Coevolutionary history of predation constrains the evolvability of antibiotic resistance in prey bacteria

Abstract

Understanding how the historical contingency of biotic interactions shapes the evolvability of bacterial populations is imperative for the predictability of the eco-evolutionary dynamics of microbial communities. While microbial predators like Myxococcus xanthus influence the frequency of antibiotic-resistant bacteria in nature, the effect of adaptation to the presence of predators on the evolvability of prey bacteria to future stressors is unclear. Hence, to understand the influence of the coevolutionary history of predation on the evolvability of antibiotic resistance, we propagated variants of E. coli, pre-adapted to distinct biotic and abiotic conditions, in gradually increasing concentrations of antibiotics. We show that pre-adaptation to predators limits the evolution of a high degree of antibiotic resistance. Moreover, lower degree of resistance in the evolved strains also incurs reduced fitness costs while preserving their ancestral ability to resist predation. Together, we demonstrate that the history of biotic interactions can strongly influence the evolvability of bacteria.

Introduction

Predicting future evolutionary trajectories of a genotype based on past adaptations has been of immense interest to both ecologists and evolutionary biologists^1–4. Evolutionary outcomes are strongly shaped by historical contingency^5–7. Accumulated mutations during adaptation to the local environment can both negatively and positively influence the emergence and maintenance of further traits that incur considerable fitness costs^8–10. However, disentangling the effects of evolutionary history on current adaptation is difficult, because it requires an assessment of the evolutionary past.

In recent years, laboratory evolution experiments have gained immense prominence for studying the effects of evolutionary history^5,7,11,12. The tractability of laboratory evolution experiments offers a distinct advantage because of the ease with which evolutionary history can be correlated with the ability to evolve in response to future challenges. Many studies have demonstrated the effects of evolutionary history on the ability of bacteria, viruses, and yeast to evolve in another environment^5,12–15. While the effect of the abiotic environment on future evolutionary trajectories has garnered a lot of attention, the effects of microbial interactions such as cooperation, synergy, and predation have not been extensively explored.

Though only a small fraction of studies have focused on the effects of microbial interactions on the ability of microbes to evolve in new environments, they have demonstrated a strong influence of eco-evolutionary history on the evolvability of microbes^6,11,16,17. For example, coevolving cross-feeding bacterial partners when exposed to antibiotics after hundreds of generations of coevolution in the absence of antibiotics, show reduced adaptability to the presence of antibiotics¹⁸. Since microbes in nature live as a network of interacting cells (both cooperative and antagonistic), understanding the effect of microbial interactions on evolvability in novel environments, such as under antibiotic exposure, is important for our basic understanding of evolutionary processes and for informing our understanding of the problem of antibiotic resistance.

One of the most prevalent microbial interactions in nature is bacterial predation. Interestingly, it is known that the bacterial predator Myxococcus xanthus can influence the frequency of antibiotic-resistant bacteria in its vicinity¹⁹. Importantly, the influence of M. xanthus on the frequency of antibiotic-resistant microbes was demonstrated in antibiotic-free pristine conditions. Given the significance of M. xanthus in the maintenance of antibiotic resistance in nature, we asked whether microbes that have coevolved with M. xanthus are more or less likely to evolve resistance against antibiotics than bacteria that adapted only to abiotic conditions.

M. xanthus is a gram-negative soil bacterium that forms spore-filled multicellular fruiting bodies when starved^20,21. Additionally, M. xanthus is a prevalent bacterial predator in soil and kills other bacteria in its environment using both contact-dependent and contact-independent mechanisms^22–25. Interestingly, a majority of M. xanthus cells die and release diffusible antimicrobial substances during the formation of spore-filled fruiting bodies^26,27. The release of such growth-inhibitory substances is responsible for the enrichment of preexisting antibiotic-resistant bacteria in natural microbial communities¹⁹. Two aspects of M. xanthus' biology suggest that pre-adaptation to M. xanthus might affect the evolvability of prey bacteria in response to abiotic stressors like antibiotics. First, several antibiotic-sensitive microbes also share their environment with M. xanthus; therefore, it is possible that the presence of M. xanthus can also influence their evolutionary trajectories. Second, although the effect of M. xanthus on the de novo evolution of non-myxobacterial species is not clear in natural settings, in laboratory settings, M. xanthus predation has been demonstrated to act as a strong selection pressure^28–30. Thus, it is likely that in nature, M. xanthus predation serves as a strong selection pressure on non-myxobacterial species.

Here, we studied the influence of pre-adaptation to M. xanthus on the extent of evolvability of E. coli strain across two clinically relevant antibiotics, nitrofurantoin and gentamicin. We selected nitrofurantoin and gentamicin for our assays primarily because both are clinically used to treat E. coli infections. Additionally, because the two antibiotics have different modes of action, they serve as two distinct stressors in our study. Further, the effect of past exposure to M. xanthus predation on the evolvability of non-myxobacterial prey species is important for two reasons. First, M. xanthus is a prevalent soil bacterium that is likely to act as a strong selection pressure in natural communities^31–33. Second, given the effect of M. xanthus on the frequency of resistant bacteria, it is important to study whether pre-adaptation to M. xanthus predation alters the evolvability of antibiotic resistance in prey bacteria. This is especially important because bacterial predators (including M. xanthus) are gaining importance as potential therapeutic agents for microbial infections³⁴.

Since the evolvability of a genotype can be defined as its ability to generate heritable genetic and phenotypic variation, thus enabling it to adapt to novel environmental stressors^35,36, evolvability can be measured primarily across three distinct parameters: (1) survival in the presence of novel stressors^12,16, (2) adaptation to increasing or alternating stressors^36,37, and (3) growth in the adapted environment^15,16. Therefore, to understand the influence of the coevolutionary history with microbial predator, M. xanthus on the prey bacterium, E. coli, we tracked the survival of the prey populations over evolutionary time at an increasing concentrations of antibiotics. We further quantified the minimum inhibitory concentrations (MICs) for the terminally evolved populations as well as their survivability at high antibiotic concentrations to determine the extent of evolvability in preys with distinct evolutionary histories.

We selected three groups of E. coli strains: first, the wild-type, and the variants that were derived from the wild-type by evolving for ~300 generations under laboratory conditions, either in the presence (coevolved) or absence (mono-evolved) of M. xanthus. To test the extent of evolvability in these strains with distinct historical contingencies, these strains were further propagated at an increasing concentrations of antibiotics for ~350 generations. We demonstrate that a coevolutionary history with a predator influences both the survival percentages of populations as well as the degree of antibiotic resistance evolved. Our results show that mono-evolved clones, which were only pre-adapted to the abiotic media conditions, not only have a higher survival percentage (in terms of the number of populations that survive increasing antibiotic concentrations) but also evolve greater antibiotic resistance relative to both wild-type and coevolved isolates. We further demonstrate that the coevolved strains, despite evolving lower resistance than mono-evolved strains, evolve higher resistances relative to the wild-type strains. However, our results suggest that depending on the antibiotic treatment, high antibiotic resistance in the mono-evolved strains may incur a greater fitness cost, making them relatively less fit compared to wild-type and coevolved strains in direct 1:1 competition with their respective ancestors. Furthermore, despite propagation in a predator-free environment, we show that resistance to predation is maintained in the coevolved preys at similar levels as their ancestors. Together, we show that pre-adaptation to microbial predators can constrain the evolvability of prey bacteria when exposed to novel and increasingly severe environmental stressors such as rising antibiotic concentrations in the environment.

