THE ANATOMICAL RECORD 00:00–00 (2016) Vertebral Development and Ossification in the Siberian Sturgeon (Acipenser Baerii), with New Insights on Bone Histology and Ultrastructure of Vertebral Elements and Scutes  AMANDINE LEPREVOST, THIERRY AZA€IS,2 MICHAEL TRICHET, AND JEAN-YVES SIRE1* Sorbonne Universites, UPMC Univ Paris 06, CNRS, Institut de Biologie Paris-Seine, Department Evolution Paris Seine, Equipe ‘Evolution et Developpement du Squelette, Paris, France Sorbonne Universites, UPMC Univ Paris 06, CNRS, Colle`ge de France, Laboratoire de Chimie de la Matie`re Condensee de Paris (LCMCP), Paris, France Sorbonne Universites, UPMC Univ Paris 06, Institut de Biologie Paris-Seine, CNRS, Service de Microscopie Electronique, Paris, France ABSTRACT In order to improve our knowledge on the vertebral development, structure and mineralization in Acipenseriformes, we undertook a study in a growth series of reared Siberian sturgeons (Acipenser baerii) using in toto clear and stain specimens, histological and ultrastructural observations, Xray micro-tomography, and solid state NMR analyses Scutes were also studied to compare the tissue structure and mineralization of endoskeletal and dermal skeletal elements This study completes and clarifies previous investigations on vertebral development and architecture in sturgeons, and brings original data on the structure of (i) the perichondral bone that is progressively deposited around the vertebral elements during ontogeny, (ii) the typical cartilage composing these elements, and (iii) the scutes In addition we provide data on the mineralization process, on the nature of the bone mineral phase, and on the growth dynamics of the vertebral elements Anat C 2016 Wiley Periodicals, Inc Rec, 00:000–000, 2016 V Key words: Acipenseriformes; Siberian sturgeon; vertebral skeleton; development; mineralization; ossification; 3D microtomography INTRODUCTION While reviewing current knowledge on the vertebral column in Acipenseriformes prior to study vertebral deformities affecting reared sturgeons in France, it appeared that knowledge on the vertebral architecture, development and mineralization in Acipenseriformes was limited and incomplete (Leprevost and Sire, 2014) It was however for long known that (i) in sturgeons the endoskeleton is mostly cartilaginous and combined with the presence of an extensive dermal skeleton (five scute rows along the body), (ii) the vertebrae are composed of several elements C 2016 WILEY PERIODICALS, INC V organized around a persistent and unconstricted notochord, and (iii) mineralization only concerns a few of these elements and occurs late in ontogeny (e.g., Kolliker, 1860; *Correspondence to: Jean-Yves Sire; E-mail: jean-yves.sire@ upmc.fr Received 17 May 2016; Accepted 14 October 2016 DOI 10.1002/ar.23515 Published online 00 Month 2016 in Wiley Online Library (wileyonlinelibrary.com) 2  LEPREVOST ET AL Goette, 1878; Hasse, 1893; Klaatsch, 1893; Gadow and Abbott, 1895; Goodrich, 1909) In sturgeons, the axial skeleton is characterized by the lack of vertebral centra, which means that the notochord is the only support for the cartilaginous elements composing the vertebrae (Arratia et al., 2001; Hilton et al., 2011; Zhang et al., 2012) In Acipenseriformes, each vertebra is composed of four elements organized around the notochord, the basidorsal and interdorsal, and the basiventral and interventral, which form the dorsal and the ventral arcualia, respectively The basidorsals carry short, fork-shaped neural spines that enclose the longitudinal ligament running along the vertebral axis (Gadow and Abbott, 1895; Arratia et al., 2001) In addition to paired ribs, the abdominal vertebrae also support a median supraneural In Acipenseridae, the supraneural series is interrupted beneath the dorsal fin (Findeis, 1997; Hilton et al., 2011) Studying Acipenser brevirostrum Lesueur, 1818, Hilton et al (2011) divided the vertebral column into two regions, the abdominal region composed of vertebrae supporting ribs and the caudal region, in which ribs are absent Within the caudal region, these authors distinguished the preural region, located between the anterior base of the dorsal fin and the base of the caudal fin, and the ural region that includes the caudal fin and its support The border between the preural and the ural regions consists of the vertebra supporting the parhypural The boundary between the abdominal and caudal regions is not well defined, especially in young specimens, in which ribs have not developed yet In adult Acipenseriformes, all elements composing the vertebrae, with the exception of interdorsals and interventrals, are supposed to be able to mineralize Several authors named this mineralization “perichondral ossification” (Grande and Bemis, 1991; Arratia et al., 2001; Hilton et al., 2011) However, to our best knowledge, the presence of perichondral bone around cartilaginous elements or of calcified cartilage was not supported in the literature by accurate histological data A single study by Meunier and Herbin (2014) displays a transverse section of a “neurapophysis” of Acipenser sp (from a material dated 1876), showing the presence of primary bone More precisely, this section shows “a crown of periosteal bone tissue surrounding a large, circular surface resembling amorphous cartilage” This bony tissue contains osteocytes and presents