Transformed notochordal cells trigger chronic wounds destabilizing the vertebral column and bone homeostasis

Notochordal cells play a pivotal role in vertebral column patterning, contributing to the formation of the inner architecture of intervertebral discs (IVDs). Their disappearance during development has been associated with reduced repair capacity and IVD degeneration. Notochord cells can give rise to chordomas, a highly invasive bone cancer associated with late diagnosis. Understanding the impact of neoplastic cells during development and on the surrounding vertebral column could open avenues for earlier intervention and therapeutics. We investigated the impact of transformed notochord cells in the zebrafish skeleton using a RAS expressing line in the notochord under the control of the Kita promoter, with the advantage of adulthood endurance. Transformed cells caused damage in the notochord and destabilised the sheath layer triggering a wound repair mechanism, with enrolment of sheath cells (col9a2+) and expression of wt1b, similar to induced notochord wounds. Moreover, increased recruitment of neutrophils and macrophages, displaying abnormal behaviour in proximity to the notochord sheath and transformed cells, supported parallels between chordomas, wound and inflammation. Cancerous notochordal cells interfere with differentiation of sheath cells to form chordacentra domains leading to fusions and vertebral clefts during development. Adults displayed IVD irregularities reminiscent of degeneration; reduced bone mineral density, increased osteoclast activity; while disorganised osteoblasts and collagen indicate impaired bone homeostasis. By depleting inflammatory cells, we abrogated chordoma development and rescued the skeletal features of the vertebral column. Therefore, we showed that transformed notochord cells alter the skeleton during life, causing a wound-like phenotype and activating chronic wound response, suggesting parallels between chordoma, wound, IVD degeneration and inflammation, highlighting inflammation as a promising target for future therapeutics. D is ea se M o de ls & M ec ha ni sm s • D M M • A cc ep te d m an us cr ip t Introduction The vertebral column is the central axis of the skeleton in all vertebrates. It is composed of segments (vertebrae) connected by joint-like structures called intervertebral discs (IVDs). In mammals, the architecture of the IVDs is made by an annulus fibrosus (AF), a collagenous layer surrounding a hydrated and gelatinous nucleus pulposus (NP) core, which contains chondrocyte-like cells derived from embryonic notochord cells (Rodrigues-Pinto et al., 2014). The disappearance of notochordal cells in mammals during development of the vertebral column has been linked to reduction of repair capacity and IVD degeneration (IVDD) (Wang et al., 2017). Occasionally, notochordal cells can cause vertebral malformations and in rare cases cell transformation lead to chordomas (Salisbury, 1993, Choi et al., 2008), a rare bone cancer of the axial skeleton and skull base (McMaster et al., 2011). With an incidence of approximately one in a million, chordomas account for about 14% percent of all primary bone malignancies and 20% of primary spinal tumours (Chugh et al., 2007). Chordomas are slow growing and highly resistant to both chemotherapy and radiotherapy, meaning that radical surgery is often the primary choice for treatment modality (McMaster et al., 2011). Unfortunately, in many cases, the proximity of chordomas to vital structures, means that local excision is rarely achieved, resulting in a recurrence rate greater than 50% (Stacchiotti et al., 2017, Barry et al., 2011). Distant metastases to lung, bone, soft tissue, lymph node, liver, and skin have been reported in up to 43% of cases (Stacchiotti et al., 2017, Barry et al., 2011). Interestingly, chordomas lead to changes in bone quality, and often appear on X-rays and computerised tomography (CT) as eroding bone lesions with associated soft tissue calcification (de Bruine and Kroon, 1988), suggesting modifications in the behaviour of the nucleus pulposus cells disrupt disc and bone homeostasis. Impairment of disc homeostasis is a hallmark of IVDD (Novais et al., 2020b), which unlike chordomas is very common, representing the most common cause of back pain (Zheng and Chen, 2015), a D is ea se M o de ls & M ec ha ni sm s • D M M • A cc ep te d m an us cr ip t symptom that 80% of the adult world population suffer from (GBD et al., 2017). How transformed nucleus pulposus cells affect the IVD and surrounding vertebrae during their development is currently unknown; and no animal models to show how transformed cells dynamically interact with and affect the IVDs and vertebral column in vivo have been described. Such models could contribute to our understanding of chordoma development, IVDD, the interaction of the nucleus pulposus with the skeletal tissues and feature possible therapeutic avenues for both conditions. Zebrafish have emerged as an advantageous animal model for a variety of human diseases, including cancer and skeletal diseases, due to their fast development, tractability, flexible genetic manipulation (transgenesis, forward and reverse genetics) and their translucency (Bergen et al., 2019). Reporter lines allow in vivo assessment of cell behaviour not only during early development but also during the later stages of skeletal formation in juveniles (Bergen et al., 2019). Zebrafish have high tissue regenerative capacity, with the ability to restore vacuolated cells of the notochord upon injury (Garcia et al., 2017). In zebrafish, notochord cells remain throughout life; they are enveloped by a sheath layer that acts as a sealing basement membrane to isolate the inner notochord vacuolated cells, and carries high potential to mineralise (Fleming et al., 2004, Stemple, 2005). The notochord sheath plays an important role in the segmentation of the vertebral column and centra primordium (chordacentra) formation (Lleras Forero et al., 2018, Pogoda et al., 2018, Wopat et al., 2018). Following genetic manipulation, mechanical injury (needle punctures) or chemical treatment (with nystatin), repair of tissue damage appears to involve a sub-population of notochord sheath cells which become activated, expressing Wilms Tumor 1 (wt1b), and migrate towards the wound, setting landmarks during notochord repair (Garcia et al., 2017, Lopez-Baez et al., 2018). D is ea se M o de ls & M ec ha ni sm s • D M M • A cc ep te d m an us cr ip t Chordoma onset has been described in larval zebrafish expressing the oncogene RAS in the notochord, using the bimodal Gal4/UAS system and activation of the oncogenic RTK/Ras pathway (Burger et al., 2014). These zebrafish chordoma models become affected within the first 3 days post-fertilisation (dpf), progressively developing notochord hyperplasia , similar to histological features of human chordomas (Burger et al., 2014). Recently, the zebrafish chordoma model was used to test genetic potential to transform the notochord in vivo, providing suggestive evidence that Brachyury (TBXT), a highly expressed gene in human chordomas (Vujovic et al., 2006), is insufficient to initiate chordomas, instead suggesting activation of members of the RTK signalling as potential players in chordoma formation (D'Agati et al., 2019). The behaviour of notochord cancer cells during zebrafish life has not yet been studied, due to early lethality of chordoma models during larval stages. It is unknown whether notochord cancer cells trigger a wound repair mechanism similar to those of notochord injury models, which activate an acute inflammatory response as is seen in other early cancers (Feng et al., 2012, Feng et al., 2010). It is also unclear whether notochord cancer cells exert control as notochordal remnants to interfere with bone formation and, later in life, with bone homeostasis. Here we studied the interactions between the notochord cancer cells within the forming vertebral column and bone homeostasis using a well-characterised transgenic line, Kita-RAS, which drives expression of HRASV12 in the notochord (and in melanoblasts, thus modelling melanoma) and survives to adulthood (Santoriello et al., 2010, van den Berg et al., 2019, Feng et al., 2012). We showed that “transformed” notochord cells destabilise the notochord sheath layer, activating a chronic wound repair response similar as those caused by induced notochord wounds previously described (Garcia et al., 2017, Lopez-Baez et al., 2018). These preneoplastic cells lead to invagination of the col9 expressing notochord sheath cells towards the wound and participation of wt1b notochord sheath sub-population. Interestingly, macrophages D is ea se M o de ls & M ec ha ni sm s • D M M • A cc ep te d m an us cr ip t and neutrophils were present in higher numbers and showed prolonged interaction with the wounded notochord sheath layer, as described in other cancers. The metameric pattern of segmentation of the vertebral column was compromised, but not abrogated, leading to vertebral fusions and clefts. Adult bone homeostasis was altered as observed by differences in vertebral bone mineral density and collagen fibre distribution. Transformed cells also compromised the adult zebrafish equivalent intervertebral disc architecture, leading to NP “scar” tissue, NP cellular disorganisation and affecting the structure of the AF, similar to IVDD. Chordoma development and skeletal defects were rescued when we partially depleted neutrophils and macrophages. In conclusion, our results indicate that transformed notochord cells cause chronic wounds leading to inflammation, vertebral abnormalities, disc and bone homeostasis impairment. Chordoma development could be controlled by limiting inflammation, revealing new avenues for therapeutics and highlighting the use of zebrafish as an animal model. Results Kita-RAS induces wound-like destabilisation of the notochord Notochord-specific Gal4 lines crossed to UAS:EGFP-HRASV12 have been previously described as powerful models for inducing chordomas in zebrafish (Burger et al., 2014). A transgenic line extensively used to induce melanoma, in which HRASV12 expression is driven by the Kita promoter in melanoblasts, goblet cells and in the notochord cells (due to the presence of an enhancer element for tiggy winkle hedgehog, twhh)(Distel et al., 2009) has the advantage over other notochord RAS expressing lines because it survives to adulthood (Santoriello et al., 2010, van den Berg et al., 2019). We used Kita-RAS-GFP and Kita-RASmCherry to study the progressive changes of the transformed notochord cells and their interaction with the forming vertebral column. In 5dpf zebrafish larvae, the outer layer of the notochord is formed by an epithelial-like sheath wrapping notochord vacuolated cells (Wopat et al., 2018) (Fig. 1A). Confocal images through the notochord, at 5dpf, showed that Kita drives D is ea se M o de ls & M ec ha ni sm s • D M M • A cc ep te d m an us cr ip t reporter expression in the notochord vacuolated cells, but not in the sheath cells (Fig. 1B). As in other chordoma RAS models, Kita-RAS led to dramatic destabilisation of the notochord vacuolated cells starting as early as 3dpf and by 5dpf affected 70% of the larvae (>200 larvae analysed). Affected larvae were considered when they displayed more than 3 lesions in the notochord. Each lesion was characterised by increased RAS expression and abnormal notochord cell morphology (Fig. 1B). At the same developmental stage (5dpf), notochord cells were interspaced by infiltration of non-vacuolated cells and accumulation of fibrous collagenous tissue (AFOG staining, red colour) (Fig. 1B-C). Furthermore, histological sections suggested local destabilisation of the notochord sheath layer at the region of collapsed vacuolated cells (Fig. 1C). To analyse cell proliferation, we treated larvae with EdU solution to be incorporated into the DNA of proliferating cells from 2 to 4dpf and followed by counting the number of EdU-positive (+) cells at 5dpf from confocal images. Notochord cells and notochord sheath cells in Kita-RAS are highly proliferative (p= 0.0002) (Fig. 1D and E). Interestingly, the organisation of the notochord in Kita-RAS fish displayed cellular characteristics reminiscent of those observed in notochord wounding models (needle puncture)(Lopez-Baez et al., 2018) (Fig. S1), suggesting that chordoma may recapitulate repair mechanisms, as has been suggested for several other cancers (Feng et al., 2010). Pre-neoplastic notochord cells trigger the notochord wound repair mechanism in zebrafish Wounds in the notochord induced by needle injury, amputation and chemical damage lead to the collapse of notochord vacuolated cells, sheath cell invasion and expression of Wilms Tumor 1b (wt1b) within a cell sub-population of the notochord sheath (Garcia et al., 2017, Lopez-Baez et al., 2018). To investigate whether pre-neoplastic notochord cells mimic a wound-like response, we crossed Kita-RAS-mCherry with Tg(col9a2:GFPCaaX), a marker for the notochord sheath layer (Fig. 2A), and to Tg(wt1b:gfp) to label the sub-population of sheath D is ea se M o de ls & M ec ha ni sm s • D M M • A cc ep te d m an us cr ip t cells that standardly respond to damage. We confirmed that at 5dpf Kita is not expressed in the notochord sheath layer, and only in vacuolated notochord cells (Fig. 2B). We observed col9a2 expression in regions of damage within the notochord, suggesting sheath cell migration towards the chordoma wounded area (Fig. 2B). Cross sections through the notochord, at 5dpf, showed col9a2 expressing cells within the notochord in connection with the notochord sheath (Fig. 2B), reinforcing the possible migration of sheath cells to the lesioned region. To check for cell abnormalities in the notochord sheath, we quantified the cell area of the col9+ cells of severely affected larvae within two regions of our Kita-RAS, woundproximal and distal (Fig. 2C). Kita-RAS showed significant reduction in cell area in wound-proximal regions (p< 0.0001), but not in wound-distal regions when compared to controls (Fig. 2C). We did not detect cell area changes in the sheath layer of less affect larvae. Therefore, wound-like lesions caused by transformed notochord cells lead to local cellular modifications in the sheath layer. Next, we analysed wt1b expression in the Kita-RAS outcrossed fish. Control fish exhibited no expression of wt1b in the notochord, whereas Kita-RAS showed strong wt1b expression by preneoplastic cells located at severe wounded regions in 100% of the cases analysed (20/20) (Fig. 2E). These findings corroborate strong parallels between cancer and wound repair (MacCarthyMorrogh and Martin, 2020). Wounded notochord sheath elicits a prolonged recruitment of innate inflammatory cells Several studies have reported that oncogene-transformed cells trigger an innate inflammatory response, with both neutrophils and macrophages recruited to the pre-cancerous tissue (Chia et al., 2018, Feng et al., 2010, Freisinger and Huttenlocher, 2014, Roh-Johnson et al., 2017). This recruitment of neutrophils and macrophages is responsible for clearing cell debris and to orchestrate tissue repair responses including wound angiogenesis and matrix deposition (Eming et al., 2017). We questioned whether oncogenic RAS expression in the notochord cells and the lesioned notochord sheath might also induce an inflammatory response in our zebrafish D is ea se M o de ls & M ec ha ni sm s • D M M • A cc ep te d m an us cr ip t chordoma model. During the first weeks of development, zebrafish do not have a functional adaptive immune system, allowing us to investigate the innate immune response on its own (Renshaw and Trede, 2012). We performed time-lapse imaging at 5dpf and analysed the interactions of neutrophils and macrophages with the notochord sheath layer. For neutrophils, we incrossed Tg(kita:Gal4; UAS:mCherry; UAS:HRASG12V-GFP;lyz:DsRed), and selected RAS-/lyz+ larvae, Tg(kita:mCherry;lyz:DsRed), and RAS+/lyz+ larvae, Tg(kita:HRASG12VGFP;lyz:DsRed), as controls and Kita-RAS fish, respectively. While for macrophages, we incrossed Tg(kita:Gal4;UAS:mCherry;UAS:HRASG12V-GFP;mpeg:FRET), selected RAS/mpeg+ larvae, Tg(kita:mCherry;mpeg:FRET) and RAS+/mpeg+ larvae, Tg(kita:HRASG12VGFP;mpeg:FRET), as controls and Kita-RAS fish, respectively. Higher numbers of neutrophils and macrophages were recruited, making a prolonged direct contact with the wounded notochord sheath in Kita-RAS in comparison with controls (Fig. 3, Fig. S2 and Movies 1 and 2), similarly to the inflammatory response previously reported in the melanoma model (Feng et al., 2010). Remarkably, we also found neutrophils and macrophages infiltrating wounded regions and in direct contact with notochord vacuolated cells (Fig. S2 and Movies 1 and 2). Together our results showed that zebrafish chordoma induces a chronic notochord inflammatory wound response with typical wound recruitment of neutrophils and macrophages. Inflammatory cells trespass the notochord sheath layer in wounded regions to form direct contact with transformed notochord cells, a similar behaviour described for other cancers (Feng et al., 2012). Depletion of neutrophils and macrophages abolishes chordoma development To further test whether the increased innate inflammatory response triggers the proliferation of neoplastic cells leading to wounds in the notochord, we transiently delayed innate immune cell development by injecting pu.1 and gcsfr morpholinos (MO) (double knockdown), at the one D is ea se M o de ls & M ec ha ni sm s • D M M • A cc ep te d m an us cr ip t cell stage embryos generated by incrosses of Kita-RAS-GFP fish (Fig. S3). Combined pu.1 and gcsfr MO injections are used to transiently arrest myeloid lineage development in larval zebrafish until at least 4dpf, therefore generating larvae lacking neutrophils and macrophages (Feng et al., 2012, Liongue et al., 2009, Rhodes et al., 2005). We confirmed the efficiency of our morpholino experiment by injecting fish carrying labelled neutrophils and macrophages at 3dpf (Tg(lyz:DsRed;mpeg:FRET))(Fig. S3B). Blocking the development of inflammatory cells in Kita-RAS resulted in a reduction of larvae exhibiting wounded (> 5 lesions) notochordal phenotype from 44.37 % (control MO) to 8.56 % (pu.1 + gcsfr MO) (p< 0.0001) at 3dpf (Fig. S3C and D). In addition, fish with affected notochord (8.56%) in the pu.1 + gcsfr MO group showed a less severe (≤ 5 lesions) phenotype in comparison to the control MO group, suggesting that incomplete ablation of inflammatory cells can ameliorate chordoma. To complement our morpholino experiment, we used CRISPR/Cas9 system to target pu.1 and gcsfr simultaneously. We were able to cause mutations with an efficiency rate of 80%, validated by fragment length analysis, for each individual genes, at 5dpf. We analysed KitaRAS larvae from morpholinos (MO) and CRISPR injections side-by-side at 5dpf (Fig. 4A). CRISPR injections led to a significant reduction in numbers of neutrophils (p= 0.0012) and macrophages (p= 0.0478), but this reduction was not as pronounced as that observed from MO injections (p<0.0001) (Fig. 4B-D). Morpholinos also led to a significant reduction in proliferation of notochord and notochord epithelium cells in Kita-RAS (Fig. 4E and F). While CRISPR injections reduced cell proliferation, they did not show statistical difference from Kita-RAS (p= 0.2422)(Fig. 4E and F). In comparison with non-affected notochords from controls, fluorescent stereomicroscopy pictures from Kita-RAS wounded notochords displayed different profiles of average pixel intensity. Notochordal lesions are detected by increased pixel intensity and enlargement of peak areas (Fig. 4G). This unbiased method allowed us to quantify the severity of notochordal wounds among the studied groups and to analyse whether we could D is ea se M o de ls & M ec ha ni sm s • D M M • A cc ep te d m an us cr ip t rescue the affected notochordal phenotype upon MO and CRISPR injections. We compared Kita (control), Kita-RAS and Kita-RAS injected with either MO or CRISPRs. Similar to our cell proliferation experiment, we detected a partial notochordal rescue with CRISPR injections and significant rescue with MO (Fig. 4H). Therefore, we have shown that the increase in neutrophils and macrophages contribute to proliferation of cancer cells in the notochord and modulation of inflammatory cells could prevent clonal expansion and chordoma development, similar to what has been previously shown for melanomas (Feng et al., 2012). Abnormal pattern of vertebral segmentation and mineralisation in Kita-RAS fish It has been demonstrated that notochord damage can lead to defective patterning of the vertebral column (Lopez-Baez et al., 2018, Fleming et al., 2004, Nguyen-Chi et al., 2014). Given that Kita-RAS cause cellular changes and a wound-like response in the notochord we questioned whether these events might have a downstream impact in the vertebral column segmentation. We crossed Kita-RAS to Tg(entpd5:kaeda), an early marker of the notochord segmentation and biomineralizing activity. Entpd5 hydrolyses nucleoside triphosphates, providing local inorganic monophosphate for biomineralization (Dallas and Bonewald, 2010, Huitema et al., 2012). During development of the vertebral column, entpd5 is expressed in alternating segments of the sheath, which will form the mineralised chordacentra; while the interdomains will develop into intervertebral discs (IVDs) (Fig. 5A and D) (Wopat et al., 2018). We analysed larvae at 8dpf, at a stage when segmentation has started but is not yet finalised. A delay in chordacentra formation was observed in Kita-RAS, compared to control of similar range of length (3.8 to 4.1 mm) (Fig. 5B). Regions in which the notochord cells were compromised in Kita-RAS coincided with mis-patterning and ectopic expression of entpd5:kaeda (Fig. 5C). Expansion of the domain of each segment was observed ectopically in the future IVD area. These results indicate that cellular abnormalities of the notochord sheath D is ea se M o de ls & M ec ha ni sm s • D M M • A cc ep te d m an us cr ip t compromise the differentiation of col9+ sheath cells towards expression of entpd5 in predetermined chordacentra domains during segmentation. Moreover, our findings suggest a role of the sheath layer and notochordal cells in domain specification. A major advantage of our Kita-RAS model in comparison with other notochord induced RAS models (Burger et al., 2014, D'Agati et al., 2019, Distel et al., 2009) is the fish survival to adult stages, beyond the stages of development that have been previously reported. This allowed us to study the effect of pre-neoplastic cells on the skeletal formation and homeostasis. To check for abnormalities in mineralized vertebral column segments, we used in vivo and ex vivo Alizarin Red S staining in 14dpf fish. We detected abnormal and uneven mineralization of the chordacentra along the whole notochord, compromising length and shape of the segments and the future IVD domains (Fig. 5E-G). We measured the length of the first seven mineralised vertebral segments from fish displaying similar sizes (5 ≤ fish length < 6 mm) (Fig. 5G). Kita-RAS showed high variability and overall reduced length of segments (Fig. 5E). Our results indicate that the presence of notochord cancer cells leads to a wounded notochord sheath which modifies vertebral column segmentation pattern through ectopic activation of entpd5 and subsequent mineralization, which ultimately may cause vertebral fusions. Transformed notochord cells lead to vertebral column fusions and clefts Next, we sought to investigate the impact of pre-neoplastic cells in the vertebral column architecture. For that, we analysed the adult vertebral column, looking for resulting bone abnormalities. We used Alizarin Red staining (controls n = 10; Kita-RAS n = 10; 6 months post-fertilisation 6mpf), X-rays (controls n = 40; Kita-RAS n = 78; 1 year old fish) and microcomputerised tomography (μCT) (controls n = 5; Kita-RAS n = 5; 6mpf) to compare KitaRAS with control fish of the same age. Vertebrae fusions were found in 100% of Kita-RAS and in 0% of controls (controls n = 40; Kita-RAS n = 78) (Fig. 6 and Fig. S4). Fusions involved two or more vertebrae along the vertebral column leading to shortening of the total fish length. D is ea se M o de ls & M ec ha ni sm s • D M M • A cc ep te d m an us cr ip t Those fish with most fusions had the most reduced lengths (Fig. 6A, E and Fig. S4). The ribs were the most severely affected region of the vertebral column. We calculated the length of six consecutive mineralised segments of the vertebral column, separated by well-defined IVDs (Fig. 6A, dashed region). Besides uncovering increased length of segments due to vertebral fusions (p= 0.0411), it highlighted high variability within the same vertebral column region of Kita-RAS fish, demonstrating that there was no common developmental pattern of fusions (Fig. 6C and D). Analysis of fish length from X-ray images reinforced the length reduction observed in Kita-RAS (p< 0.0001) (Fig. 6E and F). Kita-RAS also displayed shape abnormalities of vertebrae and arches, including enlarged regions, broadening of arches and ectopic bone growth (Fig. S4). Enlarged areas were found in 40% of Kita-RAS (Fig. S4B). Ectopic bone growth can be better visualised with higher resolution μCT (5μm) (Fig. 6G) and Alizarin Red staining (Fig. S4C). Clefts through the centra and hemicentrae were found in 70% of fish analysed. These resembled butterfly abnormalities as occasionally described in human vertebral columns (Katsuura and Kim, 2019), and those malformations involving notochordal remnants (Fig. 6G’’’) (Oner et al., 2006). When staining 1 month old (1mpf) Kita-RAS with Alizarin Red, we detected hyperplastic cells contributing to a chaotic notochord cell arrangement along the vertebral column, and failure to organise in IVDs domains, revealing regions of incomplete mineralisation, originating clefts (Fig. 6H). To visualise osteoblasts, we crossed Kita-RAS fish to Tg(osx:NTR-mCherry), an osteoblast reporter line, and analysed the vertebral column at 1mpf. While in controls the osteoblasts were distributed evenly through the arches and centra, Kita-RAS showed increased osteoblast signal and patchy distribution, with some regions displaying dense concentrations of osteoblasts while others lacked these cells. Quantification of osteoblasts was performed for two consecutive vertebrae in each fish (n= 3), confirming increase in osteoblasts in Kita-RAS (p= 0.0028) (Fig. S5D). Moreover, we detected irregular recruitment of osteoblasts to the chordacentra throughout the vertebral D is ea se M o de ls & M ec ha ni sm s • D M M • A cc ep te d m an us cr ip t column. Thus, changes in the notochord lead to abnormal osteoblast recruitment and behaviour. Next, we asked whether reduction of inflammatory cells could rescue the bone phenotype. We looked at the vertebral column of controls and Kita-RAS + CRISPR (pu.1 + gcsfr) fish at 1mpf by Alizarin Red staining. The severity of the vertebral column phenotype was scored depending on the number of fusions and clefts observed. Kita-RAS + CRISPR (pu.1 + gcsfr) partially rescue the vertebral column phenotype (Fig. 6J and K), with a subset of fish showing no fusions or clefts (Fig. 6K). Therefore, modulation of innate immune cells in our chordoma model prevents vertebral fusions and clefts. Compromised intervertebral discs and impaired bone quality in adult Kita-RAS Embryonic notochordal cells contribute to the formation of the intervertebral disc nucleus pulposus (NP), which plays an important role in regulating disc homeostasis (Choi et al., 2008). We sought to understand the impact of transformed notochord cells in the adult zebrafish intervertebral disc equivalent regions and vertebral bone. By calculating bone mineral density, we detected a significant TMD decrease in Kita-RAS (p= 0.0015) (Fig. 6A and B), indicative of impaired bone quality. We performed histological sections of the adult vertebral column and observed highly fibrotic NP, similar to IVD degeneration (IVDD), with disorganised cellularity found in enlarged vertebrae (Fig. 7A and B). Fibrosis was detected in proximity with the notochord sheath layer. AFOG and Picro-sirius red staining confirmed fibrosis and connectivity with the notochord sheath, showing increased collagen content and increased collagen fibre thickness (Fig. 7B and C). In contrast to IVDD, dehydration did not describe the phenotype of Kita-RAS NP, as an increase in glycosaminoglycans was detected (Fig. S5). Additionally, despite fibrosis and disorganisation of the NP, due to cell transformation, we did not observe intervertebral disc calcification, a feature commonly found during IVDD and ageing (Novais et al., 2020a). The outermost component of the discs, the annulus fibrosus (AF), was replaced by bone in IVDs that were compromised by fusions. The structured layers of D is ea se M o de ls & M ec ha ni sm s • D M M • A cc ep te d m an us cr ip t collagen and elastin that form the zebrafish AF were completely lost in some of the IVDs (Fig. 7D). Interestingly, disorganised and increased number of osteoblasts were detected in the IVD region, corroborating altered osteoblast activity at the endplates of adult fish. The balance between osteoblasts and osteoclasts is key in bone homeostasis and control of bone density. Moreover, osteoclasts are derived from the same cell lineage of macrophages. We performed whole-mount TRAP staining to visualise osteoclast activity. Quantification of TRAP staining revealed exacerbated bone resorption in Kita-RAS (p= 0.0026), especially in affected areas of the vertebral column (Fig. S5B and C). Picro-sirius red staining suggested a reduction in collagen fibre thickness in the bone (centra). We quantified the mean intensity of red, green and blue pixels from pictures stained with Picro-sirius red. We detected a significant reduction in red (p= 0.0004) and blue (p= 0.0016) pixels, indicating an abnormal fibre organisation and confirming bone quality impairment in Kita-RAS (Fig. 7C). We conclude that transformed cells in the notochord lead to vertebral column and intervertebral disc abnormalities affecting the NP and AF, impairing osteoblasts and osteoclasts activity, consequently altering bone homeostasis in zebrafish. Discussion “Tumours are wounds that do not heal” was postulated in a classic work published by Harold Dvorak in 1986 (Dvorak, 1986). Dvorak recognized that the composition of the tumour stroma strongly resembled healing skin wounds, suggesting activation of the wound-healing response in the host. Moreover, cancer is frequently the consequence of chronic inflammatory disease (Schafer and Werner, 2008). Given the confined nature of notochordal cells during development of the vertebral column, would pre-neoplastic notochordal cells trigger chronic inflammation as other cancers do? And what is the impact of transformed cells in disc and bone homeostasis? By demonstrating that transformed notochord cells, provoke chronic notochordal D is ea se M o de ls & M ec ha ni sm s • D M M • A cc ep te d m an us cr ip t wounds and activate wound response mechanisms in zebrafish, leading to inflammation, vertebral column abnormalities and impairment of disc and bone homeostasis, we demonstrated parallels between wound repair, cancer and IVDD in a zebrafish chordoma model. The UAS:EGFP-HRASV12 transgene has been successfully used to transform notochordal cells and melanoblasts, contributing to in vivo modelling of chordomas and melanomas (Burger et al., 2014, Feng et al., 2010, Santoriello et al., 2010, D'Agati et al., 2019). Here, we made use of the robustness of RAS expression systems to efficiently induce chordomas, using the stable line Kita-RAS, an adult melanoma model with notochordal RAS expression. Kita-RAS caused similar larval notochord morpho-pathological changes as previously described for twhh:Gal;UAS:HRASV12 and 4465:Gal;UAS:HRASV12 (Burger et al., 2014), serving as tools to investigate neoplastic notochord cells in adults. While Kita-RAS has been extensively used to study melanomas, the vertebral column can still be studied in adult fish without complications of skin tumour, as only around 20% of adult fish develop melanomas (Anelli et al., 2009). Despite unlikely, the involvement of melanocytes in the advancement of chordoma cannot be ruled out from our model. Due to melanoma active interaction with immune cells, exacerbated immune activity could possibly lead to worsening of the notochordal phenotype, and it should be further investigated. Alternatively, Kita-RAS when crossed with a pigment free line, such as casper (complete lack of melanophores and iridophores) or nacre (mutation in mitfa) (White et al., 2008) can prevent melanoma development. As for UAS:EGFP-HRASV12 chordoma models, a limitation of the melanoma model is the fact that mutations of RAS members are not common in chordoma. However, RAS-transformed cells lead to activation of downstream signalling driven by EGFR, a cell surface receptor highly involved in chordomas, and mimics upstream receptor tyrosine kinase (RTK) activation (Burger et al., 2014). D’Agati et al, recently demonstrated that while D is ea se M o de ls & M ec ha ni sm s • D M M • A cc ep te d m an us cr ip t Brachyury (tbxt) overexpression did not have a tumour-initiating potential to transform notochord cells, when the authors tested RTK, including EGFR, they were able to trigger notochord hyperplasia, suggesting RTK signalling as a possible initiating event in chordoma (D'Agati et al., 2019). Although human chordomas are thought to originate from hyperplasia of notochordal remnants, benign notochordal remnants are occasionally found and are associated with vertebral abnormalities, such as vertebral clefts and bifurcations (Oner et al., 2006). When we looked at the adult Kita-RAS we observed vertebral clefts and hemivertebra that recapitulate human vertebral column abnormalities. However, vertebral malformations might not be a direct effect from pre-neoplastic notochordal cells, but a result from abnormal notochordal cell behaviour. Recent studies have shown that notochord vacuoles function as a hydrostatic scaffold that guides symmetrical growth of vertebrae and spine formation. Vacuole fragmentation caused by mutations in dstyk (spzl mutant) resulted in vertebral centra malformation and scoliosis (Bagwell et al., 2020, Sun et al., 2020). Similar to our observations, these studies evidenced that abnormal behaviour of notochord vacuolated cells are associated with vertebral malformations like to those of notochordal remnants in human. Furthermore, hemivertebra and clefts were systematically found in another mutant, spondo, carrying a mutation in cmn (Calimmin, a teleost-specific extracellular matrix protein with weak similarity to Elastin, and expressed in the notochord sheath), due to abnormalities in the notochord sheath layer (Peskin et al., 2020). Here, we demonstrated that destabilisation of the notochord vacuolated cells also triggered cellular changes in the notochord sheath layer (Fig. 8). Hence, revealing double and overlapping routes in which notochord neoplastic cells compromise the formation of the vertebral column: the inner vacuolated cells and the outer notochord sheath cells. D is ea se M o de ls & M ec ha ni sm s • D M M • A cc ep te d m an us cr ip t Notochord damage also leads to vertebral column abnormalities, including fusions and segmentation mispatterning (Lleras Forero et al., 2018, Wopat et al., 2018, Pogoda et al., 2018). We showed that Kita-RAS mimicked notochordal damages and induced repair mechanism as demonstrated by activation and invagination of col9+ notochord sheath cells and expression of wt1b in wounded areas, as previously described for notochordal wounds (Lopez-Baez et al., 2018, Garcia et al., 2017). Our findings suggest a key role of the notochord sheath and wound repair in chordoma. Interestingly, when RAS is activated in the notochord sheath specifically with col2a1a driving RAS, it also causes chordomas (D'Agati et al., 2019), sustaining a key role of the sheath layer in zebrafish chordomas. As neoplastic cells are continuously modifying the notochord, this causes wounds that seem to progress and remain chronic or unresolved. We showed for the first time that wounding provoked by transformed notochord cells triggers the recruitment of neutrophils and macrophages. Innate immune cells not only were present in higher number but changed their behaviour by prolonging their interaction time with the notochord sheath in wounded regions; in some cases they were able to breach the sealing membrane and achieve direct contact with cancer cells. It has been recently described that inflammatory cells make use of pre-existing holes in the basement membrane to gain access and reach pre-neoplastic cells in a melanoma model (van den Berg et al., 2019). In our chordoma model, inflammatory cells were observed in direct contact with pre-neoplastic cells in regions of severe notochord sheath wounds, which similarly, may serve as breaches in the notochord sheath to allow neutrophils and macrophages to reach pre-neoplastic cells. The interaction between neutrophils/macrophages and transformed cells have been beautifully described for melanoma in zebrafish, with formation of cytoplasmic tether linking the two cell types and engulfment of transformed cells by neutrophils and macrophages (Feng et al., 2010). H2O2, a key damage signal directing recruitment of neutrophils to a wound, was also identified as the major component drawing recruitment of leukocytes to the transformed cells (Feng et D is ea se M o de ls & M ec ha ni sm s • D M M • A cc ep te d m an us cr ip t al., 2010). Remarkably, when we depleted innate immune cells using morpholinos or CRISPR, we could rescue the notochord phenotype by inhibiting the aberrant proliferation of transformed cells, as demonstrated for melanoma (Feng et al., 2010), and partially rescuing the skeletal phenotype using CRISPR, by showing reduction of vertebral fusions. Thus, highlighting parallels between cancer and wound, and suggesting that immunomodulation might be a promising treatment for chordomas. When zebrafish notochord is infected with E.coli, Nguyen-Chi et al showed strong and persistent recruitment of neutrophils and macrophages (Nguyen-Chi et al., 2014). The authors also showed that il1b is partially required for recruitment of neutrophils but not macrophages. Fascinatingly, degranulation of neutrophils led to destruction of the host tissues and adult vertebral column defects, involving clefts and fusions. il1b morphants reduced neutrophil recruitment and prevented anterior notochord lesions. Altogether, inflammation appears to play an important role in controlling notochord damage and adult bone phenotype (Nguyen-Chi et al., 2014). By showing that mosaic ablation of innate immune cells by CRISPR ameliorate chordoma and the vertebral column phenotype we highlighted potential opportunities for early intervention in the treatment of chordomas and vertebral column fusions. Kita-RAS fish displayed adult IVDs abnormalities that resembled ageing zebrafish IVDD (unpublished data) with fibrotic NP and disorganised AF. Without parallel in zebrafish, we demonstrated that abnormalities in the early notochord cells and nucleus pulposus prime IVDD. Adult discs showed compromised notochord sheath, visualised by increased thickening of collagen fibres and fibre invasion towards the NP, hence a likely involvement of wound repair mechanisms in adult discs and IVDD. Indeed, human orthologues encoding collagen type IX and collagen type XI are expressed in the notochord sheath and have been associated with IVDD in populational studies (Feng et al., 2016), which supports the involvement of the notochord sheath in IVDD in zebrafish. The inflammatory processes exacerbated by cytokines D is ea se M o de ls & M ec ha ni sm s • D M M • A cc ep te d m an us cr ip t TNF-α and IL-1β are key events in IVDD (Risbud and Shapiro, 2014), they contribute to IVDD through degradation of extracellular matrix, likewise they are implicated in wounds and cancer. NP fibrosis during degeneration mimics wounds and fibrosis in other tissues (Novais et al., 2020a). Kita-RAS also developed bone quality impairment, emphasising nucleus pulposus modifications in regulation of bone homeostasis, suggesting changes in bone metabolic markers during chordomas. We detected increased osteoclast activity and chaotic osteoblasts at the endplates, in addition to osteoblast behaviour abnormalities and abnormal bone homeostasis. Osteoclasts share a common cell lineage with macrophages, and transdifferentiation of macrophages to osteoclasts has been reported (Pereira et al., 2018), suggesting opportunities to treat the bone phenotype through modulation of inflammation. In conclusion, using zebrafish we raised equivalences between chordomas, IVDD and wound repair, highlighting inflammation as a common event for potential therapeutic intervention. Material and Methods Zebrafish husbandry and lines Zebrafish were housed as described (Westerfield, 2000). Transgenic lines included: Tg(kita:Gal4;UAS:mCherry;UAS:HRASG12V-GFP) (Feng et al. 2010; Santoriello et al. 2010) and Tg(kita:Gal4;UAS:mCherry;UAS:mCherry-HRASG12V)(van den Berg et al., 2019) were incrossed to obtain Tg(kita:Gal4;UAS:HRASG12V-GFP) and Tg(kita:Gal4;UAS:mCherryHRASG12V), here referred as “Kita-RAS”, and Tg(kita:Gal4;UAS:mCherry) as controls. Tg(lyz:DsRed) (Hall et al. 2007); Tg(mpeg:FRET) (a gift from Stephen Renshaw at the University of Sheffield); Tg(col9a2:GFPCaaX)(Garcia et al., 2017); Tg(wt1b:GFP) (Perner et al., 2007); Tg(entpd5:kaeda)(Huitema et al., 2012); Tg(osx:NTR-mCherry)(Singh et al., 2012). Animal experiments were ethically approved by the University of Bristol Animal Welfare and Ethical Review Body (AWERB) and conducted under UK Home Office project licence. D is ea se M o de ls & M ec ha ni sm s • D M M • A cc ep te d m an us cr ip t Cellular proliferation assay Cellular proliferation was quantified using the Click-iT Plus EdU Alexa Fluor 647 Imaging Kit (Life Technologies, C10640). Larvae were immersed in Danieau’s solution containing 100 μM EdU solution and were incubated for 24 h or 48 h at 28.5 ̊C before termination of the experiment at 5 days post-fertilisation (dpf). Larvae were then fixed in 4% paraformaldehyde (PFA) for 2 h at room temperature with gentle shaking, washed with PBS solution containing 0.5% Triton X-100 (PBST) and 3% (w/v) Bovine Serum Albumin (BSA), and permeabilised in PBST solution containing 1% DMSO for 1 h at room temperature. For EdU detection, larvae were washed in PBST 3% BSA and incubated with the Click-iT Plus reaction cocktail containing Alexa Fluor-647 azide for 30 minutes at room temperature, in accordance with the manufacturers protocol. For quantification, EdU positive cells within and in proximity of the notochord were counted manually through the z stacks from confocal images and similar areas or interest. Confocal imaging Live zebrafish were mounted ventrally on coverslips in 1% low-melting point agarose containing MS222 (for live samples) and imaged using a Leica TCS SP8 AOBS confocal laser scanning microscope attached to a Leica DMi8 inverted epifluorescence microscope using 10x dry lens or 20x glycerol lens. The temperature in the chamber covering the microscope was maintained at 28 ̊C. Movies were recorded at an interval time of 5.45 min or 3.75 min per frame and a total time of 60 min or 120 min for neutrophils and macrophages, respectively. Confocal post-image analysis Image processing was performed using Fiji (Schneider et al., 2012). 1Analysis of number and time of neutrophil/macrophage interactions with notochord sheath: neutrophils and macrophages were considered to be interacting with the notochord sheath when they were in D is ea se M o de ls & M ec ha ni sm s • D M M • A cc ep te d m an us cr ip t direct surface contact with the sheath layer. The number of these interactions and their duration were manually quantified from time-lapse movies in a pre-defined region of the flank above the caudal hematopoietic tissue in the zebrafish larva, from the total field of view. Neutrophils, macrophages and notochord were identified by visualisation of their fluorescence in the fluorescent channel while the notochord sheath was more accurately distinguished by visualisation in the brightfield channel. Movies were exported from Fiji as QuickTime movies to play at 3 frames per sec. 2Analysis of osteoblasts: images were converted to 32-bit, applied LUT (16 colours), flattened and then saved as tiff images. The tiff files were imported to Fiji, two consecutive vertebrae were selected using the freehand selection tool, from which the mean pixel intensity values were calculated. 3Analysis of the area of notochord sheath cells (col9a2+): Kita-RAS notochord was divided in wound-proximal and wound-distal regions. Using the freehand selection tool in Fiji, the area of 10 cells were analysed per region, using 10 fish for controls and Kita-RAS. Morpholino (MO) injections Previously described morpholinos including pu.1 MO (5′GATATACTGATACTCCATTGGTGGT-3′) (0.2 mM) (Rhodes et al., 2005), gcsfr MO (5′GAAGCACAAGCGAGACGGATGCCAT-3′) (0.3 mM) (Liongue et al., 2009) and a scrambled MO (5’CCTCTTACCTCAGTTACAATTTATA3’) (0.5 mM) (GeneTools LLC, USA) were injected into 1-cell stage embryos, as previously described (Liongue et al., 2009, Rhodes et al., 2005, van den Berg et al., 2019). CRISPR/Cas9 injections We used three synthetic gRNAs targeting each of the genes, pu.1(spi1b) and csf3r(gcsfr), ordered as crRNAs (Sigma). We used the same target sites for gcsfr as previously described (Yang et al., 2020), while for pu.1we targeted the same genomic region as previously described in a pu.1 mutant (chr7:32655153-32655197) (Yang et al., 2020). Pu.1 target sequences: pu.1 D is ea se M o de ls & M ec ha ni sm s • D M M • A cc ep te d m an us cr ip t cr1 GAGGGATGTGATGGCTACCC, pu.1 cr2 AGCTCTGTAAAGTGGCTCTC and pu.1 cr3 GCCTGGGTCCATGAAATGGC). All six crRNAs (2pg) were incubated with tracrRNA (10pg) and GeneArt Platinum Cas9 nuclease (Invitrogen) prior to injections. Injections were performed into 1-cell stage embryos as previously described (Brunt et al., 2017). To validate CRISPR efficiency, DNA was extracted from 12 individual injected larvae at 5dpf, followed by PCR amplification with FAM-M13F primer and gene-specific primers, with each forward primer containing an M13 tail (pu.1 F: TGTAAAACGACGGCCAGTCCGTGTCTAGATCACTCTTGGG; pu.1 R: AAACCAAACCATAAATGATTCGTTTT; csf3r F: TGTAAAACGACGGCCAGTGATTGCTGACGTAACTATTGTAC; csf3r R: CTCACATTTAAAGTCTTATCAG). PCRs were submitted to fragment length analysis (ABI 3500)(Carrington et al., 2015). Controls were injected with Cas9 protein and SygRNA® SpCas9 tracrRNA (10pg) (Merck). Images of the notochord were acquired at 5dpf using a Leica fluorescent stereomicroscopy (MZ10F), followed by analysis of notochord lesions. Analysis of notochord lesions Notochord images of 5dpf larvae previously injected with MO or CRISPR were analysed using custom Python scripts and by implementing three steps. First, we detected pixels of the notochord through manually setting the value of the intensity threshold. Second, we fit the pixels with a 6th order polynomial function to obtain the intensity profile along the notochord. Specifically, the intensity profile was measured along the polynomial fit inside the image, using the algorithm adapted from the scikit-image package (van der Walt et al., 2014), where we modified the function “profile line” to work with a polynomial line. The average value of the intensity profile was used as a measurement of the severity of lesions within the notochord. Finally, the average intensity from the notochord was compared among different groups. For D is ea se M o de ls & M ec ha ni sm s • D M M • A cc ep te d m an us cr ip t statistical analysis we used ANOVA and Kruskal-Wallis H-test, implemented in scipy (Virtanen et al., 2020). Dunn’s method was used for multiple comparison test, implemented in scikit-posthocs (Terpilowski, 2019), p values were adjusted with Bonferroni. Alizarin Red and Calcein staining Alizarin Red S staining was performed in fixed fish to label calcified tissues and carried out using standard protocols (Walker and Kimmel, 2007). Live Calcein or Alizarin Red S staining was carried out as previously described (Bensimon-Brito et al., 2016). 14dpf fish were fixed in 4%PFA and undergone Alizarin Red staining. Pictures of the entire fish were taken under a Leica stereomicroscope. Total fish length and the length of the first seven vertebral segments were measured using Leica LAS X Software. Vertebral column severity scoring system Alizarin Red S staining was performed in fixed samples of 1 month old fish (1mpf) (Kita control n= 47 ; Kita-RAS n = 61; and Kita-RAS +CRISPR n= 97), and pictures taken with a Leica stereomicroscope (MZ10F). The length of each fish was measured from the nose to the most posterior extremity of the vertebral column, the tail fin was not included in the measurement. Those fish in which the vertebral columns were not completely formed were excluded from our severity score analysis. The vertebral column severity scoring system was based on numbers of fusions and clefts identified in each fish. Fusions and clefts were scored independently. Score of 3: n ≥ 5; score of 2: 3<n<5; score of 1: n≤3; score of 0: n=0 (n= number of fusions and clefts).


