A three-years assessment of Ixodes ricinus-borne pathogens in a French peri-urban forest

Background Ixodes ricinus is the predominant tick species in Europe and the primary pathogen vector for both humans and animals. These ticks are frequently involved in the transmission of Borrelia burgdorferi (sensu lato), the causative agents of Lyme borreliosis. While much more is known about I. ricinus tick-borne pathogen composition, information about temporal tick-borne pathogen patterns remain scarce. These data are crucial for predicting seasonal/annual patterns which could improve understanding and prevent tick-borne diseases. Methods We examined tick-borne pathogen (TBP) dynamics in I. ricinus collected monthly in a peri-urban forest over three consecutive years. In total, 998 nymphs were screened for 31 pathogenic species using high-throughput microfluidic real-time PCR. Results We detected DNA from Anaplasma phagocytophilum (5.3%), Rickettsia helvetica (4.5%), Borrelia burgdorferi (s.l.) (3.7%), Borrelia miyamotoi (1.2%), Babesia venatorum (1.5%) and Rickettsia felis (0.1%). Among all analysed ticks, 15.9% were infected by at least one of these microorganisms, and 1.3% were co-infected. Co-infections with B. afzeli/B. garinii and B. garinii/B. spielmanii were significantly over-represented. Moreover, significant variations in seasonal and/or inter-annual prevalence were observed for several pathogens (R. helvetica, B. burgdorferi (s.l.), B. miyamotoi and A. phagocytophilum). Conclusions Analysing TBP prevalence in monthly sampled tick over three years allowed us to assess seasonal and inter-annual fluctuations of the prevalence of TBPs known to circulate in the sampled area, but also to detect less common species. All these data emphasize that sporadic tick samplings are not sufficient to determine TBP prevalence and that regular monitoring is necessary.


Background
Ticks are obligatory hematophagous arthropods and consequently, are one of the most important pathogen vectors [1][2][3]. Lyme borreliosis (LB) is the most commonly reported tick-borne disease (TBD) in the northern hemisphere and is caused by bacteria belonging to the Borrelia burgdorferi (s.l.) complex. In western Europe, Ixodes ricinus is known to be involved in the transmission of these bacteria to both humans and animals. This tick species has also been reported to be a vector for many other tick-borne pathogens (TBPs) with potentially significant consequences for human and animal health (Anaplasma, Rickettsia, Bartonella, Babesia, etc.) [4][5][6][7][8][9].
While multiple different pathogens have been identified and confirmed in I. ricinus ticks, very little is known about their seasonal and inter-annual variations. Timeseries studies are thus crucial to understanding natural variability in microbial communities over time. Over the last decade, only a handful of surveys have assessed

Open Access
Parasites & Vectors *Correspondence: thomas.pollet@vet-alfort.fr 1 UMR BIPAR, Animal Health Laboratory, INRA, ANSES, Ecole Nationale Vétérinaire d' Alfort, Université Paris-Est, Maisons-Alfort, France Full list of author information is available at the end of the article seasonal and monthly TBP variation patterns [10][11][12][13][14]. Although these results have heightened our general understanding of TBP dynamics, several of these studies were performed over short periods of less than two years, rendering it impossible to infer inter-annual discrepancies or to detect bias due to a particularly exceptional year. Only Coipan et al. [12] analysed several pathogenic genera in ticks sampled over more than two years. This study did demonstrate relationships between seasons and TBP prevalence (Borrelia, Rickettsia, Anaplasma, Neoehrlichia and Babesia) in questing tick populations. These variations were mainly attributed to the varying availability of reservoir hosts.
Tick density is also heavily influenced by the presence of suitable hosts, most notably wild ungulates that sustain adults, thus enabling tick population renewal [15,16]. However, it is important to emphasise that immediate tick survival and questing activities are highly dependent on suitable and specific environmental conditions (temperatures between 8-24 °C and humidity of up to 80%). Simultaneously, several studies have investigated whether pathogen presence influences tick behaviour. Herrmann & Gern [17,18] suggested that I. ricinus infected with B. burgdorferi (s.l.) can tolerate increased levels of desiccation, and Neelakanta et al. [19] demonstrated that I. scapularis infected with Anaplasma phagocytophilum are more resistant to cold. The presence of these TBPs could therefore enhance survival or questing activities of the infected ticks under challenging abiotic conditions, suggesting the existence of a potential link between pathogen prevalence in questing ticks and seasons.
Tick density and TBP prevalence can thus be influenced by several variables, and can therefore potentially fluctuate both seasonally and annually. Studying these dynamics is essential to better understanding and anticipating TBP risk.
Peri-urban forests containing both TBP-reservoir hosts and ticks, and which are highly frequented by people and their pets, represent a particularly interesting area to study tick and TBP dynamics. The Sénart forest, located to the south of Paris, harbours many large ungulates and abundant and diverse populations of other TBP reservoir hosts (bank voles, wood mice, Siberian chipmunks, roe deer, hedgehogs, etc.), and accommodates more than three million visitors every year. This forest is therefore particularly adapted to studying ticks and tick-borne pathogen dynamics.
In this study, we assessed the seasonal and inter-annual variability of I. ricinus-borne pathogens in the Sénart forest over three consecutive years (from April 2014 to May 2017), and determined whether any significant associations existed between these pathogens. We investigated a total of 31 pathogenic species (bacteria and parasites), belonging to 11 genera: Borrelia, Anaplasma, Ehrlichia, Neoehrlichia (only N. mikurensis), Rickettsia, Bartonella, Francisella, Coxiella, Theileria, Babesia and Hepatozoon.

