Ongoing Speciation and Gene Flow between Taxonomically Challenging Trochulus Species Complex (Gastropoda: Hygromiidae)

Geographical isolation, selection and genetic drift can cause the geographical diversification of populations and lead to speciation. Land snail species in the genus Trochulus show overlaps in geographical ranges as well as in morphology, but genetic data do not always support the species-level taxonomy based on morphological characters. Such a group offers an excellent opportunity to explore the processes involved. We have addressed the problem by determining the status of the restricted endemic T. graminicola within the larger context of Trochulus taxonomy. We used an integrated approach based on morphological features, ecological preferences and two molecular markers: mitochondrial COI sequences and microsatellites. Comparison of these results demonstrated: (i) conchological distinction of T. striolatus and T. sericeus; (ii) anatomical, ecological and genetic differentiation of T. graminicola and (iii) concordance between morphological characters and mtDNA markers in T. striolatus. Moreover, our data showed an intricate evolutionary history within the genus Trochulus, which can be best explained by: (i) recent or ongoing gene flow between taxa or (ii) their large ancestral polymorphism. Both of these hypotheses suggest that diversification within this group of snails has occurred relatively recently. The mismatches between species defined on morphology and on molecular genetics indicate the complexity of the processes involved in the diversification of this genus.


Introduction
Determining the causes of difficulties in delimiting species in many taxa has proved hard. To explain the speciation process, different evolutionary mechanisms have been proposed such as: rapid and recent phylogenetic divergence, introgression, high phenotypic plasticity, ongoing differentiation of varieties and partial barriers to gene flow between ecotypes [1]. The uncertainty and controversy about species recognition are mostly related to four different a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 morphology among Trochulus species and a significant influence of the environment on variation in their shell size and shape, which have led to the assignment of extremely distinct morphological forms to the same species [29]. Hence, morphological characters alone may be an unreliable means of distinguishing species.
Using both molecular (based on mtDNA and microsatellites) and morphological (based on shell and genitalia) methods, we revise the taxonomic status of T. graminicola and assess whether it is a true species according to typological and phylogenetic species concepts. Combining molecular, morphological and ecological data, we place this species into the larger context of Trochulus taxonomy and test its relationship with closely related and morphologically similar taxa, i.e. T. coelomphala, T. hispidus, T. sericeus and T. striolatus inhabiting adjacent areas. Additionally, we analyse several samples collected from more distant regions (Poland and Great Britain). Finally, we compare the morphological recognition of taxa with their clustering in phylogenetic trees based on mtDNA. Since the morphological identification of these species is not always certain [28,31,34], we provisionally assumed their classification provided in malacological literature based on the morphological traits of shell and genitalia [21,32,[35][36][37][38] to check the consistency of morphological and molecular approaches. Moreover, we examine the correspondence between mitochondrial and microsatellite markers to search for possible cases of introgression.

Sample collection, niche description, shell and genital morphometry
The complete list of sample sites and specimens, collected in three seasons between July and August 2011-2013 or analysed elsewhere [28,29], is shown in Table 1 and Fig 1. All the material is deposited in the Museum of Natural History in Wrocław, Poland. Trochulus graminicola was found only at the type locality with a limited number of specimens, and most of the other Trochulus populations came from adjacent areas (Fig 1). Snails from Zieleniec and Hayley Wood, determined as T. plebeius by Wiktor [39] and Paul [40], were assigned by us to T. sericeus and we use the name T. plebeius/sericeus throughout the paper. We use this terminology because the name T. plebeius should rather be reserved for similar forms from the French and Swiss Jura [29]. In addition, vegetation and dominant plant species were recorded at each site and habitat specified (Table 1). For precise niche descriptions of T. graminicola, plant nomenclature follows Oberdorfer [41].
All 397 shells were measured in standardized views [25] by the same person (MP) with a graduated eyepiece in a stereomicroscope and an accuracy of 0.1 mm. Since the systematic measurement error (with 1% error probability) does not compromise results [26], one measure of each variable on each specimen was considered sufficient. Only adult individuals (as defined in [28]) were considered. Finally, basic statistical parameters were calculated.
Each individual was measured from the side perspective for shell height (H), shell width (W), aperture height (h), aperture width (w) and body whorl height (bwH). From underneath, umbilicus major diameter (U) (i.e. the longest diameter parallel to the shell diameter, D), umbilicus minor diameter (u) (i.e. perpendicular to the umbilicus major diameter) and shell diameter (D) were taken. Finally, the number of whorls (whl) was counted according to Ehrmann's method [42]. In addition, the following coefficients of shell proportions were calculated: height/width ratio (H/W), relative height of body whorl (bwH/H), relative umbilicus diameter (U/D) (umbilicus major diameter/shell diameter ratio), ratio of umbilicus minor to its major diameter (u/U) and height aperture/width aperture ratio (h/w).
For genital morphological examinations, 114 mature snails (Table 1) were dissected and seven measurements of genitalia were taken: the length of flagellum (fl), epiphallus (ep), penis  (p), bursa copulatrix (sl), bursa duct (sd), upper vagina (uv) (i.e. distance between outlet of mucous glands and tips of inner dart sacs) and the width of bursa copulatrix (sw). The relative length of inner to outer dart sacs (is/os) was also recorded by measuring the distance between the tips of inner and outer dart sacs. Then the following proportions were calculated: flagellum/epiphallus (fl/ep), epiphallus/penis (ep/p), bursa duct length/bursa length (sd/sl) and bursa width/length (sw/sl), and included in statistical analyses. Measurements were used in canonical discriminant analysis (CDA). To reduce the number of highly correlated predictors, we performed principal component analysis (PCA) using a correlation matrix and used the first two principal components in further studies. We also computed pairwise Pearson correlation coefficients involving all variables and next reduced the number of the most correlated parameters. Based on these results, six shell variables (W, H, D, w, U and u) were removed because they were characterized by the largest mean absolute correlation. We assumed a pairwise absolute correlation cut-off of 0.9 and re-evaluated the average correlations at each step of the elimination. These procedures allowed us to reduce the number of redundant variables and leave the best predictors describing differences between the examined morphospecies. Variables contributing most to CDA, i.e. the most promising in species distinction, were further used in Kruskal-Wallis non-parametric analysis of variance (ANOVA). For statistical analysis of the data, Statistica 10 software (StatSoft, Inc. 1984-2011) was used. The elimination of correlated parameters was made by findCorrelation from the caret package in R [43].

