Periodic Reporting for period 2 - SpiL (Spillover of Leptospira in island populations of the Channel Island fox)
Reporting period: 2019-10-01 to 2020-09-30
In February 2018 Dr. Borremans was an invited participant of a workshop on spillover in Montana (US), organized by Prof. Raina Plowright. This workshop gathered a limited number of international experts for a retreat with the goal of advancing theory on the spillover of pathogens, the topic of SPIL. A consequence of this unplanned opportunity was a change in the timeline of the planned aims in the project, which has led to the publication of two articles.
The first article completes Aim 2.4 and Deliverable 2.3 which is the development of a new conceptual framework of spillover. The article presents new theory on spillover across ecosystem boundaries (the core topic of project SPIL), which is the most important way in which pathogen spillover between species occurs.
Excellent progress has been made on a California sea lion (CSL) transmission model, a key part of the model of non-stationary of risk and spillover (Aim 2.2). The transmission model uses the long-term dataset on CSL leptospirosis cases (1984-2012) to analyze the demographic and environmental causes of annual variation in outbreak size. Model performance was high, with key results being: (1) Almost 60% of variation in outbreak size can be explained by a combination of demographic and environmental factors; (2) The model can be used to predict outbreak size ahead of time, which can have significant impacts on the preparedness of stranding centers along the US West Coast; (3) Climate change simulations suggest that the average outbreak size will decrease, with occasionally more extreme outbreak sizes.
Spatiotemporal risk mapping of the transmission of Leptospira in the fox population is another key component of Aim 2.2. This work is being conducted in close collaboration with Riley Mummah, a PhD student at the Lloyd-Smith research group co-mentored by Dr. Borremans. A major challenge has been to integrate spatial data from both GPS collars and radio-telemetry tracking, for which Riley has developed new methods that will be widely applicable in the field of movement ecology, especially given the fact that it is becoming increasingly easy to gather large amounts of data from different sources. These efforts have led to the ability to generate maps of estimated home ranges of all individuals, changing over time. Additionally, using a Cox Proportional Hazards model to analyze infection risk factors, we found a highly significant effect of the amount of rainfall in the preceding season on the risk of becoming infected.
The final stage of this part of the project, which is to integrate the spatial and non-spatial models of transmission into a complete spatiotemporal model of transmission, builds on a compartmental model of transmission in which individuals are classified into different states depending on their infection status. We developed an SEICR model with compartments S (susceptibles), E (exposed but not yet infectious), I (infectious), C (chronically infectious), R (recovered/immune to reinfection). This model incorporates mortality, seasonal birth pulses, and variation in transmission linked to precipitation. The model is able to reconstruct temporal changes in the proportion of antibody- and PCR-positive individuals over time. We are currently in the process of expanding this model into a patch-based spatial model of transmission on the island.
• Development of a method to reconstruct susceptibles in a population taking into account age- and sex-specific survival. This advance moves this method outside the field of human epidemiology, and opens up applications to time series of wildlife infections, where variation in survival and lifespan are likely to significantly influence the number of susceptible individuals in a population.
• Integration of GPS and telemetry data to estimate individual home ranges. This approach advances the field of movement ecology by allowing the combined use of data sources that differ greatly in their spatial resolution and their format (spatial points vs spatial polygons).
• The quantification of model fit for the sea lion transmission models (described above) implements a new type of statistic that we currently label “feature mismatching” (FM). Model fit statistics typically provide information about how closely the predicted data match the observed data by measuring the distance between these data points and then providing a measure of residual error. Such statistics however are biased towards large differences that can occur more often in overdispersed datasets where the maximum values are much larger than the smallest values. We developed a new method that scores the mismatch between predicted and observed outbreak categories instead of absolute values.
• Development of a new approach to constructing models of antibody level decay after infection. A major challenge in the field of serological analysis has been to model how antibodies decay after infection when there are no data for which the time of infection is known and when data are sparse. These situations are typical for wildlife. Using the unique fox-Leptospira dataset, we were able to develop and test a new approach that allows the incorporation of prior information in order to construct models antibody level decay, that can be applied to other host-pathogen systems. This represents a major advance in the field of serology.