The specimen consists of two strongly fused5 premaxillae. The bony surface is relatively well preserved with the original ornamentation for a total length of 19.2 cm and an 8.2 cm maximum height. The suture between them is visible on the dorsal surface (Appendix 1). Compared with the spinosaurins Spinosaurus and Irritator ‌(Sereno et al. 1998), premaxillae are relatively short and robust. The maximal width of the rosette in ventral view is 8.8 cm at the 3rd alveolus (Fig. 2) and becomes thinner and thinner posteriorly (5.4 cm at the 7th alveolus). There is a crest on the dorsal margin of premaxillae increasingly large posteriorly (Figs 1; 4A). On more complete premaxillae of spinosaurids such as Baryonyx (Charig & Milner 1986, 1997), Suchomimus ‌(Sereno et al. 1998) and Spinosaurus ‌(Sasso et al. 2005), a complex posterior margin of premaxillae can be observed in lateral view, consisting of a bifurcation of the bone in two distinct rami: a short one below and a longer one above, forming an elongated process articulating with nasal dorsoposteriorly. It lacks this posterior portion, however, on both Cristatusaurus specimens (MNHN.F.GDF365 and MNHN.F.GDF366). But on the right bone in lateral view, we can clearly notice a concavity on the posteroventral portion marking the articulation with the maxilla (Figs 1B; 7A), just after the 7th alveolus and below the short ramus (visible on the left maxilla of MNHN.F.GDF366, too). This structure is found in Baryonyx ‌(Charig & Milner 1997) and Suchomimus ‌(Sereno et al. 1998) and we know that the external naris are located close to this junction in these taxa, unlike Spinosaurinae (Fig. 20B, C). It must have been the same for Cristatusaurus. The distance between the tip of the muzzle and the 7th alveolus is proportionally short (13.7 cm) compared with “Spinosaurinae” taxa such as Spinosaurus ‌maroccanus (Fig. 13A).

As noted by Kellner & Campos (1996), there are two bony ridges in the ventral side, which in this specimen are apparently well connected with the premaxillae. Because of the incomplete preparation of the specimens, the authors made the hypothesis these ridges are the rostral region of vomers. By examining this structure in other spinosaurids, it is clear that it is not the anterior segment of the vomers, but a part of the premaxillae themselves. For example, in Cristatusaurus ‌lapparenti (MNHN.F.GDF366), there is no bony suture between premaxillae and these processes, nor in Spinosaurus ‌maroccanus (Figs 2; 13A). Taquet & Russell (1998) assumed that these ventral bony processes did belong to premaxillae (by addition of a contact with the vomers further back, below the anterior part of maxillae as in SpinosaurusMNHN.F.SAM124).

The foramina identified by combining the external observations and the tomographic results amount to 49 on the right premaxilla and 42 on the left. Their diameter varies from 1 to 8 mm (for an average of 3 mm). Their density increases particularly on the first anterior third of the bone, from the maximum concavity of the alveolar margin to the tip of the snout, which corresponds to the rosette. Finally, we note the good preservation of which we can call a “prenarial foramen” of 3 mm in diameter on both premaxillary surfaces and which extends its furrow posteriorly towards the maxillae (Fig. 1).

Seven alveoli are present on each side, with a relative symmetry between the two bones. Replacement teeth with serrated carinae are visible in several alveoli (Fig. 2): on the 2nd right, the 2nd left, and the 4th left alveolus. We observed generally 5 serrations each millimeter (Fig. 2) and the profile of teeth is laterally compressed, as typical for Baryonychinae (Sereno et al. 1998).

Tomographic results indicate a neurovascular network among the best-preserved discovered in a dinosaur and the very first complete described in a spinosaurid (Fig. 4B; Appendix 1B, D, F, H). In addition, the presence of the two fused premaxillae makes it possible to compare the structure of this network on the bones of the same individual and to study eventual symmetry in the arrangement of the branches connected to the superficial foramina (Fig. 4; Appendix 1B, D).

The neurovascular network is described here from the posterior to the anterior part of the specimen.