Ethics statement
Samples were taken only on communal lands. No specific permission was required for crossing these areas or carrying out snail surveys. The target species are not protected by any national law or local regulations.

Genetic analysis
DNA extraction, PCR amplification and sequencing. A snail foot fragment or, in the case of juvenile specimens, the entire animals (both preserved in ethanol) were used for DNA extraction. Total genomic DNA was isolated from 56 individuals with Sherlock AX (A&A Biotechnology, Gdynia, Poland) according to the manual. Quality and quantity of isolated DNA was assessed in gel electrophoresis (1% agarose gel). The analysed DNA fragment of cytochrome c oxidase subunit I (COI) was amplified with primers according to PCR conditions described in [29]. PCR product was sequenced in both directions with an ABI 3730 sequencer. Furthermore, six microsatellite loci were amplified: TROA3, TROA4, TROA5, TROA6, TROB5 and TROB105. All microsatellite loci, their length and repeat size along with amplification conditions were previously described by Dépraz et al. [44]. Amplification was performed in three multiplex reactions (two loci in each multiplex) with a Qiagen Multiplex PCR Kit (in 10 μl volume) according to the producer's manual. To minimize genotyping errors we used aerosol-resistant filter tips, exposed all equipment to UV radiation prior to use and added negative controls to all reactions. Lengths of the microsatellite products were read using an ABI 3730 sequencer (Wyzer Biosciences Inc., MA, USA) and GeneMarker 2.4.2 software (SoftGenetics, www.softgenetics.com). All nucleotide sequences reported in this study have been deposited in GenBank with accession numbers KU556333-KU556388. Population genetic data analysis. Since we were interested in genetic structure of analysed individuals, we calculated the probability of individual membership to distinct genetic clusters, using the Bayesian approach, as implemented in Structure 2.3.4 [45][46][47]. We used correlated allele frequencies with an admixture model without informing the former about sample localities. The calculations run for 1,000,000 replicates with burn-in in the 200,000 step. After that, we analysed K clusters from 1 to 20 and ten replicates for each K. The Delta K method [48] along with standard prediction of K based on the plotted mean ln probability of K (L(K)) were used to find the most probable number of distinct genetic clusters. Both plots (Delta K and L(K)) were calculated using Structure Harvester [49]. Clumpak package [50] with implemented Clumpp [51] and Distruct [52] software were used to estimate pairwise similarities between runs and to visualize the most probable clustering scheme.
Phylogenetic inference. Phylogenetic analyses were based on two data sets created by 98 aligned sequences of mitochondrial COI. The first set (called long) included 655 positions, whereas the second one (short) contained 561 positions, after exclusion of alignment positions that contained incomplete terminal sequence in at least one taxon. Sequences newly obtained in this study were combined with 39 previously published sequences [16,29] collected from BLAST searches of the GenBank database. Sequences of three representatives from other Hygromiidae: Candidula unifasciata (Poiret, 1801), Lindholmiola girva (Frivaldszky, 1835) and Kovacsia kovacsi (Varga & Pintér, 1972) were used as an outgroup (accession numbers JX911297, EU182448 and EU182471, respectively). To infer phylogenetic trees, we applied four approaches: Bayesian in MrBayes 3.2.3 [53] and PhyloBayes MPI 1.5a [54] and maximum likelihood in TreeFinder (partitioned model) [55] and PAUP Ã 4.0b (not-partitioned model) [56] as well as maximum parsimony and neighbour-joining methods in PAUP Ã .
In MrBayes analyses, we assumed three separate mixed models for three codon positions to sample appropriate models across the substitution model space in the Bayesian MCMC analysis itself, avoiding the need for a priori model testing [57]. In addition, we applied gammadistributed rate variation across sites with five discrete rate categories for the first and third codon positions as suggested by jModelTest 2.1 [58,59]. In the analyses, two independent runs starting from random trees, each using four Markov chains, were applied. Trees were sampled every 100 generations for 10,000,000 generations. In the final analysis, we selected trees from the last 3,779,000 (for the long dataset) and 4,271,000 (for the short dataset) generations that reached the stationary phase and convergence (i.e. the standard deviation of split frequencies stabilized and was lower than the proposed threshold of 0.01). In PhyloBayes, we used the CAT-GTR model with rate variation across sites modelled by five discrete rate categories of gamma distribution. The number of components, weights and profiles of the model were inferred from the data. Two independent Markov chains were run for 100,000 generations in each of these analyses. The last 25,000 (for the long dataset) and 5000 (for the short dataset) trees from each chain were collected to compute posterior consensus trees after reaching convergence, when the largest discrepancy observed across all bipartitions (maxdiff) was below the recommended 0.1.
In TreeFinder tree reconstruction, we applied search depth level 2 and separate substitution models for three codon positions as suggested by this program's Propose Model module according to AIC, AICc, BIC and HQ criteria, for the long dataset: J2+Γ(5), HKY{3,1,1,1,1,3}+ Γ(5), TVM+Γ(5) and for the short dataset: J2+Γ(5), J3+I, TVM+Γ(5). In the case of maximum likelihood and parsimony approaches in PAUP, the final tree was searched from ten starting trees obtained by stepwise and random sequence addition followed by the tree bisection and reconnection (TBR) branch-swapping algorithm. The maximum likelihood and neighbourjoining trees were calculated using the best-fit substitution model TPMuf+I+Γ(5), as found in jModelTest among 1624 candidate models considering all four criteria: AIC, AICc, BIC and DT. Non-parametric bootstrap analyses were performed for two datasets using 1000 replicates for neighbour joining, maximum likelihood and parsimony methods. In addition to that, the number of approaches supporting a given node was calculated using consent from PHYLIP package 3.69 [60] and superimposed on MrBayes trees.
Intra-and interclade diversity was calculated as p-distance (the proportion of nucleotide sites at which two compared sequences are different) for the long dataset with the pairwise deletion option using MEGA 6.05 package [61]. Its variance was computed by the bootstrap method assuming 1000 replicates. Three outgroup sequences were excluded in this calculation.

