Among wildlife diseases, sylvatic plague is of exceptional relevance for conservation biologists and land managers, given its direct impact on species conservation (Salkeld 2017), ecological relationships in wildlife communities (Biggins and Kosoy 2001), overall ecosystem functionality (Abbott and Rocke 2012, Eads and Biggins 2015), and implications for human health (Abbott and Rocke 2012). The pathogen responsible for sylvatic plague, the bacterium Yersinia pestis, was introduced from Asia to North America in the early 1900s, and today is ubiquitous from Texas to southern Canada (Richgels et al. 2016). It typically circulates among rodent hosts and their fleas but can spill over and infect more than 200 different mammalian species including lagomorphs, carnivores, and humans (Biggins and Kosoy 2001, Abbott and Rocke 2012). Following recent outbreaks of infection in humans, plague is currently classified by the World Health Organization as a re‐emerging zoonotic disease (Richgels et al. 2016).
Prairie dogs (Cynomys spp.) are particularly susceptible to plague, and their populations can suffer high mortality (>90%), colony die‐offs, and local extinction (Cully and Williams 2001). Consequently, extensive research has been conducted in the United States on sylvatic plague ecology (Tripp et al. 2009, Friggens et al. 2010, Eads et al. 2013) and management (Biggins et al. 2010, Matchett et al. 2010, Tripp et al. 2016, Rocke et al. 2017) with respect to the conservation of black‐tailed prairie dogs (Cynomys ludovicianus, hereafter BTPD) and their endangered specialist predator, the black‐footed ferret (Mustela nigripes; Belant et al. 2015).
Black‐tailed prairie dogs in Canada are federally protected as a species at risk and have recently been up‐listed from Special Concern to Threatened, given limited population size, fragmentation, isolation, and the increasing threat posed by drought and sylvatic plague (Canada Gazette Vol. 152, N. 4;
Managing disease and extinction risk under these circumstances can seem daunting in the face of limited, disjunct sets of data, with inadequate data commonly perceived as an impediment to effective conservation (Cardoso et al. 2011). Here, we aim to demonstrate that individually limited or incomplete data need not be a barrier to evidence‐based conservation management where multiple sets of data can be integrated with appropriate consideration of uncertainties to improve our understanding of ecologically complex systems. We combined data on prairie dog population ecology, plague surveillance, and plague management collected in Grasslands National Park (Saskatchewan, Canada) during 2013–2017 to evaluate plague transmission in the northernmost BTPD ecosystem and help inform future mitigation strategies. Our analyses address the following specific objectives: (1) determine how BTPD population dynamics respond to current plague mitigation measures; (2) assess temporal patterns in the prevalence of fleas and Y. pestis; (3) identify what factors influence flea prevalence on hosts; and (4) characterize the composition of the flea community.
Our study was conducted in the West Block of Grasslands National Park (hereafter GNP), in southwestern Saskatchewan (49°07′ N, 107°45′ W). The area encompasses over 423 km2 of mixed‐grass prairie and represents the northernmost periphery of the BTPD geographic range (Stephens et al. 2018). GNP protects 17 of the 19 BTPD colonies that existed in Canada at the time of data collection (one previously extirpated colony has since been recolonized; Fig. 1). The data analyzed here were collected for alternative purposes by three different organizations, resulting in variable sampling effort and imperfect spatiotemporal overlap. An overview by year and colony of plague mitigation measures (insecticide application and oral vaccines) and data collected (visual counts, colony extent, and flea sampling from hosts and burrows) is provided in Appendix S1: Table S1, with detailed descriptions below.
Enlarge this image.1 Fig.. Black‐tailed prairie dog (Cynomys ludovicianus) colonies (yellow) in the study area shown in their larger extent during 2013–2017. Largest extent of sections treated with deltamethrin (2013–2017; shaded), colonies treated with SPV (2015–2017; red outlines), plague‐positive fleas (green dots), and plague‐infected host carcasses (prairie dogs and ground squirrels; red crosses) are displayed. Colonies inside Grasslands National Park holdings (light brown): Broken Hills (BRH), Dixon Main (DXM), Dixon North Hill (DNH), Dixon South (DXS), Dixon Southwest (DSW), Dixon West (DXW), Ecotour (ECO), Larson (LAR), Monument A (MOA), Monument B (MOB), North Gillespie (NGP), Police Coulee (PCL), Sage (SGE), Snake Pit (SNK), South Gillespie (SGP)*, Timbergulch (TIM), Walker (WLK), 70 Mile (70M) Colonies outside Grasslands National Park holdings: Dixon Community Pasture (DCP) and Masefield (MAS). Inset map shows the location of Grasslands National Park (green), Saskatchewan (SK), in relation to Canada (white in inset) and USA (gray in inset), with the international border indicated by a yellow dashed line in the main map. *The colony was extinct between 2011 and 2015 and thus not part of this study.
Since the first detection of Y. pestis in the local BTPD population in 2010 (Antonation et al. 2014), GNP has applied mitigation measures for sylvatic plague. During our study, approximately one third of BTPD colony area in GNP (4–8 colonies, mean ± SD = 340.8 ± 155.4 ha; Fig. 1, Appendix S1: Table S1) was treated with an insecticide each year by applying 4–5 g of 0.05% deltamethrin powder (DeltaDust, Bayer Crop Science, Guelph, ON, Canada) per burrow, following label instructions and previously published methodologies (Tripp et al. 2016). Selection of treatment areas was primarily driven by public health considerations, such as proximity of colony to roads and overall ease of access for park visitors. Treatment generally occurred in fall (October–November) to maximize suppression of fleas during the season of plague activity in the following year (Tripp et al. 2016). However, localized insecticide application (i.e., burrow dusting) occurred in summer 2017, in response to detection of plague in fleas or hosts (see Suspected plague outbreaks and plague‐caused mortality).
Additionally, during 2015–2017, two BTPD colonies (DXW and WLK; Fig. 1) were selected for experimental treatment with sylvatic plague vaccine, distributed at a rate of 100 baits/ha according to methodologies and timing described by Rocke et al. (2017).
Relative abundance of the BTPD population on colonies (and its inter‐annual variation) was estimated through visual counts (Severson and Plumb 1998). Surveys were conducted annually from 2010 to 2016 on 10 colonies (Fig. 1; Appendix S1: Table S1), selected to be representative of the different habitats (i.e., upland and valley bottoms) occupied by the local BTPD population. A total of 24 plots (16 plots = 3.61 ha, 8 plots = 4 ha) were surveyed over three consecutive days during June–September. Surveys started 1.5 h after sunrise and visual count scans were conducted from an 8 ft ladder every 10 min using 10 × 50 mm binoculars for a total of 3 h (18 scans). Numbers of BTPD and RGS were recorded with each scan. Density estimates were calculated from the maximum count of BTPD over the three days divided by plot area (Severson and Plumb 1998). Given challenges with detecting animals in tall grass, RGS densities were not considered reliable and therefore not included in analyses.
