A molecular test for cryptic diversity in ground water: how large are the ranges of macro-stygobionts?


Peter Trontelj, Department of Biology, Biotechnical Faculty, University of Ljubljana, P.O. Box 2995, 1001 Ljubljana, Slovenia. E-mail: peter.trontelj@bf.uni-lj.si


1. Various groundwater habitats have exceptionally high levels of endemism caused by strong hydrographical isolation and low dispersal abilities of their inhabitants. More than 10% of macro-stygobiotic species nevertheless occupy relatively large ranges, measuring from some hundred to over 2000 km in length. These species represent a challenge because their distributions disregard hydrographical boundaries, and their means to disperse and maintain long-term gene flow are unknown.

2. Based on mitochondrial and nuclear gene sequences, we examined the phylogeographic structure of six formally recognised stygobiotic species (Niphargus virei, N. rhenorhodanensis, Troglocaris anophthalmus, T. hercegovinensis, Spelaeocaris pretneri, Proteus anguinus) and searched for cryptic lineage diversity in a genus-wide phylogeny of Niphargus. Using tree-based criteria as well as comparative divergence measures, we identified cryptic lineages, which may tentatively be equated with cryptic species.

3. Fourteen analysed nominal stygobiotic species with large ranges emerged as highly diversified, splitting into 51 tentative cryptic lineages. The degree of divergence was within the range or larger than the divergence of other related pairs of sister species. A substantial part (94%) of the cryptic lineages had ranges <200 km in length. One half of them were recorded at single sites only. The largest range recorded was that of a cryptic N. virei lineage (700 km), while none of the very large traditional ranges (e.g. Niphargus aquilex– 2300 km, N. tauri– 1900 km) could be corroborated.

4. These data suggest that small ranges of macro-stygobionts are the rule, and ranges over 200 km are extremely rare.

5. The implications of this result for groundwater biodiversity assessment and conservation include a considerable increase in overall diversity at the regional and continental scale and a strong decrease in faunal similarities among regions, coupled with greater endemism.


Ground water, in particular its karstic component embedded in consolidated rocks, is known to be an environment of exceptionally high endemism. This is perhaps best illustrated by the stygobiont distribution data summarised in Botosaneanu (1986). Sket, Paragamian & Trontelj (2004) have shown for the Balkan Peninsula that whenever a region reaches 20 obligate subterranean species or more (both aquatic and terrestrial species counted), at least 40–60% of them are endemic to that region. Of the 380 stygobiotic species recorded in France, 272 species and subspecies (72%) are endemic to the country, and 96 (26%) are known from a single locality only (Ferreira, 2005). Regional groundwater faunas differ considerably in species richness and composition (Marmonier et al., 1993; Sket, 1999a,b; Christman & Culver, 2001; Gibert & Deharveng, 2002; Sket et al., (2004a); Gibert & Culver, 2005). The causes of these patterns have been amply discussed (Culver, Kane & Fong, 1995; Stoch, 1995; Sket, 1999a; Danielopol, Pospisil & Rouch, 2000; Holsinger, 2000), and these authors provide several possible explanations accounting for high levels of endemism, all involving restricted dispersal:

The tendency of many groundwater animals to have their distribution restricted to small areas is contrasted by a smaller number of nominal stygobiotic species with large ranges. In many taxonomic groups single-site endemics coexist with large-ranged species (e.g. Sket, 1999b). The question arises, how can, under the same set of ecological limitations (see above), the ability of biologically similar stygobiotic species to disperse and maintain genetic contact be so different. In other words, do the present or historical discontinuities of the groundwater habitat not apply to all species? Stoch (1995), while looking for answers to similar questions, concluded that some subterranean crustacean taxa have high dispersal abilities while others do not. However, he also suggested that widely distributed species may in fact constitute unrecognised sibling species. This idea has often been supported by data, although not always with respect to widely distributed species (e.g. Avise & Selander, 1972; Laing, Carmody & Peck, 1976; Holsinger, 1978; Magniez, 1981; Cobolli Sbordoni et al., 1990; Culver et al., 1995; Danielopol et al., 2000; Sbordoni et al., 2000; Verovnik et al., 2004; Lefébure et al., 2006a, 2007; Zakšek, Sket & Trontelj, 2007). Several authors recognised the lack of correspondence between genetic and morphological differentiation of stygobiotic populations due to convergent or parallel morphology (e.g. Culver et al., 1995; Turk, Sket & Sarbu, 1996; Arntzen & Sket, 1997; Wiens, Chippindale & Hillis, 2003; Fišer, Trontelj & Sket, 2006).

Recently, molecular tools and computational methods that can reveal both historical and recent lack of gene flow between populations have become available (e.g. Templeton, 1998; Avise, 2000; Posada & Crandall, 2001). These approaches were proposed to supplement or even replace traditional taxonomic practice (e.g. Tautz et al., 2003; Blaxter, 2004). While the term ‘‘DNA taxonomy’’ has been used for many different aspects of DNA-assisted discovery, diagnosis, identification and classification of taxa, we refer here only to the delimitation of lineages that have not been distinguished morphologically, but might qualify as species after further scrutiny.

Cryptic stygobiotic diversity may result from two different evolutionary processes: (i) genetic differentiation among allopatric cryptic sister lineages or (ii) convergent or parallel evolution of non-sister lineages. In the case of cryptic allopatric speciation (type 1 cryptic diversity), sister lineages from different sites should exhibit considerable genetic divergence along with mutual exclusion of haplotypes/alleles. This pattern is equivalent to the phylogeographic category I of Avise (2000), in which major lineages occur allopatrically. The question is at what point of genetic divergence deeply split lineages can be considered as species, although it remains debatable whether such a point exists, as does even the idea of using genetic divergence as a criterion to delineate species (e.g. Avise, 2000; Hebert & Gregory, 2005; Meyer & Paulay, 2005; Will, Mishler & Wheeler, 2005; but see Lefébure et al., 2006b). A second case of cryptic diversity (type 2) corresponds to morphological resemblance of non-sister lineages due to either lack of morphological differentiation or convergent evolution under similar ecological conditions. Type 2 cryptic diversity can be assessed from the tree topology. This criterion corresponds to genealogical discordance (e.g. de Queiroz, 1998; Avise, 2000; Sites & Marshall, 2004) and represents one of the reasons for species-level non-monophyly (Funk & Omland, 2003). Its identification is generally straightforward but requires careful phylogenetic inspections since true biological species may appear paraphyletic, too, for reasons such as introgression, hybridisation or incomplete lineage sorting (e.g. Funk & Omland, 2003).

