To determine whether Urp1 and Urp2 are required for spine morphology, we used CRISPR/Cas9 to generate zebrafish mutant lines. Urp1 and Urp2 are encoded by 5-exon genes with the final exon coding for the 10-amino acid peptides that are released by cleavage from the pro-domain (Figure 1A–B, Figure 1—figure supplement 1A–B). We used pairs of guide RNAs to induce deletions across the genetic region coding for the peptides (Figure 1A, Figure 1—figure supplement 2A–B). We refer to the resulting mutant lines as urp1^∆P and urp2^∆P because they lack the peptide coding sequence. In addition, mRNA quantitation revealed downregulation of urp1 and urp2 in their respective mutant backgrounds, indicating transcript decay (Figure 1—figure supplement 2E).

We first assessed urp1^∆P and urp2^∆P mutants for embryonic phenotypes. A previous morpholino-based knockdown study concluded that Urp1 and Urp2 are required for axial straightening, the process by which the ventrally curved zebrafish embryo straightens as the trunk elongates and detaches from the yolk (Figure 1C). Urp1/Urp2 morphants failed to undergo straightening and therefore displayed CTD (Zhang et al., 2018). Surprisingly, both urp1^∆P and urp2^∆P mutants underwent normal axial straightening and did not exhibit CTD (Figure 1Ei). By contrast, we observed CTD in both cfap298^tm304 mutants that lack cilia motility in the central canal (Bearce et al., 2022) and sspo^b1446 mutants in which the RF constituent SCOspondin is mutated, as expected (Figure 1Ei, Figure 1—figure supplement 3A–B). Notably, cfap298^tm304 and sspo^b1446 mutants maintained urp1 and urp2 expression in CSF-cNs, central canal neurons marked by pkd2l1 expression (Figure 1D). However, urp1 and urp2 transcripts were quantitatively reduced in cfap298^tm304 and sspo^b1446 mutants (Figure 1F). We reasoned that the absence of CTD in urp1^∆P and urp2^∆P mutants might reflect redundancy, since Urp1 and Urp2 peptides are highly similar, with identical hexacyclic regions (Figure 1B, Figure 1—figure supplement 1A–B). Alternatively, maternally derived urp1 and/or urp2 transcripts may function to prevent phenotypes from manifesting. However, maternal zygotic urp1^∆P;urp2^∆P double mutants also exhibited linear body axes (Figure 1Ei), ruling out redundant or maternal gene product function. This demonstrates that Urp1 and Urp2 peptide-null mutants undergo axial straightening.

To confirm this finding, we performed additional Urp1 and Urp2 loss-of-function experiments. By injecting four guide RNAs (gRNAs) along with Cas9 into wild-type embryos at the one-cell stage, we generated mosaic mutants, called crispants, that were then assessed for body shape phenotypes (Figure 1G). In positive control experiments, cfap298 and sspo crispants exhibited robust CTD, phenocopying germline cfap298^tm304 and sspo^b1446 mutants (Figure 1Eii). Quantitation of body curvature revealed that crispant generation was highly efficient, with CTD penetrance being close to 100% (Figure 1H). By contrast, urp1 and urp2 single and double crispants exhibited straight body axes that were not different to uninjected embryos or embryos injected with Cas9 only (Figure 1Eii and H). Using T7 endonuclease assays, we confirmed that high levels of insertion-deletion mutations were generated at gRNA sites in crispants (Figure 1—figure supplement 2C–D). We used the AB genetic background for the majority of our work, but we also generated and phenotyped urp1;urp2 double crispants on WIK and TU backgrounds to test for potential background effects. Normal axial straightening upon mutation of urp1 and urp2 was also observed on these backgrounds (Figure 1—figure supplement 2F). Overall, crispant results confirmed germline mutant findings. We conclude that Urp1 and Urp2 peptides are dispensable for axial straightening in embryonic zebrafish.

