Elevated periostin level in serum of adults with osteogenesis imperfecta is associated with disease severity.

A total of 61 adult patients with OI and 61 age-, sex-, and BMI-matched controls were included in the analysis. Thirty-two of the 61 OI patients (52%) were female, and 12 of these women (37%) were postmenopausal. The mean age was similar between groups (42.8 ± 15.8 years in OI vs 42.6 ± 15.5 years in controls; P = .93), as was BMI (24.8 ± 4.3 vs 23.7 ± 3.3 kg/m 2 , respectively; P = .11). However, individuals with OI were significantly shorter (158 ± 14 cm vs 169 ± 11 cm, P < .001) and lighter (62.2 ± 14.9 kg vs 68.4 ± 15.3 kg, P = .02) than controls, which is consistent with the classical morphology of OI patients ( Table 1 ). Demographic and clinical characteristics of patients with OI and matched controls Values are presented as mean (standard deviation (SD)) for continuous variables and n (%) for categorical variables. We compared the data between controls and OI with a P value calculation using a t test. T-scores, Z-scores, and BMD were derived from dual-energy X-ray absorptiometry. Z-scores are provided only for premenopausal women and men <50 years. Severe fractures defined as femur/tibia/humerus/vertebral fractures. Abbreviations: BMD, bone mineral density (g/cm 2 ); BMI, body mass index; VAS, visual analogue scale on 10-cm line. * P < 0.05. Among OI patients, 91% were classified as Sillence type I and 8% as type III. The mean age at diagnosis was 8.9 ± 13.0 years, and 58% reported a family history of the disease. Approximately 72% of patients reported resting pain, with a mean of visual analog scale (VAS) score of 3.9 ± 2.9. The average number of lifetime fractures (excluding skull fractures and fractures of the fingers, metacarpal, and metatarsal bones) was 20.4 ± 17.2, with a mean of 5.0 ± 5.0 severe fractures (defined as fractures of the pelvis, hip, humerus, or vertebrae). Nearly half (47%) had scoliosis, with a mean Cobb angle of 27.6 ± 21.0 degrees. The mean height-wingspan difference, indicative of axial skeletal disproportion, was 5.5 ± 7.1 cm. Extra-skeletal manifestations were common: dentinogenesis imperfecta was present in 26% and hearing loss in 30% of patients. Regarding bone mineral density (BMD), mean T-scores were −2.1 ± 1.5 at the total hip, −1.5 ± 1.2 at the femoral neck, and −1.7 ± 1.3 at the lumbar spine. Additionally, 61% of OI patients had received bisphosphonate therapy. No patient in our cohort had sustained a fracture within the year preceding inclusion. No patient had received bone-specific pharmacological treatment other than calcium and vitamin D supplementation in the 2 years preceding enrollment. Demographic and clinical data are presented in Table 1 . Serum periostin levels were significantly higher in patients with OI compared to age-, sex-, and BMI-matched controls. The mean periostin concentration was 796.5 ± 209 pmol/L in the OI group, vs 713.6 ± 167 pmol/L in the control group ( P = .017). These results are visually represented in Fig. 1 . Serum periostin levels in patients with OI and matched controls. Boxplot illustrating serum periostin concentrations measured by ELISA in patients with OI and age-, sex-, and BMI-matched healthy controls. The assay used was a sandwich ELISA. The horizontal lines within the boxes represent the medians, boxes denote interquartile ranges, and whiskers indicate minimum and maximum values. * P < .05 was considered statistically significant. Among OI patients, serum periostin levels varied according to several demographic characteristics. Periostin was higher in females than in males (861.4 vs 724.8 pmol/L, P = .01), and this association remained significant after FDR correction. Menopausal status was not associated with periostin (847.8 vs 887.4 pmol/L, P = .67), and no significant correlation was observed with BMI ( r = −0.086, P = .515). Smoking status was not associated with periostin in univariate analyses (809.3 vs 730.8 pmol/L, P = .18) ( Table 2 ). Comparison of Serum periostin levels according to clinical and biological characteristics in OI patients Mean serum periostin concentrations (pmol/L) are shown for each subgroup of OI patients according to key clinical and biological characteristics. Comparisons were performed using independent t -tests. To account for multiple subgroup comparisons, FDR-adjusted q -values were computed using the Benjamini–Hochberg procedure. Two-sided q -values < 0.05 were considered statistically significant. *P < .05 was considered statistically significant. Regarding skeletal involvement, periostin tended to be higher in Sillence type 3 compared with type 1 (1108.1 vs 768.6 pmol/L, P = .05). Periostin was