Within-host evolution of Burkholderia pseudomallei in four cases of acute melioidosis

Abstract

Little is currently known about bacterial pathogen evolution and adaptation within the host during acute infection. Previous studies of Burkholderia pseudomallei, the etiologic agent of melioidosis, have shown that this opportunistic pathogen mutates rapidly both in vitro and in vivo at tandemly repeated loci, making this organism a relevant model for studying short-term evolution. In the current study, B. pseudomallei isolates cultured from multiple body sites from four Thai patients with disseminated melioidosis were subjected to fine-scale genotyping using multilocus variable-number tandem repeat analysis (MLVA). In order to understand and model the in vivo variable-number tandem repeat (VNTR) mutational process, we characterized the patterns and rates of mutations in vitro through parallel serial passage experiments of B. pseudomallei. Despite the short period of infection, substantial divergence from the putative founder genotype was observed in all four melioidosis cases. This study presents a paradigm for examining bacterial evolution over the short timescale of an acute infection. Further studies are required to determine whether the mutational process leads to phenotypic alterations that impact upon bacterial fitness in vivo. Our findings have important implications for future sampling strategies, since colonies in a single clinical sample may be genetically heterogeneous, and organisms in a culture taken late in the infective process may have undergone considerable genetic change compared with the founder inoculum.

Figure 1. Maximum parsimony phylogenies of Burkholderia pseudomallei isolates derived from four Thai acute melioidosis patients.

Phylogenies for patients 19, 23, 44 and 45 are shown in Panels A-D, respectively. Multiple tissue sites from each patient were cultured (e.g. blood, urine, tracheal, wound and pus samples). Where possible, ten colonies from each tissue site were retrieved; isolates were genotyped using a 23-locus multilocus variable-number tandem repeat analysis. In each patient, the numerically dominant genotype was assumed to be the original infecting strain. Specific mutations are displayed alongside each branch (e.g. 3652k-1 refers to a one-repeat deletion at locus 3652k). A single, theoretical, intermediate genotype not observed in our isolate set (Panel D, 20k+1 and 3152k-8) is represented by a solid black square. Asterisks indicate probable founder genotypes. See Figures S1 and S2 for alternative P19 and P45 phylogenies and Table S2 for in vivo mutation rate calculations.

Figure 2. Frequency distribution of Burkholderia pseudomallei multilocus variable-number tandem repeat (VNTR) analysis mutations.

The predicted geometric distribution model (black) was compared with in vitro (gray) and in vivo (white) VNTR mutation data. The geometric distribution model fits well with observed mutations of between one and four repeats; however, this model is not appropriate for predicting 5+ repeat copy number mutations, as previously observed in Escherichia coli O157:H7 .