Next, we determined the impact of Urp1 and Urp2 loss on adult spine morphology. Outwardly, urp1^∆P mutant adults at 3 months post fertilization (mpf) appeared normal whereas urp2^∆P mutants exhibited minor body dysmorphologies and kinked tails (n=72 for urp1^∆P mutants and n=92 for urp1^∆P mutants, Figure 2—figure supplement 1A). To assess spine morphology directly, we imaged bone by X-ray microcomputed tomography (µCT). Three-dimensional reconstitutions of µCT data from 3 mpf fish showed that urp1^∆P mutants indeed exhibited overtly normal skeletal morphology (n=7) while urp2^∆P mutants showed slight sagittal curves (n=4; Figure 2—figure supplement 2A, Figure 2—videos 1–3). By contrast to these absent or mild deformities in single mutants, urp1^∆P;urp2^∆P double mutants exhibited prominent curves, with significant dorsal-ventral Cobb angles, a measure of deviation from straightness, especially in the caudal region of the spine (Figure 2B and D–F, Figure 2—figure supplements 1A and 2A, Figure 2—videos 1–4). These data indicate that Urp1 and Urp2 are essential for adult spine morphology, and that they function in a semi-redundant fashion in this context.

To assess the long-term maintenance of spine morphology in Urp1- and Urp2-deficient conditions, we aged urp1^∆P and urp2^∆P single mutants to 12 mpf then performed µCT. At this later time point, urp1^∆P and urp2^∆P mutants exhibited mild kyphosis-like curves though urp2^∆P mutants were more severe (Figure 2—figure supplement 2B, Figure 2—videos 5–7). These degenerative phenotypes demonstrate that Urp1 and Urp2 are essential for maintenance of spine morphology throughout adulthood and aging, and suggest that Urp2 plays a larger role than Urp1.

Urp1 and Urp2 peptides engage G-protein-coupled receptors (Ames et al., 1999; Chatenet et al., 2004; Elshourbagy et al., 2002; Labarrère et al., 2003; Liu et al., 1999; Nothacker et al., 1999). While a single urotensin II receptor (UT) gene is found in humans and has recently been linked to abnormal spinal curvature (Dai et al., 2021), the zebrafish genome encodes five such receptors. One of those, Uts2r3, was previously implicated in spine morphology (Zhang et al., 2018). To systematically compare the effects of Uts2r3 receptor mutation with loss of Urp1 and Urp2, we generated a uts2r3 mutant line harboring a 178-amino acid deletion after the third amino acid, significantly disrupting the protein (Figure 2—figure supplement 3). Like urp1^∆P;urp2^∆P double mutants, these uts2r3^b1436 mutants underwent normal axial straightening as embryos (Figure 1Ei) and went on to exhibit spinal curves as adults (Figure 2C, Figure 2—figure supplement 1A, Figure 2—video 8). Cobb angle measurements showed that uts2r3^b1436 mutants and urp1^∆P;urp2^∆P mutants were similar, though curves in urp1^∆P;urp2^∆P mutants were slightly more severe (Figure 2D–E). Like urp1^∆P;urp2^∆P mutants, uts2r3^b1436 mutants showed mostly caudally located curves, especially in the most rostral of the caudal vertebrae (Figure 2F). Thus, although we cannot rule out minor roles for other UT receptors, these data suggest that Urp1 and Urp2 control spine morphology largely by signaling through Uts2r3.

Next, we determined whether urotensin pathway mutants recapitulated any signs of disease present in patients. Several types of human spinal curves onset during adolescent growth (Cheng et al., 2015). To discern the stage of onset of curves in urotensin pathway mutants, we monitored urp1^∆P;urp2^∆P double mutant cohorts as they grew. Subtle curves first became apparent between 9 and 11 days pf (dpf), corresponding to a standard length between 3.9±0.7 mm and 5.9±0.4 mm (mean ± s.d.; Figure 3A–B, Figure 3—video 1), a stage when adolescents were rapidly growing (Figure 3C). By 13 dpf (standard length 6.2±0.3 mm), curves were evident in all urp1^∆P;urp2^∆P mutants and progressively worsened up to 17 dpf (8.3±0.4 mm) when we ended this analysis (Figure 3A–B). At 1 mpf, we assessed urp1^∆P;urp2^∆P mutants by µCT and found variability in curve position and amplitude (Figure 3D). Notably, at this stage, five of seven mutants exhibited a significant curve in the pre-caudal vertebrae, in addition to a caudal curve (Figure 3D, Figure 3—figure supplement 1Ai-Bii). Since pre-caudal curves were rare in mutants at 3 mpf (Figure 2B and F), this suggested that curve location is dynamic and that pre-caudal curves form then resolve or shift in some mutants as they grow to adulthood.

