Mutations in KIF7 Implicated in Idiopathic Scoliosis in Humans and Axial Curvatures in Zebrafish

Abstract

Idiopathic scoliosis (IS) is a spinal disorder affecting up to 3% of otherwise healthy children. IS has a strong familial genetic component and is believed to be genetically complex due to significant variability in phenotype and heritability. Previous studies identified putative loci and variants possibly contributing to IS susceptibility, including within ECM, cilia and actin networks, but the genetic architecture and underlying mechanisms remains unresolved. Here, we used whole exome sequencing from three affected individuals in a multigenerational family with IS and identified 19 uncommon variants (MAF <0.05). Genotyping of additional family members identified a candidate heterozygous variant (H1115Q, G>C, rs142032413) within the ciliary gene KIF7, a regulator within the hedgehog (Hh) signaling pathway. Resequencing of a second cohort of unrelated IS individuals and controls identified several severe mutations in KIF7 in affected individuals only. Subsequently, we generated a mutant zebrafish model of kif7 using CRISPR-Cas9. kif7^co63/co63 zebrafish displayed severe scoliosis, presenting in juveniles and progressing through adulthood. We observed no deformities in the brain, Reissner fiber, or central canal cilia in kif7^co63/co63 embryos, although alterations were seen in Hh pathway gene expression. This research suggests defects in KIF7-dependent Hh signaling may drive pathogenesis in a subset of individuals with IS.

Introduction

Idiopathic scoliosis (IS) (MIM# 181800) is a common pediatric spinal deformity defined as a structural lateral spinal curvature of at least 10° with spinal rotation. IS is an isolated skeletal disorder that occurs in approximately 2–3% of otherwise normal children across populations (Asher & Burton, 2006) and is known to run in families (Cowell, Hall, & MacEwen, 1972; Grauers, Einarsdottir, & Gerdhem, 2016; Riseborough & Wynne-Davies, 1973; Wynne-Davies, 1968). Severe curves occur in approximately 0.2–0.5% of children, who are predominantly female (Kane & Moe, 1970; Stuart L. Weinstein, 1994; S. L. Weinstein, Zavala, & Ponseti, 1981). Despite a significant understanding of the clinical characteristics and natural history of IS, a unifying hypothesis of etiology is lacking, reflecting both the phenotypic and genetic complexity of this disorder. Genome-wide association studies of IS (W. Gao et al., 2013; Kou et al., 2013; Kou et al., 2018; Nelson, Chettier, Ogilvie, & Ward, 2011; Ogura et al., 2015; Sharma et al., 2011; Sharma et al., 2015; Takahashi et al., 2011; Zhu et al., 2015) and familial linkage (Alden et al., 2006; Einarsdottir et al., 2017; X. Gao et al., 2007; Gurnett et al., 2009; Justice, Miller, Marosy, Zhang, & Wilson, 2003; Marosy et al., 2010; Miller et al., 2005; Miller et al., 2012; Miller et al., 2006; Raggio et al., 2009; Wise et al., 2000) and exome studies (Baschal et al., 2018; Baschal et al., 2014; Buchan, Alvarado, et al., 2014; W. Gao et al., 2017; Grauers et al., 2015; Haller et al., 2016; Li et al., 2016; Patten et al., 2015) have identified variants in multiple molecular pathways (Grauers et al., 2016) which have initiated a good foundation for functional and animal model studies. The theory that IS may be caused by multiple, variable risk loci has gained traction (Grauers et al., 2016) as only a minority of individuals with IS have been shown to possess any single known risk variants, including those within or near LBX1 or ADGRG6 (GPR126), the loci most replicated thus far across diverse populations by GWAS. Specific causal mutations may differ family by family, a concept reflected in our familial exome sequencing study, in which no single genes or variants were shared across all families (Baschal et al., 2018). It is yet to be seen whether mutations within shared molecular pathways may cause IS, even though there are no specific, individual variants responsible for the majority of IS cases.

Most recently, both human and vertebrate studies have related primary and motile cilia genes as potentially related to the instigation of IS (Buchan, Gray, et al., 2014; Grimes et al., 2016; M. Hayes et al., 2014; Konjikusic et al., 2018). The primary cilium, a microtubule-based organelle on each cell type, is an extracellular projection that functions as a cellular communicator with its environment through the regulation of key signaling pathways such as Wnt and Hedgehog (Hh), and is known to play key roles in the development of multiple body systems, including the axial skeleton and central nervous system (Malicki & Johnson, 2017). Motile cilia are specialized organelles which possess a pair of central microtubules connected to dynein arms, enabling them to beat and move fluids including CSF and respiratory mucosa (Mitchinson, 2017). Ciliopathies, conditions that affect the structure or function of the cilia, are frequently biallelic recessive conditions that result from molecular mechanisms related to cilium dysfunction (Ferkol & Leigh, 2012). Ciliopathies have overlapping phenotypic features within multiple organ systems, such as hydrocephalus, polydactyly, kidney disease, and motor and cognitive impairments (Ferkol & Leigh, 2012), thus distinguishing this group of disorders from that of IS. Scoliosis can be a phenotypic feature within this disease group, although the molecular mechanisms underlying this feature have not been elucidated (Oliazadeh, Gorman, Eveleigh, Bourque, & Moreau, 2017). The potential that a more minor ciliary dysfunction may result in an axial growth disturbance during periods of rapid growth and hormonal change such as IS is intriguing and one that bears attention. In this work, we present evidence for rare heterozygous mutations in kinesin family member 7 (KIF7), which encodes a ciliary protein that participates in the Hh pathway, as a potential contributor to the etiology of IS based on human genetic and zebrafish model data.