Introduction
The vertebral column is the central axis of the skeleton in all vertebrates. It is composed of segments (vertebrae) connected by joint-like structures called intervertebral discs (IVDs).
In mammals, the architecture of the IVDs is made by an annulus fibrosus (AF), a collagenous layer surrounding a hydrated and gelatinous nucleus pulposus (NP) core, which contains chondrocyte-like cells derived from embryonic notochord cells (Rodrigues-Pinto et al., 2014).
The disappearance of notochordal cells in mammals during development of the vertebral column has been linked to reduction of repair capacity and IVD degeneration (IVDD) (Wang et al., 2017). Occasionally, notochordal cells can cause vertebral malformations and in rare cases cell transformation lead to chordomas (Salisbury, 1993, Choi et al., 2008, a rare bone cancer of the axial skeleton and skull base (McMaster et al., 2011).
With an incidence of approximately one in a million, chordomas account for about 1-4% percent of all primary bone malignancies and 20% of primary spinal tumours (Chugh et al., 2007). Chordomas are slow growing and highly resistant to both chemotherapy and radiotherapy, meaning that radical surgery is often the primary choice for treatment modality (McMaster et al., 2011). Unfortunately, in many cases, the proximity of chordomas to vital structures, means that local excision is rarely achieved, resulting in a recurrence rate greater than 50% (Stacchiotti et al., 2017, Barry et al., 2011. Distant metastases to lung, bone, soft tissue, lymph node, liver, and skin have been reported in up to 43% of cases (Stacchiotti et al., 2017, Barry et al., 2011. Interestingly, chordomas lead to changes in bone quality, and often appear on X-rays and computerised tomography (CT) as eroding bone lesions with associated soft tissue calcification (de Bruine and Kroon, 1988), suggesting modifications in the behaviour of the nucleus pulposus cells disrupt disc and bone homeostasis. Impairment of disc homeostasis is a hallmark of IVDD (Novais et al., 2020b), which unlike chordomas is very common, representing the most common cause of back pain (Zheng and Chen, 2015), a

