In addition to the well-characterized role of the CTSK gene in osteoclasts, this gene has recently been reported to identify periskeletal progenitors (Yang et al., 2013; Debnath et al., 2018; Shi et al., 2018). In particular, the ablation of Chmp5 in mouse Ctsk-expressing cells (hereafter referred to as Ctsk^Cre;Chmp5^fl/fl mice, homozygous) caused dramatic periskeletal bone overgrowth near the joint compared to Ctsk^Cre;Chmp5^fl/+ (heterozygous) and Chmp5^fl/fl littermate controls (Figure 1A and B). Aberrant periskeletal bone growth in Ctsk^Cre;Chmp5^fl/fl mice became distinct around postnatal day 10 (d10), progressed with age, and involved multiple bones and joints throughout the body, including knee, ankle, and foot joints (Figure 1A–C and Figure 1—figure supplement 1A). With age, these animals developed severe skeletal deformities, progressive joint stiffness, reduced motility, and short stature. Despite bone overgrowth in Ctsk^Cre;Chmp5^fl/fl animals, mechanical tests showed that bone stiffness and fracture stress decreased significantly (Figure 1D), suggesting that bone mechanical properties were impaired in Ctsk^Cre;Chmp5^fl/fl mice.

To track CTSK⁺ periskeletal progenitors during the bone overgrowth in Ctsk^Cre;Chmp5^fl/fl mice, we crossed Ctsk-Cre, Chmp5^fl/fl, and Rosa26^mTmG mice to generate Ctsk^Cre;Chmp5^fl/+;Rosa26^mTmG/+ and Ctsk^Cre;Chmp5^fl/fl;Rosa26^mTmG/+ reporter animals, in which CTSK⁺ periskeletal progenitors and their descendants express membrane-localized green fluorescence protein (GFP). In Ctsk^Cre;Chmp5^fl/+; Rosa26^mTmG/+ control mice, GFP⁺ cells could be detected in the perichondrium, groove of Ranvier, periosteum, ligament, and enthesis in addition to osteoclasts in the bone marrow, but not in growth plate chondrocytes (Figure 1E). With Chmp5 deletion, GFP⁺ cells consisted of a major part of the cells in periskeletal overgrowth in Ctsk^Cre;Chmp5^fl/fl; Rosa26^mTmG/+ mice (Figure 1E, Figure 1—figure supplement 1B). Tartarate-resistant acid phosphatase (TRAP) staining did not show TRAP⁺ osteoclasts in the periskeletal lesion, while obvious TRAP⁺ osteoclasts could be detected in the bone marrow (Figure 1—figure supplement 1C). These results indicate that Chmp5-deficient periskeletal progenitors directly contribute to the bone overgrowth in Ctsk^Cre;Chmp5^fl/fl mice.

Next, we sought to confirm the role of CHMP5 in the regulation of bone formation by deleting this gene using another mouse line of osteogenic lineage Cre. We crossed Chmp5^fl/fl mice with the line of Prrx1-Cre (Logan et al., 2002), Col2a1-Cre (Ono et al., 2014) or Osx-Cre (Maes et al., 2010) mice to delete the Chmp5 gene in Prrx1-expressing limb mesenchymal progenitors (Prrx1^Cre;Chmp5^fl/fl mice), Col2a1-expressing perichondral progenitors (Col2a1^Cre; Chmp5^fl/fl mice) or Osx-expressing osteoprogenitors (Osx^Cre;Chmp5^fl/fl mice) and were unable to obtain postnatal homozygous gene knockout mice, probably due to embryonic lethality after ablating Chmp5 expression in these osteogenic lineages. Subsequently, we used the Dmp1-Cre mouse line that could target postnatal endosteal osteoprogenitors, preosteoblasts, mature osteoblasts, and osteocytes (Lim et al., 2016; Lim et al., 2017; Matic et al., 2016). Dmp1^Cre;Chmp5^fl/fl mice were born in normal Mendelian ratios and were viable. Within 3 wk of age, the skeletons of Dmp1^Cre;Chmp5^fl/fl mice began to show a distinct expansion in comparison with Dmp1^Cre;Chmp5^fl/+ and Chmp5^fl/fl littermate controls, which progressed with age and became much more severe at 1 y of age (Figure 1F). Bone expansion in Dmp1^Cre;Chmp5^fl/fl mice that occurred mainly in the endosteum was further shown in detail around the age of 10 wk by micro-CT and histological analyses (Figure 1G and Figure 1—figure supplement 1D, E).

Meanwhile, since Dmp1-expressing lineage cells also generate skeletal muscle cells (Lim et al., 2017), Dmp1^Cre;Chmp5^fl/fl mice showed a profound decrease in skeletal muscle mass with age (Figure 1F and H). Hindlimb abduction, full-limb hanging time, and front-limb pull strength tests showed that muscular functions also decreased profoundly in Dmp1^Cre;Chmp5^fl/fl compared to Chmp5^fl/fl mice (Figure 1I and J).

To further characterize the function of CHMP5 in the regulation of bone formation, we cultured and sorted Ctsk-expressing periskeletal progenitor cells (CD45^-;CD31^-;GFP⁺) from periskeletal tissues of Ctsk^Cre;Chmp5^fl/+;Rosa26^mTmG/+ and Ctsk^Cre;Chmp5^fl/fl; Rosa26^mTmG/+ mice in the early stage of skeletal lesions (P10-P14) (Figure 2A and Figure 2—figure supplement 1A, B, hereafter referred to as Ctsk^Cre;Chmp5^fl/+ and Ctsk^Cre;Chmp5^fl/fl periskeletal progenitors, respectively). Chmp5 deletion in Ctsk^Cre;Chmp5^fl/fl periskeletal progenitors was confirmed by gene expression analysis (Figure 2B). Strikingly, when cultured in the osteogenic differentiation medium, Ctsk^Cre;Chmp5^fl/fl periskeletal progenitors showed considerably enhanced osteogenesis in comparison with Ctsk^Cre;Chmp5^fl/+ control cells, as shown by alizarin red staining, von Kossa staining, and alkaline phosphatase activity assay (Figure 2C).

