Various rhizosphere bacteria are potential (micro)biological pesticides which are able to protect plants against diseases and improve plant yield. Knowledge of the molecular mechanisms that govern these beneficial plant–microbe interactions enables optimization, enhancement and identification of potential synergistic effects in plant protection. The production of antifungal metabolites, induction of systemic resistance, and the ability to compete efficiently with other resident rhizobacteria are considered to be important prerequisites for the optimal performance of biocontrol agents. Intriguing aspects in the molecular mechanisms of these processes have been discovered recently. Phenazines and phloroglucinols are major determinants of biological control of soilborne plant pathogens by various strains of fluorescent Pseudomonas spp. This review focuses on the current state of knowledge on biocontrol by phenazine-producing Pseudomonas strains and the action, biosynthesis, and regulation mechanisms of the production of microbial phenazines.
A considerable number of Pseudomonas biocontrol strains with antifungal activity produce phenazines (Thomashow et al., 1990; Becker et al., 1990; Gealy et al., 1996; Anjaiah et al., 1998; Chin-A-Woeng et al., 1998; Powell et al., 2000; Tambong & Hofte, 2001). The action of phenazines has been studied extensively over the past three decades. Several strains have already been marketed as commercial biocontrol products, such as Cedomon (BioAgri AB, Uppsala, Sweden), a seed treatment based on a phenazine-producing Pseudomonas chlororaphis strain providing protection against seed-borne diseases in barley and oats. There is much interest in the development of new biocontrol agents to extend the area of application and the range of target pathogens to be controlled.
Biocontrol microbes, either present naturally or introduced as an inoculant (e.g. by application onto the seed) interact with the surrounding organisms. The biocontrol agent must proliferate on the appropriate plant root surface and inhibit – either directly or indirectly – growth of the target pathogen to reduce or eliminate infection by the pathogen (Scher & Baker, 1980). The interaction may be a direct interaction between bacterium and pathogen or mediated through the host plant such as in case of induced systemic resistance (ISR) (van Loon et al., 1998). These two factors outlined in this paper are generally considered to be of crucial importance for biocontrol strains.
Bacterial root colonization and its role in biocontrol
Inoculant bacteria that are introduced into the soil must compete with the often better-adapted resident microflora. Microbial proliferation associated with the root is generally referred to as root colonization (Dekkers, 1997). The population of many inoculant strains declines progressively in time after introduction (Bashan, 1998).
When bacteria are applied in relatively high numbers to seeds before planting, the applied biocontrol agent can cause inhibition of soil fungi only for a short period of time, similar to the effect of a chemical pesticide. Bacterial inoculants become more powerful agents than chemical ones when they multiply on the root and colonize the root system to achieve a desired effect for longer periods of time. The ability to establish itself for several months at a high level in the rhizosphere is therefore considered to be an important factor for the success of applications for beneficial purposes and suppression of plant diseases. The analysis and understanding of traits and genes involved in rhizosphere colonization and interactions is therefore important for insight into the molecular mechanisms of biocontrol (Suslow, 1982; Lugtenberg et al., 1991; Bloemberg & Lugtenberg, 2001). Root colonization not only results in high population densities on the root system, it also functions as the delivery system of antifungal metabolites (AFMs) such as phenazines along the whole root.
Not all isolates showing pathogen inhibition in vitro provide disease control or are rhizosphere competent. Biocontrol efficacy of pseudomonads is often related to their density in the rhizosphere. The bacterial population size was indicated to be the limiting step for biocontrol (Weller, 1988). In several field trials the ability of the introduced biocontrol agent to establish itself effectively in the rhizosphere was essential for biocontrol (Schippers et al., 1987; Weller, 1988; Bull et al., 1991; Raaijmakers et al., 1995a; Raaijmakers et al., 1999). The degree of colonization ability required for a certain application may also be dependent on the mechanism by which a biocontrol agent performs its action. For a strain which acts via antibiosis one may assume that proper colonization is needed to deliver antifungal compounds along the entire root system, whereas for a strain which acts through ISR a smaller number of bacteria during a restricted period of time may be sufficient to elicit a successful response in the host plant.
Bacterial traits important for efficient colonization
Various bacterial traits contribute to the ability of a bacterial strain to colonize the rhizosphere and loss of such a trait can reduce the ability to establish itself effectively in the rhizosphere and therefore reduce its beneficial effects. Many of these traits involve the efficient use and uptake of nutrients from the rhizosphere. Important elements for the uptake of nutrients are chemotaxis and motility. A cheA− chemotaxis mutant of Pseudomonas fluorescens WCS365 is strongly reduced in competitive root colonization under gnotobiotic conditions as well as in potting soil (Weert de et al., 2002). When tested in direct competition with the wild-type, flagella-less derivatives of the good colonizing strain P. fluorescens WCS365 are severely impaired in colonization of the root tip of potato (de Weger et al., 1987; Ward, 1989) and tomato (Simons et al., 1996; Dekkers et al., 1998a; Dekkers et al., 1998b). Nonmotile Tn5 transposon mutants of the Fusarium oxysporum f. sp. radicis-lycopersici antagonistic biocontrol strain Pseudomonas chlororaphis PCL1391, producing phenazine-1-carboxamide (PCN; also known as chlororaphin) as the active metabolite, were at least 1000-fold impaired in competitive tomato root tip colonization compared with their wild type. Consequently, the colonization-impaired mutant lost its ability to suppress disease, while still being able to produce the same amounts of antifungal activity as the wild type (Chin-A-Woeng et al., 2000). Other investigators report no relationship between motility and root colonization (Howie et al., 1987; Scher et al., 1988; Moens & Vanderleyden, 1996). These unequivocal findings may be due to differences in bacterial strains, host plant, root sampling, or other experimental conditions.
The P. fluorescens strains WCS365 and WCS374 and P. putida strain WCS358 with a mutation in the O-antigen side-chain of lipopolysaccharide (LPS) are impaired in colonization (de Weger et al., 1989; Simons et al., 1996; Dekkers et al., 1998b). The colonization defect in LPS-defective strains can be explained by assuming that for optimal functioning of nutrient uptake systems an intact outer membrane is required.
