The meningeal worm (Parelaphostrongylus tenuis) is a parasite of white-tailed deer (Odocoileus virginianus) and is also a significant pathogen of moose (Alces alces) and other ungulates. Changes in climate or habitat may facilitate range expansion or increase the prevalence of meningeal worm infection in white-tailed deer, resulting in increased exposure to susceptible ungulates. We examined 3,730 white-tailed deer during 2002–05 to determine the prevalence and range of meningeal worm infection in North Dakota, US, and investigated whether these had changed since earlier surveys. We used multiple logistic regression to model potential effects of habitat and climate on prevalence in white-tailed deer. We also examined how habitat influences intermediate hosts by comparing gastropod abundance and microclimate among habitat types. Prevalence in deer was 14% statewide, and prevalence and geographic range had increased since the early 1990s. Natural woodlands provided the best habitat for intermediate hosts, and increases in prevalence of infection in deer may be due to recent patterns in growing-season precipitation. This study has redefined the geographic distribution of meningeal worm infection and increased understanding of how climate and habitat influence the prevalence and distribution of this parasite.
Parelaphostrongylus tenuis, the meningeal worm, is a nematode parasite that occurs in the dura mater, subdural space, and venous sinuses of the cranium of white-tailed deer (Odocoileus virginianus; Anderson 1963). The meningeal worm has an indirect life cycle requiring one of several species of terrestrial gastropods as intermediate hosts (Lankester and Anderson 1968). It can cause fatal neurologic disease in several species of accidental hosts, including moose (Alces alces), elk (Cervus canadensis), and mule deer (Odocoileus hemionus; Lankester 2001), although its life cycle is generally only completed in white-tailed deer (Anderson and Prestwood 1981).
Because of the requirements of the life cycle of the meningeal worm, the potential for the parasite to occur in an area depends on climate and habitat factors that influence the distribution of definitive and intermediate hosts and affect the survival of first-stage larvae (Shostak and Samuel 1984; Forrester and Lankester 1998; Lankester 2001). Gastropod abundance is linked to cool, wet microclimates in wooded habitats (Lankester and Anderson 1968; Kearney and Gilbert 1978; Raskevitz et al. 1991), and woodlands and croplands in association with woodlands promote abundant white-tailed deer populations (Cote et al. 2004; Smith et al. 2007). Prevalence of infection is also positively related to growing-season precipitation (Wasel et al. 2003); increased growing-season precipitation enhances the ability of habitats to support terrestrial gastropods (Burch 1962; Bickel 1977) and may increase meningeal worm larval survival by helping maintain a moist mucous coating on the outside of deer feces containing larvae (Shostak and Samuel 1984).
In North Dakota, US, native prairie habitats have been modified by conversion to agriculture and planting of tree rows and shelter belts (Knue 1991; Licht 1997), resulting in considerable increases in the range and abundance of white-tailed deer (Knue 1991; Smith et al. 2007). If habitat alterations also support intermediate hosts, this habitat change may have also positively influenced the prevalence of meningeal worm infection. Further, since 1993 the eastern half of the state has experienced a long-term wet climate cycle (Todhunter and Rundquist 2004), and whether this climate cycle or climate patterns at shorter or longer temporal intervals have influenced meningeal worm prevalence is unknown. We sought to better understand the current prevalence and distribution of meningeal worm infection in North Dakota by 1) estimating the prevalence and distribution of meningeal worm infection in white-tailed deer in the state and examining whether the prevalence has recently increased, 2) estimating the relative importance of climate and land cover to infection prevalence, and 3) investigating how habitat contributes to the abundance of intermediate hosts.
