FLEAS OF BLACK-FOOTED FERRETS (MUSTELA NIGRIPES) AND THEIR POTENTIAL ROLE IN THE MOVEMENT OF PLAGUE

The black-footed ferret (Mustela nigripes; hereafter ferret) was historically distributed across western North America, driven to apparent extinction, and rediscovered (Thorne and Williams 1988). The ferret is currently listed on the US Endangered Species List (US Fish and Wildlife Service 1967, 2012). A captive breeding program has been successful in re-establishing some wild populations, yet disease and availability of prey remain significant threats to its survival (US Fish and Wildlife Service 2012). Ferrets are obligate predators of prairie dogs (Cynomys spp.), and both are highly susceptible to sylvatic plague (Gage and Kosoy 2005; Godbey et al. 2006; Matchett et al. 2010), which is caused by the bacterium Yersinia pestis. Plague epizootics can reduce the number of prairie dogs within colonies by up to 100%, causing localized extirpation of both prairie dogs and ferrets (Gage and Kosoy 2005, 2006). Because plague has impacted nearly all ferret reintroduction sites (Godbey et al. 2006), successful reintroduction and establishment of self-sustaining populations of ferrets will not be possible without a better understanding of plague dynamics and the effects of sylvatic plague on ferrets and prairie dogs.

Black-tailed prairie dogs (Cynomys ludovicianus) in South Dakota, US are parasitized primarily by two flea species, Oropsylla hirsuta and Oropsylla tuberculata (Brinkerhoff 2008; Mize and Britten 2016); both are capable of transmitting Y. pestis (Wilder et al. 2008a, b). However, population genetic studies of O. hirsuta suggest black-tailed prairie dogs are not the primary disperser of these fleas within and among their colonies (Jones and Britten 2010; Brinkerhoff et al. 2011). Plague exposure has been reported in many wildlife species including badgers (Taxidea taxus), coyotes (Canis latrans), raccoons (Procyon lotor), red fox (Vulpes vulpes), skunks (Mephitis spp.; Salkeld and Stapp 2006; Brinkerhoff et al. 2009), and swift fox (Vulpes velox; Salkeld et al. 2007). However, among these species, only coyotes have somewhat of an overlap in flea fauna (<5% O. hirsuta) with prairie dogs (Brinkerhoff 2008). Northern grasshopper mice (Onychomys leucogaster) can support high numbers of O. hirsuta, which could exacerbate plague outbreaks in prairie dog populations by increasing the spread of infected fleas among prairie dogs when their populations are high (Salkeld and Stapp 2009; Stapp et al. 2009; Salkeld et al. 2010). Recent research has indicated that O. hirsuta was the most abundant ectoparasite of ferrets in the Conata Basin, South Dakota (Harris et al. 2014).

Our goal was to determine the occurrence and prevalence of flea species on ferrets in the study area. We also evaluated the influence of factors such as sex, age, colony, Julian date of collection, and insecticidal treatment on the abundance of fleas on ferrets. Fleas were also collected from prairie dogs and their burrows at the same site in 2009 and 2010 (Britten and Mize 2010). Because of their close association with prairie dogs, we questioned if ferrets might be capable of transporting infected fleas across and among colonies, thus potentially contributing to local plague epizootics. Fleas from ferrets were tested for Y. pestis in a post hoc assessment to investigate the plausibility that some ferrets could act as incidental transporter hosts of fleas infected with Y. pestis.

We conducted our study on the Lower Brule Indian Reservation (LBIR) in central South Dakota (43°59′N, 99°29′W). Ferrets were first reintroduced to colonies of black-tailed prairie dogs at this site in 2006 (US Fish and Wildlife Service 2012). We collected Y. pestis–positive fleas from burrows of prairie dogs in 2009 and 2010 (Britten and Mize 2010), and the first recorded plague epizootics occurred on the Reservation in 2011 and 2013. In 2010, there were an estimated 124 prairie dog colonies totaling 2,633 ha (range=0.5–443.1 ha, x¯=21.2 ha) on LBIR. Our study focused on 12 colonies totaling 850 ha (range=4.3–443.1 ha, x¯=70.8 ha). Deltamethrin (Bayer, Leverkusen, Germany), an insecticide used to control fleas (Seery et al. 2003; Biggins et al. 2010; Matchett et al. 2010), was applied to the prairie dog burrows of several colonies by site managers during our study to control plague (Fig. 1). The plague epizootic that occurred in 2011 reduced the number of colonies on the LBIR considerably, including some that were included in our study (Fig. 1).