Results

Evolutionary history influences the survivability of E. coli at increasing concentrations of antibiotics

To study the impact of evolving in the presence of a predator on the evolvability of prey bacteria, we selected three groups of E. coli strains. First, we selected eight clones of the wild-type strain (lab isolate MG1655) of E. coli, which were neither adapted to the media conditions nor to the microbial predator Myxococcus xanthus. Second, we selected eight isolates of mono-evolved strains that were propagated in the absence of the bacterial predator M. xanthus. Third, we selected a set of eight isolates derived from E. coli populations that had coevolved in the presence of the bacterial predator M. xanthus. The mono-evolved and coevolved strains were adapted to the respective conditions for ~300 generations in a previous laboratory evolution experiment³⁰.

Eight strains were selected from each group with distinct evolutionary histories and further subjected to gradually increasing concentrations of two clinically relevant antibiotics. Each strain was used to initiate three biological replicate lines, both in an antibiotic-containing medium and in a control treatment without antibiotics, resulting in 48 populations; 24 populations each in the presence and absence of antibiotics. The 24 populations that were propagated in the presence of antibiotics experienced gradually increasing concentrations of antibiotics for 32 transfers, with transfers every 24 h (Supplementary Tables 1 and 2).

To initiate the experiment, we determined minimum inhibitory concentrations (MICs) for the respective ancestral strains (Supplementary Fig. 1). The evolution experiment began at an antibiotic concentration equivalent to 1/8th of the MIC of the respective ancestral strains at T0 (Transfer 0) (Supplementary Fig. 1), and the concentrations were then gradually doubled every day until they reached the MIC of their respective ancestors. Following this, the antibiotic concentration was increased by 2 μg/mL at each transfer (Fig. 1 and Supplementary Tables 1 and 2). The experiment was conducted using two clinically relevant antibiotics, nitrofurantoin and gentamicin, both of which are commonly used to treat of E. coli infections^36,38.

Fig. 1. Schematic of the evolution experiment performed to study the influence of pre-adaptation to predation on the evolvability of prey bacteria.

A Shows the evolutionary history of the three strains used to initiate the experiment. First, laboratory strain of E. coli MG1655 (Referred to as wild type strain). These strains were not pre-adapted to either the predator M. xanthus, or to abiotic experimental conditions. This strain was used to initiate the first evolution experiment. Second, mono-evolved strains that were derived from the wild-type strains and adapted to minimal media conditions for ~300 generations. And third, coevolved strains that were derived from the wild-type strains and were adapted to both minimal media conditions as well as to the presence of microbial predator, Myxococcus xanthus. B Eight strains were selected per group and were further propagated in minimal media for ~350 generations at an increasing concentration of antibiotics. The experiment was initiated with antibiotic concentrations that were 1/8th of the MIC of the respective strain and were gradually increased at every transfer each day. The concentrations were doubled till they reached the MIC value of the respective ancestors, and beyond that, it was increased by 2 μg/mL at each transfer. The experiment was performed across two different clinically relevant antibiotics, nitrofurantoin and gentamicin. C Representation of the increasing concentration of antibiotics over the evolutionary timeline. The graphs represent the increasing concentrations across two antibiotics, nitrofurantoin and gentamicin, for the wild-type group. This figure was created by the authors using Microsoft PowerPoint and edited using Adobe Illustrator.

During the evolution experiment, the concentration of nitrofurantoin and gentamicin was increased at each transfer. Thus, we expected that populations would go extinct. To track the extinction events and possible evolutionary rescue, we recorded growth at every transfer. Since pre-adaptation to media conditions has been shown to have variable effects on the evolution of resistance¹², we predicted that different treatments in our experiments might not show significant differences in overall population extinction frequency or evolutionary rescue. However, contrary to our expectations, the three groups of isolates displayed distinct evolutionary patterns.

We observed that pre-adaptation to abiotic conditions, i.e., minimal media, substantially improved the evolvability of E. coli in both antibiotic treatments. This was evident from the lower frequency of extinction events in populations initiated from mono-evolved strains compared to those from wild-type strains. In the nitrofurantoin treatment, populations established from mono-evolved isolates showed a lesser tendency to go extinct than those from coevolved group (Fig. 2A).

Fig. 2. Evolutionary history influences the percentage of the population that survives increasing concentrations of antibiotics over time.

Shows the percentage of the population that survived out of the 24 populations, from each wild-type (Wt), mono-evolved (M) and coevolved (C) group. A Shows the survival percentages of the mono-evolved and coevolved groups relative to wild-type (dotted line) in nitrofurantoin treatment. (Chi-squared test between either mono-evolved or coevolved groups relative to ancestor, p^M = 0.00022 and p^C = 0.49, n = 24.) B Shows the survival percentages of the mono-evolved and coevolved groups relative to wild-type (dotted line) in gentamicin treatment. (Chi-squared test between either mono-evolved or coevolved groups with respect to ancestors, p^M = 0.36 and p^C = 0.36, n = 24.) C, D Show the percentage of population, out of the 24 established lines from each, wild-type (green line), mono-evolved (blue line) and coevolved (orange line) groups that survived throughout the evolution experiment. C The percentage of the population that survived for each group at increasing concentrations of nitrofurantoin at each transfer. D The percentage of the population that survived for each group at increasing concentrations of gentamicin at each transfer.

We further demonstrate that, although the percentage of surviving populations derived from the mono-evolved strains was significantly higher than those derived from wild-type strains (Fig. 2A, Pearson’s χ² test, p = 0.0002), there were no significant differences between populations established by coevolved strains and wild-type strains (Fig. 2A, Pearson’s χ² test, p > 0.1). Overall, in the nitrofurantoin treatment, 11 populations survived in the mono-evolved group at the terminal transfer, whereas only 5 and 4 populations survived in the coevolved and wild-type groups, respectively. Furthermore, evolutionary rescue, i.e., recovery to detectable growth at increasing antibiotic concentrations, was higher in the mono-evolved group than in either the coevolved or wild-type groups (Fig. 2C and Supplementary Fig. 2A).

