Winter is coming … How the hoarding behaviour of rodents fosters the persistence of Echinococcusmultilocularis

Abstract

This study aimed to determine whether the interspecific copro-foraging behaviour of potential intermediate host rodents can prevent the summer egg loss of the Arctic-originated cestode parasite, Echinococcus multilocularis. Red fox (Vulpes vulpes) and golden jackal (Canis aureus) scats were used as bait under camera traps to assess rodents’ approach or avoidance activity. Of the 6 detected rodent taxa, common voles (Microtus arvalis) and European wood mice (the morphologically similar species, wood mouse, Apodemus sylvaticus and yellow-necked mouse, A. flavicollis) approached and contacted the baits most frequently, primarily during autumn and winter. A higher seed content in the scat seemed to attract the common voles. Contrary to our hypothesis, autumn hoarding activity appears to foster the transmission of E. multilocularis, while the weakest link in its life cycle, the summer drought, remains less protected.

Graphical abstract

Highlights

• •Fox and jackal scats were used as bait under camera traps.
• •Six rodent species had direct contact with wild canine scat.
• •Wood mice and common vole activity and contacts were highest in fall and winter.
• •Seed-containing scats attract voles.
• •Voles' fall hoarding seems to foster the spread of E. multilocularis.

1. Introduction

Echinococcus multilocularis is a significant zoonotic parasite that causes alveolar echinococcosis (AE) in humans. The annual number of new cases is estimated to be over 18,000 globally, and without proper treatment, the fatality rate can reach 90%. For this reason, AE is considered to be the most serious parasitosis in the Northern Hemisphere (Guo et al., 2025). E. multilocularis is a cestode belonging to the Taeniidae family; it originated in Beringia approximately 2.6 million years ago and spread throughout the Holarctic Realm during the glacial cycles (Massolo et al., 2022). Its life cycle requires a carnivorous definitive host as habitat for adult worms and egg production, and a rodent intermediate host for larval development (Guislain et al., 2008; Raoul et al., 2015).

Although the parasite adapted to temperate zone hosts, such as the red fox (Vulpes vulpes) and common vole (Microtus arvalis) (Guerra et al., 2014; Beerli et al., 2017), its thick-shelled eggs are structured for long-term survival in extreme cold temperatures; conversely, they remain highly vulnerable to heat, aridity, and UV radiation (Veit et al., 1995; Shaikenov, 2006; Raoul et al., 2015).

After the Last Glacial Maximum, the parasite retreated from its former range into cold refugia, such as Alaska, the Tibetan Plateau, and the European Alps, thereby isolating certain populations (Spotin et al., 2018; Guo et al., 2025). This range reduction was probably due to climatic warming and the consequential extinction of its former hosts from its ancestral distribution area (Sommer and Benecke, 2005; Llaberia-Robledillo et al., 2025). Contrary to current global warming, the parasite's range is expanding again, likely due to the population expansion of its main definitive (Vulpes spp.) and intermediate (Microtus spp.) hosts (Cafiero et al., 2025b).

Parasites with complex life cycles utilise multiple hosts to avoid direct environmental effects and maximise their fitness to achieve the highest transmission success. However, in this mode of ontogenesis, the switch between intermediate and definitive hosts is a transmission bottleneck that is mitigated by the evolution of trophic transmission (Lafferty, 1999; Antonovics et al., 2017). A parasite-induced behaviour modification can be observed in prey species. Increased risk-taking predisposes them to predation and thus parasite transmission to the predator definitive host Lafferty (1999); Brown et al. (2001); Joly and Messier (2004); Médoc and Beisel (2011).

In the case of Echinococcus spp., the transmission from the intermediate host to the carnivore definitive host is facilitated by the behaviour-modification effect of the parasite, as larval echinococcosis causes a higher energy demand, a more frequent feeding (Martini et al., 2024), and a slower movement due to the enlarged liver (Joly and Messier, 2004; Vervaeke et al., 2006). However, less is known about the other half of the life cycle. Current theory suggests that intermediate hosts acquire infection accidently while foraging on contaminated vegetation or during grooming (Torgerson, 2016; Umhang et al., 2016; Spotin et al., 2018). In Arctic conditions, this transmission is ensured by the eggs’ cold tolerance; nevertheless, in the temperate zone, the environmental phase of the life cycle is considered the highest risk period (Matsumoto and Yagi, 2008). The dispersal of eggs from faecal material into the environment takes time, thereby increasing the risk of egg destruction primarily during hot summers (Veit et al., 1995; Raoul et al., 2015).

