Introduction

Idiopathic scoliosis (IS) is a 3D rotation of the spine without vertebral anomalies that affects about 3% of adolescents worldwide. Its etiology remained mysterious, mainly because of the lack of appropriate animal models and the complexity of its inheritance profile in large families presenting a severe scoliosis trait (Patten et al., 2015). Since the description of the first IS model in the zebrafish ptk7 gene mutant (Hayes et al., 2014), a number of mutants for genes encoding ciliary proteins were shown to develop scoliosis at late larval and juvenile stage (4 to 12 weeks post-fertilization, wpf) without any vertebral fusion or fracture, highlighting the link between cilia function and straight axis maintenance in that species (Grimes et al., 2016; Wang et al., 2022).

Cilia are microtubular organelles with sensory and/or motile functions. Zebrafish mutants in genes involved in cilia motility or in intraflagellar transport display severe embryonic axis curvature and are usually not viable beyond larval stages (Grimes et al., 2016; Rose et al., 2020). In contrast, mutants in genes implicated in ciliary gating or ciliary trafficking encoding respectively components of the transition zone (TZ) or of the BBsome complex survive to adult stage and develop scoliosis with variable penetrance (Bentley-Ford et al., 2022; Masek et al., 2022; Song et al., 2020; Wang et al., 2022). The molecular basis for these differences is still poorly understood, even if progress has been made in our understanding of zebrafish scoliosis. Cilia-driven movements of the cerebrospinal fluid (CSF) are involved in zebrafish axis straightness, both in embryos and juveniles (Boswell and Ciruna, 2017; Grimes et al., 2016) and are tightly linked to the assembly and maintenance of the Reissner fiber (RF), a SCO-spondin polymer secreted by the subcommissural organ (SCO) and running down the brain and spinal cord CSF-filled cavities (Cantaut-Belarif et al., 2018). RF loss at embryonic stage in null sspo mutants generates ventral curvatures of the posterior axis of the body and is lethal while its loss at juvenile stage in hypomorphic sspo mutants triggers scoliosis with full penetrance (Cantaut-Belarif et al., 2018; Rose et al., 2020; Troutwine et al., 2020). Signaling downstream of the RF in embryos implicates Urp1 and Urp2, two neuropeptides of the Urotensin 2 family expressed in ventral CSF-contacting neurons (CSF-cNs), which act on dorsal muscles in embryos and larvae (Cantaut-Belarif et al., 2020; Gaillard et al., 2023; Lu et al., 2020; Zhang et al., 2018). Their combined mutations or the mutation of their receptor gene uts2r3 lead to scoliosis at adult stages (Bearce et al., 2022; Gaillard et al., 2023; Zhang et al., 2018). However, whether RF maintenance and Urp1/2 signaling are perturbed in juvenile scoliotic mutants of genes encoding TZ proteins (“TZ mutants”) and how these perturbations are linked to scoliosis is unknown, especially because TZ mutants do not display any sign of embryonic or larval cilia motility defects (Wang et al., 2022). Moreover, two scoliotic models develop axial curvature while maintaining a RF showing that their curvature onset occurs independently of RF loss (Meyer-Miner et al., 2022; Xie et al., 2023). Finally, neuro-inflammation has been described in a small subset of zebrafish IS models for which anti-inflammatory/anti-oxidant treatments (with NAC or NACET) partially rescue scoliosis penetrance and severity (Rose et al., 2020; Van Gennip et al., 2018).

In this paper we dissect the mechanisms of scoliosis appearance in a novel deletion allele of the zebrafish TZ rpgrip1l gene. RPGRIP1L mutations have been found in neurodevelopmental ciliopathies, Meckel syndrome (MKS) and Joubert syndrome (JS) (Delous et al., 2007)(Arts et al., 2007) characterized by severe brain malformations. Other organs such as the kidney and liver are variably affected. Some JS patients also present with severe scoliosis, but this phenotypic trait is of variable penetrance (Delous et al., 2007)(Brancati et al., 2008). Functional studies performed in mouse and zebrafish have shown that Rpgrip1l is required for proper central nervous system patterning due to complex perturbations of Sonic Hedgehog signaling (in mice) and cilia planar positioning (in mice and zebrafish) (Andreu-Cervera et al., 2021; Mahuzier et al., 2012; Vierkotten et al., 2007). Zebrafish rpgrip1l mutants are viable and develop an axis curvature phenotype at juvenile stages with nearly full penetrance. Using a variety of approaches, we show that i) Rpgrip1l is required in ciliated cells surrounding ventricular cavities for maintaining a straight axis; ii) at juvenile stages, rpgrip1l mutants develop cilia defects associated with ventricular dilations and loss of the RF; iii) increased urp1/2 expression in rpgrip1l mutants does not account for spine curvature defects; and iv) a pervasive astrogliosis process associated with neuroinflammation participates in scoliosis onset and progression. Finally, we show that astrogliosis is conserved in another zebrafish ciliary TZ mutant, cep290, identifying a novel mechanism involved in scoliosis upon ciliary dysfunction.

Results

Rpgrip1l ^-/- fish develop scoliosis asynchronously at juvenile stage

To study the mechanisms of scoliosis appearance upon ciliary dysfunction in zebrafish, we produced a viable zebrafish deletion mutant in the rpgrip1l gene encoding a ciliary transition zone protein (rpgrip1l^Δ, identified in ZFIN as rpgrip1l^bps1) Supplementary Figure S1A). We made a 26 kb deletion by CRISPR-Cas9 double-cut, which removes 22 out of the 26 exons of the protein (Supplementary Figure S1A) and triggers protein truncation at amino-acid (aa) position 88 out of 1256. 90-100% of rpgrip1l^Δ/Δ embryos and larvae were straight (0-10%, depending on the clutch, displayed a sigmoid curvature). They did not display any additional defects found in many ciliary mutant embryos or larvae such as randomized left-right asymmetry, kidney cysts or retinal anomalies (Bachmann-Gagescu et al., 2011; Kramer-Zucker et al., 2005; Tsujikawa and Malicki, 2004)(Figure 1A and Supplementary Figure S1B, D, E). Another allele with a frameshift mutation in exon 4 (rpgrip1l^ex4, identified in ZFIN as rpgrip1l^bps2) showed a similar phenotype (Supplementary Figure S1A, B). Considering the large size of the deletion, we considered the rpgrip1l^Δ allele as a null and called it rpgrip1l^-/- thereafter. rpgrip1l^-/-animals developed scoliosis during juvenile stages. Scoliosis appearance was asynchronous between clutches, from 4 weeks post fertilization (wpf) (around 1 cm length) to 11 wpf (1.8-2.3 cm), and also within a clutch. It started with slight upward bending of the tail (the tail-up phenotype) and progressed toward severe curvature (Figure 1A, B) with 90% penetrance in adults (100% in females and 80% in males) (Figure 1C). Micro-computed tomography (μCT) at two different stages (5 wpf and 23 wpf) confirmed that spine curvature was three-dimensional and showed no evidence of vertebral fusion, malformation or fracture (Figure 1D-G and Supplementary Movie 1).

Figure 1:

Thus, the ciliary rpgrip1l^-/- mutant shows almost totally penetrant, late-onset juvenile scoliosis without vertebral anomalies, making it a valuable model for studying the etiology of human idiopathic scoliosis. Furthermore, the asynchronous onset of curvature is an asset for studying the chronology of early defects leading to scoliosis.

Reintroducing RPGRIP1L in foxj1a lineages suppresses scoliosis in rpgrip1l^-/- fish

Since rpgrip1l is ubiquitously expressed in zebrafish (Mahuzier et al., 2012), scoliosis could have different tissue origins, including a neurological origin as shown for the ptk7 and ktnb1 scoliotic fish (Meyer-Miner et al., 2022; Van Gennip et al., 2018). Indeed, no tissue-specific rescue has been performed in mutants for zebrafish ciliary TZ or BBS genes. We tested if introducing by transgenesis a tagged human RPGRIP1L protein under the control of the col2a1a (Dale and Topczewski, 2011) or foxj1a (Van Gennip et al., 2018) enhancers (Figure 1H) would reduce scoliosis penetrance and severity. The foxj1a enhancer drives expression in motile ciliated cells, among which ependymal cells lining CNS cavities (Van Gennip et al., 2018), while the col2a1a regulatory region drives expression in cartilage cells including intervertebral discs and semi-circular canals, tissues that have been proposed to be defective in mammalian scoliosis models (Hitier et al., 2015; Karner et al., 2015). After selecting F1 transgenic fish that expressed the Myc-tagged RPGRIP1L protein in the targeted tissue at 2.5 dpf, we introduced one copy of each transgene into the rpgrip1l^-/- background. The Foxj1a:5XMycRPGRIP1L transgene was sufficient to fully suppress scoliosis in rpgrip1l^-/- fish: none of the transgenic rpgrip1l^-/- fish became scoliotic while virtually all the non-transgenic rpgrip1l^-/- siblings did (Figure 1I, J). In contrast, RPGRIP1L expression triggered by the col2a1a promoter did not have any beneficial effect on scoliosis penetrance (Figure 1I, J). In conclusion, rpgrip1l expression in foxj1a-expressing cells is sufficient to maintain a straight axis.