Tick collection
Ixodes ricinus, nymphs and adults, were monthly collected during three years, from April 2014 to May 2017, in the Sénart forest in the south of Paris. Ticks were collected between 10:00 h and 12:00 h. Samplings were performed by dragging [20] on 10 transects of 10 square meters, localized on the parcel 96 (48°39′34.6″N, 2°29′13.0″E, Fig. 1). Dragging was always performed 3 consecutive times on each transect by the same persons to limit sampling bias. The presence of Dermacentor spp. was occasionally reported but no investigation has been led further. The presence of I. ricinus larvae was also sometimes noticed. Due to their small size making individual DNA extraction and analysis difficult, we chose to not collect them. After morphological identification, ticks were stored at − 80 °C. In total 1167 ticks were collected.

Tick washing, crushing and DNA extraction
Ticks were first washed once in ethanol 70% for 5 min and rinsed twice in sterile MilliQ water for 5 min. Ticks were then individually crushed in 375 µl of Dulbeccoʼs Modified Eagle Medium (DMEM) with decomplemented Foetal Calf Serum (10%) and six steel beads using the homogenizer Precellys ® 24 Dual (Bertin, Paris, France) at 5500× rpm for 20 s.
DNA extractions were then performed on 100 µl of tick crushing, using the DNA extraction kit NucleoSpin ® Tissue (Macherey-Nagel, Hoerdt, Germany), and following the standard protocol for human or animal tissue and cultured cells, from the step 2. DNA extracts were eluted in 50 µl of elution buffer and stored at − 20 °C until further use.
Two controls were performed: (i) the crushing control, corresponding to a DMEM tube in which crushing and DNA extraction were performed in the same conditions as on samples; and (ii) the extraction control, corresponding to the DNA extraction step without tick samples.

Tick-borne pathogens detection
A high-throughput screening of the most common bacterial and parasite species known to circulate in ticks in Europe was performed. This allowed us to target simultaneously 31 pathogenic species, 7 genera and 1 phylum: TBP DNA was detected using the BioMark ™ real-time PCR system (Fluidigm, San Francisco, USA), a microfluidic system allowing to perform 48 or 96 real-time PCR reactions on 48 or 96 different samples as described in Michelet et al. [21] and Moutailler et al. [22]. Briefly, each sample and primers/probe set were included in individual wells. A pressure system allowed to load them on the chip via microchannels, in individual reaction chambers of 10 nl, where each sample will be individually mixed with each primer/probe set.

Primers and probes
Primers and probes used for this analysis have been developed and validated by Michelet et al. [21] and Gondard et al. [23]. They have been designed to specifically amplify DNA from pathogens (bacteria and parasites) which are usually found in ticks in Europe. Their sequences, amplicon size, as well as targeted genes and pathogens are presented in Additional file 1: Table S1.
Please note that, due to potential cross-reactions between primer/probe combinations (i.e. design) targeting B. burgdorferi (s.s.) and B. spielmanii with respectively B. garinii/B. valaisiana and B. afzelii DNA (described in [21]), positive samples for the two former species were considered as negative when associated to the latter. Therefore, potential associations between B. burgdorferi (s.s.)/B. garinii, B. burgdorferi (s.s.)/B. valaisiana and B. spielmanii/B. afzelii cannot be detected and the co-infection percentage may be under-estimated.