Agglutination and attachment of Pseudomonas cells to plant roots via adhesins, fimbriae or pili, cell surface proteins, and polysaccharides are likely to be of importance for colonization. The number of fimbriae on bacterial cells of the biocontrol agent P. fluorescens WCS365 correlates with the degree of attachment to tomato roots (M. M. Camacho Carvajal et al., unpublished). The outer membrane protein OprF of P. fluorescens OE28.3 is involved in attachment to wheat roots (DeMot et al., 1991). A root-surface glycoprotein agglutinin was shown to mediate agglutination of Pseudomonas putida isolate Corvallis (Buell & Anderson, 1993) but had no role in long-term colonization (Glandorf et al., 1994; Knudsen et al., 1997).
Specific genes for the biosynthesis of amino acids and vitamin B1 and for utilization of root exudate components such as organic acids are important for colonization of P. fluorescens WCS365 on tomato roots (Simons et al., 1997; A. H. Wijfjes et al. unpublished). Phenylalanine, methionine, and tryptophan auxotrophic mutants of P. chlororaphis PCL1391 were at least 1000-fold less present on the tomato root tip when tested in competition with the wild type. Although the auxotrophic mutants produced the same amounts of the antifungal compound phenazine-1-carboxamide (PCN), they had lost their biocontrol abilities against F. oxysporum f. sp. radicis-lycopersici (Chin-A-Woeng et al., 2000).
Additional genes essential for competitive colonization are the nuo4 gene, which is part of a 14-gene operon encoding NADH dehydrogenase NDH-1. Pseudomonas fluorescens WCS365 possesses two NADH dehydrogenases, and apparently the absence of NDH-1 cannot be adequately compensated for by the other NADH dehydrogenase under rhizosphere conditions, resulting in suboptimal growth of these mutants on the root (Camacho Carvajal et al., 2002).
Mutants of the biocontrol strain P. fluorescens C7R12 affected in synthesis of either nitrate reductase, pyoverdine, or both, show a lower competitiveness than the wild-type strain in soil, indicating that both nitrate reductase and pyoverdine are involved in their fitness. The mutant mutated in both nitrate reductase and pyoverdine synthesis displayed the lowest competitiveness. The selective advantage given by nitrate reductase was more apparent under conditions of lower aeration (Mirleau et al., 2000; Mirleau et al., 2001).
A two-component regulatory system consisting of the colS and colR genes that have homology to sensor kinases and response regulators (Dekkers et al., 1998a) was also shown to be involved in efficient root colonization of P. fluorescens WCS365. It was concluded that an environmental stimulus is important for colonization but the nature of the signal as well as of the target genes is still to be elucidated (Dekkers et al., 1998b).
The Sss site-specific recombinase of the λ integrase gene family, to which XerC and XerD belong also, is necessary for adequate root colonization (Dekkers et al., 1998c). It was postulated that a certain bacterial subpopulation is important for competitive colonization of strain WCS365. Introduction of the sss gene into wild-type strains P. fluorescens strains F113 and WCS307 resulted into an enhanced colonization ability (Dekkers, 1997). Mutation of an sss/xerD homologue in P. chlororaphis PCL1391 also resulted in a tenfold impaired root tip colonization, both in a sterile sand system as well as in potting soil (Chin-A-Woeng et al., 2000). Although the mutant was not altered in the level of PCN production, biocontrol was abolished in an F. oxysporum f. sp. radicis-lycopersici–tomato biocontrol assay, apparently as a result of lower colonization levels (Chin-A-Woeng et al., 2000).
The production of secondary metabolites is often hypothesized to confer a selective advantage in the persistence in soil and the rhizosphere. The contribution to the ecological competence of strains was indeed shown for the phenazine-producing strains P. fluorescens 2–79 and P. aureofaciens strain 30–84 using Tn5 mutants impaired in phenazine biosynthesis. Phenazine-deficient strains have a reduced survival and a diminished ability to compete with the resident microflora (Mazzola et al., 1992). However, production of the polyketide antibiotic 2,4-diacetylphloroglucinol in P. fluorescens strain F113 did not influence the persistence of this strain in soil (Caroll et al., 1995).
Root-colonizing ability is not only important in biocontrol, but can also be predicted to be essential in other applications of microbial inoculants of seeds, such as for biofertilization, phytostimulation and phytoremediation, in which the effectiveness of the inoculant is likely to be dependent on the establishment of a minimum population size. It is also worthy of mention that many colonization traits important for biocontrol of plant diseases also play a role in colonization of animal tissues by bacteria (Drake & Montie, 1988; Weiser et al., 1990; Rijpkema et al., 1992; Chiang & Mekalanos, 1998; Lugtenberg & Dekkers, 1999).
Phenazine derivatives not only kill fungi but the phenazine derivative pyocyanin, produced by certain Pseudomonas aeruginosa strains, is also involved in killing of animal cells and tissues (Mahajan et al., 1999). The two established biocontrol traits of P. chlororaphis PCL1391, colonization and production of phenazine antibiotics, are bacterial traits that also play a role in colonization of animal tissues and in killing of animal cells, respectively. Therefore, we suggest that these bacterial traits are important in interactions with widely different eukaryotes and must have been developed early in evolution.
Antagonism against phytopathogens
The strategies through which biocontrol agents can antagonize soil-borne pathogens are generally divided into four categories (Thomashow & Weller, 1995): (1) competition for niches and nutrients (niche exclusion), (2) predation, (3) antibiosis and (4) induction of a plant defence response (ISR).
Competition and predation/parasitism
Soil microorganisms interact directly with the host plant via the root surface on which they proliferate. For growth, they are highly dependent upon nutrients present in the rhizosphere or provided at the plant root surface in the form of root exudates. One may expect heavy competition for nutrients between the biocontrol agent and the indigenous microflora in the rhizosphere of the host plant.
