Microtus arvalis and Arvicola scherman: Key Players in the Echinococcus multilocularis Life Cycle

A broad range of rodent species are described as potential intermediate hosts for Echinococcus multilocularis, a wide-spread zoonotic cestode causing alveolar echinococcosis. However, little is known about the relative contribution of these species for parasite reproduction and the maintenance of its life cycle. In a comparative study in a high endemic region in Zurich, Switzerland, we investigated prevalence rates and fertility of E. multilocularis in the most abundant vole species as well as the predation rate of foxes on these species. To ensure that the fox families had access to different vole species and that these voles were exposed to the same environmental contamination with parasite eggs, we selected eight study plots where at least two rodent species co-occurred. The parasite prevalence in Microtus arvalis [11.0%, confidence intervals (CI) 8.9–13.4] was significantly higher than in Arvicola scherman (5.3%, 3.9–7.1) and Myodes glareolus (3.9%, 2.0–6.7). None of the, only 29 individuals of, Microtus agrestis was infected (0%, 0.0–9.8) and the species was excluded for further analyses. Logistic regression models for the prevalences revealed significant differences between nearby study plots and higher infection rates for females, heavier individuals, and individuals trapped during spring, when the prevalence in M. arvalis peaked up to 65% (CI 50–79) in one plot. Furthermore, we detected significantly higher percentages of fertile infections in M. arvalis and M. glareolus than in A. scherman (OR 11.2 and 6.4, respectively) and a significantly higher protoscolex number in M. glareolus (median 100,000) than in M. arvalis (13,500) and A. scherman (4,290). The most abundant fox prey remains were of the genera Microtus (12.3%, CI 8.4–17.2) and Arvicola (11.5%, 7.7–16.3), whereas Myodes was never recorded as prey (0.0–1.3%). We conclude that M. arvalis and to a lesser extent A. scherman can be regarded as key intermediate hosts in Western and Central European high-endemic regions whereas M. glareolus and M. agrestis play a marginal role. We, therefore, postulate that distribution models of these species could contribute to predict parasite occurrence on a more detailed spatial scale than models of the distribution of foxes which have a very broad and uniform distribution.

inTrODUcTiOn Echinococcus multilocularis is a wide-spread cestode causing human alveolar echinococcosis (AE), a severe disease, with canids (mainly red foxes, Vulpes vulpes) acting as final host (1,2). A wide variety of small mammals are described as intermediate hosts (1,3). Individuals of some murid species were occasionally detected with E. multilocularis infections [e.g., one Mus musculus (4), one Rattus norvegicus (5)], but their role as intermediate hosts can be neglected (1). Regular records of rodent populations with relevant prevalences are only reported from cricetid species, e.g., Arvicola scherman (formerly A. terrestris) and Microtus arvalis in Europe (6,7); Myodes rufocanus in Japan (8,9); Ellobius tancrei and Lasiopodomys brandtii (Brandt's vole, formerly Microtus brandtii) in Central Asia (10); Peromyscus maniculatus and Microtus pennsylvanicus in north central of the United States (11,12). Due to successful rabies vaccination, increased supply of anthropogenic food resources, and changing human-wildlife interactions [e.g., urban tameness (13)], fox populations have substantially increased, especially in the densely populated and urbanized areas in many European regions, e.g., France (14), Switzerland (15), and Germany (16). There is strong evidence that these changes in the population dynamic of foxes led to a marked increase of the incidence of human AE in different regions of continental Europe during the last two decades (17,18).
In parallel to this development, the parasite spread from historically known endemic regions in Central Europe like northern Switzerland, eastern France, southern Germany, and western Austria, over large distances toward the Baltic States (19,20), Scandinavian countries (21,22), and to the west of France (14,23). In Switzerland, the southern border of the parasite distribution corresponds fairly well to the course of the Alpine crest (24). However, some case records of rodent infections south of the Alpine crest in Italy (25) and Switzerland (24,26) demonstrate that the border of the distribution area is not just a result of the harsh climatic conditions in the high Alps. In the Swiss Canton Grison, a correlation between the prevalence in foxes and the predation of foxes on cricetid but not on murid species (26) gave evidence that, beside climatic factors (27), the distribution of suitable intermediate hosts is crucial for the distribution pattern of the parasite. In contrast to many other studies that suggest a geographical spread of E. multilocularis, recent investigations in Ticino, the most southern Canton of Switzerland which is located south of the Alpine crest, revealed a stable border of the distribution of the parasite over the last 20 years. Interestingly, its local distribution and its border of distribution matched the restricted areas where the vole M. arvalis was present (24).
Throughout Switzerland, nine different cricetid species occur, which could potentially act as intermediate hosts. However, only four of these species are wide spread and occur in higher densities (28). Two of the species, i.e., A. scherman and M. arvalis, live in open fields and have been described by different authors as important intermediate hosts in Central Europe (1). The prevalence of the widespread vole M. agrestis, a species living mainly in wetland, meadows, and young forests (29,30), has less been studied. However, a recent investigation demonstrated its high susceptibility to experimental oral inoculations (31). Although no protoscoleces were found 6 weeks p.i. in this study, a recent field study in Sweden confirmed that this species can develop fertile infections (32). Myodes glareolus, the fourth widespread vole in Switzerland, lives mainly in bush lands and forests. It has been regularly described as potential intermediate host in Central Europe, but its epidemiological role for supporting or maintaining the parasite cycle remains unclear (1).
In order to effectively transmit the parasite in the European endemic area, the intermediate hosts have to develop fertile metacestodes with infective protoscoleces (33). In addition to being susceptible to the parasite, only species which share their habitat with foxes and are regularly predated by them can ensure the maintenance of the life cycle (34,35). Therefore, prevalence studies do not suffice to compare the significance of different rodent species for the maintenance of the parasite life cycle. In addition, the direct comparison between different species is hampered by the fact that transmission intensity can greatly vary in space and time (35). Correspondingly, prevalence rates in rodent populations are strongly affected by changing environmental conditions over time and the infective state of the fox individuals in local fox territories.
With our study, we wanted to elucidate the relative importance of the vole species A. scherman, M. arvalis, M. agrestis, and M. glareolus in selected study plots where several of these vole species co-occur simultaneously. This should ensure (A) that the investigated rodents were most likely exposed to the feces of the same fox family groups (and therewith to the same overall level of environmental egg contamination) and (B) that the foxes of one family group could select between the different rodent prey species. To compare the relative importance of the investigated rodent species, we estimated for each species (1) the prevalence of E. multilocularis, (2) the proportion of fertile infections, (3) the asexual parasite reproduction (number of protoscoleces), and (4) the predation frequency by foxes.