significantly higher in patients with scoliosis than in those without (860.1 vs 723.9 pmol/L, P = .01), an association that remained significant after FDR correction ( Table 2 ). In correlation analyses, periostin was positively associated with spinal curvature severity ( r = 0.496, P = .012) and with height–wingspan discrepancy ( r = 0.300, P = .023), both reflecting axial skeletal deformity; however, these correlations did not remain statistically significant after FDR correction (Table S1) ( 21 ). Periostin showed a positive correlation with femoral neck BMD T-score ( r = 0.279, P = .04), but this association was not retained after FDR correction. Periostin was not significantly associated with total fracture number ( r = 0.190, P = .143) or with severe fractures in univariate analyses ( r = 0.241, P = .06). No association was found with pain severity (VAS: r = −0.026, P = .844). Serum periostin levels did not differ significantly according to a history of fractures at the humerus, femur, pelvis, or spine, and none of these associations remained significant after Benjamini–Hochberg FDR correction (all q = 0.612; Table S6) ( 21 ). Finally, periostin was not significantly associated with extra-skeletal manifestations including dentinogenesis imperfecta (824.7 vs 764.1 pmol/L, P = .38) or hearing loss (777.3 vs 850.3 pmol/L, P = .34) in univariate comparisons ( Table 2 ). Additional analyses in controls showed no significant association between serum periostin and sex ( P = .27), smoking status ( P = .31), age ( r = 0.157; P = .28), or BMI ( r = −0.082; P = .53), whereas the patient–control difference in periostin was significant in women ( P = .02) and among non-smokers ( P = .01) (Table S2) ( 21 ). In exploratory subgroup analyses restricted to patients with Sillence type 1 OI ( n = 56), periostin levels remained higher in women than in men and tended to be lower in current smokers, in COL1A2 vs COL1A1 mutation carriers, in patients with dentinogenesis imperfecta, and in those receiving bisphosphonates, although these differences did not remain statistically significant after FDR correction (Table S3) ( 21 ). Within Sillence type 1, BMI was inversely correlated with periostin ( r = −0.32, P = .019, q = 0.27), while other skeletal parameters and fracture burden showed no robust associations (Table S4) ( 21 ). To identify independent predictors of serum periostin levels, a multivariable linear regression model was constructed, including variables selected based on their clinical relevance and prior univariate associations. Due to strong collinearity between several clinical characteristics of disease severity—including spinal curvature, height–arm span discrepancy, and number of severe fractures—we retained only the number of severe fractures in the final model to avoid redundancy. Notably, when either spinal curvature or height–arm span difference was used in lieu of severe fractures, each remained independently associated with periostin levels, reinforcing the robustness of this relationship. In the final multivariable linear regression model, serum periostin levels were analyzed after adjustment for BMI, femoral neck BMD T-score, sex, smoking status, number of severe fractures, dentinogenesis imperfecta, and hearing loss. In the final model, two variables were significantly associated with serum periostin levels. A positive association was observed with the number of severe fractures (β = 25.6, P = .041), indicating that periostin levels increased with fracture severity. In contrast, smoking was associated with significantly lower periostin levels (β = −269.8, P = .012). Hearing loss showed a trend toward higher periostin concentrations (β = 205.7, P = .083), although this did not reach statistical significance. Other variables included in the model, such as BMI, femoral neck T-score, sex, and presence of dentinogenesis imperfecta, were not independently associated with periostin levels ( Table 3 ). Multivariable linear regression model for serum periostin levels in patients with OI Linear regression model including clinically relevant variables and factors previously associated with periostin levels in univariate analyses. β (Estimate) represents the effect size expressed in pmol/L of periostin. Severe fractures defined as femur/pelvis/humerus/vertebral fractures. Abbreviations: BMD, bone mineral density; BMI, body mass index. *P < .05 was considered statistically significant. In a multivariable model restricted to Sillence type 1, only dentinogenesis imperfecta remained independently associated with lower periostin levels, whereas sex, smoking status, and the number of severe fractures were not significant predictors (Table S5) ( 21 ).

