1Parasite transmission frequently depends on an intermediate host species being subject to predation by a definitive host. We hypothesized that the population dynamics of this type of parasite would be affected by plasticity of the predation behaviour of the final host.
2The zoonotic tapeworm Echinococcus multilocularis Leuckart is transmitted to foxes (Vulpes vulpes L.) by predation on infected rodents. One possible mechanism underlying the significant decrease in the prevalence of E. multilocularis in red foxes observed towards the centre of Zurich (Switzerland) is that the relative abundance of intermediate hosts varies between different urbanized habitats.
3The water vole, Arvicola terrestris scherman L., which is the major intermediate host, was less abundant in urban compared with peri-urban areas, due to both a reduced availability of suitable habitats and lower numbers within those habitats. Stomach content analyses indicated that foxes in urban areas consumed more anthropogenic food but preyed less frequently on rodents – and those rodents taken belonged to species less susceptible to E. multilocularis infection.
4We conclude that the functional response of final hosts to changes in supply of intermediate hosts and alternative food sources can significantly affect the transmission dynamic of dixenous parasites, such as this zoonotic tapeworm.
Many macroparasites change their host species during their life cycle. Frequently, one or more asexual larval stages live in prey species and develop to an adult sexual stage in a predator species. It is assumed that a parasite enhances its transmission rate by having a larval stage in an abundant intermediate host that is a frequent prey species for a final host (Choisy et al. 2003), and that this strategy also increases the chances of finding a sexual mate (Brown et al. 2001). In addition, the different temporal and spatial dynamics of intermediate and definitive host species could enhance dispersal and increase the persistence of a parasite population (Mackiewicz 1988). However, this alternation between species makes a parasite dependent on the prey–predator interaction, and changes in food acquisition by the definitive host (which can be caused by factors such as environmental change) could adversely affect the parasite's life cycle (Mackiewicz 1988).
The prey–predator interplay is therefore a key factor in understanding the transmission dynamics of obligatory heteroxenous parasites such as cestodes. For example, the small fox tapeworm Echinococcus multilocularis Leuckart is a zoonotic parasite with a dixenous life cycle (Eckert & Deplazes 2004). The adult, intestinal stage lives in various carnivores and eggs are excreted in faeces. Intermediate hosts are infected by ingestion of these eggs and harbour the parasite's larval stage (metacestode) in their liver. To complete the life cycle of the parasite, a definitive host must ingest fully developed metacestodes by preying on infected intermediate hosts.
In east and central Europe, the main definitive host is the red fox (Vulpes vulpes L.), with arvicolid species, predominantly the common vole (Microtus arvalis Pallas) and the fossorial form of the water vole (Arvicola terrestris scherman L.), acting as intermediate hosts (Eckert 1998). Although E. multilocularis is remarkably prevalent in foxes in endemic areas (20–70% individuals are infected), prevalence rates in intermediate species are generally low (<1–6%) (Eckert 1998), and rarely exceed 20% (Gottstein et al. 1996; Henttonen et al. 2001) with infection recorded only occasionally in murid rodents (Eckert et al. 2001).
In the city of Zurich, the prevalence of E. multilocularis in foxes varies both temporally and spatially, with the highest prevalence rate (47%) occurring in winter (Hofer et al. 2000). Prevalence rates in potential intermediate host species are also variable, being remarkably high in water voles (22%), substantially lower in both common voles (4·9%) and bank voles (Clethrionomys glareolus Schreber) (2·4%), and zero in wood mice (Apodemus sylvaticus L.) and yellow-necked mice (Apodemus flavicollis Melchior) (Stieger et al. 2002; unpublished data). These different rodent species exhibit distinct differences in ecology and habitat preferences: arvicolid species (A. terrestris and M. arvalis) live mainly in meadows and pastures, whereas murid species (e.g. A. sylvaticus) prefer bushy habitats with more shelter, such as forests and hedges (Hausser 1995), but can also reach high density in urban habitats such as public parks and residential areas (Dickman & Doncaster 1987). Consequently, we would expect (1) the distribution of rodent species to vary with the size and fragmentation of these key habitat types in the urban environment; (2) that this variation would be reflected in the diet of the red fox; and (3) that this would, in turn, affect the transmission dynamics of E. multilocularis.