MATERIALS AND METHODS
Prevalence and geographic range of the meningeal worm
We estimated the prevalence of meningeal worm infection in white-tailed deer throughout North Dakota from 2002 to 2005 by examining deer heads collected from hunter check stations, from meat processors, and by the North Dakota Game and Fish Department, using the protocols of Comer et al. (1991) and Prestwood and Smith (1969). All deer were collected during the November firearms deer season. We determined age of deer from tooth wear and eruption (Severinghaus 1949) and recorded the Deer Management Unit (DMU) of origin. We estimated the prevalence of meningeal worm infection in white-tailed deer in each DMU and overall (Fig. 1) and used a chi-square test for independence to compare prevalence among fawns (6 mo old), yearlings (1.5 yr old), and adults (2.5 yr old and older).
In a 1989–91 survey, 45 of 538 (8.5%) white-tailed deer in eastern North Dakota were infected (Wasel et al. 2003). We used a Cochran-Mantel-Haenszel test to assess whether the prevalence of meningeal worm infection in white-tailed deer in North Dakota had increased between that period and 2002–05. We restricted analyses to DMUs where meningeal worms were detected in both the 1989–91 and 2002–05 surveys and those where the 1989–91 sample size was sufficient to provide >90% chance of detecting meningeal worms during that period if the actual prevalence had been the same as the 2002–05 estimate. We also evaluated the possibility that any apparent increase in prevalence was due to the lower proportion of fawns in the 2002–05 sample (9%) compared to the 1989–91 sample (24%). We calculated the infection prevalence of all fawns (8.7% prevalence) and all 1.5-yr-old and older deer (15.1% prevalence) for the 12 DMUs included in the Cochran-Mantel-Haenszel analysis and created an adjusted 2002–05 prevalence estimate based on a sample containing 24% fawns and compared this adjusted estimate to the 1989–91 estimate.
Seasonal temperature and precipitation influence the range and prevalence of meningeal worm infection via gastropod distribution and larval survival (Lankester and Anderson 1968; Shostak and Samuel 1984; Forrester and Lankester 1998; Wasel et al. 2003). We obtained data on precipitation and temperature from 160 weather stations in North Dakota for the period 1963–2005 (National Oceanic and Atmospheric Administration 2015). We combined data from all stations within individual DMUs for estimating winter (November–April) and growing-season (May–October) precipitation (cm), and mean winter and growing-season temperature (C) for each DMU. We then divided climate data into three intervals: study period (2001–05), wet cycle (1993–2005), and historic (1963–92). We chose these intervals to investigate whether current levels of meningeal worm prevalence were most closely related to current climate conditions, conditions produced by the recent wet cycle, or the longer-term climate regime of the region leading up to the wet cycle. We constructed a multiple logistic regression model for each interval with growing-season precipitation, winter precipitation, growing-season temperature, and winter temperature as independent variables and meningeal worm prevalence (number of infected deer/number sampled) as the dependent variable.
Land cover model
Woodland habitats serve as habitat for both intermediate hosts and white-tailed deer (Lankester and Anderson 1968; Raskevitz et al. 1991; Smith et al. 2007) and, in prairie-dominated regions, the wet meadow zone surrounding wetlands may serve as quality habitat for intermediate hosts (Jacques 2001), while cattails provide cover for deer (Gould and Jenkins 1993). Croplands also provide foraging habitat for deer (Cote et al. 2004; Smith et al. 2007). We used the Spatial Analyst Extension in ArcGIS 10 (ESRI Inc., Redlands, California, USA) to delineate woodlands (all artificial and natural woodlands), croplands, and wetlands (semipermanent wetlands, lakes, and rivers) from land cover data from the North Dakota Gap Analysis Project (US Geological Survey 2005). We determined percentage of cover for each habitat type by dividing the area of each cover type present in each DMU by the area of the DMU. We constructed a multiple logistic regression model (Table 1) to investigate whether land cover was related to meningeal worm infection prevalence. This model included the percentage of cover of woodlands, wetlands, croplands, and the two-way interactions for woodlands and croplands and wetlands and croplands as independent variables.
