Assessing a landscape barrier using genetic simulation modelling: Implications for raccoon rabies management

Abstract

Landscape barriers influence movement patterns of animals, which in turn, affect spatio-temporal spread of infectious wildlife disease. We compare genetic data from computer simulations to those acquired from field samples to measure the effect of a landscape barrier on raccoon (Procyon lotor) movement, enabling risk assessment of raccoon rabies disease spread across the Niagara River from New York State into Ontario, an area currently uninfected by rabies. An individual-based spatially explicit model is used to simulate the expansion of a raccoon population to cross the Niagara River, for different permeabilities of the river to raccoon crossings. Since the model records individual raccoon genetics, the genetic population structure of neutral mitochondrial DNA haplotypes are characterised in the expanding population, every 25 years, using a genetic distance measure, ϕ_ST, Mantel tests and a gene diversity measure. The river barrier effect is assessed by comparing genetic measures computed from model outputs to those calculated from 166 raccoons recently sampled from the same landscape. The “best fit” between modelled scenarios and field data indicate the river prevents 50% of attempts to cross the river. Founder effects dominated the colonizing genetic population structure, and, as the river barrier effect increased, its genetic diversity decreased. Using gene flow to calibrate the effect of the river as a barrier to movement provides an estimate of the effect of a river in reducing the likelihood of cross-river infection. Including individual genetic markers in simulation modelling benefits investigations of disease spread and control.

Introduction

Landscape genetic analyses are increasingly being recognised as a valuable approach for understanding disease–host dynamics (Rollins et al., 2006, Real and Biek, 2007). This requires extending theory explaining spatio-temporal patterns of genetic population structures of single species (e.g. meta-populations (Hanski and Gaggiotti, 2004), colonisation events (Hutchinson and Templeton, 1999), dispersal patterns (Hampton et al., 2004), and landscape barriers (Trizio et al., 2005) to disease–host systems. One of the challenges is separating the effects of these factors (Real and Biek, 2007). This can be achieved through modelling because specific aspects of the system can be explored in a controlled environment. In this study, the use of genetic data in simulation modelling is demonstrated as an emerging methodology for understanding infectious disease–host systems. Advantages of using genetic data include assessing the risk of infection in disease-free areas, areas lacking data on disease incidence or those with poor quality disease incidence records. We compare genetic data from raccoon field samples with genetic output from an individual-based spatial simulation model to investigate the effect of a major river on the spatio-temporal genetic structuring of raccoons to investigate the risk of raccoon rabies disease spread.

Raccoon rabies is a variant rabies virus specifically adapted to infect raccoons (Winkler and Jenkins, 1991). It was first detected in Florida in the 1940s, and a second major epizootic emerged in the late 1970s along the border between West Virginia and Virginia (Winkler and Jenkins, 1991), spreading northwards at a rate of 30–47 km/year (Childs et al., 2000). It reached Canada near Brockville, Ontario, in 1999 (Wandeler and Salsberg, 1999), likely, by crossing the St. Lawrence River from northern New York State (NY) and has the potential to further spread through south-central Ontario by crossing the Niagara River from western NY (Rosatte et al., 1997).

The spread of raccoon rabies has occurred as an irregular wave. The spatio-temporal variations are largely attributed to physiography (e.g. mountains and river valleys) and habitat quality of the landscape, which affect movement patterns and distribution of animals at risk (Childs et al., 2001). Quantifying the effect of various landscape barriers has, therefore, practical implications for rabies control planning in North America (Slate et al., 2005).

Rabies incidence records have been used to measure landscape barrier effects to disease flow (Smith et al., 2002). There are, however, several difficulties with this approach. Estimates of barrier effects have greater uncertainty when calculated in disease-free areas than infected areas because incidence data must be supplied from other regions, which inherently have region-specific factors influencing disease spread that do not necessarily apply to the study area. Furthermore, the highly variable quality of rabies incidence data can obscure measures of spread in both space and time. There are a number of potential sources of bias in rabies data in North America. First, rabies surveillance is typically passive and monitors incidence when and where humans are at risk. Consequently, the number of reported cases may be influenced by the density of the human population, so that many animals die undetected because they are not observed (Childs et al., 2001). Second, under-reporting is common once rabies is established in an area because people become complacent about disease presence and the surveillance system is overloaded (Wilson et al., 1997). Third, several jurisdictions are responsible for rabies surveillance (e.g. province/state, counties and townships), and these differ in their budgets, mandates and reporting procedures (Childs et al., 2001). Finally, most data have been collected on the basis of administrative units such as towns and counties that do not always correlate in size or location with environmental, ecological or biological factors affecting disease incidence (Lawson, 2001). While there are strategies for controlling the first three reporting biases (Childs et al., 2001), the quality of disease incidence data is problematic when assessing relationships of rabies spread to environmental, ecological or biological factors, and reporting units. Therefore, incidence data tend to be more valuable as an early warning system rather than allowing detailed epidemiological assessment.

Careful sample design for studies using genetic data can overcome the drawbacks in disease incidence data, and accurately and precisely represent the spatio-temporal patterns of disease spread. For instance, instead of relying on active or passive surveillance to detect rabid animals, there are many strategies for acquiring DNA for genetic analyses (e.g. sampling hair/tissues from fur houses, road mortalities and trapping programs). Consequently, data acquisition constraints are more controllable because they are a function of factors of study expenditure and time constraints, assuming that animal densities are sufficient for acquiring samples and that the samples are viable for DNA analysis.

We present the analysis of a neutral genetic marker as an appropriate data source for investigating infectious disease dynamics. Neutral genetic markers are not subject to selective pressures. New variations that arise do not affect fitness (reproductive and survival rates), thus, distribute in the population as reflected by mating and dispersal processes (Holderegger et al., 2006). Neutral markers can be used as a “tag” to identify spatio-temporal patterns resulting from these processes. We characterise the genetic population structure of raccoons from neutral markers in mitochondrial DNA (mtDNA) to calibrate a landscape barrier in a genetic simulation model to raccoon movement. This model simulates a population expansion over the barrier. Our strategy was to compare temporal “snapshots” of the simulated genetic population structure obtained from model scenarios run at different barrier permeabilities with the one derived from field data. This enabled us to quantify the effect of the barrier on raccoon movement which is beneficial for assessing the risk of raccoon rabies infection entering an area where the disease has yet to occur, therefore, is absent of disease incidence data.