Areal extent (hectares; ha) of all active BTPD colonies was measured every two years through field perimeter mapping conducted in September–October (Appendix S1: Table S1). Colony extent was assumed to be the area among and encompassing active burrows, with burrows classified as active based on observations of fresh scats within 0.5 m of the entrance, fresh digging or tracks, or a defined structure to the burrow with no obstructions (Biggins et al. 1993). GPS locations of active perimeter burrows were taken within 25 m of the previous active burrow and saved on Garmin Map60csx/Map60cx GPS units. When active burrows were located >25 m from the closest burrows, colony subsections or islands were defined and mapped as separate entities within the same colony. Waypoints of perimeter burrows were uploaded and colony extent calculated using ArcMap 10.4 (ESRI, Redlands, California, USA). For colonies treated with deltamethrin, areal extent was computed separately for dusted vs. non‐dusted sections.
Since 2013, fleas have been collected by burrow swabbing according to Seery et al. (2003), with few modifications. We randomly sampled 30 burrows/colony/session and swabbed each burrow once. Each year, between two and four swabbing, sessions were conducted from April to September in both dusted and untreated sections of 10–14 colonies (Appendix S1: Table S1). Upon arrival at the field station, plastic bags containing swabs were placed in a freezer (−4°C) for 24–48 h to kill the fleas, which were then collected with the help of a stereomicroscope and stored in 2% saline solution at −4°C until shipped for identification and diagnostic testing at the National Microbiology Laboratory (NML, Public Health Agency of Canada) in Winnipeg. Upon arrival, fleas were promptly sexed, counted, and identified to species using a dissecting microscope and the taxonomic keys of Holland (1985), and typically pooled by sex, species, and burrow before diagnostic testing. In 2017, the majority of fleas were not sexed or identified to speed up diagnostic testing given a suspected plague outbreak. Representative samples of the predominant species of flea were prepared onto slides and deposited as voucher specimens in the Wallis/Roughley Museum of Entomology, Department of Entomology, University of Manitoba, in Winnipeg.
Fleas were also collected from BTPD and RGS handled during a long‐term mark–recapture study carried out on the same colonies (see host flea collection, Appendix S1: Table S1). Individual hosts were combed 30 times with a toothbrush to dislodge fleas. All fleas collected from the animal and the handling bag were counted and then stored at 4°C until sent for identification and testing for Y. pestis (as described for those collected through burrow swabbing). All protocols were approved by the University of Calgary Life and Environmental Sciences Animal Care Committee (BI 2008‐47; BI 2007‐41), the Calgary Zoo Biological Research Review Committee (Protocol 2008‐04; 2011‐03), and the Parks Canada Research and Collection Permit System (permit n. GRA‐2013‐14017 and GRA‐2017‐24697).
From April to September, BTPD colony activity was monitored through monthly site visits aimed at identifying any dead zones or areas of low activity that might have been caused by a plague outbreak. If applicable, colony sections of low or no activity were further searched for carcasses and/or burrows were opportunistically swabbed for collection of fleas. Furthermore, any BTPD or RGS carcass opportunistically found on colony while conducting other monitoring or active management programs was immediately collected, frozen, and submitted to the Canadian Wildlife Health Cooperative (Saskatoon, Saskatchewan).
Presence of Y. pestis within the submitted flea or rodent carcasses were assessed using a real‐time PCR assay based on presence or absence of a segment of the pla gene. Fleas were briefly dipped in 70% ethanol and macerated through the use of a dounce homogenizer or addition of silica beads. DNA extraction was completed on a portion of the slurry using the Qiagen QIAamp mini kit with an overnight incubation step for tissue lysis, as described by Antonation et al. (2014). The remaining portion of the slurry was retained for culture. During 2013–2016, flea samples were tested individually, with typically 1–3 fleas per vial (max 30, but <5% testing vials had >10 fleas); however, in 2017, to maximize sensitivity of the PCR assay, fleas were systematically pooled by BTPD colony prior to DNA extraction (Cully et al. 2000), using a standardized flea pool size of 10 individuals as possible. Samples positive for the pla target were then confirmed by additional genomic signatures specific to Y. pestis and plated to selective (CIN) and non‐selective (SBA) growing media to assess for the presence of viable bacteria. For the purposes of analysis, we considered flea samples to be infected with Y. pestis when successful amplification occurred at three target regions with cycle threshold (Ct) values <38. Whole‐genome sequencing utilizing standard protocols for the Illumina MiSeq platform confirmed the presence of Y. pestis DNA signature at this cutoff and absence in flea samples that did not meet these criteria.
For BTPD and RGS carcasses, initial diagnosis was made based on gross and histological lesions suggestive of septicemia, as well as bacterial culture of three of the following tissues: lung, liver, kidney, and spleen, depending on preservation. Initial bacterial culture was performed by Prairie Diagnostic Services, Saskatoon, using blood and MacConkey agar. If bacterial morphology was consistent with Y. pestis, specimens were sent to NML for bacterial culture and molecular testing following methodologies described within this section. We considered host carcasses to be positive for Y. pestis when the bacterium was confirmed through both culture and PCR assay.
Sampling effort to determine the prevalence of fleas in burrows (i.e., the proportion of burrows with at least one flea) varied due to fluctuating operational resources (total swabs/year: min = 331, max = 2128, mean = 1293.8), and inconsistent timing of samples across years. Data on flea prevalence in burrows were therefore generally bootstrapped prior to use in analyses to account for variable sampling effort. Bootstrapping involved 1000 iterations, each randomly selecting 25 burrows per dusting treatment and each combination of month and year (monthly data), each year (annual data), or each spring (April–May, when sampling was most consistent between years, with an average of 374 ± 45 burrow swabs in spring each year).
Data regarding flea prevalence on hosts (i.e., the proportion of hosts carrying at least one flea) were also bootstrapped given the variable number of hosts captured during mark–recapture sessions (total captures/year: min = 113, max = 315, mean = 207.2 for BTPD, min = 130, max = 308, mean = 202.8 for RGS). We randomly selected 40 individual BTPD or 40 individual RGS per dusting treatment per year over 1000 iterations to determine mean annual prevalence of fleas on hosts separately for each species.
