Introduction

Understanding the evolution of biological communities is impossible without studying the variation in the composition and structure of these communities across space and/or over time. This variation is commonly defined as beta-diversity (Whittaker 1960, 1972) and represents one of the most important concepts in modern ecology. Examining patterns of beta-diversity in various communities is crucial not only from a purely theoretical perspective but also for application to subjects such as biological conservation and public health. Among these studies, investigations of beta-diversity drivers take a central place because they allow us to understand the mechanisms behind patterns of similarity or dissimilarity in species assemblages between locations and times and thus give insights into the processes shaping these assemblages.

Originally, patterns of biological diversity, including beta-diversity, were considered based on the number and identities of species (e.g., Whittaker 1960; Wilson and Shmida 1984). Biological diversity has only been recognized somewhat recently as not limited to the mere number and composition of species but as also encompassing the diversity of phylogenetic and functional entities (Tilman et al. 1997; Webb et al. 2002; Ackerly 2003; Cavender-Bares et al. 2009; De Bello et al. 2010). Indeed, species assemblages can be envisaged as assemblages of sets of certain traits. Some of these traits are inherited from common ancestors, whereas other traits arise de novo in the individual species. Compositional, phylogenetic and functional aspects (= facets) of diversity are interconnected but not equal. Consider, for example, two communities with the same number of species, but one community is mainly composed of closely related species, while the other community is mainly composed of distantly related species. Obviously, these communities will be similar in their compositional diversity but substantially different in their phylogenetic (and likely functional) diversity.

Few would deny the role of environment in shaping biological communities and species interactions therein. From the pioneering work of MacArthur and Levins (1967) onward, a plethora of studies have demonstrated environmental effects on species diversity. Furthermore, given that environment represents a combination of multiple factors including local environmental variables, as well as climate and spatial gradients, the importance of various environmental drivers may differ between compositional, phylogenetic and functional facets of diversity. However, recent studies testing the effect of environment on different biodiversity facets produced contrasting results. For example, Soininen et al. (2016) demonstrated that while the compositional diversity of stream diatoms was best explained by spatial variables, their functional diversity (in terms of guild richness) responded most strongly to environmental gradients. In contrast, Heino and Tolonen (2017) and Rocha et al. (2018) reported a similar response of the compositional, taxonomic (as a proxy for phylogenetic) and functional diversity of freshwater macroinvertebrates to local environmental and spatial variables. This suggests that the pattern of the relationships between various environmental aspects and multiple facets of diversity may manifest differently in different taxa. In particular, the effects of environment on beta-diversity may differ between free-living and parasitic species.

Similarly to free-living species, parasite communities are affected by local environmental and spatial factors (Poulin 2007; Krasnov 2008; Krasnov et al. 2019a). However, the environment for a parasite includes not only its surrounding abiotic and biotic factors but also the most important component of its environment—its host. Hosts provide parasites not only with food but also with places to live and reproduce and often serve as tools for parasite dispersal (Poulin 2007). As a result, parasite diversity has been repeatedly shown to respond to host diversity (Poulin 2007; Krasnov et al. 2007; Kamiya et al. 2014), although environmental effects on parasite diversity may also be strong (e.g. Poulin 2007; Adlard et al. 2015). Furthermore, strong associations between host and parasite phylogenies, as well as between host and parasite traits, have also been reported (Hadfield et al. 2014; Krasnov et al. 2016). Consequently, the phylogenetic and functional facets of parasite diversity can be affected by variables describing the phylogenetic and functional aspects of host species composition. In particular, this may be the case for parasite beta-diversity. Indeed, the association between host and parasite phylogenetic turnovers has been recently reported by Clark et al. (2018).

Here, we studied the effects of variation in off-host and host-associated environmental factors (the former will henceforth be referred to as environmental factors) on variation in the compositional, phylogenetic/taxonomic and functional beta-diversity of two taxa of ectoparasitic arthropods (fleas and gamasid mites) exploiting small mammalian hosts in the Palearctic, while accounting for spatial structure in all these beta-diversity facets by introducing spatial variables into the analyses. Fleas are holometabolous insects with obligatory haematophagous imagoes and non-parasitic larvae. All Palearctic species alternate between periods spent on the bodies and the nests/burrows of their hosts, although the proportion of time spent on the hosts’ bodies and in their burrows differs among flea species (see below). Pre-imaginal stages develop off-host (see Krasnov 2008). Gamasid mites demonstrate a vast range of lifestyles and feeding strategies, from endoparasitism via obligate and facultative haematophagy to predation on small invertebrates and phoresy (Radovsky 1985). Here, we focused on mites collected from host bodies that were either obligatory or facultative haematophagous. The latter spend most of their life off-host, whereas the former either are nidicolous and attack the host only to obtain a bloodmeal or prefer to stay and even to reproduce on host bodies. We asked whether (a) the roles of environmental, host-associated and spatial variables as drivers of beta-diversity differ among its multiple facets and (b) the effects of environment, hosts and space on beta-diversity differ between the two ectoparasite taxa.

