INTRODUCTION
In the temperate regions of the northern hemisphere, Lyme disease is the most prevalent vector-borne disease, with important medical and economic consequences (World Health Organization, 2004; Wormser et al. Reference Wormser, Dattwyler, Shapiro, Halperin, Steere, Klempner, Krause, Bakken, Strle, Stanek, Bockenstedt, Fish, Dumler and Nadelman2006). When left untreated, infection can lead to chronic manifestations affecting skin, joints, heart or the nervous system, causing permanent disabilities (Stanek et al. Reference Stanek, Wormser, Gray and Strle2012). This multi-systemic disorder is caused by some bacteria of the Borrelia burgdorferi sensu lato (sl.) (‘Borrelia’) species group, a complex of more than 18 distinct genospecies of which at least nine occur in Europe: Borrelia afzelii, Borrelia garinii, B. burgdorferi sensu stricto (ss.), Borrelia valaisiana, Borrelia lusitaniae, Borrelia spielmanii, Borrelia bavariensis, Borrelia turdii and Borrelia bissettii (Margos et al. Reference Margos, Hojgaard, Lane, Cornet, Fingerle, Rudenko, Ogden, Aanensen, Fish and Piesman2010, Reference Margos, Vollmer, Ogden and Fish2011; Rudenko et al. Reference Rudenko, Golovchenko, Grubhoffer and Oliver2011; Pérez et al. Reference Pérez, Kneubühler, Rais and Gern2012; Coipan et al. Reference Coipan, Fonville, Tijsse-Klasen, van der Giessen, Takken, Sprong and Takumi2013; Heylen et al. Reference Heylen, Tijsse, Fonville, Matthysen and Sprong2013). At least five of these genospecies are known to cause disease in humans (see Stanek et al. Reference Stanek, Wormser, Gray and Strle2012 and the references therein). Humans become infected with Borrelia via Ixodes ticks, and in the Western Europe, the castor bean tick Ixodes ricinus L. is the tick vector that most frequently transmits Borrelia to humans (Piesman and Gern, Reference Piesman and Gern2004). The genospecies of the B. burgdorferi sl. complex are all associated with a specific host species, or a range of hosts, and, if pathogenic, cause different disease symptoms in humans, when humans are selected as a host by the tick. Borrelia afzelii, for example, is associated with small rodents such as mice and voles and usually causes skin manifestations in humans, while B. garinii is associated with birds and can lead to neuroborreliosis (Balmelli and Piffaretti, Reference Balmelli and Piffaretti1995; Humair et al. Reference Humair, Peter, Wallich and Gern1995, Reference Humair, Postic, Wallich and Gem1998; Comstedt et al. Reference Comstedt, Bergström, Olsen, Garpmo, Marjavaara, Mejlon, Barbour and Bunikis2006). Each Borrelia genospecies can be considered as a different pathogen with its own transmission dynamics in wildlife hosts, and different clinical manifestations and pathogenicity in humans.
Transmission potential varies between hosts, with some hosts transmitting Borrelia efficiently to ticks (i.e. efficient or competent reservoir species) and others rarely or not (i.e. poor reservoirs) (Matuschka et al. Reference Matuschka, Fischer, Heiler, Richter and Spielman1992; LoGiudice et al. Reference LoGiudice, Ostfeld, Schmidt and Keesing2003; Comstedt et al. Reference Comstedt, Bergström, Olsen, Garpmo, Marjavaara, Mejlon, Barbour and Bunikis2006; Heylen et al. Reference Heylen, Matthysen, Fonville and Sprong2014). This variation in reservoir competence has led to the formulation of the dilution effect hypothesis (DEH) (Ostfeld and Keesing, Reference Ostfeld and Keesing2000b ): a high diversity of host species can lower the prevalence of Borrelia in ticks because ticks then have a higher likelihood to feed on a host species that is a relatively poor reservoir. The hypothesis assumes that species-poor host communities consist mostly of competent reservoir species, such as mice, and that with increasing biodiversity, the proportion of poor reservoir species in the host community increases and more tick bites will then be wasted – from the point of view of Borrelia transmission – on poor reservoirs (Norman et al. Reference Norman, Bowers, Begon and Hudson1999; Ostfeld and Keesing, Reference Ostfeld and Keesing2000a , Reference Ostfeld and Keesing b ). While the DEH makes predictions about the proportion of nymphs infected with Borrelia (i.e. the nymphal infection prevalence), the actual disease risk for humans also depends on the contact rate between ticks and humans, and the density of host-seeking infected nymphs (Begon, Reference Begon, Ostfeld, Keesing and Eviner2008; Ogden and Tsao, Reference Ogden and Tsao2009; Wood and Lafferty, Reference Wood and Lafferty2013). A reduction in nymphal infection prevalence (NIP) alone does not necessarily provide a decreased Lyme disease risk, because adding more species to the host community can increase tick abundance and hence tick–human contact rate and the associated disease risk (Ogden and Tsao, Reference Ogden and Tsao2009; Randolph and Dobson, Reference Randolph and Dobson2012). To counter the effect of increased tick abundances, the reduction in NIP should be substantial to be of public health significance (Schmidt and Ostfeld, Reference Schmidt and Ostfeld2001).
