Skeletogenesis in the swell shark Cephaloscyllium ventriosum

Authors


Dr B. Frank Eames, Institute of Neuroscience, 1254 University of Oregon, Eugene, OR, 97403-1254, USA. T: +1 541 346 4506; F: +1 541 346 4548; E: bfeame@uoneuro.uoregon.edu

Abstract

Extant chondrichthyans possess a predominantly cartilaginous skeleton, even though primitive chondrichthyans produced bone. To gain insights into this peculiar skeletal evolution, and in particular to evaluate the extent to which chondrichthyan skeletogenesis retains features of an osteogenic programme, we performed a histological, histochemical and immunohistochemical analysis of the entire embryonic skeleton during development of the swell shark Cephaloscyllium ventriosum. Specifically, we compared staining properties among various mineralizing tissues, including neural arches of the vertebrae, dermal tissues supporting oral denticles and Meckel's cartilage of the lower jaw. Patterns of mineralization were predicted by spatially restricted alkaline phosphatase activity earlier in development. Regarding evidence for an osteogenic programme in extant sharks, a mineralized tissue in the perichondrium of C. ventriosum neural arches, and to a lesser extent a tissue supporting the oral denticle, displayed numerous properties of bone. Although we uncovered many differences between tissues in Meckel's cartilage and neural arches of C. ventriosum, both elements impart distinct tissue characteristics to the perichondral region. Considering the evolution of osteogenic processes, shark skeletogenesis may illuminate the transition from perichondrium to periosteum, which is a major bone-forming tissue during the process of endochondral ossification.

Introduction

Despite their designation as ‘cartilaginous fish’, modern chondrichthyans such as sharks, skates and rays probably evolved from ancestral vertebrates that contained bone (Goodrich, 1930; Daniel, 1934; Romer, 1945; Ørvig, 1951; Hall, 1975; Moss, 1977; Carroll, 1988; Maisey, 1988; Smith & Hall, 1990; Janvier, 1996; Donoghue & Sansom, 2002). Indeed, fossils reveal that primitive chondrichthyans produced bone not only in the exoskeleton, or dermal skeleton (Zangerl, 1966; Hall, 1975; Moss, 1977; Maisey, 1988), but also in the endoskeleton (Coates et al. 1998). Remnants of an osteogenic programme may persist in extant chondrichthyans, as sharks and rays are argued to produce exoskeletal bone (Moss, 1970, 1977; Reif, 1980; Sire & Huysseune, 2003). The endoskeleton of living sharks also has been reported to contain bone-like tissue in the perichondrium of vertebral neural arches (Kemp & Westrin, 1979; Peignoux-Deville et al. 1982; Bordat, 1987), raising the possibility that shark endoskeletal bone is more widespread than what has been generally concluded from a few studies of fossil taxa. Moreover, because most analyses of the chondrichthyan endoskeleton have been performed on adult animals, the manner by which bone-like tissues arise during development is unknown. Here, we address two main issues surrounding chondrichthyan skeletal biology: do living sharks produce endoskeletal bone, and through what skeletogenic process does the shark endoskeleton form?

Complicating efforts to identify skeletal tissues in fishes are widespread departures from the standard classification schemes used in tetrapods. In mammals and birds, for example, clear distinctions typically are drawn between cartilage and bone, based upon histological, histochemical and immunohistochemical stains (Fig. 1; Ham & Cormack, 1987; Eames et al. 2003; Eames & Helms, 2004). In addition to various other criteria, a major distinguishing feature is the molecular nature of a skeletal tissue's extracellular matrix. Bone usually contains tightly wound bundles of Collagen type 1 (Col1), with little to no Collagen type 2 (Col2), while cartilage typically consists of loosely wound Col2 fibres, with little or no Col1 (Fig. 1; Ham & Cormack, 1987; Eames et al. 2003). However, many intermediate skeletal tissues characterize the fishes (Beresford, 1981; Benjamin, 1989; Smith & Hall, 1990; Sire & Huysseune, 2003). Chondroid bone, for example, displays a mixture of characteristics between cartilage and bone (Benjamin, 1990; Huysseune & Sire, 1990, 1992; Huysseune, 2000). Chondroid bone in the pharynx of the cichlid Hemichromis bimaculatus has a bone-like matrix (e.g. Col1-positive, Col2-negative) surrounding chondrocyte-like cells (Huysseune, 1989; Huysseune & Sire, 1990). Another skeletal tissue intermediate between cartilage and bone is fibrocartilage. In fish, fibrocartilage typically surrounds ‘true’ matrix-rich hyaline cartilage and contains many dense collagen fibres (Benjamin, 1990).

Figure 1.

Typical histological and immunohistochemical properties of vertebrate bone and cartilage. See Materials and methods for specific histological dye affinities. (A) HBQ staining of cartilage (c) and perichondral bone (b) in ceratobranchial of E15 chick embryo. (B) HBQ staining of cartilage and dermal bone in Meckel's cartilage and dentary of E14 mouse embryo. (C) Safranin O staining of ceratobranchial in E12 chick embryo. (D) Trichrome staining of Meckel's cartilage and suprangular in E11 duck embryo. (E) Col2 immunostaining of quadrate in E9 chick embryo. (F) Col1 immunostaining of suprangular in E9 chick embryo. Abbreviations: b = bone; c = cartilage. Scale bars: A,C,D = 200 µm; B,E,F = 100 µm.

