The discovery of Bulinus (Pulmonata: Planorbidae) in a Miocene palaeolake in the Balkan Peninsula

Large, sinistrally coiled gastropod species have been reported under the genus name Kosovia Atanackovi ć , 1959 from middle and late Miocene palaeolakes of central Serbia and Kosovo. Despite several papers dealing with the taxonomy and evolution of this genus, its systematic position and possible ancestry have been unresolved. Previously, it has been suggested that it may be a member of the Viviparidae, Lymnaeidae or Planorbidae, but without morphological criteria to support these assertions. In order to elucidate the systematic position of the genus, we investigated type material of the type species Kosovia matejici Pavlovi ć , 1931, which is the oldest representative and is restricted to middle Miocene deposits of central Serbia. Embryonic shell characters support membership of the Planorbidae. Based on the congru-ence of all evaluated morphological characteristics, we attribute the species to the genus Bulinus , which makes Kosovia a junior subjective synonym of Bulinus . We discuss potential ancestry of the oldest representative and relationships among the species previously attributed to Kosovia . Our study emphasizes the importance of SEM-based examination of the protoconch to clarify the systematic position of problematic freshwater gastropods, especially when molecular data are absent.


INTRODUCTION
The middle-late Miocene lacustrine deposits of central Serbia and Kosovo have yielded a number of large, sinistrally coiled, highspired gastropod species that have been attributed to the genus Kosovia Atanacković, 1959. Several papers have dealt with taxonomy and hypotheses about evolutionary processes within the group (Pavlović, 1931;Atanacković, 1959;Milaković & Milošević, 1974;Krstić, Savić & Jovanović, 2012), but its systematic position has remained uncertain. Moreover, nomenclatural issues and stratigraphic uncertainties have precluded reliable conclusions.
The peculiar morphology of Kosovia was first mentioned by Boué (1840), who mistook the shells for sinistral aberrations of a co-occurring Viviparus species. Almost a century later, Pavlović (1931) described the genus Kosovia and four new species. Unfortunately, he did not designate a type species, thus the genus name was not available from that publication (ICZN, 1999: Art. 13.3). Moreover, Pavlović did not classify the genus, but compared it with Viviparidae, as well as with marine and terrestrial genera that share the sinistral coiling or the spiral keels. Atanacković (1959) made the name Kosovia available by indicating a type species, Kosovia ornata Pavlović, 1931 and by referring to Pavlović's earlier description (Art. 13.1.2). He placed Kosovia in a separate subfamily (Kosovinae) in the Viviparidae, largely based on comparison with Recent Lanistes from Lake Malawi. Although neither Atanacković (1959) nor any later author formally described the subfamily, it is available from its original publication (ICZN, 1999: Art. 13.2.1), since it was used as valid after 1960 and before 2000 (Milošević, 1978). (Note that following ICZN, 1999: Art. 29, the correct spelling is Kosoviinae.) Zilch (1959Zilch ( -1960 and Milaković & Milošević (1974), again without any explanation, placed the genus in the planorbid tribe Camptoceratini, apparently suggesting a relationship with the sinistral, elongate Camptoceras from eastern Asia (see e.g. Walker, 1919). Most recently, Krstić et al. (2012) attributed Kosovia to the Lymnaeidae, once again without discussion.
In summary, the systematic position of Kosovia is highly uncertain. A relationship with Viviparidae seems doubtful, given that viviparids have little in common with Kosovia except for their size and, as we show here, differ in key taxonomic characters in early ontogeny. Moreover, morphological evolution in viviparids mostly involves variation of whorl cross-section and sculpture, while overall shell shape is less affected, as shown by the well-studied Plio-Pleistocene radiations in Croatia, Romania and Greece (Neumayr & Paul, 1875;Willmann, 1981;Lubenescu & Zazuleac, 1985;Mandic et al., 2015). Furthermore, sinistral representatives are rare among Viviparidae (e.g. Tiemann & Cummings, 2008) and not known for European species.
