We trapped mammals at 7 sites along a forest-to-urban gradient in Maryland and Washington, DC, USA, from early June to late September 2005 and in April 2006. The sites included 1 forested area (Smithsonian Environmental Research Center, Edgewater, MD), 2 large wooded parks (Rock Creek Park, Rockville, MD; Fort Dupont Park, Washington, DC), 2 residential neighborhoods (Takoma Park, MD; Bethesda, MD), and 2 urban areas (Baltimore, MD; Washington, DC).

We quantified the land use around each site by calculating an urbanization index (UI) within a 1,000-m radius as follows:

Impervious land and forest cover were estimated by using multitemporal (leaf-on and leaf-off) compilations of Landsat satellite images at 30-m spatial resolution, higher resolution satellite imagery, and digital orthophotography (5).

We ran trap lines of Tomahawk (models 201, 203, 204, 207; Tomahawk Live Trap Company, Tomahawk, WI, USA) and Sherman (model LFAHD; H.B. Sherman Traps, Inc., Tallahassee, FL, USA) traps for 2–5 days and nights at each site. Captured animals were chemically restrained and tagged, and age was determined by using body mass and/or reproductive characters (6). Blood samples (0.1 mL) were obtained, dispensed into tubes containing 0.9 mL BA-1 medium, and placed on ice packs until storage at –80°C. Blood samples were allowed to clot before antibody assays were run. We assayed the blood samples for neutralizing antibodies to WNV and Powassan virus (but not St. Louis encephalitis virus, which was absent in the local bird community at these sites [7]) by using the plaque-reduction neutralization test (8) at a 1:10 dilution, with 80% and 90% neutralization of plaques as cutoffs. We examined variation in WNV antibody prevalence by using binary logistic regression with species and age as categorical factors and capture date and urbanization index as covariates. We used October 15, 2005, as the capture date for the April 2006 samples because the abundance of WNV-infected mosquitoes falls precipitously after this date (9).

We obtained 244 samples from 11 mammal species (Table 1). The probability of being WNV antibody–positive varied significantly among species, was significantly higher for adults, increased with capture date for juveniles, and increased with the urbanization index (Table 2). The higher seroprevalence in samples collected in April 2006 showed that WNV exposure of juvenile eastern gray squirrels (Sciurus carolinensis) continued after the last trapping periods in September 2005 (Table 2).

Seroprevalence rates were highest (and not significantly different) in 4 peridomestic species: eastern gray squirrels, Virginia opossums (Didelphis virginiana), raccoons (Procyon lotor), and Norway rats (Rattus norvegicus) (Tables 1, 2). Eastern gray squirrels were 5.5× more likely than eastern chipmunks (Tamias striatus) to have WNV antibodies and 4.5× more likely than Peromyscus leucopus; both differences were significant (Table 2).

Previous research on the exposure of mammals to WNV has shown patterns of antibody prevalence across several states and species (2,3,10–12). However, few studies have tested for statistical differences in the factors that influence the exposure of mammals to WNV, which thus would establish their usefulness as indicators of variation in WNV transmission. We found significant effects of age, species, site, and date of capture on WNV seroprevalence.

The increase in seroprevalence with urbanization suggests that factors that increase WNV transmission, including mosquito abundance, WNV prevalence, or feeding on mammals, are higher in more urban areas. However, increased survival of mammals in urban areas could also result in increased seroprevalence associated with urbanization.

As in other studies, we found high WNV seroprevalence in 3 common peridomestic wild mammal species (D. virginiana, P. lotor, and S. carolinensis) and lower seroprevalence in T. striatus (Table 1) (3,11,12). Fatal WNV infection or low mosquito exposure for T. striatus could account for low seroprevalence in this species in areas where other species were often exposed (Table 1) (7,9,13). Additionally, we found that prevalence was significantly lower in juvenile P. leucopus in forested areas than in urbanized areas (Table 1; 0/11 vs. 2/3; Fisher exact test p = 0.032), which might explain some of the site variability found in previous studies (3). Finally, we found higher seroprevalence in rats than did previous studies (3) (4/7 vs. 2/36 [1]; Fisher exact test p = 0.004), which may have been a result of sampling rats from highly urban areas.

In addition to host death and spatial variation in prevalence and vector abundance, vector feeding preferences may also contribute to the observed variability in WNV seroprevalence. Previous studies have shown that WNV vectors do feed on S. carolinensis, P. lotor, and D. virginiana (14,15), but these studies do not show data on host abundance, so feeding preferences cannot be determined. Similarly, at our sites several mammal species, including S. carolinensis and D. virginiana, were sources of Cx. pipiens blood meals (7). Unfortunately, sample sizes of blood meals that came from mammalian hosts were too small for determining relative preferences for different mammals, and most species (except S. carolinensis) are at low enough abundances that substantial numbers of blood meals would be required to estimate feeding preferences. Mosquito preferences for different mammal species is an area for future research.

In our study, the probability of having WNV antibodies increased with capture date for juveniles but not for adults (Table 2), which suggests that juveniles experience higher exposure and would be more useful for WNV monitoring. Higher exposure of juveniles may result from increased attractiveness to mosquitoes or weaker defensive behavior.

Mammals have been proposed as sentinels for human WNV risk because infection would indicate transmission outside the enzootic bird cycle (3). Our study demonstrates that wild mammals satisfy 2 critical requirements: spatial and temporal variability in exposure. Our results also show that to estimate current year transmission at the site of capture, using wild mammals as sentinels will require adequate samples of young animals that year or a longitudinal approach (10). Our finding that mammalian WNV seroprevalence appears to be more intense in urban areas suggests that per capita risk for exposure is higher in these areas.