The majority of vertebrate endoskeletal tissues that form during embryonic development proceed through the process of endochondral ossification to various degrees. Endochondral ossification has many stages in tetrapods, typically progressing from cartilage formation and perichondral deposition of bone to cartilage degradation and endochondral deposition of bone (Patterson, 1977; Caplan & Boyan, 1994; Mundlos & Olsen, 1997; Eames et al. 2003; Kronenberg, 2003; Ortega et al. 2003; Eames & Helms, 2004; Moriishi et al. 2005). While these latter steps have never been reported in chondrichthyans, we sought to determine the extent to which chondrichthyans initiate the process of endochondral ossification.

To provide a robust characterization of chondrichthyan endoskeletal tissues and their development, we performed a histological, histochemical and immunohistochemical analysis of the entire skeleton throughout embryogenesis of an elasmobranch, the swell shark Cephaloscyllium ventriosum. In particular, we compare tissues of the neural arches with other mineralizing C. ventriosum skeletal tissues, such as exoskeletal tissue that supports oral denticles, and also Meckel's cartilage. Alkaline phosphatase activity predicts mineralization patterns in C. ventriosum, but mineralizing shark tissues differ in many histological and immunohistochemical features. Strikingly, perichondral tissue in C. ventriosum neural arches exhibits more histological and immunohistochemical characteristics of bone, including tight packing of collagen bundles and stronger Col1 immunoreactivity, than does supporting tissue of the C. ventriosum oral denticle. Both the neural arch and Meckel's cartilage resolve tissue characteristics spatially between perichondral and deep regions during C. ventriosum development. Studies of shark skeletogenesis may shed light not only on the loss of bone in extant chondrichthyans, but also on a key step in the evolution of the process of endochondral ossification in primitive vertebrates: the perichondrium to periosteum transition.

Materials and methods

Embryo collection and tissue processing

Egg cases containing swell shark (Cephaloscyllium ventriosum) embryos were collected off the coast of Santa Barbara, California, or were harvested from a captive female. Embryonic stages were assigned according to the total length from tip of snout to tip of caudal fin. Based upon external features such as gill filaments and yolk sac morphologies, embryos used in our analyses correspond to elasmobranch developmental stages (Ballard et al. 1993) as follows: 5 cm = stage 31; 9 cm = stage 32; 12 cm = stage 33; 15 cm = stage 34. These respective stages represent roughly 5, 7, 9 and 11 months of embryonic swell shark development. Embryos were decapitated, fixed in neutral-buffered 4% paraformaldehyde overnight, and dehydrated through a graded ethanol series. Specimens for tissue section analyses were demineralized for 4 days in 19% EDTA prior to dehydration, embedded in paraffin, and cut to 6 or 10 µm thickness.

Histology

Safranin O (Ferguson et al. 1998), HBQ (Hall, 1986), Trichrome (Clark & Smith, 1993) and Picro-Sirius (Junqueira et al. 1979) staining protocols were performed on tissue sections as described previously. The dyes Safranin O and Alcian blue stain sulphated glycosaminoglycans in the Safranin O and HBQ protocols, respectively (see Fig. 1). The dyes direct red and Aniline blue stain collagen fibrils in the HBQ and Trichrome protocols, respectively (see Fig. 1). Sections stained with the Picro-Sirius protocol were visualized under either regular or polarized light microscopy, in order to reveal packing of collagen bundles (Junqueira et al. 1979). Whole-mount Alcian blue/Alizarin red staining was carried out as described previously (Wassersug, 1976), apart from samples that underwent mineralization staining in the absence of the Alcian blue acid alcohol solution, which may decalcify tissues.

Histochemistry and immunohistochemistry

Alkaline phosphatase activity was determined on tissue sections as follows. Dewaxed slides were rehydrated in phosphate-buffered saline (PBS), washed for 5 min in alkaline phosphatase buffer (100 mm Tris, pH 9.5; 50 mm MgCl2; 100 mm NaCl; 0.1% Tween 20), and incubated in BM Purple (Roche, Indianapolis, IN, USA) until strong signal appeared. Sections were washed in PBS, fixed in 4% PFA for 20 min, washed twice in PBS, counter-stained in 1% acid fuchsin for 20 s, and then dehydrated and coverslipped.

Immunohistochemistry was performed on tissue sections as previously described (Eames & Schneider, 2005). Primary antibody dilutions were 1 : 15 for Collagen type 1 (R1038; Acris Antibodies, Hiddenhausen, Germany) and pro-Collagen type 1 (SP1.D8; Developmental Studies Hybridoma Bank, Iowa City, IA, USA), and 1 : 25 for Collagen type 2 (II-II6B3; DSHB, Iowa City, IA, USA). For Collagen type 1 epitope retrieval, sections underwent microwave-induced heat treatment in 0.01 m citrate buffer. For Collagen type 2 epitope retrieval, sections were treated additionally with Ficin (Zymed, South San Francisco, CA, USA), according to the manufacturer's protocol, following the microwave citrate buffer treatment.

