Introduction

A fundamental problem for applied as well as basic biodiversity research is that we often do not know to what extent morphological and genetic diversity are correlated. Phenotypic plasticity can lead to pronounced morphological differentiation despite genetic similarity. At the other extreme, morphological stasis can completely mask genetic diversification, and molecular analyses are necessary to identify distinct lineages, subspecies or species within a morphospecies. Due to their separate evolutionary histories, such genetic groups may possess unique adaptations and evolutionary potential and, hence, may be distinct evolutionarily significant units [1] relevant for conservation. This is especially true for cryptic species, which are often reproductively isolated and therefore true species under the biological species concept [2], and often millions of years old [3], but morphologically indistinguishable.

In some taxa, these issues are well understood and are considered in protection and management programs, for example in fishes, like trout Salmo trutta and Atlantic salmon Salmo salar [4]. However, many other common and widespread taxa might exhibit similarly strong cryptic diversity (e.g. [3], [5]). As cryptic species may differ in biological characteristics, correct species identification will be fundamental to ensure the comparability between studies in basic and applied research. Knowledge on the geographical distribution of cryptic taxa will be greatly beneficial in this respect, because in case phylogeographic signals are strong, it may be possible to reliably predict the species present in a particular region without time-consuming molecular analyses.

The organism we investigate in this study is the freshwater amphipod Gammarus fossarum Koch (Crustacea, Amphipoda). Amphipods are a central element of aquatic ecosystems, for example as fish prey [6] or shredders of organic material [7] and intermediate hosts for several fish and bird parasites of the phylum Acanthocephala [8]. G. fossarum is widespread in Central Europe [9], [10], especially in the upstream reaches of streams [11]. G. fossarum is vulnerable to human activities [12], [13] and is often used in ecotoxicological assays, or in ecological assessment as indicator of habitat quality, for example in the “Modul-Stufen-Konzept” for Switzerland [14].

Several studies indicate the presence of at least three cryptic species within G. fossarum. As early as 1989, Scheepmaker and van Dalfsen found large allozyme differentiation within G. fossarum in Europe [15], leading to the distinction between G. fossarum sensu stricto and G. fossarum sensu lato. This subdivision was refined by Müller [16], [17] using sequencing and allozyme data to describe three genetically distinct cryptic species - types A, B and C. Allozyme and mtDNA (16S rRNA gene) genotypes were concordant in mixed type A / type B populations, suggesting complete reproductive isolation between the types [17]. Large genetic differentiation indicates that the species split several million years ago [17], while morphological differentiation is not sufficient for species discrimination [18].

Müller found a geographical distribution pattern with type A in the east and type B in the west of his study area (mostly central and Southern Germany and surrounding areas), while type C was rare and only occurred in western populations [17]. However, as G. fossarum prefers upstream reaches of streams [11], it tends to be rarer than other Gammarus species in lowland areas like the Netherlands and many regions of Germany [13], [19], [20], [21]. Previous studies analyzing the G. fossarum cryptic species distribution have therefore not focused on the areas where G. fossarum is the dominant Gammarus species, and consequently ecologically especially relevant, like Switzerland or Austria (own observations and [22]).

In this study we compile data from previous studies analyzing the cryptic species distribution and complement this data set with new samples mainly from Switzerland, applying a novel pyrosequencing assay for rapid molecular species identification. The resulting data set allows a detailed description of the current species distribution in Central Europe and insights into the historical processes which produced it.

Results

In Switzerland, we analyzed 62 G. fossarum populations with regard to cryptic species composition. Complete 16S sequencing of 137 individuals from across Switzerland revealed all three cryptic species described by Müller [17]. Based on the full sequencing and pyrosequencing dataset, we detected 24 type A populations, 26 type B populations, and three type C populations (Table S1) in Switzerland. Nine populations contained both type A and type B at variable frequencies.

The distribution pattern in Switzerland (Figure 1) can be described as follows: In the east (parts of Rhine drainage, Danube drainage and Po drainage), only type A populations were found. Further west, there was an area where both type A and B occurred, occasionally within the same population. This contact zone was located within the Rhine drainage. In this area, type A was generally rarer than type B, i.e. there were fewer type A than type B populations, and within mixed populations, type A was mostly the rarer species. Four of the five Rhone populations contained type B, while type A was not observed in this drainage. Only three type C populations could be detected, which were all in the very west of Switzerland, but belonged to different drainages (Rhine and Rhone).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 1. Distribution of three cryptic Gammarus fossarum species (types A, B and C) in Europe.