Methods

Study Subjects

Individual study subjects were enrolled as previously described (Baschal et al., 2018; Baschal et al., 2014). A diagnosis of IS required a standing anteroposterior spinal radiograph showing ≥ 10° curvature by the Cobb method with pedicle rotation, and no congenital deformity or other co-existing musculoskeletal disorder.

Written informed consent was obtained from study subjects who were enrolled in accordance with protocols approved by the Johns Hopkins School of Medicine Institutional Review Board and the University of Colorado Anschutz Medical Campus Institutional Review Board (Colorado Multiple Institutional Review Board, Study #06–1161 and 07–0417). All procedures involving human participants were performed in accordance with the ethical standards of these institutional review boards, the 1964 Declaration of Helsinki and its later amendments, or comparable ethical standards.

We collected blood samples from all participants and extracted genomic DNA using standard phenol chloroform protocols or the QIAGEN Gentra Puregene Blood Kit. DNA quality was verified by agarose gel electrophoresis and the Qubit broad-range DNA kit.

One family was selected for exome sequencing (Baschal et al., 2018) based on the number of individuals affected with IS, the severity of the spinal curvature, and the number of individuals with available DNA samples. Subsequently, individuals III:5, IV:1 and IV:3 from the family were selected for exome sequencing based on genetic distance between family members and curve severity (see Figure 1a). Degree of spinal curvature as measured by Cobb angle is indicated in numbers below affected individuals (i.e. 20D); individuals with known double curves are indicated as such (i.e. 26/30D). Individuals confirmed as negative by physician examination are indicated as “confirmed negative,” and all others with listed degrees were determined by radiograph. Proband is indicated by an arrow. Individuals without a designation or GC or GG did not provide DNA so were not able to be genotyped. Individual III:6 was determined to be an unaffected obligated carrier due to having an affected child.

Figure 1.

a. Pedigree of KIF7 variant in IS family. Arrow indicates proband. Individuals without a GC or GG genotype were not enrolled or did not provide DNA. Individual III:6 possessed the GC genotype and was an obligate carrier for the variant. b. Sanger sequencing showing representative family members with the variant and reference genotypes at NM_198525.3:c.3345C>G.

Exome Sequencing

Exome capture was performed as previously described (Baschal et al., 2018). Exome capture was completed using 1 μg of genomic DNA from 16 individuals across five families using the Illumina TruSeq Exome kit. Samples were sequenced with a 2 × 100 bp run on the Illumina HiSeq 2000 at the University of Colorado Denver Genomics and Microarray Core Facility with three samples multiplexed per lane.

Bioinformatic Filtering

Exome reads were aligned to GRCh38 and variants were identified with FreeBayes as previously described (Baschal et al., 2018). The candidates were filtered by SnpEff (version 4.1g) (Cingolani et al., 2012) and custom scripts to retain only non-synonymous SNPs, coding indels, and variants affecting splice sites. These were also stripped of known artifacts and variants whose frequency was greater than 5% in the ExAC database (r0.3) (Lek et al., 2016). If the variant was annotated in the dbNSFP database (version 3.0) (Liu, Jian, & Boerwinkle, 2011; Liu, Wu, Li, & Boerwinkle, 2016), it was retained only if at least one of the prediction algorithms (SIFT, Polyphen2, LRT, MutationTaster) scored it as “damaging,” signifying that the resulting change to the protein had a predicted functional consequence. Variants that were not shared by all sequenced members of the family were not retained. Variants with a Minor Allele Frequency (MAF) < 0.05 that passed the filters above were retained for further analysis.

Second Exome Cohort

Whole exome sequencing data for 1221 severe adolescent idiopathic samples and 1397 controls were generated at the McDonnell Genome Institute (MGI) using IDT xGen Exome Panel V1 capture on Illumina HiSeq 4000 paired-end reads. [dbGAP accession number: phs001677]. Raw sequencing reads (FASTQ files) were aligned to the human genome reference (GRCh37) using BWA-MEM (v0.7.15). Aligned sequencing reads (BAM files) were sorted and PCR duplicate reads were marked using Picard MarkDuplicates (v2.9.0). Re-alignment of indel variants were performed with GATK RealignerTargetCreator and IndelRealigner using a known indel variant sites database (Mills and 1KG indels data from the GATK resource bundle: ftp://ftp.broadinstitute.org/bundle/b37/). Base quality score recalibration (BQSR) was then performed by GATK BaseRecalibrator and PrintReads to generate a final high quality BAM file. Variant calling of single nucleotide variants (SNVs) and insertion/deletion variants were generated first for each single sample and then combining all samples with joint genotyping method, described in GATK Best-Practices. [Genome Analysis Toolkit (GATK v3.5) https://gatk.broadinstitute.org/hc/en-us/sections/360007226651-Best-Practices-Workflows]. Variants calls were then recalibrated and filtered using GATK Variant Quality Score Recalibration (VQSR) method. A final VCF file with high quality variants was then annotated with Gencode v19 database.

Human Genotyping

DNA from additional affected and unaffected members of the family was Sanger sequenced in order to determine whether the variant segregated with the phenotype. PCR was conducted in 20 μL reactions with 10 μL Premix D (Epicentre Biotechnologies), 0.2 μL Taq Polymerase (Sigma), 20 ng genomic DNA, and 20 μM Forward and Reverse Primers. PCR reactions were run with a touchdown PCR protocol on a SimpliAmp Thermocycler (Fisher Scientific). Protocol and primer sequences are provided in Supp. File S1.

Annealing temperatures were set with the delta function to begin at 65°C and decrease by 0.3 °C each cycle, ending at 55 °C. Primers were obtained from Integrated DNA Technologies. PCR reactions were cleaned up for Sanger sequencing using ExoSAP-IT Express (Thermo Fisher). Sanger sequencing was performed by Quintara Biosciences and chromatograms were analyzed using the CodonCode Aligner v9.0 (CodonCode Corporation, https://www.codoncode.com/index.htm).