Rpgrip1l^-/- adult fish display cilia defects along CNS cavities

We observed normal cilia in the neural tube of rpgrip1l^-/-embryos (Supplementary Figure S1C). In 3 month-adults, scanning electron microscopy (SEM) at the level of the rhombencephalic ventricle (RhV) showed cilia tufts of multiciliated ependymal cells (MCC) lining the ventricle in controls and straight rpgrip1l^-/- fish (Supplementary Figure S1F-G, F’-G’). In scoliotic rpgrip1l^-/-fish, cilia tufts were sparse and disorganized (Supplementary Figure S1H, H’). Cilia of mono-ciliated ependymal cells were present in rpgrip1l^-/- fish, some of them with irregular shape (Supplementary Figure S1, F”, G’’, H’’ and H’’’).

MCC defects at SCO level correlate with scoliosis appearance at juvenile stages

We then took advantage of the asynchrony in scoliosis onset to correlate, at a given stage, cilia defects at different levels of the CNS cavities with the spine curvature status (straight or curved) of rpgrip1l^-/- juveniles. We analyzed cilia defects in three CNS regions implicated in scoliosis appearance: the spinal cord central canal (since scoliosis develops in the spine), the forebrain choroid plexus (fChP) whose cilia defects have been associated with scoliosis in katnb1^-/- fish (Meyer-Miner et al., 2022) and the subcommissural organ (SCO), which secretes SCO-spondin whose absence leads to scoliosis (Rose et al., 2020; Troutwine et al., 2020).

In the spinal cord central canal of 8 wpf juvenile fish, cilia density was reduced to the same extent in rpgrip1l^-/- fish irrespective of the fish curvature status (Figure 2 A-C, E). In addition, cilia length increased in straight mutants and was more heterogenous than controls in scoliotic mutants (Figure 2 A’-C’, D). Arl13b content appeared severely reduced in mutants (Figure 2 A”-C”). In the fChP, mono- and multi-ciliated cell populations were observed with a specific distribution along the anterior-posterior axis, as previously described (D’Gama et al., 2021). At the level of the habenula nuclei, cells of dorsal and ventral midline territories were monociliated, while lateral cells presented multicilia tufts (Supplementary Figure S2 K-K’’). rpgrip1l^-/- juveniles presented an incomplete penetrance of ciliary defects in the fChP, with no correlation between cilia defects and the curvature status (Supplementary Figure S2, L-O’’).

Figure 2

The SCO lies at the dorsal midline of the diencephalic ventricle (DiV), under the posterior commissure (PC) (Figure 2 G, H, Supplementary Figure 2A, C). CiIia in and around the SCO have not been described previously in zebrafish. We showed that, in control animals, Sco-spondin (Sspo) secreting cells varied in number along SCO length, from 24 cells anteriorly to 6-8 cells posteriorly (Supplementary Figure 2 A, A’, C) and were monociliated at all anteroposterior levels (Figure 2G-H, Supplementary Figure 2 A’). The Sspo protein produced from one copy of the transgene Scospondin- GFP^ut24 (Troutwine et al., 2020) formed cytoplasmic granules around SCO cell nuclei, under emerging primary cilia (Supplementary Figure 2A’). Conversely, MCCs lateral to the SCO did not secrete Sspo and were prominent towards posterior (Figure 2H, Supplementary Figure 2C), as observed in the rat brain (Collins and Woollam, 1979). Monocilia of Sspo secreting cells were still present but appeared longer in straight rpgrip1l^-/- juveniles and could not be individualized anymore at anterior SCO level in scoliotic fish: we observed a weaker and denser acetylated and glutamylated tubulin staining that likely corresponded to longer intertwined cilia (Figure 2 I, I’, K, K’). Lateral multicilia tufts were preserved in straight rpgrip1l^-/- juveniles (n=5) (Figure 2 J) but were missing at posterior SCO level in tail-up or scoliotic rpgrip1l^-/- fish (n=7) (Figure 2L).

Thus, rpgrip1l mutants develop both monocilia and multicilia defects at juvenile stages, but cilia defects in the spinal cord and in the fChP do not correlate with axis curvature status of the fish. By contrast, the loss of MCC cilia at the SCO strictly correlates with scoliosis onset, suggesting a causal relationship.

Rpgrip1l^-/- juveniles show ventricular dilations and loss of the Reissner fiber at scoliosis onset

Ciliary beating is an essential actor of CSF flow and of ventricular development in zebrafish larvae (Kramer-Zucker et al., 2005) (Olstad et al., 2019). Thus, ciliary defects in the brain and spinal cord of rpgrip1l^-/-fish could lead to abnormal ventricular and central canal volume and content. In the spinal cord, the lumen of the central canal was enlarged at thoracic and lumbar levels in all rpgrip1l^-/-mutant juveniles, with no correlation to the axis curvature status (Figure 2 A-C; F). To determine the timing of brain ventricular dilations relative to scoliosis, we analyzed ventricular volume at the onset of spine curvature (Figure 3A-G). Ventricular reconstruction was performed on cleared brains of 4 control and 4 rpgrip1l^-/- (3 tail-up and 1 straight) fish at 5 wpf, stained with ZO1 to highlight the ventricular border and with DiI to outline brain shape, focusing on the RhV at the level of the posterior midbrain and hindbrain (Figure 3A-D). We identified a significant increase in ventricle volume in rpgrip1l^-/-fish compared to controls, that was restricted to the ventral regions of the RhV in regions ROI4.4 (green sphere in Figure 3C) and ROI6 (green sphere in Figure 3D), as confirmed by numerical sections (Figure 3E, G). Ventricle dilations were also present, although to a lesser extent, in the RhV of the unique straight rpgrip1l^-/- fish of this study (green squares and triangle in Figure 3F, G).

Figure 3:

Defective ependymal ciliogenesis has been associated with abnormal CSF flow and defective Reissner Fiber (RF) maintenance in several scoliotic zebrafish models (Grimes et al., 2016)(Rose et al., 2020). The RF runs along the length of the body, within brain ventricles and central canal (Figure 3H). It is mainly composed of Sspo secreted by the SCO and floor plate (FP) and its loss at juvenile stages triggers scoliosis in zebrafish (Rose et al., 2020)(Troutwine et al., 2020). To visualize the RF, we used an antiserum that labels the RF in zebrafish embryos (Didier et al., 1995)(Cantaut-Belarif et al., 2018) and we confirmed the results by introducing one copy of the sspo-GFP knock-in allele, sco-spondin-GFP^ut24, (Troutwine et al., 2020) by genetic cross into rpgrip1l^-/- animals. The RF formed normally in rpgrip1l^-/-embryos, as expected given the absence of embryonic curvature (Supplementary Figure S1 C). To study its maintenance in rpgrip1l^-/- juveniles, we immunostained spinal cord longitudinal sections of 7-8 wpf curved or straight fish of the same clutch. The RF was visualized as a 1 μm diameter rod in the central canal of the neural tube and few dots were present in the apical cytoplasm of FP cells (Figure 3I) as well as in the SCO of all controls (n=8, Figure 3L, O) and in ventricles caudal to the SCO, but not rostral to it as expected (Supplementary Figure S2E, H). The RF was also present in all rpgrip1l^-/- straight fish (n=9, Figure 3J, Supplementary Figure S2I). In contrast, in rpgrip1l^-/- tail-up and scoliotic fish, the RF was absent at all axis levels and a few SCO-spondin-positive debris were present in the central canal (n=8, Figure 3K). Transverse sections at the level of the SCO and fChP as well as caudal to the SCO in severely scoliotic fish revealed abnormally packed material in the ventricles (Figure 3N, Q, Supplementary Figure S2G, J). In juveniles with a slight tail-up phenotype, SCO-secreting cells produced Sspo-GFP cytoplasmic granules as in controls (Supplementary Figure S2B, B’, n=7/7), and polymers of Sspo-GFP accumulate in the adjacent DiV, which suggests that Sspo-GFP aggregates are produced by the SCO.

Thus, our study uncovers a strict temporal correlation between the loss of MCC at posterior SCO level, impaired RF polymerization and scoliosis onset in rpgrip1l^-/-juvenile fish.

Urp1/2 upregulation does not contribute to scoliosis penetrance or severity in rpgrip1l^-/- fish

In scoliotic fish where the RF is not maintained at juvenile stage, urp1 and urp2 levels are downregulated (Rose et al., 2020). urp1 and urp2 encode peptides of the Urotensin 2 family, expressed in ventral spinal CSF-cNs (Parmentier et al., 2011; Quan et al., 2015). Their expression levels have to be finely tuned to keep a straight axis since their combined loss-of-function induces larval kyphosis that evolves into adult scoliosis (Bearce et al., 2022; Gaillard et al., 2023), while an increased dose of Urp1/2 peptides induces an upward curvature in embryos (Zhang et al., 2018) and juveniles (Gaillard et al., 2023; Lu et al., 2020). We therefore monitored via qRT-PCR urp1 and urp2 expression levels on individual straight or scoliotic rpgrip1l^-/- juvenile and/or adult fish. urp2 was the most highly expressed gene among all urotensin 2 members at juvenile stage (Figure 4A, D, E and Supplementary Figure S3A-E. urp2 expression was upregulated in 5 weeks juveniles, both in straight and scoliotic fish and its upregulation was maintained in adult scoliotic fish (Figure 4A).

Figure 4:

Given the early and long-lasting upregulation of urp1/2 in rpgrip1l^-/- fish, we attempted to rescue axis curvature by decreasing global Urp1/2 production via genetic means. To that end, we generated double (rpgrip1l^+/-; urp2^+/-) heterozygous fish using the recently generated urp2 deletion allele (Gaillard et al., 2023) and crossed them to produce double mutants. Using a curvature quantification method on whole fish (Supplementary Figure S3I and methods), we observed that the removal of one or two urp2 functional allele(s) in rpgrip1l^-/- fish did not lower their curvature index (Figure 4B, C), neither at two mpf (Supplementary Figure S3 F) nor at four mpf (Supplementary Figure S3 G-H). We observed that urp2 mRNA amounts were reduced by removing one or two functional copies of urp2, probably as a consequence of mRNA decay (Gaillard et al., 2023) (Figure 4D). We also showed that none of the other urotensin 2 family members was upregulated by a compensation mechanism upon urp2 loss (Figure 4E for urp1 and Supplementary Figure S3C-E for urp, uts2a and uts2b).

