Genotype–phenotype relationship and comparison between eastern and western patients with osteogenesis imperfecta

Patients were enrolled at the Endocrinology Department of Peking Union Medical College Hospital (PUMCH) from January 2007 to May 2022. Eligible patients were clinically suspected of having OI on the basis of a history of fracture under minor force, low BMD, with or without blue sclera, dentinogenesis imperfecta or a familial history of OI or fracture.

This study was approved by the Scientific Ethics Committee of PUMCH (JS-3545D). Informed consent was obtained from the patients or their legal guardians before participation in the study.

Clinical data for the patients were collected from medical records, including birth situation, age at diagnosis, age of initial fracture, fracture frequency and site, bone deformities (limb bending, thoracic deformity, and pelvic deformity), and extraskeletal manifestations. Body height and weight were measured using a Harpenden stadiometer (Seritex Inc., East Rutherford, NJ, USA). For patients unable to stand, body length was measured in the supine position. The height of all OI patients was converted to standard deviation score (SDS) using standardized growth charts for Chinese children and adolescents [12]. OI was classified into five subtypes according to the clinical severity: mild OI (type I), perinatally lethal OI (type II), progressive deforming OI (type III), intermediate OI (type IV) and OI with hypertrophic callus (type V) [5].

BMDs at the lumbar spine (LS) and proximal hip were measured by dual-energy X-ray absorptiometry (DXA, Lunar Prodigy Advance, GE Healthcare, USA), and appropriate pediatric software was used for measurement of BMD in children. A quality control program was conducted throughout the study, and phantom testing was completed daily using the DXA device for calibration and quality checks. Obviously compressed or deformed vertebrae were excluded from BMD analysis. LS and FN BMD results of children and adolescents were converted to age- and sex-specific Z scores according to normal reference of BMD in Asian children [13, 14].

Serum levels of β-isomerized carboxy-telopeptide of type I collagen (β-CTX, a bone resorption marker), procollagen I N-terminal peptide (P1NP, a bone formation marker), 25-hydroxyvitamin D (25OHD) and intact parathyroid hormone (PTH) were measured using an automated electrochemiluminescence system (E170, Roche Diagnostics, Switzerland). Serum levels of alanine aminotransferase (ALT), creatinine (Cr), calcium (Ca), phosphate (P) and alkaline phosphatase (ALP, a bone formation marker) were assessed using automated analyzers (ADVIA 1800, Siemens, Germany). All parameters were detected in the clinical laboratory of PUMCH.

Clinical fractures were reported by the patients or their legal guardians and confirmed by X-ray films, including nonvertebral fractures and symptomatic vertebral fractures. Vertebral compression fracture (VCF) was semiquantitatively assessed as normal (grade 0), mildly deformed (grade 1), moderately deformed (grade 2), and severely deformed (grade 3) according to Genant's classification [15]. Scoliosis was determined by anteroposterior radiography and defined as a Cobb angle > 10 degree [16]. X-ray film results were interpreted by radiologists at PUMCH.

Total genomic DNA was isolated from peripheral blood using a DNA Extraction Mini Kit (QIAamp DNA, Qiagen, Frankfurt, Germany). Clinically diagnosed OI patients underwent panel sequencing (Illumina HiSeq2000 platform, Illumina Inc., San Diego, CA, USA) using a previously described protocol [17]. The next-generation sequencing (NGS) panel covers more than 700 candidate genes of disorders related to bone, including 20 known candidate genes for OI (COL1A1, COL1A2, IFITM5, SERPINF1, CRTAP, P3H1, PPIB, SERPINH1, FKBP10, PLOD2, BMP1, SP7, TMEM38B, WNT1, CREB3L1, SPARC, MBTPS2, P4HB, SEC24D and PLS3). The overall sequencing coverage of the target regions was more than 95% at a minimum sequencing depth of 20 ×  Bioinformatics processing and data analysis were performed. The “clean reads” derived from targeted sequencing and filtering were aligned to the human genome reference (hg19) using the BWA (Burrows Wheeler Aligner) Multi-Vision software package. All SNVs and indels were filtered and estimated with multiple databases (NCBI dbSNP, HapMap, 1000 human genome dataset and database of 100 Chinese healthy adults). The deleterious effects of missense variants on the corresponding proteins were predicted by silico tools (MutationTaster, PolyPhen-2, SIFT and PhyloP). Variants were classified according to the 2015 American College of Medical Genetics and Genomics/Association for Molecular Pathology (ACMG/AMP) guidelines [18].

