PaléoEnvironnements et PaléobioSphères UMR CNRS 5125, Université Claude Bernard Lyon 1, Bâtiment GEODE, Campus de la Doua, F-69622 Villeurbanne cédex, France; e-mail: email@example.com
Abstract: The new species Acynodon adriaticus is described on the basis of remains from the Santonian–Campanian of Villaggio del Pescatore (Trieste, NE Italy). This species differs in several cranial features from Acynodon iberoccitanus, the only other Acynodon species whose cranial osteology is known in detail. The absence of maxillary and dentary caniniform teeth coupled with the presence of enlarged molariform teeth suggests that Acynodon probably fed on slowly moving hard-shelled prey. Moreover, the new materials reveal for the first time the morphology of some postcranial elements of Acynodon: in particular, medial-most paravertebral osteoderms that are characterized by two keels. A new cladistic phylogenetic analysis resolves the previously reported polytomy among the basal Globidonta: Acynodon is recognized as the most primitive globidontan. This genus may represent the geologically oldest known globidontan. The fact that Acynodon has been found only in Europe and that the outgroup of Globidonta, the Diplocynodontinae, is mainly known from Europe, suggests that globidontans may have originated in Europe and not in North America as previously supposed.
In 1997, Buscalioni et al. erected the new taxon Acynodon iberoccitanus on the basis of an isolated maxilla from the late Campanian or early Maastrichtian of Laño (Spain; Text-fig. 1). That species is characterized, among other features, by isodont maxillary teeth (absence of caniniform maxillary teeth), with the exception of a few molariform teeth placed at the back of the tooth row. A few other cranial skeletal elements from Laño were referred to A. iberoccitanus. Buscalioni et al. (1997) also referred some isolated teeth from the Maastrichtian of Quintanilla del Coco, Spain, to a second species, Acynodon lopezi. A complete description of the cranial anatomy of A. iberoccitanus was provided recently by Martin (2007) on the basis of relatively well-preserved remains from the late Campanian–early Maastrichtian of France at Massecaps (Martin and Buffetaut 2005), Quarante and Fox-Amphoux (figured by Vasse 1993; and mentioned as a member of Acynodon by Buscalioni et al. 1997).
Isolated teeth from the late Maastrichtian of the Spanish Pyrenees were explicitly referred to Acynodon (López-Martínez et al. 2001), whereas others from the approximately coeval site of Fântânele in Romania were tentatively referred to cf. Acynodon sp. (Martin et al. 2006).
Morphologically similar teeth are not rare in the European Late Cretaceous fossil record, but, in most cases are conservatively identified only to the rank of order (Debeljak et al. 2002). All the information so far available for Acynodon comes from the above-mentioned localities and concerns the cranial osteology only; no postcranial elements have been found in clear association with the cranial material so far described.
Since its original description, the phylogenetic relationships of Acynodon have been a topic of discussion. Buscalioni et al. (1997, 1999) concluded that Acynodon is closely related to Albertochampsa, Brachychampsa and Stangerochampsa, the North American short-snouted forms that they regarded as members of the Alligatoridae. Later, Brochu (1999, 2001a, b, 2003) considered Acynodon as representing an early member of the first alligatorine radiation in Europe. The phylogenetic analysis by Martin (2005, 2007) proposed, for the first time on the basis of a character matrix, that Acynodon is a basal globidontan alligatoroid.
We describe cranial and postcranial crocodylian remains from the Upper Cretaceous of Villaggio del Pescatore, north-eastern Italy. Buffetaut et al. (2001) and Dalla Vecchia et al. (2005) had previously mentioned or briefly described these remains as putative alligatoroids, but Delfino and Buffetaut (2006) recognized them as referable to the genus Acynodon. Although the bones are not completely isolated from the embedding rocky matrix (in order to keep all the skeletal elements in connection), the Italian materials provide a considerable amount of new information on the morphology of Acynodon, which allows a reconsideration of the interrelationships of the genus among basal alligatoroids.
The specimens described here, as well as all other crocodylian remains from Villaggio del Pescatore, were completely preserved in blocks of laminated carbonates. The skeletal materials have been removed partly by acid preparation, with the result that the specimens are now exposed on slabs. Therefore, the skeletal elements of the larger specimen described here are partly embedded in the rocky matrix: the entire dorsal and left lateral surfaces, and part of the ventral surface of the skull and lower jaws are visible (Text-fig. 2). The miniatures with cranial and postcranial elements, shown in some of the figures, serve to illustrate the general relationships of the skeletal elements at an earlier stage of preparation.
Derivation of name. From ‘sinus adriaticus’, the Latin name for the Adriatic Sea, on the shore of which the site of Villaggio del Pescatore is located.
Type specimen. MCSNT 57248 comprises a complete skull with lower jaws, part of the hyoid apparatus and pectoral girdle, right forelimb, the first 10 vertebrae, some ribs, and several dorsal osteoderms.
Paratype. MCSNT 57032 consists of associated osteoderms and ribs.
Type locality and age. Villaggio del Pescatore, Duino-Aurisina municipality, Trieste Province, Friuli Venezia Giulia, Italy (45°46′43′′N, 13°35′19′′S). The fossiliferous laminated carbonates are Late Cretaceous in age, either late Santonian or Santonian–Campanian (see Geological Setting section).
