Unifying the spatial population dynamics and molecular evolution of epidemic rabies virus

Abstract

Infectious disease emergence is under the simultaneous influence of both genetic and ecological factors. Yet, we lack a general framework for linking ecological dynamics of infectious disease with underlying molecular and evolutionary change. As a model, we illustrate the linkage between ecological and evolutionary dynamics in rabies virus during its epidemic expansion into eastern and southern Ontario. We characterized the phylogeographic relationships among 83 isolates of fox rabies virus variant using nucleotide sequences from the glycoprotein-encoding glycoprotein gene. The fox rabies virus variant descended as an irregular wave with two arms invading from northern Ontario into southern Ontario over the 1980s and 1990s. Correlations between genetic and geographic distance suggest an isolation by distance population structure for the virus. The divergence among viral lineages since the most recent common ancestor correlates with position along the advancing wave front with more divergent lineages near the origin of the epidemic. Based on divergence from the most recent common ancestor, the regional population can be partitioned into two subpopulations, each corresponding to an arm of the advancing wave. Subpopulation A (southern Ontario) showed reduced isolation by distance relative to subpopulation B (eastern Ontario). The temporal dynamics of subpopulation A suggests that the subregional viral population may have undergone several smaller waves that reduced isolation by distance. The use of integrated approaches, such as the geographical analysis of sequence variants, coupled with information on spatial dynamics will become indispensable aids in understanding patterns of disease emergence.

Fig. 1.

Phylogeography and spatial cluster analysis of fox rabies virus variants across southern Ontario. (A) Map of southern Ontario with sample locations for fox rabies variants indicated by the black circles. A “southern fixed point” (red arrow) was used as a fixed point geographic marker for assessing spatial distances. (B) ML tree based on 1,572-bp G gene nucleotide sequence. Relative geographic relationship of each restriction type is indicated within the province of Ontario [map adapted from Nadin-Davis et al. (16)]. (C) Cluster analysis of the 20 different G gene sequence types.

Fig. 2.

Geographic distance versus genetic distance for fox rabies virus variants indicating isolation by distance. (A) The distribution of observations in each of the 45 different geographic distance bins is log-normally distributed. (B) Correlation between the average Euclidean distance of the geographic bin and the average genetic distance within each bin (R² = 0.90, df = 44, P < 0.001).

Fig. 3.

Subpopulation structure and clinal variation in time since MRCA in fox rabies virus variants across southern Ontario. (A) Correlation between the estimated number of substitutions since MRCA and geographic distance from the sample to the “southern fixed point” (red arrow, Fig. 1 A) for the 83 mapped Ontario red fox rabies virus samples (R² = 0.47, df = 82, P < 0.001). (B) Plot of Moran's I spatial autocorrelation coefficient among the substitutions since MRCA at different lagged distance classes over the southern Ontario peninsula.

Fig. 4.

Mantel statistic computed following Epperson (17) for the two viral groups (A and B) corresponding to the eastern and southern Ontario populations and combined groups (A + B). For each computation, we kept the geographic distance matrix constant and permuted the elements of the genetic distance matrix for each group. We randomly assigned genetic distances to geographic distances 5,000 times and created a distribution of values. For both Group A and Group B and combined groups (A + B), the observed statistic value is larger than the maximum value associated with permutations (P value <0.001 for each analysis).