Ependymal polarity defects coupled with disorganized ciliary beating drive abnormal cerebrospinal fluid flow and spine curvature in zebrafish

Idiopathic scoliosis (IS) is the most common spinal deformity diagnosed in childhood or early adolescence, while the underlying pathogenesis of this serious condition remains largely unknown. Here, we report zebrafish ccdc57 mutants exhibiting scoliosis during late development, similar to that observed in human adolescent idiopathic scoliosis (AIS). Zebrafish ccdc57 mutants developed hydrocephalus due to cerebrospinal fluid (CSF) flow defects caused by uncoordinated cilia beating in ependymal cells. Mechanistically, Ccdc57 localizes to ciliary basal bodies and controls the planar polarity of ependymal cells through regulating the organization of microtubule networks and proper positioning of basal bodies. Interestingly, ependymal cell polarity defects were first observed in ccdc57 mutants at approximately 17 days postfertilization, the same time when scoliosis became apparent and prior to multiciliated ependymal cell maturation. We further showed that mutant spinal cord exhibited altered expression pattern of the Urotensin neuropeptides, in consistent with the curvature of the spine. Strikingly, human IS patients also displayed abnormal Urotensin signaling in paraspinal muscles. Altogether, our data suggest that ependymal polarity defects are one of the earliest sign of scoliosis in zebrafish and disclose the essential and conserved roles of Urotensin signaling during scoliosis progression.

Introduction Idiopathic scoliosis (IS), characterized by the abnormal rotation and curvature of the spine, is the most common spinal deformity, affecting more than 3% of children and adolescents worldwide [1]. Although more than 80% of scoliosis cases are deemed idiopathic, it is believed that genetic factors make significant contributions to the progression of the disease, based on the high incidence of scoliosis in families and twins. Currently, the pathogenesis of IS remains largely unknown due to insufficient knowledge of its etiology and subsequent disease progression.
severe scoliosis during late development [8]. These works suggest that defects in the Urotensin signaling pathway contribute to scoliosis formation. The Urotensin signaling pathway appears to be conserved in other vertebrates [24], and mutations in UTS2R, the human homolog of zebrafish uts2r3, is also associated with human scoliosis [25].
CSF is produced by the CP, a highly specialized epithelium located in the ventricles of the brain that are in close contact with ependymal cells [26]. A key feature of brain ventricle ependymal cells is the presence of multiple motile cilia in their apical surface, which need to beat in the same direction to properly propel CSF flow [27]. Planar cell polarity (PCP) signals are essential to define the distribution of these cilia and ensure the proper direction of ciliary beating [28][29][30]. Of note, ependymal cells display two types of planar polarity-rotational PCP (rPCP) and translational PCP (tPCP)-based on the orientation and positioning of basal body clusters located within cells and tissues [31,32]. Motile cilia are essential for regulating rPCP, while tPCP is established by primary cilia of radial glial cells during differentiation [28,31]. Defects in ependymal cells, including polarity defects, are often associated with hydrocephalus caused by abnormal CSF circulation [29,33].
Congenital hydrocephalus is a common phenotype that occurs in several human disorders including PCD [34]. Intriguingly, PCD patients also exhibit a high prevalence of scoliosis [35], although it remains unclear how hydrocephalus may result in the development of scoliosis. Similarly, hydrocephalus and scoliosis occur in many zebrafish ciliary mutants. Interestingly, spinal curvature did not develop in these ciliary mutants until approximately 3 weeks postfertilization, a similar stage to the beginning of scoliosis in adolescent idiopathic scoliosis (AIS) patients. The molecular mechanisms of scoliosis development at these stages remains to be elucidated. Here, we have characterized a late-onset zebrafish scoliosis mutant exhibiting loss of function of Ccdc57. Zebrafish ccdc57 mutants develop severe hydrocephalus due to defects in the coordinated beating of multiple cilia. We provide data showing that Ccdc57 regulates ependymal PCP, whose defects are likely the major cause of scoliosis formation in ccdc57 mutants. Moreover, we describe the relationship between spine curvature and abnormal Urotensin expression caused by abnormal CSF circulation, thereby providing important mechanistic clues for the formation of scoliosis.

Mutation of ccdc57 results in scoliosis in zebrafish
In an ENU-based screen for zebrafish mutants with mineralized craniofacial and skeletal tissue defects, we identified tft 168N mutants that displayed severe spinal curvatures in adults ( Fig 1A). Both Micro-CT and Alizarin red staining showed that tft 168N mutants displayed abnormal three-dimensional curvatures and deformities of the spinal vertebrae when viewed from both dorsal and lateral positions (Figs 1B and S1A). Interestingly, the severity of the spinal curvatures was comparable in male and female zebrafish mutants, in that the cobb angles measured from dorsal-ventral and medio-lateral curvature showed similar distribution patterns (S1B-S1E Fig). To identify the gene responsible for this phenotype, we performed Next Generation sequencing and identified a mutant locus at the genomic region encoding the ccdc57 gene. The ccdc57 gene contains 19 exons encoding 979 amino acids, and the tft 168N allele introduced a stop codon in the 11th exon of ccdc57 resulting in the truncation of the final 407 C-terminal amino acids (Fig 1C). To further validate this gene mutation, we generated ccdc57 mutants via CRISPR/Cas9 methods and recovered two mutant alleles, one with a 2-bp insertion (+2) and another with a 7-bp deletion (Δ7) in the first exon ( Fig 1C). Both of these two ccdc57 mutant alleles displayed scoliosis ( Fig 1D). Moreover, complementation testing between the Δ7 and tft 168N alleles showed that these two alleles failed to complement each other, confirming that mutations in ccdc57 were causative for scoliosis in these mutants (Fig 1D). Of note, when dissected from fixed vertebrae, ccdc57 mutant spinal cords also displayed curvature defects that closely mimicked those of the spinal curvatures ( Fig 1E).
To define the progression of scoliosis in ccdc57 zebrafish mutants, we conducted developmental analyses of the mutant phenotype. Zygotic ccdc57 Δ7 mutants displayed body curvature at 3 days postfertilization (dpf) (S2A Fig). Interestingly, body curvature was not apparent in 3 dpf ccdc57 tft168N mutants (S2A Fig). Such disparity may be due to the presence of truncated forms of Ccdc57 proteins in the ccdc57 tft168N mutants, as the tft 168N allele encodes the N-terminal 572 amino acids of Ccdc57 ( Fig 1C). Moreover, maternal Ccdc57 protein may contribute to early embryonic development. Indeed, maternal zygotic (MZ) ccdc57 tft168N mutants

