Phenotypic plasticity of a cooperative behaviour in bacteria


Rolf Kümmerli, Institute of Evolutionary Biology, University of Edinburgh, West Mains Road, Edinburgh EH9 3JT, UK.
Tel.: +44 131 650 7287; fax: +44 131 650 6564; e-mail:


There is strong evidence that natural selection can favour phenotypic plasticity as a mechanism to maximize fitness in animals. Here, we aim to investigate phenotypic plasticity of a cooperative trait in bacteria – the production of an iron-scavenging molecule (pyoverdin) by Pseudomonas aeruginosa. Pyoverdin production is metabolically costly to the individual cell, but provides a benefit to the local group and can potentially be exploited by nonpyoverdin-producing cheats. Here, we subject bacteria to changes in the social environment in media with different iron availabilities and test whether cells can adjust pyoverdin production in response to these changes. We found that pyoverdin production per cell significantly decreased at higher cell densities and increased in the presence of cheats. This phenotypic plasticity significantly influenced the costs and benefits of cooperation. Specifically, the investment of resources into pyoverdin production was reduced in iron-rich environments and at high cell densities, but increased under iron limitation, and when pyoverdin was exploited by cheats. Our study demonstrates that phenotypic plasticity in a cooperative trait as a response to changes in the environment occurs in even the simplest of organisms, a bacterium.


Over the last 40 years, evolutionary biology has revolutionized our understanding of animal behaviour (Krebs & Davies, 1997). Evolutionary theory predicts that natural selection will favour individuals that maximize their inclusive fitness (Hamilton, 1964; Grafen, 2006). A key factor affecting fitness of an individual is the ability to adjust behaviour in response to changes in the environment (phenotypic plasticity), rather than have behavioural strategies fixed, depending on genotype. There is a convincing body of evidence that natural selection has favoured phenotypic plasticity in a variety of vertebrate and invertebrate species (Charnov, 1992; Stearns, 1992; Scheiner, 1993; Pigliucci, 1996; Agrawal, 2001; West-Eberhard, 2003).

It has recently been suggested that micro-organisms perform a number of cooperative behaviours, including the formation of fruiting bodies and the production of extracellular factors that benefit the local group (Crespi, 2001; Velicer, 2003; West et al., 2006, 2007a; Foster et al., 2007). Cooperative behaviours add a new level of complexity to phenotypic plasticity, as they can vary in response to both the ecological and social environment. For example, in a cooperatively breeding bird, the decision to help or to breed can vary in response to the number of empty breeding sites in the population (an ecological factor) and/or to the number of helpers or the kin relationship on a given nest (social factors). Conditional adjustment of cooperative behaviour, depending upon the social environment, has been shown in a range of animals, including birds, mammals and social insects (Abbot et al., 2001; Russell & Hatchwell, 2001; Clutton-Brock, 2002; Griffin & West, 2003; Field et al., 2006; Komdeur, 2006; Ratnieks et al., 2006). By contrast, relatively little is known about the extent to which microbes exhibit phenotypic plasticity with their social behaviours; adjusting them in response to variation in social conditions (Bassler & Losick, 2006; Keller & Surette, 2006; Diggle et al., 2007a; West et al., 2007a).

We test whether Pseudomonas aeruginosa cells adjust production of iron-scavenging siderophore molecules in response to variations in the social environment. Iron is a major limiting factor for bacterial growth because most iron in the environment is in the insoluble Fe(III) form and is actively withheld by hosts during opportunistic P. aeruginosa infections (Guerinot, 1994; Ratledge & Dover, 2000; Wandersman & Delepelaire, 2004; Miethke & Marahiel, 2007). In response to iron deficiency, P. aeruginosa releases pyoverdin, the primary siderophore of this species, into the local environment (Budzikiewicz, 2001; Visca et al., 2007). Pyoverdin production is a cooperative behaviour (West et al., 2007b) as it can provide a fitness benefit to neighbouring cells, which can take up iron bound to pyoverdin produced by others. Individuals that do not produce pyoverdin can avoid the metabolic cost of its production, but potentially still exploit the pyoverdin produced by others (West & Buckling, 2003; Griffin et al., 2004; Harrison et al., 2006; Buckling et al., 2007; Ross-Gillespie et al., 2007; Kümmerli et al., 2008). Thus, pyoverdin defective mutants, which have been isolated in natural populations, are potential cheats (De Vos et al., 2001; Visca et al., 2007).

