Environmental and behavioral changes may influence the exposure of an Arctic apex predator to pathogens and contaminants

Recent decline of sea ice habitat has coincided with increased use of land by polar bears (Ursus maritimus) from the southern Beaufort Sea (SB), which may alter the risks of exposure to pathogens and contaminants. We assayed blood samples from SB polar bears to assess prior exposure to the pathogens Brucella spp., Toxoplasma gondii, Coxiella burnetii, Francisella tularensis, and Neospora caninum, estimate concentrations of persistent organic pollutants (POPs), and evaluate risk factors associated with exposure to pathogens and POPs. We found that seroprevalence of Brucella spp. and T. gondii antibodies likely increased through time, and provide the first evidence of exposure of polar bears to C. burnetii, N. caninum, and F. tularensis. Additionally, the odds of exposure to T. gondii were greater for bears that used land than for bears that remained on the sea ice during summer and fall, while mean concentrations of the POP chlordane (ΣCHL) were lower for land-based bears. Changes in polar bear behavior brought about by climate-induced modifications to the Arctic marine ecosystem may increase exposure risk to certain pathogens and alter contaminant exposure pathways.

Risk factors of pathogen exposure. Age class was most strongly associated with exposure to C. burnetii ( Table 3). The odds ratio for an age effect indicated that adult polar bears were 3.2 times more likely to be seropositive than subadults. Sex and year of capture were included in two additional models that formed the top set though offered less support (Table 3); however, 85% confidence intervals for sex class and year of capture overlapped zero, indicating these variables may be uninformative 37 (Table 4). The four models in the candidate model set accounted for 70% of all model weight, with the top model having a weight of evidence 1.5 to 2.1 times greater than other candidate models ( Table 4). All other models characterizing factors mediating exposure to C. burnetii had ΔAIC c >2 (Table 3) and no variables from any model had a p-value < 0.05 (Table 4). For T. gondii, three models were included in the top model set (Table 3), with all including an effect of habitat use (i.e., bears spending summer on shore or on the sea ice). Top models also included an additive effect of age class and sex ( Table 3). The odds ratio for habitat use indicated that the odds of polar bears using land during summer being seropositive to T. gondii were 7 times greater than the odds of bears remaining on the sea ice during summer being seropositive ( Table 4). The 85% confidence intervals for sex class and age class overlapped zero, indicating the variables may be uninformative 37 (Table 4). The three models in the candidate model set accounted for 81% of all model weight, with the top model having a weight of evidence 1.7 to 2.0 times greater than the other candidate models ( Table 3). The remaining models characterizing exposure to T. gondii had ΔAIC c >2 (Table 3).
Although >10% of individuals were seropositive for Brucella spp., we were unable to identify a top model for characterizing risk factors of exposure ( Table 3). The two top-ranked models included an effect of habitat use, with the odds ratio indicating that the odds of polar bears using land during summer being seropositive for Brucella spp. were 2.5 times lower than the odds of bears remaining on the sea ice during summer being seropositive. The 85% confidence intervals for the sex class and capture year variables in the two top-ranked models did not overlap zero, though they did in lower-ranked models. However, weight of evidence indicated no clear top model or set, with 68% of all model weight distributed among 5 of the 9 models (Table 3). No model variables had a p-value < 0.05 (Table 4). We did not characterize risk factors of exposure to pathogens to F. tularensis and N. caninum because both had <10% seropositive individuals.