Two large and parallel branches run along the suture of the premaxillary bones on their entire length. In the most posterior part, they occupy a ventral position, just above the alveolar cavities, and fill almost the entire width of the muzzle at this level, each branch having a diameter of 1.4 cm for a bone width of 2.1 cm (Figs 3; 4B).

Then, they go up anteriorly at an angle of about 30° to the level of the fourth alveolus and then form a rounding along the upper margin of the bones, just below its surface to the apex. This anterior region, which in fact represents the rosette, is clearly the most supplied in ramifications (generally pointing forward), all leading to the few large superficial foramina (Figs 1; 4A, B).

At the level of the 7th alveolus, a first important secondary branch runs dorsally (Fig. 4B; Appendix 1 B,D). It almost divides in two, at the base of the main branch. A first branch with an angle of 60° relative to the main branch goes back to the surface of the bone to lead to a prenarial foramen (Fig. 4A, B, red squares). The location of this foramen is special, because it leads to bone depression reflecting the close presence of the (not preserved) external naris at the back (Figs 4; 8; 20). A groove is also visible on the surface of the bone in continuation of this prenarial foramen (see description below). A second branch, this time, points forwards at an angle of 40° with the main branch and ends up at a substantially identical size foramen, about 3 mm in diameter. The alliance of these two branches connected at their base forms in lateral view what we will call the “neurovascular bowl-shaped structure” (Fig. 4D). Several other much smaller, more or less preserved, secondary ramifications were revealed by the scan. They also join surface foramina. Interestingly, this is where the only two rearward-pointing rami are visible in the entire scanned neurovascular network (also seen in the younger Cristatusaurus specimen). They open at the level of the prenarial foramen and are only present on this junction zone between premaxillae and maxillae. In front of the 7th alveolus, all other branches are projected forward, or for some perpendicularly, on the lateral or dorsal surfaces of the bones. Between the 7th alveolus and the base of the rosette, an increase in the number and size of the surface foramina – which is reflected in the fossil by a complexification of the underlying neurovascular network – can be observed (Figs 1; 4A, B). A branch detaches from the main right neurovascular trunk, from the back of the 7th alveolus, while on the left it does so at the level of the 5th alveolus. There is therefore a certain asymmetry here. The left and right branches both extend to the margin of the third alveolus in a foramen, rounded off. The left one is opening slightly more anteriorly. Besides the location of the start of these two branches, a second difference can be noticed: the left branch is approximately twice as thick than the right one (Appendix 1B, D). Again, a significant number of secondary rami break off from the thicker branches and emerge on the outside of the bone by foramina. We sometimes detect interconnections or anastomoses between them (Fig. 4B). The rosette extends from the point of maximum concavity of the lower premaxillary margin (from the 3rd alveolus) to the tip of the snout where this one is the widest (8.2 cm). It is here that the complexity of the network becomes the highest. The thickest and most connected branches are found on the rosette with the largest foramina visible in anterior view. A minority of these branches directly innervates the roots of the teeth, while most follow longitudinally the alveolar cavities and lead to a foramen on the lateral surface of the bone.

Though a relative symmetry between the network of the premaxillae, some notable differences should be noticed. The changes between the left and the right premaxillae are here mainly due to the great variability in the location and size of the teeth alveoli (Figs 5; 6). However, some structures present a recurrence in MNHN.F.GDF365 for example, between each dental root, a ramus is detached from each of the two main neurovascular branches to rise dorsally to large foramen. Several pairs are easily recognizable (Fig. 4A, B, red circles).

The most posterior pair is directly connected on the surface by prenarial foramina, which open into a well-preserved superficial groove on the two premaxillae and which runs along the bone dorsally. From the prenarial foramen, the pattern of these superficial furrows displays an astonishing symmetry between the two faces, with a clearly visible “cross” shape, which excludes an artefact (Fig. 4C).

