-effective process) that results in a high yield and optimal efficacy is the primary objective in fermentations designed to recover viable cells. Culture media are prepared from an eclectic assortment of ingredients and are generally nutrient rich, which does not reflect the conditions in most natural bacterial environments. This is particularly evident for biocontrol agents originally isolated from the rhizosphere or phyllosphere, where nutrients are often limiting. Considering this and the scale-up process necessary to prepare large volumes, liquid fermentation of bacterial inoculants is disturbingly similar to repeated transferring and prolonged incubation in artificial growth media, laboratory practices long known to generate spontaneous mutations in microorganisms.

Genetic and molecular analysis has demonstrated that production of various antifungal compounds is a primary mechanism of biocontrol for many strains, accounting for as much as 90% of the disease-suppressing activity (36). As more biocontrol strains are analyzed, it is becoming apparent that biosynthesis of antifungal secondary metabolites in Pseudomonas spp. is commonly controlled by a two-component system comprising the sensor kinase GacS and the response regulator GacA. gacS is the new designation for a group of conserved genes in Pseudomonas spp. that encode functionally homologous cognate sensor kinases (e.g., apdA, lemA, pheN, and repA) (19). Genetically constructed GacS and GacA mutants are less inhibitory for fungal pathogens, presumably due to loss of antibiotics and hydrogen cyanide (HCN) (35). Recent evidence indicates that a gac mutation also negatively affects other regulatory elements, including autoinducers and sigma factors (1, 28, 40).

Despite obvious potential problems with instability in gacS and gacA or other genes important in biocontrol, little if any effort has been made to document, to understand, and to control these problems during inoculant production. Here we describe accumulation of a high number of spontaneous GacS and GacA mutants in liquid cultures of the Swiss biocontrol strain Pseudomonas fluorescens CHA0. The importance of gacA in biosynthesis of the antifungal metabolites 2,4-diacetylphloroglucinol (PHL), pyoluteorin (PLT), and HCN and the role of this gene in fungal inhibition and the biocontrol activity of strain CHA0 have previously been demonstrated by using gene replacement and transposon insertional mutation (21). Less is known about the function of gacS in this strain (3). Our objectives were to phenotypically characterize the spontaneous regulatory mutants and to determine their impact on the biocontrol efficacy of bacterial inoculants. We then tried to identify the selective pressures that favor mutant accumulation during inoculant production and to develop a cost-effective approach to minimize genetic instability in P. fluorescens biocontrol strains.

(A preliminary report of this work has been published previously [8].)

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Bacterial strains, mutant derivatives, and culture media. The strains and plasmids used in this study are described in Table 1. Wild-type strain CHA0 was originally isolated in 1983 from a Swiss sandy loam soil that naturally suppressed tobacco black root rot (15). An archival sample from 1985, kept at 80°C, was used in this study. Strains CHA510, CHA89, and CHA96^rif are genetically engineered regulatory mutant derivatives of CHA0. Spontaneous regulatory mutant derivatives CHAS9, CHAS17, and CHAS45 were isolated from stationary-phase nutrient broth cultures, and mutant CHASP1 was isolated from tobacco roots that had been inoculated with wild-type strain CHA0 and grown under gnotobiotic conditions for 6 weeks. Wild-type strains PGNL1, PGNR1, and PGNR4 were isolated from tobacco roots grown in a Ghana silt loam soil that suppressed tomato root diseases; and wild-type strains PINR2 and PINR3 were isolated from tomato roots grown in an Albenga, Italy, sandy loam soil that suppressed Fusarium wilt. Spontaneous mutants of these strains were isolated from orange sectors that appeared in colonies grown for 10 to 14 days on King's medium B (KB) agar (18). Plasmids used for genetic complementation were vectored by Escherichia coli. All bacteria were stored in dilute 0.08% nutrient broth (Difco, Detroit, Mich.) supplemented with 40% glycerol at 80°C. Fresh cultures were started from glycerol stocks for each experiment by plating portions onto KB agar.

Mutant characterization. A total of 578 spontaneous mutants with a distinct orange colony phenotype were isolated from 192 NBY broth cultures of wild-type strain CHA0 that were incubated for 12 days. The mutants were analyzed to determine their genetic similarity to the wild type by using a method based on a PCR performed with randomly amplified polymorphic DNA markers. Primer D7, obtained from a series of random oligonucleotides (Operon Technologies, Almeda, Calif.), provided consistent and distinct band patterns with polymorphic markers specific to strain CHA0 (16). Bacteria were grown in wells of microtiter plates containing 50 µl of dilute (0.1×) KB broth and were incubated for 24 h at 27°C with gentle agitation. The methods used for sample preparation, PCR amplification, and gel electrophoresis were methods that have been described previously (16).