Already established (pre-emptive competitive exclusion) or aggressively colonizing biocontrol bacteria can therefore prevent the establishment and subsequent deleterious effects of a pathogen. The ability of PGPRs to establish themselves in niches or rapidly compete in the use of nutrients that are shared by the pathogen is thought to be a general mechanism for antagonistic activity. In this way, biocontrol bacteria have the function of a probiotic. Competition for nutrients such as carbon, nitrogen or iron is one of the mechanisms through which biocontrol strains can reduce the ability of fungal pathogens to propagate in the soil (Alabouvette, 1986; Buyer & Leong, 1986; Leong, 1986; Loper & Buyer, 1991; Fernando et al., 1996; Handelsman & Stabb, 1996). Also, the speed with, and extent to which, biocontrol bacteria can aggressively colonize the root system is assumed to be an important trait. Many Pseudomonas strains have a short generation time. Pseudomonas fluorescens WCS365 forms microcolonies on the tomato root (Chin et al., 1997; Bloemberg et al., 1997) already one day after seed inoculation (de Weert et al., 2002).
Bacterial colonization of the root surface is limited to a small surface area of the totally available surface, and probably corresponds to those places which are most nutrient-rich, particularly the intracellular junctions between root epidermal cells (Rovira, 1956; Bowen & Rovira, 1976; Bloemberg et al., 1997; Chin-A-Woeng et al., 1997). Failure of a pathogen to compete effectively with the biocontrol strain and use the available nutrient sources at those shared niches will disable the pathogen's propagation. A classical example of niche exclusion is the control of leaf frost injury caused by Pseudomonas syringae, which has an ice nucleation protein on its cell surface. Application of an ice-nucleation-minus mutant prevents damage caused by the establishment of the pathogenic wild type (Lindow et al., 1983; Lindow, 1983a; Lindow, 1983b).
Another well-known example of competition for nutrients is limitation by iron. Iron is an essential cofactor for growth in all organisms, but the availability of solubilized Fe3+ in soils is limiting at neutral and alkaline pH, leading to Fe3+ limitation. Most organisms, including fluorescent Pseudomonas species, take up ferric iron ions through high-affinity iron chelators, designated as siderophores, that are released from bacterial cells under Fe3+ limiting conditions. The complexed iron is taken up via specialized cell surface-located uptake systems and thus provide a route for iron uptake. The ability to produce efficient siderophores is sometimes combined with the ability to take up related siderophores from other organisms (Koster et al., 1993; Koster et al., 1995; Raaijmakers et al., 1995b; Weisbeek & Gerrits, 1999). The ability to scavenge iron under Fe3+ limitation provides the biocontrol organism with a selective advantage over pathogens or deleterious organisms that possess less efficient iron binding and uptake systems. Siderophore-deficient mutants were found to be less suppressive to pathogens than the isogenic parental strain (Bakker et al., 1986).
Microscopy analysis of a root surface colonized by Pseudomonas bacteria, including the phenazine-producing strain PCL1391, shows that they mainly colonize the root in microcolonies at the junctions of epidermal cells (Bowen & Rovira, 1976; Chin-A-Woeng et al., 1997). The fungal pathogen F. oxysporum f. sp. radicis-lycopersici colonizes the tomato root surface at the same locations, as was observed using an autofluorescent protein-marked fungus (Lugtenberg et al., 1999b; Lagopodi et al., 2002). The supposed overlap of ecological niches of biocontrol agent and deleterious microorganism presumably leads to competition for these restricted sites, as well as for nutrients present at those sites, leading to pathogen inhibition.
Plants possess inducible defence mechanisms to restrict or block the ability of microbial pathogens to produce disease. A network of interconnecting physical and chemical signalling pathways regulates these defences, resulting in partial or complete resistance to subsequent virulent pathogens. Characteristic of these induced resistances are their remote action, long-lasting resistance, host specificity, and protection against a broad range of pathogens.
Induced systemic resistance can also be activated by biotic agents such as lipopolysaccharides, siderophores, and flagella (Bakker & Schippers, 1987). The production of the phenazine derivative pyocyanin was recently shown to be involved in induction of ISR in tomato and bean against Botrytis cinerea (Leeman et al., 1995b; Audenaert et al., 2001). By contrast to its wild type, a salicylic acid or pyocyanin mutant of P. aeruginosa 7NSK2 is unable to induce resistance against Botrytis cinerea (Audenaert et al., 2001). It was hypothesized that the pyochelin precursor salicylic acid, produced in nanogram amounts on roots, is converted to the siderophore pyochelin and that pyochelin and pyocyanin act in concert to produce active oxygen species that cause cell damage and this, subsequently, leads to induced resistance (Audenaert et al., 2001).
Role of phenazine derivatives in biocontrol Phenazines encompass a large family of heterocyclic nitrogen-containing brightly coloured pigments with broad-spectrum antibiotic activity. They are synthesized almost exclusively by bacteria. Their production has been reported for Pseudomonas, Streptomyces, Nocardia, Sorangium, Brevibacterium and Burkholderia species (Turner & Messenger, 1986). Although phenazines have been known for their antifungal properties for a long time, the interest in phenazines increased with the resurgence of research in biocontrol in the 1980s, driven in part by trends in agriculture toward greater sustainability and an increased concern about the use of chemical pesticides.
Only a limited number of phenazine derivatives have been evaluated in biocontrol. Mutants in PCA production of P. fluorescens 2–79 provided significantly less control of take-all than the wild type on wheat seedlings (Thomashow & Weller, 1988). A phzB or phzH mutant of P. chlororaphis strain PCL1391, impaired in PCN biosynthesis, did not suppress disease formation in a tomato-foot and root rot system (Chin-A-Woeng et al., 1998).
The mechanisms for the action of phenazines in antifungal interactions are poorly understood. It is assumed that they diffuse across or insert into the membrane and act as an reducing agent, resulting in the uncoupling of oxidative phosphorylation and the generation of toxic intracellular superoxide radicals and hydrogen peroxide which are harmful to the organism (Turner & Messenger, 1986; Hassett et al., 1992; Mahajan et al., 1999). The higher activity of superoxide dismutases of the pyocyanin-producing P. aeruginosa, which would provide protection against the action of phenazines supports this idea (Hassett et al., 1992; Hassett et al., 1995).
