Leaf surfaces harbour bacterial epiphytes that are capable of influencing the quorum sensing (QS) system, density determination through detection of diffusible signal molecules, of the plant-pathogen Pseudomonas syringae pv. syringae (Pss) which controls expression of extracellular polysaccharide production, motility and other factors contributing to virulence to plants. Approximately 11% of the bacterial epiphytes recovered from a variety of plants produced a diffusible factor capable of inhibiting the QS system of Pss as indicated by suppression of ahlI. Blockage of QS by these interfering strains correlated strongly with their ability to limit iron availability to Pss. A direct relationship between the ability of isogenic Escherichia coli strains to sequester iron via their production of different siderophores and their ability to suppress QS in Pss was also observed. Quorum sensing induction was inversely related to iron availability in culture media supplemented with iron chelators or with FeCl3. Co-inoculation of interfering strains with Pss onto leaves increased the number of resultant disease lesions over twofold compared with that on plants inoculated with Pss alone. Transposon-generated mutants of interfering strains in which QS inhibition was blocked did not increase disease when co-inoculated with Pss. Increased disease incidence was also not observed when a non-motile mutant of Pss was co-inoculated onto plants with QS interfering bacteria suggesting that these strains enhanced the motility of Pss in an iron-dependent manner, leading to an apparent increase in virulence of this pathogen. Considerable cross-talk mediated by iron scavenging apparently occurs on plants, thereby altering the behaviour of bacteria such as Pss that exhibit important QS-dependent traits in this habitat.
Many bacterial traits have been found to be expressed in a cell-density dependent manner (Miller and Bassler, 2001; Whitehead et al., 2001). In a process known as quorum sensing (QS), bacteria coordinate gene expression in a population of cells based on the detection of their own small diffusible signal molecule (Miller and Bassler, 2001; Whitehead et al., 2001). Upon accumulation of sufficient signal molecule in their environment, a process facilitated by an increasing number of cells and/or by limitation of diffusional losses of the signal, expression of certain genes is activated or suppressed. Among the genes activated in this process are usually those involved in signal production, thereby leading to a positive feedback loop of further signal molecule production. A variety of traits including extracellular polysaccharide (EPS) production, motility, bioluminescence, genetic competence and plasmid exchange, and production of virulence factors are regulated by QS and presumably play important and context-dependent roles in the lifestyles of microbes that utilize QS.
Several plant pathogenic bacteria have traits that are regulated by QS (Loh et al., 2002; von Bodman et al., 2003). In the plant pathogenic bacterium Pectobacterium carotovorum, the CarR regulator and the signal, 3-oxo-hexanoyl homoserine lactone (HSL), generated by the carI- encoded synthase are required for the production of carbapenem antibiotics as well as exoenzymes that are involved in the maceration of plant tissues. Pantoea stewartii controls its production of the EPS virulence factor stewartan via its accumulation of 3-oxo-hexanoyl HSL and the regulator EsaR (Vonbodman and Farrand, 1995). Furthermore, the root-associated plant-beneficial bacterium, Pseudomonas aureofaciens, has two separate non-hierarchical QS systems that contribute to biocontrol of fungal pathogens; both the phzI/R and the csaI/R loci are required for the production of phenazine antibiotics and exoproteases required for successful colonization of the wheat rhizosphere (Wood and Pierson, 1996; Wood et al.,1997; Zhang and Pierson, 2001). In Rhizobium leguminosarum, a symbiont of legumes, QS induction promotes rhizosphere growth, nodule development and plasmid conjugation (Rodelas et al., 1999; Thorne and Williams, 1999; Lithgow et al., 2000).
The plant pathogen Pseudomonas syringae pv. syringae (Pss), the causal agent of brown spot disease of bean, exhibits several population size-dependent behaviours. Pss forms cellular aggregates on leaves in which the survival of desiccation stress is much greater than that of more solitary cells (Monier and Lindow, 2003). Pss exhibits QS and produces 3-oxo-hexanoyl HSL with AhlI under the control of AhlR (Quinones et al., 2004). AhlI mutants of Pss exhibit lower ability to survive the stresses of desiccation on leaves (Quinones et al., 2005). A number of traits associated with epiphytic fitness and virulence in Pss are controlled by QS. While EPS production is stimulated by QS in Pss, swarming motility is repressed (Quinones et al., 2005). The QS mutants of Pss incite many more leaf lesions after topical application, presumably because their hypermotilty increases their chances of entering the leaf through stomata or other portals (Quinones et al., 2005). A model of QS-regulation of traits in Pss posits that solitary cells would be free to move about leaves after immigrating as solitary cells, thereby increasing their chances of encountering the discrete sites on leaves where nutrient availability is high (Leveau and Lindow, 2001). After multiplying at such sites, populations can reach a size enabling QS (Dulla and Lindow, 2008), thereby stimulating EPS production which is beneficial for epiphytic survival (Yu et al., 1999) but suppressing motility, ensuring that cells would remain at such preferred sites of colonization (Dulla et al., 2005). The QS also induces internal tissue maceration but suppresses the expansion of watersoaked lesions in bean pods (Quinones et al., 2005). Thus, it appears that several aspects of the colonization of both the surface and the interior of plants by Pss is under control of QS, and that any factor that would interfere with QS could have an effect on the epiphytic fitness and apparent virulence of Pss.
