Exosome-mediated cholesterol flow drives scoliosis progression via promoting the spinal cartilage-bone positive feedback

Abstract

Adolescent Idiopathic Scoliosis (AIS) is the most common form of spinal deformity among adolescents. To explore its etiology of progression and scoliosis-modifying drugs, chondrocytic senescence was confirmed in AIS facet joint cartilage by analyzing clinical specimen. Furthermore, through 4D/480 label-free proteomics analysis, we identified an exosome-mediated positive feedback loop during scoliosis progression, which driving the elevation of cholesterol flow between spinal cartilage and vertebra. To further investigate the pathological significance of the loop in vivo, high-cholesterol flow was reconstructed in C57BL/6 J mice by injecting with recombinant adeno-associated virus rAAV9-Runx2-HMGCR. Our results confirmed the important role of the positive feedback loop in the development of scoliosis. Meanwhile, Avasimibe or/and Corylin were used to delay the scoliosis progression by targeting the key exosomal proteins APOB (Apolipoprotein B-100) or/and HSP90β (Heat Shock Protein 90-beta). This research extends the etiology of scoliosis progression and provides an alternative perspective for scoliosis non-surgical treatment.

Mechanistic modeling of scoliosis sheds light on the role of the exosome-mediated high cholesterol flow between spinal cartilage and bone tissue in the context of progression of scoliosis.

Introduction

Adolescent idiopathic scoliosis (AIS) is the most common form of spinal deformity in adolescents, with an average prevalence of 1–4% globally¹. This three-dimensional spinal curvature is diagnostically defined by a Cobb angle ≥10°. As the curve progresses, patients with AIS will suffer pain, appearance deformities, and physical and psychological disabilities. An epidemiological study encompassing over two million Chinese adolescents across four decades established the national prevalence at 1.18%². Notably, clinical stratification reveals only 4.1% of cases progress to severe deformity (Cobb angle > 40°) requiring surgical correction, while 95.9% present with mild-to-moderate curvature (Cobb angle 10°–39°). This distribution underscores the critical role of non-operative management strategies in AIS disease control. However, due to the unclear etiology of AIS initiation and progression, there is a lack of scoliosis-modifying drugs in clinical practice. Current management relies on long-term follow-ups and prolonged use of orthotic braces to delay curve progression, which, along with extended treatment duration, high risks, and subsequently a financial burden of 7–27%³. Therefore, aside from investigating the disease’s origins, it is imperative to explore key factors driving AIS progression and develop effective drug targets to delay disease advancement.

The etiology of AIS is multifactorial, besides genetic and hormonal factors¹, several recent studies have shown that the aggravation of spinal deformity is closely related to biomechanical asymmetry in the spine, and biological abnormalities of the intervertebral discs and facet joint cartilages^4–6. During the progression of AIS, as the spine becomes increasingly distorted and rotated, the facet joints passively endure unbalanced rotational stress transmitted from the adjacent vertebrae, thereby presenting pathological changes like osteoarthritis (OA), such as cartilage loss, extracellular matrix (ECM) degradation, and secretion of inflammatory factors^7–9. Some studies have demonstrated that excessive mechanical load can accelerate the senescence of chondrocytes in the knee joint and contribute to the pathological progression of OA¹⁰. However, it remains unclear whether cell senescence occurs in the facet joint cartilages during the progression of AIS, and the specific mechanisms underlying this phenomenon have yet to be elucidated.

Additionally, research has shown that AIS patients exhibit systemic bone loss, and the severity of scoliosis is proportional to the extent of bone loss¹¹. It is well established that the impaired differentiation and mineralization ability of osteoblasts can directly result in reduced bone formation. However, the precise mechanism of bone loss in AIS is still unclear at present.

It is worth noting that in terms of anatomical location, cartilage and bone tissue in the spine are adjacent and interlaced, and both exist in the same distorted and intensified mechanical system during AIS progression. Nonetheless, it is still unknown whether there is an interaction between them, as well as whether this interaction can regulate the progression of AIS. Exosomes (Exos), which are small extracellular vesicles with diameters ranging from 30 to 150 nm, are actively secreted by nearly all cell types into the extracellular microenvironment under both physiological and pathological conditions¹². They serve as essential mediators in cellular communication by facilitating the transfer of key proteins¹³. Given this, this study aims to investigate proteins carried by exosomes to elucidate the interactions between spinal cartilage and bone tissue during the progression of AIS and to explore their potential pathological implications.

In this study, we identified an accelerated senescence phenotype in the cartilage of the AIS spine. Mechanistically, we discovered an exosome-mediated positive feedback loop between cartilage and bone tissue that exacerbates two pathological processes: (1) cartilage senescence and ECM degradation, and (2) vertebral bone mass reduction, ultimately driving scoliosis progression. Targeting key exosomal proteins within this regulatory loop presents a promising therapeutic strategy to decelerate scoliosis advancement. This work not only expands the etiological framework of scoliosis progression pathogenesis but also delineates specific molecular determinants crucial for scoliosis modification. Importantly, our findings establish an alternative conceptual foundation for developing non-invasive therapeutic interventions in scoliosis management.

Results

The chondrocytes of AIS intervertebral facet joints presented obvious senescence

To verify whether cellular senescence occurs in the facet joint cartilage during the progression of AIS, six distinct inferior articular process cartilage specimens were harvested from the upper instrumented vertebra (UIV), apex vertebra (AV), and lower instrumented vertebra (LIV) located on both the concave (CC) and convex (CV) sides of the main curve in AIS patients with severe scoliosis undergoing posterior instrumentation and spinal fusion surgery (Fig. 1A). The control specimens were corresponded to AIS UIV and AV. Immunofluorescence staining with COL2A1 was employed to identify the isolated chondrocytes (Fig. 1B). Compared with chondrocytes harvested from the LIV, chondrocytes located at the UIV and AV in AIS exhibited more pronounced cellular senescence characteristics. These include elevated SA-β-Gal (senescence-associated β-galactosidase) activity (Fig. 1C, D), extensive DNA damage marked by high γ-H2AX expression (Fig. 1C, D), a substantial decrease in mitochondrial membrane potential as indicated by an increased JC-1 monomer/aggregate ratio (Fig. 1E, F), a significant rise in reactive oxygen species (ROS) levels (Fig. 1G, H), and heightened secretion of IL-6 (Fig. 1I, J). In addition, Safranin O staining revealed that the staining intensity in the chondrocytes from UIV and AV was significantly weaker compared to that in the LIV, suggesting more severe cartilage damage and extensive ECM degradation (Fig. 1C, Supplementary Fig. 1A). To sum up, the extent of chondrocytic senescence and ECM degradation correlates with their anatomical positioning. Whether on the concave or convex side, the degree of chondrocytic senescence and ECM degradation in the UIV and AV is significantly higher than that in the LIV of the main curve.