PLOS BIOLOGY
displayed body curvature defects at both 3 and 5 dpf, similar to those of MZ ccdc57 Δ7 mutants (S2A Fig). In the following studies, we focused on the ccdc57 Δ7 mutant allele to evaluate the function of Ccdc57 in spinal development and scoliosis.
In zebrafish, ciliogenesis defects are a major cause of body curvature development [7]. We therefore examined cilia development in ccdc57 mutants. Surprisingly, all of the cilia examined were grossly normal in the ccdc57 mutants as visualized using anti-glycylated tubulin antibody (S2B- S2F Fig). Although abnormal notochord differentiation can contribute to the development of congenital scoliosis, we did not detect any notochord defects in ccdc57 mutants at early stages, suggesting an independent mechanism of scoliosis progression in the absence of Ccdc57 (S2G Fig). In fact, scoliosis was first detected at approximately 17 dpf in both ccdc57 tft168N and ccdc57 Δ7 mutants, with initial bending of the mutant spine occurring in the middle portion of the trunk (S2H and S2I Fig).

Hydrocephalus in ccdc57 mutants
Recently, abnormal CSF flow has been linked to the development of scoliosis [6,8]. Therefore, to better assess the etiology of scoliosis progression in ccdc57 mutants, we dissected whole brains from ccdc57 mutants and wild type sibling controls. The ccdc57 mutant brain was larger and appeared transparent due to an expanded ependymal epithelium filled with CSF (Fig 2A  and 2B). Histological analysis of cross-sectioned brains revealed the presence of dilated ventricles in the optic tectum and rhombencephalon of ccdc57 mutants, as well as in the central spinal canal (Fig 2C-2E). In addition, micro-CT analysis confirmed the presence of dilated ventricles in ccdc57 mutant brains ( Fig 2F).
As described, CSF is produced by the CP. Therefore, we further examined the CP in wild type and ccdc57 mutants focusing on the epithelial monolayer connecting the telencephalon and optic tectum known as the forebrain CP or diencephalic CP (dChP) [36][37][38] (Fig 2G). Three-dimensional reconstructions of the dChP in wild type zebrafish revealed a chapeau-like structure with multiple folds covering the brain tissues ( Fig 2H and S1 Movie). In contrast, ccdc57 mutant dChP appeared stretched and lacked foldings, due to excess CSF accumulation ( Fig 2I and S2 Movie). Next, we investigated whether ccdc57 mutant larvae showed signs of hydrocephalus. Injection of Rhodamine-or FITC-conjugated fluorescent beads into 2 or 3 dpf zebrafish larvae revealed no obvious differences in the brain ventricle size between mutants and siblings (S3 Fig). Together, these data suggested that, similar to the late appearance of scoliosis, ccdc57 mutants developed hydrocephalus at later stages of development.

Planar polarity defects of cilia and basal bodies in ccdc57 mutant ependymal cells
Excess accumulation of CSF in cerebral ventricles is one of the major causes of hydrocephalus. In zebrafish, directed CSF flow is facilitated by the coordinated beating of motile cilia in multiciliated ependymal cells lining the brain ventricles. Focusing on the telencephalon and dChP regions, we identified multiciliated cells largely restricted to the central portion of the ependymal layer (S4 Fig). Next, we monitored motile cilia beating in these ependymal cells. In adult wild type zebrafish, all motile cilia bent to the direction of the fluid flow, and multicilia bundles beat synchronously (Fig 3A and S3 Movie). In contrast, multicilia bent in a variety of directions in ccdc57 mutants (Fig 3A and S4 Movie). Even within the same ciliary bundle, individual cilia beating appeared disorganized (S4 Movie). We next performed scanning electron microscopy (SEM) analyses to visualize the ultrastructure of the ependymal cilia bundles. Compared to the clustered distribution in wild type fish, ciliary bundles appeared highly disorganized in ccdc57 mutants (Fig 3B). Aberrant distribution of these ependymal cell ciliary bundles was further confirmed via immunostaining with anti-glycylated tubulin antibody to visualize cilia ( Fig 3C).
Cilia are anchored to the surface of ependymal cells through basal bodies. We therefore investigated basal body localization in wild type and ccdc57 mutant ependymal cells. In wild type zebrafish, basal bodies appeared localized to the same side of each ependymal cell (Fig 3D  and 3E). In contrast, basal bodies were randomly distributed in ccdc57 mutants, with many localized to the center region of ependymal cells, indicating polarity defects in the absence of Ccdc57 (Fig 3D-3F). Notably, the number of basal bodies was comparable between mutant and wild type ependymal cells ( Fig 3G). Next, we characterized the basal bodies of ependymal cells inside the ChP. In wild type zebrafish, multiciliated ependymal cells mainly localized to the center of the folds of the ChP (S5A and S5B Fig and S5  Together, these data suggested that loss of Ccdc57 resulted in abnormal distribution of multicilia in the ependymal cells, together with basal body planar polarity defects (Fig 3H).

