Introduction

Ticks are important vectors of pathogens in Western Europe (Dantas-Torres et al. 2012; de la Fuente et al. 2008). These pathogens include microorganisms (e.g., bacteria, viruses or protozoa) transmitted through the bite of infected ticks. Lyme borreliosis (LB) is the most widespread tick-borne disease in Western Europe (Gray 1998; Stanek and Reiter 2011). If untreated, an infection may result in skin, neurological, musculoskeletal, or cardiac complications (Stanek and Reiter 2011; Strle and Stanek 2009). This multi-systemic inflammatory disease is caused by spirochetes from the species complex Borrelia burgdorferi sensu lato. At least six genospecies from this complex are responsible for human diseases: Borrelia afzelii, B. garinii, B. burgdorferi sensu stricto, B. bavariensis, B. spielmanii and B. valaisiana (Stanek and Reiter 2011; Strle and Stanek 2009). In Belgium, the prevalence of B. burgdorferi s.l. in questing ticks was found to vary between 8% and 18% (Cochez et al. 2015), depending on undergrowth and landscape connectivity (Heylen et al. 2019), undergrowth (Tack et al. 2013), and year (Ruyts et al. 2017). A full list of tick infection studies in Belgium is found in Supplementary Materials. Infection prevalence in ticks collected from humans was estimated at 13.9% at the national level (Lernout et al. 2019), and the incidence of Lyme borreliosis at 103 per 100 000 inhabitants (Geebelen et al. 2019).

In Western Europe, the main tick vector, Ixodes ricinus, also transmits Francisella tularensis and Coxiella burnetii, the causative agents of respectively tularaemia and Q fever. Tularaemia, which occurs mainly in the Northern Hemisphere, starts with flu-like symptoms but can evolve towards serious clinical manifestations and significant mortality if untreated (WHO 2007). Routes of infection include skin contact with infected animals, ingestion of contaminated water, and arthropod bites (ECDC 2017; WHO 2007). Lagomorphs and rodents are the main reservoir of its causative agent, the gram-negative intracellular bacterium F. tularensis (Carvalho et al. 2014). Tularaemia is rare in Belgium, but the number of cases detected is increasing, similarly to the number of serological tests: 28 cases were detected between 1950 and 2021, of which 25 were reported between 2012 and 2021 with an unknown source of infection for most cases (Litzroth and Mori 2021).

Coxiella burnetii is a gamma(γ)-proteobacteria of the Legionella order (Körner et al. 2021; Woldehiwet 2004) that infects a wide range of hosts, including dogs, bovine, deer, rodents, birds and ticks (Cutler et al. 2007; Mori et al. 2017). In humans, it can cause acute or chronic illness, but the diagnosis is challenging because most human infections are sub-clinical (Anderson et al. 2013; Cutler et al. 2007; Mori et al. 2017). In Belgium, 15 cases of Q fever were reported in 2021 (Litzroth et al. 2021). Infections generally occur through inhalation of airborne particles contaminated with C. burnetii in milk, faeces, urine and birth products from infected ruminants (Anderson et al. 2013; Cutler et al. 2007; Duron et al. 2015). Coxiella burnetii was found in about 40 tick species (Anderson et al. 2013), including I. ricinus (Hildebrandt et al. 2011). Transmission through tick bites has been demonstrated in experimental conditions, but its role is disputed (Duron et al. 2015). Ixodes ricinus in particular was rarely found infected in the Netherlands (Sprong et al. 2012) and Switzerland (Pilloux et al. 2019), for example. Tick bites may constitute another route of transmission for human infections, but likely less important than the airborne one (Sprong et al. 2012), which is identified as the most important one (Anderson et al. 2013).