Results

Ontogeny of shark chondrogenesis and skeletal mineralization

To reveal the timing and pattern of swell shark chondrogenesis and skeletal mineralization, we performed Alcian blue/Alizarin red staining (Fig. 2A) on representative embryos throughout C. ventriosum development. Neither strong Alcian blue nor Alizarin red staining was observed in embryos measuring 4 cm or smaller from the snout to the tip of the tail (data not shown). Embryos of 5 cm clearly showed many Alcian blue-positive cartilaginous elements, such as Meckel's cartilage of the jaw and vertebral bodies of the spine (Fig. 2B,F). Two rows of tooth-like structures termed dermal denticles (also ‘placoid scales’ or ‘odontodes’; Sire & Huysseune, 2003) along the dorsal skin were slightly Alcian blue positive (Fig. 2F, arrowhead). Alizarin red staining in 5-cm shark embryos was only detected in dermal denticles of the caudal fin (Fig. 2J,K), a finding confirmed in embryos that did not undergo the acid alcohol treatment of Alcian blue staining (data not shown). In the heads of 9-cm embryos, signs of mineralization appeared in teeth of the oral cavity (Fig. 2C). In the trunk, two rows of dermal denticles along the dorsal skin were Alizarin red positive (Fig. 2G). In 9-cm embryos stained only for Alizarin red, portions of the vertebrae were also mineralized, but other endoskeletal elements, such as radials of the fins, remained unmineralized (Fig. 2H; data not shown). Apart from an increase in the number of mineralized teeth, no changes were apparent in the patterns of Alcian blue/Alizarin red staining in the head of 12-cm embryos, although the acid alcohol Alcian blue solution may have rendered some lightly mineralized elements undetectable in these specimens (Fig. 2D). By contrast, Alizarin red staining demonstrated mineralization in additional domains of the vertebrae at this and later stages of development (Fig. 2I,M). The heads of near-hatching 15-cm embryos showed numerous mineralized dermal denticles, and most cartilage elements, such as Meckel's cartilage and the palatoquadrate of the jaw apparatus, were Alizarin red positive (Fig. 2E; see also Fig. 5I). In addition to the heavily mineralized vertebrae, most elements of the 15-cm shark trunk endoskeleton were Alizarin red positive, including the pectoral and pelvic girdles and the majority of fin radials (Fig. 2I; data not shown). By contrast, no mineralization was evident in radials of the pectoral and pelvic fins (data not shown). The dermal denticles of the tail, which were the first signs of mineralization in the 5-cm swell shark embryo, persisted in the 15-cm embryo (Fig. 2L, arrowheads), and were characterized by a morphology that differed from other dermal denticles (compare Fig. 2K with Fig. 2E,I).

Figure 2.

Exoskeletal and endoskeletal mineralization during swell shark ontogeny. (A) Dorsal view of 15-cm swell shark, stained with Alcian blue and Alizarin red. (B–E) Ventral views of swell shark lower jaws, anterior to top. Meckel's cartilage (Mk) stained with Alcian blue in 5-cm (B), 9-cm (C), 12-cm (D) and 15-cm (E) embryos. Alizarin red stained teeth (t) in 9-cm, 12-cm and 15-cm oral surfaces (C–E), and dermal denticles (dd) on 15-cm head (E). (F–I) Dorsal views of swell shark spines, anterior to left. Vertebrae (v) stained with Alcian blue in 5-cm (F), 9-cm (G) and 15-cm (I) embryos; note slight Alcian blue staining of dermal elements in 5-cm embryo (F, arrowhead). Vertebrae stained for mineralization in 9-cm embryos (H) that did not undergo acid alcohol treatment (compare with G), and in 15-cm embryos (I). Alizarin red staining also marked development of dermal denticles along the dorsal surface of the spine in 9-cm (G) and 15-cm embryos (I); dermal denticles were removed from the sample in H. (J–L) Lateral views of swell shark caudal fins, dorsal to top, anterior to left. First signs of mineralization detected in 5-cm embryos (J,K); the caudal dermal denticles demonstrate a different morphology than either the initial rows along the dorsal surface or those that eventually cover the body (compare K to E, I and L), and they persist in 15-cm embryos (L). (M–Q) Transverse views of swell shark vertebrae, dorsal to left. The 15-cm embryo (M) illustrates three portions of the vertebrum: neural arches (na), centrum (c) and hemal arches (ha). Alcian blue stained neural arches of 9-cm embryo (N). Alizarin red staining was observed in neural arches of a 9-cm embryo that did not undergo acid alcohol treatment (O). Txined with Alcian blue and an outer tissue (black arrowhead) that stained with Alizarin red. Three distinct ringed layers of the centrum (c) in a 15-cm embryo (Q) were visible: inner and outer layers stained with Alcian blue, while a middle layer stained for mineralization. Abbreviations: c = centrum; dd = dermal denticle; ha = hemal arch; Mk = Meckel's cartilage; na = neural arch; v = vertebrum. Scale bars: A = 1 cm; B–D, F–H = 1 mm; E, I, J, L, M = 2 mm; K, N–Q = 0.5 mm.

Figure 5.

