Evolutionary Trade-Offs Underlie the Multi-faceted Virulence of Staphylococcus aureus

Abstract

Bacterial virulence is a multifaceted trait where the interactions between pathogen and host factors affect the severity and outcome of the infection. Toxin secretion is central to the biology of many bacterial pathogens and is widely accepted as playing a crucial role in disease pathology. To understand the relationship between toxicity and bacterial virulence in greater depth, we studied two sequenced collections of the major human pathogen Staphylococcus aureus and found an unexpected inverse correlation between bacterial toxicity and disease severity. By applying a functional genomics approach, we identified several novel toxicity-affecting loci responsible for the wide range in toxic phenotypes observed within these collections. To understand the apparent higher propensity of low toxicity isolates to cause bacteraemia, we performed several functional assays, and our findings suggest that within-host fitness differences between high- and low-toxicity isolates in human serum is a contributing factor. As invasive infections, such as bacteraemia, limit the opportunities for onward transmission, highly toxic strains could gain an additional between-host fitness advantage, potentially contributing to the maintenance of toxicity at the population level. Our results clearly demonstrate how evolutionary trade-offs between toxicity, relative fitness, and transmissibility are critical for understanding the multifaceted nature of bacterial virulence.

Fig 1. Low levels of toxicity are associated with invasive S. aureus infections across two collections of clinical isolates.

a: Individual isolates from seven time points were assayed for toxicity and their genome sequenced. Early nose cultures (ENCs) from months 1, 2, 4, 6, and 8 were all highly toxic. Late nose culture (LNC) isolates at month 12 and bloodstream isolates following bacteraemia (late blood culture, LBC) were all significantly less toxic than the ENC isolates. b and c: The toxicity of 134 USA300 isolates were assayed, and those from bacteraemia were on average significantly less toxic (measured as percent T2 and THP1 cell death (b and c, respectively)) than from carriage or SSTI (p < 0.01 ANOVA). The medians are presented as horizontal bars, with the boxes and whiskers showing the 1st and 3rd quartile and interquartile ranges. To access this data, see S1 Data.

Fig 2. The identification of six novel toxicity-affecting loci.

Transposon mutagenesis was used to functionally verify the role of loci associated with toxicity by genome-wide association study (GWAS; PLINK). Mean of six replicates are presented, error bars represent the 95% confidence intervals. The mutation in agrB is provided as a negative control. To access this data, see S2 Data.

Fig 3. Toxicity levels and disease origin distributed across the USA300 collection.

The clinical origin (carriage = C, SSTI = S, or blood = B) and the toxicity (low = green, medium = orange or high = red) of 134 USA300 isolates are mapped onto a maximum likelihood tree based on their whole genome sequence.

Fig 4. Explaining the association between toxicity and infection severity.

A range of virulence assays were performed on a subset of 10 high- and 10 low-toxicity isolates from the USA300 collection to identify the biological feature that explains the observed inverse correlation between toxicity and disease severity. (A) The toxicity of the subset of isolates is provided for comparison. (B) The ability of the bacteria to invade EA.hy926 endothelial cells. (C) The ability of the bacteria to induce the release of neutrophil extracellular traps (NETs). (D) The release of proteases by the bacteria. (E) Resistance of the bacteria to human neutrophil defensin-1. (F) The ability of the bacteria to form biofilms. The p-value for the comparison between the high and low toxicity isolates was greater than 0.05 for all traits with the exception of toxicity. The medians are presented as horizontal bars, with the boxes and whiskers showing the 1st and 3rd quartile and interquartile ranges. To access this data see S3 Data.

Fig 5. The highly toxic isolates are less fit than the low-toxicity isolates in the presence of human serum.

We quantified the relative fitness of six high- and six low-toxicity isolates from the single patient collection (A), 10 high- and 10 low-toxicity isolates from the USA300 collection (B) and an Agr wild type and mutant isogenic pair (C) in BHI with and without 5% (vol/vol) human serum. The medians are presented as horizontal bars, with the boxes and whiskers showing the 1st and 3rd quartile and interquartile ranges. p-values are indicated in the text. To access this data, see S4 Data.

Fig 6. Evolution towards increased levels of virulence.

Two competing MRSA strains with different levels of toxicity are considered, where higher toxicity is assumed to increase the transition rate from carriage to infection (SSTI) but also increases treatment-induced clearance of infections. At baseline, we assume that both strains have equal propensities to cause severe bacteraemia (σ _l = σ _h), which causes the more toxic strain (red) to be outcompeted due to its faster, virulence-induced clearance rates and leads to the dominance of the less toxic strain (green) both at carriage + SSTI (a) and at bacteraemic stage (b). By considering a negative correlation between toxicity and bacteraemia (σ _l > σ _h), the more toxic strain gains a competitive advantage, leading to higher frequencies at carriage and SSTI stages (c). However, the higher propensity of the less toxic strain, in this case maintained in the population through mutational down-regulation from highly toxic strains, maintains its high frequency at the bacteraemic stage (d). See Methods for parameter values.