Involvement of serum‐derived exosomes of elderly patients with bone loss in failure of bone remodeling via alteration of exosomal bone‐related proteins

Summary Exosomes are secreted into the blood by various types of cells. These extracellular vesicles are involved in the contribution of exosomal proteins to osteoblastic or osteoclastic regulatory networks during the failure of bone remodeling, which results in age‐related bone loss. However, the molecular changes in serum‐derived exosomes (SDEs) from aged patients with low bone density and their functions in bone remodeling remain to be fully elucidated. We present a quantitative proteomics analysis of exosomes purified from the serum of the elderly patients with osteoporosis/osteopenia and normal volunteers; these data are available via Proteome Xchange with the identifier PXD006463. Overall, 1,371 proteins were identified with an overlap of 1,160 Gene IDs among the ExoCarta proteins. Bioinformatics analysis and in vitro studies suggested that protein changes in SDEs of osteoporosis patients are not only involved in suppressing the integrin‐mediated mechanosensation and activation of osteoblastic cells, but also trigger the differentiation and resorption of osteoclasts. In contrast, the main changes in SDEs of osteopenia patients facilitated both activation of osteoclasts and formation of new bone mass, which could result in a compensatory elevation in bone remodeling. While the SDEs from aged normal volunteers might play a protective role in bone health through facilitating adhesion of bone cells and suppressing aging‐associated oxidative stress. This information will be helpful in elucidating the pathophysiological functions of SDEs and aid in the development of senile osteoporosis diagnostics and therapeutics.

of bone remodeling, which involves renewal of aged bone and repair of skeletal microdamage through processes that include enhanced osteoclast activity or decreased bone formation from osteoblast lineage cells (Henriksen, Karsdal & Martin, 2014).
Normal bone remodeling is activated by osteoclasts that are unique in their function of bone resorption and generates identifiable scalloped lacunae, followed by a constructive process in which new bone is generated by osteoblasts (Henriksen et al., 2014). The coordinated regulation of these important cell types is critical for maintaining physiological bone remodeling, which is tightly controlled by physical cell-cell interactions, secretory signals, and the endocrine system (Sims & Walsh, 2012). Osteoclast activation occurs after binding of receptor activator of nuclear factor jB (RANKL) to its receptor RANK, which is expressed in the membrane of osteoclast precursors. The signals that positively regulate osteoblast differentiation include members of the integrins, Wnts, and transforming growth factor beta (TGFb) (Matsuo, 2009). Additionally, recent studies have revealed that various key factors involved in bone remodeling are packaged in spherical bilayered membrane vesicles called exosomes. These organelles function as cell-cell communicators by transferring biologically active molecules to adjacent or distant cells (Xie, Chen, Zhang, Ge & Tang, 2017).
Various cell types secrete exosomes, which are enriched with the markers CD63 and ALG-2-interacting protein X (ALIX) (Choi et al., 2012). These organelles are formed solely within the endosomal network and released following fusion of multivesicular bodies with the plasma membrane (Yanez-Mo et al., 2015). With an average diameter of 40-150 nm, exosomes are released into the circulation and transfer the biologically active molecules contained within their lumen to target cells (Li, Kaslan, Lee, Yao & Gao, 2017). Therefore, rather than representing simple cellular debris, exosomes function as extracellular organelles with local or distant roles in intercellular signaling and communication (Hong et al., 2009). Recent reports indicate the involvement of bone-associated exosomes in regulating bone remodeling (Li et al., 2016), mainly via the transfer of critical molecules required for the regulation of osteoclasts and osteoblasts (Hao et al., 2017). However, the comprehensive changes among the proteins in serum-derived exosomes (SDEs) of aged patients with osteoporosis or osteopenia and their functions in bone remodeling remain largely unclear (Colombo, Raposo & Thery, 2014).