Tick-borne disease (TBD) occurrence depends on the probability of contact between pathogens, ticks, reservoir hosts and susceptible human hosts (Lambin et al. 2010). The life cycle of I. ricinus spans over two to six years. Ixodes ricinus takes a single blood meal by active stage (larva, nymph, and adult) on a broad range of vertebrates, including mammals, birds, and reptiles (Estrada-Peña and de la Fuente 2014). This generalist species spends extended periods off-host in the environment, for questing, rehydrating, moulting or diapausing. These activities are strongly seasonal (Dobson 2013; Dobson et al. 2011). Ixodes ricinus is found in various vegetated environments, but forests are their primary habitats and are often associated with high tick abundance at regional (e.g., Li et al. 2012; Ruiz-Fons et al. 2012; Vanwambeke et al. 2010) and local scales (e.g., Tack et al. 2012; Van Gestel et al. 2021; Vourc’h et al. 2016). Peri-urban forests are of particular interest, for both providing suitable habitats for ticks, that is, having a high level of hazard, and being intensively visited by humans, that is, generating high exposure (Dobson et al. 2011; Ruiz-Fons and Gilbert 2010; Zeimes et al. 2014). In Belgium, 35% of tick bites were reported in forests (https://epistat.sciensano.be/ticks/, last access: 7th of August 2023). Variations in tick abundance and infection in endemic areas are key components of tick-borne disease risk assessment (Horobik et al. 2006; Mysterud et al. 2013).

Tick abundance in forests is often estimated by collecting ticks by flagging or dragging on a delimited area and counting the number of ticks collected (Nyrhilä et al. 2020). Estimates of hazard drawn from ideal tick habitat are sometimes used to understand risk but do not necessarily represent well areas of intense human exposure in forests, as people may avoid bushy areas or be prevented from leaving paths as per forest regulations as often apply in peri-urban forests. Variability in tick density and pathogen prevalence is substantial in a forest, as indicated by Tack et al. (2012) and Van Gestel et al. (2021). In this study, we repetitively collected ticks throughout the period of tick activity. We assessed the spatio-temporal heterogeneity of tick density within and around a peri-urban forest, also accounting for varying degrees of potential human exposure. We hypothesize that tick density and infection prevalence are not homogeneous within a forest and are affected by within-forest heterogeneity in undergrowth. The prevalences of the three above-mentioned pathogens (B. burgdorferi s.l., F. tularensis and C. burnetii) in ticks were also assessed.

Material and methods

Study area

The Bois de Lauzelle is a peri-urban forest of nearly 200 hectares located 30 km south of Brussels (Fig. 1). This peri-urban region has a high and growing population and has a higher incidence of erythema migrans than the Belgian average (Geebelen et al. 2019). The Bois de Lauzelle is bordered by a golf course, the city of Louvain-la-Neuve, and high-speed roads. The forest belongs to and is managed by the University of Louvain (UCLouvain) since 1970. It presents a diversity of local conditions, with an alternation of loamy and sandy-loamy soils and altitudes ranging from 45 to 153 m around the Blanc Ry River. The forest is dominated by deciduous trees, mainly beech (Fagus sylvatica) and oak (Quercus petraea, Quercus ruber) (Wallonie 2021), and is legally protected by several statutes (e.g., Natura 2000, code BE31006) (EEA 2020). The mean annual temperature and rainfall of the region are 10.6 °C and 849.4 mm / year (Royal Meteorological and Institute 2023).

Fig. 1
figure 1

Location of the sites in the Bois de Lauzelle. Sites sampled in 2016 and 2018 are represented by orange stars, those only in 2018 by pink circles. IF, TF and OF stand for interior-forest, trail-forest, and outside-forest site types vrespectively

Full size image

Tick sampling

Ticks were sampled in seven sites in 2016 and an additional seven sites in 2018 (Fig. 1, orange stars and pink circles). Each site was a delimited area of 10 m², using a 1 m²-white flannel over the leaf litter, woodland scrub, or low vegetation. We included three types of sites. Interior-forest (IF) sites were forested sites with dense undergrowth and generally further from the trails (36 ± 9 m). Trail-forest (TF) sites were forested sites with low or sparse undergrowth generally closer to the trails (17 ± 13 m). Outside-forest (OF) sites were in a vegetated environment outside but close to the forest. A description of the 14 sites is provided in Supplementary files (Figures S1– S14). The categories IF and TF describe the configuration of the sites rather than specifically a typical distance from the trails. While IF sites are typically not frequented by walkers and have vegetation growing more freely, TF can be used by walkers (such as a picnic area, a bench and its vicinity), regardless of the vicinity of the trail. Typically, TF1 (edge of the scouts grounds) and TF3 (a grassy clearing used for picnic) are not directly adjacent to the trail. Further composing the slight overlap between the measured distances is the positional uncertainty of digitised trails.