In order to address the systematic position and evolutionary origin of Kosovia, we studied type material of the presumed oldest representative, K. matejici Pavlović, 1931. Earlier studies variously dated the Serbian 'Kosovia beds' as middle Miocene (Stevanović, Pavlović & Eremjia, 1977) or late Miocene (Milaković, 1983), based on biostratigraphic correlations of dreissenid bivalve faunas. Recent radioisotopic studies on nearby, presumably coeval, lacustrine deposits yielded a late Langhian (middle Miocene) age (Sant et al., 2016). Based on detailed study of the shell, particularly focusing on features of the protoconch, we show that Kosovia cannot be sufficiently distinguished from the planorbid genus Bulinus and should accordingly be considered its junior synonym.
Description: Shell sinistral, buliniform, gradate, of up to five whorls. Lectotype 15.0 × 12.0 mm. Protoconch planispiral, about 0.8 whorls, diameter 890 μm; nucleus immersed, bulbous, width 270 μm, with weakly, irregularly wrinkled surface; remaining part highly convex, densely covered by tiny, spirally arranged pits, which fade out after c. 0.6 whorls; at same time, an angulation appears between apical plane and whorl flank, forming a right angle after first protoconch whorl; shortly after onset of angulation (at c. 0.7 whorls) spiral ridges and furrows (5 in lectotype) appear on central part of apical plane, marking onset of teleoconch. Ridges and furrows increase in number and strength during ontogeny, covering entire apical plane and whorl flank after first half whorl of teleoconch. Topmost ridge on whorl flank (below angulation) is typically strongest. Growth lines cover teleoconch, producing irregularly reticulate pattern where they intersect with spiral sculpture. In adult specimens, very prominent, broad axial ribs appear (18 in lectotype, 24 in a paralectotype), extending across entire flank with almost equal strength; in apical view, ribs form regularly undulating outline. Suture deeply incised in early stages, becoming very narrow in adult specimens as coiling tightens. Aperture widely elliptical with glossy, white, sheet-like inner lip; peristome not complete in any specimen. Umbilicus narrow, slit-like, almost fully covered by inner lip.
Remarks: The protoconch of this species resembles those of Miocene species of the planorbid genera Planorbarius and Bulinus and undoubtedly classifies it in the family Planorbidae. Given the striking congruence of all the morphological characteristics assessed, we conclude that Kosovia matejici is a member of the genus Bulinus. Since it is the type species of Kosovia, the genus is herewith considered a junior subjective synonym of Bulinus. The subfamily Kosoviinae is considered a junior synonym of Bulininae accordingly.
In both Planorbarius and Bulinus, the embryonic shell consists of a similar number of whorls and bears spirally arranged tiny pits of similar size (6-7 μm, measured on c. 0.5 whorl) ( Table 1; Fig. 2G, H, J-L; see also Riedel, 1993;Bandel, 2010;Harzhauser et al., 2012Harzhauser et al., , 2014aMartínez-Ortí, Bargues & Mas-Coma, 2015). A major difference is that the pits are much more widely spaced and more regularly aligned in Planorbarius, while they are densely spaced in Bulinus, including B. matejici. Moreover, the embryonic shell of B. matejici is much larger than those of Miocene Planorbarius species, while it is similar to the size of early middle Miocene Bulinus corici Harzhauser & Neubauer in Harzhauser et al., 2012 (Table 1). Protoconch size, however, seems to vary considerably among species in both genera as there are also species with distinctly smaller embryonic shells in Bulinus (e.g. Brown, 1994) and larger ones in Planorbarius (e.g. Riedel, 1993). Bulinus matejici also shares with B. corici the typical pattern of grooves and ridges on the early teleoconch, which is restricted to the central region of the apical plane, as well as the angulation between whorl flank and apical plane emerging on the first whorl (Fig. 2L). In Planorbarius, in contrast, grooves and ridges cover the entire (visible) shell surface and the angulation is much less pronounced ( Fig. 2G, H, J, K; Riedel, 1993). The detachment of the aperture reported for several species of Kosovia (e.g. Pavlović, 1931;Atanacković, 1959) has also been documented for species of Bulinus (Brown, 1994), but not for Planorbarius. Finally, the axial sculpture on the teleoconch that is typical of B. matejici has never been found in Planorbarius, but is present, albeit weaker, in some recent African Bulinus species (e.g. Brown, 1994).