Discussion
We provide evidence: (i) of a possible increase in seroprevalence to Brucella spp. and T. gondii in SB polar bears, (ii) of the potential exposure of SB polar bears to C. burnetii, N. caninum, and F. tularensis, and (iii) that previously-documented climate-induced change in polar bear behaviors 12 our study suggests that over the subsequent eight years, overall seroprevalence of T. gondii may have increased by 81%, while overall prevalence of Brucella spp. may have increased by 27%. The reasons for these potential increases are unclear. There is some evidence to suggest that exposure to T. gondii, which has previously been reported as infecting various Arctic terrestrial mammals 39 , has become more prevalent in the marine environment. For example, there is evidence of exposure to T. gondii for ringed and bearded seals sampled in the Svalbard archipelago 27 , and in ringed seals throughout the Canadian Arctic 40 . Unlike T. gondii, Brucella spp. has long been recognized as infecting both terrestrial and marine hosts inhabiting the Arctic (e.g., caribou, muskox, pinnipeds, and cetaceans 28,[41][42][43]. Our finding of a likely increase in exposure to these pathogens for SB polar bears, concurrent with observations in a variety of Arctic species, suggests that there may be multiple routes of exposure to these pathogens. Changes that have occurred to the Arctic marine ecosystem over the last two decades have been posited as a potential factor altering host-pathogen interactions 44 . Modified environmental conditions, such as the protracted open-water period (i.e., period of time when sea ice concentration over the continental shelf is <15%) during summer and fall, have led to increases in ocean temperature and primary production 38,45,46 . These changes may be facilitating a northward range expansion for some subarctic pathogens and alterations in transmission dynamics. For example, C. burnetii, a widely distributed pathogen, has recently been detected in several species of marine mammals 33,34 , with the previous northernmost case reported for northern fur seals (Callorhinus ursinus) from the Pribilof Islands in the Bering Sea 34 . Our finding of evidence of polar bear exposure to C. burnetii, a linear distance of 1,738 km from the Pribilof Islands, marks the first detection of exposure for an Arctic marine mammal. However it remains unclear if this is a function of northward expansion of C. burnetii, attributable to long-distance migratory behavior of some polar bears 47,48 , a lack of surveillance in marine mammals of the Arctic, or detection of an unknown pathogen that cross-reacts with the test. With regard to F. tularensis and N. caninum, these pathogens may be considered terrestrial-based pathogens in that F. tularensis has long-been associated with lagomorphs and N. caninum is primarily a pathogen of cattle, some species of deer, and dogs 30 . Although antibodies to N. caninum were reported in some herbivores and canids from Alaska 32,49 , and N. caninum DNA has been detected in tissues of 11 (24.4%) of 45 European brown bears (Ursus arctos) from Slovakia 50 , this is the first report of exposure to N. caninum in polar bears in Alaska 51 . Similarly, while F. tularensis antibodies have previously been  Table 3. reported in 14-28% of grizzly bears inhabiting the Alaskan Arctic 52 , no comparable assessments have been made for polar bears in this region. The apparent detection of seropositive polar bears also represents the first evidence for potential F. tularensis and N. caninum exposure, to our knowledge, for Arctic marine mammals.
Risk factors associated with exposure to pathogens and contaminants. Since 2000, the proportion of the SB polar bear subpopulation coming ashore during summer and fall has increased from 6% to over 20%, and the length of stay on land has increased by over a month 15 . While on land, most polar bears visit sites where subsistence-harvested bowhead whale remains are aggregated, and may spend several weeks feeding on the carrion 12,20 , which may increase the risk of density-dependent inter-and intra-specific pathogen transmission 53 .
In modeling factors influencing exposure risk, we found evidence to support the notion that land use, including visiting sites of whale remains, may increase the risk of exposure to T. gondii and decrease risk of exposure to Brucella spp. Oocysts of T. gondii have high environmental resistance and have been shown to persist in soil in temperate environments 54 . However, it is unknown whether they may be able to persist in the soil where whale remains are aggregated and contribute to a chronic risk of exposure. With regards to Brucella spp., some have posited that the exposure pathway may be terrestrial, while others questioned that conclusion 26,55 . Our results suggest there may be a lower risk of exposure to Brucella spp. for polar bears that use terrestrial habitats. However, we note that while our work suggests that bears remaining on the sea ice year-round had a tendency towards a higher likelihood of exposure to Brucella spp., we did not differentiate exposures to the various strains of Brucella spp. which would be useful for identifying transmission pathways. We found no influence of age and sex classes on risk of exposure to T. gondii and Brucella spp., which is similar to previous studies in the SB but differs from studies elsewhere. For example, there was no difference by sex and age class in the prevalence of T. gondii for SB polar bears that were sampled in 2005 and 2006 38 , and for individuals exposed to Brucella spp. from 1982-1999 56 . Conversely, the prevalence of T. gondii in polar bears from the Svalbard archipelago was significantly higher in adult males compared to adult females 27 . Additionally, the prevalence of T. gondii was higher for bearded seals compared to ringed seals in Svalbard suggesting that the difference in polar bear prevalence may result from adult male bears being more likely to prey on the larger-bodied bearded seals than adult female bears 27 . We did find that adult polar bears were more likely than subadults to be exposed to C. burnetii, though the reason for the difference is not clear. Unlike the Svalbard study 27 , we lacked information on the prevalence of pathogens for prey species in the SB. Antibodies to C. burnetii may be long-lived and therefore higher prevalence in adults may reflect cumulative exposure to this bacterium through time. However, while both models from the top model set of exposure to C. burnetii contained the age class explanatory variable, the coefficient estimates were not statistically significant and biological significance should be interpreted cautiously.  Co-infections are common in wildlife populations and can influence infection risk 57 and host fitness 58 . Host immune suppression caused by one pathogen may increase risk of infection from another pathogen 59 , and pathogen-pathogen interactions may have additive effects on host health 60 . We found that possible co-exposures among the five pathogens were rare (i.e., <3%). Several lines of evidence suggest polar bears from the SB may be exposed to multiple stressors capable of undermining population health. For example, gene transcription profiles of bears sampled in the SB and adjacent Chukchi Sea during the later years of our study indicated immune function impairment for SB bears 61 and, in complimentary work involving a subset of the same individuals, a signal of enhanced viral defense was detected 62 . Additionally, the probability of fasting by SB polar bears in spring has increased over time 63 , suggestive of a greater frequency of periods of nutritional stress. Future work should focus on characterizing the co-exposure of polar bears to numerous stressors including viruses and the potential for synergistic effects of co-infection, impaired immune function, and nutritional restriction on population dynamics. Because increased contaminant burdens in marine mammals also can alter the function of physiological pathways associated with metabolism and immunological response 64,65 , they too should be included in assessments of multiple stressors on polar bear population dynamics.