Previous studies have demonstrated the underlying regulatory pathways controlling pyoverdin production. At the proximate level, pyoverdin production is regulated by the sigma factor PvdS, which in turn is regulated by the ferric uptake regulator (Fur) (Escolar et al., 1999; Hantke, 2001; Visca et al., 2002; Ravel & Cornelis, 2003; Wandersman & Delepelaire, 2004). In response to high intracellular iron concentration, the Fur protein forms a complex with iron and binds to the pvdS promoter, which represses pyoverdin synthesis (Barton et al., 1996; Leoni et al., 1996; Visca et al., 2002). Facultative expression of genes of the pyoverdin synthesis pathway has since been experimentally demonstrated in media supplemented with differing amounts of iron (Tiburzi et al., 2008). In contrast to this understanding at the genetic level, there is a lack of information on what happens at the phenotypic level, and the fitness consequences. In addition, it is not clear how pyoverdin production is adjusted in response to the social environment, and how this may interact with iron availability.

In this paper, we first show that pyoverdin production is extremely fine tuned in response to the amount of extracellular iron available confirming previous findings (Tiburzi et al., 2008). Moreover, such facultative pyoverdin production is essential, as shown by our experiments with constitutive pyoverdin-producing mutants (i.e. strains that maintain high levels of pyoverdin synthesis, irrespective of iron availability, due to mutations in the fur gene), which revert back to facultative production under high iron concentrations. We then test whether there is facultative adjustment in pyoverdin production in response to cell density and the proportion of cheats within a population – two social factors that can cause changes in extracellular iron availability and in the efficiency of pyoverdin use under natural conditions. Based on our knowledge of proximate mechanisms regulating pyoverdin production, we predict: (1) a decrease in pyoverdin production per cell with increasing cell density, as pyoverdin molecules diffuse freely in the media and are therefore shared more efficiently in the local environment at higher cell densities (Greig & Travisano, 2004; MacLean & Gudelj, 2006); (2) an increase in pyoverdin production per cooperative cell with increasing proportions of cheats, to compensate for a higher proportion of exploitative neighbours; and (3) that iron availability influences both costs and benefits of pyoverdin production, as well as the relative fitness of cooperators and cheats in mixed cultures.

Materials and methods

Bacteria strains

We used P. aeruginosa strain PAO1 (ATCC 15692), obtained from Pierre Cornelis’ laboratory (Vrije University, Brussels, Belgium; Ghysels et al., 2004), as the pyoverdin-producing wild type (wt). This strain also produces pyochelin, which is a secondary siderophore that has a significantly lower affinity for iron (Ankenbauer et al., 1985; Budzikiewicz, 2001). We also used three different deletion mutants directly derived from the specific PAO1 strain used (Ghysels et al., 2004): strain PAO1ΔpvdD, which is unable to produce pyoverdin (pvd−) as the pyoverdin synthetase pvdD is knocked out; strain PAO1ΔpchEF, which is unable to produce pyochelin (pch−) due to deletions of pyochelin synthetases pchE and pchF; and PAO1ΔpvdD pchEF, which is a double knockout defective for both pyoverdin and pyochelin production (pvd−/pch−). Finally, we used two strains (C6 and A4) that constitutively produce pyoverdin due to point mutations in the fur gene (Barton et al., 1996). A4 was derived from PAO1, whereas C6 was derived from PA6261, a PAO1mutant with a deletion in the anr gene, which controls anaerobic respiration in P. aeruginosa (Barton et al., 1996).