Diagnosis. Acynodon adriaticus differs from A. iberoccitanus and A. lopezi in the following set of characters: absence of interorbital ridge; postorbital bar extremely short and blunt with postorbital and jugal almost touching; presence of a sagittal ridge on the parietal (slightly extending to the interorbital area); dorsal surface of squamosal smooth; frontal and parietal have a concave dorsal surface; rim of the supratemporal fenestra thickened; palatines slender and sagittally grooved; palatopterygoidal suture far from the caudal angle of the suborbital fenestra; medial quadrate condyle hooked; very elongate suborbital fenestra extending anteriorly to the level of the fifth or sixth maxillary alveolus and posteriorly to the level of the largest maxillary alveolus (the penultimate alveolus); maxillary edge downturned and arched the level of the largest maxillary alveolus; posterior molariform teeth smooth, without mesiodistal apical crests horseshoe-like or simply tall.
Remarks. Due to the fact that the retroarticular region is not preserved in any known specimen of A. iberoccitanus, it is not possible to consider the presence of an anteriorly directed spine on the medial margin of the retroarticular process of MCSNT 57248 as an autapomorphy of A. adriaticus. However, the peculiar morphology of the medial margin of the quadrate condyle that characterizes the latter is likely linked to the shape of the retroarticular process and suggests that the spine may not have been present in A. iberoccitanus.
A minor diagnostic character, which could be partly influenced by the deformation of the skull, is represented by the rather acute lateral profile of the snout of A. adriaticus when compared in dorsal view with the more rounded shape of A. iberoccitanus (compare Text-fig. 2 with Martin 2007, figs 1 and 3).
Finally, the paravertebral osteoderms of A. adriaticus exhibit two keels. The number of keels, however, is unknown in the other species of Acynodon. Thus, the presence of two keels is an ambiguous diagnostic character because it may diagnose the genus Acynodon or a group within the genus. Only the recovery and description of postcranial remains referable to A. iberoccitanus and A. lopezi will help to resolve this uncertainty.
The skeletal elements are preserved in anatomical connection; the neck vertebrae form a marked bend of nearly 90 degrees (as in another crocodylian specimen, MCSNT 57031). The skull is partly deformed and several sutures are not visible, but the general morphology is well-preserved. The major effect of the deformation seems to be the shortening of the right side of the skull. Moreover the lateral side of the right maxilla and jugal are slightly ‘verticalized’ due to lateromedial compression. The skull is 15.5 cm long (as measured from the tip of the snout to the level of the posterior edge of quadrate condyles), c. 12.5 wide (as measured across the quadratojugals), approximately triangular in shape, and characterized by a distinctly short rostrum (the area anterior to the orbits is only 6.5 cm long, approximately as long as the skull table). The quadrate condyles do not project from the lateral profile of the skull, which reaches its maximum width at the level of the quadratojugals, producing a bulging aspect. The relatively acute shape of the snout is minimally affected by the deformation as indicated by its symmetry. The outline of the skull cannot be described as festooned either in dorsal or lateral view. Ridges or bosses are not developed on the snout or at the anterior corner of the orbits. The interorbital ridge (the ‘spectacle’) is absent. The skull table has almost parallel lateral margins in dorsal view, with well developed and pointed squamosal prongs (at least the left one that is not broken and not significantly deformed); in posterior view, it is only slightly concave. A weak but distinct sagittal ridge is present on the skull table: it is more developed posteriorly to the supratemporal fenestrae but it is visible anteriorly at least up to the interorbital region. The skull table is concave between the supratemporal fenestrae (some degree of deformation cannot be excluded, but the symmetry of this concavity suggests that it is natural). With the exception of the area posterior to the infratemporal fenestrae, the external surface of the skull is ornamented with irregular ridges delimiting furrows and pits. The lateral and ventral surfaces of the lower jaw shows the same pattern.
Cranial fenestrae and openings. The external naris is small and irregularly rectangular in shape; its rim is modestly ridged but the posterior sector, delimited by the nasals, is slightly collapsed ventrally as compared to the surrounding premaxillae. It is not clear whether the naris had a median septum or not. Although dorsoventral compression may have modified the general shape of the snout, it seems likely that the anterior margin of the naris is placed ventrally to the posterior margin and that therefore the naris is anterodorsally oriented. The incisive foramen, partly hidden by the dentary, is moderately large and is situated far from the premaxillary tooth row. The orbits are proportionally small, and are asymmetric due to the deformation; their rim is nearly flush with the skull surface because a modest upturning of the rim is visible in the lateral sector of the left orbit, close to the postorbital bar. The supratemporal fenestrae are almond-shaped and small relative to the skull table. The posttemporal fenestrae are small and approximately horizontal. The foramen magnum and occipital condyle are slightly shifted to the left due to compression; both are small. The condyle, not visible in dorsal view and with an evident sagittal furrow, is c. 9 mm wide and 5 mm tall. The infratemporal fenestrae are very small (each is less than half the size of the orbit) but, due to the fact that the posterior margin is slightly eroded, their original size was even smaller. The quadratojugals probably formed the posterior angle of the infratemporal fenestrae. The suborbital fenestrae are fairly small as compared with a modern alligatorid, but relatively large as compared with A. iberoccitanus. Their well-preserved, notched, posterior rims reach the level of the midpoint of the largest maxillary alveolus. The anterior rim cannot be seen with confidence, but it seems that it reached the level of the fifth or sixth maxillary alveolus. The secondary choana is small, subcircular (but slightly distorted by compression), and does not have a neck or a septum. It is located in the posterior area of the palate, c. 6 mm from the posterior edge. The choana opens in the posterior sector of the pterygoids. Two faint lineations departing in posterolateral direction from the choana do not seem to be fractures or sutures between skeletal elements.
