Ultrastructural and developmental features of the tessellated endoskeleton of elasmobranchs (sharks and rays)

Abstract

The endoskeleton of elasmobranchs (sharks and rays) is comprised largely of unmineralized cartilage, differing fundamentally from the bony skeletons of other vertebrates. Elasmobranch skeletons are further distinguished by a tessellated surface mineralization, a layer of minute, polygonal, mineralized tiles called tesserae. This ‘tessellation’ has defined the elasmobranch group for more than 400 million years, yet the limited data on development and ultrastructure of elasmobranch skeletons (e.g. how tesserae change in shape and mineral density with age) have restricted our abilities to develop hypotheses for tessellated cartilage growth. Using high‐resolution, two‐dimensional and three‐dimensional materials and structural characterization techniques, we investigate an ontogenetic series of tessellated cartilage from round stingray Urobatis halleri, allowing us to define a series of distinct phases for skeletal mineralization and previously unrecognized features of tesseral anatomy. We show that the distinct tiled morphology of elasmobranch calcified cartilage is established early in U. halleri development, with tesserae forming first in histotroph embryos as isolated, globular islets of mineralized tissue. By the sub‐adult stage, tesserae have increased in size and grown into contact with one another. The intertesseral contact results in the formation of more geometric (straight‐edged) tesseral shapes and the development of two important features of tesseral anatomy, which we describe here for the first time. The first, the intertesseral joint, where neighboring tesserae abut without appreciable overlapping or interlocking, is far more complex than previously realized, comprised of a convoluted bearing surface surrounded by areas of fibrous attachment. The second, tesseral spokes, are lamellated, high‐mineral density features radiating outward, like spokes on a wheel, from the center of each tessera to its joints with its neighbors, likely acting as structural reinforcements of the articulations between tesserae. As tesserae increase in size during ontogeny, spokes are lengthened via the addition of new lamellae, resulting in a visually striking mineralization pattern in the larger tesserae of older adult skeletons when viewed with scanning electron microscopy (SEM) in backscatter mode. Backscatter SEM also revealed that the cell lacunae in the center of larger tesserae are often filled with high mineral density material, suggesting that when intratesseral cells die, cell‐regulated inhibition of mineralization is interrupted. Many of the defining ultrastructural details we describe relate to local variation in tissue mineral density and support previously proposed accretive growth mechanisms for tesserae. High‐resolution micro‐computed tomography data indicate that some tesseral anatomical features we describe for U. halleri are common among species of all major elasmobranch groups despite large variation in tesseral shape and size. We discuss hypotheses about how these features develop, and compare them with other vertebrate skeletal tissue types and their growth mechanisms.

Introduction

Whereas the vast majority (~ 98%) of vertebrate species have bony endoskeletons, the skeletons of elasmobranch fishes (sharks, rays and relatives) are comprised largely of unmineralized hyaline‐like cartilage (Leydig, 1852; Hall, 2005; Currey, 2002; Atkins et al. 2014). As unmineralized cartilage is considerably less stiff than bone (Ashby et al. 1995), it is remarkable that sharks and rays represent such an evolutionary successful taxon, constituting some of the largest and fastest marine apex predators for more than 400 million years (Maisey, 2013; Long et al. 2015). The high performance of elasmobranch cartilage is surely linked to the fact that the majority of the skeleton is essentially armored: the uncalcified cartilaginous core of each piece of the skeleton is covered by a layer of mineralized tiles called tesserae, and then further wrapped in an outer sheath of fibrous connective tissue (perichondrium; Fig. 1; Leydig, 1852; Benzer, 1944; Applegate, 1967; Kemp & Westrin, 1979; Clement, 1992; Dean et al. 2009).

Figure 1.

Tessellated endoskeleton of elasmobranchs (sharks and rays). (A) Photograph of round stingray Urobatis halleri and insets of micro‐computed tomography (μCT)‐images showing the jaws, covered in tesserae. The final inset shows a schematic view of three tesserae (T1–3) and their non‐mineralized fibrous connections. (B) Schematic of the organization of elasmobranch tessellated cartilage: skeletal elements are comprised of an uncalcified cartilage core and armored with a hard outer layer of mineralized tesserae and unmineralized fibrous perichondrium. (C) These icons are used throughout the article to indicate the tesseral views and sectioning planes used in this study, providing consistent anatomical perspectives.

The mineralized cortex of elasmobranch cartilage has profound functional implications, reinforcing the skeleton and providing a stiff surface for muscular attachment. To cope with the mechanical demands of the adult animal, most vertebrates almost completely replace their embryonic cartilaginous skeletons with bone during development. Bone is a dynamic material where cells orchestrate removal, deposition and remodeling of mineralized tissue, allowing ongoing growth and damage repair, even under the regular loading regimes of daily life (Hall, 2005; Currey, 2002; Atkins et al. 2014). In contrast, cartilage has limited repair ability (Ashhurst, 2004; Hall, 2005) and, when mineralized, apparently cannot remodel; hence, a continuous, mineralized cartilage cortex could grow only by apposition and, therefore, only thicken. The tessellation of the elasmobranch skeleton, however, provides space for growth in between tesserae, and it is theorized that new mineralized tissue is deposited at tesseral edges during development (Clement, 1992; Dean et al. 2009). The combination of mineralized tissue and unmineralized intertesseral joints is, therefore, vital to elasmobranch skeletal biology, providing both cortical stiffness for mechanical function and room for interstitial growth (between tesserae; Clement, 1992; Dean et al. 2009), while also likely permitting flexibility under some loading conditions (Liu et al. 2010; Fratzl et al. 2016).

The growth and mechanics of tessellated cartilage skeletons rely on the maintenance of linked, but separated, tiles; however, it remains unclear how the tessellated pattern is established and maintained, how tesserae interact during growth, and how tesseral morphology changes with age. There is some suggestion that the tessellations of young animals are quite different from the abutting, polygonal tiles of adults. Embryonic elasmobranch skeletons are uncalcified, and it appears that tessellation first arises close to parturition/hatching, in the form of clusters of isolated nuclei of mineral, associated with groups of chondrocytes and alkaline phosphatase (ALP) activity, and separated from one another by unmineralized cartilage (Benzer, 1944; Ørvig, 1951; Eames et al. 2007; Maisey, 2013). However, young animal tessellations have only been anecdotally described and the steps involved in their development into adult tesserae are unknown.

As a consequence, we are missing crucial steps in our understanding of tesseral development and mineralization, largely due to research being focused either only on a narrow age range (e.g. young, pre‐tessellate animals: Eames et al. 2007; early tessellate animals: Enault et al. 2015; adult animals: Schmidt, 1952; Moss, 1968; Kemp & Westrin, 1979; Clement, 1992) or on extinct tissues, where species’ age determination is more speculative (e.g. Ørvig, 1951; Schaeffer, 1981; Maisey, 2013; Long et al. 2015). Tesserae are small (typically less than 500 μm in all dimensions) and numerous, and available data suggest there could be a great variability of adult tesseral sizes and shapes across the skeletal surface and species. However, cross‐study comparisons are nearly impossible, as there has been no standardization in the sectioning planes, skeletal elements or species investigated. As a result, our understanding of tesseral ultrastructure has also suffered and been limited entirely to two‐dimensional perspectives with little anatomical context.

Here, we aim to characterize the development of tesserae at the ultrastructural level to render a three‐dimensional concept of tesserae and define diagnostic structural features. We use polarized light microscopy (PLM), electron microscopy [scanning electron microscopy (SEM), transmission electron microscopy (TEM)] and X‐ray micro‐computed tomography (laboratory and synchrotron radiation‐based micro‐computed tomography, μCT and SR‐μCT, respectively) to describe the changes tesserae undergo during ontogeny in the round stingray Urobatis halleri. Using a standardized sectioning technique, we present sections of ontogenetic series of tesseral mats in order to define age‐ and location‐specific tesseral characteristics, including aspects of size and shape, mineral accretion processes, mineral density distribution, and the interaction of tesserae and soft tissue. Further, we investigate variation in tesserae shape across adults of a variety of shark and ray species, to highlight common structural aspects of elasmobranch tessellated cartilage in general.