In this work we examine phylogeographic relationships and lineage differentiation in several formally recognised and widely distributed groundwater species with respect to both types of cryptic diversity. Using molecular data sets unlinked to morphology, our objective was: (i) to test whether stygobionts can have large ranges and (ii) to assess the contribution of cryptic diversity to the total stygobiotic diversity.


Taxonomic and ecological scope

This paper focuses on larger groundwater animals, the so called macro-stygobionts. This size class of groundwater fauna begins where the meiofauna ends, i.e. at approximately 2 mm length. One major faunal component in ground water excluded by this size criterion are the microcrustaceans, especially ostracods and copepods (Dole-Olivier et al., 2000; Marmonier et al., 2000). They were excluded primarily because their modes of dispersal and reproduction are much more diverse, often unknown or currently unpredictable, and unconventional as compared to those of macroscopic groundwater animals. Examples include dispersal via the epikarst stratum (Gibert, 1986; Sket, Trontelj & Žagar, 2004b; Pipan, 2005), passive dispersal via surface routes and alluvial aquifers, or transport of resistant torpid individuals or eggs (Danielopol et al., 1994; Ward & Palmer, 1994).

Most of the presented cases are from the European mainland, with an emphasis on the Dinaric Karst region and eastern France. The taxonomic scope was largely determined by available distribution data and the accessibility of sampling sites, and includes five separately analysed taxonomic groups (Table 1). First, we investigated the genus Niphargus, one of the best-known and most typical groups of groundwater animals in Europe. The evaluated data consisted of two parts: (i) two extensive phylogeographic studies of Niphargus virei (Chevreux) and N. rhenorhodanensis (Schellenberg) based on nuclear and mitochondrial gene sequences (Lefébure et al., 2006a; Lefébure et al., 2007); (ii) a genus-wide phylogenetic study of Niphargus species and geographic populations based on nuclear 28S rDNA sequences. Second, we considered the European cave salamander Proteus anguinus (Laurenti), the only stygobiotic vertebrate in Europe and the only stygobiont listed on Annex II of the EU Council Directive on the conservation of natural habitats and of wild fauna and flora. Populations traversing the entire range of the species have been analysed using a portion of the mitochondrial DNA measuring up to about 1100 bp covering the control region, threonine and proline tRNA, intervening spacer plus partial cytochrome b and phenylalanine tRNA (Gorički & Trontelj, 2006). Further, we included another group of large-bodied groundwater animals with widely distributed nominal species, the decapods Troglocaris and Spelaeocaris. The phylogeny and phylogeography of these cave shrimps have been extensively studied using nuclear (28S rDNA) and mitochondrial [cytochrome oxidase I (COI), 16S rDNA] gene sequences (Zakšek et al., 2007). According to known habitat preferences, Proteus, Troglocaris, Spelaeocaris and N. virei can be considered as typical inhabitants of the karstic groundwaters, accessible through caves and karst springs. Other analysed Niphargus spp. are known from a variety of habitats, ranging from karstic and interstitial to entirely epigean.

Table 1.   Taxonomic groups and genes used in this paper to study cryptic diversity in ground water
Taxonomic groupNo. of specimens analysedGenes usedSource of data
  1. *The analysis of Niphargus included two additional specimens of N. fontanus (EF025825, DQ119323), eight N. rhenorhodanensis (from Lefébure et al., 2007) and seven N. virei (from Lefébure et al., 2006a) (n.a., not applicable).

Niphargus (Amphipoda, Niphargidae)*69n.a.28S rRNAThis study
Niphargus virei134COI28S rRNALefébure et al. (2006a)
Niphargus rhenorhodanensis51COI, 16S rRNA28S rRNALefébure et al. (2007)
Troglocaris spp. (Decapoda, Atyidae)54COI, 16S rRNA28S rRNAZakšek et al. (2007)
Proteus anguinus (Amphibia, Caudata, Proteidae)84Partial cytb and 12S rRNA, control region, tRNA-Thr, tRNA-Pron.a.Gorički & Trontelj (2006)

The distribution data for the analysis of range sizes were obtained from databases of the Department of Biology, Biotechnical Faculty, University of Ljubljana, Slovenia; of the Laboratory of the Ecology of River Hydrosystems, University of Lyon 1, France, and compiled from a literature database on Niphargidae (http://www.bf.uni-lj.si/bi/zoologija/cene_fiser/niphargus/reference.htm). The distribution of cryptic groups of N. virei and N. rhenorhodanensis is listed in Lefébure et al. (2006a, 2007). The localities of P. anguinus were taken from Sket (1997). We expanded the exploration of range sizes to some other stygobiotic taxa with well-known distributions that have not been studied by molecular methods.

Some previous studies explored subterranean range sizes expressed as surface area (Holsinger, 1978), number of U.S. counties (Culver et al., 2000), number of biogeographic regions (Sket, 1999a), or number of 400-km2 cells of a geographical grid (Ferreira, 2005). Here, we chose linear distances as a more direct measure related to a species’ ability to exchange genetic material between different parts of its range. Range sizes were thus measured as the longest possible diagonal in a geographical information system (GIS), or estimated from maps when no exact locality data were given.

