A CRISPR/Cas9-generated mutation in the zebrafish orthologue of PPP2R3B causes idiopathic scoliosis

Idiopathic scoliosis (IS) is the deformation and/or abnormal curvature of the spine that develops progressively after birth. It is a very common condition, affecting approximately 4% of the general population, yet the genetic and mechanistic causes of IS are poorly understood. Here, we focus on PPP2R3B, which encodes a protein phosphatase 2A regulatory subunit. We found that PPP2R3B is expressed at sites of chondrogenesis within human foetuses, including the vertebrae. We also demonstrated prominent expression in myotome and muscle fibres in human foetuses, and zebrafish embryos and adolescents. As there is no rodent orthologue of PPP2R3B, we used CRIPSR/Cas9-mediated gene-editing to generate a series of frameshift mutations in zebrafish ppp2r3b. Adolescent zebrafish that were homozygous for this mutation exhibited a fully penetrant kyphoscoliosis phenotype which became progressively worse over time, mirroring IS in humans. These defects were associated with reduced mineralisation of vertebrae, resembling osteoporosis. Electron microscopy demonstrated abnormal mitochondria adjacent to muscle fibres. In summary, we report a novel zebrafish model of IS and reduced bone mineral density. In future, it will be necessary to delineate the aetiology of these defects in relation to bone, muscle, neuronal and ependymal cilia function.

Scoliosis is the lateral deformation and curvature of the spine which affects approximately 4% of the general population 1 . It has a complex aetiology with many potential cellular and tissue-level causes, and broadly three forms of scoliosis can be defined. Congenital scoliosis is present at birth and may result from a primary defect of the bone that constitutes the vertebrae, and indeed some mouse models of scoliosis exhibit alterations of osteoblasts and chondrogenesis 2 . Neuromuscular scoliosis is associated with neural or muscular abnormalities such as spina bifida or muscular dystrophy, which emphasises the importance of these tissues in maintaining spinal integrity. In these cases, scoliosis may arise secondarily to defects in proprioception 3 , which refers to the body's sense of position and reaction to external stimuli. This is driven by connections between the interneurons that relay signals such as pain to motor neurons within the spinal cord, which in turn control muscle activity. Idiopathic scoliosis (IS) is not present at birth and generally develops progressively in children over the first decade of life without an overt anatomical cause. It may involve defects in the intervertebral discs and matrisome, ependymal cilia and cerebrospinal fluid flow, neuroinflammation etc. 4 . It is therefore important to define the primary cellular origins of these different forms of scoliosis.
PPP2R3B encodes the PR70 protein, a regulatory subunit of the heterotrimeric protein phosphatase 2A holoenzyme. Relatively little is known about PR70 function, although it was identified as interacting with the origin of replication complex component, CDC6, and has also been linked to regulation of autophagy [5][6][7] . PR70 has also been proposed to act as a tumour suppressor by regulating firing of DNA replication origins 8 . In this study we have used CRISPR/Cas9-mediated gene-editing to create a frameshift mutation in the zebrafish ppp2r3b gene. Homozygous mutant larvae appeared normal, while adolescents developed progressive kyphoscoliosis reminiscent of IS. PPP2R3B showed prominent expression in axial muscle in human foetuses, zebrafish embryos and adolescents. The mutant kyphoscoliosis phenotype is associated with reduced vertebral bone mineralisation and a muscular dystrophy-like phenotype. These data identify ppp2r3b as a new molecular target in the pathogenesis of IS in vertebrates. In situ hybridisation at low power showing expression of PPP2R3B within interneuron/motor neuron precursors (np), dorsal root ganglia (drg) and myotome (m) but no signal generated using a GFP negative control. (C, D) Expression of PPP2R3B in vertebral