In co-culture with the bacterivorous flagellate Poteriospumella lacustris, the prey bacterium Pseudomonas putida exhibited a characteristic succession of predation defenses. The initial and the final defense differed substantially from one another with regard to their mechanism and their population-level benefits to the bacteria.
Our results strongly indicate that the initial bacterial defense falls into the category of chemical defense, and is regulated by phenotypic plasticity. This would require P. putida to be able to sense predator density and to regulate the excretion of inhibitory substances accordingly. Because a considerable proportion of the P. putida genome is known to be involved in regulation and signal transduction allowing for very flexible responses to environmental triggers [41] both conditions are likely to be met. The filtrate exposure tests (Fig. 3) provide specific evidence for the ability of P. putida KT2440 to up- and downregulate the excretion of compounds inhibiting flagellate growth in response to grazing pressure. Previous research [25] corroborated the ability of P. putida to escape grazing from bacterivorous flagellates through induced responses like aggregation or biofilm formation.
To provide a possible characterization for the apparent bacterial toxin, the whole-genome sequences of P. putida KT2440 obtained here were aligned against the antiSMASH [42] database. The output suggests the existence of non-ribosomal peptide synthetase clusters mediating the production of pyoverdines, a particular class of siderophores. The latter are molecules released by bacteria into the environment, which enhance the uptake of essential metals like, e.g., iron under deficient conditions. Specific pyoverdines associated with P. putida KT2440 have previously been identified [43]. Recent findings have shown that the benefits from siderophore production are not limited to competitive advantages gained from enhanced resource exploitation [44]. Pyoverdines were also demonstrated to determine the virulence of Pseudomonads via the damage of mitochondria in colonized hosts [45]. Moreover, pyoverdines were shown to be involved in the inducible defense of P. putida against predatory myxobacteria [46]. Such multiple functions have been reported for a number of bacterial metabolites, especially in Pseudomonads [47], and the particular combination of pyoverdin effects would explain the observed simultaneous flagellate inhibition and promoted bacterial growth.
In contrast to the initial chemical defense of P. putida, the subsequent filamentation clearly provides an example of rapid evolution. Although the responsible mutation(s) could only be pinpointed in a few isolates so far (Table S1), there is no doubt about the genetic manifestation and heritability of the filamentous phenotype due to its demonstrated non-reversible nature.
Only recently, similar observations were made by long-term co-cultivation of Pseudomonas fluorescence with the amoeboid predator Neaglena grubei [48]. In that system, protective adaptations like enhanced biofilm formation and altered motility were traced down to mutations in two particular genes (wspF, amrZ).
From the perspective of the bacterial population, filamentation appears to be a much less efficient defense mechanism than toxin production. This is clearly reflected by the ratio of prey to predator biomass, which differed by two orders of magnitude between the initial and final defense (Table 6). It raises the question of why bacteria would abandon a highly effective form of defense in favor of a much less effective one. As demonstrated experimentally, adaptation of predators to the toxin can be excluded as a cause (Fig. 4). Moreover, it was not instantly evident how the small-sized flagellate was ultimately able to persist in large numbers given a very high proportion of completely inedible prey individuals (Fig. 1D and Fig. S2).
To develop a comprehensive understanding of the system addressing the questions raised above, we set up a semi-continuous differential equation model to simulate the dynamics of predator and prey phenotypes. The model considers seven state variables (carbon, densities of four bacterial phenotypes, flagellate density, and toxin concentration) whose dynamics are controlled by nine processes (Table 3, Fig. 2). In addition to microbial growth and grazing, the model implements a phenotypically plastic predation defense (toxin production) as well as a genetic defense (filamentation) which arises via mutation. The particular assumptions implemented in the model are as follows:
Dual effect of bacterial metabolites
In line with the above discussion on siderophore-like compounds, secondary metabolites excreted by P. putida were assumed to exhibit a dual function, both inhibiting the growth of flagellates and allowing for a more efficient exploitation of the resources by bacteria. The inhibition of predators was demonstrated directly (Figs. 3 and 4) while enhanced resource exploitation was inferred from bacterial abundances in co-cultures exceeding the carrying capacity observed in predator-free controls (Fig. 1A, day 11–18).
Metabolite production is costly
The production of bacterial metabolites was assumed to be associated with a slight fitness cost [49] since resources are diverted from reproduction, thus resulting in a lowered growth rate of toxin-producing bacteria. The assumed fitness cost of 11% (parameter cBx in Table 5) allowed for the best agreement between simulated and observed data and is in agreement with data on the cost of pyoverdine production by P. aeruginosa [50]. The cost only manifests when toxin production is upregulated.
Predator recognition and quorum sensing interact
In the model, the production of bacterial metabolites is upregulated when the two conditions of high flagellate abundance and high bacterial abundance coincide. That is, the expression of the toxin-based bacterial defense is assumed to be jointly controlled by predator recognition and quorum sensing (QS). Examples for such joint control of bacterial defenses have been reported previously [8, 26, 51]. The involvement of QS in chemical defense strategies is particularly likely as effective toxin concentrations can only be reached when producers are highly abundant. While multiple QS systems have been described for other Pseudomonads, only a single system has been identified in P. putida KT2440 so far [52, 53].
Mutation rates are conditional on stress
The emergence of mutations resulting in the filamentation of P. putida was assumed to be conditional on a high ambient concentration of bacterial metabolites. The latter was considered as a proxy for bacterial stress which can affect mutagenesis either directly or indirectly by a variety of mechanisms [54,55,56]. Without this assumption, the almost synchronous appearance of filaments in all replicates at a late point in time would be very difficult to explain. Specifically, if mutation frequencies were high, filaments would become the predominant phenotype early (Fig. S3) which contradicts observations. On the other hand, if frequencies were low but unconditional, the timing of filament appearance should vary between replicates, which is in contrast to observations either (Fig. 1B).
