Our knowledge of Neogene chelonian diversity in northern Greece is increased with the present description of a new species of Mauremys (Testudines, Geoemydidae) from the late Miocene to Pliocene of three localities in central Macedonia (Gefira‐2, Nea Silata, Allatini). This new species, Mauremys aristotelica sp. nov., is characterized by the presence of exceptionally wide vertebral scutes, a trait that is quite rare within Mauremys but has evolved independently in other pan‐testudinoid non‐testudinids. Total evidence phylogenetic analysis confirms the placement of the new species within Mauremys and reveals that its closest relative is Mauremys campanii from the late Miocene of Italy. It is also likely, under parsimony, that all geoemydids with similarly wide vertebral scutes from the Neogene of Eurasia form a clade nested within Mauremys. Our results also shed light on the evolution of geoemydids in the eastern Mediterranean during late Miocene to Pliocene times.
Until recently, the chelonofauna from Greece was poorly known. From the end of the nineteenth century, a few isolated publications presented turtle material from Greece (see historical overview in Vlachos 2015a). More recently, Georgalis & Kear (2013) gave an overview of the published information on the Greek turtles and tortoises, identifying a single pleurodiran taxon, several testudinids of both giant and small size, as well as geoemydids and emydids (Georgalis & Kear 2013 and references therein). Over the last few years, interest on the chelonian fossil record from Greece has intensified, leading to the discovery of new localities with species already known from Greece (e.g. Garcia et al. 2016; Vlachos & Tsoukala2014; Vlachos & Delfino 2016; Georgalis et al. 2018), revision of previously published specimens (Vlachos et al. 2015a), erection of new species (Georgalis et al. 2013; Vlachos et al. 2014; Vlachos & Tsoukala 2016), as well as new additions to the Greek palaeochelonofauna (e.g. the first trionychids from Greece; Vlachos et al. 2015b, Georgalis et al. 2016).
A detailed study of the vast majority of the chelonian fossils from Greece was conducted by one of us (Vlachos 2015a), showing that the Greek fossil turtle record is much more diverse and widespread than previously thought. Among the new chelonian material that has come to light through fieldwork during the last decade, we have been able to identify a new geoemydid turtle from three localities around Thermaikos Gulf, near Thessaloniki in northern Greece. This species is characterized by extremely wide vertebral scutes on the carapace, even wider than in the roughly coeval Mauremys campanii (Ristori, 1891) from the late Miocene of Italy (Chesi et al. 2009). In the past, other pan‐testudinoids from the Neogene of Europe have developed such wide vertebrals, for example, Clemmydopsis Boda, 1927, Sakya Bogachev, 1960, Sarmatemys Chkhikvadze, 1983, and ‘Melanochelys’ sakyaformis Redkozubov, 1991 (see Danilov 2005; Danilov et al. 2017; and references therein). However, the taxon described herein, as well as M. campanii, suggest as a working hypothesis that this trait has also evolved in Mauremys in the late Miocene or early Pliocene; modern Mauremys have much narrower vertebral scutes.
The primary objective of the present paper is to describe the fossil specimens attributed to this new Mauremys species, identified in three localities from Northern Greece: Gefira‐2 in Chalkidona Municipality, Nea Silata in Nea Propontida Municipality, and Allatini in Pylea–Chortiatis Municipality (Fig. 1A). We further describe some testudinid remains from the Nea Silata locality. Additionally, we conducted a total evidence parsimony analysis based on the combination of recently published matrices (morphology: Vlachos & Rabi 2017; molecules: Pereira et al. 2017) to test the generic attribution of the new species and its possible relationship with M. campanii from the late Miocene of Italy.Figure 1 Open in figure viewerPowerPoint Locality information. A, map of Greece with a close‐up of the circum Thermaikos area showing the localities discussed in the text, Gefira (Municipality of Chalkidona), Nea Silata (Municipality of Nea Propontida) and Allatini (Municipality of Pylea – Chortiatis); type locality is marked with a star. B, panoramic view of the sand pit in Gefira, type locality of Mauremys aristotelica sp. nov.; area of close‐up is indicated. C, close‐up of the sand pit section where the type specimen was recovered. D, stratigraphic interpretation. E, the specimen in situ, cut in a diagono‐transversal plane; not to scale. Legend: 1, present‐day soil; 2, poorly consolidated, medium‐to‐coarse gravels and sands; 3, well‐consolidated brownish‐ochre sandy clays; 4, well‐consolidated brownish clays; 5, well‐consolidated greyish clays; 6, consolidated dark grey sands with gravels, with cross‐bedding on the upper parts, including traces of organic material; 7, fine‐grained greyish sands; 8, poorly‐consolidated dark grey marly sands, containing the turtle fossil; 9, medium‐to‐coarse sands with some gravels; 10, coarse gravels and sands; 11, debris. Colour online.
LGPUT, collection of the Laboratory of Geology and Palaeontology of the Aristotle University of Thessaloniki, Greece.
Bony plates are indicated with lower case letters, horny scutes with upper case. ABD, abdominal; amdf, anterior musk duct foramen; AN, anal; CE, cervical; cos, costal; en, entoplastron; ep, epiplastron; FE, femoral; GU, gular; hyo, hyoplastron; hypo, hypoplastron; HU, humeral; MA, marginal; ne, neural; nu, nuchal; PEC, pectoral; pe, peripheral; PLE, pleural; pmdf, posterior musk duct foramen; py, pygal; sp, supragygal; VE, vertebral; xi, xiphiplastron.
GAS‐2, Gefira‐2 locality; PBDB, PaleoBiology Database; SLT, Nea Silata locality.Locality information
The present‐day Axios valley and Thermaikos gulf are the relics of a large, elongated, tectonic depression trending from NNW to SSE. This depression was formed during the Neogene and gradually filled up with Neogene–Quaternary sediments (Syrides 1990).
According to Koufos & Pavlides (1988), the deposits between the villages Gefira and Vathylakkos (lower Axios valley) belong to the Gefira Member of the Angelochori Formation, which consists mainly of alternating gravels and sands and overlies the Megalo Emvolon Member. The latter is exposed in the cape of Megalo Emvolon (also known as Karaburun or Karabouroun), in the western Chalkidiki peninsula near Thessaloniki, and consists of reddish marls and grey sandy marls alternated with gravels. Koufos & Pavlides (1988) mentioned that the Megalo Emvolon Member is of an early Ruscinian age (MN 14), whereas the Gefira Member is younger and terminates probably in the Pliocene (MN 15–16). In a later work (Koufos et al. 1991), the Megalo Emvolon sediments were attributed to the Gonia Formation (Syrides 1990).