We carried out all experiments in static 96-well microtitre plates in 200 μL volumes of minimal-iron media (CAA): 5 g casamino acids, 1.18 g K2HPO4·3H2O, 0.25 g MgSO4·7H2O, per litre; supplemented with 20 mm NaHCO3 (sodium bicarbonate) and 100 μg mL−1 human apo-transferrin (Sigma, Gllingham, UK) (Meyer et al., 1996; Griffin et al., 2004). Apo-transferrin is a powerful natural iron chelator that binds free Fe(III) in the presence of bicarbonate (Schlabach & Bates, 1975) and prevents nonsiderophore-mediated uptake of iron by bacteria. To prevent evaporation, we filled the outer wells of each microtitre plate with sterile water and only used the inner wells for experiments. In experiments with wt, pvd, pch− and pvd−/pch−, strains were cultured from freezer stock for 24 h prior to experimentation in 30-mL glass universals containing 6 mL of standard King’s medium B (KB) in an orbital shaker (200 rpm) at 37 °C. We measured optical density (OD) at 600 nm and diluted denser cultures prior to experimentation because strains differ slightly in their OD after a 24h growth period in KB (wt: 1.213 ± 0.006, pvd−: 1.247 ± 0.007, pch−: 1.174 ± 0.014; or pvd−/pch−: 1.116 ± 0.005). If experiments included the C6 and A4 strains, all cultures were grown from freezer stock for 24 h at 37 °C in 30-mL shaken glass universals containing 6 mL of minimal iron media (CAA). This was performed to prevent C6 and A4 reverting back to facultative pyoverdin production in iron-rich media prior to experimentation.

Unless stated otherwise, we inoculated 105 bacteria from KB cultures into CAA media and let them grow under experimental conditions for 24 h at 37 °C in a static incubator. We then measured OD at 600 nm. Pyoverdin fluoresces green and can be quantified in solution using a fluorimeter (SpectraMax M2; Molecular Devices (Workingham, UK); excitation: 400 nm, emission: 460 nm, cut off: 475 nm) (Ankenbauer et al., 1985; Cox & Adams, 1985; Prince et al., 1993). Pyoverdin is the only component measured in the culture using these excitation and emission parameters: nonpyoverdin-producing strains give readings of zero. We used the ratio RFU/OD as a quantitative measure of pyoverdin produced per unit of bacteria (i.e. per cell).

Manipulating iron availability

To test the effect of iron availability on pyoverdin production and bacterial growth, we added different quantities (range 0–500 μm) of Fe(III)Cl3 (ferric chloride; Sigma) to the CAA media. We subjected all four bacterial strains to nine different iron supplementation treatments in 12-fold replication.

To assess fitness consequences of constitutive pyoverdin production, we subjected the fur mutants A4 and C6 to nine different iron supplementation treatments, as described above, in ninefold replication and compared pyoverdin production and population growth to the wt. The fur gene is essential in P. aeruginosa (Barton et al., 1996; Banin et al., 2005), such that high iron conditions are lethal for fur mutants, and select for restoration of fur function and reversion to facultative pyoverdin production. Indeed, we found that cultures inoculated with fur mutants showed decreased levels of pyoverdin production with increased iron supplementation (see Fig. A1). Using chrome azurol S (CAS) assays (Schwyn & Neilands, 1987; Shin et al., 2001), we could qualitatively demonstrate that high iron concentrations resulted in high reversion rates to facultative pyoverdin production, presumably due to restoration of fur function. For C6, reversion was complete in treatments with FeCl3 (≥ 2 μm) such that no constitutive pyoverdin producers were detectable after a 24-h growth phase, whereas for A4 some constitutive pyoverdin producers could persist under all iron concentrations (see Fig. A2). Despite reversion to facultative pyoverdin production, our data show that pyoverdin was significantly overproduced by C6 and A4 compared with the wt in treatments with relatively low iron supplementations (0–5 μm FeCl3). Pyoverdin overproduction and processes involved in the restoration of fur function were associated with significantly lower bacterial densities and therefore reduced fitness in cultures with C6 and A4 and their revertants compared with wt cultures (see Fig. A3).

Figure A1.