Central Zurich is a densely built urban area which is ringed by residential areas with more green space; beyond the city border are rural areas dominated by forests and agriculture. Within the urban area a high supply of anthropogenic food is available for urban foxes (Contesse et al. 2004). Thus we made the following predictions: (1) the supply of suitable habitat for voles – and thus their availability – decreases from peri-urban areas towards urban areas; (2) as a functional response, foxes in urban areas would consume fewer voles and more anthropogenic food sources than in the adjacent peri-urban area; and (3) the prevalence of E. multilocularis in foxes would correspond to the predation rate of suitable intermediate hosts.
The political community of Zurich covers 92 km2, with 360 000 human inhabitants, and consists of 53% built-up area, 24% forest, 17% agricultural land and 6% water (Anonymous 2000). The built-up area and the lake are situated in the centre, while the agricultural areas and the forests are located in the urban periphery. The study area was therefore divided into three zones, with the central urban and outer peri-urban zones (39 and 20% of the area) separated by a border zone (41%; Fig. 1). The urban zone contained little green space, and we defined its limit by a line following properties where buildings were in daily use and which were no more than 100 m from the next such building. The border zone was a 500-m-wide belt, extending 250 m either side of this boundary, which consisted mostly of residential areas, allotment gardens, cemeteries, sports fields and public parks. The peri-urban zone consisted of forests, fields, pastures and meadows, which are intensively used for recreational activities.
An extensive telemetry study in the city of Zurich has demonstrated that resident foxes rarely range between the urban and the peri-urban zones (Minimal Convex Polygon (MCP) home range sizes of 28·8 ± 22·7 ha for females and 30·8 ± 11·0 ha for males; Gloor 2002). Correspondingly, significant differences in the prevalence rates of E. multilocularis in foxes have been detected over distances of just 500 m (Hegglin, Ward & Deplazes 2003).
intermediate hosts: habitat availability and density estimates
Habitat suitability for the arvicolid species A. terrestris and M. arvalis was assessed by calculating the area of meadows and pastures on-screen with the help of the GIS software package ArcView 3·3 (Environmental Systems Research Institute, Redlands, CA, USA). The relative population density of A. terrestris and M. arvalis was calculated according to Giraudoux et al. (1995) and Quéréet al. (2000), respectively. The methods are based on counting superficial signs of the activity of the two rodent species on predefined transects. The fossorial form of A. terrestris builds extended subterraneous burrow systems, which are accompanied by superficial earth tumuli. The common vole establishes clearly visible runways above the ground, where they also deposit their faeces. A total of 35 1-m-wide transects were paced out by the same observer (each 200 steps long, about 150 m) and at randomly selected locations in meadows and pastures (10 transects in the urban, 13 in the border and 12 in the peri-urban zone), and the proportion of 10-step intervals with tumuli of A. terrestris or faeces of M. arvalis was used as relative index of abundance for the two species (ranging from 0 to 100%). All transects were walked in the period 2–19 December 2003, when vegetation was low and the bare ground was visible.
definitive host: predation on intermediate hosts and consumption of anthropogenic food
The game wardens of the city of Zurich collected and recorded all foxes shot or found dead between January 1996 and June 2001 (n = 789). The foxes were categorized as urban (31%), border (49%) or peri-urban (19%) according to the site of their collection, as well as by season of collection [spring (March–June, 16% of the sample), summer/autumn (July–October, 35%) and winter (November–February, 49%)], sex (equal numbers of males and females) and age (younger than 1 year vs 1 year or older, determined according to Kappeler 1991: 55% juveniles, 45% adults.