Here, to determine the biological functions of SDEs in osteoporosis and osteopenia, we compared the proteomic profiles of exosomes purified from the serum of elderly patients with osteoporosis and low bone mass with those of aged and young normal volunteers using a combination of liquid chromatography-tandem mass spectroscopy (LC-MS/MS) analyses and tandem mass tag (TMT) labelbased quantitation. Furthermore, we analyzed the phenotypes of osteoclast precursor peripheral blood mononuclear cells (PBMCs), RAW 264.7 cells or osteoblast precursor hFOB 1.19 cells, and MC3T3-E1 cells following differentiation into mature osteoclasts or osteoblasts in the presence or absence of SDEs. Our study provides evidence that an understanding of the specific cargo of SDEs from patients might be useful for the early evaluation of senile osteoporosis and the development of novel diagnostics and therapeutics.

| Efficiency of SDE extraction and quantitative characteristics of protein profiles
We compared protein abundances in exosomes purified from serum samples of patients with osteoporosis, osteopenia, and normal volunteers using TMT-based quantitative MS (Figure 1). To confirm the exosome fractions and validate the exosomal origin of the identified proteins, we compared the expression levels of exosome markers CD63 and ALIX between exosome pellets and supernatants of pooled serum samples after ultracentrifugation (Figure 2a). The raw spectral data were interpreted using Proteome Discoverer 2.1. A total of 1,371 proteins (Score SEQUEST HT >0) were identified (1,350 of 1,371 identifiers from UniProtKB AC/ID were successfully mapped to 1,391 Entrez Gene IDs using the UniProt mapping tool). More than 50% of the identified proteins (679/1,371) exhibited Score SEQUEST HT >10 (Table S1). There was an overlap of 1,160 Entrez Gene IDs in the Exo-Carta protein list (http://www.exocarta.org, release date: July 29, 2015) ( Figure 2b). Gene Ontology (GO) classification indicated that most of the identified exosomal proteins were derived from cytoplasm and important for molecular functions of receptor activity, cell adhesion molecule activity, and GTPase activity. These proteins were also involved in the biological processes of signal transduction, cell communication, and protein metabolism, which is consistent with the reported functions of exosomes (Figure 2c-e). These data indicated that the purification of exosomes from serum was successful, and the proteins were derived from the SDEs.
The NanoSight reports demonstrated that there were no significant differences between the sizes of SDEs among the different groups. However, there were higher concentrations of SDEs in the Aged normal group compared with those in the Young normal group ( Figure S1). Based on the differential expression thresholds (0.5-fold and twofold change for down-and upregulation of expression, respectively), 585 osteoporosis differentially expressed proteins (DEPs, 131/127) and 116 osteopenia DEPs (130/127) were filtered from the results. Among 585 DEPs identified in SDEs isolated from osteoporosis patients relative to those from the aged normal volunteers, 225 proteins were upregulated and 360 proteins were downregulated. In contrast, among the 116 osteopenia DEPs, 110 proteins were upregulated and only six were downregulated. In addition, 59 aged normal DEPs (127/126) were identified. The relative abundances of these proteins are listed in Table S2. 2.2 | Different expression trends and hub regulators among exosomal proteins in the osteoporosis and osteopenia groups To reveal the differences in expression levels among exosomal proteins in the osteoporosis and osteopenia groups, heatmaps were constructed using Hierarchical Clustering Explorer 3.5 (https:// www.cs.umd.edu/hcil/hce). The clusters were identified, and pathway analysis was performed for the proteins in each cluster using  (Table S3), which are implicated in mechanosensation and signal transduction of cell regulation. However, in Cluster 2 (Figure 3a), the decreasing trends in the expression of identified proteins involved in TGF-b signaling pathways (Table S3) were similar in the osteopenia and osteoporosis groups, although the levels were relatively lower in the osteoporosis group.
In contrast, there was no marked difference in the upregulation of identified osteopenia DEPs compared with the expression levels of osteoporosis DEPs in Clusters 3 and 4 ( Figure 3a). Additional pathway analysis revealed that these osteoporosis DEPs were most highly enriched in the translation factors and selenium micronutrient network categories (Table S3). However, the increasing trends of expression levels of proteins identified from the osteoporosis group were similar to the osteopenia DEPs in Clusters 1 and 3 ( Figure 3c).