Procedures of species delimitation
For the delimitation of species, we used the long alignment of COI sequences and applied the Automatic Barcoding Gap Detection (ABGD) method [62] as well as the General Mixed Yule-Coalescent model (GMYC) [63,64]. ABGD was carried out via a web interface (http://www. abi.snv.jussieu.fr/public/abgd/abgdweb.html) using a distance matrix obtained from the phylogenetic tree inferred in MrBayes. GMYC analyses were carried out in the R environment [43] using the Splits package [65]. We applied both single [63] and multiple threshold models [64]. The input tree was obtained by the conversion of the MrBayes tree to the ultrametric one using the chronopl command from the Analyses of Phylogenetics and Evolution (Ape) package in R [66], which implements the penalized likelihood method [67].

Testing consistency of morphological, molecular and geographical data
To assess consistency of taxa delimitation by different approaches, we applied a likelihood ratio G-test of independence assuming Williams' continuity correction. We tested, in pairwise comparisons, independence of an individual's assignment to species, geographical location, morphological groups (assuming T. hispidus and T. coelomphala as one group), clades based on mitochondrial COI phylogeny and microsatellite clusters. In the last case, we considered both 12 microsatellite groupings of individuals (four clear clusters and eight mixed assignments of individuals) and five microsatellite groups, in which the eight mixed groups were considered as one cluster in addition to the four clear clusters.
Moreover, we used a Mantel test assuming 10,000 permutations to estimate the significance of correlation in pairwise comparisons of matrices including: Euclidean distances between snail samples for all shell and genitalia measurements, geographical distance and absolute differences between altitude of sites and where samples were found as well as the number of base differences per site between COI sequence and distances between microsatellite data, calculated according to the method of Nei et al. [68]. The metric features were subjected to minimum-maximum normalization before the calculation of distances. The geographical distances were calculated as orthodromic distances based on geographical latitudes and longitudes of the sites assuming a mean earth radius of 6371 km.
G-test was carried out by the likelihood.test function and Mantel test by the mantel.rtest function as implemented in R package [43]. The Benjamini-Hochberg method for correction of the obtained p-values was applied to control the false discovery rate in the two approaches.

Morphometric data
The CDA of shell morphology correctly assigned individuals to species in 90.9% of cases ( Table 2) and indicated that 93.7% of the variation could be explained by two discriminant functions. The first (Can1) and the second (Can2) functions detected 51.4% and 42.3% of variation, respectively. The first function distinguished individuals by the relative umbilicus diameter (U/D) and the second by the body whorl height (bwH) ( Table 3). Thus, morphometric analysis sustained the recognition of three distinct morphological groups of Trochulus occurring in sympatry in the study area (Fig 1). They may be attributed to three morphologically recognized species, i.e. T. graminicola, T. striolatus and T. sericeus (Figs 2 and 3). However, only T. striolatus was separated as a distinct set. Overlapping subsets were formed by T. hispidus and T. coelomphala, whose shells strongly resemble each other (Fig 4). The first of these two species also demonstrated a moderate overlap of variation ranges with T. sericeus and T. graminicola, while T. coelomphala formed an overlapping subset with T. graminicola. Trochulus graminicola, T. striolatus and T. coelomphala were well separated from T. sericeus on the basis of broader umbilicus (mean > 1.5 mm), revealing statistically significant differences (Kruskal-Wallis test: p < 0.001; Fig 5). U/D may be a discriminant character for T. graminicola compared to T. striolatus, T. hispidus and T. sericeus because the differences were statistically significant (Kruskal-Wallis test: p < 0.05). Similarly, T. hispidus and T. sericeus could also be distinguished from each other to some extent based on the latter character (Kruskal-Wallis test: p < 0.001; Fig 6). The values of U/D ranged from 0.14 to 0.27 (mean = 0.19) in T. hispidus and from 0.05 to 0.18 (mean = 0.10) in T. sericeus. Since the specimens from Ruine Waldau (RW) were difficult to assign to a certain species, they were separately studied in the CDA. Subsequently, all individuals were placed in the position between T. hispidus and T. coelomphala (Figs 3 and 4). Statistically significant differences of the RW population were recorded Ongoing Speciation and Gene Flow in Trochulus Species Complex for U and U/D but only with respect to T. sericeus (Figs 5 and 6). The shells of RW specimens were correctly classified in 42.9% of cases (Table 2).
In the CDA of genital measurements, the first discriminant function (Can1) accounted for 57.1% of the variance and the second (Can2) accounted for 22.4% (Fig 7). The highest loadings (Table 4) on the first function were for bursa copulatrix width/length ratio (sw/sl) and for the upper vagina (uv). The first function clearly separated T. graminicola specimens that were characterized by smaller values of sw/sl and longer uv (range 0.77-3.00 mm) (Figs 8 and 9). The first trait (sw/sl) was statistically significant between T. sericeus and all other groups except for T. striolatus (Fig 8). This means that T. sericeus has the most round bursa copulatrix (sw/ sl = 0.46-1), whereas T. graminicola and RW specimens the most elongated (sw/sl = 0.22-0.44 and 0.17-0.46, respectively) among all examined taxa. The RW specimens, however, could not be distinguished from T. hispidus and T. coelomphala. The highest loadings on the second  (Table 4) were ascribed to the flagellum/epiphallus ratio (fl/ep) and length of the epiphallus (ep). These features appeared to be very variable within the species/population and could not reliably separate them. For example, all specimens of T. striolatus included in the analysis intermixed either with T. hispidus, T. coelomphala or T. sericeus. Similarly, the RW specimens, T. hispidus and T. coelomphala overlapped with each other to too great an extent (Fig 7). The

