Discussion
Patterns of genetic variation often reflect spatial variation in gene flow which can be influenced in two important ways (Wang & Summers, 2013): Spatially separated populations may experience isolation-by-distance (IBD; Wright 1943) in which landscape barriers and geographical distances cause restricted gene flow; and isolation-by-environment (IBE; Wang & Summers, 2010) in which gene flow among populations inhabiting different ecological environments is limited either by selection against dispersers moving between them or by individual preference to remain in a particular environment due to local adaptation (Dobzhansky, 1937). IBE predicts a correlation between genetic divergence and environmental dissimilarity because greater environmental differences between populations are expected to be associated with stronger divergent selection and reduction in the success of dispersers (Crispo et al., 2006; Lee & Mitchell-Olds, 2011). Of course, geographical and environmental isolation do not exclude each other, and spatial genetic divergence can be associated to gene flow reduction due to both geographical and ecological factors (e.g., Coyne & Orr, 2004; Crispo et al., 2006; Thorpe et al., 2008). Furthermore, population genetic theory predicts that genetic distances among individuals will increase with increasing geographical distance (Allendorf & Luikart, 2007).
Our investigation of the genetic vs. geographic distances within species across multiple guilds of different ecological traits and taxa (95 Coleoptera families) revealed that there are only few cases in which this relationship resulted to be significant and thus predictable, i.e., for which an increase in geographic intraspecific distances was followed by an increase in genetic distances, or the other way around, and most of all, showing a sufficient sampling (in terms of number of samples to examine) to support such trends statistically. However, observed patterns of infraspecific genetic and geographical distances among most of the central European Coleoptera species and ecological guilds examined were neither uniform nor entirely different among each other. Thus, the unlikely hypothesis that all species increase their genetic diversity with the distance of their record, which could be interpreted as a signal of gradual dispersion and genetic differentiation in progress (Allendorf & Luikart, 2007), could be not universally confirmed, despite the study area suffered an almost entire biodiversity loss during the Pleistocene and was reoccupied afterwards from external founder populations (e.g., Hewitt, 2000; Hofreiter & Stewart, 2009; Abellán et al. 2011; Birks & Tinner, 2016).
Besides limited sampling, cases, in which species did not show a positive relation between genetic and geographical distances, might be explained in two ways: 1) the geographic expansion of the species was not followed by an equal genetic diversification (intraspecific DNA distances are smaller than geographic distances), i.e., occurring in species with high dispersal capabilities with continuous genetic mixing and in a well interconnected area. Or, 2) the genetic diversification in the study area was independent from the geographic scale, occurring mainly in species with a potentially different phylogeographic origin of their populations. It is well known that Central Europe experienced post-glacial recolonization events from different Mediterranean and extra-Mediterranean refugia located all across the continent (e.g., Ahrens et al., 2013; Kühne et al., 2017, Schmitt & Varga, 2012). Freijeiro & Baselga (2016) suggested that dispersal-based processes in European beetles were probably taxon-dependent, but also depended on dispersal ability and ecological traits (Gómez-Rodríguez et al., 2015). Although patterns appear not very clear due to widely lacking significance, we here found several patterns in genetic/geographic relationships among ecological preference classes which might fit more ecology-dependent dispersal and differentiation (Papadopoulou et al., 2008). Indeed, causalities are expected to depend both on environmental and ecological processes in the species range. The distribution area as well as the relation genetic distance vs. geographic distance of a species depends on several factors: the paleo-biogeographic and biogeographic history (i.e., glacial expansion dynamics, glacial refugia presence, postglacial climatic gradients, and postglacial species expansion) (Stewart et al., 2010) could have been the major cause of such genetic trend in diversification. It is widely accepted that present distribution patterns in Central Europe are related to post glacial recolonization dynamics (e.g., Schuldt & Assmann, 2009; Schmitt, 2009) beside other also important factors (Baselga et al., 2012). In this context, geographic, climatic and ecological exogenous factors (i.e., climatic gradients, habitat fragmentation or presence/absence of corridors) and ecological endogenous factors (i.e., potential niche or dispersal capabilities due to physiological properties, level of adaptation) play a crucial role to determine these patterns (e.g., Schmitt et al., 2009; Rundell et al., 2009). So far, distribution patterns and genetic differentiation have been studied for mainly selected cases in the framework of phylogeographic studies (taxa or/and study sites; e.g., Múrria et al., 2020; Garcia-Raventós et al., 2021; Domènech et al., 2022), while only some studies with wider taxonomic and geographical scope exist (e.g., Baselga et al., 2013, 2015; Joly et al., 2014; Fujisawa et al., 2014; Dapporto et al., 2019). With the upcoming barcode data, a vast amount of data is becoming available to address such questions routinely at large scale, and to uncover particularly responses at population level regarding many ecological and climatic factors which have so far been explored with limited systematic sampling.
