Morphological vs. molecular delineation of taxa across montane regions in Europe: the case study of Gammarus balcanicus Schäferna, ^† (Crustacea: Amphipoda)

Introduction

Mountainous areas are characterized by high species richness due to the dynamic patterns of speciation they hosted, associated with ecological diversification, high local extinction risk and limited dispersal between fragmented habitats (Nagy and Grabherr 2009). In temperate areas, for example in Europe, these biodiversity hot spots are also known as important glacial refugia in which fauna and flora survived and diversified along Pleistocene glacial cycles (Hewitt 2000; Schmitt 2007; Bhagwat and Willis 2008; Provan and Bennett 2008; Holderegger and Thiel-Egenter 2009).

Thus, phylogeography of mountain organisms is currently under intensive studies. Yet, most of them focus on terrestrial species (e.g. Schönswetter et al. 2003, 2004; Hewitt 2004; Schmitt and Hewitt 2004; Schönswetter and Tribsch 2005; Magri et al. 2006; Varga and Schmitt 2008; Homburg et al. 2013). Already Bănărescu and Boscaiu (1978) as well as Malicky (1983) indicated that freshwater animals cannot be grouped with the terrestrial fauna from zoogeographical and distributional aspects. According to Bănărescu and Boscaiu (1978), the distribution of rheophilous taxa without terrestrial dispersal stages coincides with the catchment areas of rivers limited by watersheds, while the distribution of rheophilous insects coincides with mountain ranges. Different is also the case of subterranean species, which may be defined by paleodrainages, crossing the borders of recent river drainages (Sket 2002). Many aquatic taxa are little studied, and more extensive data sets are available only for vertebrates such as fishes (e.g. Bănărescu 1990; Bernatchez and Wilson 1998; Doadrio et al. 2002; Schreiber 2002; Pasko and Maslak 2003; Sonstebo et al. 2007) or amphibians (e.g. Szymura et al. 2000; Babik et al. 2003; Garcia-Paris et al. 2003; Martinez-Solano et al. 2006; Hofman et al. 2007; Fijarczyk et al. 2011). Among invertebrates, most studies concerned caddisflies (Pauls et al. 2006, 2009; Balint et al. 2008), snails (Falniowski et al. 2012) and beetles (Ciamporova-Zatovicova and Ciampor 2011). Generally, less attention was paid to species inhabiting mountain streams and rivers (Pauls et al. 2006, 2009; Balint et al. 2008; Falniowski et al. 2012), which remain mostly unexplored. Aquatic crustaceans have been the object of little attention, focusing predominantly on mountain lakes (Meyran and Taberlet 1998; Vainio and Väinölä 2003; Hamrova et al. 2012). In addition, most studies were either dealing with molecular phylogeography of particular species or species groups or were based on traditional taxonomic approach. Strikingly, there is deficiency of studies employing both molecular and morphological data to uncover and delineate cryptic taxa in case of mountain aquatic species or species groups. This is surprising, as we can expect that the above-mentioned properties of mountainous areas generate substantial cryptic diversity that could be easily detected, for example, with the DNA barcoding approach. Balint et al. (2008) demonstrated that allopatric populations of montane aquatic invertebrate species may diverge in their morphometric features. However, it would be interesting to find out whether such diversification is congruent with divergence observed on the molecular level. On one side, such studies would be very useful in addressing the problem of phenotypic plasticity in montane species. On the other side, they would stimulate quest for new species-diagnostic morphological features to differentiate between molecularly distinct lineages traditionally regarded as conspecific based on other morphological features.

Good model organisms for studying diversification in mountain aquatic environments are amphipod crustaceans, being among the most abundant taxa in mountain streams. They have a wide geographical distribution, narrow range of environmental tolerance and limited dispersal capabilities (Väinölä et al. 2008).