 Mean pyoverdin production per cell (RFU/OD600) in iron-depleted media supplemented with various quantities of FeCl3 (in μm). Comparisons between wild type (filled circles) and (a) C6 fur mutant (open triangles) and (b) A4 fur mutant (open squares).

Figure A2.

 The effect of experimental treatments (FeCl3 supplementation in μm) on whether or not colonies of Pseudomonas aeruginosa produce pyoverdin on chrome azurol S (CAS) plates supplemented with no or high iron (100 μm FeCl3). Pyoverdin production causes a colour change from blue to orange in the CAS reagent, which leads to the formation of orange halos around colonies. Facultative pyoverdin producers (PAO1) should have halos on low iron but not on high iron plates, whereas constitutive pyoverdin producers (C6 and A4) should have halos on both low and high iron plates. The absence of halos around C6 colonies from experimental treatments supplemented with FeCl3≥ 2μm suggests that restoration of facultative pyoverdin production occurred.

Figure A3.

 Mean cell density (OD600) after 24-h growth in iron-depleted media supplemented with various quantities of FeCl3 (in μM). wt (closed circles), C6 fur mutant (open triangles), A4 fur mutant (open squares).

Manipulating cell density

To test whether cell density influences the quantity of pyoverdin produced per cell, we varied inoculum size by adding approximately 103, 104, 105 or 106 cells from wt cultures grown in KB to the CAA media. We carried out experiments using three different iron supplementation regimes (0, 0.5 or 50 μm FeCl3) in 24-fold replication.

If pyoverdin production per cell is density dependent, we expected a change in the pyoverdin concentration, relative to cell density, at different time points during growth. To test this, we inoculated 105 bacteria from wt cultures grown in KB into CAA media and recorded cell density and pyoverdin concentration over time (0–54 h). This also allowed us to gather information on the durability of pyoverdin molecules, which can be reused multiple times and accumulate in the media (Faraldo-Gómez & Sansom, 2003). We carried out the experiment with three different iron supplementation regimes (0, 0.5 or 50 μm FeCl3) in eightfold replication.

Manipulating wild type vs. mutant density

We competed the wt strain with the pvd−/pch− strain to test whether wt cells (cooperators) upregulate pyoverdin production in the presence of pvd−/pch− cells (cheats). We mixed wt and pvd−/pch− from KB cultures in the following proportions of wt culture volumes: 0.17, 0.33, 0.5, 0.67, 0.83 and 1; and inoculated 105 cells into CAA media under three different iron supplementation regimes (0, 0.5 or 50 μm FeCl3) in 16- to 20-fold replication. Volume ratios match cell ratios quite well, although wt cultures grow to slightly higher densities than pvd−/pch− in KB (volume mixing of 1 : 1 resulted in a proportion of pvd−/pch− cells = 0.48). We tested whether pyoverdin production per wt cell is a function of the proportion of wt and pvd−/pch− cells in the media. To account for changes in the proportion of wt cells, due to competition with pvd−/pch−, we used the average proportion of wt cells before and after the competition as the independent variable. We determined the initial and final ratios accurately by plating out dilutions of mixed cultures on KB agar and counting the number of colony-forming units (CFUs). The two strains were distinguished by their colour difference: wt colonies are green, whereas pvd−/pch− colonies are white. We then calculated the relative fitness (v) of pvd−/pch− as


where x1 is the initial proportion of pvd−/pch− and x2 is their final proportion. The fitness value of v therefore signifies whether pvd−/pch− increased in frequency (ν > 1), decreased in frequency (ν < 1) or remained at the same frequency (ν = 1) over the competitive period.

Statistical analysis

We used regression models to test for significant relationships between iron availability, cell density, time, pvd−/pch− density (independent variables) and the pyoverdin production per cell (dependent variable). We logarithmically transformed the values of pyoverdin production per cell as well as iron concentration and bacterial relative inoculation densities prior to analysis. We performed linear and polynomial (quadratic and cubic) regressions for all data sets and determined the model that best fit our data following the procedure of Crawley (2007). We used R 2.7.0 ( for all statistical computations.


