https://orcid.org/0000-0002-7864-0337
Abstract
Osteogenesis imperfecta (OI) is a rare, heterogeneous, genetic disorder characterized by bone fragility and recurrent fractures. Bisphosphonates (BPs) are the most commonly used medications for OI, but their efficacy has great variability.
We investigated the relationship of pathogenic gene mutations and responses to zoledronic acid (ZOL) in a large cohort of children with OI.
Children with OI who received ZOL treatment were included and were followed up for at least 1 year. Bone mineral density (BMD) and serum levels of β-isomerized carboxy-telopeptide of type I collagen (β-CTX, bone resorption marker) were measured at baseline and during follow-up. Causative mutations of OI were identified using next-generation sequencing and Sanger sequencing.
201 children with OI were included. They had initiated ZOL treatment at a median age of 5 years, with mutations identified in 11 genes. After 3 years of treatment, the increase in femoral neck BMD Z-score in patients with OI with autosomal dominant (AD) inheritance was greater than that in patients with autosomal recessive or X-linked inheritance (non-AD) (4.5 ± 2.9 vs 2.0 ± 1.0, P < .001). Collagen structural defects were negatively correlated with the increase in femoral neck BMD Z-score. Patients with collagen structural defects had higher incidence of new fractures (35.1% vs 18.4%, relative risk 0.52, P = .044) and less decline in β-CTX level than those with collagen quantitative reduction. Increase in lumbar spine BMD and change in height Z-score was not associated with the genotype of children with OI.
Patients with OI with non-AD inheritance or with pathogenic mutations leading to collagen structural defects may have relatively poor responses to ZOL treatment, which is possibly associated with their more severe phenotypes. New therapeutic agents are worth developing in these patients.
Osteogenesis imperfecta (OI) is a rare, monogenic, systemic, connective tissue disease characterized by bone fragility and recurrent bone fractures, limb deformities, hearing impairment, and blue sclera (1). OI is most often caused by mutations in genes encoding type I collagen, and the estimated incidence is 1 to 2 per 20 000 neonates (1). In the absence of etiological treatment, the management of OI is primarily supportive and symptomatic. At present, the mainstay treatments for OI are bisphosphonates (BPs), which can suppress bone resorption, increase bone mineral density (BMD), and reduce bone fracture risk (2-6). However, responses to BPs treatment are variable in patients with OI, and there are even reports that patients with OI type VI caused by SERPINF1 mutations had poor treatment response to BPs (7-9). To date, the mechanism of the variability in responses of patients with OI to BPs treatment is still unclear.
OI is a complex disease with great phenotypic and genotypic variability. The Sillence classification divided OI into 4 forms according to clinical severity: the mildest (OI type I), perinatal lethal (OI type II), the most severe in survivors (OI type III), and intermediate between type I and type III (OI type IV) (10). Recently, another subtype of OI has been classified as type V, which is characterized by ossification of the forearm interosseous membrane, radial head dislocation, and hyperplastic callus (11, 12). OI can be inherited as an autosomal dominant (AD), autosomal recessive (AR), or X-linked disorder. With the progress in next-generation sequencing (NGS), more than 20 pathogenic genes have been identified in patients with OI. Most often, OI is caused by pathogenic variants in either COL1A1 or COL1A2 (encoding α1 or α2 chain of type I collagen), which result in structural defects or quantitative reduction of type I collagen. OI can rise from rare gene mutations, which result in defects in post-translational modification, folding, and crosslinking of collagen, bone mineralization, and osteoblast differentiation (12). Whether there is a correlation between the genotype of patients with OI and responses to BP treatment needs to be comprehensively investigated.
We detected pathogenic gene mutations in a relatively large cohort of children with OI who received long-term treatment with zoledronic acid (ZOL) and compared the responses to ZOL treatment among different genotypical groups in order to investigate whether pathogenic gene mutations can contribute to predicting BP efficacy in children with OI.
Patients and Methods
Study Design and Participants
In this single-center, cohort study, the children included were clinically diagnosed with OI in the endocrinology department of Peking Union Medical College Hospital (PUMCH) from May 2009 to September 2020 according to the following criteria of OI: either a history of at least 1 nontraumatic or low-impact fracture and an age- and gender-adjusted areal BMD Z-score of –1.0 or less for either lumbar spine (LS) or femoral neck (FN), or an adjusted areal BMD Z-score of –2.0 or less irrespective of a history of fractures (13). We excluded patients with abnormal renal function (glomerular filtration rate <35 mL/minute), other disease history that influenced bone metabolism, prior administration of agents that affected bone metabolism, or allergic to ZOL.
All patients were administered with an intravenous infusion of ZOL at a dose of 5 mg (for children with body weight >25 kg) or 2.5 mg (for children with body weight ≤25 kg) annually and followed up for at least 1 year. Elemental calcium was supplemented as 300 to 600 mg/day according to the children’s body weight. Calcitriol of 0.25 μg was given to all patients once every other day. Patients would stop the treatment of ZOL and enter drug holiday after achieving age- and gender-specific normal BMD Z-score at both LS and FN (14).
The study was approved by the Scientific Research Ethics Committee of PUMCH, and informed consents were obtained from legal guardians of the patients.