Wild-born ferrets captured during night-time spotlight surveys were examined for fleas during fall 2009–12. Ferrets were captured using wire cage traps (91.5×10×10 cm) placed into the opening of prairie dog burrows occupied by ferrets. Each ferret was anesthetized using isoflurane in a top-opening induction chamber (41×21×27 cm) and systematically combed for fleas using 20 strokes with a standard flea comb (Gage 1999). Ferrets were combed over the induction chamber, which allowed anesthetized fleas to fall into the induction chamber where they remained anesthetized for several minutes. All fleas that fell into the induction chamber were collected for analysis. The white sides and bottom of the induction chamber aided in the detection and collection of fleas. Flea detection rates were not determined in the current study, and we are unaware of any flea detection rate studies for ferrets, but standardized methodology should have allowed for valid comparisons in our analyses (see Eads et al. 2013 for detection rates in prairie dogs using similar combing techniques). Sex and age class (juvenile or adult) of each ferret was recorded (Santymire et al. 2012). Prior to release at their capture sites, all ferrets were treated with fipronil (Merial, Duluth, Georgia, USA) to reduce flea populations and vaccinated with F1-V fusion protein vaccine for protection against sylvatic plague, with recaptures receiving a second dose (Rocke et al. 2004; Powell et al. 2005; Rocke et al. 2008). Therefore, on initial capture, fleas were collected from unvaccinated ferrets, but, on recapture, flea collections were from ferrets that had received at least one vaccination.

Individual fleas were stored in 70% ethanol and identified (Furman and Catts 1982; Lewis 2002). Whole genomic DNA was extracted from fleas using PrepGem (Zygem, Hamilton, New Zealand) extraction kits. The DNA extraction performed on fleas also extracted DNA from Y. pestis when present. The presence of Y. pestis was assessed using a nested PCR developed by Hinnebusch and Schwan (1993), later modified by Hanson et al. (2007) targeting the pla gene, which is specific to Y. pestis (Sodeinde and Goguen 1988).

We estimated the total number of fleas collected across all ferrets, prevalence of flea-infested ferrets (number of infested ferrets/total number of ferrets), prevalence of Y. pestis–positive fleas on infested ferrets (number of Y. pestis–positive fleas/total number of flea-infested ferret hosts), prevalence of Y. pestis–positive fleas collected from ferrets (number of Y. pestis–positive fleas/total number of fleas), and mean intensity of flea infestation per flea-infested ferret (number of fleas/number of infested ferrets), with appropriate 95% confidence intervals (Margolis et al. 1982) using Quantitative Parasitology v. 3.0 (Rózsa et al. 2000). Aggregation was examined for all fleas as well as fleas testing positive for Y. pestis. Degree of aggregation was estimated using a maximum-likelihood method for calculating k, the degree of aggregation of a population of organisms, where k<1 is considered an aggregated distribution (Rózsa et al. 2000; Wilson et al. 2002). The variance-to-mean ratio (σ²:x¯) of fleas collected from ferrets was also used as a measure of aggregation, with values greater than 1 representing a more aggregated distribution (Wilson et al. 2002). Both measures of aggregation were estimated using Quantitative Parasitology v. 3.0 (Rózsa et al. 2000).

We modeled factors that potentially influenced the abundance of O. hirsuta collected from ferrets with generalized linear models (GLMs) fitted with a negative binomial distribution (Shaw and Dobson 1998; Wilson et al. 2002) using program R v. 3.0.2 (R Development Core Team 2013) and the MASS package (Venables and Ripley 2002). We constructed 10 a priori candidate models representing multiple hypotheses and evaluated them using an information theoretic approach, with models ranked via Akaike's information criterion adjusted for small sample size (AIC_c; Burnham and Anderson 2002). We considered factors such as ferret sex and age, Julian date, year, the colony where individual ferrets were captured, and whether a colony had been treated with deltamethrin within the previous 6 mo prior to ferret capture. To confirm, we compared goodness-of-fit of our data to both negative binomial and Poisson distributions using a log likelihood test in the pscl package (Zuur et al. 2009). Our GLM analyses did not include fleas from recaptured ferrets because flea combing and fipronil treatment likely affected the number of fleas on ferrets during subsequent captures.

