Planar cell polarity zebrafish models of congenital scoliosis reveal underlying defects in notochord morphogenesis Free

Congenital scoliosis (CS) is a spinal deformity affecting 0.5-1 in 1000 live births. It is clinically defined as a lateral curvature of the spine (Cobb angle) of ≥10° and is characterized by the presence of congenital vertebral malformations (CVMs) that result from failure of vertebrae formation, vertebrae segmentation, or both. CS is a potentially fatal disorder due to the high risk of progressive spinal deformity and impaired pulmonary function (Hedequist and Emans, 2007). Most cases of CS are sporadic, with a reported familial incidence of ∼3% (Purkiss et al., 2002). CS is a multifactorial disease implicating genetic and environmental factors that remain poorly characterized.

Formation of the spine is mediated by two major embryonic structures: the notochord and the somites (Eckalbar et al., 2012; Wopat et al., 2018). Studies in animal models have linked embryonic defects of these two structures to spinal deformities (Bagnat and Gray, 2020; Eckalbar et al., 2012). In humans, candidate gene and genomic approaches have identified mutations in genes essential for somitogenesis in CS (Chen et al., 2020; Giampietro, 2012; Takeda et al., 2018; Wu et al., 2015). However, the major underlying pathogenic mechanisms of CS remain largely unknown, resulting in an urgent need for additional etiopathogenetic studies of this disease. Since disruption of somitogenesis is most likely insufficient to explain the complexity of CS, a closer analysis of the contribution of the notochord to CS is necessary. The notochord represents the main structural element of the anteroposterior body axis and serves as the axial skeleton of the embryo and as a scaffold for spine formation (Bagnat and Gray, 2020). Immediately after the notochord domain is specified in the dorsal mesoderm, the notochord precursor, called chordamesoderm, undergoes a dramatic morphogenetic process called convergence and extension (C&E) that converts a thick cellular array into a narrow, elongated stack of cells that defines the primary axis of the embryo (Glickman et al., 2003; Keller et al., 1989, 2000; Shih and Keller, 1992a,b). C&E is mainly mediated by the highly conserved non-canonical Wnt planar cell polarity (PCP) signaling pathway (Butler and Wallingford, 2017). As development proceeds, a large proportion of notochord cells migrate outward to form a flat notochord sheath epithelium while the remaining cells expand their volume via a vacuolation process, driving continued extension of the notochord (Yamamoto et al., 2010). Osmotic swelling of notochord vacuoles exerts pressure on the semi-rigid collagenous extracellular matrix (ECM) sheath providing a hydraulic force that gives flexural stiffness to the body axis and mechanical strength that helps straighten and extend the embryonic axis (Bagwell et al., 2020; Ellis et al., 2013). At later stages of development, the notochord serves as the site of vertebrae body condensation and is the progenitor of the nucleus pulposus in intervertebral disks (Choi and Harfe, 2011; McCann et al., 2012; Wopat et al., 2018).

Zebrafish mutants with defects in notochord sheath integrity or biogenesis of notochord vacuoles show shortening of the embryonic axis, kinking of the spine and CS (Bagwell et al., 2020; Gansner and Gitlin, 2008; Gray et al., 2014; Parsons et al., 2002; Stemple et al., 1996; Sun et al., 2020). The precise role of a defective notochord C&E in CS has been more difficult to study mainly because failure of C&E leads to severe defects, including failure of neural fold fusion, notochord folding and even complete loss of somites (Stemple et al., 1996; Wang et al., 2019). PCP mutants exhibit severe C&E defects, manifested by a shortened body axis, a widened chordamesoderm, and defects in tail notochord development, that were invariably lethal at the early juvenile stage (Hayes et al., 2014; Stemple et al., 1996). Mutation of one PCP gene called ptk7a (protein tyrosine kinase 7a) resulted in CS, making this mutant a powerful model in which the role of PCP signaling and defective C&E in CS pathogenesis can be studied (Hayes et al., 2014). Ptk7 is a highly conserved transmembrane protein with no identified kinase activity. It is essential for C&E movements during vertebrate gastrulation and neurulation (Lu et al., 2004; Peradziryi et al., 2012; Yen et al., 2009). Its expression is ubiquitous at the onset of gastrulation and is pronounced in axial and paraxial structures that undergo C&E, including notochord precursor cells and neuroepithelial cells (Hayes et al., 2013). Mutations in orthologs of ptk7a have been associated with neural tube defects (Lei et al., 2019; Lu et al., 2004, 7; Paudyal et al., 2010; Wang et al., 2015) and with both CS and idiopathic scoliosis (IS) in humans (Su et al., 2021). In zebrafish, depending on the time of loss of ptk7a function, maternal zygotic ptk7a (MZptk7a) and zygotic ptk7a (Zptk7a) mutants develop CS- and IS-like spinal deformities, respectively (Hayes et al., 2014). Loss of maternal and zygotic ptk7a RNA leads to severe C&E defects, manifested by a shortened body axis and widened midline structures, and severe CVMs at the early larval stages. A link between CVMs and embryonic defects in patterning and morphogenesis has been suggested based on the abnormal segmentation and somite patterning detected in MZptk7a mutants (Hayes et al., 2014). However, the well-established role of ptk7a as a PCP gene mediating C&E during notochord development was not investigated in that study, leaving a major potential pathogenic mechanism in CS unexplored.

