Obligate mutualistic cooperation limits evolvability

Experimental design

Consortia of auxotrophic E. coli genotypes, which previously evolved an obligate mutualistic cooperation²⁶, were used to determine how this type of interaction affects the ability of the participating individuals to respond to environmental selection pressures. To this end, two main experimental treatment groups were established. First, each of the two cooperative auxotrophs was grown as amino acid-supplemented monoculture (i.e. tyrosine and tryptophan, 100 µM each). Second, both genotypes were cocultivated in the absence of amino acid supplementation. A treatment, in which monocultures were cultivated in the absence of amino acid supplementation was not included, because auxotrophic genotypes would not grow under these conditions. Also, an amino acid-supplemented coculture was not implemented in the experimental design, because competition between both auxotrophs was likely to result in a loss of one of the two genotypes (Supplementary Fig. 1). Moreover, previous experiments showed that amino acid supplementation does not completely abolish the mutualistic interaction. Hence, the experiment compared monocultures with externally provided amino acids (i.e. no mutualism) to cocultures, which could only grow when strains reciprocally exchanged amino acids (i.e. mutualism). Replicate populations of both treatment groups were serially propagated while being subject to a stepwise increasing concentration of one of four different antibiotics (i.e. ampicillin, kanamycin, chloramphenicol, and tetracycline) (Fig. 1). These four antibiotics differed in their mode of action. In this way, not just the effect of a single stressor was probed, but rather the ability of mutualistic consortia to adapt to environmental stress in general.

Ancestral consortia differ in their growth levels and susceptibility to environmental stress

Before the actual evolution experiment was performed, both growth levels and susceptibility to environmental stress was determined in the ancestral consortia. Comparing the maximum growth rate and densities populations achieved after 72 h revealed that unsupplemented cocultures grew significantly slower (Benjamini–Hochberg correction: P < 0.05, Supplementary Fig. 2a and Supplementary Table 1) and to a lower density (Benjamini-Hochberg correction: P < 0.05, Supplementary Fig. 2b and Supplementary Table 1) than both monocultures and cocultures that have been supplemented with amino acids (Supplementary Fig. 2).

Next, the MIC of all four antibiotics was quantified for unsupplemented monocultures as well as for both mono- and cocultures, to which amino acids have been supplied. The results of this experiment confirmed consistently for all four antibiotics tested that the MICs differed significantly between treatment groups (Supplementary Fig. 3). While unsupplemented cocultures were least resistant to ampicillin and kanamycin of all four groups analysed, monocultures of the tryptophan-auxotrophic genotype showed the lowest resistance levels when the two antibiotics chloramphenicol and tetracycline were considered (Benjamini–Hochberg correction: P < 0.05, Supplementary Fig. 3c, d and Supplementary Table 2). In addition, amino acid supplementation caused a significant increase in coculture resistance levels (Benjamini–Hochberg correction: P < 0.05, Supplementary Fig. 3 and Supplementary Table 2) with the magnitude of this effect depending on the type of antibiotic considered. These results show that both the genotype (i.e. the identity of auxotrophy-causing mutation) and amino acid supplementation affected growth and antibiotic resistance levels of ancestral populations, thus providing a baseline for the results of the evolution experiment.

Obligate mutualistic cooperation limits the ability of strains to adapt to environmental stress

Analysing changes in population densities (OD_600nm) of both mono- and cocultures throughout the evolution experiment indicated that the presence of antibiotics in the growth environment had a stronger growth-reducing effect on obligate mutualistic cocultures than on monoculture controls, which were able to grow independently (Fig. 2a, b and Supplementary Fig. 4a, b). The only exception to this pattern was the monoculture of the tyrosine auxotroph that nearly went extinct upon treatment with ampicillin (Supplementary Fig. 4c).

To further analyse differences between treatment groups, the survival of cultures in the evolution experiment was compared by applying log-rank tests for each pair of cultures. This test revealed significant differences between mutualistic consortia and monoculture controls (log-rank test: P < 0.001, Fig. 2c, d, Supplementary Fig. 4c, d and Supplementary Table 3), suggesting that mutualistic cocultures were more likely to go extinct than monocultures of auxotrophs. Upon reaching sub-MIC levels after the sixth transfer, a clear difference between the two bactericidal antibiotics (ampicillin, kanamycin) and the bacteriostatic agents (chloramphenicol, tetracycline) emerged. While bactericidal antibiotics drove all cooperative cocultures to extinction (Fig. 2c and Supplementary Fig. 4c), a subset of all cultures treated with bacteriostatic agents survived until the end of the antibiotics ramping experiment (Fig. 2d, Supplementary Fig. 4d). In all four antibiotics used, cocultures showed a significantly increased death rate compared to monocultures of both tryptophan- and tyrosine-auxotrophic genotypes (log-rank test: P < 0.001, Fig. 2c, d, Supplementary Fig. 4c, d and Supplementary Table 3). When the two monocultures were statistically compared with each other, a significant difference was only observed for ampicillin-treated cultures, but not the groups treated with the other three antibiotics (log-rank test: P < 0.001, Supplementary Fig. 4c and Supplementary Table 3).