A total of 528 fleas were collected from 67 unique wild-born ferrets, some of which were recaptured (30 recaptures; Table 1) from 2009 to 2012. Of the 528 fleas collected, 98.3% (n=519) were O. hirsuta (Table 2). Sample sizes of the other flea species collected were not large enough to include in analyses. The other nine fleas collected were Epitedia wenmanni (n=1), Peromyscopsylla selenis (n=1), unknown Oropsylla spp. (n=3), fleas from Family Ceratophyllidae (n=2), and unidentifiable fleas (n=2). All 10 fleas testing positive for the presence of Y. pestis were collected from male ferrets, but our power to detect significant differences in the number of Y. pestis–positive fleas between male and female ferrets was insufficient (power=11% using NQuery Advisor [Statsols, Boston, Massachusetts, USA], based on Fisher's exact test with an α=0.05 and a two-sided test).

The number of Y. pestis–positive fleas collected from individual flea-infested ferrets ranged from 0 to 5, and the annual prevalence of Y. pestis–positive fleas ranged from 0% to 2.9% (Table 3). Annual prevalence of ferrets carrying Y. pestis–positive fleas ranged from 0% to 22.2% (Table 4) with a mean intensity of infestation ranging from 0 to 2.5 Y. pestis–positive fleas per ferret carrying Y. pestis–positive fleas (Table 4).

Both the variance to mean ratio and aggregation coefficient k showed infestation by O. hirsuta on ferrets was highly aggregated across all years (Table 2). Yersinia pestis–positive fleas on ferrets also showed an aggregated distribution. However, the variance-to-mean ratio was close to 1, indicating a low threshold of aggregation (Table 4). Aggregation could not be assessed for Y. pestis–positive fleas infesting ferrets collected in 2010 and 2011 because sample sizes were too low.

The top two candidate models, based on AIC_c scores, both included sex and year as covariates; treatment with deltamethrin was included in the second competing top model (Table 5). We selected the model that contained sex and year as the single best approximating model because it had the lowest ΔAIC_c score and the fewest variables (Table 5). We interpreted the other competing top model with ΔAIC_c<2 as a nested version of the best model and considered the additional covariate for deltamethrin treatment as an uninformative covariate (Arnold 2010) and did not consider this model further.

Our analysis indicated that male ferrets supported significantly larger infestations of O. hirsuta (χ²=7.45, P=0.006). Males had almost three times the average number of O. hirsuta than females (males x¯=8.64±14.04, females x¯=3.03±4.97). Year also had a significant influence on the number of O. hirsuta recovered from ferrets, where ferrets examined in 2010 had a lower mean abundance of O. hirsuta than in 2009, 2011, and 2012 (χ²=17.32, P<0.001). The mean number of O. hirsuta collected from ferrets ranged from 2.37 to 11.62 from 2009 to 2011 (2009: x¯=4.96±8.10; 2010: x¯=2.37±4.66; 2011: x¯=11.62±16.39; and 2012: x¯=11.57±17.06). In our study, treatment with deltamethrin did not influence the abundance of O. hirsuta on captured ferrets (χ²=2.3, P=0.13). Abundance of fleas from ferrets captured on treated colonies (x¯=5.0±11.17) were similar to those captured on untreated colonies (x¯=6.14±10.8). Additionally, there were no significant differences in the abundance of O. hirsuta between juvenile and adult ferrets, standardized Julian date, or colony of capture.

In our study, O. hirsuta was the primary flea species associated with ferrets and some fleas collected from ferrets carried Y. pestis. The O. hirsuta is one of the primary fleas of prairie dogs (Salkeld and Stapp 2008; Jones and Britten 2010; Mize and Britten 2016) and has been previously reported as a common parasite of ferrets (Harris et al. 2014). The flea species collected from these ferrets strongly resembles flea faunas found on black-tailed prairie dogs (96.8%; 447/462; O. hirsuta) and in their burrows (75.7%; 240/317; O. hirsuta) at our study site (Britten and Mize 2010). Oropsylla hirsuta is known to carry Y. pestis and has a significant role in plague epizootic dynamics (Wilder et al. 2008a, b).