The mapped results indicated that these osteopenia DEPs were predominantly involved in the cytoplasmic ribosomal protein pathways (Table S4), which are implicated in cell differentiation and highly dynamic metabolic processes. These findings indicated that the specific contents of SDEs perform multiple functions in bone remodeling processes according to their origin and bone health status.
To reveal the potential interactions among the DEPs, comprehensive interaction networks were constructed using the web-based STRING tools (http://string-db.org). In the network of the osteoporosis DEPs (Figure 3b), the downregulated proteins ITGb 1 , ITGb 3 , and hematopoietic progenitor cell antigen CD34 (CD34) represented hubs at which the protein interactions converged. These proteins potentially regulate more than six of the DEPs, most of which are F I G U R E 1 Experimental workflow. "OC" refers to osteoclast, "NTA" refers to nanoparticle tracking analysis XIE ET AL.  The six most enriched categories and the enrichment significance (Àlog (p-value), p < .05) of identified proteins in cellular components (c), molecular functions (d), and biological process (e) categories. The percentage of proteins identified in each category is indicated. "S1" refers to pooled serum from Aged normal volunteers, "S2" refers to pooled serum from Aged osteopenia patients, "S3" refers to pooled serum from Aged osteoporosis patients F I G U R E 3 Bioinformatics analysis of DEPs from SDEs based on STRING and Wiki pathway databases. The heatmap was constructed into clusters by Hierarchical Clustering Explorer 3.5 and shows contrasting or similar expression levels of osteoporosis-associated SDE DEPs compared with those identified from SDEs of osteopenia patients (MS data presented as the ratios to 131/127 and 130/127 were input into Hierarchical Clustering Explorer 3.5 with Ln transformation). The DEPs in the protein-protein interaction networks are shown as nodes (MS data presented as the ratios to 131/127 and 130/127 were matched to STRING networks with log 2 transformation). Up-or downregulation of identified proteins is indicated by colors either in the heatmap or networks (upregulated in red, downregulated in green). All identified proteins were mapped to the relevant Wiki pathway database. Proteins are represented by boxes labeled with the protein name. Relative protein expression level in SDEs of osteoporosis patients is indicated by colors (MS data presented as the ratios to 131/127 and 130/127 were matched to STRING networks with log2 transformation). Proteins in gray were not identified in this study.   Table S5). Among the osteoporosis DEPs, the upregulated proteins were predominantly enriched in pathways associated with cytoplasmic ribosomal proteins (Table S5A), which are associated with the enhancement of protein metabolism and inflammatory responses. In contrast, 199 of 257 exosomal DEPs derived from plasma membrane were downregulated in the Aged osteoporosis group (Table S3). These DEPs were mapped predominantly to integrin-mediated cell adhesion, focal adhesion, and G protein signaling pathways, which are important in cell adhesion, mechanosensation, and activation of osteoblasts (Table S5B). In addition, pathway enrichment analysis indicated that several aged normal proteins play pivotal roles in the regulation of bone remodeling by suppressing selenium-associated oxidative stress, such as intercellular adhesion molecule 1 (ICAM1), serum amyloid A-1 protein (SAA1), and peroxiredoxin 2 (PRDX2) ( Table S5D). To gain further insights, we performed analysis using Cytoscape (3.2.1) software based on the Wiki pathway database. As shown in Figure 3e and Figure  The biogenesis of ribosomes is a highly complex and energy-consuming process that is initiated in the nucleolus (Goudarzi & Lindstrom, 2016), and cytoplasmic ribosomal proteins are upregulated by RANKL-induced differentiation of osteoclasts. The pathway analysis showed a significant increase in ribosomal proteins ( Figure S5A).
