Gliridae is one of the oldest extant families of rodents with a fossil record dating back as far as the early Eocene. Eogliravus wildi, previously known from some isolated teeth from the European early Eocene, represents the oldest and most primitive glirid. The early middle Eocene oil shales of Messel, Germany, yielded an extraordinarily preserved specimen of E. wildi, exhibiting the virtually complete and articulated skeleton, the outline of the pelage, and gut contents consisting of various vegetable matter. E. wildi descended most probably from European small-sized microparamyine ischyromyids. Its protrogomorphous zygomasseteric structure indicates that the myomorphy and hystricomorphy of the extant dormice evolved convergently to other rodent clades. The postcranial skeleton compares well with extant glirids. The postcranial morphology, proportions of the limb segments, bushy tail, and small body size indicate a gracile and swift animal moving among tree branches. The very large orbits suggest nocturnal activity. E. wildi was well adapted to an arboreal or scrub environment, the major habitat of today's dormice.
INTRODUCTION
Gliridae are one of the oldest extant rodent families with a fossil record dating back to the early Eocene. As currently understood, they descended in Europe from early Paleogene ischyromyids, and most probably from advanced small-sized microparamyines (Hartenberger, 1971; Escarguel, 1999). The early and middle Eocene (MP 10–13; MP = Mammals Paleogene, see Schmidt-Kittler, 1987) genus Eogliravus Hartenberger, 1971, represents the earliest and most primitive glirid taxon. Its occlusal tooth morphology links the late Eocene/Oligocene glirid genera Gliravus Stehlin and Schaub, 1951, and Bransatoglis Hugueney, 1967, with the early Eocene (MP 8–9) ischyromyid Microparamys (Sparnacomys) chandoni Hartenberger, 1971, from the Paris Basin (Hartenberger, 1971; Vianey-Liaud, 1994; Daams and de Bruijn, 1995; Escarguel, 1999). The oldest species of Eogliravus, E. wildi, is known from a few isolated teeth from several early Eocene (MP 10) French localities and now by a complete skeleton from the early middle Eocene (MP 11) of Messel, Germany. The identification of the alleged early Eocene glirid Chaibulakomys angos Shevyreva, 1992, represented by an isolated p4 from eastern Kazakhstan (Shevyreva, 1992), appears to us not absolutely clear.
Glirids reached the highest diversity in Europe and the Near East during the late early Miocene, and predominated in many local rodent faunas of that time (Daams and de Bruijn, 1995). These localities comprised the centers of diversification throughout their fossil history (Hartenberger, 1994). In diversity and abundance, glirids replaced the Oligocene theridomyids before they themselves were partly replaced by muroids during the late Miocene. Forty living and extinct glirid genera (McKenna and Bell, 1997) and about 180 species (Daams and de Bruijn, 1995) are known. Only nine genera, five of which are monospecific, exist today in Europe, Asia, and Africa. Classification beyond the family level is controversial. The extant species are referred to three (e.g., Wahlert et al., 1993; Daams and de Bruijn, 1995; McKenna and Bell, 1997) or four subgroups (e.g., Storch, 1995). Quite recently, Holden (2005) provided a detailed discussion of intraglirid relationships.
The fossil site of Messel near Darmstadt, Germany, is renowned for the variety and completeness of its early middle Eocene fauna and flora (Schaal and Ziegler, 1992; Koenigswald and Storch, 1998; Storch, 2004). Rodents from Messel are known since the middle of the last century and include the species Ailuravus macrurus Weitzel, 1949, Masillamys beegeri Tobien, 1954, and Hartenbergemmys parvus (Tobien, 1954) (Weitzel, 1949; Tobien, 1954; Escarguel, 1999). Each of these species is presently known by several complete skeletons, most of them being still undescribed. Here we describe the first glirid from Messel. It exhibits not only details of the postcranial skeleton, including the minute phalanges and baculum, but also an amazing preservation of the soft body outline, even of single hairs. In addition, gut contents are well preserved. Previously, the specimen was only announced in two popular short notes (Storch et al., 2000, Escarguel et al., 2001). Tertiary glirids are known from dental and cranial remains, with the exception of single, fairly complete skeletons of the late Eocene Glamys priscus (Stehlin and Schaub, 1951) from Montmartre, France (Cuvier, 1824; Hartenberger, 1971; Vianey-Liaud, 1994), the early Oligocene Gliravus micio (Misonne, 1957) from Bohemia, Czech Republik (Fejfar and Storch, 1994), and the late Miocene Glirulus lissiensis (Hugueney and Mein, 1965) from the Ardèche, France (Mein and Romaggi, 1991).
