Introduction

Genetic bone dysplasia is a clinically and genetically heterogeneous group with over 400 conditions under the new nosology in 2010.1 After that, more genes involved in bone dysplasia had been identified.2, 3 Correct clinical diagnoses of a few of them which are having distinct characters are relatively easy, such as achondroplasia, osteogenesis imperfecta, Marfan syndrome, hypophosphatemic ricket and mucopolysaccharidosis, while most of the others are difficult to group for the non-expert owing to high similarities of phenotypes among themselves. Thus, collaboration between local medical staff with expert diagnostic network is necessary to reach a diagnosis for difficult cases.4

As for the molecular confirmation of genetic bone dysplasia, it is even more complex. On one side, mutations of one gene can cause multiple phenotypes, for example, COL2A1 mutations could cause a spectrum of phenotypes, ranging from premature arthritis to perinatally lethal disorder.5 On the other side, one similar clinical phenotype can result from mutations of different genes, for example, in the case of multiple epiphyseal dysplasia, autosomal dominance of them may be owing to mutations of COMP, MATN3, COL9A1, COL9A2 or COL9A3, and autosomal recessive pattern is because of SLC26A2 gene mutations.6

As a group, bone dysplasia is very common with an estimated occurrence 1:1000. Most of them manifest as short stature and limb abnormalities, which are two common reasons for referral to pediatric endocrinology.7, 8 Pediatricians treat larger number of patients with bone dysplasia than 20 years ago.9 Treatment with growth hormone for achondroplasia is approved in Japan. Bisphosphonate has been in off-label use for children with osteogenesis imperfecta in the past ten years. Statin treatment can rescue patient-specific induced pluripotent stem cell models and a mouse model of FGFR3 skeletal dysplasia, thus statin may be a new drug to treat FGFR3-related bone dysplasia.10 Accurate molecular diagnosis of the bone dysplasia is increasingly important for possible treatment options and genetic counseling.

Owing to the high number of genes and the large size of each gene associated with most of these disorders, standard diagnostic testing using Sanger sequencing are expensive and time consuming. However, this shortage of Sanger sequencing could be overcome by targeted next-generation sequencing (NGS) technologies. With genomic capturing known genes, whose mutation cause bone dysplasia leading to short stature or dysmorphology, it is possible to screen those genes in the suspected patients in high throughput by NGS.

In China, there have been no population-based studies and most bone dysplasias have been reported as case reports with mutations identified in a few of them (1.16%).11 In this study, we performed a comprehensive mutation screening of genes in a large cohort of children suspected with bone dysplasia using the targeted NGS. Our results provided a preliminary overview of the genetic etiology of bone dysplasia in Chinese children seen in the pediatric endocrinology department.

Subjects and methods

Study subjects

This study has been approved by Ethics Committee of Shanghai Xinhua Hospital (XHEC-D-2014-006). A group of unrelated patients (n=89) characterized by short stature, bone dysmorphology and X-ray abnormalities in our outpatient clinics were included. Among them, 55 cases were male, the other 34 cases were female, and the average age was 5.3 years old (range 8 months to 17 years). This cohort of individuals included one patient with achondroplasia with FGFR3 G380R mutation as positive control and six patients with confirmed diseases as negative control, for example, two patients with mucolipidosis type II/III; one with mucopolysacharidosis type I, IIIA and IVA each; and one with tyrosinemia type I who had secondary hypophosphatemia. For the rest 82 patients, mucopolysaccharidoses, GM1 gangliosidosis, and mucolipidosis type II/III have been pre-excluded by peripheral leukocytes or plasma lysosomal enzyme activity measurements. Typical patients with achondroplasia derived from FGFR3 G380R mutation, a common reason of dwarfism in humans, had been pre-excluded by Sanger sequencing with primers specific to this site. This cohort of patients included 13 patients with hypophosphatemic rickets and three patients with osteogenesis imperfecta. The rest of them had no clear and confident clinical diagnosis.

Targeted exon capturing and NGS

Custom probes with amplicon size 425 bp targeted 61 bone dysplasia-related genes according to the 2010 nosology and classification of genetic skeletal disorders (Supplementary Table 1) were designed with Illumina’s online DesignStudio software.1 As a department of pediatric endocrinology, we pay more attention to genes related with short stature. Although FBN1 gene belonged to group 30 overgrowth syndromes in the 2010 Nosology,1 it also has been demonstrated to be the causative gene of acromelic dysplasia.12 So FBN1 gene was included, along with FBN2 gene to investigate whether it is associated with other bone dysplasia besides congenital contractural arachnodactyly. As other site mutations of FGFR3 gene besides G380R can also lead to bone dysplasia, this gene was still included in our targeted gene list. A total of 1292 exons and 182 675 bases were targeted. With setting of the probe parameters, the overall coverage of gene exons is 87.5%.