Guanine nucleotide-binding proteins (G proteins) principally downregulated in SDEs of osteoporosis patients ( Figure S3B) were found to be involved in skeletal system development. Guanine nucleotidebinding protein G(s) subunit alpha isoforms short (GNAS) is important for osteoblast formation (Hsiao, Millard & Nissenson, 2016), while Kras, a downstream protein of GNAQ, is required for the control of local recruitment of osteoclasts. These findings suggested that the roles of the TGF-b pathways and G proteins in SDEs of osteoporosis patients are associated with the control of mechanosensation and activation of bone remodeling.
To identify the bone-related physiological processes implicated by the exosomal proteins, we next clustered overall DEPs (Score SEQUEST HT > 0) into GO categories using bioinformatic tools. In terms of cellular components ( Figure S4A), the upregulated osteoporosis DEPs were predominantly enriched in the cytoplasm category, while the downregulated proteins were most likely to be derived from the plasma membrane ( Figure S4D). The upregulated osteoporosis DEPs were involved in molecular functions such as structural constituents of ribosomes, protease inhibitor, and chaperone activity ( Figure S4B), which are important for the biological processes of protein metabolism and cell growth and/or maintenance ( Figure S4C). In contrast, proteins related to signal transduction, cell communication, and transport were downregulated ( Figure S4F) and were involved in the molecular functions GTPase activity, receptor activity, and cell adhesion molecule activity ( Figure S4E). The DEPs in the osteopenia group had varied characteristics in terms of GO enrichment ( Figure S5A-C). The altered proteins from SDEs of the osteopenia group showed a significant enrichment in nucleolus categories. Molecular function analysis revealed that these proteins were mainly responsible for structural constituents of ribosomes and the extracellular matrix and RNA binding, which are important for protein metabolism and cell growth and/or maintenance. These DEPs in the Aged normal group were found to be important for the biological processes of immune responses and cell adhesion, which are predominantly responsible for aging-related inflammation ( Figure S5D-F).

| Effects of SDEs on osteoclast differentiation and bone resorption
We used human PBMCs derived osteoclasts as a model. The number of TRAP-positive multinucleated osteoclasts cultured with SDEs from osteoporosis or osteopenia patients was significantly higher than the number of osteoclasts cultured with SDEs from aged normal volunteers. Also, the area of rearrangement of TRAP-positive osteoclasts cultured with SDEs from osteoporosis or osteopenia patients was significantly larger than that of osteoclasts cultured with SDEs from aged normal volunteers. However, there were no significant differences in the number and area of osteoclasts treated with either SDEs from young volunteers or SDEs from elderly normal volunteers. In addition, the bone resorption of osteoclasts cultured with SDEs from osteoporosis or osteopenia patients was significantly stronger than the resorption of osteoclasts cultured with SDEs from aged normal volunteers. We also found that SDEs derived from the elderly normal group had a slightly stronger effect on osteoclast resorption ability than the SDEs derived from younger volunteers. However, the promotion of osteoclast bone resorption by SDEs in the elderly normal group was much weaker than that induced by SDEs in the osteoporosis or osteopenia groups ( into osteoblasts (Kuroda et al., 2014). Furthermore, the integrin b 1 / Shc association leads to the activation of extracellular signal-regulated kinase (ERK), which is critical for shear induction of bone formation-related genes in osteoblast-like cells (Lee et al., 2008).
Integrin-linked protein kinase (ILK) might regulate ERK signaling LTBP1 is involved in the assembly, secretion, and targeting of TGFb 1 to sites at which it is stored and/or activated. In addition, LOF mutations of STAT3 decrease the osteogenic response following mechanical loading (Li, 2013). Thus, downregulated TGF-b signaling may lead to loss of skeletal integrity via enhancement of osteoclast bone resorption and interrupted bone formation. However, numbers of cytoplasmic ribosomal proteins were increased in SDEs of osteoporosis patients. Ribosomes are cellular machines that are essential for protein synthesis (Trainor & Merrill, 2014). The biogenesis of ribosomes is a highly complex and energy-consuming process that is initiated in the nucleolus, and cytoplasmic ribosomal proteins are upregulated by RANKL-induced differentiation of osteoclasts (Day et al., 2004). From this, we propose that SDEs of osteoporosis patients not only suppress the mechanosensation response of target cells, but also trigger signaling pathways for elevation of osteoclastic differentiation and exacerbate abnormal bone remodeling.  (Li, Liu, Zhang & Rong, 2014). All osteoclast nuclei are transcriptionally active, and the functional role is reflected by the expression of regulatory proteins that support ribosomal RNA synthesis, such as NCL, which is thought to play a role in pre-rRNA transcription and ribosome assembly (Saltman et al., 2005). Therefore, the upregulated cytoplasmic ribosomal proteins in SDEs of both osteoporosis and osteopenia patients may function as positive feedback factors in the activation of osteoclasts.

