A new fossil marine lizard with soft tissues from the Late Cretaceous of southern Italy

2.4.1. Parsimony

We performed both an equal-weight maximum parsimony (MP) and an implied weighting maximum parsimony (IWMP) analysis using TNT 1.5-beta [23–25]. The MP heuristic search was run using the tree-bisection-reconnection (TBR) algorithm, considered the best option for small datasets (27 taxa in our study) as per Goloboff et al. [23], with the number of maximum trees set to 99 999 and all the characters processed as unordered and unweighted. For the MP analysis, we applied the ‘Traditional Search’ to compute 1000 replicates of Wagner trees using additional random sequences and saving 10 trees per replicate; then, we employed a successive round of TBR branch swapping using trees from RAM, in order to increase the chance to find the actual shortest trees (see electronic supplementary material 2 and 3 for further information on analysis settings and outputs). Two optimal trees were retained after removing all the suboptimals, and the strict consensus topology is presented with relative supports in figure 6a (both optimal trees are included in electronic supplementary material, figure S11). Following Goloboff et al. [24,25], we also performed an IWMP analysis with K set to 3, and adopting the same steps described for the MP analysis. The IWMP resulted in a single optimal tree and the topology is presented in figure 6b (see also electronic supplementary material, figure S12). The nexus file used to run the analyses is provided as electronic supplementary material 3.

2.4.2. Bayesian inference

The Bayesian inference (BI) analysis was performed with MrBayes 3.2.6 [26], under the Mk(V) model for variable characters [27]. As MrBayes does not handle polymorphic scorings, we converted all the polymorphisms in the dataset used to run the parsimony analysis into uncertainties using Mesquite 3.04 [22]. We used a gamma distribution for rate heterogeneity and treated the data as a single partition. We set generations to 10 000 000, frequency of sampling to 1000, burn-in fraction to 0.25 and the temperature parameter to 0.010 (which gave the best chain mixing values). We checked the optimality of the parameters for convergence and effective sample size from both MrBayes log file, and Tracer 1.6 [28]. LogCombiner [29] and Tree Annotator [30] were used to estimate the posterior tree (maximum clade credibility tree (MCCT)). The nexus file used to run the BI is provided as electronic supplementary material 4, and further details about the settings and outputs are included in electronic supplementary material 5.

3.3.1. Cranial skeleton

The skull is dorsoventrally crushed and exposed in dorsal view (figure 2; electronic supplementary material, figure S2). Many elements are fragmentary and preserved in part as impressions, as portions of the bones were lost with the unknown counterpart. This is the case for the parietal and frontal, both of which are preserved as impressions in the area adjacent to the fronto-parietal suture. The occipital region is badly crushed, and the cavities for the semicircular canals are partially exposed as the posterodorsal portion of the braincase (i.e. part of the otoccipitals) is missing. The limit between the basioccipital and the atlas is clear, and on the right side of the atlas, forming a 45° angle with its longitudinal axis, is a thin and long bone projecting posteriorly, that is part of the hyoid apparatus and most probably represents the first ceratobranchial. Both quadrates can be easily identified: the right quadrate is better preserved and almost complete; the left is mostly present as an impression on the matrix. On the right side, almost all the original contacts for the quadrate are preserved both dorsally (with the skull roof) and ventrally (with the pterygoid and the mandible). The anterior portions of both the lower and upper jaws are hard to differentiate, while posteriorly the mandibular elements are easier to recognize at least as impressions, with the retroarticular process being particularly well developed. The anterior portion of the skull preserves the septomaxillae and portions of the nasals and premaxilla.

The three large foramina on each side at the back of the braincase are most probably areas where the cavity of the inner ear (semicircular canals) is exposed due to the breakage and removal of the dorsal portions of the otoccipitals (electronic supplementary material, figure S2a). The anterior and posterior cavities correspond to coronal sections through the anterior and posterior semicircular canals, respectively; the largest and more medially placed cavity most probably represents a section through the crus commune, i.e. the portion of the inner ear where anterior and posterior semicircular canals meet dorsally. Between these foramina is a trapezoidal element that is probably a dorsoventrally flattened supraoccipital. Anterior to the semicircular canals, in what is probably a portion of prootic, a distinct foramen is visible on the right side, and is interpreted here as an opening for the VII cranial nerve (facial nerve). These identifications are based on braincase descriptions in Russell [36], Rieppel & Zaher [37], Bever et al. [38] and Head et al. [39]. Posterior to the otooccipital, the occipital condyle is partially exposed and its articulation with the atlas is visible as a slightly convex line (electronic supplementary material, figure S2a).

The parietal is very fragmentary, especially anteriorly, but a general outline of this bone can be resolved in the dorsal view on the left side, where the posterior process is broken but almost complete, and as an impression of its ventral surface on the right side (electronic supplementary material, figure S2a). The parietal table is broadly trapezoidal, while the posterolateral process (visible on the left side) is slender and triangular in dorsal view. It is not clear whether a large gap between the parietal and the supraoccipital represents the equivalent of the space for the processus ascendens tecti synotici or is an artefact of preservation. The fronto-parietal suture is preserved mostly as an impression, but it is clear that it was fairly straight, similar to that observed in aigialosaurs or even modern monitor lizards. The outline of the pineal foramen, located anteriorly in the parietal table, can be easily recognized, and anterior to it is a distinct mid-sagittal line that divides the left and right sides of the parietal. The posterior portion of the parietal, still mostly represented by bone, indicates that the bone was divided only anteriorly (parietal notch) as is typical of juvenile monitor lizards [40] and was not paired, an adult feature observed among extant lizards only within gekkotans [41]. The incomplete ossification of the left and right counterparts of the parietal can be interpreted as either a juvenile feature or a delay in ossification, a phenomenon common in aquatic forms (see Discussion).

On the right side of the skull, in front of the prootic and lying along the right side of the parietal table, is a rod-like element that we interpret as the epipterygoid, which must have rotated 90° due to the dorsoventral diagenetic compression of the skull.

