Starting from 28 days post-fertilisation (dpf), WT were fed a regular P diet for two months. At 28 dpf, Chi/+ mutants were randomly divided in two groups and fed a ‘LP diet’ (low P content, Chi/+_LP) or a regular P diet (Chi/+) for two months. Three months old Chi/+ show severe deformities compared to WT fish on Faxitron X-ray images (Figure 1A), confirming earlier reports (15–17). Vertebral column deformities of WT, Chi/+ and Chi/+_LP were assessed by whole mount Alizarin red S staining (details are described below). Compared to WT, Chi/+ mutants have an increased frequency of kyphosis and lordosis in the abdominal region of the vertebral column. Kyphosis and lordosis are reduced in the abdominal region of the vertebral column in Chi/+_LP fish. However, compared to Chi/+ fish, Chi/+_LP fish show increased scoliosis in the caudal vertebral column. Both Chi/+ and Chi/+_LP have a high frequency of compressed vertebral bodies; examples are shown in Figure 1A, frequency of deformities is shown in Figure 1B. To assess the severity of the curvature in those specimens diagnosed with kyphosis, lordosis and scoliosis, the maximal kyphotic, lordotic and scoliotic indices were measured, respectively. Figure 1C shows the maximal curvature indices in the abdominal and caudal region of affected WT, Chi/+ and Chi/+_LP individuals. No statistical differences in the degree of curvature are detected between Chi/+ and Chi/+_LP fish at three months of age (Mann-Whitney test: non-significant).

The general metrics for the analysis of the vertebral column malformations (Table 1) show that the frequency of specimens with at least one malformation is 19% in WT, 75% in Chi/+ and 89% in Chi/+_LP. The highest average malformation load is in the Chi/+ group (3.27 malformations/deformed specimen) where a total of 49 malformations were identified. The average malformation load in the Chi/+_LP group is 2.5 with a total of 40 diagnosed malformations. These data indicate a tendency towards reduced number of malformations and average malformation load in the Chi/+_LP compared to untreated Chi/+ (Chi-square test: p=0.27).

To quantify shape variation of the caudal vertebral centra in WT, Chi/+ and Chi/+_LP animals, landmark-based geometric morphometrics was applied, based on whole mount-stained specimens (Figure 2A). The scatterplot of 2D landmarks shows differences in location of corresponding landmarks between WT, Chi/+ and Chi/+_LP animals. The scatterplot shows a high variation in the superimposition of Procrustes coordinates of Chi/+ compared to WT. Chi/+_LP display reduced variation of landmarks compared to Chi/+ and a distribution more similar to WT (Figure 2B). Principal component analysis (PCA) of superimposed landmarks shows the amount of variation between WT, Chi/+ and Chi/+_LP. Only principal component 1 (PC1) and principal component 2 (PC2) were considered. Compared to WT, Chi/+ show a statistically significant higher variance (Chi/+ versus WT, PC1 0.3321, PC2 0.1795; Chi-square test: p<0.001). On the contrary, Chi/+_LP animals show a non-statistically different variance compared to WT (Chi/+_LP versus WT, PC1 0.2484, PC2 0.2020; Chi-square test: non-significant) (Figure 2C).

Histological assessment of the vertebral column in the sagittal plane confirmed the irregular shape of mutant vertebral bodies compared to WT vertebral bodies (Figure 3). Different from WT (Figures 3A, B), in Chi/+ animals vertebral body endplates of adjacent vertebral bodies are frequently shifted against each other along the dorsal-ventral axis (Figure 3B). Moreover, scoliosis of the vertebral column in Chi/+ and Chi/+_LP animals, previously observed on whole mount-stained specimens is visible on histological sections as the absence of a sagittal middle plane (Figure 3B). Histology reveals the presence of compression fractures in Chi/+ and Chi/+_LP animals (see below for details) (Figure 3B). Vertebral centra of Chi/+ and Chi/+_LP animals display distortions of the endplates not observed in WT zebrafish (Figure 3C). Despite these distortions, the endplates retain structurally unaltered intervertebral spaces with unaltered ligaments as in WT animals (Figure 3C).

