Osteoblast‐Derived Extracellular Vesicles Are Biological Tools for the Delivery of Active Molecules to Bone

Extracellular vesicles (EVs) are newly appreciated regulators of tissue homeostasis and a means of intercellular communication. Reports have investigated the role of EVs and their cargoes in cellular regulation and have tried to fine‐tune their biotechnological use, but to date very little is known on their function in bone biology. To investigate the relevance of EV‐mediated communication between bone cells, we isolated EVs from primary mouse osteoblasts and assessed membrane integrity, size, and structure by transmission electron microscopy (TEM) and fluorescence‐activated cell sorting (FACS). EVs actively shuttled loaded fluorochromes to osteoblasts, monocytes, and endothelial cells. Moreover, osteoblast EVs contained mRNAs shared with donor cells. Osteoblasts are known to regulate osteoclastogenesis, osteoclast survival, and osteoclast function by the pro‐osteoclastic cytokine, receptor activator of nuclear factor κ‐B ligand (Rankl). Osteoblast EVs were enriched in Rankl, which increased after PTH treatment. These EVs were biologically active, supporting osteoclast survival. EVs isolated from rankl–/– osteoblasts lost this pro‐osteoclastic function, indicating its Rankl‐dependence. They integrated ex vivo into murine calvariae, and EV‐shuttled fluorochromes were quickly taken up by the bone upon in vivo EV systemic administration. Rankl–/– mice lack the osteoclast lineage and are negative for its specific marker tartrate‐resistant acid phosphatase (TRAcP). Treatment of rankl–/– mice with wild‐type osteoblast EVs induced the appearance of TRAcP‐positive cells in an EV density‐dependent manner. Finally, osteoblast EVs internalized and shuttled anti‐osteoclast drugs (zoledronate and dasatinib), inhibiting osteoclast activity in vitro and in vivo. We conclude that osteoblast EVs are involved in intercellular communication between bone cells, contribute to the Rankl pro‐osteoclastic effect, and shuttle anti‐osteoclast drugs, representing a potential means of targeted therapeutic delivery. © 2017 The Authors. Journal of Bone and Mineral Research Published by Wiley Periodicals Inc.

(osteoclasts), and the bone-sensing cells (osteocytes) have an intimate crosstalk, with coordinated and controlled coupling. This relationship induces a tightly regulated influence on the activities and functions of these cells. Since the first pivotal molecule in osteoblast-osteoclast coupling was identified, (24,25) more and more pathways and communication mechanisms between bone cells have been brought to light.
One of the most paradigmatic molecules implicated in the osteoblast-osteoclast crosstalk is the receptor activator of nuclear factor k-B ligand (Rankl). (25) Rankl is predominantly a transmembrane protein, also occasionally cleaved into a soluble form, which in bone is expressed by osteoblasts and osteocytes and which binds its receptor, Rank, on the surface of osteoclasts and monocytes. The Rankl/Rank axis is a pivotal mediator of osteoclastogenesis, osteoclast function, and osteoclast survival, and both human and murine osteopetrosis forms characterized by the absence of osteoclasts have been demonstrated to carry loss-of-function mutations of these genes. (26) Based on the complexity of the cellular cycle governing bone remodeling and the exchange of a plethora of molecules between cells in the bone microenvironment, the involvement of EVs in bone pathophysiology is a potential piece of the complex puzzle of bone cellular crosstalk. Despite the dramatic interest in EVs in many fields, their roles in bone biology are still poorly understood. Very few articles have reported the characterization of their molecular profile and activity, and it is possible that the real physiologic role of EVs in the bone microenvironment has been underestimated because only immortalized or tumor cells have been investigated so far, which exhibit intrinsic genetic abnormalities (ie, viral and pro-oncogenic genes). (27)(28)(29) In this study, we focused on mouse primary osteoblastreleased EVs and characterized some of their biologic aspects, revealing that they have a previously unrecognized role in the bone cell crosstalk, and demonstrating their relevance as a useful tool to cleverly target bone cells.