The frontal is a very elongated unpaired element, much wider posteriorly at the suture with the parietal, then strongly constricted between the orbits and finally tapering anteriorly between the nasals. As in pontosaurs and coniasaurs, the posterior end is much more broadly expanded than the anterior one [3,6,13]. The posterior half of the frontal is mostly preserved as an impression, whereas anteriorly some fragments of the bone are still present, and the most anterior fragment (tip of the frontal) is located slightly anterior to the midpoint of the prefrontal. On the impression of the posterior ventral surface of the frontal, it is possible to recognize the natural mould of the olfactory canal (figure 2; electronic supplementary material, figure S2a). Whether or not the canal was completely enclosed by descending flanges of the frontal anteriorly cannot be determined.

Laterally and in front of the orbits, the frontal articulates with the prefrontals. A fragment of bone sandwiched between the frontal and right prefrontal may represent the posterior end of the right nasal. The extension of the impression of the premaxillary internarial bar suggests that originally it came in contact with the anterior tip of the frontal posteriorly (figure 2; electronic supplementary material, figure S2a).

A fragment of the left post-orbitofrontal is preserved on the left side of the skull. It clasps the fronto-parietal suture, forms the posterodorsal margin of the orbit and has a distinct squamosal ramus projecting posteriorly (electronic supplementary material, figure S2a). The squamosal ramus is cracked longitudinally, but its slender, distally tapering shape can nonetheless be inferred. Posterolateral to the squamosal ramus of the post-orbitofrontal is the impression of a pointed and slightly recurved element that must be the anterior portion of the squamosal. The impression can be followed posteriorly as it connects to a series of thin rod-like fragments.

Medial to the head of the right quadrate, and partially overlapping it, there is another rod-like element that we interpreted as a fragment of the right squamosal. This element can be followed anteriorly into a series of other fragments and impressions that together taper into a point, very similar in thickness and length to its left counterpart (electronic supplementary material, figure S2a).

The supratemporal is preserved on the left side as a small element inserted between the otoccipital and the posterior end of the squamosal. Its right counterpart can be identified in a similar position on the opposite side of the skull, where most of it is exposed due to breakage and displacement of the posterior end of the right squamosal (electronic supplementary material, figure S2a). The extension of the contact between the supratemporal and the quadrate seems to be greater than the contact between the quadrate and the squamosal, and also seems to prevent the contact between the quadrate and the paroccipital process (at least in the dorsal view).

Both quadrates are preserved in their articular position: the right one is quite complete and exposed in posterolateral view, whereas the left quadrate is present mostly as an impression of its medial face. The anterior outline of the quadrate is quite convex, and there is a distinct, posteriorly projecting suprastapedial process that is well preserved on the right quadrate. Details of the tympanic ala and tympanic crest cannot be resolved, but a distinct lateral conch is visible anterior to the suprastapedial process (figure 2; electronic supplementary material, figure S2a). Anterodorsally, the quadrate head is still in articulation with a rod-like element that probably represents a fragment of the squamosal (rather than the supratemporal); the rest of the dorsal–posterodorsal contact was probably occupied by the parietal ramus (which is missing in the right side of the skull, but is present on the left half), and then with at least one elongated element that projects farther more posteriorly than the quadrate itself, which we interpret as the supratemporal. This last bone seems also to prevent the contact between the quadrate and the paroccipital process, the position of which is indicated by the semicircular canal openings (electronic supplementary material, figure S2a). The right quadrate overlaps the posterior process (quadrate ramus) of the pterygoid, with which it is in contact medially. It is interesting to note that, according to the geometric relationship that the quadrates and mandibles have as compressed on the slab, the quadrates would turn to be almost vertical in lateral view, more similar to the condition seen in mosasaurs, aigialosaurs or even most iguanians, rather than, for instance, in varanids [2] (I.P. 2016, personal observation).

Only the right pterygoid is partially exposed in dorsal view, and part of it is only preserved as an impression (electronic supplementary material, figure S2a). The quadrate ramus consists of a robust, plate-like bone, posteriorly recurved towards the quadrate condyle (i.e. with the concavity facing laterally). The termination of the quadrate ramus does not taper significantly posteriorly, and terminates in a blunt, sub-rectangular end. On the left side the quadrate ramus is not exposed, but anteriorly a bone that could be interpreted as the ectopterygoid process is exposed in dorsal view, and forms the floor to the anterior portion of the orbit.

On the right side of the skull is a natural mould of the prefrontal, located just in front of the orbit; the element is rotated medially, so that what we see is the mould of its lateral and posterior walls. The left counterpart instead is not clearly identifiable among a mass of bone fragments. Owing to poor preservation, nothing can be said about the contact between the prefrontal and the maxilla. The contact between frontal and prefrontal appears to have been very limited, and must have occurred at the posterior end of the external naris. The lateral wall of the prefrontal tapers anteriorly, and has a straight lateral margin, while posteriorly it is both dorsoventrally deeper and mediolaterally wider. No indication of a lacrimal foramen or notch can be observed. The posterior wall shows a gently sinusoidal ventral margin, and a weak median concavity that is facing posteriorly, while its dorsal margin is smoothly rounded (figure 2; electronic supplementary material, figure S2a).

The premaxilla is missing, and only impressions and fragments of the internarial bar are visible between the two septomaxillae, and extending posteriorly towards the anterior tip of the frontal (electronic supplementary material, figure S2a). Close to the tip of the snout, two sub-triangular and paired elements are identified as the septomaxillae. The left septomaxilla is fractured, and only its medial portion is preserved. The right septomaxilla is complete, and its shape in dorsal view is extremely similar to that of Coniasaurus gracilodens [13]. The anterolateral margin shows a sutural contact with the maxilla, while posterolaterally it was at least in contact with nasal (electronic supplementary material, figure S2a): this sutural contact between the septomaxilla and the maxilla probably prevented the maxilla from moving independently of the rest of the skull (A.P. 2016, personal observation). The external naris must have been framed by the septomaxilla anteriorly, the medial margins of the maxilla and the prefrontal laterally, and the internarial bar of the premaxilla and the nasals medially, and must have terminated posteriorly in a tapering point where the prefrontal, nasal and possibly the frontal met.