Model selection and combined models
To eliminate imprecise prevalence estimates resulting from low sampling intensity, we restricted analyses for all models to the 29 DMUs for which at least 20 deer were sampled. Although sampling fawns may result in an underestimate of prevalence (Lankester 2001), they comprised only 9% of our sample. Further, to account for whether differences among DMUs in the age structures of samples may have influenced prevalence estimates, we included the proportion of adult deer sampled in each DMU as a covariate in each model. We centered and standardized independent variables (by subtracting the mean from each value then dividing by the standard deviation) to reduce leverage of extreme values or account for differences in the scale of measurement.
To test hypotheses about which variables or combinations of variables best explained meningeal worm infection prevalence, we reduced climate (historic, wet-cycle, study-period) and habitat (percentage of cover) models using backward selection and used variables from these models (Table 1) to create a series of combined climate models, which included all possible combinations of independent variables from historic, wet-cycle, and study-period models (Table 1). We then created a second series of combined models by adding variables from the percentage of cover model to the historic, wet-cycle, and study-period models, and each of the combined climate models. We reduced models to include only significant predictors (α = 0.05). We calculated McFadden's r2 for all models and used Akaike's information criterion corrected for small sample size (AICc; Burnham and Anderson 2002) to rank models. All analyses were performed in R 2.6 (R Development Core Team 2007). We tested for spatial autocorrelation by mapping model residuals of the top-ranked combined model on to DMUs and calculating Moran's I. We considered a Moran's I value to indicate spatial autocorrelation if the z score derived from that value was >1.96.
Terrestrial gastropods and microclimate
From May to September 2003 to 2005, we used cardboard cover-board transects to sample terrestrial gastropods at 10 sites (five in 2003, 10 in subsequent years) throughout eastern North Dakota (Fig. 1). We defined five habitat types: native prairie, nonnative grassland, tree row (planted tree rows one tree wide containing no leaf litter), planted woodlot (planted tree rows or woodlots at least two trees wide containing leaf litter/woody debris), natural woodland (woodlands not artificially planted), and selected sites to permit balanced sampling of all habitat types. Within each site we nonrandomly selected patches to be sampled to ensure sampling effort was evenly distributed among habitat types. In 2003, we placed five transects in each habitat type at each site and sampled sites every 2 wk. In 2004 and 2005 we accommodated sampling a larger number of sites by placing three instead of five transects in each habitat type at each site and sampling every 4 wk. We randomly placed transects a minimum of 100 m from the habitat edge in patches >200 m wide, and we centered transects within patches <200 m wide. Each transect consisted of 10 30×30-cm cardboard squares, placed at 5-m intervals (Boag 1982). We wetted squares, covered them with clear plastic sheeting, and staked them to the ground to prevent them from blowing away (Lankester and Peterson 1996). We checked transects after 48 h and collected all gastropods adhering to cover boards beginning at first light and concluded collection within 2 h, before hot temperatures and direct sunlight drove gastropods from coverboards.
We hypothesized that gastropods would be more abundant on transects in natural woodlands than in other habitats, but that planted woodlands (woodlots and tree rows) would provide more suitable habitat (e.g., leaf litter, moist microclimates) than grassland habitats (native prairie and nonnative grasslands). Because sampling effort was uneven among sites, we pooled years and standardized the abundance of gastropods for each transect by dividing the total number of gastropods collected by the number of times that transect was sampled. We used Wilcoxon rank sum tests to conduct planned contrasts to compare gastropod abundance in naturally occurring woodlands vs. all other habitats, in planted woodlots vs. grasslands, and in tree rows vs. grasslands.
We also hypothesized that habitat types harboring greater numbers of gastropods would be relatively cool and wet compared to habitats harboring fewer or no gastropods. We assessed this hypothesis by taking hourly measurements of temperature (C) and relative humidity with 15 automatic weather recorders (Hobo data loggers; Onset Computer Corporation, Bourne, Massachusetts, USA) set 4–5 cm above ground level near the center point of three randomly selected cover-board transects of each habitat type from May to September 2004 (Fig. 1). We constructed linear mixed-effects models with habitat as a fixed effect, site as a random effect, and mean daily maximum temperature or mean daily minimum humidity from each transect as the response variable. We then used likelihood ratio tests (full models vs. null models including only the random effect) to assess the significance of the habitat factor, and conducted Tukey's post hoc tests to make pairwise comparisons of temperature and humidity between habitat types.