The Gefira chelonian material comes from two locations near the village of Gefira. To avoid confusion and a potential mixing of the fauna, we will treat these two collections as distinct, under the names Gefira‐1 or GAS‐1 (for the material of Vlachos et al. 2015b and Crégut‐Bonnoure & Tsoukala 2017) and Gefira‐2 or GAS‐2 for the material described herein. The first collection came from a fully‐exploited sandpit (PBDB no. 182685; 40°45′13″ N, 22°40′11″ E; palaeocoordinates 40.6° N, 22.5° E) and contained the pan‐trionychid material described in Vlachos et al. (2015b). The associated fauna and, especially, the association of Anancus arvernensis with hipparions, cervids and boselaphini indicate a late Pliocene age for the material in this collection (see Vlachos et al. 2015b and Crégut‐Bonnoure & Tsoukala 2017 for further information). The new turtle specimens described herein come from a second collection (JS and EV, February 2015), on the south‐western part of the vertical slopes of another sandpit in Gefira (Fig. 1B; PBDB no. 196230; 40°44′55″ N, 22°40′11″ E; palaeocoordinates 40.6° N, 22.5° E). The sediments of this section can be divided macroscopically into at least ten different layers, consisting of alternations of gravels, sands and clays of grey to brownish colour (see Fig. 1C–D for more information). The turtle was found in poorly‐consolidated dark grey marly sands (Fig. 1E), in the bottom half of the section. The sediments of GAS‐2 are similar to those of GAS‐1, situated at approximately the same altitude and quite near to each other; although a direct comparison is not possible, it is safe to assume that they represent more‐or‐less the same timeframe. Several additional field visits failed to yield more fossils in the GAS‐2 sandpit, and thus we are unable to provide a detailed chronology of the site. Taking into account all the available information, we attribute a Pliocene age to the material of GAS‐2, pending corroboration with future discoveries.
The fossil locality Nea Silata (PBDB no. 191612; 40°21′57″ N, 23°6′29″ E; palaeocoordinates 40.1° N, 22.8° E) is located in a small ravine with steep sides near Nea Silata village. The stratigraphy of the area of western Chalkidiki is well known (Syrides 1990). This region contains sediments that are arranged in several formations and consist mainly of clastic sediments (sands, silts, clays, gravels, conglomerates) that were deposited during the middle to late Miocene (conglomerates, Antonios Fm.), late Vallesian – early Turolian (redbeds, Triglia Fm.), latest Miocene (late Turolian ‘Pontian’ mollusc‐bearing clays, sands and limestones, Trilophos Fm.) During the Pliocene (Ruscinian), clastic sediments intercalating with massive marly limestones were deposited in the area of Koufalia–Pella, as well as in west Chalkidiki (Gonia Fm.) These marly limestones divide the latter formation into three members (Silata Mb., Rhodokipos Mb., Kallikratia Mb.) The sediments of the Gonia Formation indicate a fluvial–shallow‐lacustrine–marshy palaeoenvironment. Quaternary terrestrial sediments (redbeds, Moudania Fm.) formed the youngest, extensive, cover in western Chalkidiki (Syrides 1990). The deposits from Nea Silata belong to the Silata Member of the Gonia Formation. The study of both micro‐ and macromammals (Vasileiadou et al. 2003 and Koufos 2006, respectively) from the locality suggests a correlation of the fauna near the Turolian–Ruscinian boundary (MN13/14). According to the study of Koufos & Vasileiadou (2015), which is based on radiochronological and palaeontological data from northern Greece, an age of 5.4–5.23 Ma is possible for the SLT locality. Please see references cited above for further information.
The material from Allatini locality (PBDB no. 182683; 40°35′28″ N 22°58′49″ E; palaeocoordinates 40.3° N, 22.7° E) probably comes from the former clay pits (closed during the 1970s) of the homonymous private company (see Vlachos et al. 2015a). More recent residential building in the area prohibits any observation of these deposits. However, it is accepted that they formed part of the Trilophos Formation of a latest Miocene age (Syrides 1990; Vlachos et al. 2015a).Material and method
The Nea Silata locality was discovered by GS, who collected small samples during fieldwork for his doctoral thesis (Syrides 1990). More extensive sampling (focusing on micromammalian remains) was part of the master's thesis of KV. The material studied and presented herein is the result of sieving procedure and was briefly mentioned by Vasileiadou et al. (2003). EV assigned the specimen numbers during his doctoral thesis (Vlachos 2015a). Based on the compilation of the isolated fragments from Nea Silata, Vlachos (2015a) suggested that the material could belong to a new species. A few days after EV defended his thesis (February 2015), JS discovered the partial shell from Gefira‐2. Vlachos et al. (2015b) described additional material from Gefira‐1, consisting of isolated carapace plates of a pan‐trionychid. The Allatini material was fully described by Vlachos et al. (2015a), but only attributed to Mauremys sp. therein. All the material mentioned above forms part of the collections of LGPUT.