DNA pre-amplification
DNA pre-amplifiations were performed using the TaqMan PreAmp Master Mix (Applied Biosystems, Illkirch, France). Basically, the different primer pairs, used for the real time PCR, were pooled combining equal volume of primers with a final concentration of 0.2 µM. For each sample, 1.25 µl of DNA extract was pre-amplified using the Perfecta PreAmp SuperMix reagent (1×) and the 0.2× pool (0.05 µM), in a final reaction volume of 5 µl. PCR cycle comprised a first cycle at 98 °C for 2 min, followed by 14 cycles with 2 steps, the first one at 95 °C for 10 s and the second one at 60 °C for 3 min. Preamplified DNA were then diluted (1:10) by addition of 45 µl of sterile deionised water before use.

High throughput real time PCR
For each pre-amplified sample, the BioMark ™ real-time PCR system (Fluidigm, San Francisco, USA) was used for high-throughput microfluidic real-time PCR amplification using the 48.48 microfluidic dynamic array (Fluidigm, San Francisco, USA). Amplifications were performed using FAM-and black hole quencher (BHQ1)labeled TaqMan probes with TaqMan Gene Expression Master Mix in accordance with manufacturer's instructions (Applied Biosystems, Illkirch, France). Thermal cycling conditions were used as follows: 95 °C for 5 min, 45 cycles at 95 °C for 10 s, 60 °C for 15 s, and 40 °C for 10 s. Data were acquired on the BioMark Real-Time PCR system and analysed using the Fluidigm Real-Time PCR Analysis software to obtain crossing point (CP) values. Three tick species controls (I. ricinus, Dermacentor reticulatus, Dermacentor marginatus), one negative water control and one positive Escherichia coli control were included in each chip.

Nested PCR and sequencing
Samples that were positive either only for species design but not for the genus design or only for the genus design and not for the species design were all re-analysed by nested PCR. We used primer pairs allowing to target another gene that the one tested into the real-time PCR experiment. Their sequences, amplicon size, as well as targeted genes and pathogen genus are presented in Additional file 2: Table S2. Amplicons were sequenced by the Eurofins company and sequences analysed using the Bioedit software and compared to the database NCBI (National Center for Biotechnology Information) by sequence alignment using nucleotide BLAST (Basic Local Alignment Search Tool).

Statistical analysis TBP prevalences at the seasonal and multi-annual scale
Differences in TBP prevalences were tested within and between years by using a multivariable logistic regression model. We considered the calendar season level for the within-year variability. Seasons were considered as following: Winter (December to February); Spring (March to May); Summer (June to August); and Autumn (September to November). Logistic regression models were developed using the TBP status of each nymph as the outcome measure and season and year as explanatory variables. We performed four specific models for the following group/species of TBPs: (i) B. burgdorferi (s.l.) (considering B. burgdorferi (s.s.), B. garinii, B. afzelii, B. valaisiana and B. spielmanii); (ii) B. miyamotoi; (iii) A. phagocytophilum; and (iv) R. helvetica. The models were constructed from a generalized linear model option (GLM) [24] using a binomial distribution (logit link). Model assessment was based on Akaikeʼs information criterion (AIC). Results were expressed as odds ratios (OR) and 95% confidence intervals. Statistical computations were performed in R 3.5.1. [25].

Statistical modelling of tick-borne pathogen associations
We tested the associations between the TBP species that belonged to the co-infection profiles of nymphs found in this study. We used the association screening approach [26]. For a given number of pathogen species tested (NP), the number of possible combination (NC) was calculated as NC = 2 NP . Assuming similar pathogen prevalence as those observed, a simulated dataset was built as a presence/absence matrix with hosts in lines and pathogen combinations in columns. With 5000 simulations, we obtained the NC statistic distributions. We estimated a 95% confidence interval to obtain a profile that includes simultaneously all the combinations. From this profile, we inferred for each combination two quantiles, Qinf and Qsup. A global test was based on the 95% confidence envelope. When H0 was rejected, the local tests were based on the NC confidence intervals (Qinf-Qsup) [26].