In this study, we examined whether defects in notochord development caused by loss of function of ptk7a contribute to the pathogenesis of CS in the MZptk7a zebrafish model. We demonstrate that MZptk7a mutants show severe defects in C&E and in the organization of the notochord precursor cells, a significant decrease in the number of vacuolated cells as well as severely loosened inter-vacuolated cell junctions with abnormal desmosomes reflecting a change in the mechanical properties of the notochord. At later stages, MZptk7a mutants showed mineralization defects and focal kinks at the anterior part of the spine that were associated with late-stage CVMs. Brillouin analysis, a powerful technique for measuring the mechanical properties of biological tissues, confirmed changes in the mechanical properties of the mutant notochord manifested as a less-stiff ECM, which suggests reduced notochord pressure. Loss of function of another PCP gene, called vangl2, led to severe 3D curvature defects, anterior CVMs and increased notochord cell apoptosis, providing additional evidence for the role of defective PCP signaling in the pathogenesis of CS. These data support a model in which defects in PCP-mediated C&E of the notochord contribute to the pathogenesis of CS.

Ptk7a is dynamically expressed during gastrulation and neurulation, where it plays essential roles in C&E of mesodermal and neuroectodermal cells through PCP activation. A previous study of another ptk7a mutant demonstrated severe defects in axial C&E in MZ mutants. These defects were manifested by widened neural tube, notochord and somites as well as a significant reduction in length width ratio (LWR) and a defective mediolateral orientation of dorsal ectodermal cells (Hayes et al., 2013). To study the role of ptk7a in mediating C&E of the embryonic axis in further detail, we created a new CRISPR/Cas9-induced knockout allele of this gene harboring a 10-nucleotide deletion frameshift mutation in exon 1, that we designated as ptk7a^udm400 (Fig. S1). Consistent with previous reports (Hayes et al., 2013, 2014), Zptk7a^udm400 mutants exhibited an IS-like phenotype while MZptka^udm400 mutants developed severe C&E defects manifested by a significantly shortened body length (Fig. S1). A mild body curvature was noticed in MZptk7a^udm400 mutants as early as 6 days postfertilization (dpf), and around 50% of these mutants survived until adulthood and developed a scoliosis-like phenotype (Fig. S1).

Given the essential role of the notochord in body axis elongation and straightening, we proceeded to investigate this structure for the presence of developmental defects in MZptk7a^udm400 mutants. Notochord analyses throughout this study were focused on its anterior portion, at the 1-8 somite (s1-s8) level, where the process of C&E plays a more significant role in the elongation of the body axis compared with more caudal regions and where differentiation of notochord precursor cells into inner vacuolated cells and outer sheath cell epithelium begins (Bagnat and Gray, 2020; Gray et al., 2014). To analyze the dynamic changes of notochord precursor cells during notochord development, MZptk7a^udm400 were crossed into a Tg(β-actin:GFP-CAAX) background and the notochord precursor cells were examined at ∼11 hours postfertilization (hpf) at the s1-s8 level using multiphoton microscopy followed by 3D renderings. Significant defects in notochord C&E were detected in MZptk7a^udm400 embryos as manifested by a significant increase in notochord width, a significant reduction in the length-to-width ratio of notochord cells as well as a significant defect in their orientation relative to the mediolateral axis (P<0.0001) (Fig. 1). The organization of the mutant notochord precursor cells at the s1-s8 level was irregular compared with control cells (Fig. 1A), but was unaffected at the tail end (Fig. S2A).

We also detected defects in somite development, including a defective C&E manifested by a wider structure and a significantly wider chevron angle, incomplete segmentation between somites, misalignment of muscle fibers inside the somites and abnormal crossing of the myotendinous junction (Fig. S3).

We next proceeded to analyze the three major components of the notochord tissue – the vacuolated cells, the sheath epithelium and the surrounding extracellular notochord sheath – in MZptk7a^udm40 embryos. Cell trace staining followed by live multiphoton imaging of 1 dpf, 2 dpf and 7 dpf MZptk7a^udm400 did not detect any defect in vacuole biogenesis, such as fragmentation or collapse, compared with their heterozygous siblings (Fig. S4). However, multiphoton imaging at 37 hpf revealed the presence of abnormal cell clumps in MZptk7a^udm400 embryos that were absent from the notochords of their heterozygous siblings (Fig. S5). These clumps had no specific localization pattern and were later characterized to be apoptotic by NucView530 Caspase-3 live staining (Fig. 2A). To study further the temporal distribution of these apoptotic cells, in vivo time-lapse imaging of zebrafish was conducted starting at 12-14 hpf until 44 hpf (Fig. 2; see also Movies 5 and 6). At the s1-s8 region, the proportion of apoptotic cells considerably increased in MZptk7a^udm400 embryos compared with their heterozygous siblings, peaking at 16-20 hpf (Fig. 2B). Injection of p53 (tp53) morpholino (MO) in mutant embryos failed to rescue the increased apoptosis detected in their notochord, suggesting p53-independent underlying apoptotic mechanisms (Fig. S6). The number of apoptotic cells at the region from s10 to the tail end at each of 18 hpf, 20 hpf and 22 hpf in MZptk7a^udm400 embryos was similar to that of their ptk7a^udm400/+ siblings (Fig. 2C, Fig. S2B). Quantitative analyses using 3D renderings of live multiphoton imaging at 5 dpf detected a significant reduction in the number of vacuolated cells and a significant increase in their volume in MZptk7a^udm400 embryos compared with their heterozygous siblings (Fig. 2D-F).

Defects in the arrangement of vacuolated cells were detected at the s1-s8 anterior region from the beginning of vacuolation and were prominent at around 37 hpf. By 5 dpf, this disorganization was almost completely restored (Fig. 2A, Fig. S7).

