INTRODUCTION
Bovine tuberculosis (bTB) is an infectious disease caused by a bacterium, Mycobacterium bovis, affecting both domestic and wild animals worldwide. In Michigan, USA, bTB was discovered in white-tailed deer (Odocoileus virginianus) following the 1994 firearm deer season in portions of Alpena, Alcona, Oscoda, and Montmorency counties, hereafter referred to as the bTB core area [Reference Schmitt1]. In addition to deer, other wildlife that tested positive for M. bovis in Michigan include elk (Cervus elaphus), coyote (Canis latrans), raccoon (Procyon lotor), black bear (Ursus americanus), bobcat (Lynx rufus), Virginia opossum (Didelphis virginiana; hereafter referred to as opossum), red fox (Vulpes vulpes), and feral cat (Felis catus) [Reference Bruning-Fann2, Reference O'Brien3]. Between 1994 and 2010, 36 cattle (Bos taurus) farms tested positive for M. bovis in seven counties (29 beef herds, seven dairy herds) with transmission likely from a variety of sources [Reference Okafor4, Reference Walter5]. Although considerable white-tailed deer reductions have occurred since 1994, new infections and recurring infections continue to plague Michigan farmers.
Other wildlife reservoir hosts of M. bovis other than white-tailed deer, include the brushtail possum (Trichosurus vulpecula) in New Zealand and the Eurasian badger (Meles meles) in the UK indicating the potential for a multi-host system further complicating M. bovis control [Reference Hutchings6, Reference Ramsey7]. High population densities, simultaneous sharing of dens, and large social groups have contributed to the persistence of M. bovis in brushtail possums and badgers in their respective countries [Reference Ramsey7–Reference Bohm9]. While most research in Michigan has focused on white-tailed deer–cattle interactions for M. bovis transmission due to high deer densities and cattle feeding practices [Reference Garner10, Reference Miller11], little research has been conducted on the prevalence of M. bovis in opossum [Reference Bruning-Fann2, Reference Witmer12]. In addition, no information is available on the interaction of opossum in Michigan with stored feed or hay within the confines of barns or storage facilities that opossum are known to occupy [Reference Garnett13, Reference Atwood14]. Furthermore, no data are available on opossum movements and home range in the northern lower peninsula of Michigan near cattle farms.
Regardless of the efforts of management agencies in Michigan, several cattle farms test positive for M. bovis each year both inside and outside the bTB core zone as do medium-sized mammals that occupy previously depopulated farms. Previous research has documented considerable shedding of M. bovis by free-ranging European badgers but little similar information exists for medium-sized mammals in Michigan [Reference Fitzgerald15, Reference Clifton-Hadley16]. Surveillance of medium-sized mammals in Michigan for M. bovis has found that numerous mammals may be spillover hosts but little research has been conducted to understand the amount of shedding for free-ranging mammals other than white-tailed deer (but see [17]). Studies have shown that opossum are susceptible to aerosolized inoculation [Reference Fitzgerald15], but lateral transmission was not documented when infected opossum were cohabited with naive opossum for 45 days [Reference Fenton18]. Several species of mammals have been documented to use stored feed and farm structures (e.g. barns, feed silos [Reference Witmer12, Reference Atwood14, Reference Hill19]) so the potential for deposition of infective bacteria is possible. Further research on medium-sized mammals and their potential to be more than spillover hosts is warranted.
Although raccoon spatial ecology was studied in Michigan near farms that tested positive for M. bovis [Reference Atwood14], the actual presence of other species on farms would be useful in understanding the role they play in transmission of M. bovis. In addition, surveillance of farms for M. bovis in a variety of small to medium-sized mammals would further our understanding of transmission of M. bovis on cattle farms in Michigan. Our first objective was to conduct surveillance in medium-sized mammals on previously depopulated cattle farms for the presence of M. bovis. Our second objective was to capture and equip opossum with Global Positioning System (GPS) technology to assess potential differences in home range and resource selection between farms inside and outside the bTB core area that have had cattle test positive for M. bovis. Specific interest in opossum is a result of several being diagnosed as positive on farms that had cattle test positive for M. bovis.