Overall shell shape might be considered a crucial (and obvious) reason to differentiate between discoidal Planorbarius and Bulinus with a high last whorl and raised spire ( Fig. 2A-F), but the family Planorbidae is widely known for its extreme morphological variability (see Discussion). Nonetheless, some of the late Miocene species from the Kosovo Basin previously attributed to Kosovia (K. pavlovici Atanacković, 1959 andK. stevanovici Atanacković, 1959), which are characterized by non-gradate spires and less pronounced axial sculpture, closely resemble some Recent African species of Bulinus. As far as we are aware, no Planorbarius species has been reported to form a raised spire.
Classification in the families Viviparidae and Lymnaeidae, as proposed earlier, is ruled out by the strongly differing protoconchs. Viviparus, for instance, has a low domed to pointed protoconch that bears axial folds or narrow spiral threads (e.g. Riedel, 1993;Neubauer et al., 2015b). Lymnaeidae, on the other hand, bear only growth lines on the protoconch (Riedel, 1993). Similarly, members of the family Physidae, which also consist of sinistrally coiled species, have entirely smooth or weakly striate protoconchs (Riedel, 1993; 296 T. A. NEUBAUER ET AL.  Note that due to preservation issues and the smooth P/T transition, some of the numbers are approximate. Measurements are based on specimens studied by Neubauer et al. (2013) and Harzhauser et al. (2014a, b), some of which are illustrated in Figure 2. Counting of protoconch whorls follows the method proposed by Verduin (1977). 298 T. A. NEUBAUER ET AL. Appleton & Dana, 2005;De Paula & Silveira, 2005). Moreover, physid shells usually have shallower sutures and smoother surfaces (Brown, 1994). Physids are represented by a single species in the European middle Miocene, Aplexa subhypnorum (Gottschick, 1920) which differs considerably from coexisting Bulinus (including Kosovia) species in its bulbous, smooth protoconch, and in the elongate shape with well-rounded and entirely smooth whorls (Gottschick, 1920;cf. Harzhauser & Binder, 2004).

Potential ancestry of Bulinus matejici
The genus Bulinus is the only member of the Bulininae in Europe and is represented by only two rare coeval species, whereas other planorbid genera are common. Bulinus corici is endemic to the early middle Miocene Lake Groisenbach in the Aflenz Basin in Austria (Harzhauser et al., 2012). Bulinus trojanus (Neumayr, 1883) was described from freshwater deposits near Assos (presently Behramkale) in northwestern Turkey, which probably are of early middle Miocene age (for the geology of the area see Yilmaz & Karacik, 2001;Gürdal & Bozcu, 2011;Bozcu, 2015). Given their globular shapes and lack of sculpture on the teleoconch, they are not likely to have been ancestors of B. matejici. The middle Miocene of Europe and especially the regions featuring Bulinus, i.e. central Europe, the Balkan Peninsula and the Aegean-Anatolian region, are represented by numerous well-sampled faunas (see reviews by Neubauer et al., 2015a, b and references therein). We therefore do not expect a major sampling bias masking undetected Bulinus records and potential predecessors. Given the unique morphology of B. matejici and the paucity of potential relatives in the European fossil record, the ancestor might have its roots outside the European continent. Bulinus very likely originated in Africa, where its fossil record dates back at least to the early Miocene (c. 19-20 Ma; Pickford, 2008). Recent molecular analyses suggest an even older origin (Jørgensen et al., 2011). Brown (1994) postulated that the development of polyploidy greatly contributed to the success of the genus and enabled dispersal into stressful and colder environments (see also Van Damme & Van Bocxlaer, 2009). Today, Bulinus is represented by numerous species in Africa, southwestern Asia and Arabia, with one species also extending into the Mediterranean region (Brown, 1994). For it to reach remote, isolated European basins during the middle Miocene could have been possible by means of passive long-distance dispersal, probably by birds. There was a land bridge that connected Europe and Africa during the early Miocene and allowed repeated faunal exchange among mammals (c. 21-22 Ma, 19-18.5 Ma and 18-17.5 Ma; Harzhauser et al., 2007). However, this 'Gomphotherium land bridge' did not offer hydrological connections allowing active migration of freshwater gastropods. Moreover, no terrestrial pathway was available during the known geological duration of B. matejici (Popov et al., 2004;Harzhauser et al., 2007).