Acquisition of DNA sequences

For Niphargus, molecular sequence data were obtained from two previous studies on the cryptic diversity of N. virei (Lefébure et al., 2006a) and N. rhenorhodanensis (Lefébure et al., 2007). New 28S rDNA sequences for a genus-wide phylogenetic study were acquired from 67 specimens of Niphargus representing different species and/or populations, plus two outgroups. The outgroup species Gammarus fossarum Koch and Synurella ambulans (Müller) were chosen on the basis of the amphipod phylogenetic study by Englisch, Coleman & Wägele (2003), which identified gammarids and crangonyctids as potential sister families to the Niphargidae. The forward primer 5′-CAAGTACCGTGAGGGAAAGTT-3′ and the reverse primer 5′-GTTCACCATCTTTCGGGTC-3′ were used for amplification and sequencing. The sequences were deposited in GenBank under accession numbers EF617235EF617304. The data on P. anguinus are from Gorički & Trontelj (2006), and the data on Troglocaris from Zakšek et al. (2007).

Analyses of phylogenetic relationships

The identification of cryptic diversity in this study relies largely upon phylogenetic criteria. Some of the underlying taxonomic data sets have already been analysed in four separate previously published papers (Table 1). New phylogenetic analyses were performed with the 28S rDNA sequences of Niphargus (Table 1) and with selected caudatan mitochondrial DNA sequences from GenBank used to compare levels of divergence with the between-group divergence within Proteus (Fig. 3). Sequences were aligned using Muscle (http://www.drive5.com/muscle/; Edgar, 2004). The alignment is available at http://www.bf.uni-lj.si/bi/zoologija/peter_trontelj/data.modelgenerator/; Keane et al., 2006). For Niphargus, this was a general time-reversible model with gamma-distributed rate heterogeneity and no invariable sites (GTR + Γ), as suggested by Bayesian and the Akaike (I and II) information criteria. Trees were searched for under the maximum-likelihood criterion with Phyml (Guindon & Gascuel, 2003) and Bayesian inference with MrBayes (Ronquist & Huelsenbeck, 2003) assuming six discrete gamma categories. All model parameters were optimised during the searches. Branch support for the maximum-likelihood tree was assessed by applying 500 bootstrap cycles. Bayesian searches (with default prior settings) were run on four cold chains, each obtained from an independent Markov chain Monte Carlo search. In each search, four chains (one of them cold) were run for 2 × 106 generations with a sampling frequency of 100 generations. A 50% majority rule consensus tree was obtained from a combined last 5000 trees from each run.

Figure 3.

 Box plots comparing interspecific genetic divergence of established sister species of salamanders (left pane) and the divergence between six deeply divergent mitochondrial haplotype groups within Proteus anguinus (described in Gorički & Trontelj, 2006, right pane). Stars represent outliers. Divergence is measured as likelihood-estimated patristic distance. The following potential sister groups within P. anguinus are compared: (A) South-east versus south-west Slovenia; (B) Southern Coastal Region versus Lika; (C) Southern Coastal Region versus Bosanska Krajina; (D) Lika versus Bosanska Krajina; (E) Istra versus Southern Coastal Region + Lika-Krajina; (F) Istra versus Slovenia; (G) Southern Coastal Region + Lika-Krajina versus Slovenia. Salamander sequences and phylogenies were taken from McKnight & Shaffer (1997); Steinfartz, Veith & Tautz (2000); Arnason et al. (2004); Mueller et al. (2004); Storfer et al. (2004); Samuels et al. (2005); Steele et al. (2005) and Weisrock et al. (2006).

Delimitation of putative cryptic species

As outlined above, two types of cryptic stygobiotic diversity can be expected: type 1, resulting from genetic differentiation among allopatric cryptic sister lineages, and type 2, resulting from morphological resemblance of non-sister lineages due to plesiomorphic or homoplastic similarities.

In the case of type 1 cryptic diversity we used a frequently applied indicative approach in which the divergence between two potential cryptic sister species is compared to the divergence of closely related bona fide species. This approach has been applied to cryptic species of both crustaceans (e.g. Axayácatl, Fleeger & Foltz, 2001; Penton, Hebert & Crease, 2004; Lefébure et al., 2006a; Lefébure et al., 2007) and caudatan amphibians (Bonett & Chippindale, 2004), and even to describe new species of large mammals (Dalebout et al., 2002). It has been proposed as one of the criteria for assigning species rank to allopatrically occurring populations of birds (Helbig et al., 2002). Ideally, a threshold divergence would exist that is higher than the highest intraspecific and lower than the lowest interspecific divergence in a given taxonomic group (Meyer & Paulay, 2005). Lefébure et al. (2006b) examined molecular divergence threshold values for crustacean species, genera and families. The divergence level of the COI gene optimally separating intra- from inter-species divergence was 0.16 substitutions per nucleotide position measured by maximum-likelihood estimated patristic distances. Because no such data are available for 28S rDNA sequences, we calculated the average divergence of known or postulated sister taxa as landmark for the delimitation of putative cryptic species (e.g. Johns & Avise, 1998). Patristic distances for pairs of nominal sister species were extracted using the program Patristic (Fourment & Gibbs, 2006). Patristic divergences extracted from maximum likelihood trees inferred under realistic models of evolution are expected to reduce the impact of saturation when comparing divergence estimates across a wide taxonomic range (e.g. Lefébure et al., 2006b). We did so with the Niphargus data set and with various sets of salamander mtDNA sequences, for which fragments homologous to our sequences of Proteus were available in GenBank.

Type 2 cryptic diversity (among non-monophyletic lineages) can be assessed from the tree topology as lack of genealogical concordance. In principle, we looked for cases in which individuals assigned to the same nominal species did not cluster together. Wherever possible (for N. virei, N. rhenorhodanensis, Proteus, Troglocaris), we used two or more molecular markers to corroborate each inferred case of cryptic diversity (Table 1). Because only the 28S data set was available for the genus-wide phylogeny of Niphargus, we had to rely only upon measurement of phylogenetic support (i.e. non-parametric bootstrapping of maximum likelihood inferences and Bayesian posterior probabilities).