Tick temporal dynamics
From April 2014 to May 2017, a total of 1167 I. ricinus ticks were collected in the Sénart forest in the south of Paris (Fig. 1). Please note that May and June 2016 were unfortunately not sampled due to logistic issues. Collected ticks were composed of 1098 nymphs, 35 females and 34 males. Adults were sporadically detected all over the three years; due to their low total abundance (more than 10-fold less compared with nymphs), we decided to focus our temporal analysis on nymphs. The temporal dynamics of nymph densities over the three years is shown in Fig. 2. Nymph densities followed similar patterns from one year to another, with a main peak of activity observed every year in spring, a strong decrease during summer and a second peak, smaller, observed in Autumn (Fig. 2). In January and February, the average densities were less than 10 questing nymphs/100 m 2 . A clear rise was observed from March to May reaching an average peak of 95 nymphs/100 m 2 in May. We then observed a decrease in summer up to a minimum average of 5 nymphs/100 m 2 in September. The nymph densities increased slightly in October to reach an average of 13 nymphs/100 m 2 , before finally decreasing in November (2 nymphs/100 m 2 , sampled in 2015).

Detected pathogens and their prevalence in tick population
Due to technical problems, DNA was extracted and analysed only from 1044 nymphs among the 1098 previously mentioned. A total of 46 were negative for at least one positive control and thus have been removed from the analysis. From the 998 remaining DNA samples, 15.9% (95% CI: 13.7-18.3%) were positive for at least one tested pathogen, which belong to three bacterial and one protozoan genera: Anaplasma, Borrelia, Rickettsia and Babesia ( Table 1).   Pathogens belonging to the genus Anaplasma were detected in 5.4% (95% CI: 4.1-7.0%) of collected ticks. Most of them were positive for A. phagocytophilum (5.3% of all the samples) and one DNA sample was only positive for the primers/probe combination specific to Anaplasma spp. This sample was confirmed by nested PCR and the amplicon was then sequenced. The BLAST analysis on NCBI showed that this sequence matched at 99% of identity with four different Anaplasma species (A.  phagocytophilum, A. marginale, A. ovis and A. centrale). Therefore, this sample was only considered as positive for Anaplasma spp.
Two species of Rickettsia were detected in questing I. ricinus nymphs. Rickettsia helvetica was the most prevalent and was detected in 4.5% (95% CI: 3.3-6.0%) of nymphs. Rickettsia felis was detected in only one nymph (0.1%, 95% CI: 0.003-0.6%). The presence of R. felis DNA was confirmed by nested PCR and sequencing of the ompB gene. The obtained sequence (Gen-Bank: MN267050) matched with the corresponding gene sequence of R. felis (GenBank: GU182892.1) with an identity of 100% and query cover of 100%.
DNA from two species of Babesia were detected in questing nymph with the microfluidic PCR: Babesia venatorum (1.5%, 95% CI: 0.8-2.5%) of ticks and Babesia divergens (0.1%, 95% CI: 0.003-0.6%), detected in one tick. A deeper investigation of the B. divergens-positive sample, by nested PCR on the 18S rRNA gene and amplicon sequencing (GenBank: MN296295), allowed us to finally identify the DNA of B. capreoli, a species closely related to B. divergens, circulating in wild ruminants and unable to infect human and cattle erythrocytes [27].

Temporal patterns of TBPs in I. ricinus nymphs TBP prevalence at the monthly scale
In the following paragraph and corresponding figures, prevalences are those obtained for months with at least nine collected ticks.
Global infection rates fluctuated from 8% (95% CI: 3.3-15.7%) in May 2015 to 29.6% (95% CI: 18.0-43.6%) in June 2014 (Fig. 3). At the genus level, variations in TBP prevalences and the number of months for which at least one tick was positive for each tested TBPs are presented in Fig. 4 and Table 1.
DNA from pathogens belonging to both genera Rickettsia and Anaplasma were detected in 16   For parasites, DNA from the genus Babesia was detected in 9 months out of 30 sampled months. Prevalences presented the lowest variability ranging from 1.1% (95% CI: 0.03-6.1%) in April 2014 to 3.8% (95% CI: 0.5-13.0%) in March 2017 (Fig. 4). The main detected species was B. venatorum that was detected 9 times over 30 samplings and not detected during 9 consecutive sampled months, from June 2015 to April 2016.