Zebrafish Husbandry

Zebrafish (Danio rerio) were housed at the University of Colorado Anschutz Medical Campus aquatic facility. Husbandry and experimental protocols were approved under the Institutional Animal Care and Use Committee Protocol #00370. Animals were maintained in a 14 hour light/10 hour dark cycle at 28.5 °C. Animals under anesthesia in 800 μg/mL Tricaine were fin clipped for genotyping and RNA extraction. Animals were euthanized in 4 mg/mL Tricaine, followed by immersion in ice water.

In order to assess survival, three crosses of wildtype (AB), kif7^co63/+ and kif7^co63/co63 zebrafish were prepared and monitored. Approximately 100 embryos per cross were grown in screening medium in an incubator at 28.5 °C, with no more than 20–30 embryos per petri dish. Surviving embryos were counted and deceased embryos were removed on days 1, 3 and 5 past fertilization. On day 5 embryos were transferred to tanks connected to the water system at the University of Colorado Anschutz Medical Campus aquatic facility. Live zebrafish were then counted every 5 days until 60 days past fertilization, when they were euthanized.

CRISPR-Cas9

gDNA constructs were created using default parameters on the CRISPOR website (Haeussler et al., 2016). Guides for kif7 were created using the GRCz10 zebrafish genome build. Guide sequences were targeted to zebrafish kif7 exon 3 (NM_001014816.1), and are provided in Supp. File S1.

The sgRNA was constructed by annealing sense and anti-sense single stranded oligonucleotides containing 5’ Bsa1 restriction overhangs and was inserted into Bsa1 linearized pDR274 with the Quick Ligase Kit (NEB). To make the sgRNA we linearized purified pDR274 containing the guide sequence with Dra1 and used a T7 RNA polymerase for in vitro transcription (NEB). The pMLM3613 plasmid encoding Cas9 was used for in vitro transcription using the SP6 mMessage mMachine Kit (Ambion) according to manufacturer’s instructions.

Zebrafish embryos were injected with an injection mix created from 18.2 ng sgRNA, 50 ng Cas9 protein, 285 mM potassium chloride, 1 μl phenol red, and sterile water. Embryos were analyzed for mutations the day after injection by lysing, PCR and fragment analysis (FA). FA was conducted at the University of Colorado Anschutz Barbara Davis Center. FA primers are provided in Supp. File S1:

We generated F4 homozygous kif7^co63/co63 first by outcrossing CRISPR-Cas9 injected F0 to AB wild-type fish (ZIRC) to generate F1 zebrafish. F1 mutations were characterized by Sanger sequencing (see below). F1 fish were crossed to AB wild-type to generate F2 kif7^co63/+. We then incrossed F2 kif7^co63/+ to generate F3 kif7^co63/co63. We then incrossed F3 kif7^co63/co63 to generate a pool of F4 kif7^co63/co63.

DNA for genotyping was obtained by fin clip. Lobes of the tail fin of anesthetized fish were removed with razor blades sterilized with 70% ethanol. Fins were digested in 50 μl lysis buffer, incubated at 95 °C for 20 minutes, iced for 2 minutes, followed by addition of 2.5 μl Proteinase K (10 mg/mL). Lysis buffer was created from 500 μl 1M Tris pH 8.3, 2.5 mL 1M KCl, 1.5 mL 10% Tween, 1.5 mL 10% Nonidet P40, 44 mL dI water. Reactions were then incubated at 55 °C for 4 hours, followed by 10 minutes at 98 °C. Reactions were spun to pellet cellular debris and the DNA supernatant was removed for use. When both RNA and DNA were desired from fin clips from the same individuals, nucleic acids were extracted using the Zymo DNA/RNA Miniprep kit. DNA quality was assessed by Nanodrop (Thermo Scientific).

Sanger sequencing was used to characterize mutations at the basepair level. Sanger sequencing was performed with zebrafish specific primers by Quintara Biosciences, as described above for human genotyping. PCR reactions were made using 12.5 μL 2X GOTaq Green Master Mix (Promega), 1 μL gDNA, 0.5 μL 10 μM F and R primers, and 10 μL sterile water. PCR protocol and primer sequences are provided in Supp. File S1. Routine genotyping of zebrafish lines was done by PCR as above, followed by gel electrophoresis using a 3% agarose gel with a 1:1 ratio of standard agarose to MetaPhor agarose (Lonza). Gel was run at 120V for 1 hour and 45 minutes followed by UV imaging.

Two zebrafish indel lines were created from mutations resulting in 1.) a 5bp deletion/ 14bp insertion at exon 3 (RefSeq gene NM_001014816.1), Chr7:14433928 (kif7^co63) and 2.) a 4bp deletion in exon 3 (RefSeq gene NM_001014816.1), Chr7:14433924 (kif7d4), resulting in a frameshift mutation.

Zebrafish Imaging

Micro-CT: Whole-animal μCt was conducted on a Scanco μCT-50 instrument. Zebrafish adults were fixed with 10% neutral buffered formalin at room temperature overnight and rinsed with 70% ethanol. Samples were held immobilized within 19 mm sample holders packed with Kimwipes. PBS was added to the sample holder to prevent samples from overdrying.

Brain μ-CT was conducted on an Xradia microXCT scanner at the University of Texas High-Resolution X-ray Computed Tomography Facility. Zebrafish adults were fixed in 2.5% paraformaldehyde overnight and stained for ~48 hours in 25% Lugol’s solution. Images were reconstructed using the Xradia Reconstrutor at 5.30 μM resolution. Additional scan parameters are available in Supp. File S1.