Influence of mutant contamination on inoculant efficacy. Suspensions of wild-type strain CHA0, gacS mutant strain CHAS17, and gacA mutant strain CHAS33 cells were prepared from NBY broth cultures as described above. Suspensions were combined to obtain mutant concentrations ranging from 0 to 100%. Pregerminated cucumber seeds were soaked in the suspensions and planted in Pythium-infested soil. The percentages of seedlings that emerged and were standing were determined after 10 days. The treatments consisted of three replicate pots containing 15 seeds each, and the experiment was repeated once. Nonbacterized seeds served as a disease control that was not included in the analysis.

Four assays to determine the influence of mineral amendments on mutant accumulation. Unless otherwise indicated, bacteria were grown in 20-ml portions of NBY broth in 100-ml Erlenmeyer flasks and were incubated for 48 h at 27°C with shaking at 140 rpm in the dark. Filter-sterilized mineral solutions were added to autoclaved media to obtain a concentration of 1.0 mM [B(OH₃), CaCl₂ · 2H₂O, FeSO₄ · 7H₂O, LiCl, MgSO₄ · 7H₂O, Mo₇(NH₄)₆O₂₄ · 4H₂O, MnCl₂ · 4H₂O, NaCl], 0.7 mM (CuSO₄, ZnSO₄ · 7H₂O), or 0.1 mM (CoCl₂ · 6H₂O). Cultures were inoculated with 10-µl portions of overnight precultures that were diluted 1/10 so that the concentrations were approximately 10³ to 10⁴ CFU/ml. Wild-type precultures contained no detectable mutants (<1 × 10⁴ CFU/ml). Mixtures of the wild type and mutants were prepared by combining precultures and then were used to inoculate cultures. Sampling was done by plating appropriate serial dilutions onto KB agar amended with chloramphenicol (30 µg/ml) (KB^cm agar), a natural antibiotic resistance marker for strain CHA0 (38). Other P. fluorescens strains were plated onto nonamended KB agar. Colonies were enumerated, and the percentage of orange mutants relative to nonpigmented wild-type colonies was determined after 5 days.

Effect of zinc and medium dilution on mutant accumulation in other biocontrol strains. For each strain, mixtures containing 99% wild-type strain CHA0 and 1% gacA mutant were used to inoculate NBY broth, dilute (0.1×) NBY broth, and NBY broth containing 0.7 mM ZnSO₄ · 7H₂O. Mutants of each strain were HCN and protease negative and had an orange colony phenotype identical to that of CHA0 mutants, which was used to determine the percentage of mutants after 48 h. Treatments were arranged as a 5 × 3 factorial with a main plot for strain and a subplot for medium. Treatments consisted of three replicates, and the experiment was conducted twice.

Influence of minerals and medium dilution on growth of wild-type strain CHA0 and mutants. Wild-type strain CHA0 and spontaneous mutants were grown individually in NBY broth, dilute NBY broth, dilute NBY broth containing NaCl, and NBY broth containing minerals. After 48 h, the numbers of CFU per milliliter were determined by plating samples onto KB^cm agar. Treatments were arranged as a 6 × 14 factorial in a split-plot design with a main plot for bacterial strain and a subplot for medium treatment. The experiment consisted of six replicates studied over time. The growth rates of CHA0, CHAS17, and CHAS33 were determined by measuring the optical densities at 600 nm from zero time to 48 h in 150-ml portions of NBY broth, dilute (0.1×) NBY broth, dilute NBY broth containing 30 mM NaCl, and NBY broth containing 0.7 mM CuSO₄ or 0.7 mM ZnSO₄ · 7H₂O. Treatments consisted of two replicates.