MATERIAL
The present specimen was transferred from the original two complementary slabs of oil shale to two slabs of epoxy resin, slab A (Figs. 1–3) and the complementary slab B (Fig. 4). The presence of a baculum and the state of tooth wear and eruption prove the specimen to be an adult male. The specimen is housed in the Wyoming Dinosaur Centre, Thermopolis, WY (collection number WDC-C-MG202). Casts and x-ray images are kept at the Forschungsinstitut Senckenberg (SMF).
Slab A
The overall length of the fossil in situ (tip of nasals to tip of last caudal vertebra) is approximately 11 cm. The articulated skeleton is viewed from the left side in a lateral to dorsolateral position. The cheek teeth are in occlusion. The accessible parts of p4-m2 were molded in situ and SEM images prepared from casts (Fig. 6C). The mesiolabial part of m3 is visible in anterior view (not figured as close-up). The right forearm is obscured for the most part by the thorax. The right hindlimb is disarticulated at the hip and knee joints. The distal extremity of the femur is obscured by the sacral region, and the shank is crossed distally by tail vertebra. The toes of the hind feet are bent back underneath the metatarsals. The outline of the woolly fur and bushy tail, even of single hairs, is well preserved. The terminal quarter of the tufted tail, however, was complemented during preparation.
Slab B
The preserved parts of the specimen are viewed from the right side. Slab B exhibits a ventrolateral view of the cranium, including the bulla tympani (Fig. 4) and cheek teeth. Of the teeth, M2 was extracted from the fossil and M3 molded in situ for SEM documentation (Fig. 6A–B). The rostral portion of the skull is missing. Preserved parts of the skeleton include the left humerus and hand, the digits of the right hand, parts of the ribcage, the pelvic and sacral regions in ventral aspect, the left femur and distal extremity of the right femur, the right tibia and fibula, and the caudal vertebral column. The fossil contains rich gut contents in the abdominal region.
SYSTEMATIC PALEONTOLOGY
Order RODENTIA Bowdich, 1821
Suborder GLIRIMORPHA Thaler, 1966
Family GLIRIDAE Thomas, 1897
Subfamily GLIRAVINAE Schaub, 1958
P3 present (evident from its preserved strong root) and zygomasseteric structure protrogomorphous (infraorbital foramen small, ventral face of the anterior zygomatic arch flattened, anterior root of the jugal arch only slightly tilted upward beneath the infraorbital foramen).
Remarks
The taxon Gliravinae comprises the genera Eogliravus and Gliravus, protrogomorphy in the latter genus being shown in the type species G. majori Stehlin and Schaub, 1951 (Schaub, 1958; Vianey-Liaud, 1989,1994). The derived character state of “myomorphy” is first recorded in the late Eocene Glamys priscus (Stehlin and Schaub, 1951), and extant dormice are either “hystricomorph” (Graphiurus) or “myomorph” (all other taxa). Both zygomasseteric configurations evolved convergently to other major rodent clades from a protrogomorphous condition and are better termed pseudomyomorphy (Vianey-Liaud, 1975, 1994; Maier et al., 2002) and pseudohystricomorphy.
Genus EOGLIRAVUS Hartenberger, 1971
Trigon of M2 asymmetrically V-shaped, metaloph subcomplete (discontinuous between metaconule and metacone), protoloph sinuous, and conules reduced (Fig. 6A). Protocone of M2 prominent and located rather mesially, weakly connected to the anteroloph, and separated from the well developed hypocone (Fig. 6A). M3 only slightly reduced in size as compared to M2 (fig. 6B). Protoconid of p4 well individualized (Fig. 6C). Mesoconid of m1-2 slightly elongate transversely and very weakly linked to the protoconid, ectolophid discontinuous, and anterolophid not or only very weakly connected to protoconid and metaconid (Fig. 6C).
EOGLIRAVUS WILDI Hartenberger, 1971
Very small-sized cheek teeth, mesolophid of the lower cheek teeth absent or incipiently developed, and lingual sinus of M2 shallow (cf. Hartenberger, 1971, Escarguel, 1999).