Next, the notochord of MZptk7a^udm400 mutants was analyzed for the presence of ultrastructural defects by transmission electron microscopy (TEM) at 2 dpf. Strikingly, a severe reduction in caveolae number was detected in the vacuolated cells in MZptk7a^udm400 embryos. In contrast to the cave-like structure of the caveolae, the majority of the caveolae in the mutants had an abnormally spherical form (Fig. 3A-E, Fig. S8). They were also linked to an aberrant concentration of intermediate filaments on their cytoplasmic side. Mutant vacuolated cells also showed a severe loosening of their intercellular junctions along with morphologically abnormal desmosomes and an abnormal presence of extracellular vesicles (Fig. 3A-D′,F). Caveolae and desmosomes act as mechanosensors that are crucial for tissue integrity and strength against mechanical stresses such as osmotic swelling occurring during embryonic development and tissue maintenance (Dubash and Green, 2011; Garcia et al., 2017; Parton and del Pozo, 2013). The structure/morphology of the notochord outer sheath cells and the peri-notochordal basement membrane appeared normal. However, the medial layer of the basement membrane, composed of collagen fibrils, appeared generally thicker in MZptk7a^udm400 mutants compared with their heterozygous siblings (Fig. S9).

At later stages, the notochord directs centrum bone synthesis. Therefore, we next determined whether the early notochord defects caused by the loss of function of ptk7a described above affect the segmentation and mineralization of the axial centra. Live calcein staining was conducted to visualize mineralized vertebrae at 7 dpf, 9 dpf and 15 dpf. A significant delay in mineralization was detected in MZptk7a^udm400 mutants compared with their control siblings (Fig. 4A-C). Closer examination of the vertebrae at 9 dpf and 15 dpf showed various types of CVMs, including skipped mineralization, fused or wedged vertebrae and hemivertebrae, which represent the hallmarks of CS (Fig. 4D,E). Further analysis with Alizarin Red staining at 15 dpf localized these CVMs only at the anterior part of the spine and demonstrated the presence of spine kinks/bends that colocalized with the aberrant mineralization regions in MZptk7a^udm400 mutants (Fig. 4F,G). A closer examination of the mutant notochords at 2 dpf revealed a wavy appearance at the anterior region of the embryonic axis that corresponded to the region harboring the spine defects detected at later stages (Fig. 4H).

Around 50% of MZptk7a^udm400 mutants survived until adulthood and hence we were able to follow the effect of loss of function of ptk7a on spine development from the early stages of mineralization until the formation of mature vertebrae. High-resolution µCT analyses were performed for accurate characterization of the CVMs and for quantitative measurements of the 3D curvature detected in MZptk7a^udm400 mutants (Fig. 5, Movies 1-4). Consistent with the Alizarin Red segmentation defect pattern, CVMs detected in these mutants were localized exclusively at the anterior part of the embryonic axis and affected the Weberian vertebrae (V1-V4) and the abdominal precaudal V5-V9 (Fig. 5, Fig. S10).

MZptk7a^udm400 mutants exhibited a reduced number of vacuolated cells, a kinky notochord and a severe reduction in the number of caveolae and desmosomes, suggesting altered mechanical properties of the notochord. To evaluate the mechanical characteristics of the notochord in MZptk7a^udm400 mutants, we used Brillouin microscopy, which is a powerful technique for measuring the mechanical properties of biological tissues (Bevilacqua et al., 2019; Handler et al., 2023; Zhang and Scarcelli, 2021). Brillouin microscopy has been shown to be capable of revealing the thin (400 nm) peri-notochord ECM layer and measuring its mechanical properties (Bevilacqua et al., 2019). Cross-section scans at two or three points at the anterior part of the embryos (s7-s14) were conducted for each fish at 3-4 dpf (Fig. 6A). Representative images of the Brillouin shift are shown in Fig. 6B,C. The average Brillouin shift of ECM in each of the regions was quantified and the corresponding longitudinal modulus M^′ was calculated as described in Materials and Methods. The Brillouin shift did not show a significant difference between site 1 and site 3 in either group (Fig. 6D). The average Brillouin shift of different points and the correlated longitudinal modulus M′ were significantly decreased in the MZptk7a^udm400 fish (Brillouin shift: 6.474±0.027 GHz; Mʹ: 2.569±0.022 GPa) compared with their heterozygous siblings (Brillouin shift: 6.493±0.031 GHz; Mʹ: 2.584±0.025 GPa) (Fig. 6E), indicating a less-stiff ECM and a change in the notochord mechanical properties in the mutants.

Ptk7a interacts genetically with a ‘core’ PCP gene called vangl2 in mediating C&E during neurulation in mouse (Lu et al., 2004; Paudyal et al., 2010). Hence, we wanted to determine whether loss of function of vangl2 leads to CVMs. In zebrafish, similar to MZptk7^udm400 mutants, vangl2 mutants suffer from severe C&E defects during gastrulation, severe notochord folding and segmentation defects. These defects occur even in zygotic mutants, in which zygotic RNA is lost while some maternal vangl2 RNA is deposited, and are invariably lethal at early stages (Jessen et al., 2002; Stemple et al., 1996). MO knockdown of vangl2 caused severe C&E defects, manifested mainly by shortened body length and widened somites, and was invariably lethal at early stages (Jessen et al., 2002). To investigate whether partial loss of function of vangl2 can lead to CVMs, we adopted an MO knockdown strategy that allows graded knockdown of gene function. We first conducted a dose-response experiment to determine the suboptimal dose of vangl2-MO, defined as the lowest dose that causes a significant reduction in body length, but that is not lethal at later stages. The suboptimal dose was set at 0.5 ng (Fig. 7A,B). Injection of 0.5 ng of vangl2-MO in ptk7a^udm400/+ embryos caused a significant reduction in their body length, demonstrating a genetic interaction between ptk7a and vangl2 even at a low dose (Fig. 7C). Notably, 15% of embryos injected with 0.5 ng of vangl2-MO survived until adulthood and of these ∼20% developed 3D curvature defects (Fig. 7D). Calcein staining and µCT analyses demonstrated the presence of CVMs manifested as wedged first vertebrae and fused vertebrae at a low frequency (Fig. 7E-G). This is the first report of CVMs at an adult stage in a PCP zebrafish mutant other than ptk7a and this strongly supports the role of defective PCP signaling in CS pathogenesis. We next wanted to determine whether vangl2 morphants display any of the notochord defects detected in MZptk7a^udm400 mutants. Injection of 2 ng vangl2-MO caused a significant increase in apoptosis of notochord precursor cells as revealed by NucView530 Caspase-3 live staining at 19 hpf at s1-s8 (Fig. 7H,I). To demonstrate specificity of the increased notochord apoptosis to vangl2 knockdown, we conducted an apoptosis assay on vangl2^m209 mutants and we detected a significant increase in apoptosis of notochord precursor cells in vangl2^m209 mutants compared with their wild-type and heterozygous siblings at 19 hpf (Fig. 7J,K). vangl2^m209 mutants have been reported to manifest a folded notochord in the tail region and hence we analyzed the level of apoptosis in the notochord from s10 up to the tail end. A significant increase in apoptosis associated with severe defects in cell arrangement was detected in these regions (Fig. 7L, Fig. S11).