Parsimony analysis was based on the matrix of Vlachos & Rabi (2017), which is mainly based on the original testudinoid matrix of Joyce & Bell (2004) and other smaller matrices (see Vlachos & Rabi 2017 and references therein). To this original matrix of 170 characters we added a set of new characters: from the recently published matrix of Garbin et al. (2018) we selected their characters 1, 2, 4, 5, 18, 28, 29, 41, 44, 45, 47, 48, 57, 58, 59, 62, 64, 65, 68, 69, 70, 71 and 74, added as characters 171–194 to our matrix. These added characters from Garbin et al. (2018) further code the morphology of geoemydids and cover areas of the shell not represented in the Vlachos & Rabi (2017) matrix. We further added some new characters to account for the morphology of extinct geoemydids with extremely wide vertebral scutes:
Whereas the analyses of Vlachos & Rabi (2017) included a larger pan‐testudinid ingroup, our analyses herein mainly focus on Geoemydidae. In particular, we used Emys orbicularis (Linnaeus, 1758) and Testudo graeca (Linnaeus, 1758) as outgroups, and all of the original geoemydid taxa of Joyce & Bell (2004; see Vlachos et al. 2018) as the ingroup. The scorings of the characters added from Garbin et al. (2018) for the geoemydid taxa were used unchanged. We further added Mauremys aristotelica sp. nov., scored based on personal observations, and the late Miocene Mauremys campanii (Ristori, 1891) from Italy, scored based on Chesi et al. (2009). We also added other geoemydid taxa with wide vertebral scutes, including Clemmydopsis mehelyi (Kormos, 1911) from the Pleistocene of Hungary, Sakya (as a chimaera taxon with the carapace scored based on Sakya riabinini (Khosatzky, 1946) from the Pliocene of Ukraine, and the plastron based on Sakya kolakovskii Chkhikvadze, 1983 from the Pliocene of Georgia), Sarmatemys lungui Chkhikvadze, 1983 from the late Miocene of Moldavia and ‘Melanochelys’ sakyaformis from the late Pliocene of Moldavia, based on the figures of Danilov (2005) and Danilov et al. (2017), and Shansiemys latiscuta Yeh, 1963 from the late Miocene to Pliocene of China based on the figures of Yeh (1963).
We analysed the data in a total evidence analysis using the molecular matrix of Pereira et al. (2017). The matrices were constructed in Mesquite v. 3.2 (Maddison & Maddison 2017). The morphological matrix consists of 33 taxa and 209 characters. The molecular matrix consists of 24 taxa and 12 354 characters. The total evidence matrix was compiled manually as a combination of the two separate matrices. Parsimony analysis was conducted in TNT v. 1.1 (Goloboff et al. 2008), with a traditional search TBR with 1000 replicates. Consistency and retention indices were calculated with the stats.run script, Bremer support with the bremer.run script, Bootstrap and Jackknife GC frequencies (100 replicates) with the Resampling command in TNT.Systematic palaeontology
Aristotelica, from Aristotle (Αριστοτέλης or Aristotelis, 384–322 BC), a Greek philosopher and scientist born in Stageira, Chalkidiki, in the broader area where this species is distributed. The year 2016 (when we started working on this project) was declared by UNESCO as ‘Aristotle Anniversary Year’ honouring the impact of Aristotle's works on science. In his book On the Movement of Animals he made for the first time the distinction between tortoises (χελώναι) and turtles (εμύδες), introducing the name emys that has dominated the chelonian literature ever since (see Vlachos 2015b, p. 5). The name also honours the home institution of most authors of this paper, the Aristotle University of Thessaloniki, Greece, a public institution striving to promote palaeontology in Greece and abroad.
Gefira‐2 (GAS‐2) sandpit, near Gefira village, lower Axios valley, northern Greece (Fig. 1), Pliocene.
Isolated shell plates and appendicular elements from the late Miocene or early Pliocene locality Nea Silata, W. Chalkidiki (northern Greece), LGPUT SLT collection (see Appendix for specimen numbers); LGPUT GG 23, partial shell, from the late Miocene or early Pliocene locality Allatini (northern Greece), studied by Vlachos et al. (2015a).
Member of Geoemydidae based on the following combination of characters: presence of anterior and posterior musk duct foramina; neural 1 quadrangular followed by hexagonal neurals with short anterolateral sides; rib heads wide and flat, placed near the neural/costal suture on the costals; 12th marginals divided, with their sulcus expanding on the suprapygal 2 that is crossed by the posterior border of vertebral 5. Member of Mauremys sensu lato based on the following combination of characters: vertebral 1 making contact with marginal 2; entoplastron crossed by the humero‐pectoral sulcus on the posterior part; short anal scutes; deep rounded anal notch. It can be differentiated from other species of Mauremys sensu lato based on the extremely wide vertebral scutes, which almost reach the marginals, squeezing the pleural scutes to the lateral parts of the carapace; the width of these vertebral scutes is even larger than is seen in Mauremys campanii (see Chesi et al. 2009).Description
The Gefira‐2 material (LGPUT GAS 34) consists of a partial carapace, missing the nuchal and several parts of the left side, as well as an almost complete plastron missing the entoplastron and some parts of the hyo/hypoplastra. The Nea Silata material (LGPUT SLT; see Appendix for numbers) consists of numerous disarticulated plates and some appendicular elements that represent almost the entire shell of the taxon. The Allatini material (LGPUT GG 23) consists of a partial shell; given that the detailed description of this specimen has been presented recently (Vlachos et al. 2015a) it will not be repeated here. All the specimens can be attributed to the same species, based on the presence of the characteristic wide vertebral scutes, coupled with the temporal and geographical proximity. The following description is based on the entire referred material, but in some cases, we will cite specific specimens to emphasize some morphologies. The analysis of the entire available material resulted in the reconstruction of the morphology of the shell of this taxon (shown in Fig. 2D for carapace and Fig. 3D for plastron), based mainly on the Gefira‐2 material, completing some missing areas with the Nea Silata material.
The dorsal surface of the carapace is smooth with no signs of ornamentation. The development of the sutures in the shell of the holotype (LGPUT GAS 34) indicates an adult individual, with no evidence of a complete medial or any lateral keels. It is likely that a medial keel was present at least in the anterior half, but the preservation of the material does not allow confirmation of this observation. However, the presence of keels is noted in smaller specimens (e.g. Fig. 4AD–AG; Nea Silata), suggesting the presence of lateral keels in juvenile individuals. The disarticulated nature of the material of juveniles does not allow further observations on the keel morphology. The nuchal plate is hexagonal in outline. The anterior border shows a shallow nuchal notch or emargination that affects only the nuchal border but not the first peripherals. The anterolateral sides of the nuchal are slightly longer and slightly concave medially, compared to the posterolateral ones that are shorter and straight. The posterior border of the nuchal is narrow and convex anteriorly. Overall, the nuchal is slightly wider than long, making contact with the first pair of peripherals, the first pair of costals and neural 1.
Neural 1 is quadrangular with rounded sides, longer than wide, making contact with the nuchal anteriorly, the first pair of costals laterally and the second neural posteriorly. The following neurals (2–7) are all hexagonal, showing a posteriorly concave anterior border, short anterolateral and long posterolateral sides and a posteriorly convex posterior border. Neural 2 is the shortest, neural 3 the longest, and the length of the rest gradually diminishes towards the posterior part. They make contact with the anterior and corresponding pairs of costals (i.e. neural 2 contacts costals 1 and 2). Neural 8 is the smallest in the neural series, quadrangular with rounded sides, almost as long as wide. It contacts the neural 7 anteriorly, the costals 8 laterally and the suprapygal 1 posteriorly. Viscerally, all preserved neurals show remains of the attachments of the thoracic vertebrae.