Sharpey’s fibres and concentric growth ridges The few descriptions available on the axial skeleton often led to confusions between neural spines and supraneurals, and to controversy concerning their identification For example, while Arratia et al (2001) make a distinction between these two elements, Hilton et al (2011) not mention the presence of neural spines and include them in the term ‘basidorsals’, as neural spines form by dorsal growth of the basidorsals Also, Zhang et al (2012) wrongly termed the supraneural “neural spine” These few examples illustrate how limited are the data available on the axial skeleton mineralization in Acipenseriformes and support the need of new descriptions Therefore, we undertook a series of experiments using growth series of the Siberian sturgeon (Acipenser baerii Linnaeus, 1758), the most frequently reared sturgeon in France, chosen as a model species to fill these gaps We used (i) cleared and stained specimens and histological sections in a growth series to document the development of the vertebral axis in sturgeon, (ii) X-ray microtomography and mineralization rate measurements to localize and quantify the mineralization, (iii) histological and ultrastructural observations in adult specimens to describe the morphology of vertebral elements, and (iv) solid state nuclear magnetic resonance (NMR) spectroscopy to identify the nature of the mineral phase To complete these observations, the scutes were also studied in order to compare the mineralized tissue of these dermal skeleton elements to the endoskeletal elements composing the vertebrae MATERIALS AND METHODS Biological Material Sturgeons were randomly sampled in several fish farms of the company Sturgeon SCEA (Saint-Seurin-surl’Isle, Soumeras, Colombiers, Saint-Fort-sur-Gironde, France) between 2013 and 2015 Particular attention was paid to retain normally shaped specimens only, i.e., that did not show external deformities Five specimens aged 1, 2, 3, 5, and 24 years (57, 70, 80, 88, and 148 cm total length, TL, respectively) and six specimens aged years (108 cm TL in average) were used for X-ray micro-tomography The specimens aged 2, 3, 5, 7, and 24 years were also used for mineral content (MC or ash fraction) measurements, in addition to one specimen aged 20 years (144 cm TL) A growth series from to 43 days posthatching (dph) was sampled every days (two specimens for each sample, from 1.7 to 9.5 cm TL) for in toto clear and stain and for histological observations on paraffin sections Two specimens aged years (105 and 109 cm TL) were used for histological and ultrastructural observations of neural spines, supraneurals, and lateral scutes on Epon sections (1 to lm-thick and ultrathin sections) and the 105 cm individual was also used for solid state NMR spectroscopy analysis of the mineral phases The specimens aged years and more were collected after being sacrificed for caviar production in the laboratory of the company Sturgeon SCEA (Saint-Genis-de-Saintonge, France), stored at 218 8C and sent to the laboratory where they were dissected for X-ray microtomography and MC measurements Younger specimens were lethally anesthetized in MS222, then either immediately dissected and immersed in fixative (histological analyses), or immersed in PBS (solid state NMR spectroscopy) The experiments conformed to the directives of the European parliament and of the council of 22 September 2010 on the protection of animals used for scientific purposes (Directive 2010/63/EU) and the French Rural Code (Article R214-87 to R214-137, Decree no 2013-118 of 1st February 2013) As a food producer the company received the agreement n8FR 17.331.003 CE Dissections of 7-year-old specimens were performed according to the following protocol: once defrozen, the vertebral column of each specimen was roughly dissected, the head was cut between the first and second dorsal scutes, and the caudal fin was cut at its basis, anterior to the dorsal fulcrum (Fig 1) Vertebral columns were then immersed in a water bath at 65–70 8C for 7–10 Remaining soft tissues were removed carefully and VERTEBRAL OSSIFICATION IN SIBERIAN STURGEON Fig Drawing of a sturgeon in lateral view including a photograph of the vertebral axis, modified after vost et al (subm.) A and B indicate the level where the head and the caudal fin were cut, respecLepre tively The dashed line delimits the abdominal and caudal regions The arrow points to the end of the supraneural series The asterisks point to the regions sampled for X-ray micro-tomography; the numbers 1, 2, correspond to the regions sampled for mineralization rate measurements the notochord was separated from surrounding vertebral elements Vertebrae were counted from the border between the first and second dorsal scutes, at the point where the head was cut (Fig 1) stained with toluidine blue and coverslipped with R mounting medium (Sigma-Aldrich, France) EukittV Photographs were taken with an Olympus BX61 microscope equipped with a QImaging camera and Image Pro Plus software (Media Cybernetics, Bethesda, MD) In Toto Cleared and Stained Specimens X-Ray Microtomography and 3D Modelization Entire specimens were fixed immediately after death in a mixture of 4% paraformaldehyde (PFA) in PBS (1X) for 24 h at room temperature They were then rinsed in PBS and stored in ethanol 70% We used the cleared and double-stained method