Histological sections of non-demineralised vertebrae in the sagittal plane stained with Von Kossa/Van Gieson show the absence of a detectable osteoid layer (non-mineralised new bone matrix) in the growth zone of the vertebral endplates in Chi/+ animals (Figure 4A). In Chi/+_LP animals the non-mineralised osteoid layer is restored (Figure 4A).

Vertebral body endplate mineralisation was assessed quantitatively based on the extent of the mineralised and non-mineralised bone matrix in whole mount specimens stained with Alizarin red S. In comparison to WT, Chi/+ mutants show higher mineralisation of the vertebral body endplates (Chi-square test: p<0.001), but the LP diet is capable of partially reducing endplate mineralisation in some Chi/+_LP individuals (Chi-square test: Chi/+_LP versus Chi/+, p<0.001; Chi/+_LP versus WT, non-significant) (Figure 4B).

Finally, midline sections of the vertebral column were used to measure the thickness of bone structures in WT, Chi/+ and Chi/+_LP. Histomorphometry reveals that Chi/+ and Chi/+_LP bone structures are significantly thinner compared to WT. This is the case for vertebral body endplates, the central region of the vertebral body and the trabecular bone (Figure 4C, Table 2).

The analysis of whole mount-stained specimens reveals that Chi/+ and Chi/+_LP animals have vertebral compression fractures that relate to the anteroposterior compression of vertebral centra (Figure 5). Compression fractures are absent in WT animals (Figure 5A). Different severity levels of compression fractures are observed in both Chi/+ and Chi/+_LP individuals (Figures 5B–D). Some Chi/+_LP animals present compression fractures that only affect one vertebral body (Figure 5B). Some Chi/+_LP mutants show kyphosis associated with multiple compression fractures (Figure 5C). Chi/+ mutant fish display severely distorted vertebrae, as described above, and collapsed vertebral centra (Figure 5D). Regardless of the inter-individual variability among the fractures, the bone of fracture repair calli appears more dense than other bone elements when visualised with fluorescent light.

Histological sections in the sagittal plane of mutant vertebral columns confirm the presence of compression fractures in Chi/+ and Chi/+_LP animals (Figure 6). A variety of fractured sites suggests that compression factures range from being in a status of repair (as evidenced by bone remodelling, see below) to fractures that do not, or not yet, display evidence of repair. Compression fractures affect the central region of the vertebral bodies. The fractured bone trabeculae become displaced into the chordocyte-filled lumen of the notochord (Figures 6A, B). When fractures injure the notochord, the notochord strand and the notochord septum (see Figure 3A for the anatomical terms) become condensed [keratinised in response to tissue damage according to (26)] (Figures 6A, B). Analysis of serial sections from several Chi/+ and Chi/+_LP animals reveals that compression fractures display variable degrees of repair. Fractured sites show fibrocartilaginous calli at the outer surface of the compressed vertebral bodies (Figure 6C). Repair evidently continues with replacement of fibrocartilage tissue by a bone callus (Figure 6D). Repair at fracture locations is further confirmed by the demonstration of tartrate-resistant acid phosphatase (TRAP). TRAP is produced by osteoclasts and secreted onto the bone surface at the locations of bone resorption. TRAP activity is also linked to the resorption of fibrocartilage that is being replaced by a bone callus (Figure 6E) (27). Osteoclast activity is also observed at the bone trabeculae and vertebral endplates of vertebral compression fractures (Figure 6F). No signs of fracture repair or bone resorption are detected on bone elements inside the notochord (Figure 6E). This agrees with the fact that the notochord contains neither blood vessels, nor nerve fibres and no lymphatic vessels (28).

Osteoclasts and locations of bone resorption (other than locations of fracture repair) in growing juvenile individuals were identified by the demonstration of TRAP. In WT zebrafish TRAP activity is typically restricted to locations within the vertebral column that require resorption for allometric growth (27), i.e. the endosteal surfaces of enlarging neural and haemal arches (Figures 6G, H). Chi/+ and Chi/+_LP also display TRAP staining at the endosteal surfaces of the neural and haemal arches, but the enzymatic activity is increased and extends also to other bone structures such as the bone trabeculae (Figures 6I, J).