Animals
Mice were employed for cell cultures and in vivo experiments in conformity with national and international laws and policies (EEC Council Directive 86/609, OJ L 358, Dec. 12,1987; Italian Legislative Decree, Gu n.61, 14/03/2014) and the NIH Guide for the Care and Use of Laboratory Animals (8th edition. Washington, DC: National Academies Press; 2011). Their use was approved by the Institutional Review Board of the University of L'Aquila, Italy. At the end of the in vivo experiments and for primary cultures, mice were euthanized by CO 2 inhalation. Mice treated with EVs starting at 5 days of life and subjected to blood collection were euthanized by decapitation. For in vivo, ex vivo, and in vitro experiments, wild-type mice were of the CD1 and the C57/BL6 backgrounds, whereas rankl -/-(tnfsf11 -/-) mice were of the C57/BL6 background.
Mice were purchased by Charles River-Laboratories Italia S.r.l (Milan, Italy), housed in the animal facility of the University of L'Aquila, Italy, at the following conditions: temperature, 20°C to 24°C; humidity, 60% AE 5%; dark/light cycle, 12/12 hours. They had access to food and water ad libitum, and were fed with a normal diet (Mucedola code: #3KE25, Mucedola S.R.L., Settimo Milanese, Italy).
For the in vivo EV kinetics and biological effects, the sample size was calculated using dedicated software (SigmaPlot; Systat Software, Inc., San Jose, CA, USA) based on the expected differences.
For transwell experiments, primary murine osteoblasts (2 Â 10 5 cells/transwell six well) were centrifuged at 400g for 5 min and washed twice in PBS. Then, cell suspension was labeled with 1 mM PKH26 (Sigma-Aldrich; #MINI26-1KT) for 5 min and washed twice (5 min Â 2) in complete DMEM (10% FBS). A transwell system with a 1-mm pore size membrane (BD Biosciences cell culture insert, Erembodegem, Belgium) was used for culturing PKH26-labeled osteoblasts in the upper chamber and unstained cells in the lower chamber. In both chambers, cells were plated at a density of 1 Â 10 6 /cm 2 .
Primary mouse monocytes were obtained as described in the previous paragraph, without the treatment with the rhM-CSF and rhRankl.
EA.hy926 human endothelial cell lines were obtained from the American Tissue Culture Collection (ATCC, Manassas, VA, USA). EV isolation and staining EV were isolated from untreated osteoblasts or from osteoblasts treated every 12 hours with 1 Â 10 -7 M rhParaThyroid Hormone (rhPTH; Sigma-Aldrich). (30) Upon reaching 80% confluence, cells were washed in PBS and starved in serum-free DMEM to prevent FBS EV contamination. After 24 hours, culture medium was harvested and sequentially centrifuged at 4°C at 5,000g for 20 min to remove cell debris, 35,000g for 20 min to remove apoptotic bodies, and 100,000g for 70 min to spin down EVs. The pellets containing EVs were then incubated for 30 min at 37°C with either the membrane-permeant green dye 5chloromethylfluorescein diacetate (CMFDA) (Thermo Fisher Scientific, Waltham, MA, USA; #C7025) or with the SYTORNA-Select Green Fluorescent Cell Stain (Thermo Fisher Scientific; #S32703), and for 5 min at 37°C with the cell membrane labeling linker PKH26. Then, the EVs were washed in PBS by ultracentrifugation at 100,000g, at 4°C for 70 min. Finally, the pelleted EVs were resuspended in a volume of PBS according to the application, and immediately used for the experiments or stored at -80°C. All cell treatments were performed using EV suspensions derived from 2.5 Â 10 6 cells/cm 2 .

Transmission electron microscopy
For transmission electron microscopy (TEM), 5 mL of EV suspension in PBS was deposited onto Formvar-coated grids and left to air-dry for 20 min. Grids were then washed with PBS and fixed by transfer onto 1% glutaraldehyde for 5 min. After washing with distilled water, samples were contrasted with 4% uranyl-oxalate solution (4% uranyl acetate, 0.15M oxalic acid, pH 7, Sigma-Aldrich-Aldrich) for 5 min. Finally, grids were air-dried for 10 min and observed under a Philips CM 30 TEM at 80 kV.

FACS of EVs
EV pellets were loaded with CMFDA and incubated with a phycoerythrin (PE)-conjugated anti-Rankl antibody (1:100; BD Biosciences; Cat. #560295) for 30 min at 4°C, then CMFDAloaded, Rankl-positive and Rankl-negative EVs were sorted by FACS. Nanofluorescent standard particles (Spherotech, Lake Forest, IL, USA; cat# NFPPS-52-4K) were used to set dimensional gate up to 1 mm. Routine immunophenotype evaluation was performed on BD FACSCANTO II. Sorting of pure EV population was performed on BD FACSARIA II.

Gene expression
Conditioned media from osteoblasts were processed for EV isolation by ultracentrifugation as described in the "EV isolation and staining" section. Donor osteoblasts were massively washed in PBS (to eliminate any contaminants from the EVs present in the conditioned medium) and trypsinized. Osteoblast suspension was washed twice in PBS by centrifugation at 300g for 5 min at 4°C, then 1 million of these cells and the EV pellet were processed for RNA extraction by RNeasy mini kit (Qiagen, Valencia, CA, USA; cat #74104), according to the manufacturer's instructions, and quantified using a NanoDrop 2000 spectrophotometer (Thermo Fisher Scientific).
For large-scale analysis, 32 ng of total RNA were retrotranscribed by PreAMP cDNA Synthesis Primer Mix for mouse Osteoporosis, then the cDNA was ran on a mouse Osteoporosis PCR array (Qiagen; cat#PAMM-170Z).
PCR conditions were 94°C for 30 s, 60°C for 30 s, and 72°C for 30 s, replicated for 35 cycles.
EV loading with anti-osteoclast drugs Sodium zoledronate (Sigma-Aldrich) was dissolved in PBS at a concentration of 14mM, and dasatinib (Selleckchem, Munich, Germany) was dissolved in DMSO at a concentration of 10mM. EVs from 2.5 Â 10 6 osteoblasts/cm 2 were incubated with either drug for 1 hour at 37°C, then they were extensively washed in PBS and ultracentrifuged twice at 100,000g for 70 min.