MaTerials anD MeThODs study sites
The four study sites were located within and near the community of Zurich, which is situated in the Swiss midlands within a hilly landscape dominated by a mosaic of pastures, meadows, arable lands, and woodland and is characterized by a temperate climate (Köppen-Geiger climate classification Cfb, warm temperate, fully humid, warm summers) (36). Two study sites were situated along the periphery of the city of Zurich and two in rural settings in a distance of roughly 2-4 km from the border of the community (Figure 1). The altitude of the study sites is 400-600 m above sea level.

rodents Trapping and analysis
Arvicola scherman and M. arvalis were trapped in meadows and pastures which is their preferred habitat. M. agrestis and M. glareolus which live in habitats with more cover, were trapped in field verges, scrublands, and forests in the direct neighborhood  of the trapping fields for the former two species. Within each study site, one to three trapping plots were selected, which consisted of interconnected meadows, pastures, field verges, scrubland, and with adjoining forests. The borders of these plots were defined by a polygon that contained all trapped rodents. Thereby, M. agrestis and M. glareolus, which live in more covered habitats, were always attributed to the same plot as the nearest field with A. scherman and/or M. arvalis. The size of the eight study plots varied between 2 and 23 ha, and it was assumed that within a study site the same fox individuals had access to the different rodent species (home range sizes of resident foxes in Zurich according to Gloor (37): mean MCPs of 29 ha for females and 31 ha for males). which were baited with cereals (bird food), apples, and straw. These traps were set for two consecutive nights and checked always early in the morning, at noon, and late in the evening. All unintentionally trapped small mammals were released. M. agrestis and M. glareolus, which were trapped alive, were euthanized by intraperitoneal injection of T61 ® (Embutramid, Hoechst Veterinär, Unterschleißheim) after sedation. For each animal, coordinates, time, and date of capture were collected. All animals were stored at −20°C until further investigation.
Necropsy was carried out under a safety hood. Data on body length, weight (without abdominal organs), and sex were taken. The animals were categorized as reproducing or non-reproducing according to the development of the ductus deferens of the testes in males and placental scars or embryos in the uterus of females. Rodent species were determined according to Brohmer (38). M. arvalis and M. agrestis were distinguished by tooth examination (39). Livers were macroscopically examined for lesions. Suspicious lesions were isolated and investigated for protoscoleces. The number of protoscoleces was determined according to Stieger et al. (40). To compare the number of protoscoleces per individual between the investigated species, the records of fertile infections of the present study were complemented with corresponding data of two previous studies from Zurich (40,41). All liver lesions without protoscoleces were collected for PCR analysis which was carried out according to Ref.