Cell polarity defects of ccdc57 mutant ependymal cells
We next asked whether ependymal cell polarity was affected in the absence of Ccdc57. Since basal bodies rely on the cytoskeleton microtubule network to localize cilia, the misplacement of basal bodies in ccdc57 mutants may be due to abnormal organization of cytoskeletal microtubules. We therefore examined the cytoskeleton using alpha tubulin antibody staining and found that the microtubule skeleton appeared disorganized in ccdc57 mutants as compared to the highly polarized microtubule network in wild type cells (S6A and S6B Fig). The position of mitochondria also depends on the microtubule network, we further analyzed the distribution of ependymal cell mitochondria using Tom20 staining. Our results showed that ccdc57 mutant ependymal cells displayed abnormal mitochondria distribution patterns as compared with those of wild type cells (S6C and S6D Fig).
We next examined the shape of ependymal cells via Claudin-5 antibody-labeled cellular tight junctions. At 3 months postfertilization (mpf), wild type zebrafish ependymal cells appeared symmetrically distributed throughout the ependymal layer, with narrow and elongated cells located in the center and larger round cells located laterally (Fig 4A and 4B). Moreover, the elongated cell axes were parallel to the anterior-posterior axis, indicating polarity of these cells ( Fig 4C). Intriguingly, the symmetrical distribution pattern of ependymal cells was also found to be established in 17 dpf wild type zebrafish larvae (Fig 4D-4F). In contrast, polarized ependymal epithelial cell organization was severely disrupted in 3 mpf ccdc57 mutants (Fig 4B and 4C). At 17 dpf when spine curvatures are first apparent, ccdc57 mutants also exhibited abnormal-shaped ependymal cells in the central region, suggesting an earlier defect of ependymal cell polarity (Fig 4E and 4F). Notably, since multiple motile cilia formation does not occur until 1 mpf (Fig 4G) [38], these results suggested that Ccdc57 participates in the regulation of ependymal cell polarity in early zebrafish development, as well as in later developmental basal body positioning.

Ccdc57 orchestrates the synchronized beating of motile cilia in the spinal canal
During larval zebrafish development, ependymal cells are derived from radial glial cells, which also are involved in establishing their translational cell polarity [31]. Therefore, we next asked whether radial glial cilia were defective in the absence of Ccdc57. At larvae stages, spinal canal ependymal cells are a specialized type of radial glia harboring primary motile cilia that drive

PLOS BIOLOGY
CSF flow. In ccdc57 mutants, the number and length of spinal canal cilia were comparable to those of wild type larvae (S2E and S2F Fig). Further examination of cilia localization by immunostaining with anti γ-tubulin antibody showed that both wild type and ccdc57 mutant cilia were localized to the posterior apical surface of each radial glia cell in the spinal canal, with no apparent differences in basal body localization (Fig 5A and 5B).
We next examined the beating pattern of these spinal canal cilia using high speed video microscopy. By imaging cilia motility in the caudal central canal of 5 dpf wild type larvae, we found that all of the cilia beat in a similar manner, and a synchronized beating wave was easily observed throughout the field of view (Fig 5C and S7 Movie). However, the orchestrated ciliary beating was disrupted in ccdc57 mutant spinal cords, with individual cilia beating independently of one another (Fig 5C and S8 Movie). Of note, the beating frequency of individual cilia was comparable between ccdc57 mutant and wild type larvae ( Fig 5D). Ciliary beating angle measurements demonstrated that the beating angle increased significantly in ccdc57 mutants as compared to wild type siblings ( Fig 5E). We further used the bisector of each angle to evaluate the tilting direction of motile cilia and found that the tilting angles were significantly increased in ccdc57 mutants (cat. 72.11˚± 14.16˚in wild type versus 85.52˚± 5.06˚in mutants) (Fig 5F and 5G).
To further explore the beating defects of spinal canal cilia, we performed fluorescent bead tracing experiments at different stages of zebrafish development. By monitoring the movement of fluorescent beads injected into the central canal, we observed bidirectional CSF flow, as shown previously, in the central canal of wild type larvae at both 30 hpf and 3 dpf [39,40] (S9 and S11 Movies). The bidirectional CSF flow was largely maintained in ccdc57 mutants, although residential circular particle movements were also observed (S10 and S12 Movies). Strikingly, we noticed that the injected fluorescent beads were able to be transported to the end of the central canal by 6 hours postinjection (hpi) in the majority of wild type larvae, while none of the ccdc57 mutants contained fluorescent beads in the caudal spinal canal, and the fluorescent bead transport distances were significantly reduced (S7 Fig). The bead transport defects were observed in ccdc57 mutants at all stages analyzed (3 dpf, 5 dpf, and 17 dpf) (S7 Fig). Together, these data suggest that Ccdc57 orchestrates the synchronized beating of spinal canal motile cilia, whose deficiency leads to abnormal CSFflow.

Localization of core PCP components in ccdc57 mutants
The posterior tilting of spinal canal motile cilia is mainly regulated by the PCP pathway [41]. To further characterize this phenomenon, we examined the localization of two major PCP pathway components, Prickle and Dishevelled 1 (Dvl1), in wild type and ccdc57 mutant fish. Similar to their basal bodies, Dvl1 appeared localized to the apical posterior region of each radial glia cell ( Fig 5H), with Dvl1-positive vesicles localized to the region surrounding the basal bodies ( Fig 5H). Noticeably, the Dvl1 distribution angles were comparable between ccdc57 mutants and wild type control embryos (Fig 5I and 5J). The localization of Prickle was opposite to that of Dvl1 vesicles, in the anterior apical surface, as demonstrated by GFP-Prickle labeling. The Prickle localization was similar in wild type and ccdc57 mutants ( Fig 5K). These data suggested that localization of PCP components in larval zebrafish was not affected by the absence of Ccdc57. Intriguingly, in adult zebrafish, Dvl1 expressing vesicles appeared localized to one side of mature ependymal cells, corresponding to the location where multicilia formed ( Fig 5L). In contrast, in adult ccdc57 mutants, Dvl1 expressing vesicles appeared dispersed randomly throughout the entire cell, clearly indicating a late onset cell polarity defect in ccdc57 mutants ( Fig 5L).