Tissue and cell staining: Adult zebrafish were euthanized and stained in alizarin red and alcian blue as described by Sakata-Haga et al. (Sakata-Haga et al., 2018). Stained zebrafish were visualized using a Leica S9i microscope. AFRU staining of the Reissner fiber was conducted as described by Troutwine et al. (Troutwine et al., 2019). Central canal cilia were stained with a polyglutamylated tubulin antibody (GT335) and gamma tubulin (GTX113286). In both cases the samples were processed for immunofluorescence (IF) after fixation in 4% sweet PFA overnight at 4 °C and processed using standard IF protocols, except samples prepared for cilia staining were dehydrated first in ice cold acetone (20 minutes at −20 °C) prior to incubation in primary antibody. Cilia were measured using ImageJ.

RT-qPCR

At 48 hours past fertilization (hpf), 15 embryos per tube were euthanized and pooled in 1mL of Qiazol (Qiagen). Embryos were lysed using a VWR Bead Mill Homogenizer with Lysing Matrix A beads (MP Biomedicals). Samples were iced for 5 minutes, lysed at speed 5, 120 V for 45 seconds, followed by ice for 5 minutes. This process was completed three times to ensure maximal lysis. RNA was then extracted using the RNeasy Lipid Tissue Mini kit (Qiagen) using the on-column DNase digestion protocol. Zebrafish fin clips were obtained as described above. DNA and RNA were extracted from fin clips using the Zymo Quick DNA/RNA Miniprep kit.

cDNA was created from RNA with the High Capacity cDNA Reverse Transcription kit with RNase inhibitor (Applied Biosystems) using random primers. 1 μg RNA was used per reaction. RT-qPCR primer sequences are provided in Supp. File S1.

SsoAdvanced SYBR Green Supermix (Bio-Rad) was used for reactions according to manufacturer’s instructions. Reactions were run on a Bio-Rad CFX 96 instrument at 95 °C for 10 minutes followed by 40 cycles of 95 °C for 15 seconds and 60 °C for 1 minute. Results were analyzed in accordance with the 2^−ΔΔCq method (Schmittgen & Livak, 2008).

Breeding assays

Zebrafish complementation assays were performed with male homozygous kif7d4 (4bp deletion) fish (see above) X female kif7^co63/co63 (14bp insertion/ 5bp deletion) fish. Embryo survival was assessed at 24 hpf, and body curvature was assessed at 5 dpf. Results are provided in Supp. File S1. Maternal contributions were assessed by crossing female kif7^co63/co63 fish X male kif7^co63/+ fish, and female kif7^co63/+ fish X kif7^co63/co63 fish. Body curvature was assessed at 5 dpf and again at 6 weeks old.

Results

We studied a multigenerational IS family of European descent (Baschal et al., 2018). Individuals were determined to have IS based on clinical history and radiographic findings of a spinal curvature ≥ 10°. The pedigree for this family is provided in Figure 1. This family had no clinical history of other musculoskeletal abnormalities or other known genetic conditions.

Genomic DNA was extracted from blood samples of affected and unaffected individuals. Three affected individuals were selected for exome sequencing based on genetic distance and severity of the spinal curvature. We obtained between 20.9 and 26.9 million reads per sample. Filtered variants were required to be shared by all members of the family, be predicted to be damaging by at least one predictive program, and have an MAF < 0.05 according to Exome Aggregation Consortium data (ExAC).

A total of 19 variants were identified that were shared between all three affected family members [Table 1, full detail in Supp. Table S1]. Variants in cilia, actin, and microtubule functional GO terms were prioritized for further investigation, in accordance with previous findings from our laboratory (Baschal et al., 2018). This prioritization resulted in investigation of a heterozygous missense variant in KIF7 (KIF7:NM_198525.3:exon17:c.C3345G:p.H1115Q, rs142032413), which encodes a ciliary kinesin that binds microtubule-plus ends and participates within the hedgehog (Hh) signaling pathway. This mutation has a minor allele frequency of 0.01684 and may be considered “uncommon” but not “rare”. KIF7 was the only gene of the 19 genes within this family that was associated to a ciliary GO term.

Table 1:

List of filtered variants (MAF < 0.05) for exome sequenced IS family. Variant filtering was performed as described in the Methods. All variants were shared by all sequenced members of the family, predicted to have a damaging consequence on the protein product by at least one predictive algorithm, and were present at an MAF < 0.05 according to ExAc. Genotype position in hg38, reference and alternate alleles and resulting amino acids are provided. Predicted functional consequence of the protein according to PolyPhen 2 is provided.