Thus, lowering global Urp1/2 expression level in rpgrip1l mutants did not have any beneficial effect on the axis curvature phenotype, ruling out altered urp1/2 signaling as a main driver of scoliosis in rpgrip1l mutants.

Regulators of motile ciliogenesis, inflammation genes and annexins are upregulated in rpgrip1l^-/- juveniles

To get further insight into the early mechanisms driving spine curvature in rpgrip1l mutants, we obtained the transcriptome of the brain and dorsal trunk of 7 rpgrip1l^-/- juveniles at scoliosis onset (6 tail-up and 1 straight 5 wpf juveniles) and 5 control siblings (2 rgprip1l^+/+ and 3 rgprip1l^+/-). Differential gene expression (DGE) analysis was performed with the DSEQ2 software within the Galaxy environment. A complete annotated gene list is displayed in Supplementary Table 1 (Trunk) and Table 2 (Brain). Filtering the gene list to select for P-adj < 0.05 and Log2 FC > 0.75 or < −0.75 recovered a vast majority of upregulated genes in the dorsal trunk (75 downregulated vs 614 upregulated genes) and, to a lesser extent, in the brain (1 downregulated, rpgrip1l, vs 60 upregulated genes) as illustrated on the volcano plots from DGE analysis of the brain (Figure 5A) and trunk (Figure 5B). 54 out of 60 genes upregulated in the brain were also upregulated in the trunk, suggesting that, in the trunk, these genes were specific to the CNS (Figure 5D). In both the brain (Figure 5C) and the trunk (Supplementary Figure S4A), the most upregulated biological processes in GO term analysis included cilium movement and cilium assembly. Among the 20 most significant GO terms found for upregulated genes in the trunk, 8 were related to ciliary movement/assembly (Supplementary Figure S4A). The upregulation in both tissues of foxj1a, encoding a master transcriptional activator of genes involved in cilia motility (Yu et al., 2008) (Figure 5E) likely contributed to this enrichment. To confirm this hypothesis, we compared our dataset with those of targets of Foxj1 from (Choksi et al., 2014; Mukherjee et al., 2019; Quigley and Kintner, 2017). We found that 106 out of 614 upregulated genes were targets of Foxj1, among which 77 were direct targets (Supplementary Table 3). Several genes encoding regulators of embryonic axis curvature were also upregulated in rpgrip1l^-/- fish, among which pkd1b (Mangos et al., 2010), urp1 and urp2 (Bearce et al., 2022; Gaillard et al., 2023; Zhang et al., 2018) and sspo (Cantaut-Belarif et al., 2018; Rose et al., 2020; Troutwine et al., 2020) (Figure 5E). Finally, GO terms enrichment analysis also highlighted processes associated with inflammation and innate and adaptive immune responses, as described for the scoliotic ptk7 model (Van Gennip et al., 2018) sharing for example irg1l and complement genes up-regulation (Figure 5 C, E, and Supplementary Figure S4A).

Figure 5:

In order to confirm these results and to identify processes affected at the post- transcriptional level in rpgrip1l mutants, we performed a quantitative proteomic analysis by mass spectrometry of brains from 5 rpgrip1l^-/- adult fish and 5 control siblings. This analysis detected 5706 proteins present in at least 4 over the 5 replicates, of which 172 are associated with a P value < 0.05 and a Log2FC > +0.75 and < −0.75. 127 proteins were present in higher and 45 in lower quantity, as illustrated on the volcano plot (Figure 5F and Supplementary Table 4). Pathways enriched in rpgrip1l^-/- brains included regulation of vesicular trafficking and membrane repair processes, as well as proteolysis and catabolic processes, suggesting profound perturbation of cellular homeostasis (Supplementary Figure S4B). Among the proteins that were present in highest quantity in mutants, several members of the Annexin family were identified: Anxa2a, Anxa1a, Anxa5b (Figure 5F and Supplementary Figure S4C. Annexins are associated with membranes from endo and exovesicles as well as plasma membranes and involved in a wide array of intracellular trafficking processes such as cholesterol endocytosis, plasma membrane repair as well as modulation of immune system functions (Boye and Nylandsted, 2016; Grewal et al., 2016; Rentero et al., 2018). Of note, only four proteins of that list, Anxa2a, C4, C7a and Stat3, were the products of genes up-regulated in the transcriptomic study (indicated in red in Figure 5E). In agreement with the common enriched GO terms in both studies (Figure 5C and Supplementary Figure S4B), these four proteins are associated with immune response. Overall, rpgrip1l^-/- mutants at scoliosis onset present deregulated pathways in common with other scoliotic models such as activation of immune response genes, but also specific deregulated pathways such a marked activation of the Foxj1a ciliary motility program and an increase of Annexin levels.

Astrogliosis precedes scoliosis onset of rpgrip1l^-/- juvenile fish

Both transcriptomic and proteomic analysis highlighted the occurrence of innate and adaptive immune responses in rpgrip1l^-/-fish. Several upregulated genes, such as irg1l, mmp9 and anxa2, are reported to be expressed in macrophages and/or microglia but also in epithelia or neurons (Hall et al., 2014) (Kuhlmann et al., 2014) (Shan et al., 2015). We therefore investigated in which cell type(s) Anxa2, the most upregulated protein in the rpgrip1l^-/- brain proteome, was expressed. Immunostaining of brain sections revealed a strong Anxa2 staining on cells lining brain ventricles, such as SCO cells and ventral ependymal cells of the rhombencephalic ventricle (RhV) (Fig 6B’, D’) in rpgrip1l^-/- adult fish, which was barely detectable in controls (Figure 6A’, C’), and to a lesser extent on other brain regions (Supplementary Fig S5 A’-F’), suggesting a deregulation of ependymal cell homeostasis. In mutants, Anxa2 staining spread over long cellular extensions reaching the pial surface of the rhombencephalon (Figure 6D’ and Supplementary Figure S5F’). This cell shape is characteristic of radial glial cells, a cell type that behaves as neural and glial progenitors and also plays astrocytic functions in zebrafish, as anamniote brains are devoid of stellate astrocytes (Jurisch-Yaksi et al., 2020; Lyons and Talbot, 2015). To confirm the glial nature of Anxa2-positive cells around brain ventricles in rpgrip1l^-/- fish, we double-labeled brain sections with GFAP, a glial cell marker, and Anxa2. In adult rpgrip1l^-/-fish, SCO cells were strongly double-positive for Anxa2 and GFAP (Fig 6B, B’), while controls displayed much weaker labeling for GFAP in the SCO and no labeling for Anxa2 (Fig 6 A, A’). We made similar observations at rhombencephalic ventricle (RhV) and other brain levels: ependymal cells were positive for Anxa2 and strongly positive for GFAP, and GFAP labeling was much higher in mutants than in controls (Fig 6 C-D’; Supplementary Figure S5A-F’). GFAP overexpression is a hallmark of astrogliosis, a process in which astroglial cells react to the perturbed environment, as shown during brain infection, injury, ischemia or neurodegeneration (Escartin et al., 2021; Matusova et al., 2023; Zamanian et al., 2012). Thus, the strong increase of GFAP staining combined with Anxa2 overexpression in brain ependymal cells of rpgrip1^l-/- adult fish strongly suggests that these cells undergo astrogliosis as observed in other animal models following CNS injury (Chen et al., 2017; Zamanian et al., 2012).

Figure 6:

To assay whether an astrogliosis-like phenotype was already present at scoliotic onset, we immunostained GFAP on juvenile brain sections of straight and tail-up mutants and controls. Tail-up juveniles (n=3/3) presented a strong GFAP staining in almost all cells of the SCO and along the ventral rhombencephalic ventricle, higher than control levels (Figure 6 E, I, H, L, M) while the phenotype of straight fish was heterogeneous: two out of four juveniles presented GFAP-positive cells (Figure 6G, K and red dots on Figure 6M graph), while the two other did not (Figure 6F, J and green dots on Figure 6M graph) as confirmed by cell quantification (Figure 6M). As scoliosis onset is asynchronous in rpgrip1l^-/-, heterogenous GFAP staining among straight mutant fish may support an early role of astrogliosis at SCO and RhV levels in scoliosis onset.

Immune cell enrichment around the SCO and within tectum parenchyma is detected prior to scoliosis onset

Astrogliosis may be reinforced or induced by an inflammatory environment containing cytokines released by microglia or macrophages or the presence of high levels of ROS and NO (Sofroniew, 2015). We thus performed immunostaining for the LCP1/L- Plastin marker, which labels both macrophages and microglia (Figure 6N-Q’), in adjacent sections to those used for GFAP (Figure 6E-L). We found that the scoliotic fish presented a higher number of LCP1-positive cells of ameboid shape within SCO cells (Figure 6Q, Q’, R). Strikingly, the straight mutants (2/4) that presented a high GFAP staining within SCO cells (Figure 6G, K) also showed a higher number of LCP1-positive cells around the SCO (Figure 6P, P’; red dots in Figure 6R) than in controls (Figure 6N, N’), while the other straight mutants did not ((Figure 6F, J, O, O’, green dots in Figure 6M) suggesting positive regulation between astrogliosis and macrophages/microglia recruitment. LCP1+ cells were also detected in other regions of the brain (Supplementary Figure S5 G-K). In the optic tectum, where resident LCP1-positive macrophages/microglia were present in controls, a two-fold increase in the number of ramified LCP1⁺ cells was observed in straight rpgrip1l^-/-juveniles compared to controls (Supplementary Figure S5H-J). Surprisingly, scoliotic rpgrip1l^-/- fish showed similar LCP1⁺ cell number as control fish in the tectum (Supplementary Figure S5G), suggesting a transient increase of macrophage/microglia activation preceding scoliosis.

