In August 2006 and May 2007 10–20 individuals of C. alpina were sampled from 37 springs and headwaters across the Alps (Fig. 1, Table S1 in the Supplement). To gain a first insight into the genetic diversity of C. alpina only one individual from each population was chosen for sequencing. In addition, four individuals from four populations in Corsica, three individuals from three populations in Sweden, one individual from a British population and one individual of Crenobia alpina montenigrina sampled in Macedonia from a tributary of Lake Ohrid were included in our study (for more details of sampling localities see Table S1 in the Supplement). These individuals from populations outside the Alps were considered to get a first impression of the large scale genetic diversity of C. alpina in Europe. One individual of Polycelis felina sampled in Switzerland was used as outgroup. Specimens were preserved in 99 % ethanol and stored at -23 ∘C. All individuals analyzed in this study are listed in Table S1. Since C. alpina is mostly restricted to springs and headwaters each sampling location was defined as a separate population.

The DNeasy^™ Tissue Kit (Qiagen) was used to extract DNA following the protocol provided for animal tissues. For the mitochondrial cytochrome oxidase subunit I (CO I) gene the primers Pr-a2 (5'-AGC TGC AGT TTT GGT TTT TTG GAC ATC CTG AGG T-3'; Bessho et al., 1992) and CO13B (5'-AAG TGT TGN GGR AAR AAN GT-3'; Telford et al., 2000) were used for amplification. These primers increased the length of the sequenced fragment considered in our previous study (see Brändle et al., 2007) from 382 to 555 bp. 18S rRNA gene fragments with a length of 544 bp were amplified using the primers 18S1 (5'-TAC TGT TGA TCC TGC CAG TA-3'; Kuznedelov et al., 1996) and 18S2 (5'- ATT ACC GCG GCT GCT GGC ACC-3'; Kuznedelov et al., 1996). The fragments were amplified in a thermocycler (Eppendorf Mastercycler, Hamburg, Germany). The PCR conditions for the COI gene consisted of 5 min initial denaturation at 95 ∘C, followed by 31 cycles of 30 s denaturation at 94 ∘C, 50 s annealing at 50 ∘C and 1 min extension at 72 ∘C, followed by 10 min final extension at 72 ∘C. The PCR conditions for the 18S rRNA gene consisted of 5 min initial denaturation at 94 ∘C, followed by 31 cycles of 1 min denaturation at 94 ∘C, 70 s annealing at 55 and 2.5 min extension at 74 ∘C, followed by 10 min final extension at 74 ∘C. Products were purified using the MinElute^™ PCR purification kit (Qiagen) following the protocol provided. The purified PCR fragments were sequenced in both directions by a commercial company (GENterprise GmbH, Mainz, Germany). All haplotypes were submitted to GenBank (accession number KY569117-KY569175; see Table S1).

Partial COI gene and 18S rRNA gene sequences were aligned using BioEdit 7.0.5.3 (Hall, 1999). In a number of sequences particularly within the COI fragment, some positions showed ambiguities and were discarded. These exclusions reduced the length of the COI gene fragment to 450 bp and those of the 18S RNA gene fragment to 531 bp. Thus, the total length of the combined fragments used for phylogenetic inference was 981 bp. We also translated the COI gene sequences into amino acid sequences according to the flatworm mitochondrial genetic code (Nei and Gojobori, 1986) using MEGA 5.05 (Tamura et al., 2011). Except for one position in the sequence of C. alpina montenigrina that was translated into an internal stop codon, no insertions, deletions or stop codons were found. In the 18S RNA gaps were coded as missing data.

The number of substitutions and genetic distances between haplotypes of both fragments and the combined data sets were calculated with MEGA 5.05. Spatial structuring of the sampled populations was analyzed by means of correlation analysis of pairwise genetic versus pairwise geographic distances among the sample sites from the Alps (all 37 sampling sites) using patristic distances inferred from the Bayesian tree of the COI gene fragment (see below). Patristic distances were calculated with the program PATRISTIC (Fourment and Gibbs, 2006). We tested this matrix correlation using a Mantel test as implemented in the “vegan” package (Oksanen et al., 2013) in the software R (www.r-project.org).

Phylogenetic relationships among haplotypes were analyzed with MRBAYES version 3.1.2 (Ronquist and Huelsenbeck, 2003). To select an appropriate evolutionary model of nucleotide substitution for each nucleotide fragment, the approach of Minin et al. (2003) as implemented in the Perl script DT-ModSel was used. In contrast to the likelihood-ratio tests implemented in MrModeltest 2.2 (Nylander, 2004), this approach of performance-based model selection uses the Bayesian information criterion and considers the relative branch-length error as a performance measure in a consistent framework with penalty for overfitting based on decision theory (Minin et al., 2003). MRBAYES was run for 106 generations, submitting the two gene regions separately as well as both together. Since different models of sequence evolution for the different genes were detected (see results), partitions for the two genes have been specified. Trees were sampled every 1000th generation using a random tree as a starting point. Inspection of the standard deviation of split frequencies after the final run indicated convergence of Markov chains (< 0.03). In all analyses, two parallel Markov chain Monte Carlo simulations with four chains (one cold and three heated) were run. The first 10 samples of each run were discarded as burn-in. From the sampled trees, consensus trees were produced using the ALLCOMPAT command in MRBAYES.