Mitochondrial gene tree
Phylogenetic analyses based on different methods and two sets (long and short) produced very similar tree topologies (Figs 10 and 11). The sister position to the remaining Trochulus taxa is occupied by T. sericeus from Moosburg (Germany) but it is not strongly supported. In the trees, eleven main clades usually with high and very high support values could be recognized. Ten of them obtained posterior probability larger than 0.96 or bootstrap percentage greater than 96% by at least one of applied methods for the long dataset. However, none of the morphology based taxa (or morphospecies) was monophyletic in the tree. Their representatives were located in different clades. Considering deep relationships between the clades, only clades 7-11 were grouped with quite significant support. It should also be noted that members of T. hispidus complex, i.e. T. hispidus and T. sericeus distributed in clades 3, 4, 5 and 6, clustered together forming one major clade inferred by four of six applied methods/software on two datasets and strongly supported by the MrBayes approach.
Interestingly, there were several examples where sequences from a single locality of the same species were distant from each other in the trees. Trochulus sericeus from Moosburg (Germany) was located at the base of the tree and in clade 11; T. sericeus from Reischenau (Germany) was in clades 3, 6 and 11; T. sericeus from Sausteig (Austria) was in clades 6 and 11; T. hispidus from Wrocław (Poland) was in clades 1 and 3; T. hispidus from Lubawka (Poland) was in clades 3 and 11; and T. coelomphala from Kleinprüfening (Germany) was in clades 10 and 11.
Although clade 2 comprised all eight studied T. striolatus sequences from Gosheim and Untere Eichberg (Germany), it also included a sequence ascribed to T. plebeius/sericeus. The T. striolatus sequences showed low genetic divergence. Their intragroup p-distance expressed as a percentage was 1.21% ± 0.31% and their p-distance to the remaining Trochulus sequences was 14.32% ± 0.96%. On the other hand, clade 5 contained only T. hispidus whereas clade 10   Ongoing Speciation and Gene Flow in Trochulus Species Complex contained only T. coelomphala, but these species were also widely distributed in other clades. Additionally, clade 6 consisted of only T. sericeus because specimens annotated in GenBank as T. hispidus complex were morphologically attributed to T. sericeus by Duda et al. [27] and Kruckenhauser et al. [16]. Similarly, clade 7 grouped five sequences obtained from T. graminicola collected in Eichberg (Germany) but one sequence of this species was also significantly located in clade 11. The five sequences of T. graminicola showed very low within-group genetic divergence. Their pdistance was 0.34% ± 0.16%, more than 40 times less than the p-distance of these sequences to the remaining Trochulus sequences (13.79% ± 0.92%). The sixth T. graminicola sequence was remarkably diverged from the other five sequences (12.52% ± 1.26%) and clustered significantly with two sequences of T. coelomphala from Kleinprüfening (Germany) and two of T. sericeus from Moosburg and Reischenau (Germany). Our results indicate that the five T. graminicola sequences (in clade 7) are the most closely related to clade 8, including many T. coelomphala samples from different localities in Germany and two T. sericeus specimens from Ruine Farnsburg (Switzerland).
All specimens from RW formed a monophyletic well-supported clade, which was joined with very high support with T. hispidus from the closely located Schramberg in Germany (6.2 km away), forming clade 9. The position of this clade was the only difference between the tree topologies for the two datasets. In the tree for the long set (Fig 10), this clade clustered with clades 7 and 8, whereas for the short dataset (Fig 11), it was sister for clades 10 and 11. Such positions were obtained by four of six approaches/software in each of these datasets. The intragroup p-distance between snails from RW was very low, 0.31% ± 0.13%, whereas the pdistance to the remaining Trochulus taxa was very large: 13.89% ± 0.94%.