Ticks were collected between March and November. In 2016, ticks were sampled every two weeks, resulting in 17 sampling events (SE). In 2018, sites were visited every three weeks, resulting in 11 SE. The sampling design was the same as in 2018 (Table 1).

Samplings were performed on dry and non-windy days between 8:00am and 1:00pm. Sites were sampled in a random order during each SE to account for daily fluctuations in humidity and temperature with the time of the day. During each SE, we dragged a flannel on each sites several times consecutively in a rapid sequence. In 2016, we dragged 10 times consecutively, and in 2018 we reduced it to six times based on observations in 2016. Collections by drag were recorded in 2018 but not in 2016. It takes approximately 9 ± 4 min for six drags. The flannel was changed whenever it became dirty or humid and was examined for ticks at the end of each drag. Nymph and adult ticks were counted and collected after each drag, and ticks from the same SU were stored together in a vial with 70% ethanol. Larvae were not systematically sampled.

Table 1 summary of the data characteristics for 2016 and 2018
Full size table

Tick identification and preparation

Tick life stage and species were identified under a Leica EZ4 binocular (X35), based on two conventional morphological identification keys (Estrada-Pena et al. 2004; Hillyard 1996). We also used Heylen et al. (2014a) to distinguish between three similar species found in Belgium, I. ricinus, Ixodes frontalis and Ixodes arboricola. Ticks were washed in three consecutive baths with ethanol and sterilized water and crushed individually with a sterilized loop in Eppendorfs containing 200 µl of Dulbecco’s modified Eagle’s medium (DMEM) cell culture medium. Then, they were grouped in pools of four tick (50 µl/each, for a total volume of 200 µl) unless only a smaller pool was possible. Nymphs and adults collected in 2016 were pooled together. Nymph and adult ticks sampled in 2018 were pooled separately. Pooled ticks were from the same site and season. SE occurring between mid-March and mid-June were classed as spring, between mid-June to mid-September as summer, and between mid-September to the end of October as fall.

Dna extraction, sequencing, and PCR

DNA was extracted from 100 µl pool medium. The complete methods for DNA extraction and sequencing are described elsewhere (Rousseau et al. 2021). Tick genus (Ixodes versus Dermacentor) was confirmed, and DNA extraction validated using SYBR Green real-time PCR targeting 5 S and ITS2 genes. Ticks were screened for B. burgdorferi s.l., F. tularensis, and C. burnetii using: (i) two qPCR targeting the Outer Surface Protein A gene (OspA) and the Borrelia flagellin gene (Fla) (method adapted from Kesteman et al. (2010), (ii) TaqMan real-time (Light-Cycler® TaqMan® Master, Roche Diagnostics GmbH, Germany) targeting Francisella (ISFtu primers) (Versage et al. 2003)d tularensis (Tul4 primers) (Michelet et al. 2014), (iii) TaqMan real-time (Light-Cycler® TaqMan® Master, Roche Diagnostics GmbH, Germany) targeting C. burnetii insertion element IS1111 with previously described primers (Mori et al. 2013). The identification of Borrelia species was performed by Sanger sequencing on the Genetic Analyzer ABI 3730XL (Applied Biosystems, Invitrogen Life Technologies, Carlsbad, CA, USA), with the BigDye Terminator kit (Applied Biosystems). To achieve more consistent sequencing results, we used the universal M13 tailed primer attached to the locus-specific primers described by Kesteman et al. (2010). The obtained consensus sequences were analysed with the Basic Local Alignment Search Tool (BLAST) for species determination.

Tick variables

The density of nymphs (DON) and of adults (DOA) were computed as the total number of nymphs and adults sampled in the 10 or six consecutive drags in 2016 and 2018 respectively in each site and at each SE. We also computed drag-specific indices, DONi and DOAi, the cumulative number of nymphs and adults captured during the first i consecutive drags. Nymph (NIP) and adult (AIP) infection prevalences were calculated by maximum likelihood with the PoolTestR R package (McLure et al. 2021). Finally, the densities of infected nymphs (DIN) and adults (DIA) were calculated by multiplying DON and DOA by NIP and AIP respectively. For 2016, as nymphs and adults were not pooled separately, we could only compute tick infection prevalence (TIP), which may overestimate DIN and underestimate DIA as adults had more blood meals and are usually infected at higher rates (Gray 1998).