MATERIALS AND METHODS
We conducted our study in the northern lower peninsula of Michigan. The 8062 km2 study area included portions of Alcona, Alpena, Montmorency, Oscoda, and Presque Isle counties (Fig. 1). The area encompassed the majority of the cattle farms where M. bovis has been found in Michigan. We defined Deer Management Unit 452 as the bTB core area due to the high prevalence of M. bovis in free-ranging deer and the presence of M. bovis-positive cattle on farms [Reference Okafor4, Reference O'Brien20]. We also selected cattle farms outside the bTB core area that have had cattle test positive for M. bovis for comparison. Vegetation categories that were in our study areas were combined into six classes: (1) developed, including roads, development, and barren land; (2) grass, including pasture/hay fields and native grasses; (3) agriculture, including crops; (4) forest, including upland hardwood stands (Quercus alba, Acer rubrum, A. saccharum), aspen stands (Populus tremuloides, P. grandidentata), hardwood/aspen mixed stands, upland conifer stands (Pinus glauca, P. banksiana, P. resinosa), and hardwood/conifer mixed stands; (5) swamp, including lowland conifer forests/swamps (P. glauca, P. mariana, Thuja occidentalis, Abies balsamea, Latrix laricinea); and (6) farms, including structures such as animal pens, silos, barns, and residential homes. Elevations in the area ranged from 150–390 m above sea level and the mean annual temperature was 6·6°C, the mean rainfall was 72·5 cm, and there was a mean snowfall of 175 cm [Reference Hughey21].
Fig. 1. Cattle farms (stars) and capture locations (asterisks) of Virginia opossum (Didelphis virginiana) inside and outside the bovine tuberculosis core area delineated by Deer Management Unit 452 (dashed polygon) used in surveillance of mammals in the northern lower peninsula of Michigan, 2006–2010.
From April 2006 to October 2011 raccoon and opossum were targeted for surveillance sampling on farms where cattle were depopulated after a positive test for M. bovis and on arbitrarily selected farms with no documented M. bovis in cattle. All surveillance on farms that were depopulated after cattle tested positive for M. bovis occurred within 2 years post-depopulation. Other species were periodically trapped and tested due to their close proximity to barns or at the request of the farmer. Species trapped included: opossum, raccoon, red fox, coyote, woodchuck (Marmota monax), feral cat, Eastern cottontail (Sylvilagus floridanus), and striped skunk (Mephitis mephitis). A complete list of the number of each species collected can be found in Table 1. We collected all animals using cage traps, with the exception of coyotes, which were shot with a rifle. All collections were conducted under a scientific collecting permit issued by the Wildlife Division of the Michigan Department of Natural Resources (MDNR). Animals captured alive were euthanized with a CO2 chamber or by a single gunshot to the head with a .22 calibre firearm [22] and subsequently necropsied in an aseptic mobile laboratory or at the MDNR Wildlife Disease Laboratory. Due to the manifestation of M. bovis in various tissues and organs, we collected lymph node tissues (tonsillar, retropharyngeal, tracheobronchial, mediastinal, mesenteric), a small (generally at least 5 × 5 mm) section of lung tissue, and any lesions observed [Reference Palmer23, Reference Clarke24]. To increase detection, two non-lesioned pools were cultured for each animal, head and thorax tissues in one pool and abdominal tissues in a second pool [Reference de Lisle25]. Gross lesions identified on necropsy were cultured separately. All samples were kept refrigerated until they were shipped on ice to the National Veterinary Services Laboratories at Ames, Iowa, USA (NVSL) for mycobacterial isolation and identification.