Molecular delimitation of species boundaries
We applied two delimitation methods, ABGD and GMYC to determine species boundaries based on the long alignment of COI sequences (Fig 12). Both single-and multiple-threshold GMYC models fitted the data significantly better (Likelihood Ratio test, p-values 0.0007 and 4.6Á10 −5 , respectively) than the null model assuming that the entire sample derives from a single species with uniform branching, in contrast to the alternative hypothesis about several independently evolving species lineages. The number of all delimited species including singletons (species represented by a single sample) was for the single GMYC method 38 (20 clusters + 18 singletons) and for the multiple GMYC method 33 (20 clusters + 13 singletons). The results obtained with the ABGD method were similar but dependent on a priori threshold value. In total, eight partition schemes were proposed. The largest number of 35 species (22 clusters + 13 singletons) was identified for the prior maximal distance p from 0.001 to 0.002783, whereas the smallest number 19 (14 clusters + 5 singletons) for p = 0.035938. These two extremes were presented in Fig 12 for the ABGD method.
In Fig 12, we have marked by individual colours the most numerous species groups found by at least one method. There are 17 such groups consisting of 13 clusters (with at least 2 T. sericeus (KF813004) from Moosburg, which is sister to other Trochulus. Almost all 13 clusters correspond to very well-supported clades or lineages in the phylogenetic tree (Fig 10). Five of these clusters (5)(6)(7)(8)12) comprise sequences with the same species assignment. However, in cluster 8, three methods propose separation of some samples as separate species. Similarly, within clusters 2 and 10, samples assigned to the same species and grouped together in the tree are considered as separate species by some methods (usually GMYC). In clusters 2 and 3, one sequence has different taxonomic name than others: T. plebeius/sericeus (JX475100) and T. sericeus (3). The former is considered as a separate species from T. striolatus by some methods but the latter is always grouped to one cluster with T. hispidus. There are many clusters that contain a mixture of samples with various species names. Even if we take into account methods that recognized many potential species, there are still subclusters within cluster 13 that include sequences with different species affiliations.

Microsatellite clustering
Both the Delta K method and L(K) plot indicated K = 4 as the most probable number of distinct genetic clusters for microsatellites (data not shown). Thirty individuals were attributed to one of these four clusters with posterior probability larger than 0.9 while 26 individuals had mixed affiliation to two (20 individuals), three (four individuals) or four (two individuals) clusters with a probability of belonging to the clusters between 0.1 and 0.9 (Fig 13). There was no cluster that would contain representatives of only one Trochulus species. Clusters 1 and 2 were dominated by T. sericeus but also grouped one T. coelomphala individual in each of these clusters and T. sericeus also in cluster 2. Cluster 3 comprised four T. hispidus and three T. striolatus representatives. Five of six T. graminicola snails were grouped into cluster 4, which also included one T. sericeus. Many other individuals of all studied species were ascribed to mixed clusters.

Testing consistency of different approaches
The consistency of an individual's classification by morphology described by metric features, geographic location and genetic analyses was checked by G-test of independence based on appropriate contingency tables. The analyses showed significant (with p < 0.05) relationships between all studied classifications with the exception of microsatellite data when 12 microsatellite groups were considered (Table 5). However, the relationships of microsatellite data with others were significant, when four clusters and one mixed group were assumed.
We also observed many significant correlations in the Mantel test between distance matrices expressing differences in spatial distribution (geographical and altitude location), shell metric morphology, COI sequences and microsatellites of studied snail samples (Table 6). However, correlation coefficients were not high. Disregarding the relationship between geographic and altitude distance, correlation coefficients of about 0.2 were found for differences in COI sequences with distances in geographic location, altitude and shell morphology as well as between microsatellite and shell morphology distances. The two genetic distances based on mtDNA and microsatellites were weakly correlated (r = 0.137). Shell morphology also revealed weak correlation with altitude distance and a marginal correlation with geographic distance. No significant correlation was observed for microsatellites with geographical and altitude PAUP by maximal likelihood (ML), maximum parsimony (MP) and neighbour-joining (NJ) methods. Values of the posterior probabilities and bootstrap percentages lower than 0.50 and 50%, respectively, were omitted or indicated by a dash "-".   distances. We did not find a significant relationship between differences in genitalia measurements and geographic distance (r = 0.007, p = 0.420), but there was a weak correlation with altitude distance (r = 0.105, p = 0.018).