Materials and methods

Specimens

The round stingray U. halleri was chosen for this study because of the availability of full ontogenetic series, and of published developmental and structural data (Dean et al. 2009, 2010; Omelon et al. 2014), but also because the size of U. halleri skeletal elements and tesserae is particularly suitable for scanning in high‐resolution laboratory and SR‐μCT scanners. Note, throughout this article, we use the term ‘skeletal element’ to indicate a single, discrete piece of the elasmobranch skeleton, because an accepted term for elasmobranch skeletal elements is currently lacking in the literature. All specimens were donated from another study (Lyons et al. 2014), collected by beach seine from collection sites in San Diego and Seal Beach, California, USA. The animals were shipped on dry ice and stored in a freezer at −20 °C until sample preparation. Specimens were re‐thawed in warm water, before their skeletal elements (pectoral bar, propterygium and hyomandibula) were carefully removed and stored either in 75% ethanol at 4 °C or immediately processed for embedding (see below). We used these skeletal elements in our study because of their elongated and rodlike shape, facilitating sample preparation and providing large flat surfaces covered with tesserae. The first appearance of tesserae in U. halleri was previously reported for animals with a disc width (DW) of approximately 6.0 cm (Dean et al. 2009, 2010; Omelon et al. 2014). DW is a common size metric for batoid fishes and refers to the lateral dimension of the animal. Specimens for this study ranged from yolk sac embryos (1.9–5.6 cm DW), to histotroph embryos (lacking yolk sacs and feeding on intrauterine milk; 5.7–7.9 cm DW), to sub‐adults (8.0–15.5 cm DW), to adults (> 15.5 cm DW; Table 1); ontogenetic stage determination was according to Hale & Lowe (2008).

Table 1.

Developmental stages of round stingray Urobatis halleri

Micro‐computed tomography

To examine the development of tesserae across whole skeletal elements, we performed μCT scans of hyomandibulae from animals at four ontogenetic stages: 7.0 cm DW (female); 11.0 cm DW (male); 14.4 cm DW (female); 19.0 cm DW (female). Note, in this study inter‐sex comparisons are possible, because there are no appreciable differences in size between males and females of U. halleri, particularly for the sub‐adult stages (Hale & Lowe, 2008). After dissection, the skeletal elements were dehydrated in an ascending alcohol series and stored until scanning in 75% ethanol. Samples were mounted in clay, sealed in ethanol‐humidified plastic tubes and scanned with a Skyscan 1172 desktop μCT scanner (Bruker microCT, Kontich, Belgium). Scans for all samples were performed with voxel sizes of 4.89 μm, at 59 kV source voltage and 167 μA source current, over 360 ° sample rotation. Virtual cross‐sections of skeletal elements in Fig. 2 are averages of 20 (cross‐section) images (100 μm), generated in ImageJ.

Figure 2.

Armoring the endoskeleton – the development of calcified, tessellated cartilage. Micro‐computed tomography (μCT) imaging of the left hyomandibula of four different developmental stages (indicated at the top of each column) of the round stingray Urobatis halleri. The top and second rows show lateral and ventral views, respectively, of the whole skeletal element. In μCT, the earliest tesserae were observed in animals with a disc width (DW) of ~ 7 cm (histotroph animals; A), with the degree of mineralization increasing with age. Tesserae do not form simultaneously over the entire skeletal element, rather appearing first in ventral and chondocranial portions (asterisk and circle, respectively). Cross‐sections of the skeletal element (third row, section position indicated by the white lines in the second row) show an increase in number, width and depth of tesserae with age. Note the tesseral layer appears to be thicker in stark convex areas (averaged images over 100 μm = 20 images). Small tesserae amidst big tesserae suggest the development of new tesserae among existing ones (arrowhead, D bottom row). Scale bars in (A) apply also to (B–D).

To examine interspecific, gross structural variation in tesserae, as well as the shape variation of intertesseral joints, we performed SR‐μCT scans of mineralized tissue from the lower jaws of a variety of elasmobranch fishes, including nine batoid species (Dasyatis sabina, Myliobatis californica, Narcine bancroftii, Pteroplatytrygon violacea, Raja eglanteria, Raja stellulata, Rhinoptera bonasus, Torpedo californica, U. halleri) and four shark species (Heterodontus franciscii, Notorynchus cepedianus, Somniosus pacificus, Squatina californica). For each species, tissue samples, typically several centimeters square and ~ 1 cm thick, were excised from flat portions of the surfaces of the skeletal elements (mid‐shafts of the lower jaws). The perichondrium was not removed before excision; samples were, therefore, expected to contain all components of tessellated cartilage: a layer of tesserae, sandwiched between perichondrium and uncalcified cartilage. All samples were freeze‐dried for better stability in scanning, mounted with beeswax directly onto sample stubs, and scanned using high‐resolution propagation phase‐contrast X‐ray micro‐tomography at the beamline ID19 of the European Synchrotron Radiation Facility (ESRF, Grenoble, France). Samples were imaged with voxel sizes of 0.35–1.75 μm, using a multilayer monochromator, beam energy of 30 keV, propagation distance of 20–50 mm, 0.5–1.0 s exposures, in continuous mode and over 180 ° sample rotation. Anatomy and ultrastructure from all μCT scans were investigated in two‐dimensional slices and three‐dimensional volumetric reconstructions using ZIB‐Amira software (Zuse Institute Berlin, Germany).

Scanning electron microscopy

To examine the development of tesserae on the ultrastructural level, we performed SEM in combination with a backscatter electron detector (backscatter SEM) of vertical and planar sections of tesserae (Fig. 1C) from animals at four ontogenetic stages: 6.0 cm DW (female); 7.5 cm DW (female); 8.5 cm DW (female); 20.0 cm DW (male); and 24 cm DW (female; Fig. 3). Pectoral bars were dissected from U. halleri, bisected longitudinally and trimmed down to tessellated strips with little uncalcified cartilage backing. We air‐dried and simultaneously flattened the tesseral layer between Teflon plates to prevent sticking. Dried samples were cut into smaller pieces and placed in a custom‐built PMMA (plastic resin) holder, according to the desired orientation (vertical or planar). Samples were embedded in PMMA, cut in slices (300 ± 100 μm thick; Buehler IsoMet low speed saw) and mounted on a PMMA object slide using double‐faced adhesive tape. Sections were polished with sandpaper plates with descending grain sizes (Logitech PM5 Precision Lapping and Polishing Machine), and finally using a soft polishing plate with diamond spray (0.25 μm grain size).

Figure 3.

Tiling a growing surface – development of tesserae. Vertical sections (upper row) and planar sections (lower row) of tesserae from an age series of Urobatis halleri. All images are from backscatter scanning electron microscopy (SEM) and therefore show mineralized tissue and variations in mineral density. (A, B) The first, poorly formed tesserae (T) were visible, in animals with ~ 6 cm disc width (DW), as thin strips of globular mineralized tissue between chondrocytes (ch), in the uncalcified cartilage core (uc) some distance from the perichondrium (pc). As animals grow (moving from left to right in the figure), tesserae approach the perichondrium, growing larger and closer together, and tiling becomes more regular, especially after tesserae come into contact. (C, D) As tesserae grow, chondrocytes are entombed in the mineral phase, enclosed in lacunar spaces (ls). (E, F) Once tesserae come into contact, areas of higher mineral density (spokes, sp) develop at the margin of the tesserae, associated with the intertesseral contact zone (icz) of abutting tesserae. Spokes elongate as tesserae grow. Note in (E) and (G) that lacunar spaces at the perichondral side of tesserae are flatter and oriented parallel to the perichondrium, compared with the round lacunar spaces in the chondral portion of tesserae. Scale bar in (A) applies also to (C, E, G); scale bar in (D) applies also to (F, H).

Backscatter SEM, via the detection of backscattered electrons, allows visualization of differences in either tissue elemental density (e.g. mineral density) or elemental composition as variation in grayscale values. Images were acquired of polished samples in backscatter mode using a Field Emission‐Environmental Scanning Electron Microscope (FE‐ESEM, FEI Quanta 600F) in environmental mode (i.e. at low vacuum without sputtering) with an acceleration voltage of 10–12.5 kV. To determine the nature of the grayscale variation observed in backscatter SEM, we used a Tescan Vega‐3 SEM equipped with a Bruker X‐Flash 5030 energy‐dispersive spectrometer (EDS). All EDS spectra and elemental maps were acquired at 20 kV acceleration voltage at 15 mm working distance, and paired with images of the same regions of interest taken under the same conditions in backscatter mode.