Foraging on faecal materials is a well-known phenomenon in different species. This behaviour is often observed under harsh environmental conditions when faeces provide essential substances for survival. In plateau pica (Ochotona curzoniae), an intermediate host of E. multilocularis on the Tibetan Plateau, a high frequency of allocoprophagy ensures protein supplementation for successful breeding in over-grazed grassland habitats (Wu et al., 2023; Zhao et al., 2024). This behaviour was also observed in a closely related species, the collared pica (O. collaris) (Jung et al., 2026). In red foxes (V. vulpes), the utilisation of dog faeces was recorded during shortages of small mammal prey (Waggershauser et al., 2022), evidencing that energy demand occasionally overrides the predator-avoidance behaviour. For small rodents, balancing between wariness in the “landscape of fear” and seeking food in the “energy landscape” forms the behaviour of the individual (Gallagher et al., 2017; Merz and Mortelliti, 2025). Studies on predator avoidance in rodents revealed that demand for energy or water often outweighs the animals' usual caution (Staples, 2010; Merz and Mortelliti, 2025), especially in familiar environments in the proximity of the burrow system of their colony (Stryjek et al., 2018).

Our study is inspired by the research of Sasaki et al. (2023), who investigated the behaviour of potential intermediate hosts of E. multilocularis in the presence of fox faeces to assess transmission risk within the parasite's Japanese range. We hypothesize that purely chance-based transmission from definitive to intermediate hosts would impede the expansion of the parasite's range in climatic zones with hot and dry summers, contrary to current trends. Therefore, we tested whether certain faecal components could modify the predator-avoidance behaviour of rodents, resulting in a shortened environmental phase in the life cycle and thus an increased chance of transmission.

2. Material and methods

2.1. Study site

The investigation was conducted between September 1, 2023, and October 5, 2024, in an agricultural area located in southwestern Hungary (46.423°–46.425° N, 17.849°–17.850° E). Four camera traps (Cam1, Cam2, Cam3, Cam4) were deployed in a 2-m-wide strip separating two pastures at the following GPS coordinates: Cam1: 46.423248° N, 17.849596° E, Cam2: 46.423398° N, 17.850345° E, Cam3: 46.425207° N, 17.849970° E, Cam4: 46.425109° N, 17.849097° E.

Cam1 and Cam2 were approximately 60 m apart, while Cam3 and Cam4 were also approximately 60 m apart. The Cam1–Cam2 line was located about 210 m south and parallel to the Cam3–Cam4 line. These fences, bordered by 2-m-wide corridors, were selected because they are representative of the local pasture conditions and allowed effective protection of the camera traps from grazing animals.

2.2. Camera trapping

In this study, we used horizontal camera-trap orientation to detect the response of small-sized rodents to wild canid (red fox, Vulpes vulpes, and golden jackal, Canis aureus) scats. Horizontal placement of the cameras was specifically chosen to align with the predominantly ground-level movement patterns of these rodents, enabling us to capture their natural behaviour as they interacted with the scats. The cameras were placed over a grass-covered area. The equipment was fastened to planks about 120 cm above the ground and provided about a 1.2-1.5 m² field of view for detection. We used three camera models: Minox DTC 550, Spypoint Tiny WBF and EVOLVEO SAG CAM-A. We used the same configuration (photo size: 5 MB, auto infrared flash, a 5-s delay between photos) to avoid variability and bias across camera trap models (Meek et al., 2014). The cameras were checked every 2 to 10 days to inspect the equipment, change the scats and download photographs.

2.3. Data collection design

We used two types of carnivore scat during the camera trapping. We collected fox and jackal faeces from the rectums of animals dissected during other studies (Nagy et al., 2025). To avoid a significant alteration in the odour of the scat, the carcasses were stored at 2 ^oC until necropsy, which was carried out as soon as possible after harvest. However, the scat was not mixed with the perianal glands’ secretion, as can occur with naturally passed faeces. As each sample had a characteristic carnivorous odour, we presumed that this mild difference from natural faeces had no impact on findings. After collection, we labelled them individually and stored them at −80 °C for two weeks for biosafety reasons, then at −20 °C until their use. The samples were thawed at room temperature for 4 h before baiting. We recollected the used samples from under the cameras and stored them at −20 °C until further examination.

During the scat component analysis, we first dried the frozen samples at 65 °C for 48 h until their weight remained constant. We then weighed each sample individually and determined its composition. For the selection, we used a 40× magnifier. We grouped the components into the following categories: vertebrate, insect, grass, fruit, and seed. After separation, we weighed the five component categories separately. In all cases, measurements were made with an accuracy of 0.01 g. In the statistical analysis, we used the weight proportion of each category using the following equation:

$$P_{\left(ij\right)}=c_{\left(ij\right)}/td{w}_j$$

where,

P_ij = proportion of the ith component in the jth sample,

c_ij = weight of the ith component in the jth sample,

tdw_j = the total dry weight of the jth sample.

These proportions served as explanatory variables in the statistical analysis (vertebrate - VERP, insect - INSP, grass - GRAP, fruit - FRUP, and seed - SEEP.

We used the cameras in pairs, alternating between them. During the periods when Cam1 and Cam2 were in operation, no bait was placed under Cam3 and Cam4 (Fig. 1). The bait was replaced during each check (after 2-10 days). If there was fox scat under a particular camera at first, we replaced it with jackal scat during the replacement. When the fox scat and jackal scat periods for the cameras in use ended, the bait was placed under the other camera pair, where two observation periods (fox scat and jackal scat) took place, as well. Once these were over, we returned to the other pair of cameras. With this double rotation, we tried to prevent the rodents from getting used to the constant presence of bait. The scats were placed on a white plastic tray measuring 13 cm in diameter, which was securely fastened to the ground. This arrangement ensured that the baits, the animals approaching them, and the distance between them could be clearly assessed in the images.