COMPARATIVE DESCRIPTION OF THE SKELETON
Skull
Comparisons with living genera are based on specimens of the Senckenberg collections and data from Wahlert and colleagues (1993). All genera except Chaetocauda are considered. The neurocranium of E. wildi is crushed; sutures and fracture lines are not always easy to distinguish, and some bones overlap each other. The outlines of the alisphenoid (al), parietal (pa), interparietal (in), squamosal (squ) and frontal (fr) are (partially) marked on Figure 3.
The rostral part of the skull, including the diastema, is very short (Fig. 3). The orbits are large and located far anteriorly, indicating nocturnal and/or crepuscular habits. The dorsal profile of the skull was probably slightly convex. There is no evidence of sagittal or temporal crests on the neurocranium as in extant glirids. The jugal is strong and forms most of the zygomatic arch but does not contribute to the glenoid fossa. As in Glis and Graphiurus, the jugal extends anteriorly along the mediodorsal side of the maxillary zygomatic root, and its tip touches the tiny lacrymal. The posterior edge of the maxillary root rises above P3 and above P4 in extant glirids. The infraorbital foramen is not fully accessible; nevertheless, it can be described as small. The ventral side of the maxillary root of the zygomatic arch is only very slightly tilted upward. The origin of the lateral portions of the masseter muscle is well defined (slab B) by the plane ventral side of the anterior jugal arch. The nasals extend farther posteriorly than the premaxillaries. They are of similar length in extant glirids. The impression of a strong supraorbital crest of the frontal is most likely over-accentuated because of the deeply collapsed orbital wall. A strengthened supraorbital edge and a concave interorbital region occurs in Myomimus. The alleged small postorbital process of E. wildi (Fig. 3) is caused by the broken and displaced posterior end of the supraorbital crest. The alisphenoid (al) expands dorsally and separates the frontal from the squamosal (squ). In extant glirids there is a rather wide dorsal contact zone between frontal and squamosal. The interparietal (in) is large and has broad contact with the squamosals (squ). A wide interparietal is characteristic of extant glirids, in which the interparietal contacts or almost contacts the squamosal; the contact zone is usually narrow with the exception of Glis. Small paroccipital processes appear to be present. The osseous auditory bulla is of moderate size and fully ossified to the skull (Fig. 4).
The masseteric fossa of the mandible is large and deep. It is less excavated in extant glirids. Its anterior end is below the m1/m2 boundary while it is below the p4/m1 junction in extant glirids. The coronoid process of the ascending ramus is strong and rises above the condyle, which is ovoid. The angular process is slightly twisted (but not deflected medially or laterally) and its straight posteroventral border is thickened. This configuration is similar to the condition in extant glirids. The angular process is not fenestrated as in the extant genera Glis and Glirulus. It shows a fenestration in the other living genera. The mental foramen is located beneath the anterior root of p4.
Postcranials
Comparisons with extant genera are based on skeletons of Glis, Eliomys, Dryomys, and Muscardinus from the Senckenberg collections.
Vertebral Column
The sacrum consists of three fused vertebrae. The anterior sacral vertebra and the anterior extremity of the second form the sacroiliac joint via their transverse processes as in extant taxa. The tail consists of 16 vertebrae; the four proximal ones show spines and transverse processes. Extant taxa have over 20 vertebrae and stronger apophyses. Few other details of the vertebral morphology can be seen.
Forelimb
The humerus has an elevated deltoid crest which ends slightly proximal to midshaft. The supinator ridge is moderately developed (Fig. 2). The medially projecting, rather slender entepicondyle shows an entepicondylar foramen. The shafts of ulna and radius curve slightly ventrally while the short olecranon process of the ulna curves conspicuously dorsally. The lateral ulnar fossa is well excavated (Fig. 2). The radius appears to be laterally compressed over its middle part. The thumb is reduced, its accessible metacarpal is minute. The metacarpalia II–V are slightly dorsoventrally compressed and bear on their distal part two tubercles, one on the ventrolateral and one on the ventromedial edge. The terminal phalanges are high, laterally compressed, strongly curved, and sharp. The flexor tubercles are deep. Features of the forelimb are similar to those of extant species.