Gene name	chromosome	hg38_position	Reference	Alternate	Amino acid reference	Amino acid alternate	rs_dbSNP146	ExAC allele frequency	Polyphen2 HDIV predicted	IV:1 genotype	IV:3 genotype	III:5 genotype
GRHL3	chr1	24344928	A	G	N	S	rs116396279	0.03239	B	0/1	0/1	0/1
OR11L1	chr1	247841358	T	G	D	A	rs751804875	0.04886	D	0/1	0/1	0/1
ZNF717	chr3	75737791	C	A	R	I	rs2918517	.	P	0/1	0/1	0/1
DNAJC13	chr3	132492497	A	G	Y	C	rs201263331	2.55E-04	D	0/1	./.	0/1
ECE2	chr3	184276659	C	T	T	I	rs13063766	0.02194	B	./.	0/1	0/1
POM121	chr7	72925141	C	A	A	E	rs143666868	0.01247	D	0/1	./.	0/1
OR51T1	chr11	4882270	G	A	R	H	rs78644275	0.01822	P	0/1	0/1	0/1
ST5	chr11	8731051	G	C	P	R	rs61751342	0.03815	P	0/1	0/1	0/1
MYBPC3	chr11	47350047	C	T	V	M	rs3729986	0.04396	P	0/1	0/1	0/1
NDUFS8	chr11	68030914	G	A	R	H	rs77155493	0.03688	.	0/1	./.	0/1
SIX4	chr14	60723335	T	G	Y	S	rs766227737	1.65E-05	D;P	0/1	0/1	0/1
KIF7	chr15	89629547	G	C	H	Q	rs142032413	0.01684	D	0/1	0/1	0/1
ZNF213	chr16	3140689	G	A	G	D	.	.	B	./.	0/1	0/1
JPH3	chr16	87690092	T	G	S	A	.	.	B	0/1	0/1	0/1
PPM1E	chr17	58756114	C	CGAACCC						0/1	./.	0/1
ANKRD30B	chr18	14803746	A	G	T	A	rs45533337	0.003866	B	0/1	0/1	0/1
JOSD2	chr19	50511115	A	C				7.62E-04		0/1	0/1	0/1
TPTE	chr21	10569701	C	T	R	X	rs1810540	.	.	0/1	0/1	0/1
C21orf58	chr21	46314930	G	A	A	V	rs114869542	0.03269	B	0/1	./.	0/1

We then investigated whether this variant in KIF7, denoted GC, segregated with the scoliosis phenotype in additional members of the family. We sequenced the KIF7 variant in three additional affected and three additional unaffected members of the family by the Sanger method. This confirmed that the variant segregated with the IS phenotype, except for one unaffected obligate carrier who had a putative pathogenic GC genotype and was proven unaffected by radiographic examination (Figure 1).

We also investigated whether variants in KIF7 were present at higher levels in affected vs unaffected individuals in a second exome sequencing cohort of 1221 affected and 1397 unaffected individuals (Supp. Table S2). The H1115Q mutation present in our IS family was not present in this cohort in either cases or controls. Although no single variants in KIF7 passed statistical significance for association with IS (Fisher’s exact test), possibly due to sample size, and overall there was not an increased burden of missense variants in IS cases, we noted that all stopgain and frameshift variants were present in affected individuals only (Table 2). When compared to the Gnomad database, each of these severe mutations passed statistical significance. Additionally, the overall burden of predicted loss-of-function KIF7 variants in the IS cohort vs Gnomad controls was increased (0.4095% in IS vs 0.05997% in controls, p= 0.0012, Fisher’s exact test). One of these individuals had an affected parent who was also found to possess the stopgain variant (15:90174836_G>A). DNA was not available from family members of any of the other individuals possessing other stopgain or frameshift variants.

Table 2:

List of damaging KIF7 candidate variants identified in additional exome sequencing cohort of unrelated individuals.

Gene	KIF7	KIF7	KIF7	KIF7
Chr: Position, Ref>Alt (hg19)	15:90173601_G>A	15:90174836_G>A	15:90176938_GCTCATTTCTGC>G	15:90176938_GCTCATTTCTGC>G
HGVS description	NM_198525.3:c.3235C>T	NM_198525.3:c.3001C>T	NM_198525.3:c.2560_2570delGCAGAAATGAG	NM_198525.3:c.2560_2570delGCAGAAATGAG
Variant event length	0	0	−11	−11
Variant type	Stop Gain	Stop Gain	Deletion	Deletion
Allele frequency, IS cases (n= 1221)	0.00041	0.00041	0.00086	0.00086
Allele frequency, IS controls (n= 1397)	0	0	0	0
Allele frequency, Gnomad (n= 141456)	0	0.000011	0.000054	0.000054
Fisher’s Exact Test, IS vs Gnomad	p= 0.0086	p= 0.0255	p= 0.0031	p= 0.0031
Age (years)	19	19	15	Unknown
Gender	F	F	F	F
Spinal curvature (Cobb angle)	42°	65°	47°	51°
Affected family members?	Unknown	Affected father with variant	Unknown	Unknown

Zebrafish develop a degenerative scoliosis naturally with age (A. J. Hayes et al., 2013) and have previously been used as a model for IS in several studies (Boswell & Ciruna, 2017; Buchan, Gray, et al., 2014; M. Hayes et al., 2014; Van Gennip, Boswell, & Ciruna, 2018). To test the role of Kif7 in spine morphogenesis, we created a mutant model for kif7 in Danio rerio (zebrafish) using CRISPR-Cas9. We isolated a single founder (kif7^co63) which contained a complex indel mutation (5 bp deletion/ 14 bp insertion) in exon 3 of kif7 (RefSeq gene NM_001014816.1), resulting in an inframe insertion with a missense substitution (R203_H207insPGG +T208P). Additionally, we generated a second founder with a 4 bp deletion in exon 3 (kif7d4), which was used for complementation testing. Heterozygous kif7^co63/+ mutants displayed no obvious defects in spine morphogenesis, however, homozygous kif7^co63/co63 display adult-viable scoliosis without vertebral malformations, resembling IS in families (Figure 2). Homozygous kif7^co63/co63 showed an approximate 3-fold decrease of kif7 expression in embryos and adult tissue compared to wild-type adults, suggesting a mild hypomorphic defect in these mutants (Figure 3). We observed that only 37% of F4 kif7^co63/co63 zebrafish survived after 24 hpf (Supp. Table S3). Of those surviving at 24 hpf, 6–25% of kif7^co63/co63 larvae displayed a curved body phenotype, which was not observed in any wild-type fish (Supp. Table S3). All larvae with curved bodies were deceased by 1 week, and all surviving zebrafish appeared phenotypically normal until the juvenile stage, when curve progression was observed.

