Zoogeography of epigean freshwater Amphipoda (Crustacea) in Romania: fragmented distributions and wide altitudinal variability

Inland epigean freshwater amphipods of Romania are diverse and abundant for this region has a favourable geographical position between the Balkans and the Black Sea. Excluding Ponto-Caspian species originating in brackish waters and freshwater subterranean taxa, there are 11 formally recognized epigean freshwater species recorded from this country. They belong to 3 genera, each representing a different family: Gammarus (Gammaridae, 8 species or species complexes), Niphargus (Niphargidae, 2 epigean species) and Synurella (Crangonyctidae, one species). Their large-scale distribution patterns nevertheless remain obscure due to insufficient data, consequently limiting biogeographical interpretations. We provide extensive new data with high resolution distribution maps, thus improving the knowledge of the ranges of these taxa. Gammarus species display substantial altitudinal variability and patchy, fragmented distribution patterns. They occur abundantly, particularly in springs and streams, from lowlands to sub-mountainous and mountainous regions. In the light of recent molecular research, we hypothesize that the complex geomorphological dynamics of the Carpathian region during the Late Tertiary probably contributed to their allopatric distribution pattern. Contrasting with Gammarus , the genera Niphargus and Synurella exhibit low altitudinal variability, broad ecological valences and overlapping distributions, being widespread throughout the lowlands. The current distribution of N. hrabei and N. valachicus seems to be linked to the extent of the Paratethys during the Early Pliocene or Pleistocene. We further discuss the taxonomic validity of two synonymized and one apparently undescribed taxon, and provide an updated pictorial identification key that includes all taxa and forms discussed in our study. The mosaic distribution of epigean freshwater amphipod species in Romania shows that this region is particularly suitable for phylo- and biogeographical analyses of this group.


Introduction
Distribution patterns offer valuable insights towards understanding historical factors that have shaped the contemporary distributions of species (Brown et al. 1996). Freshwater amphipod crustaceans are particularly suitable for biogeographical studies because of their restricted dispersal capabilities and the fragmentary nature of freshwater habitats (Väinölä et al. 2008;Hou et al. 2011). Amphipods are predominantly aquatic benthic animals that do not possess free-swimming larval stages or resistant propagules, and thus are prone to genetic differentiation and isolation (J.L. Barnard & C.M. Barnard 1983). Furthermore, many freshwater taxa display allopatric or discontinuous distributions, frequently presumed to result from vicariant events of geological origin, such as island separation, sea level fluctuations, and continental break up, or that follow ancient drainage patterns (Hogg et al. 2006;Finston et al. 2007;Bauzà-Ribot et al. 2011. The European continent is inhabited by a relatively high number of freshwater amphipod species with diversity increasing towards the south-east (Väinölä et al. 2008). The diversity of the amphipod fauna of Romania is rich due to the favourable geographical position of the country, being situated at the edge of the Balkan Peninsula and the a benthic hand-net with a mesh size of 250 μm and were stored in either 70% or 96% ethanol, or 4% formaldehyde solution. At every sampling locality, we investigated all available microhabitats. A literature review was performed and distribution data were gathered from the relevant studies, including the most recent ones (Cărăuşu et al. 1955 and reference therein, Paraschiv et al. 2007;Petrescu 1994;1996;1997a;1997b2000;. The material from the studies of Pârvulescu (2008; was revised and incorporated into this study. Data from both the literature and from this study were taken into consideration for producing the distribution maps. Locality names, geographic coordinates and altitude are provided for each taxon in a table available from the Dryad Digital Repository: http://dx.doi.org/10.5061/dryad.fd8m9 (Copilaş-Ciocianu et al. 2014).