To investigate the effect of deltamethrin application on the relative abundance of BTPD, we used the lme4 package (Bates et al. 2015) in R (R Core Team 2018) to implement generalized linear mixed‐effects models (GLMM) with a negative binomial error structure. We used a negative binomial error structure rather than Poisson because the BTPD count data were over dispersed. The models included random intercepts on plot ID nested in colony to account for repeated measures and non‐independence of plots within a colony, respectively. Burrow dusting was tested as a fixed effect, as was environmental stress and its interaction with burrow dusting. Burrow dusting was computed either as a numeric cumulative number of years of treatment or as a binary categorical variable reflecting whether or not the plot had been dusted in the previous year. Environmental stress was coded categorically based on values of the Standardized Precipitation Evapotranspiration Index (SPEI). Standardized Precipitation Evapotranspiration Index is a temperature‐sensitive drought index that reflects the climatic water balance, calculated by taking the difference between precipitation and potential evapotranspiration and then standardizing the resulting values relative to a reference period (Vicente‐Serrano et al. 2010). Standardized Precipitation Evapotranspiration Index can be calculated for specifically relevant time windows (Vicente‐Serrano et al. 2010); we chose a six‐month period between April and September, corresponding to the local growing season for grassland vegetation, since previous research suggests that environmental stress affects BTPD primarily via impacts on forage (Stephens et al. 2018). We then used the R library SPEI (Beguería et al. 2017) to calculate SPEI values from 2007 to 2016 based on precipitation and potential evapotranspiration (Penman‐Monteith) data obtained from Agriculture and Agri‐Food Canada Drought Watch (
We ran an intercept‐only model as well as models with only burrow dusting, only environmental stress, both burrow dusting and environmental stress as additive effects, and the full model including their interaction. With the help of R package MuMIn (Barton 2018), we then used the Akaike information criterion corrected for small sample sizes (AICc) to determine the top model(s) (i.e., delta AICc < 2) and compute model weights and model‐averaged parameters (Burnham and Anderson 2002). Predictors with model‐averaged parameters whose confidence intervals excluded zero were judged most important. We assessed the fit of the top model(s) based on marginal R2 values (Nakagawa et al. 2017). Although data on the relative abundance of BTPD span from 2010 to 2016 across 24 plots in 10 colonies, data from 2010 were excluded from the model since dusting had not yet occurred (Appendix S1: Table S1). Only low and medium stress years occurred over this time frame.
To investigate the effect of deltamethrin application on colony extent, we used the lme4 package (Bates et al. 2015) in R to implement linear mixed‐effects models with biannual percent change in areal extent of treated vs. untreated colony sections as the response variable. The models tested mirrored those run for BTPD abundance, except that environmental stress was coded as a binary variable reflecting whether the two years preceding biennial measurements of colony extent included a stressful (non‐wet) year or not. We used percent change in extent rather than absolute extent as the response variable because treated sections typically were larger than untreated sections. The response variable was log‐transformed as r = log (1 + R) prior to analysis to meet the residual normality and homoscedasticity requirements of linear models. Two outliers violating assumptions of residual normality remained after transforming the data, so the models were run both with and without the outliers; inclusion of the outliers did not alter model insights, so we report on the analysis using the full dataset below. We used AICc, model averaging (Burnham and Anderson 2002) and marginal R2 (Nakagawa et al. 2017) to determine the top model(s), important predictors, and model fit, respectively. The analyzed colony extent data spanned 2011–2015 across 19 colonies and included medium (average) and low stress (wet) years.
To examine temporal patterns of flea prevalence in untreated burrows and to facilitate comparison with flea prevalence estimates in other epidemiological studies, we used Blaker's method as implemented in Epitools (Sergeant 2009) to calculate 95% confidence intervals based on raw burrow swabbing data. Blaker's method is designed to account for imperfect detection when no prior information exists on true population prevalence (Reiczigel et al. 2010). Given lack of information on how effective burrow swabbing is at correctly detecting fleas under the environmental conditions present in our study area, we defaulted sensitivity and specificity to 1. Because sampling effort varied among months and years, we also analyzed bootstrapped data to validate temporal patterns of flea prevalence.
Because drought may affect flea numbers by increasing probability of desiccation and mortality, we checked Pearson's correlations between bootstrapped flea prevalence in spring in untreated burrows and water balance (precipitation minus potential evapotranspiration) in either spring (1 April–31 May of the current year) or the previous year's growing season (1 April–30 September of the preceding year). We used water balance rather than SPEI under the assumption that standardization to a historic climatic norm is less relevant for species with short generation times.
We also investigated year‐over‐year temporal autocorrelation for each of burrow flea prevalence (percentage of burrows with at least one flea), burrow flea abundance (mean number of fleas per burrow), and burrow flea intensity (mean number of fleas per burrow harboring fleas), computing the correlation coefficient for colony‐specific spring data from untreated sections in year t vs. t − 1 for each colony that had data for all years in 2013–2017. Colony‐specific sample sizes were too small to apply bootstrapping, but spring data reflected relatively consistent sampling effort.
We checked for year‐over‐year temporal autocorrelation in host flea prevalence (percentage of hosts with at least one flea), flea abundance (mean number of fleas per host), and flea intensity (mean number of fleas per host harboring fleas) using R (R Core Team 2018) to compute the correlation coefficient for colony‐specific data in year t vs. t − 1 from hosts captured for each untreated colony section that had data for all years in 2011–2016.
Because a Shapiro–Wilk normality test indicated non‐normality of the data, we used a Wilcoxon signed rank test to determine if bootstrapped flea prevalence differed between host species.
To investigate factors influencing annual variation of flea prevalence on hosts, we used the lme4 package in R to implement GLMM with a binomial error structure and the number of individual hosts per untreated plot with and without fleas as response. The models included a random intercept on plot ID nested in colony to account for repeated measures and non‐independence of plots within a colony. Fixed‐effect covariates included a binary variable identifying the host species, BTPD density and its interaction with host species, spring water balance, and environmental stress as an ordered factor. Our reasoning for the inclusion of each covariate is outlined in Appendix S1: Table S2. All possible additive combinations of these covariates were included in the model. Numeric predictors were mean‐centered and scaled by their standard deviations to ensure comparable units among predictors. Because alternative hypotheses for factors influencing flea prevalence on hosts are not mutually exclusive, we tested all possible combinations of the predictors in an information theoretic framework, with AICc and model‐averaged parameter estimates serving to identify top model(s) and important predictors (Burnham and Anderson 2002). Model averaging for each of the two global models was conducted using the MuMIn package (Barton 2018). We also computed marginal R2 values for the top model(s) to assess model fit (Nakagawa et al. 2017).
