SMN deficiency inhibits endochondral ossification via promoting TRAF6-induced ubiquitination degradation of YBX1 in spinal muscular atrophy

GO enrichment analysis of the marker genes for Cluster 0 and Cluster 1 revealed enrichment for cellular homeostasis (Fig. S1a) and vascular development (Fig. S1b), respectively. Moreover, marker genes of SSPC^{Ptn::Palmd::Smn1} showed strong correlation with extracellular matrix organization, actin cytoskeleton, mesenchymal cell differentiation, and ossification, suggesting that SMN may be important for the differentiation of hypertrophic chondrocytes into bone (Fig. 1g). To investigate how SMN regulates endochondral ossification, we calculated the Spearman correlation coefficients (ρ) and P-values between Smn1 and other 14 715 genes. We identified 549 genes with ρ value greater than 0.3, P-value less than 0.05, and positive cell count exceeding 50. GO enrichment analysis showed that these genes were involved in autophagy, DNA repair, protein localization, proteasome-mediated ubiquitin-dependent protein catabolic process, and RNA splicing (Fig. 1h). Thus, SMN protein may influence the progression of endochondral ossification mediated by hypertrophic chondrocytes, potentially through its involvement in various biological processes, such as ubiquitin-dependent protein degradation and RNA splicing.

Consistent with previous study,²⁴ we utilized a severe SMA mouse model with a mean lifespan of 10 days (Fig. S2a). Considering that the first symptom appears at P5 with impaired motor function,¹⁹ we selected three time points: P0, P4, and P7, representing neonatal, pre-symptomatic, and symptomatic stages, respectively. No significant difference was observed in weight and body size of SMA mice compared to the Het group from P0 until P4 (Fig. S2b, c). Alcian blue and alizarin red double staining revealed the reduced length of body, humerus, femur, and tibia in SMA mice starting from P4 (Fig. 1i–l). Longitudinal limb bone development is driven by endochondral ossification controlled by the growth plates, which are divided into the resting zone (RZ), proliferative zone (PZ), and HZ based on morphology and function.²⁵ Immunofluorescence staining revealed widespread expression of SMN protein in growth plate cartilage, with particularly strong signals in HZ (Fig. 1m). To validate the reduction of SMN protein, we performed Western blot analysis and confirmed a significant decrease in SMN expression (~50%) in SMA growth plate cartilage across all three timepoints (Fig. S2d). Together with the transcriptional findings, these results support the notion that SMN is essential for growth plate–mediated endochondral ossification. SMN deficiency is therefore likely to impair bone development at the pre-symptomatic stage.

We then investigated whether restoring SMN immediately after birth could rescue defects in endochondral ossification and bone development. In line with previous work, we utilized ASO10-29, an MOE-modified antisense oligonucleotide that promotes exon 7 inclusion of SMN2 gene to increase SMN levels.²⁶ ASO10-29 was administrated to newborn pups at dose of 90 mg/kg, twice between P0 and P1 (Fig. S4a), and significantly increased SMN protein in SMA femur on P7 (Fig. S4b). Increased SMN rescued bone dysplasia in SMA mice according to the alcian blue and alizarin red double staining result (Fig. S4c). After ASO10-29 administration, both percentage and height of HZ in SMA mice were restored to levels comparable to those in Het mice (Fig. 2b, c). Moreover, the proportion of RZ area, which was reduced in untreated SMA mice at P7, was markedly increased following ASO10-29 treatment (Fig. S4d). The PZ area proportion did not exhibit any significant differences between the Het and SMA groups, regardless of whether it was at P4, P7, or after ASO treatment (Fig. S4e).

Safranin O-Fast Green (SF) staining revealed that cell size of hypertrophic chondrocytes increased and cell density decreased in HZ of Het mice from P4 to P7 (Fig. 2d–f), which is a natural phenomenon during endochondral ossification.²⁷ However, the hypertrophic chondrocytes in HZ of SMA mice exhibited smaller size and higher density compared to the Het group, a condition that was recovered by ASO10-29 treatment, indicating an increase of immature hypertrophic chondrocytes. (Fig. 2d–f). At P4 and P7, Von Kossa (VK) staining showed reduced calcification in the metaphysis, while more calcium deposition occurred in the HZ of SMA mice (Fig. 2g–i), indicating the inhibited transition from HZ to ossification zone. ASO10-29 reduced the calcification rate in HZ and increased it in metaphysis (Fig. 2g–i). Taken together, SMA mice exhibited an abnormal growth plate anatomy characterized by suppressed differentiation of hypertrophic chondrocytes. The bone dysplasia could be rescued by SMN supplementation, which consequently alleviated the restriction on long bone development.