However, in the gentamicin treatment, all populations from both the mono-evolved and coevolved groups survived by the end of the thirty-two transfers, whereas only 21 out of 24 wild-type populations survived. Our results suggest no significant differences in extinction probability between populations derived from either mono-evolved (Fig. 2B, Pearson’s χ² test, p > 0.1) or coevolved groups (Fig. 2B, Pearson’s χ² test, p > 0.1) compared to wild-type (Fig. 2D and Supplementary Fig. 2B). Additionally, although some populations in the mono-evolved and coevolved groups briefly dropped below the detection limit, evolutionary rescue (i.e., recovery to a detectable growth) and further reaching an overall 100% survival, was relatively sooner in the mono-evolved populations (by 15th transfer) than in coevolved populations (by 26th transfer) (Fig. 2D and Supplementary Fig. 2B). However, since these differences were primarily observed in a single population, they may be stochastic. Together, we demonstrate that pre-adaptation to distinct biotic and abiotic factors can influence survivability and, thereby, evolvability in response to some, but not all antibiotics.

Predation history restricts the evolution of high-level antibiotic resistance in prey bacteria

To understand whether the evolutionary history has affected the evolution of resistance, we estimated the MIC for one representative clone from each surviving population. A generalized linear model was used to test the effects of evolutionary history (i.e., wild-type, mono-evolved, and coevolved), antibiotic treatment (nitrofurantoin and gentamicin), and interactions between them. This analysis revealed that evolutionary history, antibiotic type, and their interactions can influence the evolvability of antibiotic resistance in independent populations on distinct antibiotics to different extents (Supplementary Table 3). Moreover, the overall level of antibiotic resistance was higher in mono-evolved populations than in coevolved ones. For example, in the nitrofurantoin treatment, representative isolates from 7 out of 11 surviving mono-evolved group exhibited an MIC of 160 μg/mL (the median MIC evolved across all population in the nitrofurantoin treatment) or higher, whereas 1 out of the representative isolates from 5 surviving coevolved group exceeded an MIC of 160 μg/mL. In populations propagated in the presence of gentamicin, representative isolates from 17 out of 24 mono-evolved populations exhibited MIC of 140 μg/mL (the median MIC evolved across all population in the gentamicin treatment) or higher, compared to 13 out of 24 in the coevolved group (Supplementary Fig. 3).

Our results suggest that the proportion of lineages that survive increasing antibiotic concentrations differs across experimental blocks (wild-type, mono-evolved, and coevolved). Further, based on their respective evolutionary histories, the distinct lineages also evolved different degrees of resistance. However, the intra-population diversity in the derived populations in our experiments may also play an important role in determining their evolvability. Therefore, to test the influence of the history of evolution in the presence of predator on the intra-population phenotypic diversity, we selected one population at random from the surviving populations from each block across the two antibiotic treatment regimens (Supplementary Fig. 3A, B, selected populations are marked within red circles). We picked eight random clones from each of the selected population and measured their growth at an increasing concentration of antibiotics, starting from the last concentration at which they survived, i.e., the concentration of antibiotics at terminal transfer (T32) for respective treatments. While the MIC for all mono-evolved clones across both antibiotic treatments was beyond the detection limit (90 μg/mL for nitrofurantoin and 140 μg/mL of gentamicin), i.e., the highest antibiotic concentration tested (Fig. 3B, E), the MIC for coevolved (Fig. 3C, F) and wild-type clones (Fig. 3A, D) were both within our detection limit (For most coevolved clones: 90 μg/mL for nitrofurantoin, and 140 μg/mL for gentamicin. For wild-type: 86 μg/mL for nitrofurantoin, and 130 μg/mL for gentamicin). Thus, we show that for the populations tested, mono-evolved derived isolates adapted better to increasing concentrations of antibiotics than those from coevolved or wild-type strains.

Fig. 3. Evolutionary history of predation influences evolvability of antibiotic resistance in E. coli strains.

Data shown is the MIC of clones isolated from one of the surviving populations of each wild-type, mono-evolved and coevolved group that survived over the course of the evolution experiment. Here, the OD at 600 nm is shown against increasing concentrations of antibiotics. A The growth of clones derived from the wild-type group at an increasing concentration of nitrofurantoin, n = 6. B The growth of clones derived from the mono-evolved group at an increasing concentration of nitrofurantoin, n = 6. C The growth of clones derived from the coevolved group at an increasing concentration of nitrofurantoin, n = 6. D The growth of clones derived from the wild-type group at an increasing concentration of gentamicin, n = 6. E The growth of clones derived from the mono-evolved group at an increasing concentration of gentamicin, n = 6. F The growth of clones derived from the coevolved group at an increasing concentration of gentamicin, n = 6. The horizontal line represents the limit of detection. Different colors represent distinct clones respectively. Lighter color ribbons represent 95% confidence interval.

We also show that the MICs for all mono-evolved clones, i.e., the ones that were pre-adapted to minimal media conditions, were higher than those from coevolved clones. Moreover, all mono-evolved clones grew better than coevolved or wild-type clones at the highest concentrations of the respective antibiotic treatments (Fig. 4A, B, ANOVA between the groups, followed by SNK test) that were tested in our assays. However, although coevolved clones were also pre-adapted to similar abiotic conditions, they evolved relatively lower MICs than mono-evolved clones. Overall, we show that at the highest antibiotic concentration tested, the growth of mono-evolved clones was 3.37-fold and 1.82-fold higher than coevolved clones in nitrofurantoin and gentamicin treatment, respectively. Among coevolved clones, while two out of eight clones survived at the highest concentration checked (90 μg/mL) in the nitrofurantoin treatment (Fig. 4A, ANOVA between the groups, followed by SNK test), only one out of eight clones survived beyond the highest concentration checked (140 μg/mL) in the gentamicin treatment (Fig. 4B, ANOVA between the groups, followed by SNK test). Thus, we show that in populations tested, the coevolutionary history of predation influences the growth of prey bacteria at high concentrations of antibiotics and thereby restricting the evolution of a high degree of antibiotic resistance in prey bacteria.

Fig. 4. Evolutionary history of predation influences the survivability of prey at high concentrations of antibiotics.

Shows the growth of evolved E. coli clones from each of the three groups, wild-type, mono-evolved and coevolved at the highest concentration of antibiotics tested. Here the experimental treatments are clones isolated from populations that were propagated in the presence of antibiotics, and control treatments are clones that were isolated from populations that were propagated in the absence of antibiotics in similar abiotic conditions. A Growth (OD at 600 nm) of clones isolated from different treatment regimens at 90 μg/mL of nitrofurantoin. ANOVA between different treatments, p < 0.001 followed by SNK post hoc test, n = 6. B Growth (OD at 600 nm) of clones isolated from different treatment regimens at 140 μg/mL of gentamicin. (ANOVA between different treatments, p < 0.001 followed by SNK post hoc test, n = 6.). Each dot represents a distinct clone. The horizontal line represents the limit of detection. Different alphabets represent statistically different treatments whereas same alphabets indicate that there is no statistical difference between the respective treatments.