Ptk7 was first identified in colon carcinoma cells and is often deregulated in many types of cancer. Subsequently, it was found to have several functions, including axon guidance in Drosophila and regulation of gastrulation, neural tube closure, neural crest migration and cardiac morphogenesis in vertebrates (Peradziryi et al., 2012). Ptk7 is also a target of membrane-anchored type-1 matrix metalloproteinases that control cell adhesion and migration (Golubkov et al., 2010). Ptk7 was first identified as a PCP gene in a Ptk7 mutant mouse that suffered from a severe form of neural tube defects, which is a hallmark of a defect in PCP signaling (Lu et al., 2004). Additional studies in animal models have confirmed the role of Ptk7 in mediating C&E of the axial and paraxial structures (Hayes et al., 2013; Paudyal et al., 2010; Yen et al., 2009). In this study, we investigated the role of Ptk7 in mediating C&E of the notochord in zebrafish. Loss of function of ptk7a led to severe C&E defects of the notochord precursor cells, manifested by a significant increase in notochord width and reduction in the LWR of notochord precursor cells, as well as a significant defect in their orientation relative to the mediolateral axis and organization. These findings are expected since ptk7a has been shown to affect C&E of the neural ectoderm that occurs concurrently with C&E of the axial notochordal mesoderm and the paraxial (presomitic) mesoderm during the anteroposterior elongation of the embryonic axis (Sutherland et al., 2020). At a later stage, notochord vacuolated cells in MZptk7a^udm400 mutants had a disorganized arrangement compared with the stereotypical staircase arrangement of these cells in controls. We suggest that this could be due to a shorter and widened mutant notochord that was shown to affect the organization of the vacuolated cells as they inflate and exert pressure on the ECM (Norman et al., 2018).

Strikingly, loss of function of ptk7a in MZptk7a^udm400 mutants led to a significant increase in apoptosis in the anterior notochord region prior to vacuolation, a significant decrease in the number and a significant increase in the volume of the vacuolated cells in this region, as well as significant ultrastructural defects in these vacuolated cells manifested by a severe reduction in caveolae number and loosened intercellular junctions. These defects were most likely not caused by abnormal biogenesis of vacuolation or by an aberrant ECM formation as both processes seemed unaffected in the mutants. In addition to its well-established role in regulating C&E of the embryonic axis, PCP signaling plays an important role in mediating tissue mechanical properties by regulating the formation of subcellular and supracellular structures, such as the cytoskeleton, nuclei and cell junctions (Vichas and Zallen, 2011). Hence, we propose that an increased rate of apoptosis in MZptk7a^udm400 mutants could be linked to a defective role of ptk7a as a PCP activator through at least two possible mechanisms: (1) a wider and disorganized notochord structure caused by defective C&E of notochord precursor cells that would limit access to certain trophic or signaling factors necessary for cell differentiation and survival and/or (2) defective cytoskeletal structure and weakened intercellular junctions of the mutant notochord that would lead to cell death upon mechanical stress. Generating genetic mosaics and conducting further in vivo time-lapse imaging of individual zebrafish in the background of transgenic reporter lines are needed to investigate the cellular origin of the apoptotic cells and to follow their progression during the course of spine development.

MZptka7^udm400 mutants suffered from abnormal mineralization and severe segmentation defects that later developed into CVMs in surviving adult fish. A previous study of another zebrafish mutant of ptk7a reported segmentation defects that were associated with abnormal somite patterning (Hayes et al., 2014). Since ptk7a has been shown to act as a molecular switch that activates Wnt/PCP pathway while simultaneously inhibiting the Wnt/β-catenin canonical pathway, these segmentation defects were hypothesized to be caused by abnormal signaling in both pathways (Hayes et al., 2013, 2014). Although we cannot exclude a role of defects in Wnt canonical signaling and/or somitogenesis in the pathogenesis of CVMs in MZptk7a^udm400 mutants, we favor our hypothesis that CVMs in these mutants arise mainly from defects in PCP signaling and C&E of the notochord. First, loss of function of ptk7a seems to have a more drastic effect on the formation of the anterior part of the developing spine and vertebrae compared with more-posterior regions. Anteroposterior elongation of the body axis relies heavily on C&E of the mesodermal cells only at early stages, in contrast to the formation of the caudal body axis at later stages (Mongera et al., 2019). Mutants with defective somitogenesis do not display any specific localization pattern of CVMs. In fact, the vast majority of mutations affecting somitogenesis in zebrafish (and in mice) specifically disrupt posterior somite formation (Pourquié, 2001; van Eeden et al., 1996). Second, we detected a wavy notochord and a bent or kinky spine in MZptk7a^udm400 mutants that colocalized to the region harboring the CVMs, similar to findings in other notochord mutants that developed CVMs. Loss of function of genes coding for components of the ECM sheath, including col8a1a (Gansner and Gitlin, 2008; Gray et al., 2014) col27a1a and col27a1b (Christiansen et al., 2009), or of genes essential for vacuole biogenesis, including rab11a and dstyk (Bagwell et al., 2020; Ellis et al., 2013), leads to focal notochord bends or kinks that transition into focal CVMs. Similar to dstyk mutants, MZptk7a^udm400 larvae showed delayed and skipping of mineralization that colocalized with the kinky regions of the spine (Bagwell et al., 2020; Sun et al., 2020). CVMs in dstyk and col8a1a mutants were hypothesized to result from abnormal bone deposition at abnormal notochord regions independently of defects of somitogenesis (Bagwell et al., 2020; Gray et al., 2014).