Suprapygal 1 is trapezoidal in shape with wider posterior side. The posterior border is convex posteriorly and contacts the middle part of the suprapygal 2. It contacts neural 8 anteriorly and the costals 8 laterally. Suprapygal 2 is short and much wider than suprapygal 1. It shows an inverted trapezoid shape with the anterior border wider than the posterior one. It contacts the suprapygal 1 and the posteromedial sides of the costals 8 anteriorly, the pair of peripherals 11 laterally and the pygal posteriorly. The pygal is short and wide, roughly quadrangular with a short pygal notch posteriorly. It contacts the suprapygal 2 anteriorly and the last pair of peripherals laterally.
Eight pairs of costals are present. The first pair of costals is much longer than the rest, being only slightly wider than long. They contact the nuchal anteromedially, neural 1 and 2 medially and peripherals 1–3 laterally. The anterior border is round, whereas the posterior one is straight. Costals 2–6 are similar in shape, being much wider than long. The medial side shows a long anteromedial and short posteromedial side, to articulate with the shape of the corresponding neurals. Costals 2 contact only the peripherals 4, costals 3 both peripherals 5 and 6, costals 4 only peripherals 6, costals 5 both peripherals 7 and 8, costals 6 both peripherals 8 and 9. Costals 7 are also wider than long, but the medial side is single and straight contacting only neural 7; they contact both peripherals 9 and 10 laterally. Costals 8 are narrower, as long as wide, contacting the anterior costals, the posterolateral sides of neural 7, neural 8, both suprapygals and peripherals 10–11. The axillary buttress makes a clear contact with costal 1, whereas the inguinal buttress contacts both costals 5 and 6. The buttresses extend to the middle of the costals viscerally. Although not fully preserved, the contact between the costals and the buttresses must have been sutural.
All peripherals are wider than long, and peripherals 3–8 form the bridge. On the ventral side of peripherals 3 and 7 (close to the suture with peripheral 8), musk duct foramina are present. The peripherals (excluding the peripherals of the bridge) are elongated and oval‐shaped in cross‐section.
Most probably, the presence of three carinae characterized the juvenile individuals, one in the medial plane, and one on each side. Also, it appears that the relative coverage of the vertebral scutes in the juvenile carapace (LGPUT SLT 14, Fig. 4AD–AG) was even larger than in the adult individuals (LGPUT GAS 34).
A cervical scute is present anteriorly, both dorsally and viscerally. It is only slightly longer than wide, covering approximately 25% of the nuchal on dorsal view. Viscerally it is much shorter, being wider than long. Overall, vertebrals 1–4 are extremely wide, covering more than half of the dorsal surface of the carapace. The width of vertebrals 1–4 is more than twice their length. Vertebral 1 contacts the cervical, marginals 1–2, pleural 1 and vertebral 2. It expands on the nuchal, peripherals 1–2, neural 1 and costal 1. Vertebral 2 contacts vertebral 1 anteriorly, pleurals 1–2 laterally and vertebral 3 posteriorly. It expands on costals 1–3, neural 1 and neural 3 and covers the entire neural 2. The sulcus between vertebral 1 and 2 is wavy in its medial part, in the part that is crossing neural 1. Vertebral 3 contacts vertebral 2 anteriorly, pleurals 2–3 laterally and vertebral 4 posteriorly. It expands on costals 3–5, neural 3 and neural 4 and covers the entire neural 4. The sulcus between vertebral 2 and 3 is gently convex anteriorly on the part crossing neural 3. Vertebral 4 contacts vertebral 3 anteriorly, pleurals 3–4 laterally and vertebral 5 posteriorly. It expands on costals 5–7, neural 5 and neural 7 and covers the entire neural 6. The sulcus between vertebral 3 and 4 is gently convex anteriorly on the part crossing neural 5. Vertebral 5 is the smallest vertebral, contacting vertebral 4 anteriorly, pleural 4 laterally and marginals 11–12 posteriorly. It expands on costals 7–8, neural 7 and suprapygal 2 and covers the entire neural 8 and suprapygal 1. The sulcus between vertebral 4 and 5 is gently convex anteriorly on the part crossing neural 7.
The pleurals are quite small, oval to lenticular, longer than wide, squeezed between the expanded vertebrals and the marginals. There is a tiny contact between the pleurals. Pleural 1 contacts marginals 2–4, pleural 2 marginals 4–7, pleural 3 marginals 7–9 and pleural 4 marginals 9 and 10. The sulcus between the pleural 4 and vertebral 5 coincides with the marginal 10/11 sulcus. Marginals 12 expand on the suprapygal 2 and, thus, their intersulcus completely divides the pygal.
The anterior lobe is shorter than the posterior one, and there is a shallow ventral concavity in the middle part of the plastron. The anterior border of the anterior lobe of the plastron is mostly composed by the epiplastra, which show a small constriction in the gularo‐humeral sulcus. In visceral view, there is no epiplastral lip; the region in the medial suture of the epiplastra is concave, whereas the lateral parts of the epiplastra in visceral view show the presence of thickened tuberosities, which are elongated and parallel to the gularo‐humeral sulcus. The entoplastron (observed in the referred material from Nea Silata) is rhomboidal with rounded posterior sides, and its two halves have equal or subequal lengths. There is evidence of a posterior elongation on the visceral side of the entoplastron that was certainly expanded onto the hyoplastra, but its extent cannot be observed because of the preservation state of the material. The hyoplastra contribute slightly to the anterior lobe and contact the epiplastra and the entoplastron anteriorly and the hypoplastra approximately in the middle of the plastron. The hyoplastra are wider than long and form axillary buttresses (viscerally) that contact the costal 1. The hyoplastra most certainly contacted the peripherals 3–5 in the anterior part of the bridge. The hypoplastra are longer than wide and contribute to the half of the posterior lobe. Their suture with the hyoplastra is almost straight and perpendicular to the midline, whereas the suture with the xiphiplastra is angular‐to‐rounded. The hypoplastra contact certainly the peripherals 6–7 to form the posterior part of the bridge; due to the lack of natural contact between the plastron and the carapace, it is not possible to certify whether the hypoplastra contacted also the peripheral 5. The xiphiplastra are slightly wider than long, forming the posterior part of the posterior lobe. The xiphiplastra have a constriction in the border right on the femoro‐anal sulcus and a deep, rounded, anal notch posteriorly. As such, two elongated and rounded extremities are evident in the posterior parts of the xiphiplastra. Plastral formula (across the midline): hypo>hyo>xi>en>ep.