In vitro EV biologic effects
Mature osteoclasts were starved in serum-free DMEM as described in the "Cell cultures" section and treated for 48 hours with EVs (intact or drug-loaded) derived from 2.5 Â 10 6 osteoblasts/cm 2 . Osteoclasts were then fixed in paraformaldehyde/PBS (4% vol/vol), cytochemically stained for the osteoclast marker tartrate-resistant acid phosphatase (TRAcP) (Sigma-Aldrich; #387A-1KT) and counterstained with 1mM Hoechst 33342 (Sigma-Aldrich) to identify the nuclei.

Ex vivo EV integration into bone
Seven-day-old CD1 mice were euthanized, their calvariae excised, plated in 35-mm dishes in DMEM with 10% FBS and incubated with PKH26-labeled EVs. After 48 hours, EV integration in the calvarial bones was evaluated by an Olympus Fv1000 confocal microscope (Olympus, Waltham, MA, USA).

In vivo EV kinetics and tissue targeting
Cryopreserved FACS-sorted Rankl-positive and PKH26-labeled EVs were injected intraperitoneally into 5-day-old CD1 male mice. Groups of animals (n ¼ 6 each) were euthanized after 1.5, 3, 6, 24, and 48 hours and blood, liver, spleen, kidney, and bones were collected and processed for PKH26 extraction in chloroform-methanol (2:1 vol/vol) and 0.125% sodium dodecyl sulfate. The PKH26 fluorescence intensity was measured by spectrofluorometry (excitation wavelength 550 nm; emission wavelength 567 nm) in the extracted lipid fraction.

In vivo EV biologic effects
Five-day-old C57BL/6 rankl -/male mice were injected intraperitoneally daily for 5 days with 30,000, 60,000, or 120,000 Rankl-positive EVs or with vehicle (PBS) as control. Then, mice were euthanized and tibias excised, fixed for 24 hours in 4% paraformaldehyde, dehydrated with ethanol, and embedded in methoxymethylmetacrylate without decalcification. TRAcP histochemical staining was performed to identify and quantify TRAcP-positive cells belonging to the osteoclast lineage.
Eight-week-old C57BL/6J mice were treated for 4 days with 120 mg/kg body weight (bw) of retinoic acid (Sigma-Aldrich; cat. #R2625, vitamin A acid) administered by oral gavage. (31,32) Mice (n ¼ 5/group) were treated once, at day 1 of the experiment, by intraperitoneal injection of vehicles (PBS for sodium zoledronate and 0.1% DMSO for dasatinib), 70 mg/kg bw of sodium zoledronate, 12.5 mg/kg bw of dasatinib, osteoblast EVs loaded with a solution of 14mM sodium zoledronate, osteoblast EVs loaded with a solution of 10 mM dasatinib, or naive osteoblast EVs (2.5 Â 10 6 osteoblasts/cm 2 ). One group of mice was also sham-treated by oral gavage with the retinoid acid vehicle (corn oil). At the end of the experiment, mice were euthanized. The hind-limb long bones were removed, fixed in 4% paraformaldehyde for methoxymethylmetacrylate embedding without decalcification, histochemically stained for TRAcP activity, and histomorphometrically analyzed for the osteoclast variables according to Dempster and colleagues. (33) Counterstaining with hematoxylin was performed to evaluate osteoclast apoptosis.
For the in vivo experiments, three to six mice were employed per each group of treatment, allocated after randomization in number of three mice/cage. Treatments and measurements were performed in blind. No adverse effects were observed following treatment with EVs.

Statistics
Data were reported as mean AE standard deviation (SD). Student's t test between two groups or one-way ANOVA using nonparametric Kruskal-Wallis test between multiple groups were used as comparative statistical methods, with a significant p value <0.05 (GraphPad Prism 7.00; GraphPad Software, Inc., La Jolla, CA, USA). All in vitro experiments were performed with three technical replicates for each point and repeated at least three times or using three to six mice/group.

Osteoblasts release membrane-derived structures into their surroundings
To investigate whether physiologic cell-to-cell communication in bone cells could involve EVs, we focused on osteoblasts, the active matrix-forming cell type in the bone, examining clues of intercellular communication based on EV exchange. Primary murine osteoblasts, labeled with the red fluorescent cell membrane dye PKH26, were plated in the upper chamber of 1-mm pore transwell dishes, allowing the passage of EV-sized particles, while unstained osteoblasts were plated in the lower chamber (Fig. 1A). The cultures were monitored and after 48 hours we observed the appearance of red fluorescence in the previously unstained osteoblasts in the bottom chamber (Fig. 1B), indicating the fusion of donor cell-labeled plasma membrane derivatives as well as their integration in the target cells.