Environmental variables

We investigated the effect of environmental predictors on DON measured in 2018. Environmental characteristics were not monitored at individual SE in 2016. Ixodes ricinus retreats in the leaf layer for transstadial development and when questing conditions are unfavourable. When questing for a bloodmeal, it uses vegetation as a support (Gray 1998; Van Overbeek et al. 2008). For each SU (site at a particular SE), we measured undergrowth height and categorized it as low (< 20 cm), medium (20 to 40 cm), and high (> 40 cm). Forest edge habitats are suitable for various tick hosts (Allan et al. 2003; Brownstein et al. 2005; Tack et al. 2012). In peri-urban environments, Hansford et al. (2017) found more nymphs at the woodland edge than in other habitats. We measured the distance between each site and the closest forest edge. Guerra et al. (2002) found more I. scapularis on sandy and loamy-sand soils, compared to clay and loamy soils. Soil texture was extracted from the digital soil map of Wallonia (SPW 2005).

Ixodes ricinus is vulnerable to desiccation, and temperature and relative humidity influence its survival, development rates and activity periods (Brownstein et al. 2005; Diuk-Wasser et al. 2010; Medlock et al. 2013; Ruiz-Fons et al. 2012). We measured temperature and relative humidity over the course of the study period using a HOBO U23 Pro v2 Temperature/Relative Humidity Data Logger placed at 130 cm above the ground close to site TF.3. Temperatures measured at 130 cm and 5 cm in a forest are often correlated (Ruyts et al. 2018). The recording interval was 15 min. Temperatures during sampling events ranged from 5.2 to 25.3 °C and relative humidity from 41.6 to 99.5%. Mean temperatures and relative humidity for the 14 days preceding the samplings ranged from 7.5 to 21.7 °C and from 68.6 to 89.4%.

Statistical analyses

Statistical analyses were conducted in R.4.2.1 (Core Team 2022). The dataset has multiple observations from the same locations and thus required a multi-level approach. Spatial autocorrelation was tested for DON, DOA, AIP, NIP, TIP, DIN, and DIA using Moran’s I. Due to the multilevel structure, we performed repeated measures correlation between these indicators with the rmcorr R package (Bakdash and Marusich 2023). This technique deals with repeated, nested samples and has a greater statistical power because no aggregation is necessary (Bakdash and Marusich 2017). The complete R code is available as supplementary material.

First, we used generalized linear mixed models (GLMM) to investigate differences in DON and DOA, with the type of site (IF, TF and OF), season (spring, summer or fall), year (2016 and 2018) and their one-way interaction, with site as a random effect. We used DON6 and DOA6 for 2018, and DON10 and DOA10 for 2016, as we consider they were the best proxies of true tick abundances. They were tested at SE level (resulting in 273 observations for 2016 and 2018). We had too few observations for NIP, AIP, DIN to adjust a GLMM. To assess overdispersion, we compared Poisson distributions with negative binomial and their zero-inflated versions, using Vuong’s non-nested tests (pscl R package, Jackman 2020).

DON is usually estimated with the number of nymphs captured during a single drag. We created six GLMMs exploring the effects of season and type of sites on six measures using the number of nymphs collected using different number of drags for 2018.

Negative binomial mixed models were used to assess the effects of environmental variables on DON6 in 2018, which had a complete environmental record. After standardization of the environmental predictors following Gelman (2008), the complete model included all non-collinear predictors. Collinear predictors were identified using a variance inflation factor < 3. We also computed the best-AIC model, based on Akaike’s Information Criterion (AIC). The most parsimonious of models with differences of AIC smaller than|2|, considered equivalent, was chosen. All GLMMs were performed using the glmmTMB R package (Brooks et al. 2017). Residuals were controlled for dispersion, zero-inflation, heteroscedasticity and spatial autocorrelation with the DHARMa R package (Hartig 2022).

Finally, we calculated the diversity of Borrelia genospecies with the R vegan package (Oksanen et al. 2020). We used the exponential back-transformation of the Shannon-Wiener Index to compute a biodiversity “effective number of species”, to facilitate interpretability, comparison and interpretability (Jost 2006).