Feces sampling and analysis
During the trapping periods, all fox feces encountered in the trapping fields were collected and GPS coordinates were taken. After 5 days freezing at −80°C for safety reason, feces were stored at −20°C until further analysis. Hairs were collected while sieving 2 g of the feces for taeniid egg detection (43) and also from additional 2-5 g of the remaining fecal material. After washing and drying, hairs were analyzed for rodent prey determination by microscopic investigation after Teerink (44). For hair identification, we prepared prints of the cuticle and medulla on gelatin that had been spread as a thin layer on a microscope slide, and cut cross-sections using blades. Based on these investigations we could differentiate between the three vole genera, Arvicola, Microtus, and Myodes, and the murid genera Apodemus. However, the co-occurring species of the genera Microtus (i.e., M. arvalis and M. agrestis) and Apodemus (i.e., A. sylvaticus and A. flavicollis) could not be distinguished.

statistical analysis
Prevalence rates of E. multilocularis and frequencies of infections with protoscoleces in rodents were analyzed using logistic regressions with the SPSS 22.0 (IBM) statistical software program. We chose six independent variables as possible factors for affecting prevalence rates and the proportion of fertile infections: SPECIES, REPRODUCTION, SEX, SEASON, PLOT, and WEIGHT. Weight was measured without abdominal organs and was used as a proxy of age. The comparison between the species was the main purpose of our investigations. Therefore, we used the data of a trapping plot during a specific season for the logistic regression only when a minimum of eight individuals of at least two species were available for this season. Akaike's information criterion (AIC, 76) was calculated based on the K value (degree of freedom + 2) and the −log likelihood, corrected for small sample size (AICc). The ΔAICc for all variations of the six variables was determined in order to define the best model with minimum influence between the variables. Models were only included in the selection procedure if all included variables significantly affected the model fit.
The non-parametric Kruskal-Wallis test for independent samples was used to compare the number of protoscoleces in different species. Predation rates on different rodent species were compared by calculating the exact binomial 95% confidence intervals (CI) for means of binomial variables, according to the method of Clopper and Pearson (45).

Prevalence rates
In this study, 1,918 voles were trapped and dissected, and liver lesions were detected in 430 individuals. In total, 49 of these individuals had fertile E. multilocularis infections with fully developed protoscoleces. The remaining 381 lesions were analyzed by PCR, whereof 91 samples were positive for E. multilocularis.
On a species level, the overall prevalence rates were 5.3% (95% CI 3.9-7.1, N = 810) in A. scherman and therewith less than half as in M. arvalis (11.0% CI 8.9-13.4, N = 773), whereas the prevalence of M. glareolus (3.9% CI 2.0-6.7, N = 306) was on a similar level as A. scherman. Although strong efforts have been taken to trap comparable numbers of M. agrestis, only 29 individuals were available for our dissections, and none of them was infected with E. multilocularis (0% CI 0-9.8, N = 29). We, therefore, focused in our further comparative analysis only on the other three species.
For building the multiple logistic regression models to explain prevalence rates and the frequency of fertile infections, data records of a given trapping plot and a given season were excluded if not at least data records of eight individuals per species for at least two species were available (see Materials and Methods). A total of 1,695 data records fulfilled this criterion. The best model to explain the prevalence of E. multilocularis contained all considered independent factors except the variable REPRODUCTION ( Table 1). The model confirmed the results of the univariate comparison between the three species: The infection frequency for M. arvalis was significantly higher whereas the lower infection frequency of M. glareolus did not differ significantly from A. scherman (OR 2.69 and 0.57, respectively; Table 1). These differences could also be statistically validated within a single trapping field. Thus, during spring 2014, we recorded a prevalence of 47.4% (CI 24.4-71.1) for M. arvalis and 3.5% (CI 0.4-12.1) for A. scherman within the same pasture of trapping plot 12, and a similar trend was found in the neighboring plot 11 (Figures 2A,B).
In spring season, an increased percentage of infected rodents was observed which peaked up to 65% for M. arvalis on trapping plot 11 during spring 2014 ( Figure 2B). Strong differences in prevalence rates were found between the trapping plots, with significant differences even between trapping fields in immediate neighborhood ( Table 1). For example, the trapping fields for M. arvalis and A. scherman in plots 31 and 32 are separated by a small forest and lie in a distance of less than 300 m to each other (Figures 2,C). In spite of the short distance, we recorded much higher prevalences for M. arvalis

