Temporal phylogeography of Yersinia pestis in Madagascar: Insights into the long-term maintenance of plague

Abstract

Background: Yersinia pestis appears to be maintained in multiple, geographically separate, and phylogenetically distinct subpopulations within the highlands of Madagascar. However, the dynamics of these locally differentiated subpopulations through time are mostly unknown. To address that gap and further inform our understanding of plague epidemiology, we investigated the phylogeography of Y. pestis in Madagascar over an 18 year period.

Methodology/principal findings: We generated whole genome sequences for 31 strains and discovered new SNPs that we used in conjunction with previously identified SNPs and variable-number tandem repeats (VNTRs) to genotype 773 Malagasy Y. pestis samples from 1995 to 2012. We mapped the locations where samples were obtained on a fine geographic scale to examine phylogeographic patterns through time. We identified 18 geographically separate and phylogenetically distinct subpopulations that display spatial and temporal stability, persisting in the same locations over a period of almost two decades. We found that geographic areas with higher levels of topographical relief are associated with greater levels of phylogenetic diversity and that sampling frequency can vary considerably among subpopulations and from year to year. We also found evidence of various Y. pestis dispersal events, including over long distances, but no evidence that any dispersal events resulted in successful establishment of a transferred genotype in a new location during the examined time period.

Conclusions/significance: Our analysis suggests that persistent endemic cycles of Y. pestis transmission within local areas are responsible for the long term maintenance of plague in Madagascar, rather than repeated episodes of wide scale epidemic spread. Landscape likely plays a role in maintaining Y. pestis subpopulations in Madagascar, with increased topographical relief associated with increased levels of localized differentiation. Local ecological factors likely affect the dynamics of individual subpopulations and the associated likelihood of observing human plague cases in a given year in a particular location.

Fig 1. SNP and MLVA phylogenies depicting 18 major phylogenetic subgroups identified among 773 Malagasy Yersinia pestis samples.

(A) SNP phylogeny based on 212 informative SNPs identifying 100 individual nodes (circles and stars) among 770 Malagasy Y. pestis samples (only the lineage could be identified for the remaining 3 samples, and not the specific node). Stars indicate terminal nodes defined by a sequenced strain. Circles indicate intermediary nodes (i.e., collapsed branch points) along the lineages containing groups of samples. Branch points that did not contain any samples are labeled in black italics. A dashed arrow indicates a branch leading to a single, previously identified, terminal node not represented among the samples in this analysis. Lineages (lower case letters) and nodes within lineages (numbers within circles and stars) were named as in [–19], with new letters and numbers assigned to newly identified lineages and nodes, respectively. Basal nodes d and k are represented by pie charts, indicating the presence of multiple MLVA identified subgroups within these nodes. Color shading indicates the 18 identified phylogenetic subgroups and, to the extent possible, corresponds to the subgroup colors used in reference [19]. Solid pale green and striped pale green, respectively, indicate the new SNP lineage (w) and remaining subset of samples within basal node d that were split in this analysis from a single previously identified subgroup. The number of SNPs on branches with >1 SNP are indicated in red. The single SNP differentiating between Groups I and II is indicated by a perpendicular red line on the branch between nodes d and k. (B) MLVA phylogenies of 10 and 38 Malagasy Y. pestis samples from basal SNP nodes d and k, respectively. The MLVA phylogenies consist of neighbor-joining dendrograms constructed in MEGA6 [35] using mean character based distance matrices. Bootstrap values ≥50 (generated in PAUP 4.0b10 (D. Swofford, Sinauer Associates, Inc., Sunderland, MA) based upon 1,000 simulations) supporting MLVA phylogeny branches are indicated. One and four additional phylogenetic subgroups consistent with previous analyses [19] were identified within the MLVA phylogenies of nodes d and k, respectively. In addition, 1 and 3 samples within nodes d and k, respectively, did not fall into any identified phylogenetic subgroup and were labeled with a “+” or an “*”, and classified as II.NONE and I.NONE, respectively.

Fig 2. Geographic distribution of 355 Malagasy Yersinia pestis samples from 1995 to 2000.