Results

Tick abundance and tick-borne pathogen prevalence

We collected 1642 Ixodes nymphs and 181 adults (Fig. 2 and Figure S15). Of these, 1804 were identified as I. ricinus, 17 as Ixodes frontalis and three as Ixodes ventalloi. 786 nymphs and 81 adults were collected in 2016. Based on the consecutive drags method, we obtained a DON of 6.61 ± 8.39 and a DOA of 0.68 ± 1.11 ticks per 10 m² for 2016. In 2018, 856 nymphs and 100 adults were collected, resulting in DON of 5.56 ± 9.06, and DOA of 0.65 ± 1.44 ticks per 10 m² based on consecutive drags. All sites reached their maximum abundance before July. In 2018, DON was higher in forested (IF and TF sites, 7.06 ± 10.14 nymphs / 10 m²) than in non-forested sites (OF sites, 1.80 ± 3.38). There was no statistical difference in DON and DOA between 2016 and 2018 when estimated on all sites (6.61 and 5.56 for DON and 0.68 and 0.65 for DOA) or across the different types of sites (Table S1).

Fig. 2
figure 2

Number of nymphs sampled in 2016 (A) and 2018 (B) and adult ticks sampled in 2016 (C) and in 2018 (D) by site. IF, TF and OF stand for interior-forest, trail-forest, and outside-forest respectively

Full size image

Nymphs and adults were grouped in 499 pools (221 for 2016 and 278 for 2018). Borrelia burgdorferi s.l. was detected in 90 pools (31 for 2016 and 59 for 2018). We found Borrelia positive pools in all sites, except TF.3, TF.4, OF.3. The mean TIP was 5.33%, 95%-CI [4.32–6.47], higher in 2018 (6.82%, 95%-CI [5.26–8.63]) than in 2016 (3.77%, 95%-CI [2.61–5.21]), and higher in fall (9.04%, 95%-CI [4.94–14.7]) over summer (5.60%, 95%-CI [3.95–7.62]) over spring (4.64%, 95%-CI [3.42–6.10. In 2018, NIP varied from 0 to 19% (IF.2) and AIP from 0 to 33% (IF.1 and OF.1). DIN varied from 0 to 1.25 infected nymphs per 10 m² in IF.2, and DIA from 0 to 0.20 in IF.3.

Spatial autocorrelation was only detected for DOA in 2016 (p-value = 0.04) (Table S2). We found significant repeated measures correlations between DON and DIN, DOA and DIA and DON and DOA, but not between tick prevalences and densities of ticks and infected ticks (Table 2). The strongest correlations were between DON and DIN (0.63, p-value < 0.0001) and DOA and DIA (0.45, p-value < 0.0001). Repeated correlation measures for individual years are found in Tables S3 and S4.

Table 2 Repeated measures correlations between tick indicators for 2016 and 2018. Upper part: coefficient [95% - confidence intervals]. Lower part: significance based on p-values. Degrees of freedom = 258. N.S. stands for not significant. * p-value < 0.05, ** p-value < 0.01, and *** p-value < 0.001. Significant correlations at p-value < 0.05 are in bold
Full size table

Generalized linear mixed models with site type, season and year

For all DONi indicators, the negative binomial distribution was preferred (Table S2). GLMMs expressed variations of the density of nymphs and of adults (Table 3) by year, season, and type of site. The highest estimation of DON was for spring 2016 and in IF sites: 18.62 nymphs per 10 m². It decreased in summer and in fall and in TF sites and OF sites. There were no differences between the two years and the first-level interaction between year and season was not significant. No pattern was detected in the residuals. The estimated DOA in spring 2016 and in IF sites was 0.99 adults per 10 m². There were no differences between years, seasons and types of sites, except a lower tick density in fall. The first-level interaction between year and summer was also significant. No pattern was detected in the residuals (Dharma residuals plots presented in Figure S.16 for the models with interaction).

