Two CCD patients in a Chinese family were recruited according to the criteria for the clinical diagnosis of CCD (Mundlos, 1999), together with 50 control subjects. Informed consent was obtained from the guardians of the proband and other participants. All study protocols were approved by the Ethical Committee of Peking University Health Science Center (Approval Number: IRB00001052-07100).

Peripheral blood samples were collected from the participants and genomic DNA was extracted using a TIANamp Blood DNA mini kit (TIANGEN, Beijing, China) following the manufacturer’s instructions. RUNX2 mutation was analyzed in the proband and her parents. The exons and exon-intron boundaries of RUNX2 were amplified by polymerase chain reaction (PCR) using intron–exon specific primers as described previously (Quack et al., 1999). We also used other previously reported primers (Napierala et al., 2005) and designed others using software (Primer Premier 5 software, Premier Biosoft, Palo Alto, CA, United States) to amplify the 2.0 kb region upstream of the transcription start site of RUNX2 transcript variant 1 (NM_001024630.4). To confirm IGSF10 mutation detected in WES as pathogenic, we sequenced this variant in the proband, her parents and 50 unrelated control people. IGSF10-specific primers were also designed with the same software (Primer Premier 5 software, Premier Biosoft). Primer sequences are listed in Supplementary Table S1. Genomic DNA was amplified with Premix Taq™ (Takara, Shiga, and Japan). In brief, PCR reactions were performed with a DNA Engine PTC-200 (Bio-Rad Laboratories, Hercules, CA, United States) using a program described elsewhere (Zhang et al., 2017). The amplification products were bi-directionally sequenced using an ABI 3730 XL automatic sequencer (Applied Biosystems, Foster City, CA, United States). DNA sequences were analyzed using NCBI databases and the BLASTN (BLAST nucleotide) program (http://blast.ncbi.nlm.nih.gov/).

RUNX2 copy number was determined by real-time PCR using a SYBR Green PCR kit (Roche Applied Science, Indianapolis, IN, United States). We detected the copy number of the coding region and the flanking untranslated regions (UTRs) of RUNX2 using the primers and method described previously (Ott et al., 2010). The primers were listed in Supplementary Table S2. qPCR was performed in a total volume of 12 µl containing 10 ng genomic DNA, 6 µl FastStart Universal SYBR Green PCR Master (Roche Applied Science), and 1 µl primers (0.2 µmol each). Samples were run on the ABI 7,500 Real-Time PCR System (Applied Biosystems, Foster, CA, United States) in triplicate in separate reactions to determine the variation in copy number between CCD patients and controls. Two controls were analyzed. The copy number of RUNX2 was normalized against those of the Albumin (ALB) gene and were calculated using the 2^−ΔΔCt method.

Real-time PCR was also used to evaluate the expression of osteogenic-associated genes. MC3T3-E1 cells are osteoblast precursor cells derived from mouse, we used this cell line to explore the role of IGSF10 in osteogenesis. After MC3T3-E1 cells were cultured in osteogenic medium (OM) for 7 days, total RNA was extracted using Trizol reagent (Invitrogen, Carlsbad, CA, United States) following the manufacturer’s instructions. Next, 500 ng of total RNA were used in a 10 µl reaction volume to synthesize cDNA using a PrimeScript RT reagent kit (Takara). The resultant cDNAs were amplified using specific sets of primers for Igsf10, Runx2, bone sialoprotein (Bsp), alkaline phosphatase (Alp), osteocalcin (Ocn), and osterix (Osx). The housekeeping gene glyceraldehyde-3-phosphate dehydrogenase (Gapdh) was used to normalize RNA expression levels. The primers used are listed in Supplementary Table S3. Primer specificity was confirmed with amplicon dissociation curves. The RT-qPCR was performed using 1 µl cDNA product in a 20 µl reaction volume with a SYBR Green PCR kit (Roche Applied Science) according to the manufacturer’s protocol, and the results were assessed with 7,500 Software 2.0.1 (Applied Biosystems). Relative mRNA expression was calculated using the 2^−ΔΔCt method with the formula: F = 2 − Δ Δ Ct = 2 [ Ct ( target gene ) − Ct ( GAPDH ) ] control 2 [ Ct ( target gene ) − Ct ( GAPDH ) ] treatment .