Shark Meckel's cartilage development and mineralization. (A–H) Sagittal sections of lower jaws in 5-cm (A,C,E,G) and 9-cm (B,D,F,H) swell shark embryos. Meckel's cartilage (Mk) stained with Safranin O and Alcian blue in 5-cm (A,C) and 9-cm (B,D) embryos. No staining of Meckel's cartilage with direct red of the HBQ protocol was observed in 5-cm (C) or 9-cm (D) embryos. Chondrocyte hypertrophy was apparent in Meckel's cartilage of the 9-cm embryo (B, arrows). No Aniline blue staining of the Trichrome protocol was observed in Meckel's cartilage of 5-cm (E) or 9-cm (F) embryos. No alkaline phosphatase activity was discerned in Meckel's cartilage of the 5-cm embryo (G). Intense alkaline phosphatase activity was evident in discrete groups of cells (arrowheads) near the surface of Meckel's cartilage in the 9-cm embryo (H); note alkaline phosphatase activity in the developing tooth (H, arrow). (I) Ventral view of lower jaw in 15-cm swell shark embryo. Superficial Alizarin red staining was apparent in Meckel's cartilage of a 15-cm embryo (I). Abbreviations: AP = alkaline phosphatase; Mk = Meckel's cartilage; SafO = Safranin O; Tri = Trichrome. Scale bars: A–H = 200 µm; I = 1 mm.

To decipher the exact parts of swell shark vertebrae that were mineralized, we examined trunk segments that were dissected transversely and stained. In each repeating segment of the shark vertebral column, there are three anatomical components: (1) the centrum, which is the main vertebral body surrounding the notochord, (2) two sets of paired neural arches, which are processes that extend dorsally from the centrum to encase the neural tube, and (3) a set of two hemal arches, which are processes that extend ventrally from the centrum (Fig. 2M). Alizarin red staining of near-hatching 15-cm swell shark embryos revealed two precise locations of mineralization in the vertebrae: neural arches and centrum (Fig. 2M). While neural arches of 5-cm embryos stained only for Alcian blue (Fig. 2F), two stains marked neural arches of 9-cm shark embryos: Alcian blue (Fig. 2N) and Alizarin red (Fig. 2O). The topological arrangement between Alcian blue-positive and Alizarin red-positive tissues was clarified by analyses of 15-cm embryos, which demonstrated an inner cartilaginous rod surrounded by a mineralizing tissue (Fig. 2P, arrowheads). In addition, a ring of tissue between an inner and outer layer of Alcian blue-staining tissues in the centrum showed signs of mineralization in 12-cm (data not shown) and 15-cm embryos (Fig. 2Q).

Section analyses of mineralizing shark tissues

To characterize more thoroughly the skeletal tissues within the mineralizing elements identified by our whole-mount Alizarin red stains, we performed a series of histological, histochemical and immunohistochemical stains on tissue sections of representative mineralized elements throughout swell shark development.

Neural arch

Two distinct histological and immunohistochemical tissues became apparent through analyses of neural arches during swell shark embryogenesis. No histological evidence of skeletal tissues was present anywhere in the body in representative sections of 3-cm C. ventriosum embryos, yet skeletogenic condensations were apparent in 4-cm embryos (data not shown). Growing dorsally from the centrum around the neural tube, cells of the neural arch in the 5-cm embryo were surrounded by matrix that stained with Safranin O and Alcian blue (Fig. 3A,D). Neither direct red staining, Aniline blue staining nor alkaline phosphatase activity was present in neural arches of 5-cm embryos (Fig. 3D,G,J). A Safranin O- and Alcian blue-positive cartilaginous rod extended dorsally to encompass the neural tube in neural arches of 9-cm embryos; chondrocytes appeared non-hypertrophic (Fig. 3B,E; see also Fig. 2N). A second layer of tissue in the perichondral region surrounded the central layer at this stage of shark neural arch development. This tissue did not stain for Safranin O or Alcian blue, but stained strongly for direct red and Aniline blue (Fig. 3B,E,H; note that the tissue sections in Fig. 3B,E do not include the central layer completely, but rather slice tangentially through the perichondral region). The entire perichondral tissue layer and discrete groups of cells in the periphery of the central cartilage layer showed evidence of alkaline phosphatase activity (Fig. 3K), suggesting that both tissues mineralize.

Figure 3.