Transmission electron microscopy (TEM) confirms the presence of osteoclasts, attached to the endosteal surfaces of arches in WT, Chi/+ and Chi/+_LP (Figure 7A). These cells, while attached to the bone surface, reside in resorption lacunae (as shown in Figure 7A, WT zebrafish). Actively resorbing osteoclasts exhibit a typical ‘ruffled border’, an electron-dense cytoplasm with abundant cytoplasmic vacuoles and resorption vesicles in proximity of the ruffled border (as shown in Figure 7A, Chi/+ zebrafish).

The ultrastructure of bone cells and bone matrix in the vertebral column was analysed by TEM at the level of the medio-sagittal plane. Ultrathin sections of representative specimens show that Chi/+ and Chi/+_LP have enlarged endoplasmic reticulum (ER) cisternae in osteoblasts located in the growth zone of the vertebral body endplates (Figure 7B, see Figure 3A for location). Likewise, ER cisternae in osteoblasts along the neural and haemal arches are enlarged. High magnification images show that Chi/+ ER cisternae are filled with protein (Figure 7C). Conversely, WT osteoblasts have numerous, yet not enlarged ER cisternae, indicative for high protein synthesis activity at the growth zone of the endplates (Figure 7B).

The collagen matrix appears altered in the Chi/+ mutants compared to WT. In WT, newly secreted collagen fibrils in the proximity of the osteoblasts (osteoid) are dispersed prior to maturation and assemble into larger fibres at a distance from the osteoblasts. In contrast, the Chi/+ bone matrix is characterised by densely packed collagen fibrils in the vicinity of the osteoblasts (Figure 7C). In Chi/+_LP, individual collagen fibrils are partly distinguishable in close proximity of the osteoblasts, indicative for less collagen packing and a more typical osteoid (Figure 7C).

Longitudinal sections of collagen fibres in WT zebrafish show a regular D-periodicity. The pattern is absent in Chi/+ mutants. In Chi/+_LP the collagen D-periodicity is partially visible along the fibre (Figure 7D).

Early reports on children and young patients diagnosed with OI described severe vertebral column deformities as the major complications of the disease. From very young age onwards, patients suffer from vertebral body deformities associated with progressive scoliosis, kyphosis or kyphoscoliosis (2, 4, 29). This condition is known to be age-dependent and usually worsens after the age of six (4, 30). Likewise, Chi/+ zebrafish display severe bending of the vertebral column, i.e. kyphosis and lordosis, at three months of age, as demonstrated by whole mount Alizarin red S staining and histological sections. Similar to human patients, such malformations progressively worsen in adult stages (16, 17). Vertebral body deformities, vertebral compressions and fractures represent other severe complications of young patients (4, 29). We show here for the first time that juvenile Chi/+ zebrafish suffer from vertebral body compressions in both the abdominal and caudal region of the vertebral column. Mutant vertebral bodies are distorted and characterised by increased shape variation compared to WT, as revealed by geometric morphometrics. On top of that, Chi/+ fish have highly mineralised vertebral body bone structures with no detectable osteoid (non-mineralised collagenous bone matrix), as confirmed by histology of non-demineralised specimens. Likewise, the osteoid thickness is reduced in the bone of human patients (31, 32) and murine OI models (33, 34). Moreover, hypermineralisation and bone brittleness are well documented in human patients (2, 4, 5, 29), OI mouse models (35) and OI zebrafish models (17, 36). Excess bone mineralisation related to increased dietary P intake or ageing is known to increase vertebral column malformations in WT zebrafish (22, 26). Hence, hypermineralisation alone can be considered a risk factor for bone deformities. In addition, OI bone is characterised by mutated collagen type I, which contributes to bone fragility. While mutations cannot be undone in OI patients or animal models, lowering the bone mineral content could possibly alleviate their skeletal defects. Recent findings on teleosts including WT zebrafish (22) and Atlantic salmon (23–25) have shown that low dietary P administration reduces the bone mineral content and promotes the formation of non-mineralised bone without causing vertebral column malformations. To assess the effects of reduced dietary P intake on the mutant bone phenotype, Chi/+ were fed a low P diet (LP) from one month of age. Chi/+ zebrafish under LP conditions (Chi/+_LP) have reduced incidence of kyphosis and lordosis of the abdominal region of the vertebral column, and reduced shape variation of the caudal vertebral bodies. The reduced P intake restores the shape of vertebral bodies to a condition more similar to WT vertebral bodies. Moreover, the LP diet is shown to reduce the mineralisation of vertebral body endplates in some treated Chi/+ fish and to restore the osteoid layer, that is absent in untreated mutants. It is known that the presence of an osteoid (non-mineralised collagen) has a beneficial effect on the mechanical properties of bone. The collagenous bone matrix itself is a very tough material that can withstand extreme mechanical forces and bend without fracturing. Examples for tough low mineralised bones that do not fracture are deer antlers and the bones of human infants (18), but also teleosts under reduced P intake show tough and deformable vertebral bodies with no signs of fractures (22, 24, 25). In contrast, high dietary P causes excess mineralisation and higher bone stiffness in teleosts (22, 24, 25), and results in reduced bone formation in humans (37, 38) and reduced ultimate strength (maximal load) in rats (39). Our findings suggest that the mineral content of Chi/+ bone likely was reduced by the LP diet, which could explain the reduced incidence of vertebral column deformities and the restored shape of the vertebral bodies.

