In the present study we reconstructed phylogeography and spatiotemporal history of one of the major “indigenous” American RABV lineages, SCSK. Available surveillance records demonstrate that our sampling encompassed all areas of contemporary SCSK circulation in the US [2
]. Oral rabies vaccination (ORV) of skunks has not been consistently implemented in the US to date [38
]. The lack of intervention measures means that skunk rabies can be viewed as a natural process of pathogen invasion yielding insights to disease emergence.
We demonstrated that SCSK is genetically much more divergent than was inferred previously [3
]. Evolutionary analysis did not show substantial evidence for positive selection in the dataset. In general, this is typical for RABV genes. In several previous studies a positive selection was documented inconsistently in RABV genes, and codons suggested for positive selection varied depending on datasets analyzed and methods used [13
]. Furthermore, we conclude that results of MEME analysis should be interpreted cautiously as regard to the species adaptation, even if episodic adaptive selection has been detected.
The most significant diversity of SCSK was observed in Texas, and at least one of the Texas lineages (TX1) was placed at the base of the phylogenetic tree. Such phylogenetic patterns suggest that SCSK originated in the territory of Texas, although historical introduction of an ancestral virus from Mexico still cannot be ruled out based on the available samples. Phylogenetically related viruses circulate in skunks in Mexico, the so-called central or north-central Mexican skunk RABV lineages [3
]. However, no SCSK isolates from Mexico are available in GenBank and our archives. Moreover, surveillance records indicate that SCSK has been absent in southern Texas along the Mexican border for many years [2
]. These observations suggest that arid landscapes of this area and the Rio Grande River represent barriers for virus spread.
Viruses from the SCSK lineages present in Texas perpetuate in their limited geographic areas. Studying raccoon rabies, Biek et al. [15
] found that major diversification of viruses occurred shortly after the initial infection wave. Counties sampled 5-25 years after the first wave consistently yielded the same genetic lineage that had colonized the area initially. In our study, all SCSK lineages from Texas were localized geographically during 5-20 years of sampling, with older sequences present at the bases or inside the tree clades. This observation implies that colonization of the territory had occurred at least several years before sampling. Moreover, circulation areas of the lineages had only limited overlaps, although we could not identify any obvious natural barriers restricting virus spread (). Local perpetuation of RABV is also supported by surveillance records which describe outbreaks of skunk rabies as limited foci with increased disease incidence but irregular local geographic spread [6
]. Ecological studies demonstrated that skunks reach higher concentrations and live more sedentarily in proximity to humans, where they can find a variety of food. Skunk home range is usually no larger than 0.5-1.6 km2
in urban areas, but may extend to 3-5 km2
in rural areas [45
]. In addition, these animals have demonstrated a limited mobility travelling ~2 km per night [5
]. Monitoring of radio-collared skunks during a rabies epizootic suggested that rabid animals did not exhibit increased mobility as compared to healthy individuals [46
Biek et al. [15
] estimated dispersal speed for raccoon rabies as 38.4 km/yr during the initial epizootic wave and 9.5 km/yr at later stages. Even greater rates, 35-50 km/yr, were documented for fox rabies in Europe prior to implementation of ORV programs [47
]. Our estimates suggested that the velocity of the SCSK dispersal varied between 4-12 km/yr (). Such comparatively low values are likely associated with ecological characteristics of the host (limited mobility and contact rates) and virus-host interaction traits (pathogenicity, occurrence of the virus in salivary glands, and transmissibility) which need to be addressed in further studies. As for barriers, highways and rivers do not suppress gene flow in skunk populations [49
]. We did not observe association of these landscape features with geographical compartmentalization of SCSK lineages, including the Red River and the Arkansas River. The only exceptions were the Mississippi River which limits the eastward spread of the virus and the Rio Grande River which could limit the southward virus spread to date.
