Prairie dog presence affects occurrence patterns of disease vectors on small mammals

Ecography 000: 000000, 2008 doi: 10.1111/j.2008.0906-7590.05336.x # 2008 The Authors. Journal compilation # 2008 Ecography Subject Editor: Douglas Kelt. Accepted 7 May 2008 Prairie dog presence affects occurrence patterns of disease vectors on small mammals R. Jory Brinkerhoff, Chris Ray, Bala Thiagarajan, Sharon K. Collinge, Jack F. Cully, Jr, Brian Holmes and Kenneth L. Gage R. J. Brinkerhoff (robert.brinkerhoff@colorado.edu) and C. Ray, Dept of Ecology and Evolutionary Biology, Univ. of Colorado, 334 UCB, Boulder, CO 80309-0334, USA. B. Thiagarajan and J. F. Cully, Jr, Division of Biology, 204 Leasure Hall, Kansas State Univ., Manhattan, KS 66056, USA. S. K. Collinge, Dept of Ecology and Evolutionary Biology and Environmental Studies Program, Univ. of Colorado, 334 UCB, Boulder, CO 80309-0334, USA. B. Holmes, 220 East Market Street, Meeker, CO 81641, USA. K. L. Gage, Bacterial Zoonoses Branch, Division of Vector-Borne Infectious Diseases, Centers for Disease Control, Fort Collins, CO 80523, USA. Wildlife disease is recognized as a burgeoning threat to imperiled species and aspects of host and vector community ecology have been shown to have significant effects on disease dynamics. The black-tailed prairie dog is a species of conservation concern that is highly susceptible to plague, a flea-transmitted disease. Prairie dogs (Cynomys) alter the grassland communities in which they exist and have been shown to affect populations of small rodents, which are purported disease reservoirs. To explore potential ecological effects of black-tailed prairie dogs on plague dynamics, we quantified flea occurrence patterns on small mammals in the presence and absence of prairie dogs at 8 study areas across their geographic range. Small mammals sampled from prairie dog colonies showed significantly higher flea prevalence, flea abundance, and relative flea species richness than those sampled from off-colony sites. Successful plague transmission likely is dependent on high prevalence and abundance of fleas that can serve as competent vectors. Prairie dogs may therefore facilitate the maintenance of plague by increasing flea occurrence on potential plague reservoir species. Our data demonstrate the previously unreported ecological influence of prairie dogs on vector species assemblages, which could influence disease dynamics. Dynamics and persistence of vector-borne infectious diseases depend on the communities of hosts and vectors in which they occur (Keesing et al. 2006). Theoretical studies of host-parasite systems indicate that host and parasite population dynamics can determine whether or not a parasite will establish and what conditions are necessary for its persistence (Anderson and May 1978, Holt et al. 2003). Recent investigations have demonstrated that the composition of the host community for a given pathogen or vector can influence parasite establishment (Holt et al. 2003) and disease dynamics (Ostfeld and Keesing 2000, Schmidt and Ostfeld 2001). It is less appreciated, however, that certain members of the host community may directly or indirectly influence vector prevalence and abundance on other species (Collinge et al. 2008). The black-tailed prairie dog Cynomys ludovicianus is an excellent model for exploring effects of a disease host on vector occurrence. Cynomys ludovicianus is highly susceptible to infection by the bacterium Yersinia pestis (Cully and Williams 2001), a flea-transmitted pathogen that is thought to be maintained by partially resistant rodents and their fleas (Gage and Kosoy 2005). Prairie dog presence in grassland ecosystems can influence plant and animal community composition; prairie dogs alter soil properties (Carlson and White 1987), change nutrient cycling dynamics (Whicker and Detling 1988), and influence both plant (Whicker and Detling 1988, Winter et al. 2002) and animal community structure (Smith and Lomolino 2004, Collinge et al. 2008). In particular, prairie dogs have been shown to significantly alter mammal species assemblages (Agnew et al. 1986), though these effects may not be generalizable across study systems. In some cases, small mammal abundance is higher in the presence of prairie dogs, though species richness is lower (Agnew et al. 1986, Cully et al. unpubl.). Other researchers, however, have found contrasting patterns (Ceballos et al. 1999) or have failed to find any relationship between prairie dog presence and small mammal community composition (Stapp 2007). We evaluated the effect of prairie dog presence on disease vector occurrence by comparing vector assemblages on small mammals in the presence and absence of C. ludovicianus across a wide geographical range (Fig. 1). For the purposes of this study, we define small mammals Online Early (OE): 1-OE