Two paired elements, very narrow and pointed anteriorly, are identified as fragments of the anterior ends of the nasals, and are very similar in shape and topographical location to those of Pontosaurus kornhuberi [6]. A sub-triangular fragment at the posterior end of the external naris, between the prefrontal and the frontal, is also interpreted here as a fragment of the right nasal (posterior end); this is because of its shape, size, position and the presence of an impression in the sediment that connects this element to the anterior tip of the nasal described above. If this fragment is in its natural position, then the frontal may have been excluded from the posterior margin of the external naris by the nasals and the prefrontals.

Only fragments of the left and right maxillae are present. The left maxilla is the most complete, and appears to be preserved in dorsal view. The top portion has been sheared off and displaced medially, so that the canal for the second branch of the trigeminal nerve (V₂) is exposed. Anteriorly, and located on the course of the canal for the above-mentioned nerve, is the section through a large tooth alveolus. Only one small tooth is preserved anteriorly on the maxilla. Just under the maxilla, and exposed only anteriorly, where a portion of the latter is missing, a fragment of the dentary can be observed. It bears one large tooth, complete with the root. Small fragments of dentary are also visible on the right side of the skull, but not much can be said about the general shape of this bone, except that it was probably extending posteriorly below the orbit.

Most of the lower jaws are preserved only as impressions (figure 2; electronic supplementary material, figure S2a). The condylar region appears to be exposed in dorsolateral view, as suggested by the shape of the retroarticular processes, but anteriorly the jaws are twisted somewhat more medially. On the impression of the right mandible, it is possible to identify one extensive suture line running anteroposteriorly, and starting in front of the condylar region: considering that the impression is that of the lateral face of the jaw, this suture is most probably that between the surangular and the angular (electronic supplementary material, figure S2a). On the left side this same suture can be observed dividing posterior fragments of the surangular and the angular. The retroarticular processes are preserved only as impressions, but it is possible to infer their size and shape very clearly. The retroarticular processes were broad and sub-rectangular, similar in shape and extension to those of Pontosaurus spp. and Adriosaurus suessi [3,6,42].

Scleral ossicles are visible inside both orbits, but an almost complete scleral ring is only visible on the right side. The shape of the individual ossicles cannot be established due to poor preservation, and a count of the number of elements is also not possible.

A jugal is not preserved, but the posterior extent of the orbit can be estimated by observing the posterior extension of the sclerotic ring. The orbits must have been anteroposteriorly elongate and relatively quite large, though not as large as in Pontosaurus (orbit diameter to skull length = 0.13; in P. kornhuberi = 0.19; in P. lesinensis = 0.19) (see electronic supplementary material, table S1, for measurements of the specimen).

As mentioned above, only two marginal teeth are preserved in association with the skull elements (electronic supplementary material, figure S3). They are both located towards the anterior end of the skull, and point in opposite directions. The most anterior of the two, missing the root and pointing laterally in dorsal view, is interpreted as a maxillary tooth; while the other one, almost complete and pointing medially, probably belongs to a fragment of the left dentary. Both teeth are conical, with no apparent lateral compression, and are slightly recurved posteriorly. The tooth crowns have multiple longitudinal facets separated by thin ridges (electronic supplementary material, figure S3a,b).

A slender and elongated element coming out from below the base of the skull, and exposed on the right side of the atlas and axis, is the first ceratobranchial (figure 2; electronic supplementary material, figure S2a). It is broken in two sections at mid-length and the posterior half is slightly displaced medially. This element terminates in a blunt end posteriorly, to which a thin rod of calcified cartilage is attached. The cartilaginous portion is preserved in a displaced position, and its longitudinal axis forms an angle of about 60° with that of the ossified portion of the ceratobranchial.

3.3.2. Axial skeleton

Both vertebrae and ribs show some degree of pachyostosis, which is, sensu stricto, a thickening of the perichondral bone—as defined by Ricqlès & Buffrénil [43]. The mode of preservation of most of the skeleton facilitates the observation of the thicker walls (of vertebrae and ribs especially) and the brittle-like internal organization of the bony tissue. Unfortunately, without sectioning the bones it is not possible to verify if osteosclerosis had developed in the inner bone tissue [44–47].

Both the cervical and dorsal vertebrae are elongate, and roughly rectangular in shape, while the sacral and caudal vertebrae are shorter and more square (see measurements in electronic supplementary material, table S1). The dimensions of the vertebrae are, in general, quite different from typical ophidiomorph squamates, in which cervicals and dorsals are quite short relative to their width.

The precaudal vertebral column is complete, though for the tail the most distal part is missing. The boundary between cervical and dorsal series—as defined by Hoffstetter & Gasc [48]: the first dorsal vertebra bears the first rib pair that articulates to the sternum—can be determined due to some preserved costal cartilages (figure 5d). There are four well-preserved costal cartilages on the right side of the body, which although partially covered by dorsal ribs can be highlighted using UV light. The third of these costal cartilages still preserves its articulation with one of the ribs. By following a rib to its articulation with the corresponding vertebra, it is possible to determine that the latter must represent the third dorsal, and that therefore there are 10 cervical and 22 dorsal vertebrae. The atlas is identified as the shorter element articulating with the basioccipital condyle, just before the first cervical vertebra with a rib (i.e. the axis) (electronic supplementary material, figure S2a). The number of presacral vertebrae is very different from Pontosaurus, in which the number of dorsals is above 26, and in general from most ophidiomorphs that have a presacral vertebral count greater than 35 (39 in Pontosaurus kornhuberi) [3,6,7,49]. The presacral vertebral count in Primitivus is 32, just slightly higher than Aigialosaurus dalmaticus, where there are only 7 cervicals, followed by 22 dorsals in Primitivus, for a total of 29 presacral vertebrae.

The dorsal vertebrae in MPUR NS 161 are all broken through the neural arch or slightly ventral to it (figures 1 and 5). Their centra are cylindrical and robust, slightly expanded anteriorly and bear well-developed, laterally projecting synapophyses. The last dorsal bears a pair of very small ribs.