Taxonomic position of T. graminicola and Ruine Waldau snails
The multivariate analysis of morphological variation in Trochulus taxa revealed segregation of some studied specimens into three separate species as recognized by Kerney et al. [37]. There were clear boundaries separating T. graminicola, T. striolatus and T. sericeus based on shell traits such as the umbilicus diameter and relative umbilicus diameter (Fig 3). However, the  morphometric limits of T. graminicola were less evident with comparison to T. hispidus and T. coelomphala as well as to the RW specimens. Conversely, T. graminicola could be clearly distinguished from the remaining taxa based on the reproductive system, in particular long upper vagina (Fig 9) and to a lesser extent slender bursa copulatrix (Fig 8). In agreement with that, our genetic data also showed the distinctness of almost all T. graminicola COI sequences, which formed the significantly supported clade. This clade was considered as a separate species by five of eight ABGD partition schemes. GMYC methods recognized two clusters in this clade. These specimens exhibited very low intragroup sequence divergence (0.34%) and a very large difference from the remaining Trochulus taxa (13.8%). This may imply their common and recent evolutionary history associated with allopatric fragmentation, which finally led to speciation. The genetic p-distance, 13.8% between T. graminicola and the remaining Trochulus sequences, was in the range 9.5% to 20.9% found between morphologically well-distinguished Trochulus species, i.e. T. striolatus and T. villosulus as well as T. striolatus and T. lubomirskii [16], respectively. In agreement with that, five of six T. graminicola individuals were grouped together in microsatellite cluster 4, which also included one T. sericeus. One case of the second species in this cluster may be caused by independent evolution of the same allele due to the high mutation rate of microsatellites. The sixth T. graminicola individual showed mixed affiliation to clusters 1, 3 and 4.
The current distribution area of T. graminicola remained ice-free during the Last Glacial Maximum [69], thus the species could have survived at least the last glaciation within the southern edge of Südschwarzwald. Moreover, if its limited occurrence to a single mountain is confirmed, it will suggest that T. graminicola evolved in this region and never dispersed. This may result from its strict ecological requirements. This stenoecious species inhabits a narrow ecological niche restricted exclusively to the patches of Calamagrostio variae-Pinetum association (Fig 14), a pioneer or permanent community that occurs on steep marl slopes, with shallow, lime-rich, humus-poor clay soils. This community is associated with a warmer, Atlanticinfluenced climate and known from southern Europe (Randalpen, Alpenvorlandes and Schwäbischen Alb) [70]. In southern Germany it is scarce and exists only in small stands. The site of T. graminicola is situated in the Schwarzwald-Baar region which directly borders the Schwäbischen Alb. The habitat from which we collected the snails represents the typical community but the pine trees are currently rather sparse. This is connected with the fact that the community can be a stage of succession between rocky swards and the forest [41,70]. The locality is thus well illuminated and heated, as is confirmed by the presence of many thermophilous and photophilous plants (e.g. Laserpitium latifolium, Cytisus nigricans, Anthericum ramosum, Teucrium chamaedrys, Thesium bavarum, Coronilla coronata). We found all of the plant species listed in Falkner [32] except for Norway spruce (Picea abies), European beech (Fagus sylvatica) (both probably harvested earlier) and Gentiana germanica s.l. (that is visible only in spring). This plant community is very rare [70]; as it also forms the habitat and only known location for this restricted endemic snail, it is clearly worthy of protection. Ongoing Speciation and Gene Flow in Trochulus Species Complex As shown by the very restricted ranges and habitat specialization of T. piccardi [71], T. montanus, T. biconicus, T. caelatus [38] and T. oreinos [26,72], regional speciation with local persistence seems to be common within the genus Trochulus [13,16,27,71]. In general, the poor dispersal power of snails give rise to rather clear patterns of genetic differentiation related to distance; gene flow is limited [73]. Here, however, one T. graminicola sample related more closely to other Trochulus species, which may suggest a presence of gene flow between these snails. It demonstrates that the observed differences between reproductive systems may not be sufficient to isolate reproductively the Trochulus species. Our data could also suggest the early stages of divergence of genital traits in T. graminicola that may have evolved with locally restricted but ongoing gene flow. Therefore, its ecological requirements and consequently limited geographical distribution could contribute to the evolution of reproductive isolation.
The specimens from RW were tentatively assigned to T. graminicola on the basis of shell morphology and their occurrence at a small distance (ca. 34 km in a straight line) from the type locality [33]. Our integrative studies of these snails could not, however, confirm this view. In the CDA analyses, based on both shell and genital measurements, the RW specimens were distributed among T. hispidus and T. coelomphala and, thus, separated from T. graminicola. In the phylogenetic tree (Fig 10), the RW specimens formed the significant monophyletic mtDNA group, which clustered with T. hispidus from Schramberg (Germany) but did not join directly with T. graminicola. The multiple-threshold GMYC method and three ABGD partition schemes recognized the RW specimens as one species, whereas the single-threshold GMYC and five ABGD partition schemes added to this cluster T. hispidus from Schramberg (Fig 12). In the tree based on the long alignment dataset (Fig 10), the clade was sister to the clade including T. graminicola, T. coelomphala and T. sericeus. The RW sample showed high sequence divergence (mean p-distance 12.18% ± 1.19%) from the T. graminicola clade. Additionally, the genetic cluster of the RW specimens, determined in the microsatellite analysis, also supported their separation from T. graminicola (Fig 13). They also inhabit different biotopes, despite similar elevations (Table 1). Trochulus from RW, unlike T. graminicola, originates from a castle ruin, situated close to human settlements, a typical habitat for T. hispidus and T. striolatus. In light of all the evidence, the suggestion that Trochulus from RW is conspecific with T. graminicola [33] is not justified. It formed a monophyletic clade, with high genetic distances to other clades (13.9%). This is, however, contrary to the results of microsatellite analysis that suggested a high gene flow between the RW population and many other species (Fig 13). Considering T. hispidus as a paraphyletic species complex and general high mitochondrial divergence in Trochulus species [16,28,29,31], as well as morphological similarity of the RW specimens to both T. hispidus and T. coelomphala and their low classification success of 43% and 86% in shell and genitalia, respectively, it seems that the taxonomic status of RW specimens must remain unresolved. Temporarily, it can be included in the T. hispidus complex.