Charles M. Russell National Wildlife Refuge, MT Badlands National Park, SD Thunder Basin National Grasslands, WY Wind Cave National Park, SD Boulder County, CO Comanche National Grasslands, CO Cimarron National Grasslands, KS Janos, Chihuahua, MX Figure 1. Map showing the approximate historical range of black-tailed prairie dogs (shaded region) and the locations of the 8 study areas. as non-prairie dog rodents that can be captured using 7.6 8.922.9 cm aluminum live traps. The likelihood of interand intraspecific Y. pestis transmission among mammals is dependent on the prevalence and abundance of fleas (Lorange et al. 2005) and may also be influenced by the flea species assemblage, as flea species vary in their competence as vectors (Perry and Fetherston 1997, Gage and Kosoy 2005, Krasnov et al. 2006a). When highly susceptible hosts, such as black-tailed prairie dogs, become infected, an epizootic event may be triggered (Perry and Fetherston 1997). Prairie dog burrows provide habitat and refuge for a wide variety of vertebrate and invertebrate species, including arthropods (Bangert and Slobodchikoff 2006, Davidson and Lightfoot 2007), herptiles (Shipley and Reading 2006), birds (Smith and Lomolino 2004), and mammals (Agnew et al. 1986, Shipley and Reading 2006, Collinge et al. 2008). Mammal burrows may serve to increase heterogeneity of microclimates, leading to higher overall arthropod species richness and diversity (Davidson and Lightfoot 2007). Fleas in particular may benefit from the stable microclimate within a burrow as flea development and survival are highly dependent on temperature and relative humidity (Rust and Dryden 1997, Krasnov et al. 2001). If the abundance and richness of fleas were higher on prairie dog colonies, then small mammals on colonies would be expected to carry more fleas and more flea species than identical mammal species sampled at off-colony grassland sites. Under this scenario, the transmission of flea-borne diseases would be facilitated on prairie dog colonies for at least 2 reasons. First, higher flea prevalence leads to more frequent host-switching (Bossard 2006), increasing the potential for interspecific pathogen transmission. Second, associations among flea species tend to be facilitative rather than competitive (Brinkerhoff et al. 2006, Krasnov et al. 2006b), so an increase in flea species abundance and richness should increase per-capita rates of flea-mediated pathogen transmission as hosts acquire higher and more speciose flea loads. The specific role of fleas in the transmission of plague among prairie dogs is unclear. The dominant paradigm for flea-mediated plague transmission requires that a biofilm blockage containing Y. pestis is formed in a flea s proventriculus, causing regurgitation of subsequent blood meals, thereby increasing the likelihood of Y. pestis transmission (Gage and Kosoy 2005). However, Webb et al. (2006) demonstrated with an epidemiological model that plague epizootics in prairie dogs are not driven by blocked fleas. Unblocked fleas are capable of spreading plague and earlyphase transmission by unblocked fleas may explain the epizootic patterns that are observed in nature (Eisen et al. 2006, Wilder et al. 2008). Transmission models for blocked and unblocked fleas indicate that higher flea prevalence and abundance (Lorange et al. 2005, Eisen et al. 2006) increase the likelihood of Y. pestis spread, yet the effects of prairie dogs on flea occurrence are largely unstudied. We explored the effects of prairie dogs on small mammal flea assemblages by recording flea occurrence at on- and off-prairie dog colony sites spanning the range of C. ludovicianus in the western Great Plains of North America. We examined patterns of flea prevalence, abundance, intensity, and relative species richness between site types (on- versus off-colony) at 8 study areas across the range of C. ludovicianus (Fig. 1). We also investigated the similarity of flea communities among the 8 study areas and modeled the similarity of flea assemblages as a function of geographic distance. 2-OE

Methods Flea collection We collected fleas from live-caught small mammals at 8 study areas: 1) Badlands National Park, South Dakota, 2) Wind Cave National Park, South Dakota, 3) Thunder Basin National Grasslands, Wyoming, 4) Comanche National Grasslands, Colorado, 5) Cimarron National Grasslands, Kansas, 6) Janos, Chihuahua, Mexico, 7) Boulder County, Colorado, and 8) Charles M. Russell National Wildlife Refuge, Montana (Fig. 1). Small mammal trapping at all areas occurred between May and September 2003. At each study area, trapping grids were set on active prairie dog colonies (on-colony sites) and within similar grassland sites located 5002000 m from prairie dog colonies (off-colony sites). Similar numbers of on- and off-colony sites were sampled, though replication varied among study areas; we sampled 4 pairs of grids at study areas 1 and 2, 6 pairs at areas 36, 20 pairs at area 7 and 25 on- and 30 off-colony grids at area 8. We used square trapping grids at study areas 17 consisting of 49 Sherman live-traps (7.6 8.9 22.9 cm; H. B. Sherman Traps, Tallahassee, FL) spaced at 20 m intervals. At area 8, a square trapping grid consisting of 100 traps was used with traps placed at 10 m intervals. All traps were set in the evening and checked between 06:00 and 10:00 the following morning. Grids at study areas 16 were trapped for 3 successive nights and grids at areas 7 and 8 were trapped for 4 successive nights. To facilitate handling and collection of ectoparasites, we anesthetized all captured animals using vaporized Isoflurane (Halocarbon Products Corporation, River Edge, NJ). Anesthesia was administered at areas 16 using SurgiVet vaporizer (SurgiVet, Waukesha, WI). At areas 7 and 8, the anesthetic was dispersed from wetted cottonballs placed inside a transparent anesthetizing chamber. We collected fleas both from the