The complex rostral morphology and the endoskeleton ossification process of two adult samples of Xiphias gladius (Xiphiidae)

Abstract The authors studied the morphology of the upper and lower jaws, vertebrae and dorsal‐fin rays of the teleost fish Xiphias gladius to analyse the skeletal architecture and ossification pattern. The analogies and differences among these segments were investigated to identify a common morphogenetic denominator of the bone tissue osteogenesis and modeling. The large fat glands in the proximal upper jaw and their relationship to the underlying cartilage (absent in the lower jaw) suggested that there is a mechanism that explains rostral overgrowth in the Xiphiidae and Istiophoriidae families. Thus far, the compact structure of the distal rostrum has been interpreted as being the result of remodeling. Nonetheless, no evidence of cutting cones, scalloped outer border of osteons and sequence of bright–dark bands in polarized light was observed in this study, suggesting a primary osteon texture formed by compacting of collagen matrix and mineral deposition in the fat stroma lacunae of the bone, but without being oriented in layers of the collagen fibrils. A similar histology also characterizes the circular structures present in the other examined segments of the skeleton. The early phases of fibrillogenesis carried out by fibroblast‐like cells occurred farther from the already‐calcified bone surface inside the fat stroma lacunae. The fibrillar matrix was compacted and underwent mineral deposition near the previously calcified bone surface. This pattern of collagen matrix synthesis and calcification was different from that of mammalian osteoblasts, especially concerning the ability to build a lacuno‐canalicular system among cells. Necrosis or apoptosis of the latter and refilling of the empty lacunae by mineral deposits might explain the anosteocytic bone formation.


| INTRODUCTION
Among the very large number of living fish species only those of the two families Istiophoridae and Xiphiidae are known to develop an extraordinarily long upper jaw (the rostrum), which can be 20 times or more longer than the lower jaw. This particular feature has attracted the attention to rostral morphology, biomechanics and function in hunting and feeding (Atkins et al., 2014;Carey, 1982;De Metrio et al., 1997;Domenici et al., 2014;Fierstine, 1997;Fierstine & Voigt, 1996;Habegger et al., 2015Habegger et al., , 2019Mc Gowan, 1988;Schmidt et al., 2019;Videler et al., 2016). In particular, the rostrum of Xiphias gladius (Xiphiidae) and other species of the Istiophoridae family has been extensively studied and presented as clear examples of anosteocytic bone osteonal organization in teleost fishes (Atkins et al., 2014;Currey & Shahar, 2013;Habegger et al., 2015;Poplin et al., 1976;Shahar & Dean, 2013).
Since the first observations by Koelliker (1859), the presence of bone devoid of osteocytes in living teleosts has been reported using the terms "acellular" or "anosteocytic" bone (Moss, 1961a(Moss, , 1961b(Moss, , 1963Weiss & Watabe, 1979). Nonetheless, a more recent study in the zebra fish Danio rerio has shown that both anosteocytic and cellular bone can be observed in the skeleton of this fish and also within the same skeletal element, suggesting that a transition from the cellular to the acellular bone type occurred (Weigele & Franz-Odendaal, 2016). The process underlying anosteocytic bone osteogenesis was considered to be basically similar to that of cellular bone derived from a chondroid-type tissue (chondroid osteogenesis) or from the periosteum (periosteal osteogenesis), and the development of acellularity was considered to have developed from the withdrawal of osteoblasts that do not become entrapped as osteocytes (Ekanayake & Hall, 1987;Moss, 1961a;Weiss & Watabe, 1979) or from cells dying once entrapped because conditions for survival are not given and the lacunae are refilled by the calcified matrix (Moss, 1961b;Ofer et al., 2019). In addition to the mechanism underlying anosteocytic bone formation, a number of questions concerning structure, calcification and vascular supply of the anosteocytic bone; the distribution of the two types in teleost species; and the developmental phases in the individual fish still need to be answered. Phylogeny in teleost taxa has been used to tackle the issue of cellularity distribution (Lynne & Parenti, 1986), as also extensively reported in a recent review by Davesne et al. (2019). In that paper, the authors state that "the broad consensus on the two statements (i) that cellular bone is the plesiomorphic condition for teleosts, actinopterygians and osteoichthyans in general and (ii) that acellular bone is found in 'advanced' or 'higher' teleost groups are imprecise and potentially misleading." These authors also suggested taking a phylogenetic approach to distinguish the role of adaptation from that of the phylogenetic history in the distribution of bone types among species through large-scale molecular analyses that are now available (Betancur et al., 2017;Hughes et al., 1994;Near et al., 2012).
Furthermore, fundamental knowledge of bone resorption and remodelling is also needed to understand teleost bone development (Witten et al., 2000(Witten et al., , 2001. Witten and Huysseune (2009) emphasized this point in their review, in contrast to the previous view, that teleost bone is a metabolically inactive tissue, not subjected to remodelling, a view based on old publications reporting the absence of multinucleated osteoclasts and failure to respond to parathyroid hormone stimulation (Ingleton et al., 1995;Simmons, 1971). These authors further stated the importance of accurately considering both the differences in and similarities to mammalian morphology and metabolism. This task is certainly arduous due to the large number of species in both fish and mammalians but also to the specific differences present in species of the latter class of vertebrates. In taxa with cellular bone, the osteocytes are the dominant cellular components representing up to 95% of all the cells (Hall, 2005); they are derived from the osteoblasts that have remained embedded in the matrix produced by the same cells (Franz-Odendaal et al., 2006) and can survive in the calcified matrix environment because the osteoblast-osteocyte transformation and the calcification of the extracellular matrix have progressed simultaneously with the formation of the lacunocanalicular network (Pazzaglia et al., 2010, Pazzaglia & Congiu, 2013. In this paper, the skeletal histo-morphology of X. gladius Linnaeus, 1758, was examined, with focus on the basic processes of osteogenesis and remodeling. The arbitrary choice to study this fish as an experimental model was based on the following considerations: (a) X. gladius is the only species of the Xiphiidae family and is undisputedly a bony fish; (b) its high degree of skeletal specialization such as rostral development; (c) the availability of sufficiently large-sized, full-grown subjects capable of providing proper bone samples for processing and precise orientation of histological slides and specimens for scanning electron microscopy (SEM) observation; and (d) recent reports suggesting rostral remodeling (Atkins et al., 2014;Habegger et al., 2015Habegger et al., , 2019Shahar & Dean, 2013).
The following skeletal elements were examined: the upper and the lower jaws, the vertebrae and the dorsal-fin rays. The aims of this study were (a) to highlight the histology and the mechanism underlying growth in rostral length, (b) to compare the structural layout of the upper and lower jaws, (c) to document anosteocytic bone formation mechanism in the examined skeletal segments and (d) to search for evidence of osteonal secondary remodelling in the distal rostrum.

