Each anterior (thoracic) vertebra of S. canicula includes a mineralized, hourglass-shaped, centrum (Figures 1A,B). Each centrum is dorsally overlaid by a mineralized neural arch that alternates with interneural elements, both (neural and interneural elements) appearing similar in shape but interneurals are located dorsal to the junction between two centra (Figures 1A,B). The vertebral body surface also shows thin zones of mineralization (lace-like mineralization, Figures 1A,B) that are small tesserae (see details below and in Figure 5). The neural arches and interneural elements externally appear as continuously mineralized structures (Figures 1A,B) but the virtual (microCT) and histological sections reveal that the mineralized layer encloses an inner core of unmineralized tissue (Figures 1C,E,F) of cartilaginous nature (Figures 1E,F). In the following, we used HES and PAS-AB histological staining protocols to characterize cells and their extracellular matrix in each vertebral tissue. We previously successfully used HES staining to characterize cartilaginous matrix mineralization in chondrichthyans (Enault et al., 2015; Debiais-Thibaud, 2019). Mineralized zones (stained with Alizarin Red S in Figure 1D) appear dark pink in comparison to the unmineralized matrix (Figure 1F, compare g with uc zones), probably because of an eosinophilic nature. The comparison between the virtual and histological sections (Figures 1C,F) allows identifying several types of mineralized tissue. First, there is a zone displaying cell spherical lacunae, where mineralization is less dense (lower gray level in Figure 1C), and that is located in the dorsal-most zone of the cartilaginous core (g in Figure 1). Denser mineralized zones (deeper white in Figure 1C) cover the inner unmineralized cartilage rod (uc in Figure 1). These perichondrial zones are further differentiated in the proximal perichondrium (facing the neural tube, la in Figure 1) and the distal perichondrium (facing the thoracic muscles, f in Figure 1) through the HES and PAS-AB staining protocols.

In the embryonic developing neural arches (Figures 2A,B), at a stage when mineralization initiates (Enault et al., 2016), the fibrous perichondrial tissue is characterized by flat cells enclosed in a fibrous extracellular matrix (pc in Figure 2A'), while the internal cartilaginous tissue stains positive for Alcian blue (hc in Figure 2B).

In the juvenile specimen in which structures are fully mineralized, the neural arch is organized as a central, unmineralized cartilaginous rod (uc in Figures 2C',D). This internal cartilaginous rod is no more than 100 μm wide in the juvenile specimen, which is similar to what is observed in the hatchling specimen, suggesting that this internal tissue does not significantly grow between these ontogenetic stages (Figures 2A,C). In continuity with this cartilaginous unmineralized rod, the most dorsal and ventral aspects appear mineralized (Figures 1C,E,F, 2C',D) together with the contact surface with the perichondrium (g in Figures 2C',D). The perichondrium is characterized by low to absent Alcian blue staining in the PAS-AB assay (Figure 2D) and by elongated cells engulfed in a Saffron stained extracellular matrix in the HES assay (Figure 2C'). Several differences in cell density and matrix organization can be detected when comparing the distal mineralized perichondrium (toward the trunk muscles) with the proximal mineralized perichondrium (facing the neural tube) (compare Figures 2C1,C3,D1,D3).

As a result, three mineralized histotypes can be described in the neural arch of S. canicula. The first one is superficial lamellar mineralization that arises from the perichondrium located in the proximal side of the neural arch. The mineralized matrix displays a lamellar organization, cell density appears very low, and the few cells found in the mineralized matrix are elongated (Figures 2C1,D1). The second type is globular mineralization, which occurs in the matrix surrounding chondrocytes in the dorsal part of the neural arch and the contact zone with perichondrial mineralized tissues. HES staining in the mineralized matrix shows concentric rings around the chondrocyte lacunae that we interpret as the marks of previously described Liesegang lines (Figure 2C2), typical of globular mineralization in hyaline cartilage (Ørvig, 1951; Peignoux-Deville et al., 1982; Seidel et al., 2016). The pericellular space is positive for Alcian blue (ps in Figure 2D2). The third type is a fibrous mineralized tissue in the perichondrium that is located in the distal side of the neural arch. The cells also appear elongated as expected with fibrocytes and their matrix is characterized by loosely arranged fibers that stain with Saffron (arrows in Figure 2C3). The thin pericellular space is positive for Alcian blue (Figure 2D3).