Phenazines are toxic to a wide range of organisms including bacteria, fungi, and algae (Toohey et al., 1965). The conditions at which phenazine derivatives are active can be diverse (Turner & Messenger, 1986). The PCA-producing strain P. fluorescens strains 2–79 (Thomashow & Weller, 1988) and P. aureofaciens strain 30–84 (Pierson III et al., 1995) showed no significant biocontrol activity in a F. oxysporum f. sp. radicis-lycopersici assay, whereas strain PCL1391, which produces the closely related PCN, inhibits this pathogen under identical conditions. Comparison of antifungal activity of PCN and PCA in vitro showed that the antifungal activity of PCN was at least 10 times greater at pH 5.7 or higher. Since this is probably related to the concentration of protonated PCA, this suggests that when the pH of the soil or rhizosphere increases, the anionic form of PCA will predominate (Brisbane et al., 1987; Chin-A-Woeng et al., 1998). Other differences in physical properties may influence the activities in vivo (e.g. hydroxylated phenazines are more soluble in water than their unhydroxylated counterparts).
Phenazine biosynthesis Many strains produce more than one phenazine derivative. Pseudomonas aeruginosa strains can produce pyocyanin, PCA and PCN (Chang & Blackwood, 1969). Shikimic acid acts as a precursor for phenazine biosynthesis (Ingledew & Campbell, 1969) (Fig. 1). The branch point from the aromatic biosynthetic pathway was identified at chorismate (Calhoun et al., 1972), with 2-amino-2-deoxyisochorismic acid (ADIC) as the branchpoint compound of PCA formation (Longley et al., 1972; McDonald et al., 2001); ADIC is converted to trans-2, 3-dihydro-3-hydroxyanthranillic acid (DHHA). Ring assembly by dimerization of two DHHA moieties results in the first phenazine derivative PCA. Although the mechanism of dimerization is unknown, it is speculated that the dimerization involves oxidation of two molecules of DHHA to the C-3 ketone. The molecules then react with each other by nucleophilic addition, dehydration, and tautomerization to give 5,10-dihydroanthranillic acid, which is oxidized to PCA (McDonald et al., 2001). Phenazine-1,6-dicarboxylic acid, which was previously proposed to be one of the first phenazine intermediates produced in the biosynthetic pathway (Hollstein & McCamey, 1973), was now excluded to be involved in the formation of PCA from ADIC based on feeding experiments with cell-free extracts containing phenazine biosynthetic gene products (Byng & Turner, 1976; Byng & Turner, 1977; McDonald et al., 2001).
The biosynthetic genes for production of phenazine derivatives have been identified in a number of Pseudomonas species. Pierson et al., (1995) identified a cluster of five genes, phzFABCD, required for phenazine-1-carboxylic acid production in P. aureofaciens strain 30–84. Based on homologies with enzymes in Escherichia coli, functions were assigned to the enzyme products in the basal phenazine biosynthetic pathway. The phzC gene product is homologous with 3-deoxy-d-arabinoheptulosonate-7-phosphate (DAHP) synthases and is thought to shunt precursors to the synthesis of shikimic acid, thereby bypassing the action of household DAHP synthetases. phzD and phzE are homologous to 2,3-dehydro-2,3-dihydroxybenzoate synthetases (isochorismatase) and anthranilate synthetase trpE, respectively. Chorismic acid is converted to ADIC by PhzE (Fig. 1). Subsequently, ADIC is thought to be converted to DHHA by PhzD (Pierson & Thomashow, 1992; Mavrodi et al., 1998; McDonald et al., 2001). PhzF and PhzG are thought to play a role in the conversion of DHHA to PCA via the intermediate 2,3-dihydro-3-oxo-anthranillic acid (McDonald et al., 2001).
The phenazine biosynthetic operons of P. fluorescens 2–79 and P. chlororaphis PCL1391 contain the phzABCDEFG genes. The nucleotide sequences of P. fluorescens 2–79 (Mavrodi et al., 1998), P. aureofaciens 30–84 (Pierson III et al., 1995), P. chlororaphis PCL1391, and P. aeruginosa PAO1 (Stover et al., 2000) are very homologous (70–95% identity). In P. aeruginosa PAO1, two complete sets of phzABCDEFG genes that are 99% homologous have been identified (Stover et al., 2000).
The polypeptides encoded by phzA and phzB can be found in the biosynthetic clusters of all analysed phenazine-producing organisms (Mavrodi et al., 1998; Chin-A-Woeng et al., 2001a). Homologues of phzA and phzB have also been identified in strain 30–84, where they are designated phzX and phzY, respectively (Pierson et al., 1995). Both genes encode 163-amino acid proteins, are nonessential for the biosynthesis and are involved in conversions past DHHA, possibly by stabilizing the PhzF protein. The complete set of genes is now considered to be the core genes for the biosynthesis of phenazines.
In some strains, PCA can be modified to other derivatives by specific modifying enzymes (Byng & Turner, 1977; Messenger & Turner, 1983). Downstream of phzG, the biosynthetic gene phzH was identified and appeared to be required for the presence of the 1-carboxamide group of PCN in P. chlororaphis PCL1391 (Chin-A-Woeng et al., 2001a) (Fig. 1b). PhzH belongs to the family of asparagine synthetases. The N-terminal domain of PhzH has a motif that is conserved in class II glutamine amidotransferases. In addition, the catalytic cysteine (Cys1) characteristic for the class II glutamine amidotransferase domain is also present in PhzH (Massiere & Badet, 1998). The C-terminal domain of PhzH harbours motifs characteristic for asparagine synthetases indicating that the conversion of PCA to PCN occurs via a transamidase reaction catalysed by PhzH. Transfer of the phzH gene, encoding the biosynthesis of PCN under control of the tac promoter to the PCA-producing biocontrol strains P. fluorescens 2–79 (Thomashow & Weller, 1988) and P. aureofaciens 30–84 (Pierson et al., 1995) enabled these strains to produce PCN and to suppress tomato foot and root rot more efficiently (Chin-A-Woeng et al., 2001a).