A dilution effect for Lyme disease has already been found in some regions in North America (Allan and Ostfeld, Reference Allan and Ostfeld2003; LoGiudice et al. Reference LoGiudice, Ostfeld, Schmidt and Keesing2003), but the effect may not be generalizable to other regions. In North America, B. burgdorferi ss. is the main genospecies and acts as a host generalist in this region (Kurtenbach et al. Reference Kurtenbach, Hanincová, Tsao, Margos, Fish and Ogden2006; Stanek et al. Reference Stanek, Wormser, Gray and Strle2012). Furthermore, the host community of larval ticks in North America consists mainly of one very efficient reservoir species and several alternative host species which are not, or much less, efficient in transmitting Borrelia infection. In contrast, in Europe there are multiple pathogenic Borrelia genospecies, which are associated with specific ranges of hosts, and most of the host communities probably consist of multiple competent reservoirs (Ostfeld and Keesing, Reference Ostfeld and Keesing2000b ).
In many Western Europe regions, forests are being converted from homogenous conifer plantations to more natural, structure-rich mixed forests, dominated by indigenous broadleaved trees (Olsthoorn et al. Reference Olsthoorn, Bertelink, Gardenir, Pretzch, Hekhuis and Franc1999; Spiecker et al. Reference Spiecker, Hansen, Klimo, Skovsgaard, Sterba and von Teuffel2004). Tack et al. (Reference Tack, Madder, Baeten, Vanhellemont, Gruwez and Verheyen2012a , Reference Tack, Madder, Baeten, Vanhellemont and Verheyen2013) found an elevated tick abundance after conversion of pine stands to stands dominated by oak. This is partially due to the higher shrub cover in more structurally diverse, converted stands, which leads to more suitable abiotic conditions for the ticks. Such a forest conversion will also induce changes in the host community (Brockerhoff et al. Reference Brockerhoff, Jactel, Parrotta, Quine and Sayer2008; Du Bus de Warnaffe and Deconchat, Reference Du Bus de Warnaffe and Deconchat2008; Tack et al. Reference Tack, Madder, Baeten, De Frenne and Verheyen2012b ). Structure-rich mixed forests, the intended result of forest conversion, are considered to contain a higher host species diversity than homogenous non-indigenous pine plantations (Kennedy and Southwood, Reference Kennedy and Southwood1984; Laiolo, Reference Laiolo2002; Alexander et al. Reference Alexander, Butler and Green2006; Carnus et al. Reference Carnus, Parrotta, Brockerhoff, Arbez, Jactel, Kremer, Lamb, O'Hara and Walters2006; Brockerhoff et al. Reference Brockerhoff, Jactel, Parrotta, Quine and Sayer2008; Du Bus de Warnaffe and Deconchat, Reference Du Bus de Warnaffe and Deconchat2008). According to the Dilution Effect Hypothesis, this may dilute the prevalence of B. burgdorferi sl. in the ticks. Furthermore, an increase in host diversity may lead to an increase in genospecies diversity, which, due to the association between genospecies and clinical manifestations, can change Lyme disease risk. Large-scale conversions of pine forests in Western Europe provide a unique opportunity to test the DEH in Europe in a complex setting of multiple pathogenic genospecies and increasing host diversity.
By using the different Borrelia genospecies as distinct pathogens while testing the dilution effect in Europe, this study contributes to the knowledge gap on the interaction between the Borrelia genospecies, tick vector and hosts. In addition, most studies testing the dilution effect use indirect measures of human disease risk such as NIP or tick abundance, while we use the density of infected questing nymphs, which better predicts risk (Begon, Reference Begon, Ostfeld, Keesing and Eviner2008; Ogden and Tsao, Reference Ogden and Tsao2009). Here we investigated forest stands that differed in dominant tree species and presence of shrub cover, leading to four different forest types with supposedly increasing host diversification degree: from stands without a shrub layer dominated by pine trees to stands with a substantial shrub layer dominated by oak. Pine stands with a shrub layer and oak stands without a shrub layer are intermediate in diversification degree. We expect that changing dominant tree species and forest structure leads to diversified forest communities, which entails increased host diversity, and changes Lyme disease risk. We hypothesize that forest community diversification changes the density of infected nymphs (DIN) by: (i) increasing the abundance of nymphal ticks and (ii) decreasing the NIP of B. burgdorferi sl. through a dilution effect. Furthermore, we expect the Borrelia genospecies community composition to change with forest community diversification, with nymphs from diversified forests housing a more diverse genospecies community.
