Baylisascaris procyonis is a parasitic roundworm recognized as a zoonotic threat that uses the raccoon (Procyon lotor) as its definitive host (Sorvillo et al., 2002; Wise et al., 2005; Pai et al., 2007). White-footed mice (Peromyscus leucopus) and other small rodents are the most common intermediate hosts, often becoming infected while foraging upon undigested seeds within raccoon scat, particularly if corn is present in the feces (Page et al., 2001a, 2001c). However, some exposure to B. procyonis eggs by small mammals may occur simply as a consequence of their normal movement behavior, as raccoon latrines, i.e., defecation sites, usually are located within important travel corridors for these species (Page et al., 1998). Exposure risk also may be elevated in fragmented ecosystems, where white-footed mice and raccoons often are ubiquitous and abundant, and movements of both species are concentrated in remnant forest fragments (Nupp and Swihart, 2000; Moore and Swihart, 2005; Beasley and Rhodes, 2010).

There are 2 key factors underlying the increased potential for B. procyonis infections in white-footed mice in fragmented landscapes. First, both B. procyonis in raccoons and the density of latrines are higher in fragmented landscapes than in continuously forested settings (Kazacos, 2001; Page et al., 2001b; Smyser et al., 2010). Reflecting these patterns, larval infections in white-footed mice also are elevated in fragmented landscapes relative to individuals in heavily forested ecosystems (Page et al., 2001b). Second, raccoon populations thrive in landscapes fragmented by agriculture, where densities greatly exceed those of populations in less disturbed ecosystems (Beasley et al., 2011; Houle et al., 2011). Within fragmented agricultural ecosystems, raccoons are highly dependent upon agricultural food resources, and thus their movements are primarily concentrated within forest remnants and along forest-agricultural interfaces (Rivest and Bergeron, 1981; Beasley, DeVault, and Rhodes, 2007; Beasley, DeVault, Retamosa et al., 2007; DeVault et al., 2007; Beasley and Rhodes, 2010). The combination of high densities, intensive use of forested habitats, extensive use of corn, and high prevalence of B. procyonis exhibited by raccoons in agricultural ecosystems all increase the probability that white-footed mice are being exposed to raccoon latrines with infected B. procyonis larvae (Page et al., 2001b).

Despite a growing awareness that habitat fragmentation can alter the transmission dynamics of infectious disease (Allan et al., 2003; LoGiudice et al., 2003; Brownstein et al., 2005), few data exist pertaining to factors affecting the spatial distribution of B. procyonis infection in agricultural ecosystems. Page et al. (2001b) did demonstrate that prevalence of B. procyonis increases in response to habitat fragmentation, although their study did not include samples from highly fragmented populations, where raccoon densities and roundworm prevalence are likely to be highest. Given the widespread and increasing distribution of agriculturally dominated landscapes, it is clear that the relationship between landscape attributes and the transmission dynamics of B. procyonis to intermediate hosts should be further evaluated in these systems. Thus, our goal in the present study was to determine the influence of landscape-level habitat attributes on the spatial distribution of B. procyonis infection in white-footed mice in fragmented agricultural ecosystems. To achieve this goal, we used an information-theoretic approach to develop models for predicting the prevalence, mean abundance, and mean intensity of B. procyonis infection in white-footed mice among 22 forest patches in a fragmented agricultural ecosystem in northern Indiana.

MATERIALS AND METHODS

RESULTS

In total, 353 white-footed mice were trapped, with the number of mice removed ranging from 9 to 20 among patches (x̄ = 16.05, SE = 0.73). All sites revealed evidence of larval B. procyonis infection within their respective white-footed mice populations, and prevalence ranged from 5 to 70% (x̄ = 29.25%, SE = 3.93%) among habitat patches. Mean abundance of B. procyonis ranged from 0.05 to 6.67 larvae per mouse (x̄ = 1.44, SE = 0.41), and mean intensity of infection ranged from 1 to 20 larvae per infected mouse (x̄ = 3.77, SE = 0.95).

Pearson correlation tests indicated a high degree of correlation among our infection parameters (P < 0.05). Results of our regression analyses produced 3 models with moderate support for each dependent variable (Table I). Variables included in supported models were consistent among analyses, although the ranking of these models differed slightly. All measures of B. procyonis infection status in white-footed mice (prevalence, mean abundance, mean intensity of infection) were primarily influenced by emergent properties of habitat fragmentation. In particular, forest patch size, and the availability of forest, i.e., index of habitat connectivity, immediately surrounding sampled habitat patches were the most influential variables for predicting B. procyonis infections among white-footed mouse populations based on model weights and rankings, although the availability of forest within the broader landscape surrounding sampled patches also was included in a single supported model for each analysis (Table I). The regression coefficients for each of these variables were negative for all dependent variables, indicating that the prevalence, average number of larvae per mouse, and mean intensity of infection all decreased as patch size and habitat connectivity increased (Fig. 1). No stream or anthropogenic variables were present in any supported model, i.e., ≤4 AIC_c.

