Descriptive and statistical analyses of rodent density, tick infestation and TBEV seroprevalence in rodents
Rodent density was calculated per session and per species according to the standardised closed population Schnabel method that takes into account multiple marking occasions (Schnabel, 1938). Since the captures were carried out on three consecutive days per session, we considered that the population was closed for each session. Rodent density per season was calculated using the mean of rodent density estimated for the corresponding months.
The tick infestation prevalence for small mammals was defined as the number of small mammals carrying at least one tick divided by the number of small mammals inspected. We calculated tick infestation prevalence per year, season, species and age class (juveniles vs. adults). Exact 95% CIs were calculated on the basis of binomial distribution. We described the infestation of small mammals by larvae and nymphs in season 2 (the period of highest nymph density) of 2012 and 2013 by calculating (1) the mean number of larvae and nymphs per individual infested by ticks, (2) the proportion of small mammals infested by ticks carrying nymphs only or nymphs and larvae, and (3) the proportion of larvae feeding on small mammals that were also carrying nymphs. The results for 2012 and 2013 were compared using a Mann-Whitney Utest.
The tick infestation status of small mammals was modelled using a logistic GLM (generalised linear model) with a binomial distribution and logit link, and a binary response variable (absence of ticks = 0; presence of ticks = 1). The variables considered were the small mammal species, sex, season, year, and the interactions between year and season (to allow seasonal patterns to vary between years) and between season and species (to mimic seasonal variation of small mammal community structure and density). We included the number of captures within the season as an offset term. Some animals were captured in several different seasons and constituted temporal pseudoreplication of individual hosts. We therefore ran logistic GLMs for those individuals with the tick infestation status in season 2 (respectively in season 3) as the response variable, the infestation status in season 1 (respectively season 2), species, sex and year as the explanatory variables and the number of captures in season 2 (respectively season 3) as an offset term. As the infestation status of the previous season had no effect on the infestation status of the following season, we considered that these pseudo-replications would not bias the results of the model.
We calculated the TBEV seroprevalence of small mammals per year, season and species, and its exact 95% CI based on binomial distribution. The seropositive status of captured animals was modelled using a logistic GLM as a function of the small mammal species, season and year. As few individuals were seropositive (see Results), we limited the number of variables included in the model.
Collinearity was checked in the models ensuring a variance inflation factor (VIF) < 10 (James, Written, Hastie, & Tibshirani, 2017). The backward elimination of explanatory variables was used to identify the most parsimonious model with the smallest Akaike information criteria (AIC).
Statistical computations were performed in R 3.5.0 (R Development Core team, 2018). For all statistical tests, a p-value<0.05 was considered statistically significant.
Ethical statement . The experimental protocol with small mammals complied with EU Directive 2010/63/EU and was submitted to and approved by the French Ministry of Research (APAFIS no. 2015120215112678). All efforts were made to minimize animal suffering. The species studied are not protected in France or included in the IUCN Red List of threatened species in France. The animal trapping took place with permission from the landowners.