Our data thus uncovers an early astrogliosis process initiated asynchronously at the SCO and RhV levels of straight rpgrip1l^-/- fish, favored by a transient inflammatory state, and preceding the loss of multiciliated tufts around the SCO and that of the RF, that we only detected in curved rpgrip1l^-/-fish. Astrogliosis then spreads along the brain ventricles and becomes prominent in brain of scoliotic fish. In addition, we found among the upregulated genes in dorsal trunk at scoliosis onset, a number of activated astrocyte markers (Matusova et al., 2023) such as s100b, c4b, gfap, glast/slc1a3b, viml, ctsb (Supplementary Figure S4D), showing that astrogliosis spread over the whole central nervous system.

Brain neuroinflammation and astrogliosis contribute to scoliosis penetrance and severity in rpgrip1l^-/- fish

Our data suggest that astrogliosis and neuro-inflammation could play a role in triggering scoliosis in rpgrip1l mutants. We thus aimed to counteract inflammation before curvature initiation by performing drug treatment with the hope of reducing scoliosis penetrance and severity. We used the NACET anti-oxidant and anti- inflammatory drug that successfully reduced scoliosis of sspo hypomorphic mutant fish (Rose et al., 2020). We first assayed the biological activity of our NACET batch on the ciliary motility mutant dnaaf1^-/- (Sullivan-Brown et al., 2008). Indeed, NACET was shown to suppress the embryonic tail-down phenotype of the ciliary motility mutant ccdc151^ts (Rose et al., 2020). We found that 3mM NACET treatment suppressed the dnaaf1^-/- tail-down phenotype (Supplementary Figure S6A-F). To suppress scoliosis, we treated the progeny of rpgrip1l^+/- incrosses from 4 to 12 wpf with 1.5 mM NACET as in (Rose et al., 2020).This long-term treatment reduced scoliosis penetrance from 92% to 58% and markedly reduced spinal curvature index (Figure 7A-C), measured and illustrated as shown in Supplementary Figure S3I.

Figure 7:

We then asked whether NACET treatment suppressed astrogliosis and neuroinflammation in rpgrip1l^-/- treated fish. Indeed, the intensity and number of GFAP+ cells within the SCO were reduced in slightly curved fish and alleviated in straight fish (Figure 7E, F). Furthermore, the quantification of Lcp1-positive cells in the SCO region showed a drastic reduction of these cells in NACET-treated fish, back to levels found in rpgrip1l^+/+ fish (Figure 7D, E).

Our data thus show that rpgrip1l^-/- juvenile fish display brain astrogliosis and inflammation at scoliosis onset and that an anti-inflammatory/anti-oxidant treatment of rpgrip1l^-/- juveniles is able to maintain a straight axis in approximately one third of the mutants and to slow down curve progression in the other two thirds. The beneficial effect of NACET on axis curvature correlates with the combined reduction of astrogliosis and immune cell density.

Brain astrogliosis is present in zebrafish cep290 ^fh297/fh297 scoliotic fish

To investigate whether abnormal SCO cilia function and brain astrogliosis may be a general mechanism of scoliosis development in TZ mutants, we studied the cep290^fh297/fh297 zebrafish mutant. Zebrafish cep290 mutants display variable axis curvature defects at embryonic stages as well as slow retinal degeneration, and develop juvenile scoliosis (Lessieur et al., 2019; Wang et al., 2022). Like RPGRIP1L, the human CEP290 gene encodes a TZ protein and its mutation leads to severe neurodevelopmental ciliopathies (Baala et al., 2007). All cep290^fh297/fh297juveniles developed a fully penetrant scoliosis by 4 wpf in our fish facility (n: 30/30). We analyzed cilia presence within the SCO by immunostaining on sections (Figure 8A- F’). At 4 wpf, the SCO had not reached its final adult width (14 versus 24 cells wide in the anterior SCO), and each of its cells harbored a short monocilium labeled by glutamylated tubulin (Figure 8A-C). Multicilia tufts on SCO lateral walls were not as clearly visible as in adults (Figure 2F) but patches of glutamylated tubulin staining were detected on both sides of the posterior SCO (Figure 8B-C’). In cep290^-/-, monocilia of the SCO appeared longer than in controls (N=3/3) and pointed towards an abnormally elongated diencephalic ventricle (Figure 8D-F). Lateral patches of glutamylated tubulin staining were present in cep290^-/- posterior SCO as in controls (Figure 8E-F’). Thus, cilia morphology is perturbed within cep290^-/- SCO monociliated cells, as well as the overall morphology of the diencephalic ventricle at this level.

Figure 8:

To test for the presence of astrogliosis, we quantified the number of SCO cells presenting high GFAP labeling on several sections along the A/P extent of the SCO in 3 control and 3 cep290^-/- brains (Figure 8G-K). 50% of SCO cells presented strong GFAP staining in fully scoliotic juveniles, which was not detectable in controls (Figure 8H, K, N=2/3), while a lower proportion (22%) of SCO cells was GFAP positive in the juvenile with a milder curvature (Figure 8M, N=1/3). Strong GFAP staining was also detected in some diencephalic radial glial cells (Figure 8K). On the same sections, immune cells were quantified by LCP1 staining. A mean of 4 LCP1 positive cells per SCO field was found in cep290^-/-, versus 1 in controls (Figure 8I, L and N). As in rpgrip1l mutants, abnormal monocilia, astrogliosis and immune cell enrichment were evidenced at SCO levels in Cep290 ^-/-(3/3). Furthermore, astrogliosis was observed within radial glial cells lining the RhV at cerebellar level (Figure 8O-R) and at anterior hindbrain level (Figure 8S-V), as shown in rpgrip1l^-/- scoliotic brain (Figure 6D, K, L). This indicates that astrogliosis is shared by these two zebrafish scoliotic models, even though they develop axis curvature at different stages (3-4 wpf for cep290^-/- versus 5-12 wpf for rpgrip1l^-/-).

Discussion

In this paper we dissected the mechanisms of scoliosis appearance in a zebrafish mutant for the TZ gene rpgrip1l, taking advantage of its asynchronous onset to decipher the chronology of events leading to spine torsion. Transcriptome analysis of the brain and trunk at scoliosis onset in rpgrip1l^-/- fish compared to control siblings revealed increased expression of urp1/2, foxj1a and its target genes, as well as genes linked to immune response. Genetic experiments ruled out an involvement of increased Urp1/2 signaling in scoliosis development. Cilia defects appeared asynchronously in different CNS regions, while RF loss strictly coincided with scoliosis onset. Strikingly, we observed a strong increase of GFAP staining along the diencephalic and rhombencephalic ventricles, indicative of astrogliosis. This astrogliosis preceded scoliosis onset and then spread posteriorly and increased in severity. In the region of the SCO, the appearance of astrogliosis and neuroinflammation coincided temporally, preceding the onset of RF depolymerization and spine torsion. NACET treatment showed that increased astrogliosis and inflammation participated in the penetrance and severity of scoliosis in rpgrip1l^-/- fish. Finally, we showed that another TZ gene mutant, cep290, also displayed astrogliosis at the time of scoliosis onset. Thus, the cascade of events uncovered in the rpgrip1l mutant sheds a new light on the origin of zebrafish scoliosis caused by ciliary TZ defects.

Cilia motility defects lead to RF loss and to scoliosis in several zebrafish mutants (Grimes et al., 2016; Rose et al., 2020)). Rpgrip1l encodes a TZ protein important for cilia integrity and function in many eukaryotic organisms (Andreu-Cervera et al., 2019; Gogendeau et al., 2020; Jensen et al., 2015; Wiegering et al., 2018). Our data show that rpgrip1l deficiency in zebrafish affects cilia maintenance differentially depending on the CNS territory and the developmental stage, raising the question of which ciliated cells are responsible for RF loss and scoliosis initiation in this mutant. The full rescue observed after re-introducing RPGRIP1L in foxj1a-expressing cells points towards a crucial function of the protein in cells lining CNS cavities. Foxj1a- positive cell populations encompass both monociliated and multiciliated ependymal cells, radial glial progenitors, glial cells of the SCO and the ChP, as well as CSF- contacting sensory neurons in the spinal cord (D’Gama et al., 2021; Prendergast et al., 2023; Ribeiro et al., 2017; Van Gennip et al., 2018). Recent work suggested that impairment of ciliogenesis in the fChP of the katnb1 mutant could play a role in inducing axis curvature (Meyer-Miner et al., 2022). In the case of the rpgrip1l mutant, we temporally uncorrelated fChP cilia defects from scoliosis onset. This, combined with the fact that the total loss of multicilia in the fChP of [foxj1b^-/-, gmnc^-/-] double mutants does not trigger axis curvature (D’Gama et al., 2021) suggests that another foxj1a-expressing population is crucial for axis straightness. Our results strongly suggest that cilia defects in the SCO trigger axis curvature in rpgrip1l mutants. Two cilia populations were affected in the SCO of rpgrip1l mutant juveniles at the onset of scoliosis: cilia of monociliated glial cells that secrete SCO-spondin, which appeared longer in mutants than in controls, and multiciliated tufts at SCO exit, which were lost in scoliotic mutants. Both defects could be responsible for the phenotype, by perturbing CSF flow and content (presence of SCO-spondin aggregates in ventricles) (left side on the Graphical Abstract) and leading to ventricular dilations or by perturbing primary cilia signaling within SCO cells (right side of the Graphical Abstract). The observation that multiciliated cells were not yet fully differentiated in cep290 mutants at scoliosis onset is in favor of a prominent role of monocilia. This is consistent with the literature showing that most ciliary mutant models develop axis curvature at 3-4 weeks, a stage when brain multiciliated cells are not yet differentiated (D’Gama et al., 2021). Still, as scoliosis onset in rpgrip1l mutant is asynchronous from 5 to 12 weeks, MCC loss at SCO exit could contribute to defective RF polymerization.