One of the most severe complications in OI patients are multiple compression fractures with collapse of the vertebral bodies (4, 29). Our findings show that Chi/+ zebrafish suffer from the same type of complications. Several compression fractures were identified on whole mount specimens and on histological sections. Which are the factors predicted to contribute to compression fractures? The first possible factor for compression fractures is the poor quality of OI bone, caused by mutated collagen production and excess mineralisation. As discussed above, over-mineralised bone fractures easily; both patients and Chi/+ zebrafish suffer from bone brittleness (5, 17). The second potential cause is reduced bone mass. Histological assessment of human biopsies showed that patients including children have thinner cortical bone and a reduced number of trabeculae in trabecular bone (4, 5). Likewise, two OI mice models (33, 34) and Chi/+ zebrafish have thinner bone elements compared to WT, thin vertebral body endplates, thin bone elements in the vertebral body central region and thin bone trabeculae. Thus, hypermineralised vertebral bodies with poor trabecular bone are at mechanical disadvantage and are easily subjected to compression fractures (5). Bone with low mass and excess of minerals cannot easily withstand mechanical forces exerted along the axial skeleton. The third potential factor contributing to compression fractures is the weakness of the spinal ligaments and intervertebral discs. Early reports described that OI patients lack vertebral stability because of the laxity of the spinal ligaments (2, 4) and intervertebral discs (40, 41). Similarly, the Col1a^Jrt/+ OI mouse model shows reduced cervical intervertebral space (42). Likewise, Chi/+ zebrafish show evidence of weak intervertebral ligaments. Although on histological sections the ligaments of Chi/+ fish display all the structural elements as in WT animals, Chi/+ vertebral endplates are often shifted against each other along the dorsal-ventral axis. In humans weak ligaments fail to provide sufficient support to the vertebral column and allow the progressive degeneration of the deformity into a compression fracture (4). It can be assumed that less severe complications appear first and subsequently progress into a severe compression fracture. To support this hypothesis, patients show microfractures of the vertebral body growth plates (4, 43). During growth in humans, vertebral growth plate microfractures can progress into complicated lesions such as compression fractures (29, 44).