Additionally, Kyoto encyclopedia of genes and genomes (KEGG) pathway analysis of significantly differentially expressed genes revealed that several of the top enriched pathways involve sarcomeric and cytoskeletal genes commonly associated with cardiac muscle function, including Myl2, Myl3, Myh6, and Myh7 (Fig. 3e). Given that proper cytoskeletal remodeling is essential for chondrocyte hypertrophy and growth plate morphogenesis, the enrichment of these genes suggests that hypertrophic chondrocytes in SMA mice may experience aberrant cytoskeletal dynamics, potentially disrupting their terminal differentiation during endochondral ossification. On the other hand, insulin administration has been shown to promote chondrocyte differentiation and maturation.²⁸ Both insulin deficiency and insulin resistance are known to exert deleterious effects on bone tissue.²⁹ Therefore, the enrichment of insulin resistance pathways suggests that SMN deficiency may impair chondrocyte differentiation through disrupted insulin signaling.

Moreover, enrichment of the p53 signaling pathway (Fig. 3e) suggests dysregulated cell proliferation and apoptosis.³⁰ qPCR validation showed downregulation of proliferation marker Mki67, and upregulation of Cdkn1a and the pro-apoptotic gene Bax in SMA growth plate cartilage (Fig. S5b), indicating enhanced cell cycle arrest and apoptosis. However, TUNEL staining revealed positive signals only in the periosteum, but not in chondrocytes within the growth plate (Fig. S5c). This suggests that although Bax expression is elevated, chondrocytes have not yet undergone apoptosis at this stage. Therefore, apoptosis is unlikely to be the primary mechanism underlying impaired endochondral ossification caused by SMN deficiency.

In conclusion, SMN deficiency impairs expression of genes related to proliferation and differentiation in chondrocytes.

Apart from cell proliferation and apoptosis, chondrocyte differentiation is crucial in the process of endochondral ossification and bone development. Chondrocyte differentiation is initially controlled by SOX9 and then shifts to RUNX2 during hypertrophy. Then COL X-marked hypertrophic chondrocytes regulate matrix mineralization and subsequently differentiate into SP7-dependent osteoblasts to promote bone ossification.^31,32 We examined expression of SOX9, RUNX2, COL Ⅹ, and SP7 in SMA growth plates through immunofluorescence. SOX9 showed a significant reduction throughout the growth plates of P4 SMA mice, especially in HZ (Fig. 3f, Fig. S5d, e). RUNX2, primarily expressed in hypertrophic chondrocytes, was also downregulated across P4 SMA growth plates (Fig. 3g, Fig. S5f, g). Consistent with increased HZ height, the range of COL X-positive cells expanded longitudinally in P4 SMA mice growth plate cartilages, whereas the mean fluorescence intensity (MFI) of COL X was markedly lower compared to the Het mice, reflecting restricted mature of hypertrophic chondrocytes (Fig. 3h). Moreover, the level of SP7 decreased in P4 SMA metaphysis, indicating restricted bone ossification (Fig. 3i).

To confirm that the restoration of SMN can reverse these changes, ASO10-29 was utilized to elevate SMN levels. Immunofluorescence staining of growth plate sections at P7 revealed a widespread decrease in SOX9 and RUNX2 levels throughout the growth plates of SMA mice at P7 (Fig.), consistent with the trends seen at P4. MFI of COL X and SP7 was downregulated in HZ and metaphysis of SMA mice, respectively (Fig.). In growth plates of SMA mice treated with ASO10-29, we observed marked upregulation of SOX9, RUNX2, COL X, and SP7 levels (Fig.). S6a, b S6c, d S6a–d

Overall, these results suggest that SMN loss suppresses the differentiation of mature hypertrophic chondrocytes, thereby inhibiting endochondral ossification. Notably, these defects can be reversed by early restoration of SMN protein levels.