Fig. 1.

The study site in Hungary. The four cameras were used in fixed locations during the inveAstigation. (Note: The yellow stripes indicate the area where the cameras were placed, and the blue crosses and Cam1,Cam2, Cam3 and Cam4 mark the cameras themselves.)

2.4. Identification of terrestrial small mammal species of the orders Rodentia and Eulipotyphla

Rodent and other small mammal, such as shrew, species were identified based on camera trap images using external morphological characteristics visible on the photographs, including body size, tail length, ear size, fur colouration, and overall body proportions, following standard small mammal identification guides (Kovács, 2002; Aulagnier, 2025), while the applicability of camera trapping for species-level identification was supported by previous studies (Clucas and McCluskey, 2025).

2.5. Measured indices

In the first part of the study, we treated the number of small mammal images as an event. We accepted a photograph as a separate event if at least 20 min had elapsed between two photographs. If the image contained more than one animal of the same species, it was considered a single event (Debata and Swain, 2018). Based on the counted events (photos), we determined the relative abundance index (RAI) per camera per night for the observed rodent taxa. To calculate the RAI value for each species, we multiplied the sum of all observations detected by all camera traps on each day by 100, then divided by the total number of nights of camera trapping (Debata and Swain, 2018).

In the second phase of the analysis, we examined the characteristics of the events (photographs). For this reason, we measured the distance between the rodents visible in the images and the scats placed as bait. We classified each event into hierarchical distance categories: far (if the animal was more than 20 cm away from the bait); near (if the animal was within 1-20 cm of the bait); and direct physical contact (if the animal touched the bait) (Fig. 2). If an event contained multiple images, the classification was based on the shortest distance visible across them.

Fig. 2.

Placing the bait under the camera trap. The distance between the rodents and the bait was assessed relative to the 13 cm diameter white bait station. Approach was categorised “far”, if the animal was more than 20 cm away from the bait; “near”, if the animal was within 1-20 cm of the bait; and “contact”, if the animal touched the bait. A: The yellow line and characters indicate the “far” category, the red line and characters indicate the “near” category. B: An instance of “contact” is shown.

2.6. Data analysis

During the analysis, we used negative binomial regression models given the nature of the data (with signs of overdispersion). In the first part of the analysis, we compared the occurrence of events in the observed taxa. After quantifying the data, we used the most frequently photographed species as the reference category when building the model. To account for differences in sampling effort across periods, we used the number of trapping nights as a compensation factor. This ensured that the response variable (event) was modelled as a ratio (events/trapping night). In the rest of the study, we separately analysed only the data for the two most frequently photographed species.

In the second phase of the analytical process, we examined the spatial fine structure of rodents' responses to bait using distance analysis. We analysed the three distance categories (FAR, NEAR, and CONTACT) using separate models. Because the five components are proportions of a single composition, their sum is (approximately) 1. Therefore, the parts are linearly dependent, which means four of the proportions fully determine the fifth (FRUP) (Aitchison, 1982). Including all components simultaneously would therefore induce (near-) perfect multicollinearity. This makes the full model unstable and its coefficients poorly identified (Aitchison, 1982; Pawlowsky-Glahn et al., 2015). For this reason, we omitted one component, FRUP, from the model. Thus, FRUP serves as a baseline rather than a categorical reference level. Because the proportions sum to 1, an increase in any component must be offset by a decrease in another. Each coefficient (for VERP, INSP, GRAP, SEEP) thus reflects the effect of a higher proportion of that component balanced by a correspondingly lower proportion of FRUP. The night sampling period (night) was included as a logarithmic offset term.

In the next step, we only analysed contact occurrences. For this reason, the proportions of scat components and the total number of detected events served as explanatory variables in the models. As in the distance gradient analysis, we also chose FRUP as the baseline component. To correct for sampling bias due to differences in nighttime activity duration, we added the logarithm of the number of nights as an offset term to the model, so that the results refer to contact density (number of interactions per unit time). The model estimates were performed using a log-link function. To interpret the results, we calculated average marginal effects for any significant variables in our models, which show the change in the predicted number of contacts for each bait component while holding the other variables constant.