PLOS BIOLOGY
Ependymal polarity defects drive spinal curvature in zebrafish

CCDC57 encodes a centrosomal satellite protein required for cell polarity
To reveal the mechanisms of Ccdc57 regulation of basal body positioning, we further examined the subcellular localization of CCDC57 in RPE-1 cells. Similar to previous reports, CCDC57 mainly localized to centriolar satellites in RPE-1 cells [42]. Flag-tagged CCDC57 protein colocalized with centriolar satellite proteins OFD1 and PCM1 and did not localize to cilia (ARL13B) or to the distal appendage of the basal body (CEP164) (Fig 6A-6C). Mass spectrometry analysis of pulled-down proteins using Anti-FLAG conjugated beads showed that CCDC57 interacted with multiple centrosomal proteins, including CEP170, OFD1, and CP110 (Fig 6D), and we further validated the CCDC57-OFD1 interactions via immunoprecipitation ( Fig 6E). Moreover, siRNA knockdown of OFD1 eliminated localization of CCDC57 to the centriolar satellite. In contrast, siRNA knockdown of CCDC57 gene expression had no effect on centrosomal localization of OFD1( Fig 6F).
During migration, cells can establish a front-rear polarity characterized by the polarized distribution of Golgi and centrosomes in the leading edge [43]. We therefore further tested the role of CCDC57 in establishing cell polarity in RPE-1 cells by wound-scratch assay. After scratching, the microtubule cytoskeleton of leading edge RPE-1 cells appeared polarized, with the Golgi facing toward the scratched space to direct cell migration. In control siRNA-treated RPE-1 cells, the majority of leading edge cells contained Golgi apparatus located within 60 degrees of the direction of migration relative to the nucleus (Fig 6G-6I). In contrast, the polarized orientation of Golgi apparatus was significantly compromised in CCDC57 siRNA-treated RPE-1 cells, as demonstrated by significantly increased orientation angles (Fig 6G-6I). Furthermore, migration distance was also reduced in CCDC57 siRNA-treated cells as compared to control siRNA-treated cells (Fig 6J and 6K). Together, these in vitro studies further confirmed the role of the centrosomal protein, CCDC57, in establishing cell polarity.

Abnormal RF assembly and urotensin expression in ccdc57 mutant larvae
We next sought to identify the causes of the body curvature observed in ccdc57 mutants. Epinephrine signals are essential for urotensin expression and body straightening [8]. In line with this, epinephrine treatment was also able to rescue body curvature in ccdc57 mutant embryos (S8 Fig). The RF is essential for body axis straightening through transferring the epinephrine signals to the CSF-cNs [20,21]. We found that wheat germ agglutinin (WGA) can be used to label and image the RF (S9A- S9C Fig). In various ciliary mutants, WGA-labeled RF appeared either discontinuous or absent (S9D- S9G Fig). Similarly, WGA staining of the RF was also diminished or absent in ccdc57 mutant larvae ( Fig 7A). Interestingly, body curvature severity closely correlated with the severity of RF assembly defects in ccdc57 or kif3a mutants (S9H- S9M Fig). Next, we examined the expression of urotensins in ccdc57 mutant larvae using whole-mount in situ hybridization (WISH) analysis. While urp1 expression appeared relatively normal in the anterior trunk of ccdc57 mutant larvae, the number of urp1-expressing cells was markedly decreased in the posterior trunk (Fig 7B and 7C). Together, these data suggested that Ccdc57 deficiency interrupts assembly of the RF and down-regulates the expression of urotensin genes. The fact that differences in urotensin gene expression mainly occurred in the posterior trunk may explain why ccdc57 mutant larvae exhibit only mild body curvature.

Ectopic accumulation of Urotensin neuropeptides in ccdc57 adult mutants
Next, we attempted to discover the relationship between the development of scoliosis and CSF flow defects. First, we investigated the assembly of the RF in wild type and ccdc57 mutants, based on the previously characterized roles for the RF in regulating body axis development [10,19,22]. In wild type adults, the RF appeared thick and straight (Fig 7D). In contrast, the RF appeared much thinner, discontinuous, and/or absent in adult ccdc57 mutants (Fig 7D). Of note, the severity of RF assembly defects correlated with the level of scoliosis in the mutants (S10A and S10B Fig). Next, we focused on the expression of urotensin neuropeptides. Intriguingly, we found that from a lateral view, adult ccdc57 mutant spines always contained a strong dorsal bending in the anterior trunk and a second dorsal bending in the tail region (Fig 1B and  1E). To better characterize this feature, we dissected adult wild type and ccdc57 mutant trunks into three segments (Head, Middle, and Tail) and isolated total RNA from each segment ( Fig  7E). Unexpectedly, qPCR results showed that the expression of urp1 and urp2, the major urotensins regulated by CSF signaling, was increased over approximately 10 times in the Head segments of ccdc57 mutants as compared with those of wild type control siblings (Fig 7E). In contrast, Middle trunk segments displayed lower expression levels of these genes in ccdc57 mutants, while urotensin expression again appeared up-regulated in Tail segments of ccdc57 mutant adult spines (Fig 7E). To further validate these results, we performed ISH analysis on dissected spinal cords for the expression of urp2, one of the major urotensin genes expressed at later stages. Our results showed that the expression of urp2 was dramatically increased in the brainstem region of ccdc57 mutants (Fig 7F). In contrast, urp2 expression was virtually absent in the middle part of the spine, with some urp2 expression observed in the tail region (Fig 7G and 7H). To further examine whether urp2 expression in the tail region correlated with the second dorsal bending of the spine, we dissected the spinal cord at the second bending site and further cut it at the apex into anterior and posterior fragments (S10C Fig). Interestingly, compared with wild type control siblings, urp2 gene expression was up-regulated in the posterior fragment of the dissected spinal cord, while urp2 was not detected in the anterior fragment (S10C Fig). Together, these results suggest an interesting relationship between urp2 expression and spinal curvature (Fig 7I).