Since YBX1 decreased in SMA growth plates without changes in transcription, we adopted cycloheximide (CHX) chasing assay and found that SMN deficiency accelerated the degradation of YBX1, which was prevented by proteasome inhibitor MG132 (Fig. 6c). These results suggested that YBX1 is subject to proteasome-dependent degradation in the absence of SMN. To explore the underlying mechanism, we aimed to identify the specific ubiquitin ligase responsible for YBX1 ubiquitination. Venn analysis between mass spectrometry results of SMN-interacting proteins and BioGRID dataset of YBX1³⁵ identified β-catenin, pre-mRNA-processing factor 19 (PRPF19), TRAF6, and valosin containing protein (VCP) as potential candidates (Fig. 6d). TRAF6 has been reported to induce ubiquitination of inhibitor of kappa B kinase (IKK) that binds to SMN.³⁶ Further experiments confirmed the interaction between SMN, YBX1 and TRAF6 (Fig. S10a). We then investigated the effects of SMN deficiency on TRAF6 level and detected no difference in SMA growth plate cartilages (Fig. S10b) or ATDC5 cells transfected with siSmn1 (Fig. S10c), compared to their respective control groups. These findings suggest that SMN knockdown does not directly alter TRAF6 protein levels but rather enhances TRAF6-mediated ubiquitination and subsequent degradation of YBX1. Co-IP analysis confirmed that SMN deficiency increased the interaction between TRAF6 and YBX1, leading to elevated ubiquitin tagging of YBX1 (Fig. 6e, f). Conversely, SMN overexpression reduced this interaction and ubiquitination, thereby restoring YBX1 protein levels (Fig. 6e, f). Moreover, treatment with C25-140, a selective inhibitor of TRAF6’s E3 ubiquitin ligase activity,³⁷ effectively elevated YBX1 levels, thereby underscoring the critical role of TRAF6 in mediating YBX1 ubiquitin-dependent proteasomal turnover (Fig. 6g). Together, SMN deficiency promotes YBX1 degradation through TRAF6-induced ubiquitination.

To determine whether conditional knockdown of Smn1 in chondrocytes causes bone development defects, we conducted histochemical and immunofluorescence staining to compare the anatomical structure of the growth plate and the levels of endochondral ossification-related proteins. The proportion of HZ area and HZ height were both elevated in Smn1-cKD mice, while proportion of RZ and PZ area were similar to that in control mice (Fig. 7k, Fig. S11k, l). SF staining demonstrated smaller size and higher density in hypertrophic chondrocytes of Smn1-cKD mice (Fig. 7l). VK staining revealed more calcification in HZ and delayed calcification in metaphysis of WT mice after Smn1-cKD (Fig. 7m). Moreover, the protein levels of YBX1, SOX9, RUNX2 exhibited a widespread decrease in growth plates of Smn1-cKD mice, particularly in HZ (Fig. 7o–q and Fig. S12b–d). The MFI of COL X was significantly lower in HZ, and SP7 levels decreased in metaphysis of Smn1-cKD mice compared to the control mice (Fig. 7r, s, Fig. S12e, f), indicating a retardation in the progression of bone ossification.

In conclusion, the bone development abnormalities in SMA mice are primarily caused by the absence of Smn1 in chondrocytes, independent of SMN deficiency in other tissues.

Compared to SMA mice, Smn1-cOE mice showed substantial recovery of body weight (Fig. 8b) and femur length (Fig. S13b), although these parameters remained lower than those in Het control. Micro-CT analysis demonstrated that the reductions in BV/TV, Tb.Th, and the increase in Tb.Pf observed in SMA mice were significantly ameliorated in Smn1-cOE mice (Fig. 8c–f). However, these values in Smn1-cOE mice still differed from those in Het mice (Fig. 8c–f), indicating an incomplete rescue of trabecular bone structure. Consistent with findings in Smn1-cKD mice, neither SMN deficiency nor its restoration affected BMD, Tb.Sp, or Tb.N, implying that SMN primarily influences the morphology rather than the quantity or spatial distribution of trabecular bone (Fig. S13c–e). Interestingly, cortical bone indices such as Ct.Th and Ct.BA/TA were fully restored to Het levels in Smn1-cOE mice (Fig. 8g–i), suggesting that SMN deficiency in non-chondrocytic cell types may primarily contribute to trabecular, but not cortical bone alterations.