In the lower jaw, the perfectly preserved left lateral side clearly indicates that the external mandibular fenestra is not present.
Due to the fact that only few sutures are visible, a systematic description of the single skeletal elements of the skull and the lower jaw cannot be provided. The few obvious sutures include the premaxillary-maxillary sutures, which are particularly well visible on the lateral edge of the skull; the sutures delimiting the left nasal, the lacrimal, and the prefrontal; the sutures of the parietal, including the sutures between the parietal and the right squamosal; and the sutures between the palatine, the ectopterygoid and the pterygoid. A line, approximately sagittal in alignment and located anterior to the choana, could represent part of the suture between the pterygoids.
Although most of the cranial sutures are indeterminate, many aspects of the morphology of these elements of the skull can be described. The premaxilla is wider than long. In dorsal view, the maxilla is not laterally enlarged and its lateral margins are not parallel. The lateral contact between the premaxilla and the maxilla is smooth (i.e. there is no evidence of a notch). The bone is flat and its lateral margin is almost vertical. The dorsal suture between the premaxilla and the maxilla appears to be roughly serrated. The nasals slightly enter the naris and are particularly wide anteriorly. The posterior ends of the nasals are separated by the frontal. The lacrimal is slightly elevated in front of the orbits. It forms the anterolateral edge of the orbits. The sutural contacts with other bones are partially obscured by pitting but the lacrimal seems to project farther cranially than the prefrontal. The prefrontal is very small and forms the anteromedial corner of the orbit. Only its anterior-most triangular portion contacts the nasal. The prefrontal seems to be excluded from contact with the maxilla but a clear organization is difficult to determine. No interorbital ridge is present at this level. The frontal is very wide between the orbits. Its dorsal surface is deeply concave. Farther rostrally, it is constricted between the prefrontal and finally widens before separating the posterior processes of the nasals. The frontal forms most of the medial margins of the orbits. The postorbital bar is rather massive in proportion to the overall size of the skull. The postorbital and the jugal are quite close to each other, the infratemporal fenestra being reduced to a width of 7 mm. The frontal prevents the parietal and the postorbital from meeting on the skull table. They do not meet either on the dorsal wall of the supratemporal fossa. On its dorsal surface, the parietal is medially slightly depressed and possesses a median keel. The right squamosoparietal suture is straight and occurs at the level of the notched posterior margin of the skull table. The dorsal and ventral margins of the squamosal groove are nearly parallel (this character is visible on the left side of the skull). The presence, the position, and the development of the quadratojugal spine cannot be assessed due to the erosion of the posterior rim of the infratemporal fenestrae.
The quadrate rami are particularly short and slightly sloping ventrally. The left quadrate condyle shows a probable minor anomaly because a vertical crest links the squamosal prong to the medial hemicondyle. In both cases it is not possible to assess the position of the foramen aëreum. The quadrate condyles are c. 23 mm wide, placed at the same level as the occipital condyle and characterized by the presence at their medial margin of a distinct hook-shaped spine (Text-fig. 4). The medial hemicondyle is smaller than the lateral one. The palatines are fairly narrow, with the result that suborbital fenestrae are separated by a narrow space. A median groove is present along the paired palatine rod. The sutures between the palatines and the maxillae are not visible but, by interpreting two symmetrical fractures present in this area, it can be tentatively suggested that the sutures are positioned anterior to the anterior rim of the suborbital fenestrae. At the level of the second molariform tooth, the lateral surface of the maxilla is turned down medially; the lateral and the downturned surfaces form a c. 90 degree angle. This robust and thickened area corresponds in lateral view to a distinct arch (Text-fig. 5C). The pterygoids are very elongated craniocaudally in proportion to the overall size of the skull. The anterior portions contact the palatines and participate in the palatal rod between the suborbital fenestrae. The pterygoid wings are developed in a nearly vertical plane (their ventral surface is markedly concave) and are therefore only slightly diverging laterally. The posterior wings of the pterygoids extend posteriorly almost to the level of the quadrate condyles. It is not clear how much the ectopterygoids participated in the pterygoid wings because the sutures are not clearly visible.
The anterior part of the left lower jaw is hidden by the skull, but the posterior part is nearly completely free laterally and posteriorly. It seems to be rather thin anteriorly but in the posterior two-thirds its height grows progressively, reaching a considerable size. The lower jaw is also fairly thick and with a maximum labiolingual thickness in the region corresponding to the enlarged maxillary teeth. The ventral surface of the lower jaw is distinctly flattened, but this may be partly due to lithostatic compression. The dentary symphysis is relatively broad, c. 35 mm long, and, if the disposition of the dentary teeth is similar to that of the upper teeth, this region of the dentary could accommodate seven teeth. The suture between the dentary and the splenial is well preserved and it is apparent that the splenial reaches the symphysis, probably without entering it, and that the anterior tip of the bone is ventral to the Meckelian canal. Caudally, the medial splenial wall becomes progressively thicker. It becomes thickest at the level of the largest maxillary alveolus resulting in a constriction of the space between the mandibular rami. The surangular reaches the dorsal edge of the lateral wall of the glenoid fossa and reaches the posterior tip of the retroarticular process, although it does not reach its dorsal edge. Although both retroarticular processes are posteriorly directed, rather than posterodorsally as in the majority of living crocodylians, it is not clear whether lithostatic deformation could be entirely responsible for such a marked difference. The retroarticular processes are characterized by being relatively short and medially expanded (Text-fig. 4). The right process exhibits a convex medial edge (the left one is not complete) that develops an evident anteriorly directed spine, placed at the anterior sector of the medial edge, close to the border with the glenoid cavity. Foramina aëra are not clearly visible, but the presence on the left articular of two foramina corresponding to a small dorsally directed process and a similar process developed close to a small depression on the right articular, may indicate that such foramina are slightly inset from the medial edge of the bone.