DISCUSSION

Despite a growing awareness that habitat fragmentation can have a profound influence on transmission dynamics of infectious disease (Allan et al., 2003; LoGiudice et al., 2003; Brownstein et al., 2005; Dharmarajan et al., 2012), investigation of the ways in which epidemiological processes are influenced by attributes of fragmented ecosystems remains an understudied area of research (Ostfeld et al., 2005). Here, we demonstrate that landscape attributes contribute to spatial variability in B. procyonis infections in an intermediate host, white-footed mice. Prevalence, mean abundance, and mean intensity of B. procyonis infection in mice all were highly correlated and positively associated with emergent properties of habitat fragmentation, with mice in smaller and less contiguous forest patches exhibiting increased parasite burdens. Although only 3 large forest patches were sampled in this study, i.e., >60 ha, due to their limited availability in our study area, this inverse relationship held among patches <20 ha. These data suggest that factors below the scale of landscape, such as fine-scale habitat structure or demographic and behavioral attributes of mice and/or raccoons within forest patches, likely are important drivers of intermediate host infection with B. procyonis as well (Alder and Wilson, 1987; Gortazar et al., 1998; Calvete et al., 2004; Brouat et al., 2007).

Other studies investigating spatial patterns of infectious disease have demonstrated that both host demography and attributes of landscape composition and configuration can directly affect infection parameters (e.g., Jones et al., 1998; Langlois et al., 2001; Page et al., 2001b; Allan et al., 2003; Su et al., 2009; Dharmarajan et al., 2012). For example, a similar pattern of decreasing infection prevalence with increasing patch area has been observed for larval ticks infected with Borrelia burgdorferi (Allan et al., 2003; Brownstein et al., 2005), likely due to increased densities of white-footed mice in smaller habitat patches (Nupp and Swihart, 1998, 2000). In the case of B. procyonis, elevated infections in white-footed mice inhabiting small, isolated forest fragments may reflect differences in raccoon movement behavior, and subsequently latrine distribution among forest patches varying in size. Corn comprises a significant food resource for raccoons when available, and thus raccoon movements (Pedlar et al., 1997; Dijak and Thompson, 2000; Beasley and Rhodes, 2010) and latrines (Page et al., 1998) often are concentrated along forest-agricultural interfaces in landscapes containing abundant agricultural resources. Given that the proportion of edge to interior habitat is inversely related to patch size, intermediate hosts in small forest patches may encounter raccoon latrines more frequently due to greater overlap with areas of higher raccoon activity. Moreover, densities of white-footed mice also increase as patch size decreases (Nupp and Swihart, 1998, 2000), presumably resulting in increased competition among habitat patches of varying size. Increased latrine encounter rates, coupled with elevated densities of white-footed mice in small forest patches, suggest that the importance and availability of raccoon latrines as a food resource likely increase as forest patch size decreases.

Variability in the extent of raccoon movements among forest patches also may have contributed to the patterns observed in this study. The probability of acquiring an infectious disease increases as the extent of an individual's movements increases (Rogers et al., 1998; Vicente et al., 2007). Given that raccoon home range size is negatively correlated with forest patch area in agricultural ecosystems (Beasley and Rhodes, 2010), spatial variability in B. procyonis acquisition by raccoons likely exists due to differences in raccoon movement behavior among forest patches of varying size. However, the relationship between B. procyonis recruitment and raccoon home range size currently is unknown.

Interestingly, the nature of the relationship between B. procyonis infection and patch area differed between the present study and that of Page et al. (2001b). Across 20 forest patches sampled in a moderately fragmented landscape, Page et al. (2001b) observed a positive relationship between B. procyonis prevalence in white-footed mice and forest patch area as well as the amount of forest within a 3-km buffer of sampled patches. However, their models predicting mean abundance and mean intensity of infection were only marginally significant (P = 0.06 and P = 0.09, respectively; Page et al., 2001b). Moreover, although their work was performed in a fragmented agricultural landscape, it is important to note that their study area contained ∼100% more forest habitat than the present study. These contrasting results suggest there are fundamental differences in the relationship between transmission dynamics of B. procyonis and habitat attributes among landscapes with varying levels of habitat fragmentation. Such inconsistencies likely reflect differences in landscape composition and fine-scale habitat attributes among study sites. Alternatively, differences in host (intermediate or definitive) density or movement behavior among study sites could have contributed to the disparate patterns of infection observed among studies (LoGiudice et al., 2003; Wright and Gompper, 2005). For example, average densities of mice within habitat patches in the moderately fragmented landscape sampled by Page et al. (2001b) were 41% higher (x̄ = 34.6/ha; Nupp and Swihart, 1998) than those in our study area (x̄ = 20.4/ha; Eagan, 2009).

Ultimately, both the present research and that of Page and her colleagues (2001b) support the hypothesis that the size and context of remnant habitat patches in fragmented ecosystems play important roles in the transmission dynamics of B. procyonis. However, forest remnants in fragmented ecosystems can exhibit considerable variation in both the demographic and behavioral attributes of the species that inhabit them, as well as the fine-scale structure and composition of patch-specific plant communities (Fleishman et al., 2002; Holland and Bennett, 2007; Schooley and Branch, 2009). Thus, differences in the nature of the relationship between landscape attributes and B. procyonis infection parameters among studies point to the need to examine factors affecting transmission dynamics of this parasite at finer scales of spatial and biological resolution in highly fragmented agricultural ecosystems.