Histological and immunofluorescence analyses further supported these findings. The enlarged HZ area, decreased RZ area, increased HZ height, and altered mineralization patterns in SMA growth plates were significantly rescued in Smn1-cOE mice (Fig. 8j, l, Fig. S13f, g). Notably, the size and density of hypertrophic chondrocytes were completely restored (Fig. 8k). Protein levels of YBX1, SOX9, RUNX2, and COL X in the hypertrophic zone, along with SP7 expression in the metaphysis, were all upregulated compared to SMA mice (Figs. 8m–q and Fig. S14a–e), supporting the role of SMN in regulating the terminal differentiation and function of hypertrophic chondrocytes.

These findings, taken together with the phenotypic alterations observed in Smn1-cKD mice, demonstrate that chondrocytic SMN expression is essential for the progression of hypertrophic chondrocyte-driven endochondral ossification. Meanwhile, SMN in non-chondrocytic cells appears to partially compensate for the loss of SMN in chondrocytes, thereby mitigating the severity of ossification defects in Smn1-cKD mice (Fig. 8r).

All mouse studies were conducted in accordance with the guidelines and protocols approved by the Institutional Animal Care and Use Committee (IACUC) of the Second Affiliated Hospital of Soochow University, Jiangsu, China. The study complied with Animal Research: Reporting in Vivo Experiments (ARRIVE) guidelines to minimize the discomfort and pain of the animals. The initial breeding pairs of human SMN2 transgenic mice were purchased from Jackson Laboratory (stock number 005058), and the severe SMA model (Smn1^−/−, SMN2^2TG/0) was generated as described.⁶⁷ Genotypes of mice were identified with One Step Mouse Genotyping Kit (Vazyme, PD101-01). For tissue sample collection, mice were sacrificed by CO₂ asphyxiation, and then femurs or growth plate cartilage tissues were immediately isolated, snap-frozen in liquid nitrogen for storage at −80 °C. Both male and female mice were utilized in all experiments, as no discernible sex differences were noted in any of the measured endpoints.

Mice were sacrificed by CO₂ asphyxiation and fixed in 4% paraformaldehyde for 24 h. Skin and organs of mice were removed before fixation in 95% ethanol for 48 h and then immersed into lacquer thinner for 1 week and again 95% alcohol for 2 days. Tissues were stained with Alcian blue solution (Sigma-Aldrich, B8438) for 24–48 h and destained with 95%, 90%, 40%, and 15% alcohol for 2–3 h, respectively. The ethanol was then replaced with 2% KOH clearing solution for 2 h and then stained with Alizarin red solution (Sigma-Aldrich, A5533) for 15–30 min, followed by incubation in 2% KOH until the soft tissue disintegrated for 1 week. The mice were stored long-term in glycerol.

The vertebras and femurs from WT mice injected with AAV9-shCtrl or AAV9-Col2a1p-shSmn1 were isolated and cleaned of muscles and skin. Afterwards, the femurs were fixed in 4% paraformaldehyde prior to analysis. Imaging of the tissues was conducted using a NEMO micro-CT system (PINGSENG Healthcare, NMC-200). Scanning parameters were set at 90 kV and 88 mA, with a field of view of 10 mm (voxel size of 20 μm; scanning duration of 14 min). The 3D images were visualized using the Quantum GX II’s 3D Viewer software. The region 0.5 mm below the lowest point of upper edge of femoral diaphysis was used to measure trabecular parameters, while the area 0.5 mm around the midpoint of the femoral long axis was used for cortical parameter measurements. Due to the indistinct boundary between the cortical and trabecular bone in the vertebrae, the entire vertebra was considered trabecular bone for calculations.