Vertebrate body axis straightening depends on the hydrostatic pressure generated within the notochord. As notochord cells vacuolate, they are resisted by the notochord sheath, thereby increasing the notochord internal pressure and stiffness. This mechanical force would enable the notochord to elongate and straighten without being bent by the surrounding tissues and external mechanical stresses (Bagnat and Gray, 2020; Bagwell et al., 2020; Ellis et al., 2013). As vertebral bone is formed around the notochord, the vacuoles mechanically antagonize the pushing force of invading somite-derived osteoblasts, thereby influencing the mineralization of centra. Mutant notochords with fragmented vacuoles or lacking vacuoles or with a structurally weak notochord sheath would have a defective hydrostatic pressure that will make them unable to resist these invasive forces and more easily deformable, causing kinks in the spine and asymmetrical vertebral bone growth (Bagwell et al., 2020; Gray et al., 2014).

How would a defective C&E of the notochord in MZptk7 mutants lead to CVMs? We detected in these mutants a disorganized and wider notochord along with a significantly reduced number of vacuolated cells. We also detected a dramatic reduction in the density of caveolae with abnormal accumulation of intermediate filaments, abnormal morphology of desmosomes and severely loosened junctions between vacuolated cells. Caveolae play an essential mechanosensory and mechanoprotective role in the notochord that has been suggested to be related to the increase in hydrostatic pressure as the vacuoles inflate or osmotically swell (Nixon et al., 2007; Parton and del Pozo, 2013). Hence, one could speculate that a reduction in the density of caveolae in the vacuolated region of the notochord might reflect a reduction in its internal pressure. Importantly, Brillouin microscopy analyses revealed the presence of a less-stiff ECM tissue in MZptk7a^udm400 mutants compared with their heterozygous siblings, suggesting a reduced pressure on the ECM. Based on these findings, we propose a novel model for the pathogenic mechanism underlying CVMs in MZptk7a mutants that is summarized in Fig. 8. Briefly, a defective C&E of the notochord caused by loss of function of ptk7a would lead to a wider and disorganized notochord that would affect the tissue integrity, and lead to an increased apoptosis rate of the notochord precursor cells and a decrease in the number of vacuolated cells. This would cause a decrease in the notochord pressure and a change in the mechanical properties of the notochord as manifested by a wavy notochord, a decrease in the density of caveolae and desmosomes and a less-stiff notochord ECM. This would, in turn, lead to bends and kinks in the spine axis, asymmetric bone growth and CVMs. In this model, we cannot exclude a direct effect of apoptosis on notochord folding as apoptotic cells were demonstrated to generate an apico-basal force that has a direct mechanical impact on the surroundings during tissue remodeling (Monier et al., 2015; Roellig et al., 2022).

In our study, the notochord defects detected in MZptk7a^udm400 mutants were accompanied by defects in somite development. These could be the result of a direct effect of loss of function of ptk7a on C&E of the presomitic mesoderm, as previously reported in PCP mutants (Henry et al., 2000; Yin et al., 2008), and/or on the somites as they are shaped and positioned alongside the notochord before differentiating into skeleton and muscles. Alternatively, they could result indirectly from the effect of loss of function of ptk7a on notochord development demonstrated in this study. It is well established that proper somite development depends upon the mechanical and signaling functions of the notochord (Fleming et al., 2001; Resende et al., 2010). Interestingly, studies of PCP mutants in zebrafish have provided evidence for an essential role of notochord C&E in somite patterning and differentiation during later stages. Simultaneous inactivation of the PCP genes knypek (gpc4) and vangl2 led to severe C&E defects that were shown to be responsible for malformed somites and a consequent reduction of slow muscle precursors, the adaxial cells. It was hypothesized that defective C&E movements in these double mutants impair notochord development, which in turn interferes with morphogenesis of the adaxial cells (Yin and Solnica-Krezel, 2007). Hence, it is reasonable to speculate that the severe notochord C&E defects detected in MZptk7a^udm400 mutants in this study could contribute to defects in somite morphogenesis.

Defects in PCP signaling were first demonstrated to be implicated in pathogenesis of scoliosis in the zebrafish mutant ptk7a: depending on the time of loss of ptk7a function, MZptk7a and Zptk7a mutants develop CS- and IS-like spinal deformities, respectively (Hayes et al., 2014). Loss of only zygotic ptk7a seems to affect mainly ependymal cell cilia development that would lead to reduced cerebrospinal fluid flow with subsequent reduced urotensin Urp1/2 signaling and increased neuroinflammation in the brain and spinal cord (Grimes et al., 2016; Van Gennip et al., 2018; Zhang et al., 2018). This pleiotropic effect could be due to a specific spatiotemporal expression profile of ptk7a RNA or to a dosage effect in the same pathway or to distinct functions via independent biological pathways. A later genetic study in a human cohort identified loss-of-function and hypomorphic variants in PTK7 in CS and IS respectively, suggesting a correlation between the dose of PTK7 and the type of scoliosis (Su et al., 2021).

