Human bone marrow stem/stromal cell osteogenesis is regulated via mechanically activated osteocyte‐derived extracellular vesicles

Abstract Bone formation or regeneration requires the recruitment, proliferation, and osteogenic differentiation of stem/stromal progenitor cells. A potent stimulus driving this process is mechanical loading. Osteocytes are mechanosensitive cells that play fundamental roles in coordinating loading‐induced bone formation via the secretion of paracrine factors. However, the exact mechanisms by which osteocytes relay mechanical signals to these progenitor cells are poorly understood. Therefore, this study aimed to demonstrate the potency of the mechanically stimulated osteocyte secretome in driving human bone marrow stem/stromal cell (hMSC) recruitment and differentiation, and characterize the secretome to identify potential factors regulating stem cell behavior and bone mechanobiology. We demonstrate that osteocytes subjected to fluid shear secrete a distinct collection of factors that significantly enhance hMSC recruitment and osteogenesis and demonstrate the key role of extracellular vesicles (EVs) in driving these effects. This demonstrates the pro‐osteogenic potential of osteocyte‐derived mechanically activated extracellular vesicles, which have great potential as a cell‐free therapy to enhance bone regeneration and repair in diseases such as osteoporosis.


| INTRODUCTION
Osteocytes are the most abundant cell type in bone and are known as the primary sensing and metabolism-controlling cells within the tissue.
Osteocytes are key to directing the processes of bone formation and resorption via the secretion of various signaling factors which act upon bone forming osteoblasts and resorbing osteoclasts and their progenitors, skeletal, and hematopoietic stem cells. 1 The implications of this can be seen in the highly debilitating and life-threatening disease that is osteoporosis, which has been linked to osteocyte apoptosis 2 and reduced osteocyte numbers in affected patients. 3 This results in a significant drop in quality of life, increased risk of additional complications due to immobilization, and significantly increased mortality rates due to fracture and secondary causes. 4 Not only do osteocytes have key functions in bone, but they have also been shown to be involved in a large range of other major functions throughout the body, 5 including heart, muscle and liver function, and suppressing breast cancer growth and metastasis in bone. 6 This highlights the critical role of the osteocyte in human health, and the importance of better understanding osteocyte signaling factors for the development of therapeutics to treat orthopedic and systemic diseases.
A prime example of osteocyte sensing and coordination of bone physiology is in mechanoadaptation, with mechanical loading leading to enhanced bone formation and unloading leading to bone loss. 7 In response to macroscale deformation of bone, resident osteocytes sense the micromechanical environment consisting of oscillatory fluid flow-induced shear stress and relay this biophysical signal to effector cells. 8 Mechanically stimulated osteocytes can enhance the bone forming capacity of osteoblasts via direct cell-cell contact, 9 in addition to secreted factors as demonstrated by conditioned media experiments. 10,11 Furthermore, this same mechanically activated osteocyte conditioned media was also shown to inhibit osteoclast formation. 12,13 Due to the nonproliferative state and short life span of mature bone cells, continuous bone formation requires the replenishment of the exhausted osteoblast from a stem cell or progenitor population. 14 Interestingly, the osteocyte has also been shown to coordinate bone marrow stem/stromal cell (MSC) behavior, with conditioned media from mechanically stimulated osteocytes enhancing stem cell proliferation, recruitment, and osteogenic differentiation, demonstrating the far reaching influence of this cell type, particularly in response to a mechanical stimulus. 10 The means by which osteocytes coordinate this mechanoadaptation of bone is of great interest, with several key factors identified as playing a role in this regard and, therefore, targeted as therapeutics.
There has been a plethora of studies investigating various osteocytederived factors released in response to fluid shear, including nitric oxide (NO), prostaglandin E2, ATP, RANKL, osteoprotegerin (OPG), and macrophage colony-stimulating factor (M-CSF). 1 One factor that has gained much interest is sclerostin (SOST) which is released by osteocytes and inhibits Wnt-mediated bone formation. SOST expression is inhibited following mechanical loading and inhibition of this protein via anti-sclerostin therapy has been shown in clinical trials to increase bone mineral density and reduce fracture risk. 15 To gain a greater understanding of the factors expressed by physically stimulated osteocytes, others have taken a more global approach, using microarrays to study global gene expression in osteocytes subjected to cyclic compressive forces 16 and osteocytes isolated from murine trabecular bone following vertebrae loading. 17 Furthermore, a proteomic analysis has been combined with a transcriptomic analysis of osteocytes subjected to fluid shear to investigate protein as well as gene expression information and reveal novel interactions between them. 18 These studies revealed the altered proteome of the osteocyte, due to fluid flow stimulation, and identified a range of proteins which may be involved in mechanotransduction, including nucleoside diphosphate kinase and calcyclin, which are of interest due to their roles in ATP and calcium-binding, respectively. However, to date, the full secretome protein signature of the osteocyte and how this is altered in response to mechanical stimulation is unknown.
A route of cell-cell communication which has garnered much attention of late is via extracellular vesicles (EVs). EVs are spherical proteolipids bilayer surrounded vesicles secreted from cells and are involved in cell-cell communication. EVs can transfer cargo including lipids, proteins, and nucleic acid from one cell to another, thereby influencing the recipient cell function. 19 Interestingly, it has recently been shown that bone cells release EVs and use these vesicles as a mechanism to mediate osteoblast and stem cell osteogenesis. [20][21][22][23][24] Moreover, osteocyte-derived EVs contain miRNAs known to mediate osteoblast function, highlighting a potential nonprotein based role in bone cell communication. 25,26 Bone derived EVs may also be exploited as a potential therapy for various diseases, as well as having potential for treatment of critical size bone defects. 27 In addition, the capability to load EVs with factors to guide cell behavior both in vitro and in vivo has been shown, 28 supporting their potential as a powerful drug delivery method. Interestingly, the release of EVs-and thus their content-may also be altered by mechanical loading. In fact, EV release into plasma increases following exercise, with a differential