Femurs were fixed in 4% paraformaldehyde for 24 h and then decalcified with 10% EDTA. Tissues were embedded in paraffin and 5 μm-thick paraffin sections were cut for HE, SF, and immunofluorescence staining. Fixed non-demineralized femurs were used for Von Kossa staining. For immunofluorescence staining, sections were dewaxed and rehydrated and followed with antigen retrieval by Tris-EDTA (PH 9.0) (Abcam, ab93684) in boiling water for 20 min. Sections were permeabilized with 0.3% (v/v) Triton-X100 (Sangon, 9002-93-1) for 10 min and then blocked with 10% (v/v) goat serum for 30 min. Primary antibodies including SMN (Abcam, ab314895, 1:100), SOX9 (Abcam, ab185966, 1:100), RUNX2 (Abcam, ab192256, 1:100), Collagen Type X (Abcam, ab260040, 1:100), SP7 (Abcam, ab209484, 1:500), YBX1 (Abcam, ab76149, 1:100) were incubated overnight at 4 °C. Sections were incubated with corresponding secondary antibodies conjugated to Alexa Fluor 488 (Invitrogen, A-11008) and Alexa Fluor 568 (Invitrogen, A-11004) for 1 h at room temperature. Nuclei were counter-stained with DAPI (Thermo Fisher Scientific, 62248). The images were captured with a confocal microscope (Zeiss, LSM800). The area of different zones in growth plates, cell size, cell density, and calcification area were analyzed using ImageJ.

The chondrogenic cell line ATDC5 was cultured in Dulbecco’s modified Eagle’s medium (DMEM, Gibco, 11965092) containing 10% fetal bovine serum (Gibco, 16140071) and 1% Penicillin-Streptomycin-Amphotericin B Solution (Procell, PB180121↗) at 37 °C, in a humidified atmosphere containing 5% CO₂. The ATDC5 cell line was purchased from Procell (FuHeng, FH0378) and has been authenticated by STR profiling. The siRNAs to deplete Smn1 or Ybx1 and plasmids containing ubiquitin were purchased from GenePharma and details are listed in Supplementary Table S1. According to the manufacturer’s instructions, ATDC5 cells were seeded in 6-well plates at 70% confluency, and 100 nmol of each siRNA or 0.1 ng plasmids was transfected into cells in each well, using Lipofectamine 3000 (Invitrogen, L3000001). Otherwise, ATDC5 cells were treated with 20 mol/L C25-140 (MCE, HY-120934), an inhibitor of TRAF6’s ubiquitin ligase activity, for 0, 2, 4, 6, 8 and 10 h before collection.

The growth plate tissues were pulverized in liquid nitrogen and protein extracts were separated on a 10% SDS-polyacrylamide gel and electro-blotted onto PVDF membranes (Bio-Rad, 1620184). The membrane was blocked with 5% non-fat milk for 1–2 h and was incubated overnight at 4 °C with primary antibodies including SMN (Abcam, ab314895, 1:1 000), YBX1 (Abcam, ab76149, 1:5 000), TRAF6 (CST, 67591S), Ubiquitin (CST, 20326S), and GAPDH (Proteintech, 60004-1-Ig). Membranes were incubated with secondary antibodies (Proteintech, SA00001-2, 1:1 000) at room temperature for 1 h. After three washes with TBST, immunolabeling was activated by an Omni-ECL™ Femto Light Chemiluminescence Kit (Epizyme, SQ201) and captured by a GeneGnome XRQ analyzer (Syngen). Gray value was analyzed using ImageJ normalized to GAPDH.

Co-IP was carried out using Protein A/G PLUS-Agarose (Santa Cruz, sc-2003) according to the manufacturer’s instructions. Briefly, total proteins were extracted and quantified. A total of 2 mg protein in 500 μL supernatant was incubated with 10 μg anti-SMN (Abcam, ab314895), anti-YBX1 (Abcam, ab76149), anti-TRAF6 (CST, 67591S) or anti-IgG (Abcam, ab6708) antibodies for 12 h at 4 °C. Agarose beads were washed, eluted in sample buffer, and boiled for 10 min at 100 °C. Immune complexes were subjected to Western blot and mass spectrometry analysis. Anti-IgG was used as a negative control.

In-gel digestion was performed using trypsin as previously described.⁶⁸ After digestion, peptides were extracted from the gel piece with 50% acetonitrile/0.1% formic acid. The extracted peptides were dried in SpeedVacuum concentrator and resuspended in 0.1% formic acid for LC-MS/MS analysis. Peptide samples were separated by EASY nLC-1200 (Thermo Fisher Scientific). The isolated peptides were subjected to Nano source followed by Q Exactive HF-X mass spectrometer. Mass spectra were processed and searched using Proteome Discoverer (version 2.4, Thermo Fisher Scientific) against the Swissprot protein database (release 2022_01). The mass tolerance allowed for the precursor ions is 10 x 10^-6, while the mass tolerance of fragment ions is set to 0.01 Da. Carbamidomethyl on cysteine was specified as fixed modification, while oxidation on methionine and acetyl on protein N-terminal were specified as variable modification. Peptide confidence was set at high.