Table 1. Surveillance of mammals collected on farms inside the bovine tuberculosis core area (bTB) and outside the core area (non-bTB) in the northern lower peninsula of Michigan, USA, 2006–2010. Numbers within parentheses after proportions are percent prevalence where positive samples were detected
Wildlife tissue samples received at the laboratory were stored at −20°C (or at −70°C for extended periods) until processing. Individual samples were thawed, processed and decontaminated following standard protocols [Reference de Lisle25, Reference Hines26]. Briefly, samples were trimmed, soaked in a 0·065% solution of sodium hypochlorite for 15 min, placed in individual sterilized pint jars, covered with Phenol Red broth and then mixed by securing a blade unit and gasket on the jars, inverting and blending for 30 s or until liquefied. Samples were then transferred to 50 ml centrifuge tubes and decontaminated with a 1·6% final concentration of sodium hydroxide for 7–10 min and neutralized to effect with hydrochloric acid. Samples were then spun in a refrigerated centrifuge at 4800 g for 20 min. Both BACTEC 12B (Becton Dickinson, USA) and MGIT 9360 media (Becton Dickinson) were inoculated with 500 μl of sample suspension. Media signalling positive were examined with a Ziehl–Neelsen acid-fast staining technique. Acid-fast positive bacteria were screened using AccuProbe M. tuberculosis complex nucleic acid probes (Gen-Probe, USA) to determine if the acid-fast bacteria were of the M. tuberculosis complex. Mycobacterium bovis was distinguished from M. tuberculosis isolates by spoligotyping [Reference Roring27]. Information collected on each animal sampled for M. bovis included date of death, date of sample collection, species, sex, and location.
We used the epitools package in Program R (function binom.exact; R Foundation for Statistical Computing, Austria) to calculate the prevalence, the odds ratio (function oddsratio), and 95% confidence interval (CI) to compare differences in prevalence across space, time, and demographic groups (i.e. area). To further describe the effects of exposure to farm environments that had cattle test positive for M. bovis, we also calculated the prevalence of opossum and raccoon that occupied farms that had cattle test positive for M. bovis to those that occupied farms that did not have cattle test positive for M. bovis with a one-tail probability, Z test of proportions.
We live-captured 50 opossum from 1 April to 15 July 2011, using 42 × 15 × 15 inch single-door cage traps in and around farms selected for inclusion in the study (Fig. 1). We handled all opossum without chemical immobilization using commercially available 52-inch reptile tongs to grasp the opossum and remove it from the trap, and then placed it into a cat carrier bag that fully contained the body but exposed the head and neck of the opossum. For each opossum, we recorded sex, morphometric data, marked with individually numbered metal ear tags, injected passive integrated transponder tags (PIT tags) and attached a GPS collar model G2C171 with a very high frequency (VHF) mortality sensor (model G2C171, 115 g; Sirtrack Wildlife Tracking Solutions, New Zealand). The duty cycle for GPS collars was on nocturnally for 12 h and a relocation was attempted every 3 h. To account for high mortality of captured opossum, we re-used GPS collars that were not in use for >1 month. We attempted to locate all opossum at least once per week with the built-in VHF component to determine location and source of natural mortality. We estimated survival of collared opossum using Kaplan–Meier estimators and set our period for analysis at 2 weeks due to the low survival of opossum identified in previous research [Reference Gipson28]; the survival period ended 30 September 2011.
Upon completion of the study, 1 October 2011, we initiated trapping in areas that GPS-collared opossum occupied. We trapped and euthanized opossum using a CO2 chamber [22]. If we were unable to trap GPS-collared opossum, they were located by radio-tracking, captured by reptile tongs, and euthanized in a CO2 chamber. Euthanasia and recovery of study animals was necessary to retrieve data collected in GPS collars for detailed information on movements and home range. Euthanasia also was necessary to determine the presence/absence of M. bovis in study animals by collecting tissue samples as previously described. Upon recovery of GPS-collared opossum, biological samples were collected and kept refrigerated prior to being shipped to the NVSL for M. bovis culture.
For comparison to previous studies, we estimated home range using several fixed kernel density estimators (KDE) [Reference Sunquist29, Reference Gillette30]. We used KDE with the univariate reference or default bandwidth KDE (href). Estimation with href typically is not reliable for use on multimodal datasets because it results in over-smoothing of home ranges [Reference Worton31, Reference Seaman32] so we also used KDE with least squares cross-validation (hlscv) that has been suggested instead of href [Reference Rodgers33]. Finally, in comparison to first-generation methods of KDE (i.e. href, hlscv), we present a second-generation method, the bivariate plug-in bandwidth selection (hplug-in) that performs well even when analysing dependent data common with locations collected with GPS technology [Reference Hall34]. All KDE were calculated using the adehabitatHR and ks packages in Program R (R Foundation for Statistical Computing, Austria) [Reference Calenge35]. A two-way analysis of variance was used to determine differences in size of home range between sex and bTB core area (bTB vs. non-bTB) with significance considered at P < 0·05. We did not test for difference between size of home range and estimator because contrasts among estimators were expected.