The regulation of osteogenesis by CHMP5 in skeletal progenitors was further determined in the osteogenic cell line MC3T3-E1 after removing Chmp5 using CRISPR/CAS9 technology (Figure 2D). The deletion of Chmp5 in MC3T3-E1 cells also markedly increased osteogenesis, as shown by alizarin red staining at 2 wk and 3 wk of osteogenic induction (Figure 2E). Consistently, the ingenuity pathway analysis of RNA-seq data of Ctsk^Cre;Chmp5^fl/fl vs. Ctsk^Cre;Chmp5^fl/+ periskeletal progenitors showed that multiple molecular pathways related to osteoblast differentiation were activated with the Chmp5 deletion (Figure 2F). Taken together, these results reveal a function of CHMP5 in regulating osteogenesis in skeletal progenitors.

To gain more insight into the mechanism of regulation of bone formation by CHMP5, we assumed that depletion of Chmp5 in osteogenic lineage cells also increases cell proliferation, which contributes to the bone overgrowth in Chmp5 conditional knockout mice. Unexpectedly, Ctsk^Cre;Chmp5^fl/fl periskeletal progenitors showed a significant decrease in the cell proliferation rate compared to Ctsk^Cre;Chmp5^fl/+ controls (approximately 58% decrease in the cell number on day 6 of P2 culture, Figure 3A). Because CHMP5 has been reported to regulate thymocyte apoptosis (Adoro et al., 2017), we performed cell apoptosis analyses for Ctsk^Cre;Chmp5^fl/fl and Ctsk^Cre;Chmp5^fl/+ cells. Annexin V staining showed that approximately 4.96 ± 0.81% of Ctsk^Cre;Chmp5^fl/fl periskeletal progenitors vs. 1.69 ± 0.68% of Ctsk^Cre;Chmp5^fl/+ control cells were positively stained (Figure 3—figure supplement 1A). Similarly, the in-situ terminal deoxynucleotidyl transferase dUTP nick end labelling assay (TUNEL) demonstrated that around 3.70 ± 0.72% cells in the periskeletal lesion of Ctsk^Cre;Chmp5^fl/fl mice were positively labelled (Figure 3—figure supplement 1B). These results show that cell apoptosis is activated in Ctsk^Cre;Chmp5^fl/fl periskeletal progenitors in a rather low ratio.

Notably, the low apoptotic cell ratio (approximately 5%) could not well match the decrease in cell proliferation rate (approximately 58%) in Ctsk^Cre;Chmp5^fl/fl periskeletal progenitors. We further analyzed the RNA-seq data of Ctsk^Cre;Chmp5^fl/fl vs. Ctsk^Cre;Chmp5^fl/+ cells to dig out the mechanism of CHMP5 deficiency on the fate of skeletal progenitors. The gene set enrichment analysis (GSEA) showed significant enrichment of genes in multiple Reactome molecular pathways associated with cell senescence in Ctsk^Cre;Chmp5^fl/fl relative to Ctsk^Cre;Chmp5^fl/+ periskeletal progenitors, including Hmga1, Hmga2, Trp53, Ets1, and Txn1 (Figure 3B and C). Meanwhile, GSEA of RNA-seq data also showed significant enrichment of the SAUL_SEN_MAYO geneset (positively correlated with cell senescence) and the KAMMINGA_SENESCENCE geneset (negatively correlated with cellular senescence) in Ctsk^Cre;Chmp5^fl/fl vs. Ctsk^Cre;Chmp5^fl/+ periskeletal progenitors (Figure 3—figure supplement 1C).

Furthermore, Western blot results showed that both the p16 and p21 proteins, two canonical markers of cellular senescence, were upregulated in Ctsk^Cre;Chmp5^fl/fl compared to control cells, with a higher degree of upregulation in the p16 protein (Figure 3D). However, the mRNA levels of Cdkn2a (p16) and Cdkn1a (p21) did not show significant changes according to the RNA-seq analysis (Figure 3—figure supplement 1D). Furthermore, immunostaining for another cell senescence marker γH2Ax demonstrated that there were significantly more γH2Ax⁺;GFP⁺ cells in periskeletal overgrowth in Ctsk^Cre;Chmp5^fl/fl;Rosa26^mTmG/+ mice relative to the periosteum of Ctsk^Cre;Chmp5^fl/+; Rosa26^mTmG/+ control mice (Figure 3E). Accordingly, Ctsk^Cre;Chmp5^fl/fl mice showed gross accelerated aging-related phenotypes, including hair loss, joint stiffness/contracture, and decreased motility (Figure 3F).

To further characterize the effect of CHMP5 on skeletal progenitor cell fate, we deleted Chmp5 in ATDC5 cells, which were shown to be an appropriate model of periskeletal progenitors in a previous study (Ge et al., 2016), using CRISPR/CAS9 technology (Figure 3—figure supplement 1E). Similarly, Chmp5 deletion also significantly suppressed the cell proliferation rate of ATDC5 cells and mildly increased cell apoptosis (Figure 3—figure supplement 1F, G). However, inhibition of cell apoptosis using the pancaspase inhibitor Q-VD-Oph could not reverse cell number reduction (Figure 3—figure supplement 1H), indicating that cell apoptosis was not the main cause of the decrease in the proliferation rate of Chmp5-deficient skeletal progenitors. Instead, cell cycle analysis showed a significant decrease of Chmp5-deficient ATDC5 cells in phase S, and a significant proportion of these cells were arrested in G0/G1 or G2/M phases (Figure 3G), which is a typical characteristic of cell cycle arrest in senescent cells. Taken together, these results at the molecular, cellular, and gross levels demonstrate that Chmp5 deficiency induces skeletal progenitor cell senescence.

