Wildlife Management Practices Associated with Pathogen Exposure in Non-Native Wild Pigs in Florida, U.S.

Land use influences disease emergence by changing the ecological dynamics of humans, wildlife, domestic animals, and pathogens. This is a central tenet of One Health, and one that is gaining momentum in wildlife management decision-making in the United States. Using almost 2000 serological samples collected from non-native wild pigs (Sus scrofa) throughout Florida (U.S.), we compared the prevalence and exposure risk of two directly transmitted pathogens, pseudorabies virus (PrV) and Brucella spp., to test the hypothesis that disease emergence would be positively correlated with one of the most basic wildlife management operations: Hunting. The seroprevalence of PrV-Brucella spp. coinfection or PrV alone was higher for wild pigs in land management areas that allowed hunting with dogs than in areas that culled animals using other harvest methods. This pattern did not hold for Brucella alone. The likelihood of exposure to PrV, but not Brucella spp., was also significantly higher among wild pigs at hunted sites than at sites where animals were culled. By failing to consider the impact of dog hunting on the emergence of non-native pathogens, current animal management practices have the potential to affect public health, the commercial livestock industry, and wildlife conservation.


Introduction
The One Health approach to studying disease acknowledges the interdependent relationship between humans, animals, and the environment that they share [1]. One Health perspectives further acknowledge, either implicitly or explicitly, that the ways in which landscapes are used and managed can affect the emergence and expansion of zoonotic and wildlife diseases [2][3][4][5]. While the link between changes to the physical landscape and outbreaks of infectious diseases is well-established [6-9], empirical evidence of the relationship between pathogen emergence and land use policy and management is lacking.
Hunting is a land use that has been managed in the United States since the early 20th Century. In the last several decades, hunting has been proposed as a means of wildlife disease control. exposure and PrV-Brucella spp. co-exposure, as well as individual risk of exposure to PrV, were greater in sites hunted with dogs than sites utilizing other harvest methods, but these differences were not found for Brucella spp. alone. Our results illustrate the necessity of including a One Health approach to understand the epidemiology of a disease and to incorporate effective wildlife management decisions to mitigate disease.

Materials and Methods
Data for this study included eight years (July 2007 to December 2014) of samples that were collected across Florida as part of a national wild pig disease monitoring effort led by the United States Department of Agriculture (USDA), Animal Plant and Health Inspection Service, Wildlife Services' National Wildlife Disease Program (Figure 1). We collected 35 mL of blood from euthanized wild pigs on sites where animal control work was being conducted by the USDA, or from wild pigs that were killed in dog hunts at check stations where public hunts were managed by state wildlife or land management agencies. Regardless of the carcass acquisition method, samples were collected within 5 h of death. Using body size, reproductive characteristics, and dental structure as guidelines, we estimated age as one of three classes: Juvenile (<2 months), sub-adult (2 months-1 year), or adult (≥1 year) [38]. We included a sample in our analyses if the pig was unequivocally determined to be either seropositive or seronegative for PrV and Brucella spp.; the geographic location at which the sample was collected experienced either (1) dog hunting (n = 895) or (2) other forms of nonrecreational harvest for animal control (i.e., population management by live trapping or shooting without pursuit, n = 1096); the age of the animal was known; and the sex of the animal was known. We excluded sites with recreational hunting that only permitted hunting from tree stands or by stalking (n = 42). Total sample size after omissions was 1991 wild pigs (157 juveniles, 245 sub-adults, 1589 adults), with roughly equivalent sex ratios within the sample population (1000 females, 991 males). The relative proportion of pigs in each age class did not differ significantly (α = 0.05) between sites hunted with dogs and at sites with other forms of harvest (X 2 2 = 5.35, p = 0.070). Though there were slightly more females on sites with dog hunting (0.52 of all samples) than on sites with other forms of harvest (0.48 of all samples; X 2 1 = 3.94, p = 0.047), the difference was not biologically meaningful. No animals were euthanized for the purposes of this study. The University of Florida's Institutional Animal Care and Use Committee approved this study (Protocol 201307937) on 8 May 2013.