bodies. Insets show regions magnified in C' and D'. Note expression in chondrocytes (c). PPP2R3B is also expressed in the notochord (n) whereas SOX9 is not. (E-H) Expression of PPP2R3B within Meckel's cartilage (insets magnified in E'-H'). Note expression co-localises with SOX10 and SOX9 within chondrocytes (c) but expression is absent from perichondrium (p). nt, neural tube; g, gut; k, kidney; tel, telencephalon; e, eye; ns, nasal septum; t, tongue. Scale bars are 500 µm in (A-D, A' , C'); 60 µm in (C' , D'); 1 mm (E-H); 10 µm in (E'-H'). www.nature.com/scientificreports/ a role in chondrogenesis, although it was not expressed within the perichondrium where osteoblast precursors reside prior to their migration into the cartilage matrix ( Fig. 1C' , E'). In both of these locations, PPP2R3B expression was similar to that of SOX9 on adjacent sections (Fig. 1D' , G' , H'). Within Meckel's cartilage, chondrocytes within cartilage condensations also expressed SOX10, as did the perichondrium (Fig. 1F'). SOX10 is a marker of neural crest stem cells confirming the contribution of this lineage to skeletal elements within the jaw. We note that PPP2R3B expression was quite broad, albeit with accentuation of signal in certain locations, including myotome and vertebral chondrocytes, as shown by intermediate power images (Fig. 1A' , C''). By aligning the human PR70 protein sequence to the zebrafish translated genome, we identified only a single orthologue with significant similarity, and the genomic locus encoding Ppp2r3b showed conserved synteny with their mammalian counterparts. Orthologues of neither PPP2R3B nor the adjacent gene, SHOX, are found in rodents. We therefore analysed expression of the orthologous zebrafish ppp2r3b gene by in situ hybridisation ( Fig. 2A, B). At 24 h post-fertilisation (hpf) we noted repeated chevron-shaped patterns of expression along the trunk of the embryo representing the somites which will go on to form axial muscle. Within the head region, we also noted rostral expression (Fig. 2C).
We also analysed expression in zebrafish at 36 dpf of age. We noted prominent expression throughout the region in which fast twitch muscle is thought to reside, adjacent to vertebral centra ( Fig. 2D-H). At this stage, a number of distinct muscle bundles separated by myospeta can be seen on transverse tissue sections (Fig. 2D). ppp2r3b is seen to be expressed prominently in these (Fig. 2E), and it is notable that staining intensity may vary between bundles, although this may relate to tissue sectioning quality (Fig. 2E). Within these muscle bundles, higher power imaging shows that ppp2r3b expression appears to surround individual muscle fibres (arrowheads, Fig. 2G, H) which is consistent with nuclear staining of multinucleate muscle fibres. In contrast there was only limited expression in proximity to mineralised bone-based on their morphology and location on vertebral bone surfaces, we did identify expression in what could be squamous chordoblast (osteoblast) -like cells, as previously reported 9 , although it is noteworthy that these cells were very rare (Fig. 2I). It should be noted that to definitively identify these cells as chordoblasts it will be necessary to identify specific markers for these cells in co-localisation studies in future.
Generation of ppp2r3b mutant zebrafish using CRISPR/Cas9 gene-editing. To generate a genetic model of ppp2r3b loss-of-function, we used CRISPR/Cas9 gene-editing to target this gene using a sgRNA located within exon 2 (Fig. 3A). This sgRNA was located on the forward strand and had no self-complementarity or predicted off-target sites according to the chopchop tool (http:// chopc hop. cbu. uib. no/). An Mse I restriction site was located within the sgRNA binding site which allowed us to monitor the efficiency with which indels were introduced at this location. Direct sequencing of a selection of cloned mutations from mosaic F0 embryos at 24 hpf was consistent with many previous reports in zebrafish, showing that CRISPR/Cas9 typically produces complex indels involving deletions of between 2 and 18 nucleotides (Fig. 3A).