Polarized light microscopy

To examine the orientation of collagen in and between tesserae, we performed PLM on planar sections of tesserae from U. halleri (13 cm DW, section thickness about 250 μm). The collagen fibers showed maximum birefringence at ± 45 ° relative to the position of the polarizer and analyzer. A lambda filter (or first‐order retardation plate) was used in the optical path of the microscope, to convert variations in the monochromatic birefringent signal to color, to further distinguish among different orientations of collagen fibers in the sample.

Light microscopy

Samples of the anterior shoulder girdle (propterygium) were collected from an adult male U. halleri (20.0 cm DW), immediately fixed with 4% paraformaldehyde (PFA) in phosphate‐buffered saline (PBS; 0.1 m), stored for 6 h in this solution at room temperature, then rinsed in PBS. Samples were stored in PBS (0.1 m, 0.05% sodium azide) before being processed for microscopy. The propterygium was examined because its shape offers comparatively large regions of flat skeletal surface, where tesserae are relatively uniform in their cross‐sectional shapes.

Samples were decalcified with ethylenediaminetetraacetic acid (EDTA) for 1 week, dehydrated in graded isopropanol and xylene series, and embedded in paraffin using a routine histological infiltration processor (Miles Scientific, Naperville, IL, USA). Serial cross‐sections (7 μm) were made on a HM 355S microtome (Microm, Walldorf, Germany), and three sections per slide mounted on SuperFrost^®Plus slides. Sections were stained with haematoxylin and eosin (H&E; Shandon Varistain 24‐4, Histocom Vienna, Austria).

Transmission electron microscopy

Specimens were immediately fixed in 4% PFA buffered in PBS after collection, post‐fixed in 1% osmium tetroxide in distilled water for 24 h at 4 °C, then rinsed, and decalcified as described before, dehydrated in graded ethanol series and embedded in EPON resin. Ultrathin cross‐sections (90 nm) were cut on the same microtome (see above) with an ultra‐diamond knife, mounted on dioxan‐formvar‐coated slot‐grids (#G2500C, Christine Gröpl, Elektronenmikroskopie, Tulln, Austria), and stained with uranyl acetate and lead citrate (Leica Ultrostainer, Leica Microsystem, Wetzlar, Germany). The ultrathin sections were examined with a Philips CM 120 TEM at 80 kV (FEI, Eindhoven, the Netherlands) equipped with a MORADA digital camera (Olympus SIS, Münster, Germany) using Olympus TEM Imaging Platform software.

Results

Tessellated cartilage development

Tesserae exhibited shape and size changes during development that were visible in both μCT (Fig. 2) and backscatter SEM (Fig. 3). In μCT, tesserae could be seen to be irregularly shaped in histotrophic embryos and young sub‐adult animals (Fig. 2), but more regularly geometric in older sub‐adults and adults (Fig. 2). In histotrophic animals especially, it was often difficult to detect the boundaries of individual tesserae as they typically had low densities (X‐ray attenuation) at their margins (Fig. 2). Tesserae did not develop uniformly across the skeleton in terms of their size and degree of mineralization, but rather varied considerably across different regions of the same skeletal element. In μCTs of hyomandibulae of four developmental stages of U. halleri (Fig. 2), individual tesserae were first discernible at the ventral and chondocranial edges of hyomandibulae (Fig. 2A, asterisk and circle, respectively), whereas the rest of the skeletal element appeared largely unmineralized. This was verified by virtual/digital cross‐sections through hyomandibulae, which also showed that in older animals tesserae are considerably thicker in curved regions compared with planar regions (Fig. 2, third row of images).

Backscatter SEM supported the previously mentioned observation of the increase of tesseral size with age, but also showed age‐related decreases in the distance between tesserae and the perichondrium, in the distance between adjacent tesserae, and in the homogeneity of cell distribution and mineral density in tesserae (Fig. 3). In the youngest animals investigated (yolk sac embryos, ≤ 5.6 cm DW), there was no evidence of mineralization (data not shown). There was very little mineralized tissue in embryos at the early histotrophic stage: skeletal elements are mostly a core of unmineralized cartilage, with relatively uniform chondrocyte density, wrapped in an outer perichondrium (Fig. 3A,B). Tesserae at this ontogenetic stage are poorly formed, appearing as small patches of mineralized tissue (~ 75–100 μm wide and ~ 30–50 μm deep), ~ 10–30 μm below the perichondrium, and separated from one another by ~ 50 μm gaps (Fig. 3B). The mineralized tissue is globular in appearance, forming thin and often incomplete dividers between adjacent chondrocytes. At a higher magnification, the tissue appears to be formed from conglomerations of small spherical mineralized globules (~ 1–3 μm). These are visible in particular at the margins of tesserae (Fig. 3B).

The typically described adult tesseral morphology (i.e. geometric blocks of mineralized tissue; for instance, see Bargmann, 1939; Kemp & Westrin, 1979) first appeared in our investigated specimens at the young sub‐adult stage (Fig. 3C,D). With age, tesserae increased in width (~ 75–350 μm) and depth (~ 30–250 μm; Fig. 3E–H), as shown also in our μCT data (Fig. 2); note, however, that these measurements are from two‐dimensional backscatter SEM slices (i.e. they may not capture the maximum dimensions of a tessera) and that tesserae varied considerably in size, even within a single skeletal element (e.g. see arrowhead in Fig. 2D).

As the size and shape of tesserae continued to change with age, the morphology of their contact zones changed as well. In late histotrophic animals, tesserae were generally in close contact (i.e. in contrast to the case in younger animals, Fig. 3B): tesserae had irregular margins (in both vertical and planar sections) that were in contact with adjacent tesserae for their entire length (i.e. with no visible intertesseral gap in planar sections; Fig. 3D). In older sub‐adult animals, tesseral margins became more planar, resulting in smoother intertesseral contact zones (ICZ) of adjacent tesserae and an overall more regular, polygonal tiling pattern (Fig. 3F). This sub‐adult change in tesseral morphology was also coincident with the development of hypermineralized tesseral ‘spokes’ described below.

Tesserae in the early histotrophic stage showed uniform density of lacunae, the holes in tesserae that contain living cells (Fig. 3C,D). With increasing tesseral age, lacunar density became visibly patchier. The shape and distribution of lacunar spaces varied with location in the tessera; notably, lacunae were absent from high mineral density regions associated with ICZ (see below; Figs. 3G,H and 4A,B). Lacunae located towards the perichondral edge were considerably flatter, compressed in the perichondral–chondral direction (Figs 3G and 4B). Altogether, when observed in vertical sections, the flat, perichondrally associated lacunae formed an inverted pyramid shape in the upper portion of tesserae (the ‘cap’ zone of Kemp & Westrin, 1979), with all other cell lacunae being rounder (located in the lower, ‘body’ zone portion of tesserae; Fig. 4B). The lenticular lacunar morphology appeared approximately in the sub‐adult stage, at the time when the gap between tesserae and perichondrium closed, bringing tesserae in contact with the fibrous perichondrium (Fig. 3C,E,G).

Figure 4.

Heterogenic ultrastructure – mineral density variation in tesserae [backscatter scanning electron microscopy (SEM)]. (A, B) Planar (A) and vertical (B) sections of tesserae of Urobatis halleri reveal several distinct types of mineral density variation in tesserae suggesting a variety of mineralization processes. Spokes (sp) are the most prominent features of sub‐adult and adult tesserae. They are acellular, hyper‐mineralized, laminated wedges, radiating from intertesseral contact zones (icz), but not the uncalcified fibrous zones (ifz) of intertesseral joints (itj). In (A), interspoke regions (is) (flanking spokes) are coincident with fibrous zones, contain lacunar spaces (ls) and exhibit Liesegang lines (lil). The intertesseral junction (ij), the region between three (or more) tesserae, is comprised of unmineralized cartilage containing cells and few fiber bundles. In (B), the white dashed line indicates the two portions of tesserae, with the cap zone on top and the body zone below (perichondral and chondral side, respectively). (C) Liesegang lines are bands of different mineral density following the contours of tesseral margins, and were found mostly in the chondral portion of tesserae and edges bordering intertesseral fibrous zones (ifz) between abutting tesserae. (D) Filigreed pattern in the center of tesserae, could also be found between Liesegang lines. (E) Lacunar spaces (ls), housing cells, could be found throughout tesserae, except in regions with spokes (sp). Note the variation in shape and distribution of lacunar spaces in the tessera in (B). Lacunae located towards the perichondral edge were considerably flatter and together formed an inverted pyramid shape. Filled lacunar spaces (fls) were found in the middle of older tesserae [animals > 11 cm disc width (DW)] and showed a higher mineralization compared with the rest of the tessera body. (F) Magnification of a spoke showing the oscillating mineral density of spoke laminae. (G) Rough mineralization front on the perichondral side and (H) globular mineralization front on the chondral side of tesserae.