Evaluation of the Effects of ZOL Treatment
We detected serum levels of calcium (Ca), phosphate (P), and alkaline phosphatase (ALP, a bone formation marker) using an automatic biochemical analyzer (ADVIA 1800, Siemens, Germany). Serum levels of 25-hydroxyvitamin D (25OHD), intact parathyroid hormone (PTH), and β-isomerized carboxy-telopeptide of type I collagen (β-CTX, a bone resorption marker) were measured using an automatic analyzer for electrochemiluminescence (E170, Roche Diagnostics, Switzerland). All the parameters were measured 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. Areal BMD at lumbar spine (LS, L2-L4) and proximal hip were measured by dual-energy X-ray absorptiometry using software for children (DXA, Lunar, GE Healthcare, USA). Phantom testing was conducted daily using the DXA device for calibration and quality checks to guarantee the accuracy of BMD measurements. The coefficients of variation (CVs) of DXA measurements were 1.1% and 1.7% at LS and FN, respectively. Results of X-ray films and BMD were interpreted by radiologists. Body height and weight were measured using a Harpenden stadiometer (Seritex Inc., East Rutherford, NJ, USA). Length in supine position was measured alternatively for patients unable to stand. Values of areal BMD, height, and weight were converted to age- and gender-specific Z-scores using data from a general healthy population (15-17).
All the clinical parameters of the patients were measured at baseline and every 6 months during ZOL treatment. The primary endpoints of treatment included changes in BMD at LS, FN, trochanter, and total hip, and incidence of clinical fracture during the treatment. Changes in height and bone turnover markers were used as secondary endpoints. Efficacy outcomes of the patients were assessed before they entered drug holiday or reached adulthood.
Detection of Pathogenic Mutation and Classification of OI
Genomic DNA of the patients was obtained from the peripheral leukocytes by a standard procedure using a DNA Extraction Mini Kit (QIAamp DNA, Qiagen, Frankfurt, Germany). Clinically diagnosed patients with OI underwent paneled sequencing (Illumina HiSeq2000 platform, Illumina Inc., San Diego, CA, USA) according to the previously described protocol (13). The NGS panel covered 20 known candidate genes of OI (COL1A1, COL1A2, IFITM5, SERPINF1, CRTAP, P3H1, PPIB, SERPINH1, FKBP10, PLOD2, BMP1, SP7, TMEM38B, WNT1, CREB3L1, SPARC, MBTPS2, P4HB, SEC24D, and PLS3) and 708 other genes related to bone disorders. The overall sequencing coverage of the target regions was more than 95% at a minimum of 20× sequencing depth. To confirm the mutations detected by the panel, Sanger sequencing was also completed. Polymerase chain reaction was performed to amplify the targeted fragments, which were designed to have a boundary around 100 bp away from the mutation. Then the amplicons were sequenced by a 3730xl DNA sequencer (Applied Biosystems Inc., Foster City, CA, USA).
In accordance with the inherited mode, patients with OI were divided into AD and non-AD groups. AD group included patients with the mutations in COL1A1, COLA2, or IFITM5, and patients carrying the other gene mutations were classified as non-AD groups. As for COL1A1 and COLA2 variants, based on the effects of gene mutations on type I collagen synthesis, mutations leading to amino acid substitutions in the triple helical domain of COL1A1 or COL1A2 were regarded as the collagen structural defect, while nonsense or frameshift mutations causing a premature stop codon in COL1A1 were classified as collagen quantitative reduction (13, 18). Other mutations such as splicing mutations were not evaluated here as it was difficult to predict their effects.
The efficacy outcomes were compared among the different genotype groups, including changes in BMD, incidence of clinical fractures, bone turnover markers, and height during treatment.
Statistical Analysis
The Kolmogorov–Smirnov test was performed to test normal distribution of quantitative data. Normally distributed data (including height, BMD, ALP, β-CTX, Ca, P, 25OHD, and PTH) were expressed as mean ± SD, and the Student t-test was used to compare the difference between 2 groups. Data of abnormal distribution (including number of fractures and fracture frequency) were expressed as median and interquartile range and the comparison was completed between 2 groups with the Mann–Whitney U test. The frequency of fractures was compared before and during ZOL treatment using the Wilcoxon paired test. Qualitative data were expressed as number and proportion (such as fracture incidence), which were compared using Pearson’s chi-squared test. Binary logistic regression analysis was performed to identify related factors of new fracture incidence during ZOL treatment. Multiple linear regression analysis was applied to test the factors that influenced change in BMD Z-score during 3 years of treatment.
Statistical analyses were performed using SPSS software version 25.0 (SPSS Inc., Chicago, IL, USA). Graphs were drawn using GraphPad Prism software version 8.0 (GraphPad, San Diego, CA, USA). A 2-tailed P < .05 considered to be a statistically significant difference.
Results
Pathogenic Mutation and its Relationship With Phenotype at Baseline
A total of 201 children with OI (135 boys and 66 girls) were included in this study; they had initiated ZOL treatment at a median age of 5 years (range 42 days to 17 years) for a median treatment course of 3 years (range 1-8 years). Pathogenic mutations were identified in 11 genes, and mutations in COL1A1 (113/201, 56.2%) and COL1A2 (52/201, 25.9%) accounted for 82.1% of all patients, followed by mutations in IFITM5 (11/201, 5.5%), FKBP10 (6/201, 3.0%), WNT1 (6/201, 3.0%), PLS3 (3/201, 1.5%), TMEM38B (3/201, 1.5%), CRTAP (2/201, 1.0%), P3H1 (2/201, 1.0%), PLOD2 (2/201, 1.0%), and SERPINH1 (1/201, 0.5%). Correspondently, there were 176 patients with AD inheritance, 22 with AR inheritance, and 3 with X-linked inheritance (Fig. 1). Eleven AR inherited children had homozygous variants, and the other 11 AR inherited children carried compound heterozygous variants. The X-linked inherited children were all boys with only 1 kind of pathogenic mutation in PLS3, which followed X-linked dominant inheritance (19). Among patients with defects in type I collagen, the number of patients with collagen structural defects were 74 (34 with defect in α1 chain, 40 with defect in α2 chain), and 49 patients had collagen quantitative reduction.