A remarkable feature of senescent cells is the senescence-associated secretory phenotype (SASP), which plays a critical role in mediating the pathophysiological processes of cell senescence (Birch and Gil, 2020). In particular, periskeletal overgrowths in Ctsk^Cre;Chmp5^fl/fl;Rosa26^mTmG/+ mice showed evident integration of many cells lacking evidence of Cre-mediated recombination (GFP^-) (Figure 1E), suggesting that Chmp5-deleted periskeletal progenitors might recruit neighboring wild-type cells to facilitate the bone overgrowth. Additionally, the GSEA of RNA-seq data showed a significant enrichment of genes associated with the SASP molecular pathway in Chmp5-deleted periskeletal progenitors (Figure 4A). Meanwhile, confocal fluorescence microscopy captured many extracellular vesicles shed from the plasma membrane of Ctsk^Cre;Chmp5^fl/fl periskeletal progenitors, which were rarely detected around wild-type control cells (Figure 4B). Furthermore, nanoparticle tracking analysis demonstrated higher concentrations of extracellular vesicles in the culture medium of Ctsk^Cre;Chmp5^fl/fl periskeletal progenitors compared to wild-type control cells (Figure 4C, D, Figure 4—figure supplement 1A). These results show that Chmp5 deficiency activates the SASP molecular pathway and increases the secretion of skeletal progenitors.

Functionally, conditioned medium collected from Chmp5-deficient skeletal progenitors caused a higher proliferation rate of ATDC5 cells than medium from control cells (Figure 4E). Meanwhile, coculture of wild-type with Ctsk^Cre;Chmp5^fl/fl periskeletal progenitors promoted osteogenesis of wild-type cells (Figure 4—figure supplement 1B). Simultaneously, CD45^-;CD31^-;GFP^- skeletal progenitors from periskeletal tissues of Ctsk^Cre;Chmp5^fl/fl;Rosa26^mTmG/+ mice showed increased proliferation along with enhanced osteogenic differentiation compared to the corresponding cells from Ctsk^Cre;Rosa26^mTmG/+ control mice (Figure 4F, G and Figure 4—figure supplement 1C, D). Immunostaining for the cell proliferation marker Ki-67 demonstrated an increase in the number of stained cells in the periskeletal lesion of Ctsk^Cre;Chmp5^fl/fl mice (Figure 4—figure supplement 1E). Since GFP⁺ Ctsk^Cre;Chmp5^fl/fl periskeletal progenitors showed a decreased cell proliferation rate and increased cell senescence (Figure 3), and the GFP^-Ctsk^Cre;Chmp5^fl/fl periskeletal progenitors showed an increased cell proliferation rate (Figure 4F and Figure 4—figure supplement 1C), these Ki-67⁺ cells in Figure 4—figure supplement 1E should represent GFP^--skeletal progenitors in the periskeletal lesion. Together, these results demonstrate that Chmp5-deficient skeletal progenitors could promote the proliferation and osteogenesis of surrounding wild-type skeletal progenitors.

Next, we performed a secretome analysis to characterize secretory profiles of Ctsk^Cre;Chmp5^fl/fl vs. wild-type periskeletal progenitors. Fifteen proteins were found to increase and five proteins to decrease in the cell supernatant of Ctsk^Cre;Chmp5^fl/fl periskeletal progenitors (Figure 4H and Figure 4—figure supplement 1F). In particular, all 15 upregulated proteins, including NCAM1, CSPG4, SDF4, NME2, TAGLN, OLFML2B, PXDN, COL4A1, FSTL1, TPI1, MFGE8, COL1A2, COL1A1, HSPG2, and COL3A1, have been identified in the osteoblastic cell secretome or in the regulation of osteoblast differentiation or functions (Romanello et al., 2014; Davies et al., 2019). Western blot verified upregulation of two of these proteins, COL1A1 and TAGLN, in the cell supernatant of Ctsk^Cre;Chmp5^fl/fl vs. wild-type periskeletal progenitors (Figure 4I). Notably, the secretome analysis did not detect common SASP factors, such as cytokines and chemokines, in the secretory profile of Ctsk^Cre;Chmp5^fl/fl periskeletal progenitors, probably due to their small molecular weights and the technical limitations of the mass-spec analysis. Taken together, these results demonstrate a secretory phenotype and paracrine actions of Chmp5-deficient skeletal progenitors. However, factors that mediate the paracrine actions of Chmp5-deficient periskeletal progenitors remain to be further clarified in future studies.

To affirm the role of cell senescence of osteogenic cells in contributing to bone overgrowth in Ctsk^Cre;Chmp5^fl/fl and Dmp1^Cre;Chmp5^fl/fl mice, we treated Chmp5-deficient periskeletal progenitors and animals with the senolytic drugs quercetin and dasatinib (Q+D), which have been widely used to eliminate senescent cells under various physiological and pathological conditions (Zhu et al., 2015; Hickson et al., 2019), and assessed the therapeutic efficacy. First, in vitro Q+D treatment showed that Ctsk^Cre;Chmp5^fl/fl periskeletal progenitors were more sensitive to senolytic drugs for cell apoptosis compared to control cells (Figure 5A). Ctsk^Cre;Chmp5^fl/fl animals were then treated with Q+D from the first 2 wk after birth for 7 wk and showed significantly improved periskeletal bone overgrowth and animal motility compared to Chmp5^fl/fl control animals (Figure 5B, Figure 5—figure supplement 1). Similarly, treatment of Dmp1^Cre;Chmp5^fl/fl mice with Q+D also improved musculoskeletal manifestations, including hindlimb abduction, the whole femur and cortical bone thickness, and partially restored skeletal muscle functions (Figure 5C–E). An improvement in animal motility was also found with Q+D treatment in adult mice. There were no apparent abnormalities in Chmp5^fl/fl control animals treated with the same dose of Q+D in parallel.