Figure 2.

kif7^co63 zebrafish display progressive scoliosis that starts at the juvenile period, with an absence of other gross vertebral or other bone abnormalities. A.) Representative anterior-posterior views of juvenile and adult female zebrafish, as shown by micro-CT. Curves on juvenile zebrafish are denoted by white arrows. Pictured adult kif7^co63/co63 zebrafish displays a curve of approximately 68 degrees. B.) Cartilage and bone phenotypes of zebrafish at 33 dpf, as shown by alizarin red/alcian blue staining. White arrows denote separations between vertebrae.

Figure 3.

Characterization of mutation in kif7^co63 in Danio rerio. A.) Genomic DNA sequences of kif7 at exon 3, showing a 5bp deletion/14bp insertion. B.) Kaplan-Meier curve showing survival of wild-type (AB), kif7^co63/+ and kif7^co63/co63 embryos. C.) kif7^co63/co63 adults and embryos show reduced expression of kif7 mRNA.

Analysis of cartilage and bone structures by alcian blue and alizarin red staining revealed kif7^co63/co63 adult fish had an absence of gross abnormalities in spinal tissues other than the primary curvature, although mild vertebral wedging was observed (Figure 2). Crosses between kif7^co6/co63 and kif7d4 zebrafish revealed no differences in the rate of scoliosis in trans-heterozygous fish, suggesting that the phenotype is not due to off-target effects (Supp. File S1).

Lateral spinal curvatures developed in ~20% of kif7^co63/co63 individuals at approximately 6 weeks past fertilization and progressed until ~4 months of age (2–2.5 cm average body length). Micro-CT analysis of bony structures revealed Cobb angles ranging from 50–68° in adults at 4 months of age (Figure 2). No differences in males and females were observed with regard to the severity or in the rate of progression of spinal curvature.

Specific cilia defects have been linked to IS in both humans and zebrafish models (Buchan, Gray, et al., 2014; Grimes et al., 2016; M. Hayes et al., 2014). Zebrafish models of IS caused by cilia mutations have shown some phenotypic characteristics shared with human ciliopathies, including hydrocephalus and ventricle dilation (M. Hayes et al., 2014; Sternberg et al., 2018; Zhang et al., 2018) so we investigated whether these brain abnormalities were also present in kif7^co63/co63 zebrafish. Contrast micro-CT imaging of the brain of kif7^co63/co63 zebrafish did not show hydrocephalus or dilatation of the brain ventricles (Figure 4, Supp. Movie S1). Analysis of central canal cilia, which are in contact with CSF, also appeared morphologically indistinguishable from wildtype zebrafish in both kif7^co63/+ and kif7^co63/co63 (Figure 5A–B). Cilia in kif7^co63/co63 embryos also appeared morphologically normal by staining (Figure 5E); however, we observed that cilia in mutant embryos were longer than in wild-type (kif7^co63/co63= 2.72 μM ± 0.069; wild-type= 2.032 μM ± 0.0257) (Figure 5F).

Figure 4.

kif7^co63/co63 zebrafish show a lack of abnormalities characteristic of ciliopathies. Contrast micro-CT of the brain in wildtype and kif7^co63/co63 zebrafish. Brain does not show hydrocephaly or dilation of the ventricles (arrows).

Figure 5.

Homozygous and heterozygous kif7^co63 zebrafish do not show defects in the formation of central canal structures. Confocal imaging of immunofluorescence of kif7^{co63 /+} (A-A’’, C) and kif7^{co63 /co63} (B-B’’, D) at 5 dpf demonstrating no alterations in cilia (A’, B’) in the central canal (marked by DAPI/nuclei staining (A, B) in kif7^{co63 /co63} mutants (merged in A’’ and B’’) or the Reissner fiber (C, D). Scale bar = 10 microns.

An intact Reissner fiber has recently been shown to be critical for the development of a straight body axis in vertebrate models (Cantaut-Belarif, Sternberg, Thouvenin, Wyart, & Bardet, 2018), and its disassembly is associated with the onset of scoliosis in several independent models of scoliosis in zebrafish (Troutwine et al., 2019). To test whether kif7^co63/co63 embryos and juveniles retained an intact Reissner fiber and maintained the Reissner fiber throughout development of scoliosis, we used the AFRU-antiserum, which has been used previously to stain the Reissner fiber in zebrafish (Troutwine et al., 2019). Embryonic and juvenile kif7^co63/co63 zebrafish showed intact Reissner fibers that were indistinguishable from WT (Figure 5C–D, Supp. File S1). These results suggest that kif7^co63 mutant zebrafish develop scoliosis due to defects independent of motile cilia function and cerebrospinal fluid flow physiology.

As Kif7 plays a vital role as a conserved regulator of Hh signaling, we investigated whether kif7^co63/co63 embryos displayed altered gene expression within this pathway. No differences in gene expression of Gli1, Gli2, Gli3, Ptch1, Stk36, or Sufu was observed between WT and kif7^co63/co63 embryos (Figure 6). However, genes encoding two proteins known to bind directly to Kif7, Smo and Dlg5a (Chong, Mann, Zhao, Kato, & Beachy, 2015), were altered in kif7^co63/co63 embryos. Smo (Smothened) was reduced approximately 3-fold in kif7^co63/co63 embryos, mirroring the reduced levels of kif7. Conversely, dlg5a was increased approximately two-fold in kif7^co63/co63 embryos over wild-type. These data show that decreased Kif7-dependent Hh signaling may play a role in spine morphogenesis.

Figure 6.

Gene expression for key hedgehog (Hh) pathway genes within kif7^{co63 /co63} embryos. Smo and Dlg5 appear down- and up-regulated, respectively, while the remaining genes show no differences between WT levels. * indicated p <0.05, ** indicates p < 0.01.