Flow chart describing mutations in children with OI. OI, osteogenesis imperfecta; AD, autosomal dominant inheritance; AR, autosomal recessive inheritance; XL, X-linked inheritance.
At baseline, genotype–phenotype correlations of patients with OI are shown in Table 1. Patients in non-AD groups had significantly higher β-CTX level than patients in the AD group (P < .001). Patients harboring collagen structural defects had significantly lower Z-scores of height, weight, ALP level, and BMD at LS and FN at baseline than patients with collagen quantitative reduction (P < .001, P = .023, P = .008, P = .031, P = .014). Among patients with collagen structural defects, patients with COL1A2 mutations presented with lower FN BMD Z-score than those with COL1A1 mutations (P = .003).
Baseline characteristics of patients with OI by different genetic defects
| Patients (n) | Age (years) | Gender, male n (%) | Height Z-score | Weight Z-score | LS BMD Z-score | FN BMD Z-score | ALP (IU/L) | β-CTX (ng/mL) | Fracture times (n) | Fracture frequency (n/year) | |
|---|---|---|---|---|---|---|---|---|---|---|---|
| AD OI | 176 | 6.5 ± 4.3 | 116(65.9) | –1.15 ± 1.97 | –0.19 ± 1.61 | –2.42 ± 1.83 | –3.74 ± 2.27 | 303 ± 92 | 0.850 ± 0.317 | 2 (4, 5) | 1.32 (0.73, 2.18) |
| Non-AD OI | 25 | 8.1 ± 4.6 | 19(76.0) | –1.06 ± 1.94 | 0.18 ± 1.27 | –2.56 ± 2.11 | –3.40 ± 2.26 | 275 ± 76 | 1.243 ± 0.487 | 2.5 (2, 5.75) | 1.10 (0.35, 1.50) |
| P value | .078 | .315 | .824 | .282 | .730 | .485 | .147 | <.001 | .293 | .193 | |
| Type I collagen structural defects | 74 | 7.0 ± 4.6 | 48(64.9) | –1.83 ± 2.12 | –0.48 ± 1.87 | –2.79 ± 1.65 | –4.42 ± 2.16 | 277 ± 94 | 0.784 ± 0.344 | 4 (2.5, 6) | 1.33 (0.74, 2.22) |
| Type I collagen quantity reduction | 49 | 6.2 ± 4.0 | 36(73.5) | –0.29 ± 1.41 | 0.27 ± 1.38 | –2.09 ± 1.76 | –3.39 ± 2.24 | 321 ± 82 | 0.909 ± 0.321 | 3 (2, 5) | 1.14 (0.67, 2.00) |
| P value | .355 | .315 | <.001 | .023 | .031 | .014 | .008 | .077 | .305 | .495 | |
| COL1A1 structural mutations | 34 | 6.7 ± 4.6 | 20(58.8) | –2.11 ± 2.30 | –0.29 ± 1.67 | –2.55 ± 1.78 | –3.50 ± 1.70 | 284 ± 106 | 0.789 ± 0.395 | 3 (2, 6) | 1.25 (0.66, 2.63) |
| COL1A2 structural mutations | 40 | 7.2 ± 4.7 | 28(70.0) | –1.65 ± 2.00 | –0.61 ± 2.01 | –2.96 ± 1.55 | –5.07 ± 2.23 | 270 ± 83 | 0.778 ± 0.295 | 4 (3, 6.75) | 1.37 (0.81, 2.18) |
| P value | .641 | .316 | .396 | .501 | .325 | .003 | .526 | .909 | .621 | .605 |
| Patients (n) | Age (years) | Gender, male n (%) | Height Z-score | Weight Z-score | LS BMD Z-score | FN BMD Z-score | ALP (IU/L) | β-CTX (ng/mL) | Fracture times (n) | Fracture frequency (n/year) | |
|---|---|---|---|---|---|---|---|---|---|---|---|
| AD OI | 176 | 6.5 ± 4.3 | 116(65.9) | –1.15 ± 1.97 | –0.19 ± 1.61 | –2.42 ± 1.83 | –3.74 ± 2.27 | 303 ± 92 | 0.850 ± 0.317 | 2 (4, 5) | 1.32 (0.73, 2.18) |
| Non-AD OI | 25 | 8.1 ± 4.6 | 19(76.0) | –1.06 ± 1.94 | 0.18 ± 1.27 | –2.56 ± 2.11 | –3.40 ± 2.26 | 275 ± 76 | 1.243 ± 0.487 | 2.5 (2, 5.75) | 1.10 (0.35, 1.50) |
| P value | .078 | .315 | .824 | .282 | .730 | .485 | .147 | <.001 | .293 | .193 | |
| Type I collagen structural defects | 74 | 7.0 ± 4.6 | 48(64.9) | –1.83 ± 2.12 | –0.48 ± 1.87 | –2.79 ± 1.65 | –4.42 ± 2.16 | 277 ± 94 | 0.784 ± 0.344 | 4 (2.5, 6) | 1.33 (0.74, 2.22) |
| Type I collagen quantity reduction | 49 | 6.2 ± 4.0 | 36(73.5) | –0.29 ± 1.41 | 0.27 ± 1.38 | –2.09 ± 1.76 | –3.39 ± 2.24 | 321 ± 82 | 0.909 ± 0.321 | 3 (2, 5) | 1.14 (0.67, 2.00) |
| P value | .355 | .315 | <.001 | .023 | .031 | .014 | .008 | .077 | .305 | .495 | |
| COL1A1 structural mutations | 34 | 6.7 ± 4.6 | 20(58.8) | –2.11 ± 2.30 | –0.29 ± 1.67 | –2.55 ± 1.78 | –3.50 ± 1.70 | 284 ± 106 | 0.789 ± 0.395 | 3 (2, 6) | 1.25 (0.66, 2.63) |
| COL1A2 structural mutations | 40 | 7.2 ± 4.7 | 28(70.0) | –1.65 ± 2.00 | –0.61 ± 2.01 | –2.96 ± 1.55 | –5.07 ± 2.23 | 270 ± 83 | 0.778 ± 0.295 | 4 (3, 6.75) | 1.37 (0.81, 2.18) |
| P value | .641 | .316 | .396 | .501 | .325 | .003 | .526 | .909 | .621 | .605 |
Values are given as number (proportion), mean ± SD or median (interquartile range). Bold P values indicate statistically significant differences between different genetic defects. Fracture frequency is the number of clinical fractures/disease course.