There are two sacrals, with the sacral ribs still in articulation with the posterior iliac blade. The first pair of sacral ribs is directed laterally, while the second pair is somewhat recurved anteriorly. The distal bony end of the second sacral rib bears an indentation on the posterior margin, which on both the left and right side is occupied by cartilage (figure 3c,d; electronic supplementary material, figure S4b). Such a morphology of the second sacral rib is similarly present in some iguanians, such as Iguana sp., Agama sp. and Physignathus sp., while in other lizards, such as Gecko sp., Varanus sp. and Heloderma sp., the distal bony margin of the second sacral rib appears quite square, and the posterior margin of the second sacral rib is nearly straight (I.P. 2016, personal observation). In the case of Iguana sp. or Agama sp., it looks like the posterior margin of the second sacral rib has a bony projection located about mid-length; in the case of MPUR NS 161 or Physignathus sp., this bony projection is located more distally, looking more like an indentation of the posterior distal corner of the sacral rib.

Posterior to the sacrals, there are several vertebrae with long distally tapering transverse processes that point laterally, but from about the tenth caudal vertebra the orientation of the transverse processes changes, becoming slightly posteriorly oriented. There are 19 caudal vertebrae preserved in dorsal view after the sacrum, then the tail rotates about 90° (between caudals 17 and 21) and the following caudals are exposed in left lateral view (figures 1, 6g,h and 7). The transverse processes are well developed in the first ten caudal vertebrae, then reduce in size between the 11th and 16th vertebra, where the tail begins to turn. After the curvature, there is no more evidence of transverse processes, although this might be due to preservation bias (these vertebrae are preserved in lateral view and the transverse processes may have broken off). In the portion of the tail exposed in lateral view, long haemal arches (or chevron bones) can be recognized between the 22nd and the 27th caudals (figures 6g,h and 7). After the 27th caudal vertebra, there is a gap due to matrix covering the specimen, followed by at least seven more caudals. In total, there are 37 caudals preserved, but considering that the last vertebrae present on the slab show no significant reduction in size, the tail was probably much longer. The chevron bones are slightly flattened against the vertebral centra, and most of them are disarticulated; their length is greater than the corresponding neural spines, at least for all the caudals preserved in lateral view (figure 6g,h). The caudal vertebral centra bear posteriorly a distinctive pedestal (haemapophysis) to which the chevron articulates, so the haemal arches are not fused to the haemapophyses, and the articular facet is posteroventrally oriented. Moreover, between two of the caudal vertebrae preserved in lateral view, it is possible to observe a zygosphene–zygantrum supplementary articulation (figure 6h).

The caudal neural spines in lateral view are inclined posteriorly about 45°, and narrow anteroposteriorly. Some scales preserved as both mineralizations and impressions in lateral view along the caudal region assist in determining the outline of the tail (figure 6f–h). The hypaxial portion is greatly dorsoventrally deepened in comparison to the epaxial portion; however, it is also clear that the epaxial portion of the tail must have extended dorsally beyond the neural spines. Indeed, along the dorsal edge of the tail, the scales impressed on the matrix indicate that there was some sort of caudal fin, similar to that of some modern sea snakes and sea kraits (e.g. Hydrophis platurus, Laticauda colubrina), or the water monitor, Varanus salvator. The width of the anterior portion of the tail, as inferred from the extension of the transverse processes, is quite remarkable, and is consistent with attachment for powerful caudofemoralis muscles. The depth of the posterior part of the tail suggests that it must have served as an excellent propelling organ during swimming.

Fragments of cervical ribs are preserved along the neck, but most of these ribs are either not fully exposed or too fragmentary to allow proper description. As was mentioned above, the limit between the neck and the trunk is recognizable owing to the preservation of four sternal ribs on the right side of the body, one of which (the third) retains its connection to one of the trunk ribs, which in turn is articulated to one of the vertebrae (the third dorsal vertebra).

All the ribs are single-headed. The proximal head is flared, and there is no neck-like constriction. The ribs have a thick, pachyostotic shaft, and taper distally towards the end, but expand again just before the tip, probably to form a surface for the attachment of a cartilaginous termination (figure 5d). The anterior dorsal ribs are uniformly recurved similar to Pontosaurus lesinensis or Dolichosaurus longicollis, and unlike Pontosaurus kornhuberi or Mesoleptos zendrinii, where the distal portion of the ribs is straight [3,6–8,11,14]. Some of the trunk ribs preserved as impressions on the slab also show a weak longitudinal groove, closer to the anterior margin of the shaft. The longest dorsal rib is about 81.5 mm, which indicates that the trunk must have been fairly deep compared to other squamates, as can be expected in an animal adapted to swimming [50] (electronic supplementary material, table S1). Finally, in MPUR NS 161 there are five terminal dorsal ribs considerably shorter and straight in comparison to the rest of the thoracic series. This anatomy differs from that of mosasaurids where there is a much longer series of short presacral ribs, and is similar to more basal non-ophidian pythonomorphs where this feature can be observed. In Pontosaurus kornhuberi there are at least four shortened presacral ribs, while in Adriosaurus spp. the decrease in size is quite gradual, with the last two posterior dorsal ribs being significantly shorter. The condition is variable instead in aigialosaurs, because in Aigialosaurus dalmaticus all the presacral ribs are generally shorter and decrease in length gradually as in Adriosaurus spp., whereas Aigialosaurus bucchichi is similar to MPUR NS 161.

3.3.3. Appendicular skeleton

With respect to Pontosaurus kornhuberi, Acteosaurus tommasinii and Adriosaurus suessi, the contrast in length between forelimbs and hindlimbs in MPUR NS 161 is not as pronounced, being more similar to the condition in both Aigialosaurus species. Following Palci & Caldwell [8], this can be quantified through the length ratios of the humerus and femur to the mean dorsal vertebra (mdv): the humerus to mdv ratio in MPUR NS 161 is up to 2.3, against a value of 1.3 in Acteosaurus, 2.0 in P. kornhuberi, between 1.6 and 2.2 in Adriosaurus suessi, and a similar value of 2.3 for Aigialosaurus spp.; for the femur to mdv ratio instead, the value in MPUR NS 161 is 2.9, slightly higher than that of both Aigialosaurus spp. (2.6–2.7) and Acteosaurus (2.7), and much lower in comparison to P. kornhuberi (3.3) and Adriosaurus suessi (3.3–3.6).