Table 3 Negative binomial generalized linear mixed model estimates for density of nymphs (DON) and adults (DOA). The estimates and standard errors are expressed as the exponential of the log coefficients of the models. N.S. stands for not significant. * p-value < 0.05, ** p-value < 0.01, and *** p-value < 0.001. Significant estimates at p-value < 0.05 are in bold
Full size table

Generalised linear mixed models for different estimates of don

In 2018, 34% of the nymphs (Table S6) and 51% of the adults (Table S7) were collected in the first drag. The lowest rates for both nymphs and adults were in IF sites. GLMM using DON estimated with various numbers of drags in 2018 show that the estimates of DON in spring and in IF-site increased from 4.10 nymphs per 10 m² for DON1 to 16.64 for DON6 (Table 4). As in Table 3, DON decreased from spring to fall and was higher in IF compared to OF and TF-SU. Difference in DON between site type was detected when DON was estimated with the ticks sampled from three consecutive drags or more.

Table 4 Negative binomial generalized linear mixed model estimates for the number of nymphs sampled by consecutive drags in 2018 (DONi, with i from 1 to 6). The estimates and standard errors are expressed as the exponent of the log coefficient of the model. N.S. stands for not significant. * p-value < 0.05, ** p-value < 0.01, and *** p-value < 0.001. Significant estimates at p-value < 0.05 are in bold
Full size table

Generalized mixed model for the environmental predictors of tick density in 2018

When estimating the effects of the environmental variables on 2018-DON estimators, we found an estimate of 20.10 nymphs per 10 m² in sites with high undergrowth, loamy sands and excessive drainage (Table 5). DON was also negatively affected by relative humidity and temperature of the 14 previous days and positively by temperature during the sampling. Undergrowth height was significant and included in the best-AIC model, but no other variables describing sites. Dispersion, zero-inflation, heteroscedasticity, and spatial autocorrelation were not detected in the residuals of these models.

Table 5 Negative binomial estimates for the generalized linear mixed models of the density of nymphs by the environmental variables. The estimates and standard errors are expressed as the exponential of the log coefficients of the models. The AIC for the null model was 812.9. N.S. stands for not significant. * p-value < 0.05, ** p-value < 0.01, and *** p-value < 0.001
Full size table

Borrelia genospecies diversity

Six genospecies were identified: B. afzelii was dominant (54 isolations), followed by B. garinii (24), B. burgdorferi s.s. (13) and B. valaisiana (5). Borrelia bavariensis and B. spielmanii were only isolated once, in 2016 (Fig. 3). The effective diversity number for Borrelia genospecies was 3.21, higher in 2016 (3.42) compared to 2018 (2.93). It was also higher in forested sites (4.00 for TF and 2.71 for IF) than in non-forested sites (1.51 for OF.). TF.2, TF.1, IF.4, IF.2 and IF.3 presented the highest diversity in the number of genospecies (5, 4, 3, 3 and 3 respectively), and of isolations (14, 17, 16, 13 and 14 respectively) (Table 6). OF.3 had no genospecies isolated while TF.4, TF.5 and OF.2 had only one (B. spielmanii for the former and B. afzelii, for the last two). Coxiella burnetii was detected in six pools, three in 2016: one from TF.1 in spring and two from IF.1 in summer and three in 2018: one from IF.2 in spring and two from IF.4 in spring and summer. Due to the low prevalence, these results were not analysed further. Francisella tularensis bacteria were not detected.

Table 6 Borrelia burgdorferi s.l. genospecies diversity by site. BOR stands for Borrelia burgdorferi s.l., BAF for B. afzelii, BBA for B. bavariensis, BGA for B. garinii, BSPI for B. spielmanii and BVAL for B. valaisiana
Full size table
Fig. 3
figure 3

Borrelia burgdorferi s.l. infections in pooled ticks and genospecies composition by site in (A) 2016 and (B) 2018. BAF stands for B. afzelii, BBA for B. bavariensis, BGA for B. garinii, BSPI for B. spielmanii and BVAL for B. valaisiana

Full size image

Discussion

We investigated the heterogeneity of tick density and infection prevalence in a peri-urban forest. Acarological indicators were heterogeneous within the forest, as previously reported by Vourc’h et al. (2016) in France. The mean densities of questing nymphs in forested sites, estimated at 6.61 and 7.07 nymphs per 10 m² in 2016 and 2018 respectively, were in the upper range of other studies in Belgian forests. Elsewhere, DON was estimated at of 4.05, 6.1 and 6.4 nymphs per 10 m² in 2013, 2018 and 2019 respectively (Ruyts et al. 2016; Van Gestel et al. 2021). However, direct comparison with our estimates is not straightforward, as other studies estimated DON by dragging the vegetation once. DON estimated from a single drag at the Bois de Lauzelle were 2.93 and 2.38 nymphs per 10 m² in 2016 and 2018 respectively.