Gammarus balcanicus is a freshwater amphipod inhabiting streams in various, often isolated, mountain ranges. It is widely distributed in Europe, from Southern Alps through the Carpathians and Balkan Peninsula to the Crimean Peninsula (Karaman and Pinkster 1987). Its alpha-taxonomy has been the topic of debates (Schäferna 1922; Karaman 1929, 1935; Martynov 1931, 1935; Dobreanu and Manolache 1942; Cărăuşu et al. 1955) that, in the past, led to description of many local species and subspecies. However, Karaman and Pinkster (1987) could not define any diagnostic features that would differentiate these forms and synonymized them with G. balcanicus. This led us to the assumption of a putative substantial diversification pattern. Such pattern would occur between mountain ranges but might also occur between localities within a given range. Thus, our aim is to study the morphological and molecular variation within this species (as initially delineated based on conventional fauna diagnostic features) from a set of localities representing the actual distribution of the morpho-species to reveal putative pattern of its diversification.

To achieve this goal, we will answer the following questions:

Material and methods

Sample collection and identification

Samples were collected from 13 locations from different regions inhabited by Gammarus balcanicus in Europe, including the Crimean Peninsula, Carpathian Arch and Balkan Peninsula (Table 1, Fig. 1). Most of the samples were collected during several scientific expeditions organized by the Department of Invertebrate Zoology and Hydrobiology from 2005 to 2010. The samples from Crimea were collected within the bilateral (Slovenia-Ukraine) research project of Ljubljana and Kharkiv Universities in 2006. The animals were gathered from a variety of stream and spring habitats with use of a benthic hand-net, sorted at a site and preserved in 96% ethanol. In the laboratory, the collected individuals were identified under stereomicroscope Nikon SMZ-800 according to the available literature (Karaman and Pinkster 1977a,b, 1987; Jażdżewski and Konopacka 1989).

Table 1. List of sampling localities, including collection site codes, additional information's (geographical region, river basin, altitude, country, geographical coordinates, sampling date), number of specimens used in morphological (N1) and molecular analysis (N2) as well as the GenBank accession numbers