Further, across both the antibiotic treatments, the clones isolated from the wild-type block on average had the lowest MIC (86 μg/mL and 130 μg/mL for nitrofurantoin and gentamicin respectively) compared to both mono-evolved and coevolved groups (Fig. 3A, D). This is likely because of the different degrees of adaptation to abiotic conditions in the evolutionary history of the mono-evolved and coevolved clones, which may have reduced the effective cost of evolving resistance mechanisms. Finally, in the control treatment, in which populations were not exposed to antibiotics during the evolution, no drastic increase in MIC like seen in populations that evolved in the presence of antibiotics was observed (Supplementary Fig. 4A–F).

The evolution of antibiotic resistance in strains with a coevolutionary history with M. xanthus is less costly than in strains adapted only to abiotic conditions or the wild-type isolate

The de novo evolution antibiotic resistance is known to incur fitness cost^39–42. While these resistance mechanisms provide a survival advantage in the presence of antibiotics, in an antibiotic-free environment, the cost of resistance can result in sensitive variants outcompeting the strains with costly resistance phenotypes. However, compensatory mutations that increase other aspects of fitness can further aid in overcoming this cost of resistance, allowing resistant strains to be maintained even in an antibiotic-free environment^43,44. Since all evolved clones from all three blocks evolved manifold higher MICs than their respective T0 ancestors across both antibiotic treatments, we tested whether this high degree of resistance also imposed a fitness cost. For this, the competitive fitness of the evolved clones against their respective ancestors was measured when evolved strains were mixed with ancestors in 1:1 ratio. To do so, we selected four out of the eight representative clones from each of the representative populations. The competition experiments revealed that in the nitrofurantoin treatment, resistant clones from the wild-type derived and mono-evolved groups had similar fitness to their respective ancestors (Fig. 5A, One-sample t-test ^{FDR corrected}). Thus, nitrofurantoin resistance did not impose a significant fitness cost in these strains. However, resistant isolates from the evolved populations, which had a history of adaptation to predation, were 1.09-fold fitter than their respective ancestors (Fig. 5A, One-sample t-test ^{FDR corrected}). We further demonstrate that in an antibiotic-free environment, the competitive fitness of coevolved clones, relative to their ancestors, was 1.17-fold higher than the mono-evolved clones. Thus, in an antibiotic-free medium, the cost for nitrofurantoin resistance could not be detected among the evolved strains. However, evolved strains showed similar fitness as their respective ancestors when similar experiments were performed to test the cost of resistance in the subinhibitory concentrations of antibiotics i.e., 1/8th the MIC of their respective T0 ancestors (Fig. 5B, One-sample t-test ^{FDR corrected}). Together, our experiments demonstrate that the relative cost of resistance to high concentrations of nitrofurantoin depends on the environmental concentration of nitrofurantoin.

Fig. 5. High level of resistance may incur fitness costs.

Shows the fitness of evolved E. coli clones from each of the three groups, wild-type (Wt), mono-evolved (M) and coevolved (C) relative to their respective ancestors in the presence (B, D) and absence (A, D) of subinhibitory concentrations of antibiotics (1/8th T0 MIC). A Relative fitness of clones isolated from different nitrofurantoin treatment regimens in an antibiotic-free environment. (One-sample t-test, p^Wt = 0.865, p^M = 0.108 and p^C = 0.0174 against 1 showing differences from the ancestor. ANOVA between groups, p = 0.00727, followed by SNK post hoc test, n = 3). B Relative fitness of clones isolated from different nitrofurantoin treatment regimens in the presence of subinhibitory concentrations of nitrofurantoin in the environment. (One-sample t-test, p^Wt = 0.865, p^M = 0.865 and p^C = 0.158 against 1 showing differences from ancestor. ANOVA between groups, p = 0.395, followed by SNK post hoc test, n = 3). C Relative fitness of clones isolated from different gentamicin treatment regimens in an antibiotic-free environment. (One-sample t-test, p^Wt = 0.00408, p^M = 0.02 and p^C = 0.00408 against 1 showing differences from the ancestor. ANOVA between groups, p = 0.00213, followed by SNK post hoc test, n = 3). D Relative fitness of clones isolated from different gentamicin treatment regimens in the presence of subinhibitory concentrations of gentamicin in the environment. (One-sample t-test, p^Wt = 0.00124, p^M = 0.0154 and p^C = 0.00762 against 1 showing differences from an ancestor. ANOVA between groups, p = 0.00389, followed by SNK post hoc test, n = 3). Each dot represents a distinct clone. The horizontal dotted line represents the fitness, that is equal to the respective ancestors. The asterisks indicate significant difference from 1 i.e., fitness difference from their respective ancestors. Different alphabets represent statistically different treatments whereas same alphabets indicate that there is no statistical difference between the respective treatments.

Next, estimates of the fitness of the evolved clones revealed that mono-evolved derived, wild-type derived and coevolved derived isolates from the gentamicin treatment were all less fit than their respective ancestors in an antibiotic-free environment (Fig. 5C, One-sample t-test ^{FDR corrected}). Moreover, the mono-evolved clones, which evolved the highest degree of resistance (Figs. 3E and 4B), exhibited the lowest competitive fitness relative to their own ancestors, as compared to both the wild-type as well as coevolved groups (Fig. 5C, ANOVA between the groups followed by SNK post hoc test). Overall, the relative fitness of mono-evolved clones was 1.14-fold lower than wild-type clones, while it was 1.13-fold lower than the coevolved ones, in an antibiotic-free environment. Furthermore, the competitive fitness of evolved clones was lower even in the presence of subinhibitory concentrations of gentamicin i.e., 1/8th the MIC of their respective T0 ancestors (Fig. 5D, One-sample t-test ^{FDR corrected}). Although the presence of antibiotics in the environment is expected to confer a fitness advantage to the evolved clones relative to their ancestors, we did not observe this across either treatment regimen. This is likely due to the trace amount of antibiotics used in the competition experiment, which probably does not have a substantial effect on the growth of the ancestors to affect the outcome of the fitness experiment. Thus, we show that while mono-evolved clones evolve a very high degree of resistance, the de novo evolution of these resistance mechanisms incur high fitness costs, as demonstrated in the gentamicin treatment. However, if the resistance mechanism is extremely costly for the bacterium, as reported previously for nitrofurantoin^45,46, variants that managed to mitigate the fitness effects of costly resistance are selected over evolutionary time as the selection pressure, i.e., antibiotic concentrations, keep increasing.

Anti-predatory strategies in prey bacteria are retained even when propagated in the absence of a predator

Prey-predator coevolution results in an evolutionary arms race between the defensive and offensive strategies of prey and predator, respectively^47,48. Adaptation to the microbial predator, M. xanthus in synthetic laboratory communities results in the evolution of anti-predatory strategies in E. coli strains^29,30. Moreover, we have previously shown that such adaptation to a predator also results in increased competitive fitness of the prey strains relative to their respective ancestors when competing in the presence of M. xanthus³⁰. We have further shown that the mutational spectrum of these coevolved strains is distinct from that of both their ancestors, i.e., the wild-type strains, and the control mono-evolved strains, which evolved under similar experimental conditions but in the absence of predators. Here, we show that the evolutionary histories of the three distinct groups of E. coli strains significantly influence their ability to evolve under a novel environmental stressor such as the presence of antibiotics. However, it was unclear if adaptation to increasing antibiotic concentrations would further result in the loss of anti-predatory strategies in the coevolved strains when they were propagated in the absence of predators for ~350 generations.