Parasite Fertility and Protoscolex burden
When analyzing the factors affecting whether an E. multilocularis-infected rodent had a non-fertile or fertile infection, only two of the six considered factors entered the best model. We detected a significantly higher probability for fertile infections in M. arvalis and M. glareolus than in A. scherman (OR 11.2 and 6.4, respectively; see Table 1). Non-fertile infected animals were more likely in lower weight classes than infected ones with fertile

Predation of Foxes
In total 234 fecal samples of foxes, collected in the seven trapping plots where E. multilocularis-infected rodents have been found, were analyzed for the presence of different rodent prey species.
Hairs of prey species could be found in 66 samples, wherefrom 63 were identified as rodent hairs. As only 4-7 g of each feces was used to isolate hairs, only few hairs were available per sample. This can explain why we never detected more than one rodent genus within the same sample. The most abundant rodent prey remains were of the genera Microtus and Arvicola with 29 and 27 records, respectively [12.3 (CI 8.4-17.2) and 11.5% (CI 7.7-16.3)]. Arvicola was identified in six and Microtus in five of the seven investigated study plots (Figure 4). The genus Apodemus and non-identifiably rodent hairs were recorded only in four and three samples corresponding to 1.7 (CI 0.5-4.3) and 1.3% (CI 0.3-3.7). Interestingly, M. glareolus was never recorded as prey species (CI 0.0-1.3%), although the species has been trapped by us regularly in six study plots (Figure 4).

DiscUssiOn
The emergence of AE across large regions of Europe has been associated with the increasing fox populations after the successful control of rabies in many European countries and an increased supply of anthropogenic food resources (1,15,18). However, the role and importance of different intermediate host species is under debate (1,35,46).
However, also other European cricetid species have been occasionally found infected like the sibling vole Microtus levis formerly M. rossiaemeridionalis in Svalbard [e.g., 18 (56) and one infected M. agrestis (5) and in Sweden few infected individuals of M. agrestis and A. amphibius  (40,41). Data shown in Table S1 in Supplementary Material. (32) were found, it is very unlikely that these two species play a significant role in highly endemic regions. The same applies for C. nivalis as this species has a much more restricted distribution in Western Europe than E. multilocularis. However, it is known that M. agrestis is susceptible to experimental infections (31), and it replaces M. arvalis in the Scandinavian countries where the species expands more to open habitats than in regions where it co-occurs with M. arvalis (57). Therefore, it is likely that this species together with A. amphibius maintains the life cycle at least at a low level in Scandinavian countries where surprisingly so far no infected M. glareolus have been detected (58). Although many studies have investigated different murid species for E. multilocularis infections (5,40,49,51), to our knowledge there are so far only two confirmed cases of infected murid rodents in Western Europe, namely a M. musculus which was trapped in the cellar of an inhabited house in a small village in the French Auvergne (59) and a R. norvegicus with small, nonfertile lesions (5). Experimental studies have confirmed the very high resistance of laboratory rats (60) to inoculations with high numbers of E. multilocularis eggs, but elucidated that this resistance can be reduced with immunosuppressive interventions resulting in active infections (61). Therefore, this single case and few cases from Japan cannot be regarded as indicators for the intermediate host competence of R. norvegicus. Also other non-cricetid rodent species have been reported in Europe, as the introduced Nutria Myocastor coypus (53,54) and the Eurasian beaver Castor fiber (62,63), which both can harbor fertile infections (53,62). Considering their potential to disperse over large distances and their longevity (64)(65)(66), these two species could occasionally be of some importance in the spread and the persistence of the parasite. However, both species live in low densities and are only occasionally reported in the fox diet [e.g., Ref. (67,68)].
Based on a systematic review of epidemiological studies, Oksanen et al. (46) confirmed arvicolids (including the genera Microtus, Arvicola, and Myodes) and muskrats as important intermediate hosts for E. multilocularis in Europe. However, most of the studies included in this review did not consider to which extent the different rodent species were preyed on by final hosts. As muskrats are not a frequent prey of foxes, this species is regarded by other authors more as a bioindicator for the presence of the parasite rather than a key intermediate host (51,54). Only in special cases, e.g., if trapped animals of control programs are left on river banks, the infective cadavers would be available in large numbers to foxes and boost the infection pressure (51).
By reviewing the existing literature, the three cricetid species M. arvalis, A. scherman, and M. glareolus can be regarded as the most important candidates for maintaining the parasite life cycle over large parts of its distribution area in Western Europe.  All three species are widespread, can reach high population densities (28,69), frequently co-occur in high endemic regions of Western Europe and have been regularly reported with fertile infections in the wild. In this study, we ensured by trapping different species in the same plots and during the same periods that the same fox families had access to the different rodent species. Thus, we can assume a similar overall exposition of the rodents to E. multilocularis eggs. Furthermore, foxes had the choice on which of these species the preferably prey. We also included M. agrestis in this comparative study, as this species is another common cricetid species in the Swiss midlands. However, despite of the huge trapping efforts we could catch only 29 individuals, and that none of these animals were infected. This gives evidence that this species can be neglected for maintaining the parasite life cycle in the high endemic region of the Swiss Midlands. Comparing M. arvalis, A. scherman, and M. glareolus we found significant differences among the species on several levels. Interestingly, M. arvalis had-to our knowledge-the highest prevalence ever recorded which was considerably higher than the prevalence rates for A. scherman and M. glareolus. The eminent role of this species is also underlined by a recent experimental study which demonstrated the high susceptibility of M. arvalis for fertile E. multilocularis infections (70). Thereby, it has to be considered that M. arvalis, which is known for its short generation time, can reach very high population densities of more than 2,500 individuals per hectare (28), which is higher than the peak values for A. scherman and Myodes [A. scherman: >1,000 ind./ha; M. glareolus: 50-100 ind./ha (28,69)]. Furthermore, M. arvalis are much smaller (20-35 g) than A. scherman [65-130 g (28)]. Therefore, a fox has to feed on several M. arvalis to have an equivalent of nutrition as from one A. scherman. It can also be assumed that M. arvalis is a much easier prey as it uses more superficial channels than A. scherman, which rarely leaves the tunnel system. This assumption is substantiated by the observation that M. arvalis is a preferred prey compared to other rodent species (71). Nevertheless, in our study both species have been detected in the fox feces in similar frequency.
Interestingly, our data indicate that Myodes is the best intermediate host in terms of fertility of the parasite. In experimental studies, its susceptibility to oral experimental infections was lower as compared with Microtus spp., but infected animals developed fertile infections (72). However, although M. glareolus is a widespread and common species, we recorded no predation on it by foxes in our study. This is in contrast to other studies in which foxes were shown to prey on M. glareolus (73,74). The species lives-like M. agrestis-more in covered habitats and thus probably is less susceptible to fox predation (75). Taken together, we conclude that this species plays a minor role in  ] was recorded for this species (41). It is possible that foxes prefer Arvicola when M. arvalis is not available or in very low densities. This is supported by a study of Weber and Aubry (78) in the Swiss Jura mountains where A. scherman was the main prey and recorded for 54.5% of the investigated prey items. Indeed, a sigmoid-like functional response to A. scherman density has already been described for the predation of foxes and the predation rate decreased when the density of M. arvalis increased (35,79). On the other hand, M. arvalis was consumed at a high level even when its density was very low. However, to clarify to which extent such replacement processes buffer the life cycle would need further studies.
In conclusion, our study highlights how differences between rodent species in their susceptibility, exposition to infective eggs, parasite fertility, and predation by foxes affect their relevance for the life cycle of E. multilocularis. Our results provide evidence that M. arvalis and probably to a lesser extent A. scherman distribution models could be good predictors for the distribution and abundance of E. multilocularis in Western Europe. Models on the distribution and abundance of these species, therefore, could allow to model parasite occurrence on a more detailed spatial scale than models on fox distribution, as foxes have a very broad and much more uniform distribution than the different rodent species.