Serum periostin levels were elevated in adults with OI compared to controls, and several characteristics of disease severity emerged as independent predictors of higher periostin levels. These findings suggest a potential link between periostin and disease severity.

In this study, we have evaluated for the first time serum periostin concentrations in a well-characterized cohort of adult patients with OI compared to matched healthy controls. We found that periostin levels were significantly higher in the OI group. Among OI patients, periostin levels were associated with sex and with several clinical indicators of disease severity, including scoliosis and Sillence subtype. Correlation analyses further revealed that periostin was positively associated with markers of axial skeletal deformity and BMD. In multivariable analysis adjusted for sex and other relevant covariates, the number of severe fractures emerged as an independent predictor of higher periostin levels, whereas smoking was independently associated with lower levels. The loss of a clear severity gradient when restricting analyses to Sillence type 1 suggests that the association between periostin and OI severity in the overall cohort is primarily captured by the contrast between type 1 and the more severe type 3 patients, who exhibited the highest periostin concentrations. Periostin, also known as osteoblast-specific factor 2, is an extracellular matrix glycoprotein enriched in collagen-rich connective tissues exposed to sustained mechanical loading (eg, periosteum, periodontal ligament) and contributes to connective tissue remodeling and matrix integrity ( 22 ). Periostin contains an N-terminal cysteine-rich EMI domain that binds key extracellular matrix components such as type I collagen and fibronectin, thereby supporting collagen fibril organization and anchorage within the mineralized matrix—mechanisms that may be particularly relevant in osteogenesis imperfecta ( 22 , 23 ). Beyond its structural role, periostin participates in mechanoadaptation through cell–matrix signaling and modulation of osteogenic pathways, processes likely pertinent to OI given the altered matrix mechanics and osteoblast function associated with type I collagen defects. Experimental data indicate that periostin can potentiate Wnt/β-catenin signaling (partly via suppression of sclerostin and β-catenin stabilization) and that periostin deficiency is associated with impaired Wnt signaling and reduced cortical bone formation under mechanical loading ( 24 , 25 ). In parallel, periostin is a downstream effector of TGF-β signaling and is upregulated by TGF-β1 in contexts of tissue injury and remodeling ( 6 , 26 ). Because TGF-β signaling is reported to be increased in OI bone ( 27 , 28 ), elevated periostin in OI could reflect, at least in part, enhanced TGF-β activity and a compensatory matrix-remodeling response. Finally, periostin is expressed in extra-skeletal tissues including heart and lungs, where it has been implicated in tissue repair, fibrosis, and inflammation ( 11 ), and been proposed as a biomarker of type 2 inflammation in allergic diseases ( 29 ), raising the possibility that circulating periostin may integrate both skeletal and extra-skeletal components of the OI phenotype. To date, only one study has indirectly explored a periostin–OI connection by comparing valvular involvement in oim/oim and periostin knockout mouse models ( 30 ). No study has directly investigated circulating periostin levels in patients with OI. In this context, our finding of higher serum periostin in adults with OI supports the hypothesis that periostin is altered in OI, plausibly at the interface of defective type I collagen, dysregulated Wnt/TGF-β signaling, and altered mechanobiology. Since periostin promotes type I collagen fibrillogenesis and matrix maturation, and pyridinoline—an index of mature enzymatic collagen cross-linking—may be reduced in OI ( 31 ), periostin and pyridinoline may reflect complementary aspects of bone matrix quality ( 6 , 31 ). The absence of differences between qualitative and quantitative COL1A1/COL1A2 variants further suggests that this periostin signal may represent a shared downstream pathway rather than a mutation-class–specific effect. This