Dragging is the most common method for tick sampling for its cost-effectiveness and its ease of implementation and replication among field workers (Nyrhilä et al. 2020). However, it tends to underestimate absolute tick abundance, and its efficiency also fluctuates, i.e., with the time of the day or the vegetation sampled (Bord et al. 2014; Boyard et al. 2007). Vegetation modifies the contacts between the drag and questing ticks (Tack et al. 2011). With varying capture probabilities, comparable abundance estimates are difficult to achieve (Mackenzie et al. 2002). We estimated DON based on consecutive drags, performed in fast succession during mornings to collect individuals from a closed population, which gives better proxies of the true abundances (Bord et al. 2014). The use of 10 consecutive drags in 2016 may have resulted in higher DON estimates, but 10 drags may sample a new population as ticks become active. DON and DOA did not significantly differ between our two years of sampling. Variations in DON between site types (IF, TF, OF) was only significant when it was estimated with at least three consecutive drags. Sampling the same transect repeatedly increases the probability of tick/drag contact in complex undergrowth. While vegetation structure, especially when several strata of foliage are present, would always interfere with dragging, we believe the consecutive drags method allows to bring into light differences between sites. We believe our estimate to be reliable and recommend performing DON estimations based on at least three consecutive drags when comparing areas with different vegetation types.

Tick samplings were regularly performed throughout the year to capture the temporal variation of tick density, as recommended by Dobson (2013) and Salomon et al. (2020). This limits the biases of occasional samplings caused by a variety of factors influencing questing, sometimes locally (e.g., passage of hosts before the sampling). In both years, ticks were active during the entire period of sampling, but their abundance was higher in spring over summer, over fall, as observed in other peri-urban areas of Western Europe (e.g., Hansford et al. 2017). Tick density also varied within seasons. Body inspection should therefore be recommended to forest users throughout the season of tick activity.

The density of sampled ticks was influenced by temperature and relative humidity at the time of sampling and the 14 previous days. Temperature and humidity affect desiccation and therefore I. ricinus survival and activity (Perret et al. 2000). DON was positively associated to temperature at sampling time but negatively to the 14 previous day average. The influence of temperature is complex and not easily reproducible (Boyard et al. 2011). In Belgium, temperature is likely not a limiting factor for tick survival and persistence, but may affect tick activity (Medlock et al. 2013). DON was negatively influenced by relative humidity at sampling time as observed in other studies (e.g., Hubálek et al. 2006; Kiewra et al. 2014; Li et al. 2012; Schwarz et al. 2009). We recorded temperature and relative humidity at a single forest location, above the litter, not where I. ricinus shelter. Results by Boehnke et al. (2017) suggest that on-site measurement may explain differences between sites, however, our results indicate that overall conditions also significantly affect tick activity. This would be relevant for using general weather observation as part of risk assessment.

B. burgdorferi infection prevalence in I. ricinus ticks was estimated here at 5.3%, 95%-CI [4.3–6.5], lower than prevalences in questing ticks in Belgium: 17.8% (Heylen et al. 2019), 15.6% (Ruyts et al. 2016), 12% (Kesteman et al. 2010) and 9.1% (Tack et al. 2012). The low prevalence B. burgdorferi found at the Bois de Lauzelle may also relate to the lack of connections for wildlife to other forests, or to the presence of specific hosts and host composition (Ruiz-Fons et al. 2012). To analyse the effects of hosts composition on tick abundance, B. burgdorferi prevalence and genospecies composition, further studies should consider including methods like live-trapping for small mammals or camera for large-sized mammals (Pérez et al. 2012; Ruyts et al. 2018).

We screened ticks in pools, a conventional method for arthropod vector screening. This method presents several challenges: the exact number of infected ticks in a positive pool cannot be determined and there may be a dilution effect for pools with high number of ticks (Fracasso et al. 2023). We grouped ticks in pools of four ticks, by site, season, and stage (in 2018). We estimated prevalence with the maximum-likelihood estimate of pooled prevalence, which is less influenced by pool size and infection rate of ticks than the pool positivity rate and minimum infection rate methods (Fracasso et al. 2023). This method produces robust estimates and confidence intervals of B. burgdorferi infection prevalence.

