A New Titanosaurian Braincase from the Cretaceous “Lo Hueco” Locality in Spain Sheds Light on Neuroanatomical Evolution within Titanosauria

Despite continuous improvements, our knowledge of the neurocranial anatomy of sauropod dinosaurs as a whole is still poor, which is especially true for titanosaurians even though their postcranial remains are common in many Upper Cretaceous sites worldwide. Here we describe a braincase from the uppermost Cretaceous locality of ‘‘Lo Hueco” in Spain that is one of the most complete titanosaurian braincases found so far in Europe. Although the titanosaurian Ampelosaurus sp. is known from the same locality, this specimen is clearly a distinct taxon and presents a number of occipital characters found in Antarctosaurus and Jainosaurus, which are approximately coeval taxa from southern Gondwana. The specimen was subjected to X-ray computed tomographic (CT) scanning, allowing the generation of 3D renderings of the endocranial cavity enclosing the brain, cranial nerves, and blood vessels, as well as the labyrinth of the inner ear. These findings add considerable knowledge to the field of sauropod paleoneuroanatomy in general and titanosaurian endocast diversity in particular. Compared with that of many sauropodomorphs, the endocast appears only slightly flexed in lateral view and bears similarities (e.g., reduction of the rostral dural expansion) with Gondwanan titanosaurians such as Jainosaurus, Bonatitan, and Antarctosaurus. The vestibular system of the inner ear is somewhat contracted (i.e., the radius of the semicircular canals is small), but less so than expected in derived titanosaurians. However, as far as the new specimen and Jainosaurus can be contrasted, and with the necessary caution due to the small sample of comparative data currently available, the two taxa appear more similar to one another in endocast morphology than to other titanosaurians. Recent phylogenetic analyses of titanosaurians have not included virtually any of the taxa under consideration here, and thus the phylogenetic position of the new Spanish titanosaurian—even its generic, let alone specific, identification—is not possible at the moment. Nevertheless, both the braincase osteology and the endocast morphology suggest that the specimen represents a derived titanosaurian that presumably branched further from the base of Lithostrotia, potentially even near Saltasauridae, comparable in evolutionary terms with Jainosaurus.


Introduction
During the construction of a high-speed rail track connecting Madrid with Valencia in 2007, an exceptional fossil site was discovered in the Villalba de la Sierra Formation at a locality named "Lo Hueco," near the village of Fuentes, Castile-La Mancha, Spain. Over the course of several months, a large-scale excavation yielded thousands of specimens of plants, invertebrates, and vertebrates of late Campanian-early Maastrichtian age [1]. Together with eusuchian crocodiles, titanosaurian sauropods represent the largest part of the fossil record at "Lo Hueco." Among this rich sauropod collection, only a few cranial elements were collected: two braincases and a small number of isolated teeth. These skull elements are extremely important remains as the cranial anatomy of titanosaurians is currently poorly known globally, except for a few remarkable exceptions (see [2][3][4][5][6][7]). The same holds true of the neuroanatomy of titanosaurians even though much progress has been achieved lately [8][9][10]. The aim of the present paper is to present a detailed osteological description together with a paleoneurological investigation of the most completely preserved sauropod braincase from "Lo Hueco", which is also one of the more complete from the European Upper Cretaceous as a whole. The other titanosaurian braincase from "Lo Hueco" was described previously and referred to Ampelosaurus sp. [9]. Although both braincases obviously pertain to titanosaurians, they are morphologically distinct and clearly belong to different taxa.
Sauropods show a wide disparity in skull morphology and titanosaurian braincases are easily distinguished from those of non-titanosaurian taxa (see e.g., [11,12]). The osteology of the specimen, MCCM-HUE-1667, will be contrasted with that of all the European Late Cretaceous (Campanian-Maastrichtian) titanosaurian specimens reported so far that are sufficiently complete to be informative for comparative purposes, based on the literature and, where necessary, unpublished photographic material and personal observations ( Table 1). The paleoneurology of MCCM-HUE-1667 will be compared with all the other titanosaurians for which the endocast (either physical or digital) of the cranial cavity and/or osseous labyrinth has been described ( Table 2).