To do so, we randomly selected 2 representative clones out of the 8 clones that were selected for further analysis from both the experimental and control regimens from each group across the two antibiotic treatments. With the selected clones, we performed predation assays using a laboratory strain of M. xanthus. We demonstrate that the anti-predatory strategies of all the evolved clones remain consistent relative to their respective ancestors. Further, this observation holds true for all clones, irrespective of either the degree of antibiotic resistance evolved or their evolutionary histories (Supplementary Fig. 5A, B, One-sample t-test ^{FDR corrected}). Even the clones from control lines in the coevolved group retained their ancestral levels of resistance to predation, thus showing that the anti-predatory strategies, once evolved, are somewhat stable even when the prey is propagated in the absence of a predator for hundreds of generations. Together, we show that adaptation to predation is mostly stable in the prey genome, and it influences the future evolutionary trajectories of the prey bacteria.

Discussion

Pre-adaptations to various biotic and abiotic factors influence the evolvabilities of bacteria^5,7,12,49. In natural communities, biotic interactions are abundant and play an overt role in shaping the ecological and evolutionary dynamics of the individuals within these communities^29,50,51. For example, bacterial predator M. xanthus is known to select for stress resistance mechanisms such as mucoidy in sympatrically coevolving prey bacteria and is also known to enrich pre-existing antibiotic-resistant microbes in natural soil communities. Thus, we hypothesized that pre-adaptation to M. xanthus is likely to affect the evolutionary trajectories to future stresses such as antibiotics. Here, we studied the influence of pre-adaption to M. xanthus on the evolvability of prey bacterium (E. coli) under the increasing concentration of antibiotics in the environment. To do so, we selected three groups of prey bacteria with differential evolutionary histories: First, wild-type strains with no prior adaptation to either abiotic condition or the predator, second, strains that were only pre-adapted to the abiotic experimental conditions (mono-evolved) and finally the strains that were pre-adapted to both abiotic conditions as well as predation by M. xanthus (coevolved). We demonstrate that pre-adaptation to abiotic conditions, results in the evolution of resistance to high concentrations of antibiotics. At the same time, the evolutionary history of predation restricts the de novo evolution of a high degree of resistance in prey bacteria in the tested populations. Importantly, we show that the patterns of evolved resistance can vary based on the mode of action of the antibiotic, where our results are more general across the surviving populations from distinct blocks in the nitrofurantoin treatment but show considerable variation across the evolved population in the gentamicin treatment.

Despite initiating each replicate population in the evolution experiments with an isogenic population of distinct strains, several mutations were likely to emerge and become fixed in the evolved prey population during coevolution with a predator. This resulted in the evolved population being a mixture of distinct genotypes, each of which could follow a unique evolutionary trajectory. Here, we show that a higher proportion of the strains (selected in this study) that were pre-adapted to only abiotic conditions (mono-evolved) evolved higher MICs during our experiments compared to the coevolved strains. Furthermore, within the selected populations, we observed clear differences in the evolvability of the coevolved and mono-evolved strains against two distinct antibiotics. Moreover, although individual derived populations in our experiments may show distinct patterns, our results provide an interesting proof of principle on how the history of biotic interactions, such as bacterial predation, can influence the evolvability of bacteria.

Microbial predators like M. xanthus are abundant in soil communities. Further, natural microbial communities can act as reservoirs for several pathogenic bacteria^33,52,53. Therefore, it is likely that most soil bacteria have, at some point in their evolutionary history, encountered and adapted to predation by M. xanthus. The anti-predatory strategies evolved in non-myxobacterial species can further influence their evolvabilities in an altered environment either positively or negatively. Here, we used antibiotics as a tool to determine the effect of the historical contingency of predation on the extent of evolvabilities in prey bacteria. Studying the influence of bacterial predation on the evolvability of antibiotic resistance becomes further important given the recent attention to microbial predators as either an alternative or supplementary therapeutic approach along with antibiotics³⁴. A detailed study of the interplay in the evolutionary dynamics, given the two distinct selection pressures, is essential to increase the precision in the predictability of emerging resistances.

De novo evolution of antibiotic resistance is known to influence the growth dynamics of bacteria, primarily because of either the segregation of resource allocation to facilitate the new resistance pathways or due to the downstream off-target effect of mutational alteration of genes targeted by the antibiotic^40,54,55. However, the gain of novel compensatory mutations elsewhere in the genome that enhances the growth of the individual under the respective abiotic conditions can effectively overcome the cost of resistance mechanisms^44,56. Here, we demonstrate that the populations originating from the mono-evolved variants, which were pre-adapted to the environmental conditions, had the highest total percentages that survived the increasing concentration of antibiotics (Fig. 2). In a previous study, we reported an 82 bp deletion in the rph gene consistent across the mono-evolved strains that were used to initiate the experiments in the current study. This 82 bp deletion has been earlier reported to provide growth benefits in minimal media⁵⁷. Moreover, relative to the wild-type clones, mono-evolved clones also evolve the highest degree of resistance across both antibiotics (Fig. 3B, E). Hence, we hypothesize that higher MICs in the mono-evolved clones are likely due to their adaptation to abiotic conditions that provide a growth advantage and thereby allow them to grow and survive despite the fitness costs incurred by novel resistance mechanisms.

Since both the coevolved and mono-evolved variants used in this study were derived from the lab strain of E. coli (wild-type) in a previous laboratory evolution experiment³⁰, the coevolved strains were also adapted to comparable abiotic conditions as the mono-evolved strains for a similar number of generations. However, during their course of evolution, the coevolved variants also adapted to the presence of the bacterial predator M. xanthus. We have previously shown that adaptation to predation in the lab-evolved synthetic communities included components of both intraspecies prey competition as well as anti-predatory strategies³⁰. Therefore, the total number of mutations accumulated in the genome of the coevolved clones was higher as compared to that of the mono-evolved clones. Overall, the average number of mutations per genome was 2.14-fold higher in the coevolved strains relative to the mono-evolved strains that were selected to initiate the evolution experiment. Since previous studies have already highlighted the role of existing mutational burden on the accumulation of newer mutations^10,58,59, we expected that the evolvability of the coevolved variants should be relatively lesser than the mono-evolved ones. This is simply because of the overall lesser diversity of accumulated mutations in the mono-evolved clones, which may allow them to explore a wider range of mutational landscapes. On the contrary, it is also possible for the existing mutations to have a positive epistatic effect by allowing certain novel mutations to be sustainable, which otherwise would have perished⁵. However, two aspects of our study make the latter scenario less likely; firstly, the antibiotics that were used in this study though of high clinical relevance, have a mechanism of action that is very different from the antibiotic primarily used by the lab strain of M. xanthus used against E. coli²², hence making cross-resistance less likely. Secondly, even if the pre-existing mutations allowed certain novel mutations to sustain, as the antibiotics concentrations keep increasing, the overall mutational load is more likely to restrict the sustainability of newer mutations, owing to the fitness costs and negative epistatic interactions between the existing and emerging mutations^60,61. Here, in line with our expectations, we show that compared to the mono-evolved clones, the coevolved clones had both lower percentage survivability (overall higher number of population extinction events were observed) (Fig. 2) as well as evolved lower MICs (Fig. 3).