Because of the importance of QS in various bacteria many studies have sought compounds that could block QS. Many diffusible compounds such as non-cognate acyl-homoserine lactones (AHL), AHL intermediates, di-cyclic peptides, brominated furanones and other eukaryotic exudates have the ability to interfere with QS in AHL-producing bacteria (Givskov et al., 1996; McClean et al., 1997; Bauer and Robinson, 2002). In addition, enzymes of bacterial origin such as the Bacillus sp. AiiA AHL lactonase (Dong et al., 2000), the Ralstonia sp. AiiD AHL acylase (Lin et al., 2003) and the AHL acylase PvdQ of Pseudomonas aeruginosa (Huang et al., 2003) are able to degrade a variety of AHLs. Blockage of AHL-mediated cell-cell signalling might therefore be important both to eukaryotes as well as to bacteria themselves to escape from deleterious traits expressed by bacteria that exhibit QS (Givskov et al., 1996).
As Pss is usually a member of a mixed microbial community on leaves, it seemed possible that its behaviour on plants could be influenced by the other resident microflora. In this study, we identify several epiphytic strains that are able to inhibit QS in Pss and show that they do so by a novel mechanism that involves sequestration of iron. We further illustrate that these QS interfering strains strongly influence the behaviour of Pss on plants.
Identification of bacteria interfering with QS in Pss
A total of 1117 bacterial strains were isolated from various plant species from California, Wisconsin and Minnesota and screened for their ability to interfere with QS exhibited by Pss B728a (pBQ9), an AHL biosensor containing a PahlI fusion to a promoterless gene encoding GFP, on King's medium B (KB). When this biosensor was sprayed over test colonies, their inhibitory effect on QS in Pss was apparent as zones around colonies where no GFP production (GFP production indicates QS induction in response to the production of AHL by Pss) was present (Fig. 1). A few strains that had inhibited growth of B728a (pBQ9) rather than only its ability to exhibit GFP fluorescence were omitted from further analysis. A total of 123 strains, representing 11% of the collection, were able to interfere with QS in Pss in this assay. A subset of these strains was tentatively identified by partial sequencing of their 16S rRNA. All strains characterized were found to be members of the genera Pseudomonas, Erwinia or Pantoea (Table 1). Strain 114, which was identified as a strain of Pseudomonas putida, was chosen for further study because of both its strong effect on QS as well as its ease of culture and genetic manipulation.
Table 1. Relative ability of different epiphytic bacteria to interfere with quorum sensing in Pseudomonas syringae.
Strain identity determined from partial sequence of 16S rRNA gene.
b. Quorum sensing estimated by measurement of expression of ahlI gene during growth of the bioindicator Pseudomonas syringae B728a (pBQ9) in media containing cell-free supernatants of tested strains. Shown is cell normalized GFP fluorescence.
Bioindicator strain Pseudomonas syringae B728a (pBQ9) grown in media containing cell-free supernatant of a culture of Pss B728a grown to a low cell-density.
All strains that inhibited QS in Pss produced a small diffusible compound in culture that was capable of blocking GFP production in bioindicator strain B728a (pBQ9) at some distance away from the colony (Fig. 1). Cell-free extracts from broth cultures also inhibited QS in Pss; when added to wells in agar plates to which indicator strain Pss B728a (pBQ9) was oversprayed such extracts produced large zones where GFP production, but not growth of Pss was inhibited. This suggested that the inhibitor was a small molecule. When cells of B728a (pBQ9) were suspended in such cell-free conditioned medium in which various inhibitor strains had grown, the GFP fluorescence of B728a (pBQ9) was reduced as much as fivefold as compared with the control consisting of conditioned medium in which Pss B728a itself had grown to low cell densities (Table 2).
Table 2. Mesurement of quorum sensing in Pseudomonas syringae and iron availability sensed by this species when grown near other bacteria on agar media.
QS induction estimated from level of ahlI expression as indicated by fluorescence intensity of the QS bioindicator Pss B728a (pBQ9) in cells recovered from plates proximal to colonies of the tested strain. Shown is the percentage of GFP fluorescence of cells of the bioindicator grown away from an indicator strain (denominator) with that of cells closely proximal to the test strain (numerator).
Iron availability estimated from levels of expression of an iron-repressible promoter as indicated by fluorescence intensity of the bioindicator Pss B728a (pVITIR) in cells recovered on plates far from tested strains. Shown is the percentage of GFP fluorescence of cells of the bioindicator grown away from an indicator strain (numerator) with that of cells closely proximal to the test strain (denominator).
GFP fluorescence indicative of QS induction and Iron availability measured only far from a test strain and used as a control. The results shown are from individual experiments that are representative of that obtained in multiple replicated experiments.