After dissection of the stomach of each fox, the contents were washed in a sieve and possible remains of rodents were separated (for details see Contesse et al. 2004). Stomachs that contained <2 ml were classified as empty. Rodents were identified to species level according to Niethammer & Krapp (1982), if recognizable feet, tails or teeth were present. The keys of Day (1966) and Teerink (1991) were used to identify species by their hairs. A subset of 402 of these fox stomachs were analysed for remains of food of anthropogenic origin (Contesse et al. 2004). For this study, a fox was classified as having consumed anthropogenic food when its stomach contained at least one of the following: scavenged meat (processed meat and bones), other scavenge (mainly vegetable kitchen waste, processed food), pets and domestic stock, pet food or bird seed (bird-feeding mixtures).
parasite: prevalence in definitive host
Of the 789 foxes, 582 were used for parasitological analyses. Foxes collected after April 2000 were excluded because anthelmintic baits had been distributed in parts of the study area as part of a study on the control of E. multilocularis.
Echinococcus multilocularis infection in foxes was detected by the sedimentation and counting technique (for details see Hofer et al. 2000). Briefly, the small intestine was cut longitudinally and shaken in 0·9% NaCl solution. The sediment was washed by a standardized procedure of repeated dilution, shaking and decanting in a glass bottle, then small portions (about 5–10 ml) were examined in square Petri dishes for the presence of E. multilocularis.
Density indices for A. terrestris and M. arvalis could not be normalized and were compared between the three zones using the non-parametric Kruskal–Wallis test. Dunn's post hoc tests were performed to identify significant differences between zones.
Feeding behaviour of foxes and prevalence rates of E. multilocularis in foxes can be affected not only by spatial factors, but also by temporal (season, year; Weber & Aubry 1993; Stieger et al. 2002) and individual factors (age, sex; Tackmann et al. 1998). Logistic regressions were used for modelling predation rates on rodents in general (and voles in particular), the consumption rate of anthropogenic food and the prevalence rates. The factors zone, season and age were selected as the most promising independent variables for the modelling procedure. Furthermore, the age × season interaction was considered because the age structure of a fox population varies during seasons, with most young foxes in spring and fewest in winter. The age × zone interaction was included because hunting pressure was lower in the built-up areas than in the urban periphery (unpublished data), and this may have caused differences in the age structure of the fox population between the zones.
We fitted logistic regression models using all possible combinations of the selected predictor variables (zone, season, age, age × season, age × zone), but only models that significantly explained variation in the data set (log-likelihood ratio test) were included in the model-selection process. Best models were then selected using Akaike's information criterion (AIC; Akaike 1973; Burnham & Anderson 1998) corrected for small sample sizes (AICc, Burnham & Anderson 1998). The model with the lowest AICc can be viewed as the most parsimonious. Models with ΔAICc < 2 compared with the model with the lowest AICc were selected. We then calculated Akaike model weights (Burnham & Anderson 1998) to determine the degree to which a model is supported by the data. Weights of selected models sum to 1 by definition, and higher weights indicate better explanatory power.
To evaluate the relative importance of each explanatory variable, the AICc weights were summed up over all models where that variable was present. A variable that is included in all selected models sums up to 1, whereas a variable present in just one model gets the value of the corresponding model. Subsequently, the variables were ranked according to the resulting AICc weight sums. Odds ratios and their 95% confidence interval are given for the best of the selected models. An odds ratio of 1 indicates that an event is equally likely in both groups. Therefore the only factors that were interpreted were those that comprised at least one odds ratio with a 95% CI that did not include 1.
Statistical analyses were conducted using spss 10·0 (Norusis 1986). The fit of the logistic regression models were tested by likelihood ratio and Hosmer–Lemeshow goodness-of-fit tests. The minimal P value (Pmin) of the Hosmer–Lemeshow goodness-of-fit tests and the maximal P value (Pmax) for the likelihood ratio tests are given for the selected models. To express the sensitivity and specificity of the best models, receiver operating characteristic analyses were performed and threshold probabilities were determined by minimizing the absolute value of the difference between sensitivity and specificity (Cantor et al. 1999).
intermediate hosts: habitat availability and density estimates
Meadows and pastures comprised a substantially greater proportion of the border (18·4%) and peri-urban (12·0%) zones than the urban zone (1·5%; Fig. 2a). The relative abundance of A. terrestris differed significantly between the three zones (Kruskal–Wallis test: χ2 = 10·5, df = 2, P < 0·01), being significantly higher in the border area (41·5%) than in the urban area (8·5%; Fig. 2b, Dunn's post hoc test, P < 0·05). The index of relative abundance of M. arvalis was much lower than that observed for A. terrestris and did not differ significantly between the three zones (Kruskal–Wallis test: χ2 = 0·1, df = 2, P > 0·1).