Finally, we compared the foraging strategies and degrees of specialisation of the two dominant rodent species, and we calculated Rényi entropy profiles. The analysis was based on the distribution of choices between the three scat components. To this end, we used all events recorded by the cameras, i.e., distant observations, close interest, and direct physical contact. By varying the scale parameter (α) of the Rényi entropy (H_α) (from 0 to infinity), we determined the choice diversity of the rodent species. Lower α values are sensitive to rare choices, while higher values are sensitive to dominant preferences. Thus, higher entropy values indicate a generalist foraging strategy, while steeply declining profiles indicate strong specialisation in the choice pattern of a given species (Ricotta and Avena, 2001; Chao and Jost, 2015). To characterise and compare intra-species bait selection strategies, Rényi diversity profiles were generated for the two dominant rodent taxa. These profiles provide a comprehensive view of specialisation by scaling the importance of rare versus dominant preferences through the α parameter. Statistical differences between the two species were tested at three distinct levels of the selection pattern. First, a Fisher's exact test of independence was performed to compare contact frequencies and determine whether overall bait-choice patterns differed significantly between species. Second, to test for differences in strategic complexity (uncertainty), a Hutcheson t-test was performed at the Shannon entropy point (α = 1). Third, to evaluate differences in specialisation intensity, the Berger-Parker dominance index (d) was calculated, representing the proportion of the most preferred resource. Due to the categorical nature of the experimental setup at individual sites, population-level dominance was compared using a Wilcoxon rank-sum test, with bootstrap resampling (1000 iterations) to account for sampling variance.

Statistical analyses and visualisations were performed in R (v4.4.2). Negative binomial models were fitted using the MASS package (Venables and Ripley, 2002). Predicted values and confidence intervals for marginal effects and distance gradients were estimated using the ggeffects package (Lüdecke, 2018). Rényi entropy profiles and related diversity indices were calculated using the vegan package (Oksanen et al., 2025). For plot visualisations, we applied the ggplot2 package (Wickham, 2016).

3. Results

During the 271 trap nights of the study, a total of 600 events of six rodent taxa were recorded. Species-level identification was possible for striped field mouse (Apodemus agrarius), common vole – CVOLE (Microtus arvalis), European water vole - WVOLE (Arvicola amphibius), and brown rat - BRAT (Rattus norvegicus). Individuals belonging to the genus Mus (MUS) and family Soricidae (SHREW) were identified only at genus levels, because morphological differences between these species could not be consistently distinguished from camera-trap photographs alone (Kovács, 2002; Aulagnier, 2025). Two Apodemus species, the wood mouse (Apodemus sylvaticus) and the yellow-necked mouse (A. flavicollis), cannot be identified using camera footage (Bartolommei et al., 2016); therefore, these were recorded as European wood mice (EWM). Since only one event was recorded for the striped field mouse, this species was excluded from further analysis. Detailed raw data are available in the Zenodo repository (https://zenodo.org/records/19511447).

Of all rodents, EWM and CVOLE had the highest RAI values, while SHREW had the lowest. A comparison of event counts supported this observation. EWM was the only taxon with a significantly higher number of events than the other species (Table 1).

Table 1.

The relative abundance index (RAI) and the results of comparison between the reference taxa (European wood mice: wood-mouse and yellow-necked mouse – EWM) and the other observed camera-trapped rodent taxa. (Note: common vole – CVOLE, genus Mus – MUS, brown rat – BRAT, European water vole – WVOLE, family Soricidae – SHREW; RAI – relative abundance index, β – regression coefficient).

taxon	RAI	β	p-value
EWM	103.09	reference category in the model
CVOLE	50.81	−0.84	0.038
MUS	29.09	−1.19	0.004
BRAT	17.67	−1.89	<0.001
WVOLE	15.10	−2.28	<0.001
SHREW	4.79	−3.2	<0.001

For the annual pattern of the rodents’ behaviour in the presence of predator scat, a peak occurred during autumn and winter, while the summer months passed relatively uneventfully (Fig. 3).

Fig. 3.

Seasonal trends in events and contacts during the study period for the common vole (CVOLE) and wood mouse (EWM) species. (∗The second autumn period consisted of only September 2024.)

The distance from the bait showed very similar results in the analysis of photographs of the two dominant taxa (CVOLE and EWM). The occurrence of far, near, and contact events identified in the distance gradient analysis was not influenced by the amount of specific components in the scat (Table 2).

Table 2.

Results of the distance gradient analysis models. (Note: common vole – CVOLE, European wood mice: wood-mouse and yellow-necked mouse – EWM, VERP – proportion of vertebrate-derived components in scat, INSP – proportion of insect-derived components in scat, GRAP – proportion of grass-derived components in scat, SEEP – proportion of seed-derived components in scat, β = regression coefficient with 95% confidence interval in square brackets).

