ASSOCIATION BETWEEN AGRICULTURAL LAND USE AND WEST NILE VIRUS ANTIBODY PREVALENCE IN IOWA BIRDS

Along with the use of surveillance to monitor the occurrence and frequency of disease in wild animal populations, the ability of public health and wildlife management agencies to predict where the risk of exposure may be greatest, and which wildlife populations are at greatest risk, is important for devising disease management strategies. This is particularly true for West Nile virus (WNV; family Flaviviridae, genus Flavivirus) because this virus is a significant cause of morbidity and mortality in humans, domestic animals (particularly equids), and many wildlife species (Blitvich, 2008). West Nile virus first appeared in the Western Hemisphere in New York, New York, in 1999, and has since spread to all 48 contiguous United States as well as Canada, Mexico, and parts of Central and South America (Komar and Clark, 2006; Centers for Disease Control and Prevention, 2011; Kilpatrick, 2011). This virus is primarily maintained in nature in a bird and mosquito transmission cycle and nonavian vertebrates are usually incidental hosts (Kilpatrick et al., 2007).

The incidence of WNV disease in humans has been observed to increase with urbanization and agricultural/rural land use in eastern and western regions of the United States, respectively (Eisen et al., 2010; Bowden et al., 2011; Chuang et al., 2012). DeGroote et al. (2008) found that human cases of WNV in Iowa were positively correlated with the presence of rural agricultural areas. Geographic differences in the dominant mosquito vector species, their feeding preferences, and their preferred habitats may help explain variation in human disease risk. Specifically, Culex tarsalis is more common in the western United States and is found in more agricultural habitats where it often feeds on mammals, whereas Culex pipiens is more abundant in the eastern United States and occurs more often in urban areas where it feeds primarily on birds (Tempelis et al., 1965; Turell et al., 2005; Bowden et al., 2011; Thiemann et al., 2012).

Associations between land use and WNV exposure have also been documented for birds. For example, Bradley et al. (2008) reported that landscape factors associated with urban areas, such as impervious surface coverage by streets, houses, and buildings in the metropolitan area of Atlanta, Georgia, corresponded with higher antibody prevalence for WNV in songbirds. Similarly, Hamer et al. (2012) found that birds from urban sites had marginally higher exposure to WNV than birds from forested sites in the greater Chicago area, Illinois. The higher antibody prevalence in birds in urban areas may be due to the availability of more vector breeding habitats as a result of urban development (Bowden et al., 2011) as well as trends for reduced avian diversity and abundant competent reservoir species in these areas (Ezenwa et al., 2006; Swaddle and Calos, 2008; Allan et al., 2009; Hamer et al., 2011; for an exception see Loss et al., 2009).

Although birds have been documented to have higher WNV antibody prevalence in urban areas, their exposure to WNV in agricultural areas has received less attention (but see Gibbs et al., 2006). Because birds are the primary amplifying hosts for WNV, if birds have higher WNV antibody prevalence in agricultural areas, they may be useful indicators of WNV risk to humans in these habitats. Our objective was to identify factors associated with the presence of antibodies to WNV in birds in the US Plains states. Specifically, we evaluated whether agricultural land use, sampling year, sampling round, age, and species were associated with WNV antibody prevalence in birds.

Our study took place in central Iowa in a 10.5-km radius surrounding the city of Ames in Story County (42°2′N, 93°37′W) (Fig. 1). Approximately 98% of the land cover in Story County is agricultural, primarily corn and soybeans (Otto et al., 2006). ArcGIS 9.2 (ESRI, Redlands, California, USA) was used to import the 2002 land-cover layer for Story County (courtesy of the Iowa Department of Natural Resources) and to calculate the percentages of the three major land use categories within the study area. These three categories were defined as “natural” (vegetated areas with no regular cultivation), “agricultural” (crop fields with regular cultivation), and “urban” (impervious surfaces created by streets and buildings). An “other” category consisted of land that was barren or covered by water. Ten study sites were selected to produce a gradient spanning low, medium, and high levels of each land use category (Table 1). Each site was 0.79 km², which encompassed the average breeding season home ranges of the common, peridomestic avian species in this area (Birds of North America Online, 2009). The area encompassing each site did not overlap with any of the other sites. Birds were sampled during the breeding season to increase the likelihood that the birds that were antibody-positive for WNV had become infected in the same general area as where we had captured them.