Following the same procedure, we also tested an alternate model in which we replaced spring water balance with the bootstrapped annual mean of spring flea prevalence in burrows. This alternate model covered a shorter time period (2013–2016 instead of 2011–2016) because no burrows were swabbed prior to 2013.
To gain insights into the temporal distribution of plague occurrence, we estimated the seasonal prevalence of Y. pestis in flea samples. As with flea prevalence, we calculated plague prevalence and 95% confidence intervals according to the Blaker method (Reiczigel et al. 2010), as implemented in Epitools (Sergeant 2009). To estimate the proportion of burrows potentially infected with Y. pestis, we multiplied flea prevalence (and CI limits) in burrows by plague prevalence (and CI limits) in flea samples.
In 2013–2017, a total of 3129 operator hours were spent to dust 163,553 burrows across 1818.6 ha of BTPD colonies. A total of 5660 and 809 burrow swabs were taken across untreated (i.e., no deltamethrin application) and treated sections of 15 prairie dog colonies, respectively. Following capture, a total of 1408 BTPD and 1228 RGS were processed; of these, 1243 BTPD and 1217 RGS provided data for analyses of flea prevalence on hosts, with 696 and 668 of these animals, respectively, captured on undusted plots (Appendix S1: Table S1).
The average annual bootstrapped percentage of burrows with fleas in treated sections was significantly lower than that in untreated areas (0.32% vs. 9.23%; t = −4.61, df = 2; P < 0.05; Fig. 2a). Similarly, the average annual bootstrapped prevalence of fleas on both BTPD and RGS was significantly lower in treated than untreated sections (BTPD: 0% vs. 2.80%, t = −4.40, df = 5, P < 0.01; RGS: 1.02% vs. 22.88%, t = −4.18, df = 5, P < 0.01; Fig. 2b). Model‐averaged parameters from our GLMMs identified environmental stress as the primary driver of BTPD abundance (Fig. 3a) and colony extent (Fig. 3b), exerting a negative effect on both. Additionally, a small, positive, additive effect of burrow dusting on the relative abundance of BTPD was detected when burrow dusting was measured cumulatively (Fig. 3a). Specifically, the BTPD relative abundance model set with binary burrow dusting yielded two top models (Appendix S1: Table S3), with one (Akaike weight = 0.65, marginal R2 = 0.13) containing only environmental stress, the other (Akaike weight = 0.26, marginal R2 = 0.13) environmental stress plus burrow dusting. The equivalent model set with cumulative burrow dusting yielded a single top model (Akaike weight = 0.75, marginal R2 = 0.24) that included drought and burrow dusting (Appendix S1: Table S3). Model‐averaged parameters indicated no clear‐cut effect of either binary or cumulative burrow dusting on colony extent (Fig. 3b; Appendix S1: Fig. S2b, c). The colony extent model set with binary burrow dusting yielded three top models (Akaike weights of 0.54, 0.24, and 0.22 and marginal R2 of 0.49, 0.50, and 0.47, respectively; Appendix S1: Table S3) containing environmental stress plus burrow dusting, environmental stress × burrow dusting, or only environmental stress. The equivalent model set with cumulative burrow dusting yielded two top models (Appendix S1: Table S3), with one (Akaike weight = 0.51, marginal R2 = 0.50) incorporating environmental stress × burrow dusting, the other (Akaike weight = 0.30, marginal R2 = 0.47) only environmental stress.
Enlarge this image.2 Fig.. Annual mean bootstrapped prevalence of fleas across dusted and undusted sections of black‐tailed prairie dog colonies in Grasslands National Park on (a) burrows sampled through swabbing, and (b) black‐tailed prairie dogs (black bars) and Richardson's ground squirrels (gray bars) sampled by combing. Error bars represent 95% confidence intervals. Years without data are specified as N/A.
Enlarge this image.3 Fig.. Model‐averaged parameter estimates (symbols) and 95% confidence intervals (horizontal lines) of fixed‐effect explanatory variables included in models of (a) black‐tailed prairie dog (BTPD) abundance, (b) BTPD colony extent, and (c) flea prevalence on BTPD and Richardson's ground squirrels (RGS). In each case, there were two alternative model sets. For (a) and (b), each set included environmental stress (Stress) and either a binary (Dusting‐01, triangles) or cumulative measure (Dusting‐CUM, circles) of dusting effort. For (c), the two model sets differed in using either spring water balance (spring WB, triangles) or mean flea prevalence in burrows (Burrow fleas, circles) as a proxy for potential flea exposure. The models also included prairie dog density (PD density) and a binary species indicator (Sp.), which distinguished between BTPD and RGS, with BTPD as reference level so that positive values (right of dashed vertical line) indicate that RGS have a higher flea prevalence relative to BTPD. The reference level for Dusting‐01 was no dusting, and environmental stress was ordered from low (wet) to high (dry) based on conditions in the previous one (a, c) or two (b) growing seasons. Variables that appeared in relevant top models (ΔAIC < 2) are indicated in black, those that did not in gray.
In untreated colony sections, the percentage of burrows with fleas varied significantly across months (Fisher's exact test, χ2 = 640.5174, P < 0.001), with highest prevalence in April and May (33.9% and 11.7%, respectively; Table 1). Bootstrapped data that controlled for variable sampling effort yielded patterns consistent with these results (Appendix S1: Fig. S3). Significant differences were also detected among years (Fisher's exact test, χ2 = 204.7106, P < 0.001), with spring prevalence of fleas in burrows fluctuating between a minimum of 7.5% and a maximum of 43.9% from 2013 to 2017 (Fig. 2a, Appendix S1: Table S4).
TableMonthly variation in the prevalence of fleas in burrows, prevalence of Yersinia pestis (YP) in flea samples, and estimated proportion of burrows harboring YP‐infected fleas in undusted sections of black‐tailed prairie dog colonies in Grasslands National Park in 2013–2017.| Month | Fleas in burrows | YP in fleas | YP in burrows | |||||
| Prevalence (%) | 95% CI | n | Prevalence (%) | 95% CI | n | Prevalence (%) | 95% CI | |
| April | 33.89 | 30.38–37.58 | 661 | 0.6 | 0.03–3.33 | 166 | 0.20 | 0.01–1.25 |
| May | 11.68 | 10.14–13.42 | 1481 | 0.99 | 0.27–3.54 | 202 | 0.12 | 0.03–0.48 |
| June | 3.19 | 2.39–4.24 | 1410 | 2.9 | 0.8–9.97 | 69 | 0.09 | 0.02–0.42 |
| July | 1.8 | 1.26–2.56 | 1666 | 0 | 0–17.59 | 18 | 0 | 0.00–0.45 |
| August | 1.75 | 1.04–2.91 | 802 | 0 | 0–13.8 | 24 | 0 | 0.00–0.40 |
| September | 3.09 | 1.81–5.21 | 421 | 0 | 0–20.39 | 15 | 0 | 0.00–1.06 |
| Total | 7.75 | 7.12–8.43 | 6441 | 0.98 | 0.42–2.27 | 511 | 0.08 | 0.03–0.19 |
We did not detect a significant correlation between the bootstrapped prevalence of fleas in spring and either spring water balance (r = −0.36, P = 0.55) or the previous year's growing season water balance (r = −0.15, P = 0.81) in 2013–2017. Nor did we find any evidence of significant annual temporal autocorrelation in flea prevalence, flea abundance, or flea intensity in burrows at the colony level.