The characteristics of the QS inhibitor produced Pseudomonas strain 114 was assessed as this strain interfered with QS more than most other strains. Treatment of the conditioned media of this strain with proteinase K did not reduce its ability to interfere with QS in Pss. Likewise, after partitioning of conditioned media with ethyl acetate, the inhibitor of QS remained in the aqueous phase (data not shown). The compound made by strain 114 was diffusible through membranes with an exclusion limit of 3.5 kDa (data not shown). A random transposon mutagenesis screen of Pseudomonas strain 114 was therefore performed to identify determinants of the extracellular agent that interferes with QS in Pss. Approximately 3000 mutants generated by random insertion of Tn5 were screened for QS interference using the Pss B728a (pBQ9) bioindicator as before. Several mutants were isolated which were greatly reduced, but not completely blocked, in their ability to inhibit QS in Pss. The mutant designated 114-KK1 exhibited the greatest reduction of QS interference ability and was chosen for more detailed characterization (Fig. 2A). Sequencing of the site of the transposon insertion in 114-KK1 revealed that Tn5 had disrupted a gene with 81% DNA sequence identity to corA, which encodes a putative cobalt magnesium transporter (Smith et al., 1993; Smith and Maguire, 1993) in Pss strain DC3000 (Buell et al., 2003). Sequence analysis of the regions surrounding Tn5 insertion sites in mutants 114-KK2 and 114-KK3 revealed genes most similar in sequence identity to mucA (71% in Pseudomonas fluorescens) (Schnider-Keel et al., 2001) and mvaT (82% in Pseudomonas mevaloni) respectively. In P. mevalonii, MvaT is a heteromeric transcriptional regulator of the mvaAB operon (Rosenthal and Rodwell, 1998). MvaT, a transmembrane protein with a cytoplasmic portion that binds and regulates the sigma factor AlgU in Pseudomonas aeurginosa (Mathee et al., 1997) causes increased production (up to ninefold more) of the siderophore pyocyanin than the parent strain (Diggle et al., 2002). MucA regulates the transition of the mucoid phenotype in P. aeurginosa via the regulation of AlgU which controls the production of EPSs such as alginate. In addition, global transcriptional profiling has revealed elevated expression of the pyoverdine sigma factor pvdS and other pyoverdine production genes downstream of pvdS in a MucA mutant background (Wu et al., 2004). These findings suggested that Pseudomonas strain 114 might inhibit QS in Pss via its ability to deplete metals in culture media, presumably by its production of a high-affinity siderophore that could sequester metals such as Co and Iron that might be needed for maximal QS. To test this model various metal salts such as CoCl3, FeCl3, Cadmium Acetate, ZnCl2, MgCl2 and MnCl2 were added to KB to determine if they were able to abolish the QS inhibitory effect of wild-type strain 114. The interference of QS by Pseudomonas strain 114 was suppressed when Cobalt, Iron, Zinc or Cadmium salts, but not when Magnesium or Manganese salts were added to KB on which strain 114 was growing. The interference of QS by all inhibitory strains tested was also abolished in media containing 2 µM FeCl3 (data not shown) suggesting that iron sequestration is a common mechanism of interference of QS. While both strain 114 and mutant 114-KK1 grew equally well on KB medium or on KB containing CoCl2 (50 µg ml−1) or FeCl3 (2 µM), no apparent interference of QS conferred by either strain was apparent on the metal-amended media (Fig. 2B). Furthermore, the GFP fluorescence of Pss B728a (pBQ9) on the metal-amended plates was much higher than on unamended KB (Fig. 2A and B) suggesting that the process of QS itself was enhanced in the presence of higher concentrations of these metals. In addition, Pseudomonas strain 114 apparently made iron less available to cells of Pss when they were grown together on Luria agar (LA), a medium which contains moderate amounts of iron. Cells of Pss B728a do not produce pyoverdine siderophore, evident as a diffusible yellow-green fluorescent pigment, when grown on LA. No visible diffusible fluorescent pigment was made by Pseudomonas strain 114 when grown on either KB or LA suggesting that the siderophore that it presumably produces was not a pyoverdine. Importantly, large zones of yellow-green fluorescent pigment were apparent when cells of Pss B728a were sprayed onto LB plates around colonies of Pseudomonas 114, but not around colonies of mutant 114-KK1 (Fig. 3). Presumably the pyoverdine siderophore of Pss was made only in the vicinity of Pseudomonas 114 which had sequestered iron, but at a much lower level around colonies of mutant 114-KK1 which itself was unable to sequester iron.
Further support for the role of iron sequestration by bacterial strains in interference of QS in Pss was provided by the finding that interference of QS was strongly related to the amount of iron available to Pss in co-cultures of a variety of bacteria. The iron available to Pss B728a was assessed with the GFP-based iron biosensor Pss B728a (pVITIR) (Joyner and Lindow, 2000) containing an iron-repressible Ppvd::gfp fusion; the GFP fluorescence exhibited by this biosensor is thus inversely related to the amount of ferric iron in its vicinity. The iron availability of Pss B728a (pVITIR) was compared on cells grown on KB near QS interfering bacteria with that from cells harvested from a distal part of the plate far from the interfering strain. All bacterial epiphytes that had been found to interfere with QS of Pss also reduced the availability of iron to this strain as reflected in enhanced GFP fluorescence exhibited by B728a (pVITIR) (Table 2). These strains varied in their ability to reduce iron availability, with some strains such as isolate 294 reducing iron levels on KB more than strains such as 114 and 177w (Table 2). Importantly, mutant 114-KK1 reduced iron availability much less than its parental strain 114 (Table 2). The extent of interference of QS as measured by the relative intensity of GFP fluorescence exhibited by B728a (pBQ9) in cells harvested in the vicinity of test strains with that of cells harvested at distal parts of the plate far from the interfering strain was generally inversely related to the available iron levels as measured by B728a (pVITIR) (Table 2). Strains such as 294 and 177y which inhibited QS the most also reduced iron availability more than other strains (Table 2). A significant linear correlation between the relative level of QS (y-axis) as measured by fluorescence of B728a (pBQ9) and the inverse of the relative iron availability (x-axis) as measured by fluorescence of B728a (pVITIR) compared with that of Pss B728a alone conferred by various bacterial epiphytes (Table 2) was observed: (y = −0.61x− 9.5) (R2 = 0.43) (P < 0.05).