We have now outcrossed these F0 mosaics and their progeny for more than 5 generations to achieve germline transmission and to avoid possible off-target mutations. We isolated a line of zebrafish carrying a 7 bp deletion in exon 2 of ppp2r3b which results in the frameshift mutation p.Ala31ValfsX150 (Fig. 3B). Homozygous mutants are hereafter referred to as ppp2r3b −/− . During the course of our breeding and genotyping, we noted a single nucleotide polymorphism (SNP) located within the sgRNA binding site which is not present on publicly available databases. This SNP encodes the single amino acid substitution p.Ser33Asn. In all subsequent analyses, we selected only heterozygotes whose wild-type allele encoded the reference SNP at this location in our breeding population. RT-PCR and direct sequencing of gel extracted products using primers located in exons 1-3 or 1-7 failed to identify gross alternative splicing within ppp2r3b transcripts generated from pooled 24 hpf embryos from a hetxhet incross or individual homozygous mutant animals at 48 dpf of age-these products were of the predicted size, as in wild-types, and qRT-PCR demonstrated that transcript levels in mutants were not different from wild-type ( Fig. 3C, D, Table 1). We noted that there was no deviation from expected Mendelian ratios showing that this mutation does not affect viability (Fig. 4). We also endeavoured to generate a pool of homozygous mutant adults with which to breed maternal-zygotic mutant zebrafish. This was not possible, because homozygotes never produced any eggs, and thus we conclude that they are infertile. Notably, crosses of homozygous females with wild-type males readily produced viable fry, whereas the opposite did not. ppp2r3b homozygous mutant zebrafish exhibit a fully penetrant scoliosis phenotype. We did not identify any phenotypic abnormalities in heterozygous or homozygous mutants at larval stages. At 48 dpf, we noted that homozygotes developed severe kyphoscoliosis (Fig. 5A, B). The pattern of kyphoscoliosis was very stereotypical, characterised by two ventral curves located within the precaudal and caudal vertebrae at numbers 7-9 and 25, respectively. These ventral curves flanked a dorsal curve located at approximately caudal vertebrae number 18. There was often also a sharp lateral bend within the caudal fin, although this was not as consistent. At this age, wild-type siblings never exhibited kyphoscoliosis and the spine exhibited a very gentle ventral curvature within the precaudal region only. Scoliosis is a common phenotype in old zebrafish, presenting in excess of 18 months of age in our aquatics facility but never earlier than this. This is often associated with Mycobacterium chelonae infection, however, ongoing microbiological testing confirms that this species is absent from our facility.
We monitored the onset and progression of scoliosis in ppp2r3b −/− mutants. Scoliosis was first seen at 36 dpf ( Fig. 5)-no axial defects were observed before this as we stained animals for Alizarin red and Alcian blue as early as 15 dpf (Fig. 6). At 36 dpf, the typical presentation was moderate ventral curvature within the precaudal region, with relatively little curvature of the caudal vertebral regions. However, by 48 dpf, the final pattern consisting of two ventral curves and one dorsal curve was apparent. Quantification of the proportion of animals with www.nature.com/scientificreports/ moderate or severe kyphoscoliosis confirmed that this phenotype became worse with time ( Fig. 5C). At 48 dpf, the kyphoscoliosis phenotype was present in all homozygous mutants, but not in any wild-type or heterozygous siblings. Therefore, ppp2r3b −/− mutant zebrafish exhibit adolescent onset and progressive kyphoscoliosis, which is fully penetrant and reminiscent of human IS.
To confirm that this phenotype is not the product of off-target mutations we also generated a second frameshift mutant line following microinjection of a ribonucleoprotein complex consisting of sgRNAs targeting exon 1 in complex with Cas9. This generated a 19 bp deletion, causing an out-of-frame p.L82fsX24 mutation. Notably, homozygous zebrafish for this mutation also exhibited profound scoliosis (Fig. 7).
Kyphoscoliosis in ppp2r3b mutants is associated with reduced bone mineralisation of vertebrae. To investigate this phenotype further, we analysed bone mineralisation and cartilage formation in www.nature.com/scientificreports/ ppp2r3b -/mutants. Precaudal vertebrae 5-13 include a neural spine, which projects dorsally, and two ventrally located ribs, while the caudal vertebrae include neural and haemal spines which mirror one another in size.