Correlation between urotensin expression and spine curvature in ciliary mutants
To further gauge the relationship between urotensin expression and spinal curvature, we evaluated spine curvature phenotypes in several zebrafish scoliosis mutants. Both tmem67 and ofd1 mutants displayed scoliosis. Micro-CT results showed that these mutants also displayed initial dorsal bending in the anterior spine (Fig 8A). The observed scoliosis in ofd1 and ccdc57 mutants further demonstrated the interactions between Ofd1 and Ccdc57 that were revealed by our IP pulldown experiments (Fig 6E). Of note, qPCR results confirmed the enhanced expression of urp1 and urp2 in the anterior spines of ccdc57, tmem67, and ofd1 mutants (Fig 8B).
We have previously shown that mutation of uts2r3, the major Urotensin receptor, leads to scoliosis. Noticeably, the spinal curvature of uts2r3 mutants was clearly different from that of ciliary mutants. The anterior dorsal bending phenotype appeared relatively minor in uts2r3

PLOS BIOLOGY
Ependymal polarity defects drive spinal curvature in zebrafish mutants, while all uts2r3 mutants bent to the ventral side first, and displayed strong dorsal bending in the posterior portion of the trunk (Fig 8A). To better characterize spinal bending, we measured the angle between the parasphenoid bone and the Weberian vertebrae orientation in wild type, ccdc57, and uts2r3 mutants (Fig 8C). The measured angles of ccdc57 mutants were significantly larger than those in control and uts2r3 mutants (Fig 8C-8E). Consistent with these results, urp2 expression levels were dramatically increased in the anterior spines of ccdc57 mutants, while only a slight increase was observed in uts2r3 mutants (Fig 8B), as validated via WISH assay (Fig 8F). Noticeably, the location of increased urp2 expression in uts2r3 mutant spines also correlated with the observed second bending in uts2r3 mutant spines ( Fig  8G and 8H). Together, these data provide strong evidence that spinal bending closely correlates with the activation of Urotensin signaling.

Abnormal urotensin signals in idiopathic scoliosis patients
Finally, we sought to investigate whether urotensin signaling is also involved in the regulation of spinal curvature in human scoliosis patients. Although it is difficult to obtain scoliosis patient spinal cord tissue for analysis, we were able to collect bilateral paravertebral muscle tissue from scoliosis patients during surgery. We then compared Urotensin signaling pathway gene expression in paravertebral muscle tissue harvested from the convex and concave sides of spinal curvature sites ( Fig 9A). Strikingly, although UTS2 expression appeared similar between convex and concave muscle tissue locations, we observed a remarkably asymmetric expression of UTS2R in bilateral paravertebral muscles of AIS patients (n = 46) (Fig 9B and 9C). The expression of UTS2R was significantly higher in convex side as compared to concave side muscle tissue (Fig 9B). According to the ratio of UTS2R expression in the convex versus the concave (convex/concave), we classified these AIS patients into two groups with the ratio of 2 as cutoff point. Strikingly, patients with >2-fold difference in expression (n = 26) had remarkably severe curvature magnitude as compared to patients with <2-fold difference (n = 20) (58.62 ± 9.48 degrees versus 51.35 ± 6.58 degrees) (Fig 9D and 9E). Thus, these data strongly support the hypothesis that abnormal Urotensin signaling may make a significant contribution to the severity of spinal curvature observed in AIS patients.
In summary, here we showed that Ccdc57 is a centrosomal protein required for the proper establishment of basal body and cilia polarity. Loss of Ccdc57 disrupted the planar polarity of ependymal cells, affected the polarity and beating pattern of cilia present in both radial glial and ependymal cells, and led to CSF flow defects. The CSF flow defects resulted in the up-regulation of Urotensin signals in the anterior portion of the spine, eventually leading to spinal dorsal bending and scoliosis.

Discussion
Scoliosis is one of the most common diseases diagnosed in childhood or early adolescence, although the underlying causes remain largely unknown. Recent studies suggest that in addition to environmental factors, underlying genetic factors also contribute to the incidence of spinal curvatures. The coincidence of scoliosis and hydrocephalus in several human genetic disorders suggests a potential relationship between hydrocephalus and spinal curvature development. Similarly, hydrocephalus is constantly observed in zebrafish ciliary scoliosis mutants. Remarkably, the curvature of the spine is not apparent until approximately 3 weeks fertilization in virtually all zebrafish ciliary mutants, which points to an interesting question: Why does abnormal spine curvature develop at these stages? In this paper, we present data showing that deficiency of the centrosomal protein, Ccdc57, results in scoliosis in zebrafish. Our data suggest that reduced Ccdc57 expression results in ependymal cell polarity defects as early as 17 dpf, which, in turn, affect the coordinated beating of ependymal cell multicilia and CSF flow. Moreover, the subsequent accumulation of CSF causes up-regulation of the Urotensin expression in the anterior spine, which contributes to the initial spine curvature formation.