Other bacterial strains that had not been selected for their ability to interfere with QS of Pss also exhibited a strong relationship between their ability to sequester iron and to inhibit QS. Escherichia coli AN194 which produces the high-affinity siderophore enterobactin (Langman et al., 1972) reduced QS induction of Pss by over 50% and greatly reduced the availability of iron to Pss compared with when Pss was grown alone (Table 2). In contrast, E. coli strain LG1315 which is blocked in production of enterobactin but harbours a plasmid conferring production of the siderophore aerobactin (Williams, 1979) did not make iron as unavailable to Pss as did parental strain AN194 and also did not inhibit QS as much (Table 2). Escherichia coli strain MT147, which does not produce either enterobactin or aerobactin (Ozenberger et al., 1987), had no significant effect on QS of Pss nor did it make iron less available (Table 2). A significant linear correlation between the relative level of QS (y-axis) as measured by B728a (pBQ9) and the inverse of the relative iron availability (x-axis) as measured by B728a (pVITIR) compared with that of Pss B728a alone conferred by various E. coli strains (Table 2) was observed: (y = −0.80x + 9.8) (R2 = 0.80) (P < 0.05).
Since strains that interfered with QS in Pss apparently did so by producing siderophores with higher affinity for iron than those produced by Pss, we directly tested the process of QS induction under conditions in which iron levels were varied experimentally. The QS induction in Pss B728a (pBQ9) was directly related to the amount of available iron in the culture medium. The extent of GFP fluorescence increased with the concentration of FeCl3 added to media (Fig. 4). In contrast, the extent of GFP fluorescence exhibited by this biosensor was lower on KB containing either N,N′-bis(o-hydroxybenzyl)ethylenediamine-N,N′-diacetic acid (HBED) (Fig. 4) or tannic acid (data not shown) two compounds known to bind ferric ions. The apparent suppression of QS in Pss by tannic acid and HBED was reversible by addition of FeCl3 (data not shown). This suggests that these compounds reduced QS due to their ability to chelate iron and not due inhibitory effects on the cells.
Given that the iron availability sensed by individual bacteria may vary greatly due to micro-heterogeneities in their environment (Joyner and Lindow, 2000), the extent of QS might be expected to vary between cells in culture media if iron levels are spatially variable. The fluorescence of individual cultured cells of Pss B728a harbouring pBQ9RED, containing a QS-inducible mRFP1 (Dulla and Lindow, 2008) and pVITIR, a GFP-based iron biosensor (Joyner and Lindow, 2000) was assessed microscopically. Considerable variation in GFP fluorescence, indicative of variable levels of available iron, and red fluorescence indicative of QS induction was observed in individual cells recovered from colonies of Pss grown on KB. A significant linear relationship between relative level of QS as measured by red pixel intensity (y-axis) and the inverse of the relative level of iron availability as assessed from green pixel intensity (x-axis) was observed (y = 118−1.72x, R2 = 0.405, P < 0.05) (Fig. 5).
Modulation of virulence of Pss on plants
As the process of QS, while contributing to the epiphytic fitness of P. syringae, reduces the incidence of infection of plants by suppressing the motility of Pss needed for invasion of leaves (Quinones et al., 2005), we explored whether bacteria that interfere with QS could alter the disease process. Mutants of Pss that are deficient in AHL production incite 2 to 4-fold more infections of bean leaves than does a wild-type strain (Quinones et al., 2005); therefore, it seemed possible that bacteria that interfered with QS could substantially increase disease incidence. The number of infections incited by Pss B728a when applied alone to leaves and when co-inoculated with strains that interfered with QS was compared after 7 day incubation under greenhouse conditions. The incidence of infection of plants co-inoculated with either Pseudomonas strain 120 or Pseudomonas strain 114 was more than twofold higher than on plants inoculated with Pss alone (Table 3). In contrast, there was no significant difference in the incidence of infection between plants co-inoculated with Pss and the corA mutant 114-KK1, which did not inhibit QS in vitro, and plants inoculated with Pss alone (Table 3). The epiphytic population size of Pss B728a was similar on plants co-inoculated with Pseudomonas strain 114, mutant 114-KK1 or Pseudomonas strain 120 compared with plants inoculated with Pss alone after 24 h incubation in moist conditions, as well as after several days in the greenhouse (data not shown).
Table 3. Incidence of infection of bean leaves co-inoculated with Pseudomonas syringae and bacteria varying in their ability to interfere with quorum sensing.
a. The number of disease lesions per leaf caused by Pss were counted on individual trifoliate leaves approximately 7 days after inoculation. Shown are the average number of lesions on from 30 or more leaves per treatment. Within individual experiments, treatments with the same letter designation do not differ significantly (P > 0.005) as determined by an LSD test. Representative experiments are shown; similar results were seen in repeat experiments.