definitive host: predation on intermediate hosts and consumption of anthropogenic food
Of the 789 stomachs analysed, 339 were empty. Only fox stomachs with content were considered for our analyses (N = 450). Remains of rodents were detected in 95 stomachs (21%). Of the 97 remains recorded (some stomachs contained more than one species), 56 originated from arvicolids (21, 22 and 13 remains of the genera Arvicola, Microtus and Clethrionomys, respectively); 32 from murids (26, four and two remains of the genera Apodemus, Mus and Rattus, respectively); three from the red squirrel (Sciurus vulgaris) and six from unidentified rodents.
The model-selection process to explain predation by foxes on rodents yielded eight logistic-regression models, all showing significant zonal variations (Table 1). The adequacy of all these multivariate models was supported with non-significant results from the Hosmer–Lemeshow goodness-of-fit test (Pmin = 0·60) and significant results from the log-likelihood ratio test (Pmax = 0·004). Ranking of the independent variables revealed zone as the most relevant factor (AICc weight sum = 1), followed by season (0·70), age (0·65), and the interaction terms for zone × age (0·33) and season × age (0·17). The best model comprised the two factors zone and season, while zone was the only factor with 95% confidence intervals of the odds ratios that did not include 1 (urban and border zone vs peri-urban zone; Table 2). In the urban and peri-urban zones, 16·5 and 18·8%, respectively, of stomachs contained rodents compared with 39·7% in the peri-urban zone (Fig. 2c).
Table 1. Results from logistic-regression model selection to explain (a) predation by foxes on rodents (N = 450 fox stomachs with content); (b) proportion of arvicolids in predated rodents (N = 95 fox stomachs with rodents); (c) consumption of anthropogenic food (N = 202 fox stomachs with content); (d) prevalence rates of Echinococcus multilocularis in foxes (N = 582 fox intestines)
AICc, the Akaike information criterion corrected for small sample sizes and ΔAICc, the difference in AICc to the best model, are given. Weight indicates relative support of a particular model compared with the other models, higher values indicating better support. Selected models with ΔAICc < 2 were used for model averaging (Table 2).
(a) Predation on rodents (N = 450)
Zone, season, age, zone × age
Zone, season, age
Zone, season, age, season × age
Zone, age, zone × age
Zone, season, age, zone × age, season × age
(b) Predated arvicolid rodents (N = 95)
(c) Anthropogenic food (N = 202)
(d) Prevalence rates (N = 582)
Zone, season, age, zone × age, season × age
Zone, season, age season × age
Table 2. Odds ratios (OR), 95% confidence intervals (CI), threshold probabilities, sensitivities and specificities of the best models (for N see Table 1)
(a) Predation on rodents
(b) Predated arvicolid rodents
(c) Anthropogenic food
(d) Prevalence rates
(a) Predation by foxes on rodents; (b) proportion of arvicolids in predated rodents; (c) consumption of anthropogenic food; (d) prevalence rates of Echinococcus multilocularis in foxes.
Urban vs peri-urban
Border vs peri-urban
Spring vs winter
Summer/autumn vs winter
Age (juvenile vs adult)
Zone × age
Urban × juvenile vs peri-urban × adult
Border × juvenile vs peri-urban × adult
Season × age
Spring × juvenile vs peri-urban × adult
Summer/autumn × juvenile vs winter × adult
A second model-selection process, based only on stomachs with rodent remains, was used to explain the fox predation rate on voles (N = 95), and yielded two models (Hosmer–Lemeshow goodness-of-fit test: Pmin = 0·93; log-likelihood ratio test, Pmax = 0·003). Again, zone turned out to be the most relevant factor (AICc weight sum = 1), followed by season (0·27). No other factor entered the selected models (Table 1). The best model comprised only the factor zone (Table 2). The proportion of rodents belonging to the Arvicolidae (Arvicola, Microtus and Clethrionomys) was less than half in the urban zone compared with the border and peri-urban zones (30·4 vs 68·3 and 70·4%; Figs 1a and 2d).