EVs are harvestable from osteoblast conditioned media
To confirm the involvement of EVs in the transfer of membranebound dyes, we analyzed the release of EVs from cultured osteoblasts. We collected 24-hour serum-free conditioned media from primary mouse calvarial osteoblasts. After ultracentrifugation, the pellets were assessed for protein content (Fig. 1C, left bar). To further evaluate at the molecular level whether or not our ultracentrifugation pellet contained EVs, we performed a Western blot for the commonly-reported exosome marker CD63. (3) The enrichment of CD63 in our ultracentrifugation pellets compared to the total cell lysates indicated the isolation of a concentrated population of EVs (Fig. 1D). The same evaluation performed in cryopreserved EVs demonstrated that this marker was not lost upon freezing (Fig. 1D).
To confirm the presence of intact EVs obtained by ultracentrifugation, we loaded the EVs of the isolated pellets with the membrane-permeant fluorescent probe CMFDA. This dye is cleaved intracellularly by cytosolic esterases into a nonpermeant molecule that does not leak out from intact membranes. By fluorescence microscopy, we observed in the pellets a population of green-fluorescent apparently integral vesicular structures (Fig. 1E). Finally, a deeper morphologic evaluation by TEM analysis showed the expected heterogeneous population of intact and spheroid membrane bodies (Fig. 1F), which matched the expected size distributions of EV populations. (1) Next, we questioned whether EV release was a physiologically controlled process. Because it has been reported that cells can respond to stimuli by increasing vesiculation, (34) we treated osteoblast cultures with 1 Â 10 -7 M rhPTH, a potent regulator of osteoblast activity. (30) Although the donor osteoblast cell protein content was not affected by PTH (untreated osteoblasts, 1.1 AE 0.13 mg; PTH-treated osteoblasts, 1.05 AE 0.02 mg; p ¼ 0.6), the protein content of the EV pellet obtained from the conditioned media of PTH-treated osteoblasts was higher than from control conditioned media (Fig. 1C, right bar), supporting the hypothesis that this release was a hormonally regulated process.
Finally, we investigated whether EVs were able to target cells and transfer fluorophores. We incubated the CMFDA-loaded EVs with primary murine osteoblasts and monocytes, two of the most abundant cell types in bone and bone marrow, respectively, confirming the integration of this fluorescent probe in both recipient cells (Fig. 1G,H).

Qualitative and quantitative analysis of osteoblastderived EVs by FACS
To confirm the cellular origin and integrity of the membrane structures, we next performed a FACS analysis after incubation of EV pellets with CMFDA. We noted that 16.67% AE 1.93% of events analyzed in the ultracentrifugation pellet represented CMFDA-positive structures up to 1 mm in diameter, consistent with the size of EVs (Fig. 1I,J). Given that CMFDA is a dye that does not leak out from integral membrane particles, we concluded that these structures were intact. The CMFDApositive particles were sorted and their morphology was evaluated by TEM, confirming the expected EV size distribution and apparent membrane integrity (Fig. 1K).
We next assessed whether FACS-sorted EVs retained the ability to fuse with target cells. To this aim, we incubated primary osteoblasts with FACS-sorted CMFDA-loaded PKH26-labeled EVs. Microscopic analysis showed the integration of both fluorophores into target cells ( Fig. 2A, left panels). Trypsin/ EDTA treatment of target cells did not compromise the presence of the dyes in their cytoplasm ( Fig. 2A, right panels), confirming