15.11 ± 2.56 A
14.11 ± 1.16 A
22.46 ± 2.40 A
66.87 ± 9.11 A
Pss + Pseudomonas 114
38.21 ± 5.43 B
33.27 ± 2.62 B
Pss + Pseudomonas 114-KK1
18.05 ± 1.45 A
Pss + Pseudomonas 120
35.19 ± 3.78 B
Pss + E. coli top10
48.36 ± 6.84 B
Pss + E. coli top10 (pGFD4)
28.95 ± 2.78 A
Pss + E. coli AN194
148.72 ± 23.65 B
Pss + E. coli LG1315
85.86 ± 11.75 AB
Pss + E. coli MT147
72.61 ± 9.99 A
As epiphytic bacteria that interfered with QS in Pss also reduced iron availability and increased the incidence of disease when co-inoculated with this species, we explored whether other bacteria not selected from this habitat which reduce iron availability to Pss could also increase its infectivity. The incidence of infection of beans co-inoculated with Pss and E. coli Top10 that produces the siderophore enterobactin was over twofold higher than on plants inoculated with Pss alone (Table 3). Likewise, the incidence of infection of plants co-inoculated with enterobactin-producing E. coli strain AN194, was much higher than on plants inoculated with Pss alone (Table 3). This stimulation of disease could be shown to be due to its sequestration of iron and not due to other factors by comparing the behaviour of Pss mixed with isogenic strains of E. coli varying in their production of siderophores. Escherichia coli strain LG1315, blocked in production of enterobactin but harbouring a plasmid conferring production of the aerobactin increased the incidence of infection of plants co-inoculated with Pss only slightly, but not significantly compared with that on plants treated with Pss alone (Table 3). Likewise, the incidence of disease on plants co-inoculated with E. coli MT147 which lacks production of either enterobactin or aerobactin was similar to that on plants inoculated with Pss alone (Table 3). Further evidence that QS-interfering bacterial strains enhance disease incidence specifically by altering the production of, or response to, AHL in Pss was provided by comparing disease in the presence of isogenic E. coli strains that limited iron availability to Pss but which differed in AHL production. Introduction of ahlI on plasmid GDF4 into E. coli Top10 conferred the production of 3-oxo-hexanoyl HSL (Dulla and Lindow, 2009). While co-inoculation of E. coli Top10 with Pss greatly increased the incidence of infection of bean compared to inoculation with Pss alone, mixtures of AHL-producing E. coli Top10 (pGDF4) and Pss resulted in similar number of infections as when Pss was inoculated alone (Table 3).
Direct manipulation of iron availability on leaves also affected the apparent virulence of P. syringae. Bean plants were sprayed with suspensions of Pss in water or in solutions of tannic acid. After incubation in a moist chamber for 24 h to enable bacterial populations to increase in size and for mobility of Pss across the leaf to occur, plants were placed on a greenhouse bench for 7 days to enable disease symptoms to develop. Epiphytic populations of Pss measured 24 h after inoculation were higher on plants in which Pss was applied in water alone (7.44 ± 0.05 log cells g−1) to those on plants in which inoculum was applied in 30 µM tannic acid (7.03 ± 0.12 log cells g−1). Despite the slightly lower population sizes on plants treated with tannic acid more than twice as many lesions were observed as on control plants (26.0 ± 2.3 vs. 13.9 ± 1.3 lesions leaf−1). The number of cells of Pss that had invaded leaves within 24 h was assessed by quantifying the number of culturable cells in leaves macerated after surface sterilization. About twice as many cells of Pss were found within leaves topically treated with tannic acid compared to with water alone (data not shown). Thus, it appears that infection of bean was greatly stimulated under conditions of low-iron availability where QS would be suppressed and in which motility and hence invasion of the leaves was increased.
No increase in infection of plants with a non-motile, flgK mutant of Pss was observed in the presence of iron-sequestering bacteria. While the flgK mutant incited only about 50% as many infections as the wild-type strain B728a, co-inoculation of this non-motile mutant with either E. coli strains LG1315, AN194 or MT147 resulted in similar numbers of lesions as that caused by the flgK mutant alone (Table 4). Likewise, the number of lesions on plants co-inoculated with the flgK mutant and E. coli strain LG1315 (pGFD4) which over-produces AHLs was similar to that on plants co-inoculated with the flgK mutant alone. These results suggest that the primary effect of the iron sequestering strains on Psswas to alter its motility, thereby increasing its likelihood of entering a plant.
Table 4. Incidence of disease lesions on bean co-inoculated with a non-motile mutant of Pseudomonas syringae and various E. coli strains differing in siderophore production.
a. The average number of disease lesions per trifoliate leaf (n = 30) 7 days after inoculation with Pss is shown for each treatment. Treatments with the same letter designation do not differ significantly (P > 0.005) as determined by an LSD test. A representative experiment is shown; similar results were seen in repeat experiments.
Pss B728a alone
22.55 ± 1.56 A
flgK Pss alone
9.93 ± 1.06 B
flgK Pss + E. coli LG1315
8.96 ± 0.78 B
flgK Pss + E. coli LG1315 (pGFD4)
10.16 ± 1.26 B
flgK Pss + E. coli AN194
8.62 ± 0.73 B
flgK Pss + E. coli MT147
10.97 ± 1.05 B
As QS controls a variety of traits including swarming motility that contribute to the virulence and epiphytic fitness of the plant pathogenic bacterium P. syringae, it seemed possible that bacteria that could interfere with such pathogens could have measurable effects on its behaviour on leaves. A surprisingly high proportion of bacteria recovered from plants were able to interfere with QS in Pss B728a. The finding that 11% of all strains had the potential to block QS in Pss suggests that such interference might be a common process on leaves. Similar numbers of interfering strains have been found in studies done in the rhizosphere. Pierson and colleagues found that about 7% of the culturable rhizosphere bacteria of wheat could inhibit QS in P. aureofaciens via mechanisms that remain uncharacterized (Morello et al., 2004).
While a large proportion of bacteria on leaves could interfere with QS in Pss, the number of taxa with this potential was low. The subpopulation of epiphytes that inhibited QS regulation all were members of the genera Pseudomonas, Pantoea or Erwinia. This was not surprising given that members of these genera are among the most common taxa of culturable bacteria on leaves. No obvious association between bacterial taxa and ability to interfere with QS was observed. Additional taxa would likely have been revealed had more extensive taxonomic treatment of the interfering strains or sampling from more plant species been pursued. This is particularly likely since iron sequestration, identified here as the mechanism of QS interference, should be a common trait. Iron is commonly a limiting element and one for which most bacteria have evolved high-affinity assimilation schemes (Crosa, 1989). Thus, it seems likely that bacterial taxa with high-affinity iron uptake systems will be found in most habitats, and would influence the iron-dependent behaviour of their neighbours. Since QS in Pss is apparently highly dependent on levels of iron in its vicinity, it is strongly influenced by neighbouring cells that sequester iron.