Alizarin red staining showed that the ratio between the lengths of the neural and haemal spines within the caudal region were approximately equal in length in homozygous mutants as in wild-types and heterozygotes ( Fig. 8A-C). Within the precaudal region, the ribs are approximately 2.5 times longer that the neural spines (dorsal:ventral ratio of 0.4), however, we found that the ribs were relatively shorter in ppp2r3b homozygotes as compared to wild-types or heterozygotes (Fig. 8C), suggesting a defect in patterning and/or ossification. We also noted a marked reduction in Alizarin red staining intensity throughout the vertebral body and spines/arches, which was uniform across all vertebrae in caudal and precaudal regions (Fig. 8B).
To investigate bone formation in more detail, we performed microCT scanning to compare skeletal tissue parameters of vertebrae at 36 dpf which represents the onset of scoliosis. This Alizarin red staining suggested that the gross structure of all vertebrae was normal without the characteristic wedging of vertebrae that has been reported previously 4 , even at the sites of curvature (Fig. 6A, B). Indeed, whereas 3D renders and longitudinal sections through contiguous vertebrae showed that adjacent vertebrae were closely apposed in wild-types, with a uniform and narrow intervertebral space, the intervertebral spaces were wedge-shaped in mutants, corresponding www.nature.com/scientificreports/ with the direction of curvature (Fig. 9A, B). Remarkably, we also found that multiple holes were apparent throughout the mutant vertebrae (Fig. 9A, B). Measurement of tissue mineral density (TMD), which is specific to cortical bone and is appropriate for analysis of non-trabeculated vertebral bone, was significantly reduced (Fig. 9C). However, the overall dimensions of the vertebrae, including length and diameter, were not affected suggesting a specific effect on bone mineralisation rather than morphogenesis. qRT-PCR of RNA extracted from the trunk at 36 dpf did not reveal any changes in the expression of a panel of key bone cell markers (Table 1). Histological analyses of osteoclasts and osteoblasts did not reveal differences between wild-type and mutant (Fig. 10)-Mallory's trichrome staining showed similar numbers of vacuolated chordoblasts (osteoblasts) and squamous chordoblasts within the notochord centra and sheath, respectively. Tartrate-resistant acid phosphatase (TRAP) staining labels osteoclasts and revealed similar staining in neural and haemal arches in both wild-type and mutant, although no staining was detected within the vertebrae centra which is where the reduced TMD was Table 1. qPCR analysis of bone and muscle marker genes. www.nature.com/scientificreports/ observed previously. These data are subject to technical limitations relating to the proportionally small number of bone cells in whole zebrafish tissues at this stage and the non-quantitative nature of histological methods and are therefore not conclusively negative. In future, it will be necessary to analyse these cell types in more detail.
Abnormal mitochondria associated with axial muscle in ppp2r3b mutants. Given the link between progressive spinal curvature and proprioception, and the expression of PPP2R3B/ppp2r3b that we detected in muscle and bone, we monitored muscle fibre and neuromuscular junction formation in our mutant fish. Initially we analysed skeletal muscle birefringence at 48hpf, an established indicator of muscle fibre integrity based on polarised light transmission through striated fibres. No obvious changes in birefringence signal were detected between ppp2r3b mutants and wildtype siblings (Fig. 11A). To challenge a role for PR70 in neuromuscular development, embryos were further stained for F-Actin and Acetycholine receptor (AChR), using fluorescently conjugated phalloidin and α-bungarotoxin. Muscle fibres and AChR localisation appeared organised and indistinguishable between ppp2r3b −/− embryos and wild-type siblings (Fig. 11B). Thus, PR70 is not required for normal neuromuscular development.