In visceral view, the covering of the plastral scutes on the bony plates is expanded, as in crown Testudinoidea. The gulars are longer than wide, and narrow as the gularo‐humeral sulcus forms an obtuse angle. They expand posteriorly to cover the anterior third of the entoplastron. The humerals are wider than long, and medially much shorter than the gulars; they are in fact the short medially scutes of the plastron. The pectorals are wider than long with wavy humero‐pectoral and pectoro‐abdominal sulci: they are concave medially and convex laterally. The anterior part of the pectorals covers the posterior third of the entoplastron. Most probably, the pectorals contacted the marginals 4–6. The abdominals are wider than long and the largest scutes of the plastron, showing a medially concave abdomino‐femoral sulcus. Most probably, the abdominals contacted the marginals 6–8. The femorals are wider than long, covering most of the anterior part of the posterior lobe. The femoro‐anal sulcus is angular‐to‐rounded and convex. The medial length of the anals is shorter than that of the femorals. Plastral formula (across the midline): ABD>PEC=FE>GU>AN>HU.Phylogenetic analysis
The parsimony analysis of the total evidence matrix resulted in three most‐parsimonious trees of 5146 steps (best score hit 1000 times of 1000; CI: 0.473; RI: 0.399). The topology (Fig. 6) is overall well‐resolved, having only a polytomy of the extinct species within Mauremys. The support is relatively high for all derived clades of Geoemydidae (indicated by >3 Bremer support). Most clades also show important character conflict, as indicated by the low Bootstrap/Jackknife GC frequencies. Mauremys aristotelica sp. nov. is clearly placed within Mauremys, forming the most derived clade together with the late Miocene Mauremys campanii from Italy. This clade is strongly supported by two synapomorphies in all trees (wavy humero‐pectoral sulcus; well‐developed gular protrusion) and one more in tree #2 (cervical scute longer than wide) (see Vlachos et al. 2018). Interestingly, the other extinct taxa with similarly wide vertebral scutes (namely ‘Melanochelys’ sakyaformis, Shansiemys latiscuta, Sarmatemys lungui, Sakya and Clemmydopsis mehelyi) are placed in between Mauremys caspica and the clade of aristotelica + campanii. This clade of extinct geoemydids with wide vertebral scutes shows moderate Bremer support (2) and low GC frequencies and is supported by different sets of synapomorphies in each tree, two synapomorphies in tree #0 (pygal notch; vertebral scutes more than twice wider than long), three synapomorphies in tree #1 (anal notch added to the synapomorphies in tree #0), and three synapomorphies in tree #2 (first pleural is not in contact with the nuchal; contact of the second marginal scute with the first vertebral scute; rounded posterior borders of the entoplastron) (see Vlachos et al. 2018). Although obvious, it is worth stating that the same topology of this extinct clade is recovered from the parsimony analysis of the morphological matrix alone. This topology, which will be discussed below, could have important implications for the taxonomy of these taxa, as well as the evolutionary history of these geoemydids during the Neogene of Eurasia. The inclusion of more extinct geoemydid taxa is beyond the scope of this work but is expected not to alter this result under this set of characters: that at least some of these geoemydids with wide vertebral scutes, including Mauremys aristotelica sp. nov., are within Mauremys sensu stricto.Figure 6 Open in figure viewerPowerPoint Strict consensus of the three most‐parsimonious trees recovered herein, showing the position of Mauremys aristotelica sp. nov. within Geoemydidae. Bremer support and Bootstrap/Jackknife GC frequencies are shown on the nodes. The numbers of the nodes/taxa are shown on the left. Main clade names are indicated. *Extinct taxa added to the matrix of Vlachos & Rabi (2017) for this study. Other taxa from the studied localities
Besides Mauremys aristotelica sp. nov., two other turtle species are identified in two of the three localities mentioned herein. From Gefira, Vlachos et al. (2015b) reported the presence of a pan‐trionychid turtle. From Nea Silata, some of the recovered plates are attributed to a tortoise, specifically to Pan‐Testudo sp. (see Appendix and Fig. 7). These 38 specimens (see Appendix for detailed identifications) are shell fragments, including some costal and peripheral plates that are typical of a small testudinid. In particular, the shape of the preserved costal plates (Fig. 7A–L) indicates the presence of an alternating pattern of neurals (rectangular vs octagonal) and costals (longer medially and shorter laterally). Also, the peripheral plates (Fig. 7M–Q) show a clear coincidence of the pleuro‐marginal sulci and the costo‐peripheral suture, as well as the presence of pointed tips in some peripherals. Plastral specimens (Fig. 7R–U) agree with the thickened shell of a testudinid. Last but not least, most of these specimens show a rough outer surface, with distinct growth lines as in most tortoises. All the above allow a clear attribution to Pan‐Testudo (sensu Vlachos & Rabi 2017) and a distinction from the geoemydid species.Figure 7 Open in figure viewerPowerPoint Pan‐Testudo sp. from the late Miocene – early Pliocene locality Nea Silata (Thessaloniki). LGPUT collection. A–C, SLT 194, costal in: A, dorsal; B drawn dorsal; C, visceral view. D–F, SLT 189, costal in: D, dorsal; E, drawn dorsal; F, visceral view. G–I, SLT 183, costal in: G, dorsal; H, drawn dorsal; I, visceral views. J–L, SLT 6, costal in: J, dorsal; K, drawn dorsal; L, visceral view. M–O, SLT 180, peripheral in: M, dorsal; N, drawn dorsal; O, visceral view. P–Q, SLT 193, peripheral in: P, dorsal; Q, visceral view. R–S, SLT 186, hyoplastron in: R, dorsal; S, visceral view. T–U, SLT 195, plastron fragment in: T, dorsal; U, visceral view. In drawings, natural borders are shown with thick black lines, sutures with thinner black lines and sulci between scutes in grey. Scale bar represents 5 cm. Colour online. Discussion