The mechanism by which epiphytic bacteria inhibit QS in Pss is unlike that seen in other studies. Soil-borne bacteria such as Bacillus, Ralstonia and Pseudomonas sp. are able to inhibit AHL-mediated QS via their production of AHL-degrading enzymes such as AHL lactonases and AHL acylases, production of non-cognate AHLs that block the process of QS, or other small molecules that block QS induction (Leadbetter, 2001). We did not find evidence that such mechanisms were common among the bacteria that we studied. For example, the finding that the interfering agents in culture extracts could pass through a dialysis membrane suggests that degrading enzymes were not involved in this process. Considerable circumstantial evidence pointed to iron sequestration as the mechanism for interference by Pseudomonas strain 114 as well as other strains. The requirement of the CorA homolog, a putative metal transporter, strongly suggests that metal uptake was required for interference of QS. Likewise, as discussed above, disruptions in of mucA and mvaT would likely affect iron uptake in Pseudomonas strain 114. Direct evidence that these loci were all involved in metal-dependent suppression of QS was provided by the finding that iron supplementation of culture media reversed QS inhibition conferred by Pseudomonas strain 114 as well as that of all interfering stains examined. Consistent with this observation was the finding that QS interference of all strains was much less pronounced on media such as LA, in which iron is relatively available, compared with low-iron media such as KB. It is possible that since many bacterial genes are regulated in an iron-dependent fashion, iron may have suppressed traits that led to AHL degradation. However, the production of the AHL-degrading and interfering agents noted above are not known to be regulated by iron and it seems unlikely that they would be regulated by iron in all cases.
Certain epiphytic bacteria appear to be able to sequester iron much more effectively than Pss, both in culture and on plants. It was surprising to observe that certain bacteria, such as Pseudomonas strain 114, could stimulate pyoverdine production in Pss even on culture media such as LA, which contains sufficient iron that Pss pyoverdine production is normally suppressed. This suggests that other bacteria take up more iron than Pss on such a medium, thereby depleting these metals. Direct evidence for iron depletion was provided by assessing iron availability with the iron biosensor Pss B728a (pVITIR) grown near these strains. All reduced iron levels more than did Pss itself (Table 2). Furthermore, such strains apparently produce siderophores that have a higher affinity for iron than the siderophore(s) of Pss. It is well known that bacterial siderophores differ in affinity for iron (Neilands, 1981; 1995). The stability constant for enterobactin (log Kf = 52, pH 7.4) is much higher than for aerobactin (log Kf = 22.9, pH 7.4) (Raymond and Carrano, 1979) and pyoverdine from P. syringae (log Kf = 25, pH 7) (Torres et al., 1986). This explains why enterobactin-producing E. coli strain AN194 made iron less available to Pss than did aerobactin-producing E. coli strain LG1315 (Table 2). The P. syringae B728a genome does not have any putative aerobactin receptors. Pseudomonas syringae does have two putative fepA-like genes (PSYR2290 and PSYR2892), but the strong influence of the enterobactin producing AN194 on P. syringae suggest that it cannot uptake enterobactin. The reduced levels of iron available to Pss in the presence of these strains was directly correlated to its reduced levels of QS and thus to the enhanced disease that it caused (Tables 2 and 3). While the affinity of iron sequestering agents produced by the other epiphytic bacteria studied here is unknown, the finding that a similar relationship between their abilities to withhold iron from Pss and block its QS suggests that they also produced siderophores with high affinity for iron.
We were surprised to find that low-iron availability apparently blocks QS in Pss given that in most other bacteria QS is inhibited by high-iron levels. Unlike in other bacterial species where modest changes in QS are associated with varying iron levels, in Pss the level of AHL production was dramatically lower in low-iron media compared with more iron replete media (Fig. 4). A recent report also reveals that a fur mutant of P. syringae pv. tabaci is severely diminished in AHL production, suggesting that iron levels may also influence QS in this related species (Cha et al., 2008). In other species such as Pseudomonas aeuruginosa several global transcriptional studies have linked QS to the process of iron sequestration (Whiteley et al., 1999; Schuster et al., 2003; Wagner et al., 2003). In P. aeuruginosa, QS induces siderophore production. More recently, Juhas and colleagues found that the VqsR regulator, which upregulates the production of AHLs in P. aeruginosa, increases production of the siderophore pyochelin (Juhas et al., 2004). The production of the pyoverdine siderophore in fluorescent Pseudomonads is regulated by the alternative sigma factor pvdS. Expression of pvdS has also been shown to be reduced in a vqsR mutant (Juhas et al., 2004). The pvdS is also controlled by the general iron-regulated repressor Fur (Lamont et al., 2002; Ravel and Cornelis, 2003). When Fur is bound to iron it acts as a transcriptional repressor, although other metals such as Co, Zn and Mn can also bind to Fur (Mills and Marletta, 2005).