host s body by brushing through the pelage with a toothbrush (all areas) and from the anesthetizing chamber (areas 7 and 8 only) using forceps. Fleas from each captured host were stored in a 2% saline solution containing a small amount of the surfactant Tween (polysorbate) 80. Fleas were identified to species using keys presented in Hubbard (1947). In addition to ectoparasite data, we recorded host species identification, sex, weight, and length measurements. We marked individuals with either uniquely numbered aluminum eartags (study areas 16) or with batch marks by shaving a patch of fur from the hindquarters (areas 7 and 8). Analytical methods We calculated values for flea prevalence (proportion of hosts infested), flea abundance (number of fleas divided by number of hosts), and flea intensity (number of fleas divided by number of infested hosts), as well as numbers of flea and mammal species sampled at each study area by averaging across trapping grids. To test for effects of prairie dog presence and study area on flea occurrence, we used mixed-model ANOVAs, each of which consisted of 3 independent variables: prairie dog presence or absence (fixed, 2 levels), study area (random, 8 levels), and an interaction term. We used separate tests for each dependent variable: flea prevalence, flea abundance, flea intensity, and the ratio of flea-to-mammal species richness at each grid. The denominator degrees of freedom for each F-test in the mixed-model ANOVAs were calculated using Satterthwaite s method (Satterthwaite 1946) in JMP 4.04 (SAS Inst., Cary, NC). Because latitude has been shown to influence parasite occurrence (Lindenfours et al. 2007), we used linear regression to test for influence of latitude on flea and mammal species richness across our study areas. In order to determine if flea species richness is a function of mammal trapping effort, we regressed trapping effort (number of trap-nights) against flea species richness at each study area. To determine effects of colony presence that may be independent of host species richness, we calculated the ratio of flea species to small mammal species from each of our trapping grids and compared these values both among study areas and between on- and off-colony sites. To control for differential effects of small rodent species on flea occurrence, we tested for differences in flea prevalence, abundance, and relative flea species richness on the most commonly caught mammal species, Peromyscus maniculatus, across the 8 study areas, and between on- and off-colony sites. Peromyscus maniculatus did not always occur at both on- and off-colony grids at each study area, so mixed-model ANOVA was not applied. Instead, we used 2-sample t-tests to compare flea occurrence (prevalence, abundance, and intensity) at on- and off-colony sites and 1-way ANOVA to compare flea occurrence on P. maniculatus across study areas. We used t-tests to account for variation among study areas dues to differences in sampling effort. To compare flea species assemblages among our study sites, we used Sorenson s similarity index. We used a randomized Monte-Carlo Mantel test to relate flea assemblage dissimilarity (1 S) and geographical distance. We also used non-metric multidimensional scaling (NMS) to graphically assess the similarity of flea assemblages among study areas. The ordination was generated using the Sorenson (Bray-Curtis) distance metric, beginning with a random starting configuration of study areas and minimizing final stress (lack of fit) after 50 runs with the real data set. Dimensionality was determined by comparison of the ordination generated from real data to a randomized data set generated by Monte Carlo simulation. Results We collected 2444 fleas representing 19 species from 2373 small mammals across the 8 study areas (Table 1). In addition to capturing rodents, we also trapped juvenile desert cottontails at study areas 7 and 8 (Supplementary material, Table S1). Although traps were set overnight, we trapped a number of diurnal rodent species, including 2 ground squirrel species and 1 chipmunk species (Supplementary material, Table S1). Absolute flea species richness was highest at the CM Russell study area (12 species) and lowest at the Janos study area (4 species). The number of flea species collected from on- and off-prairie dog colony grids was generally equivalent at each study area, though the flea species composition varied depending on prairie dog 3-OE

Table 1. List of flea species collected from each study area. Asterisks indicate fleas that have been found naturally infected with Y. pestis (Pollitzer and Meyer 1961). Flea species Badlands (1) Cimarron (2) Comanche (3) Janos (4) Thunder Basin (5) Wind Cave (6) Boulder (7) Russell (8) On Off On Off On Off On Off On Off On Off On Off On Off Aetheca wagneri* 54 11 2 3 158 125 133 36 353 24 395 270 Amaradix euphorbi 1 Callistopsyllus deuterus 2 Callistopsyllus terinus 2 3 Cediopsylla inaequalis 6 17 Corrodopsylla curvata 1 Echidnophaga gallinacea 14 6 Epitedia wenmanni 1 0 2 1 1 Eumolpianus eumolpi* 1 Foxella ignota* 4 9 Malareus telchinus* 6 3 27 12 22 0 13 21 Meringis arachis 5 4 Meringis parkeri 3 Orchopeas leucopus* 2 4 15 17 2 5 10 1 1 8 5 6 39 O. sexdentatus* 0 1 1 2 10 Oropsylla hirsuta* 2 17 1 31 1 4 Peromyscopsylla hesperomys* 26 5 1 1 24 6 3 1 44 4 Pleochaetis exilis* 1 160 52 70 29 8 8 1 Thrassis fotus* 