Black-footed ferrets experienced a severe population bottleneck, a process often associated with a loss of parasite diversity (Gompper and Williams 1998). There is a paucity of information in the historic record about parasites specific to ferrets before their population decline in the wild (Gompper and Williams 1998). Rhadinopsylla fraterna was collected from a ferret recorded near Jordan, Montana, in 1937; this flea species was also recorded from a suite of rodent hosts as well as long-tailed weasels (Mustela frenata, Holland 1985). Oropsylla hirsuta has also been collected from ferrets in Powder River County, Montana (Hubbard 1947) and near Chamberlain, South Dakota (Boddicker 1968). More recently, reintroduced ferrets captured in the Conata Basin, South Dakota, yielded O. hirsuta, O. tuberculata, and Pulex irritans as well as E. wenmanni and Peromyscopsylla hesperomys (Harris et al. 2014). Here we report the first collection of Peromyscopsylla selenis from a ferret, a flea documented to be associated with a variety of small rodents (Hubbard 1947).

In our study, male ferrets had a higher abundance of fleas than female ferrets. Morphological, physiological, and behavioral characteristics might explain why males harbor more fleas than females; though this difference is not unique to ferrets (Brinkerhoff et al. 2006; Krasnov 2008; Lopez et al. 2013). Male ferrets are almost twice as large as females (Anderson et al. 1986), thus males have more surface area to support a higher population of ectoparasites. Although males are usually less than half as abundant as females within a typical ferret population (Forrest et al. 1988), the home ranges of male ferrets are about twice as large as the home ranges of female ferrets (Jachowski et al. 2010; Eads et al. 2011; Livieri and Anderson 2012) and typically overlap several female ferrets' ranges (Powell 1979; Livieri and Anderson 2012; Eads et al. 2014). This pattern of space use provides males a high potential to receive fleas from prairie dogs, burrows, and other ferrets, over a relatively large area. Variation in infestation based on host sex also may be due in part to differences in host defenses and physiology, flea phenology, or environmental factors (Krasnov 2008; Krasnov et al. 2012; Kiffner et al. 2013). While all Y. pestis–positive fleas collected were from male ferrets, our power to detect significant differences in the number of Y. pestis–positive fleas between male and female ferrets was insufficient.

Our models indicated there was a difference in O. hirsuta abundance between years. Variation between years could be due to annual variation in weather such as temperature and precipitation. Fleas are sensitive to changes in microclimate of their hosts' nests or burrows such as temperature and relative humidity changes, which can alter egg production, development time, and survival rates (Marshall 1981; Krasnov 2008). In addition, fleas are also sensitive to larger scale weather conditions; Stenseth et al. (2006) suggested warmer spring temperatures and higher summer relative humidity increased flea burdens on great gerbils (Rhombomys opimus) in Kazakhstan. However, Eads et al. (2016) observed the highest flea densities on black-tailed prairie dogs during the driest year of a 3 yr study in New Mexico.

Contrary to our expectations, treatment of colonies with deltamethrin did not influence the abundance of fleas collected from captured ferrets. Deltamethrin is a water-insoluble insecticide that is commonly used to control fleas and prevent deaths of prairie dogs and ferrets from plague (Biggins et al. 2010; Matchett et al. 2010). Several factors might explain this observation. During this study, not all colonies were treated with deltamethrin. Ferrets that were captured on treated colonies could have been born and reared on untreated colonies but captured during dispersal on treated colonies carrying fleas from their natal, untreated colonies. The effectiveness of the insecticide might have been diminished by the length of time between treatment and ferret capture on specific colonies, excessive precipitation, or a combination of both. Another possible explanation might include human error during the deltamethrin application process: an inadequate amount of insecticide could have been applied during treatment. This finding reflects the difficulty in determining the spatial scale at which deltamethrin should be applied and the efficacy of treating relatively small colonies that comprise a relatively small proportion of the total landscape; targeting small islands for treatment may not effectively reduce flea infestations when the surroundings may be an ocean of plague.