TBP prevalences at the seasonal and multi-annual scale
In order to determine if the prevalence of TBPs was different within and between years, a multivariable logistic regression model was performed. The spring season and the year 2014 have been considered as references for the seasonal and yearly effect, respectively. Because some TBPs had too low prevalences in the nymph population (producing unreliable statistics), analyses were only performed on the most prevalent TBPs: A. phagocytophilum, R. helvetica, B. burgdorferi (s.l.), B. miyamotoi and B. venatorum.
Significant differences, in terms of prevalence, were observed at the seasonal scale (Table 2, Fig. 5) for R. helvetica (higher in summer compared to spring), B. burgdorferi (s.l.) (higher in autumn compared to spring) and Table 2 Multivariable logistic regression models assessing the seasonal and yearly TBP prevalence variations in nymphs. Odds ratios and their associated 95% confidence intervals obtained from the best model of TBP seasonal and yearly prevalence in questing nymphs
Eight different co-infection profiles were found ( All these associations between pathogens were tested using the association screening approach of Vaumourin et al. [26]. Compared to a random analysis, no associations were found to be underrepresented while two were overrepresented: the first one between B. garinii and B. afzelii (observation = 3; minimum expected = 0; maximum expected = 2), and the second one between B. garinii and B. spielmanii (observation = 2; minimum expected = 0; maximum expected = 1).

Ixodes ricinus density and seasonal dynamics
This three-year survey demonstrated a clear seasonal pattern in I. ricinus density, with a marked peak of questing nymphs in spring and a smaller peak in autumn. Low, but present activity was detected in winter, as has been observed in Germany [28]. In addition to these general patterns, some unexpected data were observed, the most striking being no peak activity in spring 2017 (April and May) with tick densities very similar to those recorded in March. Abiotic factors such as temperature, relative humidity and rainfall, or fluctuating host numbers in the sampling area are known to influence questing tick abundance and activity patterns [29][30][31][32][33][34] and could explain these unusual observations. It is important to note that 2017 was distinguished by an abnormally wet March, with total rainfall much higher than that recorded in previous years in the same area (71.3, compared to 11.2, 33.6 and 61.7 mm rain/month in 2014, 2015 and 2016, respectively). Interestingly, the increased March rainfall was followed by an April drought (7.9 mm of rain/month in 2017, compared to 48.4, 27.2 and 66.2 mm rain/month in 2014, 2015 and 2016, respectively) (rainfall data estimated from the Orly station, Metéo-France data; https ://donne espub lique s.meteo franc e.fr/?fond=produ it&id_ produ it=90&id_rubri que=32). These unusual meteorological characteristics could explain the stable tick density from March to May 2017. Thereby, this finding clearly shows that the bimodal tick activity pattern usually observed during this study can exceptionally change with unusual environmental conditions, reinforcing the importance of regular monitoring.

Ixodes ricinus-borne pathogen composition and prevalence over the three years
Most of the detected pathogen species corresponded to microorganisms known to circulate in the Western Palaearctic [35][36][37][38][39][40][41][42]. However, several species belonging to the genera Bartonella and Francisella previously reported in the studied area [40,43], were not detected. The most prevalent pathogen species were A. phagocytophilum (5.4% of the examined nymphs), R. helvetica (4.5%), and B. burgdorferi (s.l.) (3.7%). Both high-and low-prevalence TBPs were consistently detected in the sampling area for the duration of the study. Although prevalences varied between different TBPs, and some were not detected for long periods, they were all detected recurrently. Continued detection is consistent with the year-round presence of reservoir hosts in the sampling area (wood mice, bank voles, Siberian chipmunks, roe deer, common blackbird, European robin, song thrush, etc.) [34,44,45]. The continued presence of reservoir hosts could facilitate the circulation of dominant species, and maintain, even at low rates, less prevalent pathogen species that may not be detected by a single sampling. This does support to regularly studying TBP temporal dynamics, to improve the assessment of their prevalence.
We also detected in a single tick, the DNA of the emergent human pathogen R. felis. Its detection is particularly unexpected as this bacterium is known to be mainly transmitted from cat to cat via fleas, with human contamination arising from cat or flea bites. As we only detected DNA from R. felis, we cannot exclude that this detection could correspond to remnant DNA from the previous blood meal. Nevertheless, several studies have already detected the presence of R. felis or R. felis-like organisms in hematophagous arthropods (reviewed in [46,47]), including in ticks collected from natural environments [48], and notably in two studies performed on questing I. ricinus, including one based on RNA detection [49,50]. Rarely investigated in studies dealing with TBPs, the repeated detection of R. felis should encourage increased surveillance for this spotted fever-causing pathogen in humans. Finally, all these findings suggest that a unique sampling would certainly not facilitate the detection of this pathogen, again highlighting the importance of collecting and analysing ticks at a large temporal scale.
Babesia divergens and B. capreoli are two closely related species. During this study, we found out that the design initially used to detect B. divergens actually also enabled detection of DNA of B. capreoli. While B. divergens is responsible of babesiosis in humans and cattle, B. capreoli is only able to colonize erythrocytes from deer, therefore presenting no threat for humans or livestock [27]. These results emphasise the importance of confirmation and careful interpretations of microfluidic real time PCR [51].