Human OI patients show evidence of fracture repair, but often fractures heal with deformities (45). We show similar findings in Chi/+ zebrafish. Compression fractures in Chi/+ are also subjected to fracture repair but the vertebral bodies remain deformed. Analysis of histological sections from different specimens suggests the identification of fractures in different stages of repair and allows for the tentative reconstruction of the steps involved in vertebral fracture healing. Like in mammals, Chi/+ zebrafish fractures are initially stabilised by the formation of fibrocartilaginous calli, that subsequently become ossified. Despite the observation of fractures in different stages of repair, the design of our study does not allow to generate a timeline of fracture repair. In the goldfish (Carassius auratus), a cyprinid species like zebrafish, a bone fracture callus takes about 35 days to replace the fibrocartilaginous template (46). Thus, the presence of bone calli in compressed vertebral bodies of three months old Chi/+ indicates that mutants likely developed vertebral compression fractures before the start of the experiment. This could explain why both Chi/+ treated with the LP diet and untreated mutants show compression fractures. Compression fractures are therefore one of the most important complications during the initial stages of the disease, both in zebrafish and children affected by OI, and are crucial in the quick progression of the OI phenotype.

Bone from individuals with OI is characterised by an increased bone turnover rate and an increased number of osteoclasts, associated to an increased eroded surface compared to controls (47–49). Analogous findings derive from studies on two OI mouse models, the Oim (33) and Brtl (34) mice. Likewise, Chi/+ zebrafish exhibit increased osteoclastic activity, as demonstrated by TRAP staining on histological sections. Resorption activity is increased at the endosteal surfaces of the neural and haemal arches compared to WT animals. TRAP activity extends also to other bone structures that are not subjected to resorption in WT zebrafish, such as the vertebral body endplates and the bone trabeculae that connect the vertebral body endplates (50). Trabecular bone surfaces display high osteoclast activity also in the Oim and Brtl mice (33, 34). The bone in zebrafish can be remodelled by multinucleated osteoclasts, but different from mammals, thin bony elements are typically resorbed by mononucleated cells (27). Flat, elongated mononucleated osteoclasts occur at endosteal surfaces, for example of neural and haemal arches (27, 50), as also observed in WT specimens in this study. Bone resorption in zebrafish is primarily related to the demands of allometric growth. The common type of mononucleated osteoclasts does not create deep resorption lacunae (27). TEM from Chi/+ specimens showed osteoclasts with abundant electrondense cytoplasm characterised by several cytoplasmic vacuoles. These cells exhibit the typical ruffled border, indicative for active resorption of the bone matrix. Also osteoclasts in bone from the Oim mouse exhibited cellular changes. Oim cells have a larger diameter and exhibit three times the number of nuclei compared to osteoclasts in WT mice (51). The abundant cytoplasmic vacuoles of osteoclasts and the larger amount of TRAP-positive bone surfaces are indicative for increased resorptive activity in OI bone, both in Chi/+ zebrafish and mouse models (34, 51). While endosteal bone resorption is increased in human OI patients, osteoblasts produce less new bone, which results in a reduced osteoid thickness, as discussed above. The decrease in bone formation in human patients, reviewed by Fedarko (49), is consistent with data from the Oim and Brtl mouse models (33, 34). In vitro studies have shown that osteoblasts from human patients (reviewed by 49) and murine models (52) show decreased synthesis, processing and matrix incorporation of collagen compared to controls. The production of mutated collagen type I in OI osteoblasts results in delayed collagen folding and excess post-translational modifications, that cause intracellular retention of defective collagen and endoplasmic reticulum (ER) stress (10). At an ultrastructural level, Chi/+ zebrafish have osteoblasts in the vertebral growth zone with enlarged ER cisternae. Similar findings were previously described on osteoblasts from the caudal fin of adult Chi/+ (16). Here we show that ER cisternae in Chi/+ osteoblasts are filled with protein, likely mutated collagen type I. Intracellular collagen retention leads to collagen over modification and impairs collagen secretion in Chi/+ (16) and in OI models (52). The extracellular collagen matrix is also impaired. In close proximity of the osteoblasts, the collagen fibrils in Chi/+ fish are densely packed and lack the typical pattern of the osteoid collagen where the fibrils have a low density and are loosely arranged (53). In contrast, in Chi/+ the fibrils in the bone matrix have a high degree of compactness that is typical of fully mineralised bone (53). These findings correlate well with the lack of osteoid seams in Chi/+ and OI bone in general. It is widely accepted that insufficient osteoblast performance is at the basis of the severe OI phenotype. Thus, recent studies have targeted OI osteoblasts to relieve ER stress and improve cellular functions. The treatment with 4-phenylbutyrate, a chemical chaperone already approved by the FDA for urea cycle disorders, stimulates collagen secretion in murine OI osteoblasts in vitro (52), and improves the OI bone phenotype in vivo (16). In the present study, we show that two months of reduced dietary P administration partially rescue the OI phenotype in Chi/+. Mutants which received the LP diet show improved extracellular matrix with less dense collagen fibrils, indicative for a more typical osteoid. This finding together with the reduced incidence of vertebral column deformities and the rescue of the vertebral body shape, suggests that a reduced dietary P intake can alleviate the severe bone phenotype in juvenile Chi/+ zebrafish.