A total of 202 of the 402 fox stomachs investigated for anthropogenic food contained food remains. Remains of anthropogenic food were found in 151 of these stomachs (74·8%). Only one model (Hosmer–Lemeshow goodness-of-fit test, P = 0·99; log-likelihood ratio test, P = 0·003) with the independent factors age and zone remained after the model-selection process (Table 1). A higher proportion of fox stomachs contained food from anthropogenic food sources in the urban zone (86·8%; Fig. 2e) than in the peri-urban zone (61·7%; Table 2).
parasite prevalence in definitive host
A total of 582 fox intestines were analysed for E. multilocularis infections. Three models explained best prevalence rates in foxes (Hosmer–Lemeshow goodness-of-fit test, Pmin = 0·41; log-likelihood ratio test, Pmax < 0·001). All included the factors zone and season (Table 1). Accordingly, the ranking of the independent variables revealed zone and season as the most relevant factors (AICc weight sum = 1 for both), followed by age and season × age (0·69 for both), and zone × age (0·45). The best model comprised all factors that were added to the model-selection procedure (Table 2). However, the odds ratio 95% CIs of season and age comprised 1, and were therefore not interpreted (Table 2). The prevalence rates in foxes decreased from the peri-urban (62·9%) to the border (39·5%) and the urban zones (16·5%’Fig. 2f). The interaction term zone × age revealed that this decrease was more pronounced for adult foxes (63·1% peri-urban, 41·2% border and 7·5% urban zone) than for juvenile foxes (62·6% peri-urban, 37·6% border and 19·2% urban zone). Furthermore, seasonal changes of prevalence rates proved to be more pronounced in juvenile (13·2% spring, 20·4% summer/autumn and 56·7% winter) than in adult foxes (28·9% spring, 31·0% summer/autumn and 50·0% winter).
Parasites depending on a prey–predator system are limited to locations where the distributions of definitive and intermediate hosts intersect. Their occurrence and persistence depend on the availability of host species and the specific environmental conditions that enable the survival of free-living stages.
Our data indicate that the supply of intermediate hosts for E. multilocularis is strongly reduced in urban areas. The intermediate hosts A. terrestris and M. arvalis are strongly associated with meadows and pastures (Hausser 1995), but these habitats were 12·3 times less abundant in the urban area than in the adjacent border area (Fig. 2a). Furthermore, meadows and pastures sustained substantially lower densities of A. terrestris in the urban zone (Fig. 2b), possibly because of the increased fragmentation of these habitats, which would affect the connectivity of subpopulations, and because of the lower dispersal potential of arvicolid species relative to murid species (Dickman & Doncaster 1989). Consequently, the rodent community in the urban zone is likely to consist of a substantially greater proportion of murid species.
Predation by foxes on A. terrestris and M. arvalis correlates in general with the abundance of these two prey species (Weber & Aubry 1993; Kjellander & Nordström 2003), with foxes switching to alternative prey when voles become rare (Kjellander & Nordström 2003). Few signs of the arvicolid M. arvalis were detected in the meadows and pasture investigated here, although we recorded M. arvalis as frequently as A. terrestris in fox stomachs. Despite their dietary plasticity, different studies point out that foxes show preferences for certain prey species (Macdonald 1977; Green 2002). Microtus species appear to be more attractive than other arvicolid species, and arvicolid species in general are more attractive than murid species (Macdonald 1977). Correspondingly, we found that the decrease in predation on vole species (Fig. 2c,d) was much less pronounced than one would expect given the observed decrease in the supply of vole prey towards the city centre (Fig. 2a,b).
Nevertheless, in accordance with prediction 2, the analyses of stomach content indicate that foxes can shift their diet to murid species (which are less susceptible to E. multilocularis infection) if arvicolid species are rare (Fig. 2d). In addition, the total predation rate on rodents was significantly lower for foxes in urban areas than for foxes in the urban periphery, further reducing the total uptake of intermediate hosts (Fig. 2c). This functional response may be caused by the surplus of anthropogenic food accessible to foxes in urban areas, reflected in the increased contribution of such items to the fox diet (Fig. 2e).