The posterior vertebrae of R. clavata are composed of a centrum associated with one neural arch located dorsally, at the junction between two neural spine-interdorsal complexes (Figures 3A,B). The neural arches and interneural elements are not covered by tesserae, as opposed to the neural spines (Figure 3B). The virtual section shows that the whole depth of the posterior neural arches is mineralized (Figure 3C). In both hatchling and juvenile specimens, this architecture is similar in the posterior vertebral column (not shown). HES histological staining highlights the presence of subspherical cells in the center of these neural arches (Figures 3I',J,N',O,N2,O2). Their extracellular matrix displays Liesegang lines with HES and an Alcian blue positive pericellullar space with PAS-AB, making this zone similar to the globular mineralization observed in S. canicula (Figures 3N2,O2). This internal mineralized cartilaginous rod does not seem to differ in size between the hatchling and juvenile specimens, suggesting little growth of this internal zone between the corresponding ontogenetic stages (Figure 3, compare the g zone in I' with N').

Surrounding this mineralized cartilage core is a tissue with fibrocyte-shaped cells, engulfed in a fibrous matrix devoid of Alcian blue staining except for a thin pericellular matrix (Figures 3I',J,N',O,N1,N3,O1,O3). This mineralized fibrous tissue lines both the proximal and distal sides of the neural arch and is thicker in the juvenile than in the hatchling specimen (compare Figures 3I',N'). We assume this tissue to have a perichondrial nature, and it appears similar to the fibrous mineralization in the distal perichondrium of the S. canicula neural arches. Similar to the observations in S. canicula, perichondrial mineralization appears denser than globular cartilage mineralization (Figure 3C). In contrast to S. canicula, no lamellar organization of the extracellular matrix was observed in either the proximal or distal perichondrium of R. clavata (Figures 3N1,N3,O1,O3). In S. canicula, the architecture of neural arch mineralization is the same along the anterior-posterior axis of the vertebral column (no difference is observed between the anterior and posterior vertebrae) although the timing of centrum mineralization differs between these regions (Enault et al., 2016). In R. clavata, the mineralization architecture in the neural arches differs between anterior and posterior vertebrae (Figures 3B,F). In anterior vertebrae, we identified only interneural elements, located dorsal to the junctions between two centra and that were partly covered with tesserae (Figures 3F,H), while more massive mineralized units cover the anterior and posterior faces of the interneural elements (Figures 3F,K). A longitudinal frontal section in an interneural element shows that these massive units have a U-shape delineating the anterior and posterior faces of each interneural element (Figure 3K). As such, we considered them to be “corner tesserae” as previously proposed in the jaw of the Haller's round ray Urobatis halleri (Dean et al., 2009), although their size relative to the closest tesserae is higher than that of previously observed “corner tesserae” (Figure 3F). Histological staining showed the occurrence of mineralized globular cartilage in the center of these elements and layers of fibrous perichondrial mineralization on their proximal and distal surface (Figures 3L,M).

As a consequence, the interneural elements and neural arches of S. canicula and R. clavata in the caudal zone of the vertebral column display continuous mineralization, while the anterior vertebrae of R. clavata (immediately posterior to the synarcual) display a combination of tesserae and massive “corner tesserae”. All these mineralized structures are characterized by a mineralized perichondrial tissue that covers a partly (S. canicula) or more extensively (R. clavata) mineralized cartilaginous rod. Perichondrial mineralization occurs as lamellar mineralization only in the proximal zone of the neural arches of S. canicula and as fibrous mineralization in all other occurrences.