Wild type AB (WT) and heterozygous Chihuahua (col1a1a^dc124/+, Chi/+) zebrafish were bred in-house. The mutant Chi/+ carries a heterozygous c.2207G>A mutation in col1a1a causing a p.G736D (G574D) substitution in the α1 chain of collagen type I (15). Zebrafish embryos were kept in petri dishes in fish water (1.2 mM NaHCO₃, 0.01% instant ocean, 1.4 mM CaSO₄, 0.0002% methylene blue) at 28°C until 7 days post-fertilisation (dpf), then housed in ZebTEC semi-closed recirculation housing systems (Techniplast, Buguggiate, Italy) at 28°C, pH 7.5 and conductivity 500 μS on a 14/10 light/dark cycle. Zebrafish from 7 to 21 dpf were fed three times a day alternating commercial dry food (ZM000, Zebrafish Management Ltd., Winchester, UK) and brine shrimp (Artemia cysts, Zebrafish Management Ltd., Winchester, UK). Fish were then fed for another week three times a day with a dry regular P diet (22), see also below), until 28 dpf, to adjust them to this type of dry feed. Starting from 28 dpf, WT were fed three times a day with a regular P diet for two months. At 28 dpf, Chi/+ were randomly divided in two groups and fed three times a day with a ‘LP diet’ (low P content) or a regular P diet for two months (see below for details). Specimens were collected after two months of dietary treatment, euthanised by tricaine (3-amino benzoic acidethylester) overdose (0.3%) and fixed for further analyses as described below. The experiments were conducted in the centralised animal facility of the University of Pavia (Pavia, Italy). All animal studies were conducted in agreement with EU Directive 2010/63/EU for animals. The experimental protocol was approved by the Italian Ministry of Health (Approval animal protocol No. 260/2020-PR, 26 March 2020).

Diet composition and nutritional experiments are described in detail in (22). Briefly, the diets were formulated to have a total P content of 0.5% and 1.0%, termed low P (LP) diet and regular P diet, respectively (Table 3). Monoammonium phosphate (MAP) was used as dietary inorganic P supplement (54). In order to keep all diets equal in nutrients, except for P concentration, MAP replaced the inert filler diatomaceous earth (Diamol, Imerys, Denmark). The diets were formulated by SimplyFish AS (Stavanger, Norway, www.simplyfish.no) and produced by extrusion with subsequent crumbling to a suitable particle size by the Danish Technological Institute (Taastrup, Denmark, https://www.dti.dk). The P content of the product was verified at the University of Hohenheim (Stuttgart, Germany, https://www.uni-hohenheim.de) and determined with 5.04 g/kg diet and 9.84 g/kg diet for the low P and regular P diet, respectively (Table 3).

X-rays of WT, Chi/+ and Chi/+ under LP diet were acquired with a Faxitron Mx-20 (Faxitron, Tucson, Arizona, USA) using 25 kV for 10 sec. The Kodak DirectView Elite CR System and k-Pacs software (Kodak, Rochester, New York, USA) were used for image digitalisation.