Birds were captured at the center of each study site for 3–5 days on three separate occasions (“rounds”), separated by roughly 4 wk, from May through August in 2009 and 2010. These months were selected to reduce the likelihood of capturing migrating birds and therefore increase the likelihood of trapping birds that had been exposed to WNV locally. Nontoxic permanent markers were used to mark a small area of light-colored plumage on the inner left wing of each bird to identify recaptures. Recaptured birds within a round were not resampled, but birds that were recaptured between rounds were resampled. We used logistic regression power analysis to determine that a sample size of 90 birds per site resulted in ∼80% power to detect a 2.7-fold (or greater) difference in WNV antibody prevalence across a land-use gradient. We targeted our sample size to be able to detect this magnitude of change based on the results of a similar study in Georgia, which documented a nearly fivefold increase in WNV antibody prevalence in birds sampled across sites along a comparable land-use gradient (Bradley et al., 2008).

Feeding stations of black oil sunflower seeds and shelled peanuts were placed at each of the 10 study sites and maintained throughout the sampling season to attract birds. Birds were captured using two mist nets (12 m long, 3 m high, 38-mm mesh) and up to two walk-in traps (0.9 m long, 0.6 m wide, 0.2 m deep) per study site. Birds were removed from the mist nets and walk-in traps and placed into handling sacks to minimize their stress during processing. The species, age, sex, mass, and tarsus length of each bird was recorded. Age was defined as after-hatch-year (AHY), hatch-year (HY), and unknown (U) (Pyle, 1997). Additionally, any anomalies present on each bird (i.e., deformities, old injuries, ectoparasites, etc.) were recorded. Permission to sample birds was granted by the Iowa State University Institutional Animal Care and Use Committee (permit 4-09-6724-Q) and trapping was conducted under Federal Bird Banding (23285) and Iowa Scientific Collectors (SC 871) permits.

The brachial vein of each bird was punctured using a 26.5- or 27.5-gauge needle. Whole blood (≤0.5% of total mass of each bird) was collected and directly deposited into 75-mm heparinized capillary tubes. Capillary tubes were transported to the laboratory in a cooler and centrifuged (2 min, 8,000 × G) after which serum was collected and stored at −20 C.

Serum samples were screened for antibodies to flaviviruses by epitope-blocking ELISA as previously described (Blitvich et al., 2003). The assay was performed using the flavivirus-specific monoclonal antibody (MAb) 6B6C-1 (InBios International, Seattle, Washington, USA). The ability of the test sera to block the binding of the MAb to WNV antigen was compared with the blocking ability of control chicken serum without antibodies to flaviviruses. Data were expressed as relative percentages and inhibition values of ≥30% were considered positive for flavivirus-specific antibodies.

Sera positive for antibodies to flaviviruses by blocking ELISA were further tested by PRNT, the gold-standard serologic technique for flavivirus diagnosis. The PRNTs were performed using WNV (strain NY99-35261-11) and St. Louis encephalitis virus (SLEV; strain TBH-28) with African green monkey kidney (Vero) cells. Viruses were acquired from the World Health Organization Center for Arbovirus Reference and Research at the Centers for Disease Control and Prevention, Division of Vector-Borne Infectious Diseases, Fort Collins, Colorado. Serum titers were described as the reciprocal of the lowest dilution ratio that resulted in ≥90% reduction in the number of plaques (PRNT₉₀). A fourfold or greater difference in PRNT₉₀ titer was used for differential diagnosis.

Only birds known to breed locally based on the Iowa Breeding Bird Atlas (Iowa Department of Natural Resources, 2012) were included in the data analyses. Using the PRNT results, we calculated prevalence of antibody to WNV for each species that had at least one antibody-positive individual. Logistic regression in program R (R Development Core Team, 2011) was used to test for associations between antibodies to WNV in sampled birds and the proportion of a site in agricultural land use, as well as the categorical variables of year, round, species, and age. We did not include sex as a variable because we were unable to confidently identify the sex of approximately 20% of the birds. We evaluated all possible combinations of the above variables as well as an intercept-only model for 32 candidate models. Using corrected Akaike information criterion (AICc) scores, we ranked the candidate models and used model averaging to calculate final estimates, unconditional standard errors, and 95% confidence intervals for each variable based on the AICc weights (Burnham and Anderson, 2002). An AICc weight is an estimate of the probability a particular model is the best model given the model set and the data.