Both the dorsal projections of the hyoid are present and underlie both pterygoids. They are distinctly flat and do not appear to flare apically (Text-fig. 3).
Dentition and occlusion pattern
Because the jaws are preserved in occlusion and embedded in the matrix on their right side, only the five premaxillary teeth (on both sides), the first four teeth on the right maxilla, and nearly all those of the left one are visible. On the left maxilla, a small (c. 12 mm long) collapsed area hides the underlying teeth; it seems likely that this area corresponds to three teeth, which would bring the total number of maxillary teeth to 13. Teeth are present in all the visible tooth positions. The premaxillary and the first 10 maxillary teeth are not significantly different in shape and size, but the last three maxillary teeth are particularly large and molariform in appearance. On both sides, the fifth premaxillary tooth is slightly larger than the first few maxillary teeth (Text-fig. 5B). The tenth maxillary tooth (preceding the molariform ones) is somewhat larger than the preceding ones and slightly bulbous. The anterior teeth are labiolingually flattened and have a well-marked mesiodistal ridge, a weakly pointed apex, and a crown 4 mm high and wide (at the base). Teeth showing signs of wear tend to have a flattened apex, which is relatively narrow but larger than the mesiodistal keels. The absence of a notable constriction at the base of the crown gives a peg-like appearance to these teeth in labial view, but a hint of constriction is visible on their lingual side. The fourth and fifth maxillary teeth are approximately similar in size to those preceding and following (Text-fig. 5B). The three posterior-most, molariform teeth are not completely visible because they are partly hidden by the lower jaw. The first of these three teeth has a mesiodistal length of c. 7 mm, the second one is by far the largest, being 18 mm long, whereas the third, partly visible tooth seems to be of intermediate size. The height of these teeth is not measurable because of jaw occlusion and the presence of residual matrix, but it seems clear that at least the largest tooth has no mesiodistal keel and that it is longer that tall. The labial surface of this tooth is weakly concave and devoid of ridges or grooves. The surfaces of the premaxillary and non-molariform maxillary teeth exhibit ornamentation of small, linearly arranged granules, whereas the surfaces of the molariform teeth are finely ornamented with numerous small pustules that are not arranged in lines. A small occlusal pit is visible between the first and second alveoli of each premaxilla.
None of the dentary teeth are visible.
The pectoral girdle is represented by the right scapula and both coracoids. The distal edge of the left scapula may be exposed close to the right scapula. The entire lateral surface of the scapula is exposed. The dorsal edge of the scapular blade does not flare significantly. The deltoid crest is relatively thin. The scapulocoracoid synchondrosis is not closed. The only visible regions of both coracoids are the proximal ends (note that the articular area of the left one is visible on the inferior surface of the slab and therefore not in Text-fig. 2).
The right forelimb is nearly complete whereas the left forelimb appears to be represented only by a fragment of humerus. The right humerus is exposed in posterior aspect (Text-fig. 6). It lacks a part of the diaphysis, therefore the total length can be only estimated (c. 75 mm). The ulna is 55 mm long, and has a bent and swollen diaphysis that may indicate a healed fracture. The radius, partly hidden by the ulna and the distal end of the humerus, is c. 50 mm long. The organization of the elements of the manus is not clear due to the unfinished preparation. The radial and ulnar carpals are followed by a few metacarpals and phalanges. One digit (presumably the third) is completely exposed and consists of four phalanges, including the ungual phalanx. It is likely that phalanges of other digits are hidden under the surface of the matrix.
To the left of the right forelimb are four skeletal elements that are flattened and elongated. It is unclear whether they are all ribs or if a mixture of ribs and osteoderms exposed in internal aspect. Along the edge of the fissure separating the anterior (bearing the head) and the posterior portion of the slab, an elongated bone may be the left humerus.
The cervical vertebrae and a fragment of the first dorsal vertebra lie on the slab immediately posterior to the skull. They are exposed in right lateral aspect. The left sides of some of the vertebrae can be partially seen through a hole in the slab. Vertebrae are in articulation anteriorly, but become progressively disarticulated posteriorly, but maintain close associations with neighbouring elements. The fifth cervical vertebra is slightly separated from the following one and partly shows a condyle, indicating a procoelous condition. The neurocentral sutures of all the vertebrae are closed. Although the morphology of the anterior vertebrae is not clear because they are still partly obscured by matrix, the first two vertebrae are tentatively considered to be the atlas-axis complex. The atlas could be represented by the neural arch whereas the axis is represented by a wide fractured neural spine and by the right postzygapophysis. The length of the centra cannot be determined for most cervicals, but for the sixth vertebra we estimate a length of about 17 mm (the length of the centrum without the condyle is 14 mm). The neural spines of the following cervical vertebrae are tall and narrow, although not all are preserved; the neural spine of the eighth vertebra is 22 mm long. The ninth presacral vertebra, partly obscured by matrix, is the first dorsal element. Its neural spine is slightly anteroposteriorly longer than the previous one. No other vertebrae are visible.