Tesseral spokes and other features of mineral density variation

A pronounced change in tesseral mineral density (visualized in backscatter SEM imaging) was observed during tesseral development. This began approximately in the late histotrophic stage when tesserae first came into contact with one another and planar contact surfaces began to form, resulting in tesserae appearing to have more linear margins and overall geometric shapes (Fig. 3D,F,H). At this point, at the margins of tesserae, exclusively in the regions bordering the planar intertesseral contact surfaces, laminae of higher mineral density tissue began to form (Fig. 4A). From this stage onward, as tesserae grew wider by accretion of new mineral, new highly mineralized laminae were added at their edges, at the ICZ. At any given intertesseral joint, the most recently deposited lamina (i.e. at the tesseral edge) was the same length as the ICZ. As a result, as tesserae and their associated contact zones grew larger with age, the swaths of high mineral density increased in length (via addition of new laminae distally, at the tesseral edge) and became more wedge shaped (by each successive added lamina being wider than the previous). The length of each lamina (i.e. the width of the spoke) varied from ~ 10 μm close to the center of tesserae to ~ 50 μm at the edges of adult tesserae.

The high mineral density regions appeared in planar view like spokes on a wheel (Figs 3H and 4A). In vertical views, the shape of these ‘spokes’ was more variable, and was dependant on the nature of the points of contact of adjacent tesserae: tesserae were typically not in contact over the entire perichondral–chondral distance of the joint, rather there could be multiple points of contact with spokes radiating from each (Figs 3G and 4B). From EDS data (not shown), we verified that these contrast differences (and all other features involving backscatter SEM contrast variation, see below), were due to local differences in the extent of mineralization, rather than variation in elemental composition (e.g. the introduction of high atomic number elements that could also account for backscatter SEM contrast variation). This is further supported by the observation from EDS data that oscillations in backscatter SEM contrast could be matched to local oscillation in calcium and phosphorus content.

Adult spokes were characterized by laminae (~ 100 nm–2 μm thick) of varying mineral density, arranged in series parallel to the plane of the ICZ (Figs 4A,B,F, 5E–H and 6A–D). Brighter (higher mineral density) laminae typically bordered darker (lower mineral density) laminae, resulting in spokes having an appearance of oscillating mineral density at higher magnifications (Fig. 4F). Evidence of spokes first appear in the histotrophic stage in the form of periodic, ‘wispy’ lines (~ 100 nm) of higher mineral density, arrayed parallel to ICZ (Fig. 5A,B).

Figure 5.

Ultrastructural reinforcement linked to growth mechanisms – development of spokes in Urobatis halleri. (A, B) Spokes (sp) first appeared in animals of ~ 7.5 cm disc width (DW), coincident with intertesseral contact zones (icz) and approximately when adjacent tesserae (T) first came into contact. Faint ‘wispy’ lines more or less parallel to the tesseral margin were often visible near the tesseral edges in the periods before spokes formed. (C–F) As animals age, tesserae grow and intertesseral contact zones (icz) widen, spokes lengthen and widen at their distal ends, their distinct laminar structure of oscillating mineral density becoming more obvious. Spokes converge toward the center of tesserae; in some tesserae, adjacent spokes merge proximally (e.g. G, H), suggesting early contact of tesserae at this region in the beginning of spokes development. Filled lacunar spaces (fls), lacunar spaces (ls).

Figure 6.

Spoke laminae and the underlying soft tissue – spoke ultrastructure in adult specimens of Urobatis halleri. (A, B) Spoke (sp) laminae in adjacent tesserae mirror each other's shape (arrowheads) and suggest the shape of younger tesseral margins [backscatter scanning electron microscopy (SEM) images]. (C, D) Though the resolution is lower, micro‐computed tomography (μCT) is capable of visualizing spokes and spoke laminae. (E, F) Histological staining [haematoxylin and eosin (H&E)] and (G, H) transmission electron microscopy (TEM) of decalcified, vertical thin sections, showing the organic tissue underlying spokes, which also exhibits a laminar pattern. H&E and TEM sections show bands of thick organic tissue (arrow heads) alternating with gaps/light areas of low organic content (arrow), presumably correlating with areas of lower and higher mineral density spoke laminae in backscatter SEM, respectively.

In the regions of tesserae lacking spokes, mineral density was generally lower, and we observed additional forms of mineral density variation. For example, Liesegang lines, concentric lines following the contours of adjacent tesseral structural features (e.g. tesseral margins, lacunar spaces), were often observed and were similar in morphology to those described by previous authors for other species (e.g. Weidenreich, 1930; Bargmann, 1939; Ørvig, 1951; Kemp & Westrin, 1979). In our samples, Liesegang lines were found particularly near the chondral edge of tesserae and could in some cases be very long, following nearly the entire contour of a tessera's chondral margin (Fig. 4B). Backscatter SEM revealed that Liesegang lines were bands of variable mineral density, appearing as light (higher mineral density) and dark (lower mineral density) bands following each other in succession (Fig. 4C). In more central areas of tesserae, the mineral density variation often appeared with an even finer degree of detail, forming a series of thin, filigreed patterns (Fig. 4D). Neither Liesegang lines nor the filigreed mineral density variation were visible in the perichondral portion of tesserae, which instead had a similar fibrous appearance to the overlying perichondrium but was mineralized (Figs 4B and 8F). ‘Normal’ lacunar spaces (housing living cells) were found in all tesseral regions, except those with spokes; however, lacunar spaces located in the middle of adult tesserae (i.e. equidistant between chondral and perichondral portions) were occasionally filled, to varying degrees, with mineralized tissue (Fig. 4B). Filled lacunae and spokes represented the features of highest mineral density in adult tesserae (Fig. 4E).

The pronounced laminar morphology seen in the spokes of adult tesserae first appeared in young sub‐adults, where spokes were short and consisted of few laminae (Fig. 5C,D). The length of spokes (i.e. the number of laminae) increased with age (Fig. 5C–H). Spoke laminae close to the contact zone mirrored the shape of the tesseral margin (Fig. 6A,B). The morphology of laminae in spokes tended to change with distance from the contact zone and in a similar way in opposing tesserae: the specific appearance (e.g. contour and mineral density) of a given lamina at a particular distance from the contact zone was similar to that of a lamina the same distance away in the opposing tessera (Fig. 6B, arrowheads). The organic matrix underlying spokes, visible in demineralized paraffin histology (6 μm) and ultrathin TEM sections (90 nm), also exhibits a laminar pattern, consisting of parallel bands of thicker fibrous tissue (~ 0.5–2.0 μm wide; Fig. 6E–H, arrowheads in H) separated by gaps filled with looser, less organized fibrous material (Fig. 6H, arrow).

Intertesseral joints

In early histotrophic animals, the gap between tesserae is large and filled with unmineralized tissue (containing cells and fiber bundles; Fig. 3A,B). Tesserae come into contact in the mid‐histotrophic stage, which results in the formation of what we define as the intertesseral joint, comprised of regions of intimate abutment of the two opposing edges of adjacent tesserae (ICZ) and small gaps containing cells and fiber bundles (intertesseral fibrous zones, IFZs; e.g. Figs 7B,C, 8D and 9). There is very little space between ICZ in U. halleri (< 2 μm); however, we observed no complex or large interdigitations at these interactions (Figs 9 and 10A). As tesserae increased in size with age, the dimensions (i.e. the width and depth) of intertesseral joint regions also increased.

Figure 7.

Flexible linkage of tesserae – collagen fibers at the intertesseral joints. (A–B) Backscatter scanning electron microscopy (SEM) of a planar section of tesserae (Urobatis halleri), with a focus on the intertesseral joint and the contact zone of two abutting tesserae. A comparison of backscatter SEM (B) and environmental SEM (C) images of the same joint area illustrates the complexity of intertesseral joints (itj), including intertesseral contact zones (icz) with spokes (sp) and unmineralized intertesseral fibrous zones (ifz) containing cells and fiber bundles (fb). Strings of cells were sometimes seen continuing between the ifz and adjacent tesserae (in lacunar spaces; ls); (D, E) Intertesseral collagen fibers are arrayed perpendicular to the plane of contact zones, as shown by polarized light microscopy (PLM) of adjacent tesserae in planar view visualizing the common orientation of joint collagen fibers bundles on opposite sides of tesserae (blue and yellow arrows), with (D) and without (E) the lambda filter. (F) TEM of a joint region of a vertical section, showing that fiber bundles pass uninterrupted from tesserae (T) into the unmineralized fibrous zone of the intertesseral joint.