Osteoblast-derived EVs target and integrate into recipient cells
We next investigated the kinetics of EV integration into target cells. To this aim, we performed a time-course experiment using osteoblasts incubated with FACS-sorted EVs loaded with CMFDA and labeled with PKH26. By confocal microscopy, we observed a time-dependent integration of EV fluorescent components in target cells starting after 24 hours from EV administration and persisting until at least 48 hours (Fig. 2B). The analysis of the integrated dyes showed a typical vesicular distribution. Furthermore, z-axis evaluation enabled us to distinguish also domains of the target cells enriched only in one of the two fluorophores with no colocalization in the same areas (Fig. 2B). This observation suggests that target cells could process the individual constituents of EVs separately, transferring them to different cellular compartments.
Finally, we investigated the integration of FACS-sorted EVs with other cell types, such as endothelial cells and monocytes, with which the osteoblasts exchange important molecular information. Fluorescence microscopy confirmed that both human umbilical vein cell line EA.hy926 and primary mouse monocytes internalized the fluorochromes shuttled by EVs (Fig. 2C), suggesting also a mouse-human cross-species integration.
Osteoblast-derived EVs contain Rankl Because EVs are reported to trigger ligand/receptor interactions in target cells, (16,17) we asked whether EVs from osteoblasts could shuttle Rankl, one of the most important intercellular crosstalk molecules in bone. We therefore investigated the presence of membrane-bound Rankl by FACS in osteoblast EVs and observed that 53.95% AE 3.48% of them were Rankl-positive (Fig. 3A). To further characterize the quality of FACS-sorted EV preparations, we performed TEM analysis of Rankl/CMFDAdouble positive EVs and observed a heterogeneous population of intact spheroid bodies, with the expected EV morphology and size distribution (Fig. 3B).
Because we noted that PTH increased EV production by osteoblasts (Fig. 1C), and given that PTH is a well-known inducer of Rankl expression, (30) we evaluated the proportion of osteoblast-derived EVs which were Rankl-positive following treatment with 1 Â 10 -7 M rhPTH. FACS analysis showed that PTH increased the percent of Rankl-expressing osteoblasts as well as the mean PE-fluorescent units per osteoblast (Fig. 3C,D). Furthermore, the percentage of RANK-L-positive events increased up to 63.6% AE 4.20% of EVs in PTH-treated ultracentrifugation pellets versus vehicle-treated osteoblasts (Fig. 3E,F). Consistently, PTH also increased the total number of EVs (Fig. 3G), the number of Rankl-positive EVs (Fig. 3F), the mean PE-fluorescent units per event (Fig. 3H), and the percentage of Rankl-positive events (Fig. 3I).

Biologic effects of osteoblast EVs
To assess the biologic effect of EVs on bone cells we investigated the phenotype of osteoblasts and osteoclasts exposed to osteoblast EVs. Semiquantitative RT-PCR performed on control osteoblasts and osteoblasts treated for 48 hours with osteoblast EVs showed no modulation of Alp, Runx2, and Atf4, while osteoblast EVs significantly reduced osterix mRNA (Fig. 4A,B). Likewise, the transcriptional expression of the matrix protein, Col1a1 was not affected by osteoblast EVs, while Osteocalcin was significantly increased compared to control osteoblasts (Fig. 4A,  B). These data suggest selective regulation of osteoblast genes by means of their own EVs.
Next, we evaluated the biologic effect of EVs on osteoclasts and noted that treatment with osteoblast-derived EVs improved osteoclastic variables, increasing osteoclast size and number of nuclei/cell (Fig. 5A,B). Moreover, the EV-treated osteoclasts stained more strongly for TRAcP than vehicle-treated osteoclasts (Fig. 5A). Taken together, these data confirmed a direct biologic effect of osteoblast-derived EVs in cells of the bone microenvironment and supported the concept that EVs are a means of Rankl mediated osteoblast/osteoclast crosstalk.
To establish the involvement of Rankl in the pro-osteoclastic effect of osteoblast-derived EVs, we took advantage of the rankl -/-mouse to isolate primary osteoblasts and collect rankl -/-EVs. We treated 12-hour-starved osteoclasts for 48 hours with wild-type or rankl -/-EVs, noting that rankl -/-EVs were less efficient than wildtype EVs in supporting osteoclast survival (Fig. 5C-E). In fact, osteoclasts treated with rankl -/-EVs generally contained fewer nuclei and exhibited signs of nuclear degeneration (Fig. 5C-E). Moreover, we noted a higher percentage of retracted or disaggregated osteoclasts (Fig. 5C-E), reminiscent of apoptotic cells. EV retrieval from wild-type and rankl -/osteoblasts was similar, as shown by the protein quantification of the EV pellets from the two genotypes (Fig. 5F), ruling out that the observed differences were due to altered EV release by rankl -/osteoblasts. These results demonstrated that EVs can exert a direct effect on osteoclast survival through their Rankl cargo.

EVs are biotechnological tools to shuttle drugs to osteoclasts
EVs have raised great interest as drug carriers. (35) Because we found a clear effect of na€ ıve osteoblast EVs on osteoclasts, we investigated whether they could be manipulated to shuttle antiosteoclastic agents to target cells. To test this concept, we loaded osteoblast EVs with the clinically-approved anti-osteoclast drugs, sodium zoledronate (N-bisphosphonate) and the tyrosine kinase inhibitor dasatinib. We treated mature osteoclasts for 48 hours with EVs loaded with zoledronate ( Fig. 6A) (EV-Zol) or dasatinib (EV-Das) (Fig. 6B), which induced cell death with high efficiency (Fig. 6C,D).