A bone-like tissue in neural arches of the shark vertebrae. (A–N) Transverse sections of swell shark neural arches in 5-cm (A,D,G,J), 9-cm (B,C,E,F,H,I,K,L) and 15-cm (M,N) embryos. Cells producing a matrix that stained with Safranin O (A) and Alcian blue (D) projected dorsally from the centrum (c) around the neural tube (nt), forming the neural arch (na), in 5-cm embryos. The neural arch of 9-cm embryos had extended dorsally to encompass completely the neural tube, and contained two tissue layers; in addition to a central layer (cl) that stained with Safranin O, a surrounding layer (sl) did not stain with Safranin O (B). A tangential section of the 9-cm neural arch demonstrated that the central layer (of which only the most dorsal and ventral aspects were represented in this section) stained with Alcian blue, while the surrounding layer stained with direct red (E). Neither Aniline blue staining in the Trichrome protocol (G) nor alkaline phosphatase activity (J) was observed in the neural arch of 5-cm embryos. Aniline blue staining was strong only in the surrounding layer of the neural arch in the 9-cm embryo (H). Evidence of alkaline phosphatase was strong in both the central and the surrounding layers of the neural arch in the 9-cm embryo (K). Picro Sirius-stained neural arches of 9-cm embryos indicated that the surrounding layer contained tight bundling of collagen fibres, as revealed by red birefringence under polarized light; by contrast, the central layer showed loose collagen organization (C). The central layer of the 9-cm neural arch demonstrated strong immunoreactivity for Collagen type 2 (Col2), while the surrounding layer was Col2-negative (F). The central layer of the 9-cm neural arch showed no immunoreactivity for Collagen type 1 (Col1), while the surrounding layer was Col1-positive (I). An antibody for the precursor peptide of Col1 (pre-Col1) was reactive only at the dorsal aspect of the surrounding layer in the neural arch of the 9-cm embryo (L). High-magnifications of swell shark neural arches in 15-cm embryos revealed rounded, well-separated cells in the central layer (M) and flattened, more closely spaced cells in the surrounding layer (N). Abbreviations: AP = alkaline phosphatase; c = centrum; cl = central layer; Col1 = Collagen type 1; Col2 = Collagen type 2; na = neural arch; nt = neural tube; pre-Col1 = precursor peptide of Col1; SafO = Safranin O; sl = surrounding layer; Tri = Trichrome. Scale bars: A,D,G,J = 200 µm; B,C,E,F,H,I,K,L = 400 µm; M,N = 20 µm.

Further analyses revealed that the two tissues in the 9-cm swell shark neural arch differed clearly in collagen organization and composition, and also cellular morphology. To assay collagen organization, we analysed Picro Sirius-stained sections under polarized light (Junqueira et al. 1979). The central layer showed a loose organization of collagen (i.e. black to green birefringence), while the surrounding layer's red birefringence indicated tightly packed collagen bundles (Fig. 3C). With regard to collagen composition, the central cartilage layer was strongly immunoreactive to antibodies against Col2, whereas the surrounding layer in the perichondral region was Col2-negative (Fig. 3F). In stark contrast, the central layer demonstrated no immunoreactivity for Col1, but the surrounding layer was Col1-positive (Fig. 3I). The precursor peptide for Col1 appeared to be expressed only at the most dorsal aspect of each neural arch, suggesting that only these cells were actively secreting a Col1 matrix (Fig. 3L). No changes in these staining patterns were observed in later staged embryos (data not shown). Finally, even by analysis under light microscopy, a graded difference in cellular morphology was apparent between inner and outer tissues of the shark neural arch. Cells surrounded by Alcian blue-positive matrix, for example, deep in the central tissue layer of the neural arch of 15-cm swell shark embryos were rounded, well separated by extracellular matrix, and somewhat randomly distributed (Fig. 3M). Adjacent to this cell population towards the surface of the neural arch were cells of a similar morphology, but that were surrounded by direct red-positive matrix (data not shown). Still further superficially, cells surrounded by direct red-positive matrix were spindle-shaped, packed closer together and aligned in a lamellar conformation (Fig. 3N).

Oral exoskeleton

To provide an internal comparison for our staining procedures, we analysed tissue designated as bone that supports chondrichthyan dermal denticles (Moss, 1970, 1977; Reif, 1980; Sire & Huysseune, 2003), focusing on oral denticles (i.e. teeth) of the 15-cm swell shark lower jaw. A tissue extending from and continuous with dentine located at the base of the tooth stained with direct red and Aniline blue, but showed no staining with Alcian blue or Safranin O (arrows in Fig. 4A–D). The denticle-supporting tissue demonstrated strong alkaline phosphatase activity and immunoreacted weakly with antibodies for Col1 (Fig. 4E,F). Analysis of collagen organization under polarized light revealed that the collagen bundles were not tightly packed in the supporting tissue of the oral denticle (Fig. 4G,H).

Figure 4.

A bone-like tissue supporting the shark tooth. (A–H) Sagittal sections through the distal tip of the lower jaw in 15-cm swell shark embryo. Low (A) and higher magnification (B–F; black box in A) views revealed a tissue (arrow) supporting the tooth (t) that stained with direct red of the HBQ protocol (A,B) and Aniline blue of the Trichrome protocol (C). This tissue did not stain with Alcian blue (A,B) or Safranin O (D), but demonstrated alkaline phosphatase activity (E). While the tooth-supporting tissue showed some immunoreactivity to Collagen type 1 (F), high-magnification views (G,H; white box in F) of this tissue (outlined) in Picro Sirius-stained sections revealed that the collagen fibrils were not tightly wound. H is view of tissue in G under polarized light. Abbreviations: AP = alkaline phosphatase; Col1 = Collagen type 1 antibody; Mk = Meckel's cartilage; SafO = Safranin O; t = tooth; Tri = Trichrome. Scale bars: A = 800 µm; B–F = 200 µm; G,H = 100 µm.