Only the right pectoral girdle is clearly recognizable on the skeleton as exposed, and this is partially overlapped by the ribs at the cervical–dorsal series transition, which are now lost but have left impressions on the surface of the bones (figure 3a,b; electronic supplementary material, figure S4a). Of the left pectoral girdle, only a large cartilaginous element is visible on the left side of the trunk, and this most probably represents the suprascapular cartilage (figures 1 and 5a,b).

The scapula and coracoid are single elements, and are not fused together. Under UV light, cartilage is identified in several places around both the scapula and coracoid (figure 3a,b; electronic supplementary material, figure S4a). Overall, the pectoral girdle is quite reduced in comparison to the rest of the body, a feature typical of pythonomorphs, and its morphology resembles very much the condition seen in Carsosaurus marchesetti and Dolichosaurus longicollis.

The scapula is hourglass-shaped, much smaller than the coracoid and with both ends about the same width. This is different from the condition identified, for instance, in Adriosaurus skrbinensis, and more similar instead to that of Dolichosaurus longicollis, Carsosaurus marchesetti and Coniasaurus gracilodens.

The coracoid is crescent-shaped, similar to that of Aigialosaurus bucchichi and Carsosaurus marchesetti. There seems to be no emargination on the anterior margin of the coracoid, while a scapulocoracoid fenestra seems to be present close to the glenoid fossa, as in C. marchesetti. The presence of a coracoid foramen cannot be confirmed, due to extensive cartilage material covering the median portion of the bone (epicoracoid cartilage), surrounding both the coracoid and the scapula, and probably in contact with the suprascapula (which is also preserved as cartilage).

The forelimbs are preserved pressed to the body and pointing posteriorly, so that the hand is exposed in ventral view (flexor aspect). The overall morphology of the humerus resembles that of Carsosaurus marchesetti, rather than either Pontosaurus or Aigialosaurus. The bone is hourglass-shaped but still quite elongated in comparison to the pectoral girdle elements and the presacral column, while in both Pontosaurus and Aigialosaurus the propodial is clearly shortened relative to the overall length of the limb (see measurements in electronic supplementary material, table S1). The distal end surface of the humerus is damaged, and the bone does not show the presence of either the ectepicondylar or the entepicondylar foramina. Although lack of an entepicondylar foramen is expected, as it is an autapomorphy of Squamata [51], lack of the ectepicondylar foramen may be preservational. The epiphyses are present on both humeri, but are not fully ossified, as observed when the skeleton is exposed to UV light (figures 3 and 4; electronic supplementary material, figures S4–S7).

The radius and ulna are best preserved on the left side. It is not possible to resolve their proximal epiphyses on either side of the body; however, on the left forelimb, distal unfused epiphyses (slightly disarticulated) are clearly visible on both bones. The two bones are close together proximally, contacting each other at the articulation with the distal margin of the humerus and then strongly diverge distally, although part of the divergence is artificial, because the ulna is no longer in articulation with the ulnare and its distal end is located dorsal to the pisiform (figure 4a–d; electronic supplementary material, figure S6). Divergent epipodials are also characteristic of other non-ophidian pythonomorphs, and are considered to be associated with an aquatic lifestyle [3,36,42,52,53]. The radius is a rod-like bone, slightly hourglass-shaped, with its posterior margin more prominently recurved than the anterior one; both its proximal and distal ends are only weakly enlarged in comparison to the thin shaft. The distal end has an oblique surface to which the distal epiphysis is attached. The ulna has a more evidently constricted shaft, with a more symmetrical anterior proximal end, and a posterior proximal end characterized by a distinct olecranon process.

In both fore- and hindlimbs the autopodium is much longer than the epipodial portion, consistent with the tendency of reduction of the proximal elements of the limb found in other non-ophidian pythonomorphs; however, in MPUR NS 161 this tendency is not as strong as in both Pontosaurus and Aigialosaurus for the propodials, even if the autopodial length is more than twice the length of the epipodials.

The right manus is very poorly preserved, so the following description is based on the left, which is complete, although surface preservation is not excellent (most of the perichondral bone has been sheared off) (figure 4a–d; electronic supplementary material, figure S6). In the left manus only the proximal carpals are readily recognizable, while, of the distal carpals, only the large fourth distal carpal can be seen close to the proximal end of the metacarpal IV.

The proximal carpals consist of a large square radiale, located between the epiphysis of the radius and metacarpals I and II; a large round central element (probably the lateral centrale), located postaxial to the radiale; a large oval ulnare, postaxial to the lateral centrale; and a small, comma-shaped pisiform, sandwiched between the ulnare and the distal epiphysis of the ulna (therefore, the distal end of the ulna is clearly disarticulated and shifted somewhat postaxially).

All metacarpals are hourglass-shaped (figure 4a–d; electronic supplementary material, figure S6). Metacarpal I is the shortest and broadest of the metacarpals, followed by metacarpal V, while metacarpal III is the longest.

The phalangeal formula for the manus is 2-3-4-5-3. The shape and size of all the phalanges (excluding the unguals) are similar in all digits with a flared proximal head and a less expanded distal condyle (electronic supplementary material, figure S6b). Most ungual phalanges are quite well preserved on both manus and pes, making up a distinct claw, posteriorly recurved and bearing two tubercles for attachment of the flexor musculature: a larger one located ventroproximally, and a smaller one located on the ventral margin (electronic supplementary material, figure S8).

All the unguals appear mediolaterally compressed and taper anteriorly into a blunted distal tip. On some ungual phalanges the articulation for the penultimate phalanx is also visible, and this facet appears slightly sinusoidal in the lateral view. On the dorsolateral surface of the ungual phalanges there are two foramina: one is located proximally, and the other one more distally, at about the mid-length of the dorsal margin leading to the distal tip of the ungual. Below this second foramen, running longitudinally along the tapering distal end of the ungual phalanges, there are some parallel grooves that do not reach the proximal end of the ungual (electronic supplementary material, figure S8).

