De novo mutations mediate phenotypic switching in an opportunistic human lung pathogen

Abstract

Bacteria evolving within human hosts encounter selective tradeoffs that render mutations adaptive in one context and deleterious in another. Here, we report that the cystic fibrosis-associated pathogen Burkholderia dolosa overcomes in-human selective tradeoffs by acquiring successive point mutations that alternate phenotypes. We sequenced the whole genomes of 931 respiratory isolates from two recently infected patients and an epidemiologically-linked, chronically-infected patient. These isolates are contextualized using 112 historical genomes from the same outbreak strain. Within both newly infected patients, diverse parallel mutations that disrupt O-antigen expression quickly arose, comprising 29% and 63% of their B. dolosa communities by 3 years. The selection for loss of O-antigen starkly contrasts with our previous observation of parallel O-antigen-restoring mutations after many years of chronic infection in the historical outbreak. Experimental characterization revealed that O-antigen loss increases uptake in immune cells while decreasing competitiveness in the mouse lung. We propose that the balance of these pressures, and thus whether O-antigen expression is advantageous, depends on tissue localization and infection duration. These results suggest that mutation-driven alternation during infection may be more frequent than appreciated and is underestimated without dense temporal sampling.

Figure 1:. A new suspected Burkholderia dolosa transmission cluster among 3 adults with cystic fibrosis with connection to a historical outbreak.

Two siblings with cystic fibrosis (Patients Q and R) were recently infected with B. dolosa and suspected to be part of a three-person transmission cluster (with Patient J). Patient J was previously infected as part of a historical outbreak that infected 39 people, 14 of whom were profiled in a previous study and whose isolates are analyzed here. About 10 years elapsed between the last known transmission in that outbreak and the latest cluster. Here, we sequenced a total of 122 isolates cultured from the sputum of the new Patients Q and R over multiple timepoints (vertical bars). We also acquired 650 isolates from the lungs, lymph nodes, and spleen during autopsy of Patient J and 159 from blood obtained a week prior. Gray bars indicate survival time past infection; arrowheads on the far right indicate that the patient was alive as of January 2024.

Figure 2:. In early infection establishment, multiple B. dolosa clones are transmitted between patients.

(a) A maximum parsimony SNV phylogeny was built from whole-genome sequencing of 805 B. dolosa isolates from autopsy samples of the suspected index case (Patient J), 122 from serial sputum samples of Patients Q and R, and 112 previously published sequences of isolates from Patient J and 13 other patients a decade prior. All isolates from Patient Q and R are descended from a subset of Patient J’s diversity; a high resolution phylogeny of this clade (Methods) is shown to the right. The time of sampling of each isolate is indicated by color in the rightmost vertical bar, clades within the tree are shaded by patient, and SNVs of interest for transmission inference are indicated by colored shapes. (b) Potential transmission scenarios between Patients J, Q, and R are diagrammed, showing the need for either more than two transmission events or parallel nucleotide evolution to explain observed diversity. Arrows point in the direction of B. dolosa transfer and are colored by the SNV in panel (a) that precedes a transmission.

Figure 3:. B. dolosa lipopolysaccharide (LPS) O-antigen expression is reduced in early CF lung infection by de novo mutation, contrasting with trends in longer-term colonization.

(a) Parallel mutations in genes suspected to be involved in O-antigen synthesis are represented by circles. Each isolate contained at most one such mutation. Mutations are colored by the type of mutation and gene icon length is proportional to gene size. (b) The proportion of observed SNVs that fall in the O-antigen synthesis is significantly greater for Patients Q and R (recently infected) than for Patient J (infected for nearly a decade; P < .01 binomial proportion test). Both sets of mutations occur at a greater frequency than expected by random mutation across the O-antigen pathway (P < .01, binomial test, see Methods). Error bars indicate 95% confidence intervals. (c) Selected isolates from Patient Q and R, as well as near-isogenic samples from Patient J, were profiled for O-antigen banding pattern. Isolates are ordered by a phylogeny of SNVs and O-antigen-affecting indels (see Methods) and branches are colored by the occurrence of an O-antigen-affecting SNV, insertion, or deletion. Of 14 isolates with an O-antigen-affecting mutation, 13 show a reduction of O-antigen expression. (d) The proportion of O-antigen-disrupted isolates is significantly increased from the first to last sampled time point for both Patients Q and R (P < .01, binomial proportion test), but negligible in patient J after 10 years. Lines represented with 95% confidence intervals. (e) In a reanalysis of 112 isolate genomes from the original B. dolosa outbreak collected over a greater duration of infection, we identify a trend during extended chronic infection that contrasts with early infection. See Fig. S6 for more details. The shaded region indicates standard error of the mean.

Figure 4:. Disruption of B. dolosa LPS O-antigen expression leads to tradeoffs in lung versus spleen infection in vivo.

(a) Mice were infected with a 50:50 mixture of O-antigen-intact and disrupted phenotypes, with one B. dolosa strain marked with a lacZ cassette conferring blue colony color on X-gal-containing Bcc-selective agar. Lungs and spleens were excised to determine the bacterial load of O-antigen-intact and O-antigen-disrupted strains. Two trials of this experiment were performed for each of three near-isogenic pairs. The fraction of O-antigen-disrupted bacteria in mouse spleens and lungs is shown with error bars representing 95% confidence intervals. The O-antigen-disrupted to O-antigen-intact ratio is higher in spleens relative to lungs for all pairs (P < .002, paired T-test). (b) Macrophages were infected for 2 hours with near-isogenic B. dolosa strains with and without an O-antigen-disrupting mutation and then incubated for 2 hours with kanamycin, which kills extracellular bacteria. The number of intracellular bacteria is compared to the total bacteria obtained from an identical culture without kanamycin treatment, revealing an advantage for O-antigen-disrupted bacteria within macrophages (P < .03, Wilcoxon rank sum).

Figure 5:. Navigation of selective tradeoffs on O-antigen presentation by de novo mutation.

(a) Inferred natural history of B. dolosa LPS O-antigen presentation during an outbreak in people with CF. The initial B. dolosa outbreak was initiated by a strain lacking O-antigen due to a premature stop codon in a glycosyltransferase gene. This premature stop mutation likely occurred shortly before or during outbreak initiation, as experimental reversion successfully restores O-antigen presentation. As the outbreak spread and persisted in chronic infections, independent reversions restored B. dolosa’s O-antigen in multiple patients. Within each patient, strains with variable O-antigen presentation were recoverable and coexisted for years. Two patients recently infected by a survivor of this outbreak were initially colonized by strains expressing O-antigen. Remarkably, O-antigen-disrupted phenotypes re-emerged independently within each of these patients via a variety of de novo mutations in the O-antigen pathway. (b) Our in vivo experimental results suggest that O-antigen presence is advantageous in the lung while O-antigen absence is favored in the spleen. Combined with the historical observations, these results suggest that survival in the spleen or immune cells may be important during the early years of chronic B. dolosa lung infection.