Differences between biological communities (i.e. beta-diversity) can be partitioned into components that result from contrasting underlying processes, namely species replacement and species losses or gains (Harrison et al. 1992; Williams 1996; Lennon et al. 2001; Baselga 2007, 2010; Cardoso et al. 2014, Legendre 2014). Furthermore, the processes of replacement and loss/gain can be applied not only to species but also to phylogenetic lineages or traits. In other words, the partitioning of beta-diversity into additive components portraying different processes can be applied not only to separate species but also to separate phylogenetic lineages (for phylogenetic beta-diversity) or traits (functional beta-diversity). In this study, we used the approach of Cardoso et al. (2014) and partitioned the total beta-diversity (βtotal) into its species replacement (= turnover, βrepl) and richness difference (βrich) components. Then, we tested for the effects of environmental, host-associated and spatial variables on each of these components for each beta-diversity facet and each ectoparasite taxon.

Materials and methods

Ectoparasite and host distribution

The data on flea and gamasid mite distribution across the Palearctic were extracted from published surveys and unpublished data (for mites) that reported a given flea or mite species recorded on a given host species. It is well known that estimation of parasite species composition can be inaccurate for small samples (Poulin 2007). To avoid this inaccuracy, we considered data from surveys that aimed to study the entire flea or mite community harboured by the entire community of small mammals in a given region. Then, we selected surveys in which (a) 10 or more individuals of each host species were examined and (b) at least 500 individuals belonging to all small mammal species were captured. Parasitological examination of 10 conspecific individuals has been proved to allow reliable estimation of ectoparasite species richness and composition (Krasnov et al. 2005). In total, we used data on 202 flea and 69 mite species collected from 135 and 68, respectively, small mammal species (rodents, shrews, moles and pikas) across 45 and 30 regions, respectively (see map in Fig. S1 of Supplementary Material). For fleas, these regions were distributed from 69.80° in the North to 38.14° in the South and from 16.94° in the West to 161.21° in the East. For mites, the sampled regions were distributed from 97.74° in the North to 39.61° in the South and from 11.48° in the West to 161.21° in the East. Details on sampling methods and respective references can be found elsewhere (Krasnov et al. 2015 for fleas and Krasnov et al. 2010a for mites). In addition to 28 regional surveys referenced in Krasnov et al. (2010a), we used data reported for Norway (Edler and Mehl 1972), while the data for Tomsk region (Davydova and Belova 1972) were divided into two datasets (the middle and the southern boreal forests).

Flea phylogeny and mite taxonomy

As the backbone of flea phylogeny, we used the most recent molecular phylogeny of the order Siphonaptera (Zhu et al. 2015). This tree comprised almost all genera (but not all species) from our dataset. To determine the positions of species from our dataset that were not represented in Zhu et al. (2015), we used either their morphologically derived taxonomic affinities or additional molecular and/or morphological phylogenetic studies (see Krasnov et al. 2019b for references).

No phylogenetic tree is available for gamasid mites except for the tree of Dowling and OConnor (2010). However, this tree is poorly resolved and does not comprise the majority of the species from our dataset. Consequently, we used the Linnean taxonomic tree with four above-species hierarchical levels (subgenus, genus, subfamily, family) as a proxy for phylogeny. Mite taxonomy was taken from various taxonomic sources (Bregetova 1956; Vinarski and Korallo-Vinarskaya 2016, 2017). Sixty-nine mite species considered in this study were represented by four families, five subfamilies, eight genera and nine subgenera.