To produce a three-dimensional reconstruction of the endocast of the cranial cavity and endosseous labyrinth of the inner ear, the specimen was subjected to computed tomographic (CT) scanning on an Yxlon CT Compact (Yxlon International, Hamburg, Germany) with a voltage of 180 kV and a current of 2.8 mA. The in-plane pixel size was 0.18 mm, with an interslice spacing of 0.20 mm. Data were output from the scanner in DICOM format and then imported into Avizo 7.1 (VSG, Burlington, MA, USA) for digital extraction of the anatomical features of interest. The scan data were difficult to work with due to there being little contrast between the bone and matrix and apparently some scanner-induced artifact that produced banding; nevertheless, the scan data were adequate to segment the major features of the cranial endocast and endosseous labyrinth. The resulting 3D models were then imported into the 3D modeling software Maya 2012 (Autodesk, San Rafael, CA, USA) for removal of segmentation artifacts, final rendering, and generation of the illustrations. The 3D PDF in the Supporting Information (S1 Fig) was generated by exporting the 3D models from Maya into Deep Exploration 5.5 (Right Hemisphere, San Ramon, CA, USA) and then Adobe Acrobat 9 Pro Extended (Adobe Systems Inc., San Jose, CA, USA). The 3D PDF can be viewed with any computer on Description Osteology MCCM-HUE-1667 was discovered in the lowest part of the fossiliferous succession of "Lo Hueco" (G1; see [1] and references therein), in close association with cervical vertebrae of a partial sauropod skeleton (Fig 1). It has been slightly sheared during early diagenesis, fractured by the shrink-swell cycles of clay minerals, and is missing the left caudolateral portion, due to breakage in the field by the bucket of a mechanical digger. As in many other braincases of sauropods in general and titanosaurians in particular, the sutures are generally indistinct. In this specimen, every bone was fully ossified and most elements are fused together. Numerous fractures hinder observation of the braincase surface, but the specimen is otherwise well preserved. Overall, the braincase is short and deep. It is 101.9 mm long rostrocaudally, 127.4 mm tall dorsoventrally, and the better-preserved right side is 67.9 mm wide transversally (Figs 2 and 3, S1 Fig).
Prefrontal-Frontal. The frontal is a rostrocaudally short bone (Figs 2A, 2C, 3A and 3C, S1 Fig). We interpret a ventrally curved process (more complete on the left side) that projects Fig 1. Photograph of the indeterminate titanosaurian sauropod skeleton from the Cretaceous of Fuentes, Spain, with which the braincase MCCM-HUE-1667 was associated. The neurocranium had already been removed when the photo was taken, but the inset (arrow) shows how it appeared in the field, after partial coating with gauze. from the rostrolateral corner as an articulated prefrontal (Figs 1A-1D, 3A, 3B and 3E, S1 Fig). If this process had been preserved on the right side too, the combined prefrontals-frontals would have a U-shaped rostral margin. The dorsal surface of the frontal is not flat, but corrugated with a channel that runs from the caudomedial area toward the rostrolateral corner. It is probable that the rostromedial fraction of the supratemporal fenestra was formed by the caudal portion of the frontal (although the lack of clear sutures prevents any certainty in this regard). Ventrally, the frontal is markedly concave as it forms the orbital roof (Figs 2B, 2C, 3B and 3C, S1 Fig). The prefrontal projection is buttressed so that it has a triangular outline in cross-section. The transverse breadth of the combined prefrontal-frontal is 57.0 mm at most and extends 41.8 mm rostrocaudally.
Parietal. Because of a lack of evidence for an interparietal suture, this bone is treated here as a single entity. The parietal is a rostrocaudally short, dorsoventrally shallow but transversally wide bone ( : figs 1, 2, S1, S2, S3) and, indeed, other taxa. The latter specimen, however, does not show the flat surface that is present at the top of the parietal in MCCM-HUE-1667 but instead a concavity. Regarding the depth of the caudal concavity formed by the transverse nuchal crest (nuchal fossa), the parietal in MCCM-HUE-1667 is closer in morphology to that in FAM 03.064 ( [12]: figs 2, 5a) than to that of Ampelosaurus atacis. The parietal in MCCM-HUE-1667 is strikingly different from that in FGGUB 1007 ( [20]: fig. 15). In the latter, it is proportionally larger, does not bear obvious transverse nuchal crests, and shows laterally, near its suture with the frontal, peculiar small and ovate protuberances.