Finally, an unsupervised learning algorithm was applied to identify differences and similarities in the evolutionary trajectories of monocultures and cocultures. For this, changes in cell densities over time were used to statistically compare the different experimental groups. This analysis revealed in all cases the emergence of clusters that almost exclusively consisted of coculture replicates (Monte-Carlo resampling after n = 10⁶ permutations: P < 10⁻⁶, Fig. 2e, f, Supplementary Fig. 4e, f). This observation suggests cocultures followed a distinct evolutionary path that was significantly different from the trajectories of monoculture controls. For the monocultures treated with kanamycin, chloramphenicol, and tetracycline, the algorithm consistently detected clusters composed of a mixture of tryptophan and tyrosine auxotrophic monocultures (Fig. 2e, f and Supplementary Fig. 4f), while in the case of ampicillin, a clearly separated set of two clusters emerged (Monte-Carlo resampling after n = 10⁶ permutations: P < 10⁻⁶, Supplementary Fig. 4e). Taken together, these results clearly show that mutualistic cooperation limits the ability of obligate mutualisms to adapt to environmental selection pressures.

Strain-level differences cause the increased susceptibility of cooperative consortia to environmental stress

As a next step, we asked whether or not the reduced ability of cooperative consortia to cope with antibiotic-mediated selection was due to differences between individual genotypes. To address this issue, the MIC of the focal chloramphenicol- or tetracycline-treated mono- and cocultures was determined for both ancestral and derived consortia. Subtracting the MIC values of ancestral populations from the ones achieved by their derived counterparts provided a measure of how much this parameter has changed over the course of the evolution experiment. Even if the concentrations of antibiotics tested exceeded the levels strains have experienced in the original experiment, differences in the resistance of populations are indicative of their evolutionary potential: reaching a higher MIC requires mutants that are able to survive under these conditions.

The results of this analysis revealed that the increase in resistance of consortia that coevolved in unsupplemented medium was significantly lower than the one of monocultures that have evolved in the presence of amino acids (Benjamini–Hochberg correction: P < 0.05, Fig. 3 and Supplementary Table 4). The only exception to this was the case of the monoevolved tyrosine auxotroph, whose MIC for tetracycline increased to a significantly lower extent than was the case for the corresponding cocultures (Fig. 3b and Supplementary Table 4). Next, the MIC values from isolated individual genotypes of coevolved consortia, cultivated in the presence of the required amino acid, were compared to the ones of amino acid-supplemented cocultures. This analysis showed for both antibiotics that the MIC values of the coevolved tyrosine auxotroph increased much less than the one of the corresponding coculture, while no such difference could be detected in the case of the tryptophan auxotrophic mutants (Fig. 3a,b). Notably, the positive effect of amino acid supplementation on antibiotic resistance levels, which was observed in ancestral cocultures (Supplementary Fig. 3), disappeared over the course of the experiment in both the chloramphenicol (Fig. 3a) and the tetracycline-treated group (Fig. 3b). This finding indicates that the metabolic intertwining between cocultured auxotrophs has tightened and that they became less able to use environmentally available amino acids.

Finally, comparing how resistance levels changed over the course of the evolution experiment between monocultures of auxotrophs that did or did not evolve as part of a mutualistic consortium revealed for all four cases analysed a significantly reduced increase of the MIC in coevolved strains relative to the corresponding monoevolved cultures (Fig. 3). The consistent response of auxotrophic strains in this experiment likely explains the reduced resistance levels coevolved consortia reached when treated with antibiotics (Fig. 3). Together, these results demonstrate that differences in the ability of individual strains to adapt to environmental stress limited the survival and thus evolvability of the entire consortium.