Bootstrapped prevalence of fleas on hosts varied between species, with RGS significantly more likely to harbor fleas than BTPD (2011–2017 prevalence: RGS = 26.69%, BPTD = 3.27%; V = 21, P < 0.05; Fig. 2b). Our mixed‐effects models confirmed this difference and also suggested that greater BTPD density reduced flea prevalence on hosts, at least on prairie dogs, with model‐averaged parameters for host species and its interaction with BTPD density excluding zero from the 95% confidence intervals (Fig. 3c). Two top models emerged (Akaike weights of 0.48 and 0.27, and R2 = 0.16 in both). The first included all covariates except environmental stress; the second also omitted spring water balance (Appendix S1: Table S5).
Model‐averaged results for the alternate model that included flea prevalence in burrows instead of spring water balance yielded similar insights, with host species and its interaction with BTPD density being the only model‐averaged parameters not to include zero in their confidence intervals (Fig. 3c). Three top models emerged (Appendix S1: Table S5); all incorporated BTPD density, species, and their interaction. The second‐ranking one additionally incorporated environmental stress, suggesting higher flea prevalence on hosts following drier years. The third model instead incorporated flea prevalence in burrows. These models carried Akaike weights of 0.45, 0.25, and 0.17, respectively, and a marginal R2 = 0.19 each.
Flea prevalence, abundance, and intensity on hosts showed no autocorrelation year to year at the colony level.
A total of 587 flea samples were morphologically identified upon collection through burrow swabbing, yielding nine taxa. Oropsylla tuberculata tuberculata and Oropsylla rupestris represented 77.7% and 15.7% of the flea community, respectively (Table 2). Oropsylla t. tuberculata had the widest distribution, having been collected on 14/15 (93.3%) BTPD colonies, whereas O. rupestris and Hystrichopsylla dippiei dippiei were found on 12/15 (80%) and 7/15 (46.6%) colonies, respectively.
TablePercent composition of the flea community sampled through burrow swabbing on black‐tailed prairie dog colonies in Grasslands National Park (2013–2016) and combing (2013, 2014, 2015, and 2017) of black‐tailed prairie dog (BTPD) and Richardson's ground squirrel (RGS), and the number of flea samples testing positive for Yersinia pestis out of the total number of samples tested (n).| Year | Species | Burrows | BTPD | RGS | Plague positives/n |
| 2013 | Catallagia decipens | – | – | – | – |
| Epitedia wenmanni | – | – | – | – | |
| Hystrichopsylla d. dippiei | 5.45 | – | – | 0/6 | |
| Neopsylla inopina | – | – | – | – | |
| Oropsylla bruneri | – | – | – | – | |
| Oropsylla labis | 1.82 | – | – | 0/2 | |
| Oropsylla rupestris | 2.73 | – | 47.06 | 0/15 | |
| Oropsylla t. tuberculata | 90 | 100 | 52.94 | 4/110 | |
| Rhadinopsylla sp. | – | – | – | – | |
| Unknown | – | – | – | – | |
| n (total) | 110 | 1 | 17 | 4/133 | |
| 2014 | C. decipens | – | – | – | – |
| E. wenmanni | – | – | – | – | |
| H. d. dippiei | 4.27 | – | – | 0/5 | |
| N. inopina | – | – | – | – | |
| O. bruneri | 1.71 | – | 10 | 0/4 | |
| O. labis | – | – | – | – | |
| O. rupestris | 17.95 | 50 | 45 | 0/29 | |
| O. t. tuberculata | 73.5 | 50 | 45 | 0/94 | |
| Rhadinopsylla sp. | – | – | – | – | |
| Unknown | 2.56 | – | – | 0/3 | |
| n (total) | 117 | 2 | 20 | 0/135 | |
| 2015 | C. decipens | – | – | – | – |
| E. wenmanni | – | – | – | – | |
| H. d. dippiei | 1.25 | – | 1.19 | 0/2 | |
| N. inopina | – | – | – | – | |
| O. bruneri | – | – | – | – | |
| O. labis | – | – | – | – | |
| O. rupestris | 12.5 | 77.78 | 83.33 | 0/87 | |
| O. t. tuberculata | 55 | – | 10.71 | 0/53 | |
| Rhadinopsylla sp. | – | – | 1.19 | 0/1 | |
| Unknown | 31.25 | 22.22 | 3.57 | 0/1 | |
| n (total) | 80 | 9 | 84 | 0/144 | |
| 2016 | C. decipens | 0.72 | – | – | 0/1 |
| E. wenmanni | – | – | – | – | |
| H. d. dippiei | 2.16 | – | – | 0/3 | |
| N. inopina | 0.72 | – | – | 0/1 | |
| O. bruneri | – | – | – | – | |
| O. labis | – | – | – | – | |
| O. rupestris | 16.55 | – | – | 0/23 | |
| O. t. tuberculata | 79.14 | – | – | 0/110 | |
| Rhadinopsylla sp. | 0.72 | – | – | 0/1 | |
| Unknown | – | – | – | – | |
| n (total) | 139 | 0/139 | |||
| 2017 | C. decipens | – | – | – | 0/0 |
| E. wenmanni | 1.16 | – | – | 0/1 | |
| H. d. dippiei | 1.16 | – | – | 0/1 | |
| N. inopina | – | – | – | – | |
| O. bruneri | – | – | – | – | |
| O. labis | – | – | – | – | |
| O. rupestris | 13.95 | – | – | 0/12 | |
| O. t. tuberculata | 8.14 | – | – | 0/7 | |
| Rhadinopsylla sp. | 2.33 | – | – | 0/2 | |
| Unknown | 73.26 | 100 | 100 | 1/58 | |
| n (total) | 86 | 4 | 5 | 1/81 | |
| Total† | C. decipens | 0.22 | – | – | 0/1 |
| E. wenmanni | 0.22 | – | – | 0/1 | |
| H. d. dippiei | 3.71 | – | 0.84 | 0/20 | |
| N. inopina | 0.44 | – | – | 0/2 | |
| O. bruneri | 0.44 | – | 1.68 | 0/4 | |
| O. labis | 0.44 | – | – | 0/2 | |
| O. rupestris | 15.72 | 80 | 73.95 | 0/181 | |
| O. t. tuberculata | 77.73 | 20 | 22.69 | 4/385 | |
| Rhadinopsylla sp. | 1.09 | – | 0.84 | 0/6 | |
| Unknown | – | – | – | 1/62 | |
| n (total) | 458 | 10 | 119 | 5/664 |
Flea species most common in burrows or on hosts are indicated by figures in bold for each year with the exception of 2017, when the majority of fleas were rushed for Y. pestis testing (i.e., without morphological identification of flea species) due to a suspected plague outbreak. Total composition is adjusted by excluding fleas that were not morphologically identified.