Code	Species	Locality	Mountain/region	River basin	Altitude (m asl)	Country	Coordinates	Collection date	N1	N2	Accession number
1	Gammarus balcanicus	Close to Dwerniczek	Eastern Carpathians	Vistula	576	Poland	49.212° N, 22.739° E	26.06.2008	20	12	KF053224, KF053225
2	G. balcanicus	North of Muczne	Eastern Carpathians	Vistula	697	Poland	49.149° N, 22.715° E	26.06.2008	20	9	KF053226, KF0532267
3	G. balcanicus	Between Yaroslyak and Vorokhta	Eastern Carpathians	Danube	983	Ukraine	48.190° N, 24.558° E	22.07.2010	20	12	KF053228, KF053229, KF053230, KF053231
4	G. balcanicus	Near Ovidiopol	Black Sea Lowlands	Dniestr	24	Ukraine	46.257° N, 30.419° E	22.08.2009	20	10	KF053232
5	G. balcanicus	South East from Simferopol	Crimea	Salhir	505	Ukraine	44.846° N, 34.213° E	26.08.2006	12	12	KF053233
6	G. balcanicus	Near Kizil-Koba (Red Cave)	Crimea	Salhir	503	Ukraine	44.867° N, 34.343° E	17.08.2006	10	10	KF053234, KF053235, KF053236, KF053237, KF053238
7	G. balcanicus	Near Catrinari	Eastern Carpathians	Danube	1004	Romania	47.191° N, 25.499° E	25.08.2005	20	12	KF053239, KF053240, KF053241, KF053242, KF053243, KF053244
8	G. balcanicus	20 km south of Cârţişoara	Southern Carpathians	Danube	933	Romania	45.667° N, 24.593° E	30.08.2005	20	11	KF053245, KF053246, KF053247, KF053248, KF053249, KF053250, KF053251
9	G. balcanicus	Zărneşti	Southern Carpathians	Danube	834	Romania	45.542° N, 25.296° E	29.08.2005	20	9	KF053252, KF053253, KF053254, KF053255, KF053256
10	G. balcanicus	Close to Luncavica	Dobrogea Plateau	Danube	93	Romania	45.228° N, 28.311° E	02.09.2005	20	8	KF053257, KF053258
11	G. balcanicus	Hunedoara	Transylvanian Plateau	Danube	255	Romania	45.751° N, 22.888° E	07.09.2005	20	8	KF053259, KF053260, KF053261, KF053262, KF053263
12	G. balcanicus	Near Kolašin (locus typicus)	Dinarides	Skadar Lake	1069	Montenegro	42.836° N, 19.569° E	24.09.2006	20	10	JX899354
13	G. balcanicus	Close to Burrel	Albanian central mounatin range	Mat	224	Albania	41.584° N, 20.031° E	16.09.2006	20	12	KF053264
BOS	Gammarus bosniacus	Vrelo Bosne	Dinarides	Danube	497	Bosnia and Herzegovina	43.820° N, 18.270° E	27.08.2006	–	1	JX899355
PLJ	Gammarus pljakici	Close to Andrijevica	Dinarides	Danube	828	Montenegro	42.729° N, 19.801° E	30.08.2006	–	1	KF053265
SAL	Gammarus salemaai	Lake Ohrid, Sv. Stefan	Helenides	Ohrid Lake	694	Macedonia	41.080° N, 20.799° E	–	–	1	JX899266
SKE	Gammarus cf. sketi	Sv. Naum springs	Helenides	Ohrid Lake	697	Macedonia	40.914° N, 20.742° E	–	–	1	JX899272
MAC	Gammarus macedonicus	Lake Ohrid, Sv. Zaum	Helenides	Ohrid Lake	761	Macedonia	40.949° N, 20.774° E	–	–	1	JX899221
PUL	Gammarus pulex	Öland	–	–	–	Sweden	–	–	–	1	JF965943
FOS	Gammarus cf. fossarum	Bled	Alps	–	–	Slovenia	–	–	–	1	JF965879
LAC	Gammarus lacustris	Liubliaz	Polesia	Pripyat	141	Ukraine	51.848° N, 25.472° E	09.08.2009	–	1	JX899356
VAR	Gammarus varsoviensis	Korostianka	Polesia	Pripyat	147	Ukraine	51.770° N, 25.354° E	09.08.2009	–	1	JX899357
LEO	Gammarus leopoliensis	Between Borsa and Prislop Pass	Eastern Carpathians	Danube	837	Romania	47.626°N, 24.768° E	25.08.2005	–	1	KF053266
ROE	Gammarus roeseli	Preševo	Preševo Valley	Danube		Serbia	–	–	–	1	JF965948
STO	Gammarus stojicevici	Near Pirot	Balkan Mts.	Danube	414	Serbia	43.217° N, 22.537° E	28.08.2007	–	3	KF053221, KF053222, KF053223

Results

Morphological analyses

Among the 23 features used for the morphological analysis, 20 were differentiating populations at the statistically significant level (Table 2). The first two dimensions of the MCA ordinance analysis performed on the above-mentioned 20 morphological traits explained only 47.32% of the total observed variance within the data set. Thus, we reduced the data set based on the criterion of the discrimination measure (DM) values associated to particular features (Table S1). Only the 15 features with mean DM values above 0.1 were included in the final MCA (Fig. 3), with 61.45% of the total variance explained by the first two dimensions, and 81.13% explained by the first three dimensions. The results of this analysis are presented on the 2D graph (Fig. 2). The observed ordination pattern was affected mostly by the following features with the associated mean DM values above 0.20 (Fig. 3, Table S1), (symbols given in brackets correspond to Table 2): number of setae on peduncle of antenna 1 (A, B), number and length of setae on peduncle of antenna 2 (F, G, H), length of setae on flagellum of antenna 2 (I), the length ratio of endopodit to exopodit of uropod 3 (K), type of setae on exopodit outer margin of uropod 3 (P), presence and type of setae on endopodit of uropod 3 (S, U, V). Other analysed features contributed to a lesser extent to separation of individuals, with discrimination measures from 0.13 to 0.17 (Fig. 3, Table S1). The 2D MCA ordinance graph does not provide clear distinction of individuals associated with different geographical localities, although many statistically significant differences may be observed between particular populations (Table S2). In 31 cases, the dimension 1 significantly differentiates the populations (Table S2); the features with highest DM scores (DM > 0.47) for this dimension are as follows: number and length of setae on peduncle of antenna 2 (F, H), length of setae on second antenna flagellum (I), the length ratio of endopodit to exopodit of uropod 3 (K), presence and type of setae on endopodit of uropod 3 (U, V). The dimension 2 significantly differentiates populations in 38 cases. The features with highest DM (DM > 0.36) for this dimension include the following: number of setae on peduncle of antenna 1 (A, B), the length ratio of endopodit to exopodit of uropod 3 (K), type of setae on endopodit outer margin of uropod 3 (S) and type of setae on endopodit inner margin of uropod 3 (V). The dimension 3 significantly differentiates populations in the same number of cases as does the dimension 2. The features with a highest DV values (DM > 0.48) for the dimension 3 are number of setae on 4th and 5th peduncle segments of antenna 2 (F, G).