PLOS BIOLOGY
In mouse, multiciliated ependymal cells display planar polarity in three distinct levels: the direction of cilia beating known as rotational polarity; the displacement of cilia via the basal body positioning known as translational polarity; and the alignment of ciliary basal bodies at the intercellular level known as tissue polarity [44]. Similarly, our data suggest that multiciliated ependymal cells also displayed these three types of polarities in zebrafish. In wild type zebrafish, most basal bodies appeared localized to the same side of individual ependymal cells. In ccdc57 mutants, both basal body placement and coordinated cilia beating were disrupted in ependymal cells. Noticeably, synchronized cilia motility can occur at the cellular (each individual cilium) and intercellular (cilia bundle) levels, and both were affected in the absence of Ccdc57. The beating direction of cilia is closely related to the orientation of the basal foot, an accessory structure projecting from the side of the basal body. Within each ependymal cell, the basal feet are aligned towards the direction of cilia beating [45]. Considering the subcellular localization of CCDC57, it is conceivable that Ccdc57 deficiency may lead to misalignment of ependymal cell basal feet. Moreover, the rotational orientation of cilia is also regulated by several PCP-core modules including Frizzled, Van Gogh-like (Vangl1/2), and Prickle. PCP of ependymal cells is first established through the asymmetric localization of these core PCP components via a polarized microtubule network, which further organizes the distribution of ciliary basal bodies in a PCP-dependent manner [32,[46][47][48]. Interestingly, while the distribution of the PCP proteins Dvl1 and Prickle was initially normal in the radial glial cells of larval mutants, adult ccdc57 mutants showed disorganized basal body distribution, suggesting that Ccdc57 may function downstream of PCP components to regulate basal body positioning. Finally, ccdc57 mutant ependymal cells displayed abnormal cell polarity characterized by microtubule polarity defects, which occurred before the maturation of multiciliated cells. These data suggest that organizational defects in the microtubule network were upstream of the observed abnormal basal body distribution in ccdc57 mutants. Ccdc57 may regulate the polarity of the microtubule network via its microtubule binding domain [42]. Since the basal foot also participates in the regulation of the polarity of microtubule network [49,50], Ccdc57 may also control the basal body polarity to organize the microtubule network. In the absence of Ccdc57, the microtubule network polarity is defective, resulting in later defects of abnormal cilia beating and basal body positioning.
Defects in motile cilia of the brain ventricles are associated with hydrocephalus in humans, mice, and zebrafish. While ccdc57 mutants displayed strong hydrocephalus and multiciliated cell defects, it is surprising that brain morphology appeared relatively normal in gmnc zebrafish mutants that lack multiciliated cells [38,51] (S11A-S11C Fig). Moreover, gmnc mutants did not exhibit scoliosis, suggesting that defective multicilia were not the prerequisite for CSF flow defects and the progression of scoliosis (S11D Fig). In contrast to ccdc57 mutants, coordinated single motile ciliary beating was relatively normal in gmnc mutants (S11E Fig and S13 and S14 Movies), which, together with heartbeat and body movements, ensured sufficient circulation of CSF in the CNS [52]. In line with this, the RF also developed normally in gmnc mutants (S11F Fig). In adult gmnc mutants, although multicilia were not differentiated as demonstrated by the localization of basal bodies, the polarity displacement of primary motile cilia remained largely normal (S11G Fig). Moreover, the distribution pattern of ependymal cells also appeared relatively normal, as compared to the severely disorganized pattern observed in ccdc57 mutants (S11H Fig). Therefore, it is likely that the single motile cilia in gmnc mutant ependymal cells are sufficient to drive CSF flow. In contrast, coordinated ciliary beating is essential to prevent the development of hydrocephalus. In ccdc57 mutants, the PCP defects resulted in dampened directional CSF flow due to lack of coordinated beating of individual cilium in the ependymal cells. The CSF flow defects further resulted in CSF accumulation in the brain ventricles, leading to hydrocephalus. It is noteworthy that hydrocephalus forms from the cumulative effects of defective CSF flow, as only a small number of ccdc57 mutant larvae exhibited hydrocephalus at 17 dpf when spinal curvature was initially observed (S12A and S12B Fig). In contrast, the width of spinal canal was significantly increased in the mutants starting from as early as 3 dpf (S12C and S12D Fig), an earlier sign of CSF flow defects.
In summary, we return to our question: How does hydrocephalus result in scoliosis? We found that ependymal cell polarity defects in ccdc57 mutants first occurred at approximately 17 days after fertilization, prior to the formation of differentiated multicilia. Interestingly, scoliosis was also first apparent at this developmental stage, suggesting a potential relationship between cell polarity defects and scoliosis. We have shown previously that the Urotensin signals govern zebrafish body axis straightening through activating its receptors located in dorsal muscle fiber cells [8]. Analyses of dissected antero-posterior portions of adult ccdc57 mutant spines revealed that the expression of Urotensin neuropeptides was significantly up-regulated in the head region. In ccdc57 mutants, CSF flow defects led to hydrocephalus and excess accumulation of CSF in the brain ventricles. The observed enhanced expression of urp2 in the hindbrain may be caused by up-regulation of epinephrine signaling in the brain ventricles due to CSF accumulation. Interestingly, we found that all ccdc57 mutants developed dorsal curvature in the anterior part of the spine. Since Urotensin signals can promote the contraction of dorsal muscle fibers, it is possible that the anterior-most, first dorsal curvature was the result of enhanced urp2 expression in this region. In fact, almost all reported zebrafish ciliary scoliosis mutants displayed a first dorsal curvature phenotype [6,22,53,54]. In contrast, uts2r3 mutants that lack urotensin receptors displayed minor dorsal bending in the anterior portion. Of note, the second dorsal bending of the ccdc57 mutants observed in the posterior portion of the trunk also corresponded to enhanced urp2 expression. It is likely that these secreted neuropeptides activate the contraction of muscle fibers locally and that the uneven distribution of Urotensin neuropeptides in ccdc57 mutant results in unbalanced muscle contraction surrounding the spine, eventually leading to scoliosis.
Finally, data obtained from human AIS patients suggested that Urotensin signals were also differentially activated between the convex and concave sides of the spine in these patients. Interestingly, the observed up-regulation of UTS2R expression in AIS patients also occurred on the convex side of the spine, similar to the up-regulation of urp2 observed in the dorsal bending sites of zebrafish mutant spines. It is noteworthy that we did not observe differential expression of urotensin neuropeptides between the convex and concave spinal muscle tissues of AIS patients. We believe this result may be due to the fact that AIS patient tissues were collected from paravertebral muscles that were enriched for the expression of UTS2R, but not for the neuropeptides, which are mainly secreted from the neurons.
Altogether, our data suggest that ependymal polarity defects are the earliest sign of scoliosis development in zebrafish ciliary mutants, and we provide one explanation for how scoliosis develops in zebrafish ciliary mutants, which may provide new insight into mechanisms regulating scoliosis in humans.