Disease Models & Mechanisms • DMM • Accepted manuscript
symptom that 80% of the adult world population suffer from (GBD et al., 2017). How transformed nucleus pulposus cells affect the IVD and surrounding vertebrae during their development is currently unknown; and no animal models to show how transformed cells dynamically interact with and affect the IVDs and vertebral column in vivo have been described. Such models could contribute to our understanding of chordoma development, IVDD, the interaction of the nucleus pulposus with the skeletal tissues and feature possible therapeutic avenues for both conditions.
Zebrafish have emerged as an advantageous animal model for a variety of human diseases, including cancer and skeletal diseases, due to their fast development, tractability, flexible genetic manipulation (transgenesis, forward and reverse genetics) and their translucency (Bergen et al., 2019). Reporter lines allow in vivo assessment of cell behaviour not only during early development but also during the later stages of skeletal formation in juveniles (Bergen et al., 2019). Zebrafish have high tissue regenerative capacity, with the ability to restore vacuolated cells of the notochord upon injury (Garcia et al., 2017). In zebrafish, notochord cells remain throughout life; they are enveloped by a sheath layer that acts as a sealing basement membrane to isolate the inner notochord vacuolated cells, and carries high potential to mineralise (Fleming et al., 2004, Stemple, 2005. The notochord sheath plays an important role in the segmentation of the vertebral column and centra primordium (chordacentra) formation (Lleras Forero et al., 2018, Pogoda et al., 2018, Wopat et al., 2018.
Following genetic manipulation, mechanical injury (needle punctures) or chemical treatment (with nystatin), repair of tissue damage appears to involve a sub-population of notochord sheath cells which become activated, expressing Wilms Tumor 1 (wt1b), and migrate towards the wound, setting landmarks during notochord repair (Garcia et al., 2017, Lopez-Baez et al., 2018.