MATERIALS AND METHODS
Study species
The life cycle of I. ricinus consists of three parasitic stages: larva, nymph and adult. Each of these stages needs one blood meal (except for adult males, which do not feed) and waits in the vegetation for a suitable host to pass to which it can attach itself (i.e. questing). Ixodes ricinus becomes infected with Borrelia by taking up spirochetes from the blood from an infected host, and afterwards passes infection from stage to stage (i.e. transstadial transmission) (Gray, Reference Gray1998). An infected tick can, in its turn, infect an uninfected host. Larvae usually hatch free of infection (Barbour and Fish, Reference Barbour and Fish1993; Parola and Raoult, Reference Parola and Raoult2001). The nymphal stage is most responsible for transmitting Borrelia to humans (Barbour and Fish, Reference Barbour and Fish1993), and thus the most interesting stage to examine in the framework of public health risk. Ixodes ricinus is vulnerable to desiccation (Needham and Teel, Reference Needham and Teel1991) and therefore occurs predominantly in forests that supply enough humidity by their dense vegetation cover (Gray et al. Reference Gray, Kahl, Robertson, Daniel, Estrada-Peña, Gettinby, Jaenson, Jensen, Jongejan, Korenberg, Kurtenbach and Zeman1998). Forests moreover harbour a large enough host community for ticks to feed on (Gray et al. Reference Gray, Kahl, Robertson, Daniel, Estrada-Peña, Gettinby, Jaenson, Jensen, Jongejan, Korenberg, Kurtenbach and Zeman1998; Ruiz-Fons et al. Reference Ruiz-Fons, Fernández-de-Mera, Acevedo, Gortázar and de la Fuente2012).
Study area
In Belgium, converting structure-less pine forests to structure-rich, natural forests dominated by broadleaved tree species occurs on a large scale in the Campine region, a region with a high risk of Lyme disease (Linard et al. Reference Linard, Lamarque, Heyman, Ducoffre, Luyasu, Tersago, Vanwambeke and Lambin2007; Verheyen et al. Reference Verheyen, De Schrijver, Wuyts, Gielis, Van Gossum, Geudens, Van Herzele, De Boever and Vanhellemont2007; Vanthomme et al. Reference Vanthomme, Bossuyt, Boffin and Van Casteren2012). The Campine region is an area with nutrient-poor and acid sandy soils located in the northeastern part of Belgium and spanning the provinces Limburg and Antwerp. The region has a temperate climate with warm summers (Peel et al. Reference Peel, Finlayson and McMahon2007) with a mean annual precipitation and temperature of 800 mm and 9·0 °C, respectively (Royal Meteorological Institute of Belgium, http://www.kmi.be, accessed December 19, 2014). In the 19th and the beginning of the 20th century, the original oak–birch forests on former heathland were mainly replaced by pine plantations of Scots pine (Pinus sylvestris L) and, to a lesser extent, Corsican pine [Pinus nigra Arnold subsp. laricio (Poiret) Maire]. The current characteristic forest structures of this region are pine plantations, interspersed with deciduous forests composed of pedunculate oak (Quercus robur L.), red oak (Q. rubra L.), common beech (Fagus sylvatica L.), silver birch (Betula pendula Roth.) and downy birch (Betula pubescens Ehrh.). The shrub layer mainly consists of rowan (Sorbus aucuparia L.), black cherry (Prunus serotine Ehrh.), alder buckthorn (Frangula alnus Miller), pedunculate oak, red oak and silver birch. The most common species in the herbaceous vegetation are wavy hair-grass [Deschampsia flexuosa (L.) Trin.], purple moor-grass [Molinia caerulea (L.) Moench], broad buckler-fern [Dryopteris dilatata (Hoffmann) A. Gray], narrow buckler-fern [D. carthusiana (Vill.) H.P. Fuchs], blackberry (Rubus fruticosus L. agg.), European blueberry (Vaccinium myrtillus L.), heather [Calluna vulgaris (L.) Hull] and honeysuckle (Lonicera periclymenum L.) (Waterinckx and Roelandt, Reference Waterinckx and Roelandt2001).