Taxa were identified using the morphological delimitation criteria of the following authors: Cărăuşu et al. (1955), G. Karaman & Pinkster (1977a;1977b;1987) and Jazdzewski & Konopacka (1989). At present it is known that G. balcanicus, G. fossarum, G. komareki, G. roeselii, and G. pulex are taxonomically challenging poly/ paraphyletic cryptic species complexes (indicated by "s.l." below) (Scheepmaker 1990;Müller 1998;Hou et al. 2011Hou et al. , 2013Weiss et al. 2013). We are aware that the Romanian populations might represent distinct cryptic lineages, as has been shown for G. balcanicus (Hou et al. 2011. Without a detailed insight based on molecular data, we treated the Romanian populations as belonging to the above mentioned morphospecies. However, we took into consideration the three distinct morphs of G. balcanicus reported from Romania. Two of them, G. balcanicus dacicus Manolache, 1942 andG. balcanicus montanus S. Karaman, 1929, were formally described as subspecies. The third morph resembles G. balcanicus from the type locality (Cărăuşu et al. 1955), but the results of Hou et al. (2011) and Mamos et al. (2014) indicate that the Romanian populations are molecularly distinct from those from the type locality and, therefore, we label it G. cf. balcanicus. Although both above-mentioned subspecies are presently synonymized with G. balcanicus (G. Karaman & Pinkster 1987), it seems likely that many such synonymized taxa might actually be distinct species (Hou et al. 2011, Wysocka et al. 2014. Thus, we considered it appropriate to show the morphology and distribution of the Romanian morphs of G. balcanicus separately, to facilitate future taxonomic and systematic studies. Aside from G. balcanicus s.l., we present a putatively undescribed taxon, G. cf. kischineffensis, which is morphologically and geographically distinct although it bears some resemblance to G. kischineffensis (see below). Spatial analyses. We performed spatial autocorrelation analyses using Moran's I (Moran 1950) on latitude and longitude data in order to determine the degree of geographical clustering between the localities of different species. This analysis was performed with the software SAM v4.0 (Spatial Analysis in Macroecology) (Rangel et al. 2010) with default settings, using a significance test with 200 permutations. This analysis requires at least 30 entries to give reliable results, thus it was performed only on taxa for which we had a sufficient number of records. These were: G. cf. balcanicus, G. balcanicus dacicus, G. fossarum s.l., G. cf. kischineffensis, N. valachicus and S. ambulans.
Identification key. Based on available literature (Cărăuşu et al. 1955;G. Karaman & Pinkster 1977a;1977b;1987) and our own observations, an identification key was compiled to contain all morphologically distinct taxa of the respective genera (including the synonymized subspecies of G. balcanicus s.l. and G. cf. kischineffensis) recognized from Romanian territory. The key (provided in the Appendix) combines text with graphical depiction of all relevant identification characters based on original drawings. However, as it does not contain the taxa of Ponto-Caspian origin, the key should be used as a complement to other identification resources rather than alone.

Results and discussion
Checklist of species. The presence of 11 epigean freshwater amphipod species or species complexes previously recognized in Romania, and the two presently synonymized subspecies of G. balcanicus, was confirmed by our data and recent literature. The only species not encountered during our field survey was N. hrabei; however, a recent study reported it in south-eastern Romania (Flot et al. 2014), confirming its presence after more than 50 years.
Furthermore, one morphologically distinct form with a well delimited distribution pattern did not correspond to any of the morphological criteria or character combinations used to differentiate previously recognized taxa (for details, see Taxonomic remarks below) and therefore we treat it as a separate entity throughout this paper. It superficially resembles G. kischineffensis and we have temporarily labelled it G. cf. kischineffensis.
Distribution patterns, altitudinal ranges and habitat preferences. Below, each taxon recognized in Romania is discussed separately, in alphabetical order. Their distributions are shown in three maps: the G. balcanicus complex in Fig. 1, remaining Gammarus species except G. fossarum s.l. in Fig. 2, and Niphargus, Synurella and G. fossarum s.l. in Fig. 3. Altitudinal ranges of taxa are summarized in Fig. 4.
Gammarus arduus is a species with a south-eastern European distribution (G. Karaman & Pinkster 1977a). Altogether, literature records and new data from our study reveal only 6 localities scattered throughout Romania (Fig. 2). Previous reports indicated its presence in western and southern Romania. During this study it was collected from three additional sites in the south-eastern part of the country, in the Măcin Mountains. It is found in streams (Table 1) at altitudes below 300 m a.s.l. (Fig. 4).