Disease Models & Mechanisms • DMM • Accepted manuscript
Chordoma onset has been described in larval zebrafish expressing the oncogene RAS in the notochord, using the bimodal Gal4/UAS system and activation of the oncogenic RTK/Ras pathway (Burger et al., 2014). These zebrafish chordoma models become affected within the first 3 days post-fertilisation (dpf), progressively developing notochord hyperplasia , similar to histological features of human chordomas (Burger et al., 2014). Recently, the zebrafish chordoma model was used to test genetic potential to transform the notochord in vivo, providing suggestive evidence that Brachyury (TBXT), a highly expressed gene in human chordomas (Vujovic et al., 2006), is insufficient to initiate chordomas, instead suggesting activation of members of the RTK signalling as potential players in chordoma formation (D'Agati et al., 2019). The behaviour of notochord cancer cells during zebrafish life has not yet been studied, due to early lethality of chordoma models during larval stages. It is unknown whether notochord cancer cells trigger a wound repair mechanism similar to those of notochord injury models, which activate an acute inflammatory response as is seen in other early cancers (Feng et al., 2012, Feng et al., 2010. It is also unclear whether notochord cancer cells exert control as notochordal remnants to interfere with bone formation and, later in life, with bone homeostasis.
Here we studied the interactions between the notochord cancer cells within the forming vertebral column and bone homeostasis using a well-characterised transgenic line, Kita-RAS, which drives expression of HRASV12 in the notochord (and in melanoblasts, thus modelling melanoma) and survives to adulthood , van den Berg et al., 2019, Feng et al., 2012. We showed that "transformed" notochord cells destabilise the notochord sheath layer, activating a chronic wound repair response similar as those caused by induced notochord wounds previously described (Garcia et al., 2017, Lopez-Baez et al., 2018. These preneoplastic cells lead to invagination of the col9 expressing notochord sheath cells towards the wound and participation of wt1b notochord sheath sub-population. Interestingly, macrophages