Abbreviations: β-CTX, β-isomerized carboxy-telopeptide of type I collagen; AD, autosomal dominant inheritance; ALP, alkaline phosphatase; BMD, bone mineral density; FN, femoral neck; LS, lumbar spine; non-AD, autosomal recessive and X-linked inheritance; OI, osteogenesis imperfecta; TOTAL, total hip; TROCH, trochanter.
Baseline characteristics of patients with OI by different genetic defects
| Patients (n) | Age (years) | Gender, male n (%) | Height Z-score | Weight Z-score | LS BMD Z-score | FN BMD Z-score | ALP (IU/L) | β-CTX (ng/mL) | Fracture times (n) | Fracture frequency (n/year) | |
|---|---|---|---|---|---|---|---|---|---|---|---|
| AD OI | 176 | 6.5 ± 4.3 | 116(65.9) | –1.15 ± 1.97 | –0.19 ± 1.61 | –2.42 ± 1.83 | –3.74 ± 2.27 | 303 ± 92 | 0.850 ± 0.317 | 2 (4, 5) | 1.32 (0.73, 2.18) |
| Non-AD OI | 25 | 8.1 ± 4.6 | 19(76.0) | –1.06 ± 1.94 | 0.18 ± 1.27 | –2.56 ± 2.11 | –3.40 ± 2.26 | 275 ± 76 | 1.243 ± 0.487 | 2.5 (2, 5.75) | 1.10 (0.35, 1.50) |
| P value | .078 | .315 | .824 | .282 | .730 | .485 | .147 | <.001 | .293 | .193 | |
| Type I collagen structural defects | 74 | 7.0 ± 4.6 | 48(64.9) | –1.83 ± 2.12 | –0.48 ± 1.87 | –2.79 ± 1.65 | –4.42 ± 2.16 | 277 ± 94 | 0.784 ± 0.344 | 4 (2.5, 6) | 1.33 (0.74, 2.22) |
| Type I collagen quantity reduction | 49 | 6.2 ± 4.0 | 36(73.5) | –0.29 ± 1.41 | 0.27 ± 1.38 | –2.09 ± 1.76 | –3.39 ± 2.24 | 321 ± 82 | 0.909 ± 0.321 | 3 (2, 5) | 1.14 (0.67, 2.00) |
| P value | .355 | .315 | <.001 | .023 | .031 | .014 | .008 | .077 | .305 | .495 | |
| COL1A1 structural mutations | 34 | 6.7 ± 4.6 | 20(58.8) | –2.11 ± 2.30 | –0.29 ± 1.67 | –2.55 ± 1.78 | –3.50 ± 1.70 | 284 ± 106 | 0.789 ± 0.395 | 3 (2, 6) | 1.25 (0.66, 2.63) |
| COL1A2 structural mutations | 40 | 7.2 ± 4.7 | 28(70.0) | –1.65 ± 2.00 | –0.61 ± 2.01 | –2.96 ± 1.55 | –5.07 ± 2.23 | 270 ± 83 | 0.778 ± 0.295 | 4 (3, 6.75) | 1.37 (0.81, 2.18) |
| P value | .641 | .316 | .396 | .501 | .325 | .003 | .526 | .909 | .621 | .605 |
| Patients (n) | Age (years) | Gender, male n (%) | Height Z-score | Weight Z-score | LS BMD Z-score | FN BMD Z-score | ALP (IU/L) | β-CTX (ng/mL) | Fracture times (n) | Fracture frequency (n/year) | |
|---|---|---|---|---|---|---|---|---|---|---|---|
| AD OI | 176 | 6.5 ± 4.3 | 116(65.9) | –1.15 ± 1.97 | –0.19 ± 1.61 | –2.42 ± 1.83 | –3.74 ± 2.27 | 303 ± 92 | 0.850 ± 0.317 | 2 (4, 5) | 1.32 (0.73, 2.18) |
| Non-AD OI | 25 | 8.1 ± 4.6 | 19(76.0) | –1.06 ± 1.94 | 0.18 ± 1.27 | –2.56 ± 2.11 | –3.40 ± 2.26 | 275 ± 76 | 1.243 ± 0.487 | 2.5 (2, 5.75) | 1.10 (0.35, 1.50) |
| P value | .078 | .315 | .824 | .282 | .730 | .485 | .147 | <.001 | .293 | .193 | |
| Type I collagen structural defects | 74 | 7.0 ± 4.6 | 48(64.9) | –1.83 ± 2.12 | –0.48 ± 1.87 | –2.79 ± 1.65 | –4.42 ± 2.16 | 277 ± 94 | 0.784 ± 0.344 | 4 (2.5, 6) | 1.33 (0.74, 2.22) |
| Type I collagen quantity reduction | 49 | 6.2 ± 4.0 | 36(73.5) | –0.29 ± 1.41 | 0.27 ± 1.38 | –2.09 ± 1.76 | –3.39 ± 2.24 | 321 ± 82 | 0.909 ± 0.321 | 3 (2, 5) | 1.14 (0.67, 2.00) |
| P value | .355 | .315 | <.001 | .023 | .031 | .014 | .008 | .077 | .305 | .495 | |
| COL1A1 structural mutations | 34 | 6.7 ± 4.6 | 20(58.8) | –2.11 ± 2.30 | –0.29 ± 1.67 | –2.55 ± 1.78 | –3.50 ± 1.70 | 284 ± 106 | 0.789 ± 0.395 | 3 (2, 6) | 1.25 (0.66, 2.63) |
| COL1A2 structural mutations | 40 | 7.2 ± 4.7 | 28(70.0) | –1.65 ± 2.00 | –0.61 ± 2.01 | –2.96 ± 1.55 | –5.07 ± 2.23 | 270 ± 83 | 0.778 ± 0.295 | 4 (3, 6.75) | 1.37 (0.81, 2.18) |
| P value | .641 | .316 | .396 | .501 | .325 | .003 | .526 | .909 | .621 | .605 |