WT (n=26), Chi/+ (n=20) and Chi/+_LP (n=18) were fixed for 24 h in 4% paraformaldehyde (PFA) in 1× phosphate-buffered saline (PBS) at C and were stained with Alizarin red S according to an established protocol (55). Fish were analysed and imaged using an Axio Zoom V16 stereomicroscope (Carl Zeiss, Oberkochen, Germany) with oblique illumination equipped with a 5MP CCD camera. Classification of the deformities was performed on whole mount-stained WT and mutants as defined by Martini et al. (56). Only the abdominal and caudal region of the vertebral column were analysed. The maximal curvature indices related to the abdominal and the caudal region of the vertebral column were calculated according to the method described by Marie-Hardy et al. (57). Briefly, the maximal kyphotic and lordotic indices were calculated in the sagittal plane of Alizarin red S stained specimens as the ratio (a/b) between the perpendicular distance from the axis in correspondence of the maximal curvature (segment ‘a’ in Figure 1A) and the standard length (described as the distance from the anterior most tip of the upper jaw to the posterior edge of the caudal peduncle where caudal fin rays insert (58), segment ‘b’ in Figure 1A). The same method but in the coronal plane was used to calculate the maximal scoliotic index. Moreover, for each group the following general metrics were calculated according to Martini et al. (56): the frequency (%) of specimens with at least one malformation, the total number of malformations identified, the average malformation load (total number of malformations diagnosed in a group/number of malformed individuals per group).

Lateral images of stained fish were used to quantitatively analyse mineralisation levels of vertebral endplates of WT, Chi/+ and Chi/+_LP as described in (22). The first 5 caudal vertebral bodies (59) were considered for analysis in all specimens, following the established protocol (22). Briefly, the non-mineralised endplate was expressed as a percentage of the total non-mineralised endplate length over the total vertebral length. Vertebral endplates with a non-mineralised percentage value greater than 10% were classified as low mineralised, between 3% and 10% as intermediate mineralised, and less than 3% were considered high mineralised.

Lateral images of whole mount Alizarin red S stained specimens were used to quantify the shape variation of the first 10 caudal vertebral centra (59) in WT (n=15), Chi/+ (n=13) and Chi/+_LP (n=12) by means of landmark-based geometric morphometrics. The landmarks, defined as biologically homologous anatomical loci recognisable on all specimens in the study (60), were selected on the vertebral centra as represented in Figure 2A: landmarks number 1, 5, 6 and 10 on the vertebral endplates; landmarks number 2, 4, 7 and 9 on the anterior and posterior cone of the centrum; landmarks number 3 and 8 in the central region of the centrum. A similar configuration of landmarks has been applied in a vertebral fracture assessment study in human patients (61). 2D landmarks were extracted from digital images using Fiji (NIH, Bethesda, Maryland, USA) and digitised in the same order for all vertebral centra analysed. Procrustes superimposition of digitised landmarks and visualisation of shape variations were performed using Past4.04 software (62). Principal component analysis was performed using PCAGen8 software as described in (60).

WT (n=4), Chi/+ (n=4) and Chi/+_LP (n=5) were fixed for 24 h in 2.5% PFA, 1.5% glutaraldehyde, 0.1 M sodium cacodylate buffer (pH 7.4) and 0.001% CaCl₂ at 4°C, decalcified in 0.1 M EDTA for 14 days at 4°C and embedded in glycol methacrylate (27). Sagittal 2 μm sections were cut on a Microm HM360 (Marshall Scientific, Hampton, New Hampshire, USA) automated microtome and were stained with toluidine blue (0.5% toluidine blue, 1% Na₂B₄O₇ in demineralised H₂O (dH₂O), pH 9 for 15 sec), rinsed in dH₂O and mounted with DPX. Images were acquired using an Axio Imager-Z1 microscope (Carl Zeiss, Oberkochen, Germany) equipped with an Axiocam 503 colour camera (Carl Zeiss, Oberkochen, Germany). Bone structure histomorphometry was analysed on images of toluidine blue stained sections of the middle plane of the vertebral column. Thickness of bone structures in the endplates, in the central region and in the trabeculae of 5 to 10 vertebral centra per specimen were measured using Fiji (NIH, Bethesda, Maryland, USA) (see Figure 3A for location). The mean values were considered for analysis.