were slightly more females on sites with dog hunting (0.52 of all samples) than on sites with other forms of harvest (0.48 of all samples; Χ = 3.94, = 0.047 ), the difference was not biologically meaningful. No animals were euthanized for the purposes of this study. The University of Florida's Institutional Animal Care and Use Committee approved this study (Protocol 201307937) on 8 May 2013.   Table S1 (PrV) and Serological tests indicated the presence or absence of host antibodies to a pathogen. Because both PrV and Brucella spp. induce a lifelong infection and antibody production, a positive antibody test indicated that an animal was either exposed but not infected, or exposed and infected. Seroprevalence data for PrV and Brucella spp. have previously been used to determine relative pathogen prevalence and risk of transmission in wildlife populations (References [42,43], methodology reviewed in Reference [44]). Serological tests for PrV and Brucella spp. To account for hierarchical uncertainty in diagnostic sensitivity, we estimated the probability of antibody detection at the state, site, and assay level in order to achieve a more accurate estimate of serological prevalence [45]. Assuming the presence (not prevalence) of disease was static at each site through time (e.g., References [27,46]), we used a year-to-year capture (i.e., detection) history approach within the software PRESENCE version 12.12 [47] to estimate the probability of antibody detection at the state and site levels. Using separate site-level detection probabilities for hunted and culled populations, we also calculated corrected probabilities of occurrence of PrV and Brucella spp. at each site. Because each animal was only sampled once, we could not estimate the probability of antibody detection for a given individual. We estimated the probability of antibody detection for each serological test by averaging the sensitivities presented in the existing literature (e.g., References [48][49][50]). For the Brucella spp. assays, we averaged the sensitivities for Brucella abortus due to the absence of studies evaluating the detection of Brucella spp.-specific antibodies. Compounded probabilities of detection were calculated as the product of probabilities at all levels (state, site, and assay). Because the detection of antibodies was approximately equivalent across treatments (see "Results" for details), all subsequent analyses used the apparent, uncorrected values of serological incidence and prevalence.
We tested our hypotheses regarding apparent seroprevalence (i.e., the proportion of wild pigs exposed to a given pathogen) and individual exposure risk separately. To first compare seroprevalence in sites with dog hunting to sites with other forms of harvest, we performed a contingency table Chi-squared analysis comparing the expected and observed frequencies of pigs exposed to no pathogen, PrV only, Brucella spp. only, and both pathogens. Expected frequencies were calculated by multiplying the number of pigs in a given exposure category by the proportion of pigs in each harvest category (i.e., 0.45 from hunted populations, 0.55 from populations with other forms of harvest). We then used multi-model inference within an information theoretic framework [51] to determine the magnitude of the effect and relative importance of dog hunting (i.e., hunting = 1 and other harvest = 0), age (i.e., juvenile, sub-adult, or adult), and the interaction between dog hunting and age on the risk of PrV and Brucella spp. exposure, independently. The response variable was the serostatus of an individual pig, coded as 0 for no antibodies detected (no apparent exposure) and 1 for antibodies detected (apparent exposure) for each pathogen. All models included a random effect of site to account for stochastic spatial variation in disease prevalence, and were fit using a binomial error structure with a logit link. We included age as a predictor variable because previous studies have indicated that the seroprevalence of both PrV and Brucella spp. increases with age in wild pigs (e.g., References [27,[51][52][53]). Because most evidence suggests sex has a limited impact on either PrV or Brucella spp. seroprevalence in wild pigs [27,[52][53][54], we excluded sex from our analyses. Though we would have liked to have included population density as a covariate because of its strong association with disease prevalence [42,55,56], no local or regional estimates of wild pig density exist within the state of Florida.
Because we could not rule out any combination of variables, our model set for each pathogen included the factorial combinations of hunting status of the site where the sample was collected, age of the animal, and the interaction between hunting status and age (model set n = 4). All models were fit using the R [41] package lme4 version 1.1-13 [57]. We used the R package MuMIn version 1.40.0 [58] to rank the models by the Akaike information criterion (AIC) and to calculate the model-averaged coefficient and relative importance of each term [51]. Coefficients were averaged over all models and weighted by model likelihood. Since we used logistic regression models, we also calculated odds ratios as the exponentiated coefficients. The relative importance of each variable was determined as the sum of the Akaike weights of the models in which they appeared [51]. For each term, we used α = 0.05 for the determination of statistical significance, and α = 0.10 as the threshold for borderline significance.

Probability of Disease Detection
The analysis of hierarchical uncertainty in serological results provided insight into the sensitivity and bias in our samples (Table 1). Wild pigs tested seropositive for PrV and Brucella spp. in Florida every year, suggesting that the probability of antibody detection (p) at the state level was 1.00.