To evaluate muscle development during juvenile growth, at a point when scoliosis has manifested, transmission electron micrographs were produced from muscle biopsies. ppp2r3b −/− fish displayed normal myofibril organisation and sarcomeric assembly ( Fig. 11C and D), and consistent with this, qRT-PCR analyses showed that expression of key muscle markers was unchanged (Table 1). However, a marked reduction in mitochondrial content was observed in ppp2r3b mutants (Fig. 11C). Closer examination of the mitochondria showed general dysmorphic character in the mutants with undefined cristae and smaller overall size (Fig. 11D). Quantification of mitochondrial area adjacent to sarcomeres showing a statistically significant reduction in mutants (Fig. 11E). Taken together, these data suggest that ppp2r3b, whilst not required for the formation of muscle, is required to maintain muscle mitochondrial abundance.

Discussion
In this study, we have created a targeted mutation in ppp2r3b in zebrafish which resulted in juvenile onset, progressive and fully penetrant kyphoscoliosis closely mirroring IS, as well as abnormal mitochondria in the vicinity of axial musculature. Collectively the following pieces of information indicate that this results from a loss of PR70 protein function. While we did not see any differences in transcript levels in mutants by qRT-PCR and no specific antibody is available to detect PR70 protein, we did show that all detectable isoforms of ppp2r3b included exon 2, which carries the engineered frameshift mutation. This is located at residue 31 of the encoded protein which would therefore lead to deletion of two EF-hand motifs located within the C-terminal half of the protein which are essential for calcium binding and protein function 6 . Furthermore, protein sequence analysis does not indicate that there are any translation initiation codons downstream of the mutation which could produce protein product not involving the frameshift mutation. Another consideration relates to potential off-target mutations produced by the gene-editing process. Having bred the mutation out for a number of generations, any putative off-target mutations would be lost and certainly not be expected to segregate with the observed phenotype. By contrast, kyphoscoliosis segregated perfectly (100% penetrance) with the frameshift mutation over hundreds of animals and several generations, indicating linkage to the mutation.
The zebrafish has emerged as a powerful model to study IS 4 . By generating a zebrafish mutant in ppp2r3b, we have been able to investigate the requirement for this gene beyond larval stages, thereby revealing a requirement for normal bone mineralisation. Two key findings in relation to the potential cellular mechanism of pathogenesis (2) the identification of reduced vertebral bone density and altered intervertebral structure, as well as mitochondrial defects similar to those previously reported in muscular dystrophy models which followed normal muscle formation and occurred in the absence of muscle fibre (Z-disc) defects 10 . Both the skeletal and muscle defects that we report were degenerative in nature. Collectively, these observations suggest that kyphoscoliosis occurs secondarily either to defects in muscle function and/or because of disrupted communication between muscle and bone. The holes in the vertebrae, representing sites of www.nature.com/scientificreports/ low mineralisation were particularly striking. Whether this is a primary cause of IS or an unrelated secondary consequence of loss of spinal integrity remains to be tested. Nonetheless, resembling osteoporosis, this is likely to contribute to the overall pathology. Bending of the spine seems to correlate with intervertebral disc wedging which could be a driving force in pathogenesis, and so the role of PR70 in intervertebral disc formation and maintenance must be investigated further. Expression of PPP2R3B appeared to be mutually exclusive with that of SOX9, a marker of cartilage lineage, at CS-23 but not at CS-22. It is possible that this represents transient colocalization with chondrogenic niches during development and further studies will be required to test this. If so, this may imply that role of ppp2r3b is primarily within the intervertebral discs rather than vertebral bone and will require future investigations. Results from our electron microscopy studies suggest a defect in mitochondria adjacent to muscle fibres, which themselves appeared structurally normal. In future it will be necessary to dissect the causal relatioships between muscle function, mitochondria and vertebral bone/intervertebral disc integrity. The expression pattern of PPP2R3B in human foetuses is very broad and, at a high level, it looks to be specifically enriched in the nervous system (especially along ventricular surfaces). A growing body of evidence from zebrafish scoliosis models has implicated ependymal cell cilia, circumventricular organ system and cerebrospinal flow defects in the pathogenesis of idiopathic scoliosis. ppp2r3b mutant zebrafish embryos do not display features associated with other models of ciliary dysfunction, such as body curvature or gastrulation defects, and we did not detect neuronal defects in our own analyses. Nevertheless, in future mechanistic studies, it will be important to undertake more detailed analyses of neurophysiological origins of scoliosis.