Right ribs of the fifth, sixth and seventh cervical vertebrae are present in MCSNT 57248. They are characterized by a short shaft and short capitular and tubercular processes. Small portions of the ribs are also visible on the inferior surface of the slab. In MCSNT 57032, the distal extremities of three ribs emerge from below the osteoderms whereas four others are placed beside them; the surfaces of the latter group of ribs are damaged. Close to the set of four ribs are what appear to be several ventral ribs. The latter are small, thin, elongated, and poorly preserved structures. Another rib is visible along the section.
At least seven osteoderms are scattered in the neck region of MCSNT 57248. Some of them expose the visceral surface, but others show an external surface with a sagittal keel surrounded by several small pits. The largest osteoderm is c.16 mm long and 12 mm wide.
Much more informative are the osteoderms of MCSNT 57032 because several paravertebral elements are arranged in approximately their original anatomical position. Twenty-five osteoderms or osteoderm fragments are present (20 are present on the surface whereas five are placed below them). As shown in the reconstruction of Text-fig. 7, it is possible to identify at least four parasagittal rows (plus a probable fifth row represented by a damaged accessory osteoderm). The right medial-most parasagittal row is in a vertical position and its elements are not complete. All the osteoderms of the left medial-most parasagittal row bear two sagittally aligned keels: the medial keel is straight whereas the lateral one can be relatively straight, as in the osteoderm of the first preserved, transversal row, or laterally concave. We infer that the vertically tilted right medial-most parasagittal row had osteoderms with two keels because a laterally concave keel is regularly present in the preserved portion of the osteoderms, and the first element of the row preserves a partial keel placed medially to the concave one. The paravertebral osteoderms in the adjacent parasagittal rows have only one keel that is approximately straight, not sagittally aligned but diverging laterally in caudal direction. The external surface of all the osteoderms is characterized, next to the keels, by small pits separated by tubercles, and are sometimes united in small furrows separated by weak ridges. An anterior smooth articular surface, as wide as the osteoderm, is regularly present. The general appearance is not significantly different from that of the osteoderms of MCSNT 57248, even though the latter have a different shape, are smaller in size (they were more cranially placed), and have more pits. The osteoderms with two keels are rectangular to trapezoidal in shape; the best-preserved element is 23 mm wide and 22 mm long. The osteoderms that bear a single keel are approximately rectangular; the best-preserved element is 16 mm wide and 22 mm long.
Some unidentified skeletal elements are visible on the lateral sections of both type and the paratype.
Taxonomic remarks and phylogenetic relationships
The referral of MCSNT 57248 from Villaggio del Pescatore to the genus Acynodon is fully supported by several characters. Of particular relevance is the isodont condition of the anterior sector of the tooth row, chiefly the absence of an enlarged fourth and/or fifth maxillary tooth. A fairly large set of characters (see Diagnosis) permits the referral of this specimen to a new species, here named Acynodon adriaticus. Due to a similarity in the ornamentation, an isolated group of osteoderms (MCSNT 57032) is referred to the same taxon. Pending a full preparation, other crocodylian remains from this locality (a second specimen with a partly preserved skull and some postcranial elements, MCSNT 57031, and an isolated rib, MCSNT 57245) previously referred to Acynodon by Delfino and Buffetaut (2006) are here conservatively considered as Crocodylia indet. because the are not sufficiently diagnostic and preserve osteoderms that exhibit a morphology that may or may not belong to a different sector of the body of the same taxon.
We examined the phylogenetic relationships of A. adriaticus and A. iberoccitanus using the dataset of Salisbury et al. (2006), which includes 45 ingroup taxa and two outgroups that were scored for 176 discrete morphological characters (see Salisbury et al. 2006 for taxa and character list, data matrix, and analysis protocol). Following a personal communication from S. Salisbury, the state of character 69 of Isisfordia has been changed from 1 to 0. For both Acynodon species, the characters related to the morphology of the bones encircling the external mandibular fenestra (64, 65) have been scored as indeterminate, due to the absence of the fenestra. The matrix [see supplemental data files on the Palaeontological Association website (http://palass.org)] was processed with PAUP 4.0b10* (Swofford 2001).
Parameters for the topology summarized in Text-fig. 8 are the following: number of trees retained = 18; tree length = 507; consistency index (CI) = 0.45; homoplasy index (HI) = 0.55; CI excluding uninformative characters = 0.43; HI excluding uninformative characters = 0.57; retention index (RI) = 0.75; rescaled consistency index (RC) = 0.34.
In general terms, the strict consensus topology for alligatoroids is congruent with those of Brochu (2003, 2004a) and Martin (2007). The new analysis of the phylogenetic relationships of Acynodon, inclusive of A. adriaticus and A. iberoccitanus, clarifies the position of the genus, which was previously placed in a polytomy with Albertochampsa, Brachychampsa, and Stangerochampsa at the base of Globidonta (Martin 2007, fig. 8).