The phzO gene is localized immediately downstream of the core biosynthetic operon phzXYFABCD of P. aureofaciens 30–84. PhzO directs the synthesis of 2-hydroxyphenazine carboxylic acid (Delaney et al., 2001) (Fig. 1b). The protein belongs to the family of two-component nonhaem, flavin-diffusible bacterial aromatic monooxygenases. The subsequent conversion from 2-hydroxyphenazine carboxylic acid to 2-hydroxyphenazine appears to occur spontaneously in the absence of enzymes. The phzO gene is conserved in isolates of P. aureofaciens (which all produce hydroxyphenazines), indicating that the acquisition of this phenazine-modifying gene, and possibly also of phzH, is a fairly recent evolutionary event (Delaney et al., 2001).
In P. aeruginosa PAO1, in addition to phzH, two other phenazine-modifying enzymes, phzM and phzS, were characterized. The phzM gene is located upstream of the phzA1B1C1D1E1F1G1 operon and transcribed divergently. It encodes a 334 amino acids protein that most closely resembles O-demethylpuromycin-O-methyltransferases with a methyltransferase motif and a S-adenosyl-l-methionine (SAM)-binding domain (Mavrodi et al., 2001; Pattery et al., 2001). Functional analysis revealed that the phzM gene product is involved in the production of pyocyanin (Fig. 1b). The phzS gene is located downstream from phzG1 and encodes a 402-residue protein similar to bacterial monooxygenases. PhzS is thought to be involved in pyocyanin production as well as in the biosynthesis of 1-hydroxyphenazine in P. aeruginosa PAO1 (Mavrodi et al., 2001) (Fig. 1B).
Regulation of phenazine production in Pseudomonas
Many of the antifungal factors produced by biocontrol organisms are secondary metabolites (Gutterson, 1990). The production of secondary metabolites is often not only dependent on intracellular factors but also on environmental conditions. One mechanism that bacteria have adopted to regulate the production of antifungal factors is population density-dependent gene expression or quorum sensing. A quorum sensing system was described in the early 1970s in the bioluminescent marine symbiont Photobacterium fisheri (Nealson et al., 1970) and is often used as a model system for quorum system (Eberhard, 1972).
Quorum sensing regulation processes are often interlinked with regulation of primary metabolism, global regulatory systems such as the GacS–GacA system, and stress response (e.g. sigma factors). Quorum sensing seems to be the most important regulation mechanism for phenazine-1-carboxylic acid production in P. aeruginosa (Pierson et al., 1994) and phenazine-1-carboxamide production in P. chlororaphis (Chin-A-Woeng et al., 2001b). In these strains, quorum sensing is also dependent upon GacS and GacA.
Population density dependent regulation
Bacteria can perceive extracellular conditions using numerous sensor/kinase signal transduction pathways (Gross et al., 1989). Quorum sensing enables bacteria to regulate their gene expression in a population density-dependent way and adjust their physiology according to their environmental conditions and coordinate the behaviour of the entire cell population (Fig. 2). Autoinducer signal molecules convey population density information from neighbouring sister cells into the cell. These autoinducer molecules leave the cell but can diffuse back into the bacterial cell.
The signal molecule is believed to interact with and activate a transcriptional activator (Hanzelka & Greenberg, 1995; Stevens & Greenberg, 1997) when the intracellular signal reaches a certain threshold concentration as a result of a high population density. The accumulation of the signal may be enhanced in a closed system or confined space such as in a biofilm. In Gram-negative bacteria the most intensively investigated quorum sensing signal molecules belong to the class of N-acyl-l-homoserine lactones (N-AHLs) that modulate a diverse array of phenotypes ranging from bioluminescence to virulence and secondary metabolite production. The genes involved belong to the luxI–luxR family of response regulators (Fuqua et al., 1994). They consist of an autoinducer synthase, encoded by a luxI homologue, and a transcription activator, encoded by a luxR homologue.
Under certain circumstances, collectively acting bacteria will be more efficient than individually acting cells such as in a collective attack on other organisms, the production of a high concentration of metabolites or survival by generation of distinct cell types with different abilities to adjust to environmental changes (Shapiro, 1998).
A culture of a phenazine-producing Pseudomonas spp. strain grown under laboratory conditions shows a typical cell density-dependent production of phenazines. A lag of phenazine production is observed during early and mid-exponential growth, whereas large amounts of phenazines are produced in the late exponential growth and early stationary phase. The quorum sensing system regulating phenazine biosynthesis requires the PhzI and PhzR proteins. The phzI gene encodes an N-AHL-synthase that produces diffusible signal molecules derived from homoserine lactone (Fuqua et al., 1994; Salmond et al., 1995; Salmond et al., 1995; Ulitzur & Dunlap, 1995; Wood & Pierson, 1996; Hanzelka et al., 1997; Pierson et al., 1998; Chin-A-Woeng et al., 2001b), the response to which is mediated by PhzR, a transcriptional regulator with an N-AHL-binding and a DNA-binding domain.