Values are given as number (proportion), mean ± SD or median (interquartile range). Bold P values indicate statistically significant differences between different genetic defects. Fracture frequency is the number of clinical fractures/disease course.
Abbreviations: β-CTX, β-isomerized carboxy-telopeptide of type I collagen; AD, autosomal dominant inheritance; ALP, alkaline phosphatase; BMD, bone mineral density; FN, femoral neck; LS, lumbar spine; non-AD, autosomal recessive and X-linked inheritance; OI, osteogenesis imperfecta; TOTAL, total hip; TROCH, trochanter.
General Responses to ZOL Treatment of Patients With OI
The general responses of patients with OI to ZOL are shown elsewhere (Table S1 (20)). After 3 years of ZOL treatment, the FN BMD increased from 0.408 ± 0.170 g/cm2 (Z-score –3.70 ± 2.27) to 0.761 ± 0.156 g/cm2 (Z-score 0.61 ± 2.58), and LS BMD increased from 0.419 ± 0.163 g/cm2 (Z-score –2.44 ± 1.87) to 0.819 ± 0.416 g/cm2 (Z-score 1.98 ± 1.98). The serum level of β-CTX declined from 0.896 ± 0.362 ng/mL to 0.656 ± 0.252 ng/mL and ALP level declined from 300 ± 90 IU/L to 216 ± 78 IU/L. The frequency of clinical fractures reduced from 1.066 (887 in 832 patient-years) before treatment to 0.258 per person year (153 in 593 patient-years, P < .001). However, there was no significant increase in Z-score of height during ZOL treatment.
Correlation of Pathogenic Mutation and Response to ZOL Treatment
BMD
As shown in Fig. 2, a significant difference of percentage increase in FN BMD between AD and non-AD groups was observed at 6 months (49.8 ± 43.0% vs 24.3 ± 30.1%, P = .016, Fig. 2A) during ZOL treatment, and patients with OI with AD had greater gains in FN BMD Z-score than patients with OI with non-AD at 2 years (3.80 ± 2.84 vs 2.03 ± 2.14, P = .012) and 3 years (4.46 ± 2.92 vs 2.04 ± 0.96, P < .001, Fig. 2D) of ZOL treatment. The increase in FN BMD was similar between patients with collagen structural defect and those with collagen quantitative reduction (Fig. 2B and 2E). Among patients with collagen structural defects, there was significantly more increase in FN BMD Z-score in patients with mutations in COL1A2 than those with COL1A1 mutations at 1 year (3.29 ± 2.87 vs 1.48 ± 1.86, P = .006), 2 years (4.03 ± 3.33 vs 2.37 ± 1.80, P = .028, Fig. 2F) of ZOL treatment. However, changes in BMD of LS, trochanter, or total hip from baseline were similar between different genotypical groups during ZOL treatment (Fig. 3 and Fig. S1 (20)).
Change of BMD at femoral neck during ZOL treatment. (A) Percentage change of FN BMD in AD and non-AD OI. (B) Percentage change of FN BMD in OI with structural defect and quantitative reduction of type I collagen. (C) Percentage change of FN BMD in OI with structural mutations in COL1A1 and COL1A2. (D) Change of FN BMD Z-score in AD and non-AD OI. (E) Change of FN BMD Z-score in OI with structural defect and quantitative reduction of type I collagen. (F) Change of FN BMD Z-score in OI with structural mutations in COL1A1 and COL1A2. FN, femoral neck; AD, autosomal dominant inheritance; non-AD, autosomal recessive or X-linked inheritance; structure, structural defects of type I collagen; quantity, quantitative reduction of type I collagen; COL1A1, structural mutations in COL1A1; COL1A2, structural mutations in COL1A2. Data are shown as mean and standard error or median and interquartile range. Numbers of individuals during the follow-up were indicated.