Furthermore, to verify the efficacy of senolytic drugs in treating the periskeletal bone overgrowth in Ctsk^Cre;Chmp5^fl/fl mice, we used another senolytic drug Navitoclax (ABT-263), which is a BCL-2 family inhibitor and specifically induces apoptosis of senescent cells (Tse et al., 2008; Chang et al., 2016). Micro-CT analysis demonstrated that ABT-263 could also relieve periskeletal bone overgrowth in Ctsk^Cre;Chmp5^fl/fl mice (Figure 5F). Together, these results demonstrate that the elimination of senescent cells using senolytic drugs is effective in alleviating musculoskeletal pathologies in Ctsk^Cre;Chmp5^fl/fl and Dmp1^Cre;Chmp5^fl/fl animals, confirming that osteogenic cell senescence is responsible for musculoskeletal abnormalities in Ctsk^Cre;Chmp5^fl/fl and Dmp1^Cre;Chmp5^fl/fl mice.

Next, we wondered whether Chmp5 deficiency in skeletal progenitors affects the endocytic pathway. Notably, approximately 45% of Ctsk^Cre;Chmp5^fl/fl periskeletal progenitors contained enlarged GFP⁺ vesicles, which were rarely found in wild-type Ctsk⁺ periskeletal progenitors (Figure 6A), suggesting a possible issue with the endocytic pathway in Chmp5-deleted skeletal progenitors. Next, we used molecular markers for different components and the function of the endocytic pathway to elucidate the influence of Chmp5 deficiency on the endolysosomal system in skeletal progenitors (Figure 6B).

The LysoTracker Red DND-99 trace and the immunostaining of lysosome-associated membrane protein 1 (LAMP1) showed positive staining in many of the enlarged vesicles in Ctsk^Cre;Chmp5^fl/fl periskeletal progenitors (Figure 6C and Figure 6—figure supplement 1A, B). The quantification of the number and average size of LAMP1-stained vesicles were further performed in ATDC5 cells with or without Chmp5 deletion (Figure 6D). Next, the number and size of the vesicles stained for the late endosome marker RAB7 also increased markedly in Chmp5-deficient relative to wild-type skeletal progenitors (Figure 6E and F). Meanwhile, many of these accumulated vesicles were double positive for LAMP1 and RAB7 (Figure 6—figure supplement 1C, D), suggesting that they are terminal compartments in the endocytic pathway. Accordingly, the transmission electron microscopy demonstrated a significant accumulation of MVB-like and lysosome-like structures with different electron density in Ctsk^Cre;Chmp5^fl/fl periskeletal progenitors (Figure 6G).

To examine the function of the endocytic pathway, we used a pH-sensitive fluorescent EGF conjugate and traced the degradation of the internalized EGF receptor in Ctsk^Cre;Chmp5^fl/fl vs. wild-type periskeletal progenitors. Chmp5 deficiency significantly delayed the degradation of the EGF conjugate in Ctsk^Cre;Chmp5^fl/fl periskeletal progenitors (Figure 6H), indicating altered function of the endocytic pathway. It should be noted that the early endosomes of EEA1⁺ also increased slightly in approximately 10% of Ctsk^Cre;Chmp5^fl/fl periskeletal progenitors, while the recycling endosomes of RAB11⁺ and the TGN38⁺ trans-Golgi network were relatively normal (Figure 6—figure supplement 1E). These results demonstrate that CHMP5 deficiency in skeletal progenitors causes the accumulation of late endosomes and lysosomes and disrupts the function of the endolysosomal pathway.

To dissect the molecular mechanism of CHMP5 that regulates the endolysosomal pathway, we performed a proteomic analysis in Ctsk^Cre;Chmp5^fl/fl and wild-type periskeletal progenitors. Bioinformatic analyses of proteomic data showed that VPS4A is one of the most significantly decreased proteins in Ctsk^Cre;Chmp5^fl/fl compared to wild-type periskeletal progenitors (Figure 6I). The decrease in the VPS4A protein in Ctsk^Cre;Chmp5^fl/fl periskeletal progenitors was verified by Western blot analysis (Figure 6J), while the Vps4a mRNA did not show a significant change in the RNA-seq analysis (Figure 6—figure supplement 1F). On the other hand, when CHMP5 was overexpressed by transfection of a CHMP5 vector into HEK-293T cells, the level of VPS4A protein increased significantly and VPS4A mRNA was not significantly affected (Figure 6—figure supplement 1G, H). These results indicate that CHMP5 is essential in maintaining the VPS4A protein. Since VPS4A is a critical ATPase mediating the function of the ESCRT-III protein complex, the reduction of VPS4A protein could be responsible for endolysosomal dysfunction in Chmp5-deleted skeletal progenitors.

A remaining question is how Chmp5 deficiency and endolysosomal dysfunction cause skeletal cell senescence. Impairment of lysosomal functions could hinder the degradation of damaged mitochondria, resulting in the accumulation of reactive oxygen species (ROS) in mitochondria (Martini and Passos, 2023; Park et al., 2018). Indeed, the GSEA of RNA-seq data showed significant enrichment of genes associated with the molecular pathway of oxidative stress-induced cell senescence in Ctsk^Cre;Chmp5^fl/fl vs. Ctsk^Cre;Chmp5^fl/+ periskeletal progenitors (Figure 7A). Accordingly, the level of mitochondrial ROS and the abundance of mitochondria increased markedly in Chmp5-deficient compared to wild-type skeletal progenitors, as shown by CellRox, tetramethylrhodamine methyl ester (TMRE), and MitoTracker fluorescence stain, and expression levels of mitochondrial inner membrane proteins NDUF88, SDHB, and ATP5A (Figure 7B–D and Figure 7—figure supplement 1A, B).