The role of PCP defects in scoliosis was later confirmed by detecting segmentation defects in vangl2 zebrafish mutants and in wild-type embryos injected with dominant-negative Dishevelled that specifically disrupts PCP (the Xenopus Xdd1) (Hayes et al. 2014). Genetic chimeras in which MZptk7a or vangl2 mutant cells were transplanted into the tail bud of host embryos developed CVMs in that region at larval stages (Hayes et al., 2014). Conditional loss of vangl2 in foxj1a-positive cell lineages was shown to cause ependymal cell cilia and Reissner fiber formation defects as well as IS, similar to Zptk7a mutants (Jussila et al., 2022). A recent study has demonstrated that deletion of Vangl1 and Vangl2 in mouse causes vertebral anomalies resembling human CVMs and has identified potentially deleterious variants in these two genes in humans with CVMs (Feng et al., 2024). In our study, and similar to MZptk7a^udm400 mutants, partial MO knockdown of vangl2 has revealed an increase in notochord apoptosis, the presence of 3D spinal curvature and mild anterior CVMs. Notably, vangl2^m209 mutants that suffer from severe PCP defects and that die at early stages showed a significant increase in apoptosis in the notochord that correlated with these defects all along the embryonic axis. These observations from our group and others support a model in which loss of function of PCP genes exerts a dosage-sensitive pleiotropic effect on notochord and spine development leading to CS- and/or IS-like phenotypes.

In conclusion, our study establishes a novel model for CS pathogenesis whereby defects in notochord C&E caused by defective PCP signaling are associated with CVMs in zebrafish. Defects in notochord sheath integrity or biogenesis of notochord vacuoles were shown to cause CVMs (Bagwell et al., 2020; Gansner and Gitlin, 2008; Parsons et al., 2002; Sun et al., 2020), representing orthologs of notochord genes as promising candidates for human CS. A recent study using whole-exome or genome sequencing data from 873 probands with CVMs and 3794 control individuals followed by biological validation in single-nucleus RNA-sequencing data from human embryonic spines has demonstrated an enrichment of the burden test signals of ultra-rare variants in the notochord at early developmental stages (Zhao et al., 2024). Genetic molecular studies in additional CS cohorts as well as functional validation models will shed important insights into the role of variants in notochord genes in human CS and will extend our understanding of the role of the highly conserved and essential notochord developmental mechanism in normal and abnormal spine formation.

All fish handling and experimental procedures were in accordance with the guidelines of the Canadian Council of Animal Care and were approved by the Animal Care Committee of the CHU Sainte Justine Research Center.

In Fig. 4H, for each fish embryo, 4-6 stacks of bright-field images acquired on Leica DMi8 microscope with HC PL FLUOTAR L 20× DRY objective were subjected to a ‘mosaic merge’ with a blend of ‘smooth’ in LAS X software.

For Figs S2B, S3D, S4C,D and S11A'-D', stacks of images acquired by multiphoton microscopy with a HC PL IRAPO 25×/1.00 water objective were subjected to a ‘mosaic merge’ with ‘3D’ and ‘smooth’ in LAS X software.

Two ptk7a transcripts are reported at the Ensembl database (ENSDART00000098461.6, ENSDART00000149683.2). Gene-specific guide RNAs (gRNAs) were designed using the online tool CRISPRscan (http://www.crisprscan.org/) that target 18 nucleotides shared by both transcripts, two nucleotides downstream of the start codon. Synthesis of gRNAs was done with MEGAscript™ T7 Kit (Invitrogen, AM1334). A 1 nl drop of a mix of 100 ng/μl of CleanCap® Cas9 mRNA (Trilink, L-7206) and 30 ng/μl of gRNA was injected into one-cell-stage wild-type embryos using a Picospritzer III pressure ejector. At 2 dpf, ten embryos were genotyped to determine the indel efficiency. For DNA extraction, embryos were digested individually in 20 μl of 50 mM NaOH at 95°C for 10 min, and the solution was next buffered by adding 2 μl of 100 mM Tris HCl (pH 8). PCR amplification of the CRISPR target regions was performed using TaqHifi polymerase protocol with primers (F: 5′-AATCGCGTTCATCGA-3′; R: 5′-AACTTACTCAGCACT-3′) flanking the target site (Fig. S1A) and an annealing temperature of 58°C. PCR amplified products were sequenced by Sanger sequencing (Genome Quebec, McGill University, Montreal, QC, Canada) to confirm the presence of CRISPR-induced indels. The remaining siblings were raised to adulthood and genotyped as described above using a small clipping of the caudal fin.

F0 adult fish were crossed with wild-type fish and those that generated a higher percentage of embryos with CRISPR-induced indels were re-crossed with wild-type fish to generate an F1 heterozygous line. F1 ptk7a^udm400/+ mutants were intercrossed to obtain F2 zygotic homozygous mutants (Zptk7a^udm400). Zygotic homozygous females were crossed to heterozygous males to obtain F3 maternal zygotic homozygous mutants (MZptk7a^udm400), in which both zygotic and maternal ptk7a RNA were mutated. Mutant ptk7a fish were genotyped first with PCR amplification and Sanger sequencing as described above and later with high resolution melting (HRM). Briefly, 1 µl of 1:10 diluted genomic DNA from a digested embryo or a fish fin clip was mixed with 2.5 µl of Precision Melt Supermix (1725110, Bio-Rad) and 1.5 µl ddH2O and subjected to the following HRM assay conditions in a light cycler PCR machine (Roche, light cycler 96): preincubation (95°C, 120 s), two-step amplification (95°C, 10 s; 60°C 30 s; 45 cycles) and HRM (95°C, 4.4°C/s, 60 s; 40°C, 2.2°C/s, 60 s; 65°C, 2.2°C/s, 1 s; 97°C, 0.07°C/s, 1 s).