described by Hanken and Wassersug (1981) Briefly, after fixation, specimens were eviscerated and the eyes removed They were washed in distilled water several times for at least days, and cartilage stained with alcian blue Then the specimens were dehydrated in absolute ethanol, rehydrated through a graded series of ethanol, and macerated in a mixture of a saturated solution of sodium borate and a solution of 0.5% potassium hydroxide (used as a clearing agent instead of trypsin) This step can last from a few days to a few months according to specimen size Bone was then stained with alizarin red and specimens were finally transferred into glycerol via a graded series of glycerol in a 0.5% potassium hydroxide solution A few crystals of thymol were added to glycerol for storage Photographs were taken with a SZ X12 Olympus binocular magnifier equipped with a Spotflex camera and Spot software (Media Cybernetics, Bethesda, MD) Three segments including two vertebrae were collected (i) in the abdominal region around vertebra 15, (ii) in the caudal region around vertebra 30, and (iii) between the end of the dorsal fin and the caudal fin The samples were dehydrated in a graded series of ethanol and stored in butanol for at least days Butanol improves the contrast of cartilaginous elements and of the notochord on X-ray microtomography images The vertebrae were placed in plastic tubes and wedged with polystyrene pieces, and the atmosphere inside the tubes was saturated with butanol in order to avoid desiccation Ribs could not be kept as they were often lost during skeletal preparation X-ray microtomography images were obtained in the AST-RX platform of the Museum National d’Histoire Naturelle (Paris) with a microtomograph reference v|tome|x 240 L (GE Sensing and Inspection Technologies Phoenix X|ray), equipped with a RX microfocus 240 kV/320 W tube 3D modelizations were performed using ImageJ and Avizo softwares Cartilage diameter and bone thickness (mm) were measured on supraneurals from year onwards The virtual section orientation was re-adjusted with ImageJ in order to obtain frontal sections, and measurements were performed with Adobe Photoshop CS6 Histological Analyses on Paraffin Sections Until 40 dph the specimens were not dissected before fixation; in older specimens, the head and the caudal fin were removed These biological materials were fixed using the same protocol as described for the cleared and double stained method They were gently decalcified in acetic acid (10%) for one month, the solution being changed every week Samples were then dehydrated through a graded series of ethanol, shortly immersed in toluene and embedded in Paraplast (Sigma-Aldrich, France) The 10 mm-thick sections were obtained in the abdominal and caudal regions with a Leica RM2245 microtome and deposited on Superfrost slides (Fisher Scientific, France) The sections were dewaxed in toluene, rehydrated through a decreasing series of ethanol, Mineralization Rate Measurements The vertebral axis located between the head and the posteriormost vertebra supporting a supraneural was divided into three parts and three vertebrae were collected in the central region of each part (Fig 1) Neural spines and supraneurals located above the neural canal were cut and separated from the rest of the vertebra The samples were dehydrated in a graded series of ethanol (70, 95, and 100%, 24 h each), and the lipids were removed in acetone (48 h) and in trichloroethylene (24 h), according to Deschamps et al (2009) Then, the samples were dried for 48 h at room temperature under a laboratory hood, for 24 h at 37 8C and then weighed  LEPREVOST ET AL (Wdry) at the nearest mg They were incinerated for h at 800 8C in a muffle furnace (reference LE 14/11, Nabertherm), the ashes weighed (Wash) at the nearest mg, and mineral content (ash fraction) (MC, %) was calculated according to the formula: MC (Wash/Wdry) 3100 Control was made on muscle samples in order to evaluate MC in an unmineralized tissue The MC value of muscles being 1.80% in average, samples having MC  1.80% were considered unmineralized Histological and Ultrastructural Observations Samples of neural spines, supraneurals, and scutes were fixed in a mixture of PFA 1.5% and glutaraldehyde 1.5% in PBS 13, overnight, at room temperature Tissues were then decalcified for one month in the same fixative, to which EDTA 4% was added The samples were postfixed in 1% osmium tetroxide for h, rinsed in PBS, dehydrated through a graded series of ethanol then immersed in propylene oxide prior embedding in Epon 812 (EMS); mm thick, transverse sections were obtained using a Reichert OMU-3 ultramicrotome, deposited on a glass slide, stained with toluidine blue, mounted, and photographed These sections also enabled to select the region to be studied at the ultrastructural level using transmission electron microscopy (TEM) Ultrathin, 80 to 90 nm-thick sections were cut with a Leica Ultracut ultramicrotome and collected on 200 Mesh copper grids (Agar) Sections were stained with 2.5% aqueous uranyl acetate (Prolabo) and 1% lead citrate (Agar) TEM observations were carried out at 80 kV in a Zeiss 912 Omega, equipped with side-mounted 2k 2k Veleta (Olympus) CCD camera controlled with iTEM (Olympus) software Solid State NMR Spectroscopy Samples of neural spines and scutes were dissected (about mm long mm wide) after slaughter, then stored in PBS 13 Solid state NMR analyses were rapidly performed to avoid tissue degradation and dehydration 1H and 31P solid state NMR experiments were performed in an Avance 300 Bruker spectrometer operating at mL(1H) 300.13 MHz and mL(31P) 121.5 MHz Intact samples were placed in-between two Teflon spacers into a mm zirconia rotor and analyzed under magic angle spinning (MAS) at mMAS kHz 1H and 31 P chemical shift was referenced (d ppm) to TMS and 85 wt.