Mitochondrial functions were also compromised in Chmp5-deficient skeletal progenitors as shown by decreased mitochondrial respiratory capacity, increased extracellular acidification rate, and reduced tolerance to galactose (Figure 7E and F). Furthermore, mitochondrial dysfunction and DNA damage response usually form a feedback loop to drive senescent cell phenotypes (Gorgoulis et al., 2019; Korolchuk et al., 2017; Birch and Passos, 2017). The GSEA of the RNA-seq data also showed enrichment of genes associated with DNA damage-induced cell senescence in Ctsk^Cre;Chmp5^fl/fl vs. Ctsk^Cre;Chmp5^fl/+ skeletal progenitors (Figure 7—figure supplement 1C). Upregulation of two critical DNA damage-responsive genes H2afx and Trp53 was verified by qPCR in Ctsk^Cre;Chmp5^fl/fl relative to Ctsk^Cre;Chmp5^fl/+ skeletal progenitors (Figure 7—figure supplement 1D). Importantly, N-Acetylcysteine (NAC) antioxidant treatment could reverse the reduced activity of cell proliferation in Chmp5-deficient skeletal progenitors (Figure 7G). Therefore, the accumulation of dysfunctional mitochondria and mitochondrial ROS is responsible for cell senescence in Chmp5-deficient skeletal progenitors.

In this study, two Chmp5-flox mouse strains were used, and similar phenotypes were observed. The Chmp5^tm2.1Gho strain was previously described (Greenblatt et al., 2015). The alternative Chmp5-flox strain was established by CRISPR/Cas9-mediated genome engineering. Similarly to the Chmp5^tm2.1Gho strain, exons 4 and 5 were selected as the conditional knockout region in the newly created strain, and deletion of this region results in a frameshift of the Chmp5 gene. To engineer the target vector, homologous arms, and the conditional knockout region with 5’ and 3’ loxP sites were generated by PCR using the BAC clone. Cas9, gRNA, and targeting vectors were co-injected into fertilized eggs for gene-edited mouse production, and pups were genotyped by PCR followed by sequencing analysis. Mice bearing the targeted allele were backcrossed with C57BL/6 J mice up to the F8 generation.

Ctsk^tm1(cre)Ska (Ctsk^Cre), B6N.FVB-Tg(Dmp1-cre)1Jqfe/BwdJ (Dmp1-Cre), B6.Cg-Tg(Prrx1-cre)1Cjt/J (Prrx1^Cre), B6;SJL-Tg(Col2a1-cre)1Bhr/J (Col2a1-Cre), and B6.Cg-Tg(Sp7-tTA,tetO-EGFP/cre)1Amc/J (Osx1-GFP::Cre) mice have previously been reported (Greenblatt et al., 2015; Logan et al., 2002; Lu et al., 2007; Rodda and McMahon, 2006; Nakamura et al., 2007). While the Ctsk-Cre strain is a knock-in mouse line, other strains (Dmp1-Cre, Prrx1-Cre, Col2a1-Cre, and Osx-Cre) are transgenic lines. The Gt(ROSA)26Sor^{tm4(ACTB-tdTomato,-EGFP)Luo} (Rosa26^mTmG) mice were purchased from Jackson Laboratories. To generate experimental animals, Chmp5^fl/fl mice were crossed with Ctsk^Cre;Chmp5^fl/+ and Dmp1^Cre;Chmp5^fl/+ mice, respectively. To generate reporter mouse lines, Ctsk^Cre;Chmp5^fl/+ mice were mated with Chmp5^fl/fl;Rosa26^mTmG mice, and Ctsk^Cre mice were crossed with Rosa26^mTmG animals.

All mice were kept in a C57BL/6 or a C57BL/6/129S6 mixed background and housed in the standard animal facility on a 12 hr light/dark cycle with ad libitum access to water and standard chow. In most situations, the littermates were compared otherwise as indicated. All animals were used in accordance with the Guide for Care and Use of Laboratory Animals (NIH, United States) and were handled according to protocols approved by the IACUC at the University of Massachusetts Medical School (Worcester, United States) and Capital Medical University (Beijing, China). The biological and technical replicates (n numbers) for each animal experiment are reported in the figure legends.

In this study, both sexes of mice were used, and unless otherwise stated, no gender differences were observed in the reported phenotypes. No animals were excluded from all analyses unless they had unexpected deaths or illnesses unrelated to this study. Due to the overt musculoskeletal phenotypes of Ctsk^Cre;Chmp5^fl/fl and Dmp1^Cre;Chmp5^fl/fl mice, the experimenters could not be blinded to animal genotypes. All data were analyzed by individuals with no knowledge of the experimental groups.

Micro-CT analyses were performed using the Scanco μCT-35 or μCT-40 scanner (Scanco Medical). Images were processed using the Inveon Research Workplace (Siemens Medical Solutions). X-ray images were acquired using the X-Ray MX-20 Specimen (Faxitron) following the default programme.

All samples were fixed in 10% neutral formalin or 4% paraformaldehyde overnight for premature samples or 2 d for mature skeletal samples. Samples were decalcified with 14% EDTA and serial sections were cut for all samples. At least five slides were selected at a certain interval (80–100 μm) from each sample for histological evaluation. Fate-mapping images were acquired using a Leica TCS SP5 II laser scanning confocal microscope.