Change of BMD at lumbar spine during ZOL treatment. (A) Percentage change of LS BMD in AD and non-AD OI. (B) Percentage change of LS BMD in OI with structural defect and quantitative reduction of type I collagen. (C) Percentage change of LS BMD in OI with structural mutations in COL1A1 and COL1A2. (D) Change of LS BMD Z-score in AD and non-AD OI. (E) Change of LS BMD Z-score in OI with type I collagen structural defect and quantitative reduction. (F) Change of LS BMD Z-score in OI with structural mutations in COL1A1 and COL1A2. LS, lumbar spine; AD, autosomal dominant inheritance; non-AD, autosomal recessive or X-linked inheritance; structure, structural defects of type I collagen; quantity, quantitative reduction of type I collagen; COL1A1, structural mutations in COL1A1; COL1A2, structural mutations in COL1A2. Data are shown as mean and standard error or median and interquartile range.
Bone turnover markers
Compared with patients with collagen structural defects, the serum level of β-CTX of patients with collagen quantitative reduction declined more significantly after 6 months (–22.7 ± 30.3% vs –43.1 ± 20.8%, P = .003), 1 year (–13.2 ± 37.7% vs –36.4 ± 21.9%, P < .001), and 3 years (–6.8 ± 35.0% vs –33.0 ± 22.9%, P = .003, Fig. 4B) of ZOL treatment. Such a difference was not observed between other groups (Fig. 4A and 4C). There were no significant differences in changes of ALP level between different genotypical groups (Fig. 4D-4F).
Change of bone turnover markers from baseline during ZOL treatment. (A) Percentage change of β-CTX in AD and non-AD OI. (B) Percentage change of β-CTX in OI with structural defect and quantitative reduction of type I collagen. (C) Percentage change of β-CTX in OI with structural mutations in COL1A1 and COL1A2. (D) Percentage change of ALP in AD and non-AD OI. (E) Percentage change of ALP in OI with structural defect and quantitative reduction of type I collagen. (F) Percentage change of ALP in OI with structural mutations in COL1A1 and COL1A2. β-CTX, β-isomerized carboxy-telopeptide of type I collagen; ALP, alkaline phosphatase; AD, autosomal dominant inheritance; non-AD, autosomal recessive or X-linked inheritance; structure, structural defects of type I collagen; quantity, quantitative reduction of type I collagen; COL1A1, structural mutations in COL1A1; COL1A2, structural mutations in COL1A2. Data are shown as mean and standard error.
New fracture incidence
During ZOL treatment, 89 patients had new fractures. The proportion of patients with a new fracture during the first 2 years of ZOL treatment was higher in the collagen structural defect group (35.1%) than in the collagen quantitative reduction group (18.4%, relative risk 0.52, P = .044), but the difference was not statistically significant during the 3 years of ZOL treatment (40.5% vs 24.5%, relative risk 0.61, P = .066, Fig. 5B). There were no significant differences in fracture incidence between other groups (Fig. 5A and 5C).
Incidence of new fracture and related factors during ZOL treatment. (A) Percentage of patients with new fracture in AD and non-AD OI. (B) Percentage of patients with new fracture in OI with structural defect and quantitative reduction of type I collagen. (C) Percentage of patients with new fracture in OI with structural mutations in COL1A1 and COL1A2. (D) Related factors of new fracture during ZOL treatment in patients with type I collagen mutations. AD, autosomal dominant inheritance; non-AD, autosomal recessive or X-linked inheritance; structure, structural defects of type I collagen; quantity, quantitative reduction of type I collagen; COL1A1, structural mutations in COL1A1; COL1A2, structural mutations in COL1A2; OR, odds ratio; CI, confidence interval.
Height and weight
We found that height and weight increased during ZOL treatment, but observed no significant increase in height Z-score (Table S1 (20)). We failed to observe any significantly different change in Z-score of height or weight between the different genotypical groups (Fig. S2 (20)).
Related Factors of new Fracture Risk and Change in BMD Z-score
As shown in Fig. 5D, starting ZOL treatment earlier (OR 0.83, 95% CI 0.74-0.94, P = .002) and quantitative reduction in type I collagen (OR 0.31, 95% CI 0.10-0.92, P = .034) were predictors of fewer new fractures during ZOL treatment in patients with type I collagen defects.
Multiple linear regression analysis revealed that change in FN BMD Z-score after 3 years of ZOL treatment was positively correlated with collagen quantitative reduction (β = 1.14, P = .030), but negatively correlated with age at ZOL initiation (β = -0.38, P < .001) and FN BMD Z-score at baseline (β = –0.77, P < .001, Table 2). However, change in Z-score of LS BMD after 3 years of ZOL treatment was not associated with collagen defect type (P = .851, Table S2 (20)).