Long-term adaptation to an antibiotic-free environment is known to alter the sensitivity of bacterial strains to different antibiotics and influence the evolvability of antibiotic resistance in an idiosyncratic fashion, as demonstrated by previously using the Lenski Long-term Evolution Experiment (LTEE) strains¹². While these strains were tested for their evolvabilities after ~50,000 generations in a constant environment, natural bacterial communities are expected to undergo a much shorter number of generations in such a constant environment. In line, with the previous observations, although we show that populations derived from ancestral strains with similar pre-adaptations (i.e., within the same group), might have an idiosyncratic evolutionary outcome, where some populations thrive, whereas others don’t, at the same concentrations of antibiotics, between groups the evolutionary outcomes are distinct in our experiments. Our results suggest that a coevolutionary history of predation restricts the evolvability of antibiotic resistance against clinical antibiotics. In contrast, pre-adaptation to the abiotic environment alone can allow the evolution of a high degree of antibiotic resistance. Further, the outcomes of the de novo evolution of resistance were influenced by the identity of the lineages and based on the antibiotics used. Together, we show that despite the idiosyncrasies within each group across the two antibiotic treatments, the evolutionary outcomes have some degree of predictability based on their distinct historical contingencies.

It is further expected that the gain of novel resistance mutations will incur fitness costs in an antibiotic-free environment. However, the effect of these costly mutations can be ameliorated by gaining compensatory mutations elsewhere in the genome^43,44. In this study, we used nitrofurantoin, as one of the antibiotics, which is routinely used to treat Urinary Tract Infections (UTIs)⁶². The mechanism of nitrofurantoin resistance, reported in both laboratory experiments as well as in studies using clinical isolates, includes sequential loss of function mutation in two oxygen-insensitive nitroreductases nfsA and nfsB^45,63. These genes are important in maintaining the redox state within the E. coli cells, thereby preventing ROS (reactive oxygen species) mediated damage to essential pathways⁶⁴. Therefore, loss of function in these genes has a high fitness cost, thus making gain of resistance to nitrofurantoin rather difficult. However, we show that the strains derived from the coevolved ancestor has a higher competitive fitness relative to its ancestor compared to the mono-evolved or the wild-type strains in an antibiotic-free environment (Fig. 5A). Interestingly, despite the high fitness cost of resistance, we do not observe a reduction in the competitive fitness of the evolved clones from either the wild-type or mono-evolved groups relative to their ancestor, in the nitrofurantoin treatment. This is most likely due to the gain of compensatory mutations. Although a deeper understanding of the specific mutations, that can completely nullify the fitness effect of such a costly resistance mechanism is interesting in the clinical context, it lies beyond the scope of this study.

Both epistatic interactions and pleiotropic effects have been shown to influence evolvabilities^58,65. The effects of past pre-adaptations can influence the ability of microbes to adapt to novel stressors both positively as well as negatively. Epistatic interactions of novel mutations with the prevailing mutations can positively influence the evolvabilities by allowing individuals to access new mutational landscapes that are accessible only in the background of specific existing mutations⁶⁶. It could further have a negative influence on the evolvabilities by constraining the overall accessible mutational landscapes¹⁰. Alternatively, it is also possible that antagonistic pleiotropic influence the evolvabilities, where a mutation that has beneficial effects in one background is deleterious in another⁶⁷. While several studies have now demonstrated the effects of existing mutational background on the emergence of novel mutations, it is imperative to understand the eco-evolutionary factors that impart fitness advantage to distinct genotypes and hence shape the evolvability of a genotype.

Although the influence of abiotic interactions on evolvabilities has been studied elaborately, the influence of historical contingencies of antagonistic interactions is relatively less explored. Since predation plays a major role in shaping evolutionary trajectories, a better understanding of the influence of predatory interaction on the evolvabilities of the prey is critical. Together, we have demonstrated that the coevolutionary history of predation can influence the survival, adaptability, and growth at an increasing concentration of antibiotics and thereby restrict the likelihood of the evolution of high antibiotic resistance in the prey bacteria. However, if such resistance evolves in the bacteria that has coevolutionary history with M. xanthus, then the resistance is generally associated with lesser fitness cost and thus might spread more rapidly than in the genetic background that evolved in the absence of M. xanthus.

Materials and methods

Strains and culture conditions

The wild-type E. coli MG1655 was acquired from Dr. Deepa Agashe’s laboratory at the National Centre for Biological Sciences, Bangalore, India, as described in earlier studies. Three distinct groups of E. coli strains were used for the evolution experiments. The wild-type and two additional variants were derived from the wild-type in an earlier experimental evolution conducted in the lab³⁰. In the previously conducted evolution experiment (graphically demonstrated in Fig. 1A), wild-type E. coli was grown in minimal media condition (M9 + 0.2% Glucose + 1.5% Bacto agar) either in the presence of M. xanthus (coevolved lines) or in the absence of M. xanthus (mono-evolved lines). Since M. xanthus cannot use glucose as a carbon source, in these experiments, the only available carbon source for M. xanthus was E. coli, which acted as a prey bacteria. After 4 days of incubation, 5% of the population was transferred to fresh media for both coevolved and mono-evolved treatments. Sixteen such transfers were done (~300 generations for E.coli in our experiments). From the terminally evolved population, E. coli clones were segregated from M. xanthus by dilution-plating the evolved population in LB soft agar (Luria broth + 0.5% Agar agar + 0.5% NaCl), supplemented with 0.5% NaCl. This is because M. xanthus is sensitive to high salt concentrations, thus, LB agar supplemented with NaCl acts as a selective media for E. coli. The isolated E. coli clones were phenotypically and genotypically characterized in an earlier study³⁰.

From the above-mentioned study, clones/strains of E. coli with two distinct evolutionary histories were selected and used for the experiments reported in this study. First, the ones that coevolved with M. xanthus for ~300 generations in minimal media conditions were labeled as coevolved strains. Second, the strains from the control lines that evolved under similar environmental conditions but in the absence of M. xanthus, were labeled as mono-evolved strains. Though both the strains evolved under similar conditions, either in the presence or absence of M. xanthus, and also originated from a common ancestor, during their evolution, different sets of mutations were accumulated in their genomes, respectively. Thus, they diversified from their respective common ancestors. The range of mutations on each of these strains has been reported previously³⁰. In this study, eight clones were selected from both the coevolved and mono-evolved groups respectively. The identities of the coevolved strains that were selected for this study were V1, V4, V5, V7, V9, V12, V14 and V16 from the vertically coevolved and vertically mono-evolved group of clones, respectively, from the earlier study³⁰. The M. xanthus strain used for the predation assay is the GJV1 strain, acquired from Dr. Gregory Velicer’s laboratory at ETH Zurich, Zurich, as described in earlier studies³⁰.