Fig. 1. The chondrocytes of AIS intervertebral facet joints presented obvious senescence.

A Anatomical positioning of the inferior articular involved in the study. B Identify the isolated primary chondrocytes by COL2A1. C, D β-Galstaining, Safranin O staining, γ-H2AX immunofluorescence staining, and corresponding statistical analyses. E, F Mitochondrial membrane potential assay by JC-1 fluorescence probe and corresponding statistical analyses. G, H ROS assay by DCFH-DA fluorescence probe and corresponding statistical analyses. I, J IL-6 concentration in chondrocytic supernatant. CCU, the UIV of the concave side, CCA, the AV of the concave side, CCL, the LIV of the concave side, CVU, the UIV of the convex side, CVA, the AV of the convex side, CVL, the LIV of the convex side, CH, chondrocytes, γ-H2AX, phospho-Histone H2AX-S139, CCCP, a positive control for inducing mitochondrial membrane potential depolarization, Rosup, a positive control reagent for inducing ROS. All data were presented as mean ± standard deviation $\left(\overline{x}\pm s\right)$, n = 5. One-way analysis of variance (ANOVA) was used, followed by Tukey’s post hoc test to determine the statistical significance. * p < 0.05, ** p < 0.01, *** p < 0.001; ns, no significance.

Chondrocytic senescence corresponds with reduced osteoblast differentiation and low bone mineralization in vertebrae

Next, Safranin O-Fast Green staining and immunohistochemical analysis for the senescent marker proteins p16 and p21 were performed on the intervertebral facet joint cartilage tissues. The results demonstrated that, compared with cartilage derived from the LIV, there was significant cartilage loss and ECM degradation in the UIV and AV of the main curve (Fig. 2A, Supplementary Fig. 2A). Moreover, the proportion of p16- and p21-positive cells increased markedly, suggesting a prominent senescence-related pathological phenotype in vivo (Fig. 2A, B). To investigate the pathological changes in osteoblasts at corresponding scoliosis positions, we first isolated and cultured primary osteoblasts, which were subsequently identified by immunofluorescence staining with the osteoblast-specific marker Runx2 (Fig. 2C). Alkaline Phosphatase (ALP) and Alizarin Red S (ARS) staining assays were employed to evaluate osteogenic differentiation and bone mineralization, respectively. Our results revealed that the bone formation of osteoblasts in AIS vertebrae was closely related to their anatomical positioning. Specifically, the differentiation and mineralization of osteoblasts derived from the UIV and AV regions of both sides were significantly reduced compared to those from the LIV of the main curve (Fig. 2D, E). Taken together, chondrocytes exhibiting aggravated senescence and ECM degradation correspond with osteoblasts displaying weakened osteogenic differentiation and mineralization. The consistency in the severity of pathological changes in these two cell types hints at a possible interaction between them.

Fig. 2. Chondrocytic senescence corresponds with reduced osteoblast differentiation and low bone mineralization in vertebrae.

A, B Safranin O-Fast Green staining and immunohistochemical analysis for p16 and p21 on the intervertebral facet joint cartilage tissues and corresponding statistical analyses (n = 5). C Identify the isolated primary osteoblasts by RUNX2. D, E ALP, ARS staining and corresponding statistical analyses (n = 3). OB, osteoblasts. All data were presented as mean ± standard deviation $\left(\overline{x}\pm s\right)$. One-way analysis of variance (ANOVA) followed by Tukey’s post hoc test was used to determine the statistical significance. * p < 0.05, ** p < 0.01, *** p < 0.001; ns, no significance. Red arrows indicate staining-positive cells.

The positive feedback regulation exists between AIS intervertebral facet joint chondrocytes and vertebral osteoblasts

To investigate the interaction between AIS intervertebral facet joint chondrocytes and vertebral osteoblasts, we performed a co-culture experiment using conditioned medium (CM). Firstly, the supernatant of intervertebral facet joint chondrocytes, derived from the UIV (CH-CCU-CM or CH-CVU-CM) and LIV (CH-CCL-CM or CH-CVL-CM) of the main curve, was used as conditioned medium for co-culture with control osteoblasts. Subsequently, ALP and ARS staining analyses were performed, along with the detection of expression levels of osteogenesis and bone mineralization markers RUNX2 (Runt-Related Transcription Factor 2) and OCN (Osteocalcin) (Fig. 3A–C, Supplementary Fig. 3A). The results revealed that, regardless of concave or convex side, the supernatant of chondrocytes from the UIV, which exhibited aggravated senescence and ECM degradation, significantly suppressed the osteogenic differentiation and mineralization of control osteoblasts compared to the supernatant of chondrocytes from the LIV.

Fig. 3. The positive feedback regulation exists between AIS intervertebral facet joint chondrocytes and vertebral osteoblasts.

A The supernatant of intervertebral facet joint chondrocytes, derived from the UIV or LIV of the main curve, was used as conditioned medium for co-culture with control osteoblasts, ALP and ARS staining (n = 3). B Corresponding statistical analyses based on (A). C Western blot assay detected the expression level of RUNX2 and OCN (n = 3). D The supernatant of osteoblasts, derived from the UIV or LIV of the main curve, as a conditioned medium for co-culture with control chondrocytes, SA-β-Gal and Safranin O staining (n = 5). E Corresponding statistical analyses based on SA-β-Gal staining in (D). F Western blot assay detected the expression level of p16 and p21 and corresponding statistical analyses (n = 3). G GW4869 (10 μΜ) was used as an inhibitor of exosome secretion. ALP and SA-β-Gal staining indicated that exosomes play an essential role in the process of mutually exacerbated pathological changes between chondrocytes and osteoblasts (n = 3). CM, conditioned medium. All data were presented as mean ± standard deviation $\left(\overline{x}\pm s\right)$. One-way analysis of variance (ANOVA) followed by Tukey’s post hoc test was used to determine the statistical significance. * p < 0.05, ** p < 0.01, *** p < 0.001.

On the contrary, we employed the supernatant of osteoblasts, derived from the UIV (OB-CCU-CM or OB-CVU-CM) and LIV (OB-CCL-CM or OB-CVL-CM) of the main curve, as a conditioned medium for co-culture with control chondrocytes. SA-β-Gal and Safranin O staining analyses were performed, along with the detection of expression levels of senescence markers p16 and p21 (Fig. 3D–F, Supplementary Fig. 3B, C). The results revealed that, regardless of concave or convex side, the supernatant of osteoblasts from the UIV, which exhibited weakened osteogenesis and bone mineralization, significantly accelerated the senescence and ECM degradation of control chondrocytes compared to those from the LIV.