Previous coadaptation to environmental stress limits the evolvability of mutualistic consortia

Given that obligate metabolic cooperation can constrain adaptive evolution (Figs. 2 and 3 and Supplementary Fig. 4), we asked how a previous coevolutionary history in a stressful environment affects the ability of a mutualistic consortium to adapt to the same environmental challenge. We hypothesised that a shared coevolutionary history should enhance the ability of derived cocultures to cope with a previously experienced environmental stress when individuals interact with each other, yet limit their ability to tolerate increased stress levels when the otherwise obligate interaction is experimentally uncoupled. To test this hypothesis, pairwise consortia consisting of either the two coevolved auxotrophs or the two corresponding monoevolved genotypes that have been previously exposed to the two antibiotics tetracycline and chloramphenicol were again subjected to continuously increasing concentrations of the same two antibiotics. This time, however, the experiment was performed by cultivating both coevolved and monoevolved strains as cocultures (i.e. CO_EVO and CO_AUX, Supplementary Table 5) in both the absence and presence of environmentally-supplied amino acids. This experimental design allowed us to experimentally disentangle the effect of a shared coevolutionary history (i.e. coevolved versus monoevolved) from effects emanating from the interaction itself (i.e. with versus without environmentally-supplied amino acids).

This experiment showed that in the absence of amino acid supplementation, coevolved cocultures of obligately cooperating genotypes were significantly better able to cope with the antibiotic to which they have been previously exposed (linear mixed model for chloramphenicol and tetracycline: P < 0.001, Fig. 4a, d and Supplementary Table 6) than cocultures of strains that previously had adapted individually to the corresponding antibiotic (CO_AUX, Supplementary Table 5). However, when the obligate dependence between bacterial partners was relaxed by externally providing the required amino acids, the observed pattern changed to the opposite. Under these conditions, individually evolved strains reached significantly higher population densities than coevolved strains (linear mixed model for chloramphenicol and tetracycline: P < 0.001, Fig. 4b, e and Supplementary Table 6). In other words, coevolved cooperators were better off when survival depended on a metabolic interaction between both strains, while monoevolved strains had an advantage when the need to interact was experimentally removed.

Statistically comparing the growth response of consortia of coevolved genotypes when cultivated in the absence and presence of amino acids suggested that coevolved bacteria did not benefit from an external supplementation with amino acids (linear mixed model for chloramphenicol: P = 0.067 and for tetracycline P = 0.74, Fig. 4 and Supplementary Table 6). In contrast, the growth of individually evolved auxotrophs in the presence of elevated antibiotic concentrations strongly increased upon amino acid supplementation (linear mixed model for chloramphenicol and tetracycline: P < 0.001, Fig. 4 and Supplementary Table 6). A control experiment, in which all experimental groups were cultivated without any antibiotic treatment, showed a similar pattern, yet with less pronounced effects (Fig. 4c, f). In the absence of amino acid supplementation, both cocultures reached comparable cell densities, while growth levels of cocultures of monoevolved strains exceeded the ones of coevolved genotypes when amino acids were externally supplied (Benjamini–Hochberg correction: P < 0.05, Fig. 4c, f and Supplementary Table 6). Together these findings imply that coevolution curtailed the ability of strains to exist outside the interaction. This result is consistent with the interpretation that an increased dependence among genotypes coupled their evolutionary fate, thus limiting their evolvability.

Increasing environmental stress can destabilise obligate mutualistic cooperation

In situations where it is costly to obligately interact with another individual, for example when growth and survival depends on the amount of amino acids that the corresponding interaction partner provides, natural selection should favour individuals that evolve metabolic independence. To assess whether this also happened in the course of the antibiotics ramping experiment, the population-level proportion of mutants that evolved metabolic autonomy (i.e. reverted to a prototrophic phenotype) was quantitatively determined. For this, terminal populations of both coevolved and monoevolved auxotrophs, which have been exposed to increasing concentrations of bacteriostatic antibiotics, were randomly chosen to assess the population-level fraction of reverted phenotypes. This screening revealed that a subpopulation of the initial tyrosine auxotrophic genotypes regained the ability to grow without tyrosine supplementation. Interestingly, the rate of phenotypic reversion was significantly increased in strains that evolved in coculture relative to the respective monocultures (two-sided Pearson χ2 test for chloramphenicol TYR revertants: P < 0.001, χ² = 110.63, df = 1 and for tetracycline TYR revertants: P < 0.001, χ² = 119.18, df = 1, Fig. 5). In contrast, no revertants were detected among 540 screened colonies derived from 60 populations of monocultured and cocultured tryptophan auxotrophic strains. Particularly striking was the observation that no tryptophan-auxotrophic mutants could be detected in any of the tetracycline-treated cocultures (Fig. 5). This, indicates that the newly evolved phenotypic revertants either outcompeted their interaction partner or simply outnumbered them (lower detection limit: 2.5 × 104 cells ml⁻¹). Together, these results confirm that under conditions that increase the cost of mutualistic cooperation, natural selection will favour autonomous types that abandon the obligate interaction. As a result, cooperative interactions are lost from populations, thus favouring metabolically autonomous types.

Source: Ecology - nature.com