Supraoccipital. The supraoccipital is small and rounded, forming, along with the parietals, a low sagittal nuchal crest ( Otoccipital. This bone results from the early co-ossification of the exoccipital and opisthotic, which is a common feature of adult archosaurs and other diapsids (e.g., [24]: 7). The foramen magnum is dorsoventrally elongate (20.9x29.5 mm) and, as a result, it is bordered almost entirely by the otoccipitals, in that both the supraoccipital dorsally and the basioccipital ventrally contribute much less (Figs 2E and 3F, S1 As revealed by the CT scan data, the fenestra cochleae (= fenestra pseudorotundum, foramen perilymphaticum) opens into the rostral wall of metotic foramen, as is typically the case in sauropods. The oval window (fenestra ovalis = fenestra vestibuli) opens at the dorsal extremity of the depression, where it is bordered by the otoccipital and prootic; this area of the fossil itself is not particularly clear via visual observation, but these features are visible in the CT scan data ( Fig 3D, S1 Fig). The paroccipital process is aliform, much taller dorsoventrally than long rostrocaudally. It strongly arches ventrally and curves medially back towards the braincase at it distal end (Figs 2E and 3F, S1 Fig). The dorsolateral notch in the paroccipital process for the posttemporal foramen is relatively broad and shallow (Figs 2E and 3F, S1 Fig). The base of the paroccipital process is embedded in matrix rostrally so that it is difficult to assess the configuration of the crista otosphenoidea (= crista prootica), but the CT scan suggests that the crest is comparable to that of other sauropods. There is a distinct crest on the rostral surface of the distal portion of the paroccipital process, which appears dorsally near the mid-line but merges with the medial border of the bone ventrally (Figs 2D and 3E, S1 Fig).
In contrast with the condition in MCCM-HUE-1667, the foramen magnum of Ampelosaurus atacis ( [18]: fig. 4.2) and Ampelosaurus sp. ([9]: figs 1, 2, S1, S2, S3) is not dorsoventrally elongate. In these taxa, the paroccipital process is not vertical and hanging laterally as in MCCM-HUE-1667, but rather is more dorsoventrally oblique. It is also oriented laterally in caudal view, at least proximally, whereas that in MCCM-HUE-1667 arches strongly ventrally from the proximal portion ( [23] believed they could perceive the limit between the exoccipital and the opisthotic) near the foramen magnum. This protuberance was compared with that seen on the Dongargaon sauropod braincase ( [26]: fig. 2) and correctly interpreted similarly as an articular facet for the proatlas (the specimen from India is now referred to as of the Isisaurus morph [27,28]). Two small depressions lateral to the protuberance were also noted. No such protuberance and depressions associated with the proatlas are noticeable on the otoccipital of MCCM-HUE-1667 (Figs 2E and 3F, S1 Fig). The paroccipital process in the unnumbered MNHN partial braincase ( [19]: fig. 2, pls 5-6) is not as strongly pendant and vertically oriented as in MCCM-HUE-1667. Admittedly, both specimens exhibit some deformation but not to a sufficient extent to account for the difference in the two specimens. The paroccipital process of the unnumbered MNHN specimen also dilates distally, whereas that of MCCM-HUE-1667 tapers distally (Figs 2D, 2E, 3E and 3F, S1 Fig).
Basioccipital. The basioccipital is a relatively dorsoventrally tall but rostrocaudally short bone (Figs 2E and 3F, S1 Fig). It probably constituted most of the occipital condyle (Figs 2E and 3F, S1 Fig). The latter has been damaged through excavation work, although it was likely hemispheric in morphology and transversally wider than the foramen magnum, as seen in a variety of sauropods. A subcondylar recess separates the occipital condyle from each basal tuber, which progressively arises with an irregular outline from the ventrolateral border of the caudal surface of the basioccipital (Figs 2E and 3F, S1 Fig). The basal tubera are relatively close to one another (Figs 2E and 3F, S1 Fig), which is consistent with the configuration in most sauropods, in which the basal tubera are close to one another and best seen in a caudal view of the braincase (Spinophorosaurus nigerensis is a notable exception to this pattern; [29]: figs 1-3, S1, S2, S3). The basioccipital is perforated caudolaterally by the hypoglossal canal ( Fig 3D, S1 Fig).
The bone is 55.3 mm tall in the midline and 52.8 mm wide at the level of the basal tubera.