To investigate whether inhibition of exosome secretion could prevent the progression of AIS, we utilized GW4869 (10 μM) as an inhibitor of exosomal release. Subsequent ALP and SA-β-Gal staining demonstrated that suppression of exosome secretion significantly attenuated the mutually exacerbating pathological interactions between chondrocytes and osteoblasts (Fig. 3G, Supplementary Fig. 3D).

To summarize, during the progression of AIS, a pathological positive feedback loop is established, in which the chondrocytes within the intervertebral facet joints and the osteoblasts of the vertebrae reciprocally exacerbate each other’s dysregulation.

Exosomes mediated the positive feedback regulation between chondrocytes and osteoblasts

To investigate the mediators involved in the interaction between chondrocytes and osteoblasts, we successfully isolated exosomes from chondrocytes of the AIS intervertebral facet joint (CH-Exos) and osteoblasts of the vertebrae (OB-Exos). TEM revealed that both types of exosomes exhibited characteristic exosomal morphology (Fig. 4A). NTA indicated that the average diameters of CH-Exos and OB-Exos were approximately 112.5 nm and 118.6 nm (Fig. 4B), with concentrations of 1.2 × 10⁸ particles/ml and 1.5 × 10⁸ particles/ml, respectively. To assess the purity of the extracted exosomes, we assay the exosomal protein content using the BCA Protein Concentration Assay Kit (Beyotime Biotechnology, Shanghai, China). As a result, the protein concentration of CH-Exos and OB-Exos was 1041.75 μg/ml and 1241.5 μg/ml. Subsequently, the ratio of protein content to the number of exosomes was calculated. For CH-Exos, the ratio is 8.68 pg/particle; for OB-Exos, the ratio is 8.28 pg/particle. Western blot analysis confirmed the expression of the exosomal positive markers HSP70 and CD81, as well as the negative marker Calnexin (Fig. 4C). Additionally, confocal microscopy observed that OB-Exos or CH-Exos were effectively internalized by chondrocytes or osteoblasts, respectively (Fig. 4D).

Fig. 4. Exosomes mediated the positive feedback regulation between chondrocytes and osteoblasts.

A TEM was used to observe the exosomal morphology. B NTA was employed to assay the particle size, distribution, and concentration of exosomes. C The exosomal positive markers HSP70 and CD81, as well as the negative marker Calnexin, were detected by western blot. D Confocal microscopy observed that OB-Exos or CH-Exos were effectively internalized by primary chondrocytes or osteoblasts, respectively. E ALP staining demonstrated that the mediator regulating osteogenic differentiation is the chondrocyte exosome (n = 3). F SA-β-Gal and Safranin O staining showed that the osteoblast exosomes mediated the chondrocytic senescence and ECM degradation (n = 5). Exos, exosomes. All data were presented as mean ± standard deviation $\left(\overline{x}\pm s\right)$, and one-way analysis of variance (ANOVA) followed by Tukey’s post hoc test was used to determine the statistical significance. * p < 0.05, ** p < 0.01, *** p < 0.001.

To investigate the effects of CH-Exos on osteoblast differentiation, we incubated control osteoblasts with the following: control chondrocyte supernatant (CH-Con-CM), supernatant from chondrocytes of the UIV (CH-CVU-CM), exosomes derived from chondrocytes of the UIV (CH-CVU-Exos), and exosome-depleted supernatant from chondrocytes of the UIV (CH-CVU-CM (non Exos)). Following incubation, the ALP staining results (Fig. 4E) demonstrated that, compared to the CH-Con-CM, osteogenic differentiation was inhibited when cultured with CH-CVU-CM, and the exosomes isolated from this co-culture conditioned medium exhibited a stronger inhibitory effect. Moreover, the inhibitory effect was absent in the exosome-depleted conditioned medium, indicating that the primary mediator regulating osteogenic differentiation is the chondrocyte exosomes. On the contrary, SA-β-Gal and Safranin O staining (Fig. 4F, Supplementary Fig. 4A) showed that the primary mediator regulating chondrocytic senescence and ECM degradation is the osteoblast exosomes. Taken together, exosomes mediated the positive feedback regulation between chondrocytes and osteoblasts during AIS progression.

The chondrocyte exosomal protein HSP90β suppresses osteogenic differentiation while enhancing cholesterol synthesis and efflux

To further investigate exosomal proteins that regulate the interaction between chondrocytes and osteoblasts during the progression of AIS, we employed exosomes derived from the LIV as a control group, and used 4D/480 label-free proteomics to analyze the enriched proteins in exosomes isolated from the corresponding UIV and AV of the same patient (pathologically aggravated group). This analysis was designed to identify highly circulating exosomal proteins involved in the communication between chondrocytes exhibiting exacerbated senescence and ECM degradation and osteoblasts with impaired osteogenic differentiation and mineralization. These proteins may represent key molecules driving the progression of scoliosis.

Bioinformatics analysis revealed that, on both the concave and convex sides, HSP90β (Heat Shock Protein 90-beta) was significantly upregulated in the CH-Exos derived from the UIV and AV of the main curve (Fig. 5A–C). We confirmed the involvement of HSP90β in AIS-CH-Exos (Supplementary Fig. 5A) and the up-regulation of it in CH-CCU-Exos by western blotting (Fig. 5D). Additionally, treatment of hFOB1.19 cells with human recombinant HSP90β protein (700 ng/mL) resulted in impaired bone formation and mineralization, as assessed by ALP and ARS staining (Fig. 5E). This was further corroborated by the decreased expression levels of RUNX2 and OCN (Fig. 5F). We identified CDK4 (cyclin-dependent kinase 4) as an interacting protein of HSP90β via protein-protein interaction network analysis (Fig. 5G). Subsequently, we demonstrated that the protein level of CDK4 in AIS osteoblasts isolated from the UIV was significantly reduced (Fig. 5H). Furthermore, supplementation of hFOB1.19 cells with recombinant human HSP90β protein markedly decreased the CDK4 level (Fig. 5I). Treatment of hFOB1.19 cells with the CDK4 inhibitor LEE011 (10 μM, Fig. 5J) or transfection with a CDK4 overexpression plasmid (Fig. 5K, Supplementary Fig. 5B) further confirmed that HSP90β suppresses bone formation through the downregulation of CDK4.

Fig. 5. The chondrocyte exosomal protein HSP90β suppresses osteogenic differentiation while enhancing cholesterol synthesis and efflux.