We selected farms for equipping opossum with GPS collars based on their proximity to forested habitat that provided similar landscape characteristics preferred by opossum (i.e. proximity to forest patches), to maximize species density in our study design [Reference Atwood14], and that had active cattle operations. Because we were interested in understanding potential differences in use of farms inside and outside the bTB core area, we selected some farms based on previous detection of M. bovis in cattle herds and white-tailed deer; however, it was not possible to select all farms positive for M. bovis because some operations closed and no longer had cattle after a previous positive test for M. bovis. We attempted to trap two opossum near structures (i.e. barns, silos) and two opossum in the surrounding forested habitat at each farm to assess potential for use of structures by residents or visitors to these structures. Each farm served as a replicate in our study design.
We estimated a population-level resource selection function (RSF) using a mixed-effects logistic regression model. For RSF analysis, we used the lmer function (family = binomial) from the lme4 package in Program R (R Foundation for Statistical Computing, Austria). We compared GPS locations to the same number of random points drawn from within individual opossum KDE home range estimated with href in a type III study design [Reference Manly36]. Using logistic regression with use–availability data presents some problems because predicted values are not scaled between 0 and 1 and generally do not reflect true probabilities of resource selection [Reference Manly36, Reference Keating37], but logistic regression can provide an informative and unbiased method for ranking habitat use and for comparing relative probability of use [Reference Keating37, Reference Johnson38]. We used individual opossum as a random-intercept effect in our mixed-effects logistic regression analysis to address issues associated with autocorrelation and uneven sample sizes between individuals [Reference Gillies39]. We created seven a priori models and an intercept model using Akaike's Information Criterion for model selection [Reference Burnham40].
RESULTS
We sampled mammals on 14 farms inside the bTB core area and 16 farms outside the bTB core area (Fig. 1). M. bovis was not isolated from mammals surveyed other than raccoon and opossum (Table 1). Overall, 13 medium-sized mammals (eight opossum, five raccoon) tested positive for M. bovis inside the bTB core area and only one outside the bTB core area. Due to the low overall number of opossum and raccoon positive for M. bovis, odds ratios could not be estimated. Prevalence in opossum on farms inside the bTB core area was comparable (5·6%, 95% CI 2·0–11·0) to prevalence in raccoon (4·0%, 95% CI 1·0–9·0, P < 0·001). Only one opossum tested positive for M. bovis (1%, 95% CI 0·03–5·8) but no raccoon tested positive on farms outside the bTB core area (Table 1). Prevalence in opossum occupying farms that had cattle test positive for M. bovis was higher (6·4%) than opossum occupying farms that never had cattle test positive for M. bovis (0·9%, P = 0·01). Prevalence in raccoon occupying farms that had cattle test positive for M. bovis was similar (1·9%) to raccoon occupying farms that never had cattle test positive for M. bovis (1·7%, P = 0·45). None of the eight opossum equipped with a GPS collar and euthanized at the completion of the study were positive for M. bovis.
We captured and equipped 48 individual opossum with GPS collars and PIT tags from April to mid-July 2011. Due to slipped collars, lost signals, and unrecoverable carcasses, we were not able to document sources of mortality for all opossum and estimates of survival were conducted with censored data. The Kaplan–Meier estimate of survival was 35% for opossum with known fates for the duration of the study period. Mortality sources included drowning in a farm catchment pond, roadkill, harvest by human, and one possibly predation by a black bear. Two opossum appeared to have died for unknown reasons in a subterranean den because a collar signal was detected, but the collar was not recovered despite hours of effort.
Five female and five male opossum met our minimum sample size requirement of ⩾25 locations and being monitored for ⩾30 days for estimates of home range inside and outside the bTB core area (Table 2, Fig. 2). Mean (±s.d.) size of home range for 50% estimates with href, hlscv, and hplug-in were 39·6 ± 28·5 ha, 10·9 ± 11·2 ha, and 14·7 ± 12·5 ha, respectively. Mean (±s.d.) size of home range for 95% estimates with href, hlscv, and hplug-in were 201·3 ± 179·5 ha, 48·1 ± 44·7 ha, and 65·4 ± 49·1 ha, respectively. Mean (±s.d.) size of home range for 50% and 95% estimates was similar by sex (P = 0·791) and bTB zone (inside or outside the bTB core area, P = 0·218), with no interaction of sex and bTB zone (P = 0·365) regardless of the estimator used (Fig. 2). Results of resource selection analysis identified farms and bTB zone as having the most support of all models evaluated. Vegetation and zone provided 56% of model weights and the addition of sex provided an additional 22% (Table 3). Parameter estimates indicated that opossum selected farms over any other habitat and selection of habitat by opossum differed between farms inside and outside the bTB core area (Table 4).
