Anthropogenic environmental change can significantly alter availability and quality of food resources for reservoir hosts and impact host-pathogen interactions in the wild. The state of the host’s nutritional reserves as time of infection is a key factor influencing infection outcomes by altering host resistance. Here we combine experimental and model-based approaches to better understand how an environmental stressor affects host resistance to West Nile virus (WNV). Using American robins (Turdus migratorius), a species considered a superspreader of WNV, we tested the effect of acute food deprivation immediately prior to infection on host viremia. Our results show that robins food deprived of food for 48 hours prior to infection, developed higher virus titers and were infectious longer than robins fed normally. To gain understanding about the epidemiological significance of food stressed hosts, we developed an agent-based model that simulates transmission dynamics of WNV between the avian reservoir host and the mosquito vector. When we simulated a nutritionally stressed host population, the mosquito infection rate rose significantly reaching levels that represent an epidemiological risk. An understanding of the infection disease dynamics in wild populations is critical to predict and predict and mitigate zoonotic disease outbreaks.

Analysis of the Agent Based Model Output for mosquito infection rate (MIR). 

Using both daily and weekly aggregated data generated by the simulation, we built models for both model fit and computational efficiency. These models were for viral loads with food stress levels and time (with interaction) as independent variable. AIC model selection criterion was used to compare and determine the best model, which we subsequently report in this study.

Phase 1. Visualization of the data (both by day and by week), we notice that the robins exposed to the four stress levels, exhibit an upwards linear trend in MIR with time  and heteroscedasticity. Further, with time the variance of MIR increases where each stress level also shows linear trend as time increases and the variances increases as time increases.

We approach the analysis of the hypothesis that with low levels of circulating virus and a constant number of robins, early in the season the effect of varying stress levels of the robin population on MIR will be minimal, using a linear model that takes into account the inherent heteroscedasticity.

Phase 2. During the second phase of the season, virus transmission is amplified due to large influxes of susceptible robins on two different dates (19 September [day 50] and 10 October [day 71]). We tested the hypothesis that during this period with increasing numbers of susceptible robins, the effect of food stress will magnify the between-group (i.e. different host stress levels) differences in the MIR.

Visualization of the data, we notice that the four stress levels, exhibit a quadratic trend, rising during through day 60 before stabilizing and decreasing. Having only 5 data points for the weekly data did not adequately exhibit these quadratic trends, though the trend was exhibited more clearly using daily data. Subsequently, we conducted the analysis for Phase 2 with daily data. Multiple models for viral loads were run for the daily data and AIC model selection criterion was used to determine the best model.

All ABM analysis was conducted using Systat statistical software, R statistical software, with the glmmTMB package used to account for heteroscedasticity.

Viral Titer Data

Viral titer data in log/PFU per 0.1ml of whole blood for American robins (Turdus migratorius) that were inoculated subcutaneously in the cervical region with 0.1ml of 10⁵ log PFU/ml of infectious West Nile virus (strain WN02, 1986). We collected whole blood (0.05ml) daily through 6 DPI, dispensed it into tube with BA-1 (M199 medium with Hank’s salts, 1% bovine albumin, TRIS base (tris [hydroxymethyl] aminomethane), sodium bicarbonate, 2% fetal bovine serum, and antibiotics) and quantified viral titers using the Vero cell plaque assay as described by Blitvich et al. 2003.

The birds are all hatch years that were captured at one location in mid-Michigan during fall migration of 2018. The food-stressed birds were deprived of food for the 48 hours prior to inoculation and the normal birds' food access remained unchanged. All birds were given ad libitum access to water.

Viral titer data: The limit of detection for assay is 0.7 PFU/0.1ml. Hence, a bird is considered to be uninfected (0 titer) if they fell below this limit. There are no missing data.