The intensity of pathogen shedding in reservoir hosts across space and time is a key component of the processes leading to pathogen spillover. However, quantifying spatiotemporal variation in pathogen shedding and its association with hypothesized environmental drivers is challenging, as the underlying transmission process is often unknown (e.g., density or frequency dependence, whether infection is lifelong or temporary) and as many replicate host populations must each be repeatedly sampled over time. We here capitalize on a three-year study of Hendra virus (HeV), a lethal bat-borne zoonosis, across six flying fox (Pteropus spp.) colonies in southeast Australia. We first use generalized additive mixed models (GAMMs) to test for seasonal pulses in HeV shedding and if these differ between new and established overwintering roosts. We next fit GAMMs to each time series to derive a roost-level metric of cumulative shedding intensity, the area under the epidemic curve (AUC), and use generalized least squares models to test how roost covariates are associated with long-term shedding intensity. We lastly use generalized linear mixed models to assess whether AUC is predictive of realized HeV spillover events.
Results/Conclusions
History of roost establishment during a food shortage was the best predictor of seasonal HeV shedding, explaining 47% of the spatiotemporal variation in flying fox urine pool prevalence. While long-established overwintering camps showed little seasonal trend in HeV shedding, camps newly established in urban habitats in response to food scarcity displayed a sharp peak in prevalence during winter (July to August). Similarly, cumulative shedding intensity (AUC) was also greatest in roosts with history of food scarcity and in roosts with lower relative abundance of native dietary plants. These analyses over different timescales support the idea that the loss of native foraging habitat drives nutritionally stressed flying foxes into urban habitats with poor-quality anthropogenic resources, facilitating greater viral shedding through immune impairment. Lastly, AUC was predictive of realized HeV spillover events within 20 km per roost, though in a non-linear fashion. The number of spillover events showed a concave association with AUC, with spillover maximized at intermediate cumulative HeV shedding intensity. This suggests that other parts of the spillover process, such as environmental persistence of virus and recipient host exposure, must be linked with spatiotemporal shedding to fully predict HeV spillover risks.