Next, we assessed whether spinal curves in urotensin pathway mutants were caused by congenital defects of vertebral patterning or structure. Staining of juveniles with the vital dye calcein revealed no defects in vertebral patterning or spacing in urp1^∆P;urp2^∆P mutants at 10 dpf (4.5–6 mm standard length [Figure 3C]) (Figure 3G). We then assessed µCT data from 3 mpf fish and quantified vertebral body shape for vertebrae 11–16, where curves occurred in urp1^∆P;urp2^∆P mutants (Figure 2F). Vertebral body length and height aspect ratios were 1.00±0.007 (mean ± s.d.) and 0.99±0.007, respectively, for wild type (n=4; Figure 3F). By contrast, urp1^∆P;urp2^∆P mutants exhibited more variable vertebral length aspect ratios (0.97±0.03, p=0.022; Bartlett’s test for equal variances, Figure 3F). Vertebral height aspect ratios in mutants (1.00±0.006) were not significantly different to controls (p=0.745, Figure 3F). These data are consistent with subtle vertebral shape defects at the points of curvature in urp1^∆Purp2^∆P mutants. However, we do not observe vertebral fusions or missing or transformed appendages, suggesting that vertebral defects do not underlie spinal curves; instead, small changes in vertebral shape are likely due to the presence of curves themselves rather than causative of curves in the first place. This is similar to what occurs in non-congenital forms of human scoliosis (Cheng et al., 2015).

Additionally, we parsed our phenotypic data for sex since spinal curves often show sex bias in severity in humans (Cheng et al., 2015), something which has been recapitulated in some zebrafish spinal curve mutants (Marie-Hardy et al., 2021). However, in both urp1^∆P;urp2^∆P and uts2r3^b1436 mutants, we found no significant differences in curve penetrance or severity between males and females (Figure 2E’).

We next compared the phenotypes of urotensin pathway mutants to other mutant lines that exhibit spinal curves. The cfap298^tm304 line harbors a temperature-sensitive mutation in cfap298, a gene required for cilia motility in several organisms including humans (Austin-Tse et al., 2013; Bearce et al., 2022; Jaffe et al., 2016). cfap298^tm304 mutants exhibit reduced cilia motility in the central canal and, if the resulting CTD is embryonically rescued by temperature shifts, develop adolescent-onset spinal curves (Grimes et al., 2016). These curves were argued to model an adolescent idiopathic scoliosis (AIS)-like condition (Figure 4A–B, Figure 4—video 1; Grimes et al., 2016; Marie-Hardy et al., 2021). Importantly, both urp1 and urp2 transcripts were significantly downregulated in cfap298^tm304 mutants (Figure 1F) suggesting that spinal curves in cfap298^tm304 might be the result of reduced Urp1/Urp2 expression. To systematically compare cfap298^tm304 mutants with urotensin-deficient mutants, we raised cohorts of urp1^∆P;urp2^∆P mutants, uts2r3^b1436 mutants, and temperature-shift-rescued cfap298^tm304 mutants alongside one another in the same aquatics facility after backcrossing all lines to the AB strain for multiple generations. At 3 mpf, we performed µCT scanning and three-dimensional reconstitutions. First, we calculated dorso-ventral Cobb angles, which revealed that cfap298^tm304 mutants were more severely curved (average total Cobb angle: 260.4±32.7°) than either urp1^∆P;urp2^∆P mutants (197.5±38.9°) or uts2r3^b1436 mutants (146.1±47.4°) (Figure 4E, Figure 2E, Figure 2—figure supplement 1D–E). Second, cfap298^tm304 mutants showed prominent dorsal-ventral curves in the pre-caudal as well as caudal vertebrae, a distinct pattern compared with the predominantly caudal curves in urp1^∆P;urp2^∆P and uts2r3^b1436 mutants (Figure 4A–C, Figure 2—figure supplement 1C). Third, cfap298^tm304 mutants exhibited significant lateral curvature of the spine, often with spinal twisting, a hallmark of AIS-like curves (Figure 4Fi–Fii, Figure 4—figure supplement 1). By contrast, urp1^∆P;urp2^∆P and uts2r3^b1436 mutants showed planar curves, with very minor or no lateral deviations (Figure 4Fi–Fii, Figure 4—figure supplement 1). These results demonstrated that cilia motility mutants and urotensin-deficient mutants exhibit distinct spinal curve phenotypes. As such, the causes of spinal curves in cfap298^tm304 mutants can only be partially explained by reduced Urp1/Urp2 expression.