Tick infection prevalence with B. burgdorferi was variable, but the small number of pools and the broad confidence intervals prevented us from analysing the differences in space and time. Borrelia burgdorferi infection prevalence was also found to be highly variable in Belgium and the Netherlands, ranging from 0 to 2% to 20–25%, over short distances (Hartemink et al. 2021; Kesteman et al. 2010). Tick infection prevalence should therefore not be estimated from a single location in a forest on a single timeframe. Hartemink et al. (2021) and Ruyts et al. (2017) did not find inter-annual variation in NIP and considered that DIN is mostly determined by DON. Forests present a diversity of micro-environments influencing tick density and infection prevalence. We found a weak correlation between DON and TIP (0.23), which is common in endemic areas (Randolph 2001), and has been reported elsewhere (James et al. 2013; Jouda et al. 2004; Ruyts et al. 2017; Vourc’h et al. 2016).

Borrelia afzelii was the dominant genospecies, followed by B. burgdorferi s.s. and B. garinii, which was consistent with other studies in Belgium. A recent meta-analysis of Borrelia prevalence in I. ricinus questing ticks in Western Europe also identified a composition of 46.6% for B. afzelii, 23.8% for B. garinii, 11.4% for B. valaisiana, 10.2% for B. burgdorferi s.s., and B. bavariensis, B. spielmanii rarely detected (Strnad et al. 2017). The effective number of Borrelia species, calculated as the exponential back-transformation of the Shannon Index, was higher in the forest, especially in sites with open areas and under-developed undergrowth. Borrelia bavariensis and B. spielmanii, identified once in this study, are rare genospecies in Belgium. Borrelia afzelii and B. spielmanii are commonly associated with small mammals, while B. garinii and B. valaisiana are hosted by sea birds and songbirds. B. burgdorferi s.s. is generalist (Comstedt et al. 2006; Gray 1998; Hanincova et al. 2003; Heylen et al. 2014b; Kurtenbach et al. 2002; Pedersen et al. 2020). The dominance of B. afzelii over B. garinii may suggest that rodents (e.g., Apodemus sylvaticus) are the most important feeding hosts for larval ticks in the Bois de Lauzelle.

Francisella tularensis and Francisella-like bacteria were not detected in this study. Their prevalence in ticks is usually low. In the Sénart Forest (France), F. tularensis was found in one out of 69 ticks (1.45%) in 2008, and not detected between 2009 and 2014 (Paul et al. 2016). Coxiella burnetii was detected in three pools in 2016 and three in 2018 and its prevalence in questing I. ricinus is also usually low, 0.2% in a study from the Netherlands (Sprong et al. 2012). The prevalence of these pathogenic agents in ticks is currently not well known in Belgium. The public health significance of pathogens, present at very low prevalence, is difficult to assess with the number of pools tested in this study. Other tick-borne pathogens would warrant investigation in Belgium as well, such as Anaplasma phagocytophilum, Babesia spp., Borrelia miyamotoi, Neoehrlichia mikurensis and Rickettsia helvetica (Lernout et al. 2019).

Conclusion

We detected ticks in all locations sampled in and around a periurban forest. DON was assessed with the consecutive dragging method, which offers a representative and comparable estimate of tick abundance in areas with different undergrowth. DON was higher within than outside the forest and heterogeneous within the forest. Forest undergrowth type and height are good indicators of the level of tick abundance in a forest, but other factors such as host abundance likely also affects abundace. Tick abundance variability was associated to two types of factors: (i) the micro-environment of the sampling site providing suitable habitats for ticks and their hosts; (ii) the sampling method and weather at the time of sampling that influence tick capture efficiency and tick activity respectively. Borrelia infected ticks were foundeverywhere, but at relatively low prevalence compared to other studies in Belgian forests. Six different Borrelia species were identified, with a typical composition for Western European forests. Coxiella burnetii was present but rare, and F. tularensis was not detected. Further understanding of tick abundance variability and its determinants may help identify high-hazard areas and, in heavily visited forests, assess the feasibility of managing vegetation to limit tick abundance in the most visited areas.