For this study, we primarily used the same stands as Tack et al. (Reference Tack, Madder, Baeten, Vanhellemont, Gruwez and Verheyen2012a ). These stands are dominated by either oak (pedunculate or red oak) or pine (Scots or Corsican pine), hereafter called oak and pine stands, respectively. The stands belong to one of four forest types, namely oak or pine stands with (>50% of the forest ground covered with shrub layer) or without (<25%) a well-established shrub layer, and represent a gradient of forest community diversification from low diversified pine stands without a shrub layer to highly diversified oak stands with a well-established shrub layer. The pine stands with a shrub layer and the oak stands without a shrub layer are less straightforward to rank mutually according to forest community diversity, but we can reasonably assume they are both intermediate between the high and low diversity forest types. A total of 93 stands were selected in 20 different forest sites: 20 pine stands without shrub layer, 32 pine stands with shrub layer, 19 oak stands without shrub layer and 22 oak stands with shrub layer. We attempted to select at least one stand of each type in every forest site. Due to the strong association between host species and tree species, the host community may differ between the two oak species. In our study, we selected pedunculate oak stands, as well as stands of non-indigenous red oak, but investigated their (indirect) effect on the ticks and Borrelia infection later during analysis. Forest stands were on average 1 ha large, ranging from 0·5 to 4 ha, and were distributed between 51°16′N, 4°29′E in the northwestern part of the Campine region and 50°54′N, 5°39′E in the southeast.
Data collection
Each forest stand was visited once when the vegetation was dry between June and September 2013. Questing ticks were sampled by sweeping a white flannel blanket (1 × 1 m2) attached to a wooden pole over the herbaceous vegetation (‘flagging’). Six 25 m parallel transects were performed in the centre in a representative part of each stand. The sampled area was calculated as the product of distance flagged (6 × 25 m = 150 m2) and width of blanket (1 m); nymphs per 100 m2 was used to as our measure of relative nymphal population density. The composition of the herbaceous layer was comparable between the different stands so we believe that it did not influence the density estimates by impeding the sampling technique (Tack et al. Reference Tack, Madder, Baeten, De Frenne and Verheyen2012b ). For that reason, also, sampling in dense shrubbery and bracken fern [Pteridium aquilinum (L.) Kuhn] was avoided. After each transect, nymphs were removed from the blanket with forceps and stored in vials containing 70% ethanol. Nymphs were pooled per forest stand and 35 individuals were selected randomly from each pool. For one stand, only 34 individuals were analysed due to a low number of nymphs caught. DNA extraction from each of these ticks was done by alkaline lysis in ammonium hydroxide, as described previously (Schouls et al. Reference Schouls, Pol, Rijpkema and Schot1999). For the detection of B. burgdorferi sl., a duplex qPCR, targeting ospA and flaB genes, was used. For the sequences of primers and probes we refer to Appendix A. The qPCR was performed using the iQ Multiplex Powermix PCR reagent kit (Bio-Rad Laboratories, Hercules, USA), in a LightCycler 480 Real-Time PCR System (F. Hoffmann-La Roche, Basel, Switzerland). The reaction mix consisted of iQ multiplex Powermix, 100 nm of the B-FlaB-Rc and B-FlaB-Rt primers, 200 nm of the B-FlaB-F, 400 nm of the B-OspA_modF and B-OspA_borAS primers, 100 nm of the B-OspAmodPatto probe, 200 nm of the B-FlaB-Patto probe, and 3 µL of template DNA in a final volume of 20 µL. Cycling conditions included an initial activation of the iTaq DNA polymerase at 95 °C for 5 min, followed by 60 cycles of a 5 s denaturation at 95 °C, followed by a 35 s annealing-extension step at 60 °C (ramp rate 2·2 °C s−1 and a single point measurement at 60 °C) and a cooling cycle of 37 °C for 20 s. Analysis was performed using the second derivative calculations for crossing point values. For each run positive and negative controls and blank samples were included. For confirmation of presence of B. burgdorferi sl. DNA, the positive samples of the qPCR were further submitted to PCR targeting the variable 5S–23S intergenic spacer region (IGS). The PCR was performed according to the protocol described in Coipan et al. (Reference Coipan, Fonville, Tijsse-Klasen, van der Giessen, Takken, Sprong and Takumi2013). Identification of Borrelia genospecies was done based on the DNA sequence of IGS. PCR products were sequenced using an ABI PRISM BigDye Terminator Cycle sequencing Ready Reaction kit (Perkin Elmer, Applied Biosystems). Sequences were confirmed by sequencing both strands (Sanger et al. Reference Sanger, Nicklen and Coulson1977). Storage and analysis of the IGS sequences were performed in BioNumerics version 7.0 (Applied Math, Belgium). Borrelia genospecies were assigned based on sequence identity with reference DNA sequences from GenBank (http://www.ncbi.nlm.nih.gov).