Prevalence and geographic range of meningeal worm infection
We examined 3,730 white-tailed deer originating from every DMU in North Dakota for meningeal worms. Infection was detected in 26 of 38 DMUs, including 14 DMUs where it had not been reported previously. Overall prevalence of infection was 14.5% (95% confidence interval [CI] = 13.3–15.6%), which was higher than previously reported (8.5%; Wasel et al. 2003). Prevalence was lower in fawns (8.7%, 95% CI = 6.3–11.7%) and yearlings (10.7%, 95% CI = 9.0–12.5%) than in 2-yr-old and older deer (16.6%, 95% CI = 15.1–18.2%; χ22 = 32.4, P<0.001). For DMUs with meningeal worm infection, prevalence ranged from 0.7% to 35.1%. Prevalence declined east to west, and infection was not detected in the southwest corner of the state (Fig. 1). We detected meningeal worm infection in two deer from DMUs 3B2 and 3E2, west of the Missouri River. Meningeal worms were previously unknown for areas this far west in North Dakota (Fig. 1). Prevalence was higher in 2002–05 than in 1989–91 (χ21 = 8.9, P = 0.001; Table 2), and this difference does not appear to be the result of the greater proportion of fawns in the 1989–91 sample, as the prevalence estimate from the adjusted 2002–05 sample with 24% fawns was 13.5%, compared to 8.4% in the 1989–91 sample. (A statistical comparison of these two samples was not conducted because of the unknown variance component in the simulated sample.)
The reduced historic, wet-cycle, study-period, and percentage of cover models identified several candidate variables to be included in combined models (Table 1). The top-ranked combined model included positive relationships between prevalence and wet-cycle growing-season precipitation; wet-cycle growing-season temperature; percentage of cover of woodlands, wetlands, and croplands; two-way interaction of woodlands and croplands, and proportion of adult deer sampled. The model also included an inverse relationship between prevalence and historic winter temperature (Table 3). The five best models all included wet-cycle growing-season precipitation, and this variable also had the largest effect size in all models (Table 3). There was no evidence of spatial autocorrelation in the best combined model (Moran's I = 0.01, z = 0.27, P = 0.79).
Terrestrial gastropods and microclimate
We collected 2,778 gastropods from nine of 10 sites sampled. Nine of 15 species detected (13 snail and two slug species) were known intermediate hosts for meningeal worms (Table 4). Deroceras laeve was the most common and widespread gastropod encountered (detected at nine sites). Gastropods were more abundant in woodlands than all other habitats (W = 1,070.5, P<0.001) and were marginally more common in planted woodlots than in grasslands (W = 215.5, P = 0.064). However, gastropod abundance did not differ between tree rows and grasslands (W = 183.5, P = 0.38). Mean daily maximum temperatures (±SD) varied among habitat types (likelihood ratio χ24 = 26.6, P<0.001) and were significantly cooler in natural woodlands (21.4±1.3 C), followed by planted woodlands (22.6±1.2 C) and were warmest in native prairie (36.1±2.7 C) and nonnative grasslands (31.6±1.4 C). Habitat type also influenced humidity (likelihood ratio χ24 = 28.2, P<0.001). Natural woodlands (76.0±6.0%) and planted woodlots (75.5±3.8%) had the highest mean daily minimum humidity, whereas tree rows (54.3±13.8%), native prairie (51.5±2.9%), and nonnative grasslands (38.6±6.0%) were drier.
