Non-target effects of bacterial biological control agents suppressing root pathogenic fungi

Abstract

Microorganisms have knowingly been used during the last century to control plant diseases. During the last decades, research and application of biological control agents (BCAs) as a pest control strategy have gained increasing attention. This review focuses specifically on non-target effects of bacterial BCAs that are used to suppress root pathogenic fungi. It attempts to critically evaluate the strengths and weaknesses of non-target effect studies published to date and relate them to the success of the BCA in fungal pathogen control. Significant non-target effects of BCAs have indeed been observed, but these are generally small in scale and limited to a growth season, and have not been proven to affect soil health. We discuss these studies and point out what we believe are notable deficiencies. Among the modes of disease suppression by BCAs, antibiotic production is believed to be of major importance. But assurances that in situ antibiotic production actually occurs in environmental samples are lacking in the non-target effect studies. Also the effectiveness of the BCA on the target pathogen, the absence of appropriate controls for inoculation effects, and the presence of pathogenic fungi are missing in most studies. In future non-target effect studies we recommend focusing on proven effective BCAs and clearly distinguishing effects of antimicrobial compounds from effects of general microbial activity.

1 Introduction

The use of microorganisms to control plant diseases has been practised during the last century. This strategy as an environmentally friendly alternative to chemical pesticides has gained attention especially during the last decades. Bacteria, fungi and protozoa have been suggested to combat pests in below- and above-ground plant material, and several biological control agents (BCA) are on the market today in the USA (http://www.epa.gov/pesticides/biopesticides/) and in other countries (http://www.agrobiologicals.com) [1].

Non-target effects can be defined as effects of the introduced BCA on organisms other than the target organisms or on biogeochemical cycles. For environmental risk analysis, direct or indirect non-target effects mediated through a chain of events or interactions among organisms will cause concern. Likewise, immediate effects or cumulative long-term effects will cause concern [2]. An important parameter is the duration of the effects, as generally large temporary fluctuations can be accepted [3]. Non-target effects can be categorized in the same way as Domsch et al. [3] categorized the side effects of agrochemicals on soil microorganisms, as negligible, tolerable, and critical, depending on the magnitude of the effect and the time needed for the system to recover. The definition of ‘negligible’, ‘tolerable’ and ‘critical’ is, of course, subjective, and prone to societal and/or political decisions. A prerequisite for introducing the BCA into the environment is generally that, in addition to effective disease suppression, the effects on non-target organisms should be at least tolerable, if not negligible. Since most target effects are not uniquely specific to pathogens, non-target effects can be expected. Consequently, considerable effort has been put forth to characterize and quantify these non-target effects. In practical terms, intolerable non-target effects should persist beyond the time of crop harvest and be significantly different from changes due to growth season and agricultural practices.

This minireview focuses specifically on non-target effects of bacterial BCAs that are used to control root pathogenic fungi. It attempts to critically evaluate important published non-target effect studies, their usefulness in risk analysis, and their relationship to the success of the BCA in pathogen control. Finally, we suggest future research approaches for non-target effects of BCA.

2 Mode of disease suppression by BCAs

Bacterial antagonists of root pathogenic fungi use a variety of suppressive and inhibitory mechanisms including (a) competitive advantage for resources (e.g. iron), physical space, or nutrients supplied by seed or root, (b) production of one or more antibiotics (e.g. 2,4-diacetylphloroglucinol (DAPG), viscosinamide, pyoluteorin (Plt), zwittermicin A, kanosamine, phenazine-1-carboxylic acid (PCA)) or other toxic substances (e.g. cyanide), (c) degradation of pathogenicity factors produced by the pathogen, (d) production of enzymes that degrade fungal cellular components thus inactivating the pathogen (e.g. chitinolytic and cellulolytic enzymes), and (e) induction of resistance in the target plant against the pathogen [4]. The BCAs subjected to thorough non-target studies are listed in Table 1 and encompass those considered in our review. A comprehensive list of bacteria applied to seeds or roots to control fungal plant pathogens is provided by Whipps [4], while Raaijmakers et al. [5] give a list of antifungal compounds produced by BCAs. Despite the extent of mechanisms for biocontrol, the effectiveness of many BCAs is reported to depend on metabolites with antimicrobial activity in combination with other mechanisms [5,6]. As a consequence, research into non-target effects primarily addresses the effects of these antagonistic substances. Establishment of antibiotic-overproducing mutants (e.g. CHA0/pME3424 [7] and F113(pCUGP) [8]), or mutants deficient in antibiotic production (e.g. F113G22 [9]) (Table 1), has facilitated investigations of non-target effects of antagonistic substances.