% aqueous H3PO4, respectively Direct acquisition 31P MAS experiments were performed with recycle delay RD 30 s Two dimensional 1H–31P CP MAS experiments were recorded through the HetCor (Heteronuclear Correlation) sequence 1H–31P HetCor parameters were as follow: RD 3.5 s, CT ms, 400 transients for each 128 t1 increments were acquired High power proton decoupling was applied during acquisition (SPINAL-64, mRF(1H) 60 kHz) for each 31P experiments RESULTS Development of the Vertebral Elements During development, cartilage anlages initiate above and below the notochord They will give rise to the basidorsals and basiventrals, respectively (Fig 2A) Then, the two series extend in opposite directions: the basidorsals develop anteroposteriorly and the basiventrals develop postero-anteriorly (Fig 2A,B) At 25 dph each vertebra possesses paired basidorsals and paired basiventrals, while neural spines start to develop anteroposteriorly through dorsal extension of the basidorsals (Fig 2C) At 31 dph, the interdorsals, interventrals, and ribs have initiated and they develop in the same directions as basidorsals and basiventrals (Fig 2D) Supraneurals have also started to develop anteroposteriorly above the neural spines They form independently from the other vertebral elements (Fig 2D) then extend dorsally in an oblique direction (Fig 2E,F) Vertebrae display different morphologies along the vertebral column (Fig 2D–F) Abdominal vertebrae are tall and easily recognizable as they support ribs, neural spines, and supraneurals The supraneural series is interrupted beneath the dorsal fin In the caudal region, ribs are absent, and neural spines become shorter, so that the caudal vertebrae look rounder In toto alizarin red staining did not enable to visualize accurately the onset of mineralization of the vertebral elements, because from 10 months onward the muscles could not be cleared enough to see the skeleton It is the reason why we followed the mineralization using X-ray microtomography Localization of the Mineralization in the Vertebral Axis Mineralization of the vertebral axis in A baerii is limited to a few vertebral elements that become perichondrally mineralized at a relatively late ontogenetic stage Supraneurals are the first elements being mineralized at the age of year (Fig 3A), then neural spines start to mineralize around years (Fig 3B) The mineralized matrix completely surrounds the neural spines in the abdominal region of the specimen aged years (Fig 3C) Mineralization of neural spines and supraneurals follows the growth of these elements, then extends to basidorsals and parapophyses in older specimens (Fig 3D–F) The mineralization is regionalized along the vertebral axis and progresses anteroposteriorly during development In our growth series, mineralization of caudal vertebrae, both anterior and posterior to the dorsal fin, was only detected in the specimen aged 24 years (Fig 3F–H) The anteroposterior progression of the mineralization along the vertebral column was confirmed when measuring mineral content (MC) of neural spines and supraneurals (Fig 4) In all specimen studied, MC decreased anteroposteriorly and in a given region of the vertebral column, MC increased with age For example, in the sampling zone 1, between years and 24 MC increased from 14.2 to 23.4% in these two elements, while in sampling zone 3, it increased from to 17.5% In order to determine the growth process of these vertebral elements, we measured the cartilage diameter and bone thickness in different regions of the supraneural on virtual sections obtained with X-ray microtomography (Table 1) In our growth series, the diameter of the cartilage in the medial region of this element was almost constant during ontogeny (except in the oldest specimen), whereas it increased in distal and proximal regions Bone thickness was always higher in the medial region, and increased in all regions with ageing It VERTEBRAL OSSIFICATION IN SIBERIAN STURGEON Fig Development of the vertebral elements in a growth series of Acipenser baerii Left columns: In toto cleared and stained specimens; cartilage in blue, mineralized tissues in red Only the scutes are mineralized in this series (dorsal row first, then lateral and ventral rows) Right columns: Transverse, paraffin sections through the abdominal (A2, B2, C2, D3, E3, F3) and caudal regions (C3, D4, E4, F4) The levels of section are shown on A1 Specimens of same age as on the left column Toluidine blue staining A: dph, both 1.7 cm TL; B: 15 dph, both 2.0 cm TL; C: 25 dph, 2.6 and 2.8 cm TL; D: 31 dph, 5.6 and 5.0 cm TL; E: 37 dph, 7.4 and 6.5 cm TL; F: 43 dph, 8.1 and 9.5 cm TL bd: basidorsal, bv: basiventral, dsc: dorsal scute, id: interdorsal, iv: interventral, lsc: lateral scute, nc: neural canal, no: notochord, ns: neural spine, p: parapophysis, r: rib, sn: supraneural, vsc: ventral scute Asterisks: posteriormost vertebra