The year-to-year probability of PrV antibody detection was p = 0.96 for sites with dog hunting and p = 0.72 for sites with other forms of harvest (overall p = 0.81, Table S1). For Brucella spp., site-level antibody detection was p = 0.73 for hunted sites and p = 0.56 for sites with other forms of harvest (overall p = 0.63, Table S2). Despite the variation in detection probabilities between dog-hunted and other harvest sites, the corrected probabilities of occurrence were approximately 20% higher for PrV and 30% higher for Brucella spp. in sites with dog hunting compared to sites with other forms of harvest (Table S1, Table S2). For PrV assays, comparative studies have found that by the second week of exposure, both LAT and ELISA have near 100% sensitivity ( Table 1). The probability of antibody detection varied more strongly among Brucella spp. serological tests, with FPA exhibiting the highest levels of sensitivity (Table 1). Given the general superiority of FPA, we used the FPA results for analysis in the rare cases where more than one test was implemented for the same sample. The compounded detection probabilities suggested that we identified 8 out of 10 PrV antibody-positive individuals, regardless of serological test. We detected approximately 1 out of every 3 Brucella spp. antibody-positive wild pigs screened by RIV, 1 out of every 2 screened by card, and 2 out of every 3 screened by FPA.

Effect of Dog Hunting on the Prevalence of Disease
The contingency table analysis indicated that seroprevalence differed significantly between wild pigs that were hunted with dogs versus those that were harvested without dogs ( 2 3 = 59.33, p < 0.001; Table 2). The frequency of pigs with zero pathogen exposure was 15% lower than expected at sites with dog hunting, and 12% higher than expected at sites with other forms of harvesting. Conversely, the frequency of exposure to PrV was 18% higher than expected for dog-hunted populations, and 14% lower than expected for populations with other forms of harvesting. Frequency of exposure to both PrV and Brucella spp. was 40% higher than expected at sites with dog hunting, and 34% lower than expected at sites with other forms of harvest. Observed exposure to Brucella spp. alone did not deviate from expected frequency for either treatment.

Effect of Hunting on the Odds of Pathogen Exposure
The model that best fit the data for PrV exposure included the effects of harvest method, age, and the interaction between harvest method and age (Table 3). These three variables explained 20% of the variation in the data, and the random site effect explained another 30% of the variation. Independently, harvest method and age were equally important in predicting PrV exposure (Table 4). Dog hunting caused a roughly twelvefold increase in the odds of exposure to PrV (95% CI = 7.01, 16.61) compared to other harvest methods. Adult pigs were ten times more likely to be exposed to PrV than juveniles (95% CI = 6.31, 13.71), but the odds of exposure did not differ between juveniles and sub-adults (95% CI = −1.56, 5.76). The effect of age did not vary significantly by harvest treatment. No antibodies, pseudorabies virus (PrV) antibodies only, Brucella spp. antibodies only, and the occurrence of both antibodies in populations that were hunted with dogs and populations that were harvested by other means. Direction of arrow indicates whether observed frequencies (provided) were higher (↑) or lower (↓) than expected. Significance: ns = not significant;~= borderline significant (0.05 < p < 0.10); * p < 0.05; ** p < 0.01; *** p < 0.001.
Dog-hunted 426 ↓*** 326 ↑** 42 ns 101 ↑*** Not dog-hunted 694 ↑*** 291 ↓** 53 ns 58 ↓** Notes: K is the total number of parameters (including intercept, random site effect, and residual error); ∆AIC is the difference in the Akaike information criteria of model i and the best-fitting model; w i is the Akaike weight of model i; R 2 m is the marginal R 2 (i.e., the proportion of total variance explained by fixed factors only); and R 2 c is the conditional R 2 (i.e., the proportion of total variance explained by both fixed and random factors). Notes: Juvenile was the reference age class to which adult and sub-adult exposure odds were compared. SE OR is the standard error of the odds ratio. Coefficients were averaged over all models, weighted by model likelihood.
In models that did not contain a given variable, the coefficient of that variable was assumed to be zero.