Cells adjust pyoverdin production in response to iron availability

The pyoverdin-producing strains (wt and pch−) increased pyoverdin production as iron was increasingly limited (Fig. 1a, linear regression on log-transformed values for wt: r2 = 0.964, F1,106 = 2888, < 0.00001; for pch−: r2 = 0.965, F1,106 = 2922, < 0.00001). Under high iron supplementation (FeCl3≥ 50 μm), pyoverdin production ceased completely. These results are consistent with a similar pattern of facultative pyoverdin production in response to iron supplementation described by Tiburzi et al., (2008).

Figure 1.

 (a) Pyoverdin production per cell (RFU/OD600) and (b) cell density (OD600) after 24-h growth in iron-depleted media (CAA + 20 mm NaHCO3 + 100 μg mL−1 apo-transferrin) supplemented with various quantities of FeCl3 (in μm). Values are given as mean ± 95% confidence interval (CI) across 12 replicates. Pyoverdin-producing strains (solid lines): wt (filled circle) and pch− (open circle). Pyoverdin-defective strains (dashed lines): pvd− (filled square) and pvd/pch− (open square).

Pyoverdin production had a dramatic effect on population growth and absolute fitness (Fig. 1b). At relatively low levels of iron supplementation (0–5 μm), the wt and pch− strain grew to significantly higher densities than pyoverdin-defective strains (pvd− and pvd/pch−) (contrast comparisons following anova: 10.8 < t232 < 21.1, all < 0.0001). Under these conditions, strain pvd/pch− hardly grew at all, whereas strain pvd− that is able to produce pyochelin grew to intermediate densities, illustrating that pyochelin can only partly compensate for the lack of pyoverdin. Population growth of pyoverdin-defective strains (pvd− and pvd/pch−) increased at relatively high levels of iron supplementation (20500 μm), although they grew to significantly lower densities than pyoverdin-producing strains (contrast comparisons following anova: 5.6 < t185 < 7.1, all < 0.0001).

Cells adjust pyoverdin production in response to cell density

In iron-poor environments (0 and 0.5 μm FeCl3), pyoverdin production per cell decreased significantly with an increasing density of inoculum (Fig. 2, quadratic regression for 0 μm FeCl3: r2 = 0.593, F2,92 = 69.5, < 0.00001; linear regression for 0.5 μm FeCl3: r2 = 0.624, F1,94 = 158.6, < 0.00001). In the iron-rich environment (50 μm FeCl3), bacteria produced no or very low amounts of pyoverdin.

Figure 2.

 Pyoverdin production per wt cell (RFU/OD600) as a function of the bacteria inoculum size. Values are given as mean ± 95% confidence interval (CI) across 24 replicates after 24-h growth in iron-depleted media supplemented with 0 μm FeCl3 (squares), 0.5 μm FeCl3 (circles) or 50 μm FeCl3 (triangles).

Our time-series analysis revealed that the amount of pyoverdin per cell in the media varied significantly over the growth period (Fig. 3a) and therefore as a function of cell density (Fig. 3b). The relationship between the amount of pyoverdin in the media and time (length of growth period) was best described by cubic regressions (Fig. 3a, 0 μm FeCl3: r2 = 0.798, F3,76 = 105.3, < 0.00001; 0.5 μm FeCl3: r2 = 0.794, F3,76 = 102.2, < 0.00001; 50 μm FeCl3: r2 = 0.633, F3,76 = 46.3, < 0.00001). Hence, the amounts of pyoverdin per cell increased rapidly after the lag phase (after 6 h) and peaked at 9 h (50 μm FeCl3), at 12 h (0.5 μm FeCl3) or at 24 h (0 μm FeCl3) and decreased thereafter. Pyoverdin molecules can be reused multiple times (Faraldo-Gómez & Sansom, 2003); however, our time-series analysis shows that pyoverdin detected at 12 h had completely disappeared after 24 h in the iron-rich environment (50 μm FeCl3), suggesting a limited durability.

Figure 3.