supporting a supraneural Scale bars: A1, B1, C1, D1–D2, E1–E2, F1–F2 mm; A2–A3, B2–B3, C2–C3, D3–D4, E3–E4 100 lm; F3–F4 200 lm 6  LEPREVOST ET AL Fig A–H: 3D modelization (lateral view) of two vertebrae in a growth series of Acipenser baerii The mineralized vertebral elements (in red) were revealed using X-ray microtomography Anterior to the left Vertebra drawings are modified from Gurtovoy (1976) I: Growth dynamic scheme in a vertical section of a supraneural (red: bone; blue: cartilage) The numbers 1, 2, indicate the region where the measurements of table I were taken A: 1-year-old (y.o.), 57 cm total length (TL); B: y.o., 70 cm; C: y.o., 80 cm TL; D: y.o., 88 cm TL; E: y.o., 108 cm TL; F, G, H: 24 y.o., 148 cm TL A–F: abdominal vertebrae; G: caudal vertebrae of the region where the supraneural series is interrupted (around vertebra n830); H: caudal vertebrae between the posterior base of the dorsal fin and the caudal fin bd: basidorsal, bv: basiventral, l l: longitudinal ligament, n c: neural canal, no: notochord, n s: neural spine, p: parapophysis, r: rib, sn: supraneural Scale bars cm increases of about 22 mm per year in the proximal region, of 51 mm per year in the medial region, and of 45 mm per year in the distal region, between years and 24 These findings indicate that the supraneurals extend in length by means of cartilage deposition at both extremities, allowing also extension of the diameter of the cartilage matrix Once bone is deposited around the cartilage, the latter can no longer extend in diameter and supraneural growth in diameter is then ensured by bone apposition as shown in the medial region (Fig 3I) Histological Analyses of Vertebral Elements and Comparison with Scutes Fig Anteroposterior distribution of neural spine and supraneural mineral content (MC %) in a growth series of Acipenser baerii The numbers 1, 2, correspond to the sampled regions of the vertebral axis defined in Fig In order to compare the microstructure and organization of the endoskeleton (neural spines and supraneurals) with those of an element of the dermal skeleton, the same histological analyses were performed on bone tissues of vertebral elements and of scutes Histological observations of the vertebral elements using light microscopy and TEM revealed that the bone matrix directly lines the cartilage, which typologically corresponds to perichondral bone This matrix is entirely composed of randomly distributed bundles of collagen fibers, a feature that characterizes fibrous (woven- VERTEBRAL OSSIFICATION IN SIBERIAN STURGEON TABLE Measurements (mm) of the cartilage diameter and bone thickness in different regions of the supraneural in a growth series of Acipenser baerii Base (1) y.o y.o y.o y.o y.o 24 y.o Median (2) Top (3) Cartilage Ø Bone thickness Cartilage Ø Bone thickness Cartilage Ø Bone thickness 2.02 2.04 2.45 3.21 4.86 4.33 0.11 0.16 0.17 0.30 0.33 0.63 0.96 0.79 0.87 0.65 1.15 1.41 0.23 0.29 0.43 0.48 0.94 1.40 2.01 1.28 1.83 0.93 1.30 0.79 0.11 0.16 0.15 0.29 0.36 1.14 Location of measurements (1, 2, and 3) are shown in Fig fibered) bone (Fig 5A–D) A few osteocytes are located within the bone matrix, but no vascular canals were identified Taken altogether these findings allow to define the mineralized matrix of supraneurals as a perichondral, avascular, cellular woven-fibered bone Supraneurals and neural spines (not shown) display the same features The organic matrix of the scutes is also cellular and similarly organized as in the vertebral elements (Fig 5E–H) However, the bone matrix of the scutes is formed in absence of a cartilage anlage and is vascularized In both, vertebral elements and scutes, an irregular layer of scattered osteoblasts lines the bone surface, roughly delimiting the interface between the unmineralized bone matrix, osteoid, and the surrounding mesenchyme, which is largely composed of layers of oriented bundles of collagen fibrils (Fig 5B,F) The osteoblasts are elongated and the nucleus occupies a large part of their volume The cytoplasm does not house a large number of rough endoplasmic reticulum cisternae and of Golgi apparatus that are generally associated to an active protein synthesis In contrast, the osteoblasts exhibit a large number of cytoplasmic extensions facing the forming bone matrix and surrounding patches or collagen bundles (Fig 5B,F,G) These features indicate that these bone cells are rather modeling the pre-existing bundles of collagen fibrils of the mesenchyme than synthesizing new collagen matrix The modeled collagen bundles progressively organize into a woven-fibered bone matrix, a process which characterizes metaplastic ossification Some cells are entrapped into this forming bone matrix and become osteocytes (Fig 5C,D,H) The osteoid tissue resulting from this process is first unmineralized, then is mineralized As samples were decalcified, the mineral phase is not visible on these sections, but the region where the collagen matrix was mineralized is characterized by the electron dense, thin extrafibrilar matrix, probably deposited by the osteoblasts prior to mineral deposition (Fig 5C) The only structural difference between the vertebral elements and scutes is that the latter are anchored to the dermis by large collagen bundles (Fig 5E,F) In supraneurals and neural spines the interface between the perichondral bone and the cartilage matrix