Data analysis
All analyses were conducted in R version 3.2.0 (R Core Team, 2015). Graphs were made with the package ggplot2 (Wickham, Reference Wickham2009). We defined NIP as the proportion of nymphs infected with B. burgdorferi sl. and the density of nymphs (DON) as the number of nymphs caught per 100 m2. The DIN is the product of NIP and DON. First, we fitted effects for tree species (‘pine’ or ‘oak’), the presence of a shrub layer (‘yes’ or ‘no’) and the interaction between tree species and the presence of a shrub layer on DON, NIP and DIN using generalized linear models (glm). DON and DIN resembled overdispersed count data, so we applied a negative binomial error distribution using the package MASS (Venables and Ripley, Reference Venables and Ripley2002). For NIP we used a glm with binomial error distribution for proportional data. Significances in all model fits were assessed through analysis of deviance with Chi-square (χ 2) test. We checked for heterogeneity of the residuals following the approach described in Zuur et al. (Reference Zuur, Ieno, Walker, Saveliev and Smith2009).
Second, to examine the impact of the different types of forest composition and structure on the Borrelia genospecies community compositions, we fitted a multivariate generalized linear model (manyglm) to our data (Warton et al. Reference Warton, Wright and Wang2012) as implemented in the R package mvabund (Wang et al. Reference Wang, Naumann, Wright and Warton2012). The fitted model assumed a binomial distribution of the prevalence (proportion) data. Predictor variables included tree species, the presence of a shrub layer and the interaction between tree species and presence of a shrub layer. An analysis of deviance for multivariate glm fits with likelihood ratio test was employed to determine which variables contributed significantly to the differences between the community compositions. Afterwards, we considered the Borrelia genospecies as distinct pathogens, and analysed the effect of the predictor variables on the prevalence of the individual genospecies the same way as described above for the response variable NIP with a binomial glm.
Finally, we calculated the diversity of each genospecies community (exponent of Shannon index or equivalent species numbers) with the function diversity in the R package vegan (Oksanen et al. Reference Oksanen, Blanchet, Kindt, Legendre, Minchin, O'Hara, Simpson, Solymos, Stevens and Wagner2015) and tested for the effect of the forest type variables with a glm.
To test the impact of the ecologically relevant difference between the stands of indigenous pedunculate oak and non-indigenous red oak, we conducted all above-mentioned analyses on the whole dataset of 93 stands and afterwards on a limited dataset of the oak stands. The outcomes of the analysis on the dataset of the red oak stands were then compared with those of the pedunculate oak stands. No difference was found for any response variable, at the level of B. burgdorferi sl. or the genospecies, between the indigenous pedunculate oak stands and the non-indigenous red oak stands, so we considered the category ‘oak’ as robust.
RESULTS
In the 93 forest stands, a total of 9554 I. ricinus nymphs were caught, with a mean density (±s.d.) of 40·5 nymphs (±34·8) per 100 m2. 3254 nymphs were examined for Borrelia spirochetes, and 508 were infected. This corresponds with a mean (±s.d.) NIP of 15·6% (±8·1) over all forest stands. Average DIN was 5·9 nymphs (±6·1) per 100 m2. Borrelia-positive nymphs were found in all forest stands examined. We identified six different genospecies in 462 (91%) infected nymphs, namely B. afzelii, B. garinii, B. burgdorferi ss., B. valaisiana, B. spielmanii and B. bavariensis. For the other 46 infected nymphs, genospecies could not be defined because of the lower sensitivity of the conventional PCR in comparison with the qPCR. Only one genospecies was identified in each infected nymph. From the six detected genospecies, B. afzelii was the most prevalent genospecies, occurring in 10·6% of the nymphs (±7·0) or 73·8% (±27·4) of the infected nymphs. The mean prevalence of B. garinii and B. burgdorferi ss. in nymphs was 1·5% (±2·5) and 1·2% (±2·1), respectively. The mean prevalences of the other genospecies were lower than 1%.
The glm showed a significant effect of tree species on DON (P = 0·02) and a marginally significant effect of the presence of a shrub layer (P = 0·05) with lowest densities in pine stands without a shrub layer (Fig. 1). The interaction between tree species and the presence of a shrub layer had a marginally significant effect on DIN (P = 0·09). We found no significant effect of any predictor variable on NIP (Fig. 1).