The best-fitting model for Brucella spp. exposure included only the main effects of harvest method and age. These main effects explained 13% of the variation in the data, and the random site effect explained an additional 32% of the variation (Table 3). In contrast to PrV, age was more important than harvest method for predicting the seroprevalence of Brucella spp. (Table 4). The harvest method did not significantly affect exposure to Brucella spp. (95% CI = −1.34, 7.53), but adults were about 12 times more likely to be exposed to Brucella spp. than juveniles (95% CI = 7. 46, 16.15). Risk of exposure was also about five times higher for sub-adults than for juveniles, but the effect was only borderline significant (95% CI = −0.04, 9.15). As with PrV, the effect of age did not vary significantly between dog-hunted populations and populations harvested by other means.

Discussion
Our study detected higher disease exposure in populations that were hunted by dogs than in those that were harvested by other methods, regardless of pig age. Contrary to classical theories identifying hunting as a means of wildlife disease control [10][11][12][13][14][15][16], these findings support the predictions of a more recent harvest-disease dynamics model, which proposed that hunting increases the prevalence of disease in species with a high reproductive output, such as wild pigs [24]. This positive association between dog hunting and disease exposure could be the result of several phenomena or the interaction between them: (1) Density-dependent demography [13,24], (2) immunosuppression due to elevated stress levels [20][21][22], and/or (3) increases in contact rates due to the disruption of animal movement and social structure [64,65]. Though this observational study was statistically robust, we did not have the experimental data necessary to conclude that dog hunting caused heightened disease exposure: However, these results identified a need to consider the elevated prevalence of disease as a side effect of this particular harvest practice.
The higher prevalence of PrV and PrV-Brucella spp. co-exposure among wild pigs in sites hunted with dogs compared to other harvest methods is not well explained by density-dependent responses to harvest or by systematic differences in population densities. Population-level responses to animal removal, either by recreational hunting or by animal control, could induce elevated birth rates in both scenarios, which increases the supply of disease-susceptible individuals [13,24]. A previous work on European wild boar proposed that these demographic responses to hunting allow the persistence of diseases such as classical swine fever in free-ranging wild boar populations [17,18] and that removal of up to 90% of the adult population is required per annum to keep populations from increasing in density [66]. However, because all sites had some level of removal, we could not attribute the apparent effect of hunting with dogs observed here to density-dependent demography. In addition, it seems unlikely that there were systematic differences in population densities between sites that allowed hunting with dogs and other sites that harvested pigs. Dog hunting in Florida is not preferentially prescribed for areas with higher densities of pigs. At sites that allow dog hunting, it is primarily a form of recreation, with population control as a secondary objective. Indeed, most land and wildlife managers employ baited trapping as preferred forms of targeted population control, particularly in areas with high densities of pigs that create conflicts with land use mandates such as nonhunting outdoor recreation (e.g., county or state parks) or highly endangered habitats (e.g., national wildlife refuges, state wildlife refuges, national parks).
Nondemographic factors, such as stress, may have contributed to the harvest-induced disease response for PrV, an alpha herpesvirus. Previous studies have linked harvest activities to elevated levels of stress in other wildlife populations [20][21][22], and elevated stress has been linked to immunosuppression [23]. Elevated stress has been shown to induce recrudescence of latent herpesvirus infection and intensify viral shedding in cattle with bovine herpesviruses [67] and mice with herpes simplex virus [68], and stress has been the suggested cause of Macacine alphaherpesvirus 1 recrudescence in rhesus macaques [69] and PrV recrudescence in swine [70,71]. Recrudescence of latently infected wild pigs would increase the risk of exposure and transmission via multiple routes of contact [31,70,[72][73][74][75]. Though Brucella spp. may respond to general immunosuppression associated with chronic stress, it does not exhibit the stress-induced recrudescence of PrV, which may explain why we found no difference in Brucella spp. exposure between populations hunted with dogs and populations that were culled by other harvest methods. We did, however, find a positive response to harvest method in co-exposed animals. If PrV infection occurs during periods of low immune response, it would leave swine more vulnerable to secondary infections such as Brucella spp. [48]. This additional decline in immunity associated with PrV infection may explain why dog-hunted pigs were more likely to be co-exposed to both PrV and Brucella spp., but not more likely to be exposed to Brucella spp. alone. Testing this hypothesis explicitly would require measuring stress hormone levels concurrently with immune function markers and pathogen exposure data in populations with and without dog hunting.