The pelvic girdle is flattened on the slab and both sides are exposed in medial view, with the individual bones still in articulation or just slightly dislocated (figure 3c,d; electronic supplementary material, figure S4b). All the pelvic elements are preserved: pubes and ischia are complete, while the ilia are present part as actual bone and part as an impression in the matrix. Although tightly connected, the individual pelvic elements are not fused together.

In dorsomedial view, the ilium is characterized by an elongated and well-developed rod-like, posteriorly oriented process—here referred to as the posterior iliac or post-iliac process—and a long, thin anteroventrally oriented preacetabular process overlapping the pubis (electronic supplementary material, figure S4b). The ilium is still connected to both sacral ribs, and this articulation is visible on the medial aspect of the right post-iliac process, although most of the iliac process is preserved only as an impression. The contact between the second sacral rib and the ilium is intact, with a strip of cartilage completing the termination of the rib onto the posterior iliac process (figure 3c,d; electronic supplementary material, figure S4b). The posterior end of the post-iliac process is blunt in the mediolateral view, and on the left side it partially overlaps the left transverse process of the first caudal (pygal?) vertebra. The presence of the anterodorsally oriented supra-acetabular process, found in many terrestrial lizards (e.g. varanoids, iguanians) as well as mosasaurids, cannot be verified due to preservational factors: both left and right anterior iliac portions are flattened against the head of the femur, and their dorsal margin, where the supra-acetabular anterior iliac tubercle might be, is not clearly exposed. The two facets for articulation with the pubis and the ischium on the iliac head have about the same length: the right ilium is still weakly articulated to the ischium more posteriorly, while the left ilium is still articulated to the pubis but only partially with the ischium.

The left pubis is particularly well preserved, with a broad proximal head that is greatly expanded posteriorly in lateral view (resulting in the typical hatchet-like shape for this bone) (figure 3c,d; electronic supplementary material, figure S4b). The distal end of the right pubis is hidden underneath the last dorsal vertebra, and only its proximal head remains visible. Owing to the poor preservation of the bony surface, it is not possible to determine the location or presence of a pubic foramen. The anterior pubic process (or tubercle) is very inconspicuous, and appears only as a swollen eminence along the anterior margin of the proximal head of the pubis, not far from the acetabulum. The articular facet for the ilium occupies most of the dorsal and posterodorsal margin of the pubis in medial view; while the facet for the ischium is located posteriorly on the pubic head. The ventromedially directed pubic shaft is significantly narrower than the proximal head, and ends distally in a square termination, quite weakly expanded; still attached to the distal end, there is also a fragment of cartilage that is most probably part of the pubic symphysis. On the medial side of the pubic shaft, there is a well-preserved and dorsoventrally elongated, teardrop-shaped surface for the attachment of muscle tissues: considering the position (medial view), the surface was probably for the insertion of the musculus puboischiofemoralis internus [54] (figures 3c,d and 5e,f; electronic supplementary material, figure S4b).

The ischium is the shortest element of the hip. Both ischia are slightly dislocated from their original position in connection with the other pelvic bones, and are partially covered by the sacral vertebrae and ribs (electronic supplementary material, figure S4b). The ischium is strongly recurved along its anterior margin, and its proximal head is narrower than the distal end. The ischial expansion opens posteriorly right below the ischial neck to form a steep angle and then continues along the posterior margin of the ventromedially directed shaft, almost until the distal end. The distal termination of the ischium, where the element would contact its counterpart along the midsagittal plane, is fairly straight and at least twice the size of the distal extremity of the pubis. The proximal head of the ischium articulates with the pubis along an anterior facet, and with the ilium along its dorsal margin.

Unlike in Aigialosaurus spp., in MPUR NS 161 the femur is still quite long relative to the axial skeleton, and the morphology of the pes is not significantly modified in comparison to a terrestrial lizard, as observed in Pontosaurus [3,6,7]. However, the epipodials in the hindlimbs, similar to those described for the forelimbs, are strongly divergent distally, a feature that is associated with swimming [3,42].

As in the humerus, the epiphyses of the femur are not completely ossified (figure 3c,d; electronic supplementary material, figure S5b). The shaft of the femur is quite long and robust, almost twice the length of the epipodials (electronic supplementary material, table S1). The proximal epiphyses are partially overlapped by the anterior portions of the ilia. The proximal head of the femur is expanded, but less than the distal end; a gently rounded condyle for articulation with the pelvic girdle is visible at least on the left femur, especially when the element is exposed to UV light. On both sides of this condyle, in posterior view, there are two weak trochanters, with the internal one being slightly lower than the external (figure 3c,d; electronic supplementary material, figures S4b and S5b). For the articulation with the tibia and fibula, the distal femoral end in posterior view bears a more prominent mesial (tibiofibular) condyle, and a less prominent but wider lateral (tibial) condyle.

As a result of the flattening of the hindlimbs, with most of the axial skeleton visible in dorsal view, both pairs of epipodials are exposed on the slab in posterolateral view (electronic supplementary material, figure S5b). The epiphyses of both the tibia and fibula are not fully ossified, nor fused to the diaphyses. Tibia and fibula are in close contact proximally, at the articulation with the femur, but then diverge distally, as described for the radius and ulna.

The tibia is more robust than the fibula, and its proximal head is larger than the distal end (electronic supplementary material, figure S5b). On the right tibia, both medial and lateral condyles can be recognized, and still in contact with the articular cartilage of the distal end of the femur. The distal end of the tibia is fan-shaped, and articulates with both the astragalus and a small preaxial element that is identified as part of the tibial epiphysis. The tibial shaft is strongly constricted at mid-length and its internal margin is more prominently recurved than the external one.

The fibula is preserved mostly as an impression on both sides (only proximal and distal extremities are represented by bone). It is more gracile than the tibia, and both proximal and distal terminations are about half as wide as those of the latter bone (electronic supplementary material, figure S5b). The internal articulation of the left fibula with the medial condyle of the tibia is particularly well preserved, showing the close proximal contact of the two bones. The distal end of the fibula is somewhat irregular, due to the greatly expanded articular cartilage, and it contacts both the calcaneum (distally) and the astragalus (preaxially).