The antibiotics produced by BCAs differ in structure and mode of action. Phenazine compounds are heterocyclic compounds, while DAPG and Plt both are phenolic compounds. Viscosinamide is a cyclic depsipeptide with surfactant and antifungal properties and is produced by Pseudomonas fluorescens DR54 [10]. Viscosinamide is hypothesized to inhibit growth by forming ion channels in the fungal membrane [11], resulting in higher intracellular Ca²⁺ levels, which can trigger fungal encystment. In vitro tests for antagonism or toxicity tests of the antibiotics produced by the BCAs indicate that many BCAs not only inhibit a narrow group of target fungi, but may also be suppressive to other microorganisms present in soils [12–15] or on different crops [16–18]. These broad inhibitory capabilities of the antibiotics obviously call for the design of application procedures that limit non-target effects.

3 Non-target effects of BCAs

One of the most important concerns associated with the use of BCAs in plant production is the possible disruption of microbial processes that are essential to general soil ecosystem functioning. This includes effects on microorganisms and on carbon, nitrogen and phosphorus cycling, as well as general impacts on soil ecosystem functioning. Effects on non-target microorganisms (bacteria, fungi and protozoa), and on the biochemical functions they catalyze, are frequently studied in the rhizosphere where the BCA is introduced. This has led us to review non-target effects on microbial abundance and community structure, enzyme activities and available nutrients, microbial indicators, and other trophic levels such as protozoa, nematodes and plants.

3.1 Abundance of closely related bacteria

Numerous studies with P. fluorescens BCA candidates have screened for non-target effects on microorganisms that are closely related to the BCA itself, with the hypothesis that these organisms are most likely to be affected by the BCA due to competition for the same niches and resources. Indeed, effects are commonly seen, but only temporarily. Microcosm experiments with F113 and the antibiotic-overproducing mutant F113(pCUGP) generally showed either no effect or an inhibiting effect on indigenous pseudomonads 17–21 days after the BCA was introduced on pea seeds, while the antibiotic-producing negative mutant F113G22 had a stimulating effect [19–23]. In contrast, Brimecombe et al. [24] reported a decrease in indigenous pseudomonads in 17-day-old pea and wheat rhizospheres when inoculated with F113 or F113G22. In field studies with sugar beet, inoculation with F113 resulted in a reduction of the indigenous culturable fluorescent pseudomonads in the rhizoplane after 19 days, without affecting the rhizosphere pseudomonads [25]. Studies with CHA0 and its antibiotic-overproducing strain, CHA0/pME3424, showed a transient suppressive effect in cucumber microcosms [26] and either no effect or a transient effect in lysimeter effluent water [27]. The effects of DR54 have been studied in barley rhizospheres in microcosms [28]. Based on colony-forming unit (CFU) counts and quantification of 16S rDNA copies, the indigenous P. fluorescens population was reduced for approximately 2 weeks after introduction of DR54 compared to the untreated controls and a pesticide treatment. In 5-week-old wheat microcosm experiments with and without Pf29A and the fungus Gaeumannomyces graminis var. tritici (Ggt), indigenous pseudomonads and Pf29A increased. Pf29A became the dominant pseudomonas strain, and the indigenous P. fluorescens population was reduced [29]. Finally, SBW25, which is not known to produce any antifungal metabolites, had variable effects on indigenous pseudomonads, either reducing [23,30] or not affecting [19,23,31] the populations.

From these studies, it appears that introduced pseudomonads, with or without potential production of antimicrobial compounds, sometimes temporarily affect the abundance of indigenous pseudomonads in soil in different directions. Effects were not dependent on the inoculum density. The results and conclusions are, however, biased by the unavoidable inclusion of the inoculants in the measurement of the number of pseudomonads present, since the indigenous number of pseudomonads is calculated as the difference between the inoculant concentration and the total number. The variation and transient nature (less than 3 weeks) in reported effects point to the absence of any significant adverse impact on indigenous pseudomonads.

Stimulating effects of an introduced BCA may be caused by the indigenous population proliferating on the inoculated bacteria and/or any substrates added along with them. For example, an increase in microbial abundance as a result of a 10³-fold increase in inoculation density of CHA0 to lysimeter effluent water was explained by bioavailable carbon leaching from decaying CHA0 cells [27]. This proliferation may occur especially from inoculation of the DAPG-negative mutant F113G22 or from inoculation of SBW25, which is not known to produce antimicrobial compounds, but may also occur if the antimicrobial production in situ is hampered. The effects of the fungal pathogen on the plant and on the BCA itself may also stimulate the introduced BCA [29,32].