We captured and sampled 1,936 birds. Four of these birds were of migrant species that are not known to breed locally (two Purple Martins [Progne subis], one Swainson's Thrush [Catharus ustulatus], and one Tennessee Warbler [Oreothlypis peregrine]), and were not included in data analyses. The remaining 1,932 birds comprised 37 species, 19 families, and three orders (Table 2). Of these, 990 (51.2%) were sampled in 2009, and 942 (48.8%) were sampled in 2010. The PRNT analysis detected 56 (2.9%) birds with flavivirus-specific antibodies, of which 44 (2.3%) were antibody-positive to WNV, two (0.1%) had antibody to SLEV, and 10 (0.5%) had antibodies to an undetermined flavivirus (or flaviviruses). The antibody prevalence for WNV in 2009 was 3.7% compared to 0.7% in 2010 (Table 3).

Of the 11 species with antibodies to WNV, Baltimore Orioles (Icterus galbula), Northern Cardinals (Cardinalis cardinalis), Mourning Doves (Zenaida macroura), and American Robins (Turdus migratorius) had the highest prevalences at 25, 19, 10, and 9%, respectively (Table 2). However, given the small sample size of Baltimore Orioles (n = 4), this percentage may not accurately reflect actual exposure risk in this species.

The species variable in our models created quasi-complete separation of the data since there were many species with very few individuals sampled or no antibody-positive individuals. To avoid this problem, we compared the effects of species for which we a priori expected higher antibody prevalence for WNV based on existing literature (American Robins, Gray Catbirds [Dumetella carolinensis], House Sparrows [Passer domesticus], Mourning Doves, and Northern Cardinals; Komar et al. 2005; Beveroth et al., 2006; Loss et al. 2009) with remaining species grouped into an “other” category.

The AICc values indicated that the top five models accounted for 99% of the model weights (Table 4). The subsequent model-averaged estimates indicated that bird species and age as well as sampling year were related to the probability of a bird being antibody-positive for WNV, whereas agricultural land use and sampling round were not (Table 5). Consistent with the model-averaged results, these three variables were consistently represented in the top ranking models. Specifically, bird species was included in the top 16 models accounting for >99.9% of model weights, sampling year was included in the top eight models accounting for 99.9% of the model weights, and age was present in the top four models accounting for 97.5% of the model weights. Within species, American Robins, Mourning Doves, and Northern Cardinals were more likely than birds in the “other” category to have antibodies to WNV (Table 5). With respect to age, AHY birds (41/1,462) were more likely than HY birds (1/363) to have antibodies to WNV. Birds sampled in 2009 were more likely to have antibodies to WNV than birds sampled in 2010.

West Nile virus antibody prevalence in bird species sampled in our study (2.3%) was lower than documented in several previous studies (5.3–15.5%; Ringia et al., 2004; Komar et al., 2005; Beveroth et al., 2006). The timing of sampling relative to the emergence of WNV may explain the differences in results. Our study took place roughly 8 yr after WNV first appeared in Iowa, whereas the studies cited above that documented higher prevalences took place in the first few years after the emergence of WNV. Investigators working in the greater Chicago area also found relatively low WNV antibody prevalence in birds (3.5%; Hamer et al., 2012).

Our finding that agricultural land use was not associated with WNV antibody prevalence in peridomestic birds differs from findings that agricultural land use was associated with increased WNV disease in humans (DeGroote et al., 2008; Liu et al., 2008), especially west of the Mississippi River. The difference may be due to the fact that these studies measured WNV disease rather than antibody prevalence (Bowden et al., 2011). These differences may also be attributable to habitat differences between ornithophilic and mammalophilic mosquito vector species. DeGroote et al. (2008) reported higher proportions of Cx. tarsalis, a species that feeds on both mammals and birds, in rural agricultural areas of Iowa, which is where the incidence of WNV in humans was also highest. In contrast, the relatively more ornithophilic Cx. pipiens (Turell et al., 2005; Farajollahi et al., 2011) is most abundant in urban areas. Another potential explanation is variation in the species of birds captured as a function of land use. Although the species with higher WNV antibody prevalence in other studies (i.e., American Robins, Gray Catbirds, House Sparrows, Mourning Doves, and Northern Cardinals) were captured in almost all sites, there tended to be fewer caught in areas with high agricultural land use compared to areas with less agriculture (Table 1). The lack of relationship between agricultural land use and WNV antibody prevalence in birds indicates that birds are not likely to be useful indicators of WNV activity in agricultural areas despite human risk being highest in those areas.