In addition to elevating stress and birth rates, hunting with dogs may increase disease transmission by altering animal movement and social structure, which in turn may increase contact rates and transmission among individuals. For example, multiple groups of wild pigs may concentrate in areas inaccessible to dogs during the hunting season. Indeed, a study of wild pig movement in Alabama, U.S., found that during the hunting season, wild pigs had a significantly smaller home range and exhibited differential preference in habitat use [64]. However, other studies have found no such effect [76]. Such concentration of typically disparate social groups could temporarily increase local pig densities and associated contact rates, facilitating disease transmission [65]. Future research tracking swine movement and contact rates during and after the hunting season would elucidate the role of movement and social structure in the harvest-disease dynamic in feral swine observed here.
The possibility of confounding environmental variables was possible but not probable, as our sampling of both harvest treatments occurred across the geographic breadth of the state and primarily on public lands. As habitat generalists, wild pigs are found in virtually every habitat type in Florida, but tend to concentrate in areas with canopy cover [37] and in areas that offer supplemental sources of feed (e.g., agriculture [66,77,78]) or temporary refugia from hunting [64,79]. If habitat type were a confounding variable that influenced population density and therefore disease prevalence, then we would expect land cover associated with increased density (e.g., agriculture) to have been higher in dog-hunted sites. However, sites without dog hunts had significantly more agricultural cover within 6 km of the collection site (p < 0.001, Hernández et al. in prep. [80]), suggesting that this land cover type was not confounding the observed pattern. In addition, our inclusion of site as a random variable in our mixed effect model accounted for stochastic spatial variation in our samples. Indeed, the detection of a significant effect of hunting with dogs, despite the large variation among sites (Table 3), bolstered the support for the association between harvest method and disease dynamics.
The finding that hunting may modulate disease dynamics of PrV and Brucella spp. has agricultural, public health, and conservation implications. PrV infection is rapidly fatal in cattle and carnivores [33] and has been implicated in multiple deaths of the endangered Florida Panther (Puma concolor coryi) [81]. Although PrV is not considered to be a disease agent in humans, Brucella spp. is a major zoonotic pathogen in Florida and the U.S., producing nonspecific, flu-like symptoms in humans with the potential for serious, lifelong musculoskeletal, cardiovascular, genitourinary, and neurological complications if untreated [30,82]. In Florida, human infection with Brucella spp. is the second-highest reported zoonotic infection after Borrelia burgdorferi, the causative agent of Lyme disease [83]. Commercial livestock in the U.S. are considered free of both PrV and Brucella spp., yet both pathogens are widespread in wild pig populations throughout the country, which creates the potential for reinfection of commercial herds [26,27,56].
By using a One Health approach to consider land use effects in the potential for transmission of pathogens that affect humans and animals, we are better able to make management recommendations about the control of disease in wild pigs. Considering the growing accumulation of evidence against hunting as an effective population control method in density-dependent species [24,[84][85][86][87], a shift away from public hunts for the purposes of reducing population size warrants strong consideration. Targeted trapping, sharpshooting, and/or aerial gunning have been shown to reduce a larger proportion of the population, which is necessary to overcome density-dependent increases in reproduction, but they require greater resources to implement [83,[88][89][90]. For other dog-hunted species, such as white-tailed deer (Odocoileus virginianus) and raccoons (Procyon lotor), managers should consider the potential to increase the risk of disease when making management decisions. Sustainable wildlife management generally assumes that harvest and infection rates are independent of one another [24]. However, recent models [24] and the results of this study suggest significant covariance between disease risk and hunting method. Harvesting policies that do not account for hunting-induced increases in disease prevalence and associated mortality may underestimate overall mortality rate and the probability of local extirpation of harvested populations.
Our One Health approach extends beyond the management of this invasive species and into conservation and public health policy. By identifying a land use (i.e., legal hunting with dogs) associated with higher disease prevalence and exposure risk, we have also identified areas with increased cross-species transmission potential. PrV is the third leading cause of mortality in Florida panthers (Mark Cunningham, pers. comm.). This population is expanding northward into the Kissimmee River Valley of Florida, where multiple publicly managed lands allow dog hunts. This vulnerable expanding population may benefit from a change in land use management policy to remove hunting with dogs as a hunting option in these areas. These areas are also hotspots for coinfection with Brucella spp., and thus wild pig hunters have a higher likelihood of exposure to this zoonotic pathogen. Increased hunter education about carcass handling and meat processing would be warranted in these public lands that allow dog hunts.