The autoinducer molecules that accumulate in the culture medium during growth are different in phenazine-producing P. aureofaciens 30–84, P. fluorescens 2–79, and P. chlororaphis PCL1391 strains, despite the high homology observed in their phzI genes. The N-AHLs autoinducer molecules produced by a diverse range of Gram-negative bacteria differ in acyl chain length, ranging from C4 to C14, and the nature of the substituent on the carbon-3 position of the acyl chains. Pseudomonas aureofaciens 30–84 produces N-butanoyl homoserine lactone (C4-HSL) and N-hexanoyl-l-homoserine lactone (C6-HSL). Pseudomonas fluorescens 2–79 produces at least five autoinducers signals, among which are N-(3-hydroxyhexanoyl)-l-homoserine lactone (3-OH-C6-HSL), N-(3-hydroxyoctanoyl)-l-homoserine lactone (3-OH-C8-HSL), N-(3-hydroxydecanoyl)-l-homoserine lactone (OH-C10-HSL), N-octanoyl-l-homoserine lactone (C8-HSL) and C6-HSL, although not all of these signals are necessarily products of PhzI. The pyocyanin-producing P. aeruginosa PAO1 strain produces C4-HSL. The signals alleged to be produced by PhzI in P. chlororaphis PCL1391 are C6-HSL, C4-HSL and C8-HSL. The quorum sensing signals of strain PCL1391 can be detected with bioassays based on the Chromobacterium Cvi (McClean et al., 1997), Photobacterium Lux (Swift et al., 1993; Winson et al., 1998; Devine & Shadel, 2000), Agrobacterium Tra (Cook et al., 1997) and Pseudomonas Las (Pearson et al., 1994) quorum sensing systems. The extensively used biosensor bacterium Chromobacterium violaceum strain CV026 has a defective N-AHL synthase gene, cviI. As a result, the response regulator CviR only activates purple violacein pigment production when autoinducer is provided exogenously (McClean et al., 1997).
The phzI–phzR system differs from the Photobacterium quorum sensing model in that the N-AHL synthase gene is not part of the biosynthetic operon. PhzR controls both phzI and the biosynthetic operon. In the Photobacterium system, the N-AHL synthase and the response regulator are also present in two divergent transcriptional units but the N-AHL synthase gene and the genes for bioluminescence, luxCDABEG, are localized in a single operon (Shadel et al., 1990).
The phzI and phzR genes of P. aureofaciens 30–84, P. fluorescens 2–79, and P. chlororaphis PCL1391 are very homologous. They are located directly upstream of the phenazine biosynthetic core genes. In P. aeruginosa PAO1 the genes for regulation of phenazine biosynthesis are not located upstream of the core biosynthetic genes but elsewhere in the genome. In this strain, expression of the phz genes is regulated by two sets of LuxI–LuxR homologues, RhlI–RhlR and LasI–LasR (Passador et al., 1993; Latifi et al., 1995).
PhzI homologues are approximately 200 amino acids long. Protein sequence analysis shows that there is about 25–35% similarity between LuxI homologous, with a small number of residues conserved in all sequences. Two regions of a LuxI homologue seem to be essential for enzyme function. The C-terminal region has been proposed to play a role in acylated ACP (acyl carrier protein) selection whereas the N-terminal domain contains the active site of the enzyme (Hanzelka et al., 1997). PhzI homologues produce more than one N-AHL molecule, suggesting that they can use more than one fatty acid type as a substrate. PhzI in P. chlororaphis PCL1391 produces C6-HSL, C4-HSL and C8-HSL. Although minor products may not interfere with the function of the main compound, one should take into account the fact that there may be crosstalk if autoinducer molecules produced by one organism reach, and have an activity in another. One should also consider the fact that sensitivity to the distinct signals differs from one organism to another. This is reflected in the concentrations needed to activate gene expression (Schaefer et al., 1996).
LuxI homologues such as LasI and TraI also confer the ability to produce the corresponding signal molecules in E. coli. Experimental evidence was provided that the homoserine lactone moiety of N-AHL is derived from SAM or directly from homoserine lactone pools (Eberhard et al., 1991; Huisman & Kolter, 1994; Hanzelka & Greenberg, 1996). The variable acyl chains are thought to be the products of either fatty acid biosynthesis or degradation (Cao & Meighen, 1989).
Using radiolabelled 3-oxo-C6-HSL in P. fisheri and E. coli it was shown that autoinducers can diffuse between the cytoplasm and the environment (Kaplan & Greenberg, 1985). Additional experimental data have indicated that active transport may be involved in the translocation of long-chain N-AHLs in P. aeruginosa (Pearson et al., 1999). A P. aeruginosa PAO1 mutant lacking the MexAB–OprM efflux pump accumulated autoinducer to a higher internal level than the wild type. These multidrug efflux pumps are known to cause efflux of certain hydrophobic compounds (Nikaido, 1996). The N-AHL signalling systems using long-chain molecules may therefore be dependent upon the function of membrane efflux pumps and the proton-motive force needed to drive these pumps.
PhzR homologues show around 20–35% similarity with other members of the LuxR group. The C-terminal DNA-binding domain present in all LuxR homologues has a helix-turn-helix motif, and is accordingly classified as belonging to the larger LysR superfamily of transcriptional regulators (Henikoff et al., 1988; Fuqua et al., 1994). The N-terminal domain of LuxR homologues is considered to contain the N-AHL binding domain (Sitnikov et al., 1995; Swift et al., 1996). Analysis of E. coli strains containing wild-type and mutant luxR alleles suggest that LuxR functions as a homomultimer, and a region of the LuxR protein is required for multimerization. Promoters that are thought to be binding sites for LuxR-type proteins show sequences with dyad symmetry, suggesting that some LuxR-type proteins bind as dimers (Choi & Greenberg, 1992). Inverted repeat sequences were identified in promoters of the phzI and phzA genes (Chin-A-Woeng et al., 2001b). Such a sequence, commonly referred to as a lux-box-like sequence, are supposed to be binding sites for PhzR (Shadel et al., 1990; Fuqua et al., 1994; Chin-A-Woeng et al., 2001b). The lux-box-like element in the promoter of phzA in P. aeruginosa was shown to be involved in quorum sensing-controlled transcription (Whiteley & Greenberg, 2001).
Immunoprecipitation using antibodies against LuxR showed that LuxR is located in membranes and not in the soluble pool of cytoplasmic proteins (Kolibachuk & Greenberg, 1993). Since LuxR homologues do not have a characteristic α-helical transmembrane motif, it was suggested that LuxR is an amphipathic protein associated with the inner leaflet of the cytoplasmatic membrane, with the C-terminal domain extending into the cytoplasm (Kolibachuk & Greenberg, 1993).