is devoid of cells and no remnants of pre-existing cells were present The boundary between the two tissues is rather sharp (Fig 5D) These features indicate that (i) the cartilage was deposited first by chondroblasts located in the perichondrium, (ii) these cells stopped depositing cartilage, (iv) either the chondroblasts disappeared from the cartilage surface or they differentiate into osteoblasts (perichondrium to periosteum transition), and (v) the latter start to form bone tissue at the cartilage surface using pre-existing bundles of collagen fibers of the surrounding mesenchyme In addition to the measurements presented in Table 1, these observations lead to our interpretation of dynamic growth schematically presented in Fig 3I In the two vertebral elements studied the cartilage matrix has an unexpected appearance: the matrix is dense compared to various cartilage types and the embedded chondrocytes are scarce and not hypertrophied, looking like rounded osteocytes (Fig 5A,D) In addition, the number of chondrocytes appears higher in young that in adult specimens (Fig 2) Nature of the Mineral Phase in Vertebral Elements and Scutes To complete our comparison between endoskeletal and dermal elements of the sturgeon skeleton, we analyzed the mineral phase from neural spines and scutes using 31 P solid state NMR The samples were processed as rapidly as possible after being dissected from the animal in order to preserve the native hydration of the samples Indeed, dehydration could lead to the precipitation of unwanted calcium phosphate phases from ionic precursors present in the biological medium 31P quantitative MAS (direct acquisition) spectra revealed that the mineral phase in these samples is similar to bone mineral from mature ewe bone (2-year-old) used as control, which is representative of mammalian bone hydroxyapatite (Fig 6A) Indeed, for neural spines and scutes, the 1D 31P MAS spectra exhibit one single broad resonance centered around 2.9 ppm that is similar in terms of position and line width to the 31P resonance of bone apatite from ewe bone The 2D {1H}31P HetCor spectrum of scute (Fig 6B) is similar to ewe bone and displays two different spectral regions that correspond to two different mineral domains as already demonstrated for bone mineral (Wang et al., 2013): (i) the apatitic core of the crystals where orthophosphate ions correlate with hydroxyl ions (d(1H) ppm), and (ii) the hydrated disordered surface, for which ions correlate with water molecules (d(1H) 4.8 HPO22 ppm) Despite our efforts, a partial dehydration of the sample occurred in the rotor and the 2D {1H}31P HetCor spectrum of scute reveals the spectral signature of prosurface ions (d(1H) 6–16 ppm) 31P tons from HPO22 resonances corresponding to phosphate ions from the  LEPREVOST ET AL VERTEBRAL OSSIFICATION IN SIBERIAN STURGEON Fig Solid state NMR analyses of mineralized elements of a specimen of Acipenser baerii aged years (108 cm TL) A: Direct acquisition 31P quantitative MAS spectra of neural spines and scutes compared to mature ewe bone (control) Asterisk: phosphate ions from the phosphate buffer solution, in which samples were stored B, C: 2D {1H} 31P HetCor spectra of scute and ewe bone highlighting the two different domains present in the apatite crystals: (i) the apatitic region and (ii) the hydrated, amorphous calcium phosphate surface two domains possess the same position (d(31P) 2.9 ppm), but differ from their line width: 31P resonance from the surface domain is twice larger than the one from the apatitic core The similarity of the spectral fingerprints between scute and ewe bone of similar hydration degree (Fig 6C) emphasizes that the mineral phase in scute is similar to bone mineral and possesses the same core/layer organization However, the proportion of the apatitic domain seems to be lower for scute Indeed, the correlation resonance corresponding to the apatitic phase is less intense for scute compared to ewe bone This result might be related to a lower degree of maturity/crystalinity of the apatite crystals in scute (Wang et al 2014) DISCUSSION Previous Descriptions are Confirmed, but Largely Completed and Clarified Our study largely completes and clarifies previous fragmented observations available on the vertebral column of sturgeons and paddlefishes (Gadow and Abbott, 1895; Grande and Bemis, 1991; Bemis et al., 1997; Findeis, 1997; Arratia et al., 2001; Hilton et al., 2011; Zhang et al., 2012) In the past, most data were obtained in various species and in a few individuals, which did not allow a clear understanding of the growth and mineralization of the axial skeleton By using growth series in a single species, the Siberian sturgeon Acipenser baerii, Fig Transverse, epon sections of a supraneural (A–D) and of a lateral scute (E–H) in Acipenser baerii (7 y.o., 105 cm TL) A, E: to lmthick sections, toluidine blue staining B–H: TEM micrographs of 80 to 90 lm-thick sections B: Interface between the mesenchyme and the osteoid tissue; C: Transition area between osteoid and maturing bone; D: Interface between bone and cartilage; F: Interface between the dermis and the scute; G: Detail of osteoblasts at the scute surface Cytoplasmic extensions encircling collagen fibers; H: Detail of the maturing bone tissue of a scute ab: anchoring bundle, b: bone, c: cartilage, ch: chordocyte, ep: epidermis, ld: loose