Immunofluorescence for γH2AX was performed using a Phospho-Histone H2A.X (Ser139) antibody (clone 20E3, CST). Immunohistochemistry for Ki-67 was performed using the VECTASTAIN Elite ABC HRP Kit (Peroxidase, Rabbit IgG; Vector Laboratories). Cells for immunofluorescence were cultured in 35 mm MatTek glass bottom dishes (No. 1.5 coverslip, MatTek Corporation). Briefly, tissues or cells were fixed in 4% PFA for 15 min, blocked in 1% BSA for 30 min, and incubated with primary antibodies (mouse anti-EEA1 monoclonal antibody, rabbit anti-RAB7 monoclonal antibody, rabbit anti-RAB11 monoclonal antibody, rat anti-CD107a monoclonal antibody, rat anti-TGN38 monoclonal antibody) for 2 hr and then fluorescence-conjugated secondary antibodies for 1 hr. Images were acquired using the Leica TCS SP5 II laser scanning confocal microscope. The quantification of intracellular vesicles was performed using FIJI software following the particle analysis procedure (https://imagej.net/imagej-wiki-static/Particle_Analysis, NIH).

The femurs of Ctsk^Cre;Chmp5^fl/fl and Chmp5^fl/fl mice at 6–7 wk of age were subjected to mechanical testing using the DMA Q800 mechanical test system with the 3-point bending mode following the manufacturer’s procedure (TA Instruments, USA). The femurs were placed in a direction of the anterior surface downward and the posterior aspect of the femoral condyles facing upward. The span length (L) was 10 mm and the load was applied in the middle of the bone. Five mice per group were used and bone stiffness and ultimate stress (fracture stress) from male mice were reported. Similar changes were observed in both sexes.

The four-limb hanging test was performed using the top grid of a mouse cage (Bonetto et al., 2015). The grid was set at a height of about 35  cm, and the measurement was repeated for 3–5 times for each mouse at a rest interval of about 3–5 min. The maximum hang time was recorded for each animal. Both male and female animals were included in this experiment, and there was no significant sex difference regarding the four-limb hanging time.

The forelimb grip strength test was carried out using an electronic scale as described with slight modification (Bonetto et al., 2015; Takeshita et al., 2017). A mouse was allowed to grasp the gauze attached to the scale. The scale was reset to 0  g after stabilization and a researcher slowly pulled the mouse’s tail backward. Five consecutive measurements were performed for each mouse and the peak pull force in grams was recorded. A significant sex difference was observed in the forelimb grip strength. Both male and female mice were included in this study, and the same trend of changes was found in both genders.

The periskeletal progenitors were isolated from the hindlimbs of Ctsk^Cre;Rosa26^mTmG/+, Ctsk^Cre;Chmp5^fl/+;Rosa26^mTmG/+, and Ctsk^Cre;Chmp5^fl/fl;Rosa26^mTmG/+ mice (n=10 animals per genotype) at 2 wk of age according to the previously established method (Ge et al., 2016). Briefly, the periskeletal tissues were dissected, minced into small pieces, and digested with 1 mg/ml collagenase type I (Worthington), 1 mg/ml collagenase type II (Worthington), and 1.5 mg/ml dispase (Roche) for 30 min at 37 °C. Cells were cultured in skeletal stem cell medium (Joint Therapeutics, Beijing, China) for 1 wk, stained with anti-CD31 and anti-CD45 antibodies, and sorted using a FACSAria II cell sorter (BD Biosciences) for further analyses.

In all experiments, Ctsk^Cre;Chmp5^fl/fl and control cells from littermate mice were cultured, sorted, and re-seeded in parallel for subsequent analyses. The coculture experiment was carried out by directly mixing 90% wild-type skeletal progenitors with 10% Ctsk^Cre;Chmp5^fl/fl or control periskeletal progenitors to simulate the in vivo context of cell-cell contact in periskeletal overgrowth. The cocultures were induced in osteogenic differentiation medium for 4 wk. The biological and technical replicates (n-numbers) for each cellular experiment are reported in the figure legends.

Pairs of CRISPR guide RNA oligos (mouse Chmp5 single guide RNAs [sgRNAs] targeting GGCTCCGCCACCTAGCTTGA and GTTTCGCTTTTCCGAAGAAT on exon 1, respectively) were annealed and cloned into the BsmBI sites of the lentiCRISPR V2-puro vector (plasmid 52961, Addgene). CRISPR lentiviral plasmids and lentiviral packaging plasmids (pMDLg/pRRE, pRSV-Rev, and pMD2.G; Addgene) were transfected into HEK 293T cells. The supernatants were harvested and filtered through a 0.45 μm filter 2.5 d after transfection. MC3T3-E1 and ATDC5 cells were infected with CRISPR lentivirus and selected with puromycin (4.5 μg/ml, Clontech) for 7 d. The depletion of Chmp5 was confirmed by Western blotting.

For osteogenic differentiation, 1.0×10⁵ cells were seeded in each well of 24-well plates and cultured overnight. Cells were changed into osteogenic media (Joint Therapeutics) the next day and induced for the indicated times. Osteogenesis was determined by alizarin red and von Kossa staining and by examining the activity of alkaline phosphatase.

The cell number was counted with a Countess II FL Automated Cell Counter (Thermo Fisher Scientific, USA) at the indicated time intervals. The AlamarBlue cell viability assay was performed as previously described (Ge et al., 2016). The cell cycle was determined using the APC-BrdU Flow Kit (BD Biosciences). In vitro apoptosis analysis was carried out with Annexin V-PI staining (ThermoFisher Scientific), and in situ cell apoptosis was determined with TUNEL staining (Abcam).