In the case of observed negative effects, this can be due to competition or antimicrobial effects of secondary metabolites. However, some studies have shown that many indigenous fluorescent pseudomonads are resistant to antagonistic pseudomonads [14] or high concentrations of the components active in antagonism (DAPG) [26]. Thus, competition for nutrients, substrates, and space is the most probable explanation for the reduction in indigenous pseudomonads.

3.2 Community structure of closely related bacteria

Effects on community structure of closely related organisms have been studied for some of the P. fluorescens BCAs. In a field study of 19-day-old sugar beets, F113 caused a shift in the phenotypic structure of the Pseudomonas rhizoplane community, while there was no effect on the genetic structure, as determined by amplified rDNA restriction analysis (ARDRA) and random amplification of polymorphic DNA of 498 isolates [25]. No effect was observed on the corresponding Pseudomonas rhizosphere communities. For CHA0 and CHA0/pME3424 the phenotypic, but not the genotypic, diversity of fluorescent pseudomonads present in the rhizosphere was altered due to inoculation for 10 days and only the CHA0-treated soil community had recovered after 52 days in microcosms. However, the microbial changes during 52 days of plant growth had a greater impact on the phenotypic diversity than the inoculations [26]. Likewise, the genotypic diversity of fluorescent pseudomonads was decreased after 5 weeks by Pf29A inoculation in the presence of Ggt, but not in the absence of Ggt [29]. Whether the observed changes in community structures have any importance for the ecological functions of the pseudomonads is unclear.

3.3 Microbial abundance

Assessment of the effects of BCAs on microbial abundance generally involves measuring changes in the number of culturable bacteria or fungi in soil, rhizosphere and/or rhizoplane. In vitro tests for antagonism or toxicity of purified metabolites from BCAs have clearly shown not only inhibition of a narrow group of target fungi, but also inhibition of the majority of indigenous soil bacteria and saprophytic fungi tested [13,14,27,33]. However, results of microcosm and field experiments are less clear. The Pseudomonas BCAs (F113, CHA0, DR54, SBW25, Q2-87, and WCS358r) and the Bacillus cereus BCA UW85 have been studied in various test systems, including microcosms with pea [21,23,24], wheat [24,29,30], cucumber [13,34], and barley [28,35] plants, soil lysimeter effluents [27], and field studies with sugar beets [36], wheat [30,31,37,38], and soybean [39]. The majority of non-target effects on microbial abundance seem to occur more or less randomly and show no relationship to experimental methods used or to the plant rhizospheres and BCAs tested. On the other hand, the inoculation of F113, CHA0, SBW25 and Q2-87 on peas in microcosms did result in a significant reduction of culturable rhizosphere and rhizoplane bacteria and rhizosphere fungi in the presence of Pythium[23]. In the absence of Pythium, however, F113 and Q2-87 had stimulating effects on the rhizoplane bacteria and F113, CHA0 and Q2-87 had stimulating effects on rhizoplane fungi, while F113 and CHA0 had inhibitory effects [23] or no effect on the rhizosphere bacteria [21]. The presence of Pythium thus seems to affect the non-target effects of the BCAs. This may be due to interaction between Pythium and the BCA, since pathogenic fungi are known to affect bacteria, including introduced BCAs [32,40–42]. But it may also be due to target effects of the BCA on the fungi, which may in turn affect the plant root exudation. However, the majority of the reported non-target studies of BCAs point to no substantial effects on bacterial abundance.

Fewer reports are available on non-target effects on abundance of soil fungi. Repeated growth cycles of CHA0- or CHA0/pME3424-treated cucumbers had no effect on the fungal CFUs present in the rhizosphere [34]. However, a significant reduction in the number of culturable filamentous fungi, specifically Fusarium spp., was observed following the field introduction of the high PCA-producing BCA, Pseudomonas putida WCS358r (derivative GMM 8), into wheat rhizospheres [37]. No reduction was observed in rhizospheres exposed to the wild-type non-PCA-producing strain, WCS358r. The actual detection of PCA in the rhizosphere inoculated with the PCA-producing strain and not in the rhizosphere inoculated with the wild-type WCS358r is a unique aspect of this study. It provides the strongest evidence to date that adverse non-target effects on microbial abundance can be observed if the secondary metabolite is produced. In a subsequent study, WCS358r and derivatives that were modified to produce either PCA or DAPG had no effect on abundance of either bacteria or fungi [38]. Other studies have shown effects on target fungi in rhizosphere soil, but without measurement of the in situ antibiotic production (e.g. [29,43,44]).