Among Geoemydidae, there are other taxa presenting wide vertebrals, as summarized by Danilov (2005, fig. 80). The Pliocene Clemmydopsis mehelyi from Hungary (see Danilov 2005, fig. 80A–B) lacks the first pleural scutes, showing quite wide vertebrals 1–3 as a result, in contrast with the situation in Mauremys aristotelica sp. nov. Also, the plastron of Clemmydopsis mehelyi shows a short anal notch, which is different from the condition seen in extant Mauremys and the new species described herein. The Pliocene Sakya kolakovskii from Georgia (see Danilov 2005, fig. 80C–D) also shows wide vertebrals that reach the lateral parts of the costals, in contact with narrow pleurals. In this case, however, the number of vertebrals is higher than normal, showing 10 wide and narrow vertebral scutes. This character suggests that Mauremys aristotelica sp. nov. cannot be associated with Sakya. The Miocene Sarmatemys lungui from Moldavia (see Danilov 2005, fig. 80E–F) shows some similarities with Mauremys aristotelica sp. nov. in the shape of the vertebrals and the shape of the anal notch. However, the two species differ in many other aspects: in S. lungui there is no nuchal notch in the carapace, the cervical and marginals 1 have the same length, there are no inter‐pleural sulci crossing the costals as the vertebrals extend up to the costo‐peripheral sutures, the configuration of the costal 7/8 in relation to the sulci is different from the diagonal crossing of the vertebro‐pleural sulcus, the pleurals do not overlap the peripherals (only in the anterior area), the entoplastron is not overlapped by the gular scutes and the humero‐pectoral sulcus is rather straight (instead of sinuous as in Mauremys). ‘Melanochelys’ sakyaformis from the late Pliocene of Moldavia (see Danilov et al. 2017) is generally similar to Mauremys aristotelica sp. nov., but its pleural 1 contacts the marginal 1, and the sulcus between the vertebral 1/2 is straight in the part crossing neural 1. On the other hand, Shansiemys latiscuta from the late Miocene to Pliocene of China (Yeh 1963) is strikingly different, as the pleural scutes appear to be missing entirely and the vertebral scutes contact directly the marginals.
Extant species of Mauremys (e.g. Mauremys rivulata (Valenciennes in Bory de Saint‐Vincent, 1833), Mauremys leprosa (Schweigger, 1812)) always have narrow and long vertebrals. Among the known fossil forms of Mauremys, the only taxon in which the vertebrals approach a width comparable with that observed in Mauremys aristotelica sp. nov. is Mauremys campanii (Chesi et al. 2009), but even in this species the vertebrals are still narrower. Other fossil Mauremys species from Italy (Chesi 2008) have narrower and longer vertebrals as well. The vertebrals of Hardella thurjii Gray, 1831 are wider than in other extant geoemydids, but still almost twice longer than in Mauremys aristotelica sp. nov. Whereas the carapaces of the above‐mentioned taxa are radically different from other geoemydids (and even within this group the type and configuration of the wide vertebral scutes exhibits enormous variation) the morphology of the plastron is quite conservative and does not allow documenting important differences.
In Mauremys aristotelica sp. nov. the humerals are medially very short, increasing the area of the entoplastron that is covered by the gulars and pectorals. This condition is found only in some species of Mauremys (i.e. Mauremys caspica (Gmelin, 1774), Mauremys leprosa, Mauremys sarmatica (Purschke, 1885), Mauremys gaudryi (Depéret, 1885); see character 49, state 4, and fig. 3.21 in Hervet 2003), whereas the gulars and pectorals cover less of the entoplastron in Mauremys portisii, Mauremys pygolopha (Peters, 1868) and some species of Palaeochelys (see Hervet 2003, 2004). Another character supporting the taxonomic placement of Mauremys aristotelica sp. nov. within Mauremys is the narrow and deep anal notch. The anal notch of Mauremys aristotelica sp. nov. is somewhat rounded as in M. sarmatica and M. pygolopha, but more similar to the angular notch seen in M. caspica and M. leprosa. However, the anal notch is wider in these two extant taxa than in Mauremys aristotelica sp. nov. Compared to Mauremys sp. from San Giovanni di Sinis, Italy (Chesi et al. 2007), Mauremys aristotelica sp. nov. shows a deeper and narrower anal notch. Compared to the late Miocene Mauremys campanii (comb. nov. in Chesi et al. 2009), Mauremys aristotelica sp. nov. shows a deeper nuchal notch, and a more posteriorly located humero‐pectoral sulcus. Finally, Mauremys aristotelica sp. nov. is distinct from Mauremys sp. from the early late Miocene of Plakias (Crete), described by Georgalis et al. (2016) as the Plakias Mauremys, it most certainly had narrow vertebrals. The presence of this new species indicates that the late Miocene to early Pliocene diversity of geoemydids in the eastern Mediterranean was higher than previously thought.
Our initial working hypothesis was that the appearance of the wide vertebral scutes occurred several times in the evolution of geoemydids. This hypothesis was based not only on the radical differences on the carapace morphology of these taxa, but also on external evidence, such as their temporal and geographical position. Finally, the fact that they have been attributed to several distinct extant and extinct genera added some points to our initial hypothesis. However, the phylogenetic analysis herein suggested as a more parsimonious conclusion that all these should be considered as members of a monophyletic clade, situated in a derived position within Geoemydidae.
If so, then it seems that the presence of such wide vertebral scutes occurred once in Geoemydidae, and led to further, extreme, specializations. Also, the same morphology has evolved at least once more in pan‐testudinoids, in the peculiar pan‐emydid turtle Acherontemys heckmanni from the middle Eocene of North America (see Hay 1899, 1908; Vlachos 2018).