Data analysis. Bacterial CFU data were transformed by using the logarithmic base 10, and percentage data were transformed by using the arcsine of square roots prior to analysis of variance. Unless indicated otherwise, treatments were arranged in a randomized complete block design, and experiments were repeated two to four times. Data from repeated trials were pooled after we confirmed in a preliminary analysis that the trial-main effects interaction was not significant and/or that variances between trials were homogeneous as determined by an F test or Bartlett's test. For most experiments, main effects and interactions were further analyzed for significance with the SAS general linear models procedure (Statistical Analysis Systems Institute, Cary, N.C.), with the mean comparisons performed by using Fisher's protected least-significant-difference (P = 0.05) (LSD_0.05) test. SAS regression procedures were used to determine relationships between mutant content and inoculum efficacy and between zinc concentration and mutant accumulation.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Mutant phenotypic, genotypic, and biochemical characterization. Spontaneous mutants appeared at a high frequency (approximately 1%) in stationary-phase cultures of CHA0. Mutants were easily distinguished from the wild type in dilution-plated samples based on the unusual appearance of colonies (i.e., orange color; flattened; expanded; often transluscent; surrounded by a more intense, diffusible, yellow, fluorescent pigment, whose intensity increased over a period of 5 days). The correlation between the orange colony color and the loss of HCN, protease, and TSO production was approximately 98%. Orange mutants were indistinguishable from the wild type by PCR-randomly amplified polymorphic DNA analysis. A total of 49.7% of 205 orange mutants were restored to a wild-type phenotype with a gacS clone, and 48.2% were restored with a gacA clone. Of the remaining 2.1% pleiotropic mutants not restored with either of the single-gene clones, none required both clones for complementation. Generally, GacS and GacA mutants behaved similarly in all tests throughout this study, and spontaneous mutants were indistinguishable from genetically engineered derivatives.

5

Spontaneous mutation compromised the biocontrol efficacy of inoculants. Wild-type strain CHA0 inhibited the growth of Pythium spp. on KB agar with or without supplemental iron (Table 2). Added iron actually improved the inhibitory activity of CHA0, probably by stimulating antibiotic biosynthesis. In contrast, GacS and GacA mutants lost the ability to inhibit Pythium growth on KB agar amended with iron, which repressed production of diffusible fluorescent pigments typical of pyoverdine siderophores. In the absence of added iron, mutants overproduced fluorescent pigment, and their inhibitory activity was identical to that of the wild type.

View larger version (17K): [in this window] [in a new window]   FIG. 1.   Influence of mutant contamination on biocontrol efficacy of CHA0 inoculants. Cucumber seeds were treated with suspensions of wild-type CHA0 inoculum contaminated by adding a gacS (CHAS17) or gacA (CHAS33) mutant at a range of concentrations from 0 to 100%. Seeds were grown in soil infested with P. ultimum. The percentages of seedlings that emerged and were standing after 10 days are shown. The values are means based on six replicates; the error bars indicate standard errors.

Certain mineral amendments and medium dilution improved the genetic stability of CHA0 cultures. Four approaches were used to evaluate the influence of culture conditions on the accumulation of regulatory mutants in NBY broth. First, we determined the effects of copper amendment, medium dilution, and electrolyte concentration on the appearance of mutants in wild-type cultures that contained no detectable mutants at the start of the experiment. All treatments significantly reduced the accumulation of orange mutants compared to normal-strength (1×) NBY broth, and dilute (0.1×) NBY broth provided the best control of mutant accumulation (Table 3). The validity of using orange colony color to identify mutants was supported by the fact that nearly identical results were obtained when randomly sampled colonies from each treatment were tested to determine whether they produced HCN, protease, and TSO. In NBY broth, approximately 1% of the colonies were negative for these metabolites. In comparison, no negative colonies were observed in dilute NBY broth, and only 0.2 and 0.3% of the colonies were negative in copper-amended NBY broth and dilute NBY broth containing NaCl, respectively. Total bacterial growth was less for all treatments than total bacterial growth in normal NBY broth (Table 3).

View larger version (18K): [in this window] [in a new window]   FIG. 2.   Mutant accumulation from contaminated seed cultures through three stages of scale-up from 20 to 100 to 500 ml (Scale 1 to 3). The inoculation and sampling methods used are described in Materials and Methods. The lines indicate origins of inocula. Cultures were grown for 48 h in NBY broth (solid triangles), dilute (0.1×) NBY broth (open triangles), and NBY broth containing 0.7 mM CuSO4 (grey triangles). The values below the symbols are the average percentages (standard errors) of mutants based on 14 replicate broth preparations.

View larger version (19K): [in this window] [in a new window]   FIG. 3.   Effects of minerals on competition between CHA0 and spontaneous orange mutants. NBY broth (control), dilute (0.1×) NBY broth, dilute NBY broth containing NaCl, or NBY broth containing minerals was inoculated with a low but detectable level of orange mutants (approximately 0.3%). After 48 h, the percentage of mutants was determined. The bars indicate means based on 16 cultures; error bars indicate standard errors. Fisher's protected LSD0.05 value was 6.71%.