3.4 Microbial community structure

Analysis of non-target effects on community structure has been performed with culture-dependent and culture-independent methods, including both genetic and physiological assays. The rate of colony appearance on agar plates was reduced in the presence of F113, compared to F113G22 and SBW25, in pea rhizoplane microcosms [19]. A shift of indigenous bacteria toward K-strategists and/or an increase in stressed and starving cells in the rhizosphere may explain this observation. Also, CHA0 or CHA0/pME3424 affected the BIOLOG fingerprint (after 10 days, but not 52 days), and ARDRA analysis of dominating genotypes (among 2500 isolates) showed significant differences [13]. Significant treatment effects of CHA0 and CHA0/pME3424 were found when analysing more than 11 000 fungi, belonging to 105 species and 50 asexual morphotypes, by ARDRA [34]. The fungi were isolated from cucumber microcosms incubated for one to five growth cycles of 32 days each. On the other hand, DR54 did not affect functional and genetic diversity, based on BIOLOG and polymerase chain reaction-denaturing gradient gel electrophoresis (PCR-DGGE), of barley rhizosphere microbial communities [35]. However, phospholipid fatty acids (PLFA) analysis showed a transient but significant effect. After 123 consecutive introductions through nightly irrigations of turfgrass, TX-1 established itself in the soil and rhizosphere, over-wintered and was detected 230 days after the last introduction [45]. Despite this, genetic diversity based on PCR-DGGE was unaffected. Remarkably, the TX-1 was not itself detectable as any specific band in the PCR-DGGE community pattern, suggesting that PCR-DGGE may not have been very sensitive in the conditions used. In a 2-year experiment the potential BCA P. putida 06909 was added either once a year or at weekly irrigations during 7 monthsocuses specifically on non-target effects of bacterial BCAs that are used to control root pathogenic fungi. It attempts to critically evaluate important published non-target effect studies, their usefulness in risk analysis, and their relationship to the success of the BCA in pathogen control. Finally, we suggest future research approaches for non-target effects of BCA.

3.5 Enzyme activities and available nutrients

Naseby and Lynch [19–21] and Naseby et al. [22,23] have extensively studied the effects of F113 on soil enzyme activities in pea microcosms. Generally, soil enzyme activities related to the carbon cycle (β-glucosidase, β-galactosidase) decreased following introduction of F113, while enzyme activities involved in the cycling of phosphorus (alkaline phosphatase, phosphodiesterase), sulfur (sulfatase) and nitrogen (urease) increased. In the presence of Pythium, the effects of F113, CHA0, SBW25, and Q2-87 were reversed: the N-, P- and S-cycle enzymes were decreased compared to an untreated control [23]. The indicator of fungal activity, N-acetyl glucosaminidase, was also decreased, both with and without Pythium, as would be expected based on the antifungal effect of DAPG. SBW25 generally had the same effect as F113 on enzyme activities in the presence of Pythium while in the absence of Pythium, only β-glucosidase activity was affected [23]. In contrast, SBW25 changed some enzyme activities in wheat rhizospheres [50]. Field experiments lasting for 6 months showed that F113 had no detectable effect on soil enzyme activities and was itself below the detection limit [51].

The difference in response between plant species might be more related to inoculum preparation than to plant differences. In the pea experiments, SBW25 cells were scraped off agar plates and suspended in quarter-strength Ringer's solution, while in the wheat experiments, cells were cultured to late exponential phase in broth before being coated onto the seeds in a guar gum solution. The concentration of produced antibiotics may vary depending on culturing conditions, as production of some metabolites (e.g. DAPG) is higher when cells are grown on solid media. Both ways of preparing the inoculum have the potential of introducing spent media that could affect microbial activities. F113 introduction also resulted in higher amounts of organic acids in microcosms than did F113G22 introduction, while both strains led to increased contents of water-soluble carbohydrates, organic acids, and other water-soluble carbon sources [22]. This was probably due to carry-over of spent media from the inoculation of cells scraped off agar plates or from nutrients released from lysis of the introduced BCA, in addition to effects on roots and microbiota.

In the field, some reductions in available nutrients have been observed to result from F113 inoculation, but the magnitude of their effects was smaller than the heterogeneity among soil samples, and comparable to the effects of agronomic practices [52]. Until further work is performed, the usefulness of effect studies on available nutrients is probably limited.