Whereas the occipital condyle was probably knob-like in MCCM-HUE-1667, that of Ampelosaurus atacis ( [18]: fig. 4.2) and Ampelosaurus sp. ([9]: figs 1, 2, S1, S2, S3) may have been distinctly wider transversely than dorsoventrally tall. Although it cannot be confirmed, the entire basioccipital of Ampelosaurus atacis and Ampelosaurus sp. may have been less tall dorsoventrally than that in MCCM-HUE-1667. In Lirainosaurus astibiae ( [30]: figs 2-4), the basioccipital forms a generally hemispheric occipital condyle with a short neck, as was possibly the case in MCCM-HUE-1667. However, in that taxon the surface of the basioccipital between the occipital condyle and the basal tubera is not excavated as in MCCM-HUE-1667 and lacks the subcondylar recess or foramen. In Lirainosaurus astibiae, the basal tubera are rounded and prominent and, therefore, unlike those in MCCM-HUE-1667. Díez Díaz et al. [30] noted that the distal surface of each basal tuber is pierced by a foramen and regarded this configuration as autapomorphic of Lirainosaurus astibiae. Nevertheless, it is probably homologous to the notch observed at this place in Jainosaurus septentrionalis ([28]: fig. 6; [31]: fig. 11). In the unnumbered MNHN braincase ( [19]: fig. 2, pls 5-6), the occipital condyle appears to have had a relatively longer neck than that in MCCM-HUE-1667. Similarly, the basal tubera are much more rounded and prominent caudally in the unnumbered MNHN braincase than in MCCM-HUE-1667 and the zone of the basioccipital between the occipital condyle and the basal tubera is more excavated in the former specimen than in the latter.
Parabasisphenoid. As is generally the case in sauropods, as well as in most dinosaurs and crocodilians and indeed many other reptiles [32], the basisphenoid and parasphenoid are indistinguishably fused with each other (at least in adults). They are, therefore, described here as a single unit, the parabasisphenoid. It forms the ventral extremity of the braincase, just beneath the basal tubera (Figs 2E and 3F, S1 Lateral to this base of the parasphenoidal rostrum, the parabasisphenoid is slightly concave. The abducens canal opens on the dorsalmost extension of the rostral surface of the parabasisphenoid, very close to the ventral limits of the laterosphenoid, and about equidistant from the ventral midline and lateral edge of the braincase. Neither of the two external openings of the internal carotid artery's cerebral branch canals are visible in the fossil specimen, but the CT scan data shows that they open lateral to the base of the parasphenoidal rostrum on the rostral surface of the parabasisphenoid (Fig 3D, S1 Fig). As preserved, the parabasisphenoid is only 13.4 mm long and 40.9 mm wide at the level of the base of the basipterygoid processes.
The body of the parabasisphenoid in MCCM-HUE-1667 is relatively much shorter rostrocaudally than that of Lirainosaurus astibiae ( [30]: figs 3, 4C, 6, 7B, 8B). Based on their proximal sections, the basipterygoid processes were possibly more robust in Lirainosaurus astibiae than in MCCM-HUE-1667. The parabasisphenoid of Lirainosaurus astibiae is pierced by well-characterized foramina for the entrance of the cerebral branch of the right and left internal carotid arteries ventrally and the exit of the right and left abducens nerve (cranial nerve [CN] VI) rostrally. The preservational state of MCCM-HUE-1667 obscures the location of these foramina externally, but their positions can be determined by the internal canals visible in the CT scan data (Fig 3D, S1 Fig). In the poorly preserved braincase described by Le Loeuff et al. [23], these authors assumed the presence of a strong basipterygoid process, but this is actually a remnant of the right basal tuber: nothing from the basisphenoid and parasphenoid seems to have been preserved in that specimen. The parabasisphenoid of the unnumbered MNHN specimen ( [19]: fig. 2, pls 5-6) does not show the well-defined triangular concavity visible on the ventral surface of this bone complex in MCCM-HUE-1667. Comparisons of the parabasisphenoid of MCCM-HUE-1667 with that of FAM 03.064 ( [12]: figs 3-5) are made difficult by the incomplete preservation of the latter specimen ventrally. In the peculiar braincase reported by Weishampel et al. ([20]: fig. 15), the basisphenoid and parasphenoid form a much rostrocaudally longer complex than in MCCM-HUE-1667. In comparison with the braincases of titanosaurians illustrated by Curry Rogers and Wilson [25: fig. 6] in caudal views, the subcondylar region of the basicranium in MCCM-HUE-1667 most closely resembles that of Jainosaurus septentrionalis in both its overall relative proportions and the shape of the basal tubera.
Prootic. The prootic is largely concealed because the adductor chamber is partially filled with matrix on both sides. The chamber is very short rostrocaudally (as is the supratemporal fenestra) and the prootic is even shorter. It inserts as a wedge between the otoccipital caudally and the laterosphenoid rostrally and it is limited rostrodorsally by the frontal. The left side shows that the ventral vertex of the prootic is pierced by the large trigeminal foramen (for CN V), which is caudally bordered by the crista otosphenoidea (Fig 3C and 3D, S1 Fig). The singular foramen indicates an extracranial position of the trigeminal nerve, as in most sauropods (see e.g., [33]). Its outline is somewhat like a triangle with convex sides pointing ventrally ( Fig  3C and 3D In between the latter bone and the capitate process, it constitutes the rostral limit of the trigeminal foramen (Fig 3C and 3D, S1 Fig). At the level of the trigeminal foramen, the laterosphenoid extends rostrocaudally only 7.1 mm, but it is about 40 mm tall.
