Evolution of increased virulence is associated with decreased spite in the insect-pathogenic bacterium Xenorhabdus nematophila

Abstract

Disease virulence may be strongly influenced by social interactions among pathogens, both during the time course of an infection and evolutionarily. Here, we examine how spiteful bacteriocin production in the insect-pathogenic bacterium Xenorhabdus nematophila is evolutionarily linked to its virulence. We expected a negative correlation between virulence and spite owing to their inverse correlations with growth. We examined bacteriocin production and growth across 14 experimentally evolved lineages that show faster host-killing relative to their ancestral population. Consistent with expectations, these more virulent lineages showed reduced bacteriocin production and faster growth relative to the ancestor. Further, bacteriocin production was negatively correlated with growth across the examined lineages. These results strongly support an evolutionary trade-off between virulence and bacteriocin production and lend credence to the view that disease management can be improved by exploiting pathogen social interactions.

1. Introduction

The degree of damage caused by a pathogen to its host, i.e. the pathogen's virulence, is a complex trait determined by various factors. Key among these are the social interactions between the pathogens co-inhabiting a host [1,2]. Pathogens exhibit myriad forms of social behaviours, such as cooperation and cheating [3,4]. How social behaviours affect virulence is assumed to depend on how a behaviour influences pathogen growth within the host [5]. While not always true, it is commonly assumed that faster growing pathogens are more virulent [6,7]. Therefore, social traits that promote faster pathogen growth, such as the production of cooperative public goods, may be positively correlated with virulence [8,9]. On the other hand, social behaviours such as spite that decrease pathogen growth may reduce virulence [10,11]. Importantly, if these relationships have an underlying genetic basis, then the evolution of virulence may affect the evolution of social traits as well.

A common social trait among bacteria is the production of anticompetitor toxins, called bacteriocins, which can kill closely related competitor strains without killing clone-mates of the producer cells [12–14]. Bacteriocin production is metabolically costly and many Gram-negative bacteria require cell lysis for bacteriocin release [12,15,16]. These negative fitness effects on both the producers and their sensitive competitors classify bacteriocin production as a spiteful trait [17]. The growth costs of bacteriocin production as well as the killing effect of bacteriocins may lead to reduced pathogen densities and thus reduced virulence. Theoretical models examining bacteriocin production assume such growth costs as well as within-host killing of bacteriocin-sensitive cells [6,11]. These models demonstrate that the conditions that favour increased bacteriocin production also favour reduced virulence [6,11].

Empirical studies have shown that bacteriocin-mediated competitive interactions can reduce virulence during mixed infections. For instance, during co-infections with a sensitive competitor strain, bacteriocin-producing Pseudomonas aeruginosa lineages result in lower pathogen densities and thus reduced virulence relative to bacteriocin non-producing mutants [11]. Similarly, within-host bacteriocin-meditated interactions among insect-pathogenic Xenorhabdus spp. lead to reduced virulence [10,18,19]. However, very few studies have examined the relationship between the evolution of spiteful traits and virulence evolution. Garbutt et al. [20] found that evolution in mixed infections can result in reduced virulence and increased competitive antagonism relative to evolution in single infections in the insect-pathogen Bacillus thuringiensis [20]. Specifically, the increased antagonism occurred without a significant increase in absolute growth rates, suggesting that the observed antagonism was likely bacteriocin-mediated. Nonetheless, it is currently unclear if levels of bacteriocin production can be directly affected by selection on virulence.

Here, we examine the evolutionary interplay between virulence, growth and spite in the insect-pathogenic bacterium Xenorhabdus nematophila. Using an experimental evolution approach, we ask whether the evolution of increased virulence is associated with changes in spiteful bacteriocin production. We also examine the growth of the evolved and ancestral lineages to investigate the underlying correlations between growth, virulence and bacteriocin production.