3.6 Microbial indicators

Certain soil and rhizosphere bacteria are especially important for key processes in soil geochemical cycles and can be considered microbial indicators. Non-target effects on these indicator bacteria are very important, as such effects could affect soil ecosystem functioning. Nodule formation on legumes, for instance, is an important parameter for nitrogen availability to certain plants and for nitrogen cycling in general. BCAs tested to date have shown contrasting effects on nodule formation: reduced formation in sterile microcosms, increased formation in non-sterile microcosms and no effect in the field. The natural heterogeneity in the field thus seems to be larger than any effects of the BCAs. Typically, nitrogen-fixing strains (such as Sinorhizobium, Rhizobium) are exposed to the BCA in microcosms and then cell densities, activity and nodulation followed. Sinorhizobium meliloti strains with different sensitivities to Plt, but insensitive to DAPG, grew to the same densities following treatment with CHA0 and CHA0/pME3424 in gnotobiotic microcosms. However, these densities were 10 times lower than in control treatments. Nodulation was reduced in the presence of CHA0 for the most Plt-sensitive strain [53]. In non-sterile pea microcosms, nodulation by Rhizobium leguminosarum R1112 was increased in the presence of F113, but unaffected by F113G22 [54]. At the same time, R1112 and F113 added separately to soil containing indigenous rhizobia stimulated nodulation, while F113G22 did not [55]. This increased nodulation was possibly due to increased root exudation caused by increased root surface or by an effect of DAPG on root permeability [54], however, no data are available to support this hypothesis. In contrast, root nodulation and activity were reduced in the presence of two potential P. fluorescens BCAs on bean roots [56]o other microorganisms present in soils [12–15] or on different crops [16–18]. These broad inhibitory capabilities of the antibiotics obviously call for the design of application procedures that limit non-target effects.

3.7 Protozoa and nematodes

Protozoa and nematodes are important microfauna in the soil ecosystem as they play key roles in recycling of organic matter. Some of these soil-dwelling organisms also have the potential to suppress plant pathogens, while others are plant pathogens. Hence, bacterial BCAs targeting pathogenic fungi could potentially also affect protozoa and nematodes. A variety of non-target studies of bacterial BCAs have been performed with these organisms and the potential effects are worth noting.

F113, as well as synthetic DAPG, is known to increase egg hatching of the root pathogenic nematode Globodera rostochiensis, but to reduce the mobility of the juveniles [61]. The reduced juvenile mobility suggests that survival conditions are suboptimal, while increased egg hatching would potentially increase nematode attack on plants. Contrasting effects upon the nitrogen uptake in pea and wheat shoots and roots after F113 and F113G22 application have been reported. In pea shoots and roots the uptake was increased while in wheat shoots it was decreased [62,63]. This discrepancy was probably caused by increased protozoa and nematode populations in the pea rhizosphere in the presence of F113 and F113G22. In wheat rhizosphere, these strains had no effect on the protozoa but did lead to an increased number of nematodes [24]. These results were tentatively explained by production of a nematicidal compound by the germinating pea and subsequent catabolism of this compound by the introduced pseudomonads, thus relieving the negative pressure on the protozoa and nematodes. Germinating wheat should not produce a nematicidal compound. Since this hypothesized nematicide has not been isolated and purified, the ability of F113 to metabolize it is still open to question. Further studies on the effects of F113 and F113G22 inoculation on nematode community structure in pea rhizosphere have shown that the number of bacteria-feeding nematodes increased, presumably as a result of the nematodes feeding on the introduced bacteria [64]. Fungal feeding and plant parasitic nematodes initially declined in all treatments but later increased, especially in the F113-treated microcosm. This was probably due to higher root biomass in the F113-treated plants.

Negative effects of CHA0 and CHA0/pME3424 on the ciliate Tetrahymena pyriformis[65], and the nematode Meloidogyne javanica[66] have been shown in vitro. Andersen [67] has shown negative effects of the BCAs P. chlororaphis MA342, and CHA0, DR54 and UW85 on the growth of the amoeba Hartmanella vermiformis compared to a control bacterium (Enterobacter aerogenes) in in vitro tests. When the same bacteria were tested on protozoa extracted from soil, only CHA0 and DR54 significantly inhibited the growth of protozoa. In mesocosm experiments with non-sterile soil, the addition of DR54 to bulk soil and rhizosphere soil resulted in a higher number of fast-growing protozoa throughout the 130 days of barley growth. In contrast, the total number of protozoa only increased in the bulk soil during the first 18 days, leading to a change in the relative distribution of fast-growing versus slow-growing protozoa as a result of DR54 addition. Using signature fatty acids, Gagliardi et al. [47] found significant effects of introduced P. chlororaphis 3732 on eukaryotes, but they were unable to assign the effects to any specific group of organisms (protozoa, nematode, fungus, or plant).

It thus seems evident that BCAs can affect soil protozoa and nematodes, but the differentiation between toxicological effects of the antimicrobial compound and feeding effects remains unclear. Whether such effects are able to significantly change the in situ population size and function of protozoa and nematodes over a prolonged time still needs to be shown.