Meckel's cartilage

Similar to the neural arch in 5-cm swell shark embryos, Meckel's cartilage stained with Safranin O and Alcian blue at this young stage, but did not stain with direct red or Aniline blue (Fig. 5A,C,E). In addition, Meckel's cartilage did not show signs of alkaline phosphatase activity in 5-cm C. ventriosum embryos (Fig. 5G). In 9-cm embryos, the perichondral region of Meckel's cartilage demonstrated alkaline phosphatase activity in discrete groups of cells that would correspond to the chondrichthyan pattern of tesserae (Kemp & Westrin, 1979; Summers et al. 1998; Dean & Summers, 2006), but otherwise showed the same staining patterns as observed in 5-cm embryos (Fig. 5B,D,F,H). In addition to the difference in alkaline phosphatase activity between cells in the perichondral region of 9-cm Meckel's cartilage and those inferior to them, the morphology of chondrocytes in these two regions varied. For example, chondrocytes appeared more hypertrophic and spaced apart in central regions of Meckel's cartilage (arrows in Fig. 5B) when compared with the spindle-shaped, more closely packed chondrocytes in the perichondral region. Also, the perichondral chondrocytes were surrounded by matrix that stained more deeply for Alcian blue (Fig. 5D). No changes in these results were observed in Meckel's cartilage of 12- or 15-cm shark embryos (see Fig. 4B–E), despite the presence of a thin layer of Alizarin red staining in the perichondral region of Meckel's cartilage in 15-cm shark embryos (Fig. 5I). For example, neither direct red, Aniline blue nor Col1 staining was observed in perichondral regions of Meckel's cartilage in 15-cm swell shark embryos (see Fig. 4B,C,F). In addition, analysis of collagen organization under polarized light revealed that the collagen bundles were not tightly packed in Meckel's cartilage of 15-cm embryos (see Fig. 4G,H; data not shown).

Centrum

Finally, the chondrichthyan centrum contains an Alizarin red-positive ring of tissue (see Fig. 2M,Q) that is thought to be a unique form of mineralization, termed ‘areolar calcification’ (Dean & Summers, 2006; after Ørvig, 1951). Therefore, we included it with our analyses of mineralizing tissues in the swell shark embryo. In 5-cm embryos, the centrum consisted of two layers of cells whose matrix stained with the dyes Safranin O and Alcian blue (Fig. 6A,C). Cells of the inner layer had a flattened morphology compared with those in the outer layer. No staining with direct red or Aniline blue was observed in these tissues (Fig. 6C,E), and no alkaline phosphatase activity was demonstrable (Fig. 6G). In 9-cm embryos, three cell layers were apparent in the centrum, which is in agreement with Daniel (1934). Inner and outer layers stained with Safranin O and Alcian blue, while a middle layer stained with direct red and Aniline blue (Fig. 6B,D,F). The inner layer displayed a more aberrant cellular morphology than the outer, containing some hypertrophic chondrocytes (see inset Fig. 6B′, which is high-magnification view of the boxed area in Fig. 6B). Alkaline phosphatase activity extended from the notochord into all three layers (Fig. 6H). The middle layer of the centrum showed yellow birefringence when Picro Sirius-stained sections were viewed under polarized light, while surrounding layers typically appeared green (data not shown), suggesting that the middle layer contained somewhat more tightly packed collagen fibres. Slightly greater Col1 immunoreactivity and lower Col2 immunoreactivity were present in the middle layer, compared with the inner and outer layers, but this middle tissue layer appeared to be both Col1-positive and Col2-positive (data not shown).

Figure 6.

A bone-like tissue in the shark vertebral centrum. (A–H) Transverse sections of swell shark centra in 5-cm (A,C,E,G) and 9-cm (B,D,F,H) embryos. Safranin O staining (A) and Alcian blue of the HBQ protocol (C) distinguished two tissue layers in the centrum of vertebrae in the 5-cm shark embryo: lighter, more diffuse staining of an inner layer (il) of tissue adjacent to the notochord (n), and deeper, more concentrated staining in an outer layer (ol) of tissue. In 9-cm embryos, the centrum was divided into three tissue layers: inner (il) and outer (ol) layers stained with Safranin O, while a middle layer (ml) remained unstained by Safranin O (B). Note hypertrophic cells in the inner layer of the centrum (inset B′, which is a high-magnification of cells in the boxed area in B). Similar to Safranin O staining of the 9-cm centrum, Alcian blue stained inner and outer layers, but the middle layer stained slightly with direct red of the HBQ protocol (D). No Aniline blue staining in the Trichrome protocol was apparent in the 5-cm centrum (E), while only the middle layer of the 9-cm centrum stained slightly with Aniline blue (F). No alkaline phosphatase activity was present in the 5-cm centrum (G), but evidence of alkaline phosphatase was detected in all three layers of the 9-cm centrum (H). Abbreviations: AP = alkaline phosphatase; il = inner layer; ml = middle layer; n = notochord; ol = outer layer; SafO = Safranin O; Tri = Trichrome. Scale bars: A,C,E,G = 100 µm; B,D,F,H = 200 µm.

Discussion

Shark bone?

Classifying shark tissues as bone is challenging, because many intermediate tissues characterize the fish skeleton (Beresford, 1981; Benjamin, 1989; Smith & Hall, 1990; Sire & Huysseune, 2003). In fact, a survey of immunohistochemical properties of cartilage and bone in numerous teleost species revealed some histologically identified cartilages to lack Col2 and even a few histologically identified bones to contain abundant Col2 (Benjamin & Ralphs, 1991). As pointed out by Huysseune (2000), ‘It is nevertheless important to attempt to classify the different skeletal tissues as a workable tool for studying form and function in the fish skeleton.’ In searching for signs of an osteogenic programme during C. ventriosum skeletogenesis, we focused on those skeletal elements that have been reported to form bone in sharks, such as neural arches of the endoskeleton (Kemp & Westrin, 1979; Peignoux-Deville et al. 1982; Bordat, 1987) and supporting exoskeletal tissue of dermal denticles (Moss, 1970; Moss, 1977; Reif, 1980; Sire & Huysseune, 2003).

