To study changes of host type in Ebola we used whole genome Ebola sequences from 78 patients recently obtained and aligned with sequences from previous outbreaks [39 (link)]. The authors of this study investigated the phylogenetic relationship of samples within or between Ebola outbreaks. We applied the three phylogeographic methods presented above to infer the contribution of zoonotic events to Ebola spread. We used the same alignment provided in [39 (link)] for the BEAST2 analysis, including sampling dates, but we also added information regarding host type. We defined two subpopulations, human and animal reservoir, and we allowed lineages to transmit forwards in time from the animal reservoir to a human host, but not vice-versa. So our phylogeographic model had two locations (respectively human and animal reservoir) but migration was only assumed to occur in one direction. This results in a structured coalescent model with three phylogeographic parameters for MTT and BASTA (one migration rate and two effective population sizes), but only two parameters for DTA, as only a single general effective population size can be defined in that model. A peculiarity of these analyses is that no samples from one of the two considered populations were available. While this might seem an impassable limitation, previous studies have shown that the structured coalescent can provide meaningful estimates even in the absence of samples from one populations (i.e. “ghost deme”, see [47 ]), suggesting that it is possible to perform statistical inference of zoonosis rates in this scenario.
Since the inclusion of no animal samples is unusual, we considered a second, more typical, analysis in which we included genetic sequences from bats. Relatively little sequencing has been performed in potential animal reservoirs, so we were able to include only partial Ebola virus sequences from a 265 bp region of the polymerase (L) gene from seven bats collected in [48 (link)]. In this analysis, it was necessary to allow a small but non-zero rate of migration from humans to the animal reservoir to avoid predetermining inference of the ancestral location of the root. Therefore we constrained the migration rate from humans to animals at a rate 105 times lower than the animal to human rate. This preserves the ability of the model to infer ancestral locations in either of the two subpopulations, once samples from the animal reservoir have been included.
Since the inclusion of no animal samples is unusual, we considered a second, more typical, analysis in which we included genetic sequences from bats. Relatively little sequencing has been performed in potential animal reservoirs, so we were able to include only partial Ebola virus sequences from a 265 bp region of the polymerase (L) gene from seven bats collected in [48 (link)]. In this analysis, it was necessary to allow a small but non-zero rate of migration from humans to the animal reservoir to avoid predetermining inference of the ancestral location of the root. Therefore we constrained the migration rate from humans to animals at a rate 105 times lower than the animal to human rate. This preserves the ability of the model to infer ancestral locations in either of the two subpopulations, once samples from the animal reservoir have been included.
Free full text: Click here