It is very important to note at this point that this prenarial foramen is also visible and preserved on the other specimens of Baryonychinae (Fig. 20), namely on the premaxilla of the European Baryonyx ‌walkeri (Charig & Milner 1986, 1997) and that of Suchomimus ‌tenerensis from Niger (Sereno et al. 1998). We now know that this important foramen marks the exit of the posterior branch of the “bowl structure” described above and recurrent in the two specimens of Cristatusaurus lapparenti, MNHN.F.GDF365 and MNHN.F.GDF366. It indicates in a way the separation between the branches pointing forwards (in the direction of the rosette) and those pointing backwards (in the direction of the external nostrils). This new character will be examined in the discussion section, as will the interpretation of the prenarial grooves (Figs 1B; 4C). The right premaxilla shows another well-preserved nerve branching imprint just anterior to that described above (Fig. 4C). Grooves and channels of a similar nature have been described on maxillae of Tyrannosaurus ‌rex ‌Osborn, 1905 and interpreted by the authors as natural molds for the rami of the maxillary nerve (Brochu 2003).

All teeth have typical apical curve found in most theropods. The first pair of alveoli has proportionally rather small teeth compared to those usually found there in Spinosauridae: 1.75 cm long on the left and 2.25 cm on the right. They had not yet protruded while the animal was alive, either because they were not sufficiently developed (possible ontogenic reasons), or because they are replacement teeth in formation, the main ones having been lost during decomposition or the fossilization process itself (Fig. 5B). We can also note the presence of an orphan fragment of about 1 cm long at the tip of the right tooth, forming an angle of 90° with the axis of the tooth. Small replacement teeth are sometimes partially or even completely returned in the alveolar cavity they occupy, probably due to movements and compression of the sediment over time (Fig. 6). The second pair of alveoli reveals a tooth of 5.3 cm long on the left side and 6.4 cm on the right. The latter appears to be the most complete in the specimen, from the root (indeed very close to the bone margin) to the apex. A small replacement tooth of about 1 cm is present medially to the left main tooth. The scan shows us a very intriguing structure: a sort of half-moon sheath around this tooth in labial position (Fig. 5C). We initially thought of an artifact, but an in-depth study of the density of this lunula shows us that it is indeed a vestige of the previous main tooth, which occupied this alveolus. This second right alveolus therefore offers us both the vision of three dental generations in the same cavity but also a very good example of the last stage of the demineralization process of a tooth. Multiple levels of bone sections in axial view were highlighted by CT scans (Fig. 5). The third pair contains replacement teeth of 3.2 cm on the right and 2.3 cm on the left. The principal tooth is lacking on the right side and the left one is incomplete. Each third alveolus has a second small replacement tooth, like the one seen on the first pair (the right one is completely turned with an angle of 180°). The fourth pair preserved a relatively complete 4.8 cm tooth only in the left alveolus. Here a fragment of the apex of a tooth is identifiable. It could be a small replacement tooth. Its angle and position could also match with the apex location of a mandibular tooth when the animal had its jaws almost closed. The fifth pair reveals a tooth 1.5 cm on the left (probably a replacement tooth) and 1.9 cm on the right with again a small “characteristic” second replacement tooth, this time pointing up. The sixth pair retained only a small tooth of about 1 cm on the right. The seventh and final pair of alveoli is free of preserved teeth (Fig. 6).

While the apex of only three teeth is visible superficially (one, incomplete, on the 2nd right alveolus, two on the 2nd and the 4th left alveoli), the CT-scans reveal a total of 12 clearly identified fossilized teeth inside the bones (Fig. 6). From a general point of view, the replacement teeth were formed medially, along the premaxillary groove, to gradually occupy a more labial position as they grew, corresponding to the alveolar openings visible on the surface. These replacement teeth developed in resorption crypts, lingually to the principal tooth, as described in crocodilians (Edmund 1960; Peyer & Zangerl 1968) and several theropods such as Tarbosaurus ‌Maleev, 1955a (Hanai & Tsuihiji 2019), Allosaurus ‌Marsh, 1877 (LeBlanc et al. 2017a), Gorgosaurus ‌Lambe, 1914 (LeBlanc et al. 2017a) or Coelophysis ‌Cope, 1889 (Fong et al. 2016). The new tooth recovered minerals from the old one for its own construction during this gradual demineralization. This process had never been observed or described in Spinosauridae. We therefore observe a similar tooth ontogeny in the two specimens of Cristatusauruslapparenti and also in Spinosaurusmaroccanus (Figs 5; 6; 13; 17A).