Variance within most localities is at a similar level. Three populations, one inhabiting Fagaras Mountains in Southern Carpathians (site 8), second from the Bjelasica Mountains in the Dinarid range on the Balkan Peninsula (site 12) and third from the Crimean Peninsula (site 6), are the exceptions (anova, F_12,229 = 12.13, p < 0.001; unequal N HSD Tukey test, Table S3). Within the localities 8 and 12, the morphological variability is at the highest level, while it is the lowest in the site 6 (Table 3). Geographical distances between sites do not seem to be a very important factor structuring the observed variability (Mantel test, one-tailed p-value = 0.07), however, there is some weak yet statistically important positive correlation between pairwise group centroid distances (Table S2) and geographical distances among sampled localities (Spearman rank order correlation, R = 0.284, p = 0.006).While populations from such distant sites as those from Albanian Central Mountain Range (site 13) and Eastern Carpathians (site 1, 3, 7) show overlap in the range of morphological variability, no significant differences could be found for most of the analysed features (Fig. 2, Table 2, Table S2). In addition, some populations from mutually close localities within Eastern Carpathians (site 2, 3) or Southern Carpathians (site 8, 9) show less overlap in morphology and differ significantly in several features (Fig. 2, Table 2, Table S2).

Table 3. Molecular and morphological diversity for each population of Gammarus balcanicus (see Table 1 for localities)

Code	H	D	SE	M
1	2	0.0006	0.0004	0.207
2	2	0.0004	0.0004	0.225
3	4	0.0012	0.0006	0.345
4	1	n/c	n/c	0.227
5	1	n/c	n/c	0.232
6	5	0.0015	0.0007	0.103
7	6	0.0018	0.0007	0.248
8	7	0.0032	0.0011	0.448
9	5	0.0193	0.004	0.316
10	2	0.0009	0.0005	0.313
11	5	0.0096	0.0021	0.337
12	1	n/c	n/c	0.657
13	1	n/c	n/c	0.208

• H, number of haplotypes; D, mean molecular distances (p-distance); SE, standard errors for D; M, average morphological distance of individuals from group centroid; n/c, not calculated (only one haplotype present at a site).