The Gammarus balcanicus complex is widely distributed throughout south-eastern Europe and Asia Minor (J.L. Barnard & C.M. Barnard 1983;G. Karaman & Pinkster 1987;Özbek & Ustaoğlu 2006;Özbek et al. 2009). It is also the most widespread amphipod in Romania (Petrescu 1994). We present data for all three reported morphs of this taxon from the country (Fig. 1). The most common is G. cf. balcanicus, phenotypically similar to the morph from the species' type locality, which is widespread throughout the Carpathians and reaches the Dobrogea region in south-eastern Romania. Intriguingly, its occurrence decreases considerably in the central part of the country and it is absent in the south (Fig. 1). It has the widest altitudinal range among freshwater amphipods from Romania, being recorded between 16 and 1530 m, with the majority of localities situated between 300 and 600 m (Fig. 4). It mostly occurs in springs and brooks and occasionally in caves or in rivers (Table 1). It was occasionally found coexisting with G. balcanicus dacicus, G. fossarum s.l. and G. roeselii s.l.  The morphologically distinct G. balcanicus dacicus is likewise well represented in the Romanian territory but differs in distribution from G. cf. balcanicus. It is concentrated mostly in central and southern Romania where it replaces G. cf. balcanicus and reaches the western lowlands of the country (Fig. 1). The altitudinal range of this taxon extends from 37 to 1072 m, mostly between 170 and 600 m (Fig. 4). It coexists with the same taxa as G. cf. balcanicus and inhabits small brooks, springs and lowland rivers (Table 1). G. balcanicus montanus has been recorded from only a few localities in the Southern Carpathians (Fig. 1). It inhabits only high altitudes that range between 880 and 1930 m (Fig. 4).
The Gammarus fossarum complex has a wide distribution area that spans western, central and south-eastern Europe and reaches northern Anatolia (G. Karaman & Pinkster 1977a; J.L. Barnard & C.M. Barnard 1983;Özbek & Ustaoğlu 2006). In Romania, it occurs in the western part of the Carpathians in two isolated regions, one in the north-west and the other in the south-west (Fig. 3). Its altitudinal distribution ranges from 47 to 860 m, mainly between 300 and 550 m (Fig. 4). The populations from south-western Romania also occur in rivers in the lowlands while the north-western populations are confined to springs and streams in sub-mountainous regions (Table 1, Fig.  3). In some localities G. fossarum s.l. coexists with G. cf. balcanicus, G. balcanicus dacicus and G. roeselii s.l.
Gammarus kischineffensis has a discontinuous distribution encompassing two distinct areas. One is restricted to north-eastern Romania, Moldova and south-western Ukraine and the other is limited to the eastern half of Turkey (G. Karaman & Pinkster 1977a, J.L. Barnard & C.M. Barnard 1983, Özbek & Ustaoğlu 2006). Thus, it is possible that the latter represents a distinct lineage. On the Romanian territory, G. kischineffensis occurs only in the north-eastern part of the country, being limited to the Siret and Prut River catchments and never reaches the inner Carpathian basins (Fig. 2). Nevertheless, morphologically distinct populations, which we treat separately (see below), are found throughout south-western Romania. G. kischineffensis is restricted to altitudes below 460 m and inhabits springs, streams and rivers (Table 1, Fig. 4). Gammarus cf. kischineffensis is a newly recognized form. Due to its morphological distinctness and allopatric distribution, it seems likely that it is a species separate from G. kischineffensis in a strict sense (see Taxonomic remarks). It is encountered in the south-western part of the country, in the Almăjului, Aninei and Semenic Mountains (Fig. 2). Altitudinally, it was found between 140 and 860 m in springs and streams (Table 1, Fig. 4). It occasionally coexists with G. fossarum s.l. and G. cf. balcanicus.
The Gammarus komareki complex has a distribution range that extends from Bulgaria and northern Greece, throughout the northern half of Turkey into the north-western part of Iran (G. Karaman & Pinkster 1977a;Grabowski & Pešič 2007;Özbek & Ustaoğlu 2006;Zamanpoore et al. 2011). It is known only from three localities in south-eastern Romania from the Dobrogea Region (Fig. 2). It occurs at altitudes below 100 m and inhabits slow flowing rivers with rich vegetation, coexisting with G. pulex s.l. (Copilaş-Ciocianu 2013) (Table 1, Fig. 4).
Gammarus leopoliensis occurs only in the northern half of the Carpathian region in Poland, Ukraine, Slovakia, Hungary and Romania (Grabowski & Mamos 2011;Papp & Kontschán 2011). Its distribution in Romania is confined to the northern part of the country (Fig. 2), to streams at high altitudes, from 600 to 1150 m (Table 1, Fig.  4) (Papp et al. 2008).