3.8 Plants

During the development of effective BCAs, the ability to reduce disease incidence and increase plant yield are major criteria for success. In spite of this, reports of negative effects of BCAs or their antibiotics on plant growth do exist. For example, DAPG [16] and Plt [17] show phytotoxic effects. CHA0/pME3090 showed variable phytotoxic effects depending on plant species [17], while DR54 had a slight phytotoxic effect on barley growth in microcosm studies [18]. In contrast, Pf29A showed positive effects on plant growth [29] and SBW25, which is not known to produce antimicrobial substances, showed positive effects [28] or no effect [19]. Numerous reports on F113 tested in sandy soil pea microcosms without addition of fungal pathogens consistently show that root growth is increased, often leading to significantly reduced shoot/root ratio (e.g. [19,21,22,55]). The DAPG-deficient mutant F113G22 generally did not have this effect. Since increase in root growth can be a sign of stress as well as a sign of healthy plants, it may be difficult to distinguish between positive and negative effects of DAPG. For wheat, however, no such effect on the shoot/root ratio of F113 and F113G22 was found [63]. In the presence of added fungal pathogens, the BCAs F113, CHA0, CHA0/pME3090, DR54, Pf29A, SBW25, and Q2-87 reduced the abundance of the pathogens on the roots and in soil and partly restored plant growth [17,18,23,29]. Generally, BCAs tend to relieve the pathogen effect on the plant, though not always completely as the BCAs themselves can inhibit plant growth.

4 Discussion

The literature on non-target effects of bacterial BCAs that are intended to control root pathogenic fungi reveals a multitude of experimental designs with diverse results. Generally, non-target effects could be observed. These were often small and transient, but a few were shown to persist beyond the growth season. However, the majority of the non-target studies raise a number of concerns related to experimental design, especially with the controls used in the experiments. For environmental risk analysis appropriate controls are needed for evaluation of observed non-target effects of the BCAs. Such controls include tests of BCAs functioning in soil, target suppression, and production of important antimicrobial compounds. Furthermore, non-target effects should be compared to controls with microbial inocula without biocontrol activity, to agricultural practices including the chemical counterpart of the BCA, and to environmental variability. These requirements are not always met, and this has led us to review the parameters that should be considered when designing a study of non-target effects of bacterial BCAs.

4.1 BCA inoculum

In evaluating the risks of BCAs it is anticipated that the effects should result from biological action of the BCAs after inoculation and not from co-introduced metabolites accumulated in the growth media, although a few studies [37,49,68] in fact stress this possibility. However, the target and non-target studies reported for the BCAs reviewed here are generally not consistent in their inoculation procedures, and often this complicating factor is not considered in evaluation of test results. For example, at field scale TX-1 was administered in irrigation water along with the growth medium [45], while MA342 (the active agent in the product Cedomon, http://www.bioagri.se) was applied to seeds in high numbers in spent medium followed by drying of the seeds prior to sowing [68]. As the antifungal component (2,3-deepoxy-2,3-didehydrorhizoxin) [69] very likely is present in the liquid, drying will result in a high concentration of antifungal components on the seed surface. Inoculating washed MA342 cells indeed had a lower antifungal effect than inoculation of unwashed cells [68]. Similarly, all the F113 non-target studies performed in microcosms and the UW85 field experiments apparently used cells suspended directly from the agar substrates on which they were cultured, giving opportunities for co-introducing a variety of non-specific metabolites. In contrast, all non-target effect studies with CHA0 (except [23]) and DR54, as well as F113 non-target field studies, used washed cell suspensions of broth cultures. The reported non-target effects of F113 were primarily found in the microcosm studies, thus raising the possibility that the effects seen were due to co-addition of metabolites. The DAPG production of F113 is known to be higheither PCA or DAPG had no effect on abundance of either bacteria or fungi [38]. Other studies have shown effects on target fungi in rhizosphere soil, but without measurement of the in situ antibiotic production (e.g. [29,43,44]).