A–C The chondrocyte exosomal proteins were analyzed by 4D/480 label-free proteomics. D Western blotting assay confirmed the upregulation of HSP90β in CH-CCU-Exos compared to CH-CCL-Exos. E Human recombinant HSP90β protein inhibited bone formation and mineralization, assessed by ALP and ARS staining. F HSP90β decreased the expression levels of RUNX2 and OCN. G Protein–protein interaction network analysis predicted the interaction between HSP90β and CDK4. H CDK4 in AIS osteoblasts isolated from the UIV was significantly reduced compared to that from the LIV. I HSP90β markedly decreased the CDK4 level in hFOB1.19 cells. J CDK4 inhibitor LEE011 suppressed bone formation. K CDK4 overexpression rescued the HSP90β-induced bone loss. L The cholesterol levels in the osteoblasts from AIS. M The levels of HMGCR, APOA1, and ABCA1 in the osteoblasts from AIS. N Lova rescued the HSP90β-induced-HMGCR and bone loss in hFOB1.19 cells. O, P HSP90β enhanced the expression of both APOA1 and ABCA1in hFOB1.19 cells. Lova, lovastatin. All data were presented as mean ± standard deviation $\left(\overline{x}\pm s\right)$, n = 3. Two-tailed unpaired Student’s t-tests (E, F, H–J, M, P) and one-way analysis of variance (ANOVA) followed by Tukey’s post hoc test (K, N) were used to determine the statistical significance. * p < 0.05, ** p < 0.01, *** p < 0.001.

Research has shown that HSP90β can enhance cholesterol synthesis in cells¹⁴. Consequently, upon quantifying the concentrations of total cholesterol, free cholesterol, and cholesterol ester, we observed significantly higher cholesterol levels in the AIS osteoblasts at the UIV compared to those at the LIV (Fig. 5L, Supplementary Fig. 5C), which was consistent with increased expression of cholesterol biosynthesis marker HMGCR (3-Hydroxy-3-Methyl-Glutaryl-CoA Reductase) (Fig. 5M). Since elevated intracellular cholesterol concentrations are known to suppress osteoblast bone formation¹⁵, we treated the hFOB1.19 cells with lovastatin (Lova, Supplementary Fig. 5D), an HMGCR inhibitor. The results demonstrated that HSP90β inhibits osteoblast differentiation and mineralization by promoting cholesterol biosynthesis (Fig. 5N).

Cholesterol homeostasis is maintained through the regulation of cholesterol synthesis, influx, efflux, and metabolism¹⁶. Notably, exosomes derived from AIS chondrocytes are also enriched in APOA1 (Apolipoprotein A1, Fig. 5C), which interacts with ABCA1 (ATP-Binding Cassette Transporter A1) to facilitate intracellular cholesterol efflux¹⁷. Our experimental results demonstrated that the expression levels of APOA1 and ABCA1 were significantly up-regulated in primary AIS osteoblasts isolated from the UIV compared to those at the LIV (Fig. 5M). Furthermore, HSP90β was found to enhance the expression of both APOA1 and ABCA1 in hFOB1.19 cells (Fig. 5O, P). Therefore, it can be inferred that AIS osteoblasts export intracellularly synthesized cholesterol into the extracellular microenvironment via the HSP90β-APOA1/ABCA1 pathway.

To summarize, chondrocyte-derived exosomal protein HSP90β inhibits osteoblast differentiation and mineralization by down-regulating CDK4 expression and promoting cholesterol biosynthesis, while facilitating cholesterol efflux into the extracellular microenvironment via the HSP90β-APOA1/ABCA1 pathway.

The osteoblast exosomal protein APOB promotes cholesterol influx, and the increased cholesterol further accelerates chondrocytic senescence and ECM degradation

To investigate whether cholesterol in the extracellular microenvironment enters chondrocytes at the same anatomical location, we measured the concentrations of total cholesterol, free cholesterol, cholesterol esters, and the expression of HMGCR in intervertebral facet chondrocytes of AIS patients. The results demonstrated that with senescence and ECM degradation, the cholesterol concentration in chondrocytes increased while their synthesis decreased (Fig. 6A–C, Supplementary Fig. 6A). This suggests that the elevated cholesterol levels in chondrocytes are primarily derived from the influx from the extracellular microenvironment. Furthermore, label-free proteomic analysis (4D/480) revealed that AIS osteoblast-derived exosomes transport an amount of APOB (Apolipoprotein B-100), a critical protein involved in cholesterol influx¹⁸, to chondrocytes (Fig. 6D, E). We confirmed the involvement of APOB in AIS-OB-Exos (Supplementary Fig. 6B) and the upregulation of it in OB-CCU-Exos by western blotting (Fig. 6F). Concurrently, the expression level of OLR1 (Lectin-type Oxidized Low-density Lipoprotein Receptor 1), another marker protein associated with cholesterol influx¹⁹, was significantly up-regulated in AIS chondrocytes isolated from the UIV compared to those at the LIV (Fig. 6B). These findings confirm that the cholesterol in the microenvironment inflows into chondrocytes promoted by osteoblast exosomal protein APOB.

Fig. 6. The osteoblast exosomal protein APOB promotes cholesterol influx, and the increased cholesterol further accelerates chondrocytic senescence and ECM degradation.

A Concentrations of total cholesterol of AIS chondrocytes. B, C The expression levels of HMGCR, APOB, and OLR1 in intervertebral facet chondrocytes of AIS patients. D, E The enriched exosomal proteins in AIS osteoblasts were analyzed by 4D/480 label-free proteomics. F Western blotting assay confirmed the upregulation of APOB in OB-CCU-Exos. G–I mTORC1 activity, cholesterol metabolic axis CH25H-CYP7B1-RORα level, and the expression levels of corresponding senescence and catabolism markers in AIS chondrocytes. J, K Rapa rescued cholesterol-induced mouse primary chondrocytic senescence by suppressing mTORC1 activity. L si-CYP7B1 rescued cholesterol-induced ECM degradation through inhibiting the CH25H-CYP7B1-RORα axis. Rapa, rapamycin, Chol, cholesterol. All data were presented as mean ± standard deviation $\left(\overline{x}\pm s\right)$, n = 3. Two-tailed unpaired Student’s t-tests (C, H, I) and one-way analysis of variance (ANOVA) followed by Tukey’s post hoc test (K, L) were used to determine the statistical significance. * p < 0.05, ** p < 0.01, *** p < 0.001; ns, no significance.