Significance statement
Bone regeneration requires the osteogenesis of stem/stromal progenitor cells. A potent stimulus driving this process is physical loading. Osteocytes are mechanosensitive cells which play fundamental roles in coordinating bone mechanoadaptation. However, the exact mechanisms are poorly understood. This study demonstrates that osteocytes subjected to fluid shear secrete a distinct collection of factors that enhance stem cell recruitment and osteogenesis.
Moreover, this study identified that these factors are delivered via extracellular vesicles (EVs), demonstrating a novel mechanism of osteocyte-stem cell communication. Therefore, osteocyte-derived mechanically activated EVs hold great potential as a novel cell-free therapy to enhance bone regeneration.
protein cargo in EVs from subjects after exercise compared to those at rest. 29 In summary, the importance of the role EVs play in osteogenesis can be seen, and due to the dynamic nature of this tissue, indicate potential further mechanisms involving EVs in bone mechanoadaptation.
While changes in several factors in and released by osteocytes have been shown via proteomics analysis, the specific composition and factors implicated in mechanically mediated osteocyte paracrine signaling are yet to be elucidated. Thus, the aim of this study was to further investigate the means by which osteocytes mediate bone mechanoadaptation, with this being achieved by constructing, for the first time, an extensive map of the osteocyte secretome protein signature. This map comprises the relative expression of hundreds of proteins in the osteocyte proteome under static and dynamic conditions in addition to providing information on how they interact with one another. We first validated the ability of the osteocyte secretome to induce a chemotactic and osteogenic response in human bone marrow stem/stromal cells (hMSCs) using a parallel plate flow chamber approach to mechanically stimulate osteocytes. We then conducted a proteomic analysis on the osteocyte secretome via mass spectrometry, to identify proteins released by cells under both static and dynamic culture conditions. Enrichment of gene ontology terms was investigated to elucidate the primary cellular components and processes with which the osteocyte secretome is involved, with further analysis comparing the altered protein release and most differentially expressed proteins released by mechanically stimulated cells.
This led to the discovery of MA-EVs. Specifically, EVs were subsequently separated from the secretome of mechanical-activated osteocyte; characterised; and found to elicit similar trends in MSC recruitment and osteogenesis to that seen with conditioned media (ie, whole secretome). We further demonstrated how MA-EVs drove enhanced later term osteogenesis as assessed by alkaline phosphatase (ALP) activity. This demonstrated a key role for osteocyte EVs in mediating hMSC behavior, identifying a potential novel mechanism by which osteocytes coordinate loading-induced bone formation.