Hindlimb
The ilium of the pelvis is straight, long, and slender. Its cranial end is only slightly deflected laterally, similar to Muscardinus and Dryomys. The iliac spine in front of the acetabulum is prominent. Femur, tibia, and fibula are long, slender, and straight. The head and greater trochanter of the femur are of subequal height, the trochanteric fossa and the lesser trochanter are of moderate size, the small third trochanter is located rather proximally on the shaft (Fig. 5, left hind limb), and the patellar groove is large. Tibia and fibula are not synostosed distally but tight-fitting over about one fifth of the length of the tibia (Fig. 5, right hind limb). The distal synovial tibiofibular joint is a plesiomorphic character; the early Oligocene Gliravus micio has a very short distal synostosed fusion (Fejfar and Storch, 1994), and living glirids have a fusion for about one-fourth to slightly more than one-third of the tibial length. The tibial crest is moderately high. The shaft of the fibula is gracile and of even width. Both calcanei are exposed in plantar view (Fig. 5). The tuber calcanei is well developed and its distal extremity is slightly expanded and concave. The peroneal tubercle and the sustentaculum protrude in an opposite position. The morphology of the metatarsalia II–V and terminal phalanges is similar to the corresponding elements of the hand. Except for the non-synostosed tibia and fibula, the hind limb agrees morphologically with living glirids.
Baculum
It is exposed ventrally to the ischium (Fig. 5). It is a simple, rod-like bone with a sigmoid apex, not unlike the penis bone in most extant dormice (Storch, 1978:fig. 46).
CONCLUSIONS
Locomotor Behavior
It is generally concluded from the scansorial habits of the majority of the extant dormice that fossil glirids were either forest dwellers and indicative of closed habitats or, to a lesser extent, rock dwellers and pointing to more open habitats.
The Messel specimen strongly suggests skillful climbing capabilities. In addition to the long and bushy tail, a combination of postcranial characters indicates climbing specializations. Morphologically, these features include (1) the short and distinctly anteriorly curved olecranon process of the ulna, (2) the unreduced metatarsals I and V, (3) the dorsally convex first phalanges of hands and feet being equipped with ventrolateral and ventromedial flexor tubercles, and (4) the high, strongly curved, laterally compressed, and sharp terminal phalanges of hands and feet showing strong flexor tubercles.
Limb proportions also corroborate climbing habits. Tibia and fibula are long and straight. The crural index (tibia : femur × 100) of E. wildi is high (118) and compares generally with living arboreal rodents (e.g., mean of 117 in the red squirrel, Sciurus vulgaris). Among extant glirids, the index value of E. wildi lies between the data of the arboreal species Dryomys nitedula (114–116) and Muscardinus avellanarius (120–125) (Fejfar and Storch, 1994). The intermembral index (humerus + radius : femur + tibia × 100) is also high (75) and takes a similar position among arboreal species as the crural index and again lies between D. nitedula (70–71) and M. avellanarius (77–79) (Fejfar and Storch, 1994).
The first and second phalanges of digits II–V of hands and feet are relatively long and suitable for climbing because they enhance grasping abilities of the clawed digits on twigs and allow for a greater circumference of broader vertical supports. The phalangeal index of the 5th tarsal digit of E. wildi (proximal + intermedial phalanx as a percentage of the metatarsal) is high (94) and compares well with values for arboreal rodent species (specimens from Senckenberg collection).
Feeding
Preserved gut contents comprise vegetable matter only, consisting of food items such as lumps of hard sclerenchymatic tissue from seeds (Fig. 7) and soft plant tissue (Fig. 8). Apparently, seeds, fruits and buds formed major components of the diet of the present E. wildi specimen, food items which are preferred by living species such as the edible dormouse (Glis glis) and the hazel dormouse (Muscardinus avellanarius) (Storch, 1978). The gut contents are not necessarily indicative especially for woodland but fit such a paleoenvironment.
The postcranial morphology, the proportions of the skeleton and limb segments, and the small body size of E. wildi imply in combination with the “paratropical” former Messel rain forest, a gracile and swift animal moving among branches. The large orbits indicate a nocturnal or crepuscular life style. This very early member of the family Gliridae appears thus well adapted to an arboreal or scrub environment, a major habitat of today's dormice.
Acknowledgments
We are grateful to B. Pohl for access to the specimen. We thank Zhe-Xi Luo and two anonymous reviewers for providing helpful comments on an earlier version of this paper. We extend our gratitude to G. Escarguel for discussions of tooth morphology of early Paleogene rodents. K. Krohmann and S. Tränkner provided photographic assistance.