Figure 8.

Micro‐computed tomography (μCT) imaging of Urobatis halleri tesserae. (A) Adjacent tesserae in planar view. (B) Rough perichondral surface of a tesserae and (C) smoother chondral surface of a tesserae. (D) Lateral view of the joint face of a tessera from the abutting tessera's perspective, showing the complex arrangement of ICZ surface (ics, brighter area) surrounded by fibre attachment surfaces (fas) where fibers tether the two tesserae together. (E) Planar section through the middle of a tessera [from the sectioning plane shown with a dashed line in (D)] showing acellular regions in the periphery corresponding to spokes and the inter‐tesseral contact surface (ics), and the rough fibre attachment surface (fas). (F) Vertical section through the center of a tessera showing a dashed line separating the cap zone (cz, perichondral portion, exhibiting an inverted cell pyramid with its base on the perichondral side) from the body zone (bz, chondral portion), containing cells in the center and acellular regions in the periphery, which correspond to spokes and ICZ.

Figure 9.

Intertesseral joint morphology of Urobatis halleri. Intertesseral joint morphology varies considerably with the sectioning plane through a tessera. The same tessera, sectioned in different orientations and planes, is shown in all images in the figure, highlighted in yellow; inset icons indicate section orientation and location. (A–D) Planar virtual sections through synchrotron radiation‐based micro‐computed tomography (SR‐μCT) scans of adjacent tesserae at different depths. (E, F) Morphological comparison of opposing intertesseral joints of a tessera. (G–J) Comparison of serial vertical sections through an intertesseral joint of abutting tesserae. Where tesserae are in contact, the contact zones (icz) are rather flat, showing no interdigitations. Abutting tesserae, however, are never in contact over their entire lateral edges; instead cells and intertesseral collagen fibers are maintained in fibrous zones (ifz) adjacent to contact zones. For clarity, in all images, surrounding material (e.g. perichondral and chondral tissue) has been replaced with gray background post‐scan.

Figure 10.

Tesserae shape variation across different shark and ray species. Virtual sections of synchrotron radiation‐based micro‐computed tomography (SR‐μCT) scans through the centers of adjacent tesserae in vertical (upper) and planar view (lower row). Despite a great variability in tesseral morphology, we observed commonalities in the morphology of the intertesseral joints, the arrangement and size of lacunar spaces, and the presence of spokes (sp). Lacunar spaces (black dots inside tesserae) are ~ 5–10 μm wide in all images. (A) Urobatis halleri, (B) Pteroplatytrygon violacea, (C) Dasyatis sabina, (D) Raja stellulata, (E) Raja eglanteria, (F) Myliobatis californica, (G) Rhinoptera bonasus, (H) Narcine bancroftii, (I) Torpedo californica, (J) Heterodontus franciscii, (K) Squatina californica, (L) Notorynchus cepedianus.

In the fibrous zones of intertesseral joints (i.e. where adjacent tesserae are not in direct contact), unmineralized fiber bundles span the distance between tesserae (Figs 6A,E,F and 7A–C). These fibers bundles were aligned in dense, parallel arrays, linking non‐spoked regions of adjacent tesserae. Strings of several cells in end‐to‐end series were typically visible between intertesseral fibers (Fig. 6E,F), visible continuing into adjacent tesserae as strings of lacunae oriented toward the tesseral center (Fig. 7B,C). Under polarized light, planar sections of tesserae show patches of coloration associated with intertesseral joints (indicating areas of aligned birefringent material). Intertesseral joints on opposite sides of the same tessera exhibited similar coloration (e.g. blue arrows in Fig. 7D, indicating similar alignment), and joints at acute angles to each other exhibited different coloration (e.g. blue vs. yellow arrows in Fig. 7D,E). The birefringence indicates that the alignment of fibrous material in intertesseral joints is perpendicular to contact zones of adjacent tesserae, with fibers extending into both tesserae that share a joint. TEM images of demineralized intertesseral joint regions show joint fiber bundles extending uninterrupted from the joint space into tesserae (Fig. 7F).

In two‐dimensional sections of joints (either vertical or planar), the proportions of IFZ and ICZ depend on the location of the cutting plane (Figs 7B,C and 9). This is due to the spatially complex, three‐dimensional interactions of fibrous and contact zone regions (Fig. 8). In vertical views of tesserae (i.e. from the perspective of a neighboring tessera), the two regions of the joint are distinguishable by their appearance: where tesserae are in contact (ICZ) the tesseral sides exhibit a smoother surface texture (intertesseral contact surface), whereas the surrounding IFZ exhibit rougher walls, pock‐marked with cellular lacunae (fiber attachment surface, Fig. 8D). The fibrous zones are typically slightly recessed and concave with respect to the contact surfaces (Fig. 9). The arrangement of the zones is mirrored in the adjoining tessera, such that the smooth contact surfaces of adjacent tesserae are touching and forming a bearing surface, whereas the opposing concave fibrous zones form a cavity filled with fibers spanning the joint space (Fig. 9E–J). Virtual sections of μCT‐scanned tesserae (Fig. 8E) confirm our backscatter SEM observations (e.g. Figs 4, 5, 6) that acellular and cellular areas are associated with intertesseral contact surfaces and fibrous zones, respectively.

Interspecies comparisons

Many common features were observed among the tesserae of adult U. halleri and those of other elasmobranch species investigated by SR‐μCT (Fig. 10) and backscatter SEM (Fig. 11). With the exception of the sleeper shark (S. pacificus; not shown), lower jaw cartilages from all examined species exhibited tesserae. Lacunar spaces were visible in great numbers in all species’ tesserae, except for those of the sevengill shark N. cepedianus, which contained either very few or no lacunar spaces (Fig. 10L). In vertical sections through the center of tesserae, the flat, perichondrally associated lacunae formed an inverted pyramid shape, similar to what we observed in U. halleri (Fig. 10A–E,H,K). As in U. halleri, aligned collagen fibers span the IFZ between tesserae, and spokes were present at points/zones of intertesseral contact in most of the examined species (Figs 10 and 11).

Figure 11.

Mineral density variation and tissue reinforcement – spokes in other elasmobranch species. (A, B) Scyliorhinus retifer, (C, D) Leucoraja naevus, (E, F) Carcharias taurus. Note that spokes are present in all species, and that all spokes lack cells and exhibit the same hyper‐mineralization and laminar structure as the spokes of Urobatis halleri.

All types of mineral density variation reported for U. halleri tesserae (i.e. Liesegang lines, spokes and filled lacunae) were also observed in other species’ tesserae using SR‐μCT (Fig. 10), most clearly though in backscatter SEM (Fig. 11). The most prominent difference among species was the variation in tesseral shape and size: tesserae ranged from stellate (Fig. 10B) to nearly circular (Fig. 10H) in planar view, and from rather flat discs (Fig. 10I) to rectangular blocks (Fig. 10D) in vertical view.

Discussion

General growth and ultrastructure concepts

Our data show that the characteristic tessellated pattern of the adult elasmobranch skeleton is absent before birth in U. halleri. In fact, in yolk sac embryos of U. halleri, tesserae could not be detected using μCT and backscatter SEM. At the early histotroph stage, the tessellated mineralization of the cartilage surface appeared patchy in μCT, being most developed along the ventral margin of skeletal elements towards the chondocranium (Fig. 2A). Further, the existing tesserae differed in the degree of mineralization, which suggests that the formation of tesserae is locally controlled (see below). In later developmental stages, all investigated skeletal elements were completely sheathed in mineralized tesserae, which varied in thickness but appeared to have a similar degree of mineralization (Fig. 2B–D). In adult specimens, the size and shape of tesserae varied between regions on the skeletal element, with generally thicker tesserae in strongly convex areas (e.g. the dorsal and ventral margins of hyomandibulae; Fig. 2C,D). Tesserae continued to increase in both depth and width with age, which agrees with previous observations of U. halleri (Dean et al. 2009) and one other elasmobranch species (Squalus acanthias: Benzer, 1944), and supports the idea that accretion (i.e. deposition of mineralized tissue at tesseral margins) is a central mechanism of mineralized tissue growth in this system.