Here, deeper going conclusions lack statistic support since barcode data/ libraries are generally not designed to explore phylogeographic patterns in the context of species ecology. At this stage we expect sampling bias since data were generated with the scope of collecting and barcoding as many species as possible for future species identification. However, the amount of available data on central European beetle species was good enough to start to enquire the relationships between ecological properties of the species and their intraspecific patterns of genetic differentiation and to look at patterns that go beyond a single guild or species group (Baselga et al., 2013, 2015; Fujisawa et al., 2014). In fact, we faced severe problems due to the available number of sampling localities and of individuals per species. Therefore, we included all the sampling variables in the linear models excluding all the species with poor number of specimens and sampling sites. Furthermore, we investigated for the role of the sampling variables, such as number of individuals per species and number of localities, on the final results of statistic and significance scores as well as for an eventual implication of geographic distances of arbitrary chosen sample sites.
PCA and NMDS techniques captured different information compared to the Mantel test. Thus, the PCA technique was the least efficient in describing the relation of genetic and geographic distances in terms of amount of resulting significant species, followed by NMDS method and Mantel test. Results of both were similar but partially different from those obtained from Mantel test (Figure S5). This discordance limited more general conclusions. It is likely that the different algorithms behind the analytical techniques behaved differently in the presence of high level of noise in the data caused by spatial autocorrelation (Diniz-Filho et al., 2003; Legendre et al., 2015; although not tested here) and lacking sufficient geographical sampling (due to financial constraint of the Barcoding initiative, which did not allow higher samples numbers). Because different methods may emphasize different aspects of the data, using different data analyses techniques (Figure 2) may reveal more aspects of the data structure than a single method (Kenkel & Orloci, 1986). The Mantel test was considered to handle the limited number of available samples per species best which was disadvantageous for the ordination technique, which better read and converted the data matrices in more readable and efficient row data (Legendre et al., 2015) for the further Procrustes analysis. Ordination techniques are known to work better when dealing with big amount of data. On the other hand, minimal sampling size in our data was below the generally suggested amount to robustly investigate phenomena depending on spatial scale (at least 20 sampling localities; Dale & Fortin, 2014). Being first applied in population genetics by Sokal (1979), the Mantel test is currently one of the most commonly used methods to evaluate the relationship between geographic distance and genetic divergence (Mantel, 1967; see Manly, 1985, 1997; Diniz-Fhilo et al., 2013) – despite recent controversy and criticism about its statistical performance (e.g., Harmon & Glor, 2010; Legendre & Fortin, 2010; Guillot & Rousset, 2013; Castellano & Balletto, 2002) and the existence of more sophisticated and complex approaches to analyze spatial multivariate data (Diniz-Fhilo et al., 2013). In our case study, the mean number was only five sampling localities per species. Even though ordination methods are better suited, less prone to type I error and better in describing patterns (Legendre & Fortin, 2010; Legendre et al., 2015; C. Wang et al., 2010; I.J. Wang et al., 2013), results were not congruent with those of the Mantel tests. Nevertheless, PCNM methods combined to genetic information should be considered an alternative to the Mantel test and further analysis on a richer dataset could then possibly lead to clearer ecological conclusions.