hypothesis is further supported by the observed association between elevated periostin levels and markers of axial skeletal severity in OI patients. Specifically, periostin concentrations were significantly higher in individuals with scoliosis and positively correlated with both the degree of spinal curvature and the arm-span to height discrepancy. These two parameters are commonly used to assess axial skeletal deformity and are likely indicative of underlying vertebral compression fractures and progressive vertebral remodeling. The relationship between periostin and these deformities suggests that periostin may be upregulated as part of a tissue-level response to abnormal mechanical stress and microarchitectural instability in the axial skeleton. This aligns with prior findings in other bone fragility contexts, where periostin has been shown to be upregulated in response to mechanical strain and fracture repair stimuli ( 6 , 32 ). One study suggests that discal degeneration is associated with an upregulation of periostin in addition to increased expression levels of type I collagen, which could also help explain the association with scoliosis ( 33 ). Interestingly, our results revealed higher periostin levels in female patients with OI compared to males, possibly due to hormonal influences on periosteal remodeling, although the existing literature on sex-related differences in periostin expression remains divergent ( 34 , 35 ). In our multivariable analysis, smoking was independently associated with lower periostin levels which is consistent with literature ( 36 ). This is in line with prior studies showing that tobacco use negatively impacts bone metabolism and extracellular matrix composition, potentially through oxidative stress and impaired osteoblast function, which may downregulate periostin expression. After adjusting for sex, BMI, and smoking status, the disease severity, reflected by the number of severe fractures, emerged as an independent predictor of periostin levels. This reinforces the hypothesis that periostin may be a marker of chronic skeletal remodeling activity in OI, particularly in the context of cumulative bone damage and could be associated with the OI severity. Similar associations between periostin and fracture burden have been observed in postmenopausal osteoporosis and in patients with fibrous dysplasia, where periostin has been proposed as an index of bone turnover and skeletal fragility ( 14 , 15 ). Importantly, none of the patients in our cohort had experienced a recent fracture (within the year prior to inclusion), suggesting that the observed elevation in serum periostin is unlikely to reflect an acute or subacute post-fracture remodeling response. This study is the first to investigate circulating periostin levels in adults with OI, highlighting its potential link with skeletal severity. Among its strengths are a well-characterized and clinically relevant cohort, an appropriate control group matched for age, sex, and BMI, and the use of a robust and reproducible ELISA assay. Despite the OI rarity, the sample size remains substantial. The statistical methodology was rigorous, with adjustments for relevant confounding factors. A key limitation is the absence of replication in an independent OI cohort; therefore, these findings should be viewed as hypothesis generating and require confirmation in larger external datasets. The sample size, although reasonable for a rare disease, limits the power of subgroup analyses. The number of correlations tested may increase the risk of type I error, although findings remained consistent with biological hypotheses. Importantly, periostin is expressed in multiple tissues beyond bone, including the lung and heart valves—both potentially affected in OI ( 37 ). The origin of increased periostin levels in OI therefore remains uncertain. We did not record menstrual cycle phase at the time of sampling, which could influence circulating periostin levels according to previous reports ( 12 ). This limitation should be considered when interpreting the observed sex-related differences. Finally, while periostin was associated with fracture severity in our cohort, its predictive value in children or early disease stages remains to be explored.