Accepting this topology would imply some important taxonomic changes to the generic combination of these extinct species; in fact, they should be transferred to Mauremys. These changes would pose no problem from a temporal point of view, as this clade has existed since the Oligocene, and the clade of Mauremys caspica (the sister group of the extinct clade herein) since the beginning of the Miocene, based on a recent molecular clock estimate (Pereira et al. 2017). Therefore, the existence of this clade would be corroborated from a chronostratigraphic point of view. However, we would have to accept that at the beginning of the Miocene a new clade of Mauremys with extremely wide vertebral scutes appeared and was distributed all over Eurasia. It is worth noting that even extant Mauremys shows a similar distribution (TTWG 2017 and references therein). This clade soon became quite diverse and widespread, but it became extinct at the end of the Neogene. These results offer an intriguing evolutionary scenario for these species, but they should be further tested by future analyses that include even more extinct geoemydid taxa; this is beyond the scope of this paper. However, even if changes occur to the composition of this peculiar clade of extinct turtles, we feel confident that the new species erected herein will remain within Mauremys as this position seems to be well‐supported.
Chesi et al. (2009) discussed the effect of the Messinian Salinity Crisis on the herpetofaunal communities, suggesting that turtle faunas were not affected; as shown by the survival of Mauremys and Trionyx after the Messinian. Although the Greek late Miocene fossil record for these clades believed to be incomplete, as only a few localities with turtle remains of these clades are known, a similar pattern emerges. Both Mauremys and pan‐trionychids are present in the early late Miocene of Plakias (Georgalis et al. 2016), and these clades continue to exist in the latest Miocene and the Pliocene as well. These occurrences support the suggestion of Chesi et al. (2009) regarding the presence of Mauremys in southern Europe in pre‐ and post‐Messinian times. In particular, the presence of Mauremys aristotelica sp. nov. shows that the clade with M. campanii was more diverse and more widely distributed. As M. campanii is certainly known from pre‐Messinian times (Chesi et al. 2009), there is a possibility that Mauremys aristotelica sp. nov. represents a post‐Messinian dispersal event to the southern Balkans; Vamberger et al. (2015) have already shown that the tolerance of Mauremys to salty water would allow it to overcome potential biogeographical barriers in the region and thus transoceanic dispersal. This tolerance was further confirmed by Mantziou & Rifai (2014), who highlighted that Mauremys rivulata could overcome extreme heat that could cause drying of the water bodies.
Chesi et al. (2009) further reported that Testudo sensu lato was present in Italy only after the Messinian, possibly correlated with new peri‐ or post‐Messinian dispersal events. Small‐sized testudinids, however, were already present in pre‐Messinian times in Greece (Vlachos & Tsoukala 2016; Vlachos 2015a; Paraskevaidis 1955; Georgalis & Kear 2013; and references therein) and Asia Minor (Staesche et al. 2007 and references therein). Concerning the presence or absence of taxa above‐species‐level, as far as we can tell, turtle faunas of Greece were not affected by the changes during the Messinian. But our efforts need to be increased, both from a taxonomic, as well as from a stratigraphic and fieldwork point of view, in order to add to the fossil record and be able to discuss those changes in diversity in a species level.
Extant Mauremys rivulata (the species that is distributed in Greece at present) usually inhabits slow‐moving water bodies, such as lakes, ponds and lagoons, but mountain ridges appear to be a major barrier to its distribution (Mantziou & Rifai 2014). Based on the sedimentological context of the fossil localities, a similar environment could be inferred for Mauremys aristotelica sp. nov. Its presence in at least three different localities around Thermaikos Gulf implies that this species was capable of overcoming boundaries imposed by the conditions of water bodies and temperature.
The presence of wide vertebral scutes on this species is a peculiar character that allows for some speculation regarding its potential utility. Mantziou & Rifai (2014, p. 080.4) reported that the carapace of adult M. rivulata is ‘usually uniformly brown, or olive green to green, with faint or no pattern … [but younger individuals] exhibit obvious, reddish‐brown or sometimes yellowish light and irregular reticulate patterns, that extend beyond the scutes to the adjacent scutes on a green to dark background.’ They further explained how adults and juveniles of this species often do not share the same place in the water bodies; juveniles are usually in shallower and muddier water compared to adults. As such, the patterns of juveniles would provide a better camouflage. Finally, the juvenile turtles always exhibit wide vertebral scutes in relation to their size. Based on all the above, and if we extend those observations to the new extinct taxon following an actualistic approach, we could speculate that the presence of the wide vertebral scutes in Mauremys aristotelica sp. nov. is a retention of a juvenile character (heterochrony/paedomorphosis). If, together with the width of vertebral scutes, the carapace of this species also retained the juvenile pattern, it would provide camouflage and protection against predators for a longer period of time (Fig. 8).Figure 8 Open in figure viewerPowerPoint Reconstruction of Mauremys aristotelica sp. nov. showing the inferred colouration of the wide vertebral scutes that could act as camouflage (by Jorge Gonzalez). Colour online.
Mantziou & Rifai (2014 and references therein) highlighted that competition between extant Mauremys and other emydid species, especially the imported Trachemys scripta, could pose a threat for the survival of Mauremys. Our current knowledge of the fossil Greek turtle record (Vlachos 2015a; Vlachos & Delfino 2016) indicates that emydids were have only been present in the south Balkan Peninsula since Pleistocene times. On the other hand, Mauremys‐like geoemydids are known at least from the beginning of the late Miocene (~9.9 Ma; see Georgalis et al. 2016). This would have allowed geomydids to thrive and diversify in the Balkan Peninsula for several million years before the emydids arrived.Conclusion
Over the last few years, our knowledge of the diversity, composition and evolution of the palaeochelonofaunas in the Neogene of northern Greece and the circum‐Thermaikos Gulf area has increased dramatically. Here, we describe a new species, Mauremys aristotelica sp. nov., from the late Miocene to Pliocene of the Thermaikos Gulf area, and, specifically, from three localities: Gefira (Pliocene), Nea Silata (late Miocene – early Pliocene) and Allatini (latest Miocene). The new species is closely allied with Mauremys campanii from the late Miocene of Italy, and probably with all known extinct geoemydids with similarly wide vertebral scutes. Our phylogenetic analysis strongly supports the hypothesis that all of these taxa form a monophyletic clade nested deeply within Mauremys. This result would require numerous changes in the taxonomy of the species involved. The new species establishes the presence of geoemydids in the area of Thermaikos Gulf, where at least four different turtle clades are known during the late Miocene to Pliocene interval. Aquatic taxa are represented by the new geoemydid described herein and the pan‐trionychid from Gefira‐1 (Vlachos et al. 2015b). Also, at least two distinct terrestrial turtles are known, including the small‐sized Testudo graeca (see Vlachos et al. 2015a) and the giant tortoise Titanochelon bacharidisi (Vlachos et al., 2014). This emerging composite palaeochelonofauna from circum‐Thermaikos Gulf stands between the Italian peninsula and Asia Minor, and appears to share common elements with both of these areas. On the one hand, there is a clade of Mauremys with wide vertebral scutes present in both Italy and Greece that is so far absent in Asia Minor (this work and Chesi et al. 2009). On the other hand, giant tortoises are known from both Greece and Asia Minor but are so far absent from the Italian fossil record (Vlachos et al. 2014 and references therein). Soft‐shelled turtles, mostly indeterminate, are known from the Neogene of the entire eastern Mediterranean (see Georgalis & Joyce 2017 for review and references therein). Our work highlights that the area of the present‐day Thermaikos Gulf is crucial for our understanding of the evolution of turtles during the final parts of the Neogene.Acknowledgements
We thank Prof. E. Tsoukala for her support during our entire study and research. We thank A. Pereira for providing the molecular matrix used in the total evidence analysis, the paleoartist J. Gonzalez for the reconstructions of the new species, and M. Delfino for valuable comments during various parts of this project. We deeply thank the editor R. Benson, the technical editor S. Thomas and the reviewers I. Danilov and G. Georgalis, for editorial work and comments that greatly improved this manuscript. EV was partially supported by a Research Committee Grant (50141; 2013–2014), Aristotle University of Thessaloniki, Greece. The Willi Hennig Society sponsors the use of TNT software.