This specimen has two separate premaxillae, with a smooth surface. This indicates that the bones are more eroded than on the mature specimen (MNHN.F.GDF365). Small parts are lacking on the very anterior portion. Some foramina are there strongly denatured, sometimes even fused with their neighbors. Furthermore, dark streaks on this area indicate the typically tubular structure of the internal bone with lamellae and osteons canaliculi (Müller 2009).

The right premaxilla is 11.8 cm long and 6.6 cm high. The left one is 11.4 cm long and 6.5 cm high. The ventral view shows a total of 7 alveoli containing the principal teeth on both premaxillae (Fig. 9C). Secondary alveoli are visible on the 3rd alveolus for the right premaxilla and on the 3rd and the 7th alveoli on the left one (where replacement teeth apexes are visible). All principal teeth are crushed or lacking (Fig. 9). The structure of laterally compressed teeth of the Baryonychinae type (Sereno et al. 1998) can, however, be demonstrated by the observation of the tooth in the second alveolus of the left premaxilla; the denticulate carina is clearly recognizable there. Once again, the analysis of the CT scans made it possible to refine the number, size and exact distribution of the foramina spotted on the surface of the specimen. Some of them, questionable, could be retained in this description because their direct link with underlying neurovascular preserved branches. We counted 50 foramina on the right premaxilla and 54 on the left one (Fig. 7). Their diameter varies from less than 1 mm to 4 mm (for an average of 1 mm). It also made it possible not to consider certain depressions that could have been mistaken for foramina without internal analysis of the fossil. The density of foramina is greater on the front of the muzzle, at the level of the rosette (Fig. 7; Appendix 2).

A subcircular depression of about 1 cm in diameter was noted on the posterior portion of the right premaxilla, 1.5 cm above the alveolar margin. This is difficult to interpret but it seems to be the result of any trauma during the animal’s lifetime, like an abscess, shock or bone callus that has left a scar (Fig. 7B).

The difference in density between the bone and the sediment was lower than with MNHN.F.GDF365, probably for taphonomic reasons. The result of the scans nevertheless allows revealing a complex strongly branched network. We note again the presence of the two main parallel branches in each bone that run along the premaxillary suture, with a diameter corresponding again to almost the entire bone in its posteroventral part (Appendix 2B, D). These branches then rise rapidly to reach the dorsal edge of the premaxillae at the level of the 3rd alveolus (Appendix 2F, H). They then branch very strongly in the rosette into numerous smaller secondary branches and open out via the foramina. These general observations coincide well with the arrangement described in MNHN.F.GDF365; no striking difference was noted. The notable differences between the two specimens relate to the length of each branch rather than their general arranging and could therefore be related to ontogeny.

Although the dorsoposterior part of the premaxillae is lacking, the neurovascular complex is not affected. As in MNHN.F.GDF365, the peculiarity of this area is the presence of the “bowl-shaped structure” with vascular ramifications pointing for the first time backwards (prenarial foramen). It marks the anterior beginning of the bony depression leading to the nasal cavity (Fig. 8). On the right main branch, behind the 7th alveolus, two secondary branches with the same base stand out, one pointing towards the back and the other laterally towards the edge of the bone, perpendicularly. On the left main branch, the structure is more confused but is nevertheless present (Fig. 8B, C).

Just behind the 6th alveolus of the two premaxillae, another secondary branch is detached with an opening angle of approximately 20° with respect to the main branch, then which further subdivides into two smaller rami opening out to the surface of the bone at the 5th and 4th alveoli. In the part starting at the front of the 7th alveolus up to the maximum concavity point of the lower margin of the bone (base of the rosette), the preservation of the network is very choppy and incomplete in the right premaxilla but much better preserved in the left one (Appendix 2B, D). There is the birth of a secondary branch which detaches from the main branch at the same angle as those back (approximatively 20°). It begins between the 4th and 5th teeth to move towards the surface of the bone leading to large foramina, mostly in the rosette (Fig. 8A).

From the base of the rosette (the middle of the 4th alveolus) to the apex of the bone, the internal conservation of the left premaxilla is more complete. As in MNHN.F.GDF365, the network here is much more branched, with most secondary branches (5 of the 6 in all raised on the left) leading to large foramina in this zone. Most of the branches in turn subdivide before reaching the surface of the bone. Finally, the diameter of the foramina is on average larger on the front of the rosette: 3 or even 4 mm compared to 2 or 1 mm further behind (Fig. 7).