Equimolar amounts of entry vectors (299: p5E-bactin2; 383: pME-EGFP; and 302:p3E-polyA) and destination vector (395: pDestTol2CG2) were combined with LR Clonase II Plus enzyme mix (Thermo Fisher Scientific, 12538-120) and incubated overnight at 25°C. One microliter of 2 μg/μl Proteinase K solution was added to terminate the reaction. Samples were next incubated at 37°C for 10 min and transformation was carried out in 50 µl of One-Shot MAX Efficiency™ DH5α-T1R Competent Cells Phage-Resistant Cells (Invitrogen, 12297016). Ampicillin-resistant clones were verified by restriction digests with SalI. Tol2 transposase mRNA was synthesized from plasmid 396:pCS2FA-transposase using the mMESSAGE mMACHINE™ SP6 Transcription Kit (Thermo Fisher Scientific). One nanoliter of a solution containing 25 ng/μl of each of Tol2 transposase mRNA and actb2:EGFP construct was injected into embryos before the one-cell stage. Injected embryos were screened at 3 dpf for GFP distribution in the heart. Embryos with strong and even GFP distribution were selected for raising to adulthood. Adult fish were outcrossed to select and keep the embryos with robust, uniform, and broad GFP expression.

MZptk7a^udm400 mutants were generated in the background of Tg(actb2: EGFP). Embryos at 12-14 hpf were manually dechorionated and mounted in 1% low-melting-point agarose in egg water. Imaging was performed on multiphoton microscopy with HC PL IRAPO 25×/1.00 water objective and LAS X software (Leica Wetzlar, Germany). For notochord width, six points along the notochord in the whole field view were chosen with Fiji [ImageJ 2.1.0/1.53c /Java 1.8.0_172 (64-bit)]. The width of the notochord was determined by measuring the widest distance across all the z-stacks. Notochord cell renderings were created using the cells features of Imaris software, v.9.0.0 and v.9.1.2 (Bitplane USA). Cell BoundingBoxOO Length B (length of the axis perpendicular to the longest axis) and Cell BoundingBoxOO Length C (length of the longest axis) were taken from these renderings. LWR was calculated as Cell BoundingBoxOO Length C divided by Cell BoundingBoxOO Length B. Cell orientation relative to the x-axis is indicated with the degrees from medial lateral angles; the angle value was calculated with parameter ‘Cell Ellipsoid Axis C_X’ using the ‘ABS’, ‘ACOS’ and ‘DEGREES’ functions in Excel.

Cell internal membranes were stained with vital dye GFP counterstain BODIPY TR Methyl Ester (C34556, Invitrogen) (MED) at a concentration of 100 μM for 1 h and then washed three times with egg water. After the staining, 1 dpf or 2 dpf embryos were embedded in 1% low melting agarose and immersed in egg water. Imaging was performed on multiphoton microscopy with HC PL IRAPO 25×/1.00 water objective and LAS X software.

For studies of notochord cell organization, staining with MED was carried out at 17 hpf and imaging was performed at 18-22 hpf with multiphoton microscopy with HC PL IRAPO 25×/1.00 water objective and LAS X software.

NucView530 Caspase-3 Substrates (Biotium, 10408) was diluted to 1:1000 in egg water and applied to the agarose-embedded embryos. In vivo time-lapse imaging with multiphoton microscopy with HC PL IRAPO 25×/1.00 water objective and LAS X software (Leica) was conducted starting from 12-14 hpf (8-10 ss) until 12 h or 36 h later.

For apoptosis detection at the posterior part of the notochord, NucView530 Caspase-3 substrate was added to embryos 1 h before the indicated time point. Embryos were embedded laterally and subjected to imaging with multiphoton microscopy as described above.

p53 MO (5ʹ-GCGCCATTGCTTTGCAAGAATTG-3′) (Langheinrich et al., 2002) was ordered from Gene Tools (Oregon, USA). Embryos injected with 4 ng of vangl2 MO or embryos obtained from crossings of female ptk7a^udm400 mutant and male ptk7a^udm400/+ were injected with 2-6 ng of p53 MO before the four-cell stage.

Fish embryos at 2 dpf were fixed in 4% paraformaldehyde for 2 h at room temperature (RT), washed three times for 10 min each wash with PBS solution and a tip of the tail was cut for genotyping. They were next fixed in a solution containing 2.5% glutaraldehyde and 2% paraformaldehyde in PBS for 1 h at RT followed by three 10 min washes with PBS solution. A solution containing potassium ferricyanide (1.5%) and osmium tetroxide (1%) in water was prepared and samples were immersed for 60 min at RT. Following three 5 min washes in distilled water, samples were then dehydrated in a series of ethanol (30%, 50%, 70%, 80%, 95%, 2×100%, 60 min for each concentration). LR White resin Hard Grade (LW) was used for embedding the tissue with infiltration steps at 50%, 75% LW to ethanol for 60 min at 4°C with agitation followed by two 60 min washes with LW. Samples were finally embedded in LW pure at 60°C for 3 days. Blocks were sectioned on a Leica EM UC7 ultramicrotome at 70 nm and mounted on Formvar/Carbon Support Film Hexagonal 200 mesh NI Grids. Thin sections were viewed on a FEI Tecnai T12 (Eindhoven, The Netherlands) A transmission electron microscope equipped with a LaB6 filament and operated at an acceleration voltage of 80 kV. TEM imaging was carried out at the Electron Imaging Facility, Faculty of Dental Medicine, University of Montreal, Canada.

Electron micrographs of the notochord plasma membrane septal regions were taken at random at magnifications of 26,000× or 52,000× from 2 dpf MZptk7a^udm400 and their sibling's embryos (n=3 for each genotype). A total of three to six images were selected for each embryo. Along the plasma membrane, all the caveolae invaginations were measured with Fiji [ImageJ 2.1.0/1.53c /Java 1.8.0_172 (64-bit)] for both ‘neck’ width and depth. To count the caveola number, we first analyzed the caveolae morphology in the controls and defined caveolae within 95% of neck width (13.26598-78.39107 nm) and depth (27.43025-90.43325 nm) distribution as normal caveola. The caveolae number along a membrane length unit was calculated after membrane length standardizing.