West Nile virus antibody prevalence in 2009 (3.7%) was greater than in 2010 (0.7%). The initial WNV epidemic occurred in Iowa in 2002, and subsequent year-to-year differences may be best explained by environmental variability. The percentage of WNV-infected Culex spp. mosquito pools collected in Iowa in 2009 was also higher than in 2010 (2.5 and 1.3%, respectively; Bartholomay, unpubl. data). The higher mosquito infection rate in 2009 does not appear to be related to mosquito abundance, as there was more than three times the number of Cx. pipiens and four times the number Cx. tarsalis captured per trap event in Story County (where our study took place) in 2010 compared to 2009 (Bartholomay, unpubl. data).

We observed no significant trends in antibody prevalence for WNV in birds according to the round (month) of sampling in 2009 or 2010. These results are consistent with those of Bradley et al. (2008), who reported no significant difference in avian antibody prevalence by month, but differ from reports of seasonal differences in WNV infections in humans, mosquitoes, and crows (DeGroote et al., 2008; Liu et al., 2008; Ludwig et al., 2010; respectively). Although both antibody and virus isolation indicate WNV activity, these data differ in temporal resolution. The viremic period in birds may last up to a week after exposure (Komar et al., 2003), which allows a short period of time for detecting actual infection. However, isolation of virus from an infected human or animal indicates recent WNV exposure. In contrast, neutralizing antibodies to WNV are usually produced within 1–2 wk postexposure (Komar et al., 2003; Styer et al., 2006) and can be long-lasting (Komar et al., 2003; Gibbs et al., 2005), making it difficult to determine when infection occurred. All but one of the antibody-positive birds in our study were AHY birds, and WNV antibody prevalence may reflect exposure in a previous year.

Our finding of high antibody prevalence in Northern Cardinals is similar to other findings that identified this species as having high antibody prevalence for WNV (e.g., Beveroth et al., 2006; Gibbs et al., 2006; Hamer et al., 2012). We also found that Mourning Doves (the only member of Columbidae sampled) and American Robins (the only antibody-positive member of Turdidae sampled) both had higher antibody prevalence than other species, which is consistent with the findings of Beveroth et al. (2006). Mosquito feeding studies have reported preferences of Cx. pipens for American Robins, Mourning Doves, and Northern Cardinals in Chicago, Illinois (Hamer et al., 2009), and of Cx. tarsalis for American Robins in Colorado (Kent et al., 2009). In the Washington, DC, area, Kilpatrick et al. (2006) also found American Robins were overrepresented in mosquito blood meals, but they found that Northern Cardinals were poorly represented. Although we cannot account for WNV-associated mortality in our study, the higher WNV antibody prevalence we found in American Robins, Mourning Doves, and Northern Cardinals may be a function of higher exposure to the virus. The high WNV antibody prevalence combined with the widespread distribution and high abundance of Northern Cardinals, Mourning Doves, and American Robins makes these taxa potentially useful sampling targets for monitoring WNV activity in the US Plains states.

This study was supported by a grant from the Iowa Science Foundation. We thank the field and lab assistants: Katie Patrick, Claudette Sandoval-Green, Stacy Beyer, Kelsey Hoeppner, and Alejandra Navarro. We appreciate the use of laboratory equipment made available by Carol Vleck and the Iowa State University Hybridoma facility. Thanks to Stephen Dinsmore and Jaymi Lebrun for providing mist-netting and bird-handling training. Additionally, we thank Stephen Dinsmore, Dason Kurkiewicz, and Erik Osnas for statistical consultation; Todd Hanson for GIS assistance; Jesse Randall for assisting with manuscript preparation; and Lyric Bartholomay for sharing mosquito surveillance data. Finally, we thank the very gracious landowners for allowing access to their properties for field sampling.