A high amount of anthropogenic food available enables foxes to fulfil their requirements within small home ranges encompassing not more than 30 ha (Gloor 2002). We have shown in a preceding experimental study that this small-scale spatial organization of urban foxes is reflected in a low spatial dynamic of the life cycle of E. multilocularis. In six areas, each 1 km2, foxes were dewormed regularly by the delivery of anthelmintic baits over 2 years. However, the effect of this treatment was detectable only within a range of 500 m around the baiting areas (Hegglin et al. 2003). This finding corresponds to the manifold decrease of E. multilocularis prevalence within a distance of not more than 500 m along the urban periphery (prediction 3; Fig. 2f). Hence we believe that the low supply of suitable intermediate hosts and the high levels of anthropogenic food sources available, combined with the resulting small-scale spatial organization of urban foxes, are responsible for the pronounced differences in the prevalence of E. multilocularis over small distances.
Conversely, however, anthropogenic food sources could enhance the transmission of E. multilocularis by increasing host densities. In urban areas, fox densities can exceed 10 adults per km2, much higher than observed in rural areas (Baker et al. 2000; Gloor 2002). Thus the transmission of E. multilocularis could be especially intense in the transition zone between urban and peri-urban areas where high fox population densities are sustained by high anthropogenic food resources and suitable habitat for arvicolid species is abundant (Deplazes et al. 2004).
Our results suggest that the temporal and spatial dynamics of E. multilocularis is also affected by the age structure of fox populations. In juvenile foxes, the spatial variation of the prevalence rates was less pronounced than in adult foxes. This may reflect the spatial behaviour of juvenile foxes, which frequently show exploratory behaviour and/or disperse before they establish a territory. Furthermore, dispersion occurs mainly during late autumn and winter (Trewhella 1988), the period when the prevalence in juvenile foxes was highest in our study. Thus juvenile foxes may have best access to voles, and their roaming behaviour may play an important role in the spread of this parasite.
The larval stage of E. multilocularis can cause human alveolar echinococcosis, a severe helminthic zoonosis which is fatal if left untreated (Ammann & Eckert 1995). However, low incidence rates of human alveolar echinococcosis have been recorded in Europe to date (Kern et al. 2003). Apart from individual risk factors (e.g. owning dogs, living in a farmhouse; Kern et al. 2004), ecogeographical factors such as a high vole densities appear to increase the risk for human alveolar echinococcosis (Viel et al. 1999). The growing fox population in Western Europe (Chautan, Pontier & Artois 2000), and the newly observed colonization of urban habitats in many central European cities (Gloor et al. 2001), raise concern about the risk of a possible increased infection pressure for this disease in densely populated areas (Eckert & Deplazes 2004). Therefore knowledge about the factors affecting the population dynamic of this zoonotic parasite is relevant to public health.
This study confirmed our predictions that the relative supply of arvicolids does correlate with the predation of foxes on these rodent species and, as a consequence, corresponds with the spatial differences in the prevalence of E. multilocularis in foxes. Our results support the hypothesis that the supply of prey species, acting as intermediate hosts, and the abundance of alternative food sources, can significantly affect the population dynamics of dixenous parasites and confirm the crucial role of plasticity in predation behaviour of definitive hosts for this type of parasite. Hence the functional response of final hosts to a varying supply of intermediate hosts and alternative food sources can be considered as a key factor for the transmission dynamics of parasites that depend on a predator–prey interplay.
We thank Christian Stauffer and the game wardens of the city of Zurich for their generous collaboration. Valuable comments of Alexander Mathis, Michael Schaub, Paul Torgerson, Nigel Yoccoz and an anonymous referee greatly helped to improve the manuscript. This investigation was carried out in the context of the Integrated Fox Project, an interdisciplinary research project on the dynamics of the fox population in Switzerland. The study was supported by the Swiss Federal Office of Veterinary Medicine, Berne (Projekt No. 1·99·03) and the Swiss Federal Office for Education and Science, Berne (EchinoRisk Projekt QLK2-CT-2001-001995/BBW No. 00.0586-2).