FAR model	CVOLE		EWM
	β [CI95%]	p-value	β [CI95%]	p-value
intercept	−1.80 [−2.81, −0.80]	<0.001	−1.36 [−2.31, −0.42]	0.005
VERP	−0.84 [−2.65, 0.97]	0.361	−0.63 [−2.24, 0.98]	0.444
INSP	−9.28 [−27.07, 8.51]	0.307	−5.26 [−17.10, 6.58]	0.384
GRAP	1.41 [−0.68, 3.50]	0.187	1.29 [−0.74, 3.31]	0.213
SEEP	0.90 [−1.29, 3.11]	0.420	0.85 [−1.27, 2.98]	0.432

NEAR model	CVOLE		EWM
	β [CI95%]	p-value	β [CI95%]	p-value
intercept	−1.21 [−2.07, −0.36]	0.005	−0.45 [−1.12, 0.22]	0.192
VERP	−1.03 [−2.54, 0.49]	0.183	0.10 [−0.98, 1.17]	0.862
INSP	−1.25 [−10.36, 7.86]	0.788	−2.37 [−9.65, 4.90]	0.523
GRAP	1.02 [−0.82, 2.86]	0.278	0.14 [−1.38, 1.65]	0.860
SEEP	0.47 [−1.47, 2.41]	0.636	0.44 [−1.11, 1.98]	0.580

CONTACT model	CVOLE		EWM
	β [CI95%]	p-value	β [CI95%]	p-value
intercept	−2.62 [−3.65, −1.61]	<0.001	−0.90 [−1.63, −0.16]	0.017
VERP	−0.47 [−2.27, 1.33]	0.609	−0.44 [−1.67, 0.79]	0.485
INSP	0.25 [−10.25, 10.74]	0.963	−0.62 [−8.33, 7.08]	0.874
GRAP	1.04 [−0.97, 3.05]	0.309	0.10 [−1.57, 1.77]	0.910
SEEP	1.37 [−0.64, 3.38]	0.180	0.09 [−1.63, 1.80]	0.921

The contact models revealed clear differences in the two rodents' contact analysis. We observed that common vole's contacts were positively affected by the number of events (β = 0.08, incidence rate ratio (IRR) = 1.09, 95% CI of IRR = 1.05 – 1.14, p < 0.001). Furthermore, the number of contacts was also positively associated with the proportion of seed-derived components (SEEP) in scats (β = 1.06, IRR = 2.9, 95% CI of IRR = 1.03–8.19, p = 0.041) (Fig. 4).

Fig. 4.

Negative binomial model predicted effects of seed content (SEEP, A) and number of events (B) on the common vole's contact. The dark orange line indicates the average marginal effects, and its light orange ribbon shows the 95% confidence interval.

In the case of the EWM, both explanatory variables proved significant. The remaining components (VERP, INSP, and GRAP) showed no significant association with contact numbers (all p > 0.05; see Table 2). The number of visits also showed a positive association with contacts (β = 0.11, IRR = 1.12, 95% CI of IRR = 1.09–1.15, p < 0.001). Unlike CVOLE, the SEEP showed a negative correlation with contact numbers (β = −0.84, IRR = 0.43, 95% CI of IRR = 0.198 – 0.89, p = 0.025) (Fig. 5).

Fig. 5.

Negative binomial model predicted the effects of seed content (SEEP, A) and the number of events (B) on the European wood mice's (wood mouse and yellow-necked mouse) contact. The dark blue line demonstrates the average interaction effect, and its light grey ribbon shows the 95% confidence interval.

For common voles (CVOLE), we recorded 26 contacts. These were distributed as follows: 9 (34.62%) occurred with seeds (SEEP), 9 (34.62%) with fruit (FRUP), 5 (19.23%) with grass (GRAP), and 3 (11.54%) with vertebrate remains (VERP). No contacts were recorded with arthropod-rich (INSP) components. In the case of wood mice (EWM), the total number of contacts was considerably higher (n = 87). The animals most frequently came into direct contact with fruit-rich (FRUP) baits (n = 41, 47.13%), followed by those containing seeds (SEEP: n = 29, 33.33%) and vertebrate remains (VERP: n = 17, 19.54%). No contacts with grass (GRAP) or arthropod (INSP) components were observed for EWM.

The Rényi entropy profiles revealed different selection strategies for the two species. Although both curves showed a similar decline, the EWM curve was lower than the CVOLE curve at higher α values (Fig. 6), indicating a stronger dominance pattern in EWM. The qualitative composition of bait choices did not differ significantly between the two species (Fisher's exact test, p = 0.241). The CVOLE showed a balanced preference for seeds (SEEP) and fruit (FRUP), each accounting for 34.6% of its total contacts, while EWM primarily targeted the fruit component (47.1%).

Fig. 6.

Rényi diversity profiles of bait selection strategies for common vole (CVOLE) and European wood mice (Wood-mouse and yellow-necked mouse, EWM). The curves represent the continuum of diversity indices from species richness (α = 0) to dominance (α → ∞).

The Hutcheson t-test revealed no significant difference in Shannon diversity between the species (p = 0.746). However, the Berger-Parker index indicated a divergence in dominance patterns. The EWM population showed a marginally higher dominance level (p = 0.054) compared to the CVOLE.

4. Discussion

E. multilocularis possesses cold preference due to its Arctic legacy; in contrast, it cannot tolerate heat and aridity during the environmental phase of its life cycle. In Central Europe, the maximum survival time between −15 °C and 27 °C is 240 days, while extremely high temperatures over 40 °C or exposure to 25% relative humidity shorten this period to hours (Veit et al., 1995). Despite these facts, the parasite spreads unimpeded throughout Europe, even in the Mediterranean climatic zone. Although humid and cool conditions of wet habitats support the survival of eggs in the environment (Shaikenov, 2006; Moloi et al., 2023), the current success in range extension toward warm climatic zones, such as the Balkans (Balen Topić et al., 2023) and Italy (Cafiero et al., 2025a), suggests a stronger protection in the parasite's transmission cycle.