dermis, m: mesenchyme, osb: osteoblast, osc: osteocyte, ost: osteoid, vc: vascular canal Scale bars: A, E 100 lm B, C, D, F, H 10 lm; G 5 lm 10  LEPREVOST ET AL our study brings new and accurate information to our knowledge As previously observed by Hilton et al (2011) in A brevirostrum, our observations in A baerii confirm that the elements of the vertebral axis develop in two opposite directions The basidorsals develop anteroposteriorly and the basiventrals develop postero-anteriorly, which is an original process among vertebrates The paired neural spines develop by dorsal extension of basidorsals as in all actinopterygians and enclose the longitudinal ligament, as noticed by Arratia et al (2001) The supraneurals develop independently from the other cartilaginous elements and articulate with the neural spines Neural spines and supraneurals are therefore distinct elements and their identification should not be controversial, even though some fusions between adjacent cartilaginous elements can occur, especially in the caudal region (Hilton et al., 2011) Our study brings new information concerning the mineralization of the vertebral axis in A baerii In Acipenseriformes, the mineralization of the vertebral column was known to be restricted to a few vertebral elements (Gadow and Abbott, 1895; Grande and Bemis, 1991; Bemis et al., 1997; Findeis, 1997; Arratia et al., 2001; Hilton et al., 2011) Here, for the first time we described the timing of mineralization in a sturgeon The supraneurals are the first elements to mineralize, then neural spines start to mineralize on their outer lateral sides, before mineralization extends to their inner sides besides the longitudinal ligament Basidorsals and parapophyses are the last elements to mineralize but late in ontogeny, as seen in a 24-year-old specimen Basiventral mineralization was not detected in our oldest specimen, but according to Hilton et al (2011), it can occur late in ontogeny in A brevirostrum These findings confirm that only a few elements mineralize in this Acipenseriformes, and also indicate that the mineralization process starts late then progresses slowly during ontogeny Mineralization of the vertebral column starts in the abdominal region and extends anteroposteriorly during ontogeny, a finding that is in accordance with the previously described timing of mineralization of each vertebral element This process results in an anteroposterior regionalization of mineralization in all specimens In addition to morphological differences that exist between abdominal and caudal vertebrae, this feature reinforces the differentiation between the abdominal and the caudal region of the vertebral column in Acipenseriformes Previous authors (Grande and Bemis, 1991; Arratia et al., 2001; Hilton et al., 2011) used the term “ossification” to describe the mineralization observed in the vertebral column of various sturgeons However, in the literature there was no histological demonstration of the nature of the mineralized tissue surrounding the vertebral elements, except one image of a “neurapophysis” of Acipenser sp dated 1876 (Meunier and Herbin, 2014) Here, we confirmed that the perichondral, mineralized tissue deposited around the supraneurals and neural spines consists of bone, and for the first time, we provided TEM ultrastructural data of this tissue, with comparison to the bone tissue of a scute of the dermal skeleton Vertebral Elements and Scutes Possess a Similar, but Unexpected Bone Tissue In all sturgeon species, the reduced mineralization of the endoskeleton, mostly in the form of “perichondral” bone, is combined with the presence of an extensive dermal skeleton consisting of five rows of scutes extending along the trunk (Bemis et al., 1997; Findeis, 1997) Various types of bone tissues often occur in a same species, particularly in osteichthyans, which are characterized by the possession of a wide range of skeletal tissues (Meunier and Huysseune, 1991) Our study in A baerii of the morphology and ultrastructure of the perichondral bone of vertebral elements and the dermal bone composing the scutes revealed that in both bone tissue is entirely woven-fibered bone According to Meunier and Huysseune (1991), all living, non-teleost fishes, including Acipenseriformes, possess cellular bone, a condition that was confirmed by our study Scutes have been the subject of a few studies, most of them describing their development and morphology, rather than their histological features (Sewertzoff, 1926; Jollie, 1980) Nevertheless, Meunier et al (1978) carried out a histological and micro-radiographic study of various “scales” in several non-teleost extant actinopterygians In A sturio they found that the scutes have a homogenous structure, composed of cellular bone with parallel fibers, not much vascularized On their deep surface, large resorption cavities were sometimes observed, on the surface of which secondary bone with woven fibers had deposited Periodic differences of mineralization were visible on X-ray images, and the ornamentations appeared more mineralized Here, in A baerii our observations revealed that the structure of the bone matrix was not parallel-fibered, but entirely woven-fibered (fibrous) The bone tissues of neural spines, supraneurals, and scutes have not only the same morphological aspect but also the same mineral phase, i.e., mainly