Flea and mite traits

Traits for both ectoparasite taxa included (a) mean abundance on the principal host; (b) the degree of host specificity in terms of the numbers of hosts exploited in a region (corrected for the confounding effect of the sampling effort); (c) phylogenetic diversity of a host spectrum within a region or (d) across all regions where a given ectoparasite species was recorded; (e) body size; and (f) the degree of sexual dimorphism. The rationale and details of the calculations of the mean abundance and variables describing the degree of host specificity can be found elsewhere (Poulin et al. 2011; Krasnov et al. 2013, 2015; Surkova et al. 2018b). In brief, we substituted the original values of the number of hosts exploited by a flea or a mite in a region with the residuals of the regression of their log-transformed values against the log-transformed number of hosts examined within this region. We then averaged these values across regions for each flea or mite species. Phylogenetic host specificity was measured as the Δ+ index of Clarke and Warwick (1999) and Warwick and Clarke (2001) modified to take into account phylogenetic rather than taxonomic distances and calculated using the function “taxondive” of the package “vegan” (Oksanen et al. 2018) implemented in the R 3.5.1 Statistical Environment (R Core Team 2018). The construction of the phylogenetic tree of hosts used to calculate phylogenetic host specificity is described below. The body size of an ectoparasite species was estimated via either maximal body length (fleas) or the midline length of the dorsal shield (mites). Details of body size measurements are described in Krasnov et al. (2013) and Surkova et al. (2018a, b). Due to substantial sexual size dimorphism in both fleas and mites, we used the median of the average male and average female body size as a characteristic of the body size of each ectoparasite species. The degree of sexual dimorphism was estimated as the logarithmic female-to-male size ratio (see Smith 1999 for rationale and explanations).

In addition, we characterized each flea species by (a) seasonality (whether the main reproductive period was during the warm season or the cold season or year round) and (b) microhabitat preferences. The latter trait is associated with the proportion of time a flea spends on either the body of its host (“body species”) or in its burrow (“nest species”) or equally (“both body and nest species”). Detailed explanations and references for both traits can be found elsewhere (see Krasnov 2008 and references therein). In gamasid mites, a particular feeding mode is a characteristic species trait (e.g. Radovsky 1985). We designated each mite species as (a) an obligate exclusive haematophage (feeds solely on the host’s blood), (b) an obligate non-exclusive haematophage (feeds on both the host’s blood and small nidicolous arthropods) or (c) a facultative haematophage based on information from Bregetova (1956) and Zemskaya (1973).

Environmental variables

We determined the latitudinal and longitudinal positions of the centre of each region with ArcGIS 10.6. Environmental variables for each region included altitude (mean, maximal and minimal), amount of green vegetation [normalized difference vegetation indices (NDVI)], four variables describing air temperature (mean, maximum, annual and monthly ranges) and precipitation. NDVI and precipitation were calculated separately for autumn, winter, spring and summer. The sampling area of each region was calculated using data from an original source. Environmental variables for a region were averaged within this area across 30 arc-second grids (ca. 1 km resolution). Data on altitude were obtained using ArcGIS 10.6; data on NDVI were obtained from the VEGETATION Programme (http://free.vgt.vito.be), whereas data on air temperature and precipitation were obtained from the WORLDCLIM (BIOCLIM) 2.0 package (Fick and Hijmans 2017). The majority of environmental variables correlated with one another (r = 0.33–0.92, p < 0.05 for all). To avoid multicollinearity, we applied principal component analyses (PCA) of environmental variables separately for flea and mite data and substituted the original values with the scores of the three (for fleas) or four (for mites) first principal components, which were then used as environmental variables (E1–E3 or E1–E4, respectively). These principal components explained 76.02% and 85.76% of environmental variation for flea and mite data, respectively. For flea data, E1 represented mainly an increase in summer precipitation and summer and autumn NDVI, and E2 correlated negatively with altitude, whereas E3 reflected an increase in mean and maximal air temperatures and spring and winter NDVI (see Table S1 in Supplementary Material). For mite data, E1 was associated mainly with an increase in maximal air temperature and E2 with an increase in mean air temperature, precipitation (fall, spring and summer) and a decrease in the annual range of air temperature, while E3 reflected an increase in mean and maximal altitude, and E4 correlated positively with NDVI (see Table S2 in Supplementary Material).

Host-associated variables

The host-associated environment was characterized from the perspective of host species composition, host phylogenetic diversity and host trait diversity (see data analyses) separately for flea and mite data. We took the data on host species composition from the original sources on ectoparasite species composition. In other words, we considered only those host species on which fleas or mites were recorded because hosts that do not support ectoparasite communities are unlikely to affect their diversity. We transformed host species composition into continuous predictors for further analyses as follows. First, we constructed distance matrices (separately for flea and mite hosts) of pairwise community dissimilarity between regional host assemblages using the Raup-Crick dissimilarity index (Raup and Crick 1979) with the function “vegdist” of the R package “vegan”. Then, we applied principal coordinate analyses on these distance matrices using the function “cmdscale” of the R base package “stats” and extracted five (Hc1–Hc5) vectors for the composition of host species harbouring either fleas or mites. These vectors explained 71.25% and 71.22% of the variation and were thus continuous variables representing differences in host species composition for fleas and mites, respectively, among regions.