All the E. coli strains were grown in 1.2 mL minimal liquid media (1x M9, 0.2 mM MgSO₄, 0.1 mM CaCl₂ supplemented with 0.2% Glucose) in 2.5 mL 96 deep-well plates. The strains were incubated at 32 °C and 200 rpm for growing to mid-log phase before the experiments were set up. The OD was adjusted to 0.1 OD with TPM buffer (Tris-Cl Phosphate buffer, pH 7.6) prior to setting up all experiments.

Evolution experiment

The evolution experiment was set up using eight strains each, from each wild-type, mono-evolved and coevolved group. From each strain, three biological replicate lines were established hence giving a total of 24 population lines per group (eight strains with three replicates/strain). Increasing concentrations of two distinct clinically relevant antibiotics, nitrofurantoin and gentamicin were used for two different treatment conditions. Each treatment condition further consisted of six different blocks, wild-type, mono-evolved and coevolved blocks in the presence of increasing concentrations of antibiotics as well as in the control lines originating from the same ancestors, propagated in similar abiotic conditions but in the absence of antibiotics. To start the experiment 10 μL of 0.1 OD adjusted cultures were inoculated in 1 mL of liquid minimal media (1x M9, 0.2 mM MgSO₄, 0.1 mM CaCl₂ supplemented with 0.2% Glucose) containing a concentration of antibiotics that was equivalent to 1/8th the MIC of the respective T0 ancestor strains (Supplementary Tables 1 and 2), in 2.5 mL 96 deep-well plates. At every transfer, 100 μL populations were transferred to fresh 1 mL of liquid minimal media (1x M9, 0.2 mM MgSO₄, 0.1 mM CaCl₂ supplemented with 0.2% Glucose) every 24 h, while gradually increasing the concentration of antibiotics at each transfer and 32 such transfers were done. The concentrations of antibiotics were doubled till the concentration in the liquid media was equivalent to the MIC of the respective ancestral strain (Supplementary Fig. 1), and then incremented by 2 μg/mL at every transfer (Fig. 1 and Supplementary Tables 1 and 2). The cultures were incubated at 32 °C and 200 rpm. The increase in antibiotic concentrations followed the similar pattern across both antibiotic treatments and were contingents on the MIC of the respective ancestral strains in each group. and the populations were stored at regular intervals over the course of their evolution as glycerol stocks at −80 °C.

Percentage of survival and evolutionary rescue over time

The growth of the independent populations was tracked at every transfer by recording the OD of 100 μL culture at 600 nm, using Thermo Scientific Varioskan Lux plate reader. Populations at each transfer were classified as growing or non-growing based on the respective OD values, depending on whether the OD values were below or above the instrument detection limit. The percentage survival was calculated based on the fraction of the growing population out of the 24 population lines in each group across antibiotic treatments, at every transfer. A population was considered dead if the OD at a given transfer was below the detection limit. However, limitation of OD based measurement of growth is that it does not account for all viable counts of cell.

$$Percentagesurvival=\left(\frac{Numberofgrowingpopulations}{Totalnumberofpopulations}\right)\times 100$$

Clone isolation and MIC analysis

Out of the populations that survived the increasing concentration of antibiotics over the 32 transfers, one clone was taken as a representative clone from each population. The clone was picked by streaking the terminally evolved (T32) population on M9+Glucose hard agar (1x M9, 0.2 mM MgSO₄, 0.1 mM CaCl₂ supplemented with 0.2% Glucose + 1.5% Agar agar). The MIC (Minimum Inhibitory Concentration) for the clones were estimated by inoculating 10 μL of 0.1 OD adjusted cultures in 200 μL of media, supplemented with increasing concentrations of the respective antibiotics, i.e., nitrofurantoin or gentamicin. The experiments were performed in 96 well flat bottom microtiter plates and were incubated at 32 °C static conditions. Post 48 h incubation, the OD was recorded at 600 nm using Thermo Scientific Varioskan Lux plate reader. The MIC assays were done for all surviving populations of gentamicin and nitrofurantoin. However, out of the reported surviving population, only one wild-type population could be retrieved for clone isolation in the wild-type block of the nitrofurantoin treatment. The highest concentration that was checked for nitrofurantoin was 160 μg/mL and that for gentamicin was 260 μg/mL. For any clone showing growth until or beyond these concentrations, the highest concentrations of the respective antibiotics used was considered as the MIC value for analysis.

To further test the within-population phenotypic diversity, one population was chosen at random from the experimental blocks of each wild-type, mono-evolved (Population 4.2, which is derivative of lineage 4 for nitrofurantoin and population 7.3, which is a derivative of lineage 4 for gentamicin; Supplementary Fig. 3) and coevolved (Population 6.2, which is a derivative of lineage 6 for nitrofurantoin and population 1.3, which is a derivative of lineage 1 for gentamicin; Supplementary Fig. 3) for both the antibiotic treatments. Further, the sister populations arising from the selected populations in the control block were also selected for clonal isolation. The mono-evolved and coevolved populations that were selected from the nitrofurantoin treatment were derivatives of the CV1 and MV14 ancestor, respectively, while the coevolved and mono-evolved populations chosen from the gentamicin treatment were derived from the CV12 and MV7 ancestor, respectively. Since all the wild-type populations were derived from biological replicates of the same strain, they have a common ancestor. The selected populations were streaked on minimal hard agar media (1x M9, 0.2 mM MgSO₄, 0.1 mM CaCl₂ supplemented with 0.2% Glucose + 1.5% Agar agar), and eight isolated colonies were picked at random, grown till log phase in liquid media and stored at −80 °C, as clones. MIC assays were further performed with these clones as explained above.

Competition assay

To test the influence of the evolved resistances on the competitive fitness of the clones evolved from strains with differential historical contingencies, we selected 4 out of the 8 clones for which the MICs were determined from each experimental block per antibiotic treatment and competed them with their respective ancestral strains in 1:1 ratio either in the absence of antibiotics or in the presence of subinhibitory concentrations of antibiotics (1/8th of ancestral MIC). The assay was performed in 1 mL of liquid minimal media (1x M9, 0.2 mM MgSO₄, 0.1 mM CaCl₂ supplemented with 0.2% Glucose) with or without supplementation of the subinhibitory concentrations of antibiotics. Five μL of 0.1 OD adjusted cultures of the respective ancestors were mixed with 5 μL of 0.1 OD adjusted cultures of the distinct evolved strains. The experiments were performed in 96 deep-well plates and were incubated at 32 °C and 200 rpm for 24 h. The frequency of the ancestor and evolved clones was determined by dilution plating in LB soft agar (0.5% agar agar) with and without the respective antibiotics. Since MIC of evolved clones were manifold higher than their ancestors in both antibiotics, no additional markers were necessary to differentiate between the two and, hence, estimate the relative counts for each. The concentration of the distinct antibiotics used for the plating post-incubation of the competition experiment was 63 μg/mL nitrofurantoin and 60 μg/mL gentamicin, respectively. These concentrations were used since they were the concentrations of the terminal transfer that the populations survived for the respective antibiotics.