Monophyly of Acynodon is supported by the following set of characters: 22-1: sides of scapular blade subparallel; 36-1: dorsal osteoderms segmented sagittally into rectangular paravertebral osteoderms and square to round accessory osteoderms; 41-1: splenial lacks rostral perforation for mandibular ramus of cranial nerve V; 43-1: splenial excluded from the mandibular symphysis and splenial rostral tip passing ventral to the Meckelian groove; 46-2: coronoid obliterates the medial intermandibular foramen at maturity; 50-0: retroarticular process projects caudally; 51-0: surangular extends to the caudal end of the retroarticular process; 62-0: external mandibular fenestra absent; 70-0: postorbital bar massive; 89-3: (anterior) maxillary alveoli homodont; 90-0: lateral edges of palatines parallel caudally; 98-1: caudomedial processes of pterygoid are small and project caudoventrally; 105-1: lateral edge of suborbital fenestra bowed medially; 112-0: quadrate with small, ventrally-reflected medial hemicondyle; 113-0: basisphenoid expanded rostrocaudally ventral to the basioccipital; 116-1: pterygoidoectopterygoidal flexure remains throughout ontogeny; 119-0: basisphenoid not broadly exposed ventral to the basioccipital at maturity in occipital aspect and pterygoid dorsoventrally short ventral to median eustachian opening; 147-1: lateral eustachian canals open medial to eustachian canal; 165-1: caudal maxillary alveoli mediolaterally compressed.
Acynodon adriaticus is characterized by the following autapomorphies: 79-0: naris projects rostrodorsally; 85-1: palatopterygoidal suture far from caudal angle of the suborbital fenestra.
Therefore, even if the character list used for this analysis permits a distinction between A. adriaticus and A. iberoccitanus, the distinction is actually based on two characters only, one of which (the orientation of the naris) could possibly be influenced by preservation. Several additional characters should be defined in order to better characterize A. adriaticus on a cladistic basis.
In A. adriaticus, the characteristic morphology of the osteoderms, unknown in A. iberoccitanus and A. lopezi, is of particular relevance (Text-fig. 7). MCSNT 57032 indicates that in A. adriaticus the two medial-most osteoderms from each transverse row of the paravertebral shield had a medial straight keel and a lateral concave keel; these double-keeled osteoderms are laterally flanked by at least one (there is possible evidence of additional accessory osteoderms) osteoderm with a slightly oblique keel, pointing posteriorly in lateral direction. It is worth noting that double keeled osteoderms are not exclusively present in A. adriaticus but have been described at least for two non directly related taxa: Bernissartiafagesii (Dollo 1883; Buffetaut 1975, pl. 4; Norell and Clark 1990) and Pristichampsus rollinati (Rossmann 2000, fig. 20j–k). However, at least Bernissartia has a biserial paravertebral shield (see also Salisbury and Frey 2001, figs 27–29; Salisbury et al. 2006, fig. 5).
In order to test the phylogenetic relationships of Acynodon with a slightly different character matrix, another analysis has been run adding B. fagesii and Hylaeochampsa vectiana as outgroups to the data matrix of Piras and Buscalioni (2006). This matrix is actually a slightly modified version (same character list but with some differences in the scorings of some characters) of the one by Brochu (2004a), used by Martin (2007) for a first analysis of the relationships of A. iberoccitanus. The position of Acynodon as the basal-most globidontan is confirmed, but a series of polytomies result within the clade of alligatorids while the relationships among the Diplocynodon species are well resolved [see supplemental data files on the Palaeontological Association website (http://palass.org)].
The snout and tooth morphology of Acynodon is so peculiar and remote from modern crocodylian analogs that its diet is difficult to determine. All the attempts to discuss a possible diet for this genus concentrated on the presence of the large posterior globular teeth (see Martin 2007; and literature therein), but the absence of caniniform teeth, a unique character among short-snouted crocodylians, is probably the key element for the proper assessment of the diet of Acynodon. Short-snouted crocodylians without maxillary caniniform teeth are very rare in the fossil record. In this respect, Acynodon resembles Iharkutosuchus makadii?si et al. 2007; a basal hylaeochampsid eusuchian from the Upper Cretaceous of Hungary whose extremely heterodont dentition shows posterior multicusped molariform teeth. Acynodon and Iharkutosuchus share an extreme brevity of the snout, the presence of posterior molariform teeth in the maxilla (the penultimate is by far the biggest in both the cases), the shortness of the quadrates, the absence of external mandibular fenestra and the posteriorly tall and massive lower jaw. However, conspicuous differences are represented by the absence in Iharkutosuchus of supratemporal fenestrae and the presence of multicusped teeth, both reflecting different biomechanical solutions for food processing and, probably, food type (which has been suggested to consist of very fibrous plant structures for Iharkutosuchus; ?si et al. 2007). Several other extinct crocodylians developed large molariform posterior teeth. It is the case, among others, of Cretaceous taxa such as Bernissartia, Brachychampsa, Albertochampsa, and Stangerochampsa (the latter two may possibly be synonyms), but all retain enlarged caniniform teeth in the anterior region of the maxilla and the dentary (Erickson 1972; Buffetaut and Ford 1979; Norell et al. 1994; Wu et al. 1996; Brochu 1999). All extant short-snouted crocodylians bear a caniniform tooth in the maxilla (usually the fourth in alligatoroids and the fifth in crocodyloids). From observations on alligators, Busbey (1989) reported that large prey are seized at the level of the anterior region of the rostrum. Therefore, such caniniform teeth might greatly increase the chance of securing prey during attack. The functional meaning for the absence of caniniform teeth in Acynodon is difficult to assess but may reflect the fact that Acynodon did not hunt large vertebrates. Occlusion, or clipping action on molluscs, plants, or crustaceans are all equally likely to be responsible for the marked wear facets on the anterior peg-like teeth, but the diet of Acynodon will remain a matter of speculation until stomach contents are recovered.