Previous studies suggest that a high concentration of cholesterol can contribute to the catabolism of OA cartilage²⁰. This includes accelerating cellular senescence by sustaining mTORC1 (Mechanistic Target of Rapamycin Complex 1) activity²¹, as well as facilitating ECM degradation via the cholesterol metabolic pathway CH25H (Cholesterol 25-Hydroxylase)— CYP7B1 (25-Hydroxycholesterol 7α-Hydroxylase)—RORα (Retinoic Acid-related Orphan Receptor Alpha) axis¹⁹. Consequently, we firstly validated the two pathways and their corresponding senescence and catabolism levels in AIS chondrocytes (Fig. 6G–I). Subsequently, by treating with rapamycin (Rapa) to suppress mTORC1 activity and inhibiting the cholesterol metabolic axis CH25H-CYP7B1-RORα using si-CYP7B1, we demonstrated that elevated cholesterol levels in mouse primary chondrocytes accelerate cellular senescence by maintaining mTORC1 activity, while promoting ECM degradation through the CH25H-CYP7B1-RORα axis (Fig. 6J–L, Supplementary Fig. 6C, D).

Taken together, our findings confirm that the cholesterol in the microenvironment inflows into chondrocytes mediated by osteoblast exosomal protein APOB, and the increased cholesterol further accelerates chondrocytic senescence and ECM degradation.

Targeting the key exosomal proteins HSP90β or APOB can block the positive feedback regulation loop between chondrocyte and osteoblast

Our studies have demonstrated an exosome-mediated positive feedback regulation loop characterized by high cholesterol flow between chondrocyte and osteoblast during the progression of AIS (Fig. 7A). Under abnormal stress, chondrocytes exhibit senescence and ECM degradation while secreting HSP90β-rich exosomes to nearby osteoblasts. HSP90β promotes the biosynthesis of cholesterol to weaken osteoblast osteogenic differentiation and mineralization, and upregulates APOA1/ABCA1 to facilitate cholesterol efflux into the extracellular microenvironment. Osteoblasts with abnormal osteogenic differentiation and mineralization also transport APOB into adjacent chondrocytes through exosomes. APOB facilitates the influx of extracellular microenvironment cholesterol, and the resultant elevation in cholesterol levels further exacerbates the senescence and ECM degradation of chondrocytes. At this point, a positive feedback loop is established, exacerbating the pathological deterioration of chondrocytes and osteoblasts.

Fig. 7. Targeting the key exosomal proteins HSP90β or APOB can block the positive feedback regulation loop between chondrocytes and osteoblasts.

A Schematic diagram of the exosome-mediated positive feedback regulation loop characterized by high cholesterol flow between chondrocytes and osteoblasts during the progression of AIS. B The correlation analysis between the Cobb angle and blood lipid parameters of AIS patients, including total cholesterol, APOA, APOB, HDL-C, and LDL-C (n = 38). C Effect of different concentrations of Cor on cell viability (n = 5). D, E Cor effectively reversed the reduced bone formation induced by CH-CCU-Exos in osteoblasts from AIS CCU (n = 3). F Effect of different concentrations of Ava on cell viability (n = 5). G, H Ava successfully mitigated senescence and ECM degradation caused by OB-CCU-Exos in AIS chondrocytes (n = 3). Ava, Avasimibe, Cor, Corylin. All data were presented as mean ± standard deviation $\left(\overline{x}\pm s\right)$. Pearson correlation analysis (B) and one-way analysis of variance (ANOVA) followed by Tukey’s post hoc test (C, E, F, H) were used to determine the statistical significance. * p < 0.05, ** p < 0.01, *** p < 0.001; ns, no significance.

To determine whether the high cholesterol flow we identified is systemic or region-specific, we examined the correlation between the Cobb angle and blood lipid parameters in AIS patients, including total cholesterol, APOA, APOB, high-density lipoprotein cholesterol (HDL-C), and low-density lipoprotein cholesterol (LDL-C) (Fig. 7B). The results indicated no significant correlation between the Cobb angle and blood lipid parameters in patients with AIS. These data support that systemic cholesterol homeostasis does not directly associate with the spinal curvature severity, the evaluated cholesterol conditions were confined to the microenvironment of spinal cartilage and bone tissue. Therefore, the exosome-mediated cholesterol dysregulation is proposed as a primary driver of spatially patterned disease progression along the scoliotic spine.

To demonstrate that the exosomal proteins HSP90β and APOB serve as critical targets for disrupting the high-cholesterol microenvironment, we employed Corylin (Cor) to inhibit the activity of HSP90β¹⁴. Western blot analysis revealed that Cor effectively reversed the reduced bone formation induced by CH-CCU-Exos (Fig. 7C–E). Meanwhile, we utilized Avasimibe (Ava) to enhance intracellular degradation of APOB²². Western blot results indicated that Ava successfully mitigated AIS chondrocytic senescence and ECM degradation caused by OB-CCU-Exos (Fig. 7F–H). Collectively, these findings suggest that the key exosomal proteins HSP90β and APOB are pivotal targets for interrupting the positive feedback loop between chondrocyte and osteoblast.

To sum up, during the progression of AIS, a positive feedback regulation mediated by exosomes and characterized by high cholesterol flux exists between chondrocytes and osteoblasts. Exosomal proteins HSP90β and APOB are pivotal targets for interrupting this process in vitro.

Targeting the key exosomal proteins HSP90β and/or APOB delays the progression of scoliosis in vivo

Here, we constructed the recombinant adeno-associated virus rAAV9-RUNX2-HMGCR to reconstruct the high-cholesterol flow between cartilage and bone tissues, combined with bipedal standing to establish an experimental scoliosis mouse model. Mice were treated with Ava and/or Cor through intragastric administration (Fig. 8A). At the end of the experiment, we observed that specific overexpression of HMGCR in osteoblasts (Fig. 8B) significantly accelerated the typical thoracic scoliosis deformity (Cobb angle ≥ 10°). Furthermore, treatment with Ava and/or Cor rescued the increased Cobb angle without systemic toxicity (Fig. 8C, D, Supplementary Fig. 7A). In addition, previous studies have demonstrated that HSP90β could promote osteoclastogenesis¹⁴. In our study, the vertebrae of scoliosis mice exhibited not only suppressed osteoblast differentiation (Fig. 8E) but also enhanced osteoclast activation (Fig. 8F), ultimately leading to bone loss. That is consistent with the systemic osteopenia observed in AIS patients. In addition, the intervertebral discs at the UIV and AV of scoliosis also showed senescence and ECM degradation (Fig. 8G). Ava or/and Cor therapy reversed the imbalance between osteogenesis and osteoclastogenesis, while preventing intervertebral facet joint cartilage senescence and ECM degradation (Fig. 8E–G). Taken together, our results demonstrated the critical role of the exosome-mediated positive feedback regulation loop between chondrocyte and osteoblast in the progression of scoliosis, as well as the therapeutic effects of targeting exosomal proteins HSP90β and APOB to block the high cholesterol flow in vivo.