To characterize the host-associated environment from the perspective of host phylogenetic diversity, we used the global phylogenetic supertree of Bininda-Emonds et al. (2007) and established the topology of branches within orders, families and genera using various sources (see references in Krasnov et al. 2019b) (we also used this tree calculating the degree of phylogenetic host specificity of ectoparasites; see above).The lengths of all branches were set to an equal length of 1, and the tree was arbitrarily ultrametricized using the program Mesquite 3.51 (Maddison and Maddison 2018). Then, we constructed distance matrices (separately for flea and mite hosts) of pairwise phylogenetic community dissimilarity between regional host assemblages (Ives and Helmus 2010) using the R packages “ape” (Paradis et al. 2004) and “picante” (Kembel et al. 2010). We applied a principal coordinate analysis to these distance matrices as described above (i.e. using function “cmdscale”) and extracted five vectors of host phylogenetic composition (Hp1–Hp5) for either flea or mite hosts. Similarly to host species composition factors, the vectors of host phylogenetic composition represented among-region variation in the phylogenetic composition of hosts exploited by fleas and mites and explained 54.35% and 41.58% of this variation, respectively.

To depict the host-associated environment from the perspective of host traits, we characterized host species by quantitative and ordinal traits that could affect either flea or mite parasitism or the species composition and richness of ectoparasite assemblages, namely body mass, basal metabolic rate, skin thickness, hair density, geographic range, population density, degree of sociality, shelter depth and complexity, and hibernation pattern (see rationale, explanations and details in Krasnov et al. 2016). Trait values were extracted either from the PanTHERIA database (Jones et al. 2009) or from various sources (see references in Krasnov et al. 2016). To obtain variables representing among-region variation in host trait composition, we first constructed among-host trait distance matrices (separately for flea and mite hosts) using the Gower distance coefficient with the function “gowdis” implemented in the R package “FD” (Laliberté and Legendre 2010). From these matrices and host species × region matrices, we then built matrices of trait dissimilarities among regional host assemblages based on trait dissimilarities between host species using a functional dissimilarity measure based on a generalized Sørensen dissimilarity (Ricotta et al. 2016) with the function “PADDIS” implemented in the R package “adiv” (Pavoine 2018). These trait dissimilarity matrices were transformed into continuous predictors via principal coordinate analyses with the R function “cmdscale” (see above). This resulted in two (Ht1 and Ht2; for flea hosts) and three (Ht1–Ht3; for mite hosts) vectors reflecting among-region variation in the trait composition of host assemblages and explaining 64.45% and 69.32% of this variation.

Spatial variables

To obtain spatial variables for the main analyses (see below), we applied the approach of Moran Eigenvectors Maps (MEM) [= the principal coordinates of neighbouring matrices (PCNMs)] (Dray et al. 2006), which allows the evaluation of spatial structures over a range of scales with the first to the last eigenvectors representing ever-decreasing spatial scales (Borcard and Legendre 2002). We built a truncated Euclidean distance matrix from geographic coordinates of the regions (separately for fleas and mites) and then extracted eigenvectors with positive eigenvalues (Borcard and Legendre 2002). These spatial variables (further referred to as PCNMs) were extracted using the function “pcnm” implemented in the R package “vegan”. Then, we tested each PCNM for spatial autocorrelation using Moran’s I test implemented in the R package “ape” and retained for further analyses only the PCNMs that were characterized by a significant spatial autocorrelation (13 of 25 PCNMs for flea data and seven of 15 PCNMs for mite data) (Borcard et al. 2018).

Data analyses

We produced three dissimilarity matrices (βtotal, βrepl and βrich) based on each of the following: (a) flea or mite species incidence (compositional beta-diversity), (b) phylogenetic (for fleas) or taxonomic (for mites) information (phylogenetic or taxonomic beta-diversity, respectively) and (c) information on flea or mite species traits (functional beta-diversity) using the function “beta” implemented in the R package “BAT” (Cardoso et al. 2014, 2015, 2018). Average indices of compositional, phylogenetic/taxonomic and functional beta-diversity (βtotal) and its components (βrepl and βrich) were calculated using the function “beta.multi” from the package “BAT” separately for fleas and mites.