Disease Models & Mechanisms • DMM • Accepted manuscript
and neutrophils were present in higher numbers and showed prolonged interaction with the wounded notochord sheath layer, as described in other cancers. The metameric pattern of segmentation of the vertebral column was compromised, but not abrogated, leading to vertebral fusions and clefts. Adult bone homeostasis was altered as observed by differences in vertebral bone mineral density and collagen fibre distribution. Transformed cells also compromised the adult zebrafish equivalent intervertebral disc architecture, leading to NP "scar" tissue, NP cellular disorganisation and affecting the structure of the AF, similar to IVDD. Chordoma development and skeletal defects were rescued when we partially depleted neutrophils and macrophages. In conclusion, our results indicate that transformed notochord cells cause chronic wounds leading to inflammation, vertebral abnormalities, disc and bone homeostasis impairment. Chordoma development could be controlled by limiting inflammation, revealing new avenues for therapeutics and highlighting the use of zebrafish as an animal model.

Kita-RAS induces wound-like destabilisation of the notochord
Notochord-specific Gal4 lines crossed to UAS:EGFP-HRASV12 have been previously described as powerful models for inducing chordomas in zebrafish (Burger et al., 2014). A transgenic line extensively used to induce melanoma, in which HRASV12 expression is driven by the Kita promoter in melanoblasts, goblet cells and in the notochord cells (due to the presence of an enhancer element for tiggy winkle hedgehog, twhh)  has the advantage over other notochord RAS expressing lines because it survives to adulthood , van den Berg et al., 2019. We used Kita-RAS-GFP and Kita-RAS-mCherry to study the progressive changes of the transformed notochord cells and their interaction with the forming vertebral column. In 5dpf zebrafish larvae, the outer layer of the notochord is formed by an epithelial-like sheath wrapping notochord vacuolated cells (Wopat et al., 2018) (Fig. 1A). Confocal images through the notochord, at 5dpf, showed that Kita drives reporter expression in the notochord vacuolated cells, but not in the sheath cells (Fig. 1B). As in other chordoma RAS models, Kita-RAS led to dramatic destabilisation of the notochord vacuolated cells starting as early as 3dpf and by 5dpf affected 70% of the larvae (>200 larvae analysed). Affected larvae were considered when they displayed more than 3 lesions in the notochord. Each lesion was characterised by increased RAS expression and abnormal notochord cell morphology (Fig. 1B). At the same developmental stage (5dpf), notochord cells were interspaced by infiltration of non-vacuolated cells and accumulation of fibrous collagenous tissue (AFOG staining, red colour) ( Fig. 1B-C). Furthermore, histological sections suggested local destabilisation of the notochord sheath layer at the region of collapsed vacuolated cells (Fig. 1C). To analyse cell proliferation, we treated larvae with EdU solution to be incorporated into the DNA of proliferating cells from 2 to 4dpf and followed by counting the number of EdU-positive (+) cells at 5dpf from confocal images. Notochord cells and notochord sheath cells in Kita-RAS are highly proliferative (p= 0.0002) ( Fig. 1D and E).
Interestingly, the organisation of the notochord in Kita-RAS fish displayed cellular characteristics reminiscent of those observed in notochord wounding models (needle puncture) (Lopez-Baez et al., 2018) (Fig. S1), suggesting that chordoma may recapitulate repair mechanisms, as has been suggested for several other cancers (Feng et al., 2010).

Pre-neoplastic notochord cells trigger the notochord wound repair mechanism in zebrafish
Wounds in the notochord induced by needle injury, amputation and chemical damage lead to the collapse of notochord vacuolated cells, sheath cell invasion and expression of Wilms Tumor 1b (wt1b) within a cell sub-population of the notochord sheath (Garcia et al., 2017, Lopez-Baez et al., 2018. To investigate whether pre-neoplastic notochord cells mimic a wound-like response, we crossed Kita-RAS-mCherry with Tg(col9a2:GFPCaaX), a marker for the notochord sheath layer ( Fig. 2A), and to Tg(wt1b:gfp) to label the sub-population of sheath Disease Models & Mechanisms • DMM • Accepted manuscript cells that standardly respond to damage. We confirmed that at 5dpf Kita is not expressed in the notochord sheath layer, and only in vacuolated notochord cells (Fig. 2B). We observed col9a2 expression in regions of damage within the notochord, suggesting sheath cell migration towards the chordoma wounded area (Fig. 2B). Cross sections through the notochord, at 5dpf, showed col9a2 expressing cells within the notochord in connection with the notochord sheath

Wounded notochord sheath elicits a prolonged recruitment of innate inflammatory cells
Several studies have reported that oncogene-transformed cells trigger an innate inflammatory response, with both neutrophils and macrophages recruited to the pre-cancerous tissue (Chia et al., 2018, Feng et al., 2010, Freisinger and Huttenlocher, 2014, Roh-Johnson et al., 2017. This recruitment of neutrophils and macrophages is responsible for clearing cell debris and to orchestrate tissue repair responses including wound angiogenesis and matrix deposition (Eming et al., 2017). We questioned whether oncogenic RAS expression in the notochord cells and the lesioned notochord sheath might also induce an inflammatory response in our zebrafish Disease Models & Mechanisms • DMM • Accepted manuscript chordoma model. During the first weeks of development, zebrafish do not have a functional adaptive immune system, allowing us to investigate the innate immune response on its own (Renshaw and Trede, 2012). We performed time-lapse imaging at 5dpf and analysed the interactions of neutrophils and macrophages with the notochord sheath layer. For neutrophils, we incrossed Tg(kita:Gal4; UAS:mCherry; UAS:HRASG12V-GFP;lyz:DsRed), and selected RAS-/lyz+ larvae, Tg(kita:mCherry;lyz:DsRed), and RAS+/lyz+ larvae, Tg(kita:HRASG12V-GFP;lyz:DsRed), as controls and Kita-RAS fish, respectively. While for macrophages, we incrossed Tg(kita:Gal4;UAS:mCherry;UAS:HRASG12V-GFP;mpeg:FRET), selected RAS-/mpeg+ larvae, Tg(kita:mCherry;mpeg:FRET) and RAS+/mpeg+ larvae, Tg(kita:HRASG12V-GFP;mpeg:FRET), as controls and Kita-RAS fish, respectively. Higher numbers of neutrophils and macrophages were recruited, making a prolonged direct contact with the wounded notochord sheath in Kita-RAS in comparison with controls ( Fig. 3, Fig. S2 and Movies 1 and 2), similarly to the inflammatory response previously reported in the melanoma model (Feng et al., 2010). Remarkably, we also found neutrophils and macrophages infiltrating wounded regions and in direct contact with notochord vacuolated cells ( Fig. S2 and Movies 1 and 2).
Together our results showed that zebrafish chordoma induces a chronic notochord inflammatory wound response with typical wound recruitment of neutrophils and macrophages. Inflammatory cells trespass the notochord sheath layer in wounded regions to form direct contact with transformed notochord cells, a similar behaviour described for other cancers (Feng et al., 2012).