Fig. 1. Density of nymphs (A), NIP (B) and DIN (C) in the different forest types (mean ± s.d.). n, number of stands used per forest type. Note the difference in scaling of the y-axis, and the units used.
The community analysis of Borrelia genospecies revealed a significant effect of tree species on the diversity of Borrelia genospecies communities (Fig. 2, Appendix B), with more diverse communities in nymphs from oak stands (t = −2·21, P = 0·03). No significant effect of tree species or the presence of a shrub layer was found on the genospecies community composition, although the effect of tree species was marginally significant (Dev = 4·44, P = 0·07). When we considered the Borrelia genospecies as separate pathogens in the glm, we could not detect a significant effect of the forest type variables on the NIP of the genospecies. Nevertheless we can clearly see in the genospecies community composition (Fig. 2) a trend towards a higher prevalence of B. afzelii in nymphs from pine compared with oak stands, and for B. garinii and B. burgdorferi ss. a trend towards a higher prevalence in oak compared with pine stands. The genospecies community in the infected nymphs appeared to be dominated by B. afzelii in all forest types, but it was more pronounced in the pine stands [83·9% (95% confidence interval [71·4, 91·6]] of the infected nymphs) compared with the oak stands (62·4% [46·9, 75·8]) (Fig 2, Appendix B). B. garinii and B. burgdorferi ss. occurred in 6·7% [2·4, 17·6] and 4·9% [1·4, 15·3] of the infected nymphs from the pine stands compared with 12·5% [5·4, 26·5] and 16% [7·6, 30·5] in the oak stands, respectively. The share of B. valaisiana, B. spielmanii and B. bavariensis in the genospecies community in nymphs was relatively low, namely 1% [0·1, 13·3], 3% [0·6, 13·3] and 0·4% [4·8×10−5, 2·4] in pine, and 5·7% [1·6, 18·5], 2·9% [0·5, 15·6] and 0·3% [1·9×10−5, 3·9] in oak stands, respectively.
Fig. 2. The community composition of Borrelia genospecies in field collected nymphs (mean proportion + 95% confidence interval) in pine and oak stands, with or without shrub layer. n, number of stands used per forest type.
DISCUSSION
This study tests the DEH for Lyme disease in Europe while focusing on the different Borrelia genospecies as distinct pathogens, which, to our knowledge, has hardly been performed. Even though we did not test the dilution effect directly, since we did not empirically quantify the host communities, the association between planted forests and biodiversity has been covered sufficiently in literature to make reasonable assessments about the host diversity in the examined forest types (Laiolo, Reference Laiolo2002; Carnus et al. Reference Carnus, Parrotta, Brockerhoff, Arbez, Jactel, Kremer, Lamb, O'Hara and Walters2006; Brockerhoff et al. Reference Brockerhoff, Jactel, Parrotta, Quine and Sayer2008; Du Bus de Warnaffe and Deconchat, Reference Du Bus de Warnaffe and Deconchat2008). We assumed the host diversity to be higher in the oak stands with a substantial shrub layer than in the pine stands without a shrub layer. Further research which empirically quantifies the host communities, however, is needed to confirm this statement. We show that the dilution effect must be nuanced for Europe and should consider the response of multiple pathogenic genospecies. The DIN and associated human health risk will not necessarily decline with increasing biodiversity or host diversity, but will instead largely depend on the species assemblages of the host community. The different Borrelia genospecies can thus best be considered as separate pathogens, each with a separate possible dilution effect.