Whatever its food may have been, it is tempting to express some further considerations about A. adriaticus. As already stated by Buscalioni et al. (1997), Acynodon can be considered as a specialized form of crocodylian. Even though the presence of molariform teeth has been considered as evidence for both specialization and versatility (for crocodylians see Buffetaut and Ford 1979; and literature therein; for other reptiles see Pregill 1984; Estes and Williams 1984), their association with isodont premaxillary and anterior maxillary peg-like teeth clearly indicates a strong specialization. The arched lateral profile of the maxilla, with the arch centred on the largest molariform tooth, adds further information for A. adriaticus. The shape of this downturned region, arched, wide and robust, may suggest that a particularly high stress developed in the area because of the action of the enlarged molariform teeth and that this arched morphology could disperse the stress more efficiently than a straight lateral profile of the skull. Further support for hypothesizing an apomorphic mastication style in A. adriaticus is offered by the hook-shaped process developed on the medial side of the quadrate condyle and of the retroarticular process of both sides of the skull (Text-fig. 4), both of which may testify for the presence of an unusual muscular link between the skull and the lower jaw.
Ontogenetic stage and size
MCSNT 57248 represents an adult specimen because the cervical vertebrae have a closed neurocentral suture; therefore, the apparently open scapulocoracoid synchondrosis visible in the same specimen could indicate that the closure was realized very late in ontogeny (Brochu 1995, 1996). A further indication of adulthood is suggested by the observation that the skull is so ossified that most of the sutures are hardly visible. Martin (2007) reported at Cruzy the presence of a small and therefore presumably young A. iberoccitanus specimen (ACAP-FX1) with clearly visible sutures that contrast with the undetectable sutures that usually characterize larger specimens of the same species. It seems therefore that Acynodon, like other crocodylians, tended to develop during the last ontogenetic stages a marked ossification and ornamentation that obscured the sutures. Assuming that MCSNT 57248 is an adult specimen, on the basis of its skull size it is possible to suggest that adult A. adriaticus reached a total length slightly larger than one metre.
The fact that skeletal elements of MCSNT 57248 are articulated with osteoderms clearly indicates that burial preceded the complete decay of the soft tissues, and that transportation was relatively limited. Both the best preserved crocodylian specimens from Villaggio del Pescatore, A. adriaticus MCSNT 57248 and the indeterminate crocodylian MCSNT 57031, as well as at least one of the dinosaurs, show a neck with a dorsal concavity of about 90 degrees. Flexure of the vertebral column, especially in the cervical region, has often been reported in fossil vertebrate skeletons, and was investigated in great detail by Weigelt (1927), who described modern instances of this phenomenon in carcases of Alligator mississipiensis in Texas that are of special relevance to our topic. Cervical flexure frequently occurs when a carcase dries and becomes naturally ‘mummified’. Shortening of the neck musculature and tendons brought about by desiccation results in marked curvature of the vertebral column, which can reach 90 degrees. Flexure of the vertebral column is therefore not a pathological phenomenon occurring at the time of death, as had been supposed (Moodie 1923), but a post-mortem event linked to desiccation of the carcase. In the case of Acynodon and hadrosaur specimens from Villaggio del Pescatore, a likely explanation of the flexed condition of the skeleton is that the carcases spent some of time out of the water, whereupon desiccation led to extreme flexure of the vertebral column, before they were eventually transported into a lagoonal environment where they were buried in carbonate mud. As underlined by Salisbury et al. (2003), the desiccation process, and therefore the hardening of the soft tissues, would facilitate the association of the skeletal elements during heavy transport by water. Anatomical connection that can be partly lost if the carcase, once in the water, is not buried quickly and the previously hardened soft tissues start to soften due to decomposition.
According to our results (Text-fig. 8), Acynodon can now be considered the basal-most globidontan and the North American short-snouted forms as more crownwards in Alligatoroidea. A previous analysis of Acynodon highlighted the fact that North America was not unambiguously ancestral for basal Globidonta (Martin 2007). This new analysis of Acynodon brings complementary data for the origin of that group. Even though the Late Cretaceous crocodylian fossil record has been growing fairly rapidly in recent years and new discoveries could quickly change the present interpretations, according to this new phylogenetic scenario, Globidonta may have originated in Europe. The presence of fragmentary remains of a probable alligatoroid in the Santonian of Hungary (Rabi 2005) is congruent with these results. It is noteworthy that the outgroup to Globidonta, the Diplocynodontinae, is nearly exclusively known from Europe: all the non-European evidence for this group is based on wrongly identified remains or on very fragmentary material that needs to be revised (Brochu 1999). The possible Santonian age of Acynodon adriaticus would make it the oldest recorded member of Globidonta and one of the oldest alligatoroids but, admittedly, this putative primacy is likely related to the present incompleteness of the mid-Cretaceous crocodylian fossil record, and the study of new undescribed fossil remains from the Upper Cretaceous of North America could change the biogeographic scenario presented here.