To further explore the relationship between cilia motility and urotensin peptides, we generated adult cfap298^tm304;urp1^∆P;urp2^∆P triple mutants that were embryonically rescued by temperature shifts. Triple mutants exhibited significant curves, similar to cfap298^tm304 single mutants (Figure 4D–E, Figure 2—figure supplement 1C–E, Figure 4—video 2; average total Cobb angle: 284.4±33.5°). The curves of triple mutants were three-dimensional in nature, with both dorso-ventral and medio-lateral deviations, as well as incidences of spinal torsion (Figure 4Fi–Fii, Figure 4—figure supplement 1); curves were present in both pre-caudal and caudal vertebrae, as opposed to being more restricted to caudal vertebrae in urp1^∆P;urp2^∆P double mutants (Figure 2—figure supplement 1C). Overall, this suggests that motile cilia contribute to urotensin-dependent and urotensin-independent pathways controlling spine morphology.

Pkd2l1 is a polycystin family ion channel expressed in CSF-cNs, the same cell type which expresses Urp1 and Urp2 (Figure 1D; Quan et al., 2015). Pkd2l1 is responsible for flow-induced Ca²⁺ signaling in CSF-cNs (Böhm et al., 2016; Sternberg et al., 2018). While pkd2l1^icm02 mutants exhibited normal early axis development, they went on to develop mild kyphosis-like curves upon aging (Sternberg et al., 2018), a result we recapitulated after raising pkd2l1^icm02 mutants on the same genetic background and under the same conditions as urotensin-deficient mutants for direct comparison (Figure 2—figure supplement 2A–B). Notably, pkd2l1^icm02 mutants showed very subtle curves in the pre-caudal vertebrae and absence of lateral deviation at 12 mpf but no obvious curves of any kind at 3 mpf (Figure 2—figure supplement 2A-B, Figure 4—videos 3–4). This mild kyphosis-like phenotype was therefore also highly distinct from urp1^∆P;urp2^∆P and uts2r3^b1436 mutants.

Overall, our phenotypic data show that urp1^∆P;urp2^∆P mutants develop adolescent-onset curves without significant vertebral structural defects or sex bias. Coupled to the consistently caudal location of curves at 3 mpf as well as the lack of lateral deviation, we suggest that urotensin pathway mutants most closely reflect a lordosis-like condition. In agreement, urotensin mutants were phenotypically distinct from established AIS-like (cfap298^tm304) and kyphosis-like (pkd2l1^icm02) models.

Given the links between motile cilia, the RF and urp1 and urp2 expression (Figure 1D and F; Cantaut-Belarif et al., 2020; Lu et al., 2020; Zhang et al., 2018) as well as the requirement for proper RF assembly to prevent spinal curves (Rose et al., 2020; Troutwine et al., 2020), we assessed Sspo, the major component of the RF, in spinal curve mutants. To visualize Sspo localization, we used the sspo-GFP^ut24 line in which GFP coding sequence is fused to the endogenous sspo locus, producing Sspo-GFP protein (Troutwine et al., 2020). First, we assessed Sspo localization in the central canal of cfap298^tm304 mutants at 28 hr pf (hpf). In sibling controls, Sspo localized into an RF throughout the central canal (Figure 5A). By contrast, cfap298^tm304 mutants raised at restrictive temperatures, which exhibited reduced central canal cilia motility and CTD (Figure 1E; Bearce et al., 2022), lacked RF. Instead, Sspo was diffusely localized in the central canal in these mutants (Figure 5B), in agreement with previous work showing that cilia motility is required for RF assembly in embryos (Cantaut-Belarif et al., 2018).