Conclusions
Climate change-driven declines in sea ice habitat appear to be linked to increased use of land 15 , which may influence exposure to some pathogens and contaminants by SB polar bears. For example, we found that exposure to Brucella spp. and mean concentrations of ΣCHL were lower for bears that used terrestrial habitat. As noted above, the reason for reduced exposure to Brucella spp. is unknown, but reduced concentrations of CHLs in land-based bears likely are mediated by a shift from foraging predominantly on ringed seals to scavenging bowhead whale carcasses 12,66 . Polar bears from the Beaufort Sea region have historically preyed primarily on ringed and bearded seals and, to a lesser degree, beluga whales, which all have diets that include both fish and invertebrates 24,67 . As a result, seals and beluga whales occupy a relatively high trophic position and are vulnerable to the bioaccumulation and biomagnification of harmful contaminants like POPs. By contrast, bowhead whales filter feed zooplankton which occupy a lower trophic position than fish and invertebrates. Thus bowheads should be a less contaminated food source with respect to exposure to strongly biomagnifying POPs like CHLs for polar bears. Lower concentrations of other OC pesticides may be less expected in onshore bears as some of the OCs do not biomagnify to the same extent as CHLs, and have been shown to be at a similar or even higher concentration (e.g., ClBzs, HCHs, DDTs) in bowhead whales relative to ringed and bearded seals in this region 68 .
The sources and consequences of exposure of polar bears to the pathogens and contaminants we surveyed remain largely unknown. Both C. burnetii and Brucella spp. are associated with reproductive pathology in a wide range of species 69 . The population-level effects of toxoplasmosis in wildlife are unknown though it has been linked to compromised immune function 70 , as has recently been observed in SB polar bears 61 . While there is no evidence of population-level effects of F. tularensis and N. caninum in large carnivores 71,72 , tularemia epizootics are capable of regulating small mammal populations 73,74 and neosporosis can cause high abortion rates in cattle and neurological disease in dogs 75 . Thus, all of the infectious agents we surveyed have the potential to threaten polar bear population health, which is a concern given their status as "vulnerable" under the International Union for the Conservation of Nature (IUCN) Red List Assessment. Causal investigations are needed to understand the potential impacts that these agents may have on polar bear population dynamics.

Methods
Study area and data collection. The study area was the Alaska portion of the SB, ranging from Demarcation Point (69°N, 141° W) at the US−Canada border in the east to Point Barrow (71°N, 156° W) in the west (Fig. 1). The Beaufort Sea is characterized by a narrow continental shelf, beyond which is some of the deepest water in the Arctic Ocean 76 . The character and spatio-temporal extent of sea ice in the Beaufort Sea has changed substantially since 1979 (i.e., advent of satellite observation of sea ice). Historically, sea ice was persistently available over or adjacent to the continental shelf. However, since the early 2000s, the duration of the open-water period has increased at a rate of ~9 days per decade, which is among the largest rates of increase for the seas of the Arctic Ocean 77,78 . The SB coastal region is characterized by an industrial footprint associated with oil and gas exploration and extraction, generally concentrated along the central coast 79 . There are three communities within the study area that harvest bowhead whales in the fall: Utqiagvik (formerly known as Barrow), Nuiqsut, and Kaktovik (Fig. 1). Remains from the harvest have been sporadically aggregated at Point Barrow (up to 2012) and consistently aggregated at Cross Island and Barter Island (Fig. 1). These locations have served as focal attractors for polar bears during fall 15,17,21 .