Molecular analyses

A total of 42 haplotypes have been identified of the 135 individuals sequenced. No haplotype was shared among populations. Nine of 13 locations showed more than one haplotype (Table 3). The p-distances between sequences within most sampled localities were <0.0032 (Table 3). The highest observed values still remained quite low, being 0.0193 and 0.0096, respectively, in Southern Carpathians and Transylvanian Plateau (sites 9 and 11). This pattern contrasts with the high overall mean molecular distance (p-distance: 0.153, SE: 0.009; K2P: 0.183, SE: 0.012) between localities (Table S4a and S4b). Most of these distances (94%) were high, ranging from 0.14 to 0.22 p-distance and from 0.16 to 0.27 K2P distance (Fig. 5a, Table S4a and S4b). They are in the same range as between-species distances calculated for several Gammarus species well delineated in morphological terms (Fig. 5b, Table S5a and S5b). However, there are also a few cases with between-locality distances below 0.11. Two cases concern the Carpathians (sites: 1, 3, 11; sites: 2, 7) and one Crimean Peninsula (sites: 5, 6). Interestingly, the distance between two other localities from the Southern Carpathians (sites: 8, 9) being in geographical proximity (57 km) is 0.18 p-distance and 0.21 K2P. Even between two sites remote by only ca. 7 km (sites: 1, 2) and 11 km (sites: 5, 6), the genetic p-distance is 0.19 and 0.10, respectively (Table S4a and S4b). The genetic distance between localities is not related to the geographical distance (Mantel test, one-tailed p-value = 0.54; Spearman rank order correlation, R = −0.026, p = 0.411). All the analysed populations of G. balcanicus and other species from G. balcanicus group (G. bosniacus, G. sketi, G. macedonicus, G. salemaai) cluster together within one clade on the neighbour-joining tree with 77% bootstrap support. The observed grouping pattern did not support monophyly of G. balcanicus versus other morphologically well-defined species. Interestingly, inside this clade, we can also find G. stojicevici belonging to the G. roeseli morphological group and G. leopoliensis belonging to the G. pulex group as defined by Karaman and Pinkster (1977a,b, 1987). The other species from G. pulex group (G. pulex, G. fossarum, G. lacustris, G. varsoviensis) and G. roeseli serve as an outgroup. Generally, inside the G. balcanicus clade, the branching order is poorly resolved. Only the localities from Crimean Peninsula (5 and 6) or Eastern and Southern Carpathians (1, 3 and 11) grouped with a 100% support (Fig. 4).

Discussion

In his original description of G. balcanicus based on individuals from several montane and submontane locations representing both western and central parts of the Balkan Peninsula (Montenegro, Hercegovina, Bulgaria), Schäferna (1922) indicated that many morphological traits of this species were largely variable. He attributed this variation to the rather wide distribution of the species. Subsequently, similar forms were found in numerous montane or submontane localities in Europe ranging from South-Eastern Alps through the Balkan Peninsula, Carpathian Arch to Crimean Peninsula and described as separate species or subspecies (Schäferna 1922; Karaman 1929, 1935; Martynov 1931, 1935; Dobreanu and Manolache 1942; Cărăuşu et al. 1955). Later, Karaman (1977) and Karaman and Pinkster (1987) claimed lack of geographical pattern in the morphology of G. balcanicus in Europe, as well as instability of features defining its subspecies and some of the taxa close in morphology but erected to a species status. Therefore, they synonymized most of them with G. balcanicus. Our formal analysis of a large number of morphological characters revealed high level of morphological polymorphism yet without a clear geographical pattern. Most variability ranges of the analysed features did not differ between localities. In some cases, a substantial overlap in morphology could be observed, even between sites from geographically distant areas, such as western Balkan Peninsula and Southern Carpathians. Interestingly, that conforms to the opinion of Karaman (1977) and Karaman and Pinkster (1987) drawn only from simple observations without any statistical support. In contrary to the species we studied, the montane caddisfly Rhyacophlla aquitanica (McLachlan, 1879) showed clear diversification among montane ranges in Europe (Balint et al. 2008).

The relatively high morphological variability of G. balcanicus joined by high molecular diversity is not a surprising phenomenon as it was observed in many other taxa (e.g. Omland 1997; Kim et al. 2010; Seligmann 2010). However, in our case, the high morphological diversity within sites is contrasting with low molecular distances. Similar pattern was already observed in the cichlid fishes from the ancient lake Victoria (Seehausen 2006) as well as in gammarid crustaceans from the European ancient Lake Ohrid (Wysocka et al. 2013). Apparently, in case of the ancient lakes, the morphological diversification with incomplete sorting of genetic lineages may be related to rapid events of sympatric speciation caused by availability of free ecological niches (Elmer et al. 2010). Nevertheless, ancient lakes are ‘long-living’ and spatially complex systems (Delvaux et al. 1995; Albrecht and Wilke 2008). Many mountain rivers hold substantial endemic diversity (Balian et al. 2008; Tockner et al. 2009 and references therein) what can suggest that these structures are also ‘long living’ in spatiotemporal terms (Malicky 1983). Our study has shown that the levels of morphological variability between and within sites are similar. Strangely, akin limits of this polymorphism are kept even in populations from remote sites in different mountain ranges. Thus, we may suspect that high morphological polymorphism may be of some adaptive significance, and as such is maintained even in allopatry. Similar results were found in case of the terrestrial snail Cepaea nemoralis Linnaeus, 1758; however, the cause of this phenomenon has not been discussed (Ożgo 2005; Cameron et al. 2011). It is impossible to explain this mechanism in G. balcanicus, based on the results of our study, yet, we may suspect that it may be supported in some yet unknown way by a range of microhabitats available for the species in alpine streams (Nagy and Grabherr 2009).

