Common Adaptive Strategies Underlie Within-Host Evolution of Bacterial Pathogens

Abstract

Within-host adaptation is a hallmark of chronic bacterial infections, involving substantial genomic changes. Recent large-scale genomic data from prolonged infections allow the examination of adaptive strategies employed by different pathogens and open the door to investigate whether they converge toward similar strategies. Here, we compiled extensive data of whole-genome sequences of bacterial isolates belonging to miscellaneous species sampled at sequential time points during clinical infections. Analysis of these data revealed that different species share some common adaptive strategies, achieved by mutating various genes. Although the same genes were often mutated in several strains within a species, different genes related to the same pathway, structure, or function were changed in other species utilizing the same adaptive strategy (e.g., mutating flagellar genes). Strategies exploited by various bacterial species were often predicted to be driven by the host immune system, a powerful selective pressure that is not species specific. Remarkably, we find adaptive strategies identified previously within single species to be ubiquitous. Two striking examples are shifts from siderophore-based to heme-based iron scavenging (previously shown for Pseudomonas aeruginosa) and changes in glycerol-phosphate metabolism (previously shown to decrease sensitivity to antibiotics in Mycobacterium tuberculosis). Virulence factors were often adaptively affected in different species, indicating shifts from acute to chronic virulence and virulence attenuation during infection. Our study presents a global view on common within-host adaptive strategies employed by different bacterial species and provides a rich resource for further studying these processes.

Keywords: infectious diseases; microbial genetics; microbiology; within-host adaptation; within-host evolution.

Fig. 1.

Overview of the study. (a) Number of strains of each bacterial species in the database. (b) Fraction of strains in each species isolated from patients undergoing different clinical scenarios (infection/carriage). PA, Pseudomonas aeruginosa; TB, Mycobacterium tuberculosis; SA, Staphylococcus aureus; KP, Klebsiella pneumoniae; ST, Salmonella enterica subsp. enterica serovar Typhimurium; EC, Escherichia coli; EFci, Enterococcus faecium; HI, Haemophilus influenzae; CJ, Campylobacter jejuni; CD, Clostridioides difficile; AB, Acinetobacter baumannii. (c) Analysis pipeline for extracting genes undergoing either loss of function or any change (including loss of function) in a strain. We constructed assemblies for the different isolates using their respective sequencing reads and determined the reference genome closest to the isolates of each strain using QUAST. We utilized TRACE to assess the phylogenetic relations between the different isolates to extract progenitor–progeny isolate pairs (Materials and Methods). Breseq was then used to compare the sequencing reads of each isolate with the reference genome, and positions that differed from the reference genome and varied between the progeny and progenitor were determined. These variations were analyzed by SnpEff to define: 1) genes that underwent loss of function (based on high-impact variations appearing in the progeny isolate and not the progenitor isolate, leading, e.g., to deletion of a gene or to a stop codon); 2) changed genes (based on variations of moderate or high impact appearing in the progeny isolate and not the progenitor isolate, where moderate impact refers, e.g., to amino acid substitutions). The genes determined as undergoing loss of function make up a subgroup of the genes determined as changed.

Fig. 2.

Different species vary in the number of genes that underwent change or loss of function per strain. (a) Upper panel: Number of changed genes per strain in the data for each bacterial species. Lower panel: Number of genes that underwent loss of function per strain in the data for each bacterial species. Error bars represent standard error. Note that the group of genes that underwent loss of function is a subgroup of the changed genes. (b) The same, with the result for each species normalized by the number of proteins in its reference proteome, scaled to 1,000 coding sequences. Names of bacterial species are as in figure 1b. The number to the right of the bar represents the number of strains included for that bacterial species. CJ was excluded from (b) to maintain scale.

Fig. 3.

The same genes of a bacterial species repeatedly undergo change or loss of function during infection by different strains. Our statistical framework defines genes that undergo change or loss of function repeatedly in a statically significant number of strains as undergoing adaptive change or loss of function, respectively. (a) Upper panel: Boxplots of the fractions of adaptively changed genes out of all changed genes in the different strains of each species. The number above the bar represents the number of strains included for that bacterial species. Lower panel: The mean similarity between the lists of genes undergoing change or loss of function in different strains of each species as measured by the Mean Jaccard Index (Materials and Methods). The number above the bar represents the number of studies included for that bacterial species. (b) Upper panel and lower panel as in (a), for genes undergoing loss of function. For most strains, there is no relationship between the fraction of genes undergoing adaptive change/loss of function and the similarity between strains of a species. Names of bacterial species are as in figure 1b.

Fig. 4.

Adaptive loss of function is mostly mediated by frameshift mutations. (a) The number of adaptive loss of function events due to different loss of function mechanisms, normalized by the number of progenitor–progeny pairs in each species. (b) %GC in insertions or deletions leading to frameshift events underlying adaptive loss of function, normalized by the %GC of the entire genome of each species. Asterisks denote %GC that statistically significantly differs from the %GC of the entire genome by binomial test. *P ≤ 0.05, **P ≤ 0.01, and ***P ≤ 0.001. Names of bacterial species are as in figure 1b. The numbers above the bars represent the number of strains included for that bacterial species.

Fig. 5.

Convergent adaptation of different pathogens is evident at the pathway and functional levels (a) Flagella pathway. A schematic illustration of the bacterial flagella. Parts that include genes that underwent adaptive change or loss of function in different bacterial species are denoted. (b) Rifampicin resistance. Locations of mutations in the RRDR of RNAP in TB, PA, and SA strains are denoted. The numbers denote the corresponding residues on the RpoB of TB, which were affected by the mutations. Color indicates rifampicin treatment: dark green—confirmed treatment by rifampicin; bright green—likely treated by rifampicin based on conventional medical indications for the use of the drug; light orange—likely not treated by rifampicin based on conventional medical indications for the use of the drug; dark orange—not treated by rifampicin. (c) Glycerol-3-phosphate synthesis. Genes involved in glycerol-3-phosphate synthesis that underwent change/loss of function in a large fraction of the strains in different bacterial species. X marks the gene encoding the relevant protein that underwent adaptive change or loss of function, with the color of X indicating the species. PRPP: phosphoribosyl pyrophosphate. Names of bacterial species are as in figure 1b. Shapes represent chemical compounds: orange circle—phosphate, blank circle—glycerol, blank square—glycerone, blue pentagon—imidazole, blank triangle—ribose or ribulose, and blank ellipse—formimino group + AICAR group.

Fig. 6.

Selected KEGG pathways and GO annotations that were overrepresented among genes undergoing change or loss of function in different bacterial species. (a) Each row represents a different pathway: Rows with light orange background—overrepresented pathways among changed genes; rows with light blue background—overrepresented pathways among genes undergoing loss of function. Dark boxes indicate that the pathway was overrepresented in the species appearing on the column. The rightmost column indicates whether the pathway was overrepresented when considering all strains of all species together. (b) As in (a) for GO annotations. Names of bacterial species are as in figure 1b.