4 10 5 7 7 15 15 1 Total 86 24 202 77 113 45 27 28 207 156 164 60 380 32 466 377 Study area total 110 279 158 55 363 224 412 843 presence or absence (Table 1). Although variable among our study areas, trapping effort was not a significant predictor of number of flea species sampled among our study areas (linear regression, r 2 0.34, p0.1). The variation in flea collection techniques at study areas 7 and 8 did not result in abnormally high flea prevalence or abundance (prevalence: t 0.66, p0.51; abundance: t 0.89, p 0.38). Flea prevalence and abundance, as well as the ratio of flea species to small mammal species, differed significantly among study areas (prevalence: F 7, 7 5.78, p0.017; abundance: F 7, 7 4.61, p0.031; flea-to-mammal species ratio; F 7, 7 5.94, p0.016) and between on- and offcolony trapping locations (prevalence: F 1, 13 9.64, p 0.008; abundance: F 1, 13 13.16, p0.003; flea-to-mammal species ratio; F 1, 13 7.92, p0.015. Mean flea intensity did not vary by site (F 7, 7 1.88, p0.21) or as a function of prairie dog presence (F 1, 13 1.05, p0.33). Flea prevalence and abundance, as well as the ratio of flea species to small mammal species, were all significantly higher in the presence of prairie dogs (Fig. 2). The interaction between study area and prairie dog presence did not significantly affect any dependent variable (prevalence: F 7, 138 1.06, p0.39; abundance: F 7, 138 1.03, p0.41; intensity: F 7, 138 2.0, p0.06; flea-to-mammal species ratio; F 7, 138 1.17, p0.33). Latitude was not a significant predictor of mammal species richness among our study areas (adjusted r 2 0.0, p0.6) but was a significant predictor of flea species richness (adjusted r 2 0.44, p 0.045). The flea-to-mammal species richness ratio, however, was not significantly influenced by latitude (adjusted r 2 0.37, p0.07). Peromyscus maniculatus occurred at all of our study areas and at 136 of 159 trapping grids (65 on-colony grids and 71 off-colony grids). The relative abundance of P. maniculatus ranged from 0 to 95% among study areas (Table 3). Averaged across study areas, flea prevalence, abundance, and intensity were significantly higher on P. maniculatus sampled at prairie dog colony sites (prevalence: t 2.26, p 0.026; abundance: t2.53, p 0.013; intensity: t2.13, p0.035); the number of flea species collected from P. maniculatus, however, did not vary depending on prairie dog presence (number of flea species; t 0.64, p0.52). Study area significantly was a significant predictor of flea prevalence (F 7, 128 3.64, p0.001), abundance (F 7, 128 3.19, p0.004), intensity (F 7, 128 3.32, p0.002), as well as the number of flea species sampled (F 7, 128 6.16, pb0.001) from P. maniculatus. Sorenson s similarity indices were lowest (0.11) between the CM Russell (Montana) and Cimarron (Kansas) study areas and were highest (0.92) between the Comanche (Colorado) and Badlands (South Dakota) study areas. A randomized Monte Carlo Mantel test indicated a significant positive relationship between flea assemblage dissimilarity and geographical distance (r 0.67, p0.005; Fig. 3). Non-metric multidimensional scaling (NMS) resulted in a 2-dimensional representation of flea assemblages by study area (Fig. 4). Ordination stress of the real data set was reduced substantially after 16 iterations and reached its minimum value of 2.89 after 39 iterations; ordinations with stress values B5 are considered highly robust representations of the dataset (McCune and Grace 2002). R-squared 4-OE

Figure 2. Differences in flea prevalence (a), flea abundance (b), flea intensity (c), and flea-to-mammal species ratio (d) between on-colony and off-colony sites averaged across all trapping grids at all 8 study areas. Bars represent mean values; error bars represent standard error. Statistical significance as described in the text is generated from mixed-model ANOVA analysis. values between Sorenson s distance and axes 1 and 2 were 0.179 and 0.754, respectively, and explain a total of 93% of the variance between distance in the ordination space and distance in the original data set. Discussion The presence of prairie dogs is significantly associated with differences in metrics of flea occurrence on small mammals; fleas were more prevalent and present in higher abundance on small mammals associated with black-tailed prairie dog colonies. Furthermore, relative flea species richness, measured as the ratio of flea to mammal species at a sampling grid, was significantly higher in the presence of prairie dogs. Given that plague, the disease caused by the bacterium Yersinia pestis, is reliant on high levels of flea prevalence and abundance for successful transmission (Lorange et al. 2005), the presence of prairie dogs in grassland ecosystems may affect disease dynamics in ways that increase the probability of epizootic events. Dozens of flea species have been found to be naturally infected with Y. pestis (Pollitzer and Meyer 1961, Gage and Kosoy 2005), including several which are present in our study system (Table 1), and increased flea abundance may be associated with higher rates of plague transmission (Krasnov et al. 2006a). Thus, the ecological effect of prairie dog colonies on small mammal fleas may ultimately lead to higher probability and rate of plague transmission to and among prairie dogs. Potential causes of increased flea occurrence on small mammals The