Avian dispersal is readily underestimated as an important vector and has been demonstrated for several species of gastropods living today (e.g. Green & Figuerola, 2005;Kappes & Haase, 2012;van Leeuwen & van der Velde, 2012;van Leeuwen et al., 2012avan Leeuwen et al., , b, 2013. Recently, it was suggested as dispersal mode for a species of Bulinus in Nigeria (Salawu & Odaibo, 2013). Dispersal by waterfowl has also been invoked as factor explaining distributions of fossil freshwater snails (Wesselingh, Cadée & Renema, 1999;Harzhauser et al., 2016). One case even involves the isolated occurrence of a presumed descendant of B. matejici in the late Miocene of the Turiec Basin in Slovakia, found at a distance of c. 700 km from its apparent origin in the Metohia Basin (Neubauer et al., 2015b; see below).

Late Miocene to Pliocene evolution in palaeolakes Kosovo and Metohia
In addition to the type species, several species from late Miocene strata of the Kosovo and Metohia basins in Kosovo have been ascribed to the genus Kosovia by earlier authors (Pavlović, 1931;Atanacković, 1959;Milošević, 1978;Atanacković & Stevanović, 1990; Table 2; Fig. 3). Compared with B. matejici, these taxa cover a distinctly wider morphological spectrum, ranging from slender to stout forms and from coarsely ribbed or keeled to striate, almost smooth surfaces (Pavlović, 1931;Atanacković, 1959;Milošević, 1978). Many of the described species co-occur in the same localities and differ only in intensity of sculpture or height of the spire and might merely represent local phenotypes of a single species.
The poorly resolved stratigraphy of the deposits in the Kosovo and Metohia basins remains a major impediment to unravelling the evolutionary pattern among late Miocene forms and the connection to their presumed ancestor B. matejici. Based on biostratigraphic data from molluscs, ostracods and diatoms, the lacustrine deposits have been traditionally correlated with the regional stratigraphic stages of the Pannonian and Dacian basins, which correlate to the late Miocene to Pliocene (e.g. Pavlović, 1903;Atanacković, 1959;Milošević, 1966;Popović, 1970bPopović, , 1974Milaković, 1983;Atanacković & Stevanović, 1990;Ognjanova-Rumenova, 2006, 2014Elezaj, 2009;Elezaj et al., 2010;Krstić et al., 2012). Contrary to the middle Miocene SLS, no absolute age data are available for the lacustrine deposits in the two basins as yet. However, if the biostratigraphic correlations are correct, Bulinus populations from central Serbia and from the Kosovo and Metohia basins would have been separated by several million years. Although time-displaced phenotypic convergence cannot be ruled out, the high similarity between B. matejici and the Kosovan species (Pavlović, 1931;Atanacković, 1959) suggests that they belong to the same lineage. Until revised stratigraphic data and a detailed character analysis of all species are available, no firm conclusion can be drawn. Bulinus bouei (Pavlović , 1931) late Miocene x x Bulinus matejici (Pavlović , 1931) middle Miocene x Bulinus ornatus (Pavlović , 1931) late Miocene x x Bulinus pavlovici (Atanacković , 1959) late Miocene x Bulinus stevanovici (Atanacković , 1959) late Miocene x Bulinus striatus (Milošević , 1978) late Miocene x Popovicia compressa (Pavlović , 1931) late Miocene x Popovicia levantica (Popović , 1964) early Pliocene x Popovicia turriculoidea (Popović , 1964) early Pliocene x *Note that the species name 'Kosovia praepontica' mentioned by Krstić et al. (2012) is a nomen nudum, referring to a description in an unpublished manuscript by Milošević . Re-examination of the respective specimens from the middle Miocene of the Metohia Basin showed that they do not represent a Bulinus but are juveniles of an unidentified Planorbarius species.