Fig. 8. Targeting the key exosomal proteins HSP90β and/or APOB delays the progression of scoliosis in vivo.

A Study overview of establishing an experimental scoliosis mouse model and treatment with Ava and/or Cor. B The expression of HMGCR in the vertebrae of scoliosis mice. C, D Treatment with Ava and/or Cor rescued the increased Cobb angle of scoliosis in mice. E Treatment with Ava and/or Cor restored the suppressed osteoblast differentiation of scoliosis mice. F Treatment with Ava and/or Cory reduced the enhanced osteoclast activation of scoliosis mice. G Treatment with Ava and/or Cory relieved the aggravated senescence and ECM degradation of intervertebral facet joint cartilage of scoliosis mice. All data were presented as mean ± standard deviation $\left(\overline{x}\pm s\right)$, n = 5. Two-way analysis of variance (ANOVA) (D) and one-way analysis of variance (ANOVA) followed by Tukey’s post hoc test (E–G) were used to determine the statistical significance. * p < 0.05, ** p < 0.01, *** p < 0.001; ns, no significance. Red arrows indicate staining-positive cells.

Discussion

This study confirmed the senescence phenotype of AIS chondrocytes. The exosome-mediated positive feedback regulation loop between spinal cartilage and bone tissue, which was characterized by high cholesterol flow, can exacerbate changes in cartilage and osteopenia of vertebrae, thereby promoting the progression of scoliosis. Targeting key exosomal proteins HSP90β or APOB within this positive feedback loop represents an effective strategy to retard the progression of scoliosis. This research identifies the critical factors driving the aggravation of scoliosis, thereby extending the etiological theory of AIS progression. Furthermore, it also provides an alternative perspective for non-surgical treatment options for scoliosis in clinical practice.

In our study, intervertebral facet joint cartilage senescence and ECM degradation increased with scoliosis progression in mice (Fig. 8G). This phenomenon is consistent with the observations made in the intervertebral facet joint cartilage of patients with AIS. Notably, in addition to intervertebral facet joint cartilage, the spine also contains intervertebral discs, which transfer 70–80% of the spinal load²³, as well as vertebral growth plate cartilage. Studies have shown that facet joint and intervertebral disc cartilage have a high concordance in the pathological grades of degeneration²⁴. Therefore, we observed cellular senescence and ECM degradation in the intervertebral disc and vertebral growth plate cartilage from the UIV and AV of scoliosis of our scoliosis mice; the pathological alterations were found to be consistent with those observed in the intervertebral facet joints (Supplementary Fig. 8A, B). Based on these data, we presume that exosome-mediated positive feedback regulation characterized by high cholesterol flow is present between all spinal cartilage and bone tissues.

Currently, the research on AIS predominantly focuses on its etiology of initiation, while fewer studies address the causes contributing to the progressive worsening of AIS⁴. In this study, by integrating cellular senescence, exosome-mediated communication, and metabolic regulation, a pathological framework of “cholesterol flow between cartilage and bone tissues” was proposed. This framework elucidates the shared progression mechanism underlying cartilage degeneration, bone mass loss, and the resultant scoliosis, representing a theoretical foundation for understanding scoliosis progression.

Clinically, abnormal cholesterol biosynthesis may present with the symptom of scoliosis^25,26. A missense variant in SLC39A8 (p.Ala391Thr, rs13107325) has been associated with severe AIS²⁷. SLC39A8 has previously been linked to cholesterol synthesis and circulation^28–35, suggesting the importance of cholesterol in AIS. In addition, many clinical studies have indicated an association between biomechanical factors and the incidence and pathological progression of AIS⁴. Changes in cholesterol level affect both mechanosensitivity and basal channel function; for example, cholesterol regulates mechanosensitive piezo channels by altering their microenvironment³⁶. An in-frame mutation of piezo1 in zebrafish led to fully penetrant juvenile-onset scoliosis, resembling non-congenital scoliosis symptoms in humans³⁷. Piezo2 has shown a connection with clinical proprioceptive tests in subjects with idiopathic scoliosis³⁸. Especially, in progressive aggravation of AIS, stress upregulates piezo1 in vertebral growth plate chondrocytes and accelerates pathological scoliosis^4,39,40. To investigate the piezo channels in our scoliosis model mice, we performed immunohistochemical staining for Piezo1, a major skeletal mechanosensor⁴¹, on the facet joint, intervertebral discs, and vertebral growth plate cartilage at the UIV and AV of scoliosis. As shown in Supplementary Fig. 8C, the expression level of Piezo1 in the scoliosis group was significantly higher than that in the control group. Meanwhile, when the high cholesterol flow between spinal cartilage and bone tissue was inhibited by Ava or Cor treatment, the expression level of Piezo1 was rescued. These results indicate that mechanical loading undergoes substantial alterations during scoliosis progression in the mouse model. In summary, the high cholesterol flow between spinal cartilage and bone tissue associates with not only spinal cholesterol metabolism but also mechanical loading, playing an essential role in the progression of scoliosis.

A universally applicable animal model for AIS research is currently lacking^42–44. Our research shows that exosome-mediated cholesterol synthesis, efflux, and influx were involved in the interaction between cartilage and bone tissue. If consistent cholesterol flow can be replicated in mice with a result in Cobb angle of ≥10°, these mice could subsequently serve as experimental scoliosis model mice for future studies. To realize this concept, we constructed the recombinant adeno-associated virus rAAV9-RUNX2-HMGCR. rAAV9 infects osteoblasts effectively^45,46, while the RUNX2 promoter ensures HMGCR expression in osteoblasts. In addition, the bipedal standing mouse model utilized in this study mimics the vertical stress pattern observed in humans while inducing degeneration of both intervertebral disc and intervertebral facet joint cartilage^47–49. In our study, after the injection of rAAV9-RUNX2-HMGCR in C57BL/6 J mice, combined with bipedal standing, a typical thoracic scoliosis deformity was observed, with a Cobb angle of ≥10°. The development of this model mouse is based on etiological interpretations derived from the analysis of clinical samples, thereby ensuring that in vivo drug treatment studies maintain clinical relevance.