mechanisms by which prairie dog presence increases flea prevalence, abundance, and relative species richness are unclear, but may be related to the role of prairie dogs as ecological engineers. Prairie dog burrows provide habitat and refuge for other vertebrate species and the presence of prairie dogs increases arthropod species richness in grassland ecosystems (Davidson and Lightfoot 2007). Flea development and survival is highly sensitive to ambient conditions such as temperature and relative humidity (Rust and Dryden 1997, Krasnov et al. 2001) and the burrow network created by prairie dogs may mediate microclimatic conditions in ways that are favorable to fleas (Krasnov et al. 1997). Presence of prairie dogs may increase (Shipley and Reading 2006) or decrease (Agnew et al. 1986) local small mammal species diversity, although small mammal abundance in this study system is higher in the presence of prairie dogs (Collinge et al. 2008). In this study system, the higher overall abundance of rodents, including prairie dogs, at prairie dog colony sites could result in higher flea prevalence and abundance; increased overall host availability could lead to higher flea fitness and therefore greater flea populations. However, this hypothesis contradicts previous studies that show a negative or no relationship between small mammal abundance and flea prevalence and abundance (Stanko et al. 2002, 2006). Such variation in results could be due to differences in focal mammal communities; social animals may be more likely to share ectoparasites than non-social animals (Brown and Brown 1996), potentially resulting in higher parasite prevalence and abundance. The propensity of fleas to switch hosts is dependent on both flea and host species identities and fleas are more likely to switch among ecologically and taxonomically similar hosts (Traub 1985). Thus, the species composition of the local mammalian host community is likely to influence the rate at which fleas parasitize non-characteristic hosts, potentially altering flea prevalence and abundance. In this study system, small mammal species richness generally is lower at on-colony sites than at off-colony sites (Collinge et al. 2008). Relative to small rodent species richness, we have shown that flea species richness is significantly higher in the presence of prairie dogs (Fig. 2). One potential explanation for this phenomenon is that small rodents at prairie dog colony sites acquire prairie 5-OE

6-OE Table 2. Numbers of rodents and fleas sampled at each study area. Flea occurrence metrics (prevalence, abundance, and intensity) are averaged across all on- or off-colony grids at each study area. Badlands (1) Cimarron (2) Comanche (3) Janos (4) Thunder Basin (5) Wind Cave (6) Boulder (7) Russell (8) On Off On Off On Off On Off On Off On Off On Off On Off Number of rodents sampled 103 40 86 91 57 62 22 68 150 155 140 37 525 191 248 398 Number of rodent species sampled 5 5 9 9 7 6 4 7 4 5 1 4 4 7 4 8 Number of flea species sampled 6 4 4 6 6 5 3 4 6 7 4 6 4 4 6 12 Number of fleas collected 86 24 202 77 113 45 27 28 207 156 164 60 380 32 466 377 Mean flea prevalence 0.5 0.2 0.58 0.5 0.61 0.4 0.39 0.1 0.57 0.4 0.37 0.6 0.35 0.14 0.64 0.4 Mean flea abundance 0.84 0.5 1.88 0.7 1.84 0.8 1.36 0.4 1.31 0.9 1.17 1.5 0.73 0.17 1.77 1 Mean flea intensity 1.64 1.6 3.26 1.6 2.82 1.9 1.81 4.2 2.38 2 3.19 2.3 2.1 1.12 2.5 2 Table 3. Average flea prevalence, abundance, and intensity on Peromyscus maniculatus at each study area. Badlands (1) Cimarron (2) Comanche (3) Janos (4) Thunder Basin (5) Wind Cave (6) Boulder (7) Russell (8) On Off On Off On Off On Off On Off On Off On Off On Off Number of P. maniculatus sampled 78 22 19 27 1 33 0 2 48 55 51 22 497 117 220 342 Relative abundance of P. maniculatus 0.76 0.6 0.22 0.30 0.02 0.53 0 0.03 0.32 0.35 0.36 0.59 0.95 0.61 0.89 0.86 Mean flea prevalence 0.53 0.18 0.17 0.39 1.0 0.30 0 0 0.53 0.36 0.39 0.75 0.32 0.19 0.60 0.41 Mean flea abundance 1.02 0.38 0.36 0.74 1.0 0.64 0 0 1.32 0.94 1.26 1.54 0.68 0.22 1.54 0.89 Mean flea intensity 1.89 1.0 32.0 2.0 1.0 1.66 0 0 2.5 2.5 3.55 2.04 1.7 1.18 2.15 1.89

Figure 3. Relationship between linear distance and flea community dissimilarity among the 8 study areas. Dissimilarity was calculated as 1-S, where S is the Sorensen s similarity index value for each pair of study areas. dog-specific fleas, which they would otherwise not encounter. Our data indicate that this is a plausible explanation; small rodents sampled from on-colony grids at 4 of our study areas harbored a total of 54 Oropsylla hirsuta, the most common warm-season flea of C. ludovicianus (Table 1). Oropsylla hirsuta is rarely collected from nonprairie dog mammal species (Brinkerhoff unpubl.) and prairie dogs rarely acquire fleas typical of other mammal species (Brinkerhoff et al. 2006). However, it is possible that the relatively high mammal biomass associated with prairie dog colonies makes these unusual host-switching events slightly more common. Because host sampling effort may confound measures of parasite species richness (Guegan and Kennedy 1996), Axis 2 1.0 0-1.0-2.0 Boulder Russell Thunder Basin Wind Cave Badlands Comanche Janos 1.0 0-1.0 Axis 1 Cimarron Figure 4. NMS ordination of pairwise similarity values generated from flea species presence/absence data for each of the 8 study areas. Study areas that are closer geographically