Orbitosphenoid. The orbitosphenoids have fused completely into a single, median unit, which constitutes the rostroventral extremity of the braincase (Figs 2B-2D and 3B-3E, S1 Fig). This element forms an acute ventral carina (Figs 2B, 2D, 3B and 3E, S1 Fig). As a result, it presents a short and wide Y-shape in transverse section. The carina is in line with the base of the parasphenoidal rostrum described above (Figs 2D and 3E, S1 Fig), and indeed the cartilaginous interorbital septum attached to both these structures. The orbitosphenoids form the floor of the olfactory aperture (Figs 2D and 3E, S1 Fig), which is roofed by a rostral rising of the frontals, and enclose the laterally directed optic foramina (Fig 3D, S1 Fig). They also border rostrally the oculomotor and trochlear foramina (for CN III and CN IV, respectively), which are oval and slit-like, respectively (Fig 3D, S1 Fig). The average diameter of the ocumolotor foramen (5.4 mm) is in between that of the relatively large optic aperture and that of the diminutive trochlear opening (Fig 3D, S1 Fig). The united orbitosphenoids are 31.7 mm wide transversely at the middle part and about 24 mm tall dorsoventrally. Comparisons of the orbitosphenoid of MCCM-HUE-1667 with that of other European titanosaurians is especially difficult as this bone is often poorly preserved or even lacking altogether.

Neuroanatomy
Although Marsh ([34]; see also [35]: pl. 8; [36]: fig. 34) initiated the paleoneurological study of sauropod dinosaurs, it was not until more than three decades later that a physical endocast made from a specimen of this group was figured in detail ( [37]: fig. 16A). The first paleoneurological data on titanosaurian sauropods were provided by Huene ( [21]: pl. 28) from the braincase foramen pattern of Antarctosaurus wichmannianus. A few years later, Huene and Matley ( [38]: fig. 6) figured for the first time a "braincast" of a titanosaurian, "Antarctosaurus" septentrionalis. In the present case, the CT scan data permitted a comprehensive rendering of the cranial endocast and endosseous labyrinth (Figs 4-7, S1 Fig). The completeness of the braincase allowed reconstructing the course of every cranial nerve of the left side as they traverse the braincase wall, which has not been possible so far in most sauropod taxa. Below, we refer to the reconstructed digital casts of bone-bounded spaces that housed soft-tissue structures as if they were the structures themselves (e.g., "trigeminal nerve" rather than "digital cast of trigeminal foramen").
Brain endocast. The olfactory tract is extremely short (Figs 4, 5C and 5D, S1 Fig), which is related to the small extent of the frontal rostrocaudally (Fig 3D, S1 Fig). As far as it can be observed, the olfactory bulb is also short (Figs 4, 5C and 5D, S1 Fig). There is little doubt that the olfactory region of the nasal cavity, where the olfactory nerve (CN I) originated, was situated very caudally. This is consistent with shortened (or highly modified) nasal bones and, therefore, greatly retracted or broadened external nares as in other sauropods [39]. The same European Titanosaurian Paleoneurology configuration of the olfactory lobe is observed in other titanosaurians ([8]: figs 1-6; [9]: figs 3, S1, S2, S3; Fig 6).
The longitudinal axis of the olfactory tracts is almost in line with the medulla oblongata ( Fig  4, S1 Fig). This stands in sharp contrast with the condition in the basal titanosauriform Giraffatitan brancai, in which the pontine and cerebral flexures of about 50°each result in an endocast that appears sigmoid or "flexed" in lateral view ( [40]: figs 1, 2A, B). An endocast that is sigmoid to varying degrees in lateral view is, in fact, widely distributed in non-titanosaurian sauropods  fig. 6.9)), which suggests that a straighter (less sigmoidal or "flexed") morphology, as seen in MCCM-HUE-1667, is the derived state. The total volume of the endocast is 41.63 cm 3 , over 6% of which is devoted to the pituitary fossa.
Cranial nerves. The optic nerve (CN II) is short but stout (Figs 4 and 5A, S1 Fig). Each leaves the braincase through its own aperture (Fig 3D, S1 Fig). It compares well in most respect to the conformation in the South American taxa examined by Paulina Carabajal ([8]: figs 2A, 3, 5A, 6A, C, 8B). As in most titanosaurians, the optic nerve canal is directed almost directly laterally (Figs 4, 5A and 5C, S1 Fig) rather than rostrolaterally as in most other dinosaurs and more basal sauropodomorphs (see e.g., [42]), presumably reflecting the retraction and telescoping of the structures caudal to the expanded narial region, rotating the orbits far laterally and compressing the adductor chamber.