| Exosomes in the circulation of patients with osteopenia could result in a compensatory elevation of bone remodeling
In contrast to SDEs of osteoporosis patients, those of osteopenia patients promote osteoblastic bone formation. Our analysis indicates that versican core protein (VCAN) and connective tissue growth factor (CTGF) are upregulated in SDEs of osteopenia patients and act F I G U R E 4 SDEs regulate osteoclast differentiation and bone resorption. (a) The number and area of rearrangement of TRAP-positive multinucleated osteoclasts cultured with SDEs from osteoporosis or osteopenia patients were significantly higher than that of TRAP-positive multinucleated osteoclasts cultured with SDEs from aged normal volunteers. Moreover, the bone resorption of osteoclasts cultured with SDEs from osteoporosis or osteopenia patients was significantly greater than the resorption of osteoclasts cultured with aged normal volunteers. In addition, the bone resorption of osteoclasts cultured with SDEs from aged normal volunteers was significantly greater than the resorption of osteoclasts cultured with SDEs from young normal volunteers. However, there was no significant difference in the number and area of rearrangement of between the TRAP-positive osteoclasts treated with either SDEs from young or aged normal volunteers. (b) RAW264.7 cells cultured with RANKL had a similar effect on osteoclast differentiation and bone resorption compared with the effects of SDEs from different groups; however, the bone resorption of osteoclasts cultured with SDEs from aged normal volunteers was no difference comparing to the resorption of osteoclasts cultured with SDEs from young normal volunteers. Representative photographs are shown in the up panel.
Quantification of cells is shown in the down panel. All values are representative of at least two independent experiments with similar results and are displayed as mean AE SD. **p < .01, *p < .05. "OC" refers to osteoclast as hubs in regulating the function of osteoblasts. VCAN is an ECM component that plays an essential role in transformed cell behavior.
Increases in VCAN have been confirmed in human MSCs undergoing osteoblast differentiation (Foster et al., 2005). CTGF serves as an adhesive substrate for promoting cell spreading via cytoskeletal reorganization as well as enhancing osteoblast adhesion via integrin a v b 1 . Consistently, integrin-mechanosensation-related proteins were slightly upregulated (relative abundance of proteins as ratio of 130/ 127 > 1 but <2). Additionally, fibrillin-1 (FBN1) was constitutively upregulated in SDEs of osteopenia patients. FBN1 mediates cell adhesion via its binding to cell surface receptors integrin a v b 1 and selectively blunts expression of osterix, which is a transcriptional regulator of osteoblast maturation (Nistala et al., 2010). In terms of the adhesion and activation of osteoclasts that are accompanied by (a)

(b)
F I G U R E 5 SDEs regulate bone formation by osteoblasts. (a) In the human osteoblastic cell line Hob (hFOB 1.19) cultured on day 21, the alkaline phosphatase (ALP) levels confirmed that higher levels of mineralization of cells cultured with SDEs from osteopenia patients than that of cells cultured with SDEs from aged normal volunteers. However, the level of mineralization in hFOB 1.19 cells cultured with SDEs from osteoporosis patients was significantly lower than that of cells cultured with SDEs from aged normal volunteers. In addition, the mineralization level of cells cultured with SDEs from aged normal was higher than that of cells cultured with SDEs from young normal volunteers. (b) At 15 days after induction of MC3T3-E1 cells by osteogenic media containing vitamin C and b-glycerol phosphate, mineralization nodules were first observed in the group cultured in the presence of SDEs from osteoporosis patients. Three days later, mineralization nodules were observed in all the groups, and the OD values of Alizarin Red staining in the Aged osteoporosis and osteopenia groups were higher than those of the Aged normal group. In the later stages of matrix mineralization on day 21, the OD value in the Aged osteoporosis group was lower than those in both the Aged normal and osteopenia groups. Moreover, the OD value in the Aged osteopenia group was higher than that in the Aged normal group. The OD value in the Young normal group was lower than that in the Aged normal group. The OD values of Alizarin Red staining in all these four groups were in accordance with the ALP levels detected in human osteoblastic cells. Representative images are shown in the left panel. Quantification of cells is shown in the right panel. All values are representative of at least two independent experiments with similar results and are displayed as mean AE SD. **p < .01, *p < .05 elevated bone formation, we believe that these DEPs from exosomes in the circulation of patients with osteopenia are involved in a compensated elevation of bone remodeling, which could result in a relatively low level of bone destruction and bone mass reduction.