In opposition to morphology that shows only a weak geographical pattern, the molecular divergence between sites is always very high, even in case of geographical proximity. Exclusive presence of private haplotypes in populations inhabiting different sites suggests that they are in complete allopatry (i.e. that no gene flow occurs). The high molecular distance between these allopatric populations indicates that their actual divergence is at least on a species level. In fact, the average frequency distribution of COI distances between the analysed localities of G. balcanicus falls within the range observed between other species of Gammarus well defined in morphological terms (Fig. 5b). This may support Hou et al.'s (2011) conclusion about the very old origin of the present gammarid diversity in freshwaters of Europe, with the oldest lineages possibly dating back to the Tethys regression in Tertiary. Besides, G. balcanicus appears to be polyphyletic as some of its populations group with such species as G. leopoliensis and G. stoicevici which are prominently different in their morphology and classified respectively to G. pulex and G. roeseli groups by Karaman and Pinkster (1977a,b). For example, the G. roeseli group was defined based on presence of sharp dorsal processes absent in most species of Gammarus. This points out the need for re-evaluation of morphological features used in the taxonomy of gammarids.

There were already many attempts to define limits between species based on DNA data, one of them being the DNA barcoding approach. Hebert et al. (2003) found out that in insects, molecular p-distance among most congeneric morpho-species was >0.03 for COI. For the same molecular marker, Costa et al. (2007) defined interspecific molecular distance among congeneric marine Gammarus species as ranging from ca. 0.06 to 0.31 K2P, while the intraspecific range was up to 0.03 K2P. Similarly, Hou et al. (2009) as well as Hou and Li (2010) observed at least 0.15 uncorrected pairwise divergence for a number of morphologically distinguishable freshwater species of Gammarus from China. Witt et al. (2006) proposed a different approach to the problem using a species screening threshold (SST) set at 10 times the average intrapopulation COI haplotype divergence.

It is important to mention some ambiguity associated with species definition through such DNA barcodes and problems according to the species concept used. For example, neither Hebert et al. (2003) nor Hou et al. 2009 nor Costa et al. (2007) did refer to any species concept and did not discuss this issue. Witt et al. (2006) were discussing on the level of the phylogenetic species concept. Various problems posed by implementation of DNA barcoding in taxonomy have been already widely discussed (e.g. Collins and Cruickshank 2013). The necessity of combining various sorts of data, including molecular divergence, ecological requirements, hybridization experiments and thorough morphological studies, for the proper definition of gammarid species was pointed already by Meyran et al. (1997) and further by Hou and Li (2010). Nevertheless, in practice, such multiproxy approach is difficult to use. Usually, only preserved specimens are available what limits the study only to their morphology and molecular diversity. Thus, such handy and simple approach as DNA barcoding is necessary for at least provisional definition of species, whose validity may be further tested by other scientists.