Osteoblast-derived EVs shuttle RNAs to recipient cells
EVs are reported to be involved in cell reprogramming via the shuttling of RNAs, (5,10,36) so we examined whether osteoblastderived EVs contained RNAs. To address this issue, we stained EV  pellets with SYTO RNASelect, a cell-permeant dye that exhibits bright green fluorescence only when bound to single-stranded nucleic acids. We then treated both primary osteoblasts and monocytes with SYTO-stained EVs, noting the integration of green fluorescence into target cells, thus indicating the shuttling of RNAs (Fig. 7A,B). FACS analysis revealed that most EVs (97.1% AE 0.26%) were SYTO-positive and contained RNAs (Fig. 7C,D). Treatment of primary osteoblasts with SYTO-positive FACS-sorted EVs confirmed that green-stained RNAs were transferred into target cells (Fig. 7E).
With this data in hand, we characterized the mRNA content of EVs. To achieve this goal we focused on a panel of 84 genes, known to be involved in osteoblast/bone physiology, using a Real-Time Profiler PCR array. We found that EVs from osteoblasts were enriched especially in Col1a1, Col1a2, Sparc, and Spp1 genes, typically involved in osteoblast function and bone formation (Table 1, Fig. 7F). We confirmed the expression of the most abundant gene, Col1a1, normalized versus Gapdh, by RT-PCR (Fig. 7G). A further bioinformatics comparison of transcriptome profile between donor osteoblasts and released EVs showed no significant differences between them (Fig. 7H), suggesting that in our conditions the EV transcriptome mirrors the osteoblastic transcriptomic profile.

Osteoblast-derived EVs show marked ex vivo osteotropism and in vivo bone biodistribution
We next investigated whether osteoblast-derived EVs were able to integrate into the bone tissue. To this aim, we incubated mouse calvariae ex vivo with PKH26-labeled osteoblast EVs for 48 hours. Confocal microscopy showed  integration of PKH26 into the inner part of the bone (Fig. 8A,  B), supporting the hypothesis of an intrinsic osteotropism and integration of osteoblast EVs. It is known that peritoneal vessel allow EVs to pass into the circulation and reach target organs. (37,38) Therefore, in order to demonstrate that EVs could target the bone tissue in vivo, we performed intraperitoneal injections of 30,000 FACS-sorted Rankl positive EVs, labeled with PKH26, into 5day-old CD1 pups. Kinetic evaluation of PKH26 distribution in tissues revealed a fast uptake of PKH26 in bones, which peaked at 1.5 hours from EV injection, declining thereafter to a plateau within 24 hours (Fig. 8C). PKH26 fluorescence was also observed in blood cells, liver, spleen and kidney, but with lesser intensity or slower kinetics (Fig. 8C). In liver, a peak of PKH26 fluorescence was observed after 24 hours from EV injection (Fig. 8C), suggesting clearance of the dye from the other tissues. Confocal microscopy on optimal cutting temperature (OCT)-embedded liver sections showed the presence of PKH26 spots (Fig. 8D,E), confirming the accumulation of PKH26 in the liver parenchyma.

Osteoblast-derived EVs shuttle specific information to bone in vivo
To further investigate the effect of EVs on bone physiology, we focused on the osteoclast lineage and performed in vivo experiments in rankl -/mice, a murine model of severe osteopetrosis characterized by the lack of osteoclasts and the consequent absence of the osteoclast biomarker TRAcP in the bone. We intraperitoneally injected increasing amounts of FACS-sorted Rankl-positive osteoblast EVs into rankl -/mice. Histochemical analysis of tibial sections revealed the presence of TRAcP-positive cells in all EV treatments. In contrast, TRAcP positivity was totally absent in vehicle-injected mice (Fig. 8F-K). The total number of TRAcP-positive cells did not vary between the EV treatment regimens (Fig. 8J), but the total TRAcP-positive area increased with the number of injected EVs (Fig. 8K). These results suggest that Rankl-positive EVs exhibit a direct in vivo osteoclastogenic potential.
Osteoblast EV-encapsulated drugs inhibit osteoclast activity in vivo To investigate whether osteoblast EV-encapsulated drugs could be used in vivo to inhibit osteoclast activity, we induced an acute osteoclast overactivation by administration of retinoid acid (31) and treated the mice once, at day 1 of the experiment, with the free drugs or with EVs isolated from the conditioned media of osteoblasts, collected from one 125-cm 2 flask/mouse (2.5 Â 10 6 osteoblasts/cm 2 ) and loaded with vehicle, zoledronate, or dasatinib. After 4 days, mice were euthanized, tibias were histochemically stained for TRAcP and evaluated histomorphometrically for the osteoclast variables. As expected, mice receiving retinoic acid underwent a significant reduction of body weight (Fig. 9A), indicating the efficacy of the treatment. (31,32) They also showed an increase of the bone resorption marker, collagen type 1 C-telopeptide (CTX) (Fig. 9B), and of the osteoclast number (Fig. 9C,D) and surface (Fig. 9D) over bone surface. The increase of these variables was efficiently counteracted by treatment with free zoledronate or dasatinib, while EVs alone had no effect. Remarkably, osteoblast EV-encapsulated drugs reduced osteoclast overactivation (Fig. 9A-E) and, in the case of EV-Zol, induced apoptosis (Fig. 9F,G).