Genomic DNA was extracted from peripheral blood according to the manufacturer’s protocol with the RelaxGene blood DNA isolation kit (Tiangen Biotech, Beijing, China). DNA with a concentration of 300 ng μl−1 was hybridized with probes. A library was established following the standard Illumina protocols, which was then sequenced on a MiSeq sequencing system (Illumina, San Diego, CA, USA). Data were analyzed by MiSeq Reporter. Reads were aligned to the NCBI37/hg19 assembly using the BWA Multi-Vision software package with single-nucleotide polymorphisms and indels identified using the SOAPsnp software and the GATK Indel Genotyper, respectively.

PCR and Sanger sequencing

Potential novel mutations identified by NGS were further verified by polymerase chain reaction with site-specific primers and Sanger sequencing. Parents of probands were also checked whether they carried the same mutations.

Pathogenicity evaluation of novel variants

The novel variants were interpreted according to the 2015 American College of Medical Genetics and Genomics standards and guidelines.13 Glycine substitutions in the triple-helical domain G-X-Y of the COL2A1 gene were presumed to be a moderate evidence of pathogenicity. Matching of genotypes with its phenotypes and co-segregation with disease status in family gene analysis are two required supporting evidence of pathogenicity. Null mutations including insertion, deletion, splicing and stop codon changes were regarded as a very strong evidence of pathogenicity. MutationTaster (http://www.mutationtaster.org/) and PolyPhen-2 (http://genetics.bwh.harvard.edu/pph2/index.shtml) for amino acid substitution were applied to evaluate the variation pathogenicities. The predicted result ‘disease-causing’ by MutationTaster and ‘probably/possibly damaging’ by PolyPhen-2, indicated that these variations had two lines of computational evidence supporting a pathogenic effect. Missense variants are common mechanism of X-linked recessive Chondrodysplasia punctata.14 Thus, a missense variant of the ARSE gene is a supporting evidence of pathogenicity. Furthermore, the presumed pathogenic or likely pathogenic variants should not be a previously reported single-nucleotide polymorphism or polymorphisms.

Results

The average sequencing depth was 96.8 and the sequencing depth over the region of interest was 84.7. As expected, no pathogenic variations had been identified in the targeted 61 genes in the six patients with confirmed other bone diseases, which were regarded as negative controls in this pilot study. In the achondroplasia patient, who was treated as positive control, G380R mutation in FGFR3 gene was identified, consistent with his previous diagnosis by Sanger sequencing.

Totally 44 of the 82 patients (54%) had confirmed mutations, including nine patients with type II collagenopathy (Table 1), nine with PHEX-mutated X-linked hypophosphataemic rickets (Table 2), eight with COMP gene mutations (Table 3), four with TRPV4 gene mutation, four with FBN1-related acromelic dysplasia15 and three with COL1A1-mutated osteogenesis imperfecta. In addition, mutations of COL11A1, TRAPPC2, MATN3, ARSE, TRPS1, SMARCAL1 and ENPP1 gene were identified in one patient each (Table 4). In all, mutations in 13 different genes were found from our targeted list involved in bone dysplasia. It has been the first time that Chinese cases of group 8, TRPV4 group, and group 21, chondrodysplasia punctata, are being reported.11

Table 1 Patient characteristics with confirmed COL2A1 gene mutations in our study
Table 2 Patient characteristics with confirmed mutations in PHEX gene
Table 3 Patient characteristics with confirmed COMP gene mutations in our study
Table 4 Genes with a small number of patients having mutations

There were only two patients with an autosomal recessive mode, which were hypophosphataemic rickets caused by biallelic mutations in ENPP1 gene and Schimke immuno-osseous dysplasia caused by a homozygous mutation in SMARCAL1 gene (Table 4). Most cases were sporadic. In only seven cases (7/44=16%), the mutation were familial, of which two cases with COL2A1 mutations were derived from their respective father, one case with COL11A1 mutation from the proband’s mother, one case with an ARSE mutation derived from the proband’s mother and three cases with PHEX mutations from their mother, respectively. Except the mother with heterozygous ARSE mutation being asymptomatic, the other parents with dominant inheritance mode were symptomatic and showed very similar clinical phenotypes with the proband.