The secondary branches opening dorsally in pairs described in MNHN.F.GDF365 are found in this specimen, too (Fig. 8A). They are relatively symmetrical, except for the most posterior pair. This area corresponds to the anterior foramina of the “bowl-shaped” structure. The ramus is atrophied and/or badly preserved on the left side. There seems to be fewer interconnections between the secondary branches than in the more mature specimen (Figs 4B; 8A). However, the quality of preservation of the neurovascular network being of lower quality here, it is possible that these connections are not visible. The grooves described on the surface in MNHN.F.GDF365 running along the internal roof surface of the bone and connecting different foramen is partly found in MNHN.F.GDF366 but is much less preserved. The superficial breaks and the absence of the posterodorsal portion of the bones make its follow-up difficult, but we were able to highlight a trace of a groove over 2 cm from one of the foramina of the “bowl” structure, as in the more mature specimen (Figs 4; 8).

Surprisingly, we were able to count a total of 35 main and replacement teeth on both premaxillae: 18 on the right and 17 on the left (Fig. 9). Apart from the seventh left alveolus, all others have principal teeth preserved but broken above the alveolar margin. Each alveolus contains a replacement tooth in addition to the main tooth. We sometimes even notice a second replacement tooth, always small and placed on the lingual side of the first ones. Each of the alveoli of the first pair contains a main tooth and two replacement teeth. The second pair has a main tooth in each alveolus as well as a replacement tooth. A second replacement tooth is visible on the right side. The third pair of alveoli shows two main teeth and two replacement teeth. The main tooth on the left, however, is reduced and not sufficiently developed to come out of the gum line. It was growing back following the loss of a mature tooth in this location (indeed, each main tooth was quickly replaced when it broke or fell out). The fourth pair has a main tooth as well as two replacement teeth in each alveolus. The fifth pair illustrates a good example of a tooth replacement stage. The replacement tooth here present indeed begins its lateral demineralization action, with a second replacement tooth in the making. The sixth and seventh pairs both have main teeth accompanied by developing replacement teeth (a second replacement tooth is visible in the sixth right alveolus).

The sample is a portion of a right maxilla with the anterior and posterior parts missing. It is 10.6 cm long and approximately 3.5 cm high. There are five alveoli on the specimen. The first coincides with the anterior fracture of the bone and highlights what seems the pulp cavity of a main tooth 3.8 cm long (probably 0.5 cm more with the dentin and enamel layers). The three more posterior alveoli have the same diameter as the first (1.4 cm) but the teeth are broken there. They nevertheless have well-preserved enamel around their dentin. The fifth is on the posterior fracture of the bone and is not complete. The tooth there is broken longitudinally. The thickness of the dentin layer covering the pulp cavity is approximately 2 mm on each tooth. The lateral surface of the maxilla has 11 clearly identifiable foramina (Fig. 10A). Two of them open forward at a 45° angle and are associated with furrows in the bone, as observed in Baryonyx ‌walkeri ‌(Charig & Milner 1997), a sign that they were probably innervating the lips below the alveolar margin (Reisz & Larson 2016). The lingual surface is difficult to describe due to the bad preservation of the fossil.

In ventral view, we identify lingually a bony lamina running along the maxilla, separated with alveoli by the maxillary groove. As can be seen in other Spinosauridae, this lamina gradually grows in width anteroposteriorly (Sereno et al. 1998; Dal Sasso et al. 2005) with 1 cm wide at the front of the sample to reach 1.3 cm at the rear (Fig. 10B). Conversely, the alveolar zone decreases (1.8 cm wide at the front and 1.4 cm at the rear). Thanks to the breaking of the bone coupled with the information of the CT scans, it can be confirmed that the maxillary groove does not mark the separation of two bones and is only an invagination of the surface of the maxilla, containing pits of resorption of the teeth. It shows that the maxillary lamina does not belong to the vomer but to the maxilla itself (Dal Sasso et al. 2005). The interdental plates are also clearly visible between each alveolus (Fig. 10B).