The Gammarus pulex complex has a western, central and northern European distribution with patchy populations being encountered in south-east Europe, Asia Minor and throughout Asia (G. Karaman & Pinkster 1977a; J.L. Barnard & C.M. Barnard 1983;Özbek & Ustaoğlu 2006). It is encountered only in south-eastern Romania, in the Dobrogea Region (Fig. 2), in groundwater, springs, and streams, at altitudes that do not exceed 100 m (Table 1, Fig. 2), cohabiting with G. komareki s.l. in some sites (Fig. 2). The Gammarus roeselii complex is distributed across western, central and south-eastern Europe as well as the western part of Turkey (G. Karaman & Pinkster 1977a; J.L. Barnard & C.M. Barnard 1983;Jazdzewski & Roux 1988;Özbek & Ustaoğlu 2006). It is present in the western and southern parts of Romania in a few distinct patches (Motaş et al. 1962;Pârvulescu 2009) (Fig. 2). It is a typical lowland taxon, occurring mostly at altitudes below 200 m (Fig. 4). G. roeselii s.l. is ecologically the most plastic gammarid in Romania, being found in springs, streams, rivers, and occasionally in lakes and swamps (Motaş et al. 1962 , Table 1). It can co-occur with G. cf. balcanicus, G. balcanicus dacicus and G. fossarum s.l.
Niphargus hrabei occurs in central and south-eastern Europe from the Small Hungarian Plain to the Danube Delta (Cărăuşu et al. 1955; J.L. Barnard & C.M. Barnard 1983;Meijering et al. 1995;Nesemann et al. 1995). It is found only in the south-eastern lowlands of Romania and the Danube Delta in springs, streams, ponds and swamps at altitudes below 350 m (Table 1, Figs 3-4).
Niphargus valachicus has a large and fragmented range, spanning from the Pannonian Plain along the Lower Danube, and reaching the Danube Delta (Cărăuşu et al. 1955; J.L. Barnard & C.M. Barnard 1983). It is also present in Turkey along the southern shore of the Black Sea and reaches the south of the Caspian Sea in Iran (Fišer et al. 2009;Hekmatara et al. 2013). In Romania, it is a common species in the lowlands, being found in swamps, canals, temporary ponds, and large rivers in sympatry with S. ambulans (Cărăuşu et al. 1955;Copilaş-Ciocianu & Pârvulescu 2012; Table 1). It inhabits the western and southern plains of Romania (Fig. 3), being encountered between 0 and 360 m with most localities ranging around 100 m (Fig. 4). It often coexists with S. ambulans and G. balcanicus dacicus, and occasionally with G. roeselii s.l. and N. hrabei (Motaş et al. 1962).
Synurella ambulans is widespread in central, eastern and southern parts of Europe (G. Karaman 1974;Sidorov & Palatov 2012). In Romania it has a distribution similar to N. valachicus, co-occurring in the same habitats and at the same altitudes (Table 1, Figs 3-4). Biogeographical patterns. Freshwater amphipods in Romania have patchy and often non-overlapping distribution patterns. A distinction can be made between the distributions of Gammarus and Niphargus /Synurella. Gammarus species exhibit high altitudinal variation and allopatric distributions, contrasting with Niphargus and Synurella which are sympatric and restricted to the lowlands.
The distribution patterns of the analysed species are significantly non-random (I > 0, p ≤ 0.05). However, the positive autocorrelation distance values varied between genera. Maximum significant positive autocorrelation distances ranged between 19 and 180 km for different Gammarus taxa (p ≤ 0.01) and reached 327 km and 296 km for N. valachicus (p = 0.03) and S. ambulans (p = 0.005), respectively (Fig. 5). Thus, the spatial autocorrelation analyses further emphasize this patchy distribution pattern by revealing that populations of Gammarus are significantly autocorrelated for shorter distances than N. valachicus and S. ambulans. This means that Gammarus species have more geographically clustered distributions than the latter two.