In view of the novelty of mutations, we could see that nearly half of them (23/44=52.3%) were reported previously and the other half (21/44=47.7%) were novel. The novel variants were evaluated as pathogenic or highly pathogenic according to the joint consensus recommendations for the interpretation of sequence variants of American College of Medical Genetics and Genomics.13 For example, p.G711R (No 3, Table 1) was a de novo variant based on family-based gene analysis, which was a strong evidence of pathogenicity. Furthermore, it was located with the well-established functional domain glycine of G-X-Y, and was absent in 1000 Genomes Project, which were two moderate evidences of pathogenicity. In addition, two lines of computational analysis (MutationTaster and PolyPhen-2) that supported p.G711R had a deleterious effect and the patient’s phenotype was highly specific for spondyloepimetaphyseal dysplasia, Strudwick type (provided in the Discussion), which were two supporting evidences of pathogenicity. Totally, this variant has one strong, two moderate and two supporting evidences of pathogenicity, which support it pathogenically, according to the rules for combining criteria to classify sequence variants.13 The evaluation of other novel variants was provided as Supplementary Table 2 except for three FBN1 variants, which were reported previously.15

Discussion

In this pilot study, type II collagenopathy is a common disorder in our cohort including seven cases with spondyloepiphyseal dysplasia congenita (SEDC) and two cases with spondyloepimetaphyseal dysplasia, Strudwick type. COL2A1 mutations may give rise to a spectrum of phenotypes, spanning from the mild late-onset premature arthritis to the severe intrauterine achondrogenesis.5 The prevalent COL2A1-related SEDC of this study may be explained by our unique subject collection in pediatric endocrinology, where a large number of patients seek help for short stature. Otherwise, it may be explained that COL2A1 gene mutations usually cause SEDC, as two-thirds were classified to SEDC in a large multinational study where 93 patients with type II collagenopathy were reported.16 There were five novel mutations and four known mutations with seven of them involved with glycine substitution in the triple-helical domain, which was consistent with previous result that glycine substitutions are the most common types of mutation.16, 17

Thirteen patients with hypophosphatemic rickets were recruited in this study. Mutations in PHEX gene were identified in nine of them, and mutations on both the alleles in ENPP1 gene were identified in one patient. No mutations were identified in FGF23, DMMP1, SLC34A3 or CLCN5, which indicated that PHEX gene mutation was also the predominant cause of hypophosphatemic rickets in Chinese patients. A girl without family history and another girl whose mother has similar bone dysplasia, indicating an apparent dominant trait, revealed no mutations of six known genes related to hypophosphatemia, for example, PHEX, FGF23, DMP1, ENPP1, CLCN5 and SLC34A3. As a multiplex ligation-dependent probe amplification analysis was an indispensable supplementary to detect larger deletions/duplications in PHEX or FGF23 gene,18 these two girls should be further analyzed by multiplex ligation-dependent probe amplification method. In one male child, the hypophosphatemia was secondary to kidney involvement of tyrosinemia type I. In another female child, the hypophosphatemia was secondary to renal tubule pathology of kidney.

For the COMP-related bone dysplasia, seven out of eight subjects were assigned to pseudoachondroplasia with one assigned to multiple epiphyseal dysplasia (MED). At the molecular level, there are three known mutations in four patients and four novel mutations with one occurrence each, for example, T529A, D509V, D475H and D439G. At all these four codons, different nucleotide transversions or transitions causing different amino acid substitutions had been reported with consistent clinical diagnosis.19, 20

For unknown reasons, only a few of Chinese MED patients were reported.11 Mutation p.R121Q in the MATN3 gene was identified in one subject assigned to MED. At the same codon of the MATN3 gene, another mutation R121W has been reported to be causative to MED.21 Thus, our cohort of patients contained only two patients with confirmed MED at the molecular level, caused by COMP and MATN3 gene, respectively. The rarity of MED in this cohort of study could be partially explained given that most people with MED have normal or mildly short stature. It was also possible that there is a low occurrence of MED in Chinese ethnics.

Furthermore, there were four cases with TRPV4 gene mutation, which constituted 9% (4/44) of individuals with confirmed gene mutation, indicating TRPV4 gene is a relatively common cause of bone dysplasia in Chinese. In accordance with previous findings,22 two of these four patients carried the hot mutation R594H of TRPV4 gene. On the basis of the clinical and radiological characters, three of them were diagnosed with spondylometaphyseal dysplasia Kozlowski type, and one with metatropic dysplasia.

Only one patient with type 11 collagen group has been identified. As most mutations were reported on COL11A1 gene,23 this novel mutation c.1245+1G>A also altered the splicing consensus sequences.

As stated, most confirmed cases in this cohort had been diagnosed only after the availability of genetic analysis by targeted NGS, such as ARSE-associated X-linked recessive chondrodysplasia punctata, TRPV4-related spondylometaphyseal dysplasia, type II collagenopathy and FBN1-associated acromelic dysplasia. We previously have reported the clinical characteristics of patients with FBN1-associated acromelic dysplasia.15 Here we provided the clinical and radiographic characteristics of some patients with other confirmed genetic diagnosis.