Our model for scoliosis appearance in the rpgrip1l mutant is presented in the graphical abstract. Since we have observed several straight rpgrip1l^-/- fish with astrogliosis and increased immune cells presence around the SCO, we propose that rpgrip1l^-/- SCO and astroglial cells experience altered signaling that activates immune cell recruitment, producing an inflammatory environment. This would promote multicilia loss as a secondary event (Graphical Abstract, orange arrow) (Lattke et al., 2012) leading to RF depolymerization and spine curvature. However, as sspo ^-/- hypomorphic mutants which present aggregates within ventricles have been shown to present an inflammatory response (Rose et al., 2020), it is equally possible that SCO astrogliosis is a downstream consequence of the presence of these proteoglycans aggregates in contact with primary cilia or ventricular cell surfaces (Graphical Abstract, green arrow). These two scenarios may also act in parallel to reinforce scoliosis onset and ependymal cilia loss. Future experiments will help to discriminate between these two scenarios.

Our transcriptome analysis and qRT-PCR data challenges a model proposed for axis straightness of zebrafish embryos in which the loss of RF leads to a down-regulation of urp2 and urp1 expression (Zhang et al., 2018) since we observed an upregulation of both genes before and after RF loss. As forced expression of urp1/2 induces a tail- up phenotype in embryos and juveniles (Zhang et al., 2018) (Gaillard et al., 2023; Lu et al., 2020), we attempted to rescue rpgrip1l^-/- axis curvature by down-regulating urp1/2 expression. No beneficial effect on scoliosis penetrance or severity was observed, indicating that increased urp1/2 expression level does not significantly contribute to axis curvature in rpgrip1l^-/-fish.

In this study, we found an unexpected and strong defect characterized by enhanced GFAP and Anxa2 expression, which we identified as astrogliosis (also called reactive astrogliosis). Astrogliosis is a reaction of astroglial cells to perturbed homeostasis of the CNS, characterized by molecular and phenotypic changes in these cells (Escartin et al., 2021; Matusova et al., 2023). In zebrafish, radial glial cells are endowed with both neurogenic and astrocytic functions, and are thus also called astroglial cells (Jurisch-Yaksi et al., 2020). Astrogliosis in rpgrip1l mutants arose in a subdomain of Foxj1a-positive cells, within the DiV and ventral ependymal cells lining the RhV. The observation of astrogliosis in some straight juvenile mutants suggests an early role of this defect in scoliosis onset.

Astrogliosis along the ventricular cavities could arise in response to local mechanical or chemical perturbations caused by cilia motility defects, abnormal CSF flow and content and ventricular dilations. In two murine models of primary ciliary dyskinesia, decreased CSF flow was associated with gliosis at juvenile stage (Finn et al., 2014). A response to CSF pertubation in rpgrip1l mutants is also suggested by the strong upregulation of Foxj1a and its target genes. foxj1a expression is also reported to be upregulated in the CNS of three other scoliotic models in addition to rpgrip1l, namely in ptk7, sspo and katnb1 mutants (Escartin et al., 2021; Meyer-Miner et al., 2022; Van Gennip et al., 2018), suggesting that it constitutes a common response to similar ventricular and CSF defects. Interestingly, foxj1a expression is also upregulated upon zebrafish embryonic (Cavone et al., 2021; Hellman et al., 2010) or adult (Ribeiro et al., 2017) spinal cord injury. The up-regulation in the rpgrip1l mutant proteome of the membrane repair module Anxa2-S100a10-Ahnak (Bharadwaj et al., 2021) (Figure 4C) together with the Foxj1-induced motility module detected in the transcriptomic analysis may indicate ongoing reparation attempts of CNS ventricles and associated neural tissues.

Moreover, the presence of few Lcp1+ cells around the mutant SCO suggests that microglia/macrophages could participate in astrogliosis establishment and reinforcement. This dialogue between immune Lcp1+ cells and astroglial cells is evidenced during early stages of regeneration after acute injury or cellular damages (Cavone et al., 2021). In our bulk RNAseq analysis performed at scoliosis onset, we have indications of the upregulation of microglia and macrophage markers (p2ry12 and mpeg1) at trunk levels and of numerous markers of activated astrocytes characterized in several mammalian astrogliosis models (Matusova et al., 2023). A similar dialogue is detected in polycystic kidney disease models where compromised primary cilia signalling leads to uncontrolled cytokines secretion by epithelial cells, a situation that favors local immune cells recruitment and proliferation (Viau et al., 2018).

What are the intracellular mechanisms involved in astrogliosis induction in ciliated ventricular cells? Candidate pathways emerge from our multi-omics studies. Proteomic data showed a 2.4-fold increase in GSK3bb amount in rpgrip1l^-/- mutant brains (P value < 0.001, Supplementary Table 3). In several neurodegenerative murine animal models, GSK3b enzymatic activity was shown to promote inflammation and gliosis (Jorge-Torres et al., 2018; Medina and Avila, 2010; Mines et al., 2011). Another potential trigger of astrogliosis may be a defective mitochondrial activity as the amounts of four proteins involved in electron transport chain activity (gpd1b, mt-nd6, pdia5 and dmgdh) were reduced by 40% to 86% in the analysis of mutant brain proteome (Supplementary Table 4). We think this is of particular interest in light of a recent report demonstrating that impaired mitochondrial activity within ciliated astrocytes leads to the induction of the Foxj1 ciliary motility program, the elongation and distortion of astrocyte primary cilia as well as reactive astrogliosis (Ignatenko et al., 2023), three phenotypes also observed in rpgrip1l^-/- brains.

Finally, our observation of widespread astrogliosis in two TZ gene mutants, rpgrip1l and cep290, suggests that it could represent a general mechanism involved in scoliosis downstream of cilia dysfunction. Of note, astrogliosis markers such as glast/slc1a3b, vim, s100b, c4, timp2b, gfap and ctssb.2, are also upregulated in the transcriptome of ptk7 mutants (Van Gennip et al., 2018). We propose that sustained astrogliosis might impair neuronal survival and activity, crucial for proper interoception and locomotor control, thus leading to axis curvature in the context of a rapidly growing axial skeleton. A similar context of astrocytosis associated with spinal cord dilation has been observed in human patients with syringomyelia. This population with a high incidence of scoliosis can present hyper-signals on T2 MRI scans, which were proposed to reflect local astrocytosis and could be validated by biopsies in rare cases (Sherman et al., 1987). Further imaging studies will need to be performed to validate a potential link between developing astrogliosis and the onset and progression of idiopathic scoliosis in humans.

Acknowledgements

We are grateful to the IBPS aquatic animal, imaging and bioinformatics facilities and to the ICM sequencing facility for their technical assistance. We thank MichaeLJl Trichet from the IBPS electron microscopy facility for participating in the MEB experiments, the TACGENE facility for providing the Cas9 protein; the TEFOR Paris- Saclay facility for the brain clearing experiment; Thierry Jaffredo and Pierre Charbord for their precious help in transcriptome analysis; Nicolas Baylé for initial characterization of the rpgrip1l mutant; Claudia Hoffman for providing Cep290 fixed samples; Brian Ciruna for sharing the foxj1a expressing constructs and detailed NACET treatment protocol; Ryan Gray for sharing the scospondin-GFP^ut24 zebrafish line. This work was supported by funding to SSM from the Fondation pour la Recherche Médicale (Equipe FRM EQU201903007943) and the Fondation Yves Cotrel.

Note

This reviewed preprint has been updated to correct the order of the figure images.

Material and methods

Zebrafish

Wild-type, rpgrip1l^ex4 and rpgrip1l^Δ zebrafish embryos and adults were raised, staged and maintained as previously described (Kimmel et al., 1995). All our experiments were made in agreement with the european Directive 210/63/EU on the protection of animals used for scientific purposes, and the French application decree ‘Décret 2013-118’. The projects of our group have been approved by our local ethical committee ‘Comité d’éthique Charles Darwin’. The authorization number is APAFIS #31540-2021051809258003 v4. The fish facility has been approved by the French ‘Service for animal protection and health’ with approval number A750525. All experiments were performed on Danio rerio embryos of mixed AB/TL background. Animals were raised at 28.5°C under a 14/10 light/dark cycle.