Polar bears were encountered from a helicopter opportunistically on the sea ice from mid-March to mid-May, 2007-2014, and on land in August and October of 2008 and 2009. Bears were immobilized with the drug tiletamine hydrochloride plus zolazepam hydrochloride (Telazol ® , Fort Dodge and Warner-Lambert Co.) using a projectile syringe fired from a dart gun. Captured bears were uniquely identified using ear tags and a corresponding lip tattoo. A subset of adult females was fitted with global positioning system (GPS) satellite radio collars. Age at first capture of subadults and adults was determined by analysis of cementum annuli from a vestigial premolar, while dependent young were aged visually 80,81 . Ear punch samples for all captured individuals were genotyped at up to 20 microsatellite loci by Wildlife Genetics International (Nelson, British Columbia, Canada).
Blood samples from subadult and adult bears were drawn into additive-free and EDTA-treated evacuated tubes by venipuncture of either the femoral or jugular veins. All samples were centrifuged at 3,500 rpm for 5 minutes on the day of collection to derive sera and plasma 38 , aliquoted into 2 mL cryovials, stored at −20 °C, and then stored at −80 °C upon return from the field. Each serum sample was tested for antibodies to Brucella spp., C. burnetii, T. gondii, F. tularensis, and N. caninum. Plasma samples from a subset of individuals that had antibodies to one or more of these pathogens were used to quantify exposure to PCBs and OC pesticides.
The buffered Brucella antigen card test and standard plate test (SPT) were performed at the National Veterinary Services Laboratory in Ames, IA. A sample was considered positive if either the card test or SPT was positive. C. burnetii indirect fluorescence assay (IFA) was conducted 33 at Colorado State University in Fort Collins, CO, with a cut-off titer of 1:128 considered positive for prior exposure.
Serological testing for T. gondii and N. caninum was performed at the Animal Parasitic Diseases Laboratory (APDL), Beltsville Agricultural Research Center, Beltsville, MD. The modified agglutination test (MAT) was used for the detection of T. gondii with a cut-off titer of 1:25 82,83 . The MAT is considered highly specific and sensitive, can be used for all host species, and has been extensively validated using isolation of T. gondii for several species, including black bears 84 . For N. caninum testing, a similar agglutination test was used (Neospora agglutination test, NAT) 85 .
To test polar bears for prior exposure to F. tularensis, a commercial slide agglutination test was performed at the U.S. Geological Survey Alaska Science Center, Anchorage, AK, according to the manufacturer's protocol (Becton, Dickinson and Company, Sparks, MD), using serial dilutions of 1:20-1:320. F. tularensis antisera were tested in parallel with all polar bear sera to serve as positive controls. Polar bear sera reacting to antigen at a 1:20 or higher dilution were considered as putatively positive for prior exposure to F. tularensis. Thirty-three percent of samples were run in duplicate as quality control.
Sample extraction and analysis to assess exposure to PCBs and OCs were performed at the Center for Environmental Science and Engineering (CESE) at the University of Connecticut, Storrs, CT, according to established protocols 86,87 , with modifications as described. Approximately 1.5 g of blood plasma was weighed and spiked with the following deuterated surrogate standards: 1,2,4,5-tetrachlorobenzene (1,2,4,5-tetraClBz), and CBs 9, 116 and 156. Samples were mixed and allowed to equilibrate for 30 min., 1 mL of 6 M hydrochloric acid (HCl) was added and mixed, then 3 mL of 2-propanol was added and mixed. A 6 mL aliquot of a 1:1 mixture of methyl t-butyl ether (MtBE):hexane was added, mixed, and followed by ultrasonication for 20 min., then centrifugation at 1000 rpm for 10 min. The organic phase was collected, and the extraction was repeated two further times. To the combined extracts, 6 ml of 1% potassium chloride was added and mixed, proceeded by centrifugation for 5 min @ 1000 rpm and subsequent collection of the organic phase. The extract was then concentrated and subject to further clean up by solid phase extraction (SPE) 86 . Finally, the eluate was exchanged into trimethylpentane, spiked with the internal standards, o-terphenyl and m-xylene, and concentrated to a final volume of 400 μl.