Distribution areas of morphologically defined stygobiotic species

The largest known range of any macro-stygobiont is that of Niphargus aquilex (Schiödte). From its type locality in England to the Louros River spring in Greece it measures more than 2300 km. Other Niphargus species spanning large parts of Europe and the Middle East are Niphargus tauri (Schellenberg) (1900 km), N. longicaudatus (Costa) (1500 km), N. caspary (Pratz) (1350 km), N. nadarini (Alouf) (1100 km), N. pachypus (Schellenberg) (1000 km), N. pupetta (Sket) (1000 km), N. bajuvaricus (Schellenberg) (900 km), N. serbicus S. Karaman (900 km), N. virei (850 km), N. longidactylus (Ruffo) (550 km), N. grandii (Ruffo) (500 km), N. rhenorhodanensis (about 500 km), and N. hebereri (Schellenberg) (350 km). Because of the inaccuracy of some site coordinates, range sizes were rounded to the nearest 50 km. There are 34 additional taxa (species or subspecies) recognised by current taxonomy with ranges of 200–500 km in length. A complete list can be found in the electronic supplement.

Among other large-ranged stygobionts with well documented distributions are some holodinaric species (sensuSket, 2005), such as the European cave salamander P. anguinus (530 km) and the cave shrimps Troglocaris anophthalmus (Kollar) (590 km) and Troglocaris hercegovinensis (Babić) (530 km).

There are other macro-stygobionts in Europe with large distribution areas, particularly among isopods (Magniez, 1968; Henry, 1971; Henry & Magniez, 1988, 2003). Their ranges measure from several hundred up to about 1000 km. Examples include Proasellus cavaticus (Leydig) (1060 km), P. walteri (Chappuis) (470 km), P. lescherae Henry & Magniez (310 km), Bragasellus lagari Henry & Magniez (550 km) and Stenasellus virei virei Dollfus (350 km). Several North American amphipods, such as Stygobromus allegheniensis (Holsinger) (550 km), S. mackini Hubricht (400 km), Bactrurus mucronatus (Forbes) (700 km), B. hubrichti (Shoemaker) (400 km), appear to be distributed over areas of approximately the same size (Holsinger, 1978; Koenemann & Holsinger, 2001).

Nevertheless, these large-ranged taxa represent a relatively small percentage of the entire stygiobiont diversity. For example, in the Niphargidae (n = 318) single-site endemics account for 49% of the species. Only 18% exhibit ranges measuring more than 200–300 km across the longest diagonal.

Distribution areas of molecular lineages

Niphargus (Amphipoda: Niphargidae) The N. virei study revealed that the taxon previously regarded as a single widespread species consists of three distinct lineages, one from Benelux, one from the Jura and one from a wider north–south stretch in east-central France (Table 2). This result is corroborated by data from two independent molecular loci – a fragment of the mitochondrial gene encoding COI and a partial sequence of 28S rDNA, a nuclear gene. The species N. virei as perceived based on morphological traits was apparently monophyletic and thus could be regarded as an example of type 1 cryptic diversity. Molecular divergences between COI haplotypes of different lineages ranged from 0.29 to 0.52 substitutions per site. The most diversified of the cryptic lineages has the largest distribution – about 700 km from Metz in north-eastern France to the Montpellier area in southern France (Fig. 2).

Table 2.   Cryptic molecular diversity found in widespread Niphargus species
Nominal morphological speciesDistributionSize (km) Cryptic taxon*DistributionSize (km)Source of data
  1. *Taxa are named as in Fig. 2. Cryptic lineages are indicated by ‘‘cf.,’’ except for lineages sampled at or close to the type locality.