composed of bone-like apatite Ultrastructurally, the formation of this bone tissue resembles that of a primordium of fibrous dermal bone, characterized by mesenchymal metaplasia First, osteoblasts appear to remodel pre-existing collagen fibers of the mesenchyme, then secrete non collagenous extracellular components that fill the interfibrillar spaces, then initiate mineralization Some osteoblasts can be trapped within the matrix and differentiate into osteocytes, while new osteoblasts can be recruited from undifferentiated pre-osteoblasts located in the surrounding mesenchyme During the formation of a typical dermal bone in an actinopterygian, the initial woven-fibered osseous metaplasia is followed by the deposition of newly synthesized collagen fibrils from osteoblasts organized around the elements This results in the formation of parallel-fibred bone then of lamellar bone later in ontogeny (Sire and Huysseune, 2003) In A baerii, the bone matrix either around the vertebral elements or in the scutes was woven-fibered type in its entire thickness, which strongly suggests that the same metaplastic process is conserved throughout ontogeny, a process that was never reported elsewhere to our knowledge In all vertebrates, bone is composed of three basic components: cells [osteocytes, osteoblasts, bone-lining cells (inactive osteoblasts at the bone surface) and osteoclasts], an organic matrix (predominant network of type VERTEBRAL OSSIFICATION IN SIBERIAN STURGEON I collagenous fibers and proteoglycans) and a mineral phase (apatite nano platelets) Moreover, bone is frequently vascularized, and submitted to resorption and reconstruction, the so-called bone modeling or remodeling processes (Meunier and Huysseune, 1991) However, in teleost fish, bone tissues can lack entrapped cells The histological features of bone tissues can vary according to the species studied and to the bones considered (Meunier and Huysseune, 1991) In A baerii the bone tissue surrounding the supraneurals and neural spines (i) was never found vascularized, a feature that could be related to the relatively thin bone matrix deposited at the cartilage surface, and it contrasts with the thicker bone matrix of the scutes that is crossed by some vascular canals; (ii) it displayed always the same woven-fibered, cellular structure whatever the region examined, an organization also observed in the scutes and that is uncommon in vertebrate bones; (iii) did not reveal resorption pictures related to osteoclast activity, although scutes are frequently subjected to osteoclastic resorption (pers obs.), and (iv) possesses a mineral phase closely related to bone apatite Two types of bones, belonging either to the endoskeleton or the dermal skeleton, are generally distinguished: endochondral bone (forming secondarily from cartilaginous precursors) and intramembranous bone (forming de novo in the mesenchyme) (Hall, 2005; Hirasawa and Kuratani, 2015) Bone types were classified by Francillon-Vieillot et al (1990) according to the organization of the collagen fibers, the vascular patterns and their ontogeny They differentiate cellular bone, acellular bone, aspidin (in jawless hetereostracans), avascular, vascular, woven-fibered, parallel-fibered, lamellar (plywood-like pattern), laminar, plexiform, reticular, radiating fibro-lamellar bone, hyalodentine, and finally, chondroid bone The latter, usually found beneath articular surfaces, has an appearance intermediate between cartilage and bone It exhibits large chondrocyte-like cells embedded in a mineralized woven-fibered bone-like matrix and forms de novo from osteogenic precursors, not by metaplasia (Huysseune and Verraes, 1986; Huysseune and Sire, 1990; Meunier and Huysseune, 1991; Benjamin et al., 1992) The type of bone described here in the mineralized elements of A baerii corresponds exclusively to the woven-fibered (fibrous) type and its collagenous matrix seems to result from metaplasia of the mesenchyme, like intramembranous bone Although such type of bone was expected in scutes, in the vertebral elements, its presence directly at the cartilage surface is quite surprising and, to our knowledge, does not correspond to any existing description of bone tissue A Cartilage Showing Also Unexpected Aspects The cartilage forming the core of the supraneurals and neural spines in A baerii displayed unexpected features, at least in adults In the youngest specimens studied, the cartilage matrix looked like a classic hyaline cartilage, with numerous hypertrophic chondrocytes embedded in a loose collagen matrix However, in the same locus but years later, in adult specimens, the chondrocytes are small and scarce, and the collagen matrix, although exhibiting the classical metachromatic staining, is particularly dense Such features suggest that important changes occurred in chondroblast 11 functioning prior the deposition of the bone matrix at the cartilage surface Studying many teleost species, Benjamin (1990) then Benjamin et al (1992) described the various cartilage tissues found in the head, trunk and fins They classified them into different types: hyaline-cell cartilage and its subtypes (fibro-, elastic- and lipo-hyaline), Zellknorpel, fibro/cell-rich cartilage, elastic/cell-rich cartilage, cellrich hyaline cartilage (>50% of the cartilage volume occupied by cells) and matrix-rich hyaline cartilage (