Discussion

In this work, we combine human exome data from a multiplex family (Baschal et al., 2018) and an unrelated IS cohort to suggest that mutations in KIF7 may contribute to the IS phenotype. We then further support these findings with a unique zebrafish model, kif7^co63/co63, which develops juvenile-onset spinal curvatures without obvious vertebral malformations or hydrocephalus, thus mirroring the IS phenotype. Our results collectively implicate hypomorphic mutations in KIF7 as contributory to IS pathogenesis in a subset of individuals in IS.

KIF7 encodes a broadly conserved kinesin ciliary protein that localizes to the axonemal tip of primary cilia and binds to the plus-ends of microtubules (He et al., 2014). The protein functions as a scaffold protein and acts as both a negative and positive regulator of the hedgehog (Hh) signaling pathway, an evolutionary conserved molecular pathway central to embryonic development, limb patterning and musculoskeletal maintenance (He et al., 2014), including formation and maintenance of the intervertebral disc (Rajesh & Dahia, 2018). Perturbations in both KIF7 and SUFU, both with overlapping coordinated functions in the Hh pathway, have been shown to recreate a spectrum of limb defects in mouse (Zhulyn & Hui, 2015). Interestingly, SUFU has also been associated with IS via linkage (Marosy et al., 2010). Loss of function mutations in KIF7 have been linked to both Joubert and the rare Acrocallosal syndromes (OMIM #611254), two ciliopathies with many overlapping, system-wide defects including developmental disability, skeletal abnormalities and kidney disease. Scoliosis has been observed in 5–33% of Joubert syndrome cases (Bachmann-Gagescu et al., 2015; Brancati, Dallapiccola, & Valente, 2010; Elhassanien & Alghaiaty, 2013), potentially related to early hypotonia, whereas the scoliosis prevalence in Acrocallosal syndrome has not yet been described.

Our findings are consistent with previous studies suggesting that cilia mutations may contribute to IS etiology, which has been observed within both human genetic studies (Baschal et al., 2018) and within vertebrate models (Buchan, Gray, et al., 2014; Grimes et al., 2016; Hassan et al., 2019; M. Hayes et al., 2014; Konjikusic et al., 2018). Damaging variants in POC5 were found in multiple IS families (Patten et al., 2015; Xu et al., 2018), caused spine deformity in zebrafish (Patten et al., 2015), and led to cilia length disruptions and cell cycle defects in vitro (Hassan et al., 2019). Recent findings from our laboratory found that IS families may bear an increased burden of uncommon variants in ciliary genes (Baschal et al., 2018). At least 55 of the 303 cilia genes in the SYSCILIA database have been associated with a clinical syndrome displaying scoliosis (Oliazadeh et al., 2017), although it is important to note that IS individuals are otherwise normal individuals without the phenotypic features associated with ciliopathies. In zebrafish, the kinesin kif6 was found to be necessary for proper spine morphogenesis (Buchan, Gray, et al., 2014; Patten et al., 2015), and mutations in motile cilia genes including ptk7 were shown to be sufficient to induce spinal curvatures (Grimes et al., 2016; M. Hayes et al., 2014). Defective flow of the cerebrospinal fluid (CSF) was then determined to underlie scoliosis in motile cilia models of IS (Grimes et al., 2016; Zhang et al., 2018), and both CSF-contacting neurons (Sternberg et al., 2018; Van Gennip et al., 2018) and the Reissner fiber, a threadlike structure secreted into the CSF (Rodriguez, Rodriguez, & Hein, 1998), were found to be necessary for proper vertebrate spine morphogenesis (Troutwine et al., 2019).

Following the discovery of KIF7 mutations in IS patients, we used CRISPR-Cas9 technology to create a mutant kif7 line in Danio rerio. kif7^co63/co63 zebrafish develop isolated spinal curvatures as juveniles with no evidence of abnormalities in brain morphology or hydrocephaly, or morphological changes to the central canal motile cilia other than length. CSF-contacting cells in the spinal canals of both homozygous and heterozygous kif7^co63 zebrafish appear to be normally ciliated (Figure 5) suggesting an absence of defects in CSF-flow. Together with our other findings, this suggests that the development of abnormal spine morphogenesis in kif7^co63/co63 may be caused by factors distinct from abnormal CSF flow. To date, aberrant CSF-flow has been linked to the development of scoliosis in several zebrafish models (Grimes et al., 2016). In ptk7 mutant zebrafish, scoliosis was found to be caused by neuroinflammatory signals downstream of defects in the CSF flow (Van Gennip et al., 2018). It is unknown whether individuals with IS experience any CSF flow disturbances. A single pilot study of 11 individuals with idiopathic scoliosis found possible mild CSF flow disruptions through the vertebral column in some individuals (Jean-François Catanzariti, 2017). We suggest the mutations in kif7 may represent a pathway parallel or downstream of motile cilia and the CSF control of spine morphogenesis in zebrafish. This phenotype appears to mirror IS in humans, who have not yet shown evidence of ciliopathy-associated defects.

The Reissner fiber (RF) is a dynamic structure of secreted glycoproteins extending from the subcommissural organ in the brain through the CSF to the end of the spinal cord (Rodriguez et al., 1998). It is essential to the early stages of axial development in most vertebrates (Cantaut-Belarif et al., 2018), although it is unknown what its specific functions are in spinal neurodevelopment, or whether it is present in humans past fetal stages (Galarza, 2002). Most recently, a study has shown the RF to be defective in multiple zebrafish models of IS (Troutwine et al., 2019). Our work has shown that the kif7^co63/co63 zebrafish appear to have normal development of the Reissner fiber (Figure 5), suggesting that kif7^co63/co63 zebrafish develop scoliosis through an alternate mechanism either downstream or independent of RF formation.