2. Methods

(a). Study system

Xenorhabdus nematophila is a species of insect-pathogenic bacteria that forms a mutualistic symbiosis with the nematode Steinernema carpocapsae [21–23]. Free-living juvenile nematodes carry the bacteria in a specialized receptacle of the intestine until they locate an insect host [22]. Within the insect, the nematode and bacteria separate, replicate, and quickly kill the insect host [24,25]. Bacteriocin-mediated interactions between co-infecting Xenorhabdus strains are known to occur within the insect host [26]. When host resources are depleted, nematodes re-associate with the bacteria and emerge to seek a new insect host [22].

(b). Bacterial strains

Experimental evolution was conducted in a previous study [27], where 16 independent lineages of X. nematophila derived from a single ancestral colony were passaged through Galleria mellonella caterpillar hosts. After 20 host passages, each bacterial lineage was injected into insect hosts and the mean time to host death was used as a measure of virulence (as shown in fig. 2a of [27]). Pairwise comparisons between each evolved lineage and the ancestor revealed that 14 of the 16 lineages showed significantly faster host-killing than the ancestor (figure 1a). These 14 evolved lineages were used in this study. The experimental evolution protocol is briefly described below.

Figure 1.

Increased virulence (a) of the 14 evolved lineages (light grey bars) relative to the ancestor (dark grey bar) is shown as a lower mean (±s.e.) time to host death. The evolved, more virulent lineages show (b) reduced bacteriocin production and (c) faster growth relative to the ancestor. Reduced bacteriocin production is shown as a lower mean (±s.e.) relative lag time, indicating that the bacteriocin extracted from the evolved lineages was less inhibitory than the bacteriocin extracted from the ancestor. Growth is shown as mean (±s.e.) OD₆₀₀ after 4.5 h culturing in vitro.

During every round of selection, each bacterial lineage was passaged by infecting 15 larval G. melonella caterpillars each with approximately 4000 colony-forming units (CFU). Depending on the treatment, the caterpillars were either infected with nematodes that carried the bacteria (M+), or directly injected with the appropriate bacterial dose by the experimenters (M−). Selection for faster host-killing (S+) was imposed by choosing only the fastest-dying insect host for subsequent passaging. By contrast, in the ‘S–’ treatment, a dead host was chosen at random. Emerging juvenile nematodes were used for subsequent passaging in the ‘M+’ treatments. For ‘M−’ treatments, bacteria were extracted from the dead caterpillar hosts, serially diluted, and plated on selective NBTA agar plates with 50 µg ml⁻¹ ampicillin (nutrient agar supplemented with 0.0025% (w/v) bromothymol blue and 0.004% (w/v) triphenyltetrazolium chloride, pH = 8). After 36 h of growth, 20 colonies were picked and suspended in 1 ml phosphate-buffered saline to repeat the infection process. A 20-colony slurry of the ancestral population was cryopreserved prior to experimental evolution, and similar 20-colony slurries of every experimental lineage were cryopreserved at the end of 20 host passages. Slurries were grown overnight in Luria–Bertani broth (LB) at 28°C, and preserved as 20% (v/v) glycerol solutions at −80°C. Additional details are available in the original paper [27].

(c). Bacteriocin extraction

Pure cultures of the ancestral and evolved lineages were grown in 5 ml LB until OD₆₀₀ = 0.5 at which point 0.5 µg ml⁻¹ mitomycin C was added to chemically induce bacteriocins. After overnight incubation at 28°C, bacteriocin extracts were collected by centrifuging cultures at 1620g for 5 min and filtering the supernatant through 0.45 µm filters. This procedure allows the bacteriocin to pass through while eliminating any cells in the extract. Bacteriocin extracts were collected from one colony for every evolved lineage, and four colonies of the ancestral population. These extractions, as well as the growth measurements and bioassays described below, were conducted in four experimental blocks, such that four evolved colonies and one ancestral colony were processed simultaneously.

(d). Growth measurements

The pure cultures used to extract bacteriocins were monitored for growth via periodic OD₆₀₀ measurements on a spectrophotometer. OD₆₀₀ values for all cultures after 4.5 h of growth were used to compare growth rates. This time point was chosen as it was after all cultures reached exponential phase (OD₆₀₀ > 0.2) and before mitomycin C was added to the fastest growing culture.