Distribution of morphospecies in the phylogenetic tree
Species recognised on morphological criteria (morphospecies) did not correspond to monophyletic clades identified by molecular data but were scattered over the phylogenetic tree. The species delimitation methods used identified at least 14 potential species. Of these, five clades in the phylogenetic tree contained two morphospecies (clades 1, 2, 3, 4 and 8) and clade 11 consisted of four mixed taxa: T. hispidus, T. sericeus, T. graminicola and T. coelomphala. The significant grouping of sequences from different species suggests some gene flow, hybridization or incomplete lineage sorting between them. Alternatively, they may represent a smaller number of species or populations showing great morphological variation related to environment. Gene exchange could happen between different species inhabiting the same region, i.e. between: T. hispidus and T. sericeus in Lower Silesia (clade 1) and T. hispidus and T. plebeius/sericeus in southern Great Britain (clade 4), as well as T. sericeus with T. hispidus, T. graminicola and T. coelomphala in Germany (clade 11). Similarly, hybridization could have occurred between T. coelomphala and T. sericeus snails in neighbouring regions in Germany and Switzerland (clade 8). Interestingly, gene flow over longer distances seems indicated by similarities between German T. striolatus and Polish T. plebeius/sericeus (clade 2), German T. sericeus and Polish T. hispidus (clade 3) and Austrian T. sericeus and Polish T. hispidus (clade 11). In clades 1, 3 and 11, no species delimitation method, recognized all morphospecies as separate clusters. In clade 4, most of these methods also considered different morphospieces as one cluster, which suggests that these morphospecies should not be considered as separate species.
Trochulus hispidus collected from distant locations in Downside (Great Britain) and Moosburg (Germany) were significantly clustered in clade 5 with exclusion of other English and German T. hispidus. This could also indicate some gene flow between the English and continental populations or a common ancestry because all species delimitation methods consider the Downside and Moosburg populations as a separate species.

Integrating morphological and molecular analyses of other taxa
Most of the individuals of T. coelomphala placed in the ordination area overlapped T. hispidus in both shell and genital analyses (Figs 3 and 7). Likewise, this evidence was confirmed using the CDA method. Based on this analysis, a posteriori classification of shell morphology indicated that only 69% of a priori-identified specimens of T. coelomphala were correctly classified (Table 2). There was no perfect classification success rate in genital morphology (86%). It appears that the interspecific boundary was disturbed either due to high morphological variation and/or scarce taxon sampling in the present work. In our phylogram (Fig 10), all specimens tentatively determined as T. coelomphala based on their geographic origin (the Bavarian Danube valley [36]) were separated into three clades, with only one being monophyletic (clade 10) and consisting of five snails coming from two localities examined in this study. All these specimens were recognized as one species by all methods. The remaining two clades (8 and 11) also contained other Trochulus morphotypes from distant geographic regions such as Baselland (Switzerland), Swabia, Upper Bavaria, Schwarzwald-Baar (Germany), Bregenz Forest (Austria) and Central Sudetes (Poland). The clustering of T. coelomphala with T. sericeus in these clades would suggest gene flow between these snails or incomplete mitochondrial lineage sorting, if we assume that they are separate species. However, no delimitation method considered them as separate species in clade 11 and in clade 8. Most methods grouped T. coelomphala with at least one T. sericeus into one cluster. The analysis of microsatellite markers revealed a low degree of genetic differentiation among the examined T. coelomphala populations, and no cluster contained exclusively T. coelomphala individuals. Each snail was classified into two clusters, 1 and 2, six were assigned simultaneously to clusters 1 and 2, and one individual showed mixed affiliation to clusters 1 and 3, whereas the other was grouped to clusters 1, 2 and 3. Both here and in the results obtained by other authors [27], the taxonomic status of T. coelomphala remains unclear.
Individuals morphologically resembling T. hispidus and T. sericeus should be considered as T. hispidus complex [27] because none of them meets the criterion of monophyly [16]. In these two parallel studies, nine highly differentiated mitochondrial clades were identified, also revealing no morphometric differentiation between the clades. Species delimitation methods also clustered T. hispidus and T. sericeus together. We identified three main T. hispidus complex clades in the phylogenetic tree (1, 3+4+5 and 11). Clades 1, 3, 4 and 5 consisted exclusively of both morphotypes. Clade 1 was clearly separated from clades 3, 4 and 5, which grouped together and formed one big clade. The third separated clade (11) besides T. hispidus and T. sericeus also contained two T. coelomphala sequences and one T. graminicola sequence, suggesting some gene flow to them from members of the T. hispidus complex or incomplete lineage sorting of mtDNA. At least five cases of hybridization between T. hispidus and T. sericeus can be assumed. Two of them concern geographically distant populations in Poland and Germany/Austria. Interesting instances are T. hispidus from Downside in Great Britain and Moosburg an der Isar in Germany that form the strongly supported clade 5. This could indicate that the populations from the British Isles and continental Europe maintained genetic similarity, e.g. during and shortly after the last glacial period when the sea level was low and these two regions were not separated by the English Channel but connected by the Weald-Artois anticline and Doggerland, which was flooded by rising sea around 6,500-6,200 BC [74]. Nevertheless, the genetic differences between these populations are so large that all species delimitation methods considered them as separate species.
Interestingly, clade 6 contained individuals with a small relative diameter, thus belonging to T. sericeus [28] and corresponding to clade 8 in trees of other authors [16,27]. In the study by Duda et al. [27], clade 8 containing exclusively T. sericeus was also the only one in which both morphological and genetic data were consistent. In our trees, four snails from Karlsruhe, i.e. those of clade 8 [27], were nested within our clade 6. Species delimitation methods recognized at least three potential species within clade 6, but did not separate T. hispidus and T. sericeus.
In contrast to that, T. hispidus and T. sericeus were never grouped together into welldefined microsatellite clusters, which may result from smaller sampling of microsatellite than mitochondrial data. Individuals of T. sericeus dominated in two microsatellite clusters (1 and 2); in addition, one each of T. coelomphala and T. striolatus snails were included in cluster 2. The presence of these single members of other species in these clusters may result from the high degree of microsatellite mutation and development of non-homologous alleles of the same size by chance.
Studies on shell morphology showed only a limited overlap between T. hispidus and T. sericeus morphotypes (Fig 3). The major umbilicus diameter (U) and relative umbilicus diameter (U/D) best described the differences between these two ecophenotypes (Figs 5 and 6). However, this morphometric separation does not agree with the mtDNA trees, in which these taxa are grouped together in many clades. This is because these two different morphospecies, T. hispidus and T. plebeius/sericeus, from the same geographical region, are closely related [28], and can interbreed giving fertile offspring [75]. Moreover, under constant laboratory conditions the average shell shape of T. hispidus changed significantly from flat with a wide umbilicus to more globular or even elevated with a narrower umbilicus, thus resembling T. sericeus. Therefore, this phenotypic plasticity could explain the evolution of sympatric polymorphism in this species complex [75]. The absence of significant anatomical differences between these morphotypes recorded in this and other studies [27,28] provides additional evidence for the lack of reproductive barriers as indicated by genetic studies.
The coexistence of distinct gene pools in sympatry strongly indicates the presence of separate species. In allopatry, however, it is harder to assess the significance of observed differences in terms of the conventional biological species concept. While the efficiency of microsatellites in data to identify species has been recently demonstrated for various groups of plants and animals [1,[76][77][78][79][80], our studies did not show the clear separation of morphological species into microsatellite clusters. Almost half of the snails studied showed a mixed classification.
Considering the phylogenetic species concept, there are three possible explanations of the discrepancy between genetic and morphological data regarding the recognition of the two morphospecies T. hispidus and T. sericeus: (1) the morphospecies may have diverged recently and a complete barrier to gene flow between them has not yet been established; (2) they belong to the same species but display substantial morphological variation; (3) they were represented by too few individuals (i.e. insufficient sampling), which prevented the identification of separate gene pools. Our results illustrate difficulties in assessing species status as well as problems associated with underlying definitions of species concepts in snails.
The differentiation of T. striolatus from the T. sericeus and T. graminicola examined in this study was clear based on U/D, being intermediate (0.14-0.22) between the two latter taxa (0.05-0.18 and 0.21-0.29, respectively). Despite our limited sampling of T. striolatus (n = 42) these values stayed in accordance with those calculated based on extensive material [25,29]. Therefore, this trait, besides a specific riffle pattern on the shell surface [27], may be useful to discriminate T. striolatus. Other shell and genital morphological features separated T. striolatus only to some extent (Figs 5, 8 and 9) because of its extremely high intraspecific shell size differences [27,29], which may be consequences of its ecophenotypic variation. For example, our recent study revealed correlations between the morphological variation of T. striolatus and altitude as well as other climate variables, which might be explained by this plasticity. As a result, we found no sufficient justification for its infraspecific classification (Proćków et al. unpublished work). Admittedly, a slight morphological differentiation in the cross section of the penis between subspecies (T. striolatus striolatus, T. s. danubialis and T. s. juvavensis) was detected but it was based on a limited sampling dataset (n = 26 [27]). In agreement with that, the three different subspecies were not clearly separated on the basis of the COI mtDNA data [16]. On the other hand, T. striolatus appears to be a good species based on morphology and COI, and its monophyly was also evidenced in this and previous studies [16,29]. Clade 2 in mtDNA trees was the only one that included all the studied snails of one species, namely T. striolatus, although it also contained one sequence ascribed to T. plebeius/sericeus from Zieleniec (Poland), suggesting gene flow between these taxa or incomplete lineage sorting. However, not all species delimitation methods recognized all T. striolatus samples as one species. Trochulus striolatus did not form pure microsatellite clusters either but mixed with other species. This may result from incomplete sorting of microsatellite alleles into these lineages or independent evolution of the same microsatellite alleles in these species. Patterns of divergence with gene flow were observed in both invertebrate [81,82] and vertebrate species [83][84][85].