View larger version (30K): [in this window] [in a new window]   FIG. 4.   Effects of minerals on competition between CHA0 and gacA mutant CHAS9 (A) or CHAS45 (B) or gacS mutant CHASP1 (C) or CHAS17 (D). Broth preparations (see the legend to Fig. 3) were inoculated with a bacterial mixture containing 90% wild-type strain CHA0 and 10% mutant. After 48 h, the percentages of mutants were determined. The bars indicate means based on eight cultures; the error bars indicate standard errors. Fisher's protected LSD0.05 values were 10.4% (A), 8.2% (B), 11.3% (C), and 6.8% (D).

View larger version (23K): [in this window] [in a new window]   FIG. 5.   Relationship between zinc concentration and mutant accumulation. NBY broth preparations amended with a range of zinc concentrations were inoculated with a mixture containing 90% wild-type strain CHA0 and 10% spontaneous mutant CHAS17 (A) or CHAS45 (C) or genetically engineered mutant CHA510 (B) or CHA96rif (D). After 48 h, the percentages of mutants were determined. The values are means based on six cultures; the error bars indicate standard errors. The regression lines approximate 6x2  26x + 25 (P  0.0001; r2  0.93).

Reduction of spontaneous mutation in other biocontrol strains. Spontaneous mutants defective for HCN, protease, and the antibiotics PHL and PLT were readily recovered from five wild-type biocontrol pseudomonads isolated from tobacco roots grown in soil from Ghana (PGNR1, PGNR4, PGNL1) and Italy (PINR2, PINR3). Orange translucent sectors composed of regulatory mutants appeared in colonies grown for an extended period (about 10 days) on KB agar. Over time, mutants eventually overgrew the wild type. These orange mutants were phenotypically identical to those observed with strain CHA0 and were complemented with a gacA or gacS clone. When a wild type was combined with a corresponding GacA mutants at a concentration of 10%, mutant accumulation was reduced by zinc amendment and medium dilution (P = 0.0289) compared to mutant accumulation with full-strength NBY broth (Table 4). This was true for all five strains. However, a significant strain-medium interaction (P = 0.0113) indicated that some strains responded better than others to zinc amendment and medium dilution. In zinc-amended NBY broth, the reductions in mutant accumulation compared to nonamended NBY broth ranged from 4.6-fold for PGNR4 to 17.2-fold for PINR2. In diluted NBY broth, the reductions ranged from 4.3-fold for PINR3 to 19.8-fold for PGNL1. It was also evident that some strains (e.g., PINR3) were more susceptible to mutant accumulation than others regardless of medium, which indicated that genetic stability varied among biocontrol isolates (Table 4).

Mineral and medium dilution effects on mutant accumulation were independent of any effect on growth of individual bacteria. A significant mineral-strain interaction and significant main effects (P = 0.0001) indicated that medium treatments had differential effects on the yield of culturable bacteria after 48 h of growth. Generally, for any given medium there were no consistent differences between the wild type and mutants (Table 5) and no apparent relationship between medium effects on growth and mutant accumulation (Fig. 4). For example, in zinc-amended NBY broth growth of one GacA mutant, CHAS9, was slightly reduced compared with growth in nonamended NBY broth, but growth of another GacA mutant, CHAS45, was not affected (Table 5). In copper-amended NBY broth, the reverse was true, and growth of CHAS45 but not growth of CHAS9 was reduced. Zinc and copper did not affect growth of either of the GacS mutants or the wild type. However, both zinc and copper reduced accumulation of all mutants in competition experiments (Fig. 4). Furthermore, cobalt, which also reduced mutant accumulation (Fig. 4), did not affect growth of any of the mutants but reduced growth of the wild type (Table 5). Growth of all strains was reduced in dilute NBY broth and dilute NBY broth containing NaCl, but there were generally no differences among strains (Table 5).

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

Our results document for the first time that maintaining genetic stability during liquid fermentation is critical for production of high-quality biocontrol inoculants (32, 39). The gacS and gacA global regulatory genes of several P. fluorescens strains were very unstable in certain media. Even though all strains were susceptible, the degrees of instability of strains varied. The ability of an inoculant to protect cucumber from disease was significantly reduced by as little as 10% mutants and was essentially abolished when the level of contamination reached 50%. Such levels occur because mutants multiply exponentially when processes are scaled up to increasingly larger fermentor volumes. Our method for identifying gac mutants by dilution plating and visual assessment of distinct colony phenotypes (e.g., darker pigmentation, sprawling, hyperfluorescence) offers an inexpensive and simple option for routinely monitoring the quality of inoculant fermentations. Mutants of our biocontrol strains had the same orange-brown color, compared to beige for the wild type, but in other strains the color to watch for may be different (4, 11).