 (a) Quantities of pyoverdin per wt cell in the media and (b) cell density (OD600) at different time points during the growth period. Values are given as mean ± 95% confidence interval (CI) across eight replicates after growth in iron-depleted media supplemented with 0 μm FeCl3 (squares), 0.5 μm FeCl3 (circles) or 50 μm FeCl3 (triangles).

Cells adjust pyoverdin production in response to the presence of cheats

In iron-limited environments (0 and 0.5 μm FeCl3), wt cells (cooperators) significantly increased pyoverdin production in the presence of higher proportions of nonpyoverdin-producing cheats (pvd/pch−) (Fig. 4a, linear regression for 0 μm FeCl3: r2 = 0.461, F1,108 = 94.2, < 0.00001; for 0.5 μm FeCl3: r2 = 0.871, F1,88 = 594.4, < 0.00001). A negative relationship might be expected simply due to the density effect observed in Fig. 2 (i.e. higher proportions of pvd/pch− mean lower wt densities). However, the effect of pvd/pch− cheat frequency on pyoverdin production was four times (0 μm FeCl3) and three times (0.5 μm FeCl3) stronger (measured as the maximal difference in pyoverdin production per cell between treatments) than the effect seen in the density experiment, and occurred at a cell density range 1000 times smaller than in the density experiment. In the iron-rich environment (50 μm FeCl3), pyoverdin production was close to zero for all cooperator-to-cheat ratios.

Figure 4.

 Competition between nonpyoverdin-producing pvd/pch− (cheats) and pyoverdin-producing wt (cooperators) in various mixing ratios and its effect on (a) the pyoverdin production per cooperator cell, and (b) relative fitness of cheats compared with that of cooperators. Values with an asterisk are significantly different from 1 (< 0.05). Competitions took place over 24 h in iron-depleted media supplemented with 0 μm FeCl3 (squares and open bars), 0.5 μm FeCl3 (circles and grey bars) or 50 μm FeCl3 (triangles and black bars).

The nonsiderophore-producing pvd/pch− strain (cheat) exploited wt cooperators efficiently in the environment where no iron was added: cheat fitness values were significantly higher than one (Fig. 4b, two-tailed t-tests: 2.25 < t15–19 < 5.56, all < 0.05). This fitness advantage disappeared in the environments where iron was added (0.5 and 50 μm FeCl3), where pvd/pch− fitness was either not significantly different from one or significantly lower than one, depending on the cooperator-to-cheat ratio (Fig. 4b). The proportion of cheats had a marginally significant effect on relative cheater fitness only in the 0 μm FeCl3 treatment (anova on log-transformed data: F4,89 = 2.50, = 0.048), whereas there was no significant effect in the 0.5 μm FeCl3 (anova: F4,73 = 2.25, = 0.072) and 50 μm FeCl3 (anova: F4,73 = 0.84, = 0.50) treatments. In the 0 μm FeCl3 treatment, cheats with a proportion of 67% cooperators had significantly lower fitness values than cheats in cultures with 50% (contrast comparisons: = 0.013) and 83% (= 0.028) cooperators. This observation does not have an obvious biological explanation.


We have demonstrated that the bacterium P. aeruginosa facultatively adjusts the production of its primary siderophore molecule, pyoverdin, depending upon the social environment. Specifically, we have shown that wild-type cells produce more pyoverdin at lower population densities and when a high proportion of the population is composed of individuals that do not produce pyoverdin (cheats). Furthermore, facultative adjustment of pyoverdin production in response to iron availability significantly affected the competitive dynamics between cheats and wild-type cells, such that the relative fitness advantage that cheats experienced in iron-depleted media disappeared in iron-rich environments as pyoverdin production ceased.