Coexisting with Bulinus during the late Miocene, another group of sinistrally coiled gastropods appeared in the Metohia Basin (Fig. 3). Representatives of that group have been previously attributed to the genus Metohia Popović, 1964, which was recently shown to be a junior homonym and, thus, invalid. Neubauer & Harzhauser in Neubauer et al. (2015b) introduced Popovicia as a replacement name, with Metohia levantica Popović, 1964 as type species.
Compared with the still uncertain phylogenetic relationship of late Miocene Kosovan Bulinus species, the evolution of the genus Popovicia is better understood. The oldest known species, P. compressa (Pavlović, 1931), has an almost entirely planispiral coiled shell. Only in latest ontogeny does the shell start to grow slightly in an adapical direction (Popović, 1964(Popović, , 1968Neubauer et al., 2015b). The shape of this species forms the basis for the morphological evolution of broad, occasionally quite high-spired, pseudodextral shells with bulbous, weakly flattened apex (Popović, 1964). Each of the three species ascribed to the Popovicia lineage is confined to a distinct horizon in the stratigraphic succession of the Metohia Basin. Following the stratigraphic classification of Elezaj et al. (2010), P. compressa appeared in the latest Miocene, while the succeeding P. levantica (Popović, 1964) and later P. turriculoidea (Popović, 1964) are confined to the Pliocene. Despite considerable discussion on evolutionary patterns within Popovicia (Popović, 1964(Popović, , 1968(Popović, , 1969Milošević, 1967Milošević, , 1970, little has been published about its potential ancestry. The only comprehensive work dealing with this is an unpublished manuscript by Milošević from 1982 (who died before he could publish the rich information it contained). He provided numerous illustrations that convincingly demonstrate the continuous morphological transition between sinistral 'Kosovia', through planispiral forms to pseudodextral Popovicia. The extreme morphological variability and continuum between Kosovan Bulinus and Popovicia raises considerable doubt about the taxonomic separation of the two genera. However, more material, close examination of the protoconchs and detailed knowledge about the stratigraphic succession are required to properly address the evolutionary patterns in palaeolakes Kosovo and Metohia, and their relation to B. matejici.

Morphological disparity of the Planorbidae
Planorbids are known for their enormous morphological disparity in both Recent and fossil clades, exceeding those of most other freshwater gastropod families. Apart from the ubiquitous spirally coiled (e.g. Anisus, Gyraulus, Planorbarius, Planorbis) and limpet-like shapes (e.g. Ancylus, Ferrissia), they have produced buliniform (e.g. Bulinus, Glyptophysa, Kessneria, Miratesta), corkscrew-like (e.g. Gyraulus), slightly to almost entirely uncoiled (e.g. Gyraulus) to dentaliiform (Orygoceras) shells, as well as all kinds of shell sculpture (e.g. Baker, 1945;Nützel & Bandel, 1993;Finger, 1998;Walker & Ponder, 2001;Albrecht, Kuhn & Streit, 2007;Neubauer, Mandic & Harzhauser, 2013Rasser, 2013;Figure 3. Topographic map of middle (left) and late (right) Miocene records of Bulinus (formerly Kosovia) (circles) and Popovicia (triangles) from Serbia and Kosovo as derived from the literature (Pavlović, 1931(Pavlović, , 1932(Pavlović, , 1935Atanacković, 1959;Popović, 1964;Milošević, 1965Milošević, , 1967Milošević, , 1978Milaković & Milošević, 1974;Atanacković & Stevanović, 1990). The list of occurrences forming the basis of this figure is provided in the Supplementary Material. Palaeolake outlines are tentative estimations based on available reconstructions, distributions of lacustrine sediments and localities with mollusc faunas, as well as present topography (see Popović, 1970a;Atanacković, 1990;Dumurdzanov, Serafimovski & Burchfiel, 2005;Neubauer et al., 2015a). Contrary to this rather conservative reconstruction, Krstić et al. (2012) considered the middle Miocene lacustrine systems of central to southern Serbia, Kosovo, Macedonia, Bulgaria and northern Greece to belong to a single, huge 'Serbian Lake', but this is not reliably supported by data. Mollusc faunas indicate little similarities between Serbian, Kosovan and Macedonian systems (Burgerstein, 1877;Pavlović, 1903Pavlović, , 1931Pavlović, , 1932Pavlović, , 1933Pavlović, , 1935, while the presumed connection to Bulgaria and Greece is based on outdated stratigraphy (Neubauer et al., 2015a). Note that the northern boundary of the Serbian Lake System is unknown, but probably did not extend north of the Belgrade region. 