Rpgrip1l mutant generation and genotyping

Guide RNA preparation and microinjection

CRISPR target sites were selected for their high predicted specificity and efficiency using the CRISPOR online tool (Haeussler et al., 2016). Real efficiency was assessed on zebrafish embryos by T7E1 test. The two most efficient guides (Rpgrip1l_x4_G1: GCTTACGGTCCTTCACCAGACGG and Rpgrip1l_x25_G3: CCTCAGTTGACAGGTTTCAGCGG) respectively situated 24 nt from the beginning of exon 4 and 82 nt downstream of exon 25 were kept for further experiments. sgRNA transcription templates were obtained by PCR using T7_Rpgrip1l-x4_G1_Fw primer (5′-GAAATTAATACGACTCACTATAGGCTTACGGTCCTTCACCAGAGTTTTAGAGCTAGAAATAGC- 3′) or T7_Rpgrip1l-x25_G3_Fw (5′- GAAATTAATACGACTCACTATAGGCCTCAGTTGACAGGTTTCAGGTTTTAGAGCTAGAAATAG C-3′) as forward primer and sgRNA_R universal primer (5′- AAAAGCACCGACTCGGTGCCACTTTTTCAAGTTGATAACGGACTAGCCTTATTTTAACTTGCTA TTTCTAGCTCTAAAAC-3′) as reverse primer. sgRNAs were transcribed using Megascript T7 Transcription Kit (Thermo Fisher Scientific, Waltham, MA) and purified using NucleoSpin® RNA Clean up XS kit (Macherey Nagel, DuLJren, Germany). sgRNA:Cas9 RNP complex was obtain by incubating Cas9 protein (gift of TACGENE, Paris, France) (7.5 μM) with sgRNA (10 μM) in 20 mM Hepes-NaOH pH 7.5, 150 mM KCl for 10 min at 28 °C. 1–2 nl was injected per embryo. For deletion, Rpgrip1l_x4_G1 and Rpgrip1l_x25_G3 RNP complexes were mixed half and half.

We kept two F1 fish, one with a 15 Kb deletion between exon 4 and 25 of rpgrip1l, that we called rpgrip1l^Δ and that we submitted to ZFIN under the name rpgrip1l ^bps1and a second one with a stop codon within exon 4 called rpgrip1l ^ex4, that is submitted to ZFIN under the name rpgrip1l ^bps2.

Screening and genotyping

Injected (F0) fish were screened for germline transmission by crossing with wild type fish and extracting genomic DNA from resulting embryos. For genomic DNA extraction, caudal fin (juveniles/adults) or whole embryo DNA were used. Genomic DNA was isolated with Proteinase K (PK) digestion in 40 of lysis buffer (100 mM Tris- HCl pH 7.5, 1 mM EDTA, 250 mM NaCl, 0.2% SDS, 0.1 μg/μl Proteinase K (PK)) for embryos (300 μl for adult fin) overnight at 37°C with agitation. PK was inactivated for 10 min at 90°C and a five-fold dilution was used as template for PCR amplification. Combined genotyping of wild type and mutant alleles (Supplementary Figure S1A’) was performed by PCR using 3 primers, a common reverse primer for both alleles : Rpgrip1l_ex25_R3 (GTTGTGTCTCTGCCATATATTG), a specific forward primer for the deleted allele: Rpgrip1l-ex4-del (CCCACACTGCATACGCACTC) and a specific forward primer for the wild type allele : Rpgrip1l_ex25F (AGTGTGCGGTACATCTCCAA) at an annealing temperature of 60°C for 35 cycles. Expected size of the amplified fragment is 364 nt for the wild type allele and 544nt for the mutant allele. The deleted allele presents a frameshift at position N77 leading to the generation of a STOP codon after 11 aas, thus leading to a N-terminally truncated protein at position 88/1256 aas.

For genotyping the urp2 mutant line, caudal fins from adult fish were extracted in 300µl of lysis solution and PCR was performed as above to detect exon 5 deletion using the following primers: urp2-F1 (TGATTACTAGCCCTGTCCCAAC) and urp2-R1 (AGGTACAGTACACACGTCACAG).

μCT scans

The samples were scanned on a Bruker micro scanner (Skyscan 1272) with a resolution of 8.5 μm, a rotation step of 0.55° and a total rotation of 180°. For the acquisition of adult fish (2.5 cm), a 0.25 mm aluminium filter was used, for a voltage of 50 kV and an intensity of 180 mA, for juvenile fish (0.9 to 1.2 cm) the filter was omitted, and a voltage of 60 kV was used with an intensity of 166 mA. Each image contained 1008 x 672 pixels and was based on the average of 3 images. The 3D reconstruction by backprojection was carried out by the NRecon software and the 2D image overlays were then cleaned by the CT Analyser software. The Dataviewer software allowed all fish skeletons to be oriented in the same way taking the otoliths as landmarks. The CTvox software then allowed 3D visualizing of the samples. Morphometric analysis was performed with the CT Analyser software.

Scanning electron microscopy

3 month fish (3 controls, 5 mutants) were euthanized using lethal concentration of MS222 (0.028 mg/mL). The brains were quickly dissected in 1.22 X PBS (pH 7.4), 0.1 M sodium cacodylate and fixed overnight with 2% glutaraldehyde in 0.61 X PBS (pH 7.4), 0.1 M sodium cacodylate at 4°C. They were sectioned along the dorsal midline with a razor blade to expose their ventricular surfaces. Both halves were washed four times in 1.22 X PBS and post-fixed for 15 minutes in 1.22 X PBS containing 1% OsO4. Fixed samples were washed four times in ultrapure water, dehydrated with a graded series of ethanol and critical point dried (CPD 300, Leica) at 79 bar and 38 °C with liquid CO2 as the transition fluid and then depressurized slowly (0,025 bar/s). They were then mounted on aluminum mounts with conductive silver cement. Sample surfaces were coated with a 5 nm platinum layer using a sputtering device (ACE 600, Leica). Samples were observed under high vacuum conditions using a Field Emission Scanning Electron Microscope (Gemini 500, Zeiss) operating at 5 kV, with a 20 μm objective aperture diameter and a working distance around 3 mm. Secondary electrons were collected with an in-lens detector. Scan speed and line compensation integrations were adjusted during observation.

Histological sections of juvenile and immunofluorescence on sections

Juvenile and adult zebrafish were euthanized using lethal concentration of MS222 (0.028 mg/mL). Pictures and size measurements were systematically taken before fixation. Fish were fixed in Zamboni fixative [35 ml PFA 4 %, 7.5 ml saturated picric acid (1.2 %), 7.5 ml 0.2M Phosphate Buffer (PB)] [55] overnight at 4°C under agitation. Fish were washed with Ethanol 70% and processed for dehydration by successive 1 h incubation in Ethanol (3 x 70% and 2 x 100%) at room temperature under agitation, then for paraffin inclusion. 14 μm sagittal paraffin sections were obtained using a Leica RM2125RT microtome. Sections were deparaffinized and antigen retrieval was performed by incubation for 7 min in boiling citrate buffer (10 mM, pH 6). Immunofluorescence staining was performed as described previously (Andreu-Cervera et al., 2019). The following primary antibodies were used: anti-RF, anti-Acetylated Tubulin, anti-Glutamylated Tubulin, anti-Arl13b, anti-LCP1, anti- AnnexinA2. Corresponding primary and secondary antibodies are described and referenced in the Resources Table.

Immunofluorescence on whole embryos

Embryos from 24 to 40 hpf were fixed 4 hr to overnight in 4% paraformaldehyde (PFA) at 4°C. For Reissner fiber staining, larvae at 48 and 72 hpf were fixed 2 hr in 4% PFA and 3% sucrose at 4°C, skin from the rostral trunk and yolk were removed. Samples from 24 to 40 hpf embryos were blocked overnight in a solution containing 0.5% Triton, 1% DMSO, 10% normal goat serum and 2 mg/mL BSA. For older samples (48 to 72 hpf) triton concentration was increased to 0.7%. Primary antibodies were incubated one to two nights at 4°C in a buffer containing 0.5% Triton, 1% DMSO, 1% NGS and 1 mg/mL BSA. All secondary antibodies were from Molecular Probes, used at 1:400 in blocking buffer, and incubated a minimum of 2.5 hr at room temperature. The primary antibodies chosen for in toto immuno-labeling against Reissner fiber, Acetylated-tubulin, Myc, Arl13b and Gamma-tubulin as well as the corresponding secondary antibodies are referenced in the Key Resources Table. Whole mount zebrafish embryos (dorsal or lateral mounting in Vectashield Mounting Medium) were imaged on a Leica SP5 confocal microscope and Zeiss LSM910 confocal microscope, both equipped with a 63X immersion objective. Images were then processed using Fiji (Schindelin et al., 2012).

Whole-mount brain clearing

Brains were dissected from 4-5 wpf size-matched (0.9-1.2 mm length) zebrafish after in toto fixation with formaldehyde. Whole–mount tissue clearing was performed following the zPACT protocol (Affaticati et al., 2017). In brief, brains were infused for 2 days in hydrogel monomer solution (4% acrylamide, 0.25% VA- 044, 1% formaldehyde and 5% DMSO in 1X PBS) at 4°C. Polymerization was carried out for 2h30 at 37°C in a desiccation chamber filled with pure nitrogen. Brains were transferred into histology cassettes and incubated in clearing solution (8% SDS and 200 mM boric acid in dH2O) at 37°C with gentle agitation for 8 days. Cleared brains were washed in 1X PBS with 0.1% Tween-20 (PBT) for 3 days at room temperature and kept in 0.5% formaldehyde, 0.05% sodium azide in PBT at 4°C until further processing. Brains were subsequently placed for 1h in depigmentation pre-incubation solution (0.5X SSC, 0.1% Tween-20 in dH2O) at room temperature. The solution was replaced by depigmentation solution (0.5X SSC, Triton X-100 0.5%, formamide 0.05% and H2O2 0.03% in dH2O) for 45 minutes at room temperature. Depigmented brains were washed for 4h in PBT and post-fixed (2% formaldehyde and 2% DMSO in PBT) overnight at 4°C.