| Mechanical stimulation and conditioned medium collection
Forty-eight hours prior to fluid shear application, 75 × 38 mm glass slides were coated with 0.15 mg/mL type I collagen (Sigma C3867) for 1 hour and washed with phosphate buffered saline (PBS), after which osteocytes were seeded at a density of 1.16 × 10 4 cells/cm 2 . Glass slides were transferred to custom-made parallel plate flow chambers (PPFCs) as previously described. 32 Each glass slide was assembled within an individual PPFC under sterile conditions and incubated at 37 C and 5% CO 2 . Cells in PPFCs were either subjected to a fluid shear stress of 1 Pa at a frequency of 1 Hz, or maintained in the PPFC under static conditions, with each condition completed in quadruplicate. After 2 hours of treatment, slides were transferred to culture dishes, washed with PBS, and 2.5 mL of serum-free medium was applied. A control group consisting of collagen-coated glass slides with no cells was also incubated with 2.5 mL of serum-free medium. All culture dishes were incubated for 24 hours and medium was collected from cells which had undergone fluid shear (CM-F), statically cultured cells (CM-S) and from cell-free slides with collagen coating (Medium). Samples were centrifuged at 3000g for 10 minutes at 4 C to remove debris, after which the supernatant was collected and stored at −80 C prior to use (Figure 2A).

| MS data analysis
Raw data from MS analysis was processed using MaxQuant software 33,34 version 1.5.5.1 and spectra searched using the built in Andromeda search engine 35 with the Uniprot FASTA validated Mus musculus database being used as the forward database and the reverse for the decoy search being generated within the software. A minimum six amino acid length criteria was applied and the false discovery rate (FDR) for MS data analysis was set to 1% at the peptide and protein level. Cysteine carbamidomethylation was included as a fixed modification and oxidation of methionine and protein N-terminal acetylation were set as variable modifications for the peptide search.
The "match between runs" algorithm was used to transfer peptide identifications between MS runs where possible to increase total number of protein hits. At least one unique or razor peptide was required per protein group for identification. Label-free quantification (LFQ) was carried out using the MaxLFQ algorithm 36 within the software, with Fast LFQ being disabled. Other settings were kept as default in the software.

| EV isolation from conditioned media
Medium from statically and dynamically cultured osteocytes was collected and centrifuged at 3000g for 10 minutes to remove debris.
Medium was then filtered through a 0.45 μm pore filter. Medium was subsequently ultracentrifuged at 110 000g for 75 minutes at 4 C, using an SW32.Ti swing bucket rotor. Collected EV pellets were washed in PBS and the ultracentrifugation process was repeated.

| Transmission electron microscope imaging
EV imaging was conducted via a JEOL JEM1400 transmission electron microscope (TEM) coupled with an AMT XR80 digital acquisition system. Samples were physiosorbed to 200 mesh carbon-coated copper formvar grids and negatively stained with 1% uranyl acetate.
Secondary antibodies were incubated for 1 hour at room temperature and developed using Immobilon Western Chemiluminescent horseradish peroxidase (HRP) substrate (Millipore, Massachusetts).

| Quantification of EV content in conditioned medium
As a surrogate of EV quantities, protein contents were measured using a BCA protein assay kit (Thermo Scientific, 23 227). Bovine serum albumin (BSA) standards (10 μL) were added to a 96 well plate after which 200 μL of working reagent was added (50:1 ratio of reagents A and B). EV samples were diluted in CST lysis buffer (Cell Signaling Technology, 9803), vortexed, and incubated for 1 hour on ice. Ten microliters of sample lysates were added to the plate and mixed with 200 μL of working reagent. The plate was incubated for 30 minutes at 37 C and absorbance read on a spectrophotometer at 562 nm. BCA assay results combined with the volume of the isolate were used to calculate the total quantity of protein in the EV isolates and this value was used to calculate the original concentration of EV protein in the conditioned medium.