According to the criteria specified above, most of the studied populations attributed provisionally to G. balcanicus deserve a separate species status. Only the populations 3 and 11 from Eastern Carpathians and Transylvanian Plateau, respectively, as well as populations 2 and 7 from the Eastern Carpathians are divergent by 0.04 K2P and 0.05 p-distance and do not show statistically significant differences in morphology. Although, the values are greater than the 0.02 threshold set up by Hebert et al. (2003), they fall in between intraspecific and interspecific distance values given by Costa et al. (2007). Yet, taking into account the Witt et al. (2006) SST, these populations also differ enough to be called different species. Even taking into account the high interspecific divergence level observed by Hou and Li (2010), only a few population pairs diverged by <0.15. Two are the already mentioned populations 2 and 7. The interpopulation distance for three populations (1, 3, 11) from the Eastern Carpathians and Transylvanian Plateau is from 0.04 to 0.09 (p-distance), what is to be appreciated in the face of the lack of significant differences in morphology. The Crimean populations are mutually divergent by 0.11 (p-distance) and also do not differ significantly in morphological terms.

The molecular divergence between the studied populations of G. balcanicus is at the same level as between other European Gammarus species. Also, there are significant differences in case of some of the studied morphological features. Thus, we can assume that at least some of the numerous species and subspecies described years ago and later synonymized with G. balcanicus (see above) might in fact be distinct species, not even closely related in phylogenetic terms. We can conclude that the name G. balcanicus sensu Schäferna 1922 applies only to the population from type locality near Kolasin, Montenegro. In this paper, we neither aim to define the key diagnostic morphological features for identification of the analysed populations nor to perform formal taxonomic revision of the species. That is a goal for a separate study upon taxonomy of G. balcanicus – a difficult one taking into account that many of the original collections and type specimens were lost, and in many cases, the exact type localities are unknown. Yet, at present, we can conclude which features show the greatest variability among sites that could be used in further taxonomic studies upon G. balcanicus. These are as follows: number of setae on peduncle of antenna 1 (A, B), number and length of setae on peduncle and length of setae on flagellum of antenna 2 (F, G, H, I) and on third uropod: the length ratio of endopodit to exopodit (K), type of setae on endopodit outer margin (S), presence and type of setae on endopodit (U, V). Interestingly, all the analysed features localized on antennas were classified by Karaman and Pinkster (1987) as characters stable within species, while the armature of third uropod was recognized as showing intraspecific variability. Once again, it leads to the conclusions that the taxonomy of gammarids based on morphological characters should be definitely re-evaluated. In our opinion, a wider variety of phenotypic characters should be studied for the sake of delimitation of these, so far, cryptic species. Cuticle ultrastructure revealed by SEM is a particularly promising marker (Khalaji-Pirbalouty and Sari 2004, 2006). On the other side, implementation of further molecular markers such as 28S, 18S, EFα1 will help to illustrate the phylogenetic relationships among these putative taxa and their evolutionary history.

Proper identification of such cryptic species is crucial also for practical reasons. Various widely spread Gammarus species, including G. balcanicus, are commonly used in biomonitoring of waters and particularly for toxicological tests (e.g. Krupa and Guidolin 2003; Felten et al. 2008; Parvulescu and Hamchevici 2010; Maazouzi et al. 2011; Sroda and Cossu-Leguille 2011). Distinct species, even not separable based on morphology but distant in molecular terms, may have different tolerance ranges towards a variety of factors (e.g. heavy metals). For example, Scheepmaker (1990) reported that even very similar and closely related species of Gammarus differ in the minimum water temperature required for reproduction by 2–4°C. Recognition of the level of genetic diversity and its maintenance is also one of the key targets for nature conservation (WWF 2012). Taking into account the already mentioned (see 1) properties of montane areas, it may be expected that other aquatic species, widely distributed in European mountain ranges and poorly defined in morphological terms, will also reveal deep within-species molecular divergence and wide cryptic diversity. Generally, all this makes the recognition of cryptic diversity in aquatic taxa an important goal for future studies, notably in case of those distributed across various mountain ranges.

In conclusion, (1) there is a high morphological variability within populations of G. balcanicus, however, this variability shows no geographical pattern across European mountain ranges; (2) in contrary, the level of molecular divergence within sites is very low, while it is very high between different localities yet also not geographically structured; (3) analysis of molecular data suggests that G. balcanicus is a polyphyletic group of allopatric cryptic species; (4) a number of morphological features show substantial variability and help to separate some of these putative species in multiple correspondence analysis, however, so far we could not define any of them as being diagnostic.