| SDEs from aged normal volunteers involved in
cell adhesion and oxidative stress might function as protective regulators of bone health and the aging process The aging skeletal system undergoes a progressive decline in the renewal of bone tissue; therefore, we investigated the changes in protein expression of SDEs from elderly volunteers without bone loss compared with those from young normal volunteers. The 59 aged normal DEPs were found to be important for the biological processes of immune responses, which are predominantly responsible for alleviation of aging-related inflammation ( Figure S5F).
Pathway enrichment analysis indicated that several selenium-associated proteins play pivotal roles in the regulation of bone remodeling by inhibiting oxidative stress. Selenium is beneficial to total BMD (Beukhof et al., 2016). Selenite protects MSCs against H 2 O 2 -induced inhibition of osteoblast differentiation through suppressing oxidative stress and ERK activation (Liu, Bian, Liu & Huang, 2012). Furthermore, selenium deficiency is detrimental to bone microarchitecture by increasing bone resorption, possibly through decreasing antioxidative potential (Cao, Gregoire & Zeng, 2012). The representative proteins, such as ICAM1 and SAA1, were upregulated in the SDEs of elderly volunteers recruited in this study, while PRDX2 was downregulated. ICAM1 provides high-affinity adhesion between osteoblast and osteoclast precursors, thereby facilitating osteoclast activity and inhibiting bone formation (Cheung, Simmons & You, 2012). SAA1, a major acute phase protein, stimulates MMP9 production for bone The data also demonstrated that there was no significant difference between the male and female patients in each group. (c) Analysis of the correlation between the expression level of ITGB1 and ITGB3 in SDEs and BMD. "BMD" refers to bone mass density resorption (Paret, Schon, Szponar & Kovacs, 2010). In contrast, peroxiredoxin 2 (PRDX2), a member of the antioxidant enzyme family, is protective against oxidative damage during osteoclastogenesis (Park et al., 2015).
As shown in the heatmap ( Figure S7C), 33 overlapping DEPs were identified in the Aged normal and Aged osteoporosis groups.
Of these, 27 were highly expressed in the Aged normal group but were expressed at low levels in the Aged osteoporosis group (Table S2). Among these proteins, we found that CD44, ICAM, and ITGA1 play pivotal roles in resisting age-related bone loss via mediation of cell-cell and cell-matrix adhesion (Zu, Liang, Du, Zhou & Yang, 2015). However, owing to the small number of overlapping DEPs (both <10%; 33 of 574 in the osteoporosis group and nine of 112 in the osteopenia group. Figure Table S6).