Perichondral tissue of the C. ventriosum neural arch had more bone-like characteristics (Fig. 1; Ham & Cormack, 1987; Eames et al. 2003; Eames & Helms, 2004) than exoskeletal tissue supporting the oral denticle, but neither should be called bone. Both the denticle-supporting tissue and neural arch perichondral tissue had similar histological properties, such as Aniline blue and direct red staining of collagen, and were also alkaline phosphatase-positive. However, the denticle-supporting tissue did not immunoreact as strongly with antibodies for Col1, and did not display tight organization of collagen bundles, features that are at odds with the perichondral tissue in C. ventriosum neural arches. Notably, the finding of abundant Col1 in homogenized shark cartilage (Rama & Chandrakasan, 1984; Sivakumar & Chandrakasan, 1998) either is due to a perichondral bone-like tissue or reflects a feature of adult shark cartilage not yet attained during C. ventriosum embryogenesis.

Regarding cellular aspects of shark bone-like tissues, the denticle-supporting tissues of C. ventriosum and other elasmobranchs are acellular (Fig. 4; also Moss, 1970, 1977), while the perichondral neural arch tissues of C. ventriosum and other elasmobranchs are cellular (Fig. 3; also Peignoux-Deville et al. 1982; Bordat, 1987). While vertebrate bone can be cellular or acellular, the two types are rarely observed in the same specimen (Huysseune, 2000). Addressing this potential conundrum, we note that the tissue supporting C. ventriosum oral denticles is continuous with a tissue agreed to be dentine. Indeed, Miyake et al. (1999) were apprehensive in identifying supporting tissue of dermal denticles in the chondrichthyan Leucoraja erinacea as bone, noting many similarities to dentine. Neural arch Col1 staining was not as intense as seen in tetrapod bone controls (compare Fig. 3I with Fig. 1F). Although this may represent a technical problem of antibody affinity for tetrapod Col1 fibres, our ontogenetic series revealed another aberration from typical bone tissue formation. The creation of cellular bone involves a developmental transition whereby osteoblasts at the developing bone's surface become osteocytes surrounded by bone matrix (Franz-Odendaal et al. 2006), and no ‘osteoblast-like’ cells are obvious at the surface matrix of the bone-like neural arch tissue even in early stages of C. ventriosum development.

If not bone, then what should the C. ventriosum endoskeletal tissue be called? Features of chondroid bone in the cichlid Hemichromis bimaculatus (Col1-positive, Col2-negative matrix surrounding chondrocyte-like cells; Huysseune, 1989; Huysseune & Sire, 1990) are extremely similar to some aspects of the perichondral tissue in C. ventriosum neural arches, but the two tissues differ in anatomical location. In both cranial and axial skeletons, teleost chondroid bone is located invariably at sites of articulation (Benjamin & Ralphs, 1991; Benjamin et al. 1992; Huysseune & Sire, 1992), which differs from the bone-like tissue in C. ventriosum neural arches. Another tissue type that shares features with those observed in the bone-like tissue of C. ventriosum neural arches is fibrocartilage, as both tissues surround ‘true’ matrix-rich hyaline cartilage (Figs 2 and 3; Benjamin, 1990). Also, fish fibrocartilage contains many dense collagen fibres (Benjamin, 1990), which would cause similar histological staining properties as those reported here in shark vertebrae. In contrast to typical cuboidal osteoblasts, cells in the C. ventriosum neural arch perichondral region displayed a flattened and fibroblastic morphological appearance. Our new data thus support Ørvig (1951), who classified the perichondral tissue of shark neural arches as ‘some kind of calcified fibrous cartilage’.

On the evolution of endochondral ossification

Examination of the fossil record reveals that all stages of the process of endochondral ossification did not appear simultaneously during vertebrate phylogeny. Perichondral bone deposition, for example, preceded cartilage degradation and endochondral bone deposition, given that the former feature can be found in primitive jawless vertebrates, such as galeaspids and heterostracans, while the latter features first appear in primitive jawed vertebrates, such as placoderms (cited in Donoghue & Sansom, 2002) or osteichthyans (Hall, 1975; Smith & Hall, 1990; Janvier, 1996). What remains unclear, however, is how the perichondral region of cartilage elements evolved an osteogenic programme in the first place. To achieve some clarity on this topic, two points would need to be addressed. First, distinct tissue characteristics would have to be spatially resolved between perichondral and deep regions of a cartilage element. Second, the perichondrium must somehow acquire osteogenic potential. As extant agnathans do not mineralize, let alone ossify, their endoskeleton, chondrichthyan endoskeletal development may offer the best insight into this important evolutionary transition in vertebrate skeletogenesis.