This study is a secondary analysis of the miROI study, which was originally designed as a prospective observational study aimed at identifying biomarkers associated with OI, focusing primarily on microRNA dysregulation. Here, we have conducted a matched case–control study design. The present periostin analysis was pre-specified and hypothesis-driven. Periostin was not selected through an untargeted screening of multiple protein biomarkers. Patients were prospectively recruited from October 2019 to October 2021, with an extension until April 2022. The detailed study design and initial objectives have been previously published ( 17 ). A total of 61 adult patients with OI carrying a confirmed COL1A1 or COL1A2 mutation were included in the present analysis. Control participants were carefully matched for age, sex, and BMI from three established population-based cohorts: OFELY, MODAM, and STRAMBO ( 18-20 ). The study protocol received approval from a French ethics committee (CPP Ile-de-France VI, registration number 2019-A00521-56) and was conducted in compliance with French national ethical guidelines and the ethical standards set forth by the revised Helsinki Declaration. The study is registered at ClinicalTrials.gov under the identifier NCT04009733 . The recruitment was performed either during hospitalization or during consultation in the reference center for bone diseases at the Edouard Herriot University Hospital in Lyon, France (OSCAR network). Written informed consent was obtained from each participant. Further methodological details can be found in the original publication ( 17 ). Blood samples were collected in the morning after an overnight fast, centrifuged at 2000 g at 4 °C within 1 hour of collection for serum preparation, aliquoted, and stored at −80 °C until analysis. Macroscopic visual analysis of the blood samples was performed to assess the quality of the serum, including the presence of hemolysis and fibrin clotting. Serum periostin levels were measured using a commercial sandwich enzyme-linked immunosorbent assay (ELISA) (Biomedica Cat# BI-20433, RRID: AB_3720966, Biomedica, Vienna, Austria). This assay detects all known human periostin isoforms using a monoclonal mouse anti-human periostin antibody as the capture antibody (binding an epitope near the N-terminus, in the FAS4 domain, AA542-549, Uniprot ID Q15063 , isoform 1) and a polyclonal goat anti-human periostin antibody as the detection antibody. All procedures followed the manufacturer's instructions. The assay demonstrates high precision, with an intra-assay coefficient of variation below 3% and an inter-assay variation below 6%. Results are reported in picomoles per liter (pmol/L). The range for the assay was 20 to 4000 pmol/L Clinical information was collected both by a clinical questionnaire filled in by the physician with the patient. Patients’ clinical data were obtained from their standardized electronic medical records (from 2009 to today, DPI Easily®). The Sillence Classification and the assessment of the disease severity were made by a rheumatologist of the reference center for rare bone diseases. Lifetime history was captured from the electronic medical record and the standardized OI follow-up note template routinely used in our center, recording both the total number of fractures and fracture sites. This way we also accessed radiographs and surgical reports. All details can be found in the original publication ( 17 ). The primary endpoint of the study was the comparison of serum periostin levels between patients with OI and matched controls. Continuous variables were summarized as means with standard deviations. Categorical variables were described as counts and percentages. For the primary comparison, group differences in periostin levels were evaluated using Student's t -test. Secondary analyses aimed to explore clinical and biological correlates of serum periostin levels within the OI cohort. Pearson's correlation coefficients were computed to assess associations between periostin and continuous variables. For categorical variables (binary or nominal), comparisons of periostin levels were performed using independent t -tests. We adjusted P -values for multiple comparisons using the Benjamini–Hochberg false discovery rate (FDR) procedure. Variables that showed significant associations in univariate analysis, as well as clinically relevant variables, were included in a multivariable linear regression model to identify independent predictors of serum periostin levels. Multicollinearity was assessed using the variance inflation factor (VIF). All tests were two-sided. For the primary endpoint (OI vs matched controls), a P -value < .05 was considered statistically significant. For secondary analyses involving multiple comparisons (subgroup comparisons and correlation analyses), P -values were adjusted using FDR, and q -values < 0.05 were considered statistically significant. Unadjusted P -values are provided alongside q -values for transparency. Graphics and figures were performed using GraphPad Prism version 8.0.0 for Windows (GraphPad Software, San Diego, CA, USA, http://www.graphpad.com ).