Multiplex ligation-dependent probe amplification (MLPA) was performed with the SALSA MLPA Probemix P080-C2 kit according to the manufacturer’s instructions (MRC-Holland, Amsterdam, Netherlands) by the Department of Medical Genetics, Peking University Health Science Center (Beijing, China). All the peak areas were normalized, and a ratio between 0.8 and 1.2 was considered normal. A heterozygous duplication was identified with a ratio between 1.3 and 1.7, and a heterozygous deletion between 0.3 and 0.7.

Karyotype analysis was performed by Giemsa-Trypsin (GTG) banding using standard procedures by the Department of Medical Genetics, Peking University Health Science Center (Beijing, China). Twenty G-banded metaphase cells were examined.

Whole-exome sequencing (WES) was performed using genomic DNA from the proband and her parents. DNA libraries were prepared using the Agilent SureSelect system (Agilent Technologies, Santa Clara, CA, United States) and sequenced on the Illumina HiSeq 2000 platform (Illumina, San Diego, CA, United States) to generate 100 bp paired-end reads. Bioinformatics analyses included production of clean data, alignment to human reference genome hg19 using Burrows–Wheeler Aligner (BWA) (Li and Durbin, 2009), variant calling, and annotation using Samtools, Genome Analysis Tool Kit, and Annovar (Hintzsche et al., 2016). We then used in-house AnnoDB software to annotate the confident variant results. Quality control was present throughout the whole procedure. Next, steps were performed to identify potential causal variants. First, the variants were segregated by trait (variants present in affected individuals but not in unaffected family members). Second, variants were further filtered by minor allele frequency (MAF <0.5%) in the context of heterozygous genotype in the 1,000 Genomes database (1000G_ALL and 1000G_EAS) and the NHLBI exome variant server (ESP_ALL). Third, deleterious variants were picked based on SIFT, Polyphen-2, and variant impact. Then, the functions of those genes retained after filtering were reviewed, and genes related to bone development, especially those in the RUNX2 pathway, were selected for further study.

Multiple sequence alignment for IGSF10 was performed using ClustalW in MEGA-X software to analyze the conservation of affected amino acids (Kumar et al., 2018). To analyze the effect of the mutation on the molecular structure of IGSF10, the three-dimensional (3D) structures of wild-type and mutant IGSF10 were predicted in silico using I-TASSER (https://zhanglab.ccmb.med.umich.edu/I-TASSER/) (Yang et al., 2015).

MC3T3-E1, a pre-osteoblast murine cell line, was purchased from the Cell Bank of Chinese Academy of Sciences and cultured in a proliferation medium containing α-MEM (Gibco, Grand Island, NY, United States), 10% fetal bovine serum (Gibco), 100 U/ml penicillin (Gibco) and 100 μg/ml streptomycin (Gibco) in an incubator at 37°C with 5% CO₂. For osteogenic induction, MC3T3-E1 cells were cultured in proliferation medium plus 100 nM dexamethasone (Sigma-Aldrich, St Louis, MO, United States), 10 mM β-glycerophosphate (Sigma-Aldrich), and 50 μg/ml l-ascorbic acid (Sigma-Aldrich).

The pCMV5-Runx2 plasmid expressing full-length wild-type RUNX2 was provided by the School of Dentistry, University of Michigan (Ann Arbor, MI, United States). Plasmid pcDNA3-CMF608FL (Segev et al., 2004), containing a full-length open reading frame for expression of Flag-tagged rat IGSF10 protein, was provided by Quark Pharmaceuticals Inc. (Newark, CA, United States/Ness Ziona, Israel). The sequences of wild-type Runx2 and Igsf10 were confirmed by DNA sequencing. Empty pCMV5 and pcDNA3.1 vectors served as transfection controls. Three small interfering RNAs (siRNAs) for Igsf10 and Runx2 were purchased from Ribo Bio Technology (Guangzhou, China). The most efficient siRNA for each gene was selected for the subsequent experiments: si-Igsf10, 5′-GCA​CCT​TCC​TGA​TTT​CAA​A-3′; and si-Runx2, 5′-GCA​CGC​TAT​TAA​ATC​CAA​A-3′. Scrambled siRNA was used as a negative control.

For transfection, MC3T3-E1 cells were seeded at a density of 1×10⁵ cells/well in six-well plates. When cells reached 80% confluency for plasmids and 30–50% confluency for siRNAs, transfection was performed with Lipofectamine 3,000 (Invitrogen, Shanghai, China) in accordance with the manufacturer’s instructions. Cells were collected for subsequent experiments 48 h after transfection.