The mesopodium of MPUR NS 161 is one of the most complete found in dolichosaurs (figure 4e,f; electronic supplementary material, figure S7). Both distal and proximal rows of tarsals are preserved on the hindlimbs, though better exposed in the right pes, which is not overlapped by the tail. The proximal row consists of a bowtie-shaped astragalus, and a trapezoidal calcaneum, which shows a weak proximal concavity for articulation with the fibula (electronic supplementary material, figure S7b). The two bones are clearly not fused together. The astragalus articulates proximally with the epipodials, and postaxially with the calcaneum.

In the right pes, distal to the astragalus and calcaneum, a large centrale and four distal tarsals are preserved. The first three distal tarsals increase gradually in size postaxially, while the fourth is the smallest (electronic supplementary material, figure S7). There is no fifth distal tarsal. The centrale, which lies between the astragalus and the fourth distal tarsal is a large trapezoidal element, roughly as large as the calcaneum.

In all metatarsals the epiphyses are poorly ossified, especially distally (figure 4e–h; electronic supplementary material, figure S7). Metatarsals I and V are shorter and stouter in comparison to the elongated and slender metatarsals II to IV (electronic supplementary material, table S1). Metatarsal I is the only element of the metapodium with a distal termination that is smaller than the proximal one, and it also has a greatly recurved preaxial margin. Metatarsal IV is the longest element of the metapodium, and like metatarsals II and III, is characterized by a relatively thin shaft. Metatarsal V is broadly expanded proximally, and slightly hooked.

Common to both metatarsals and phalanges is the presence of a flexor groove on the ventral aspect: this structure is visible on the distal half of the autopodial elements of the right pes that are preserved as impressions in the matrix (figure 4e,f; electronic supplementary material, figure S7).

The phalangeal formula for the pes shows the typical primitive condition seen in lepidosaurs, corresponding to 2-3-4-5-4, so there is no reduction of the fifth digit, contrary to what is observed in aigialosaurs and mosasaurs. The overall shape of the phalanges of the pes is similar to that of the same elements in the manus: they have a broad square proximal end, a constricted shaft and a smaller distal end (electronic supplementary material, figure S7b). The terminal phalanges consist of well-developed claw-like unguals, and the same description given for the ungual phalanges of the manus also applies to the pes (electronic supplementary material, figure S8).

3.3.4. Cartilage

With the use of UV light, it was possible to distinguish between bone and cartilage, and even identify cartilaginous elements that are not readily visible under natural light. These elements include: tracheal and bronchial rings, calcified sternal ribs, epicoracoid and suprascapular cartilages, and all the unossified appendicular epiphyses (figure 5). On the right side of the neck, starting from the fifth cervical vertebra—at the beginning of the bend in the neck—a set of tracheal half-rings are visible under UV light (figure 5c). Unlike Pontosaurus kornhuberi, where complete rings are preserved, in MPUR NS 161 we only see incomplete rings, which appear as narrow rod-like fragments of cartilage compressed against the cervical vertebrae. The rings disappear next to the eighth cervical, where the cartilaginous portion of the pectoral girdle covers up most of the right side of the anterior region of the trunk. Under UV light, beneath both the ribs and vertebrae of the anterior trunk, several cartilaginous elements are observed (figure 5d). These include both the cartilaginous sternal ribs and the bronchial rings. The bronchial rings are well exposed along the right side of the trunk: they appear as tiny rod-like fragments among the anterior dorsal ribs, and extend anteriorly up to a point just posterior to the right coracoid. Four pairs of cartilaginous sternal ribs, each formed by at least two segments, can be recognized, and are especially evident on the right side of the body. The sternal ribs can be distinguished from the other trunk ribs because of their greatly expanded distal ends. The third cartilaginous rib still retains its articulation with one of the dorsal ribs (articulating with the 13th presacral vertebra). On the left side of the trunk, close to the proximal head of the left humerus, another broad cartilaginous element was UV-illuminated, which we interpreted as the suprascapula (figure 5a,b). The element is trapezoidal in shape, with the distal margin greatly expanded. Under UV light, the epicoracoid cartilage appears preserved in connection with the coracoid on the right side, and surrounds most of the scapula (figure 3a,b; electronic supplementary material, figure S4a). This cartilage was crushed underneath some trunk ribs and the rest of the pectoral girdle; it is hard to tell if the right suprascapula is preserved; however, the cartilaginous portion visible along the anterior margin of the scapula is quite extensive, so it is likely that the epicoracoid cartilage and suprascapula were in contact. With regard to the appendicular skeleton, all the epiphyses are unossified: the lack of ossification becomes more evident when the bones are exposed to UV radiation and the epiphyses are emphasized in a much brighter white colour, in comparison to the diaphysis, indicating a contrast in density and mineralization (figures 1–4; electronic supplementary material, figures S4–S7).

3.3.5. Integument

Different types of squamation are recognizable together with the skeletal remains. Even accounting for some post-mortem flattening of the body during compaction of the sediments, the scales appear to be preserved in their original position (figures 1 and 6). However, the external morphology of the scales is not preserved anywhere along the body, suggesting that it is not the outer layer of the epidermis (keratinized stratum corneum) that is preserved, but rather an inner layer of the integument, resulting in the reproduction of the scale outlines (here referred to as ‘scale ghosts’) and not the actual epidermal scales.