Pathogenic variants identified by NGS were confirmed by Sanger sequencing. Targeted primers were designed, PCR was performed, and the amplicons generated using the primers designed were sequenced with a 3730 DNA analyzer (Applied Biosystems, Foster City, CA, USA). Segregation analysis was performed if DNA was available from the family members. COL1A1 and COL1A2 variants were categorized based on the effects of gene mutation on type I collagen synthesis as glycine substitution, nonglycine substitution, haploinsufficiency or splicing mutation [19]. Compound heterozygous or homozygous mutation patterns were classified as biallelic variants [20].

We searched for previously published large cohorts of OI from Western and Eastern countries in PubMed, Embase, and Medline databases using “osteogenesis imperfecta”, “cohort”, “children with bone fracture”, and “children with osteoporosis”. We collected data on age at OI diagnosis, family history, height, fracture frequency, BMDs, classification of OI and gene mutations in these cohorts of OI, and differences between Eastern and Western OI cohorts were compared.

Continuous variables were tested for normal distribution using the Kolmogorov‒Smirnov test. Normally distributed data (BMD Z-scores) are presented as the mean ± standard deviation (SD). Differences in BMD Z-scores between groups were compared with one-way ANOVA. Nonnormally distributed data are expressed as medians (range), including age, age at first fracture, height Z-score, frequency of peripheral fracture, and serum PTH and 25OHD levels, and the Mann‒Whitney U test was used to compare these parameters between groups. Categorical data are presented as frequencies and percentages (%), Fisher's exact test was utilized to compare these categorical variables (positive family history of fracture, peripheral fracture, VCF, long bone deformity, scoliosis, etc.) between groups. The association between the position of glycine substitution and the frequency of peripheral fracture, LS and FN BMD Z-score, or height Z-score was evaluated using Spearman rank correlation coefficient analysis.

A total of 671 patients with a clinical diagnosis of OI were included in this study. Of these, 560 patients from 427 unrelated families were identified as carrying OI pathogenic mutations, and the rate of disease-causing gene mutation positivity was 83.5%. The pathogenic mutations of 427 probands identified in this cohort are listed in Supplementary Table 1. Mutations in 15 OI candidate genes were identified, with COL1A1 (n = 308, 55%) and COL1A2 (n = 164, 29%) being the most common (Fig. 1a). Among these mutations, 275 (49%) are missense, 69 (12%) nonsense, 118 (21%) frameshift, 101 (18%) splicing and 1 (0.2%) a chromosome translocation. Among the COL1A1/COL1A2 variants (n = 472), 190 (40.3%) are glycine substitutions, 42 (8.9%) nonglycine substitutions, 149 (31.6%) haploinsufficiency, 90 (19.1%) splicing mutations and 1 (0.2%) a chromosome translocation (Fig. 1b). Additionally, c.769G > A (p.G257R) of COL1A1 and c.1009G > A (p.G337S) of COL1A2, which were the most common variants of COL1A1 and COL1A2, were identified in 10 and 6 unrelated families, respectively, suggesting that these two mutations are hotspot mutations in the Chinese OI cohort. Among biallelic variants, mutations in SERPINF1 (21.2%) and WNT1 (19.2%) were most frequent (Fig. 1a).

Detailed clinical data were available for 414 OI probands, including 328 children and 86 adults (Table 1). A total of 197 patients (48.3%) had a positive family history of OI. The median age of initial visit and age at first fracture among the OI patients were 10.0 years (range: 0.2–64.0 years) and 1.5 years (range: 0.0–59.0 years), respectively. The most common clinical manifestation was peripheral fracture (96.6%), and total frequency varied greatly, ranging from 0 to 70 times, with a median time of 4.0. (Table 1). A total of 2887 peripheral fractures were noted in this cohort, with femurs being the most commonly affected sites (34.8%). (Fig. 2a). VCF was also common in this study, as noted in 180 (43.5%) OI patients. A total of 815 VCFs were recorded from C7 to L5, with the most common sites at L1 (10.8%) and T12 (9.8%) (Fig. 2b). Overall, 251 OI patients (60.6%) developed bone deformities, including long bone and ribcage deformities and scoliosis (Table 1). Based on clinical phenotype, 202 (48.8%) patients were classified as OI type I, 70 (16.9%) as OI type III, 121 (29.2%) as OI type IV, and 21 (5.1%) as OI type V (Table 1). Additionally, among 111 patients with no detectable OI mutations, 81 (73.0%) were classified as OI type I, 9 (8.1%) as OI type III and 21 (18.9%) as OI type IV.