For IGSF10 knockdown, we used a lentiviral RNAi small hairpin RNA (shRNA)-encoding system using the pHBLV-U6-Puro lentiviral RNAi vector (Hanbio, Shanghai, China). The targeting sequence of the Igsf10 shRNA was 5′-GCA​CCT​TCC​TGA​TTT​CAA​A-3′. To establish stable IGSF10-knockdown cell lines, MC3T3-E1 cells were transduced with the lentiviral RNAi vector at a multiplicity of infection (MOI) of approximately 100 in the presence of 10 μg/ml polybrene (Hanbio), and infected cells were selected with 2 μg/ml puromycin (Sigma-Aldrich). The empty lentivector was used as a negative control.

MC3T3-E1 cells were cultured in OM 12-well plates for 7 days. Then, the cells were fixed with 4% paraformaldehyde for 15 min. After washing in double-distilled water (ddH₂O) three times, the cells were stained using an ALP histochemical staining kit (Cwbiotech, Beijing, China) in accordance with the manufacturer’s protocol.

For ALP activity testing, MC3T3-E1 cells were collected after osteogenic induction for 7 days. The cells were lysed in RIPA buffer containing protease and phosphatase inhibitors (Huaxing bio, Beijing, China). Lysates were centrifuged at 13,523 × g at 4°C for 30 min and the supernatants were collected for analysis. Protein concentration was determined using a Pierce BCA protein assay kit (Thermo Fisher Scientific, Waltham, MA, United States). ALP activity was measured with an ALP activity kit (Nanjing Jiancheng Biotech, Nanjing, China) in accordance with the manufacturer’s instructions.

MC3T3-E1cells were fixed in 4% paraformaldehyde after osteogenic induction. Alizarin red staining was performed using 2% alizarin red (Sigma-Aldrich). For quantitative assessment of the degree of mineralization, the stain was dissolved in 10% cetylpyridinium chloride (Sigma-Aldrich) and the absorbance at 562 nm was measured.

After MC3T3-E1 cells were cultured in OM for 7 days, total protein was extracted and quantified with the method described in the ALP activity assay. Samples with equal amounts of denatured protein were analyzed by sodium dodecyl sulphate polyacrylamide gel electrophoresis. Proteins were transferred to polyvinylidene difluoride membranes (Millipore, Bedford, MA, United States) and blocked in 5% skimmed milk. Then, membranes were incubated with primary antibodies against RUNX2 (CST, Danvers, MA, United States), BSP (CST), OSX (Abcam, Cambridge, United Kingdom), and GAPDH (Huaxing bio) at 4°C overnight. After washing in TBS-T, the membranes were incubated with HRP-conjugated secondary antibody (Huaxing bio) for 1 h at room temperature and visualized using an enhanced chemiluminescence blotting kit (Cwbiotech).

All experiments were repeated at least three times and data are expressed as the mean ± SD. Differences between groups were determined by statistical analysis with GraphPad Prism v8.02 (GraphPad Software, Bethesda, MD, United States). Normal distribution and variance homogeneity tests were carried out. For the expression of IGSF10 in the osteogenesis of MC3T3-E1 cells, one-way analysis of variance (ANOVA) was performed and followed by Bonferroni’s post hoc test. Comparison between two groups was performed using unpaired Student’s t-test. p < 0.05 was considered statistically significant.

A Chinese family was studied. The proband, a 12-year-old girl, showed typical features of CCD, including patent fontanelles and bilateral hypoplastic clavicles on skull and chest radiography (Figures 1A–C). Furthermore, multiple dental anomalies commonly seen in CCD were also found, such as retained deciduous teeth, delayed eruption of permanent teeth, and crossbite (Figures 1D–G). A narrowing upper dental arch and high vault were also present (Figure 1F). Notably, supernumerary teeth were absent.

The pedigree of the family showed that the proband’s mother was also a CCD patient (Figure 1H). The phenotype of the proband’s mother was similar to the proband (Figure 1I–L), except that supernumerary teeth were present in the proband’s mother (Figure 1L). However, the number of supernumerary teeth was uncertain because the tooth extraction history of the mother was not reliable. No obvious signs of any other health problems were present in the two patients. After a detailed inquiry of the family medical history, we found that the proband’s two uncles and grandparent had no obvious CCD phenotypes. We were unable to confirm the genotype because they refused to carry out genetic testing.