This study was based on the hypothesis that rodents can come into direct contact with definitive hosts' faeces, hence shortening the environmental phase of the parasitic life cycle. Contrary to our preliminary assumption, the highest contact rate was experienced on humid and cool days, primarily during the peak activity period of rodents between August and December (Esther et al., 2014; Jacob et al., 2014; van Rosmalen et al., 2022). Summer drought days were the least busy periods regarding both rodent activity and contact rate. Investigating the potential effect of faecal compositions, seed content showed a significant influence on the behaviour of both common vole and wood mice species; however, the influence of seed content manifested differently in the two species. While voles were attracted by seeds, mice seemed to avoid them. This apparent avoidance resulted from the statistical approach that the influence of all faecal components was compared to fruit content. Therefore, the increased attraction of mice to the fruit content misleadingly implied that seed content deterred them. Rényi's diversity analysis of the two dominant rodent taxa (CVOLE, EWM) also revealed that the CVOLE is a seed-specialist, while the diet of EWM is more diverse. This finding partly aligns with other studies on rodent food choice, as most studies demonstrate a dominant preference for seed across all investigated species (Fischer and Türke, 2016; Mühlböck et al., 2024).

The finding that most faecal contacts happened during fall and early winter contradicts our preliminary hypothesis that rodents' food-seeking activity during summer drought remarkably increases the transmission rate among the hardly tolerable climatic conditions. Conversely, the rodents' behaviour appeared to support the parasite's survival rate in those periods, which are ab ovo suitable for long-term environmental survival of eggs. Therefore, after July and early August with highly adverse climatic conditions, the period between late August and December was ecologically over-insured, providing a good survival rate in the environment and an extremely high rate of potential infection acquisition.

However, the increased risk of parasite acquisition during the autumn results in an annual peak of infective protoscoleces in voles between mid-winter and early spring (Burlet et al., 2011; Beerli et al., 2017). This period of the year is, in parallel with the breeding season of foxes (Farstad, 1998), a period characterised by territorial fights and high levels of the animals' blood cortisol (Gese et al., 2023). Besides social stress, the trade-off between reproduction and the immune system also contributes to decreased resistance to parasitic infection (Martin et al., 2008). For females, this dip in immune function lasts until weaning of the pups. During summer, the pups also begin to hunt for small prey, such as rodents. Due to the parasite-induced behaviour, like more frequent feeding (Martini et al., 2024) and slower movement (Joly and Messier, 2004; Matsumoto and Yagi, 2008), infected rodents are easy prey for less skilled hunting pups. In parallel, juvenile foxes are parasitised with Toxocara spp. nematodes. In contrast with the small E. multilocularis, these larger parasites cause remarkable nutrient shortage (Saichenko et al., 2021), elevated inflammatory markers (Zheng et al., 2019) and alteration in the intestinal microbiome (Sieng et al., 2023) in the host. As a result, roundworm infection forces young foxes to seek food continually, utilising any available energy sources. These factors contribute to the highly abundant E. multilocularis infection in juvenile foxes in specific European populations (Otero-Abad et al., 2017; Citterio et al., 2021).

From mid-July, the vixens with post-weaning poor body condition and the juveniles begin to shed a large amount of parasite eggs into the environment. However, camera footage suggested that this excessive parasite load could not reach the intermediate hosts directly. In the middle of the summer drought, both activity and contact rate of the rodents remained very low. This finding partly aligns with previous observations regarding the seasonal activity of rodents. Although primarily spring is determined as the main activity season of rodents (Borowski, 2002; Eccard and Herde, 2013), novel research revealed that food hoarding increases the activity of certain individuals during autumn (Zhang et al., 2024). In addition, the well-known population peak in autumn (Esther et al., 2014; Jacob et al., 2014; van Rosmalen et al., 2022) might also contribute to seemingly heightened activity.

Although summer time possesses the highest risk of egg loss, the parasite's life cycle appears to rely on chance encounters between intermediate hosts and parasite eggs. The only direct risk-mitigating factor seems to be a specific behaviour of foxes, as they tend to scat-mark the core of their territories, such as den areas and feeding sites with high prey densities (Dell'Arte and Leonardi, 2020). Consequently, the boundaries of extensively used burrow systems of vole colonies are highly contaminated with fox faeces (Robardet et al., 2011). In rural habitats, these spots can be found in field margins (Fukui et al., 2024) and in permanent grassland ecosystems (Viel et al., 1999; Giraudoux et al., 2013, 2019); while within human settlements, high vole (Microtus spp.) densities are associated with wastelands (Robardet et al., 2011). This parallel food source of anthropogenic subsidies and an abundant rodent community facilitates the formation of dense fox societies (Porteus et al., 2024). Although the excess of anthropogenic food reduces rodent biomass in the carnivore diet, the dominance of voles, particularly Microtus spp., means that the risk of a rodent-borne infection remains high even in synanthropic habitats (Raoul et al., 2010). Therefore, synanthropic microfocus formation of E. multilocularis poses a high public health risk, whereas wild carnivores compete with free-roaming domestic dogs for potential prey, thereby fostering a peri-domestic reservoir of the parasite (Raoul et al., 2015). Consequently, microhabitats that are markedly suitable for the maintenance of large vole populations can contribute to the trophic transmission of E. multilocularis between its intermediate and definitive hosts (Giraudoux et al., 2019; Raoul et al., 2015; Robardet et al., 2011).