Trizol (Invitrogen, 15596026) was used to extract total RNA from growth plate cartilages of mice and ATDC5 cells after pulverizing them in liquid nitrogen with mortar and pestle. One microgram of total RNA was reverse transcribed with Hiscript III Reverse Transcriptase (Vazyme, R302-01). Semi-quantitative real-time PCR was performed with 2× Taq Plus Master Mix (Vazyme, P212-01) in PTC-200 (BIO-RAD, CA, USA). Quantitative real-time PCR was performed with ChamQ SYBR qPCR Master Mix (Vazyme, Q311-02/03) in QuantStudio 5 (Thermo Fisher Scientific) according to the manufacturer’s instructions. The primers for all genes are supplied in Supplementary Tables S2 and S3. Gene expression levels were analyzed relative to Gapdh.

Neonatal mice at P4 were euthanized, and the femurs were rapidly harvested in cold PBS. Under a stereo microscope, surrounding soft tissues and metaphyseal bone were removed with micro-scissors. The epiphysis was bisected longitudinally to expose the growth plate and cartilage was carefully dissect with fine forceps. Total RNA was isolated from the growth plate cartilages of the femurs on P4. RNA-seq was performed using an Illumina Genome Analyzer, with an average of 44–52 million mapped reads per sample. The raw data were subjected to QC analyses using FastQC v0.11.7 software. The transcripts were normalized to those of the control group and transcripts with low variance (<0.1) across samples were removed. Statistically significant changes in gene expression were calculated using the DESeq2-package. Genes with FC > 2 and Q-value < 0.05 were considered as differentially expressed genes and the DAVID Bioinformatics Resources 6.8. rMATS software was used to calculate the inclusion of a given differentially expressed exon as percent spliced-in (PSI) and assess the fraction of a gene’s mRNA.

The scRNA-seq data were downloaded from two deposited dataset (GSE190616↗ and GSE179148↗). We excluded cells with fewer than 200 detected genes, more than 7% mitochondrial genes, or those expressed in ≤5 cells. Cells were categorized into different clusters as described in Long et al., 2022. and marker genes for each cluster were identified using the R package of Seurat 5.^69–73 Monocle analysis was performed (num_dim = 50) using the Monocle3 R package.^74–77

ATDC5 cells were transfected with negative control or siSmn1 for 4 h and then treated with MG132 (20 μmol/L) or CHX (50 μg/mL) for 0, 2, 4, 6, 8 or 10 h. Cells were collected at the indicated time points and subjected to Western blot analysis. All experiments were performed at least three times independently.

MOE-modified ASO10-29 (5′-ATTCACTTTCATAATGCTGG-3′) with phosphorothioate backbone and all 5-methylcytosines were synthesized by Genepharma. The oligonucleotide solution was injected subcutaneously twice at 90 mg/kg between P0 and P1 and tissue samples were collected at P7. AAV9-shCtrl, AAV9-Clo2a1p-shSmn1 and AAV9-Col10a1p-oeSmn1 were synthesized by Genechem. The rAAV9 virus was diluted in 50 µL of saline at a concentration of 2 × 10¹¹ vg/g and administrated via retro-orbital injection at P1 or P3. Blanching of the superficial temporal vein indicates successful injection.

The Shapiro–Wilk test was performed to evaluate the normality of the numerical data. Normally distributed data are presented as the mean ± standard deviation (SD), while non-normally distributed data are presented as the median (25th percentile, 75th percentile). A two-tailed unpaired Student’s t test was used for comparisons between 2 groups obeying normal distribution. One-way ANOVA followed by Tukey post hoc test was applied for comparisons among several groups obeying normal distribution. In condition of an abnormal distribution or a small sample size (n < 6), Wilcoxon rank-sum test was applied for comparisons between 2 groups and Kruskal–Wallis test was applied for comparisons among several groups. A P-value of <0.05 was considered statistically significant. GraphPad Prism 9.0 software was used for all statistical analyses.