  2. Single site data or data from insufficiently sampled areas.

  3. The taxon is not a strict stygobiont; epigean populations may occur.

Niphargus aquilex SchiödteEurope between U.K. and Greece2300N. aquilexSouth-east England, U.K.45Fig. 1
N. cf. aquilex 1Spring of Louros River, Vouliasta, Ioannina, Greece0Fig. 1
N. cf. aquilex 2Megara Pećina Cave, Podgorica, Montenegro0Fig. 1
N. cf. aquilex 3Huda Luknja, Velenje, Slovenia0Fig. 1
N. cf. aquilex 4Podutik, Ljubljana, Slovenia0Fig. 1
N. cf. aquilex 5Wavreilles, Rochefort, well 400 m west of Lavreilles, Belgium0Fig. 1
Niphargus arbiter G. KaramanNorth-western Dinaric region300N. arbiterNorth-western Dinarides, between Lika, Croatia, and Lušci, Bosnia Herzegovina110Fig. 1
N. cf. arbiter 1Kvarner islands Krk and Rab, HR26Fig. 1
Niphargus fontanus BateWestern and Central Europe1400N. fontanusSouth-eastern England and western France295Fig. 1
N. cf. fontanus 1Central France0Fig. 1
N. cf. fontanus 2Germany0Fig. 1
Niphargus illidzensis dalmatinus SchäfernaCentral Balkan Peninsula430N. illidzensis dalmatinusSpring Biba, Vrana, Zadar, Croatia0Fig. 1
N. illidzensis cf. dalmatinus 1Between Central Bosnia Herzegovina, and Belgrade, Serbia200Fig. 1
Niphargus krameri (Schellenberg)Istra, CRO, also ITA, SLO73N. krameriPazin flysch area, Istra, Croatia47Fig. 1
N. cf. krameri 1Čičarija, Slovenia and Croatia31Fig. 1
Niphargus longicaudatus (Costa)South-east Europe and Italy1470N. longicaudatusVicinity of Napoli, Italy0Fig. 1
N. cf. longicaudatus 1Spring on Stražbenica, Velebit, Croatia0Fig. 1
N. cf. longicaudatus 2Central Italy, between Napoli and Gargano130Fig. 1
N. cf. longicaudatus 3Islands Krk and Cres, Croatia40Fig. 1
N. cf. longicaudatus 4Spring Škrip, Brač, Croatia0Fig. 1
Niphargus longidactylus (Ruffo)Northern Italy and north-western Balkans540N. cf. longidactylus 2Hyporheic zone of Sava River, Ljubljana, Slovenia0Fig. 1
N. cf. longidactylus 1Želimlje, Ljubljana, Slovenia0Fig. 1
Niphargus rhenorhodanensis (Schellenberg)Eastern France450N. cf. rhenorhodanensis ABCPre-Alps and eastern pre-Alps, France150Lefébure et al. (2007)
N. cf. crhenorhodanensis DEFrench Jura and Massif Central, France170Lefébure et al. (2007)
N. cf. rhenorhodanensis FGFrench Jura (interstitial habitats)160Lefébure et al. (2007)
N. rhenorhodanensis HFrench Jura,0Lefébure et al. (2007)
N. cf. rhenorhodanensis IFrench Jura0Lefébure et al. (2007)
N. cf. rhenorhodanensis JKSouth-western Alps, France0Lefébure et al. (2007)
Niphargus salonitanus S. KaramanSouth-eastern Dinaric region260N. salonitanusSouth-eastern Dinaric region along Adriatic coast, Croatia195Fig. 1
N. cf. salonitanus 1Bikovica Cave, Pirovac, Zadar, Croatia0Fig. 1
Niphargus tatrensis WrzesniowskyCentral-eastern Europe430N. tatrensisLodowe Zrodlo, Koscieliska valley, Zakopane, Poland0Fig. 1
N. cf. tatrensis 1North-eastern Slovenia33Fig. 1
Niphargus tauri (Schellenberg)Balkans and Turkey1880N. cf. tauri 1Eskisehir, Yesiltepe district, Turkey0Fig. 1
N. cf. tauri 2Iška, Ljubljana, Slovenia0Fig. 1
N. cf. tauri 3Prepadna Jama Cave, Kočevje, Slovenia0Fig. 1
N. cf. tauri 4Koblarska jama, Kočevje, Slovenia0Fig. 1
N. cf. tauri 5Sitarjevec, Litija, Slovenia0Fig. 1
Niphargus virei ChevreuxEastern France, Benelux840N. cf. virei AEast-central France700Lefébure et al. (2006a)
N. virei BFrench Jura, France180Lefébure et al. (2006a)
N. cf. virei CBenelux100Lefébure et al. (2006a)
Figure 2.

 Ranges of some cryptic Niphargus species and/or lineages. Single-site cryptic endemics are represented on the map by a single mark, whereas cryptic species and/or lineages with larger distributions are drawn on white background.

What has been considered as N. rhenorhodanensis based on standard morphological characters proved to be a complex of at least six highly divergent lineages with interspersed, partly sympatric distributions. Both type 1 and type 2 cryptic diversity were observed among them, because the assemblage was only partly monophyletic. Range sizes of the cryptic species were all small compared to the distribution of the species complex, with two of the cryptic species occurring at single sites only, and the largest spanning less than 200 km. The genetic separation of these lineages was supported by complete genealogical consistency and mutual exclusivity of haplotypes and alleles. Furthermore, the genetic divergence between these groups measured by COI patristic distances equalled or exceeded the general crustacean threshold separating interspecific from intraspecific distances (0.16–0.32 substitutions per sites).

The 28S rDNA tree (Fig. 1) revealed considerable discordance between species delineated by morphological criteria and molecular lineages. Of 18 nominal species or subspecies represented by more than one population, five displayed type 2, up to seven indications of type 1, and two showed signs of both types of cryptic diversity. Type 1 cryptic diversity was considered by reference to molecular divergences observed between undisputed sister species and in the two previously discussed Niphargus cases (N. virei and N. rhenorhodanensis). This consideration is only tentative because there was a great overlap between inter- and intraspecific 28S patristic distances (0.0109–0.1470 and 0.0017–0.0494, respectively), preventing definition of a threshold value. The divergences between putative cryptic sister lineages were distributed around the median interspecific divergence or higher, ranging from 0.0259 (between the type lineage of Niphargus fontanus Bate and N. cf. fontanus 1 from Central France) to 0.0743 between N. virei cryptic lineages from Benelux and east-central France.

Figure 1.

 Maximum likelihood tree depicting phylogenetic relationships and cryptic diversity in the genus Niphargus based on 28S rDNA sequences. Species (in one case a subspecies) are identified and named according to valid taxonomic descriptions and identification keys currently in use. Shaded boxes denote cryptic diversity between putative sister lineages (type 1 cryptic diversity), tentatively identified based on the degree of sequence divergence. Cryptic taxa within a single nominal ‘‘morphological’’ species that are scattered across the tree (type 2 cryptic diversity) are indicated by symbols (see legend). Individuals from the type locality or a nearby site to it are named as nominal species. All putative cryptic new species are marked by ‘‘cf.’’ preceding the species epithet. Numbers next to nodes represent their likelihood bootstrap values and Bayesian posterior probabilities, respectively, in per cent.

In the most extreme case of N. aquilex, the 2300-km range broke apart into seven narrow areas, mostly single sites, spread over the entire range of the nominal species defined by morphological criteria (Fig. 2). Furthermore, none of the large species-ranges was corroborated by the 28S rDNA phylogeny, indicating mostly very narrow distributions, reaching 300 km only in N. fontanus from north-western Europe (Table 2, Fig. 2). In some cases, exact assessment of the type of cryptic diversity was impossible because of the lack of node support (e.g. Niphargus cf. aquilex 1 and 2 in Fig. 1). In such cases the degree of divergence was used as an additional criterion.

It has to be noted that not all Niphargus species analysed can be considered as strict stygobionts. Some have been observed in ditches, small streams or springs, although their exact habitat preferences are often not well known. Among the taxa displaying cryptic diversity, Niphargus krameri (Schellenberg), N. rhenorhodanensis, N. longicaudatus, and Niphargus illidzenzis dalmatinus (Schäferna) have been recorded from one or more epigean habitats. Dispersal of some of these taxa (e.g. N. krameri) may depend mostly on surface waters (Fišer, Sket & Stoch, 2006).