Seasonal and inter-annual dynamics of I. ricinus-borne pathogens
Improving the prevention of TBDs requires a better understanding of their temporal, and in particular, their seasonal dynamics. However, only a few studies have addressed these issues during a minimum three-year period [12,13]. As ticks were collected monthly for over three years in this study, we detected significant seasonal or annual infection rate fluctuations for four TBPs: R. helvetica, B. burgdorferi (s.l.), B. miyamotoi and A. phagocytophilum. Note that the statistically significant highest prevalence of B. miyamotoi in winter is only due to the detection of one positive tick sampled during winter in 2017. In our opinion, this result alone is insufficient to presume that B. miyamotoi have an increased winter prevalence. However, we can observe that even if very few ticks are questing during these periods, they may carry TBPs.
While significant seasonal and annual differences were observed for B. miyamotoi and A. phagocytophilum, respectively, the presence of R. helvetica and B. burgdorferi (s.l.) varied significantly according to both seasons and years. None of these microorganisms followed a similar pattern. Comparing our results to the multi-annual studies previously mentioned, we observed that only R. helvetica presented similar seasonal patterns [12]. This finding again emphasises how the season, the year or the sampling area can influence TBP presence and prevalence in questing tick populations.
The most common explanation for temporal variations in TBP prevalences is the variable availability of reservoir hosts during tick previous feeding stage. This hypothesis was already suggested by Coipan et al. [12] who observed that several micro-organisms, assumed to share the same reservoir host, also presented similar seasonal patterns. Because the tick life-cycle is fundamentally linked to its host, any changes to the available host spectrum will undoubtedly influence TBP prevalence in the tick community [52]. However, because the entire tick life-cycle is multi-annual, it is difficult anyway to know if nymphs questing at the same time did perform their previous blood meal at the same period. The same generation of questing nymphs could come from larvae that would have fed at a different moment and thus potentially upon different host species. An alternative hypothesis, based on both the presence of pathogens and tick physiology, could also explain these patterns. Carrying certain TBPs was shown to improve tick resistance to challenging abiotic conditions. Herrmann & Gern [17,18] demonstrated that ticks carrying Borrelia species exhibited higher survival rates in desiccating conditions and a lower tendency to move to favourable conditions for maintaining water balance than non-infected ticks. This was associated to a higher reserve of energy in Borrelia-infected ticks [53] which would therefore exhibit higher resistance capacities to hydric stress notably. This hypothesis would explain the higher prevalence of Borrelia-infected questing ticks observed during or after the summer period in the present study and in those of Coipan & Takken [12,13]. Similarly, Neelakanta et al. [19] demonstrated a higher expression of the iafgp gene, coding for an antifreeze glycoprotein, in A. phagocytophilum-infected ticks. This thus conferred to ticks a stronger resistance to cold that could lead to higher prevalence of A. phagocytophilum-infected questing ticks during or just after the winter. This hypothesis was not consistent with our data as A. phagocytophilum was not observed in greater prevalence during the cold seasons of our study.
Our results, in combination with those from the literature, support the hypothesis that TBP prevalence is influenced by both biotic and abiotic factors, and reiterate that sporadic samplings are insufficient to assess it.
In this longitudinal three-year study, two TBP associations were significantly overrepresented compared to a random distribution: the first one was between B. garinii and B. afzelii, as has been previously observed in studies using similar detection tools [22,42], or different methods (16S rRNA gene sequencing [64]); the second one was between B. garinii and B. spielmanii. Interestingly, these findings contrast with published results on I. ricinus TBPs. While performing a meta-analysis on data published from 2010 to 2016, Strnad et al. [65] observed a negative correlation between B. garinii and B. afzelii. Similarly, Herrmann et al. [66] also detected a negative co-occurrence between these two species following an analysis of 7400 nymphs collected over three years. These results are coherent considering the host specificity of these Borrelia species. Indeed, B. garinii does not share the same reservoir host (birds) than B. afzelii or B. spielmanii (wood mice and bank voles, or hazel and garden dormice) [67][68][69][70][71][72], and none of these species are known to be transmitted transovarially.