4.2 Antibiotic production in situ

The majority of the published literature on non-target effects of BCAs is based on the assumption that antibiotics produced in situ are the key to suppression of fungal pathogens, and correspondingly to the non-target effects. Indeed, through measurements of antibiotics in soil and rhizosphere, in situ production of antibiotics by various BCAs has been shown in a few studies (for review see [5,70]). For example, 12 days after introduction of the BCA WCS358r, PCA could be detected in the rhizosphere, suggesting the in situ production of this metabolite [37]. Viscosinamide has also been measured in the sugar beet rhizosphere in increasing amounts 5 and 7 days after sowing of DR54-inoculated seeds [18]. A significant positive correlation was found between inoculated Q2-87 and DAPG concentration in wheat rhizosphere, resulting in a constant DAPG production per population unit of 0.62 ng DAPG 10⁵ CFU⁻¹[71]. Bonsall et al. [72] further showed in situ DAPG production by Q2-87 of up to 2.4 μg g⁻¹ root plus rhizosphere soil. This is within the same order of magnitude as the concentration of 1 μg ml⁻¹ DAPG reported to be inhibitory to soil bacteria [13]. On the other hand, they saw no effects on the frequency of DAPG- or Plt-sensitive bacteria in microcosms inoculated with CHA0 or CHA0/pME3424, which indicates that the antibiotics had an insignificant impact on the soil bacteria in situ. This again may be due to physical factors such as a lack of contact between the antibiotics and the bacteria, or adhesion of antibiotics to inert material. Additionally, in situ production and function of the antibiotics are sensitive to abiotic and biotic factors [5,73]. Hence, great care has to be taken to ensure that in situ antibiotic production occurs, when results are explained by antibiotic effects. This is especially the case when the pathogen is not present and any target effect by the BCA thus is immeasurable.

The use of antibiotic-over- or non-producing BCA mutant strains is a valuable technique in studies of the possible role of antibiotic production in non-target effect studies [5,70]. However, it may only give indirect evidence of the role of antibiotics in non-target effects. For example, comparative studies using the overproducers F113(pCUGP) and CHA0/pME3424 and the respective wild-types concluded that no effects could be attributed to production of antibiotics, while the non-producer F113G22 generally showed different effects than did the wild-type. However, in a single case where a significant effect was seen, pleiotrophic effects of the CHA0/pME3424 derivative could not be excluded [26]. The overproducers F113(pCUGP) and CHA0/pME3424 were both constructed by insertion of a plasmid carrying the genes for antibiotic production [7,8]. CHA0/pME3424 further carries an extra copy of the rpoD gene, a sigma factor essential for regulation of primary metabolism. Also, the potential BCA WCS358 was modified to produce either PCA or DAPG by insertion of the respective genes in mini-Tn5 transposons into the chromosome [33,37,38]. The non-producer F113G22 was constructed by Tn5 mutagenesis and, apart from the lack of DAPG production and inhibition of test fungi, F113G22 showed lower production of a brown pigment than F113 [9]. Thus, one should be aware that other factors might be affected by the genetic modifications thereby affecting cell physiology involved in target suppression [7], as well as non-target effects [26]. This especially applies to genetic modifications of regulatory genes as for the CHA0 mutants [7].

4.3 Target efficiency of BCA

The focus on the role of the antibiotics in most non-target studies could lead to overlooking other parameters that may be equally important for biocontrol in natural soils. The majority of the tested BCAs produce additional compounds with antifungal activities, such as enzymes, siderophores and cyanide (Table 1), which may enhance target effects. Survival characteristics of the BCA strains may also be important. This is particularly evident in the light of the improved BCA effectiveness of P. fluorescens Q8r1-96, which has a comparable DAPG production capability but superior survival and root-colonizing capabilities compared to other potential BCAs [74]. The BCA effectiveness of this strain is hypothesized to be partly due to its utilization of different carbon sources, such as trehalose and benzoate, which are known to attract specific bacteria and believed to act as inducers of antagonism in the rhizosphere. Also, the effective SBW25 is not known to produce any antimicrobial compounds but still demonstrates some transient non-target effects. The BCA effectiveness is most certainly dependent on the likelihood of the BCA encountering the pathogen. This is, of course, affected by a variety of physiological conditions and the microscale location of the BCA in the rhizosphere.

4.4 Presence of pathogen

The use of pathogen-free soils as the medium for non-target studies is often argued to be optimal because it avoids the responses of the plant (e.g. exudation), which will mask non-target effects [34]. In agricultural practices, addition of the BCA to soil in situations without disease outbreaks is certainly a possibility. However, only a few studies [23,29] have included addition of the target pathogen as a positive control in non-target effect studies. The studies found significant differences between treatments with and without added pathogen, using plant growth and bacterial abundance as indicators of effects. The majority of non-target effect studies, however, lack the positive control that demonstrates significant target-suppressive effects of the BCA. Thus, lack of a significant impact of the BCAs in many of the non-target studies may be due to target ineffectiveness of the BCAs under the experimental conditions. Hence, we believe that the presence of the BCA together with a pathogen imposing a disease pressure is essential to verify effective target inhibition by the BCA under the same experimental conditions as the non-target effect study. This includes verifying the survival of BCA and production of any antimicrobial compounds. In this respect, the pathogen can be naturally present in infested soil or be inoculated together with the BCA. The pathogen will, of course, potentially exert detrimental effects on the plant, probably resulting in increased root exudation. Further, adding the pathogen results in the addition of two microorganisms instead of just the BCA and can complicate the system by further interactions with the indigenous microorganisms. Thus, including the pathogen will enlarge and change the non-target effect studies in several ways, requiring additional treatments and controls to be considered. But this must be considered necessary to obtain useful results for risk analyses. Even though the pathogen is not added or known to be present in the soil, it may occur, and give rise to reduced seed emergence or lesions on plant roots. In our opinion, some measure of fungal attack and target inhibition should be included in future non-target studies to be able to verify the target and non-target effects of the BCA.