The tomography shows us that the configuration of the neurovascular complex in the maxillae differs from that encountered in the premaxillae (Appendix 3B, D, F, H). The main branch no longer overhangs the alveolar roots, but labially runs along the alveolar cavities on their first upper third by projecting secondary branches under an angle of about 45° between each tooth and leading to surface foramina (Fig. 11A, B). The teeth show a very good quality of preservation and two to three generations of teeth are present in each of the five alveoli (Fig. 11C). Only the first two main teeth do not have their apex broken. The first alveolus (the most anterior) contains this famous tooth of 3.8 cm partially visible externally due to the breakage of the bone with a replacement tooth in formation of 1.5 cm; the second alveolus contains a main tooth of 2.8 cm, a replacement tooth of 0.4 cm and a very reduced vestige of the old main tooth almost completely demineralized on the labial margin of the alveolar cavity. The third alveolus shows an incomplete tooth of 3.2 cm and a replacement tooth of 1.2 cm. The fourth one has three generations of teeth with the main (incomplete) 3.1 cm tooth already hollowed out laterally by a robust 2.2 cm replacement tooth accompanied by a second tiny 0.5 cm replacement tooth. The fifth alveolus contains a broken main tooth of 2.6 cm and still a replacement tooth of 1 cm. It should be noted that the replacement teeth always begin their formation at mid-depth of the alveolar cavity, on the lingual surface and occupying more and more space in the basal direction before growing distally to finally “come out” of the alveolus, as if the tooth takes the time to ensure good root strength before exposing itself outside the bone.

The premaxillae are incomplete, particularly on the left side of which the dorsal part of the rosette is missing (Fig. 12A). We are in the presence of a mature animal considering the “strongly fused” aspect of the two bones (Taquet & Russell 1998). The length of the right premaxilla, the most complete, reaches 24.5 cm.

Overall, there is no notable difference between these premaxillae and those belonging to the species Spinosaurus ‌aegyptiacus, except for the number of alveoli: seven in S. ‌maroccanus and six in S. ‌aegyptiacus (Fig. 13A). However, intraspecific or ontogenic variations in the number of alveoli appear to occur in dinosaurs and even on the same specimen, as on the Baryonyx holotype (six alveoli on the left and seven on the right (Charig & Milner 1997). The number of dental alveoli is therefore generally not considered a relevant feature in the diagnoses. For some authors, significant differences between the two species were found by comparison of cervical vertebrae (Taquet & Russell 1998) and some others consider Spinosaurus ‌maroccanus as a nomen dubium (Dal Sasso et al. 2005; Ibrahim et al. 2014). In lateral view, the anterodorsal margin of the rosette forms a typical curve for Spinosaurinae. The curve of the alveolar margin is proportionally more ovoid than in the genera of the Baryonychinae types (as in Cristatusaurus). The posterior branches of the premaxillae are not preserved. Total length of existing bones are approximatively 24.5 cm and a maximum height in frontal view of 10.4 cm. There are 28 foramina on the left premaxilla and 36 on the right one with a diameter of less than 1 to 9 mm, but most have a diameter between 4 and 6 mm (Fig. 12). The scan showed the location of several additional foramina, either inconspicuous on the surface or questionable to the naked eye. The numerical difference between the left and the right is due to the incompleteness of the left premaxilla, with a missing area of approximately 25 cm2 on the top of the rosette (Fig. 12A).