have more similar small mammal flea assemblages. Axes represent NMS scores for each study area. within-species measures of parasite richness and occurrence may be more reliable than measures summed across multiple host species (Stanko et al. 2002). Given that sampling effort was not a predictor of flea species richness, we are confident that our samples accurately represent the flea species assemblages at each study area. However, mammal species vary in their suitability as hosts for fleas (Krasnov et al. 2004), so it is likely that variation in mammal communities among our sites influences flea occurrence. Analysis of the most frequently encountered host in our study system, P. maniculatus, demonstrated that flea prevalence and abundance on this species were significantly higher on prairie dog colonies than at offcolony grassland sites even though the relative abundance of P. maniculatus was highly variable across study areas (Table 3). The number of flea species collected from P. maniculatus, however, did not differ between on- and off-colony sites. Although both prairie dog presence and study areas were significantly associated with differences in flea occurrence on P. maniculatus, the probabilities associated with these factors were lower for the latter than the former; additionally, flea prevalence at 2 study areas (Cimarron and Wind Cave) was greater on off-colony P. maniculatus, suggesting that flea occurrence on this species is more strongly influenced by locality than by the presence of prairie dogs. Consequences of high flea prevalence and abundance Parasite intensity is positively related to consequences of infestation. Although fleas tend not to cause direct mortality of their hosts (Traub 1985), they are vectors of a variety of pathogens (Shaw et al. 2004). High flea prevalence leads to higher flea species exchange among hosts (Bossard 2006) and a higher rate of flea species exchange could increase the probability of inter-species pathogen transmission. Because Y. pestis is a highly virulent pathogen and vector competence often is very low, transmission is thought to require high flea prevalence and abundance, even when many susceptible hosts are available (Lorange et al. 2005). Krasnov et al. (2006a) determined that there is a positive relationship between the abundance of a flea species on its host and its efficacy as a plague vector. Thus, the rapid spread of plague among prairie dogs could be at least partly due to the fact that flea prevalence and abundance on small mammals at on-colony sites is inflated relative to off-colony grassland sites. Most North American flea species have traditionally been considered to be poor vectors of plague because they rarely form proventricular blockages (Pollitzer and Meyer 1961, Gage and Kosoy 2005). However, a number of flea species, including Oropsylla hirsuta, have recently been found to transmit Y. pestis without forming blockages (Eisen et al. 2006, Wilder et al. 2008). Thus, traditional assumptions about which fleas are effective plague vectors may preclude complete understanding of plague transmission dynamics. It is likely that a number of small rodent flea species may be competent plague vectors, spreading the bacterium by way of early phase transmission (Eisen et al. 2006). If this is the case, increased flea prevalence and abundance, irrespective of flea species identities, may lead to 7-OE

higher rates of plague transmission within and among mammal species. The impact of plague on most wild mammals in North America is unknown (Gage and Kosoy 2005), but it has dramatic negative effects on prairie dogs. Plague spreads quickly among prairie dog colonies, results in nearly 100% mortality (Biggins and Kosoy 2001, Cully and Williams 2001), and reduces prairie dog genetic variation (Trudeau et al. 2004). Cynomys ludovicianus recently was removed from the candidate list of endangered species (Anon. 2004) but it is still a species of conservation concern given that the historical range occupied by this species has been diminished by 98% in the last 100 yr (Miller and Cully 2001). Regional variation in flea occurrence Flea prevalence, abundance, and relative flea species richness on small mammal hosts varied significantly among our study sites indicating that factors other than presence of prairie dogs are important determinants of flea occurrence. The mammal communities among the 8 study areas are highly variable (Collinge et al. 2008; Supplementary material, Table S1) and this variation is likely to account for some of the spatial heterogeneity in flea assemblages given that fleas tend to be strongly host-specific. However, geographical distance also significantly influences flea species assemblages in western North America (Fig. 3). The fact that the NMS ordination closely mirrors the geographical dispersion of the study area locations (Fig. 1, 4) suggests that factors associated with geography may be determinants of flea species assemblages; flea species assemblages are known to vary spatially due to factors such as environmental conditions (Rust and Dryden 1997, Krasnov et al. 1997) and host geographic range (Krasnov et al. 2005). Latitude was a significant predictor of flea, but not mammal, species richness in this study system. This result contrasts with the finding that primates at lower latitudes showed higher diversity of protozoan