Making a Pathogen? Evaluating the Impact of Protist Predation on the Evolution of Virulence in Serratia marcescens

Abstract

Opportunistic pathogens are environmental microbes that are generally harmless and only occasionally cause disease. Unlike obligate pathogens, the growth and survival of opportunistic pathogens do not rely on host infection or transmission. Their versatile lifestyles make it challenging to decipher how and why virulence has evolved in opportunistic pathogens. The coincidental evolution hypothesis postulates that virulence results from exaptation or pleiotropy, i.e. traits evolved for adaptation to living in one environment that have a different function in another. In particular, adaptation to avoid or survive protist predation has been suggested to contribute to the evolution of bacterial virulence (the training ground hypothesis). Here, we used experimental evolution to determine how the selective pressure imposed by a protist predator impacts the virulence and fitness of a ubiquitous environmental opportunistic bacterial pathogen that has acquired multidrug resistance: Serratia marcescens. To this aim, we evolved S. marcescens in the presence or absence of generalist protist predator, Tetrahymena thermophila. After 60 d of evolution, we evaluated genotypic and phenotypic changes by comparing evolved S. marcescens with the ancestral strain. Whole-genome shotgun sequencing of the entire evolved populations and individual isolates revealed numerous cases of parallel evolution, many more than statistically expected by chance, in genes associated with virulence. Our phenotypic assays suggested that evolution in the presence of a predator maintained virulence, whereas evolution in the absence of a predator resulted in attenuated virulence. We also found a significant correlation between virulence, biofilm formation, growth, and grazing resistance. Overall, our results provide evidence that bacterial virulence and virulence-related traits are maintained by selective pressures imposed by protist predation.

Keywords: bacteria; evolution; exaptation; opportunistic pathogen; predation; virulence.

Fig. 1.

Confirmation of grazing activity. a) Phase contrast image of T. thermophila after 24 h coculture with fluorescently labeled S. marcescens KZ19 in Neff media with 180 µg/mL spectinomycin, b) Cy5 filter image showing the E2 crimson–labeled S. marcescens KZ19 cells inside T. thermophila, and c) composite overlay of the phase contrast and fluorescent images. Images were taken at 20× on a Keyence BZ-X700 series all-in-one fluorescence microscope and overlayed in ImageJ.

Fig. 2.

Characterization of the mutations present in the evolved populations. a) Percentage of nonsynonymous, synonymous, intergenic, nonsense, and RNA mutations detected in each evolved population (ME1 to ME6 and PE1 to PE6) and compared with the percentage of each mutation type expected at random (expected). b) Number of observed parallel (gene- and site-specific) mutation events observed in ME, in PE, and in both ME and PE (ME–PE) populations compared with random expectation (expected vs. observed). Significantly more gene-specific and site-specific parallel mutations were observed than would be expected to occur by chance (χ² test, P < 0.0001). The percentage of KEGG (Kanehisa and Goto 2000) functional categories of c) all genes in the ancestral genome, d) all genes that contained mutations in each group (ME, PE, or ME–PE), and e) the genes in which parallel events occurred across populations. The functional categories of all genes that contained mutations in ME, PE, and ME–PE populations d) did not significantly differ from the representation of functional categories in the ancestral genome (χ² tests, P > 0.5). The functional categories of genes that contained parallel mutations in ME, PE, and ME–PE populations e) significantly differed from the representation of functional categories in the ancestral genome (χ² tests, P < 0.0001).

Fig. 3.

Frequency of mutations present in the evolved populations. a) Plot of all the detected mutations and their frequencies in each population highlighting the parallel “fixed” mutations. b) STRING v12 (Szklarczyk et al. 2023) predicted protein–protein interaction network containing seven of the ten genes in which we identified “fixed” parallel mutations. c) Heatmap showing the frequency of mutations in or adjacent (indicated by *) to the seven genes from the same protein–interaction network within each evolved population.

Fig. 4.

Genotypes and phenotypes of evolved isolates. a) Mutations present in each isolate grouped by nonsynonymous, synonymous, and promoter (intergenic mutation in a promoter region), indicating the mutation type (SNV, nonsense mutation, duplication, or deletion) and the gene (or gene upstream) affected. b) Mean growth rate of PE_i and ME_i isolates and the ancestral strain based on 600 nm OD readings every hour for 24 h in Neff media. Each point represents the mean of three biological replicates, and the error bars indicate the standard deviation. c) Mean number of colony-forming unit per milliliter of PE_i and ME_i isolates and the ancestor after 18 h of growth on Neff agar. Each point represents a biological replicate. d) Mean biofilm formation of ME_i and PE_i isolates and the ancestral strain based on 590 nm OD readings after 24 h of growth in Neff media. Four assays were performed in triplicate, and each point represents a biological replicate e) Mean grazing (predation) resistance of PE_i and ME_i isolates and the ancestral strain based on 600 nm OD readings of bacterial filtrate after 48 h in coculture with T. thermophila. Two assays were performed in triplicate, and each point represents a biological replicate. f) Percent death of honey bees 5 d after exposure to the ancestral, ME_i, and PE_i isolates (control group average percent death = 2.3%; see supplementary fig. S4, Supplementary Material online). Three replicate assays were performed with 5replicates of 20 bees each per isolate per assay (15 replicates total per isolate). Each data point represents a biological replicate of 20 bees. g) Percent cytotoxicity of the ancestor, ME_i, and PE_i isolates to murine macrophages (RAW264.7) based on LDH release. b to g) Significance was tested by comparing each evolved isolate with the others and the ancestor using one-way ANOVA with Dunnett's multiple comparisons test. *P < 0.01; **P < 0.001; ***P < 0.0001. Dashed lines represent the overall mean when combining all ME_i or PE_i isolates.

Fig. 5.

Correlations between mean a) biofilm production and grazing resistance, b) biofilm production and virulence in bees, c) grazing resistance and virulence in bees, and d) growth (colony-forming unit per milliliter) and virulence in bees. Dashed lines represent a simple linear regression, and the correlation strength and significance were tested using the Pearson correlation coefficient.