A number of studies investigated the specificity of autoinducer molecules for LuxR using N-AHL analogues. In these studies, the cognate autoinducer molecule was found to be the most active inducer of the reporter systems. Acyl chain length was shown to be an important factor determining activity, allowing deviation into a slightly shorter or longer acyl chain, albeit with a lower activity (Schaefer et al., 1996).
Quorum sensing in the rhizosphere
By now it is clear that many rhizobacteria produce signal molecules for quorum sensing regulation, although the phenotypes or genes regulated by these signals is sometimes unknown. A survey of 137 soilborne and plant-associated Pseudomonas strains revealed that 40% of these strains produced an N-AHL detectable using one of the available N-AHL reporter systems. Remarkably, N-AHL-producing strains were only found among strains isolated from plants; none of the soil-borne strains produced detectable N-AHLs (Elasri et al., 2001).
Many bacteria on the root surface are present in microcolonies forming a biofilm and these locations are likely to be the places where bacteria can reach high cell densities on the root and where a cell density signal can accumulate (Chin-A-Woeng et al., 1997). The production of PCA in the wheat rhizosphere (Thomashow et al., 1990) and expression of the PCN biosynthetic genes of P. chlororaphis PCL1391 in the tomato rhizosphere (Chin-A-Woeng et al., 1998) indicate that quorum sensing systems must be active in the rhizosphere. Coinoculation of P. aureofaciens 30–84, which produces C6-HSL, restored phenazine gene expression in a phzI mutant to wild-type levels in the rhizosphere (Wood et al., 1997), showing that a quorum sensing signal is indeed required.
Since two or more bacterial species can use identical or the same class of signal molecules for interspecies communication, crosstalk may occur between different signalling systems. Microcolonies on the root surface can indeed consist of more than one population or strain (Bloemberg et al., 2000; Dekkers et al., 2000). In mixed biofilms of P. aeruginosa and Burkholderia cepacia in the lungs of cystic fibrosis patients unidirectional N-AHL signalling between the two strains was detected using GFP-based biosensors (Riedel et al., 2001).
Using an N-AHL-dependent indicator strain of P. aureofaciens 30–84 Wood et al. (1997) showed that a bacterial subpopulation on the wheat root induced expression of the reporter. Some isolates were able to block phenazine gene expression, and coinoculation of these negative strains even reduced the disease suppression by P. aureofaciens 30–84 in biocontrol of G. graminis var. tritici. Positive and negative interpopulation signalling on the plant root may therefore play a substantial role in the final efficiency of the use of biocontrol strains (Pierson et al., 2002).
Global regulation of secondary metabolites and quorum sensing systems
Many quorum sensing systems are not only population dependent but are also regulated at additional levels. One particular quorum sensing modulon may be regulated by another quorum sensing system. Bacteria can possess multiple quorum sensing units in which one unit resides on top of one or more other modules and in this way regulate gene expression in a cascaded way. In the phenazine-producing P. aeruginosa PAO1, the lasI–lasR regulatory pair controls the expression of the rhlI–rhlR, also termed vsmI–vsmR, quorum sensing genes (Latifi et al., 1995).
The control of gene expression at the transcriptional level appears to be the major mechanism for quorum sensing systems to modulate the production of secondary metabolites. The stationary phase factor RpoS or σS is involved in the regulation of expression of over 30 genes that function during, or in the transition to, stationary phase. RpoS affects many regulatory genes, some of which are also involved in the regulation of the production of antifungal compounds. An rpoS mutation resulted in the increased production of pyocyanin and the siderophore pyoverdine in P. aeruginosa PAO1 (Suh et al., 1999).
The GacS–GacA global regulatory system regulates the production of several secondary metabolites including 2,4-diacetylphloroglucinol, HCN and pyoluteorin in P. fluorescens CHA0 (Laville et al., 1992; Heeb & Haas, 2001), and pyrrolnitrin, chitinase, 2-hexyl-5-propyl resorcinol and HCN in P. fluorescens BL915 (Gaffney et al., 1994). In P. aureofaciens 30–84 and P. chlororaphis PCL1391 GacS–GacA regulates phenazine biosynthesis (Pierson & Pierson, 1996; T. F. C. Chin-A-Woeng et al. unpublished). GacA controls cyanogenesis in P. aeruginosa via a transcriptional activation of the rhiI gene and, in addition, via a post-transcriptional mechanism involving a recognition site overlapping the ribosome binding site of the hcnA gene (Pessi & Haas, 2001).
Such a post-transcriptional mechanism of regulation has been elucidated for a number of secondary metabolites. The translation of HCN biosynthesis and protease genes appears to operate via a translational repressor protein RsmA/PrpA. RsmA and the homologous CsrA were found to bind to cognate regulatory RNA molecules RsmB (previously AepH), a 259-nucleotide regulatory RNA in E. carotovora (Liu et al., 1993) and CsrB, a 350-nucleotide regulatory RNA identified in E. coli (Liu et al., 1997). It was proposed that binding to RsmB and CsrB antagonizes the regulatory activity of CsrA and RsmA, respectively. The expression of RsmA/PrpA is dependent on the GacS–GacA system.
The expression of PrrB, encoding a regulatory RNA is also dependent on the GacS–GacA system. It was postulated that this noncoding regulatory RNA PrrB sequesters an RsmA-like repressor protein similar to RsmB, which in turn appears to be positively regulated by the GacS-GacA system (Aarons et al., 2000).
There is also experimental evidence that rpoS expression is regulated directly by both quorum sensing and GacS–GacA two–component regulatory systems. RhlR appeared essential for activation of an rpoS–lacZ reporter in response to C4-HSL in E. coli (Latifi et al., 1996). In addition, 3-oxo-C12-HSL triggered stationary-phase phenomena such as repression of cell growth, and change in cell shape and size when added during exponential phase growth (You et al., 1998). Transcription of rpoS in biocontrol strain P. fluorescens Pf5, assessed with an rpoS–lacZ transcriptional fusion, was positively influenced by GacS and GacA during the transition between exponential growth and the stationary phase (Whistler et al., 1998). Recently, the psrA (Pseudomonas sigma regulator) gene product of P. putida WCS358 was indicated to induce rpoS expression (Kojic & Venturi, 2001), while psrA expression is positively regulated by the GacS–GacA system (T. F. C. Chin-A-Woeng et al. unpublished). Increased expression levels of the phzI/phzR genes and consequently of the phz biosynthetic operon, resulting in a tenfold increased PCN production in a psrA mutant of strain PCL1391 may therefore be caused by a reduced rpoS expression.