Establishing domains within shark skeletal elements

Subanatomical features became apparent through our analyses of skeletal elements during C. ventriosum ontogeny. A shared feature of the Chondrichthyes (Janvier, 1996; Moss, 1977), extant sharks and skates mineralize the surface of their cartilages using prisms of crystalline calcium phosphate in discrete blocks termed tesserae (Kemp & Westrin, 1979; Dingerkus et al. 1991; Summers et al. 1998; Dean & Summers, 2006). How are these tissue characteristics spatially resolved? Our ontogenetic series of C. ventriosum demonstrates for the first time that during shark skeletogenesis, as described in other vertebrates, alkaline phosphatase activity predicts this mineralization pattern. For example, small groups of alkaline phosphatase-positive cells are in perichondral regions of the unmineralized Meckel's cartilage in 9-cm C. ventriosum embryos. In addition to mineralization markers, other features distinguish perichondral and core regions of C. ventriosum cartilage elements. In contrast to tissue in the perichondral region of Meckel's cartilage, core chondral tissue stained intensely for the Col2 antibody, for example, but stained less strongly for Alcian blue. The core chondral tissue also contained somewhat hypertrophic cells. Whether these microanatomical features of shark hyaline cartilage are under genetic or epigenetic control remains unclear, but could be tested using in vitro organ culture systems.

Through what cellular mechanisms are core and perichondral regions of C. ventriosum elements patterned? In agreement with Kemp & Westrin (1979) and Huysseune & Sire (1992), we favour the interpretation that a given population of perichondral scleroblasts secretes both matrices in succession during ontogeny. Huysseune & Sire (1992) documented gradual changes to extracellular matrix during ontogeny of Meckel's cartilage in the cichlid H. bimaculatus. Our developmental series demonstrated that perichondral cells of the 5-cm C. ventriosum neural arch abutted Safranin O-, Alcian-blue-positive matrix, while perichondral cells of the 9-cm neural arch abutted direct red-, Aniline blue-positive matrix. Indicating conserved homology of process, previous studies in chick and mouse suggest such a transition from prechondroblast to preosteoblast during endochondral ossification (Kahn & Simmons, 1977; Fang & Hall, 1997; Nakashima et al. 2002). The embryonic origins of perichondral and deep tissue populations in chondrichthyan skeletal elements can be identified by injecting perichondral cells with lineage tracer, for example, and tracking the fates of their progeny through development.

Evolution of the periosteum

If tissue characteristics can be resolved spatially between perichondral and core regions of a cartilage element, then the phylogenetic transformation of the perichondrium to a periosteum in primitive jawless vertebrates (Hall, 1975; Maisey, 1988; Smith & Hall, 1990; Janvier, 1996; Donoghue & Sansom, 2002) may only have involved the acquisition of osteogenic potential in the perichondral region. Skeletal studies in avians and mammals have revealed key molecules for specifying preosteoblasts in the perichondrium, such as transcription factors (e.g. Runx2 and Osx; Komori et al. 1997; Nakashima et al. 2002; Eames et al. 2004) and growth factors (e.g. Ihh and Wnt's; St-Jacques et al. 1999; Mak et al. 2006; Rodda & McMahon, 2006). Therefore, future studies on such molecules during development of C. ventriosum may identify genetic pathways that underlie the perichondrium-to-periosteum transition.

What would generally contribute much to this discussion is sufficient knowledge about the ‘ancestral periosteum’. The peculiar cellular and extracellular matrix features in perichondral vs. deep regions of mineralized cartilages and neural arches of C. ventriosum may either be neomorphic adaptations during evolution of extant chondrichthyans, or they may represent unique, primitive conditions. That is to say, if primitive sharks indeed formed bone in the perichondrium – a feature lost in living sharks – then perhaps significant characteristics of the C. ventriosum endoskeleton are derived. On the other hand, having diverged some 440 million years ago from all other living vertebrates that make a mineralized endoskeleton, chondrichthyans may be our best window into the ancestral condition. Palaeontology may provide the final answer to the issue of the ‘ancestral periosteum’, but comparative anatomy should help. In this regard, our studies on C. ventriosum neural arches agree with those on other elasmobranchs (Kemp & Westrin, 1979; Peignoux-Deville et al. 1982; Rama & Chandrakasan, 1984; Bordat, 1987; Sivakumar & Chandrakasan, 1998), but whether they represent the endoskeleton of extant chondrichthyans would be addressed by similar studies on chimaeras. Furthermore, if C. ventriosum endoskeletal development reflected that of primitive vertebrates, then comparable staining patterns may be found in endoskeletal elements of basal extant actinopterygians and sarcopterygians.

Acknowledgements

Many thanks to Shane Anderson of UC-Santa Barbara for his selfless efforts to provide shark embryos to interested scientists and educators, and to Drs Brian Hall, Ann Huysseune, Philippe Janvier, Charles Kimmel, Moya Smith and Eckhard Witten for their insightful comments on this project and manuscript. Col1 and Col2 antibodies were obtained from the DSHB, maintained by the University of Iowa under the auspices of the NICHD. This study was funded, in part, by an NIH fellowship F32 DE016778-01 to B.F.E.; by R01 DE00432 and R01 DE13017 from the NIDCR to J.A.H.; and by R03 DE014795-01 and R01 DE016402-01 from the NIDCR, Research Grant 5-FY04-26 from the March of Dimes Birth Defects Foundation, and UCSF Academic Senate and REAC grants to R.A.S.

Ancillary