To test for the effects of the environment, host-associated factors and space on variation in each component of each of the three facets of flea and mite beta-diversity (compositional, phylogenetic/taxonomic and functional), we applied a distance-based redundancy analysis (db-RDA; Legendre and Legendre 1998; Legendre and Anderson 1999; Borcard et al. 2018). A distance-based redundancy analysis allows one to perform constrained ordinations on data using non-Euclidean dissimilarity measures. Given that our response matrices (pairwise total beta-diversity and its components) consisted of a matrix of distances between “objects” (pairwise between-region differences in species compositions and richness) which may be generated by non-linear measures, application of db-RDA allowed us to determine how well environmental, host-associated and spatial parameters can explain the variation among these “objects” in the matrix. Importantly, db-RDA allows testing for the effect of each separate factor or gradient on the variation in beta-diversity as well as for shared effects of any combination of these factors/gradients. A possibility of estimation of the pure and shared effects of separate explanatory variables represents one of the advantages of db-RDA over other methods of beta-diversity studies such as, for example, multiple regression on distance matrices (MRM; Lichstein 2007). Technically, db-RDA first applies a principal coordinate analysis on the response distance matrix and converts it back to a rectangular table containing rows of objects by columns of “principal coordinates”. This set of “principal coordinates” represent the dissimilarities in a Euclidean space and are then related to additional environmental parameters using a classical redundancy analysis which, in turn, is a method allowing extraction and summarisation of the variation in a set of response variables (i.e. resulted) principal coordinates as explained by a set of explanatory variables (i.e. environmental, host-associated and spatial gradients). In addition, using db-RDA allowed our results on drivers of beta-diversity of ectoparasitic arthropods being directly comparable with those obtained by Rocha et al. (2018) on drivers of beta-diversity of free-living insects.

In our analyses, variables describing host-associated environment were represented by (a) host species composition for compositional beta-diversity, (b) host phylogenetic composition for phylogenetic (for fleas)/taxonomic (for mites) beta-diversity and (c) host trait composition for functional beta-diversity. The db-RDAs were carried out with the function “capscale” from the R package “vegan” with the option “sqrt.dist” set to “TRUE” to correct for negative eigenvalues (Legendre 2014; Rocha et al. 2018). To understand the pure and shared effects of the off-host environment, host-associated environment and space on each of the three measures of compositional, phylogenetic/taxonomic and functional beta-diversity, we used variation partitioning via the function “varpart” from the R package “vegan”. This function partitions the variation in the response table (in our case pairwise dissimilarity matrix of beta-diversity) into components accounted for explanatory variables and their combined effects. Then we tested for the significance of the pure effects using the function “anova” of the “vegan” package. The latter function was also used to test the significance of separate explanatory variables (Oksanen et al. 2018). Prior to running db-RDAs, we selected the final sets of predictors using the function “ordistep” of the package “vegan” with 499 permutations. This function allows variable selection for constrained ordination using permutation p values (Borcard et al. 2018; Oksanen et al. 2018). The selection using “ordistep” resulted in the retention, for the main analyses, of only those variables which did not show collinearity, namely all principal components of environmental variables and all predictors associated with composition, phylogeny and traits of hosts, but only five (S1–S5) of 13 PCNMs for flea data and two (S1–S2) of seven PCNMs for mite data.

Results

In fleas, the βrepl component was higher than βrich independently of whether beta-diversity was determined by compositional, phylogenetic or functional dissimilarities (Table 1). In mites, indices of all facets of beta-diversity (including βtotal) were lower than those of fleas with the contribution of the βrepl component to the βtotal of either beta-diversity facet being approximately the same as that of βrich (Table 1).

Table 1 Indices of compositional, phylogenetic and functional beta-diversity and its fractions (i.e. species replacement and richness difference) of flea and gamasid mite communities harboured by small mammalian hosts in the Palearctic
Full size table

The distance-based redundancy analyses of flea beta-diversity demonstrated that variation in compositional, phylogenetic and functional βtotal, as well as the respective βrepl components, was significantly associated with variation in all three groups of factors (Table 2, Fig. 1). The same was true for functional βrich. Nevertheless, environmental variables taken separately from other predictors produced significant axes for phylogenetic βrich (Table 2). Furthermore, correlations between ordination axes for compositional βrich and one host-associated factor (Hc1) and one spatial variable (S1) were significant (p < 0.05 for both) (Fig. 1). The results of separate significance tests for each explanatory variable and correlations between individual predictors and the first two ordination axes are presented in the Supplementary Material (Tables S3–S5 and Tables S6–S8, respectively).