The distinct lineages recognized in the present study may represent different species. In general, species are considered cryptic if they have been erroneously classified as a single species because of a lack of distinct morphological features. DNA barcoding has revealed cryptic diversity in a variety of taxa (Nieberding et al., 2005; Bickford et al., 2006; Witt et al., 2006; Lefébure et al., 2007; Pauls et al., 2010). Lefébure et al. (2006b) showed that clades of crustaceans diverging by more than 0.16 substitutions per site of the CO I gene as measured by patristic distances are likely to present distinct species. Based on the threshold distance suggested by Lefébure et al. (2006b) a minimum of 10 distinct species of C. alpina may be present across the Alps. Further evidence for putative species was found by analyzing uncorrected pairwise sequence divergences. Uncorrected p-distances of the CO I gene ranged from 0.1 to 14.6 % (±0.03 SD), with an average of 11 %. In a study of the cryptic diversity of Tetramorium ants in the western Palearctic, Schlick-Steiner et al. (2006) found interspecific divergences for the CO I gene ranging from 4.0 to 10.6 %. Liu et al. (2003) investigated cryptic species diversity within the western American springsnail Pyrgulopsis micrococcus and found uncorrected p-distances of the CO I gene among lineages from 3.69 to 10.6 %. Our phylogenetic analyses provided some evidence for distinct species also within C. alpina. Descriptions of new species in flatworms, however, will require studying the internal anatomy in particular the copulatory apparatus of planarians. But note that body form, reproduction mode and morphology of the reproduction apparatus of the Turbellaria vary considerably depending on environmental conditions and food supply (e.g., Steinböck, 1942). As a consequence, classical taxonomy may have overlooked species in the past leading to an underestimation of the diversity of lineages (see also Rader et al., 2017). In that regard we hope that genetic characterization could guide future morphological studies.

Pfenninger and Schwenk (2007) showed that cryptic species are evenly distributed among the major metazoan taxa. This suggests that a simple body plan is not a sufficient explanation for an increased occurrence of cryptic species within a taxonomic group. Some authors (e.g., Bickford, 2006) have suggested a link between extreme environmental conditions and cryptic species. Extreme environments impose stabilizing selection on changes in morphology during speciation, resulting in morphological stasis (Rothschild and Mancinelli, 2001; Nevo, 2001; Lefébure et al., 2006a; Bickford, 2006; Lefébure et al., 2007). C. alpina is in general restricted to extreme habitat conditions of cold springs and crenal streams. Harsh conditions occurring in such environments and emerging biological properties, such as general inability to tolerate temperatures above 13–15 ∘C and preference of temperatures slightly above 0 ∘C (Steinböck, 1942; Reynoldson, 1953; Beauchamp and Ullyott, 1932; Köhler, 1932, in Wright, 1974), might limit morphological divergence. Evidence for morphological stasis of groundwater amphipods (Lefébure et al., 2007; Trontjl et al., 2009) and spring-inhabiting snails (Benke et al., 2009) has been found. The concept that extreme environmental conditions favor the evolution of cryptic species still requires further evaluation. Pfenninger and Schwenk (2007) have already shown that there are no differences in the occurrence of cryptic species among different biogeographical regions, which raises doubt about the importance of environmental conditions for cryptic diversity.

Given that we propose the existence of cryptic species and since our sampling scheme is based on one individual per site only, we cannot detect multiple species at one location. In fact, it is surprising that we still found moderate spatial structuring in our data. According to our phylogenetic analysis of C. alpina, roughly five biogeographic groups could be distinguished: (i) western Alps I, (ii) southwestern Alps, (iii) eastern Alps (iv) western Alps II and (v) southern Alps (Fig. 1). In the central parts of the Alps, we found no clear geographical structuring of C. alpina. The detected spatial structuring fits with previous findings for plants (Schönswetter et al., 2005) and animals (e.g., Pauls et al., 2006) and thus might be a promising hint for future studies with more samples per site and additional markers. With such approaches it would be possible to retrieve a more complete and convincing picture of the inter- and intraspecific divergence and to test whether closely related cryptic species show vicariant patterns as documented between flatworm genera (Wright, 1974; Snook and Milner, 2001; Roca et al., 2006). Furthermore, due to the strong environmental link to cold springs it is expected that little gene flow occurs between flatworm populations (Brändle et al., 2007). The moderate geographic structuring revealed in our study may support this finding. However, the well-known potential for asexual reproduction in C. alpina (Reynoldson and Young, 2002) and founder effects may also have a strong influence on the geographic structuring.

Our study provides some evidence that the orogenic history of the alps and presumably the climatic variability during the Quaternary promoted genetic diversity of C. alpina. Although our study represents only a first insight into the evolutionary history of C. alpina in the Alps, we demonstrated that it is well worth having a closer look at species that apparently show less variability in terms of morphology, habitat choice and ecology (Rader et al., 2017). Judging whether the detected genetic lineages represent different species will be a challenge of future studies.