From the same micrographs for caveola quantification, the gap between neighboring vacuolated cells was measured with Fiji [ImageJ 2.1.0/1.53c /Java 1.8.0_172 (64-bit)]. The gap area along a membrane length unit was calculated after membrane length standardization.

Calcein solution (Sigma-Aldrich; C0875) was prepared at 0.2% in egg water, and the pH was restored to 7.0. Zebrafish embryos were immersed in calcein solution for 10 min at RT. The embryos were rinsed three times, 3 min each, in egg water, anesthetized with 0.016% tricaine, and embedded in 0.8% low melting agarose. Imaging was carried out with 2.5× and 10× objectives using a Leica DMi8 microscope.

Anesthetized fish were incubated overnight at 42°C in a fixative solution (5% formalin, 5% Triton X-100, 1% KOH), then in an enhancement solution (20% ethylene glycol, 5% Triton X-100, 1% KOH) for 24 h at 42°C. Fish were next put in the staining solution (0.05% Alizarin Red S, 20% ethylene glycol, 1% KOH) for 30 min at RT and then in the clearing solution (20% Tween 20, 1% KOH) for 24 h at RT with four water changes. Specimens were next immersed in increasing glycerol concentrations of 50% to 90% and then kept in 100% glycerol. Stained fish were visualized with a stereomicroscope (M205FA, Leica, Germany), and imaging was performed with an MC 190 HD camera (Leica).

Samples were fixed with 4% paraformaldehyde overnight and kept in ethanol. Scanning was performed on a SkyScan 1176 μCT system (Bruker, Kontich, Belgium) with an X-ray source voltage and current of 30 kV and 360 μA, respectively. Over 2800 projection images were generated over 360° with a 0.5° rotation step and three averaging frames, and isotropic resolution was 8.76 μm. 3D reconstruction was carried out with Imaris. Snapshots of lateral and dorsal views of zebrafish were acquired from μCT 3D reconstructions in Imaris. Cobb angles in different views were measured using Surgimap (Nemaris, Inc.). Lines were drawn parallel to the top and bottom-most displaced vertebrae. The same measurements were carried out as on the Alizarin Red-stained fish.

We used a confocal Brillouin microscope for mechanical imaging of the zebrafish (Zhang and Scarcelli, 2021). The light source was a 660 nm continuous wave laser. The incident power for zebrafish experiment was 15 mW. A 40× objective lens with 0.6 NA was used to focus the beam into the zebrafish. The step size of imaging was 0.5 μm and the exposure time of the camera at each pixel was 50 ms. Zebrafish embryos were divided into 17 groups, and each group had one ptk7a^udm400/+ zebrafish and one MZptk7a^udm400 zebrafish. Embryos were embedded in egg water with 1% low-melting point agarose and 0.016% tricaine at 27°C. The Brillouin imaging was performed group by group. Cross-section scans of half neural tube (three scans for the first five groups and two scans for the following 12 groups) were taken for each embryo at different locations at yolk extension.

To quantify the average shift of ECM, the region of ECM was identified first. Since the ECM area has a distinctive Brillouin width, which represents the sample's viscosity, we determined the ECM region based on a threshold selection of Brillouin width. Here, we used 0.975 GHz as the threshold. The average Brillouin shift of the selected region was used to represent the tissue modulus. The longitudinal modulus of the sample was calculated based on Eqn 1. Here, λ=660 nm and θ=180°. n and ρ were obtained from published data with n=1.3398 and ρ=1.0103 g/cm³ (Schlüßler et al., 2018).

A previously characterized antisense MO oligonucleotide targeting the start site of vangl2 (5′-GTACTGCGACTCGTTATCCATGTC-3′; Gene Tools) (Park and Moon, 2002) was injected into embryos at the one-cell stage. ImageJ software was used to measure the body length, and the body length ratio was calculated by dividing the length of each injected fish by the average length of uninjected fish.

Vangl2^m209 mutants were kindly provided by Brian Ciruna (The Hospital for Sick Children, SickKids, University of Toronto, Canada). Heterozygous fish were crossed and progenies were subjected to NucView530 Caspase-3 Substrates apoptotic assay (Biotium, 10408) and MED staining to view cell organization. After image acquisition, each embryo was subjected to DNA extraction as described for ptk7a mutants and Sanger sequencing of PCR reaction with the primer pairs (F: 5′-ACCTCTTCCTGTGCGTGTTT-3′; R: 5′-GATAAACTCCTCCCCCAGGT -3′).

Sample sizes are stated in the figure legends. Animals were allocated to experimental groups based on genotype. All statistical analyses were conducted using GraphPad Prism software (v.6.04). Data are represented as mean±s.d. All statistical tests used are listed in the figure legends. P-values are reported in the figure legends and statistical significance was set at P<0.05.

We thank Brian Ciruna for providing the vangl2^m209 mutants and for helpful comments on the manuscript. We acknowledge Kessen Patten (INRS- Centre Armand-Frappier Santé Biotechnologie) for helpful advice on the study design. We also thank the Platform for Imaging by Microscopy of the Centre de recherche Azrieli du CHU Sainte-Justine for assisting with multiphoton imaging, the Faculty of Dental Medicine Electron Imaging Facility of the University of Montreal for conducting TEM imaging, and the Integrative Biosciences Center (IBio) of Wayne State University for conducting Brillouin microscopy imaging.

The peer review history is available online at https://journals.biologists.com/dev/lookup/doi/10.1242/dev.202829.reviewer-comments.pdf"