Table 2 Results of the significance tests of distance-based redundancy analyses of the total compositional (C), phylogenetic/taxonomic (P/T) or functional (F) beta-diversity (βtotal) of flea and mite communities and its components [species replacement (βrepl) and richness difference (βrich)]. Constrained intertia (ConstI) is the proportion of total variance uniquely explained by the explanatory variables (Borcard et al. 2018). The significance of all explanatory variables, as well as the pure effects of the off-host environment, host-associated environment (compositional, phylogenetic or trait-bases) and space, was tested using the function “anova” in the R package “vegan” (see text for explanations)
Full size table
Fig. 1
figure 1

Biplots of distance-based redundancy analyses of the compositional, phylogenetic and functional beta-diversity (total beta-diversity and its species replacement and richness difference components) of fleas. The length of each arrow is proportional to the correlation between the ordination and the respective variable. E, H (Hc, Hp and Ht) and S are environmental, host-associated (compositional, phylogenetic and trait-based, respectively) and space variables, respectively. Numbers denote the individual numbers of variables. Only variables significantly correlated with the ordination axes are shown

Full size image

The ordination axes of the entire model db-RDAs for the βtotal of all three facets of beta-diversity of mites and their βrepl components were significant (Table 2, Fig. 2). However, this was not the case for the pure fraction of host-associated variables and taxonomic beta-diversity (Table 2). Although the variation in compositional, taxonomic and functional βrich represented by axes of the entire model db-RDAs did not differ from random, the pure fraction of host-associated variables was significant for taxonomic and functional βrich (Table 2, Fig. 2). In addition, some individual environmental, host-associated and/or spatial variables correlated positively with ordination axes for the βrich of all beta-diversity facets (Fig. 2; see also Tables S11–S14 of the Supplementary Material).

Fig. 2
figure 2

Biplots of distance-based redundancy analyses of the compositional, taxonomic and functional beta-diversity (total beta-diversity and its species replacement and richness difference components) of gamasid mites. The length of each arrow is proportional to the correlation between ordination and the respective variable. E, H (Hc, Hp and Ht) and S are environmental, host-associated (compositional, phylogenetic and trait-based, respectively) and space variables, respectively. Numbers denote the individual numbers of variables. Only variables significantly correlated with the ordination axes are shown

Full size image

Venn diagrams based on variation partitioning and demonstrating variation in flea beta-diversity as explained by unique and shared effects of environmental, host-associated and spatial predictors are presented in Fig. 3. The largest fractions of variation in compositional βtotal and βrepl were almost equally explained by host-associated and spatial predictors. Variation in compositional βrich was jointly explained by all three groups of predictors with the unique effect of environment being important as well. Variation in host phylogenetic composition explained substantially more variation in phylogenetic βtotal and βrepl than other groups of predictors or their joint actions. Similarly to compositional βrich, variation in phylogenetic βrich was best explained by the shared contribution of all groups of predictors followed by the unique contribution of environment. The unique contribution of spatial variables appeared to be the most important in explaining variation in functional βtotal and βrepl, whereas variation in host trait diversity explained the largest fraction of variation in functional βrich.

Fig. 3
figure 3

Venn diagrams based on variation partitioning of the variation in the compositional, phylogenetic and functional beta-diversity of fleas as explained by unique and shared effects of environmental (E), host-associated (Hc, Hp and Ht for species composition, phylogeny and trait composition, respectively) and spatial (S) predictors. Negative fractions and zero (see Legendre 2008) are not shown

Full size image

The variation partitioning of the ordination of mite data is presented in Fig. 4. Host-associated predictors either alone or jointly with environmental predictors explained the largest portion of the variation in compositional βtotal, βrepl and βrich. In contrast, these predictors were the least important in explaining the variation in taxonomic beta-diversity. Instead, the most important drivers of the latter were environmental predictors, which also acted jointly with spatial variables in explaining most of the variation in taxonomic βrepl. The variation in functional βtotal and βrepl was best explained by the unique contributions of environmental predictors followed by host-associated variables, whereas the fractions representing the shared effects of either the two latter or all three groups of explanatory variables significantly explained the variation in the βrich component of this beta-diversity facet.

Fig. 4
figure 4

Venn diagrams based on variation partitioning of the variation in the compositional, taxonomic and functional beta-diversity of gamasid mites as explained by unique and shared effects of environmental (E), host-associated (Hc, Hp and Ht for species composition, phylogeny and trait composition, respectively) and spatial (S) predictors. Negative and zero fractions (see Legendre 2008) are not shown

Full size image

Discussion

The most important finding of this study is that the pattern of the relative importance of environmental, host-associated and spatial variables is widely variable between beta-diversity facets, their components and the two ectoparasite taxa. This is in contrast to the results of some, but not other, earlier studies of beta-diversity variation (Heino and Tolonen 2017; Rocha et al. 2018 vs Soininen et al. 2016, respectively). Contrasting results of ecological studies dealing with similar questions and carried out on different taxa or even the same taxon can, sometimes, be attributed to differences in analytical approaches (e.g. Poisot et al. 2017 vs. Krasnov et al. 2019a). Here, we intentionally followed the analytical methodology applied by Rocha et al. (2018) for freshwater macroinvertebrates, but the pattern of the relative effect of each group of predictors on each beta-diversity facet was found to be similar in Rocha et al. (2018) and different in our study.

