We inoculated waxmoth larvae (Galleria mellonella
] with single clones of a wild-type, pyoverdine-producing strain of P. aeruginosa
(cooperator), an isogenic mutant of this strain that does not produce pyoverdine (cheat) [14
], or both. We found that insects were on average killed by cooperator infections after 12 hours, 2 hours sooner than for cheat infections, and that mixed infections resulted in an intermediate time to death (Figure ; Kruskal-Wallis test; H
= 42.76, P
< 0.0001; pair-wise comparisons (Mann-Whitney) showed significant differences between all groups: P
< 0.01 in all cases). In separate experiments, we measured the growth rate of single- and mixed-clone infections over 8 hours, and this showed the same pattern as for virulence: cooperators grew faster than cheats, and mixed infections showed an intermediate growth rate (Figure , ANOVA: F(2,55)
= 5.76, P
= 0.005; pairwise comparisons: P 2
0.055). These data demonstrate that the presence of siderophore cheats can reduce the growth rate of a bacterial population, and hence reduce virulence.
Time to death in hours of waxmoth larvae inoculated with pure cooperator, pure cheat or mixed clone infections of P. aeruginosa. Mixed inocula contained cheats and cooperators in a 1:1 ratio. Error bars show ± one standard error.
The effect of single versus mixed cooperator and cheat infections on the number of bacterial doublings (per gram fresh weight). Mixed inocula contained cheats and cooperators in a 1:1 ratio. Error bars show ± one standard error.
We next addressed whether cheats were more likely to be favoured in mixed versus single-clone infections. As predicted, cheat populations grew more rapidly in mixed as opposed to single-clone infections, while the opposite pattern was observed for cooperators (Figure ; ANOVA on log-transformed data shows a significant interaction between number of infecting clones and cheat/cooperator fitness; F(1,75)
= 4.97, P
< 0.03; main effects: cheat/cooperator fitness F(1,75)
= 11.22, P
< 0.002; number of clones F(1,75)
= 0.26, P
= 0.610). Mixed-clone infections are therefore more likely to favour the evolution of siderophore cheats than are single-clone infections, a result consistent with previous in vitro
] and theoretical [8
The number of doublings (per gram fresh weight) or cooperator (black) and cheat (white) clones in single and mixed infections. Mixed inocula contained cheats and cooperators in a 1:1 ratio. Error bars show ± one standard error.
We found that cheats had a slight selective disadvantage when in direct competition with cooperators at 1:1 ratios (Figure ; relative fitness of cheats was less than 1: t
= 2.30, P
< 0.02). Both theory [19
] and experiments on other microbial systems [23
] suggest that the fitness of cheats should be a negative function of their frequency, as a rare cheat will have relatively more cooperators to exploit. To investigate frequency-dependent selection of siderophore cheats, we competed cooperators and cheats in caterpillars at a range of initial cheat frequencies (between 0.03 and 0.9). We found that the relative fitness of cheats increased as their initial frequency decreased (Figure ; F(1,33)
= 9.34, P
< 0.005), such that the fitness of cheats and cooperators was not significantly different at cheat frequencies of 0.1 or lower (cheat relative fitness not significantly different from 1; t
= 1.43, P
The relative fitness of the cheat in mixed infections decreases as its frequency in the inoculum increases. Error bars show ± one standard error.
However, the relative fitness of cheats is noticeably lower in caterpillars than it is in vitro
]; further in vitro
assays carried out simultaneously with the in vivo
work confirm this (data not shown). There are two likely explanations for this. The first is the greater spatial heterogeneity within a caterpillar compared with a media-filled tube. This is likely to increase the relatedness of immediate neighbours [19
], reducing the chances for direct cooperator-cheat interactions and so bestowing a net benefit on cooperators. The second possible explanation is the longer periods of time bacteria spend at high densities in tubes compared with caterpillars; tube assays reached densities of approximately 108
], while insect assays reached densities of approximately 107
cells/ml. The higher the population density of bacteria, the greater the likelihood that cheats will come into contact with cooperators [19
]. Siderophore cheats have been observed at appreciable frequencies in chronic, clinical P. aeruginosa
], and pyoverdine production has been known to decrease over the course of chronic infection [25
]. These observations strongly suggest that cheats can enjoy a selective advantage in longer-term, high-density infections. The extent to which this apparent advantage is frequency dependent is not known. However, as most patients are initially colonised by a single, environmental clone [26
], any cheats present will most likely have arisen within the patient and so must have invaded from an initially low frequency. We are currently carrying out studies investigating the de novo
evolution of siderophore cheats over longer-term scales in this system to confirm this.