Although the associations we identified were statistically "overrepresented", in fact we only observed one additional association than the fixed overrepresentation threshold (i.e. observed associations = 3 and 2; minimum expected = 0 and 0; maximum expected = 2 and 1; for B. garinii + B. afzelii and B. garinii + B. spielmanii associations, respectively). This indicates that caution should be applied when drawing conclusions about permanent associations between these different bacteria in ticks. Several different hypotheses could potentially explain these associations in the same nymph. First, hosts are likely to carry several adjacent feeding ticks. This phenomenon, known as co-feeding, could promote pathogen exchange between ticks even in the absence of systemic host infection [73]. Secondly, as discussed by van Duijvendijk et al. [74], when bloodmeals are disrupted due to host grooming, immune response or death, ticks may feed on more than one host to completely engorge, and consequently be exposed to several pathogens. Thirdly, despite these TBP species segregating between bird and rodent hosts, all of them have been detected in hedgehogs [75,76], and B. afzelii and B. garinii have been simultaneously detected in one Siberian chipmunk [45]. Both of these mammals were found to host a large number of tick larvae [45,77], and Siberian chipmunks have been reported to induce higher B. burgdorferi (s.l.) infection rates in nymphs, compared to bank voles and wood mice [45] in the Sénart forest. A last hypothesis might be that our analytical methods do not enable distinguishing the rodent-circulating B. garinii OspA serotype 4 (corresponding to B. bavariensis) [78] from other B. garinii serotypes.
Associations between B. garinii and B. valaisiana are frequently reported, which is not surprising as these species share the same reservoir host [79]. This association was the most common TBP association in a meta-analysis of literature published between 1984 and 2003 [80], and has been reported several times recently [11,66,81]. While we observed this association twice, both times in association with a third Borrelia species, either B. afzelii or B. spielmanii, it was not significantly overrepresented compared to a random distribution. Among the three previously mentioned studies, only Herrmann et al. [66] demonstrated that this association was overrepresented when compared to a randomly sampled analysis. However, our study was performed on a much smaller dataset (998 vs 7400 analysed nymphs), with a halved co-infection percentage (1.3 vs 3%), indicating that our statistical analysis may be less powerful, which could explain why this association was not detected.
These contrasting tick pathogen association results highlight the complexity in clearly identifying pathogen associations in field-collected ticks. Several other parameters can also potentially influence pathogen association (host spectrum within the studied area, sample size influencing analytical statistical power, identification bias, etc.). In this context, performing investigations under controlled conditions (suitable TBP growing and tick breeding systems, etc.) will be a crucial future step to experimentally test these different associations and improve our knowledge of TBP co-occurrences.

Conclusions
This three-year study of I. ricinus-borne pathogens (i) identified several TBPs previously reported in the area, consistent with reservoir host availability; (ii) allowed the surprising detection of R. felis DNA, a microorganism rarely reported in questing ticks; (iii) highlighted significant variations in seasonal and inter-annual pathogen prevalence; and (iv) identified several unexpected co-occurrences between pathogens belonging to the B. burgdorferi (s.l.) complex. All these results represent another step towards understanding the TBP ecology and emphasise the need to perform longitudinal studies, particularly because the main factors that are supposed to influence tick and TBP ecology could change in the next few years with climate changes. Associated to other factors such as host information or meteorological measures, this kind of data is crucial to allow a better understanding of TBP ecology and TBD epidemiology.
Additional file 1: Table S1. Targeted genes, amplicon size, primers and probe sequences used for TBP and tick species detection.
Additional file 2: Table S2. Targeted genes and primer sequences used for results confirmation.