| Particle sizalysis
Particle size analysis was performed on EV samples using a NTA NS500 system (NanoSight, Amesbury, UK). EV samples were diluted 1:50 in PBS and injected into the NTA system, which obtained 4 × 40 second videos of the particles in motion. Videos were then analyzed with the NTA software to determine particle size.

| Statistical analyses and bioinformatics
Statistical analysis on recruitment and gene expression data was carried out using one-way analysis of variance (ANOVA) and Bonferroni's multiple comparison post-test (*P < .05, **P < .01, ***P < .001.  The results of this reveal the presence of 97 proteins which have significant differential expression in CM, indicated in red in Figure 3C, and listed in Table 1. Within these proteins, significant enrichment (enrichment factor > 1.7, P < 10 -4 ) of several "extracellular" GOCC terms was shown in comparison to the total 393 identified proteins using Fisher's exact test, with enrichment of UniProt keywords "secreted" and "signal" (enrichment factor > 1.6, P < 10 -5 ) also occurred ( Figure 3D). This validates the successful isolation of proteins released by the osteocyte into their surrounding environment, with evidence for further downstream signaling functions. Functional enrichment within CM proteins of GOCC terms with reference to the whole Mus musculus genome further reported the significant enrichment of membrane-bound vesicles and exosomes in the secretome (Table S2). This suggested a potential role for EVs, and in particular exosomes (FDR < 10 -40 ), in transporting signaling factors, either protein to RNA based, released by osteocytes. Functional enrichment of GOBP, GOMF, and Pfam terms was also investigated, showing significant roles for these proteins in mechanosensensing and mechanosignaling, as evidenced by the most significantly enriched terms "response to stress" (FDR < 10 -6 ) and "protein complex binding" (FDR < 10 -8 ). The interaction network between identified proteins in the osteocyte secretome reveals a highly significant degree of protein-protein interaction (P < 10 -16 ) as illustrated in Figure S1. Enrichment analyses was also conducted on proteins more abundantly expressed in the control Medium samples using Fisher's exact test ( Figure S3) and functional enrichments (Table S4), revealing enrichment of muscle and cytoskeletal terms. These associations are likely due to the incorporation of proteins from rat tail collagen type 1 used for coating glass slides.

| Mechanical stimulation alters the protein release characteristics in osteocytes
Subsequent analysis separating the CM-S and CM-F groups showed that different proteins were released from statically cultured and mechanically stimulated osteocytes, highlighting the role of external mechanical forces in regulating the osteocyte secretome. The more F I G U R E 2 Outline of experiment procedure. MLO-Y4 cells were seeded to collagen coated glass slides and cultured for 48 hours (A), before being transferred to parallel plate flow chambers for dynamic (OFF, 1 Pa, 1 Hz, 2 hours) or static culture. The slides were then transferred to culture dishes and 2.5 mL of serum free medium was applied, with a control group being present with collagen coated glass slides without cells. The serum-free medium was collected and centrifuged to remove debris. One milliliter of each sample was collected, and proteins were precipitated and digested in solution before being purified via C18 stage tips (B). Samples were analyzed via liquid chromatography-mass spectrometry and tandem mass spectrometry (LC-MS/MS), and label-free quantification was carried out in MaxQuant before a bioinformatic analysis was completed in Perseus ( &&& P < .001 vs Medium using one-way analysis of variance (ANOVA) and Bonferroni's multiple comparison post-test) (C). Pearson correlations between technical replicates, biological replicates, and sample groups were determined, with correlations between biological replicates with combined technical replicates shown ( is associated with osteogenic growth peptide (OGP) and known to stimulate osteoblast activity. 40 Subsequently, functional enrichment in differentially secreted proteins between CM-F and CM-S was investigated to help further elucidate their collective biological relevance in mechanically mediated osteocyte signaling (Table S3). The top four enriched GOCC terms: extracellular region, membrane-bounded vesicle, extracellular region part, and extracellular exosome are associated with EV proteins with a highly significant FDR (<10-10). 65% to 76% of all differentially secreted proteins were associated with these terms. This confirms that EVs are not only implicated in the osteocyte secretome, as demonstrated above, but are a key component of mechanically mediated signaling. Also, of substantial interest is the enrichment of the top two GOMF terms "calcium ion binding" (FDR < 0.01) and "phosphoserine binding" (FDR < 0.05), revealing the potential role of mechanically activated osteocyte EVs as sites of mineralization via binding of calcium and phosphate components, which has been previously postulated. 20,24 A String DB network was constructed to further investigate F I G U R E 3 Proteomic analysis of the osteocyte secretome. Hierarchical clustering of all samples with imputed data (A) and hierarchical clustering in the control samples without imputation of data (B). Volcano plot illustrating proteins significantly upregulated proteins marked in red in CM-S and CM-F groups compared to the control (C). Enrichment analysis of GOCC terms and Uniprot keywords in upregulated proteins using Fisher's exact test represented as a word cloud (D). The size of the word represents enrichment of terms, while color represents FDR corrected P value. All terms with a minimum of 0.5 enrichment factor and 0.05 FDR corrected P value were included. FDR, false discovery rate any potential interactions between proteins associated with EVs ( Figure 4H) revealing a significant degree of protein-protein interaction (P < 10-3). Interestingly, there are several interactions between positively and negatively regulated proteins, including an interaction path between Anxa5 and Ywhab/Ywhae which are associated with calcium ion binding and phosphoserine binding, respectively. Between these nodes are gelsolin and cofilin, the former of which is calciumsensitive and both of which have been shown to regulate changes in the actin cytoskeleton, 41 as well as occurring in vesicles from mineralizing osteoblasts. 42