We demonstrate that adult U. halleri tesserae exhibit highly heterogeneous calcification, forming features such as spokes, Liesegang lines and filled (hypermineralized) lacunae (Figs 4, 5, 6, 7, discussed each in turn below). We believe that Liesegang lines and spokes reflect successive, accretive mineralization events, following the contours of nearby features (e.g. lacunar spaces, tesseral edges), and in this way are structural records of the former shapes and locations of tesseral borders (mineralization fronts). In this growth model, the oldest portions of tesserae are closest to the center of each tile. Our ultrastructural data support the theory that new layers are built on top of older ones without remodelling: we never observed intersections of consecutive Liesegang lines or abrupt cessation of the pattern that might suggest local restructuring of tissue, supporting previous assertions that elasmobranchs lack the ability to repair skeletal damage (Clement, 1986 1986; Ashhurst, 2004; Huber et al. 2013). This is in contrast to, for example, osteonal bone where newer osteons intersect and interrupt older ones, providing a visual record of remodelling activity (Atkins et al. 2014). Calcium‐labeling of developing skeletons (e.g. with calcein) may help to determine how these patterns relate to mineral deposition.

Liesegang lines as records of accretive growth

Liesegang lines have been observed in the tesserae of both extant and extinct elasmobranch species (e.g. plates 6–7 in Ørvig, 1951; figs 2–7 in Applegate, 1967; fig. 16 in Kemp & Westrin, 1979; fig. 10 in Peignoux‐Deville et al. 1982; fig. 2 in Takagi et al. 1984; fig. 6 in Bordat, 1988; plate 5 in Clement, 1992; fig. 7F in Johanson et al. 2010). Our backscatter SEM data demonstrate that Liesegang lines are bands of varying mineral density and not simply homogeneously mineralized layers added on top of previous layers. This is supported by the few other published backscatter SEM images of tesserae, which depict similar features (from U. halleri: fig. 2A in Omelon et al. 2014; fig. 1F in Dean et al. 2015; from an unnamed species and sectioning plane: fig. 7F in Johanson et al. 2010). Kemp & Westrin (1979) suggested that Liesegang lines in tesserae might be caused by periodic secretion of enzymes or other cell products in the uncalcified extracellular matrix (ECM); however, it is possible that these structures are formed without such control. The phenomenon of periodic and concentric precipitation bands – variously called Liesegang ‘lines’, ‘rings’ or ‘bands’ – has been recognized in a variety of biological, chemical and geological systems for over a century (Liesegang, 1907; see Stern, 1967 for a bibliography of Liesegang rings). Although there is still no unified theory as to their formation, it is generally accepted that the Liesegang line pattern can be created by an oscillating chemical reaction between components diffusing in a medium (Liesegang, 1907; Thompson, 1942; Henisch, 2005; Kuz'min et al. 2013). The interaction results in periodic precipitation and depletion events that need not be biologically regulated. This implies that if components for mineralization are available at the mineralization front, then the concentric Liesegang bands characteristic of tesserae could form passively from cycles of nucleation and depletion. In this scenario, regulation would primarily be necessary in the processes that deliver materials for mineralization to mineralization fronts (e.g. potentially via vesicular transport; Kemp & Westrin, 1979; Clement, 1986 1986; Bordat, 1988; Takagi et al. 1984) and in those that prepare them for assembly (e.g. the interaction of the enzyme ALP with polyphosphates, liberating inorganic phosphate for skeletal mineralization; Omelon et al. 2014).

Our data suggest that Liesegang lines are associated with mineralization events where the concentrations of organic and inorganic components vary inversely. In decalcified sections of tesserae examined with TEM (Fig. 6H), Liesegang lines in the body zone continued to be visible as collagenous bands of varying electron density that inversely mirrored the local mineral density variation (Blumer et al. 2015). There is also likely concomitant variation in non‐collagenous matrix organic components (e.g. proteoglycans), as Liesegang lines exhibit varying degrees of basophilia (affinity to the hematoxylin stain in H&E histology; Blumer et al. 2015). This may represent cyclic variation in the concentrations of negatively charged glycosaminoglycans (and proteoglycans) in the mineralizing ECM. Elasmobranch proteoglycans have been shown to be effective inhibitors of hydroxyapatite crystal formation (Gelsleichter et al. 1995); Liesegang lines could reflect the degree to which proteoglycans in successive bands are dismantled to facilitate mineralization. In this way, the relationship between mineralization processes and the ultrastructure, mineral density and staining of Liesegang bands, stands to greatly inform our understanding of growth and mineralization regulation processes of tesserae.

Proposed phases of tesseral growth and skeletal mineralization

Liesegang lines and the filigreed pattern between them are characteristic of the chondral portion of adult tesserae, but are not found in the cap zone or the lateral portions of tesserae, neighboring ICZs (Figs 4B and 8F). We believe these differences in ultrastructure between the chondral portion and the rest of a tessera reflect the history of growth, how and when the tessera came into contact with surrounding tissues (e.g. perichondrium, adjacent tesserae) during ontogeny. From this, we propose several different phases in the developmental progression of tessellation.

The first phase we propose for U. halleri cartilage calcification (Fig. 12A–D) constitutes the early development of tesserae in histotrophic animals, when tesserae are still separated by uncalcified cartilage matrix (Fig. 3A,B). Tesserae of young animals of other species have also been reported as separate (non‐abutting) mineralized islands (Bordat, 1988; Maisey, 2013), suggesting this may be common young morphology among elasmobranch tessellations. Tesserae in U. halleri are apparently first formed by the accretion of numerous small globules of mineralized cartilage, filling the interstices between cells (Fig. 3A,B; Benzer, 1944; Ørvig, 1951; for other species see also Bordat, 1988); these isolated tesserae apparently continue to accrete mineral at their margins, resulting in the Liesegang lines described above. We saw no evidence in U. halleri to support Bordat's (1988) hypothesis that tesserae start as separate perichondral and chondral structures that grow together, suggesting that Bordat's observations for S. canicula could have been artifacts of sectioning young tesserae with deeply fluted edges. Based on our ultrastructure and ontogeny data and the general model of accretionary growth proposed above, we believe these ‘proto‐tesserae’ correspond to the central region (body zone) in adult tesserae. Unlike in older tesseral mats (> 10 cm animal DW), we saw no evidence for the presence of an organized collagenous network, comprised of parallel aligned fiber bundles, connecting the early centers of tesserae (or scaffolding the early growth). Therefore, based on PLM, the lack of birefringent collagen fiber bundles between tesserae in very young U. halleri skeletons suggests that other organizing mechanisms, apart from a collagen network, define the initial tessellation pattern. Eames et al. (2007) demonstrated discrete zones of activity of ALP in embryonic, pre‐tessellated swell sharks Cephaloscyllium ventriosum that appeared to predict the locations of future tesserae. ALP has been localized to mineralization fronts in adult U. halleri tesserae (Omelon et al. 2014), and is also known to be present in adult elasmobranch blood (Urist, 1961, 1962, 1967). It is reasonable to assume that ALP is active in early (phase I) tesserae, in U. halleri and other elasmobranch species; however, this has yet to be demonstrated. The source of ALP in elasmobranch skeletons and the mechanisms coordinating its localized, punctate expression in pre‐tessellated skeletons also remain to be investigated.

Figure 12.

Hypothesis of the development of tesserae and spoke laminae. (A–C) Schematic of the tesseral development when tesserae are separated from each other. (A) Minute mineralized globules develop between chondrocytes, associated with localized alkaline phosphatase (ALP) activity (Eames et al. 2007). We observed no evidence for specific orientation of collagen fibers at this stage. (B) ‘Proto‐tesserae’ (~ 100 μm wide) are agglomerations of mineralized globules. The obvious collagen fibers of the developing intertesseral joints appear to develop around this age. (C) Tesserae grow by accreting new mineral at their margins, forming Liesegang lines of varying mineral density (see Fig. 4). (D) The second growth phase of tesserae (D–H) begins when adjacent tesserae come into direct contact at the intertesseral joints, coincident with the appearance of spokes. (E–H) Abutting tesserae of sub‐adult and adult animals, illustrating the development of spoke laminae in a cycle of two alternating growth periods: a high‐extracellular matrix (ECM)/low mineral content growth period, forming dark bands (F–G), and a low‐ECM/high mineral content growth period, forming bright bands (H). We observed neither fibrous matrix nor cells in the contact zones between tesserae. (F) In the high‐ECM/low mineral content period, we propose that small amounts of distance are created in contact zones by the breaking down (perhaps by matrix metalloproteinases, MMPs) and rebuilding of collagen in the adjacent fibrous zones. ECM from the fibrous zones is directed to fill the space (F) and is then mineralized (G) In the low‐ECM/high mineral content period (H), tesseral edges bordering the contact zone accrete hyper‐mineralized tissue with little to no ECM.