The maintenance of the parasitic cycle is also facilitated by the risk-taking behaviour of certain rodent individuals who do not avoid carnivore scat, as this study has recorded. This attitude increases the vole's likelihood of acquiring resources; however, it also increases the risk of both parasite intake (Kortet et al., 2010; Santicchia et al., 2019) and predation (Borowski, 2002; Eccard and Herde, 2013; Kortet et al., 2010; Santicchia et al., 2019). Bold behaviour often results in high life-history productivity despite its elevated mortality risk, since bolder individuals mostly exhibit better body condition and immune function (Kortet et al., 2010). However, other findings suggest the opposite, as exploratory behaviour and greater body mass are associated with higher infection intensity due to decreased immunity in older, thus larger, animals (Santicchia et al., 2019).

In E. multilocularis, individual differences in immune reactivity fundamentally determine the disease outcome in intermediate hosts. In humans, at least 20% (Autier et al., 2023) or even 99% (Gottstein et al., 2015) of exposed patients are estimated to remain clinically healthy. In addition, individual variation in faecal egg shedding has been observed in foxes, with 1/3 of infected individuals depositing 2/3 of the egg-contaminated faeces (Da Silva et al., 2024). This documented resilience of both intermediate and definitive hosts of E. multilocularis appears to be a factor undermining maintenance and transmission. However, this self-limitation might keep the parasitism at a level that does not pose an extinction risk to the rodent population of a microfocus. Bolder, large-bodied, resilient voles with slow-growing lesions might contribute to the overwintering of a protoscolex biomass sufficient to maintain the next season's parasite load of the local foxes. While limited egg shedding of the fox population might guarantee a less contaminated environment, in which only the bold, more explorative rodent individuals are at high risk of infection, and a vast majority of the rodent population can serve as a food source for foxes and recruitment to compensate for losses due to parasitism and predation. The continuously increasing vole density (Eccard and Herde, 2013; Esther et al., 2014; Jacob et al., 2014; van Rosmalen et al., 2022) and the parallel parasite load of pup-rearing fox societies during the summer are likely to yield a significant protoscolex biomass by late autumn (Otero-Abad et al., 2017; Citterio et al., 2021).

During winter, common voles do not hibernate; however, their reproduction is rare and depends on specific conditions, such as abundant food reserves and deep snow cover (Jacob et al., 2014). Although winter decimates the rodents' populations, the immune function of individuals in good body condition improves due to the changing photoperiod and consequential elevation of melatonin level (Nelson and Drazen, 2000). In experimental infection, protoscoleces develop in rodents in 10-12 weeks, causing severe, often fatal disease (Woolsey et al., 2015; Martini et al., 2024). Probably, natural infections with fewer numbers of eggs at the highest levels of immune reactivity result in slower disease progression and prolonged parasite carriage. Furthermore, rodents died during wintering are not necessarily lost for sustaining the parasitic life cycle. Decaying tissues can serve as sources of viable protoscoleces for weeks in winter conditions (Ohnishi et al., 1984; Laurimäe et al., 2020).

Although our findings suggest that summer is a period with a negligible risk of parasite transmission from the definitive to the intermediate host, climate-induced egg loss is probably offset by the aforementioned behavioural, immunological, and population-biological factors. In addition, an adverse human activity could also contribute to the epidemiological dynamics. In natural and semi-natural ecosystems, micro-local hotspots of E. multilocularis are often formed as a result of an anthropogenic habitat transformation, like deforestation and agricultural encroachment, which creates not just an altered plant community but also reduces the small mammal species richness (Giraudoux et al., 2013, 2019), weakening the ecosystem's health.

Both naturally occurring and human-induced small-scale changes in the epidemiology of E. multilocularis contribute to large-scale processes through the dispersion of definitive hosts, as has been observed in Europe over the past decades (Simoncini and Massolo, 2024). In the absence of stable, large-carnivore top-down regulation, mesopredators, such as the red fox, adopt territorial behaviour primarily driven by food availability (Cavallini, 1996). Locally abundant food sources contribute to the population expansion of foxes (Porteus et al., 2024), resulting in alterations in social behaviour, dispersal patterns (Iossa et al., 2009), and health status (Gortázar et al., 2003). While low-density populations tend to be rather philopatric, high population density forces a male-biased dispersion during autumn and winter (Iossa et al., 2009; Dorning and Harris, 2019), the period when resource shortage suppresses the foxes' immune function, thus increasing their susceptibility to infection (Wang et al., 2025). Consequently, micro-local hotspots formed by abundant populations of vole prey and its parasite-infected fox predator can facilitate not only the local circulation of E. multilocularis (Raoul et al., 2015) but also its range expansion (Simoncini and Massolo, 2024).