We have shown that pyoverdin production per wild-type cell decreases with cell density (Fig. 2). We predicted this because, at higher densities, pyoverdin molecules are shared more efficiently, as fewer molecules are lost due to random diffusion. Consequently, investment in pyoverdin production can be reduced because fewer molecules per cell are required to guarantee a sufficient supply of iron. This finding further demonstrates that pyoverdin production has social consequences for both the focal cell that produces pyoverdin and other, neighbouring cells and selection for pyoverdin production can therefore be influenced by both direct and indirect fitness consequences (Griffin et al., 2004; Kümmerli et al., 2008). At higher cell densities, the relative importance of indirect fitness consequences increases (Greig & Travisano, 2004). In addition, the observation that P. aeruginosa downregulates pyoverdin production in the presence of higher numbers of cooperative conspecifics contrasts with observations from between-species competitive interactions, where P. aeruginosa cells upregulated pyoverdin production in the presence of competing Staphylococcus aureus bacteria (Harrison et al., 2008).

The reduction of pyoverdin production with increased cell density contrasts with the production of a number of extracellular products that are controlled by quorum sensing (QS) and only released above a certain cell density threshold (Diggle et al., 2007a, Bassler & Losick 2006). There is some controversy over whether or not the regulatory pathways of QS and siderophores are interlinked. Although some studies showed evidence for such links (Stintzi et al., 1998; Whiteley et al., 1999; Cornelis & Aendekerk, 2004; Juhas et al., 2004; Oglesby et al., 2008), others could not confirm them (Schuster et al., 2003; Wagner et al., 2003) or provided alternative explanations (Bredenbruch et al., 2006; Diggle et al., 2007c; Dubern & Diggle, 2008). It has been suggested that QS coordinates release of extracellular products at high population densities, when their production will provide the greatest benefit (Diggle et al., 2007b; Sandoz et al., 2007). We suggest that variation in the fitness consequences of producing different extracellular products could account for the differences in their production in response to cell density. If a product is essential for growth, then it should be released at all population densities, and hence less will be required per cell at higher population densities due to more efficient sharing. By contrast, costly extracellular products that are beneficial, but not essential for growth, should only be produced at higher population densities when their production cost is diminished (Brown & Johnstone, 2001). Formal theory addressing this difference would be extremely useful (Nadell et al., 2008).

How does facultative adjustment of cooperative behaviour affect the relative fitness of cooperators and cheats? Facultative adjustment of pyoverdin production, as opposed to fixed levels of production, benefits cheats in low iron concentrations because cooperators are stimulated into increasing production of pyoverdin to compensate for exploitation (Fig. 4a). However, as iron supplementation is increased, we have shown that wild-type cooperator cells reduced (0.5 μm) or switched off (50 μm FeCl3) pyoverdin production (Fig. 1a). Thus, the total cost of pyoverdin production was reduced when it was not needed, which increased the fitness of the cooperators relative to the cheats when grown in mixed populations (Fig. 4b). The overall effect of phenotypic plasticity in environments with fluctuating iron availabilities is therefore hard to predict, because, in mixed populations, it provides a relative advantage to cheats at low iron concentrations, but a disadvantage at high iron concentrations. This emphasizes the importance of examining both the ecological and social structure of natural populations. Additionally, a mechanism to avoid exploitation by cheats, and more generally by other strains, would be to make pyoverdin molecules more specific such that their uptake is limited to close relatives (e.g. clonemates). In agreement with this idea, evidence for diversifying selection at the pyoverdin locus has been described (Fuchs et al., 2001; Smith et al., 2005).

Monocultures of pyoverdin-defective strains (pvd− and pvd/pch−) grew to significantly lower densities than wild-type cultures under high iron availability (Fig. 1b) suggesting that the knocked out genes (coding for nonribosomal peptide synthetases) have other fitness-related functions. This seems plausible because nonribosomal peptide synthetases are multifunctional enzymes generally involved in the syntheses of a variety of exoproducts (Finking & Marahiel, 2004; Grünwald & Marahiel, 2006).

To conclude, our results illustrate two points. First, we have demonstrated phenotypic plasticity in response to changes in the social environment in a cooperative trait in bacteria. Phenotypic plasticity is also likely to affect selection on the production of other extracellular bacterial products that have been proposed as potential cooperative public goods (West et al., 2007a). Secondly, the production of extracellular molecules such as siderophores is often associated with virulence effects in the host (Meyer et al., 1996; Rumbaugh et al., 1999; Takase et al., 2000; West & Buckling, 2003; Harrison et al., 2006; Harrison, 2007; Miethke & Marahiel, 2007). Thus, the consequences of phenotypic plasticity and the relative fitness of mutants compared with wild types may play a key role in their within-host evolution.