The synarcual of H. colliei is a post-cranial element pierced by dorsal and ventral nerve foramina (Figures 4A,B) that develops from the fusion of successive embryonic vertebrae and that does not display discrete neural arch or interneural elements (Johanson et al., 2015). The equivalent zone of elasmobranch neural arches in the synarcual is considered to be the zone dorsal to a ventral foramen (Goodrich, 1930; Eames et al., 2007; Criswell et al., 2017). We hypothesized a similar situation for chimaeras, and focused on the portion of the synarcual located dorsally to a ventral foramen, although our observations were similar in all other parts of the synarcual (including in the vertebral body, see Figures 4A,D,E,G). Previous studies showed poorly developed cartilaginous tissue in the posterior vertebrae of juvenile H. colliei, where histology displayed no sign of modified cartilaginous matrix that could be linked to cartilage mineralization (Debiais-Thibaud, 2019). However, more recent publications pointed out the thin layer of mineralized matrix in the synarcual of the elephant shark Callorhinchus milii (Pears et al., 2020) and recognized thin tesserae in the adult C. milii and rabbit fishes Chimaera monstrosa (Pears et al., 2020; Seidel et al., 2020). In our H. colliei sample, we detect a similar thin layer of mineralized tissue in the synarcual of a hatchling and a juvenile specimen (Figures 4A,G). This pattern is found as a peripheral delineation of the ventral part (corresponding to the vertebral body) and the dorsal part (corresponding to the neural arches) of the synarcual, but also in the inner contact with the notochord (Figures 4A,D,G). Most tesserae are not more than 10 μm wide in their chondral-perichondrial axis (Figures 4E',F,G',H). The line of modified cartilaginous matrix can be observed at the hatching stage and is more obvious in the juvenile specimen (Figures 4E–H). The matrix stains deep purple with HES (Figures 4E',G') and does not stain with Alcian blue (Figures 4F,H), suggesting a difference in the glycosaminoglycan content at this mineralization location. This site of mineralization is located in the subperichondrial zone of the cartilaginous unit. The cells located on the perichondrial side of this line are enclosed in lacunae in a matrix of cartilaginous nature [as supported by positive Alcian blue staining (Figures 4F,H)], although the cells are flattened (Figures 4E',G'). Moreover, no cells can be observed fully embedded in the mineralized cartilaginous matrix, either at the hatching or juvenile stage and the mineralized layer appears discontinuous (Figures 4G',H), supporting the presence of aligned tesserae as described in other holocephalan species (Pears et al., 2020).

To better compare our observations with literature data on tesserae, these mineralized structures were also illustrated in selected species (Figure 5). As previously noted, focal cartilage mineralization is detected with microCT outside the neural arches and surrounding the vertebral body of S. canicula (Figures 1A,B). Such sites also appear on histology sections, they are stained dark pink with HES and purple with PAS-AB and occur in a subperichondrial location within the cartilaginous matrix (Figures 5A,B). Although poorly developed, these sites can be interpreted as thin tesserae engulfing chondrocytes and display intertesseral joint regions (Figures 5A',B), as previously described (Seidel et al., 2016). The tesserae are separated from the perichondrium by a subperichondrial and unmineralized cartilaginous layer (Figures 5A',B). They are comparable to the discontinuous pattern of mineralization described in H. colliei (Figures 4G',H) but are thicker (up to 50 μm) and engulf chondrocytes in the mineralized matrix (Figure 5A') in S. canicula.