A remaining concern is that, in the human data, pathological changes are described regionally, suggesting spatial specificity along the curve. In contrast, the animal model is based on systemic manipulation of cholesterol metabolism and exosome signaling. So, how is such a systemic perturbation proposed to generate regionally distinct pathological outcomes? In our conceptual framework, we identified cholesterol metabolism dysregulation as a central determinant of scoliosis progression. Given that biomechanical loading is inherently intensified during curve progression, the observed cholesterol flow between spinal cartilage and bone tissue cannot be regarded as an isolated phenomenon; rather, it is coupled to altered mechanical stress. Indeed, our investigation originated from this biomechanical premise: with the progressive increase in segmental spinal stress, we hypothesize that intervertebral facet joint cartilage undergoes stress-associated senescence. This hypothesis was subsequently supported by experimental evidence of cartilage cellular senescence and ECM degradation. Consequently, the cholesterol metabolic perturbation described herein is interpreted as a stress-responsive pathological adaptation. This interpretation aligns with our murine findings: systemic modulation of cholesterol metabolism induced scoliotic deformity exclusively in the spine under the double hindlimb standing model, which is a paradigm explicitly designed to amplify axial mechanical load on the vertebral column. Thus, our data demonstrated that the positive feedback regulation loop mediated by exosomes between cartilage and bone tissues and characterized by high cholesterol flow serves as a contributing or permissive factor that exacerbates scoliosis severity under specific mechanical conditions during scoliosis progression.

We must acknowledge the limitations of our study, including that the exosome concentrations used may not reflect the physiological joint microenvironment, and no in vivo data directly demonstrate that these cargos are altered in AIS or required for scoliosis development. Furthermore, the potential involvement of other upregulated or downregulated exosomal proteins in the pathogenesis and progression of scoliosis through alternative mechanisms was not addressed in this study, which indicates potential future research directions for this work.

In conclusion, our study demonstrated a senescence phenotype in AIS spinal cartilage. The exosome-mediated positive feedback regulation loop, characterized by high cholesterol flow between spinal cartilage and bone tissue, plays an essential role in the progression of scoliosis. Additionally, the double-stimulated mice used in this research can serve as experimental non-invasive models for scoliosis. Finally, exosomal proteins HSP90β or APOB have been identified as promising therapeutic targets for the clinical management of mild-to-moderate scoliosis progression.

Methods

Clinical tissue specimens

Facet joint cartilage and vertebral tissue specimens were obtained from 19 patients with severe scoliosis (mean age: 14.0 ± 0.5 years; male: female = 3:16) who underwent posterior instrumentation and spinal fusion surgery at the Spine Surgery Center, The First Affiliated Hospital of Hainan Medical University, China. Control specimens (n = 5, age 27.0 ± 2.7, 4 males and 1 female) were collected from the bilateral articular process resected due to trauma-induced fractures requiring posterior spinal fusion surgery. All specimens were then divided either for histological staining (in 4% paraformaldehyde) or for primary cell isolation (in DMEM/F12 medium supplemented with 1% penicillin–streptomycin). The specimen collection process was an integral part of the surgical procedure and did not affect the treatment or prognosis of the patients. This study protocol was approved by the Ethics Committee of Hainan Medical University (HYLL-2024-090). Informed consent was obtained from all participants and their legal guardians.

Cell preparation

Human facet joint chondrocytes: after washing in germ free phosphate-buffered saline (PBS, 0.01 M, Servicebio, Wuhan, China), intervertebral facet joint cartilage tissues were cut into fragments about 1 mm³ in size and sequentially digested in 0.25% trypsin (37 °C, 30 min; Servicebio, Wuhan, China) followed by collagenase P/DMEM-F12 (1:1, 37 °C, 4 h; Roche, Basel, Switzerland). The digested suspension was then filtered through 40 μm cell strainers (Biosharp, Beijing, China) and subjected to centrifugation at 1000 rpm for 5 min. After resuspending in complete DMEM/F12 (Servicebio, Wuhan, China) medium involving 10% fetal bovine serum (FBS; Gibco, NZ, Australia), the cell pellets were placed into 6-well microplates for culture.

Human vertebral osteoblasts: vertebral tissues were rinsed with PBS, mechanically dissociated into fragments, and directly resuspended in DMEM/F12 medium containing 0.5 ng/mL D-valine (Sigma-Aldrich, MO, USA) and 10% FBS. After transferring to culture dishes, the cell suspension was maintained for 1 week under 37 °C and 5% CO₂ conditions.

Primary mouse chondrocytes: rib cartilage from neonatal mice (1–7 days old) was dissected, and the membranous tissues were meticulously removed. The cartilage was minced and digested with collagenase P/DMEM-F12 (1:100, 37 °C, 4–6 h). After filtering through 40 μm cell strainers, the suspension was subjected to centrifuging and PBS washing. Thereafter, DMEM-F12 complete medium was used to culture the isolated cells.

Human osteoblast hFOB1.19: The hFOB1.19 cell line was purchased from the Cell Bank, Type Culture Collection, Chinese Academy of Sciences (CBTCCCAS) and then cultured in DMEM/F12 complete medium supplemented with 0.3 mg/mL G418 (Beyotime Biotechnology, Shanghai, China) at 33 °C.

Isolation and characterization of exosomes

Ultracentrifugation was performed to isolate exosomes from the cell supernatant of intervertebral facet joints, chondrocytes, and vertebral osteoblasts. Briefly, conditioned media from cultures maintained in exosome-free FBS (VivaCell) were sequentially centrifuged at 4 °C: first at 800×g for 10min and then at 2000×g for 30min. The clarified supernatant was passed through a 0.22-μm filter and subjected to ultracentrifugation at 120,000×g for 2 h (Optima XPN-100, Type 70 Ti rotor; Beckman Coulter). The exosome precipitate was washed in a large volume of PBS, recovered by repeated ultracentrifugation, and finally resuspended in PBS and stored at −80 °C. The morphological assessment of exosomes was conducted via transmission electron microscopy (TEM). Imaging was performed using a Hitachi HT-7800 TEM at 100 kV. Nanoparticle tracking analysis (NTA) was used to detect the average diameters and concentrations of exosomes. In addition, HSP70 and CD81 were employed as exosomal positive markers and Calnexin as a negative marker. To verify the exosomal internalization by primary chondrocytes or osteoblasts, labeling of exosomes was implemented as per the protocol of the PKH67 Green Fluorescent Cell Linker Kit (Sigma, USA). Fluorescence microscopy (Olympus IX71, Tokyo, Japan) imaging of the plasma membrane was accomplished using Biotium’s CF®-labeled wheat germ agglutinin (CF®555 WGA).

Proteomics analysis

A 4D/480 label-free exosome proteomic analysis was conducted by Genechem (Shanghai, China). Samples (n = 4) were analyzed using a timsTOF Pro mass spectrometer (Bruker, Bremen, Germany) operated in PASEF mode with data-dependent acquisition. The raw mass spectrometry data were processed using MaxQuant software (version 1.6.17.0) and searched against a database (determined by uniprot_homo_20231008_20427_9606_swiss_prot). A global false discovery rate (FDR) threshold of 0.01 was applied at both the peptide and protein levels. Protein abundance was quantified using normalized spectral protein intensity (LFQ intensity). Differentially expressed proteins were defined as those showing a fold change >2 or <0.5 with a p-value < 0.05 (Student’s t-test). The mass spectrometry proteomics data have been deposited in the ProteomeXchange Consortium via the PRIDE repository with the dataset identifier PXD070507.