parasites than primates at higher latitudes (Nunn et al. 2005) but is consistent with the idea that flea assemblages vary spatially depending on host community (Stanko et al. 2002), local environmental conditions (Krasnov et al. 1997), or other factors (Krasnov et al. 2005). It is possible that the variation in flea collection techniques led to significant differences in flea occurrence metrics, though substantial variation in flea prevalence, abundance, and intensity was found among study areas where methods were identical (Table 2). Given the sensitivity of fleas to temperature and humidity, it is reasonable to expect that temporal variation in sampling could influence flea occurrence. Indeed, significant variation among study areas in flea occurrence could stem from local climatological processes rather than inherent differences associated with each study area. However, with one exception, each study area was sampled for a minimum of 7 weeks, which should have accounted for day-to-day fluctuations in temperature and humidity. Study area 4 (Janos) was sampled for only 1 week and, as a result, the sample size from this study area is relatively small (Table 2). However, the flea occurrence metric values from this site are within the range of values from all sites. We have demonstrated heretofore unrecognized associations between prairie dog presence and occurrence patterns of small mammal fleas. The mechanistic link between prairie dog presence and flea occurrence is unclear, though we suggest that higher flea prevalence and abundance on small mammals at prairie dog colony sites could affect transmission dynamics of Y. pestis and other flea-borne pathogens. Higher flea prevalence and abundance among small mammals on prairie dog colonies may lead to increased rates of host-switching by fleas (Bossard 2006) and higher probability of plague transmission (Krasnov et al. 2006a). In some systems, prairie dogs decrease small mammal species diversity but increase overall small mammal abundance (Agnew et al. 1986, Ray and Collinge 2006, Collinge et al. 2008). Given that the relative abundance of particular pathogen hosts or reservoirs can influence the force of pathogen transmission (Ostfeld and Keesing 2000, Schmidt and Ostfeld 2001), the cumulative effects of prairie dogs on small mammals and their fleas serves to alter the dynamics of plague transmission and lead to epizootic events. Acknowledgements We thank the land management agencies that granted us permission to perform the field work associated with this study. These include the National Park Service, Boulder City Open Space and Mountain Parks, Boulder County Parks and Open Space, Jefferson County Open Space, and the Colorado Division of Wildlife. In particular, we thank B. Kenner, D. Roddy, D. Foster, D. Augustine, A. Chappell, T. Byer, C. Lockerman, M. Brennan, B. Pritchett, and C. Richardson for their logistical support. We would also like to thank our many field assistants for their hard work, including W. Davis, C. Pope, T. Johnson, B. Erie, M. Campbell, D. Stetson, D. Conlin, and A. Markeson. This research was supported by grants from the National Center for Environmental Research STAR program of the US-EPA (R-82909101-0), the NSF/NIH joint program in Ecology of Infectious Diseases (DEB-0224328), and the USGS Species at Risk Program. Comments made by 3 anonymous reviewers greatly improved this manuscript. References Agnew, W. et al. 1986. Flora and fauna associated with prairie dog colonies and adjacent ungrazed mixed-grass prairie in western South Dakota. J. Range Manage. 39: 135139. Anderson, R. M. and May, R. M. 1978. Regulation and stability of hostparasite population: I. Regulatory processes. J. Anim. Ecol. 47: 219247. Anonymous 2004. Finding for the resubmitted petition to list the black-tailed prairie dog as threatened. Federal Register 69: 5121751226. Bangert, R. K. and Slobodchikoff, C. N. 2006. Conservation of prairie dog ecosystem engineering may support arthropod beta and gamma diversity. J. Arid Environ. 67: 100115. Biggins, D. E. and Kosoy, M. Y. 2001. Influences of introduced plague on North American mammals: implications from ecology of plague in Asia. J. Mammal. 82: 906916. Bossard, R. L. 2006. Mammal and flea relationships in the Great Basin desert: from H. J. Egoscue s collections. J. Parasitol. 92: 260266. Brinkerhoff, R. J. et al. 2006. Abundance patterns of two Oropsylla (Siphonaptera: Ceratophylidae) species on black-tailed prairie dog (Cynomys ludovicianus) hosts. J. Vector Ecol. 31: 355 363. 8-OE

Brown, R. C. and Brown, W. B. 1996. Ectoparasitism as a cost of coloniality in cliff swallows (Hirundo pyrrhonata). Ecology 67: 12061218. Carlson, D. C. and White, E. M. 1987. Effects of prairie dogs on mound soils. J. Soil Sci. Soc. Am. 51: 389393. Ceballos, G. et al. 1999. Influence of prairie dogs (Cynomys ludovicianus) on habitat heterogeneity and mammalian diversity in Mexico. J. Arid Environ. 41: 161172. Collinge, S. K. et al. 2008. Effects of disease on keystone species, dominant species, and their communities. In: Ostfeld, R. et al. (eds), Infectious disease ecology: effects of ecosystems on disease and of disease on ecosystems. Princeton Univ. Press, pp. 129144. Cully, J. F. and Williams, E. S. 2001. Interspecific comparisons of sylvatic plague in prairie dogs. J. Mammal. 