Fig. 2. Size of home range for GPS-collared Virginia opossum (Didelphis virginiana) in the northern lower peninsula of Michigan inside and outside the bovine tuberculosis (bTB) core area, April–September 2011. Home range was reported for 50% and 95% fixed kernel density estimation using the reference bandwidth (href), least-square cross validation (hlscv), plug-in (hplug-in) bandwidth selection.
Table 2. Size of home range (hectares), days on air, and number of locations used to estimate size of home range of GPS-collared Virginia opossum (Didelphis virginiana) around farms in Michigan, USA, April–September 2011. Home range was reported for 50% and 95% fixed kernel density estimation using the reference bandwidth (href), least-square cross validation (hlscv), plug-in (hplug-in) bandwidth selection
Table 3. Models, along with the intercept-only model, identified by Akaike's Information Criterion (AIC), ΔAIC, and AIC weights (wi) from the mixed-effects logistic regression analysis of Virginia opossum (Didelphis virginiana) resource selection in the northern lower peninsula Michigan, USA, April–September 2011
Table 4. Parameter estimates, standard error, and P values for the model with the most support for Virginia opossum (Didelphis virginiana) resource selection in the bovine tuberculosis (bTB) and non-TB core areas of the northern lower peninsula Michigan, USA, April–September 2011
DISCUSSION
We identified two species of medium-sized mammals that were positive for M. bovis on farms after the cattle tested positive for M. bovis. Although overall sample size limited statistical significance, 13 positive medium-sized mammals were from within the bTB core area while only one was from outside the bTB core area. While numerous studies on captive inoculated medium-sized mammals have stated that small- and medium-sized mammals likely do not shed enough M. bovis to be considered as reservoirs for M. bovis [Reference Fitzgerald15, Reference Fenton18], medium-sized mammals such as brushtail possum in New Zealand and badgers in the UK and Ireland are considered primary reservoirs of M. bovis [Reference Corner41, Reference Jackson42]. Furthermore, prevalence of M. bovis in free-ranging white-tailed deer has held steady for the past 4 years at just below 2% [Reference O'Brien43] which is lower than prevalence for opossum and raccoon in our study. Changing management strategies resulting in less dense populations of white-tailed deer, minimizing deer use of farm environments in Michigan [Reference Walter5, Reference O'Brien44], and results from our surveillance efforts suggest that the role of medium-sized mammals as spillover hosts or potential routes of transmission to cattle deserves considerable attention.
Survival of opossum in northern latitudes long enough to ingest and shed enough M. bovis to be considered a reservoir for M. bovis is not known. Survival of opossum has been documented to range from 17% to 63% in various regions of the USA [Reference Gipson28, Reference Sunquist29] and has been difficult to study. Low survival rates in our study (i.e. 35%) appear to be comparable to others with mortality sources ranging from drowning to predation by a black bear. Considerable movements by males has also been documented to contribute to mortality and difficulty in tracking opossum with radio-collars due to dispersal and movements over large distances in a short period of time [Reference Gillette30, Reference Fitch45]. Opossum survival rarely exceeds 2 years and few survive more than a single summer, further complicating research on exposure to and transmission of M. bovis [Reference Sunquist29, Reference Fitch45]. Although our study was not designed to monitor annual survival, opossum and raccoon that were culture-positive for M. bovis on farms previously depopulated of M. bovis-positive cattle suggests direct contact with cattle or an indirect means of transmission of M. bovis.