Mutants lost the ability to produce all of the key antifungal metabolites (e.g., PHL, PLT, hydrogen cyanide). Increased siderophore production, which compensated for the loss of these compounds in in vitro inhibition assays, did not compensate in the rhizosphere, where sufficient iron may have been available (24). There was a significant inverse relationship between inoculant efficacy and mutant contamination which reflected a dose-response relationship like that defined in various biocontrol systems. A threshold population density of bacteria is required for significant disease suppression, and relatively small decreases in the size of a population can dramatically reduce the level of protection (14). We extended this idea by specifying that a threshold population of biocontrol-active cells is needed for effective disease suppression. Thus, mutant contamination compromises inoculant performance by diluting the dose of biocontrol-active cells in the final product.

Having identified gacS-gacA instability as a problem, we set out to understand why mutants accumulate during inoculum production and to develop an approach to manage this problem. The results of scale-up experiments, in which small amounts of mutants were repeatedly transferred into fresh media, indicated that mutant accumulation resulted primarily from greater mutant competitiveness rather than increased frequency of mutational events. Further evidence that competition is the mechanism behind mutant accumulation comes from the fact that when mutants and wild-type strain CHA0 were grown in various media individually, the numbers of CFU at the end of fermentation were similar. However, we cannot eliminate the possibility that minerals and medium dilution have different effects on growth rates. Other workers have recently sequenced gacA alleles from several mutants, which revealed a surprising diversity of mutational events (point, deletion, and frame-shift mutations involving as few as 3 bp) (C. T. Bull and D. Haas, unpublished data). This indicates that selection is for the Gac phenotype rather than for a particular mutational hot spot. The main selective factor for gac mutants appears to have been the nutrient content of the media because accumulation was far greater in full-strength NBY broth than in 0.1× NBY broth. However, because the beneficial effect of medium dilution was largely negated by raising the medium conductivity to the same level as the conductivity of NBY broth, the real culprit is more likely higher osmotic potential and not excess nutrient availability per se.

It is still possible that a gac mutation may be a starvation response akin to the GASP phenotype described for Escherichia coli and other bacteria, in which subpopulations which have a growth advantage during the stationary phase are selected (42). Even if our media retained an excess of most nutrients, certain key factors (possibly minerals) were gradually depleted during culture growth, which may have resulted in a shift towards mutants. This would explain why gac mutants typically appeared in older cultures. Compensating for such a deficiency may be the mechanism by which mineral amendments repressed mutants during the stationary phase, but this does not explain how minerals worked when they were added to fresh media. Mutation did not appear to increase the sensitivity to toxic metals like zinc, because growth characteristics were similar for mutants and wild-type strain CHA0 when they were inoculated individually. Overproduction of metal-chelating siderophores may have provided a competitive advantage to mutants as trace minerals became scarce, an advantage that was lost when excess amounts of the metals were supplied. In mycopathogenic Pseudomonas tolassii, gacS mutants are thought to appear because the gene is downregulated under starvation conditions, something like gac atrophy (11). Interestingly, most of the minerals that we found which repressed mutants have previously been shown to stimulate biosynthesis of antifungal metabolites regulated by gacS-gacA (7). Another unusual, but possible, explanation is that a compound in the media triggered mutation, in the same way that the plant phenolic compound acetosyringone selects for nonpathogenic variants of the pathogenic bacterium Agrobacterium tumefaciens (10) or benzyl alcohol selects for Tol variants in the bioremediation bacterium Pseudomonas putida (22).

Factors that influence the fitness of gac mutants could have long-term implications after inoculant application. Although in nature we have only rarely seen gac mutants like our tobacco root isolate CHASP1, such mutants appear to be at least as competitive as the wild-type strain in certain environments. Natsch et al. (26) found that a gacA insertion mutant of CHA0 was slightly less competitive in bulk soil, equally competitive in the rhizosphere, and more competitive on the rhizoplane and in the root interior compared to the wild type. Similar results were observed with a gacS insertion mutant of the phytopathogenic organism P. syringae pv. syringae which exhibited reduced colonization on bean leaves in the field but was equally competitive on germinating bean seeds (12). Understanding the conditions that modulate mutation and/or mutant accumulation may enable us to prevent gac mutants that contaminate inoculants from displacing biocontrol-active cells after application to plants. On the other hand, identifying environmental signals that are conducive to gac mutants may inspire new disease control strategies that can be used against pathogenic bacteria which also have very unstable gac genes and which are rendered generally avirulent by mutation (9, 11, 23, 29, 34, 41).