Factors related to change in FN BMD Z-score after 3 years of ZOL treatment among patients with collagen defects
| Factors | Multiple R = 0.76, P < 0.001 | |
|---|---|---|
| Partial regression coefficient (95% CI) | P value | |
| Age at ZOL initiation | –0.38 (–0.48, –0.28) | <.001 |
| Female vs male | –0.71 (–1.52, 0.10) | .086 |
| Quantity reduction vs structural defect of type I collagen | 1.14 (0.11, 2.16) | .030 |
| Mutations in COL1A2 vs COL1A1 | 0.25 (–0.77, 1.28) | .624 |
| FN BMD Z-score at baseline | –0.77 (–0.97, –0.57) | <.001 |
| Factors | Multiple R = 0.76, P < 0.001 | |
|---|---|---|
| Partial regression coefficient (95% CI) | P value | |
| Age at ZOL initiation | –0.38 (–0.48, –0.28) | <.001 |
| Female vs male | –0.71 (–1.52, 0.10) | .086 |
| Quantity reduction vs structural defect of type I collagen | 1.14 (0.11, 2.16) | .030 |
| Mutations in COL1A2 vs COL1A1 | 0.25 (–0.77, 1.28) | .624 |
| FN BMD Z-score at baseline | –0.77 (–0.97, –0.57) | <.001 |
Abbreviations: BMD, bone mineral density; FN, femoral neck. Bold P values indicate significantly related factors.
Factors related to change in FN BMD Z-score after 3 years of ZOL treatment among patients with collagen defects
| Factors | Multiple R = 0.76, P < 0.001 | |
|---|---|---|
| Partial regression coefficient (95% CI) | P value | |
| Age at ZOL initiation | –0.38 (–0.48, –0.28) | <.001 |
| Female vs male | –0.71 (–1.52, 0.10) | .086 |
| Quantity reduction vs structural defect of type I collagen | 1.14 (0.11, 2.16) | .030 |
| Mutations in COL1A2 vs COL1A1 | 0.25 (–0.77, 1.28) | .624 |
| FN BMD Z-score at baseline | –0.77 (–0.97, –0.57) | <.001 |
| Factors | Multiple R = 0.76, P < 0.001 | |
|---|---|---|
| Partial regression coefficient (95% CI) | P value | |
| Age at ZOL initiation | –0.38 (–0.48, –0.28) | <.001 |
| Female vs male | –0.71 (–1.52, 0.10) | .086 |
| Quantity reduction vs structural defect of type I collagen | 1.14 (0.11, 2.16) | .030 |
| Mutations in COL1A2 vs COL1A1 | 0.25 (–0.77, 1.28) | .624 |
| FN BMD Z-score at baseline | –0.77 (–0.97, –0.57) | <.001 |
Abbreviations: BMD, bone mineral density; FN, femoral neck. Bold P values indicate significantly related factors.
Discussion
To our knowledge, this is the first study to comprehensively evaluate the responses to ZOL treatment and pathogenic mutations in a relatively large cohort of children with OI. We found that AD patients with OI had more increase in FN BMD Z-score than patients with other inheritance patterns during ZOL treatment. Compared with patients with collagen structural defects, patients with collagen quantitative reduction were likely to have lower incidence of new fractures during ZOL treatment. In addition, regression analysis revealed that collagen quantitative reduction and starting ZOL treatment earlier was correlated with better efficacy of ZOL in patients with OI.
BPs are widely used off-label in the treatment of patients with OI with the intention of increasing BMD and reducing the risk of fracture. ZOL, a third-generation BP, is relatively potent in suppressing bone resorption by disrupting osteoclast formation, survival, and cytoskeletal dynamics (21). Studies have shown that most patients with OI have low bone mass, which is mirrored by decreased bone cortical width and cancellous volume at the microscopic scale (22). Despite the fact that BPs can hardly improve bone tissue properties of patients with OI, they can increase trabecular volume and cortical thickness, and reduce cortical porosity, and there is strong evidence that BPs can increase bone mineral density in patients with OI (23, 24). However, the effects of BPs on fracture occurrence have been inconsistent. Several meta-analyses failed to support significant effects of BPs on fracture prevention in patients with OI (25, 26). As OI is a genetically and clinically heterogeneous disease with at least 20 pathogenic genes; therefore, complex genotypes and phenotypes led to inconsistent outcome of BPs in reducing fractures of patients with OI (27). It was assumed that studies with more mild forms of patients with OI were likely to get positive results, in contrast with studies enrolling more moderate to severe OI, which could hardly conclude effects of BPs in reducing fracture incidence (2).
In this study, we found that responses of patients with OI to ZOL treatment were closely associated with the pathogenic gene mutations. After 3 years of ZOL treatment, the increase in femoral neck BMD Z-score in patients with OI with autosomal dominant inheritance was greater than that in patients with autosomal recessive or X-linked inheritance. Collagen structural defects was negatively correlated with the increase in femoral neck BMD Z-score during ZOL treatment. The difference of responses to ZOL treatment between AD and non-AD patients with OI may be associated with the various defects in collagen processing, bone mineralization, and osteoblast development (12). Genes such as P3H1, CRTAP, and TMEM38B are related to post-translational modification of type I collagen. Mutations in FKBP10, PLOD2, and SERPINH1 can result in defects in collagen folding and crosslinking. WNT1 is closely implicated in osteoblast differentiation through the WNT pathway. Genotype–phenotype correlation analyses have revealed that AR OI is usually associated with more severe phenotypes, which implies that defects in the above collagen metabolic process could have a more obvious impact on bone tissue properties, thus resulting in poor response to BP treatment (22, 28).