Quality control steps included method blanks and spiked clean dog plasma extracted with each batch of samples. Method blanks were below the detection limit for PCBs and OCs. Accuracy and precision was shown by the n = 4 spiked dog plasma replicate ΣPCB and ΣOC concentrations of 92 ± 7% (±SD) and 84 ± 5%, respectively, of the spiked values (excluding methoxychlor, for which there appeared to be contamination or interference in the dog plasma, but not in the samples or blanks). In the samples, surrogate standard recoveries were 82 ± 12% for 1,2,4,5-tetraClBz and 83 ± 7% the PCBs. Instrument blanks, recovery standards and calibration standards were also included at the start of each run and after every 15 samples. Statistical analyses. We used multiple methods to differentiate between polar bears that spent time on shore (i.e., "onshore" bears) and on the sea ice (i.e., "offshore" bears) in the summer and fall prior to their spring capture. We used location data from radio-collared adult females to confirm land or sea ice use in summer and fall 15 . Additionally, individuals were classified as onshore bears if they were detected (by genetic identification) at hair-snags erected in the fall around the remains of bowhead whales at Point Barrow (2010-2011) 20 or Kaktovik (2012-2014) or from biopsy-darting during fall coastal surveys from 2010-2013. Last, we classified bears as onshore based on the detection of bowhead whale stable isotope (SI) and fatty acid (FA) signatures in hair and fat samples collected during spring capture, 2007-2012 12,66 . Hair and fat biopsy samples were genotyped at 20 microsatellite loci and compared to genotypes from captured bears to identify individuals sampled at hair-snags and during aerial surveys. We used a threshold of >5% bowhead whale composition of the diet to classify an individual as being onshore. Bowhead whales of the Bering-Chukchi-Beaufort population typically winter in the Bering Sea, migrate along the coast northward to the Chukchi and Beaufort seas in spring, and migrate back to the Bering Sea in the fall 88 . As a result, bowhead whales, which are too large for polar bears to hunt, are only regularly available as a food source to polar bears scavenging on land in summer and fall. Thus detection of a SI or FA signature in spring is assumed to be indicative of bowhead whale consumption in the prior fall 12 . A bear was considered to be onshore if it was identified using any of the three previously-mentioned methods. There were no instances in which the classification techniques (i.e., radio-collar, hair-snag, and diet) yielded conflicting results.
For seroprevalence calculations, individuals that were captured multiple times during the study were only counted once and considered seropositive for a given infectious agent if any sample from an individual was positive. This approach was used to minimize overestimates of seroprevalence resulting from long-lived antibodies in individuals recaptured during the course of this study 89 . Observed seroprevalence and associated 95% confidence intervals (CI) were calculated using the package "epiR" 90 in the R software, version 3.2.1 (R Development Core Team 2015).
For pathogens for which seroprevalence was ≥10% of individuals tested, we characterized factors influencing exposure using generalized linear mixed models (GLMM) with a binomial (logit) link. Prior exposure (positive or negative) of individuals to each pathogen was used as the response variable and year of sampling, sex, age (adult and subadult), and summer habitat use (onshore or sea ice) as explanatory variables. We included individual identity as a random effect to control for repeated observations of some individuals. We created a priori candidate models derived from combinations of explanatory variables and compared them using Akaike's Information Criterion corrected for small sample size (AIC c ) to aid in determining top models. We used AIC c to rank and compare models based on ΔAIC c and normalized Akaike weights w i and considered models with ΔAIC c values >2.0 to differ in information content 91 . When faced with model uncertainty, we reported all models, calculated 85% CI for parameter estimates, and considered parameters whose 85% CI overlapped zero to be uninformative 37 .
Exposure to PCBs and OCs were characterized from plasma samples collected in 2013 and 2014 from a subset of individuals used to calculate seroprevalence. Contaminants like PCBs and OCs can compromise the function of the immune response and increase host susceptibility to infectious pathogens 92,93 . We hypothesized that contaminant concentrations, and perhaps susceptibility to some pathogens, would be lower for individuals that spent time on land feeding on bowhead whale remains. To investigate this, we summed individual concentrations of PCBs and OCs and tested for differences in mean concentrations relative to sex class, habitat use, and their interaction using 2-way factorial ANOVAs. Statistical significance was accepted at α ≤ 0.05. All contaminant concentrations were log(x + 1) transformed to approximate normal distributions.
This research was approved under the Marine Mammal Protection Act and Endangered Species Act with U.S. Fish and Wildlife Service permit number MA690038. Capture protocols were approved by the U.S. Geological Survey Institutional Animal Care and Use Committee. All methods were performed in accordance with relevant guidelines and regulations.