Depletion of neutrophils and macrophages abolishes chordoma development
To further test whether the increased innate inflammatory response triggers the proliferation of neoplastic cells leading to wounds in the notochord, we transiently delayed innate immune cell zebrafish until at least 4dpf, therefore generating larvae lacking neutrophils and macrophages (Feng et al., 2012, Liongue et al., 2009, Rhodes et al., 2005. We confirmed the efficiency of our morpholino experiment by injecting fish carrying labelled neutrophils and macrophages at 3dpf (Tg(lyz:DsRed;mpeg:FRET))( suggesting that incomplete ablation of inflammatory cells can ameliorate chordoma. To complement our morpholino experiment, we used CRISPR/Cas9 system to target pu.1 and gcsfr simultaneously. We were able to cause mutations with an efficiency rate of 80%, validated by fragment length analysis, for each individual genes, at 5dpf. We analysed Kita-RAS larvae from morpholinos (MO) and CRISPR injections side-by-side at 5dpf (Fig. 4A).
CRISPR injections led to a significant reduction in numbers of neutrophils (p= 0.0012) and macrophages (p= 0.0478), but this reduction was not as pronounced as that observed from MO injections (p<0.0001) ( rescue the affected notochordal phenotype upon MO and CRISPR injections. We compared Kita (control), Kita-RAS and Kita-RAS injected with either MO or CRISPRs. Similar to our cell proliferation experiment, we detected a partial notochordal rescue with CRISPR injections and significant rescue with MO (Fig. 4H). Therefore, we have shown that the increase in neutrophils and macrophages contribute to proliferation of cancer cells in the notochord and modulation of inflammatory cells could prevent clonal expansion and chordoma development, similar to what has been previously shown for melanomas (Feng et al., 2012).

Abnormal pattern of vertebral segmentation and mineralisation in Kita-RAS fish
It has been demonstrated that notochord damage can lead to defective patterning of the vertebral column (Lopez-Baez et al., 2018, Fleming et al., 2004, Nguyen-Chi et al., 2014. Given that Kita-RAS cause cellular changes and a wound-like response in the notochord we questioned whether these events might have a downstream impact in the vertebral column segmentation. We crossed Kita-RAS to Tg(entpd5:kaeda), an early marker of the notochord segmentation and biomineralizing activity. Entpd5 hydrolyses nucleoside triphosphates, providing local inorganic monophosphate for biomineralization (Dallas andBonewald, 2010, Huitema et al., 2012). During development of the vertebral column, entpd5 is expressed in alternating segments of the sheath, which will form the mineralised chordacentra; while the interdomains will develop into intervertebral discs (IVDs) ( Fig. 5A and D) (Wopat et al., 2018).
We analysed larvae at 8dpf, at a stage when segmentation has started but is not yet finalised.

Transformed notochord cells lead to vertebral column fusions and clefts
Next, we sought to investigate the impact of pre-neoplastic cells in the vertebral column architecture. For that, we analysed the adult vertebral column, looking for resulting bone abnormalities. We used Alizarin Red staining (controls n = 10; Kita-RAS n = 10; 6 months post-fertilisation -6mpf), X-rays (controls n = 40; Kita-RAS n = 78; 1 year old fish) and microcomputerised tomography (μCT) (controls n = 5; Kita-RAS n = 5; 6mpf) to compare Kita-RAS with control fish of the same age. Vertebrae fusions were found in 100% of Kita-RAS and in 0% of controls (controls n = 40; Kita-RAS n = 78) ( Fig. 6 and Fig. S4). Fusions involved two or more vertebrae along the vertebral column leading to shortening of the total fish length.

Disease Models & Mechanisms • DMM • Accepted manuscript
Those fish with most fusions had the most reduced lengths (Fig. 6A, E and Fig. S4). The ribs were the most severely affected region of the vertebral column. We calculated the length of six consecutive mineralised segments of the vertebral column, separated by well-defined IVDs Next, we asked whether reduction of inflammatory cells could rescue the bone phenotype. We looked at the vertebral column of controls and Kita-RAS + CRISPR (pu.1 + gcsfr) fish at 1mpf by Alizarin Red staining. The severity of the vertebral column phenotype was scored depending on the number of fusions and clefts observed. Kita-RAS + CRISPR (pu.1 + gcsfr) partially rescue the vertebral column phenotype ( Fig. 6J and K), with a subset of fish showing no fusions or clefts (Fig. 6K). Therefore, modulation of innate immune cells in our chordoma model prevents vertebral fusions and clefts.

Compromised intervertebral discs and impaired bone quality in adult Kita-RAS
Embryonic notochordal cells contribute to the formation of the intervertebral disc nucleus pulposus (NP), which plays an important role in regulating disc homeostasis (Choi et al., 2008).
We sought to understand the impact of transformed notochord cells in the adult zebrafish intervertebral disc equivalent regions and vertebral bone. By calculating bone mineral density, we detected a significant TMD decrease in Kita-RAS (p= 0.0015) ( Fig. 6A and B), indicative of impaired bone quality. We performed histological sections of the adult vertebral column and observed highly fibrotic NP, similar to IVD degeneration (IVDD), with disorganised cellularity found in enlarged vertebrae ( Fig. 7A and B). Fibrosis was detected in proximity with the notochord sheath layer. AFOG and Picro-sirius red staining confirmed fibrosis and connectivity with the notochord sheath, showing increased collagen content and increased collagen fibre thickness ( Fig. 7B and C). In contrast to IVDD, dehydration did not describe the phenotype of Kita-RAS NP, as an increase in glycosaminoglycans was detected (Fig. S5).
Additionally, despite fibrosis and disorganisation of the NP, due to cell transformation, we did not observe intervertebral disc calcification, a feature commonly found during IVDD and ageing (Novais et al., 2020a). The outermost component of the discs, the annulus fibrosus (AF), was replaced by bone in IVDs that were compromised by fusions. The structured layers of Disease Models & Mechanisms • DMM • Accepted manuscript collagen and elastin that form the zebrafish AF were completely lost in some of the IVDs (Fig.   7D). Interestingly, disorganised and increased number of osteoblasts were detected in the IVD region, corroborating altered osteoblast activity at the endplates of adult fish. The balance between osteoblasts and osteoclasts is key in bone homeostasis and control of bone density.
Moreover, osteoclasts are derived from the same cell lineage of macrophages. We performed whole-mount TRAP staining to visualise osteoclast activity. Quantification of TRAP staining revealed exacerbated bone resorption in Kita-RAS (p= 0.0026), especially in affected areas of the vertebral column ( Fig. S5B and C). Picro-sirius red staining suggested a reduction in collagen fibre thickness in the bone (centra). We quantified the mean intensity of red, green and blue pixels from pictures stained with Picro-sirius red. We detected a significant reduction in red (p= 0.0004) and blue (p= 0.0016) pixels, indicating an abnormal fibre organisation and confirming bone quality impairment in Kita-RAS (Fig. 7C). We conclude that transformed cells in the notochord lead to vertebral column and intervertebral disc abnormalities affecting the NP and AF, impairing osteoblasts and osteoclasts activity, consequently altering bone homeostasis in zebrafish.

Discussion
"Tumours are wounds that do not heal" was postulated in a classic work published by Harold Dvorak in 1986(Dvorak, 1986. Dvorak recognized that the composition of the tumour stroma strongly resembled healing skin wounds, suggesting activation of the wound-healing response in the host. Moreover, cancer is frequently the consequence of chronic inflammatory disease (Schafer and Werner, 2008). Given the confined nature of notochordal cells during Although human chordomas are thought to originate from hyperplasia of notochordal remnants, benign notochordal remnants are occasionally found and are associated with vertebral abnormalities, such as vertebral clefts and bifurcations (Oner et al., 2006). When we looked at the adult Kita-RAS we observed vertebral clefts and hemivertebra that recapitulate human vertebral column abnormalities. However, vertebral malformations might not be a direct effect from pre-neoplastic notochordal cells, but a result from abnormal notochordal cell behaviour. Recent studies have shown that notochord vacuoles function as a hydrostatic scaffold that guides symmetrical growth of vertebrae and spine formation. Vacuole fragmentation caused by mutations in dstyk (spzl mutant) resulted in vertebral centra malformation and scoliosis (Bagwell et al., 2020, Sun et al., 2020. Similar to our observations, these studies evidenced that abnormal behaviour of notochord vacuolated cells are associated with vertebral malformations like to those of notochordal remnants in human. Furthermore, hemivertebra and clefts were systematically found in another mutant, spondo, carrying a mutation in cmn (Calimmin, a teleost-specific extracellular matrix protein with weak similarity to Elastin, and expressed in the notochord sheath), due to abnormalities in the notochord sheath layer . Here, we demonstrated that destabilisation of the notochord vacuolated cells also triggered cellular changes in the notochord sheath layer (Fig. 8). Hence, revealing double and overlapping routes in which notochord neoplastic cells compromise the formation of the vertebral column: the inner vacuolated cells and the outer notochord sheath cells.