Ethics statement
All zebrafish studies were conducted according to standard animal guidelines and approved by the Animal Care Committee of Tufts University and Ocean University of China. Human tissue collection was approved by the Ethics Committee of the Nanjing University Medical School Affiliated Nanjing Drum Tower Hospital, China (No. 2019-066-01).

Spinal cord and brain dissection
To dissect zebrafish spinal cords, humanelyAU : PleasenotethatasperPLOSstyle; use}euthanize}onlywhe killed adult zebrafish were fixed in 4% of paraformaldehyde (PFA) overnight at 4˚C. After washing 3 times with 1X PBS (phosphate buffered saline (PBS)) for 20 minutes each, the skin, muscle, and neural arches were carefully removed with tweezers, and the spinal cord was separated from the vertebral bodies. For brain dissection, zebrafish were first humanely killed using Tricaine in ice-cold water, then the skulls were removed with tweezers in 1X PBS to expose the brain. The dissected brain and spinal cord were further processed for WISH according to standard protocols.

Alizarin red and Calcein staining
Adult zebrafish were fixed in 4% PFA at 4˚C for 1 week. After a 1-hour wash with PBST (0.1% Tween20), fixed zebrafish were stained with 0.01% Alizarin red (Sigma) in 1% KOH for another week. Next, the samples were washed in PBST for 3 days (1 wash/day) at room temperature and further cleared with 0.5% trypsin digestion for 24 hours at room temperature. After multiple washes with PBST, the skin was manually removed and the skeleton was imaged on ZEISS stemi 508 microscope. For Calcein staining, live zebrafish were incubated in 0.2% Calcein (Sigma) (pH 7.5) for 15 minutes and then washed twice with system water. In vivo stained zebrafish were anesthetized with 0.01% tricaine methanesulfonate (MS222), mounted in 3% methyl cellulose, and imaged using a fluorescent Leica M165FC microscope.

Micro-CT and vibratome sectioning of brain tissue
To visualize mineralized skeletal structures via Micro-CT, zebrafish were first anesthetized with 0.01% tricaine methanesulfonate (MS222). Micro-CT images were captured using a Per-kinElmer Quantum GX2 microCT scanner. For histological analysis, the dissected brain was fixed in 4% PFA overnight at 4˚C and then embedded in 3% low melting agarose. Transverse serial sections through the brain were collected using Leica vibratome vt1000s at a thickness of 90 μm each and imaged using Leica M165FC microscope.

High-speed video microscopy
To record the motility of ependymal cell cilia, the brain was dissected from the adult zebrafish, and the ependymal cell layer was gently peeled away from the top of the telencephalon with tweezers. The ependymal tissues were positioned on a glass cover slip containing 1X PBS and placed upside down in the center of a depression slide. Cilia movement was recorded using a 100X oil objective on a Leica Sp8 confocal microscope equipped with a high-speed camera (Motion-BLITZ EoSens mini1; Mikro-tron, Germany). Recordings of in situ cilia movement in the posterior portion of the spinal canal of 5 dpf larvae were conducted as previously described [58]. In all experiments, cilia movement was captured at a rate of 500 frames per second, and playback was set at 25 frames per second. Image processing was performed using ImageJ software (National Institutes of Health, Bethesda, MD, USA).

Scanning electron microscopy
Ependymal epithelia were dissected from the brain, fixed in 2.5% glutaraldehyde overnight at 4˚C, then washed 3 times with 1X PBS. Next, samples were serially dehydrated to 100% ethanol and transferred into isoamyl acetate. After critical point drying, samples were sputter coated with gold-palladium alloy before imaging. Images were acquired using Hitachi-3400N scanning electron microscope (Hitachi, Tokyo, Japan).

Immunofluorescence assay
Immunostaining of whole-mount larvae was performed using standard protocols [59]. For immunofluorescence of ependymal cells, the brain was first dissected from adult zebrafish and then fixed in Dent's fixative (80% methanol and 20% dimethylsulfoxide). Fixed brains were incubated with primary antibodies overnight at 4˚C, followed by washing 3 times with PBST at 4˚C, then incubated with secondary antibody overnight at 4˚C. Immunostained adult zebrafish ependymal cell layer tissues were carefully collected with tweezers for imaging. The ependymal cell tissues of zebrafish less than 1 month old were imaged directly and did not require dissection from the brain.
For RF staining, 48 hpf zebrafish larvae were fixed in Dent's fixative overnight at 4˚C, washed 3 times with 1 X PBS, then incubated with WGA dissolved in PBS (1:200) overnight at 4˚C. After 4 times wash with 1X PBS, the samples were imaged with Leica Sp8 confocal microscope. For staining in adult zebrafish, the spinal cord was first dissected, fixed in 4% PFA overnight at 4˚C, and further processed for WGA staining. After staining, the tissues surrounding the spinal cord were removed to expose the RF for imaging.

Immunofluorescence microscopy of cultured cells
The methods used were as previously described [60]. Briefly, cells grown on coverslips were fixed with 4% PFA for 10 minutes at room temperature and then permeated by cold methanol for 10 minutes at −20˚C. After washing with PBS, the cells were incubated with the primary antibodies at room temperature for 1 hour. After washing twice with PBS, cells were incubated with secondary antibodies and 4 0 , 6-diamidino-2-phenylindole (DAPI) for 1 hour in dark. After washing 3 times with PBS, the coverslips were mounted with mounting buffer (S36963, Invitrogen). All antibodies were diluted in blocking buffer (1% BSA in PBS). Images were acquired using laser scanning confocal microscope (FV3000, Olympus) with a 40× oil-immersion objective. ImageJ software was used for quantification. For IF, the following antibodies were used: mouse anti-flag (1:1,000, Sigma-Aldrich); rabbit anti-PCM1(1:1,000, Cell Signaling Technology); rabbit anti-CEP164 (1:4,000, Proteintech); and OFD1 (1:5,000, homemade).
Proteins were eluted from FLAG M2 beads using 0.2 mg/mL FLAG peptides (F3290, Sigma-Aldrich), mixed with SDS sample buffer and resolved using SDS-PAGE. Finally, protein bands were excised from the SDS-PAGE gels, digested with trypsin, and the extracted peptides were analyzed using mass spectrometry.