The maximum concavity of the rosette on the alveolar margin is 8 cm from the apex of the snout (at the level of the 3rd alveolus). This is also where the rosette is the widest in ventral view at 13.5 cm (Fig. 14A, B). The diameter of the alveoli varies from 1.4 cm for the smallest to 4.4 cm for the largest. It should be noted that the first pair of alveoli subsequently has additional sub-cavities separated by a bony part with “autonomous” replacement teeth (Fig. 13B1). We could think of interdental plates, as has already been observed in many dinosaurs such as some hadrosaurs (LeBlanc et al. 2016), heterodontosaurs (Butler et al. 2012) and theropods like juvenile dromaeosaurids (Norell et al. 1994), Coelophysis ‌Cope, 1889 (LeBlanc et al. 2017a), tyrannosaurs (Hurum & Sabath 2003; Funston et al. 2021; Sattler & Schwarz 2021) or Archaeopteryx ‌von Meyer, 1861 (Wellnhofer 1994). It seems to be a primitive character in archosaurians for some authors (Elzanowski & Wellnhofer 1993). Such interdental plates have also been described in other taxa such as mosasaurs (Caldwell et al. 2003; Rieppel & Kearney 2005; LeBlanc et al. 2017b) or even snakes (Budney et al. 2006; LeBlanc et al. 2017b). Our CT scans have not detected any notable difference in densities between the premaxillary bone and this potential interdental plate (Fig. 13B1-B3). Given all of these indications, we believe that this bone separation could well be an interdental plate, insofar as the two teeth takes root in the same place (Fig. 13B3). More pushed comparisons with other Spinosauridae are necessary to support this hypothesis.

This potential interdental plate separating a replacement tooth from the main tooth in the first alveolus is not found both in either small (MNHN.F.GDF366) or more mature (MNHN.F.GDF365) Cristatusaurus individuals (Fig. 13C1-C3). The tomographic results indicate in ventral view that the replacement teeth at this location, mostly invisible on the surface, developing very clearly in the same base alveolus as the main tooth, without any bony separation (not observed on the surface for MNHN.F.GDF365 because of the sediment but visible on CT scans).

The internal configuration in Spinosaurinae is necessarily partly different from that of Baryonychinae, such as Cristatusaurus. Indeed, the posterior bone processes of the premaxillae are more developed in Spinosaurinae and their external nostrils occupy a much more posterior position compared to Baryonychinae. We can clearly observe that the neurovascular branches take an increasingly ventral position posteriorly, eventually to emerge below the external nostrils in Cristatusaurus (see internal descriptions of MNHN.F.GDF366 and MNHN.F.GDF365). This is not the case in Spinosaurus where the two main neurovascular branches remain in a dorsal position all along the premaxillae to operate a descent very probably just before the opening of the external nostrils, at the level of the posterior suture between premaxillae and nasals (Fig. 14).

The left premaxilla being very incomplete, the underlying network is impossible to reconstruct, except at the apex of the rosette. The backward pointing secondary branches and the bowl structure described in Cristatusaurus specimens are not observed here perhaps due to lack of preservation here. However, it is very likely that such a structure was also present in Spinosaurus but further back, to invade the upper area of the outer nostrils while the main branches were to dip below them, as in Cristatusaurus.

On the posterior half of the specimen, the secondary branches extending from the two main branches are generally thin (in the order of 1 mm) and emerge more or less perpendicular to the surface. From the 3rd alveolus and up to the apex of the muzzle, the teeth widen considerably (Fig. 14A, B). The secondary branches also widen in this area, to reach a diameter of several millimeters (up to 6 mm for the widest); Furthermore, the orientation of the branches change, and they all point towards the front of the muzzle (Fig. 14B). Even though teeth are missing, the neurovascular network is once again found to follow their shape from the root to the alveolar bone margin. Globally, we find the same neurovascular arrangement as in Cristatusaurus. The widest branches directly give rise to a series of six foramina, which run along both sides of the bony symphysis between the premaxillae on the front of the rosette, in a relatively symmetrical way (Fig. 14A; Appendix 4B).

Internal scans reveal seven teeth still preserved out of the total 14 premaxillary alveoli (Fig. 14B). On the left premaxilla, there is a first tooth 2.2 cm long in the first alveolus, pointing backwards at an angle of approximately 45° with respect to the alveolar margin and probably in a transitional situation between replacement tooth and main tooth. A residue is also visible on the scan in the apical position of the tooth (probably the broken tip). Still on the left premaxilla, the largest preserved tooth of the specimen is housed in the second alveolus. Although broken on the surface, it measures 6 cm in length (probably more than 10 during the animal’s lifetime) and 2.6 cm in diameter at the root. Since the dentin has dissolved, the enamel thickness can easily be measured here: 5.1 mm.