†Includes a total of 32 fleas for which collection year is unknown.
Of 684 fleas found on hosts (34 taken from 24 BTPD, 650 from 188 RGS), a total of 143 samples were processed for identification. Of these, 16 samples taken from BTPD comprised 2 taxa, whereas 127 samples taken from RGS comprised 5 taxa (Table 2). The flea community composition on hosts contrasted significantly with that observed in burrows (Fisher's exact test, χ2 = 172.9066, P < 0.001), with O. rupestris constituting 80% and 74% of identified fleas on BTPD and RGS, respectively, whereas O. t. tuberculata made up only 20% and 22.7% (Table 2).
Of 664 flea samples tested, five (0.75%) were positive to Y. pestis. No differences in estimated prevalence of Y. pestis were found among months (Fisher's exact test, χ2 = 3.7590, P = 0.71), while a statistical difference was found among years (Fisher's exact test, χ2 = 12.1935, P < 0.05), with highest prevalence recorded in 2013 (3.7%). Given the prevalence of Y. pestis across flea samples and prevalence of fleas within burrows, the estimated percentage of burrows harboring plague‐infected fleas was 0.08% overall and highest between April and June (0.09–0.20%; Table 1). No infected fleas were detected during July–September.
Although no significant differences in Y. pestis prevalence were detected among the nine flea taxa sampled (Fisher's exact test, χ2 = 4.1188, P = 0.94), only O. t. tuberculata was confirmed to carry plague in our study area (Table 2).
Of 18 host carcasses (15 prairie dogs and 3 ground squirrels) collected on prairie dog colonies (i.e., excluding roadkills), all found in 2016–2017, necropsy was conclusive for nine. Of these, three (2 RGS, 1 BTPD; 33.3% of diagnostic specimens) were confirmed to have died of sylvatic plague. All three were collected from the same colony (i.e., BRH) within 30 d (20 June–14 July 2017).
In 2013–2017, plague activity was documented in GNP on two separate occasions. In May 2013, a severe decline in the BTPD population was noted at the DNH colony. Fleas were opportunistically sampled and found in 19/30 (63.3%) burrows, with 2/19 (10.5%) flea samples testing positive for Y. pestis by PCR. However, no BTPD carcasses were found on this colony. In July 2017, sylvatic plague was confirmed as the cause of mortality for one BTPD and two RGS found on the BRH colony, according to post‐mortem examination, bacterial culture, and PCR assay. However, no colony contraction or population crash was observed in the following months, and the colony remained active until the end of the season.
Paucity of data is a common challenge in managing rare or imperiled species (Cardoso et al. 2011). Through our analyses, we have demonstrated how even imperfect data can offer insights to inform species conservation and refine active management strategies. By integrating information on BTPD population dynamics, flea prevalence, flea diversity, and prevalence of Y. pestis, while accounting for uncertainties during statistical analyses, our results provided preliminary insights into plague transmission ecology in the threatened Canadian BTPD ecosystem, helped inform management strategies, and identified critical knowledge gaps to be prioritized by future research.
In documenting infection of fleas with Y. pestis, as well as plague‐induced mortality in both BTPD and RGS, we confirmed the presence of sylvatic plague in the Canadian BTPD ecosystem, as first reported by Antonation et al. (2014). With the exception of one suspected event (in May 2013), for which plague‐positive carcasses could not be confirmed despite the occurrence of plague‐positive fleas, we encountered no evidence of colony collapse or widespread mortality. The lack of noticeable die‐offs is compatible with the circulation and maintenance of sylvatic plague at an enzootic level, whereby the disease results in a low number of occurrences, affects only a small proportion of the host population, but contributes to chronic mortality rates (Biggins et al. 2010), as previously recorded in other sites across the species range (Biggins et al. 2010, Matchett et al. 2010). Based on our data, contributing factors likely include relatively low flea prevalence in burrows and low prevalence of Y. pestis infections in fleas, coupled with a short season of flea activity and relatively low BTPD density (rarely exceeding 20 individuals/ha). However, limitations associated with the collection of fleas through burrow swabbing in our and other studies (Salkeld and Stapp 2008, Eads, 2017), the imperfect detection of Y. pestis using standard diagnostic tests, and the occurrence of plague epizootics in populations characterized by low prairie dog densities (Collier and Spillett 1975) suggest that other unknown factors may contribute to limiting plague activity in GNP, at the northern range limit of both BTPD and Y. pestis distributions (Maher et al. 2010).
Oropsylla t. tuberculata and O. rupestris were the predominant flea species in our study area. These species are characterized by intermediate pathogen acquisition efficiency (i.e., percentage of fleas infected after feeding on an infected host) and intermediate to high vector efficiency (i.e., percentage of infected fleas that transmit Y. pestis; Eisen et al. 2009). Oropsylla spp. are deemed efficient vectors of Y. pestis through early‐phase transmission (Wilder et al. 2008a, b, Eisen et al. 2009), a mechanism that can explain and support rapidly spreading plague epizootics (Eisen et al. 2006). Although only O. t. tuberculata was confirmed as carrying Y. pestis in our study area, several of the other eight flea species observed have been confirmed as plague vectors elsewhere (e.g., O. rupestris, Oropsylla labis and Neopsylla inopina; Jellison 1945, Anderson and Williams 1997).