Filamentation is associated with a fitness cost
Measurements of growth rate constants revealed a significant fitness disadvantage of filamentous isolates in comparison to single-celled, undefended isolates (p < 0.05, Wilcoxon rank sum test). Under the given experimental conditions (wheat grass medium, 19 °C) the growth rate of filaments was 5% lower than the growth rate of single-celled, undefended competitors.
Filaments do not produce metabolites at toxic levels
When flagellates were exposed to sterile filtrate harvested from cultures in the filamentous stage, no inhibition of growth was observed (Figs. 3 and 4). Owing to this empirical result and the questionable benefit of a redundant defense, the model does not consider metabolite production by filaments.
Filaments undergo asymmetric division
Upon growth, bacterial filaments are assumed to divide regularly at their tips [57]. Specifically, the model is built on the assumption that elongation occurs with a certain probability p while single-celled, undefended offspring are produced with probability 1-p. Division is obviously necessary for filaments to persist in liquid cultures subject to continuous dilution.
The model was able to qualitatively and quantitatively reproduce all major patterns of the observed predator-prey dynamics (Fig. 5), confirming that the above set of assumptions are a plausible explanation for the observed experimental dynamics. In particular, the model correctly predicted the shift from toxin production, resulting in a high-amplitude cyclic dynamic, to a stable steady state where filamentous bacteria are the predominant phenotype and toxin production has been abandoned. But the real strength of the model lies in the opportunity to analyze the specific contributions of individual processes which can hardly be disentangled from empirical observations alone. The simulations offer an explanation for the persistence of high predator numbers in a world of oversized, inedible prey: the flagellate numbers are mainly supported by the continuous release of single cells during asymmetric filament division (Fig. S4).
Colors indicate different initial values of flagellate abundance (see legend of Fig. 1).
The model also confirms that the filamentous genotype only emerges during the experiment and could not be just a result of selection from a heterogeneous inoculum. Any tiny seed of mutants provided to the model would result in the early predominance of filaments in simulated co-cultures (Fig. S3). Furthermore, the model predicts the emergence of filamentation as the final bacterial defense strategy, even when it was assumed to come at a much higher cost than toxin production (Fig. S5) and despite this strategy being far less efficient from a population point of view (Table 6). This counter-intuitive result can be explained by the fact that selection acts at the level of individuals and does not necessarily maximize population fitness [58]. Thus, the ultimately superior strategy is determined by the fitness of filaments in comparison to the fitness of toxin-producers in a mixed population. This, in turn, depends on the relative costs and benefits of each defense strategy. The model results make it clear that filamentation is the superior strategy due to its relative benefit, i.e. the level of grazing protection (Fig. S6). Excretion of toxins represents a concerted action of the bacterial community mediated by cell-to-cell signaling which benefits the population as a whole, acting as a public good [17]. In contrast, filamentation provides grazing protection only to the individuals carrying this trait and cannot be exploited by non-filamentous individuals. This distinction is crucial for determining the outcome of selection: since filamentous bacteria benefit from both forms of defense, they always have a stronger grazing protection than toxin-producing bacteria, which can offset even relatively high costs.
In addition to its higher fitness, other factors are likely to further promote the persistence of filaments. First, owing to its genetic manifestation, the filament-based protection is unconditionally stable whereas induced defenses undergo reversal when trigger conditions are, even temporarily, no longer met. It has been suggested that inducible defenses may be particularly favorable precisely because they can be switched off when they are not needed, thus allowing the individual to economize on the costs of defense [59]. However, the reversible nature of the initial defense appears to be a disadvantage in our system, as it allows the recovery of the flagellates to high density and thereby sets the stage for the inedible filamentous bacteria to become dominant. Second, there is a self-stabilizing effect associated with the filamentous phenotype. In the late phase of the experiment, the majority of the available resources are consumed by the filamentous bacterial genotype (Fig. S4), preventing undefended single cells from reaching the critical density required to trigger the initial, community-driven defense once again.
Unlike individual-based strategies, community defenses based on public goods are vulnerable to “cheating” [17] in the sense that a sub-population emerges that benefits from cooperation but does not pay the costs. Siderophore production in Pseudomonas aeruginosa is a classic model system for resistance to cheating and the maintenance of cooperation [15, 60]. The results of these previous studies suggest that, in our experimental system, due to the lack of spatial structure which may promote the maintenance of cooperation [15, 17, 61], non-toxin-producing cheaters would eventually displace toxin-producers. However, additional model simulations showed that the observed breakdown of the community defense in our experiments is unlikely to be a result of cheating. Specifically, model simulations that incorporate the emergence of a cheater mutant indicate that, while such cheaters will inevitably rise to dominance and cause the collapse of cooperative toxin production, this will take far longer than the rapid dominance of filamentous bacteria (Fig. S7) because the fitness advantage of cheaters is relatively small. This finding provides a highly relevant new angle to the question of the maintenance of cooperation, which has been extensively investigated with a focus on resistance against cheating. Our results indicate that cooperation may be far more vulnerable to the evolutionary invasion of individual-level strategies that provide the same benefit.
Overall, the unique succession of bacterial defenses observed in our long-term co-cultures demonstrates that inducible traits and rapid evolution not only impact population dynamics but also interact with each other. In particular, our results demonstrate that an inducible defense can pave the way for an evolved defense, which in turn suppresses the original one.
Source: Ecology - nature.com