At the molecular level, PPP2R3B encodes the DNA origin of replication complex (ORC) regulator, PR70 5,6,8 . Core ORC components, such as CDC6, are mutated in Meier-Gorlin syndrome which features a variety of skeletal defects including scoliosis 11,12 . However, to our knowledge, nothing is known about the molecular or cellular function of the ORC in skeletogenesis. A zebrafish cdc6 mutant has been reported, but post-larval phenotypes www.nature.com/scientificreports/ are potentially confounded by severe early embryonic defects and no skeletal features were reported 13 . There was a recently identified role in autophagy as well 7 . The mouse homologue Ppp2r3a (mgi 2442104) is reported to have decreased bone mineral content. The zebrafish model that we report here provides a basis to investigate the molecular and cellular mechanisms of pathogenesis in future.

Materials and methods
Zebrafish. Zebrafish (Danio rerio) embryos were obtained from a wild-type strain and raised in E3 Medium until the desired developmental stage is reached. This solution contains 5 mM NaCl, 0.17 mM KCl, 0.33 mM CaCl 2 , 0.33 mM MgSO 4 and 0.1% Methylene Blue. This medium serves as a control medium and will not have any effect on patterning.

Whole-Mount in situ hybridisation on zebrafish.
Chorions were removed manually prior to fixation in 4% paraformaldehyde (PFA) in 1 × PBS overnight at 4 °C. Fixed embryos were dehydrated in 100% Methanol (MeOH) for 15 min. at room temperature. Embryos were cooled to -20 °C for at least 30 min for permeabilization prior to proceeding with in situ hybridization experiments. Whole-mount in situ hybridization was performed following the protocol from ZFIN. Westerfield 14 .
In situ hybridisation on human tissue. 7  Template for PCR amplification of PPP2R3B and SOX10 was human cDNA. The SOX9 probe has been reported previously 15 . Human embryonic samples were fixed in 10% (w/v) neutral-buffered formalin solution (Sigma-Aldrich) and embedded in paraffin wax before sectioning. ISH was performed in 300 mM NaCl, 5 mM EDTA, 20 mM Tris-HCl, 0.5 mg/mL yeast tRNA, 10% dextran sulfate, 1 × Denhardt's solution, and 50% formamide with digoxigenin-incorporated riboprobes at 68 °C. Posthybridization slides were incubated with anti-digoxigenin conjugated with alkaline phosphatase (Roche) diluted 1:1000 in 2% newborn calf serum. Expression patterns were visualized with a Nitro-Blue Tetrazolium   . RNP complexes were established following an adapted protocol from Kroll et al. 16 . In brief, 2 nmol crRNA for each target was resuspended in 10 µl duplex buffer (IDT, #11-01-03-01). 5nmole tracrRNA (IDT, #1072532) was resuspended in 25 µl Duplex buffer. Each crRNA was annealed to the tracrRNA by combining 1 µl of crRNA with 1 µl of tracrRNA and 1.5 µl of duplex buffer, components were incubated at 95 °C in a thermocycler. The Cas9 was then complexed to the annealed guide RNAs by adding 1 µl of 10 µg/µl Cas9 endonuclease (Alt-R S.p. HiFi Cas9 nuclease, IDT, #1081060) to each of the 3.5 µl guide RNA solutions and incubated at 95 °C for 5 min. All three complexed RNPs were then pooled, mixed, aliquoted and stored at − 20 °C until required. 0.5 nl of pooled RNPs were intracellularly microinjected into wildtype (AB × TupLF) embryos at the one-cell stage and embryos allowed to develop until adulthood.

Data availability
All data sets and raw data are provided in this manuscript and are available to all readers. Please contact the corresponding author for further information if required.