Discussion
The involvement of EVs in the communication between bone cells is very poorly understood. Authors have reported vesiclelike structures arising from osteocytes, (39) but with no data on their composition and function. Some EV protein characterization has been reported for osteoblast cell lines, (28,40) and "malignant" osteosarcoma cell EVs have been described to transport osteoclastogenic instruction. (41) Despite these reports, no functional data are available to understand the function of EVs in bone biology.
In our study, we used primary murine osteoblasts and assessed their release of EVs into the environment. We noted in vitro that osteoblasts shuttled lipophilic dyes to distant cells through transwell membranes with porosity up to 1 mm. These results are consistent with the shuttle of EVs from donor cells to distant recipient cells. We confirmed the involvement of EVs in this process, isolating EVs from the conditioned media of osteoblasts, and showing their ability to directly uptake, activate, and retain cytoplasmic dyes, such as CMFDA. These results highlighted that our vesicular structures were cellderived and that they showed integrity, as shown by the presence of CD63 and by the TEM analysis. These EVs were able to integrate into target cells, including osteoblasts, monocytes, osteoclasts, and endothelial cells, presumably transferring their cargoes. Confocal microscopy analysis showed different patterns of dye accumulation in target cells, suggesting that the integration of EV components is a dynamic and specific process. Different localization of dyes, with or without colocalization in different cell compartments, suggested that EV constituents are recycled in their target cells, likely in a programmed and regulated way, according to their needs.
It has been reported that EVs can exert their effect by ligand/ receptor interaction. (16,17,27) Here we found evidence that in bone this type of intercellular communication includes the Rankl/Rank axis. In fact, FACS analysis showed that osteoblastderived EVs contain Rankl on their surface. Previous reports have shown Rankl presence on EVs from the stromal/osteoblast cell lines UAMS-32P and SV-HFO. (27,28) Although these cells are   established osteoblast-like lines, we believe that they cannot totally mirror the physiology of primary osteoblasts due to acquired genetic aberrations, being similar to tumor cells or viral-gene manipulated cells. (27) We found Rankl on osteoblast-derived EVs in basal conditions, and this expression was upregulated by PTH. Our results showed that Rankl-positive EVs affected osteoclast function in vitro, increasing their size and number of nuclei. Moreover, the TRAcP staining appeared more intense in EV-treated osteoclasts, suggesting more metabolically active cells. The effect seems to be Rankl-dependent, because rankl -/-EVs did not preserve osteoclast vitality, although the rate of EV retrieval was similar in wild-type and rankl/ -/osteoblasts. Many osteoclasts showed signs of sufferance under rankl -/-EV treatment, with cellular retraction and loss of nuclei. These data are consistent with the deprivation of the well-known osteoclast pro-survival function of Rankl. (42) Interestingly, osteoblast EVs appear also to affect the expression of osteoblast-specific genes, suggesting selective autocrine regulation of osteoblast gene expression by means of their own EVs.
A key evaluation of the effect on the osteoclast lineage was represented by the in vivo administration of wild-type osteoblast EVs in rankl -/mice. The rankl -/mouse is a model of human osteoclast-poor autosomal recessive osteopetrosis. (26) Like its human counterpart, this mouse model is characterized by the absence of the osteoclast lineage, and their bones are totally TRAcP-negative. We found an EV density-dependent appearance of TRAcP-positive cells in tibias of rankl -/mice treated with escalating numbers of our EVs, suggesting a biologic effect of the EV-shuttled Rankl. Ex vivo tissue-specific integration experiments in calvariae confirmed the ability of EVs to integrate into the bone cells and in vivo distribution experiments established that osteoblast-derived EVs are preferentially taken up very rapidly by the bone, further suggesting a tissue targeting specificity of the EVs released by osteoblasts.
EVs are known to contain and shuttle RNAs. Many reports over recent years have described the molecular content of EVs isolated from different cells, tissues, and biologic samples. (35) Web compendia about these data, such as ExoCarta (http:// exocarta.org/) and Vesiclepedia (http://microvesicles.org/), display growing reporting lists of lipids, RNAs and proteins recognized in various classes of EVs.