Bacterial strains that limit iron availability to Pss had a surprisingly strong effect on the behaviour of this pathogen on leaves. Since these strains are able to sequester iron on leaves they reduced QS and apparently made Pss behave as if it were blocked in AHL production. Mutants of Pss B728a that are blocked in AHL production are hyper-swarmers and are able to invade the leaf much quicker than the wild-type strain. Apparently because those cells were more likely to enter the leaf interior they more frequently caused disease (Quinones et al., 2005). Interfering strains 114 and 120 which reduced iron availability to Pss both blocked QS and doubled the number of brown spot lesions on bean leaves when co-inoculated with Pss (Table 2). We presume that this effect was also due to increased swarming motility of Pss and consequent invasion into the tissue to cause disease since the immobile flgK mutant strain of B728a was unaffected by the iron-limiting strains (Table 4). The striking difference of the ability of strain 114 and its mutant 114-KK1 into alter the ability of Pss to cause disease suggests that cross-talk involving iron competition can be substantial on leaves. It is unlikely that the various QS interfering epiphytes and E. coli strains that enhanced disease in this study colonized the interior of the plants since they are non-plant pathogens that do not multiply well within plants unlike pathogens such as Pss (Alfano and Collmer, 1996). It thus appears that the interactions with Pss that suppressed QS would have occurred primarily on the surface of the plant.
The presence of large numbers of bacteria on plants that can compete with Pss for iron suggests that complex interactions that affect the process of QS occur in this habitat. Although interactions mediated by iron scavenging has been shown in studies done in the plant rhizosphere displaying competition (Raaijmakers et al., 1995) and cooperation (Loper and Henkels, 1999), no such study has been done in the phyllosphere which is a physiologically different environment. There has been much interest in the use of beneficial organisms for the biological control of disease. The application of such organisms to plants at appropriate times often leads to large changes in the composition of microbial communities compared with plants where normal successional processes dictate microbial community composition (Lindow and Suslow, 2003). The properties that differentiate successful biological control organisms from those that do not control disease are poorly understood. A poor correlation between antibiotic production and biological control activity is commonly observed, and other processes such as competition for limiting resources are often operative (Wilson and Lindow, 1993). This study suggests that disease will be favoured given a particular population size of Pss on plants harbouring other microbes capable of reducing the availability of iron. It might thus be expected that the most successful biological control agents applied preemptively for the control of diseases caused by this pathogen would not themselves compete well with Pss for iron. In addition, their production of 3-oxo-hexanoyl HSL, the cognate AHL of Pss might additionally suppress the motility of the pathogen and reduce the likelihood of disease (Dulla and Lindow, 2009). Insight into the processes leading to disease should enable the rational development of such biological control strategies.
Quorum size is typically defined as the population size threshold where chemical signals reach a sufficient concentration to modulate expression of density-dependent genes. The presence of bacteria that block the production of AHLs or the response of producers to such signal molecules would affect the size of populations needed to achieve a ‘quorum’. The quorum size of Pss was found to be often quite small (less than 30 cells aggregate−1) on leaves, especially under relatively dry conditions where the AHL could not disperse (Dulla and Lindow, 2008). This study suggests that the quorum size of a pathogen such as Pss will be strongly context-dependent since neighbouring bacteria can influence the process of QS by altering iron levels. The prevalence of bacterial strains on leaves that can interfere with QS in Pss suggests that there is much cross-talk among plant-associated bacteria. More study of the role of iron in QS in Pss is warranted given the strong effects of iron observed here.
Leaves used for isolation of epiphytes were collected from various plant species (other than bean) from California, Wisconsin and Minnesota as before (Dulla and Lindow, 2009). Individual leaves were placed in bags with 10 ml of 10 mM KPO4 buffer (pH 7.0) (phosphate buffer) and sonicated for 7 min to remove bacteria from leaves. Appropriate dilutions of leaf washings were plated on KB and the plates incubated at 28°C for 3 days. Individual colonies differing in size and appearance were spotted on KB, LB, and KB containing 10 mM FeCl3 and screened for their ability to interfere with AHL production of Pss by overspraying the inoculated plates with a dense suspension (109 cells ml−1) of Pss B728a (pBQ9) with an artists' airbrush and incubating them at 28°C overnight. The Pss B728a (pBQ9) AHL biosensor contains a PahlI fusion to a promoterless gene encoding GFP. Inhibitors of QS were identified as colonies that permitted the growth of Pss B728a (pBQ9) but blocked GFP fluorescence exhibited by this indicator strain in their vicinity as visualized with a handheld UV lamp.
The identity of QS interfering strains were determined by partial sequence analysis of cloned 16S ribosomal RNA genes. Bacterial strains were grown on KB, scraped from plates after 24 h growth at 28 C, and washed once in sterile water to generate a dense cell suspension for use as a template for amplification of their 16S rRNA gene. A 1.5 kb region of the 16S rRNA gene was amplified using the universal 16S sequencing primers 1510R (5′-GTGCTGCAGGGTTACCTTGTTACGACT-3′) and 6F (5′-GGAGAGTTAGATCTTGGCTCAG-3′) (van der Meer et al., 1998). Cycling conditions were as follows: 95°C for 3 min for initial denaturation, 32 cycles of 94° for 30′, 54°C for 30′ and 72°C for 1 min. Tentative phylogenic placement of strains was made using comparisons to the ribosomal DNA sequence using BLAST search with the NCBI Gen Bank database (Altschul et al., 1997).