Environmental effects

We found that environmental variation contributed mostly to variation in (a) the compositional and phylogenetic βrich of fleas and (b) the taxonomic βtotal, βrepl and βrich and functional βtotal and βrepl of mites. The effect of external environment (especially air temperature and moisture) on ectoparasite communities (species richness, species composition, relative abundances) has been repeatedly documented (e.g. Krasnov et al. 2010a, b for fleas; Vinarski et al. 2007 for gamasid mites). Obviously, this effect is realized via responses of individual species (Marshall 1981; Krasnov 2008; Walter and Proctor 2013) and thus is reflected by the species composition of an assemblage. However, the species composition of flea communities was earlier found to be, to a great extent, the result of filtering by host species composition, whereas the abiotic environment played a secondary role (Krasnov et al. 2015). Nevertheless, environmental variation did not adequately explain the variation in the compositional and phylogenetic βtotal of fleas, but did explain their βrich components quite well, thus reflecting the losses/gains of species or phylogenetic lineages. The effect of environment on compositional βrich can be explained by, for example, species loss due to harsh conditions in northern regions, whereas this effect on phylogenetic βrich can be associated with evolutionary/historical patterns of dispersal in different phylogenetic lineages (e.g. Medvedev 2014).

In contrast to flea communities that are strongly influenced by host-associated variables (Krasnov et al. 2015), ectoparasitic gamasid mites are highly sensitive to both environmental conditions and host species composition (Vinarski et al. 2007). For example, the similarity in mite assemblages has been found to decrease with both increasing environmental dissimilarity and increasing dissimilarity in host species composition (Vinarski et al. 2007). This is the most likely reason for the variation in the compositional beta-diversity of mites being linked with the variation in both environment and hosts. However, the variation in the taxonomic and functional beta-diversity of mites was almost solely explained by the variation in environmental, rather than host-associated, predictors (except for functional βrich). On one hand, a strong effect of environment on the taxonomic and functional beta-diversity of mites may arise because of the differential responses of mites belonging to different taxonomic groups or possessing different traits due to environmental factors. For example, the relationship between body size and environmental tolerance has been shown in some arthropod taxa (e.g. Kaspari et al. 2015), although this has never been experimentally tested in mites. Nevertheless, indirect observations implied that the linkage between body size and a certain degree of environmental tolerance may be the case for gamasids (Krasnov et al. 2013). On the other hand, the general lack of effect of the variation in the phylogenetic or trait composition of host communities on the variation in the taxonomic or functional, respectively, beta-diversity of mites can be explained by an extremely broad host generalism (both from the phylogenetic and trait-based perspective) of many mites either belonging to the same taxonomic group or possessing the same trait. For example, all Androlaelaps mites are obligate non-exclusive haematophages and are found on host species belonging to different genera, families, orders and even classes and differing in a variety of their ecological, morphological and physiological traits (Vinarski and Korallo-Vinarskaya 2017). Similarly, many (albeit not all) species of Haemogamasus are facultative haematophages recorded on multiple rodent and shrew hosts (Vinarski and Korallo-Vinarskaya 2017).

Host effects

Variation in the compositional, phylogenetic or trait structure of host communities was an important factor responsible for variation in (a) the compositional and phylogenetic βtotal and βrepl of fleas and the compositional βtotal and βrepl of mites and (b) the functional βrich of both taxa. Strong links between host and parasite species or phylogenetic lineage composition have often been reported (Krasnov et al. 2012; Dallas and Presley 2014; Maestri et al. 2017). This is not surprising because, from an ecological perspective, the species composition and phylogenetic structure of any community are determined by the availability of resources utilized by the species composing this community, while from an evolutionary perspective, the species of a given community either originated locally or immigrated. In parasite communities, the availability of resources is represented by host availability, as parasites cannot exist without exploiting their hosts, which has itself an important evolutionary component. As a result, the species composition of a parasite community is partly due to the identities of the host species and partly due to the evolution and historical biogeography of these hosts. Although parasite and host phylogenies are rarely fully congruent because of host switching and other co-evolutionary events (Poulin 2007), strong relationships between host and parasite evolutionary histories cannot be refuted (Hadfield et al. 2014). Therefore, parasite species and lineage turnover, but not species or lineage richness, appeared to be strongly influenced by host species and lineage turnover.