Next, we took advantage of the temperature-sensitive nature of the cfap298^tm304 mutation to determine whether RF is also disrupted at later timepoints, during adolescent stages when true spinal curves begin to develop. To do so, we initially raised cfap298^tm304 mutants at permissive temperatures, allowing RF to correctly form and embryos to fully straighten. At 6 dpf, we transitioned larvae to restrictive temperatures (Figure 5G), which led to the appearance of curves by 11–14 dpf (standard length 6.5±0.2 mm), then assessed Sspo localization at 12 dpf. As in embryos, Sspo localized into a defined RF in the central canal in sibling controls (arrow head in Figure 5D) but was diffuse in temperature upshifted cfap298^tm304 mutants, both in mutants that had yet to develop curves and those exhibiting subtle curves (Figure 5E, Figure 5H). This demonstrates (1) that cilia motility is not only required for the initial formation of the RF but also for its maintenance; and (2) breakdown of the RF precedes curve onset in cilia motility-deficient mutants. This supports a model in which continued cilia motility maintains the RF structure, and loss of the RF causes AIS-like curves to develop in cilia motility mutants.

Next, we imaged Sspo-GFP localization in the central canal of uts2r3^b1436 mutants to determine whether RF breakdown could be causative in urotensin-associated spinal curves. In agreement with that lack of axial straightening phenotype in uts2r3^b1436 mutant embryos (Figure 1Ei), we found that Sspo localized normally into the RF in the central canal at 28 hpf in uts2r3^b1436 mutants (Figure 5C). At 12 dpf, as spinal curves were beginning to manifest (Figure 5I), Sspo still formed an intact RF in uts2r3^b1436 mutants (arrow head in Figure 5F). The fact that RF is present in uts2r3^b1436 mutants as curves form suggests that urotensin-associated curves are not caused by defective RF formation. This result coheres with the distinct spinal phenotypes exhibited by cfap298^tm304 mutants and uts2r3^b1436 mutants.

While imaging Sspo-GFP at 12 dpf, we noted that in both cfap298^tm304 mutants and uts2r3^b1436 mutants, large central canal cells with the appearance of CSF-cNs became GFP-positive (arrows in Figure 5E–F), something we rarely observed in control fish (Figure 5D). Indeed, the GFP-positive central canal cells in uts2r3^b1436 give the effect of making the RF appear comparatively smaller/dimmer (compare RF in Figure 5F and Figure 5D). We suggest that CSF-cNs may endocytose Sspo-GFP monomers in cfap298^tm304 mutants where the RF has broken down and in uts2r3^b1436 mutants where RF is likely to be making increasing numbers of contacts with CSF-cNs (Orts-Del’Immagine et al., 2020) owing to the onset of spinal curves.

AB, TU, and WIK strains of D. rerio were used. Zebrafish lines generated were sspo^b1446, urp1^b1420 (called urp1^∆P), urp2^b1421 (called urp2^∆P), uts2r3^b1436 as well as previously published lines: cfap298^tm304 (Jaffe et al., 2016), pkd2l1^icm02 (Sternberg et al., 2018), and sspo-gfp^ut24 (Troutwine et al., 2020). Experiments were undertaken in accordance with research guidelines of the International Association for Assessment and Accreditation of Laboratory Animal Care and approved by the University of Oregon Institutional Animal Care and Use Committee (Protocol number 21–45). All zebrafish strains and other materials are available upon request.

Generation and genotyping of mutant lines sspo^b1446, urp1^b1420, urp2^b1421, and uts2r3^b1436 were generated using CRISPR/Cas9. gRNA oligos were designed using CRISPRscan (Moreno-Mateos et al., 2015). gRNA templates (IDT) were assembled by annealing and extension with Bottom strand ultramer_1 (Key resource table) using Taq Polymerase (NEB, M0273) with cycling parameters of 95°C (3 min), 95°C (30 s), 45°C (30 s), 72°C (30 s), and 72^oC (10 min) with 30 cycles of the middle three steps. PCR product was purified (Zymo DNA Clean and Concentrator Kit, D4013) and then used for in vitro RNA synthesis using a MEGAshortscript T7 Transcription Kit (ThermoFisher, AM1354). Synthesized gRNAs were purified (Zymo RNA Clean and Concentrator Kit, R1013), then 150 pg along with 320 pg/nl Cas9 (IDT, 1081058) were injected into one-cell stage fertilized eggs. The mosaic mutant fish resulting from these injections (F₀ fish) were raised and outcrossed to AB wild-types, and DNA was extracted from the resulting F₁ embryos. Mutant alleles were screened by PCR coupled with restriction enzyme digestion and/or Sanger sequencing (GeneWiz). Embryos from F₁ clutches harboring mutations were raised to adulthood and outcrossed to AB wild-types to generate F₂ families which were screened for mutations and raised. The nature of mutations was identified by sequencing DNA extracted from adult fin clips of F₂ heterozygous fish using CRISP-ID to deconvolute (Dehairs et al., 2016) and confirmed by sequencing DNA of F₃ homozygous embryos.