Troglocaris and Spelaeocaris (Decapoda: Atyidae) The phylogenetic analysis revealed that the morphological characters originally used to define both large-ranged species T. anophthalmus and T. hercegovinensis are not suitable to distinguish between them. Some northern populations showing characters from the original description of T. hercegovinensis (Babić, 1922), clearly belonged to the T. anophthalmus clade. Our careful subsequent morphological inspection revealed differences in the structure of male pleopods I-II that were not taken into account in standard descriptions. Within the actual T. anophthalmus clade, four geographically separated cryptic lineages were identified, dissecting the old 590 km-range into four smaller ones, measuring 235 km at most (Adriatic clade; Table 3). A new taxon of Spelaeocaris (Matjašič), (provisionally called para-pretneri clade) discovered by the molecular phylogenetic analysis revealed, after morphological inspection, distinctive characters separating it from all other groups of Dinaric cave shrimps. It was composed of two genetically and morphologically well-distinguishable lineages with relatively small distribution areas.

Table 3.   Cryptic and overlooked molecular diversity found in Troglocaris sensu lato species from the Dinaric Karst*
Nominal morphological speciesDistributionSize (km)Cryptic taxonDistributionSize (km)
  1. *Data and taxa names from Zakšek et al. (2007)

  2. Single site data or data from insufficiently sampled areas.

  3. An entirely new taxon discovered by molecular analysis, occurring sympatrically with other species, morphologically distinct. n.a., not applicable.

Troglocaris anophthalmus (Kollar)Dinaric region590Western Slovenian cladeSouth-western Slovenia and Istra, Croatia75
Eastern Slovenian cladeSouth-eastern Slovenia and Istra, Croatia61
Adriatic cladeSouth-eastern Croatia and southern Herzegovina, along Adriatic coast235
Bosnian cladeNorth-western Bosnia, Bosanska Krajina13
Troglocaris hercegovinensis (Babić)Dinaric region545Troglocaris hercegovinensisBetween southern Herzegovina and southern Montenegro99
Spelaeocaris pretneri Matjašič plus ‘‘Para-pretneri clade’’n.a.Southern lineageCave Obod in southern Herzegovina, and cave Obodska Pećina in Montenegro94
Northern lineageCave Suvaja Pećina, Sanski most, Bosnia-Herzegovina0

The genetic separation of cryptic and overlooked Troglocaris lineages was supported by genealogical consistency and mutual exclusivity of haplotypes from one nuclear (28S rDNA) and two mitochondrial genes (COI, 16S rDNA). The degree of COI divergence between cryptic sister lineages ranged from 7.4% to 27%. While the genetic isolation between the least divergent groups was unequivocally supported by unlinked loci, the COI divergence was below the 16% threshold optimally separating inter- from intraspecific COI distances in crustaceans (Lefébure et al., 2006b).

Proteus anguinus (Caudata: Proteidae) The phylogenetic analysis grouped the haplotypes into six distinct, well supported groups. All haplotype lineages were geographically defined with no overlap in ranges. Likewise, there was no sharing of haplotypes between lineages. The ranges covered by the cryptic lineages were much smaller than the range of the nominal species, measuring 205 km at most (Southern Coastal Region clade; Table 4). Interestingly, the non-cryptic taxonomic subdivision of the species into two subspecies, the troglomorphic nominotypical P. a. anguinus and the dark-pigmented P. a. parkelj Sket & Arntzen, was found to be the matter of a local intra-lineage diversity nested within the south-eastern Slovenia clade. Such a phylogenetic position of P. a. parkelj was suggested previously by allozyme polymorphism (Sket & Arntzen, 1994). The cryptic lineage from the Istra Peninsula was one of the most divergent and the most likely candidate for the sister group to all other Proteus lineages (Gorički & Trontelj, 2006). Comparative analyses showed that the degree of divergence between cryptic Proteus sister lineages was similar to or exceeded the divergence between established salamander sister species (Fig. 3). A very conservative interpretation of these results would imply a tripartite cryptic structure, composed of the Istra lineage, the Southern Coastal Region + Lika-Krajina lineage, and the two Slovenia lineages (comparison G in Fig. 3). The presented ranges of cryptic lineages are based on a slightly less conservative subdivision into six lineages (comparisons A through F in Fig. 3).

Table 4.   Cryptic molecular diversity found in Proteus anguinus occurring in the Dinaric region*
Nominal morphological speciesSize (km) Cryptic taxon*DistributionSize (km)
  1. *Data and taxa names from Gorički & Trontelj (2006)

  2. Single site data or data from insufficiently sampled areas.

  3. A wider distribution on the Istrian Peninsula is documented by old samples only; no recent records are known from other sites.

Proteus anguinus Laurenti530Istra cladeIstra, Croatia0
Bosanska Krajina cladeBosanska Krajina, Bosnia Herzegovina13
Lika cladeLika, Croatia18
Southern Coastal Region cladeSouth-eastern Croatia and Southern Herzegovina, along Adriatic coast205
South-western Slovenia cladeSouth-western Slovenia41
South-eastern Slovenia cladeSouth-eastern Slovenia49

Range reduction due to cryptic diversity

Even with some of the evidence based on a single gene only, each of the 14 analysed nominal stygobiotic species with a distribution area measuring more than 200 km proved to be subdivided into two or more cryptic lineages. Even within the relatively small-ranged and often epigean N. krameri (70 km) a second, cryptic lineage was recorded. The degree of divergence between these lineages was within the range of divergence of other related pairs of sister species or above it. As a result, 15 nominal stygobiotic species as identified by currently valid and operational morphological criteria might be split into as many as 51 cryptic lineages. A large part (94%) of these lineages were found to occupy areas measuring about 200 km or less in diameter (Fig. 4). One half (48%) of the cryptic lineages were recorded at single sites only. There were only two cryptic taxa whose ranges considerably exceeded 200 km, Niphargus cf. virei from east-central France (700 km), and the type lineage of N. fontanus from Great Britain and the French mainland (300 km).