| EVs are present within the osteocyte secretome and EV morphology and size distribution is not altered by mechanical stimulation
Given the identification of EV-associated proteins within the osteocyte secretome, we next investigated whether osteocytes release EVs and, if so, whether EV characteristics were altered by mechanical stimulation. EVs were successfully separated from osteocyte CM using filtration and ultracentrifugation, with the presence of EVs confirmed by TEM imaging and immunoblotting. TEM imaging confirmed Differentially expressed proteins between CM-S and CM-F groups the presence of EVs of typical morphology and size ( Figure 5A,B). The Long-term study of influence of EVs on human bone marrow stem/stromal cell (hMSC) osteogenesis. A, Summary schematic of experimental procedure and outputs. Intracellular ALP activity at day 7 (B), with a comparison of EV groups with EV depleted conditioned medium (C). Extracellular ALP activity at day 3 (D) with a comparison of EV groups with EV depleted conditioned medium (E). Extracellular ALP activity at day 7 (F) with a comparison of EV groups with EV depleted conditioned medium (G). All data n = 6. Statistical analysis using one-way analysis of variance (ANOVA) and Tukey's multiple comparison post-test (*P < .05, **P < .01, ***P < .001, &&& P < .001 of indicated group vs all other groups). ALP, alkaline phosphatase; EVs, extracellular vesicles particle size distributions between EV-S and EV-F ( Figure 5E); however, no changes in average particle size was detected, with values of 177 nm and 183 nm, respectively ( Figure 5F).