For individual validation, 72 additional serum samples (8-10 males and 8-10 females in each group) were collected (information shown in Table S7). All these serum samples were collected at the same time of day and under fasting conditions. All the patients and volunteers involved in the study were without diabetes, thyroid diseases, autoimmune diseases, or tumors and had not received treatment for osteopenia or osteoporosis prior to this study; their information was obtained from the PLAGH. BMD-based bone assessments were measured by DXA (OsteoSys EXA 3000 Bone Density Device) at the one-third radius site. In postmenopausal women and men aged ≥50, osteoporosis and osteopenia were diagnosed according to the application of the WHO diagnostic T-score criteria to BMD measurements (with T-scores at À1.0 or above defined as "normal"; between À1.0 and À2.5 defined as "osteopenia"; and at À2.5 or below defined as "osteoporosis"). However, in premenopausal women, and men <50 years of age, the ethnicity or race adjusted Z-scores were recommended by the International Society for Clinical Densitometry (ISCD) instead of T-scores in the diagnostic classification (with Zscores of À2.0 or lower defined as either "low bone mineral density for chronological age" or "below the expected range for age" and those above À2.0 defined as "within the expected range for age").
Informed consent was obtained from all patients and normal volunteers. The study was performed with the approval of the Ethics Committee of the PLAGH.

| Isolation of exosomes from human serum
Exosomes were isolated from pooled serum samples belonging to different groups by ultracentrifugation and using the Total Exosome Isolation Reagent as previously reported (Chen et al., 2017). In brief, serum was first diluted with an equal volume of PBS. For ultracentrifugation, the diluted serum was centrifuged at 10,000 9 g for 30 min at 4°C followed by ultracentrifugation at 110,000 9 g at 4°C for 90 min using a Beckman Optima L-100XP Ultracentrifuge. The pellet was washed with 1 ml PBS and then dissolved with 40 ll 8 M urea (Kim, Tan & Lubman, 2015
The differential expression threshold was defined as a twofold change for both down-and upregulation of expression. The mass spectrometry proteomics data have been deposited to the Pro-teomeXchange Consortium via the PRIDE (Vizcaino et al., 2016) partner repository with the dataset identifier PXD006463. GO analysis was performed with FunRich software (2.1.2) (Pathan et al., 2015). The Entrez Gene IDs retrieved from UniProtKB accession numbers were mapped to cellular components (CC), molecular functions (MF), and biological processes (BP) items using default statistical parameters (threshold: count 2, ease 0.1). The UniProtKB accession numbers were uploaded directly using the WEB-based GEne SeT AnaLysis Toolkit (http://bioinfo.vanderbilt.edu/webgesta lt/) for pathway mapping. Pathways with at least three target genes and a p-value of <.05 were considered to be statistically significant.
For Hierarchical clustering analysis, the MS proteomics data presented as the ratios to 131/127 and 130/127 were input into Hierarchical Clustering Explorer 3.5 with log transformation (natural), and a heatmap was then constructed using the complete linkage method.
At the beginning of the process, each element is in a cluster of its own. The clusters are then sequentially combined into larger clusters until all elements are combined as a single cluster. In each step, the two clusters separated by the shortest distance are combined (Brian All STRING network analyses were performed using UniProt accession numbers as the input and with "Experimental" and "Database" evidence at medium (0.4) confidence level. The networks were downloaded as tab-delimited text files, and visualized, and reorganized using Cytoscape (3.2.1) software. In addition, the MS proteomics data presented as the ratios to 131/127 and 130/127 were matched to these STRING networks with log 2 transformation.

| Statistical analysis
Data are expressed as the mean AE SD of triplicate experiments, unless otherwise indicated. Statistical differences among groups were analyzed by one-way analysis of variance with a post hoc test to determine group differences in the study parameters. All statistical analyses were performed with SPSS software (version 24.0 for Windows; Armonk, NY, USA) and Prism software (GraphPad prism for Windows, version 6.01; Nashville, TN, USA). Unless otherwise indicated, data were analyzed using Student's t test, and p < .05 was considered to indicate statistical significance.

ACKNOWLEDG MENTS
We thank the technical staff from Institute for Advancing Translational Medicine in Bone & Joint Diseases, Hong Kong Baptist University, for providing critical comments and technical support.

CONFLI CT OF INTEREST
The authors report no conflict of interest relevant to this article.