| Preparation and selection of anatomical specimens
The rostrum was cut transversally with a band saw into five segments of c. 40 cm in length, with short 5 cm long segments between the longer samples. The proximal segment included the proximal rostral cone and part of the frontal bones. X-rays of the long segments of rostrum were taken in antero-posterior (a-p) and lateral projections; the short segments were further sectioned in slices of 5 mm thickness to obtain X-ray images in the transverse projection. Computed tomography (CT) was performed separately on the five long rostral segments from the basal cone to the tip using a NewTom Cone Beam CT equipment (New Tom, Verona, Italy). The two branches of the lower jaw were first divided and separated from the tip; then X-rays were taken in the a-p and lateral projections. The first five proximal vertebrae were dissected, radiographed and examined with CT. The dorsal-fin rays were radiographed and then sectioned at the base.

| Macromorphology and histo-morphology
The bone specimens of the upper and lower jaws, vertebrae and fin rays were reduced to smaller specimens and sectioned for histology using a low-speed, diamond circular saw (Buheler Ltd., Lake Bluff, Illinois, USA). Subsequently, 2 mm thick sections were processed for SEM observation; 500 μm thick sections were further reduced to c. 300 μm by manual grinding; the sections were then polished, ultrasonicated and stained and undecalcified with toluidine blue or using the Von Kossa method before being examined in reflected and transmitted light Using a low-power stereomicroscope (Olympus SZX 7, Japan). Other specimens were decalcified in a solution of acetic and hydrochloric acid (2% CH 3 HCOOH/2% HCl) for 30 days, dehydrated in increasingly concentrated ethanol-water solutions and embedded in paraffin. Transverse and longitudinal 7 μm thick sections were cut with a sledge microtome and stained with haematoxylin-eosin, May-Grunwald-Giemsa, periodic acid-Schiff, Osmium and Alcian blue.

| Morphometry
Examination using circularly polarized microscopy was carried out to evaluate the number and density of the following osteon types: (a) full dark, (b) full bright and (c) alternate sequence of bright and extinct bands (Bromage et al., 2003). Transverse, decalcified thin sections (stained with haematoxylin-eosin) were cut at three levels of the rostrum at a distance of À10, À20 and À30 cm from the tip. Fifty randomly selected rectangular fields (434.52 Â 325.04 μm) were acquired for each level at 100Â magnification using a Colour View IIIb digital camera (Soft Imaging System GmbH, Munster, Germany) in circularly polarized light microscopy. Only osteons not intersected by the field borderline were counted and assigned to the osteon typology as reported earlier. Statistical analysis could not be applied to the distribution of osteon typology because all resulted in the "all dark" class.
The number and density of osteons among the levels were compared with MedCalc programme (MedCalc Software, Ostend, Belgium). Differences between groups were assessed using t-test; a probability of P < 0.05 was considered statistically significant.