The neural spines of R. clavata, as well as the neural arches of the bowmouth guitarfish Rhina ancylostoma are covered by tesserae (Figures 3B, 5C–F). The tesserae are separated by unmineralized intertesseral joints (itj in Figures 5C',D,E',F) and display internal spokes (sp in Figures 5C',D,E',F) as defined by Seidel et al. (2016). Two zones can further be distinguished in the tesserae that correspond to the previously defined body and cap zones of a tessera (Kemp and Westrin, 1979; Seidel et al., 2016, 2017). In both species, the cap zone is characterized by flat cells, a higher content in collagen fibers, and a matrix poorly stained by Alcian blue except in the pericellular layer (Figures 5D,F). The cells in the body zone are subspherical and surrounded by an extracellular matrix displaying Liesegang lines with HES staining (Figures 5C',E'). In both batoids (R. clavata and R. ancylostoma), the maximum width of a tessera (tangential axis) is about 200 μm and does not seem to depend on the specimen body size as R. clavata is a young juvenile of 13 cm DW and 21 cm TL, while R. ancylostoma is 147 cm long. However, the depth of the tesserae (along the chondral-perichondrial axis) is less than 100 μm in R. clavata, while it is more than 200 μm in R. ancylostoma (compare Figures 5D,F). There are no obvious Sharpey's fibers radiating from the tesserae of R. clavata, whereas they are well-defined in the tesserae of R. ancylostoma (Figures 5C',D,E'F). Sharpey's fibers are thick collagenous fibers previously described in tesserae (Kemp and Westrin, 1979; Seidel et al., 2017).

Most of the neural arch of the smooth butterfly ray Gymnura micrura and Torpedo sp. is covered by tesserae that are 250 μm wide in their tangential axis (Figures 6A,B,E,F). Two larger mineralized elements are found on the anteriormost and posteriormost surfaces of each neural arch, similar to the corner tesserae described in R. clavata. These elements are more than 500 μm in their longer dimension (Figures 6A,E). The virtual sections made through the center of the neural arches show regular-sized tesserae on the neural arch surface (Figures 6B,F). The histological sections made through the corner tesserae (Figures 6C,G) show a central zone of globular mineralized cartilage covered by fibrous mineralization (Figures 6C',D,G',H).

The neural arches of the shovelnose guitarfish Pseudobatos productus are covered by classically organized tesserae with no massive corner tesserae (Figure 6I). The tesserae of the neural arch reach 400 μm long (in their tangential axis) and 100 μm wide (along the chondral-perichondrial axis, Figure 6J) and display intertesseral joints (Figures 6K',L). The cap zone of a tessera contains elongated cells in a fibrous (Saffron-positive) matrix with little acidic glycosaminoglycan content (Figures 6K',L), while cells in the body zone are similar in shape to chondrocytes in the unmineralized matrix (Figures 6K',L). The neural arch of R. ancylostoma also is covered by standard tesserae only (Figures 6E,F) without enlarged elements (data not shown).

In both the velvet belly Etmopterus spinax and the longnose velvet dogfish Centroscymnus crepidater, the neural arches display thin zones of mineralization dorsal to the centra (Figures 7A,B,E,F). Virtual sections passing through the neural arches show a thin, discontinuous layer of mineralization (Figures 7B,F). The histological staining results show that the thin mineralization layer is both proximal (facing the neural tube) and subperichondrial (Figures 7C,D,G,H). Although the majority of the neural arch is composed of unmineralized cartilage matrix containing round chondrocytes (Figures 7D,H), the HES staining shows thin purple-stained blocks within the cartilaginous matrix that do not embed any cells (Figures 7C',G'), similar to the thin tesserae described in H. colliei. The tesserae are 10 μm wide (in their chondral-perichondrial axis, Figures 7C',G') and are separated from the neural tube by a layer of cartilaginous matrix in C. crepidater (Figure 7G'), while it occurs in contact with the very thin perichondrium in E. spinax (Figure 7C').

In the Pacific angelshark Squatina californica, the dorsal neural arches and interneural elements appear as continuously mineralized structures that are contiguous with the tesserae covering the neural spine and the basidorsal and basiventral elements (Figure 8I). The virtual section of the S. californica vertebra shows an internal variation in the mineral density of this structure, with denser peripheral mineralization than in the core (Figure 8J). On the histological sections, the cartilaginous tissue core contains round chondrocytes in a modified matrix (Figures 8K',L), similar to our previous description of globular mineralization. This central cartilaginous zone is covered proximally and distally by a fibrous mineralized tissue that contains flattened cells, similar to what we already described for fibrous mineralization of the perichondrium (Figures 8K',L).