From densely populated fox societies, young foxes disperse to new habitats in autumn, carrying their parasites with them (Citterio et al., 2021). Although more than 30% of the migrating juveniles die before their adulthood (Soulsbury et al., 2008), their potential contribution to the parasite's range expansion is not negligible due to the well-documented high parasite load of young and subadult carnivores (Sugden et al., 2023; Torgerson, 2006; Otero-Abad et al., 2017). The dispersal of young foxes coincides with the peak of rodent food hoarding for winter storage (Zhang et al., 2024; Esther et al., 2014; Jacob et al., 2014; van Rosmalen et al., 2022), creating a hazard for the expansion of the original geographical range of E. multilocularis.

The camera footage analysis did not directly support that the weakest link in the life cycle, the mid-summer, was considerably strengthened by food or humidity-seeking behaviour of potential intermediate hosts, the rodents. Literature on the ecological demands of E. multilocularis agrees that the environmental survival chance of eggs limits the distribution area; furthermore, this factor could impede its release from the historical foci, such as the European Alps, the Tibetan Plateau, and Alaska for thousands of years (Veit et al., 1995; Massolo et al., 2022). Interestingly, the most extensive European range expansion has been taking place since the 1990s, in parallel with the gradually warming and drying climate of the continent (Guerra et al., 2014; Massolo et al., 2022).

In Arctic conditions, the safeguarding of the parasite's overwintering proved to be a successful evolutionary strategy, since Arctic summers are short enough to prevent significant egg loss (Massolo et al., 2022). Among these circumstances, large overwintering protoscolex biomass is the basis of the next year's parasite generations, and thus the guarantee for the survival of the species in cold places. Notably, this winter-focused survival strategy was also observed in Europe (Burlet et al., 2011; Beerli et al., 2017), and it proved to be successful even in sub-Mediterranean (Citterio et al., 2021; Omeragić et al., 2022; Balen Topić et al., 2023; Cafiero et al., 2025b) parts of the continent.

Paradoxically, this survival mechanism seems to leave the weakest link unprotected. On closer inspection, however, the apparent contradiction between cold tolerance and range expansion towards warm climate zones highlighted the trophic ecological aspects of parasite transmission in micro-local hotspots, where an extremely high density of intermediate hosts is utilised as the primary food source of a dense social group of foxes (Raoul et al., 2015), which in turn shows a strong dispersal tendency (Iossa et al., 2009).

Although analyses of camera footage suggest that the summer leaves a gap in disease transmission, our method failed to detect other potential factors that drive microscale dynamics in E. multilocularis. Our study showed that during certain periods of the year, the acquisition of parasite eggs is not accidental but the result of active contact with carnivore scats. Nevertheless, our findings raised further questions that warrant investigation in the future.

5. Conclusion

This study investigated the behaviour of potential intermediate hosts of E. multilocularis in the presence of fox and golden jackal faeces. The aim was to assess the chance of transmission from definitive to intermediate hosts. We hypothesised that during harsh summer drought, rodents give up their wariness due to their single-minded resource seeking, and consequently, they frequently encounter carnivores' faeces as potential sources of water and food. Contrary to our preliminary hypothesis, we found that summer time transmission success was not directly facilitated by the behaviour of potential intermediate hosts; rather, the autumn and winter seem the most important periods of infection acquisition of the intermediate host, in the same way as in the Arctic. However, our findings suggest that during autumn hoarding, the common vole, the most important intermediate host of E. multilocularis in Europe, has a significantly increased preference for carnivore faeces with higher seed content.

Ethics approval

Not applicable.

Consent for publication

Not applicable.

Availability of data and materials

The datasets used and/or analysed during the current study are available in the Zenodo repository (https://zenodo.org/records/19511447).

Funding

Not applicable.

CRediT authorship contribution statement

Eszter Nagy: Conceptualization, Data curation, Writing – original draft, Writing – review & editing. Árpád Bokor: Data curation. Boróka Bárdos: Investigation, Methodology. Izabella Zám: Investigation. Rebeka Ráhel Nagy: Investigation, Writing – original draft. Tamás Tari: Investigation, Supervision, Writing – original draft. Julianna Bokor: Investigation. Sibusiso Moloi: Investigation, Writing – original draft. Bawan Mustafa Abdalrahman: Investigation. Gábor Nagy: Conceptualization, Data curation, Funding acquisition, Investigation, Methodology, Supervision, Writing – original draft. Ágnes Csivincsik: Conceptualization, Writing – original draft, Writing – review & editing.

Declaration of competing interest

The authors declare no competing interests.