| Scanning electron microscopy
Two different preparation techniques were applied to examine the laminar structures on the outer surface of the lower jaw (a) and the compact tissue of the distal rostrum (b): (a) Bone specimens were kept for 30 days in a 40% oxygen peroxide solution to remove the epidermis and the underlying soft tissues and then were repeatedly washed and sonicated in a bath of buffered saline solution (pH 7.4). They were then subjected to a slight, superficial decalcification in a 6% Na 3 PO 4 solution (pH 9.1) for 1 min at 15 C to display the collagen scaffold, and then they were routinely processed for SEM observation.  All the osteons examined at levels À10, À20 and À30 cm from the rostral tip were of the "all dark" type. The differences between the mean osteon number and field among these fields were not significant, whereas the mean osteon density was significant (P < 0.001) between levels À30/À20 and À20/À10 (Table 1).

| Lower jaw
The lower jaw was formed by two elongated branches converging and merging in an anterior pointed tip (Supporting Information Figure S7c). The transverse, undecalcified thick sections were stained by Von Kossa; the corresponding decalcified thin sections showed a bean-like shape with a concave inner and a convex outer surface: the inner surface was covered by jaw muscles, with the Meckel cartilage

| Dorsal-fin rays
The first and second rays of the dorsal fin were short and underdeveloped; the third was the longest one and with a single calcified axis. The other fin rays from the fourth onwards had a single basement which fanned out in four calcified filaments covered only by tegumental tissue (Supporting Information Figure S10a). The transverse section of the finray basal segment showed four distinct ossified nuclei separated by fat cellular stroma. Each of these was formed by a system of densely packed laminae springing from a compact base and oriented outwardly. Shortest and thick/pointed apophyses emerged from a compact base towards the mid-longitudinal line of the fin ray (Figure 5b).
Histology showed a texture of parallel, densely packed and pointed laminae. A row of aligned small, circular or occasionally long and narrow lacunae suggested a structural layout similar to that observed in the dorsal ossification columns of the rostrum, in the low jaw and in the vertebrae.   (Figure 6a). The latter was formed by a compact network of collagen fibrils as substantiated by the documented evidence of the period at higher magnification F I G U R E 3 Xiphias gladius. Lower-jaw left branch. (a) Toluidine blue in reflected light, transverse 500 μm thick section, bar = 0.5 mm. Bean-like shape of the bone showing the inner and outer bone wall: the inner with a compact texture, the outer consisting of parallel, calcified laminae connected by thin transverse septa. The Meckel cartilage (arrows) and muscles are evident on the inner surface; the bone grows at the upper and lower poles with the formation of new laminae with the same appearance of the latter; (b) details of the central outer part (sector B of A, 500 μm thick section, toluidine blue in reflected light, bar = 0.05 mm). The outer surface is covered by the subepidermal fat tissue ( §). The inner wall presents a more compact texture of the calcified matrix. The external laminae and the thin septa delimit lacunar spaces filled by a cellular-adipose stroma. (c) Decalcified section, 10 μm thick, haematoxylin-eosin, bar = 50 μm. Random distribution of "resting" and "active" zones in the cellular-adipose stroma of lacunae; the latter show hypertrophic capillaries and fibrillogenesis (asterisks). The bone of laminae and septa is anosteocytic; the lamina far down still shows empty cellular lacunae (arrows)

| DISCUSSION
The higher degree of skeletal specialization of X. gladius is represented, in particular, by the considerable development of length of the T A B L E 1 All the osteons at the three examined levels of the distal rostrum (À30, À20 and À10 mm from the tip) correspond to the "dark" type in circularly polarized light microscopy. mean number and density (S.D) among levels is significantly higher < 0.001) between levels À30/À20 and