Four Gammarus taxa have well-delimited, wide distributions in Romania; these are G. cf. balcanicus, G. balcanicus dacicus, G. fossarum s.l. and G. kischineffensis. The former two have an intertwined, complementary distribution pattern that is mostly non-overlapping, while the latter two are restricted to the western and eastern parts of the country, respectively (Figs 1-3). The remaining taxa, G. arduus, G. leopoliensis, G. komareki s.l., G. pulex s.l. and G. roeselii s.l. are not so widespread, most likely due to the fact that the territory of Romania only marginally overlaps with their distribution areas. The full distribution of G. cf. kischineffensis is at present unknown, it is possible that it extends into neighbouring Serbia.
The allopatric distributions displayed by Gammarus species in Romania are typical for the genus (G. Karaman & Pinkster 1977a;1977b;1987;Väinölä et al. 2008). Molecular phylogenetic analyses indicate that the frequent allopatry observed in many Gammarus species is the result of geological vicariant events of Tertiary age and that the majority of extant freshwater species originated in the Late Tertiary (Hou et al. 2007;2013;Wysocka et al. 2014). During this period, the Carpathian Mountain range was a geomorphologically highly dynamic archipelago surrounded by the shallow Central Paratethys Sea (Popov et al. 2004). This constantly changing topography was characterized by different timings of landmass uplift and drastic variations in sea level (Harzhauser & Piller 2007, Kováč et al. 2007. Sea level fluctuations and the vicariance they create are considered to be of significant importance in the evolution of many, especially subterranean, freshwater amphipod taxa (Notenboom 1991;Holsinger 1994). We hypothesize that the dynamic geomorphology of the Carpathian region during the Late Tertiary has left its footprint on the contemporary ranges of Gammarus species in Romania. Niphargus valachicus, N. hrabei and S. ambulans are unusual amongst their congeners because of their predominantly epigean lifestyle, wide distributions and ecological plasticity (Straškraba 1972;Meijering et al. 1995;Nesemann et al. 1995;Sidorov & Palatov 2012). Both Niphargus species are sympatric with S. ambulans throughout their distribution range, sharing the same ecological requirements (Motaş et al. 1962;Meijering et al. 1995;Nesemann et al. 1995), and in many cases coexisting in the same habitat (Motaş et al. 1962;Straškraba 1972;Akbulut et al. 2001;Juhász et al. 2006;Copilaş-Ciocianu & Pârvulescu 2012). The co-existence of Niphargus and Synurella seems to be quite old, since both genera are known from Baltic amber that dates back to the Eocene, i.e., is at least 35 million years old (Jażdżewski & Kupryjanowicz 2010;Jażdżewski et al. 2014).
Molecular data indicate that Niphargus colonized south-eastern Europe at the beginning of the Oligocene (~25 million years ago) (McInerney et al. 2014). Although N. valachicus was not included in that study, its distribution reflects the extent of the Paratethys Sea during the Early Pliocene (~5 million years ago), when-as it has been hypothesized-it may have colonized available freshwater habitats through coastal lagoons (Sket 1981). Straškraba (1972) suggested that the present-day distribution of N. valachicus is linked with the more recent Pleistocene extent of the Paratethys. However, during the Pliocene and Early Pleistocene, the extent of the Paratethys was more or less the same, without major fluctuations (Popov et al. 2004). Thus, it is possible that the hypothesized freshwater colonization may have taken place anytime during this time frame. The lowland regions of Romania, where N. valachicus is nowadays present, fit to this general pattern since they were continuously submerged under the waters of the Paratethys during the Pliocene/Pleistocene (Popov et al. 2004). Due to its large and fragmented range, it is possible that this taxon might harbour independently evolving lineages (Fišer et al. 2009). Straškraba (1972) also suggested a connection between the distribution of N. hrabei and the extent of the Paratethys during the Pleistocene. Since this species is morphologically, ecologically and biogeographically similar to N. valachicus (Straškraba 1972;Nesemann et al. 1995), it is possible that it follows the same pattern. It is considered that N. hrabei is expanding its range at present (Ketelaars 2004). Both species are sympatric in the Lower Danube basin in Romania and coexist in some instances (Motaş et al. 1962). S. ambulans, although having a wide distribution in Europe, has a problematic taxonomy and probably represents a species complex, as suggested by its morphological variability and ecological plasticity (Meijering et al. 1995;Konopacka & Blazewicz-Paszkowycz 2000;Sidorov & Palatov 2012). It belongs to the crangonyctoid group which is considered one of the most ancient groups of freshwater amphipods (J.L. Barnard & C.M. Barnard 1983). Its unclear taxonomy and lack of molecular data restrain biogeographical interpretations for this taxon.