To the best of our knowledge, no studies to date have investigated the mRNA profile of EVs from bona fide primary osteoblasts. Here we found that our EVs contain a set of transcripts related to osteoblast activity. Although we are aware that our analysis is limited only to a very partial gene panel and does not measure the absolute quantity of transcripts, we believe that these results could have important biologic relevance. The 15 most abundant mRNAs in osteoblast EVs comprise a set of genes clearly associated with bone metabolism and osteoblast function. Col1a1 (ID: 12842) encodes the alpha-1 subunit of type I collagen, the most abundant extracellular matrix protein, mainly contained in bone, skin, and tendon. (43) Col1a2 (ID: 12843) encodes the alpha-2 subunit of the same type I collagen and, in association with the alpha-1 subunit, forms heterotrimeric type I procollagen for fibril formation. (44) Bglap (ID: 12096) encodes osteocalcin, one of the most abundant noncollagenous proteins in the mineralized matrix of bone. (44) Osteocalcin is known to exert endocrine activities, regulating energy metabolism, neural development, male fertility, and muscle function. (45)(46)(47) Our observation that osteocalcin transcriptional expression is modulated in osteoblasts by their own EVs suggest that EVs could also contribute to the control of the osteoblast endocrine function. Sparc (ID: 20692) encodes for the secreted acidic cysteine-rich glycoprotein, or osteonectin, that in bone is involved mainly in the interaction between mineral crystals and collagen, probably triggering the nucleation of hydroxyapatite. (48) Secreted phosphoprotein 1, or osteopontin, is the product of the spp1 gene (ID: 20750). It plays a role as an integral part of the mineralized matrix and is involved in important cell-matrix interactions. (49) In fact, it has been reported that osteopontin works as a "bridge" between extracellular matrix and cells, acting as a ligand of integrin heterodimers, such as avb3 or avb5, (50)(51)(52)(53) and driving osteoblast differentiation. (54) A set of four transcripts are from structural and housekeeping genes, such as Gapdh (ID: 14433, glyceraldehyde-3-phosphate dehydrogenase), Actb (ID: 11461, actin beta), B2m (ID: 12010, b-2 microglobulin) and Hsp90ab1 (ID: 15516, heat shock protein 90 alpha [cytosolic] class B member 1). A further comparison between EV and cellular transcriptomes did not reveal any preferential sorting or enrichment of mRNAs in EVs; therefore, the osteoblast EVs seem to mirror the transcript profile of their donor cells. This is at variance with what has been described in other cells, in which selective sorting of molecules into EVs was observed. (36) Due to their ability to shuttle specific cargoes and because of their targeted uptake in tissues, EVs are under intense investigation for various clinical applications. They could be used as "biomarkers" for diagnostic purposes, because they can reflect the molecular content of the donor tissues and are released into the circulation. Therefore, they could be useful in "liquid biopsies" to monitor the progression of diseases or scrutinize the success of therapies. (1,5) Furthermore, they can be loaded with tracking molecules and drugs, widening the therapeutic index, narrowing the median effective dose (ED 50 ) of drugs and reducing the risk of side effects because of their targeted delivery. (8) Here, we showed that osteoblast-derived EVs can be loaded with anti-osteoclastic agents. In particular, we loaded osteoblast-derived EVs with dasatinib, a clinically approved inhibitor of tyrosine kinases for malignancies, (55)(56)(57)(58)(59) or with zoledronate, a bisphosphonate clinically approved for osteoporosis. (60)(61)(62) Both agents are known to be very effective against osteoclasts. When encapsulated in EVs, both dasatinib and zoledronate still exerted their biologic effects. Das-EVs induced massive retraction and collapse of cells due to the inhibition of crucial tyrosine kinases, such as c-Src, involved in osteoclast cytoskeletal remodeling and actin ring formation. (31,63) Zol-EVs induced vacuolization and apoptosis of osteoclasts, consistent with inhibition of the mevalonate and lipid pathways known to be disrupted by N-bisphosphonates. (64,65) The EV-encapsulated drugs maintained their efficacy, as shown by the degeneration of osteoclasts similar to that induced by the free drugs. Most importantly, this antiosteoclastic activity was observed also in vivo in an acute mouse model of osteoclast over-activation, thus supporting their in vivo use to shuttle pharmacological agents to bone cells.
The proof of principle that EVs can have a potential therapeutic effect in bone pathology was further uncovered by our in vivo results in Rankl-deficient mice. Mice harboring mutations in the tnsf11a gene are the genocopy/phenocopy of human autosomal recessive osteopetrosis 2 (OMIM: #259710, OPTB2). (26) Presently, the patients are untreatable, have a very poor quality of life and a short life expectancy. They could take advantage of substitutive therapy with soluble Rankl, (66) but a pharmacological formulation of Rankl is not under development. On this basis, the hypothesis of the use of clinicalgrade EVs from healthy donors or engineered from patient cells to carry the wild-type protein, could help setting a new strategy for treating, or at least alleviating the symptoms, of this lethal disease. More widely, we can image EVs as a shuttle for delivery of membrane-bound molecules, cytokines, and growth factors like the Rankl reported in this study.
In conclusion, this report demonstrates that osteoblast EVs play a physiologic role in the interaction with other cells in the bone microenvironment, especially monocytes and osteoclasts. Moreover, we reported a proof of concept for the use of osteoblast-derived EVs for the treatment of bone diseases, because they are able to efficiently target the bone and induce clear biologic events, both by endogenous (Rankl) and exogenous (bisphosphonates and tyrosine kinase inhibitors) molecules, opening an avenue for their biotechnological use to treat skeletal disorders.

Disclosures
All Authors declare no conflict of interest.