Quantification of QS inhibition and pyoverdine production
To quantify inhibition of Pss QS by interfering strains, each putative interfering strain was grown to stationary phase overnight at 28°C in KB broth. Cells were separated from conditioned media via centrifugation and the supernatant filter-sterilized with a 0.2 µm filter. Cell-free conditioned media (1 ml) was amended with an equal volume of fresh KB broth, inoculated with Pss B728a (pBQ9) to a concentration of 105 cells ml−1 and grown overnight at 28°C. GFP fluorescence was then measured and normalized per cell (determined by OD600). Alternatively, 107 cells of an inhibitor isolate was spotted onto a KB plate and grown overnight at 28°C. Plates were then oversprayed with a dense suspension of Pss B728a (pBQ9) using an artists' airbrush and incubated overnight at 28°C. Pss B728a (pBQ9) cells proximal to the inhibitor isolate were harvested from the plate and GFP fluorescence was determined on a TD-700 fluorimeter (Turner Designs; Sunnyvale, CA) and normalized per cell (determined by OD600).
Iron availability in the vicinity of inhibitor strains was estimated using the iron biosensor Pss B728a (pVITIR) in which Ppvd is fused to a promoterless gene encoding GFP which exhibits GFP fluorescence that is inversely proportional to the available iron in culture media (Joyner and Lindow, 2000). A total of 107 cells of a given inhibitor strain were spotted onto a KB plate and grown overnight at 28°C. Plates were then oversprayed with a dense suspension of Pss B728a (pVITIR) using an artists airbrush and incubated overnight at 28°C. Pss B728a (pVITIR) cells proximal to the inhibitor isolate were harvested from the plate with a wire loop and their GFP fluorescence determined in a fluorimeter as above. Pyoverdine production of Pss B728a was also visualized as the diameter of blue-green fluorescence exhibited by this indicator strain when a dense suspension was oversprayed on colonies of the inhibitor strain grown on LA overnight at 28°C as above.
Transposon mutagenesis of interfering strain Pseudomonas isolate 114 was performed by introducing mini-Tn5, harboured on suicide plasmid pRL27, via conjugation from E. coli BW20767 (Larsen et al., 2002). Donor and recipient strains were grown on LA, suspended in phosphate buffer, mixed in a 1:1 ratio, plated on KB and incubated at 28°C overnight. Mating mixtures were harvested, re-suspended in phosphate buffer and plated on KB containing kanamycin and rifampicin. A subset of the resultant mutants were spotted on KB and scored for ability to interfere with QS of Pss as above. The identity of the genes into which Tn5 had inserted was determined by cloning the genomic DNA proximal to the inserted transposon followed by sequencing of the regions flanking the inserted Tn5 as described before (Larsen et al., 2002).
Pss harbouring pBQ9RED (Dulla and Lindow, 2008) and pVITIR were grown on KB agar overnight, scraped off and re-suspended in 1 mM KPO4 buffer (pH 7.0). The cell suspension was air dried on ProbeOn Plus (Fisher Scientific; Pittsburg, PA) glass microscope slides and mounted with Aqua-Polymount (Polysciences, Warrington, PA). Cells were viewed at 1000× magnification with a Laser Scanning Confocal Microscope, Zeiss 510 UV/Vis Meta. GFP fluorescence was visualized by excitation using a single line 488 nm laser and emissions filtered with a 505–550 nm bandpass. The mRFP1 was excited with a single line 543 laser and emissions collected with filters set to a 560–615 nm bandpass. Pixel intensity was determined using Adobe Photoshop (Adobe Systems, San Jose, CA).
Bacterial strains were grown on KB, harvested with a loop, washed in phosphate buffer and re-suspended in phosphate buffer at a concentration of 106 cells ml−1. The cell suspension was then sprayed to wetness onto leaves of bean plants (Phaseous vulgaris cv. Bush Blue Lake 274) with fully expanded trifoliate leaves (approximately 2 weeks old). Plants were then placed in a humid chamber at approximately 25°C and periodically exposed to a fine mist of sterile water to maintain high relative humidity and to maintain a thin film of water on the leaf surface. Artificial light was maintained for 16 h periods within the 24 h cycle. After 1 day, plants were taken out of chamber and placed on a greenhouse bench for 7 days to allow disease development. The number of brown spot lesions was then counted on each trifoliate leaflet. Epiphytic populations were determined on trifoliate leaves as previously described (Quinones et al., 2004; 2005).
To determine the ability of Pss to invade leaves, cells were inoculated similar to above and plants placed in a humid chamber and sampled at various times. Individual trifolate leaflets were surface sterilized with 15% hydrogen peroxide for 10 min, rinsed with sterile distilled water and then treated with a catalase solution (1 mg ml−1) to remove any residual hydrogen peroxide. Bacterial populations within leaves that had escaped surface sterilization were quantified by plating of appropriate dilutions of leaves macerated in phosphate buffer on appropriate selective media.
A one-way ANOVA using Statistica (Statsoft, Tulsa, OK) was used to determine effects of treatments on brown spot disease on bean. When necessary, data were log (X + 1)-transformed for normality. Multiple comparisons among treatment effects were made using a LSD test. A simple regression analysis using JMP IN 5.1 (SAS Institute, Carey, NC) was used to test the linear relationship between QS inhibition of Pss and iron limitation of Pss caused by the various isolates or treatments.
We thank Drs Linda Kinkel and Susan Hirano for supplying field-grown plant materials. We thank Bianca Quinones with assistance with isolations and greenhouse work. Thanks to Dr Ryan Shepard and Kavitha Mattaparthi for assistance with strain identification and Juliana Cho for assistance in measuring iron-dependent QS. We also thank Dr Joyce Loper for providing isogenic E. coli strains differing in siderophore production. This research was funded in part by United Stated Department of Agriculture National Research Initiative Grant 2004-35319-14145 and an Environmental Protection Agency STAR fellowship awarded to G.D.