The boy with ARSE-associated X-linked recessive chondrodysplasia punctata (No. 11 in Table 4) presented to our clinics for disability to raise head and macrocephaly at 5 months of age. He was the second baby of an unconsanguineous young couple. His old sister was healthy. As a fetus, he was found to have a comparatively larger head by ultrasound. He was spontaneously delivered at term with birth weight 3.7 kg (1 s.d.) and birth length 52 cm (1 s.d.) with head circumference 39 cm (4 s.d.). The physical examination on presenting to our clinic included a squared skull with the head circumference 46.5 cm (4 s.d.) and the anterior fontanel 4x4 cm, and length 66 cm (−1 s.d.) and weight 7 kg (−1.5 s.d.). Brain MRI revealed a small posterior fossa. Interestingly, retrospective review of the images of spine, pelvic, humeri and femurs taken at the age of 5 months did not show characteristic stippled epiphyses (Figure 1). However, imaging of distal phalanges was not properly taken, where chondrodysplasia punctata were frequently noted.24 When re-examined at the age of 27 months, a depressed nasal bridge, a small nose and chin and ichthyosis on the dorsal skin of forearm and hand were observed. The anterior fontanel was still open with size 2.5 × 2 cm. He could not walk independently yet, neither could he utter meaningful words.

Figure 1
figure 1

Radiographs of a boy with X-linked chondrodysplasia punctata. The bone epiphyses (a, lateral view of spine; b, upper limbs and chest; c, pelvis and lower limbs) taken at the age of 5 months did not show signs of stipples.

One patient with R594H mutation in TRPV4 gene was a girl aged 4 years 5 months (No. 3 in Table 4), who had growth retardation in the past 2 years. As the lateral view of the spine was misinterpreted as bullet-shaped and ribs were likely oar-shaped (Figures 2a and b), she was referred to us for mucopolysaccharidosis-related lysosomal enzymatic assay. On physical examination, she was 99 cm (−2 s.d.) in height and 15.5 kg in weight (−1 s.d.). Neither did she have a coarse face, nor a tight or relaxed joint. Her teeth were also normal. However, she had a mild protruding sternum and a mild lumbar lordosis. All the tested lysosomal enzyme activities were normal. After identification of the R594H mutation of TRPV4 gene, her X-rays were reviewed retrospectively, and platyspondyly (Figure 2a), delayed carpal bone ossification (Figure 2c) and short femoral neck (Figure 2d) were identified, which was consistent with the characteristics of spondylometaphyseal dysplasia, Kozlowski type. Without the result of NGS, we could not correctly diagnose this patient.

Figure 2
figure 2

Radiographs of a girl with TRPV4-related spondylometaphyseal dysplasia. (a) Lateral view of the spine showed platyspondyly; (b) chest radiography showed mild oar-shaped ribs; (c) delayed carpal bone ossification in the left hand; (d) broad and short femoral neck indicated metaphyseal dysplasia. All the pictures were taken at the age of 4 years 5 months.

The girl (No. 3 in Table 1) with G711R mutation in COL2A1 gene went to our clinic at the age of 5 years for short stature and wadding gait. On physical examination, she was 91.3 cm (−5 s.d.) in height and 14 kg in weight (−2 s.d.) and had a mild pectus carinatum. She had visited an orthopedic surgeon for shortness of left leg compared with her right leg at the age of 2 years when a pelvic radiogram indicating ossification, absence of epiphyseal center of left proximal femurs (Figure 3a). The paddle-shaped ribs (Figure 3b) and abnormal lateral view of spine (Figure 3c) had prompted the patients to take the lysosomal enzyme tests, which were unremarkable. He was then followed at the age of 8 years (Figures 3d–f). On the whole, the radiograph of this patient was in line with spondyloepimetaphyseal dysplasia, Strudwick.

Figure 3
figure 3

Radiographs of a girl with COL2A1-related spondyloepimetaphyseal dysplasia, Strudwick type. Hip at the age of 2 years showed ossification absence of epiphyseal center of left proximal femurs (a); imaging taken at 5 years old showed oar-shaped ribs (b) and the marked thoracic vertebral dysplasia in the lateral view of spine (c); imaging taken at 8 years old showed small and fragmented epiphyseal centers of both proximal femurs (d), and irregular metaphyses of left femur, tibia, fibula (e) and radius and ulna (f).

In this study, nearly half the patients were without a clear diagnosis, which may be further investigated by whole-exome sequencing. With more patients getting precise molecular diagnosis, we can have a more accurate disease spectrum on genetic bone dysplasia.

In conclusion, we made a pilot study to investigate the genetic epidemiology of bone dysplasia in Chinese children. In the context of pediatric endocrinology, type II collagenopathies, PHEX-related X-linked hypophosphataemic rickets and COMP-related pseudoachondroplasia are the most common diseases in patients with bone dysplasia when excluding the lysosomal storage disorders and achondroplasia. Targeted next-generation sequencing may be adopted clinically to investigate the defective genes of difficult cases of genetic bone dysplasia, which are very critical for the patient’s prognosis prediction and genetic counseling of the family.