In MCCM-HUE-1667, the course of the abducens does not brush past the pituitary fossa as it does in Ampelosaurus sp. ( [9]: figs 3, S1, S2, S3). The more medial path seen in the latter is closer to the plesiomorphic condition in which the nerve penetrates the pituitary fossa (see e.g., the basal titanosauriform Giraffatitan brancai ( [40]: fig. 1)).
The facial nerve (CN VII) emerges from the brainstem immediately caudal to the trigeminal nerve (Figs 4 and 5C, S1 Fig) in close association with the vestibulocochlear nerve (CN VIII), as in other amniotes. The facial nerve runs alongside the trigeminal nerve and leaves the braincase just caudally to it, passing through a canal located fully within the prootic on the caudal side of the crista otosphenoidea (Fig 3D, S1 Fig).
As its name implies, the vestibulocochlear nerve (CN VIII) has two branches. The vestibular branch enters the vestibular labyrinth in the region of the ampulla of the rostral semicircular canal, which is typical in extant diapsids, as well. The more ventral cochlear branch could not be traced in the CT scan data, but is assumed to have been present.
The glossopharyngeal and vagoaccesory nerves (CN IX-XI) are combined together with other structures into a slightly oblique, dorsoventrally elongated, and remarkably slender metotic canal (Figs 4, 5B and 5C, S1 Fig). As noted above, the perilymphatic duct opens from the labyrinth into the rostral wall of the metotic canal. Therefore, the course of the glossopharyngeal nerve (CN IX), as in most other titanosaurians ([8]: figs 2, 3, 5, 6; [28]: fig. 7), does not run through a separate aperture. Ampelosaurus sp. ( [9]: figs S1, S2, S3) has a slightly altered arrangement in which the glossopharyngeal nerve appears to have diverged from the rest of the contents of the metotic canal at its the lateral terminus. A narrow canal caudal to the metotic foramen in a fragmentary braincase from Argentina of presumed titanosaurian affinity (MPCA-PV-80) was identified by Paulina Carabajal ([8]: fig. 7A) to be an independent accessory nerve (CN XI). However, in sauropsids the accessory nerve combines intimately with the vagus nerve (CN X) and, indeed, it has long been remarked that in modern archosaurs these two nerves leaves the cranial cavity as one ( [50]: 45). The course of this canal is parallel to a hypoglossal nerve (XII) and we speculate it could actually be a reduced rostral branch of the hypoglossal.
The hypoglossal nerve (CN XII) has a single root (Figs 4, 5B and 5C, S1 Fig) as in many other titanosaurians [8,9] fig. 4A, B) also identified only one hypoglossal root in the dicraeosaurid diplodocoid Amargasaurus cazaui, but we conservatively reinterpret the duct infilling they designed as an accessory nerve as a rostral root of the hypoglossal. The hypoglossal nerve of MCCM-HUE-1667 may be homologous with the most caudal of the hypoglossal roots of non-titanosaurian sauropods, which is usually the largest (see [42]: figs 6.8C, 6.9C), although it is also very possible that there was no real variation in the actual nerve roots in life, but rather only variation in whether those nerve branches passed through one or two bony canals. The single canal in MCCM-HUE-1667 emerges from the medulla oblongata and extends laterally and slightly ventrally and caudally (Figs 4, 5B and 5C, S1 Fig). In contrast with other titanosaurians ([8]: figs 2, 3, 5, 6; [9]: figs 3, S1, S2, S3; [28]: fig. 7; Fig 6), the root of the hypoglossal nerve is very close to the metotic group in MCCM-HUE-1667. We suggest that this may be an autapomorphic feature.