Figure 4.

 Size of distribution ranges of cryptic stygobiotic lineages plotted against the range sizes of their original species identified by morphology. Crosses represent a hypothetical case of no cryptic diversity. Black dots stand for ranges established by comprehensive geographic sampling, white dots for single-site data or insufficiently sampled areas. Taking into account cryptic diversity, 95% of all studied stygobiont distributions were smaller than 230 km. Based on Niphargus data presented in this paper, and on published data on N. virei (Lefébure et al., 2006a), N. rhenorhodanensis (Lefébure et al., 2007), Proteus anguinus (Gorički & Trontelj, 2006) and Troglocaris (Zakšek et al., 2007).


The fallacy of large ranges in groundwater fauna

The discovery of an apparent upper size limit of stygobiont ranges prompts for an explanation. The 100–200 km range we observed for some of the cryptic lineages disagrees with the traditional low-dispersal hypotheses presented in the Introduction, but there was also not a single case supporting the notion of an extensive continuous hypogean habitat, known as the interstitial highway hypothesis (Ward & Palmer, 1994). Instead, many of the studied taxa appear to be able to disperse and yet be confined by some boundaries beyond which dispersal is impossible. It is unclear, however, what exactly these boundaries are. Present-day catchments do not or only partly match the distribution areas of cryptic lineages of N. virei, N. rhenorhodanensis, T. anophthalmus and Proteus. Nevertheless, the upper size of present-day catchments in karst massifs appears to correspond to the size of the established ranges. For example, the main catchments of the Dinaric Karst (Cetina, Kolpa/Kupa, Krka, Ljubljanica, Neretva, Reka, Trebišnjica, Una) measure approximately 55–150 km across their longest diagonal. It thus seems likely that the ranges of stygobionts reflect historical rather than current hydrogeological conditions. The cladogenetic events that gave rise to the stygobiotic taxa studied here occurred several million years ago (Verovnik et al., 2004; Lefébure et al., 2006a; Lefébure et al., 2007; Zakšek et al., 2007). The phylogeographic patterns found in these studies can be best explained by pre-Pleistocene geological and hydrogeological events (Sket, 1994). Consequently, although the restricted size of karstic catchments may be a factor limiting the ranges of macro-stygobionts, catchment size needs to be considered from a historical perspective.

Most large-scale geographic patterns of groundwater fauna have partially remained obscure because of the difficulty of sampling subterranean habitats (Gibert & Deharveng, 2002; Culver et al., 2006). The presented studies are no exception. The phylogenetic analysis of the Niphargus was not directly designed to measure the distribution of cryptic lineages as we did not expect to find such a high degree of cryptic diversity at the outset of the investigation. Therefore, the possibility to underestimate range sizes due to insufficient geographic sampling cannot be ruled out. Lefébure et al. (2007) analysed 12 sites with N. rhenorhodanensis and identified six cryptic taxa. Thus, more intensive sampling is likely to reveal additional cryptic taxa and may clarify range limits.

Nevertheless, the sampling coverage for N. virei, Troglocaris and Proteus was quite comprehensive. At least for these taxa it seems unlikely that any widely distributed cryptic lineage has been missed. Supported by previous considerations (e.g. Holsinger, 2000; Sbordoni et al., 2000), the results of our study suggest that it should be the large-ranged stygobiotic species that need justification through morphological and molecular scrutiny, not the endemics. In a similar way as the proportion of for instance Holarctic species among cyclopoid copepods has decreased from 68% to 28% due to improved taxonomic approaches (Reid, 1998), the share of widely distributed stygobiotic species is expected to decline substantially as additional molecular data become available.

The need for morphological re-investigation

The discovery of cryptic diversity can help to solve the paradox of widely spread versus endemic stygobionts, but should not be an end in itself. Stygobiotic species with large ranges tend to be traditional, ‘‘old’’ species. A substantial proportion of them were described before 1900, while only a fraction of the currently known endemics and medium-ranged species were known before 1900 (e.g. Botosaneanu, 1986). Often we are dealing with taxa that need taxonomic revision according to present-day standards. The formal taxonomic changes that ought to follow might rely on molecular data only using so called DNA taxonomy, as outlined in the Introduction, but in many cases additional diagnostic characters are expected to emerge after morphological re-investigations. The case of Troglocaris demonstrates the importance of molecular systematics to pinpoint cases of ‘‘overlooked’’ diversity. All three caves with Para-pretneri representatives have been biologically sampled several times before, and this taxon has never been detected – most probably due to taxonomic inertia caused by its co-occurrence with other species of Troglocaris sensu lato. To find two or more closely related decapod species sharing the same habitat in the same cave was utterly unexpected.

On the other hand and despite the fact that N. virei lineages have been estimated to have diverged at least 14 million years ago (Lefébure et al., 2006a), extensive morphological investigations in pre-molecular times revealed no taxonomically distinct breaks in the morphological variation. An allometric analysis (Ginet, 1960) based on 550 individuals from populations belonging to both the Jura and eastern France clades (as defined in Lefébure et al., 2006a) and considering 11 measured characters, revealed almost no differences with the exception of whole-body size and allometric rates of two characters (gnathopods and antennae). As Ginet's (1960) work also revealed that such allometric variation could be observed in certain rearing conditions, he stated that new taxonomic units could not be created based on such subtle character variations. While further morphological examinations of N. virei are now warranted in view of the present results, the extent to which molecular diversity can be explained by ‘‘overlooked’’ diversity remains to be evaluated.