Although our study did not set out to document direct observations of cattle and opossum, only one direct contact of cattle with a raccoon but none with opossum was found in a separate study in Michigan [Reference Hill19]. A raccoon was identified with infrared cameras coming face-to-face at about 3 m with cattle on one occasion in another study on farms in Michigan [Reference Witmer12]. Although direct transmission of M. bovis from cattle to opossum seems unlikely, our study shows that opossum are more likely to be infected with M. bovis if they are potentially exposed to infected cattle. Opossum being infected with M. bovis could be from an indirect source, but we cannot eliminate the possibility that opossum were infected through direct contact with live or dead white-tailed deer or cattle. Although most farms are depopulated when cattle test positive for M. bovis, complete removal of stored feed and sanitization of feed storage facilities are not required. Opossum were observed and captured in barns, silos, and under storage facilities that contained stored feed.
Estimates of home range of opossum in our study identified sizes comparable to studies in other regions with opossum potentially using more than a single farm (Fig. 3). Several opossum appeared to use more wooded areas in their home range while others concentrated their home range on farms or farm structures. Upon study completion, we captured one opossum inside a barn used to store old household and farm equipment where the owner fed open bags of dry food to feral cats. Other daytime relocations of radio-collared opossum included inside tree root cavities, an underground cavern in forested habitat, a slash pile of wood and debris, and under the floor of an abandoned shed; the latter two locations were within 50 m of farm structures. Although all opossum were not documented to use farms and farm structures exclusively, use of farms was documented on a regular basis for several of our GPS-collared opossum (Fig. 3).
Fig. 3. Locations of GPS-collared Virginia opossum (Didelphis virginiana) that occupied forests and farms on a regular basis in the northern lower peninsula of Michigan inside the bovine tuberculosis core area, April–September 2011.
Third-order resource selection confirmed that opossum used farms and farm structures on a regular basis identifying the potential for exposure to M. bovis through contaminated sources. Although farms were not sampled for contaminated sources in this study, three farms that had cattle test positive for M. bovis in Michigan had 455 samples of soil, water, feed, hay, pasture grass, livestock faeces, deer faeces, and carnivore faeces tested with none positive for M. bovis [Reference Witmer12]. Using infrared cameras in barns, 17·4% of observations were in barns that cattle occupied; however, raccoons were the most common species observed in barns [Reference Witmer12]. Although we did not specifically monitor barn use by our GPS-collared opossum, it is reasonable to infer that resource selection near farms could lead to a greater potential for ingestion of M. bovis on a regular basis for both sexes of opossum. Until opossum are trapped on farms or in natural settings and tested for M. bovis and subsequently tested for shedding of M. bovis in saliva, urine, or faeces, assertions that opossum are not able to shed enough bacteria to act as vectors for M. bovis cannot be conclusively established. Furthermore, until Michigan becomes a bTB-free state and no farms have cattle that test positive for M. bovis, all possible avenues of transmission and survival of M. bovis should be explored.
In conclusion, surveillance identified two wildlife species that tested positive for M. bovis on farms in the northern lower peninsula of Michigan. Although surveillance of mammals and monitoring of GPS-collared opossum were not done on the same farms during the same period, valuable insight was gained from both portions of the study. Opossum infected with M. bovis were more likely to be found in the bTB core area and on contaminated farms (i.e. with cattle positive for M. bovis) suggesting that opossum acquire M. bovis by direct or indirect means when they occupy such environments. Although detection of M. bovis in environmental samples is difficult [Reference Witmer12], use of farms by GPS-collared opossum and recovery of live opossum on farms suggests that indirect transmission of M. bovis to opossum is possible. Further sampling designs to test environmental samples on farms for M. bovis from areas known to attract opossum on a regular basis would further our understanding on the potential for indirect transmission to medium-sized mammals in Michigan. Given the ongoing infections of cattle with M. bovis in Michigan, research should be continued to further our understanding of opossum, raccoon, and white-tailed deer resource selection and infection patterns on and around farms to provide further elucidation of the ecology of M. bovis in the region.
ACKNOWLEDGEMENTS
Funding was provided by the United States Department of Agriculture, Animal Plant and Health Inspection Service, Wildlife Services and Veterinary Services, National Wildlife Research Center. We thank Tony Aderman, Dane Williams and Rex Schenk for planning and assistance in surveillance and trapping efforts. We also thank the Michigan Department of Natural Resources for use of their Wildlife Disease Laboratory for conducting necropsies.
DECLARATION OF INTEREST
None.