Endocranial vasculature. The cerebral carotid arteries (which enter the ventral end of the pituitary fossa remarkably close to one another) are by far the main vascular structures of the endocast (Figs 4, 5A and 5B, S1 Fig). However, CT data also reveal the existence of a few small veins draining the forebrain that course through the braincase wall, but were not reconstructed. One of these leaves the neurocranium on the laterosphenoid-orbitosphenoid suture. Although more gracile, it is clearly homologous to the orbitocerebral vein that is present in some sauropods like Diplodocus longus ( [42]: fig. 6.9) and Camarasaurus lentus ( [42]: fig. 6.8). In addition, two very delicate veins appear to run ventrolaterally from the endocast on the laterosphenoidfrontal interface or approximately so. Whether they are diploic veins, as in Diplodocus longus ( [42]: fig. 6.9A, D, F), or not is unclear. It is unquestionable that the main venous drainages run off through apertures shared with nervous structures (e.g., foramen magnum, trigeminal foramen). In Bonatitan reigi, a "middle cerebral vein", which extends essentially dorsolaterally, is present well dorsal to the trigeminal nerve ( Inner ear. The left endosseous labyrinth could be reconstructed (Fig 7, S1 Fig). The steplike transition between the dorsal two-thirds and the remaining ventral portion (Fig 7A and  7B, S1 Fig) probably reflects the functional separation of the inner ear into two main regions (vestibular system and lagena), as discussed by Witmer et al. ([42]: p. 80) in Diplodocus, Camarasaurus, and other sauropods. In fact, the vestibule-lagena transition should be approximately coincident with the tangent line that passes through the dorsal top of the oval window. Thus, the length of the lagena is 10.7 mm.
The vestibular system of the inner ear is somewhat contracted (i.e. the radius of the semicircular canals is small), but less so than in other titanosaurians (Fig 8), such as Bonatitan reigi  fig. 4A-C), is much shorter (Fig 8). Although our knowledge of the morphology of the vestibular system among sauropod taxa is still deficient, this equalization of the length of the two vertical semicircular canals may well turn out to be a titanosaurian synapomorphy. We have previously brought up the reasons behind the reduction of the semicircular canal radii of curvature in most sauropods [9,29,42], and quantitative analyses are underway. In birds, the slenderness and great length of the semicircular canals is related inter alia with their level of exposure to abrupt displacements during flight [53]. The reduction of the relative length of the semicircular canals in titanosaurians (an observation in line with the above-mentioned reduction of the floccular recess) and other sauropods might be linked to a restricted head rotation range.
The lagena (= cochlea) is slightly arched medially (Fig 7B, S1 Fig), following the curvature of the internal surface of the braincase cavity at this place. It is irregularly ellipsoid in cross section; its ventral extremity is blunt (Fig 7A and 7B, S1 Fig). The length of the lagena is relatively greater than in many other titanosaurians, being much longer than in CCMGE 628/12457 ( [10]: fig. 4A-C; Fig 8) and Jainosaurus septentrionalis ([10]: fig. 4D-F). The oval window is unevenly round in outline (Fig 7A, S1 Fig). It faces laterally and lies at the medial end of a relatively long duct, which reflects the thickness of the bony wall separating the labyrinth from the middle ear (Fig 7A and 7B, S1 Fig). At about the same height as the oval window, but more medially, the perilymphatic duct heads caudally (Fig 7B, S1 Fig), towards the metotic canal, which it reaches after a short course. From there on, it cannot be distinguished from the other structures that pass across the metotic foramen (see above). The perilymphatic duct is ellipsoid in outline as it enters the metotic group (Fig 7B, S1 Fig).

Discussion
Skeletal remains of sauropods are generally common fossils in the continental Campanian and Maastrichtian of Europe, from Spain to Romania, but braincases, which are delicate bone complexes, are rare components of this record. Even though the number of species represented remains unclear, our knowledge of the diversity of European latest Cretaceous sauropods has improved recently. For instance, no less than eight "forms," possibly representing as many species, have been distinguished in the Late Campanian-Late Maastrichtian interval of southwestern Europe based on femoral morphology [54]. This record, which is intriguing from a biogeographical and ecological perspective, includes in the Late Campanian alone Lirainosaurus astibiae [55], Ampelosaurus atacis [56], Atsinganosaurus velauciencis [57], and an indeterminate taxon. Even if neurocranial and postcranial elements are not found preserved together or in articulation for all these forms in further excavations, a pattern of co-occurrence should materialize in the long run, allowing the matching of dental, cranial, and postcranial elements.
Much progress has been accomplished recently regarding the phylogeny of basal titanosauriforms [58][59][60]. Unfortunately, this does not hold true for more derived taxa. This situation is exacerbated for both the European species, such as Lirainosaurus astibiae and Ampelosaurus atacis, as well as Jainosaurus septentrionalis, which are rarely taken into account in cladistic analyses. Thus, even the supertree depicting the phylogenetic relationships for titanosauriforms produced by Souza and Santucci [61], which is based on source trees selected to maximize the number of taxa in the analysis, does not include any of these three species as terminal taxa. As a result, there is no suitable recent phylogeny onto which the characters of MCCM-HUE-1667 could be optimized. Nonetheless, MCCM-HUE-1667 shows features that are informative for determining its relationships.