Western blot analysis
Proteins were separated by 10% SDS-PAGE and transferred to NT nitrocellulose membranes (66485, BIOTRACE). After blocking with 5% milk, membranes were incubated with anti-FLAG (1:5,000; F1804; Sigma-Aldrich) and anti-OFD1 (1:30,000) for 1 hour at room temperature. After washing with PBS 3 times, the membranes were incubated with secondary antibodies for 1 hour at room temperature. The ChemiDoc Touch Imaging System (Bio-RAD) and Odyssey Laser Imaging System (LI-COR) were used for image acquisition.

Wound healing assays
RPE1 cells transfected with negative control or CCDC57 siRNAs were seeded on glass coverslips in 12-well plates. When reaching 100% confluence, the culture medium was replaced with serum-free media. After 24-hour starvation, 1 mL pipette tips were used to generate scratches on the glass coverslips, and PBS was used to wash away the detached cells.
Subsequently, cells were cultured in serum-free medium for an additional 6 hours and then fixed with 4% PFA. Golgi apparatuses were stained with GM130 (1:500; M179-3, MBL). Cells with Golgi apparatuses located in the sector facing the wound were considered positive for migration. Three groups (>100 cells per group) were used for quantification. GraphPad Prism software was used for statistical analysis. Quantitative data were presented as mean ± standard deviation (SD), and Student t tests were performed to determine statistical significance.

Quantitative PCR
The expression levels of urotensin genes were evaluated using qRT-PCR analysis. Guided by the curvature of the spines, scoliosis mutant trunks were cut into three segments to include the highest or lowest portions of each spinal curve. Total RNA was isolated from dissected larvae or adult tissues using Trizol (Takara). cDNA was synthesized using PrimeScript 1st strand cDNA Synthesis Kit (Takara). qPCR was performed on the Step One real-time PCR system (Thermo Scientific) using the Eva-Green Master Mix (ABM). The following primers were used for PCR analysis: urp1 forward: 5 0 -TCTGGCGGTGCTCTACATTC-3 0 , reverse: 5 0 -AGCAGGACAGGAAGCACAGT-3 0 ; urp2 forward: 5 0 -CCGGAGAACCAGATGCCTTT −3 0 , reverse: 5 0 -ATTTGGGCTGCTTGTTGCTG-3 0 . Relative gene expression levels were quantified using the comparative Ct method (2 −ΔΔCt method) based on Ct values for target genes and zebrafish β-actin.

Fluorescent beads tracing experiments
To evaluate hydrocephalus in zebrafish larvae, Rhodamine-or FITC-conjugated fluorescent dye (70 kDa) were injected directly into the brain ventricles at 2 or 3 dpf. Briefly, ccdc57 mutants and control siblings were first anesthetized using 0.01% tricaine and then incubated with 20 mM 2,3-butanedione monoxime (BDM, Sigma) for 6 minutes to stop the heart beating. The larvae were then placed on the surface of a 1% agarose plate and further injected with fluorescent dye. Fluorescence images were collected using Leica M165FC fluorescent microscope. The size of each brain ventricle was measured and quantified using ImageJ software.
For fine particle movement analysis, we injected 20 nm or 100 nm fluorescent beads (F8888, Thermo Fisher Scientific) into the central canal of 30 hpf (20 nm) or 3 dpf (100 nm) zebrafish larvae. The injection methods were similar to previously reported [39,40]. Briefly, 30 hpf or 3 dpf larvae were anesthetized using 0.4 mg/ml tricaine and then mounted in 1.5% low melting point agarose in the lateral position. The 20-nm or 100-nm fluorescent beads were injected into the center of the diencephalic ventricle. At 1 hour after injection, time-lapse images were acquired at room temperature on an inverted Leica DMI8 spinning disk confocal microscope equipped with an Andor iXon Life 888 EMCCD using a 40X water immersion objective (N.A. = 1.1). These images of the fluorescent beads in the rostral part of central canal were acquired at a frame rate of 10 Hz using Fusion software. The data were further analyzed with ImageJ software.
To examine the migration of fluorescent beads in central canal, we microinjected 100 nm fluorescent beads into the central canal at the position above the end of the yolk extension. The migrated distance of fluorescent beads was captured by THUNDER Imager Model Organism and quantified using ImageJ software.

3D reconstruction of the choroid plexus
The adult brain was dissected and fixed in 4% PFA overnight at 4˚C. After immunofluorescent staining described above, the dChP was imaged using a 40X water objective on a Leica Sp8 confocal microscope. For 3D reconstruction of the CP, we first performed image binarization for acquired image stacks using ImageJ software (version 1.52p) and then performed surface rendering of the binary stack using Imaris software (version 9.6.0) to obtain 3D model.

Cell polarity analysis
ImageJ software was used to quantify cell polarity. For basal body angle analysis, angles were measured between a line made between the centers of the cell and basal bodies and the horizontal axis, as illustrated in Fig 3E. The displacement distance ratio was calculated by dividing the distance between the center of the cell and the center of the basal body by the distance from the cell center to the cell membrane. Cell orientation and centers were calculated by Mor-phoLibJ plugins, and the raw data were analyzed by Microsoft Excel software to generate graphs.

Human tissue collection
Paraspinal muscles were collected from 46 female AIS patients with main thoracic curve during corrective surgery. Bilateral deep paraspinal muscle biopsies of 1.5 × 1.5 × 1.5 cm 3 were collected at the apical vertebral of the main curve for all the subjects. All patients or their guardians provided informed consent for the tissue collection.