The second proposed phase of tesserae growth (Fig. 12D–H) begins when tesserae grow into contact with the perichondrium and into direct contact with one another, coincident with the first signs of hyper‐mineralized tesseral spokes (see below). At this point, the cap zone begins to form. The presence of thicker, Type I collagen fibers from the perichondrium (Blumer et al. 2015) appears to alter the nature of the interaction between mineral and organic materials, as the ultrastructure of cap zone tissue is visibly different from that of body zone tissue (Figs 4G and 8B; Kemp & Westrin, 1979; Clement, 1992). Our backscatter SEM images and SR‐μCT reconstructions agree with Kemp & Westrin (1979), who showed that in macerated samples of adult tesserae, the rough perichondral tesseral surface was covered with aligned hydroxyapatite crystallites whereas the opposing chondral surface showed globular calcification (Figs 4H and 8C; Kemp & Westrin, 1979; Clement, 1992). The overall mineral densities of U. halleri cap and body zones, however, do not differ significantly according to our backscatter SEM imaging (e.g. Fig. 4B).

Spoke ultrastructure and development

Under backscatter SEM imaging, spokes are the most prominent feature of tesserae in sub‐adult and adult U. halleri. Spokes were never observed in the isolated (non‐abutting) tesserae of histotroph animals, and so we hypothesize that their formation is dictated by the interaction of adjacent tesserae at contact zones. Backscatter SEM imaging shows that spoke laminae from two abutting tesserae with similar distance to the ICZ mirror each other's morphology, even those laminae that are far from the joint (Fig. 6B). Therefore, we propose that spokes’ laminae, like Liesegang lines, depict stages of growth in tesserae, and because the appearance of spokes is coincident with the development of the intertesseral joint, spokes’ laminae reflect the shapes of former contact zones of abutting tesserae.

As far as we are aware, spokes have never been described as features of tesserae, although we believe they appear in the figures of multiple previous studies of a variety of elasmobranch species (e.g. Fig. 13 in Kemp & Westrin, 1979; Fig. 2 in Lee et al. 1984; Fig. 1 in Bordat, 1988; Fig. 3H in Ortiz‐Delgado et al. 2006; Fig. 6E in Johanson et al. 2010; and perhaps in the oldest tessellated shark fossil, Fig. 8B,C in Long et al. 2015). Our backscatter SEM and SR‐μCT data for multiple shark and batoid species further verify that spokes are common features of tesserae (Figs 10 and 11). The reason for spokes being overlooked by previous authors surely relates, to a large degree, to the dearth of published backscatter SEM images of tesserae, but is also likely a function of ultrastructural aspects of the spokes themselves and their response to sample preparation. Spokes are characterized by alternating laminae of high and low mineral content. TEM imaging of decalcified sections through abutting tesserae revealed that the organic matrix underlying spokes is similarly patterned, where the density of organic tissue is the inverse of the mineral density (Fig. 6H; Blumer et al. 2015). The high mineral density laminae have such low organic content, that when samples are decalcified and mineral is removed, these bands are nearly devoid of supporting tissue (Fig. 6E–H). As a result, spokes appear merely as a series of gaps in the tissue in decalcified sections, and therefore could easily be taken as artifacts (e.g. ‘knife stutter’ in sectioned tissue).

The consistent location of spokes, their lack of cells, and their distinct laminated structure (in particular the high mineral density ‘bright’ laminae), imply that their development may differ from that of the periodic Liesegang lines. We hypothesize a specific growth mechanism for spokes, concurrent with the lateral enlargement of tesserae in the second proposed phase of tesseral growth. The model is predicated on several observations from this study: (i) there is very little room for growth between tesserae at contact zones; (ii) the narrow interstitial spaces at contact zones appear to be devoid of cells and low in matrix content; (iii) spoke laminae from two abutting tesserae with similar distance to the ICZ mirror each other's morphologies; and (iv) spokes are comprised of alternating high and lower mineral density bands that appear to be patterned on very low and higher organic tissue frameworks, respectively.

Based on these observations, we hypothesize a two‐period growth cycle for spokes, beginning at the first contact of adjacent tesserae (i.e. the start of the second phase of tesseral growth): (i) a high‐ECM/low mineral content growth period; and (ii) a low‐ECM/high mineral content period (Fig. 12). At the start of the high‐ECM/low mineral content growth period, tesserae are in close contact and tethered together by intertesseral fibers (Fig. 12E); therefore, intertesseral growth must involve an elongation of intertesseral collagen fibers, linking two adjacent tesserae, to ‘make room’ for the deposition of new mineralized material (i.e. at the distal ends of spokes). However, as intertesseral collagen fibers extend uninterrupted into tesserae (i.e. with no exposed ends for new tissue addition; Fig. 7F; see also Blumer et al. 2015), fiber elongation must involve a severing of individual fibers (e.g. via matrix metalloproteinases or other enzymes) in order to introduce interstitial length (Fig. 12F). As this ‘slack’ is added into the system, new ECM, perhaps produced and directed by joint‐adjacent cells (see below), would fill the gap between the two tesserae and mineralize (Fig. 12G). This would result in a comparatively low mineral density and high organic content mineralized layer, a dark spoke lamina in backscatter SEM. The mechanisms underlying the second growth period (Fig. 12H), the subsequent deposition of a low‐ECM/high mineral content band (bright spoke lamina), are less obvious. Perhaps apatite continues to nucleate on the exposed, distal edge of dark spoke laminae after the available ECM has been mineralized. At this stage, these hypotheses of spoke growth are based purely on the laminated morphology of the tissue; high‐resolution in situ analyses of enzyme and cellular activity at contact zones are required for support of these hypotheses.

The apparent lack of cells in the interstitial space between tesseral edges at contact zones of U. halleri (Figs 7A–C, 9 and 10) is counterintuitive, as there is growing evidence that cells are involved in tesseral mineralization (Trivett et al. 2002; Egerbacher et al. 2006; Ortiz‐Delgado et al. 2006; Omelon et al. 2014). Also, the apparent lack of ECM in the ICZ goes against our observations that dark spoke bands are based on organic material (e.g. the collagen in the uncalcified cartilage matrix). We, therefore, posit that the fiber‐ and cell‐rich IFZ at the joints (Figs 8D and 9E–J) are involved in the regulation of spoke growth (e.g. via expression of enzymes and growth factors) and the determination of where and when mineralization occurs at joints. In this way, the cell‐rich fibrous zones adjacent to contact zones may also play a role in maintaining the patency of intertesseral joints (i.e. the non‐fusion of adjacent tesserae), which is critical to skeletal growth. Although we have never observed fusions of adjacent tesserae, previous authors have reported these in other species (Applegate, 1967; Maisey, 2013), suggesting that the inhibition of joint mineralization can either break down or be relaxed.

The role of cells, micropetrosis and interspecific variation in tesserae

The variation we report in lacunar shape among different tesseral regions (cap and body zone) is likely tied to their different tissue associations. Whereas the rounder body zone lacunae were present from the inception of tesserae, the flatter cap zone lacunae did not appear until the second phase of growth. The shape distinction between the cap zone and body zone lacunae and the inverted pyramid shape formed by the former was observed in a variety of species in this study (Fig. 10), implying that the growth phases proposed above could be shared among different taxa. Some species, however, showed strikingly different morphologies, exhibiting tesserae with nearly only flattened lacunae (i.e. similar to cap zone lacunae; Fig. 10D: R. stellulata) or rounded lacunae (i.e. body zone‐like; Fig. 10I: T. californica), and/or tesserae with comparatively few lacunae (Fig. 10J: T. californica) or almost no lacunae at all (Fig. 10L: N. cepedianus). Despite some apparent commonalities among tesserae (e.g. flat, non‐interdigitating contact zones, spokes and filled lacunae, see below), the observed shape variation in lacunae and whole tesserae (Figs 10 and 11) suggests that there may be species‐level variation in tesseral growth mechanisms. The tessellated cartilage system may, therefore, provide a naturally diverse palette for investigating cellular and ECM associations in vertebrate skeletal mineralization processes.