4.5 Experimental size

The differences between in vitro, microcosm, and field studies are substantial (scale, duration, cultural conditions, inoculum preparation, indigenous soil microflora, etc.) and these differences may affect the results of non-target studies. The majority of non-target effect studies have used microcosms, while field studies are considerably fewer (F113 [25,36,51,52], TX-1 [45], 06909 [46], WCS58 [33,37,38], UW85 [39,49]). The major difference between microcosm and field studies is the higher level of experimental control in microcosm experiments. As a result, potential effects are easier to detect in well-defined microcosm experiments, thus providing a higher probability of explaining the interactions and mechanisms of BCAs. Experiences from microcosms should ideally be transferable to field situations. However, effects found in microcosm experiments may be of a much smaller scale than the oscillations commonly observed to result from agricultural practices and natural variations. For example, Glandorf et al. [37] found no impact of changed community structure on various soil functions in situ.

A striking difference between the published microcosm and field experiments that could have been avoided is the choice of plant species tested. Plants like pea, cucumber, and tobacco are usually tested in the microcosms while plants like wheat and sugar beet are generally tested in the field. This complicates extrapolation of microcosm-based knowledge to the field.

4.6 Time scale

The environmental concern related to non-target effects of BCAs will obviously depend on the time needed for the ecosystem to recover. Even severe non-target effects can be accepted for short time periods (days–weeks), if disturbances or non-target effects decrease with time [3]. Soil type, agricultural practices, season and plant species heavily influence the size and functions of microbial communities [75–77], and this needs to be considered in interpreting non-target effects. Non-target effects of BCA inoculation are usually confined to one growth season, when effects would likely be smaller than effects caused by plant aging or agricultural practices. An exception is the non-target effect studies of WCS358r in wheat fields [33,37,38], which showed yearly and seasonal fluctuations. From the point of environmental risks, large effects could be tolerable within the growth season, while only small non-target effects would be tolerable if the effect extended beyond a growth season. However, the chances of detecting effects are higher shortly after introduction of the BCA. Observing non-target effects within the first 3 weeks after inoculation thus merits looking for non-target effects over longer periods. If non-target effects are not detectable in the short term, then long-term non-target effects are clearly even less likely. In the typical non-target effect field study, samples are taken several times during the growth season, especially within the first 3 weeks, which opens the opportunity to study successional responses during plant growth in addition to non-target effects.

5 Concluding remarks

In spite of the considerations pointed out about the information and validity of the reported experiments to test for non-target effects, we conclude that effects on non-target members of the food web in soil and rhizosphere have been identified. Though non-target effects are often small and transient, effects extending beyond the growth season have also been observed. Whether the non-target effects are caused by introduced antimicrobial compounds or by other activities of the BCAs is, however, not clearly distinguished. Thus, it is relevant to perform non-target studies using biocontrol strains showing target effects that can be explicitly explained by activity of the inoculum in situ. Direct introduction of antimicrobial compounds could be considered as an alternative fungicide treatment and would have the advantage of limited long-term effects due to turnover of the antimicrobial compound. However, apart from the production of antimicrobial compounds, the majority of BCAs have important additional features, e.g. siderophores or cellulolytic enzymes, that in combination and in situ might be more effective than the compound alone [6].

Only limited data indicate that significant disease suppression can result from addition of those BCAs that have been subjected to comprehensive non-target studies (e.g. F113 [23], CHA0 [78,79], and DR54 [18]). On the other hand, non-target studies of marketed BCAs (e.g. MA342 [68], Streptomyces griseoviridis K61 [80]) are absent or are very limited in number. Clearly, more non-target effect studies are needed on the marketed BCAs with less emphasis on the candidates without a high and well-documented biocontrol efficacy.

Acknowledgements

Drs. Christoph Keel and Niels Bohse Hendriksen are acknowledged for critical comments on earlier versions of the manuscript. A.W. and S.J.B. were partly financed by the ‘Center for Effects and Risks of Biotechnology in Agriculture’ under the Danish Environmental Research Program (SMP2).

References