As clearly shown by Markwick (1998) and Buscalioni et al. (2003), palaeontological evidence indicates that crown-group crocodylians radiated in Europe and North America during the Late Cretaceous. Crocodylian crown taxa gradually increased in numbers during the Late Cretaceous and their expansive radiation apparently did not involve a geographical shift of more basal crocodylians. According to our current knowledge of the fossil record, all three extant crocodylian clades appeared before the end of the Cretaceous: crocodyloids appeared in the Maastrichtian, and both gavialoids and alligatoroids in the Campanian (Brochu 2003, 2004b). However, the three clades were present in different proportions, with alligatoroid genera representing at least 50 per cent of the crocodylian crown genera during the latest Cretaceous (Buscalioni et al. 2003). Despite having been only recently described, the alligatoroid genus Acynodon has been reported from several middle to late Campanian–Maastrichtian localities mostly located in western Europe but also in Romania (see Text-fig. 1). Moreover, the isolated molariform crocodylian teeth reported from the Gosau beds of Muthmannsdorf in Austria and the French locality of Champ-Garimond (also known as Fons; Buscalioni et al. 1999) could also belong to Acynodon. Debeljak et al. (2002) described from the Slovenian Upper Cretaceous (Campanian or Maastrichtian) site of Kozina (just a few kilometres away from Villaggio del Pescatore), teeth similar to but slightly different from those of the Acynodon species known at that time, that could possibly represent A. adriaticus.
From a geographical point of view, the Acynodon remains from Villaggio del Pescatore fill the wide gap between the French materials from Fox-Amphoux and the possible fragmentary Acynodon fossils from the Ha?eg Basin in Romania. From a chronological point of view, they may represent the earliest known evidence for Acynodon if the late Santonian age suggested by some authors (see Geological Setting) is confirmed.
Due to the complex geodynamic history of what is now the Mediterranean area during the Late Cretaceous, one that is far from being known in detail, the presence of remains of continental vertebrates such as Acynodon and the associated dinosaurs, as well as the remains from Kozina or the abundant dinosaur footprints on the Adriatic-Dinaric carbonate platform, is remarkable when the absence of evidence of terrestrial vertebrates reported until a few years ago is taken into consideration. In fact, there is growing evidence for the presence on the carbonate platforms (partly exposed above sea level) of terrestrial environments large enough to host vertebrates that in absolute size can be considered as relatively large (from one to several metres in total length). As for the size of the emerged lands, if the reduced size of the dinosaurs inhabiting the post-Barremian peri-Adriatic carbonatic platforms reported by some authors (Dalla Vecchia 2003a, b; and literature therein) could suggest evolution in a relatively small archipelago, the possible influence of taphonomical biases, as suggested by Le Loeuff (2005) for the Transylvanian ‘dwarf’ dinosaurs, does not permit any approximate estimation. Interestingly, Nicosia et al. (2007, p. 87) recently suggested, on the basis of the vertebrate fossils and ichnofossils of the periadriatic region, the presence of a ‘pre-Campanian link from Istria to southern Europe’.
The specific differences among the Acynodon inhabiting different emerged lands is difficult to interpret in terms of evolutionary history and dispersions. The picture of Acynodon evolution remains incomplete in space and time and inferring local biogeographic scenarios requires more complete record of the different species of Acynodon in western, central, and eastern European localities. However, the presence of more than one species of Acynodon further supports the hypothesis suggested by Rage (2002) that European latest Cretaceous vertebrate assemblages were not necessarily homogeneous. Whether these differences are simply ‘slight geographic and/or ecological variations’ (Le Loeuff 1991, p. 96) related to evolution in separated, and possibly different ecosystems, or are related to more diverse evolutionary histories, remains to be clarified.
Acknowledgements. We are grateful to ‘Soprintendenza per i Beni Archeologici per il Friuli Venezia Giulia’ and Sergio Dolce (Museo Civico di Storia Naturale, Trieste) for entrusting us with the study of the crocodylians from Villaggio del Pescatore. Giovanni Pinna (Milano) and Nevio Pugliese (Trieste) coordinated the ‘Villaggio del Pescatore project’. Deborah Arbulla and Nicola Bressi, curators of the Trieste Museum, assisted the authors during their work at the museum. Flavio Bacchia and his team (Stoneage s.r.l.) conducted the acid preparation of the material. M. Delfino wishes to thank Nicola Bressi for logistic support during his recurrent visits. Fabio Marco Dalla Vecchia (Monfalcone), Cristiano Dal Sasso and Simone Maganuco (Milano), Attila Ösi (Budapest), Paolo Piras (Roma), and Antonio Romano (Latina) kindly provided literature and suggestions. Steve Salisbury (Brisbane) and one anonymous referee reviewed the manuscript and provided relevant comments and corrections. The handling editor, Sean Modesto, carefully revised the manuscript and significantly improved it. M. Delfino’s research was supported by University of Florence ‘Fondi di Ateneo’ (coordinator L. Rook). J. E. Martin’s research supported by Le Conseil général de l’Allier and his visit to the Trieste Museum was founded by UFR Sciences de la Terre, Lyon 1; this is his contribution UMR5125-07.062.