We found an effect of host phylogenetic composition on the phylogenetic beta-diversity of fleas, but not on the taxonomic beta-diversity of mites. One of the reasons behind this may be the availability of the phylogenetic tree for fleas but not for mites. The mite taxonomic tree used in our study was poorly resolved and contained only one superfamily, four families (among them Macronyssidae with one species recorded in one region only), five subfamilies and eight genera. It seems that this did not allow us to distinguish successfully among different mite assemblages by their taxonomic composition, so that mite taxonomic composition varied substantially more weakly along axes that portrayed host phylogenetic composition than along environmental gradients.

Variation in the trait composition of host communities was the main factor explaining the total functional beta-diversity, as well as its richness difference component, in both taxa. In addition, the effect of host traits was important as a driver of functional βrepl in mites. The link between the traits of a consumer species and a consumed species is well known for free-living organisms (Rezende et al. 2007; Spitz et al. 2014). Moreover, the importance of trait complementarity between parasites and their hosts for understanding the ecology and evolution of host-parasite interactions has been recognized (e.g. McQuaid and Britton 2013). Indeed, a study that specifically tested the association between parasite (fleas) and host (small mammals) traits reported that fleas possessing certain traits exploit mostly hosts possessing certain traits (Krasnov et al. 2015 for fleas). The complementarity of parasite and host traits is predominantly related to the resource acquisition by parasites, including both extracting the resource from a host and avoiding or minimizing the effect of its defence efforts. In haematophagous ectoparasites, the extraction of the resource (blood) can be achieved due to morphological complementarity between parasite mouthparts and host skin thickness and/or depth of cutaneous blood vessels. The host’s anti-parasitic grooming impairs ectoparasite feeding and causes their mortality (e.g. Hawlena et al. 2007). Consequently, ectoparasites exploiting hosts with higher grooming abilities are usually characterized by developed anti-grooming tools such as combs and setae (Marshall 1982; Krasnov 2008). These examples suggest that the disappearance of a host trait (represented by species possessing this trait) from a location may lead to a loss of a complementary trait represented by ectoparasite species. Similarly, a turnover of host traits may be coupled with a turnover of respective ectoparasite traits. These processes may underlie the effect of variation in the trait composition of host communities on variation in the functional beta-diversity of parasites.

Spatial effects

Spatial variation, jointly with host-associated or environmental factors, was the main driver of the compositional and functional βtotal and βrepl of fleas, while space played a minor (if any) role as a driver of mite beta-diversity. The lack of any substantial effect of space on either facet of mite beta-diversity suggests that, in general, biodiversity of this taxon is scale-invariant. The reason for this can be that mites use their hosts not only as a food source but also as an important tool for dispersal (Radovsky 1985; Walter and Proctor 1999). As a result, and despite their limited abilities for active dispersal, the passive dispersal of mites (i.e. phoresy on various mammals and even birds and insects) allows them to overcome long distances. In other words, mites are broadly spread by their hosts, so that their beta-diversity (= dissimilarity among locations) does not depend on distance between locations (see also Vinarski et al. 2007).

By contrast, the contributions of spatial variables were important in explaining the variation in flea beta-diversity, especially in its species replacement component. This suggests that flea assemblages are spatially structured with this structure manifested mainly at the large scale (because the PCNMs that correlated significantly with ordination axes were the first three or four; Tables S6–S8 of Supplementary Material). In part, this is because of the great spatial extent of our study (mean and maximal distance between pairs of regions were 2880.5 km and 7939.3 km, respectively). Biologically, this shows substantial differences in species composition but not species richness between flea assemblages situated far away from each other. For example, flea assemblages of both south-eastern Azerbaijan and the Mongolian Central Khangay were composed of the same number of species (21), but there were no shared species between these assemblages. In contrast, although the flea communities of Slovakia and Poland differed slightly in their species richness (24 and 20, respectively), 19 species were present in both regions.

In conclusion, the pattern of the relative effects of environmental, host-associated and spatial factors on beta-diversity is context-dependent and may differ among different facets of beta-diversity and their different components. This suggests that compositional, phylogenetic/taxonomic and functional beta-diversity can be governed by different rules (e.g. Soininen et al. 2016), similarly to the case of alpha-diversity (Krasnov et al. 2010a, b). In addition, this pattern may also differ among communities of different taxa depending on their biological affinities.