(e). Bacteriocin bioassay

The inhibitory activity of the bacteriocin extracts derived from the evolved and ancestral lineages were compared using a growth inhibition bioassay [28]. Briefly, a bacteriocin extract (or culture medium in the ‘no bacteriocin’ controls) and a starting culture of sensitive cells (approx. 10⁶ CFU ml⁻¹) were mixed in 1 : 5 ratio by volume. Each mixture was pipetted into four replicate wells on a 100-well plate and incubated in a Bioscreen optical plate reader at 28°C, with shaking at medium amplitude for 24 h (GrowthCurves USA). Control wells with un-inoculated medium were included to rule out contamination across the plate. OD₆₀₀ measurements were taken every 30 min. These readings were used to calculate the lag times of the cultures growing in each well with the software GrowthRates 3.0 [29].

The lag time values of the cultures growing in the presence and absence of bacteriocin extracts were used to derive a metric of bacteriocin activity we call ‘relative lag time’. The relative lag time is a ratio of the lag time of a sensitive culture growing in the presence of bacteriocin extracts and the lag time of the respective ‘no bacteriocin’ control. Since the metric is a ratio, a value of 1 indicates no bacteriocin activity was detected. For every bacteriocin sample, the growth inhibition bioassay was conducted across two distinct strains, Photorhabdus luminescens TT01 and Xenorhabdus bovienii MC19, which are both sensitive to X. nematophila bacteriocin.

(f). Statistical analyses

All statistical analyses were performed using SAS 9.4. An analysis of variance was performed to conduct pairwise comparisons of ‘mean time to host death’ between the ancestor and the evolved lineages using Proc Mixed with ‘lineage’ as a fixed effect. Bacteriocin activity (relative lag time) was compared using mixed-model analyses of variance with ‘treatment’ (ancestral or evolved), ‘sensitive strain’ (P. luminescens or Xenorhabdus bovienii) and their interaction as fixed effects; ‘lineage’ and ‘experimental block’ were included as random effects. Growth (OD₆₀₀ at 4.5 h) was analysed similarly, with treatment and experimental block as factors. Pearson's correlation coefficients between bacteriocin production, virulence and growth were determined using Proc Corr. To evaluate the significance of these correlations, while accounting for the repeated bacteriocin measures (i.e. tested against two sensitive strains), Proc Mixed was used with ‘relative lag time’ or ‘mean time to host death’ as the dependent variable and ‘OD’ as a fixed effect; ‘experimental block’ and ‘lineage’ were included as random effects.

3. Results

The evolved lineages used here exhibited significantly faster host-killing, and thus greater virulence relative to their ancestor (figure 1a). Bacteriocin production was tested against two sensitive strains using a growth inhibition bioassay. The evolved, more virulent lineages exhibited a significant decrease in ‘relative lag time’ (F_1,15 = 15.30, p < 0.001) compared with the ancestral population (figure 1b), indicating reduced bacteriocin production evolved over time. There was no significant effect of sensitive strain (F_1,15 = 0.03, p = 0.864) or interaction between treatment and sensitive strain (F_1,15 = 0.04, p = 0.839). Additionally, the evolved, more virulent lineages also show significantly higher OD₆₀₀ values after 4.5 h of growth (F_1,13 = 4.73, p = 0.049) relative to the ancestral strain (figure 1c). Across all strains, this metric of growth tended to be associated with host-killing (F_1,13 = 3.60, p = 0.08, figure 2a), consistent with the assumption that faster growing pathogens are more virulent. Moreover, relative lag time and growth were negatively correlated (figure 2b), with greater growth coming at the expense of lower bacteriocin activity (F_1,16 = 9.26, p = 0.008). Thus, the significant correlation between virulence and bacteriocin production (F_1,16 = 14.76, p = 0.001) is consistent with predictions based on their relationships with growth.

Figure 2.