urp1^b1420 mutants contain a 279 bp deletion and 1 bp insertion that were genotyped by PCR amplification with urp1_geno_1 and urp1_geno_2 primers which generate a 460 bp band from wild-type DNA and a 184 bp band from mutant DNA. urp2^b1421 mutants harbor a 61 bp deletion and were genotyped by PCR amplification with urp2_geno_1 and urp2_geno_2 primers followed by gel electrophoresis to distinguish the 283 bp wild-type band and the 226 bp mutant band. uts2r3^b1436 mutants contain a 534 bp deletion and were also genotyped by PCR, using uts2r3_geno_1 and uts2r3_geno_2, in which wild-type sequence led to an 832 bp band and mutant sequence a 298 bp band.

The nature of the sspo^b1446 mutation was determined by whole genome sequencing. DNA was extracted from mutant embryos using a phenol/chloroform procedure. Libraries were prepared using the FS DNA Library Prep Kit for Illumina sequencing (NEB, E7805). DNA was digested into 150 bp fragments, and paired-end sequencing was performed using a NovaSeq 6000 Sequencing System. Trimmomatic (version 0.36; ILLUMNIACLIP: TruSeq3-PE-2.fa:2:30:10:1:true LEADING:3 TRAILING:3 SLIDINGWINDOW:5:20 MINLEN:42 AVGQUAL:30) was used to remove Illumina adaptor sequences from paired-end reads. Illumina short-read sequences were then aligned to the GRCz11 reference sequence of chromosome 24 using BWA-MEM (version 0.7.01). SAMtools (version 1.8) was used to sort and index reads. Aligned reads in BAM format were analyzed in IGV (version 2.13.1). Mutants were routinely genotyped by PCR amplification with oligos sspo_geno_1 and sspo_geno_2, followed by BsaI-HFv2 (NEB, R2733) restriction digestion to produce 300 bp and 99 bp bands from wild-type DNA and a protected 399 bp band from mutant DNA.

Four gRNA oligos per gene for cfap298, sspo, urp1, and urp2 were chosen from a look-up table (Key resource appendix; Wu et al., 2018). gRNAs were synthesized from oligos in multiplex. After being pooled at 10 µM, oligos were annealed and extended with Bottom strand ultramer_2 using Phusion High-Fidelity PCR Mastermix (NEB, M0531) with Phusion High-Fidelity DNA Polymerase (NEB, M05030) using incubations: 98^oC (2 min), 50°C (10 min), and 72^oC (10 min). Assembled oligos were purified and used as templates for in vitro RNA synthesis, as described in ‘Generation and genotyping of mutant lines’ section. For mutagenesis, 1000 pg of gRNAs along with 1600 pg/nl Cas9 (IDT, 1081058) were injected into one-cell-stage embryos. To assess rates of mutagenesis, DNA was extracted from 1 dpf crispants and subjected to T7 endonuclease I assays (NEB, E3321).

Zebrafish larvae at 28–30 hpf were imaged using a Leica S9i stereomicroscope with integrated 10-megapixel camera. Body angles were calculated using ImageJ (Schindelin et al., 2012) as described in Bearce et al., 2022.