The sorted Ctsk^Cre;Chmp5^fl/fl and Ctsk^Cre;Chmp5^fl/+ periskeletal progenitors were subjected to RNA-seq. For library preparation, 100 ng of total RNA from each sample was subjected to rRNA depletion with the NEBNext rRNA Depletion Kit (New England Biolabs) according to the manufacturer’s manual. Subsequently, the rRNA-depleted RNA was used to build the RNA-seq library with the NEBNext Ultra II Directional RNA Library Prep Kit (New England Biolabs) according to the manufacturer’s manual. The index of each RNA-seq library was introduced using NEBNext Multiplex Oligos for Illumina (New England Biolabs). Each RNA-seq library was analyzed by the fragment analyzer (Advanced Analytical Technologies) and quantified by the KAPA Library Quantification Kit (KAPA Biosystems) according to the manufacturer’s manual. An equal amount of each RNA-seq library was mixed for sequencing in one lane of Illumina HiSeq3000 in paired-end 150 base mode (Illumina).

The raw Illumina pair-end reads were first aligned with ribosomal RNA (GenBank ID BK000964.1) with Bowtie2 v2.2.6 (Langmead and Salzberg, 2012). Reads that failed to map to ribosomal RNA were aligned with the mouse reference genome mm10 with STAR v2.5.3 (Dobin et al., 2013). Differential analysis was performed with Cuffdiff v2.2.1 (Trapnell et al., 2013). The raw data were analyzed by two independent statisticians and similar results were obtained.

Pathway analyses were performed for differentially expressed genes using the Reactome online analysis tool (Fabregat et al., 2017) and the Ingenuity Pathway Analysis software (Qiagen). Gene ontology term over-representation analysis was run using PANTHER (version 14.1) with all genes expressed in a tissue or cell type as background and the differentially expressed gene list as input. The P-values of Fisher’s exact tests were corrected by controlling the false discovery rate (FDR). Gene set enrichment analyses were performed using GSEA or ConsensusPathDB. For GSEA-based analysis, gene sets of interest were downloaded from the MsigDB, while all built-in gene sets were used for ConsensusPathDB-based analysis.

Label-free relative protein quantification analyses were performed to identify differentially expressed proteins in the supernatant or cell lysis of Ctsk^Cre;Chmp5^fl/fl versus control periskeletal progenitors (n=3 animals per group). For cell supernatants, the same amounts of samples went through 3 kD ultrafiltration tubes and then were processed for reduction and alkylation. Subsequently, the samples were digested in trypsin overnight, and the peptides were desalted prior to mass spectrometry analysis. For cell lysis, samples were precipitated using acetone after reduction and alkylation. The samples were then subjected to digestion, desalination, and mass spectrometry analysis.

Quantitative data were collected on an Orbitrap Eclipse Tribrid mass spectrometer (Thermo Fisher Scientific) coupled to a nanoLC system (EASY-nLC 1200, Thermo Fisher Scientific). All peptide mixtures were analyzed using the same chromatographic conditions: 5 μl of each sample were fractionated in a 150 μm×15 cm in-house made column packed with Acclaim PepMap RPLC C18 (1.9 μm, 100 Å, Dr. Maisch GmbH, Germany) working at a flow rate of 600 nl/min, using a linear gradient of eluent B (20–0.1% formic acid in water +80% acetonitrile) in A (0.1% formic acid in MilliQ water) from 4 to 10% for 5 min, 10 to 22% for 80 min, 22 to 40% for 25 min, 40 to 95% for 5 min, and 95 to 95% for 5 min. MS/MS analyses were performed using Data-Dependent Acquisition (DDA) mode: one MS scan (mass range from 350 to 1500 m/z) was followed by MS/MS scans up to the top 20 most intense peptide ions from the preview scan in the Orbitrap, applying a dynamic exclusion window of 120 s.

The raw MS files were analyzed and searched against the Mus_musculus_reviewed protein database using MaxQuant (1.6.2.10). The parameters were set as follows: the protein modifications were carbamidomethylation (C) (fixed), oxidation (M) (variable), acetyl (Protein N-term) (variable); the enzyme specificity was set to trypsin; the maximum missed cleavages were set at 2; the precursor ion mass tolerance was set at 20 ppm; and the MS/MS tolerance was 20 ppm. Only highly confident identified peptides were chosen for downstream protein identification analysis.

Western blot was performed using 4–20% Mini-PROTEAN TGX precast protein gels (Bio-Rad). The following antibodies: anti-CHMP5 polyclonal antibody (Shim et al., 2006), anti-COL1A1 monoclonal antibody (clone E8F4L, Cell Signaling Technology, CST), anti-TAGLN polyclonal antibody (cat#: AB14106, Abcam), anti-p16 INK4A monoclonal antibody (clone E5F3Y, CST), anti-p21 Waf1/Cip1 monoclonal antibody (clone E2R7A, CST), anti-VPS4A monoclonal antibody (clone A-11, Santa Cruz Technology), total OXPHOS antibody cocktail (Abcam), anti-GAPDH monoclonal antibody (clone 14C10, CST), and anti-β-Actin monoclonal antibody (clone D6A8, CST), were used.