On the right premaxilla, the first alveolus also contains a 1.3 cm tooth, which this time can reasonably be described as a replacement tooth (the main tooth not being preserved). The second right alveolus contains the remains of an extremely fragmentary tooth but suggests a size comparable to that which occupies the second left position. The third, fourth and fifth alveoli only contain replacement teeth with the respective sizes of 2 cm, 1.9 cm and 1.2 cm (not visible on the surface).

The maxillae are incomplete, especially on the dorsal part, but allow a global and direct view of the inner cavity of the bones (Fig. 16A; Appendix 4A). This is why we decide here to describe the exterior and the interior of the fossil in the same section.

The right maxilla is preserved on a length of 32.5 cm and the left one on 28 cm. The two bones are well fused together and the distribution between the left and right alveoli is relatively symmetrical, both in position and in size (Fig. 15). The left bone is caudally broken at the level of the eighth maxillary alveolus while the right extends a little bit more to the ninth one. The first two pairs of alveoli are relatively small in diameter (1.4 cm for the 1st pair and 1.8 cm for the 2nd one) compared to those further back: 3.5 cm for the 3rd pair, 4.1 cm for the 4th pair (the widest), then they gradually decrease in diameter for the following ones (3.7, 3.4, 3.2, 3 and 2.9 cm for the 9th pair). This configuration is identical to that observed on the specimen of Spinosaurus ‌MSNM V4047 from the Kem Kem beds, southern Morocco (Dal Sasso et al. 2005). Ventrally, proximal processes of the premaxillae are visible and strongly sutured with rostromedial processes of the maxillae at the level of the 3rd maxillary alveolus on the first anterior 12 cm (Appendix 4C).

There are 17 neurovascular foramina superficially, on the right side, and 23 on the left side. Some are very thin (less than 1 mm in diameter) and others much thicker (up to 9 mm in diameter). Of course, some could not have been detected without the tomographic data, but the roof of the bones is here missing and, consequently, many other foramina must have been present, as evidenced by the numerous grooves preserved on the inner face of the bone (Fig. 15).

Indeed, several furrows of respectable diameters (from 1 to 4 mm) and showing a certain anatomical regularity are visible on the labial internal walls of the bones, distal to the dental alveoli (Fig. 16B-D). The conjunction of the trajectories of these furrows with those of the scanned neurovascular rami allows us to assume that the main branch must have labially followed the alveolar cavities. The proximity between dental cavity, bone and neurovascular network described in the premaxillae is found in the maxillae as well. The branches are interconnected, cross the bone, then follow the dental wall in a groove (partially in the alveolar lumen) to “plunge” back into the interdental bone, and then join the next alveolus and/or the surface of the maxilla by one or more foramina (Fig. 16D).

The branches revealed by the tomography are the largest of the scanned specimens (between 0.4 and 0.8 cm in diameter) and lead outwards at an angle of approximately 45° with respect to the surface of the bone for those located between the 1st and 6th pairs of alveoli, the most posterior branches tend to pierce the bone perpendicularly or even to point backwards, like a “neurovascular fan”. Moreover, they systematically occupy the interalveolar bone zones and are the extension of a very complex and interconnected network whose arrangement can be sketched by observing the grooves on the internal walls of the bones, labially and along the alveolar cavities (Figs 16; 17B). All these elements show, again, that these rami were probably connected to two main neurovascular branches that should run along the maxillae labially to alveolar cavities on their first upper third, like what was described in the right maxilla fragment of Cristatusaurus ‌lapparenti.

The left maxilla still contains four alveoli with broken and/or fragmented teeth. In the 2nd alveolus is the best-preserved tooth with its 3.2 cm and such visible characteristics of Spinosaurinae such as a rather circular section with a slight carina without serrations and typical longitudinal striations (Fig. 17Aa1). A first-generation replacement tooth of 2.1 cm is in the 4th alveolus. The 5th alveolus houses a main tooth 3.4 cm long. The 6th alveolus contains a main tooth 4.1 cm long and in the process of lateral demineralization by a well-preserved replacement tooth of 3 cm with a second-generation replacement tooth of 0.5 cm (Fig. 17Aa2).

The right maxilla reveals a main tooth broken in two longitudinally in the 3rd alveolus as well as a replacement tooth fragment of barely 0.5 cm in the 6th alveolus.