Cholesterol quantification assay

After seeding 3 × 10⁴ cells into each well of 12-well microplates, they were cultured for 48 h at 37 °C with 5% CO₂. Total and free cholesterol levels were assessed following the instructions of Amplex® Red Cholesterol Assay Kit (Beyotime, Shanghai, China).

Transfected with an overexpression plasmid or siRNA

For CDK4 overexpression, the Lipofectamine 2000 reagent (Invitrogen, Carlsbad, CA, USA) was employed to transfect the CDK4 coding sequence (CDS)-containing plasmid to hFOB1.19 cells. The plasmid overexpressing CDK4 was constructed by Genechem (Shanghai, China). For CYP7B1 knockdown, the siRNA-CYP7B1 (sense: 5′-UUCUUCGAACGUGUCACGUTT-3′, antisense: 5′-ACGUGACACGUUCGGAGAATT-3′) was used. The transfection efficiency and target gene expression were validated through a Western blotting assay.

Establishment of Experimental Scoliosis Mouse Model

We have complied with all relevant ethical regulations for animal use. C57BL/6 J Mice (total number is 30) were purchased from Tianqin Biotechnology Co., LTD., Changsha, Hunan Province, China, and housed in SPF-rated environments (22 ± 1 °C, 55 ± 5% humidity). Two-week-old mice (half male and half female, total number is 10) were intraperitoneally injected with the recombinant adeno-associated virus rAAV9-Runx2-HMGCR (7.5 × 10¹³ vg/kg; n = 5; GenePharma, Suzhou, China) or control virus (rAAV9-Runx2-NC, 7.5 × 10¹³ vg/kg; n = 5). HMGCR (mouse) gene ID: 15357. Two weeks post-injection (at 4 weeks of age), mice were subjected to bipedal aquatic standing for 8 h/day to simulate vertical loading, continuing until 16 weeks of age. Spinal curvature progression was monitored every 2 weeks via X-ray imaging, and Cobb angles were quantified using Surgimap software. For each experimental group, all mice have been included in the analysis. Since the progression of AIS studied in this experiment does not fall under the category of an acute process, the order of treatments and measurements does not affect the experimental results.

Treatment with Ava and/or Cor in scoliosis model mice

To evaluate the therapeutic effects of targeting exosomal proteins APOB or/and HSP90β on delaying scoliosis progression, rAAV9-Runx2-HMGCR-induced scoliosis mice were randomized into the following four groups (n = 5/group, total number is 20): HMGCR overexpression group (Con); HMGCR+Ava group: Avasimibe (15 mg/kg; Beyotime Biotechnology, Shanghai, China); HMGCR+Cor group: Corylin (30 mg/kg; MedChemExpress, NJ, USA); HMGCR+Cor+Ava group: Combination therapy (30 mg/kg Corylin + 15 mg/kg Avasimibe). Mice were intragastrically administered with 200 μL of saline, Ava, or Cor every 2 days while subjected to bipedal aquatic standing conditions. Cobb angles were serially measured every 2 weeks until the experimental endpoint (16 weeks of age). At the end of the experiment, all animals were humanely euthanized by an overdose of chloral hydrate at a dose of 24 mL/kg via intraperitoneal injection. Cessation of respiration and heartbeat was monitored for at least 10 min to ensure complete and irreversible death. Spinal tissues were collected for histopathological analysis, while systemic toxicity was assessed by hematoxylin and eosin (H&E) staining of the organs (lung, heart, kidney, liver, and spleen). For each experimental group, all mice have been included in the analysis.

X-ray imaging

After weighing experimental mice, 5% chloral hydrate (Macklin Biochemical, Shanghai, China) was intraperitoneally administered at a dose of 8 mL/kg body weight for anesthesia. The anesthesia depth was confirmed by a stable respiratory rhythm and the absence of the pedal withdrawal reflex. Mice were secured in a prone position on a foam platform using medical adhesive tape, with the head, neck, trunk, and tail aligned along the longitudinal axis perpendicular to the sacral horizontal plane to standardize spinal anatomical positioning. Posteroanterior radiographs of the entire spine were acquired using a calibrated digital X-ray system with standardized parameters. Immediately post-imaging, mice were transferred to a warm environment until full consciousness was regained. Radiographic images were analyzed using Surgimap software for Cobb angle quantification, with three independent blinded measurements performed to ensure reproducibility.

Statistics and reproducibility

The entire experimental data are described as means ± standard deviations $\left(\overline{x}\pm s\right)$. Prism 9.0 (GraphPad, San Diego, CA, USA) was applied for statistically analyzing these data. Two independent groups were compared by an unpaired Student’s t-test for statistical significance evaluation (two-sided). When comparing three or more groups, Bartlett’s test for homogeneity of variances and subsequent one-way analysis of variance (ANOVA) were conducted. Two-way ANOVA was used when studying the effects of two categorical factors on continuous dependent variables (Fig. 8D). Post-hoc Tukey’s test was implemented when ANOVA revealed significant discrepancies (*p < 0.05). Pearson correlation analysis was used to quantify the strength and direction of the linear relationship between two continuous variables. All experiments were replicated independently a minimum of thrice, with representative data presented. Statistical significance thresholds were delineated as follows: * p < 0.05, ** p < 0.01, *** p < 0.001; ns no significance.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

Min Zuo, Haixia Xu, and Yuying Yang: Investigation, Validation, Data curation, Formal analysis. Wenbo Wang, Jianghong Zhou, and Guojun Li: Validation, Data curation, Resources. Jing Zhao: Validation, Data curation. Rangru Liu: Writing—original draft. Huanxiong Chen: Resources, Funding acquisition. Hua Wang: Conceptualization, Writing-review & editing, Supervision, Project administration, Funding acquisition.

Peer review

Peer review information

Communications Biology thanks Po-Chih Shen and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Primary Handling Editor: Ophelia Bu. A peer review file is available.

Data availability

The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the iProX partner repository (accession: PXD070507, https://www.iprox.cn/page/PSV023.html;?url=177280350316<div class="pi">show [QJ]3kMce, Password: HqjF). The newly generated plasmid ID number is GV657 from GENECHEM. The source data underlying the graphs presented in the main figures can be obtained from Supplementary Data 1. Uncropped and unedited western blot images can be obtained from Supplementary information file (Supplementary Figs. 9–33). All other data are available from the corresponding author on reasonable request.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s42003-026-09960-w.