Environmental conditions regulating biosynthesis of phenazines
The expression of the phenazine biosynthetic genes is also influenced by growth and environmental conditions. In many quorum sensing systems activity is also determined by the physiological status of the cell. For cells of P. chlororaphis PCL1391 grown in a liquid culture, the availability of certain carbon sources and amino acids, including major root exudates components, metal ions, and oxygen status affect phenazine production positively (Chin-A-Woeng et al., 2000; E. T. van Rij et al. unpublished). Amounts of autoinducer and consequently, phenazine production can vary considerably according growth medium (Séveno et al., 2001).
In the P. fisheri quorum sensing system, factors known to affect lux gene expression are iron, oxygen concentration and other regulatory responses to stress conditions. luxR expression requires cAMP and the cAMP-receptor-protein (CRP), for which a binding site is present in the lux operator (Ulitzur & Dunlap, 1995). The involvement of catabolite repression is also indicated by regulation of expression of lasR by Vfr, a CRP-homologue (Albus et al., 1997). The stringent response induced by the amino acid analogue serine hydroxamate also leads to premature production of 3-oxo-C12-HSL, independently of population density (van Delden et al., 2001).
The qscR gene of P. aeruginosa PAO1 dictates the timing of quorum-sensing-controlled gene expression by repression of lasI. The gene is a homologue of LasR and RhlR and could serve to ensure that quorum sensing-controlled genes, such as phenazine biosynthetic genes, are not prematurely transcribed or expressed under less prevailing conditions (Chugani et al., 2001). The combination of signals such as relayed by the GacS sensor kinase allows integration of population density information with other environmental or cellular information, thereby enabling bacteria to fine-tune their activities in line with prevailing conditions.
The phenazine biosynthetic pathway and biosynthesis of aromatic amino acids share a common pathway. In many strains, phenazine compounds are produced in larger quantities than most other endproducts of the aromatic pathway. Studies indicate a more complex form of regulation of the first enzyme in the aromatic pathway in phenazine-producing strains than in strains that do not produce phenazines (Levitch, 1970) and that the regulation of phenazine biosynthesis may differ between phenazine-producing strains (Korth, 1974). Growth rate and pyocyanin production could be enhanced by tyrosine and phenylalanine in P. aeruginosa A237 (Labeyrie & Neuzil, 1981). Phenazine production by P. aeruginosa was inhibited on media with glucose as the only source of carbon (Korth, 1973a). In P. aureofaciens and P. chlororaphis carbon sources with glyceraldehyde phosphate as metabolic intermediate gave rise to PCA and PCN production (Korth, 1973b). In P. aureofaciens, iron enhanced, while oxygen decreased PCA production (Korth, 1971). In liquid cultures of P. fluorescens 2–79 PCA production could be influenced by culture pH, temperature, carbon sources supplied, and addition of zinc sulphate, ammonium molybdate and cytosine (Slininger & Jackson, 1992; Slininger & Shea-Wilbur, 1995). The addition of adenine deterred PCA accumulation, while nitrogen source had no effect on production (Slininger & Jackson, 1992).
In addition, some strains which produce more than one phenazine derivative produce a certain derivative in higher amounts under a distinct set of culture conditions (Kanner et al., 1978). The identification of environmental conditions influencing phenazine production is important since natural environments can vary considerably in these factors and may, for this reason, explain the inconsistency in plant protection displayed by some phenazine-producing strains.
Biopesticides continue to be a promising alternative for the use of environmentally unfriendly chemical fertilizers and pesticides. Crop plant resistance and biological control methods based on natural disease-suppressing organisms are regarded as the main alternatives. A significant number of bacterial biocontrol products based on Pseudomonas, Bacillus, Streptomyces and Agrobacterium species have already been marketed and commercial preparations based on rhizobacteria are steadily increasing. However, biotechnological efforts to give novel and better-performing products are still in need to give them the status of being practical substitutes to chemical pesticides.
Multiple microbial interactions involving bacteria and fungi in the rhizosphere provide enhanced biocontrol, in many cases, in comparison with biocontrol agents used singly. Insight into the extreme complexity of interactions that can occur in the rhizosphere is therefore essential. When developing screening programmes for antagonists, modes of action involved in each type of interaction need to be assessed with particular emphasis on rhizosphere competence, antibiosis, competition/parasitism and induced resistance whenever possible.
Little is known about the regulation of phenazine biosynthesis in the rhizosphere and the exact mechanism by which these phenazines inhibit growth of pathogens. Phenazines such as pyocyanin not only have an antimicrobial action but also appear to play a role in induced systemic resistance. The use of transcriptomics and proteomics promises to greatly speed up the identification of novel genes involved in the complex mechanisms of rhizosphere colonization and plant protection, including the regulation of genes for the production of secondary metabolites. Identification of promoters that are specifically expressed in the rhizosphere will allow the engineering of biopesticides with enhanced performance. Research into the role of plant genes involved in hosting beneficial plant-associated microbes will provide greater insight and should provide new tools to improve plant health. In the future, biological control of diseases could predominate over the use of chemical pesticides, in the same way as biological control of greenhouse insects already predominates.
The research described here was supported by numerous grants from the Netherlands Organization for Scientific Research, either directly for Crop Protection and in cooperation with the Russian Federation or through the Foundations for Earth and Life Sciences, Chemical Sciences, and Technological Sciences. Moreover, grants were obtained from the European Community through the programmes BIOTECH (BIO2-CT93.0053, BIO2-CT93.0196, BIO4-CT96.0027, BIO4-CT96.0181 and BIO4-CT98.0254) and Marie Curie fellowships (ERBFMBI-CT98-2930 and HPMF-CT- 1999–00377).