It has been proposed that the ratio of the mediolateral width of the basal tubera to that of the occipital condyle may present a useful phylogenetic character [62], a high ratio (1.5 or greater) possibly being synapomorphic of Lithostrotia [59]. Characters based on ratios with arbitrary cut-offs are problematic in some ways, and thus we do not want to place too much importance on it, but nevertheless, in MCCM-HUE-1667, this ratio is estimated to be about 1.88. This value is close to that of Jainosaurus (1.91; [62]: tab. 1), and may support lithostrotian affinities for MCCM-HUE-1667. Lithostrotia is the least inclusive clade containing Malawisaurus dixeyi and Saltasaurus loricatus ( [63]: 311) and, therefore, represent a large group of derived titanosaurians. However, MCCM-HUE-1667 also shows a number of characters that are not widely distributed in the data matrix of Curry Rogers [64] and may suggest more refined, less-inclusive groupings. For example, MCCM-HUE-1667 has a supraoccipital that broadly contacts the caudal border of the parietal as in Antarctosaurus and Jainosaurus, which is not true for Nemegtosaurus, Quaesitosaurus, Rapetosaurus, and others. Likewise, MCCM-HUE-1667 has a flat occiput similar to that of Ampelosaurus and Jainosaurus, but it is not the case in Malawisaurus, Nemegtosaurus, Quaesitosaurus, Rapetosaurus, Saltasaurus, and others. Curry Rogers [64] scored Antarctosaurus as having the occipital region rostrocaudally deep with the paroccipital processes oriented caudolaterally. However, the paroccipital processes of Antarctosaurus ( [21]: pl. 28 fig. 2; [22]: fig. 64 fig. e) extend much more laterally than caudally, conferring a fairly flat occiput to this taxon. In fact, in caudal view the basicranium of MCCM-HUE-1667, including the basal tubera and basipterygoid processes, resembles quite remarkably that of Antarctosaurus ( [21]: pl. 28 fig. 1; [22]: fig. 64 fig. a) and other titanosaurian specimens from Argentina, such as the indeterminate braincase (MUCPv-334) of possible Santonian age described by Calvo and Kellner ([65]: fig. 1). Furthermore, the resemblances between MCCM-HUE-1667 and Jainosaurus ( [28]: figs 3-7) are so marked, both in neurocranial osteology (e.g., shape and orientation of the paroccipital process) and endocast morphology (e.g., outline in lateral view), that a close phylogenetic proximity is suggested. Jainosaurus is now considered as pertaining to a clade of advanced titanosaurians together with other Gondwanan taxa such as Antarctosaurus, with which it shares a number of putatively derived characters such as a sinuous parietal-supraoccipital contact [28]. The latter character is also present in MCCM-HUE-1667. We surmise that MCCM-HUE-1667 would represent a derived titanosaurian probably allied to Jainosaurus and Antarctosaurus.
It is difficult to draw firm conclusions from the variety observed in the titanosaurian braincases found so far in the Upper Cretaceous of Europe, because of challenges associated with teasing apart interspecific from intraspecific variations, as well as the major problem of associating isolated braincases with other skeletal specimens. Caveats aside, analyzing and publishing these isolated specimens has the benefit of providing the basis for testing future hypotheses of association as more complete specimens are discovered. Undoubtedly, more than a few species dwelled that area at that time. Although the large number of titanosaurian elements from "Lo Hueco" (many of which are in articulation) are yet to be fully prepared and described, preliminary observations suggest that at least two species are present, which can be differentiated by the robustness of their skeleton and a variety of dental and postcranial characters (see in particular [66]). The examination of the neurocranial material ( [9]; this work) supports this view as several features of the new specimen (e.g., ellipsoid foramen magnum) clearly show that the differences between the two titanosaurian braincases from "Lo Hueco" are not a product of the certain postmortem deformation suffered by MCCM-HUE-1667, but instead that the latter does not pertain to the same species as Ampelosaurus sp. The braincases alone reveal that Ampelosaurus-like and Jainosaurus-like forms were present in the "Lo Hueco" area around the Campanian-Maastrichtian boundary. The latter might represent a more derived form than Ampelosaurus sp. if we are to judge from the degree of offsetting of the course of the abducens nerve. Indeed, the more medial route of this nerve in Ampelosaurus sp. compared to MCCM-HUE-1667 is closer to the primordial condition in titanosauriforms and other sauropods in which this nerve passes through the pituitary fossa.