Mauremys aristotelica sp. nov (LGPUT collection, 200+ specimens): SLT 1, nuchal; SLT 2, peripheral; SLT 3, first neural; SLT 4, neural; SLT 5, pygal; SLT 7, peripheral; SLT 8, peripheral fragment; SLT 9, costal fragment; SLT 10, peripheral fragment; SLT 11, peripheral fragment; SLT 12, costal fragment; SLT 13, peripheral; SLT 14, costal, juvenile; SLT 15, peripheral; SLT 16, peripheral; SLT 17, left costal; SLT 18, peripheral; SLT 19, peripheral; SLT 20, left costal 7; SLT 21, right costal 7; SLT 22, peripheral fragment, juvenile; SLT 23, peripheral; SLT 24, nuchal fragment; SLT 26, costal fragment; SLT 28, costal fragment; SLT 29, costal fragment; SLT 32, peripheral fragment; SLT 34, peripheral, juvenile; SLT 36, peripheral; SLT 38, costal fragment; SLT 41, peripheral; SLT 42, peripheral; SLT 43, costal fragment; SLT 44, peripheral; SLT 45, plastron fragment; SLT 47, peripheral fragment; SLT 48, left hypoplastron fragment; SLT 49, left xiphiplastron; SLT 50, left xiphiplastron; SLT 51, right epiplastron; SLT 52, entoplastron; SLT 53, right hyoplastron; SLT 54, right xiphiplastron; SLT 55, right xiphiplastron; SLT 56, left costal 8; SLT 57, left epiplastron fragment; SLT 58, left hypoplastron fragment; SLT 59, right hypoplastron fragment; SLT 60, right epiplastron fragment; SLT 61, hyoplastron fragment; SLT 62, hyoplastron fragment; SLT 63, hyoplastron fragment; SLT 64, left xiphiplastron fragment; SLT 65, right hyoplastron fragment; SLT 66, right epiplastron fragment; SLT 67, left xiphiplastron fragment; SLT 68, right epiplastron; SLT 69, left hypoplastron fragment; SLT 70, left epiplastron fragment; SLT 71, left hyoplastron fragment; SLT 73, peripheral; SLT 75, hypoplastron fragment; SLT 79, hyoplastron fragment; SLT 85, nuchal fragment; SLT 86, peripheral; SLT 88, peripheral; SLT 89, peripheral; SLT 91, costal fragment; SLT 94, costal fragment; SLT 98, costal fragment; SLT 100, costal fragment; SLT 103, peripheral fragment; SLT 108, costal fragment; SLT 110, peripheral; SLT 119, costal fragment; SLT 123, suprapygal 1; SLT 127, costal fragment; SLT 134, peripheral plate, juvenile; SLT 174, costal fragment, juvenile; SLT 179, costal juvenile. Indeterminable shell fragments (SLT): 96, 109, 113, 122, 123, 128, 132, 140, 141, 145, 146, 148, 149, 150, 151, 152, 154, 155, 156, 159, 160, 161, 163, 164, 165, 166, 167, 168, 169, 170. Indeterminable carapace fragments (SLT): 25, 27, 46, 90, 92, 97, 101, 104, 106, 107, 111, 114, 115, 118, 125, 127, 138, 130, 143, 144, 158, 162, 171, 172, 177, 178. Small fragments of peripherals (SLT): 30, 31, 33, 37, 87, 95, 102, 112, 117, 120, 121, 122, 131, 139, 142, 158, 175. Indeterminable plastron fragments (SLT): 40, 72, 74, 76, 77, 78, 80, 81, 93, 99, 105, 116, 124, 126, 147. Appendicular elements (SLT, no numbers): humerus diaphysis (2); phalanx (3); cervical vertebra (3).
Testudo sp. (LGPUT collection, 38 specimens): SLT 6, costal fragment; SLT 35, peripheral fragment; SLT 111, shell fragment; SLT 180, peripheral; SLT 182, shell fragment; SLT 183, costal fragment; SLT 184, shell fragment; SLT 185, shell fragment; SLT 186, hyoplastron; SLT 187, costal fragment; SLT 188, peripheral; SLT 189, costal; SLT 190, shell fragment; SLT 191, plastron fragment; SLT 192, plastron fragment; SLT 193, peripheral; SLT 194, costal fragment; SLT 195, plastron fragment; SLT 196, peripheral; SLT 197, shell fragment; SLT 198, plastron fragment; SLT 199, plastron fragment; SLT 200, right hyoplastron fragment; SLT 201, plastron fragment; SLT 202, costal fragment; SLT 203, peripheral fragment; SLT 204, shell fragment; SLT 205, shell fragment; SLT 206, shell fragment; SLT 207, shell fragment; SLT 208, shell fragment; SLT 209, shell fragment; SLT 210, shell fragment; SLT 211, shell fragment; SLT 212, shell fragment; SLT 213, shell fragment; SLT 214, shell fragment; SLT 215, shell fragment.