Aside from historical factors, we presume that the contrasting patterns (allopatric vs. sympatric) observed between Gammarus and Niphargus/Synurella are due to the fact that these taxa have different ecological preferences. Romanian Gammarus species, except for G. roeselii s.l., are seemingly more restricted in their habitat preferences, favouring springs and streams along an extensive altitudinal gradient (Table 1, Fig. 4). Therefore, their spread might be limited by the availability of these suitable habitats. Competition and interspecific predation, common between Gammarus species (e.g. MacNeil & Dick 2012), might also be important factors contributing to their distribution patterns. On the other hand, Niphargus and Synurella are more euryoecious, preferring a wide spectrum of habitats ranging from groundwater and springs to stagnant waters or temporary ponds, but along a much narrower altitudinal gradient (Table 1, Fig. 4). This is probably due to the fact that niphargids and crangonyctids seem to be more tolerant to low oxygen levels than gammarids (Dick et al. 1997;Simčič & Brancelj 2006). Thus, they can occupy a wider variety of habitats and prevail at lower altitudes, in which they probably face lower competition pressure from gammarids. The mechanisms that permit these two genera to coexist are unclear, given the fact that freshwater amphipods often exhibit high levels of intra-guild predation (MacNeil et al. 1997;1999;Luštrik et al. 2011).
Taxonomic remarks. We present the morphological differences between the three morphs of G. balcanicus reported from Romania, as pointed out by Cărăuşu et al. (1955). G. balcanicus dacicus morphologically differs from G. cf. balcanicus by the longer endopod of uropod 3 (up to 90% the length of the exopod), the pointed inferoposterior corners of epimere 2 and the presence of long setae (as long as the urosome spines) on the dorso-posterior side of the metasome segments (Fig. 6). It also has a very distinct distribution that is complementary to that of G. cf. balcanicus and in some localities they are known to coexist and still maintain their morphological distinctness (Fig.  1). Based on this evidence, we consider that this taxon is a distinct entity and should be 'resurrected' if not even elevated to a specific status, an issue that remains to be resolved by further molecular research.
In the case of G. balcanicus montanus, the situation is less clear. Its main morphological differences from G. cf. balcanicus are its smaller size and a short endopod of uropod 3 (about 50% the length of the exopod) (Fig. 6). Its distribution is not geographically separated but it is supposedly ecologically distinct, inhabiting only high altitude springs or brooks. It might represent an ecomorph of G. cf. balcanicus, although both substantially overlap altitudinally. Molecular analyses are needed to resolve the taxonomic status of G. balcanicus montanus.
The distinct form G. cf. kischineffensis is morphologically different from its presumed relative G. kischineffensis (Fig. 7). Geographically, their ranges are separated by ca. 300 km (Fig. 2). The main distinctive features of G. cf. kischineffensis are 1) pointed infero-posterior corners of the 2 nd and 3 rd epimeres that are straight in G. kischineffensis (Fig. 7a); 2) the inner side of the telson lobes that has one spine; this spine is absent in G. kischineffensis (Fig. 7b); and 3) the presence of long setae (longer than or as long as the width of the underlying segment) in the upper quarter of the external margin of the 3 rd uropod exopodite (these are present along the upper half in G. kischineffensis) (Fig. 7c). Based on its morphological and geographical distinctness, we propose that this morph might be an undescribed species, a hypothesis that will be tested by further morphological and molecular studies.

Conclusions
Romania has a diverse fauna of epigean freshwater amphipods and further taxonomic studies are needed to truly recognize this diversity. Morphological variation in the G. balcanicus complex and G. cf. kischineffensis suggest the presence of undescribed taxa, and it is likely that molecular studies will reveal additional cryptic diversity (as it has been the case with G. cf. balcanicus). The distributions of the epigean freshwater amphipods in Romania are characterized by their patchiness and altitudinal variability. Coupled with their low dispersal abilities and the heterogeneous topography/geology of this region, these animals constitute a suitable model system for studying biogeography and phylogeography at a fine scale, with implications for further research in ecology, adaptation and speciation of freshwater amphipods.