Over the course of the 3-year study, 11,771 prairie dogs were captured, marked, and released (Table S1) with a total of 22,059 captures recorded from all pairs. Excluding prairie dogs that were recaptured between years, 10,249 unique animals were recorded. Samples of hair, whiskers, and fleas (if present) were collected from 6744 unique animals, some more than once if recaptured in another year. Of the 5996 animals sampled with an identifiable age (excluding HEUT-1, 2, and 4), bait uptake rates were lower for juveniles (63%, 95% C.I. 61–65) versus adults over all years (77%, 95% C.I. 76–78; X
2 = 159.40, p < 0.001). Bait uptake was similar between vaccine and placebo plots for adults and juveniles in 2013 (Table 2) but lower on vaccine plots in 2014 (adults X
2 = 5.10, p = 0.02; juveniles X
2 = 6.04, p = 0.01). In 2015, juvenile bait uptake rates were lower on vaccine plots (X
2 = 21.28, p < 0.001) compared to placebo plots, but adult rates were similar.
Over the 3-year study, 45 prairie dog carcasses (9 BTPDs, 4 GPDs, 10 WTPDs, and 22 UPDs) were submitted for necropsy and testing to NWHC and 28 (14 BTPDs and 14 GPDs) to CPW. Yersinia pestis was detected in tissues from 22 carcasses at NWHC by culture and/or PCR and 10 at CPW by PCR only, providing evidence that cause of death was plague. Twenty of the plague-positive carcasses were found on vaccine plots; of those, four had consumed bait just 12–21 days prior, seven were found prior to or on the day of baiting, one was not tested, and the rest were negative for bait uptake. Only one of the 20 had been caught in years previous, and it was negative for bait uptake at that time. Other causes of prairie dog mortality were predation (5), trapping or handling mortality (16), and vehicle collision (2); cause was undetermined in 13 carcasses too decomposed for analysis. A total of 5206 and 4734 flea pools from prairie dogs were tested for Y. pestis DNA by PCR at NWHC and CPW, respectively, and it was found in 70 (1.3%) and 106 pools (2.2%). Yersinia pestis was detected in at least one prairie dog carcass or one flea pool from prairie dogs at 14 of the 29 study pairs in one or more years. In 9 cases, Y. pestis was detected on both members of the pair (BTCO-1, BTCO-2, BTCO-3, RBTX-1, GUCO-3, CBUT-1, CBUT-2, HEUT-1, HEUT-2). In 3 cases, Y. pestis was detected on the vaccine plot but not the placebo plot (ERAZ-1, GUCO-2, HEUT-3), and in 2 cases, Y. pestis was detected on the placebo plot but not the vaccine plot (GUCO-1, PRWY-1).
Relative Abundance and Apparent Survival of Prairie Dogs
Although similar between plots within a pair, trapping efforts varied considerably among the 26 study pairs included in our analysis of relative abundance (Table S1). On seven study pairs, plague was confirmed as the cause of death of one or more animals, and obvious declines (>50% decrease) were noted in prairie dog relative abundance on one or both of the paired plots: BTCO-1, BTCO-2, BTCO-3, ERAZ-1, GUCO-3, CBUT-2, and HEUT-3 (Table 3). The positive plots within these 7 pairs were classified as “plague confirmed,” starting from the first year it was detected. Shortly after baiting in 2013, plague was confirmed at one study pair, BTCO-2, and complete colony collapse (>90% decline in CPUE) occurred with few animals (<1/ha) captured on either plot by 2014. Complete colony collapse also occurred on the BTCO-3 and BTCO-1 placebo plots in 2015, although the vaccine plots remained occupied. At two pairs (ERAZ-1 in 2014 and HEUT-3 in 2015), plague was confirmed on the vaccine plots, along with >50% declines in CPUE. Although plague was not detected on corresponding placebo plots, >50% declines in CPUE were noted for both in 2014.
Five study pairs were classified as “plague suspect.” At these pairs, one or more Y. pestis positive flea pools were detected by PCR (Table 3), but no plague-positive carcasses were found. At the 14 remaining pairs, all carcasses and fleas tested negative for Y. pestis, and these were classified as “plague not detected.”
Relative abundance as measured by CPUE was variable among pairs, years, and species (Fig. 2). Our best model included plague status, year, treatment, species, and treatment by year interactions and was >2 AIC points away from the intercept-only (no covariate) and the second best model (Table S2). Results indicated that vaccine treatment had an overall positive effect (p = 0.012) on CPUE all 3 years (Table 4, Fig. 3) that was significantly higher (p < 0.001) in 2014 than the other years. The odds of capture were 1.10 (95% C.I. 1.02–1.19) times higher per trap day on vaccine-treated plots than placebo plots in 2013, 1.47 (95% C.I. 1.41–1.52) times higher per trap day in 2014 and 1.19 (95% C.I. 1.13–1.25) times higher per trap day in 2015 on pairs with the same plague status (confirmed, suspect, and not detected) and the same species (Fig. 3). Removing the two pairs that started baiting in 2012 (BTCO-3 and GUCO-1) from the analysis eliminated the significant difference in CPUE in 2013 (results not shown), indicating they were responsible for the observed effect in 2013, but their removal had no effect on results for 2014 and 2015. Both confirmed and suspect plague negatively affected CPUE (p < 0.001). On average, odds of a capture were 0.34 (95% C.I. 0.30–0.38) and 0.64 (95% C.I. 0.59–0.71) times lower per trap day on pairs with plague confirmed and suspect, respectively, than pairs without plague detection (Table 4, Fig. 3). The model also included a species effect, indicating that CPUE was lower for WTPDs and UPDs than BTPDs and GPDs. Pearson’s correlation coefficient was 0.83 for fitted compared to observed values.
Between year capture/recapture data included a total of 3464 animals captured in 2013, 3791 animals captured in 2014, and 3940 animals captured in 2015; 774 (22%) of the animals captured in 2013 were recaptured in 2014, and 381 (11%) were recaptured in 2015 (Table S3). Of the 3017 animals newly captured in 2014, 583 (19%) were recaptured in 2015. Two-hundred and twenty-four animals were removed from our survival analyses due to uncertain aging at first capture.
The best survival model according to DIC included the effect of trapping effort on detection probability (the probability of capturing an animal if it is present; Table S4). On pairs where plague was detected, annual odds of apparent survival were 1.76 (95% B.C.I. 1.28–2.43) times higher on vaccine plots than placebo plots for adults and 2.41 (95% B.C.I. 1.72–3.33) times higher for juveniles (Fig. 4, Table S5). On pairs where plague was not detected, odds of survival were similar for juveniles between vaccine and placebo plots (0.93, B.C.I. 95% 0.77–1.45) but lower on vaccine plots for adults (0.68, B.C.I. 95% 0.57–0.82). Sampling effort was negatively associated with detection probability (i.e., pairs with more trapping days had lower detection probabilities), but this was likely an artifact of longer trapping sessions on some pairs with plague die-offs. On average, the probability of detection was higher on placebo plots (0.54; B.C.I. 95% 0.52–0.55) compared to vaccine plots (0.50; B.C.I. 95% 0.49–0.51). The odds of detection on vaccine plots were 0.95 (B.C.I. 95% 0.91–0.98) compared to placebo plots.