The cells within tesserae provide another indication that elasmobranch chondrocytes can control mineralization of their local environment. As tesserae thicken during ontogeny, chondrocytes from the underlying uncalcified cartilage are incorporated alive into tesserae via a process of encapsulation, whereby globular mineralized tissue engulfs cells into lacunar spaces in the mineralized matrix (Tretjakoff, 1926; Halstead, 1974; Kemp & Westrin, 1979; Takagi et al. 1984; Bordat, 1988; Dean et al. 2008, 2009, 2010). A surrounding layer of uncalcified cartilage is engulfed with each chondrocyte, and fills the space between the cell and lacunar walls (Kemp & Westrin, 1979; Bordat, 1988; Dean et al. 2010). Intratesseral cells likely remain alive due to the continuity of the uncalcified ECM in canalicular networks (i.e. inter‐lacunar passages), which would permit the transmission of nutrients, but also enable communication between cells (Dean et al. 2010). This arrangement is quite different from the calcified cartilage of other vertebrates (e.g. anywhere endochondral ossification occurs, such as at the end of long bones or the growth plate) where chondrocytes undergo apoptosis during mineralization, resulting in a largely acellular mineralized matrix (Kirsch et al. 2003). The intratesseral cell network is, therefore, in a manner, more similar to the lacunar–canalicular network connecting living, entombed cells (osteocytes) in bone. Osteocytes are believed to monitor their mechanical environment through this network, allowing them to respond to the changes in bulk tissue strain that would result from new or extreme loading regimes or bone fracture (Currey, 2002; Atkins et al. 2014). The intratesseral cellular network could, therefore, provide a similar basis for monitoring the skeleton's mechanical environment. However, the tendril‐like cell processes that link bone cells for cell–cell communication are apparently missing in intertesseral chondrocytes and chondrocytes in general (Dean et al. 2009, 2010). If elasmobranchs are indeed incapable of repairing their skeletons (Clement, 1986 1986; Ashhurst, 2004), then monitoring the health of deep intratesseral tissue may be irrelevant and intratesseral cells may instead survive only to provide some assistance (e.g. via synthesis of ECM and/or collagen synthesis) for cells ‘downstream’ in the lacunar–canalicular network at the mineralization fronts at the periphery of tesserae. Recent identification of lacunar spaces within the calcified cartilage of the chondricthyan fossil Gogoselachus, believed to possess an early transitional type of tesserae, suggests that further studies of lacunar morphology may provide insight also into the evolution of tesserae and the skeletal biology of extinct taxa (Long et al. 2015).

The intratesseral lacunar–canalicular network relies on the incorporation and maintenance of open, matrix‐filled spaces in tesserae, and likely requires the continued inhibition of mineralization local to incorporated chondrocytes (Dean et al. 2015). It is plausible that, as chondrocytes age, a breakdown of the inhibitory process that maintains the capsular zone leads to the progressive mineralization and decrease in size of lacunar spaces. This is supported by our observations and observations of other authors (Bordat, 1988; fig. 7F in Johanson et al. 2010) that mineral‐filled lacunar spaces were located primarily in the middle of tesserae (Fig. 4B,G), which we believe to contain the oldest entombed cells. The mineralization of lacunae could be triggered by cell death, distal blockages in the lacunar–canalicular network, and/or tesserae reaching a critical size that limits central cells’ access to nutrients. There are few available data on the biology of elasmobranch chondrocytes; however, assuming a no‐remodeling, accretive growth model for tesserae, it may be possible to use the occurrence of hypermineralized lacunae to estimate the lifespan of elasmobranch chondrocytes. We first noted hypermineralized lacunae in sub‐adult animals (11 cm DW), which, based on available age and growth data for U. halleri, could indicate that elasmobranch chondrocytes survive for approximately 1 year (Hale & Lowe, 2008). This would suggest a much shorter lifespan than the ~ 20 year half‐life proposed for mammalian chondrocytes (Stockwell, 1967; Bobacz et al. 2004). Backscatter SEM revealed that the mineralized, filled lacunar spaces in U. halleri are highly mineralized compared with the rest of the tesseral body, which we assume is in part due to the low organic content of the non‐mineralized matrix in lacunar spaces before they mineralize. We observed filled lacunae in backscatter SEM images of tesserae from several other genera as well (e.g. Amblyraja, Leucoraja, Negaprion, Raja and Scyliorhinus), indicating that this feature is not unique to U. halleri.

Tesseral lacunar hypermineralization is curiously similar to the phenomenon of ‘micropetrosis’ in bone, whereby osteocyte lacunae are filled with high mineral density material (Frost, 1960; Remaggi et al. 1996; Carpentier et al. 2012). This has been observed in humans more than other mammalian taxa (Frost, 1960) and seems to be linked to lifespan, with the incidence of micropetrotic lacunae increasing with age (Remaggi et al. 1996; Busse et al. 2010; Carpentier et al. 2012). Human micropetrosis results in an occlusion of the lacunar–canalicular network (Frost, 1960; Carpentier et al. 2012), but it is unknown whether the phenomenon is the result of an active cellular process (e.g. via matrix vesicle‐mediated mineralization) or a byproduct of cell death (e.g. via removal of mineralization inhibitors and/or apoptotic‐body facilitated mineralization; Kirsch et al. 2003; Busse et al. 2010). Human micropetrotic material is comprised of accretions of hypermineralized spherites (Carpentier et al. 2012) that appear similar in form to the globular mineralization we observed along tesseral chondral edges and sometimes lining the walls of intratesseral lacunae (data not shown), suggesting some mineralization processes may be shared between bone and tesseral lacunar hypermineralization.

Conclusions

In summary, our analysis of the ontogeny of U. halleri tesseral ultrastructure has shown that, although the tessellated pattern is established early in development as an array of mineralized islands, the characteristic tessellated morphology of elasmobranchs does not form until these nodes grow together to form abutting, geometric tiles. Although adjacent adult tesserae are in close contact, at the ultrastructural level they are not entirely flat‐edged: the intertesseral joint space is a complex arrangement of spatially discrete, planar contact zones with no interdigitations, interspersed with concave pockets of fibrous/cellular material (Fig. 13). The combination of mineralized bearing surfaces and fibrous attachments are likely the structural bases for intertesseral joints functioning effectively in both compression and tension loading (Liu et al. 2010, 2014; Fratzl et al. 2016).

Figure 13.

Schematic of two abutting adult tesserae in planar view, summarizing diagnostic ultrastructural features defined in this study. Compare with the previous, more simplistic notion of tesserae in Fig. 1. The two predominant portions of the intertesseral joints are displayed: (1) intertesseral fibrous zones, containing cells and fibrous tissue connecting adjacent tesserae; and (2) intertesseral contact zones, where the tesserae abut against one another. Spokes are coincident with contact zones and are acellular, whereas the rest of the tesserae body contains lacunar spaces, which house cells (not shown). bl, bright laminae in spokes; dl, dark laminae in spokes; fas, fiber attachment surface with fb, fiber bundles spanning the ifz, intertesseral fibrous zone at the itj, intertesseral joint; ics, intertesseral contact surfaces at the icz, intertesseral contact zone border the iss, interstitial space; is, inter‐spoke area; ja, joint axis; jc, joint‐adjacent cells; sp, spokes; tc, tesserae; ij, intertesseral junction.

All edges of tesserae appear to serve as surfaces for mineral deposition during ontogeny, leaving telltale marks in tesseral ultrastructure. Particularly under backscatter SEM imaging, these features may provide reliable determination of ontogenetic stage, but also clues – via aspects of cell shape and spoke presence and morphology – to the developmental mechanisms of less‐studied taxa, which exhibit different tesseral shapes and/or lacunar arrangements. The hypermineralized features we have described have implications for the mechanics of the tesseral mat (e.g. load‐channelling through joints and spokes to avoid cell damage), but also suggest some commonalities with bony skeletons (e.g. micropetrosis). This could indicate that some mineralization processes involved in the generation and maintenance of the tessellated morphology are deeply conserved, both among vertebrate taxa and across skeletal tissue types.

Authors contribution

R Seidel, MN Dean: data acquisition, data analyses, illustrations, writing of the manuscript. K Lyons, M Blumer, P Zaslansky, P Fratzl, JC Weaver: data acquisition, reviewing of the manuscript.