Meningeal worm infection is less common in North Dakota than in eastern North America, where prevalence ranged from 44% to 90% in Michigan, Maine, Minnesota, New York, and Ontario (Nankervis et al. 2000; Lankester 2001). Age-specific prevalence was similar to that reported in others studies, with fawns and yearlings having lower prevalence than older animals (Garner and Porter 1991; Wasel 1995; Banks and Ashley 2000). More importantly, the prevalence of meningeal worm infection in North Dakota may have increased since the early 1990s, and this parasite is more widespread than previously thought. We found no evidence of infection in the southwestern corner of the state; however, the small number of deer we examined from some of the DMUs in this area prevents us from definitively concluding that the parasite is absent there.
The highest-ranked combined model identified several factors that help explain the current prevalence of meningeal worm infection in North Dakota. Of these, growing-season precipitation during the wet cycle appears to have the strongest influence. Additionally, all of the top five models included wet-cycle growing season precipitation as an independent variable, and this variable had the largest effect size in all of the top-ranked models. These results support those of Wasel et al. (2003), who found that areas of North Dakota, Manitoba, and Saskatchewan where meningeal worm infection was present had higher summer and fall precipitation. It also suggests that the recent wet cycle in the eastern half of the state was the most important factor in the observed increase in prevalence. The inverse relationship between prevalence and historic winter temperature had the second highest effect size in the top-ranked model. These results also support those of Wasel et al. (2003) and suggest that the longer-term climate regime may influence transmission conditions. Finally, although not as influential in our top-ranked model, the positive relationship between growing-season temperature and prevalence suggests that warmer average temperatures in summer and fall did not limit larval survival and gastropod availability.
Although habitat variables had smaller effect sizes than climate variables in our top-ranked model, the inclusion of habitat variables suggests that habitat does influence prevalence. The positive relationship between percentage of woodland cover and prevalence may be related the importance of woodlands in providing habitat for gastropods (Lankester and Anderson 1968; Kearney and Gilbert 1978; Raskevitz et al. 1991). Gastropod collections and microclimate measurements support this idea, as wooded habitats harbored the most gastropods and had cooler and more humid microclimates. Further, the inclusion of the two-way interaction of woodland and crop cover in the top-ranked model may be related to the ability of these habitats to support white-tailed deer. Additionally, the positive relationship between wetland cover and prevalence may be due to the importance of these habitats to support deer (Gould and Jenkins 1993) as well as gastropods (Jacques 2001). The proportion of adult deer sampled was retained in three of the top five models, but the effect size for this variable was small, and its presence did not appear to obscure the effects of climate and habitat variables.
This study demonstrated the combined importance of climate and habitat as determinants of prevalence of meningeal worm infection in the northern Great Plains. Although habitat likely impacts prevalence via its influence on the abundance of both deer and gastropods, the most important determinant of the distribution and prevalence of meningeal worm infection appears to be climate, especially growing-season precipitation. Periodic climate cycles likely affect prevalence by influencing the ability of habitats to support gastropods and first-stage larvae. However, despite increased prevalence, the range of meningeal worm infection still appears limited to slightly beyond the Missouri River. Thus, climate and habitat in the Great Plains may still impose some limit to the westward expansion of this parasite, supporting the hypothesis proposed by Jacques and Jenks (2004) and Samuel and Holmes (1974).
We thank the North Dakota Experimental Program to Stimulate Competitive Research, the North Dakota Game and Fish Department, the US Bureau of Reclamation, the US Fish and Wildlife Service, the North Dakota chapter of The Wildlife Society, the University of North Dakota Biology Department, and the Wheeler Scholarship for funding this project. We also thank W. Jensen, R. Johnson, and S. Peterson for allowing us to collect samples from deer hunters and for permitting us to conduct work on North Dakota Game and Fish lands. Additionally, we are grateful to R. Newman, B. Rundquist, and J. Vaughan for their helpful input on this manuscript. Finally, we thank J. Smith, E. Pulis, T. Manuwal, B. Bly, S. Bertie, C. Brewer, K. Pulis, and S. Milne-Laux for assistance in the field and lab.