The SCSK lineages from northern and western territories are younger and less divergent than in Texas. In fact, the GP lineage from vast territories of northern Oklahoma, Kansas, western Colorado and Nebraska is genetically very homogenous. We hypothesize that skunks in rural areas of Great Plains have larger home ranges and greater mobility, so that an epizootic wave could cover broad area during a relatively short time, with a limited number of virus passages. This is supported with the greater wavefront velocity in the northward direction () and is concordant with the “surfing mutation” model [50
]. Based on the phylogeny, a possibility exists that viruses from lineage 1T were delivered from Great Plains back to Texas. Such reverse spread might occur via a secondary epizootic wave or via a long-distance translocation (although we did not find suggestions for such event in available literature). Conversely, this lineage might evolve in Texas. Geographic representation of viral clusters I-III within lineage GP () supports the hypothesis that SCSK was moving to Great Plains from the south. In this context we can speculate that ancestral viruses from lineage Pre-GP, available in our study from western Kansas and Colorado only, historically circulated more broadly, including the southern areas. Accordingly, if Pre-GP viruses circulated also in Texas, the lineage 1T could evolve locally rather than being introduced back to Texas from Great Plains.
We reconstructed the spatiotemporal history of the SCSK epizootic using phylogenetic diffusion models. The robustness of such an approach was verified previously on examples of other epizootics, where sampling covered the initial infection waves and a number of years thereafter [16
]. Our model suggests that the MRCA of the present SCSK originated ~170 years ago. We do not know which RABVs were present in skunks in Arizona, Kansas and Colorado during the 19th century [9
]. That could be SCSK (in such case our model does not cover the initial epizootic wave) or other RABV variants. As indicated by Charlton et al. [8
], skunk rabies almost disappeared in these states during the early decades of 20th century, with new cases increasingly documented since the 1940s.
A time-homogenous process of dispersal is likely to be an unrealistic approximation of viral dispersion, and implementation of RRW models significantly improves robustness of the spatial inference [16
]. According to our estimations, the dispersal rate of SCSK increased (likely corresponding to the initial infection wave) until ~1930, following by a period of stability during 40-50 years, and subsequently by a new increase since ~1980, which continues to date (). Moreover, after a long period of stability, the wavefront velocity of SCSK increased during the last decade as well. Our sampling covers the period of the 1980s-2000s, and our model for this period of time is most precise. Our model implies that SCSK exhibits a second large-scale expansion at the present time. A similar dynamic was observed for the mid-Atlantic raccoon rabies epizootic [15
]. We do not know the reasons for such large-scale changes in epizootic process, but they may be associated with global variations in host populations, associated with, for example, climate changes, human population expansion, agricultural developments, or other large-scale factors. Alternatively, they can be caused by changes in the pathogen. But we did not identify diversifying selection in SCSK G genes sampled during the last >30 years, and it is unlikely that changes in other viral genes (coding for internal proteins) might have such significant effect on virus-host interactions.
The SCSK circulation range is greater than circulation ranges of other skunk-associated RABV variants. Studying patterns of sympatric circulation of SCSK and north-central skunk (NCSK) RABV in the territory of the Great Plains, Barton et al. [14
] suggested that SCSK may have an increased pathogenicity compared to the NCSK. However, such suggestions should be corroborated via appropriate biological experiments. The NCSK seems younger than SCSK, as is suggested by our preliminary estimates (Figure S1
), but this hypothesis needs additional corroboration from further studies. At this time, we do not have a sufficiently representative dataset (per space and time) of NCSK gene sequences to perform a meaningful evolutionary comparison.
Our results demonstrate that the SCSK epizootic follows the same major patterns that were implied for the well documented mid-Atlantic raccoon rabies epizootic [15
] albeit with a lower dispersion speed, and suggest that SCSK will extend its range during the near term. The most significant spread is expected in the northern direction. Eastward expansion is also possible, once the virus crosses the Mississippi River. As the dispersal of SCSK is relatively slow, and viral lineages have occupied the same territories for decades without significant geographical spread, such territories could be targeted by local or step-wise ORV programs. One strategic area for initiation of such a program might be the Mississippi River valley.
We believe that our approach and findings can be extrapolated to other rabies reservoirs, particularly considering a potential risk of re-introduction of canine-mediated rabies in the US [38
], host shifts of RABV [13
], or invasion of canine rabies in wildlife populations elsewhere [52
], and used as a tool for investigation of epizootic patterns and planning interventions towards disease elimination.