Osteon type
Osteon mean density n mm -2 AEs:d: Level from tip (cm) À30 cm À20 cm À10 cm The rostral fat stroma of the swordfish was vascularized by a loose network of capillaries which was supported by large vessels running inside the two main parallel tunnels along the whole length of the rostrum together with myelinic nerves. Due to rostral stiffness, the bundles of myelinic fibres might not have served to innervate muscles but rather belonged exclusively to sensory nerves with only a sensory function and for which terminal receptors have not been described so far.
Despite the high degree of skeletal specialization of the swordfish rostrum, the shape of and structural differences in the lower jaw and also the differences among the other bones analysed in this teleost tissue. In all the papers dealing with the rostral structure and function, this compact, mixed tissue was interpreted as being secondary osteons resulting from active remodelling (Atkins et al., 1986(Atkins et al., , 2014Poplin et al., 1976;Shahar & Dean, 2013). Nonetheless, they better fit with the definition of "primary osteon" given in the textbook of Hall (2005)  ter turned out to be the more diffuse type in the higher orders of teleosts, whose histology and mode of formation were extensively studied by Moss (1961aMoss ( , 1961b. Moss also introduced the term "anosteocytic bone" and suggested that "cell withdrawal" theory explains the absence of osteocytes. Then, the coexistence of cellular and acellular bone was extensively discussed in later papers based on the phylogenesis (Lovejoy et al., 2004;Lynne & Parenti, 1986;Meunier, 1987Meunier, , 1989. Observations of anosteocytic and cellular bone in the same fish skeleton or in a single bone were reported by Weigele and Franz-Odendaal (2016). Hughes et al. (1994) in a study conducted in the teleosts Sparidae and Perciformes hypothesized that acellularity was determined by an apparent lack of osteocytic lacunae and that small osteocytes lie along the walls or in the lumen of tubules but without any clear evidence of these cells "metabolic activity." Several questions concerning phylogenesis and osteogenesis were recently reviewed by Davesne et al. (2019) and the remodelling of teleosts bone by Witten and Huysseune (2009).
The ossification progression documented in this X. gladius study occurred in the fat stromal tissue which surrounded the already- the "withdrawal theory" by Moss (1961bMoss ( , 1963 and rather suggesting more that the fate of the embedded cells is necrosis or apoptosis. Recently, Ofer et al. (2019) hypothesized that in D. rerio and Oryzias latipes anosteocytic osteogenesis is based upon apoptosis of the entrapped cells ending up as part of the mineralized bone matrix. Nonetheless, the phenomenon of cellular lacunae or bone canal refilling is not a specific feature of fish anosteocytic bone, as it has also been documented that changes in circulation dynamics suppressed blood flow in sectors of the haversian bone system and sealed osteons were formed in mammalian bones (Congiu & Pazzaglia, 2011;Pazzaglia et al., 2010). Nonetheless, according to the latter hypotheses, it cannot be excluded that some osteoblasts might withdraw, leaving traces of fissures on the bone surface.
The question of bone resorption and remodeling in fish was extensively discussed in the review of Witten and Huysseune (2009) with reference to the mammalian-like, multinucleated osteoclasts observed in Cyprinids and Salmonides (Domon et al., 2004;Witten et al., 2000) as well as with reference to the small, mononucleated osteoclasts observed in the early skeletal development of all teleosts and in later developmental stages of advanced teleosts with acellular bone. That review specified that these cells cannot be detected by using standard histological techniques and that a smooth (non-lacunar) type of bone resorption is carried out by the small mononucleated osteoclasts, in contrast to the lacunar type of the multinuclear cells. In the present study, two mature specimens of X. gladius were examined using histological techniques not suited for visualizing small mononuclear osteoclasts and, therefore, no conclusions on resorption pattern in the studied swordfish could be derived from the current results. Nonetheless, in such a large-size teleost bone, remodeling would mean substitution of a bone mass equivalent to the volume of all the rostral osteons as presently stated for the Xiphiidae and Istiophoridae families. Therefore, the mass of the removed tissue would have been considerable, and the actual remodeling should have left evidences of resorption pits detectable by SEM (Ali et al., 1984;Chambers, 1985), the lack of which supports the theory of unremodelled primary osteons, as put forth in this paper.
To the best of authors' knowledge, this is the first report of a primary plexiform type of anosteocytic bone ossification in a teleost. Laminar bone was also observed in some mammals such as rabbits, pigs, wild boars, cows, calves and sheep, characterized by a mixed plexiformhaversian texture and transition with ageing to the prevalence of the haversian model (Martiniakova et al., 2006;Mori et al., 2003). The relationship between the structural and histo-compositional characteristics of mammalian long bones (degree of laminarity, collagen fibre orientation and mineral content) and the adaptation to the specific features of the strain have been extensively investigated in mammals (Currey, 1959;Skedros et al., 2003). Due to their size and shape, billfish rostra are one of the best anosteocytic bone models for correlating structural patterns and strain features (Cohen et al., 2012). This study therefore provides a basic set of morphological data useful for future biomechanical analyses comparing laminar and haversian bone.

AUTHOR CONTRIBUTIONS
U.E.P., M.S. and R.M. developed the initial ideas and acquired the specimens and data; U.E.P. and Marcella Reguzzoni performed histological study and Marcella Reguzzoni and Mario Raspanti SEM study; Marcella Reguzzoni and G.Z. performed statistical analysis; U.E.P., Marcella Reguzzoni and R.M. contributed to preparation of the manuscript.

AKNOWLEDGEMENTS
The study was carried out using a SEM microscope of the University of Insubria and the Light Microscopy facilities of the University of Brescia, thanks to a scientific research agreement between the two universities. The senior author (U.E.P.) is retired professor of Orthopaedic Surgery of the University of Brescia.