Deeply imbricated, small sub-circular scale ghosts are visible on both sides of the specimen next to the trunk and limbs (figure 6a–e). Between the right forelimb and the trunk, the permineralized integument follows the bend of the anterior portion of the skeleton, so that the scales look compressed against each other, highlighting the position of the original body outline. These small scales must have covered the sides and part of the belly of the animal, at least from the pectoral girdle to the sacral region (figure 6c). Larger diamond-shaped scales are preserved between the trunk ribs, and in particular on the left side of the posterior trunk region (figure 6c,d). These scales most probably covered the dorsal region of the body, because they appear to overlap the ribs (where present) and are abruptly interrupted where the bone is missing leaving only natural moulds. Diamond-shaped scales are also present in lateral view along the tail, where in the ventral region there are at least four rows of rhomboidal scales before the beginning of the broader, transversely expanded, subcaudal scales. Dorsal caudal scales are also preserved, but are not as clear as the subcaudals (figure 6g). Impressions of rhomboidal scales are clearly visible above the neural spines of the caudals in lateral view (figure 6f), suggesting the presence of a fin-like dorsal expansion along the top of the tail, probably similar to that of modern sea snakes and sea kraits (e.g. Hydrophis platurus and Laticauda colubrina), or the water monitor, Varanus salvator. One of the most important features characterizing the new marine lizard is the presence of transversally expanded scales, visible along the ventral margin of the last preserved caudal vertebra, where the tail is exposed in left lateral view (figures 1, 6g,h and 7). Among extant squamates such scales are typical of some snakes [55], while among fossils a similar squamation has been found in another basal pythonomorph, Pontosaurus kornhuberi [6,7]. Lee & Scanlon [55] and Lee [2] refer to these scales as subcaudal scales—or simply subcaudals—and in snakes they can be present as singles (i.e. one row) or pairs (i.e. two adjacent rows). In MPUR NS 161, either the tail is exactly in the left lateral view, with the scales visible along the ventral edge representing part of the right-side counterparts (like the paired condition in snakes), or due to twisting and compression of the soft tissues—as suggested by the irregularity of the tail base outline—the long scales represent complete, transversally expanded ventral scales (like the single condition in snakes). According to this second interpretation, the partially exposed scale ghosts along the ventral edge would belong to the other flank (right) of the tail (figures 6g,h and 7). The subcaudals in MPUR NS 161 are anteroposteriorly shorter than the caudal centra, and the length of one centrum corresponds to that of about two scales, similar to Pontosaurus kornhuberi; in snakes this ratio is highly variable. The scale ghosts are preserved owing to permineralization of the integument, or as impressions in the sediment. Being only the outline of the scales present, it is difficult to determine if the scales were smooth or keeled, as in the mosasauroids Tylosaurus proriger and Ectenosaurus clidastoides, or in Pontosaurus kornhuberi [6,56,57]. In MPUR NS 161, the most complete scales are found in the posterior trunk region and around the hindlimbs (figure 6c–e), but even in these cases the permineralization of the soft tissue does not allow full assessment of their original external morphology. Finally, in some parts of the specimen there are extensive portions of permineralized soft tissues where no distinct squamation is recognizable. Here the scale ghosts are not discernible, and there is no fibre-like arrangement either, suggesting that no mineralized muscle fibres are present: what is exposed may be the dermis (with the epidermal layer being degraded) or even the superficial fascia (or hypodermis), which lies between the integument and the external musculature (figure 6a,b).

3.3.6. Muscles

Portions of permineralized epaxial and hypaxial musculature are preserved along the trunk and tail (figures 1, 5e,f and 8). Collagen fibres and muscle bundles are visible around the pelvic girdle and in close association with the anterior caudal vertebrae. To assess the identity of the muscles in MPUR NS 161, we made comparisons with studies on the musculature of several squamates [54,58–62]. On the right lateral posterior region of the trunk, where the body curves, small portions of permineralized muscles are preserved between the last dorsal ribs (figure 5e,f), with the fibres being obliquely oriented relative to the vertebral column. These muscles must have been part of the more internal layers, because they attach to the lateral surface of the vertebral centra and seem to arise from the ventrolateral aspect of the trunk (hypaxial musculature), and are overlapped by both ribs and vertebrae. Based on their position, they may be part of the musculus (m.) obliquus internus, or of the m. transversus abdominis [60,62]. More muscle tissue is visible along the anterior margin of the ischia: these muscle fibres are so well preserved that the single myomeres can be easily distinguished with the naked eye (figures 3c,d and 8a–c). Considering that the hips are visible in medial view, and that the muscles appear to attach to the anterolateral surface of the ischia, most probably they represent a mineralized portion of the m. puboischiofemoralis externus, which usually originates from the lateral surface of the ischium and inserts onto the proximal portion of the shaft of the femur [54,58]. Along the anterior caudal region, before the tail bend, a large portion of permineralized muscle tissue is preserved on the left side of the vertebral column (figure 8d–g; electronic supplementary material, figure S9). This part of the tail is exposed in dorsal view, and according to the relationship between vertebrae and soft tissues, and the changes in orientation and organization of the muscle fibres, it is possible to identify at least two different types of fascicles probably belonging to different epaxial muscles: (i) anteroposteriorly oriented, thin and tubular-like fibres overlapping the transverse processes of some caudal vertebrae, interpreted as part of the m. transversospinalis (figure 8d,f,g); (ii) bundles of broad and flat muscle sheets, positioned more laterally than the previous type (further away from the vertebrae), and oriented obliquely relative to the long axis of the body, interpreted to be part of the m. iliocaudalis [54,61,62] (figure 8d,e).

3.3.7. Gastric and gut contents

In the posterior trunk region, near the transition to the sacrum, a large mass of permineralized soft tissue is preserved that is not observable under natural light (figures 1 and 5e,f). Under UV light, some parts of this mass acquire a pink-to-purple colouring, typically assumed by the soft tissues, but most of the mass remains white, indicating the large presence of bony material in the gut (figure 5f). Several tiny, rod-like fragments of bones are visible under UV light, and although their identity cannot be clearly assessed, this suggests that the animal was feeding on small vertebrates (e.g. fish). More anteriorly, on the left side of the mid-trunk, a small bone is visible between the dorsal ribs of the specimen (figure 6d): the small and slightly recurved element does not seem to be consistent with the rest of the skeleton, because its shape and size do not match those of any other bones from this body region, and it is clearly overlapped by a trunk rib. Considering its position in the trunk, and its etched surface, we can confidently interpret this element to be a partially digested bone (possibly from a fish) that was present in the stomach at the time of death.