Since evolved resistance was used as a marker to distinguish the evolved strains from their respective ancestors, it is possible that some of the cells from the population of the evolved E. coli clones may lose resistance during the assay and thus be counted as an ancestor. Hence, we further performed productivity assays to check whether the plating efficiencies of the evolved clones were similar when plated in media containing the respective antibiotics as compared to when plated in an antibiotic-free medium. Our results demonstrate that while all clones isolated from the nitrofurantoin treatment had similar plating efficiencies (Supplementary Fig. 6A, One-sample t-test ^{FDR corrected}) in the media with and without antibiotic, the clones from mono-evolved gentamicin treatment had lower plating efficiency in the media supplemented with antibiotic (Supplementary Fig. 6B, One-sample t-test ^{FDR corrected}). However, the 0.019 decrease in the plating efficiency of the mono-evolved clones from the gentamicin treatment does not explain the fold-change recorded in the competition experiment (Fig. 5C, 0.205-fold decrease from 1). Hence, we demonstrate that the mono-evolved clones indeed has lower relative fitness compared to wild-type and coevolved clones across both the antibiotic treatments.

Productivity assay

To test the plating efficiencies of the evolved clones on the respective antibiotics, we selected the same 4 out of the 8 clones as selected for the competition assays, from each experimental block per antibiotic treatment. We inoculated 10 μL (0.1 OD) of each evolved clone in 1 mL of antibiotic-free liquid minimal media (1x M9, 0.2 mM MgSO₄, 0.1 mM CaCl₂ supplemented with 0.2% Glucose). The culture conditions were similar to that in the competition assay where the experiments were performed in 96 deep-well plates and were incubated at 32 °C and 200 rpm for 24 h. The plating efficiencies of the evolved clones was determined by dilution plating in LB soft agar (0.5% agar agar) with and without the respective antibiotics. The concentration of the distinct antibiotics used for the plating, and post-incubation for the productivity assay were 63 μg/mL nitrofurantoin and 60 μg/mL gentamicin respectively. These concentrations were used since similar concentrations were used for the plating in the competition assay.

Predation assay

To test for the anti-predatory strategies, in the evolved clones relative to their ancestors, 2 out of 8 clones were randomly selected from sister populations in each experimental and their control blocks across the two antibiotic treatments. For this assay, clones number 4 and 8 were selected from the coevolved block, 4 and 5 from the monoevolved block, and 2 and 6 from the wild-type block for nitrofurantoin treatment. Similarly, clones number 3 and 4 were selected from the coevolved block, 1 and 7 from the monoevolved block, and 1 and 8 from the wild-type block for gentamicin treatment. To set up the predation assay, the M. xanthus was grown in 8 mL of CTT liquid media⁶⁸ in 50 ml flask at 32 °C and 200 rpm. The predation assays were performed on 10 mL M9 hard agar beds (1x M9, 0.2 mM MgSO₄, 0.1 mM CaCl₂ supplemented with 0.2% Glucose and 1.5% Bacto agar) in 50 mL flasks. Fifty μL of 0.1 OD adjusted E. coli cultures were added along with 50 μL of 5*10⁹ cells/ml density adjusted M. xanthus cultures on the agar and spread evenly with the help of ~10 sterile glass beads. The cultures were incubated at 32 °C for 4 days under static conditions. Post-incubations, the cultures were harvested by the addition of 4 mL TPM buffer and shaking at 200 rpm for 30 min to dislocate the cells from the structured biofilm. The cultures were then diluted and plated on LB soft agar to obtain the E. coli counts. Since M. xanthus is sensitive to high salt conditions, LB media acts as a selective media for E. coli alone. The relative survival of the evolved E. coli strains was calculated relative to their respective ancestors when cocultured with M. xanthus.

Statistical analysis

All analyses were done using R version 4.2.2. Each experimental assays were performed in replicate blocks of three or more. Shapiro–Wilk test was done to determine normality of the data and Levene’s test was performed to check the homogeneity of variances. Appropriate statistical tests were performed wherever necessary. Details of all the statistical tests done are elaborately described in the respective figure legends.

Generalized linear model

We analyzed the MICs of 24 coevolved and mono-evolved clones each (one clone isolated per evolved population; see Fig. S3A for MICs) and eight clones of the wild-type for the gentamicin treatment. For, the nitrofurantoin treatment, we considered the MICs of clones isolated from each surviving population (see Fig. S3B for MICs). Using a generalized linear model, we tested whether the regimen and antibiotic treatment (independent variables) were predictors for the MICs evolved (dependent variable). In our statistical model, the regimen (evolutionary history: wild-type, mono-evolved, and coevolved) and the antibiotics were considered independent predictors of the resistance evolved, which was the response variable. The null hypothesis was that the predictors do not affect the response variable, i.e., that the coevolutionary history (regimen) and the identity of the antibiotics used (gentamicin/nitrofurantoin) do not affect the MICs evolved. However, a p value below 0.05 would suggest otherwise, indicating that both the regimen and the type of antibiotic have significant effects on the MICs evolved (see Supplementary Table ST3 for the table listing the outcome of GLM). We found that both the regimen and the antibiotic treatments have effects on the MICs evolved.

We used R version 4.2.2 to perform this analysis. The code for our analysis is also provided in the R script along with the dataset. We used a GLM with a Poisson error distribution and a log link function to account for the non-normal distribution of MIC data. The model included ‘regimen’ (evolutionary history: wild-type, mono-evolved, or coevolved) and ‘antibiotic’ (gentamicin or nitrofurantoin) as fixed effects, and their interaction. The following R code was used to run the analysis.

> x<-glm(formula=mic^~Regimen*Antibiotics, data=[dataset], family = poisson())

> summary(x)

Model significance was assessed using p-values derived from summary output of the model. GLMs are commonly used statistical tools to determine whether a certain predictor variable has a significant effect on a response variable^69,70.

Author contributions

S.P. conceived the project. S.P., S.S. and S.S.K. designed experiments. S.S., S.S.K. and VS performed the experiments and analyzed the data. S.S. performed the genomic analysis. S.S. wrote the first draft of the manuscript. S.P. edited the manuscript. All authors amended the manuscript.

Data availability

The primary data for this article and the original codes for the statistical analysis are available on Dryad. 10.5061/dryad.jm63xsjkg.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s44259-025-00111-5.