It is known that the occupied habitat type has significant effects on both extent of the species range and latitudinal distribution (Ribera & Vogler, 2000; Hof et al., 2006, Fujisawa et al., 2014). This extends by some aspects the results of Fujisawa et al (2014) who found infraspecific genetic variation of COI in water beetles positively correlated with occupancy (numbers of sites of species presence; i.e., a similar but not identical measure to geographical distance) and negatively with latitude, whereas substitution rates across species (which we did not examine here) was influenced mainly by habitat types; specialized species of more stable environments, such as running water, had the highest rate. Baselga et al. (2015) expected dispersal to be high in aquatic beetles (of standing waters) because of the need for movement between ephemeral water bodies, while dispersal of leaf beetles do not require long-range movement for population persistence due to more stable conditions in vegetation. This is also reflected by our findings for species using vegetation as habitat. Our data thus seem to confirm the habitat stability hypothesis (Ribera et al., 2003) which sees in Pleistocene glacial events and the following climatic stability the major causes in producing equilibrium conditions, either with environmental factors due to niche-based processes or with spatial distributions from long-term stochastic dispersal.
Our data suggested higher dispersal tendencies and lower infraspecific variation of mtDNA for more ephemeral food resources (dung, dead animals), or habitats (fungi/ mushrooms). However, low number of species in these guilds and a similar pattern for eurytopic species (Tab. 1) might indicate that this observed pattern could be also a result of sampling bias. Specimens’ body size does not provide an answer to this question, as generally divergent patterns of infraspecific genetic vs geographical distances between smaller (x_s, s, m) and larger species (l, x_l) (Figure 3) are contrasted by rather uniform correlation statistics between the size classes (Figure 4). Studies on ground beetles have shown a generally higher genetic diversity across larger species independent from their sample site distance (Assmann et al., 2010; Schuldt & Assmann, 2011) which explain this pattern by lack of interconnection among populations due to their very specific habitat requirements, the habitat quality, and respective morphological adaptations (e.g., wing reduction; Jelaska & Durbešić, 2009). According to Freijeiro & Baselga (2016), the presence or absence of wings is an important factor for a better understanding of the geographical/genetic scale relationship. Indeed, habitat fragmentation is considered a major factor limiting gene flow in ground beetle populations (Liebherr, 1988).
Our results, beside yet still enormous sampling gaps and underrepresentation of many species, indicate that ecological niches and preferences may play a major role in geographic dispersal and genetic differentiation within species even though we did not consider here environmental, climatic factors or longitudinal\latitudinal gradients which are also known to have a fundamental role in explaining population dynamics (Rosindell et al., 2011; Baselga et al., 2013, 2015; Frejeiro & Baselga, 2016). These results can be extremely helpful to further develop conservation strategies, from a simple species conservation approach towards the conservation of genetic diversity in habitats or landscapes (e.g., Hedrick, 2001; Vellend et al., 2014). Therefore, further molecular screening would be needed, with particular focus on more geographical sampling to cover more in detail the genetic variation within the study area and to uncover causalities of such patterns (i.e., extending the Barcoding towards population level). Indeed, our results showed: an increase in number of sampling localities was usually followed by a related increase in statistic score and thus increase the explanatory power of barcode data to explain infraspecific genetic patterns among different ecological guilds.
The screening at the diversity patterns of the entire entomological fauna in such vast territory as Central Europe is a demanding task which request efficiency and great sampling efforts, but in the light of the emergency of current trends of insect decline (e.g., Hallmann et al., 2017; Wagner et al., 2021) it becomes an important issue for deeper understanding of its causes. The German Barcoding of Life campaign (Hendrich et al., 2015; Rulik et al., 2017) and resulting database contributed to resolve these issues. Nevertheless, a denser geographic sampling that will also result from future monitoring studies or metabarcoding projects will enhance the number of sampled localities and specimens, while concerted actions would be desirable. This will strengthen statistic results and allow bolder conclusions regarding biodiversity in a study area rather than simple species numbers.