82: 894905. Davidson, A. D. and Lightfoot, D. C. 2007. Interactive effects of keystone rodents on the structure of desert grassland arthropod communities. Ecography 30: 515525. Eisen, R. J. et al. 2006. Early-phase transmission of Yersinia pestis by unblocked fleas as a mechanism for explaining rapidly spreading plague epizootics. Proc. Nat. Acad. Sci. USA 103: 1538015385. Gage, K. L. and Kosoy, M. Y. 2005. Natural history of plague: perspectives from more than a century of research. Annu. Rev. Entomol. 50: 505528. Guegan, J. F. and Kennedy, C. R. 1996. Parasite richness/ sampling effort/host range: the fancy three-piece jigsaw puzzle. Parasitol. Today 12: 367369. Holt, R. D. et al. 2003. Parasite establishment in host communities. Ecol. Lett. 6: 837842. Hubbard, C. A. 1947. Fleas of western North America. Iowa State College Press. Keesing, F. et al. 2006. Effects of species diversity on disease risk. Ecol. Lett. 9: 485498. Krasnov, B. R. et al. 1997. Host-habitat relations as an important determinant of spatial distribution of flea assemblages (Siphonaptera) on rodents in the Negev Desert. Parasitology 114: 159173. Krasnov, B. R. et al. 2001. Effect of air temperature and humidity on the survival of pre-imaginal stages of two flea species. J. Med. Entomol. 38: 630637. Krasnov, B. R. et al. 2004. Geographical variation in host specificity of fleas (Siphonaptera) parasitic on small mammals: the influence of phylogeny and local environmental conditions. Ecography 27: 787797. Krasnov, B. R. et al. 2005. Nested patterns in flea assemblages across the host s geographic range. Ecography 28: 475484. Krasnov, B. R. et al. 2006a. Ecological characteristics of flea species relate to their suitability as plague vectors. Oecologia 149: 474481. Krasnov, B. R. et al. 2006b. Are ectoparasite communities structured? Species co-occurrence, temporal variation and null models. J. Anim. Ecol. 75: 13301339. Lindenfours, P. et al. 2007. Parasite species richness in carnivores: effects of host body mass, latitude, geographical range, and population density. Ecol. Biogeogr. 16: 496509. Lorange, E. A. et al. 2005. Poor vector competence of fleas and the evolution of hypervirulence in Yersinia pestis. J. Infect. Dis. 191: 19071912. McCune, B. and Grace, J. B. 2002. Analysis of ecological communities. MJM Press. Miller, S. D. and Cully, J. F. 2001. Conservation of black-tailed prairie dogs (Cynomys ludovicianus). J. Mammal. 82: 889 893. Nunn, C. L. et al. 2005. Latitudinal gradients of parasite species richness in primates. Divers. Distrib. 11: 249256. Ostfeld, R. and Keesing, F. 2000. Biodiversity and disease risk: the case of lyme disease. Conserv. Biol. 14: 722728. Perry, R. D. and Fetherston, J. D. 1997. Yersinia pestis etiologic agent of plague. Clin. Microbiol. Rev. 10: 3566. Pollitzer, R. and Meyer, K. F. 1961. The ecology of plague. In: May, J. F. (ed.), Studies of disease ecology. Hafner Publ., New York, USA, pp. 433501. Ray, C. and Collinge, S. K. 2006. Potential effects of a keystone species on the dynamics of sylvatic plague. In: Collinge, S. K. and Ray, C. (eds), Disease ecology: community structure and pathogen dynamics. Oxford Univ. Press, pp. 202217. Rust, M. K. and Dryden, M. W. 1997. The biology, ecology, and management of the cat flea. Annu. Rev. Entomol. 42: 451 471. Satterthwaite, F. E. 1946. An approximate distribution of estimates of variance components. Biometrics 2: 110114. Schmidt, K. A. and Ostfeld, R. 2001. Biodiversity and the dilution effect in disease ecology. Ecology 82: 609619. Shaw, S. E. et al. 2004. Pathogen carriage by the cat flea Ctenocephalides felis (Bouché) in the United Kingdom. Vet. Microbiol. 102: 183188. Shipley, B. K. and Reading, R. P. 2006. A comparison of herpetofauna and small mammal diversity on black-tailed prairie dog (Cynomys ludovicianus) colonies and non-colonized grasslands in Colorado. J. Arid Environ. 66: 2741. Smith, G. A. and Lomolino, M. V. 2004. Black-tailed prairie dogs and the structure of avian communities on the shortgrass plains. Oecologia 138: 592602. Stanko, M. et al. 2002. Mammal density and patterns of ectoparasite species richness and abundance. Oecologia 131: 289295. Stanko, M. et al. 2006. Relationship between host abundance and parasite distribution: inferring regulating mechanisms from census data. J. Anim. Ecol. 75: 575583. Stapp, P. 2007. Rodent communities in active and inactive colonies of black-tailed prairie dogs in shortgrass steppe. J. Mammal. 88: 241249. Traub, R. 1985. Coevolution of fleas and mammals. In: Kim, K. C. (ed.), Coevolution of parasitic arthropods and mammals. Wiley, pp. 295437. Trudeau, K. M. et al. 2004. Sylvatic plague reduces genetic variability in black-tailed prairie dogs. J. Wildl. Dis. 40: 205211. Webb, C. T. et al. 2006. Classic flea-borne transmission does not drive plague epizootics in prairie dogs. Proc. Nat. Acad. Sci. USA 103: 62366262. Whicker, A. D. and Detling, J. K. 1988. Ecological consequences of prairie dog disturbances. Bioscience 38: 778785. Wilder, A. P. et al. 2008. Transmission efficiency of two flea species (Oropsylla tuberculata cynomuris and Oropsylla hirsuta) involved in plague epizootics among prairie dogs. Ecohealth, in press. Winter, S. L. et al. 2002. Vegetation of prairie dog colonies and non-colonized short-grass prairie. J. Range Manage. 55: 502508. Download the Supplementary material as file E5336 from Bwww.oikos.ekol.lu.se/appendix. 9-OE