300 T. A. NEUBAUER ET AL. Clewing et al., 2015). However, little is understood about the causes of this extreme disparity and of the individual contributions of ecophenotypic versus genetic diversity (e.g. Reif, 1985;Rasser, 2013;Clewing et al., 2015). Such uncertainty complicates taxonomic classifications, especially in the absence of molecular data. One of the most famous examples of phenotypic evolution in Planorbidae is the Gyraulus species flock in middle Miocene Lake Steinheim. Within a geologically short time, a single founder species gave rise to a diverse set of morphologies, many of which coexisted in the small lake (Hilgendorf, 1867;Mensink, 1984;Reif, 1985;Gorthner, 1992;Nützel & Bandel, 1993;Rasser, 2013). Tens of species-and subspecies-level taxa have been introduced over the past 150 years to categorize this vast morphological diversity and many of them have already been considered junior synonyms (e.g. Rasser, 2013). A comparable example of high morphological disparity is provided by the Gyraulus populations of Recent Lake Bangong on the Tibetan Plateau (Clewing et al., 2015). There, typical planispiral forms co-occur with corkscrew-like and partly open-coiled morphologies and a series of transitional shapes (Clewing et al., 2015). Molecular analyses indicate that phenotypic evolution in that clade is not genetically determined, but likely caused by the rapid ecophenotypic response of a single species to ecological stress.
Examples of scalariform specimens of the genera Anisus or Biomphalaria have been reported from breeding experiments in the laboratory (Boettger, 1949;Basch, 1968), emphasizing once more that morphological disparity does not necessarily have a genetic basis. A case similar to that documented here is the evolution of high-spired forms in North American Planorbella, which seem to have appeared at least twice independently in that genus during Pliocene to Recent times (Burch, 1982). In fact, the high-spired morphologies closely resemble B. matejici, particularly in the marked angulation between whorl flank and apical plane. Planorbella is, however, unknown from fossil European deposits; it has been introduced artificially into European waterbodies with aquarium plants (Glöer, 2002).
A case of extraordinary convergence with present B. matejici is shown by Miratesta P. & F. Sarasin, 1898, a genus endemic to ancient Lake Poso (Walker & Ponder, 2001). The shell of its only representative, M. celebensis P. & F. Sarasin, 1898, has a very similar shape and shows the same type of ornamentation, featuring broad axial ribs and numerous thin spiral keels (P. Sarasin & F. Sarasin, 1898). A marked difference is the wide anterior emargination of the peristome.
Co-occurring with B. matejici in the SLS, the genus Orygoceras is a representative of uncoiled Planorbidae (Papp, 1962;Harzhauser, Kowalke & Mandic, 2002;Neubauer, Mandic & Harzhauser, 2011;Neubauer et al., 2016). The dentaliiform-to-curved morphology probably originated in one of the slightly older lakes of the nearby Dinaride Lake System in Croatia and Bosnia and Herzegovina. Earlier authors have attributed the genus variously to valvatids, planorbids and even the marine Caecidae (Harzhauser et al., 2002). In this case, again, the protoconch has shed light on its systematic classification. The presence of widely spaced faint spiral striae on its planispiral protoconch suggests placement in the Planorbidae (Harzhauser et al., 2002;Neubauer et al., 2011).
The present study demonstrates and further expands our knowledge of the extraordinary range of morphological disparity of the Planorbidae. Moreover, the present taxonomic revision reveals the only phenotypic radiation of the genus Bulinus on the European continent. This conclusion was only possible because of the detailed assessment of the embryonic shell. Especially in the absence of molecular data, SEM-based examination of the early shell should be preferred over morphological assessment of the highly variable characteristics of the adult shell when trying to unravel the systematic position of problematic freshwater gastropods.

SUPPLEMENTARY MATERIAL
Supplementary material is available at Journal of Molluscan Studies online.