Cleared brain immunostaining of whole adult brains

Whole-mount immunolabeling of cleared brains was performed as described in the zPACT protocol with slight modifications. Briefly, brains were washed in PBT for one day at room temperature and blocked for 10h in 10% normal goat serum, 10% DMSO, 5% PBS-glycine 1M and 0.5% Triton X-100 in PBT at room temperature. Brains were washed again in PBT for 1h prior to incubation with anti- ZO-1 antibody (ZO1-1A12, Thermofisher, 1:150) in staining solution (2% normal goat serum, 10% DMSO, 0.1% Tween-20, 0.1% Triton X-100 and 0.05% sodium azide in PBT) for 12 days at room temperature under gentle agitation. Primary antibody was renewed once after 6 days of incubation. Samples were washed three times in PBT and thereafter incubated with Alexa Fluor 488-conjugated secondary antibody (A11001, Invitrogen, 1:200) for 10 days in the staining solution at room temperature under gentle agitation. Secondary antibody was renewed once after 5 days of incubation. Samples were washed three times in PBT prior to a counterstaining with DiIC18 (D282, Invitrogen, 1μM) in the staining solution for three days at room temperature with gentle agitation. Samples were washed in PBT for 3h before mounting procedure.

Mounting and confocal imaging

Samples were placed in 50% fructose-based high-refractive index solution (fbHRI, see (Affaticati et al., 2017) /50% PBT for 1h, then in 100% fbHRI. Brains were mounted in agarose-coated (1% in standard embryo medium) 60 mm Petri dish with custom imprinted niches to help orientation. Niche-fitted brains were embedded in 1% phytagel and the Petri dish filled with 100% fbHRI. fbHRI was changed three times before imaging until its refractive index matched 1.457. Images were acquired with a Leica TCS SP8 laser scanning confocal microscope equipped with a Leica HC FLUOTAR L 25x/1.00 IMM motCorr objective. Brains were scanned at a resolution of 1.74x1.74x1.74 μm (xyz) and tiled into 45 to 70 individual image stacks, depending on brain dimensions, subsequently stitched, using LAS X software.

Volumetric analysis of the posterior ventricles

The volumes of the posterior ventricles were segmented, reconstructed and analyzed using Amira for Life & Biomedical Sciences software (Thermo Fisher Scientific). In essence the ventricles volumes were manually segmented in Amira’s segmentation editor and subsequently refined by local thresholding and simplification of the corresponding surfaces. Volumes, which were open to the environment were artificially closed with minimal surfaces by connecting the distal-most points of their surface to the contralaterally corresponding points using straight edges. Due to the biologic variability of the sample population the overall size of the specimens needed to be normalized to keep the measured volumes comparable. For this spatial normalization one of the specimens was randomly selected from the wildtype population as template (1664, grey) and the ‘Registration’ module in Amira was used to compute region-specific rigid registrations for the other specimens, allowing for isotropic scaling only [details in supp. mat.]. For excluding the influence of the ventricular volumes, the registration was computed on the basis of the independent reference stain (DiIC18) (details in supp. mat.). Region specific volume differences between the mutant and wildtype population were evaluated on seven subvolumes of the posterior ventricles (details in supp. mat.).

RNA extraction for transcriptome analysis and quantitative RT-PCR

Juvenile and adult zebrafish were euthanized using lethal concentration of MS222 (0.28 mg/mL). Their length was measured and their fin cut-off for genotyping. For transcriptomic analysis, brain and dorsal trunk from 1 month juvenile (0.9 to 1.0 cm) zebrafish were dissected in cold PBS with forceps and lysed in QIAzol (QIAGEN) after homogenization with plunger pistons and 1ml syringes. Samples were either stored in QIAzol at −80°C or immediately processed. Extracts containing RNA were loaded onto QIAGEN-mini columns, DNAse digested and purified in the miRNAeasy QIAGEN kit (Cat 217004) protocol. Samples were stored at −80°C until use. RNA concentration and size profile were obtained on the Tapestation of ICM platform. All preparations had a RIN above 9.2. For Q-PCR from juveniles, whole fish minus internal organs were lysed in Trizol (Life technologies) using the Manufacturer protocol.

Quantitative PCR from individual juvenile or adult fish

The cDNA from the isolated RNAs was obtained following GoScript Reverse Transcription System protocol (Promega), using 3 to 4 μg of total RNA for each sample. The 20 μL RT-reaction was diluted 4-fold and 4μl was used for each amplification performed in duplicates. The primers used for qPCR were the following: urp2 (F: ACCAGAGGAAACAGCAATGGAC; R: TGAGGTTTCCATCCGTCACTAC), urp1 (F: ACATTCTGGCTGTGGTTTGTTC; R: CTCTTTTGCACCTCTCTGAAGC), urp (F: GCGGAAAAATGTCATGCCTCTTC; R: TTGAGCTCCTTTTGAAGCTCCTG), uts2a (F: CACTGCTCAACAGAGACAGTATCA; R: CCAAAAGACCACTGGGAGGAAC), uts2b (F: TACCCGTCTCTCATCAGTGGAG; R: TTTTCCAGCAAGGCCTCTTTTAC) where all of them are from Gaillard et al. 2023, rpgrip1l (F: CAGACACCTGCTGGAGTTACA; R: TCCTGACTCACATCAAACGCA), irg1l (F: TCCCTAAGAGGTTCCATCCTCC; R: AAGATGCAGCGATGGCCAAA), stat3 (F: CAGGACTGGGCGTATGCG; R: GAAGCGGCTGTACTGCTGAT), c4b (F: GGGTGTTCTTATGGCGGTGG and R: GCGCACAACAAGCTTTTCTCATC) lsm12b (F: GAGACTCCTCCTCCTCTAGCAT and R: GATTGCATAGGCTTGGGACAAC). The Q-PCR were performed using the StepOne real-time PCR system and following its standard amplification protocol (Thermo Scientific). Relative gene expression levels were quantified using the comparative CT method (2^-ΔΔCT method or 2^-ΔCT) on the basis of CT values for target genes and lsm12b as internal control.

RNA-sequencing and analysis

RNA-seq libraries were constructed by the ICM Platform (PARIS) from 100 ng of total RNA using the KAPA mRNA Hyperprep kit (Roche) that allows to obtain stranded polyA libraries, and their quality controlled on Agilent Tape station. The sequencing was performed on a NovaSeq 6000 Illumina for both strands. Sequences were aligned against the GRCz11 version of zebrafish genome using RNA Star function in “Galaxy” environment. The 12 Brain libraries reached between 22 to 26 Millions of assigned reads, and the trunk libraries between 26 to 35 Millions of assigned reads. Expression level for each gene was determined using Feature counts function from RNA Star BAM files. DESeq2 function was used to calculate the log2 Fold change and adjusted P values from the Wald test for each gene comparing the 7 mutants raw count collection to the 5 controls raw counts collection. Volcano plots were generated using the Log2 fold change and P adj. value of each Gene ID from the DESeq2 table. A threshold of 5.10 E-2 was chosen for the Padj. values and a log2 fold change > to +0.75 or < to −0.75 to select significantly up or down-regulated regulated genes. Both gene lists were used as input to search for enriched GO terms and KEGGs pathways using Metascape software (Zhou et al., 2019) as well as to compare up-regulated genes in the dorsal trunk versus the brain using Galaxy.

Quantitative proteomics

Sample preparation: Adult zebrafish were euthanized using lethal concentration of MS222 (0.28 mg/mL). For proteomic analysis, brain from 3 months zebrafish were dissected in cold PBS with forceps and immediately flash frozen in liquid nitrogen and stored at −80°C. Zebrafish brain samples were lysed in RIPA buffer (Sigma) with 1/100 antiprotease and sonicated for 5 minutes. After a 660 nm protein assay (Thermo), samples were digested using a Single Pot Solid Phase enhanced Sample Preparation (SP3) protocol (Hughes et al., 2014). In brief, 40 µg of each protein extract was reduced with 12mM dithiothreitol (DTT) and alkylated using 40 mM iodoacetamic acid (IAM). A mixture of hydrophilic and hydrophobic magnetic beads was used to clean-up the proteins at a ratio of 20:1 beads:proteins (Sera-Mag Speed beads, Fisher Scientific). After addition of ACN to a final concentration of 50%, the beads were allowed to bind to the proteins for 18 minutes. Protein-bead mixtures were washed twice with 80% EtOH and once with 100% ACN. The protein-bead complexes were digested with a mixture of trypsin:LysC (Promega) at a 1:20 ratio overnight at 37°C. Extracted peptides were cleaned-up using automated C18 solid phase extraction on the Bravo AssayMAP platform (Agilent Technologies).

NanoLC-MS-MS analysis

The peptide extracts were analysed on a nanoLC-TimsTof Pro coupling (Bruker Daltonics). The peptides (200ng) were separated on an IonOpticks column (25cm X 75µm 1.6µm C18 resin) using a gradient of 2-37% B (2%ACN, 0.1%FA) in 100 minutes at a flow rate of 0.3 µl/min. The dual TIMS had a ramp time and accumulation time of 166 ms resulting in a total cycle time of 1.89s. Data was acquired in Data Dependent Acquisition-Parallel Accumulation Serial Fragmentation (DDA-PASEF) mode with 10 PASEF scans in a mass range from 100 m/z to 1700 m/z. The ion mobility scan range was fixed from 0.7-1.25 Vs/cm². All samples were injected using a randomized injection sequence. To minimize carry-over, one solvent blank injection was performed after each sample.