While we did not examine individual hosts for exposure to Y. pestis, we did provide evidence that ferrets carried Y. pestis–positive fleas before major plague epizootics occurred in prairie dog populations in 2011 and 2013. Prevalence of Y. pestis–positive fleas collected from ferrets at Lower Brule ranged from 0% to 2.8% during our study, which was similar to Y. pestis prevalence of 3% in fleas collected from prairie dog burrows in 2009 and 2010 (Britten and Mize 2010).

To the best of our knowledge, there has been only one other study to examine prevalence of Y. pestis in fleas collected from ferrets. Mouse inoculation and single-reaction PCR tests failed to detect Y. pestis in about 700 fleas collected from ferrets (n=70) between 1996 and 2005 at the U. L. Bend National Wildlife Refuge (NWR), Montana (Matchett et al. 2010). However, in 2006–07, Y. pestis was detected in fleas (n=373) collected from two male and one female ferrets with a prevalence of ferrets infested with Y. pestis–positive fleas of 9.4% (3/32) at U. L. Bend NWR using the same nested PCR protocol as our study (Matchett et al. 2010). Failure to detect Y. pestis in fleas collected from ferrets between 1996 and 2005 at U. L. Bend NWR may be due, in part, to a difference in the methods used to detect Y. pestis, because the nested PCR assay is more sensitive than mouse inoculation or single-reaction PCR assays (Hanson et al. 2007; Matchett et al. 2010).

Our data provided evidence that ferrets are capable of transporting Y. pestis–infected fleas within and among prairie dog colonies. It appears plausible that vaccinated ferrets could contribute to the spread of Y. pestis by promoting connections among prairie dog coteries in a similar manner as has been reported for northern grasshopper mice (Stapp et al. 2009; Salkeld et al. 2010; Kraft and Stapp 2013) and by promoting connections among colonies. Ferrets, whose territories can range between 56 and 137 ha (Jachowski et al. 2010; Fagerstone and Biggins 2011; Livieri and Anderson 2012), would be expected to move frequently among coteries of black-tailed prairie dogs that range in size from 0.05 to 1.01 ha (Hoogland 1995). Additionally, Richardson et al. (1987) reported that during the winter months, nightly movements of ferrets averaged 1406 m, were nonlinear, and increased during the breeding season, and Biggins et al. (2006) reported that ferrets are capable of traversing their entire home ranges in 12 hr. Male ferrets may be particularly important in the movement of fleas and Y. pestis across the landscape because males have larger home ranges than females (Jachowski et al. 2010; Eads et al. 2011; Livieri and Anderson 2012) and males disperse farther and almost always leave their natal areas (Clark et al. 1986; Forrest et al. 1988), whereas females usually disperse short distances, if at all (Biggins et al. 1986; Forrest et al. 1988). These space use patterns are noteworthy because, as we have reported here, male ferrets harbor larger infestations of fleas.

Movements among prairie dog colonies by ferrets are not uncommon (Biggins et al. 1985; Richardson et al. 1987; Forrest et al. 1988) but would be much less frequent than movements among coteries. Intercolony movements would likely be influenced by colony size, distance between colonies, season, sex, and perhaps intra- and interspecific competition. At our study site, intercolony movements most often occurred by juveniles during dispersal (September–November; Forrest et al. 1988) and adult males during the breeding season (March–April). During epizootics, as areas of each affected colony succumb to plague and prey disappears, ferrets might move to other areas of the colony or to other colonies seeking prey, potentially carrying fleas and Y. pestis with them and infecting additional areas, exacerbating disease transmission.

Because of the rarity of ferrets, they are not likely to be significant agents in dispersing Y. pestis on a large scale. However, given their close association with prairie dogs, overlapping flea faunas, and movements within and between colonies, ferrets might have an incidental but potentially important role in transporting infected fleas, possibly leading to plague epizootics on local scales.

We thank the Bureau of Indian Affairs, US Fish and Wildlife Service, and South Dakota Department of Game, Fish, and Parks Wildlife Diversity Small Grants Program for providing funding. Additional resources were provided by the Lower Brule Sioux Tribe and the University of South Dakota. We are grateful for the significant contribution to the field effort for this study made by personnel with the Lower Brule Sioux Tribe Department of Wildlife, Fish, and Recreation. We thank G. Kolenic from the Center for Statistical Consultation and Research at the University of Michigan. Comments by D. Biggins, M. Dixon, R. Matchett, and an anonymous reviewer greatly improved this manuscript.