Highest densities of nymphs were found in oak stands and lowest in pine stands. Hence, our results show that changing forest composition could lead to increased tick abundances, which confirms the findings of former studies (Tack et al. Reference Tack, Madder, Baeten, Vanhellemont, Gruwez and Verheyen2012a , Reference Tack, Madder, Baeten, Vanhellemont and Verheyen2013). This could be due to the fact that more host species are present in oak forests, as suggested by Laiolo et al. (Reference Laiolo, Caprio and Rolando2004) and Tack et al. (Reference Tack, Madder, Baeten, De Frenne and Verheyen2012b ). We found an average NIP of 15·6% which is higher than the 10·1% mean NIP usually found in Europe (Rauter and Hartung, Reference Rauter and Hartung2005). This finding, together with the fact that the forests in the Campine region are readily visited for outdoor activities, suggests that it is an area of potentially high Lyme disease risk, as already proposed by Linard et al. (Reference Linard, Lamarque, Heyman, Ducoffre, Luyasu, Tersago, Vanwambeke and Lambin2007) and Vanthomme et al. (Reference Vanthomme, Bossuyt, Boffin and Van Casteren2012). The NIP did not differ between forest types. The lack of a decline in NIP does not allow us to confirm the DEH. This might indicate that even though host diversity is likely higher in oak forests, the relative proportion of non-competent hosts is probably not substantially large enough to reduce infection prevalence, e.g. because more competent than non-competent species are added to the host community with increasing host diversity (a possible scenario recognized by Ostfeld and Keesing, Reference Ostfeld and Keesing2000b ; Wood and Lafferty, Reference Wood and Lafferty2013). We could also not detect a significant effect of tree species or the presence of a shrub layer on the DIN. According to Begon (Reference Begon, Ostfeld, Keesing and Eviner2008) and Schmidt and Ostfeld (Reference Schmidt and Ostfeld2001), a dilution effect only applies to situations where tick bites are wasted on less competent reservoir hosts and when the compensatory increase in vector abundance, caused by the increase in host diversity, is limited. Regardless of whether the host species in question are competent reservoirs or not, an increase in host abundance, and thus feeding opportunities for the ticks, will increase tick abundance (Ogden and Tsao, Reference Ogden and Tsao2009; Randolph and Dobson, Reference Randolph and Dobson2012). This indeed appears to be the case in our study, with higher densities of nymphs in oak forests.
The most common genospecies found in questing nymphs in Europe are B. afzelii and B. garinii but the prevalence of the different genospecies varies between regions (Rauter and Hartung, Reference Rauter and Hartung2005). In our study region, we found that B. afzelii and B. garinii constituted 74·5 and 9·3% of the genospecies communities, respectively. Since B. afzelii is associated with small rodents, and the majority of infected nymphs were infected with this genospecies, we can assume that rodents such as mice and voles are important feeding hosts for I. ricinus larvae and important transmission hosts of Borrelia infection. B. garinii (and also B. valaisiana) is associated with passerine birds (Heylen et al. Reference Heylen, Tijsse, Fonville, Matthysen and Sprong2013), and birds appear to be less important reservoir hosts in the Borrelia transmission cycle of I. ricinus ticks than rodents, since the infection prevalence of questing nymphs with B. garinii and B. valaisiana is much lower than that with B. afzelii. This lower prevalence of infection can be due to the fact that passerine birds are less important feeding hosts for the larvae of I. ricinus or that they are less efficient at transmitting Borrelia infection or both (Brunner et al. Reference Brunner, LoGiudice and Ostfeld2008). A recent meta-analysis demonstrated that birds are generally more important feeding hosts for nymphs than for larvae (unpublished results, although this strongly depends on bird species, see Comstedt et al. Reference Comstedt, Bergström, Olsen, Garpmo, Marjavaara, Mejlon, Barbour and Bunikis2006; Marsot et al. Reference Marsot, Henry, Vourc'h, Gasqui, Ferquel, Laignel, Grysan and Chapuis2012), so that questing larvae are less likely to obtain infection with B. garinii than questing nymphs. In the nymphs in our study area, we also found an unusually high prevalence of B. burgdorferi ss., namely 9·8% (Rauter and Hartung, Reference Rauter and Hartung2005; Burri et al. Reference Burri, Cadenas, Douet, Moret and Gern2007; Bingsohn et al. Reference Bingsohn, Beckert, Zehner, Kuch, Oehme, Kraiczy and Amendt2013). This is the highest reported prevalence of this genospecies in questing nymphs in Western Europe as far as we know. B. burgdorferi ss. is most probably associated with Eurasian red squirrel (Sciurus vulgaris) but because this association is not yet strictly determined (Humair and Gern, Reference Humair and Gern1998; Pisanu et al. Reference Pisanu, Chapuis, Dozières, Basset, Poux and Vourc'h2014, but see Kurtenbach et al. Reference Kurtenbach, De Michelis, Etti, Schäfer, Sewell, Brade and Kraiczy2002), it is plausible that another host species, or range of hosts, is responsible for the transmission of B. burgdorferi ss. to the ticks in our study area. It is however not yet clear which one.
The genospecies community composition in questing nymphs did not significantly differ between forest types, but tree species had a marginally significant effect on the community composition. There was a clear effect of tree species on the diversity of the genospecies communities, with more diverse communities in oak stands compared with pine stands. This increase in Borrelia genospecies diversity in the ticks from diversified forest communities can pose serious health risks. Borrelia burgdorferi ss. and B. garinii are most often associated with Lyme arthritis and neuroborreliosis, respectively, while an infection with B. afzelii commonly causes skin manifestations such as erythema migrans or acrodermatitis chronica atrophicans (Strle and Stanek, Reference Strle and Stanek2009). These conditions can be regarded as less severe, compared to damage of joints or the nervous system caused by B. burgdorferi ss. and B. garinii. This indicates that human health risk can increase with increasing host diversity, when the contribution to the genospecies community of the more ‘dangerous’ Borrelia genospecies increases.