The Rhine drainage, which contains the contact zone between type A and type B, is shaded in grey. The small map on the top right indicates the position of the study area within Europe. The large map contains data from six previous publications (see text) and the present study. The method used for species identification is indicated (square = allozyme analysis; circle = sequencing), as the distinction between type B and C is not possible based on allozymes. For some sites, arrows indicate the flow direction of the sampled stream. The smaller map contains only the Swiss samples collected for this study and shows the exact species compositions of the samples in pie diagrams. Pie diagram size corresponds to sample size (n between 2 and 57 per population; total n = 1337). To prevent overlap of the pie diagrams, some positions are slightly shifted. The exact positions are shown in the large map.

https://doi.org/10.1371/journal.pone.0023879.g001

We compiled data from 6 studies which had analyzed 117 G. fossarum populations in Western and Central Europe (Figure 1; Table S2). The general pattern that emerged again was that type A dominated in the east (Danube drainage), while types B and C were more common in the west (Meuse, Rhone, Weser, and four smaller French river systems draining into the Atlantic or Mediterranean). Type C was observed in only two populations (Rhine and Meuse drainage) out of the 21 populations where a distinction between type B and C was possible. Again, all three cryptic species were observed within the Rhine drainage, and 13 populations within this contact zone contained both types A and B/C (Figure 1).

Sequencing of 64 precopula pairs (pairs formed prior to mating) from a mixed population in Glovelier revealed 10 type A and 53 type B pairs and only a single heterospecific pair consisting of a type B male and a type A female.

Discussion

The synthesis of the data from this study and earlier publications clearly shows that, in many cases, the geographical location alone will be an accurate predictor of the cryptic G. fossarum species present at a particular site. Because of these mostly non-overlapping geographical distributions, the range of each cryptic species will be considerably smaller than that of the entire morphospecies. We found relatively clear distribution patterns: In Switzerland, type A is more common in the east, while type B is more frequent in the west (Figure 1). This pattern recurs on a larger scale across Central Europe: type A occurs in the eastern drainages (Danube and Po), while type B is common in the west (Meuse, Rhone, Weser, and four smaller French river systems). A long contact zone spans the Rhine drainage from central Germany down to the Alps in Switzerland, covering the whole latitudinal distribution range of the species complex at this longitude. This contact zone does not coincide with any obvious environmental clines, for example forests streams with stones as well as grassland streams with more macrophytes were available across the entire sampling region. There is also no major geological cline which could explain the differential distribution of the species. This suggests that the species distribution is heavily shaped by past recolonization processes, leading to mainly geographical instead of ecological sorting of the species.

It is currently unclear if type C is generally very rare, or if the centre of its distribution range was not covered in this survey. A small number of type C populations was observed along both sides of the Western border of the Rhine drainage. Only five samples where types B and C can be distinguished are available from further west in the Rhone and Meuse systems. Interestingly, all of these are type B, which would speak against a widespread occurrence of type C in Western Europe.

The finding that, outside the contact zone, geographical sampling location is highly predictive of species composition will be of interest in applied fields of research where correct species identification is important. Members of the G. fossarum cryptic species complex are regularly used in ecotoxicological studies (e.g. [23], [24]). They are sensitive to anthropogenic acidification and pollution [12], [13], which may cause local extinctions with potentially serious impacts on ecosystem functioning [13]. In such studies, species identification is normally limited to the level of the morphospecies even though the cryptic species are known to differ slightly in their ecological requirements [25], [26]. Consequently, they may also react differently to environmental change, and results or predictions obtained from one species need not hold for another.

For situations where molecular species identification is still necessary (i.e. when samples are taken from within the contact zone), pyrosequencing proved to be a rapid and cost-effective method for the large-scale identification of cryptic species from multiple populations. The assay is technically easy to implement and requires very little optimization. The efficiency of molecular species identification based on pyrosequencing could, in principle, be further increased through the analysis of bulk samples, i.e. mixtures of DNA from multiple individuals (e.g. [27]). However, preliminary tests showed that estimates of the relative frequency of types A and B in mixtures of known composition were unreliable. A possible explanation could be the large sequence divergence between the cryptic species, which could lead to unequal amplification efficiency in the PCR.