For mineral detection on histological sections, WT and mutant zebrafish were selected based on X-rays to be representative for the phenotype. WT (n=1), Chi/+ (n=1) and Chi/+_LP (n=1) were fixed as described above and embedded in glycol methacrylate without carrying out decalcification. Sections of 2 μm were stained according to the Von Kossa/Van Gieson protocol (63). Images were acquired using an Axio Imager-Z1 microscope (Carl Zeiss, Oberkochen, Germany) equipped with a 5MP CCD camera.

WT (n=1), Chi/+ (n=1) and Chi/+_LP (n=1) were selected based on X-rays to be representative for the phenotype, fixed and decalcified as described above for histology, and embedded in epon epoxy medium (64). Semi-thin 1 μm sagittal sections were cut on a Microm HM360 microtome (Marshall Scientific, Hampton, New Hampshire, USA), stained with toluidine blue at pH 9 for 2 min, rinsed with dH₂O and mounted with DPX. For TEM analysis, ultrathin sections (about 70 nm) of the middle plane of the vertebral column were prepared on an UltracutE ultramicrotome (Reichert-Jung, Buffalo, New York, USA), contrasted with uranyl acetate and lead citrate and analysed with a Jeol JEM 1010 transmission electron microscope (Jeol Ltd., Tokyo, Japan) operating at 60 kV. Microphotographs were taken with a Veleta camera (Emsis, Muenster, Germany).

WT (n=2), Chi/+ (n=2) and Chi/+_LP (n=2) were fixed in 4% PFA in 1× PBS, pH 7.4, for 1 h at RT and decalcified in 4% PFA, 10% EDTA, pH 7.4 for 14 days at 4°C. Specimens were embedded in glycol methacrylate (27). Sagittal 5 μm sections were cut on a Microm HM360 (Marshall Scientific, Hampton, New Hampshire, USA) automated microtome and demonstration of tartrate-resistant acid phosphatase (TRAP) was adapted from (27). Briefly, sections were pre-incubated at 37°C for 45 min in 50 mL acetate buffer (0.1 M sodium acetate, 50 mM L(+) di-sodium tartrate dehydrate, pH adjusted to 5.5 with acetic acid) to which is added 0.5 mL of enzyme substrate solution (2% Naphtol AS TR phosphate dissolved in ethylene glycol mono-butyl ether). Shortly before use, 1 mL of pararosaniline solution (4% pararosaniline chloride (CI. 42500) in 7% HCl solution) was mixed to 1 mL of fresh 5% sodium nitrite and, after hexazotiation, was added to the acetate-enzyme substrate solution. Enzymatic reaction took place in 30-60 min at 37°C. Subsequently, slides were rinsed in dH₂O, counterstained with Meyers haematoxylin for 10 min, rinsed in running tap water for 10 min, flushed in dH₂O, dried at 40°C and mounted with DPX.

Quantitative variables are expressed as mean ± standard deviation, categories are expressed as percentages. Statistical analysis was performed using Past4.04 software (62). Differences in the occurrence of vertebral column deformities and in bone mineralisation levels were evaluated by means of Chi-square test followed by Bonferroni correction. Comparison of the maximal curvature indices was based on the non-parametric Mann-Whitney test. Differences in the thickness of bone structures were evaluated by means of Mann-Whitney non-parametric test followed by Bonferroni correction. For principal component analysis of geometric morphometrics data, significant differences in principal component 1 and principal component 2 were obtained using PCAGen8 software as described in (60) using Chi-square test (paired tests, WT-Chi/+ and WT-Chi/+_LP). A p value less than 0.05 was considered significant.