Additionally, according to serum 25OHD level, vitamin D sufficiency (25OHD level > 30 ng/mL), insufficiency (21–29 ng/mL) and deficiency (< 20 ng/mL) were found in 22.9% (80/349), 29.8% (104/349) and 47.3% (165/349) of OI patients, respectively. A total of 5.2% (18/349) of OI patients exhibited secondary hyperparathyroidism caused by insufficiency or deficiency of vitamin D. Furthermore, the levels of 25OHD were negatively correlated with age of OI patients (Supplementary table 2 and Supplementary Fig. 1).

Phenotypic differences in OI patients with COL1A1, COL1A2, and biallelic mutations were compared. Patients carrying COL1A1 mutation showed the mildest skeletal phenotypes, whereas those carrying COL1A2 or biallelic mutation had more severe phenotypes, including shorter height Z-score (− 0.8 vs. − 2.2, P < 0.001; − 0.8 vs. − 1.6, P = 0.004), more long bone deformity (41.9% vs. 60.4%, P = 0.001; 41.9% vs. 59.2%, P = 0.027) and ribcage deformity (7.0% vs. 17.1%, P = 0.004; 7.0% vs. 26.5%, P < 0.001), poorer mobility (31.3% vs. 45.9%, P = 0.008; 31.3% vs. 57.1%, P = 0.001) and higher OI type III frequency (13.2% vs. 24.3%, P = 0.010; 13.2% vs. 26.5%, P = 0.020) (Table 2, Fig. 3a).

Furthermore, phenotypic differences in OI patients with glycine substitution of COL1A1 or COL1A2, haploinsufficiency and biallelic variants were compared. Among these 4 subgroups, patients with haploinsufficiency of collagen type I α chains exhibited the mildest skeletal phenotypes. In contrast, individuals with glycine substitution of COL1A1 or COL1A2 or biallelic variants showed more severe phenotypes, characterized by lower height Z-score, more long bone and ribcage deformities, poorer mobility and lower BMD (all P < 0.05) (Table 3). Patients with glycine substitution of COL1A1 or COL1A2 or biallelic variants had similar bone involvement (Fig. 3b).

We screened 5 relatively large OI cohorts from Western countries, including Canada (n = 598) [21], Turkey (n = 150) [20], Italy (n = 364) [22], America (n = 544) [23] and Sweden (n = 223) [24] (Table 4). The annual fracture rate of OI patients seemed to be similar among Turkish, American, Swedish and Chinese OI cohorts (Table 4), indicating that the phenotypic severity of Eastern and Western OI patients might be roughly similar.

Moreover, the methods of genetic testing differed among countries. OI cohorts from Italy and Sweden were identified solely by Sanger sequencing or Sanger sequencing combined multiplex ligation-dependent probe amplification (MLPA) covering COL1A1/COL1A2 genes. OI cohorts from Canada, Turkey and China were identified by NGS panels or WES, which covered currently known candidate genes of OI. The detection rate of disease-causing gene mutation was similar between Turkish (83.6%), Italian (84.9%), Swedish (87.0%), and Chinese (83.5%) cohorts, though it seemed to be higher in Canadian (97.8%) and American (92.7%) cohorts (Table 4). Although the distribution of gene mutations differed among countries, COL1A1 and COL1A2 were still the dominant mutation genes in all cohorts. Furthermore, we found that SERPINF1 and WNT1 were the most common biallelic pathogenic genes in our cohort (21.2%; 19.2%), which was distinct from studies of Canadian, Turkish and American OI patients. In Canadian and American cohorts, SERPINF1 and CRTAP were the most frequently affected biallelic pathogenic genes; in Turkish cohorts, FKBP10 and P3H1 were the most frequently affected genes. Interestingly, we found a notable increase in the identification of biallelic variants from 2015 to 2022, which is attributed to the widespread use of NGS technology.