However, the prevalence of fleas in both burrows and on hosts in GNP was generally lower than observed elsewhere across the species range where plague epizootics have been recorded (Tripp et al. 2009, Friggens et al. 2010, Tripp et al. 2017). Moreover, the peak of flea activity in our study area appeared shorter (April–May) than observed at lower latitudes, where warmer climate provides favorable conditions February through November (Tripp et al. 2009), allowing flea prevalence on hosts to remain high for longer periods (e.g., June–September; Eads et al. 2013). Finally, it is worth noticing that the prevalence of Y. pestis in our study site was considerably lower (<1%) than that observed in some areas known for plague epizootics, such as central and southern U.S. locations, where up to 20–50% of fleas (or flea samples) tested positive for the bacterium (Cully et al. 1997, 2000, Stevenson et al. 2003, Salkeld and Stapp 2008). While such cases may be exceptions to a generally lower detection rate of Y. pestis, we estimated that in our study area less than 0.2% of the burrows in a prairie dog colony are likely to harbor Yersinia‐infected fleas between April and June, after which the estimated probability is close to zero.
While the GNP flea community hence includes competent vectors for Y. pestis, their limited abundance and seasonally restricted activity may curb the intensity of plague transmission. In particular, the low and declining flea prevalence in burrows from June onwards (i.e., 0–5%) may render flea‐mediated plague transmission in GNP less likely, or at least limit it from reaching the intensity and geographical extent recorded at more southern locations (Friggens et al. 2010, Tripp et al. 2017). Based on these data, it is reasonable to infer that, under current conditions, sylvatic plague operating primarily at enzootic levels is likely not by itself an immediate threat to the short‐term persistence of the BTPD population in GNP. It may, however, cause chronic mortality, reduce population growth, and hence limit BTPD resilience to other stressors (Eads and Biggins 2015). The positive effect of burrow dusting on BTPD abundance seems to support this hypothesis and provide further evidence for enzootic circulation of sylvatic plague (Biggins et al. 2010). In interaction with small population size and climate change, sylvatic plague therefore poses a sufficiently substantial long‐term threat to warrant vigilance and mitigation.
Our results confirmed environmental stress as a key constraint for the Canadian BTPD population, in which survival and reproduction are primarily driven by weather‐mediated effects on food availability and quality (Stephens et al. 2018). BTPD appear to be vulnerable even under average climatic conditions and thrive only in relatively wet years, relative to our climate reference period (1953–1982). Additionally, some top models for flea prevalence on BTPD and RSG hosts included environmental stress among explanatory variables. Drier conditions are likely to increase the intensity of flea infestation indirectly through an effect on food availability, which can lower host body conditions and consequently affect both immune system and behaviorally mediated ectoparasite defenses negatively (Eads et al. 2016). Higher intensity of flea infestation can further diminish BTPD body condition (Beldomenico and Begon 2010), as well as potentially elevate the risk of plague‐related mortality—thus creating positive feedback cycles (Biggins and Eads 2019)—even if this does not reach the magnitude of an outbreak. Given indications that burrow dusting positively affects BTPD abundance, our data support the value of reducing flea prevalence through insecticide application, particularly during drier periods. In contrast to other studies (Eads et al. 2016), however, our analyses did not signal a strong relationship between water balance during spring or the previous year's growing season on the prevalence of fleas on hosts or in burrows, possibly because the climate indices used in our study were too coarse‐grained to reflect the micro‐climate experienced by burrow‐dwelling fleas.
While burrow dusting is the recommended choice in case of plague detection (i.e., plague‐infected fleas or host carcasses) or a suspected epizootic (Tripp et al. 2016), its use as a preventative measure to control flea exposure must be carefully considered. Deltamethrin affects multiple taxa within the invertebrate community and reduced invertebrate abundance may have consequences for other imperiled, dependent species such as mountain plover (Dinsmore et al. 2005) and burrowing owl (Dechant et al. 2002). Furthermore, while the targeted flea species occur widely and associate with multiple host species (Lewis 2002), BTPD burrows may also harbor host‐specific and geographically restricted, vulnerable invertebrate taxa (e.g., Linognathoides cynomyis; Kim et al. 1986). Finally, fleas have been known to develop genetic resistance to deltamethrin following regular exposure (Eads et al. 2018). Based on our findings, prioritizing dusting to dry years and/or to colonies in which BTPD body conditions are poorer may be one cost‐effective option to mitigate flea parasitism while also reducing the risks of broader preventative use (Eads et al. 2016). Orally administered insecticides have also returned encouraging results in controlling flea populations on BTPD and other rodent hosts (Borchert et al. 2009, Poché et al. 2017, Eads et al. 2019), and although further testing is required to verify their safety, their gradual implementation may reduce impacts on non‐target invertebrates and lower costs of plague management (Borchert et al. 2009).
Dusting should be timed to precede the peak of flea prevalence (Wilder et al. 2008a, Tripp et al. 2016), which in GNP appears to be April and May. A spring peak in flea prevalence is consistent with previous studies conducted on RGS in Manitoba (Lindsay and Galloway 1997). In GNP, snow cover, soft ground following snowmelt, and the presence of sensitive species (e.g., federally listed migratory birds) hinder insecticide application in early spring, but the 40‐fold decrease in flea numbers observed in treated areas suggests that deltamethrin application in the fall is highly effective in this ecosystem.
Although we encourage their prudent use, our analyses also suggest that applications of deltamethrin or oral insecticides may need to reach beyond BTPD colony boundaries to be most effective, particularly in relation to the potential role played by RGS. Few studies have investigated plague ecology with respect to interactions between ground squirrels and prairie dogs and confirmed the potential for flea transmission between host species (Fitzgerald 1970, Anderson and Williams 1997, Cully and Williams 2001). Insights from our analyses, however, suggest that RGS may play an important role in the maintenance and circulation of plague in GNP, despite their overall low density. While the flea communities infesting BTPD and RGS were similar, flea prevalence on RGS was eight times higher than on BTPD. Moreover, along with their congenerics, Columbian ground squirrels (Urocitellus colombianus), RGS are considered the primary host for O. rupestris (Lewis 2002), the flea species most commonly infesting BTPD in our study. Importantly, flea‐harboring RGS—but never BTPD—were trapped in dusted colony sections in 2011–2013 (Fig. 2), suggesting that RGS movements are not restricted to BTPD colony boundaries and thus potentially facilitate transfer of infected fleas from other reservoir hosts (Anderson and Williams 1997). In GNP, these could include other small mammals, such as the grasshopper mouse (Stapp et al. 2009, Salkeld et al. 2010). Extending insecticide applications to include a buffer zone around colonies might curb disease transfer from other hosts and thus reduce plague exposure for BTPD. Buffers zones should be substantial, however, as male RGS can traverse more than 200 m over the early spring breeding season, with daily movements commonly exceeding 50 m (Michener 1983).