In all three Carchariniformes considered in the following, neural arch and interneural mineralization builds up continuous structures (Figures 8A,E,I). Neural arch mineralization is ventrally contiguous with tesserae from the vertebral body in the tope shark Galeorhinus galeus and the scalloped hammerhead Sphyrna lewini (Figures 8A,B,E,F).

The continuously mineralized elements of G. galeus and S. lewini include neural arches and interneural elements (Figures 8A,E). In both species, neural arch mineralization encloses an unmineralized core and appears denser on the proximal and distal faces than on the dorsal zone (Figures 8B,F). The dorsal zone and the contact zone with the mineralized perichondrium are composed of globular mineralization (g in Figures 8C',D,G',H). The perichondrial layer of the mineralized tissue is similar to the fibrous mineralization sites described in the distal side of the S. canicula neural arch, with enclosed fibrocyte-shaped cells and no observable lamellae (Figures 8C',D,G',H).

In the blue shark Prionace glauca, neural arch and interneural mineralization sites alternate dorsal to the large centrum and no tesserae can be observed on the whole vertebral unit (Figure 8I). Mineralization is very restricted to the dorsoventral axis of the neural arch. The neural arch mineralization is superficial to only one location of the unmineralized cartilaginous neural arch (Figure 8K). The mineralized tissue includes globular mineralization in its innermost zone, together with outer perichondrial layers on both proximal and distal sides of the neural arch (Figures 8K',L). These perichondrial layers are not homogeneous because the cells undergo a morphological transition from flat—in the contact zone with globular mineralization—to subspherical—in the external zone (Figures 8K',L). Besides, the extracellular matrix of the external zone shows HES and Alcian blue staining similar to unmineralized cartilage, suggesting the presence of hyaline cartilaginous matrix in this perichondrial zone (Figures 8K',L). However, this external layer also displays numerous bundles of collagen fibers similar to Sharpey's fibers (Figure 8K'). The progressive centrifugal transition from a fibrous to a cartilaginous nature of the outer layers corresponds to a progressive decrease of mineralization, as shown in the virtual section (Figure 8J).

In the horn shark Heterodontus francisci and the brownbanded bambooshark Chiloscyllium punctatum, alternate neural arches and interneural elements display a fully mineralized surface (Figures 9A,E). Tesseral mineralization is also located ventral to the neural and interneural elements of H. francisci (Figure 9A). In both species, the neural arch displays an unmineralized central core (Figures 9B,F) made of cartilage (Figures 9C',G'). However, globular mineralization is detected in the cartilaginous core in contact with the mineralized perichondrial layers (Figures 9C',G'). In H. francisci, the perichondrial mineralized layers are highly fibrous, in particular in the distal layer where numerous Sharpey's fibers are observed and incorporate numerous flattened cells (arrows, Figures 9C',D). On the contrary, the proximal perichondrium of the C. punctatum neural arch contains very few cells and the matrix is arranged in lamellae (Figures 9G',H). Overall, the organization and composition of the C. punctatum neural arch are very similar to what is observed in the S. canicula neural arch (compare with Figures 1C, 2A',B,C',D).

The neural arch of the bull shark Carcharias taurus displays superficial mineralization in the form of standard size tesserae, while the core of the neural arch remains unmineralized (Figures 9I,J) and cartilaginous (Figure 9K). The tesserae are about 300 μm long (tangential axis) and 150 μm wide (along the chondral-perichondrial axis, Figures 9K',L). The cap and body zones cannot be properly distinguished from our sample, probably because of a tilted section plan (Figures 9K',L). However, we observe Sharpey's fibers anchoring the tesserae to the distal fibrous layer of the neural arch (arrows, Figures 9K',L) and unmineralized intertesseral joints (Figures 9K',L).