Data interpretation

Raw files were converted to .mgf peaklists using DataAnalysis (version 5.3, Bruker Daltonics) and were submitted to Mascot database searches (version 2.5.1, MatrixScience, London, UK) against a Danio rerio protein sequence database downloaded from UniProtKB-SwissProt (2022_05_18, 61 732 entries, taxonomy ID: 7955), to which common contaminants and decoy sequences were added. Spectra were searched with a mass tolerance of 15 ppm in MS mode and 0.05 Da in MS/MS mode. One trypsin missed cleavage was tolerated. Carbamidomethylation of cysteine residues was set as a fixed modification. Oxidation of methionine residues and acetylation of proteins’ n-termini were set as variable modifications. Identification results were imported into the Proline software (version 2.1.2, http://proline.profiproteomics.fr) (Bouyssié et al., 2020) for validation. Peptide Spectrum Matches (PSM) with pretty ranks equal to one were retained. False Discovery Rate (FDR) was then optimized to be below 1% at PSM level using Mascot Adjusted E-value and below 1% at protein level using Mascot Mudpit score. For label free quantification, peptide abundances were extracted without cross assignment between the samples. Protein abundances were computed using the sum of the unique peptide abundances normalized at the peptide level using the median.

To be considered, proteins must be identified in at least four out of the five replicates in at least one condition. Imputation of the missing values and differential data analysis were performed using the open-source ProStaR software (Wieczorek et al., 2017). Imputation of missing values was done using the approximation of the lower limit of quantification by the 2.5% lower quantile of each replicate intensity distribution (“det quantile”). A Limma moderated t-test was applied on the dataset to perform differential analysis. The adaptive Benjamini-Hochberg procedure was applied to adjust the p-values and FDR. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository (Perez-Riverol et al., 2022), with the dataset identifier PXD042283.

Embryos drug treatment

27 hpf embryos from dnaaf1^tm317b/+ incross were dechorionated manually. N-acetyl cysteine ethyl ester (NACET) (BIOLLA Chemicals #59587-09-6) was prepared extemporaneously at the final concentration of 3 mM in pure water (pH: 7.2; conductivity 650 µS). Embryos were treated between 27 hpf to 60 hpf and embryos were anesthetized before being imaged laterally.

Juvenile drug treatment and preparation of treated samples

Two hundred larvae were housed off-system in 1.8 liter tanks with 20 fish per tank, being fed twice a day throughout the experiment. 100 were treated with NACET (BIOLLA Chemicals #59587-09-6), which was prepared extemporaneously at 1.5 mM (286.5mg/L) in fish water (pH 7.2, Conductivity 650 µS) which was changed once per day. Fish were treated from 30 dpf to 84 dpf and monitored for curvature onset once per week. At 85 dpf, fish were euthanized, imaged to measure the curvature index. Half of the fish were fixed for histology analysis in zamboni fixative [35 ml PFA 4 %, 15 ml saturated picric acid (1.2 %)] overnight at 4°C under agitation, while the other half was processed for RT-qPCR analysis. To prepare dorsal trunk RNA, internal organs were removed in cold PBS 1X, and the remaining tissue was cut in small pieces before lysis in QIAzol (QIAGEN). Homogenization was achieved using plunger pistons and passages through 1ml syringes.

Transgenesis

The Tol2 -5.2foxj1a:5xmyc-RPGRIP1L plasmid was produced using the Gateway system by combining four plasmids. P5E-foxj1a enhancer was kindly donated by [3]. 5xmyc-RPGRIP1L cDNA was amplified from the plasmid pCS2-5xmyc-RPGRIP1L (Mahuzier et al., 2012) using CloneAmp HiFi PCR Premix (Takara # 639298) using primers : forward 5’- GGGGACAAGTTTGTACAAAAAGCAGGCTGCAGGATCCCATCGATTTAAAGCT-3’ and reverse 5’ GGGGACCACTTTGTACAAGAAAGCTGGGTCTCCGAGCTCGGTTCAAGCCTCCAAGTCATCTCTGT-3. The PCR product was gel purified using QIAEX II gel extraction kit (QIAGEN, #20021) . BP recombination (Invitrogren “Gateway BP Clonase II Enzyme Mix” #11789-020) was performed into pDONR221 #218 to generate pMe- 5xmyc-RPGRIP1L. The 3’ entry polyadenylation signal plasmid was the one described in [61]. The 3 plasmids were shuttled into the backbone containing the cmcl2:GFP selection cassette (Kwan et al., 2007) using the LR recombination (Invitrogen “Gateway LR Clonase II Plus Enzyme Mix” #12538-120). rpgrip1l^+/- X AB embryos outcrosses were injected at the one cell stage with 1nl of a mix containing 20 ng/µl plasmid and 25 ng/µl Tol2 transposase RNA and screened at 48 hpf for GFP expression in the heart. Some F0 founders produced F1 zebrafish carrying one copy of the transgene that were further checked for RPGRIP1L specific expression in foxj1a territory, using Myc labelling at the base of FP cilia in 2,5 dpf embryos.

The Tol2-1.7kb col2a1a:5xMyc-RPGRIP1L plasmid was produced using the Gateway System by combining four plasmids. Col2a1a enhancer was amplified from the plasmid −1.7kbcol2a1a:EGFP-CAAX (Dale and Topczewski, 2011) using CloneAmp HiFi premix (Takara #639298) with the primers: forward 5’-GGG GAC AAC TTT GTA TAG AAA AGT TGG CCC TCT GAC ACC TGA TGC CAA TTG C-3’ and reverse 5’- GGG GAC TGC TTT TTT GTA CAA ACT TGC TTG CAG GTC CTA AGG GGT GAAAGT CG-. The PCR product was gel purified and BP recombination was performed into pDONR_P4-P1R-#219. The middle entry plasmid and the 3’ plasmid were the same as those used to generate the Tol-5.2foxj1a: 5xmyc-RPGRIP1L plasmid. F1 transgenic were first selected for GFP expression in the heart and then on Myc labelling within notochordal cells at 2,5 dpf within their embryonic progeny.

Skeletal preparations, Alizarin staining and imaging

Animals were euthanized using a lethal concentration of MS222 (0.28 mg/mL), skin was removed and fixation was performed with 4% paraformaldehyde overnight at 4°C. After evisceration, samples were incubated in borax 5%, rinsed several times in 1% KOH and incubated in solution composed with Alizarin 0.01% (Sigma, A5533) and KOH 1% during 2 days. Fish were rinsed several times in KOH 1% and incubated in trypsin 1% and borax 2% for 2 days until cleared. Samples were rinsed several times in distilled water and transferred in glycerol 80% (in KOH 1%) using progressive dilutions. Samples were kept in glycerol 80% until imaging.

Cobb angles measurements

To quantitatively evaluate the severity of spine curvature in rpgrip1l^+/-; urp2^+/- incrosses, we used the zebrafish skeletal preparations stained with Alizarin. We drew parallel lines to the top and bottom most displaced vertebrae on lateral and dorsal views. The Cobb angle was then measured as the angle of intersection between lines drawn perpendicular to the original 2 lines. This was conducted using FIJI software. For each fish, the total Cobb angle was calculated by summing all Cobb angles.

Curvature index measurements

We implemented a novel procedure to quantify curvatures on both axes by drawing a line along the body of the fish as shown in Figure supplementary 4I and curvature was calculated in MATLAB (code available on demand). We decomposed the line in a series of equidistant points and we measured the curvature at each of these points using the LineCurvature2D function. We then added the absolute values of these measures, which are expressed as an angle (in radian) per unit of length. The higher the sum is for a given line, the more this line is curvated. We verified the accurancy of this metric by comparing it with the visual assessment of curvatures by 3 independent observers. For rpgrip1l^+/-; urp2^+/- incross and NACET experiment, a line following the body deformation was drawn from the mouth to the base of the tail following the midline of the fish in lateral position and another one along the dorsal axis. Curvature index of both curves were summed for each fish. Curvature analysis was performed blinded to fish genotype. rpgri1l^+/+ or rpgrip1l^+/- siblings were used as controls.

Data acquisition for body-curvature analysis at embryonic stage

Zebrafish embryos were anesthetized and imaged laterally with the head pointing to the left. The angle between the line connecting the center of the eye to the center of the yolk and the line connecting the center of the yolk to the tip of the tail was measured to evaluate the body curvature of the embryos. For quantitative analysis, the angles of each embryo were put in the same concentric circles represented with a Roseplot, with 0° pointing to the right and 90° pointing to the top. Each triangle represents a 30° quadrant, and its height indicates the number of embryos within the same quadrant.

Quantification, statistical analysis and figure preparation

For all experiments the number of samples analyzed is indicated in the text and/or visible in the figures. Statistical analysis was performed using the Prism software. ****: P value < 0.0001; ***: P value < 0.001; **: P value < 0.01; *: P value < 0.05. Graph were made using Prism and Matlab and figures were assembled using Photoshop software.

Reagent and resource table

Supplementary figure legends

Supplementary figure S1:

Supplementary figure S2:

Supplementary figure S3:

Supplementary Figure S4:

Supplementary figure S5:

Supplementary figure S6:

Supplementary Movie 1: Micro-CT scans of rpgrip1l^Δ/Δ juvenile vertebral axis compared to controls

Superposition of 3D-reconstructed spines of 5 wpf juvenile fish, 1 control (white), 3 rpgrip1lΔ/Δ of different severities: 2 tail-up (yellow, blue) and 1 scoliotic (pink)