Sequencing analysis was first performed for the coding region, exon–intron boundaries, and the 2.0 kb region upstream of the transcription start site of RUNX2 by PCR and direct sequencing to explore the genetic basis of the familial case. However, no abnormality was found (data not shown). Because copy number variation of RUNX2 can cause CCD (Lee et al., 2008), we screened for intragenic deletions and duplications of RUNX2 by real-time PCR and MLPA. However, we detected no causative copy number variant of RUNX2 (Figures 2B,C). Chromosomal abnormalities involving RUNX2 can also result in CCD (Purandare et al., 2008), and therefore karyotype analysis was performed for the proband. A normal karyotype was shown for the proband (Figure 2A); therefore, we speculated that genetic variations in genes other than RUNX2 may lead to the disease.

To identify the genetic cause of the familial case, we performed WES and bioinformatics analyses for the proband and her parents. WES identified 166,648 variants (Figure 3A) and a total of 67 candidate genes were identified after filtering for a MAF <0.5%, predicting deleterious variants, and segregating with the trait in affected family members (Supplementary Table S4). These 67 candidate genes were assessed by a literature review. Unexpectedly, no variants were detected in genes related to RUNX2 signaling. Therefore, we predicted that variation in an unknown gene may contribute to CCD in this family. We found that, of the 67 candidate genes, only IGSF10 was related to osteogenesis based on current knowledge. IGSF10 was finally selected for further investigation because of its potential role in bone development (Segev et al., 2004).

A heterozygous two-base deletion in IGSF10 (c.6001_6002delCT, p.Leu2001Valfs*24) was detected in the proband and her mother (Supplementary Table S4). 50 unaffected controls were analyzed by PCR and Sanger sequencing to verify the mutation as pathogenic. Consistent with the WES data, a heterozygous two-base deletion of IGSF10 was detected only in the proband and her affected mother, but not in normal controls or her unaffected father (Figure 3B). This novel mutation was in exon 8 of IGSF10 (NM_178,,822.5) and resulted in a frameshift from codon 2001 to the resultant premature stop codon 2024. This was predicted to encode an IGSF10 protein truncated at the sixth immunoglobulin-like domain (Ig domain) with the loss of 600 amino acids from the C-terminus (Figure 3C).

Partial amino acid sequence alignment of the IGSF10 protein from 12 species showed that the affected amino acid, L2001, has a high level of evolutionary conservation among species (Figure 3D). Furthermore, the detected mutation was predicted by I-TASSER to alter the three-dimensional structure of the last ten Ig domains of IGSF10 (Figure 3E), leading to a loss of function of the predicted protein.

The role of IGSF10 in osteogenic differentiation was further explored. We evaluated the expression dynamics of Igsf10 mRNA after osteogenic induction in MC3T3-E1 cells and found that the levels of Igsf10 mRNA decreased upon OM treatment (Figure 4A). We then constructed an MC3T3-E1 cell line with stable IGSF10 knockdown. IGSF10 knockdown efficiency was first assessed by real-time PCR. The mRNA level of Igsf10 was decreased by more than 60% in MC3T3-E1 cells compared with control cells (Figure 4B). Regrettably, the level of IGSF10 protein could not be determined because a reliable IGSF10 antibody was not available.

We incubated stable IGSF10-knockdown cells and control cells (named si-Igsf10 and si-Control, respectively) in OM medium and analyzed their osteogenic ability. ALP activity (Figures 4C,D) and the ability to form mineralized nodules (Figures 4C,E) were obviously impaired in IGSF10-knockdown cells compared with control cells.

RUNX2 is a key protein in osteogenesis; therefore, we asked whether the role of IGSF10 in osteogenesis is mediated by RUNX2. Unexpectedly, the level of Runx2 expression did not change, irrespective of overexpression (Figure 4F) or knockdown of IGSF10 (Figure 4G). Similarly, Igsf10 expression was not affected by RUNX2 overexpression (Figure 4H) or knockdown (Figure 4I).

The effect of IGSF10 knockdown on other osteogenic-associated genes was subsequently explored. The level of Bsp mRNA decreased by 30% after IGSF10 knockdown (Figure 4M). The levels of RUNX2 and BSP proteins were similar to their mRNA levels after knockdown of IGSF10, with the level of BSP protein showing a slight reduction (Figures 4J,K). The mRNA levels of Alp, Ocn, and Osx showed no significant change after IGSF10 knockdown in MC3T3-E1 cells (Figures 4N–P).