As stated, KIF7 encodes a highly conserved motor protein that is localized at the distal end of axonemal microtubules and is a regulator of the Hh signaling pathway. Through its interactions, KIF7 controls axonemal length (He et al., 2014). While it is unknown how KIF7 recognizes microtubule ends, its unique position allows interaction with essential elements of the Hh signaling pathway critical for cellular functions including the transmembrance protein Smoothened (SMO), Discs large homolog 5 (DLG5), and the Gli family of transcriptional regulators (Chong et al., 2015). SMO is a broadly conserved transmembrane receptor upstream of KIF7 that facilitates Hh signaling following binding of the Hh ligand to the PTCH1 receptor. Dlg5 is required for Hh-induced enrichment of the KIF7-GLI2 complex at the ciliary tip, and disruption of SMO-DLG5 interactions lead to a diminished Hh pathway response (Chong et al., 2015). Using our kif7^co63/co63 mutants, we observed reduced expression of kif7 and smo and increased expression of dlg5. This is concordant with a downregulation of Kif7-dependent Hh signaling, presenting a possible mechanism by which abnormal spine morphogenesis may occur. Unexpectedly, expression of Hh target genes ptch1 and gli1 was not affected in kif7^co63/co63 embryos. Therefore, it is unclear whether smo and dlg5 dysregulation ultimately leads to changes within the Hh pathway.

We also observed that kif7^co63/co63 embryos possessed somewhat longer motile cilia within the central canal than those in wild-type controls, a finding consistent with cellular culture studies of KIF7 function in primary cilia (He et al., 2014). KIF7 function has been previously studied within primary cilia, rather than motile cilia, although our results suggest a possible role for KIF7 within motile cilia as well. Cilia length is tightly controlled by multiple cytoskeletal factors, particularly within the actin cytoskeleton, and abnormally long cilia have been observed in multiple developmental diseases (Miyoshi, Kasahara, Miyazaki, & Asanuma, 2011). The functional implications of cilium length differences are still being unraveled, although increased ciliary length has been associated with increased mechanosensation sensitivity, decreased efficiency of intraflagellar transport (IFT), and a reduced amount of force required to bend the cilium and trigger downstream signaling (Dummer, Poelma, DeRuiter, Goumans, & Hierck, 2016). Interestingly, longer primary cilia have also been observed on osteoblasts isolated from IS individuals (Oliazadeh et al., 2017). However, we do not yet know if cilia length is directly affecting cilia function in kif7^co63/co63 zebrafish, or, importantly, if abnormal length of either primary or motile cilia may ultimately contribute to IS etiology.

The apparent absence of secondary phenotypic features in kif7^co63/co63 zebrafish is novel for the field thus far and presents a unique resource for the study of abnormal spine development. It is important to note that IS is indeed “idiopathic” and does not appear to be associated with any other specific phenotypic features, including those associated with ciliopathies. The inverse is partially true, in that patients with certain ciliopathies present a modestly increased incidence of scoliosis over the general population (Oliazadeh et al., 2017).

It is important to note that the specific KIF7 mutation found within our family was also present in an unaffected individual. This may indicate incomplete penetrance or that this mutation is a disease modifier, rather than a sole causal agent. Furthermore, KIF7 mutations overall were not present at a high frequency in IS individuals, although severe mutations were present at higher frequencies in IS individuals than controls. It is likely that KIF7 mutations represent one of many possible IS risk alleles.

In summary, we present human and animal model data suggesting that KIF7 mutations contribute to the pathogenesis of IS in a subset of individuals. Our animal model indicates that defects in Kif7-dependent Hh signaling may be driving pathogenesis. We recommend further study to determine the mechanisms underlying abnormal spine morphogenesis in this novel animal model. This model may help to uncover subtle molecular changes present in individuals with IS. Delineating specific mechanisms that influence neuroaxial development, architecture and stability will lead us to a more specific understanding of the spectrum of spine deformities such as IS and enhance therapeutic interventions.

Web Resources

The following online resources were used for the completion of this research: CRISPOR (http://crispor.tefor.net/), UCSC Genome Browser (https://genome.ucsc.edu/), dbGap (http://www.ncbi.nlm.nih.gov/gap), DAVID Bioinformatics Resource v6.3 (https://david.ncifcrf.gov/), PANTHER Classification System (http://www.pantherdb.org/), GnomAD browser v2.1.1 (https://gnomad.broadinstitute.org/), SysCilia Gold Standard (http://www.syscilia.org/goldstandard.shtml), NCBI ClinVar (https://www.ncbi.nlm.nih.gov/clinvar/), OMIM (https://omim.org/), NCBI Primer-BLAST (https://www.ncbi.nlm.nih.gov/tools/primer-blast/index.cgi), IDT Oligo Analyzer Tool (https://www.idtdna.com).

Supplementary Material

SUP T1

List of filtered variants (MAF > 0.05) for exome sequenced IS family, additional detail.

SUP T2

List of all KIF7 variants detected in validation cohort of unrelated affected and unaffected individuals. Positions listed are in genome assembly hg19.

SUP T3

Summary of survival and rates of scoliosis in kif7^co63 and WT zebrafish.

SUP

Additional methods and results. Quantification of complementation assay, suggesting that off-target effects are not causing the scoliosis phenotype, maternal effect results, and additional staining results of the Reissner fiber.

V1

Contrast micro-CT movie of kif7co6/co633 and age-matched WT zebrafish brains.

V2

Micro-CT movie showing 360° view of kif7^co63/co63 skeleton.