RNA was extracted using a Zymo Direct-Zol RNA Miniprep kit (Zymo Research, R2051). cDNA was prepared from 25 ng of RNA using oligoDT primers in a 20 µl reaction using a High Capacity cDNA Reverse Transcription Kit (ThermoFisher, 4368814). qRT-PCR reactions were performed in real time using 5 µl PowerUp SYBR Green Master Mix (ThermoFisher, A25741), 0.8 µl of 10 µM forward and reverse primers, 1.4 µl of nuclease-free water, and 2 µl of diluted cDNA. PCR was performed using a QuantStudio Real Time PCR System (Applied Biosystems) with cycling parameters: 50°C (2 min), 95°C (10 min) then 40 cycles of 95°C (15 s), and 60°C (1 min). Each reaction was performed in quadruplicate. Quantitation was relative to rpl13 and used the ∆∆C_T relative quantitation method in which fold changes are calculated as 2^−∆∆CT. The efficiency of amplification was verified to be close to 100% with a standard curve of RNA dilutions.

Larvae were incubated in water containing 0.2% calcein (Sigma-Aldrich, C0875) for 10 min then rinsed two to three times in water (5 min per rinse). Larvae were immobilized with 0.005% tricaine, mounted in 0.8% low melt agarose, and imaged with a Leica THUNDER stereoscope.

Embryos were fixed in 4% paraformaldehyde at 4°C overnight, washed with phosphate buffered saline (PBS) then serially dehydrated to 100% methanol, and stored at –20°C. Embryos were rehydrated, washed with PBS containing 0.1% Tween-20, incubated in hybridization buffer (Molecular Instruments), then incubated in 2 pmol of probes at 37°C overnight in a total volume of 500 µl of hybridization buffer. Embryos were washed in wash buffer (Molecular Instruments), washed twice in 5× SSCT (sodium chloride sodium citrate with 0.1% tween-20), and then incubated in amplification buffer (Molecular Instruments) for 1 hr. RNA hairpins designed to bind either pkd2l1, urp1, or urp2 were prepared by heating 10 pmol of each to 95°C for 90 s then snap-cooled in the dark for 30 min. Embryos were then incubated overnight in 500 µl of amplification buffer containing 30 pmol hairpins at room temperature in the dark. Embryos were washed five times in 5× SSCT, stored at 4^oC, and then mounted for confocal microscopy. Images were acquired using a Zeiss LSM880 using either a ×20 air or ×40 water objective. Acquisition settings were derived using wild type embryos and then applied to all embryos. Images were exported to IMARIS 9.5.0 (Oxford Instruments). A Gaussian filter of width 0.42 µm (×20) or 0.21 µm (×40), and a rolling ball background subtraction of 10 µm was applied.

Scans were performed using a vivaCT 80 (Scanco Medical) at 18.5-µm voxel resolution (for 3 mpf and 12 mpf fish) or 10-µm voxel resolution (1 mpf fish) as previously described (Bearce et al., 2022). Digital dissections of the spinal column were performed in 3D Slicer (Kikinis et al., 2013) using the Segmentation Editor. A threshold of 3200 was used to mask 1 mm tube (Draw Tube function) around the spine in the axial slice view, beginning between the otic vesicles rostral to the first vertebrae and ending at the split of the tail. The center of the tube was set at the narrowest ‘hollow”’ within the lumen of centra.

Quantitation of lateral spine curvature was performed in ImageJ by orienting isolated spine images in a dorsal view with heads to the left and with the otic vesicles and first Weberian vertebrae parallel to the x-axis. Landmarks were assigned to the narrowest point of each centrum rostral to caudal; where maximum projection resulted in hidden or overlapping vertebrae, the appropriate number of points was added in closest approximation. The y-value from each landmark was subtracted from the point rostral to it, resulting in a map of local deflections where positive values indicate rightward displacement, and negative values indicate leftward displacement.

Embryos (28 hpf) and larvae (12 dpf) were anesthetized in tricaine until touch response was abolished and then embedded in 0.8% low-melt agarose laced with tricaine in inverted imaging chambers (14 mm #1.5 coverslips, VWR cat no. 10810–054). In larvae, care was taken to align the posterior body close to the coverslip to the RF within the working distance of the objective. A Nikon Ti2 inverted microscope equipped with Plan Apo ×40 and ×60 WI DIC (1.2 NA) objectives, a Yokogawa Spinning Disk and pco.edge sCMOS camera were used to capture 512×256 images in time series. Exposure time varied with age (100 ms–300 ms) as Sspo-GFP brightened in intensity over time; exposure, camera settings, and laser power were kept constant between age-matched individuals. Images were cropped, rotated, and intensity-adjusted in ImageJ (Schindelin et al., 2012).