Disease Models & Mechanisms • DMM • Accepted manuscript
Notochord damage also leads to vertebral column abnormalities, including fusions and segmentation mispatterning (Lleras Forero et al., 2018, Wopat et al., 2018, Pogoda et al., 2018. We showed that Kita-RAS mimicked notochordal damages and induced repair mechanism as demonstrated by activation and invagination of col9+ notochord sheath cells and expression of wt1b in wounded areas, as previously described for notochordal wounds (Lopez-Baez et al., 2018, Garcia et al., 2017. Our findings suggest a key role of the notochord sheath and wound repair in chordoma. Interestingly, when RAS is activated in the notochord sheath specifically with col2a1a driving RAS, it also causes chordomas (D'Agati et al., 2019), sustaining a key role of the sheath layer in zebrafish chordomas. As neoplastic cells are continuously modifying the notochord, this causes wounds that seem to progress and remain chronic or unresolved. We showed for the first time that wounding provoked by transformed notochord cells triggers the recruitment of neutrophils and macrophages. Innate immune cells not only were present in higher number but changed their behaviour by prolonging their interaction time with the notochord sheath in wounded regions; in some cases they were able to breach the sealing membrane and achieve direct contact with cancer cells. It has been recently described that inflammatory cells make use of pre-existing holes in the basement membrane to gain access and reach pre-neoplastic cells in a melanoma model (van den Berg et al., 2019). In our chordoma model, inflammatory cells were observed in direct contact with pre-neoplastic cells in regions of severe notochord sheath wounds, which similarly, may serve as breaches in the notochord sheath to allow neutrophils and macrophages to reach pre-neoplastic cells. The interaction between neutrophils/macrophages and transformed cells have been beautifully described for melanoma in zebrafish, with formation of cytoplasmic tether linking the two cell types and engulfment of transformed cells by neutrophils and macrophages (Feng et al., 2010).
H2O2, a key damage signal directing recruitment of neutrophils to a wound, was also identified as the major component drawing recruitment of leukocytes to the transformed cells (Feng et Disease Models & Mechanisms • DMM • Accepted manuscript al., 2010). Remarkably, when we depleted innate immune cells using morpholinos or CRISPR, we could rescue the notochord phenotype by inhibiting the aberrant proliferation of transformed cells, as demonstrated for melanoma (Feng et al., 2010), and partially rescuing the skeletal phenotype using CRISPR, by showing reduction of vertebral fusions. Thus, highlighting parallels between cancer and wound, and suggesting that immunomodulation might be a promising treatment for chordomas. When zebrafish notochord is infected with Kita-RAS fish displayed adult IVDs abnormalities that resembled ageing zebrafish IVDD (unpublished data) with fibrotic NP and disorganised AF. Without parallel in zebrafish, we demonstrated that abnormalities in the early notochord cells and nucleus pulposus prime IVDD. Adult discs showed compromised notochord sheath, visualised by increased thickening of collagen fibres and fibre invasion towards the NP, hence a likely involvement of wound repair mechanisms in adult discs and IVDD. Indeed, human orthologues encoding collagen type IX and collagen type XI are expressed in the notochord sheath and have been associated with IVDD in populational studies (Feng et al., 2016), which supports the involvement of the notochord sheath in IVDD in zebrafish. The inflammatory processes exacerbated by cytokines

Disease Models & Mechanisms • DMM • Accepted manuscript
TNF-α and IL-1β are key events in IVDD (Risbud and Shapiro, 2014), they contribute to IVDD through degradation of extracellular matrix, likewise they are implicated in wounds and cancer.
NP fibrosis during degeneration mimics wounds and fibrosis in other tissues (Novais et al., 2020a). Kita-RAS also developed bone quality impairment, emphasising nucleus pulposus modifications in regulation of bone homeostasis, suggesting changes in bone metabolic markers during chordomas. We detected increased osteoclast activity and chaotic osteoblasts at the endplates, in addition to osteoblast behaviour abnormalities and abnormal bone homeostasis. Osteoclasts share a common cell lineage with macrophages, and transdifferentiation of macrophages to osteoclasts has been reported (Pereira et al., 2018), suggesting opportunities to treat the bone phenotype through modulation of inflammation. In conclusion, using zebrafish we raised equivalences between chordomas, IVDD and wound repair, highlighting inflammation as a common event for potential therapeutic intervention.

Confocal imaging
Live zebrafish were mounted ventrally on coverslips in 1% low-melting point agarose containing MS222 (for live samples) and imaged using a Leica TCS SP8 AOBS confocal laser scanning microscope attached to a Leica DMi8 inverted epifluorescence microscope using 10x dry lens or 20x glycerol lens. The temperature in the chamber covering the microscope was maintained at 28˚C. Movies were recorded at an interval time of 5.45 min or 3.75 min per frame and a total time of 60 min or 120 min for neutrophils and macrophages, respectively.

Confocal post-image analysis
Image processing was performed using Fiji (Schneider et al., 2012). 1-Analysis of number and time of neutrophil/macrophage interactions with notochord sheath: neutrophils and macrophages were considered to be interacting with the notochord sheath when they were in Disease Models & Mechanisms • DMM • Accepted manuscript direct surface contact with the sheath layer. The number of these interactions and their duration were manually quantified from time-lapse movies in a pre-defined region of the flank above the caudal hematopoietic tissue in the zebrafish larva, from the total field of view. Neutrophils, macrophages and notochord were identified by visualisation of their fluorescence in the fluorescent channel while the notochord sheath was more accurately distinguished by visualisation in the brightfield channel. Movies were exported from Fiji as QuickTime movies to play at 3 frames per sec. 2-Analysis of osteoblasts: images were converted to 32-bit, applied LUT (16 colours), flattened and then saved as tiff images. The tiff files were imported to Fiji, two consecutive vertebrae were selected using the freehand selection tool, from which the mean pixel intensity values were calculated. 3-Analysis of the area of notochord sheath cells (col9a2+): Kita-RAS notochord was divided in wound-proximal and wound-distal regions.
Using the freehand selection tool in Fiji, the area of 10 cells were analysed per region, using 10 fish for controls and Kita-RAS.

Analysis of notochord lesions
Notochord images of 5dpf larvae previously injected with MO or CRISPR were analysed using custom Python scripts and by implementing three steps. First, we detected pixels of the notochord through manually setting the value of the intensity threshold. Second, we fit the pixels with a 6th order polynomial function to obtain the intensity profile along the notochord.
Specifically, the intensity profile was measured along the polynomial fit inside the image, using the algorithm adapted from the scikit-image package (van der Walt et al., 2014), where we modified the function "profile line" to work with a polynomial line. The average value of the intensity profile was used as a measurement of the severity of lesions within the notochord.
Finally, the average intensity from the notochord was compared among different groups. For

Disease Models & Mechanisms • DMM • Accepted manuscript
statistical analysis we used ANOVA and Kruskal-Wallis H-test, implemented in scipy (Virtanen et al., 2020). Dunn's method was used for multiple comparison test, implemented in scikit-posthocs (Terpilowski, 2019), p values were adjusted with Bonferroni.

Alizarin Red and Calcein staining
Alizarin Red S staining was performed in fixed fish to label calcified tissues and carried out using standard protocols (Walker and Kimmel, 2007). Live Calcein or Alizarin Red S staining was carried out as previously described (Bensimon-Brito et al., 2016). 14dpf fish were fixed in 4%PFA and undergone Alizarin Red staining. Pictures of the entire fish were taken under a Leica stereomicroscope. Total fish length and the length of the first seven vertebral segments were measured using Leica LAS X Software.

Vertebral column severity scoring system
Alizarin Red S staining was performed in fixed samples of 1 month old fish (1mpf)

Disease Models & Mechanisms • DMM • Accepted manuscript
Fish lengths were measured using Fiji (Schindelin et al., 2012) (in pixels), using images that were acquired under the same conditions.

Micro-computed tomography (µCT)
Six month old fish (6mpf) were fixed in 4% PFA for 14 days, followed by sequential washes in ethanol and maintained in a 70% ethanol solution. Micro-computed tomography (µCT) was performed using a Nikon X-TEK 225 HT CT scanner under an X-ray source of 130 kV, 53 µA without additional filters. Whole fish were scanned at voxel size of 20 μm, and selected spine regions rescanned at 5 μm. Images were reconstructed using CT Pro 3D software (Nikon).
Amira 6.0 (FEI) was used to generate 3D volume and surface renders for image acquisition.
For calculations of tissue mineral density (TMD), defined as measurement restricted to within the volume of calcified bone tissue (Bouxsein et al., 2010), the centrae were segmented and the mean grey values retrieved. Grey values were calibrated with phantoms of known densities (0.25 and 0.75 g.cm 3 of CaHA), and used for density calculations, as previously described . Three fish from each group were used for TMD calculation.

Histology
Adult fish (3mpf, control, n = 3; Kita-RAS, n = 3) were fixed in 4% PFA for 14 days, then decalcified in 1M EDTA solution for 20 days at room temperature. Larvae (control, n = 3; Kita-RAS, n = 4) were fixed for 2 hours. Samples were dehydrated in ethanol, embedded in paraffin and sagittal sections were taken at 8 µm thickness. Selected slides were de-waxed and stained with Toluidine Blue (Kague et al., 2018), Alcian Blue, AFOG or Picro-sirius red, as performed elsewhere (Hayes et al., 2013). Images were acquired on a Leica DMI600 inverted microscope, using 20X and 40X oil objectives, LAS software and a DFC420C colour camera.
Quantification of thickness of collagen fibre was performed using Fiji (Schindelin et al., 2012),

Disease Models & Mechanisms • DMM • Accepted manuscript
by selecting an area of interest within the bone, followed by measurement of mean intensity of red, blue and green pixels.

Tartrate-resistant acid phosphatase (TRAP) staining
TRAP staining was performed in whole-mount 3 month old fish (3mpf) (control, n = 4; Kita-RAS, n = 5) using Acid Phosphatase, Leukocyte (TRAP) kit (Merck, cat 387A) and following the instructions provided by the manufacture. Fish were fixed overnight in fixative solution (provided). Samples were washed for 15 min in distilled water, followed by permeabilization using 1% trypsin in 30% borate solution at 37˚C overnight. Fish were incubated in TRAP staining solution (provided) at 37˚C for 6h in the dark, followed by two washes of 10 min each in distilled water. Pigmentation was removed by incubating the specimens in 3% H2O2. Pictures were taken from dissected spines placed in 70% glycerol under a Leica stereomicroscope.
Quantification of TRAP signal was performed using Fiji (Schindelin et al., 2012). Images were converted to 32-bit, and LUT (physics) applied. We inverted the LUT, flattened the images and calculated the mean of red pixels, correspondent to high TRAP signal.

Statistical Analysis
GraphPad