We thank Pierre Cornelis and Michael Vasil for providing us with wild-type and siderophore knockout strains and fur mutant strains respectively. We also thank Steve Diggle, Freya Harrison, Adin Ross-Gillespie and Kevin Foster for discussion or helpful comments on the manuscript, Gavin Ballantyne for help in the laboratory and Kevin Foster for suggesting the use of fur mutants. This work was funded by the Royal Society, the Leverhulme Trust, the BBSRC, the Swiss National Science Foundation and a Marie-Curie Intra-European fellowship.


We used two strains (C6 and A4) of Pseudomonas aeruginosa that constitutively produce pyoverdin due to point mutations in the fur gene (Barton et al., 1996) to compare their performance with that of facultatively pyoverdin-producing wt bacteria (PAO1). The fur gene is essential in P. aeruginosa (Barton et al., 1996; Banin et al., 2005), such that high iron conditions are lethal for fur mutants, and selects for restoration of fur function and reversion to facultative pyoverdin production.

Indeed, we found that pyoverdin production in cultures inoculated with C6 and A4 significantly decreased with increasing quantities of iron supplementation – a pattern that is very similar to the facultative pyoverdin production pattern shown by the wt strain (Fig. A1, linear regression on log-transformed values, for C6: r2 = 0.953, F1,79 = 1616, P < 0.00001; for A4: r2 = 0.945, F1,79 = 1351, P < 0.00001; for wt: r2 = 0.958, F1,79 = 1846, P < 0.00001).

To determine whether the facultative pyoverdin-production patterns by C6 and A4 were due to bacteria strains having restored fur function, we compared phenotypes of cultures prior to and after the experiment on CAS plates (Schwyn & Neilands, 1987; Shin et al., 2001). The production of pyoverdin causes a colour change from blue to orange in the CAS reagent, which leads to an orange halo forming around the colonies on the agar plate. Accordingly, colonies of a facultative pyoverdin producer should have no halos on CAS plates supplemented with iron (100 μm FeCl3), whereas halos should appear on CAS plates with no iron supplementation. By contrast, colonies of constitutive pyoverdin producers should have halos, both on iron-rich and iron-poor CAS plates. As expected, wt cultures plated onto iron-rich and iron-poor CAS plates showed facultative pyoverdin production patterns no matter whether cultures were plated out prior to or after the experiment (Fig. A2). Moreover, as expected, cultures of C6 and A4 developed halos, both on iron-rich and iron-poor plates when plated out prior to experimentation (Fig. A2). However, for C6 this pattern changed when cultures where plated out after the experiment. On iron-rich CAS plates, there were no longer halos around colonies from cultures supplemented with FeCl3 (≥ 2μm) illustrating that in these experimental treatments only bacteria with mutations restoring fur function could survive and grow (Fig. A2). For A4, halos appeared on iron-rich CAS plates around colonies from all iron supplementation treatments. This indicates that some proportion of constitutive pyoverdin producers survived in all treatments (Fig. A2). However, the strong similarity of pyoverdin production patterns between C6 and A4 in response to iron supplementation (Fig. A1) suggests that mutations restoring fur function emerged and spread in A4 as well.

Despite reversions to facultative pyoverdin production, our data show that pyoverdin was significantly overproduced by C6 and A4 when compared with the wt in treatments with relatively low iron supplementations (0–5 μm FeCl3, contrast comparisons following anova, C6 vs. wt: t126=7.3, < 0.00001, A4 vs. wt: t126 = 10.6, P < 0.00001). Pyoverdin overproduction and restoration of fur function were associated with significantly lower bacterial densities, and therefore reduced fitness, in cultures with C6 and A4 and their revertants when compared with wt cultures (contrast comparisons following anova: C6 vs. wt: t232 = 7.4, P < 0.00001, A4 vs. wt: t232 = 13.8, P < 0.00001, Fig. A3).