Compared with the average home range of some host species, the forest stands we investigated are of relatively small size (Verkem et al. Reference Verkem, De Maeseneer, Vandendriessche, Verbeylen and Yskout2003) and often imbedded in a mosaic of stands of a different forest type. While ranging, a host species may cross different unsuitable habitats and spend some of its time there. As ticks are disseminated while being attached to their host, it is plausible that nymphs that have fed as a larva on a particular host species are occasionally found in a habitat that is not suitable for that host. Thus ticks, and so also typical host specific Borrelia genospecies, can occur in a forest type that is not the favourable habitat of the host. Hosts will, however, spend more of their time in their preferred habitat, so that the chance that a tick drops off in that habitat type instead of in an unfavourable habitat type is higher. Moreover, despite the small size of forest stands, Tack (Reference Tack2013) found that populations of bank vole and wood mouse differed between small neighbouring stands of different forest types, with different populations being composed of different individuals. Most larvae will thus probably be disseminated by their host to a forest stand that is of a favourable forest type for the host, and so the association between Borrelia genospecies and forest type remains reasonable. We believe, however, that our estimates of Borrelia genospecies prevalence and diversity are only conservative estimates. In a more homogenous landscape, with larger stands of the same forest type, the pattern in Borrelia genospecies occurrence we see now would be even clearer and the differences between the forest types even larger.
From our results we may conclude that the public health risk associated with Lyme disease in Europe will not only depend on the NIP of B. burgdorferi sl. or on DIN, but also on the prevalences of the distinct genospecies. Even if a more diverse host community causes a dilution effect to occur and the DIN declines, disease risk can increase if the prevalence of a more ‘dangerous’ genospecies increases. As already suggested by Kurtenbach et al. (Reference Kurtenbach, Hanincová, Tsao, Margos, Fish and Ogden2006), and mentioned above, the interaction between Borrelia, ticks and hosts in Europe appears to be much more complex than the situation observed in North America, because of the existence of multiple pathogenic and specialist Borrelia genospecies in Europe. We can now confirm the contrast between these two regions and suggest that increasing host diversity, or adding species to the host community, can increase or decrease the prevalence of individual genospecies, depending on the response of the associated host species. Most likely in Europe the prevalence of B. afzelii will decline with increasing forest community diversification due to a dilution effect on small rodents, while the prevalence of the more ‘dangerous’ pathogens B. garinii and B. burgdorferi ss. will increase. It is then still possible for a dilution effect to occur in Europe on the level of B. burgdorferi sl., when hosts that are added to the host community are less efficient in transmitting B. burgdorferi sl. infection to ticks than hosts already occurring in a species poor community (Ostfeld and Keesing, Reference Ostfeld and Keesing2000b ; LoGiudice et al. Reference LoGiudice, Ostfeld, Schmidt and Keesing2003), which decreases the chance that a tick feeds on a competent reservoir species. This scenario is similar to the trend we observed in our study. It is therefore not increasing biodiversity in itself that will decrease Lyme disease risk. The identity of the hosts that are added to the host community and their interactions with each other and the ticks will determine if a dilution effect will occur (Ostfeld and Keesing, Reference Ostfeld and Keesing2000b ; Randolph and Dobson, Reference Randolph and Dobson2012), both at the level of B. burgdorferi sl. and of the Borrelia genospecies. These new insights introduce an extra dimension to the DEH for Lyme disease in Europe and pose important implications for the role of biodiversity in disease ecology. This study therefore strongly emphasizes the need to consider the different Borrelia genospecies as distinct pathogens and to study the species assemblages of the host community in high risk areas to assess human disease risk.
ACKNOWLEDGEMENTS
The authors thank K. van Rooijen and H. Yilmaz for the technical laboratory assistance.
FINANCIAL SUPPORT
This work was supported by IWT-Flanders, the Institute for the Promotion of Innovation through Science and Technology in Flanders (to S.R.).
Appendix
Appendix A. Primers and probes used in the PCR analyses to detect Borrelia burgdorferi sl. and Borrelia genospecies infections
Appendix B. Proportions of Borrelia genospecies detected in the genospecies community in nymphs from the Campine region and the diversity (exponent of Shannon index) of the genospecies communities, per forest type (mean + s.d.)
