Genomic analysis of the emergence of 20th century epidemic dysentery

Abstract

Background: Shigella dysenteriae type 1 (Sd1) causes recurrent epidemics of dysentery associated with high mortality in many regions of the world. Sd1 infects humans at very low infectious doses (10 CFU), and treatment is complicated by the rapid emergence of antibiotic resistant Sd1 strains. Sd1 is only detected in the context of human infections, and the circumstances under which epidemics emerge and regress remain unknown.

Results: Phylogenomic analyses of 56 isolates collected worldwide over the past 60 years indicate that the Sd1 clone responsible for the recent pandemics emerged at the turn of the 20th century, and that the two world wars likely played a pivotal role for its dissemination. Several lineages remain ubiquitous and their phylogeny indicates several recent intercontinental transfers. Our comparative genomics analysis reveals that isolates responsible for separate outbreaks, though closely related to one another, have independently accumulated antibiotic resistance genes, suggesting that there is little or no selection to retain these genes in-between outbreaks. The genomes appear to be subjected to genetic drift that affects a number of functions currently used by diagnostic tools to identify Sd1, which could lead to the potential failure of such tools.

Conclusions: Taken together, the Sd1 population structure and pattern of evolution suggest a recent emergence and a possible human carrier state that could play an important role in the epidemic pattern of infections of this human-specific pathogen. This analysis highlights the important role of whole-genome sequencing in studying pathogens for which epidemiological or laboratory investigations are particularly challenging.

Figure 1

Shigella dysenteriae type 1 phylogeny. a. Maximum likelihood phylogeny of the 56 Shigella dysenteriae type 1 strains, relative to E. coli K12, O157, O55, Shigella flexneri and Shigella sonnei. The root of the tree was determined using the more distant relative E. coli UTI89. The tree was constructed using 78,267 variable positions found over 1,859 genes with 1,000 Bootstrap repetitions (values displayed on the branches). Sd1 strains are framed in orange. A neighbor-joining tree based on the rate of synonymous substitutions (dS) between each pair showed the same topology as the maximum likelihood tree, with similar relative branch lengths (data not shown). b. Maximum likelihood phylogeny based on a concatenation of the 919 polymorphic positions identified between the 56 Shigella dysenteriae type 1 strains in the common genes that are single-copy and not subjected to recombination (see Methods). The root (E. coli ancestor) was approximately placed based on the E. coli outgroups (Figure  1a). Geographic distribution of lineages or sub-lineages: A. Tennessee, B. China, C1. Guatemala, C2. Zambia, C3. Cameroon and an unknown location, D1. Bangladesh, D2. Central African Republic, D3. Bangladesh, D4. Bangladesh and India, D5. Bangladesh and D6. India and Thailand.

Figure 2

Similarity of genes distributions across COG functional categories. a. Distribution of genes in COG functional categories for the entire pan-genome, the genes that were lost in some strains and the genes modified by non-synonymous substitutions (expressed in percentage of genes in each group versus total number of genes). Genes for whom no function could be assigned were left out (lost: 17.18%, non-synonymous: 7.92%, pan-genome: 18.75%). b. Genes lost and genes modified with non-synonymous substitutions tend to distribute across functional categories in a similar fashion as genes overall (Pearson correlation = 0.9097 and 0.9758 respectively) indicating that the loss or modification of genes are selected at random. Only one functional category contains more lost genes than expected: lipid metabolism. In contrast, the functions of genes that were gained do not reflect the distribution of genes overall across functional categories (Pearson correlation = 0.5087) (data not shown).

Figure 3

Distribution across COG categories of the genes lost. Genes that are lost in some strains for which a COG functional category is available are represented in this heatmap, color-coded by the type of genetic change resulting in the loss. The dendrogram is the topology of the maximum likelihood tree from Figure  1b.

Figure 4

Distribution across genomes of the genes gained over the course of Sd 1 evolution. a. Distribution of genes gained based on their presence and absence in the genomes contributing to the pan-genome. The dendrogram is the topology of the maximum likelihood tree from Figure  1b. b. Distribution across genomes of the acquired genes identified as conferring antibiotic resistance. Antibiotic resistance genes were identified by sequence comparison (blast) with the Antibiotic Resistance Database (ARDB). The dendrogram is the topology of the maximum likelihood tree from Figure  1b.