Collagen defect classification was relatively common in genotype–phenotype correlation analysis in OI with COL1A1 or COL1A2 variants. Mutations leading to structurally abnormal collagen was related to severe OI, whereas quantitative deficiency of structurally normal collagen tended to cause milder forms (12). Interestingly, herein we also observed that patients with collagen structural defects had higher incidence of new fracture and less decline in serum β-CTX level than those with collagen quantitative reduction during ZOL treatment. The increased susceptibility of fractures in patients with OI was closely associated with aberrant extracellular matrix, characterized by disordered arrangement of collagen fibrils and more nonenzymatic crosslinks. These properties resulted in reduced resistance to crack propagation in mouse models (22, 29). Collagen structural defects may have more severe consequences for bone matrix structure than quantity reduction of collagen (22), which can partially explain that patients with collagen structural defects were more likely to have a new fracture than those with quantity reduction of collagen during ZOL treatment. Another likely explanation is that bones with quantity reduction of collagen still provide a relatively normal microenvironment for BPs to deposit, where they induce apoptosis of osteoclasts leading to BMD increase (21). By contrast, fibers formed by structurally abnormal collagen could be disorganized and inconsistent in shape, which made it difficult for BPs to deposit and play roles, leading to a poor response to BPs treatment (22, 24).
Moreover, although LS BMD was used as a standard endpoint for BP studies in children with OI (23), which had pronounced gains during ZOL treatment in this study, intriguingly no significant differences were found in the response of LS BMD to ZOL treatment among different genotypical groups. Previous studies showed that BPs had better effects on increasing volume of trabecular bone than cortical bone. Gene defects may have little effects on the response of spine to ZOL treatment, and we speculate that it may be related to the abundant trabecular bone in the spine (30). However, the outcome of response to BPs in different skeletal sites was inconsistent in different studies. A 2-year clinical trial showed that BPs tended to have better efficacy in improving BMD Z-score of total body in mild patients with OI (31). Another study reported that the increase in LS BMD was significantly higher in patients with OI type III than with OI type IV during pamidronate therapy (32). We observed better response to ZOL treatment in patients with OI with AD inheritance and with collagen quantitative reduction, which was often associated with mild OI phenotypes (18, 33, 34). Consistent with our findings, a 2-year prospective study found a more pronounced motor development improvement in AD children with OI than AR children with OI during ZOL treatment, but they did not analyze the change of BMD (35). Two recent studies did not find association between the change of LS BMD and collagen defect type during pamidronate treatment (4, 36). Unfortunately, pharmacogenetic studies to investigate the association between pathogenic mutation and responses to BPs treatment in patients with OI were very scarce.
The progress of molecular diagnosis has promoted the precise drug treatment of OI. For patients with OI with poor responses to BPs, the efficacy of other therapeutic drugs is worth in-depth research, such as teriparatide and monoclonal antibodies to RANKL, sclerostin and transforming growth factor-β (37-40). For instance, OI type VI patients caused by SERPINF1 mutations had poor response to BPs treatment, but denosumab seemed to have superior efficacy in these patients (41, 42).
In this relatively large-sample study in children with OI, we comprehensively evaluated the relationship of pathogenic mutation and responses to ZOL treatment. Our findings may help to predict BP efficacy in patients with OI on the basis of pathogenic mutations. In our study, all patients were administered with ZOL, which could avoid bias of various BPs types. All parameters of this study were detected in a single center, thus minimizing measurement bias. However, there were a series of limitations in this study. There were 82.1% patients carrying mutations in COL1A1 and COL1A2, and the sample size was insufficient, so it was difficult to analyze the relationship of rare gene mutations and response to ZOL. It was also difficult to rule out the effects of different rates of puberty on the results of BMD and bone turnover biomarkers. Additionally, the influences of pathogenic gene mutations on bone microstructure and biomechanical properties deserve in-depth study to reveal the possible mechanism of different gene mutations affecting BP efficacy.
In conclusion, responses to ZOL treatment are possibly associated with genotypes of patients with OI. Patients with OI with non-AD inheritance or with pathogenic gene mutations leading to collagen structural defects may have relatively poor responses to ZOL treatment, which is possibly associated with their more severe phenotypes. New therapeutic agents are worth developing in these patients.
Abbreviations
- 25OHD
25-hydroxyvitamin D
- β-CTX
β-isomerized carboxy-telopeptide of type I collagen
- AD
autosomal dominant
- ALP
alkaline phosphatase
- AR
autosomal recessive
- BMD
bone mineral density
- BP
bisphosphonate
- CV
coefficient of variation
- FN
femoral neck
- LS
lumbar spine
- NGS
next-generation sequencing
- OI
osteogenesis imperfecta
- PTH
parathyroid hormone
- ZOL
zoledronic acid
Funding
This study was supported by National Key R&D Program of China (2021YFC2501704), CAMS Innovation Fund for Medical Sciences (CIFMS) (2021-I2M-C&T-B-007, 2021-I2M-1-051), National Natural Science Foundation of China (nos. 81873668, 82070908), and Beijing Natural Science Foundation (7202153).
Author Contributions
M.L. contributed to conceptualization and study design of the research. L.S. collected the clinical data, analyzed data, and drafted the manuscript. J.H. and J.L. contributed to collecting clinical data and performing the experiments. All the authors contributed to revising and approving the final version of the manuscript.
Disclosures
The authors declare that there is no conflict of interest regarding the publication of this paper.
Data Availability
Some or all datasets generated during and/or analyzed during the current study are not publicly available to preserve patient confidentiality but are available from the corresponding author on reasonable request.
