Age-dependent relationships among diet, body condition, and Echinococcus multilocularis infection in urban coyotes

Abstract

Urban coyotes (Canis latrans) in North America increasingly exhibit a high prevalence of Echinococcus multilocularis, a cestode of recent and rising public health concern that uses rodents as intermediate hosts and canids as definitive hosts. However, little is known about the factors that drive the high urban prevalence of this parasite. We hypothesized that the diet of urban coyotes may contribute to their higher E. multilocularis infection prevalence via either (a) greater exposure to the parasite from increased rodent consumption or (b) increased susceptibility to infection due to the negative health effects of consuming anthropogenic food. We tested these hypotheses by comparing the presence and intensity of E. multilocularis infection to physiological data (age, sex, body condition, and spleen mass), short-term diet (stomach contents), and long-term diet (δ13C and δ15N stable isotopes) in 112 coyote carcasses collected for reasons other than this study from Edmonton, Alberta and the surrounding area. Overall, the best predictor of infection status in this population was young age, where the likelihood of infection decreased with age in rural coyotes but not urban ones. Neither short- nor long-term measures of diet could predict infection across our entire sample, but we found support for our initial hypotheses in young, urban coyotes: both rodent and anthropogenic food consumption effectively predicted E. multilocularis infection in this population. The effects of these predictors were more variable in rural coyotes and older coyotes. We suggest that limiting coyote access to areas in which anthropogenic food and rodent habitat overlap (e.g., compost piles or garbage sites) may effectively reduce the risk of infection, deposition, and transmission of this emerging zoonotic parasite in urban areas.

Fig 1. Map of the study area, with inset showing the location of Edmonton in central Alberta, Canada.

Samples were collected from areas within (“urban”) and outside (“rural”) the city, as highlighted in the map. Roads (purple) and water bodies (blue) are shown for reference.

Fig 2

E. multilocularis prevalence and intensity compared between sampling locations (urban vs. rural, left column), age classes (juvenile vs. adult, center column), and health classes (above or below the median health score, right column). Prevalence is represented as the percentage of the sampled population, and infection intensity is represented as worm counts with a log-transformed y-axis. Juvenile and adult coyotes were distinguished as being younger or older, respectively, than the median age of our sample (1.78 yr). Significant differences (p < 0.05 in regression models; see Methods and Table 1) are represented with asterisks (*).

Fig 3. Interactions between age, location, and spleen mass (normalized by body mass, g kg^-1) as predictors of E. multilocularis infection status and intensity.

Infection status (a) and intensity (b) both declined significantly with age in rural coyotes but less so in urban coyotes. (c) Spleen mass exhibited a significant three-way interaction with age and location. The figure shows predicted relationships between spleen size and infection for urban and rural coyotes at three different ages (the mean age and ±1 standard deviation). Infection status (probability of infection) and intensity (natural log-transformed worm counts) were both adjusted to appear on the same y-axis.

Fig 4. Dietary drivers of E. multilocularis infection status.

(a) Three-way interactions between age, location, and each focal diet item (stomach contents [solid lines, measured in ml] and stable isotope values [dashed lines]). As in Fig 3, relationships are plotted separately for urban and rural coyotes at three different ages. All diet components were rescaled so that the x-axis extends from the minimum to maximum value for each item (digestible anthropogenic food, 0–360 ml; indigestible anthropogenic food, 0–160 ml; rodents, 0–325 ml; δ¹³C, -24.3–-19.2, δ¹⁵N, 5.2–10.8). Refer to Table 1 for the results of likelihood ratio tests indicating the significance of these relationships. (b) Model-averaged coefficients for the top models predicting infection intensity. Each main effect is plotted followed by its two- and three-way interactions with other variables (indicated with an ‘x’). The coefficient for location reflects urban relative to rural coyotes. Coefficients were standardized by the partial standard deviation and weighted based on model weight prior to averaging. Thick and thin lines indicate 50% and 95% confidence intervals, respectively.

Fig 5. Coyote diet and stable isotope measures in relation to coyote infection status, age, and location (urban or rural).

Note the different scales on each panel. In the bar graphs, error bars indicate the standard error. In the boxplots, the box indicates the median and interquartile range, the line represents the 95^th percentile, and the dots represent outliers. Coyotes were classified as juvenile or adult based on whether they were younger or older, respectively, than the median sample age (1.78 yr).

Fig 6. Dietary drivers of E. multilocularis infection intensity.

(a) Three-way interactions between age, location, and each focal diet item (stomach contents [solid lines, measured in ml] and stable isotope values [dashed lines]). As in Figs 3 and 4, relationships are plotted separately for urban and rural coyotes at three different ages. All diet components were rescaled so that the x-axis extends from the minimum to maximum value for each item (see Fig 4). Refer to Table 1 for the significance of these relationships. (b) Model-averaged coefficients for the top models predicting infection intensity. Each main effect is plotted followed by its two- and three-way interactions with other variables (indicated with an ‘x’). The coefficient for location reflects urban relative to rural coyotes. Coefficients were standardized by the partial standard deviation and weighted based on model weight prior to averaging. Thick and thin lines indicate 50% and 95% confidence intervals, respectively.