The evolutionary distinctness of the different G. fossarum types is underlined by the extent of reproductive isolation observed between them. Although the species were found to be partly interfertile under laboratory conditions ([28]; references therein), our small sample of precopula pairs indicates a preference for conspecific mates. This is in agreement with the results from genetic analysis, which suggested the absence of hybrids in mixed populations [17]. In the field, assortative mating could be promoted for example if the species sort into different microhabitats [26], which is not possible under lab conditions.

The observed geographical distribution of type A and types B/C is not only interesting for applied research, but also allows the inference of past recolonization processes. The cryptic G. fossarum species probably split well before the onset of the Pleistocene glaciations [17]. The fact that we observe a clearly separate geographic distribution of type A and types B/C indicates persistence of the species in different refugia during the Pleistocene glaciations. Müller [16] suggests that type A survived in an eastern and type B in a western refugium. The contact zone would have formed when the colonization fronts met in Central Europe. The eastern refugium was probably located in the Danube basin, where only type A occurs. The Danube basin also played a central role as a refugium for multiple fish species [29], [30]. From there, recolonization of Central Europe by type A might have taken place in a north-western direction to the west of the Danube drainage, as well as the Po and Rhine drainages. The location of the refugium of type B is more difficult to infer. One possibility is a south-western European refugium, as suggested by Müller [17], for example in France or even further south on the Iberian peninsula, one of the classical Mediterranean refugia [31]. In accordance with this idea, G. fossarum still occurs as far south as the Pyrenees [9] and the South of France seems to harbour high genetic diversity in the form of several endemic Gammarus species [32]. During the less severe glaciations in the late Pleistocene, more northern, secondary refugia may also have been available [33]. Such cryptic northern refugia have also been suggested for another freshwater crustacean Asellus aquaticus L. [34].

All probable recolonization patterns would have required repeated crossing of watersheds, for example of type C across the Rhine-Rhone boundary or a spread of type A out of the Danubian drainage (Müller [17] and Figure 1 of the present study). Movements across watersheds may be explained by past landscape changes which produced connections between drainages, for example the formation of post-glacial meltwater lakes (see e.g. Figure 5 in [35]), which might also explain the geographical distributions of genetic diversity in several fish species [35], [36], [37]. More recently, transport by waterfowl [38] or anthropogenic introductions (e.g.[39]) might occasionally have led to exchanges across watersheds.

Contact zones between cryptic species can complicate research and conservation efforts because the species at a given site cannot be inferred without molecular data. However, contact zones are interesting insofar as they can enhance our understanding of the coexistence of similar species, a topic that has been debated for decades [40]. Types A and B co-occur in a contact zone within the Rhine drainage that is quite narrow especially in the more northern part (Germany). Within the contact zone, we find remarkably few mixed populations where types A and B coexist (Figure 1), suggesting that interference or exploitative competition probably prevents long-term coexistence. Gammarus species often are strong intraguild predators [7]. If one species showed a stronger cannibalistic tendency, this could dramatically affect the probability of coexistence [41]. Additionally, or alternatively, one species could be better adapted to a particular local environment. Indeed, it has been shown that type A prefers stony habitats with leaves, while type B prefers plants and muddy substrate [26]. Such mechanisms would translate into unequal population growth rates and, ultimately, displacement of the poorer competitor. The few mixed populations we do find might have evolved ecological differentiation pronounced enough to allow coexistence [42] or they might represent transitional stages before the outcompetition of one species. It is possible that one species is currently expanding its range at the cost of the other [16]. Consistent with this scenario, type B populations do not show a pattern of isolation by distance, and a decrease in the mean number of allozyme alleles towards the contact zone in the German study area [16].

Methods

Supporting Information

Acknowledgments

Data analyzed in this paper were generated in the Genetic Diversity Centre of ETH Zürich, Switzerland.

We would like to thank Aria Minder and Christian Rellstab for technical support, and Maria Alp, Andreas Bruder and Michael Drees for providing Gammarus samples. Several people from the Eawag Aquatic Ecology department helped us with field work. Michael Döring and Rike Stelkens made comments that improved this manuscript.