eThics sTaTeMenT
Trapping of animals was performed under the direct supervision of a veterinary specialist, and according to the Swiss law, the guidelines on Animal Welfare and the specific regulations of the Canton of Zurich (permit number 17/2013) by the Veterinary Office and the Ethics Committee of the Canton of Zurich (Kantonales Veterinäramt Zürich, Zollstrasse 20, 8090 Zürich, Switzerland).

aUThOr cOnTribUTiOns
All authors listed have made substantial, direct, and intellectual contributions to the work, and all approved its content for publication.

acKnOWleDgMenTs
The authors would like to thank Kathrin Hirsbrunner for her valuable support in trapping the voles and the landowners who have made their fields available to us. The authors would like to thank the SWILD team, especially Lucretia Deplazes, Anouk Taucher, and Dominique Waldvogel for their precious help in the field work. The authors are grateful to Maria Teresa, Armua-Fernandez, and Francesca Gori for their instruction and help in the lab and to Alex Mathis for revising the manuscript. This work represents the dissertation of the first author Olivia Beerli, veterinarian, at the University of Zürich.

FUnDing
This work was supported by the Swiss Federal Food Safety and Veterinary Office (FSVO), by the EMIDA-ERA NET framework, and is within the scope of the EMIRO project "The significance of rodent communities for the distribution of Echinococcus multilocularis: ecological and experimental investigations" (grant number 1.12.18 EMIDA EMIRO).