| Osteocytes regulate human MSC recruitment and osteogenesis in response to fluid flow shear via MA-EVs
To determine whether murine osteocyte-derived EVs could be taken up by human MSCs, we labeled EVs with PKH26. Following 24 hours treatment, labeled-EVs were preferentially located within the cytoplasm, indicating uptake of EVs by hMSCs ( Figure 5G). Control samples are illustrated in Figure S4. A high density of EVs can be seen around the nuclear region in particular with minimal detection within the nucleus.
Upon verifying EV uptake, the cellular response of hMSCs sub- To further investigate the role of EVs in later stage differentiation, hMSCs were cultured up to 7 days, with intracellular ALP being investigated at day 7 and extracellular ALP being investigated at day 3 and day 7. This was conducted on both EVs and EV depleted medium to investigate how other non-EV associated factors within the medium affect osteogenesis ( Figure 6A). Intracellular ALP activity was significantly enhanced at day 7 with EV-F vs EV-S in addition to all other groups ( Figure 6B), while it was also demonstrated that EV-F significantly enhanced ALP activity compared to both static and flow EV depleted medium ( Figure 6C), which is consistent with trends seen with earlier gene expression. Extracellular ALP was also assessed from the medium change at day 3, in addition to the end of the experiment at day 7. There were no significant differences between any groups at day 3 ( Figure 6D,E). However, at day 7 it can be seen that EV-F is significantly enhanced compared to EV-S ( Figure 6F). Interestingly, while there is an increase in extracellular ALP in the EV-F compared to the EV depleted CM-F, this is not significant ( Figure 6G). Taken together, this data demonstrate that EVs isolated from mechanically activated osteocytes can enhance the osteogenesis of hMSCs. Annexin A5 has been shown to increase at the cell membrane in addition to Ca2+ levels in osteoblasts under fluid flow. The disruption of annexin A5 inhibits Ca2+, implicating its role in calcium signaling, 55 with its knockdown impairing osteoblast function. 56 One downregulated protein of interest which we have identified is thrombospondin 2. The knockdown of this protein in mice has been shown to increase angiogenesis 57 and endosteal bone formation 58 while thrombospondin 2 null mice demonstrate enhanced callus bone formation, vascularity and MSC proliferation following tibial fracture. 59 In addition to the above, many of the proteins we have identified are known to be EV cargo, with many also having previously been identified in osteoblast EVs. 42,60 The content of the osteocyte secretome, in addition to a possible delivery mechanism via EVs, provides us with a database of information via which to develop targeted therapeutics for specific bone diseases such as osteoporosis.
The role of EVs in osteocyte signaling to MSCs was investigated to further investigate the mechanisms behind our previously discussed findings. EVs were separated from osteocyte conditioned media, where interestingly, we did not detect any changes in EV morphology or quantity between static and dynamic groups. This is in contrast to previous work which has demonstrated an upregulation in EV number following fluid shear stimulation. 61 We have however Given the similar concentrations of EVs between groups, it is expected that this pro-osteogenic effect is a result of EV content changing in response to mechanical stimulation. EVs are also known to act as sites of mineral nucleation, as has been demonstrated in osteoblasts. 20,62 We have also seen evidence for this, with enrichment of calcium ion binding (such as annexin A5) and phosphoserine binding (such as 14-3-3 proteins Ywhae and Ywhab) proteins in our proteomic analysis. In our protein interaction network, Annexin A5 is linked to the calcium sensitive protein gelsolin, 41 which in turn is linked to the 14-3-3 proteins via the phosphate regulating cofilin. 63 In addition to the known role of calcium ions in mineralization, negatively charged amino acids such as phosphoserine are also known to play a key role in hydroxyapatite nucleation and growth. 64 Therefore, osteocyte EVs may promote mineralization via delivery of calcium and phosphate interacting proteins through interaction with gelsolin and cofilin, respectively. Another likely mechanisms in the pro-osteogenic capabilities of EVs are RNAs, specifically miRNAs, with a previous study demonstrating altered miRNA expression in EVs isolated from the plasma of osteocyte ablated and wild-type mice. 25 The potency of EVs in mediating cell behavior has led others to exploit them for the development of therapeutics. For example, EVs have been investigated as a potential therapy for root canal treatments, 65 for the functionalization of TE scaffolds to enhance bone regeneration, 66,67 and as delivery vehicles for the loading of drugs for osteoporosis therapies. 28 Taken together, we have shown that MA-EVs are a key mechanism by which osteocyte communicate chemotactic and osteogenic signals to osteoprogenitors in response to loading, and present themselves as a potential cell free therapy to mimic the beneficial effect of loading and enhance bone formation.
In summary, this study presents evidence that the mechanically stimulated osteocyte secretes factors which coordinates MSC recruitment and osteogenesis demonstrating a mechanism required for loading-induced bone formation. Importantly, for the first time, we have mapped the osteocyte protein secretome and determined how this is altered in response to mechanical stimulation generating a database of potential factors mediating this mechanism. Last, this study also demonstrates the presence and fundamental role of MA-EVs released by osteocytes in coordinating MSC recruitment and osteogenesis, identifying a novel mechanism by which osteocytes coordinate bone mechanobiology. Moreover, these pro-osteogenic osteocyte derived MA-EVs represent a potential cell-free therapy to enhance bone regeneration and repair in diseases such as osteoporosis.

ACKNOWLEDGMENTS
We would like to acknowledge funding from European Research

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from the corresponding author upon reasonable request.