Total RNA was isolated from cells using Trizol reagent (Qiagen). RNA samples were treated with the TURBO DNA-free Kit (Thermo Fisher) and equal amounts (1–2 μg) were used for reverse transcriptase reaction using an iScript Reverse Transcription Supermix (Bio-Rad). Quantitative PCR was run using the SYBR Green Master Mix on a CFX Connect Real-Time PCR Detection System (Bio-Rad). Gene expression levels were analyzed relative to the housekeeping gene Gapdh and presented by 2^-△Ct (Schmittgen and Livak, 2008). The primer sequences were used for qPCR are as follows: Chmp5, forward, 5’-ATGAGAGAGGGTCCTGCTAAG-3’, reverse, 5’-CCGTGGTCTTGGTGTCCTTTA-3’; H2afx, forward, 5’-GTGGTCTCTCAGCGTTGTTC-3’, reverse, 5’-CGGCC TACAGGGAACTGAA-3’; Tp53, forward, 5’-GTCACAGCACATGACGGAGG-3’, reverse, 5’-TCTTCCAGATGCTCGGGATAC-3’; Gapdh, forward 5′-TGCCAGCCTCGTCCCG TAGAC-3′, reverse 5′-CCTCACCCCATTTGATGTTAG-3′; VPS4A, forward 5′- AGAACCAGAGTGAGGGCAAGGG-3′, reverse 5′- GCACCCATCAGCTGTTCTTGCA-3′; CHMP5, forward 5′-CCAGCCTGACTGACTGCATTGG-3′, reverse 5′- TCGCAAGGCTTTCTGCTTGACC-3′; GAPDH, forward 5′- GGAGTCCACTGGCGTCTTCAC-3′, reverse 5′-GAGGCATTGCTGATGATCTTGAGG-3′.

The concentration and size distribution of extracellular vesicles were analyzed using the NanoSight NS300 following the manufacturer’s protocol (Malvern Instruments). Briefly, cells were seeded in 24-well plates at the density of 1×10⁵ cells per well and cultured overnight. The next morning, cells were changed to serum-free medium and incubated for 9 hr. The medium was collected and subjected to sequential centrifugation at 300 g for 10 min, 2000 g for 20 min, and 10,000 g for 30 min. The resulting supernatant was manually injected into the instrument and run in the ‘standard measurement’ module with five captures per sample and the data were processed with a detection threshold of 2.

The wild-type and Ctsk^Cre;Chmp5^fl/fl periskeletal progenitors were incubated on ice for 10 min before adding fresh medium with 2 ug/ml pHrodo red EGF conjugate (Thermo Fisher). Cells were incubated at 37 °C for 30 min and subsequently washed and changed to fresh medium. Live-cell images were obtained at 1, 12, and 24 hr using the Leica TCS SP5 II laser scanning confocal microscope.

Wild-type and Ctsk^Cre;Chmp5^fl/fl periskeletal progenitors were fixed in culture plates overnight at 4 °C using 2.5% glutaraldehyde in 0.1 M Na cacodylate-HCl buffer (pH 7.2). After washing in 0.5 M Na cacodylate-HCl buffer (pH 7.0), the cells were post-fixed for 1 hr in 1% osmium tetroxide (w/v) at room temperature. After post-fixation, the culture plate with adherent cells was en bloc stained (20 min) with 1% aqueous uranyl-acetate (w/v). Culture plates were washed again in the same buffer and dehydrated through a series of graded ethanol to 100% and transferred through two changes of 50/50 (v/v) SPIpon resin (Structure Probe, Inc)/100% ethanol and left overnight to infiltrate. The following morning, the cell culture plates were transferred through three changes of fresh SPIpon resin to finish the infiltration and embedding and finally, the plates were filled with freshly prepared SPIpon resin and polymerized for 48 hr at 70 °C.

Once fully polymerized, the plate was cut apart, and each well was plunged into liquid nitrogen to separate the SPIpon epoxy block with the embedded cells from the culture dish. The round epoxy disks with the embedded cells were then examined under an upright light microscope, and areas of cells were cut from the disks and glued onto blank microtome studs and trimmed for ultramicrotomy. Ultrathin sections (70 nm) were cut on a Reichert-Jung ultramicrotome using a diamond knife. The sections were collected and mounted on copper support grids and contrasted with lead citrate and uranyl acetate, and examined on a FEI Tecnai G2 Spirit transmission electron microscope at 100 kV accelerating voltage and images were recorded at various magnifications using a Gatan 2 K digital camera system.

Mitochondrial respiration was examined in Ctsk^Cre;Chmp5^fl/fl and wild-type periskeletal progenitors using a Seahorse XF^e96 Extracellular Flux Analyzer (Agilent) as previously described (Fiorese et al., 2016). Briefly, 10,000 cells per well were seeded in XF 96-well culture plates and cultured overnight. The next day, cells were changed to assay medium and incubated at 37 °C w/o CO2 for 1 hr. The assay was carried out using the Seahorse XF Cell Mito Stress Test kit with sequential injections of Oligomycin (1  µM), FCCP (1  µM), and Rotenone plus Antimycin A (0.5  µM) following the manufacturer’s protocol.

For in vitro cell treatment, Ctsk^Cre;Chmp5^fl/fl and control cells were seeded in 96-well plates at a density of 6000 cells/ well and cultured overnight. Quercetin and dasatinib were added to the culture medium at a final concentration of 50 μM and 500 nM, respectively. Apoptotic cells are labeled using Incucyte Annexin V Red Dye and are quantified in real time using the Incucyte Live-Cell Analysis System. Six replicates were used for each group and the experiment was repeated twice using cells from two different animals per group.

For in vivo treatment, the senolytic drugs quercetin and dasatinib were administered intraperitoneally weekly at 50 µg/g and 5 µg/g body weight, respectively. Senolytic treatment in Ctsk^Cre;Chmp5^fl/fl and littermate control mice started from the first week after birth, and mice were collected at 7–8 wk of age for skeletal analyses. For the Dmp1^Cre;Chmp5^fl/fl strain, treatment began the second week after birth, and the animals were monitored for 16 wk. ABT-263 was administered intraperitoneally to Ctsk^Cre;Chmp5^fl/fl and littermate control mice weekly at 10 μg/g body weight, starting from the first week after birth, and mice were collected at 8 wk of age for micro-CT analyses. In these in vivo studies, mice were randomly assigned to receive the corresponding senolytic treatment and evaluated blindly to experimental conditions. There was no attrition of animals during these experimental procedures.