Conclusions
General comparisons of results for shell and genitalia metrics and spatial distribution as well as genetic analyses revealed that these features are not completely independent and show some correlations. Altitude showed greater influence on shell and genitalia morphology than geographic distance. Mitochondrial genetic variation revealed a stronger relationship with shell morphology and spatial distribution than microsatellite data. However, correlation coefficients between these features were at best weak although significant. This indicates that geographic and altitude location cannot fully explain shell, genitalia and genetic variation. The same constraint applies to comparisons of shell morphology with genetic variation and to the comparison of mitochondrial with microsatellite data. Most species delimitation methods recognise T. graminicola as a distinct species, based on its consistent recognition as a separate mitochondrial clade, its domination in one microsatellite cluster and the distinction from other species in genitalia characteristics. Trochulus sericeus created a few separate clades in the COI phylogeny, dominating in two microsatellite clusters and differing from other Trochulus in shell morphology. Trochulus striolatus was grouped into one clear mitochondrial clade and showed different shell morphology in comparison to other species. However, both mitochondrial and microsatellite data indicated that there is no clear separation of all Trochulus species; samples of the different morphospecies studied were mixed, suggesting continuing gene flow between them. Alternatively, they are descendants of an ancestral widespread population, from which the recently separated populations inherited the same markers and differentiated morphologically due to local environment. Additional research is required to understand the evolutionary history of Trochulus well. Our studies constitute an important step towards the comprehensive revision of the genus, in which taxonomic status of many species remains poorly understood.
Supporting Information S1 Dataset. Shell and genitalia measurements of all specimens examined. (XLSX)