Our work provides a histological examination in an unprecedented taxonomic range of elasmobranch fishes. We uncover the occurrence of fibrous mineralization in continuously mineralized neural arches in an Heterodontiform, a Squatiniform, most Carcharhiniformes examined in this study, and a Rajiform (Figure 10). Fibrous mineralization is also identified in the perichondrial zone of both corner and standard tesserae (Table 2). Here we wish to discuss the hypothesis of fibrous mineralization to be naturally mineralizing fibrocartilage, as previously suggested by Ørvig (1951) and Eames et al. (2007). Mineralized fibrocartilages were described in tetrapod entheses that join tendons to bones (Benjamin and Ralphs, 1998) and surrounding rib cartilage (Claassen et al., 1996). Fibrocartilage histology shows strong variation, ranging from almost hyaline to highly fibrous (similar to tendons) in mammals (Wachsmuth et al., 2006) and bony fishes (Benjamin, 1990). Similar variation is observed in our sample, where fibrous mineralization resembles more hyaline-type cartilage crossed by fibers in P. glauca (Figures 8K,L). This hypothesis would also explain the presence of type II collagen and a thin layer identified as “supra-tesseral cartilage” in the perichondrial surface of tesserae in some species (Seidel et al., 2017). Further histochemical, histological, and cellular characterization is therefore needed to better define the nature of this tissue and compare it to tetrapod tissues.

Beyond the nature of this tissue, we show that perichondrial fibrous mineralization is widely present in skeletal tissues of elasmobranchs. Ancestral character state reconstruction by a most-parsimonious model (Supplementary Material 1 and summarized in Figure 10) suggests that this histotype is ancestral to elasmobranchs (as previously hypothesized by Atake et al., 2019). This scenario implies a secondary loss of fibrous mineralization in C. crepidater and E. spinax. Considering the secondary loss of mineralization observed in extant holocephalans (Maisey et al., 2020; Pears et al., 2020), it is also plausible that fibrous mineralization is ancestral to all chondrichthyans but lost in holocephalans. Fibrous mineralization may then be a remnant of forms of mineralization found in early vertebrates. As a consequence, a review of the paleontological data is needed, especially in stem gnathostomes and agnathan taxa where several forms of perichondrial bone (cellular or acellular, as described in placoderm taxa Ørvig, 1951 or galeaspid and osteostracan taxa, as reviewed by Donoghue and Sansom, 2002) may be revised in light of this fibrous mineralization that covers cartilaginous skeletal units. Further comparison with extant and extinct taxa of bony fishes may also uncover similar tissues outside of chondrichthyans. Only this information may shed light on the evolution of fibrous mineralization and its significance to the evolution of a mineralized skeleton in vertebrates.

The continuous architecture always includes fibrous mineralization in the distal part of the neural arch but we identified lamellar, instead of fibrous, mineralization in the proximal side of the neural arch of two galeomorph species (Table 2). The fibrous and lamellar mineralizations histologically differ in that the former contains numerous cells, as opposed to the latter where only very few cells are observed, and by the arrangement of fibers. Another difference is the lack of a thick unmineralized connective tissue surrounding the lamellar mineralization (see Figure 2C', as opposed to the situation in other galeomorphs, see Figures 9G',K'). Bone-like tissue was previously reported in the neural arches of Carcharhiniformes (Peignoux-Deville et al., 1982; Eames et al., 2007) and Rajiformes (Atake et al., 2019). These previous studies did not distinguish the cellular from almost acellular tissues and the histological staining results in the work of Atake et al. (2019) better support the presence of fibrous mineralization in both the proximal and distal sides of the neural arches of L. erinacea and B. eatonii. As a consequence, the lamellar mineralized pattern as defined in our work has only been described so far in S. canicula and most probably in C. ventriosum (Eames et al., 2007) (both Carcharhiniformes), and in C. punctatum (Orectolobiformes). Considering these results, ancestral character state reconstruction with a most-parsimonious model suggests parallel evolution of this lamellar pattern in Carcharhiniformes and Orectolobiformes (Supplementary Material 2, summarized in Figure 10). Still, our observations and hypothesis need to be confirmed with broader sampling among Orectolobiform species, including older specimens. As a broader conclusion, our results support a recent evolution of this trait, making it unlikely to be homologous to any bone type found in early vertebrates.