Time to host death (a) and bacteriocin production (b) are negatively correlated with growth (OD₆₀₀ after 4.5 h culturing in vitro). Shorter time to host death indicates greater virulence. Bacteriocin production is shown as the relative lag time value exhibited by all evolved and ancestral lineages against both sensitive strains.

4. Discussion

Increasing evidence suggests that the virulence of pathogenic microbes may be affected by their social traits. Here, we examined how virulence, growth and a spiteful trait, bacteriocin production, are evolutionarily linked using experimentally evolved lineages [27] of the insect-pathogenic bacterium, Xenorhabdus nematophila. We found that the evolved, more virulent lineages (figure 1a) exhibited reduced bacteriocin production and faster growth in vitro relative to their ancestor (figure 1b,c). Further, bacteriocin production across all lineages showed a significant negative correlation with growth (figure 2b), demonstrating that bacteriocin production is costly at the population level. We also found that faster growing lineages tended to be more virulent (figure 2a). Overall these results strongly support an evolutionary trade-off between virulence and bacteriocin production, and highlight the importance of competitive social interactions in the evolution of pathogens [30,31].

It is important to note that the populations used in this study did not encounter any bacteriocin-sensitive competitors during the course of experimental evolution. In the absence of competitors, selection for the maintenance of bacteriocin production should be weak, which can facilitate the evolution of faster growth via reduced investment in bacteriocin production. We found that faster in vitro growth tended to be associated with greater virulence (figure 2a), suggesting that the observed trade-offs between virulence and bacteriocin production may be mediated by their opposite correlations with growth. However, some bacteria do not exhibit population-level growth costs of bacteriocin production [32]. In such systems, or in systems where virulence is not correlated with growth, bacteriocin production would not be expected to show a trade-off with virulence. Thus, our results support two key assumptions made in theoretical models: that bacteriocin production imposes growth costs, and that growth is positively correlated with virulence [6,11].

There has been considerable interest in exploring the relationship between virulence and social behaviours in pathogens for the possibility of exploiting it for therapeutic application [4,33,34]. Cooperative bacterial traits such as public goods production and quorum sensing [8,35] have received particular attention in this regard, as exploiting the ability of ‘social cheats’ to invade cooperative pathogen populations may reduce pathogen fitness and virulence [9,36]. The trade-offs between virulence and spite reported here, and as seen in reports of mixed infections [10,18,20], could suggest that maintaining selection for spite may help constrain the evolution of virulence. Akin to using ‘social cheats’ as therapeutic agents, perhaps the use of live competitors to maintain selection for bacteriocin production could help constrain pathogen growth and thus reduce virulence. Genetically engineered probiotics that are being developed for therapeutic use [37] may serve as potential candidates for such treatment strategies.

To conclude, we investigated how the evolution of increased virulence is associated with spiteful bacteriocin production and growth in insect-pathogenic X. nematophila. We found that the evolution of increased virulence is associated with reduced bacteriocin production, and faster growth in vitro. Such trade-offs between virulence and spite may offer novel opportunities to exploit social traits in pathogenic populations for disease management. These results provide some of the earliest direct insights into how the evolution of virulence and bacteriocin production are correlated and demonstrate how the social lives of pathogenic bacteria may be intricately linked with disease.

Data accessibility

All data and code for statistical analyses are available on the Dryad Digital Repository at: https://doi.org/10.5061/dryad.dn34vn2 [38].

Authors' contributions

A.B. co-conceived and co-designed the study, conducted the experiments and analyses and wrote the manuscript. V.C.T.D. conducted the experiments with A.B., helped with data analysis and commented on the manuscript. L.T.M. conducted experimental evolution, advised with statistical analyses and edited the manuscript. F.B. co-conceived and co-designed the study, helped with statistical analyses and with writing the manuscript. All authors approve the final version of the manuscript and agree to be held accountable for the content therein.

Competing interests

We declare we have no competing interests.

Funding

This study was supported by Indiana University via the College of Arts and Sciences and the Center for the Integrative Study for Animal Behavior and by the National Science Foundation DBI 1460949.