Reliance on molecular taxonomy only is unlikely to lead to a solution of the cryptic diversity problem. Cryptic species are commonly defined as species indistinguishable by morphology; hence, morphological scrutiny is an essential element in the assessment of cryptic diversity. It is also clear that not only morphology, but also molecular analyses are likely to neglect phylogenetically young species (Meyer & Paulay, 2005). While molecular data are likely to be less sensitive to convergent evolution, they still can be influenced by a number of factors obscuring the true relationships. For example, assessment of type 1 cryptic diversity could be misled by an extreme variation in the rate of molecular evolution among lineages. However, variation among closely related lineages is expected to be readily detectable on a phylogram. False cryptic diversity of type 2 can be indicated by the survival of an ancestral molecular polymorphism across multiple speciation events (i.e. incomplete lineage sorting). Here only analyses of additional loci can serve as a test. Thus, even if one line of evidence (28S rDNA) indicates an impressive level of cryptic diversity for the niphargids as a whole, we regard our current estimate of this diversity as provisional until putative cryptic lineages are corroborated by data on additional loci and individuals. Our results of the other taxa (N. virei, N. rhenorhodanensis, Troglocaris, Proteus) are much more robust, although divergence thresholds are always somewhat subjective.

Cryptic diversity and conservation biology

Hidden diversity is not peculiar to groundwater habitats. A search in scientific literature databases (http://isiknowledge.com/, http://www.scirus.com/, http://scholar.google.com/) in April 2007 revealed about 500 titles dealing with cryptic diversity and over 2000 dealing with cryptic species. Judging on the basis of numerous published cases, the existence of cryptic species appears to be a common phenomenon in taxonomically poorly studied and widespread epigean species with limited gene flow. However, the overall species richness is much higher in epigean aquatic ecosystems than in ground water (Marmonier et al., 1993; Sket, 1999b; Danielopol et al., 2000). Consequently, the expected proportion of cryptic taxa is likely to be much higher in the latter. In the case of Proteus, for example, taking into account cryptic lineages would lead to a sixfold increase in the regional diversity. As spectacular as this may sound, such a development has been predicted for many taxonomic groups (e.g. Danielopol et al., 2000).

What was not explicitly predicted is the substantial change in the geographic pattern of diversity and endemism, following the breakup of nominal species defined by morphological criteria into narrowly distributed cryptic taxa. The consequences of this finding for biodiversity assessment and conservation planning go beyond a simple numerical correction. The foreseeable implications for groundwater biodiversity not only are considerable increases in diversity at the regional and continental scale but also a decrease in faunal similarities among regions, coupled with a rise of endemism. These effects ought to be considered in future developments of conservation policies for groundwater ecosystems.

Two points are particularly important: First, since virtually all stygobiotic macrofauna is endemic to small or medium-sized hydrographic groundwater units, ground water as a habitat deserves greater conservation attention. Although groundwater sites cannot compete with epigean sites as hotspots of overall diversity, some of them clearly represent European endemism hotspots. Second, the near complete lack of faunal overlap between regions renders ineffective conservation strategies based on the protection of most important sites. For example, trying to protect Proteus in Slovenia by safeguarding its stronghold in the Postojna-Planina Cave System would neglect the cryptic lineage from the south-eastern Slovenian karst. Any conservation strategy aiming at preserving Proteus by focusing on the strongest populations would miss the small, isolated and divergent taxon from the Istra Peninsula. For effective conservation of macro-stygofauna at the European level, each groundwater catchment or karstic hydrological unit will need to become a site of general concern, since any regional extinction would mean a biodiversity loss at the European scale.

In conclusion, our molecular survey indicates that only a small fraction of stygobiotic species with supposedly large ranges have large ranges in reality, and the sizes of the actual ranges are only fractions of the sizes previously supposed. Molecular systematics has helped solve the paradox of widely distributed groundwater species by unearthing considerable cryptic diversity within several species. These results lend credence to a few commonly invoked hypotheses (see Introduction) explaining the common view that groundwater species have narrow distributions. The opposite view – that of high dispersal potential (e.g. Ward & Palmer, 1994) and abilities (e.g. Danielopol et al., 1994) – appears to be less relevant for macro-stygofauna. Interstitial species have been traditionally considered to have much broader distributions than karstic stygobionts (Sket, 1994; Danielopol et al., 2000), and their habitat was described as being spatially continuous at the continental scale (Ward & Palmer, 1994). However, as indicated by the cases of the interstitial N. longidactylus (Fig. 1) and generalist N. rhenorhodanensis (Lefébure et al., 2007), this generalisation might not hold for all species living in interstitial groundwater habitats.


Our thanks go to (in alphabetical order) Uroš Arnuš (Maribor, Slovenia), Elzbieta Dumnicka (Krakow, Poland), Andreas Fuchs (Landau, Germany), Reinhard Gerecke (Tübingen, Germany), Jos Notenboom (Bilthoven, the Netherlands), Tonči Radja (Split, Croatia), Mustafa Tanatmiş (Eskişheir, Turkey), Paul J. Wood (Loughborough, U.K.), who furnished Niphargus specimens. We are grateful to a number of friends and colleagues who helped us with the fieldwork: Gregor Bračko, Andrej Hudoklin, Katarina Jazbec, Franc Kljun, Simona Kralj-Fišer, Ivan Kos, Slavko Polak, Simona Prevorčnik, Rudi Verovnik, Maja Zagmajster (all Slovenia), T. Datry, J. Issartel, J. Lips and M. Meyssonier, J. Arnaud, L. Bruxelles, G. Deflandre, T. Dubourget, B. Hamon, M. Meyssonier, G. Michel, S. Ranchin, M. Wiemin and F. Malard (all France). We are further grateful to Jožica Murko-Bulič (Ljubljana, Slovenia), Laure Granger, and Catherine Lerondelle (Lyon, France) for their indispensable help in the laboratory, to Gregor Bračko for gathering distribution data, and to Maja Zagmajster for her help with GIS work. The work was financially supported by the Slovenian Research Agency and through contract n° EVK2-CT-2001-00121 (PASCALIS) of the Fifth Research and Technological Development Framework Programme of the European Union.