The map of Madagascar in the upper left indicates elevation, all of the geographic points in this study (small black points), and the portion of Madagascar represented in the other panels (rectangle). The geographic distribution of identified subgroups is presented in separate panels for each year. Circles in the panels represent the locations of the fokontany (i.e., villages) or commune centroids (when the fokontany was unknown) where samples were collected. In some cases where separate circles were too close together to be visibly distinguished at this scale, a single circle indicating the overlapping circles was substituted. This occurred primarily for fokontany within communes Mahajanga and Mahabibo within district Mahajanga I, and for fokontany within the various arrondissements (i.e., administrative divisions) within district Antananarivo Renivohitra, but also occasionally occurred at other locations. Colors of the mapped circles indicate identified subgroups and correspond to the subgroup color designations in Fig 1. Divisions within circles indicate that multiple subgroups were found at that location in that year. Unaffiliated Group I and II samples (i.e., I.NONE and II.NONE) are indicated by a “*” and a “+”, respectively.

Fig 3. Geographic distribution of 373 Malagasy Yersinia pestis samples from 2001 to 2008.

The geographic distribution of identified subgroups is presented in separate panels for each year, with symbols and colors as in Fig 2.

Fig 4. Geographic distribution of 45 Malagasy Yersinia pestis samples from 2009 to 2012.

The geographic distribution of identified subgroups is presented in separate panels for each year, with symbols and colors as in Fig 2.

Fig 5. Geographic distributions of identified SNP nodes within Group I.

Four map panels showing the geographic distributions of identified SNP nodes within Group I phylogenetic subgroups j (A); l, y, and z (B); q (C); and s (D) are presented. A full map of Madagascar in the lower right indicates the portion of Madagascar included in each of the expanded map panels, with each expanded section indicated by a colored rectangle corresponding to the color associated with one of the phylogenetic subgroups depicted in the corresponding expanded panel. Circles within the expanded map panels represent the locations of the fokontanys (i.e., villages) or commune centroids (when the fokontany was unknown) where samples were collected. In some cases where separate circles were too close together to be visibly distinguished at this scale, a single circle indicating the overlapping circles was substituted. This occurred primarily for fokontany within communes Mahajanga and Mahabibo within district Mahajanga I, and for fokontany within the various arrondissements (i.e., administrative divisions) within district Antananarivo Renivohitra, but also occasionally occurred at other locations. Colors within the mapped circles correspond to the subgroup color designations in Fig 1. Divisions within circles indicate that multiple SNP determined nodes were found at that location. Numbers within circles and pie chart slices indicate the node within a given subgroup from Fig 1A that the mapped samples belong to. Overlapping circles and pie chart slices representing the same node were merged together and identified with a single label to simplify the maps. The specific node could not be definitively determined for two subgroup q samples and one subgroup s sample, so no number is shown.

Fig 6. Geographic distributions of identified SNP nodes within Group II.

Four map panels showing the geographic distributions of identified SNP nodes within Group II phylogenetic subgroups v (A), h (B), t (C), and u, w, and x (D) are presented, with symbols and colors as in Fig 5. A full map of Madagascar indicates the portion of Madagascar included in each of the expanded map panels, with each expanded section indicated by a colored rectangle corresponding to the color associated with one of the phylogenetic subgroups depicted in the corresponding expanded panel.

Fig 7. Variation in sampling frequency over time.

The number of samples identified each year in this analysis for each Yersinia pestis subgroup found predominantly in the well-sampled Betafo and Moramanga regions.

Fig 8. Maximum clade credibility phylogeny, reconstructed in BEAST, with mean divergence times for whole genome sequences from 32 Malagasy Yersinia pestis strains and reference strain CO92.

A maximum clade credibility phylogeny based upon SNPs identified among the 31 Malagasy Y. pestis strains sequenced here and the previously sequenced Malagasy strain IP275 and reference strain CO92. Malagasy strain branches are labeled with the SNP nodes from Fig 1, the strain IDs from S1 Table, and the year of isolation of the strain. Colors of the clades and/or branches indicate identified subgroups and correspond to the subgroup color designations in Fig 1. A timeline indicates the estimated mean divergence times for the various branch points. Estimated mean divergence times for various nodes of interest are also indicated on the phylogeny. Yellow circles indicate the posterior probabilities for each of the clades, where larger circles indicate higher confidence.