Greater flea prevalence on RGS than BTPD may be due to behavioral differences between the species. Although differences in relative size and pelage characteristics (i.e., length and smoothness) between the two species may make fleas easier to find on RGS than BTPD, BTPD are also more social and therefore more likely to engage in allogrooming (Hoogland 1995), a direct cause of flea mortality (Marshall 1981). Additionally, increased opportunities for self‐grooming with enhanced group vigilance among BTPD could explain the inverse relationship we observed between BTPD densities and flea prevalence on BTPD, although parasite overdispersion might also contribute to this pattern, with a large proportion of an overall small number of fleas concentrated on few hosts (Eads et al. 2013). Contrary to theoretical expectations (Krasnov et al. 2002), therefore, BTPD population growth might not result in increased disease or parasite transmission, at least as long as food resources do not limit individual body conditions and anti‐parasite defenses (Eads et al. 2016).
Moving forward, careful targeting of insecticide applications would be facilitated by greater understanding of the factors that locally influence flea prevalence. We found no evidence of flea prevalence in one year influencing flea prevalence in the following year. Additionally, the prevalence and species of fleas detected in burrows poorly matched the composition of flea species infesting hosts. As burrow swabbing typically samples only the first 2–3 m of the burrow system, it will miss flea species with different microhabitat requirements. Similarly, temporal fluctuations in flea numbers detected through swabbing might reflect changes in temperature and humidity in the upper portion of the burrows rather than actual vector abundance (Seery et al. 2003, Salkeld and Stapp 2008), with fleas perhaps retreating to greater depths during summer. In this sense, our findings stress the importance of including both burrow swabbing and live animal combing in plague surveillance plans and warrant prudence in interpreting seasonal fluctuations of flea prevalence recorded in our study area.
Additionally, surveillance might be helped by technological advances (e.g., pipe crawler ROVs or snake robots; Sanfilippo et al. 2016, Lattanzi and Miller 2017) that could facilitate deeper swabs and pair swabbing data with finer‐scale measurements of temperature and humidity.
Given the overall low prevalence of fleas in burrows and Y. pestis in fleas, sampling effort (i.e., burrow swabbing) should be increased to improve detection. Epidemiological calculators applied to field prevalence data to compute sampling requirements are helpful in this regard (Appendix S1: Table S6; Liccioli et al. 2015). We also acknowledge that the prevalence of Y. pestis we reported may be underestimated due to limitations associated with suboptimal storage of fleas and could further be confounded by the variable number of individual fleas in diagnostic samples. Furthermore, the specificity observed for an assay targeting the pla gene must take into account homogeneity to other species (Hänsch et al. 2015). In the absence of successful bacterial culture, it is recommended that alternative targets or methods confirm results utilizing a one target approach to avoid overestimation of prevalence.
The apparently low prevalence of plague in GNP presents an opportunity to continue evaluating the effectiveness of large‐scale distribution of bait‐delivered sylvatic plague vaccine (SPV) in protecting prairie dogs against plague (Rocke et al. 2017) while reducing the potential impact on non‐targeted invertebrate communities. In this regard, RGS and their movements might also be considered, and SPV possibly implemented in strategic combination with burrow dusting (Tripp et al. 2017). For example, central portions of BTPD colonies could be treated with SPV while dusting could be applied in their periphery (and in a surrounding buffer zone), where relative abundance of RGS appears higher based on our observations.
Although the current ecological conditions in GNP have not supported detectable plague epizootics, continued plague surveillance is pivotal to detecting any changes in the system, especially in consideration of a warming climate, which has the potential to shift or extend vector life cycles, alter flea community composition and host–parasite interactions (Eads and Hoogland 2017), and overall shift the geographic range of plague in accord with climatic patterns (Nakazawa et al. 2007).
Despite limitations associated with our datasets, our study yielded pertinent insights on plague ecology in the northernmost BTPD ecosystem. Our study thus illustrates the value of employing multiple, complementary approaches to analyze limited data, quantify uncertainty, understand complex ecological systems, and inform management strategies for species at risk. It also serves as an example of a proactive approach to understanding and mitigating the risk of an invasive disease that can have devastating effects on a rare and threatened ecosystem.
Funding for this work was provided by Parks Canada Agency and Western College of Veterinary Medicine Interprovincial Scholarship Fund. We thank all GNP and Calgary Zoo staff, students, volunteers involved with collection of field samples, and in particular Austin Baron, Lauren Beaulieu, Sarah Champagne, Justin Crowe, Heather Facette, Samantha Fischer, Laura Gardiner, Travis Houston, Nils Lokken, Jacqueline Menzies, Chris Reed, Nathan Young. We thank Jeff Lane and his laboratory (Colleen Crill, Jillian Kush, and numerous field technicians and volunteers) for aiding disease surveillance and data collection at the Walker colony. We are grateful to Kenneth Gage and David Eads, for invaluable discussions and comments offered throughout the data analysis and writing process. The Canadian Wildlife Health Cooperative and Prairie Diagnostic Services provided important diagnostic capacity for plague diagnosis in sylvatic rodents. The authors have no competing interests to declare.
© 2020. This work is published under http://creativecommons.org/licenses/by/3.0/ (the “License”). Notwithstanding the ProQuest Terms and Conditions, you may use this content in accordance with the terms of the License.
Details
; Stephens, Tara 2
; Wilson, Sian C 2 ; McPherson, Jana M 2
; Keating, Laura M 2 ; Antonation, Kym S 3 ; Bollinger, Trent K 4 ; Corbett, Cindi R 3 ; Gummer, David L 5 ; L. Robbin Lindsay 6 ; Galloway, Terry D 7 ; Shury, Todd K 8 ; Moehrenschlager, Axel 2 1 Grasslands National Park, Parks Canada Agency, Val Marie, Saskatchewan, Canada
2 Centre for Conservation Research, Calgary Zoological Society, Calgary, Alberta, Canada
3 Bioforensics Assay Development and Diagnostics, National Microbiology Laboratory, Public Health Agency of Canada, Winnipeg, Manitoba, Canada
4 Department of Veterinary Pathology, Canadian Wildlife Health Cooperative, Saskatoon, Saskatchewan, Canada
5 Natural Resource Management Branch, Parks Canada Agency, Calgary, Alberta, Canada
6 Zoonotic Diseases and Special Pathogens, National Microbiology Laboratory, Public Health Agency of Canada, Winnipeg, Manitoba, Canada
7 Department of Entomology, Faculty of Agricultural and Food Sciences, University of Manitoba, Winnipeg, Manitoba, Canada
8 Parks Canada Agency, Saskatoon, Saskatchewan, Canada