Exosomes derived from osteogenic tumor activate osteoclast differentiation and concurrently inhibit osteogenesis by transferring COL1A1‐targeting miRNA‐92a‐1‐5p

2.1. Cell culture

In this study, we used four human PCa cell lines (MDA PCa 2b, C4‐2, PC3, and LNCap AI+F) and normal immortalized cells from the prostate (RWPE‐1 cells). MDA PCa 2b, C4‐2, PC3 and RWPE‐1 cells were obtained from American Type Culture Collection (ATCC, USA). LNCap AI+F cells (Xu et al., 2010) were a kind gift from Prof. Zhihua Tao at Zhejiang University, China. MDA PCa 2b cells were cultured in F12K medium (ATCC) containing 20% fetal bovine serum (FBS), according to ATCC instructions. C4‐2 and PC3 cells were cultured in RPMI 1640 medium (Gibco, USA) containing 10% FBS. RWPE‐1 cells were cultured in H‐DMEM (Gibco) containing 10% FBS. LNCap AI+F cells were cultured in F12K medium (ATCC) containing 10% charcoal stripped FBS (Gibco). The Raw264.7 cell line was maintained in H‐DMEM (Gibco) supplemented with 10% FBS. The MC3T3‐E1 clone 4 cell line was maintained in α‐MEM (Invitrogen, USA) containing 10% FBS. All culture media were replaced with fresh medium every 3 days and supplemented with 100 U/ml penicillin and 100 μg/ml streptomycin (HyClone, USA).

2.2. Exosome isolation, qualification and characterization

Exosomes were isolated from cell culture supernatants with 10% exosome‐depleted FBS (VivaCell Biosciences, China) through differential centrifugation, as described (Lötvall et al., 2014). After 48 h of culture, when cells were 90% confluent, 210 ml of the supernatant was harvested and centrifuged at 300 × g for 10 min, 2000 × g for 10 min, and then 10,000 × g for 30 min. Next, the supernatant was centrifuged at 120,000 × g for 70 min (Type 45 Ti Rotor, Beckman Coulter, USA), the pellet was resuspended in 35 ml of Phosphate buffered saline (PBS), and the sample was centrifuged again at 120,000 × g for 70 min. The supernatant was carefully removed, and the sample was resuspended in 200 μl of PBS. Exosomes isolated from human serum samples were obtained from six diagnosed PCa patients. Informed consent was obtained from all patients. Patient serum exosomes were isolated as described above and resuspended in 100 μl PBS. A BCA Protein assay kit (Pierce™, Thermo Fisher Scientific, USA) was used for exosome quantification, and exosome morphology was analysed using transmission electron microscopy (TEM) (Tecnai, USA), as previously described (Shelke et al., 2014). To examine exosome size distribution and particle concentration, nanoparticle tracking analysis (NTA) was performed using a ZetaView instrument (Particle Metrix, Germany).

2.3. Western blotting

Exosomes (10 μg) or proteins (30–50 μg) extracted from cells were diluted 1:4 (v/v) in a loading buffer (Beyotime, China), heated at 95°C for 10 min, loaded onto 10% gels (1.5 × 10 wells; Genscript, China), electrophoresed at 180 V for 50 min, and transferred to membranes by using the iBlot (Invitrogen) 8‐min program. For all subsequent antibody‐incubation and washing steps, a rocking platform was used. After blocking with a blocking buffer (Odyssey, USA) for 60 min, the membranes were incubated with primary antibodies at 4°C overnight, washed in TBST (3 × 5 min), incubated with secondary antibodies at room temperature for 60 min, washed with TBST, and then exposed to a detection reagent (Chemiluminescent HRP Substrate, Millipore, USA) for visualization. The primary antibodies were as follows: CD9 (Cell Signaling Technology, USA, 1:1000), Alix (Cell Signaling Technology, 1:1000), Annexin V (Cell Signaling Technology, 1:1000), HSP70 (Cell Signaling Technology, 1:1000), GM130 (Cell Signaling Technology, 1:1000), COL1A1 (Cell Signaling Technology, 1:1000), GAPDH (Cell Signaling Technology, 1:1000), β‐Actin (Cell Signaling Technology, 1:1000), and RANKL (Proteintech, China, 1:3000).

2.4. Uptake of exosomes

Exosomes (10 μg) were labelled using a PKH67 Green Fluorescent Cell Linker Kit (Sigma‐Aldrich), as per the manufacturer's protocol. Briefly, exosomes in PBS were added to 0.5 ml of Diluent C, and 4 μl of PKH67 was added to 0.5 ml of Diluent C; the two reaction mixtures were incubated in the dark at room temperature. After 4‐min incubation, 2 ml of 0.5% BSA was added to prevent excessive labelling, and then the labelled exosomes were resuspended in 35 ml of PBS and centrifuged 120,000 × g for 70 min to remove the residual dye. The recipient cells were mixed with the labelled exosomes and incubated in the dark. After aspirating the cell supernatant, the cells were washed with PBS once and fixed with 4% paraformaldehyde (PFA) for 10 min, and then the nucleus and the cytoskeleton were stained with DAPI and fluorophore‐conjugated phalloidin (to visualize F‐actin), respectively. Stained samples were examined using a confocal microscope (Nikon, Japan).

2.5. Conditioned medium (CM) preparation

CM was harvested as described (Liu et al., 2015). Briefly, 5 × 10⁶ MDA PCa 2b cells were cultured in 75‐cm² cell‐culture flasks in F12K medium containing 20% FBS for 12 h, and then the medium was changed to 10 ml of F12K plus 0.5% FBS with/without 20 μM GW4869 (MCE, USA). After 48 h, the supernatant was collected and centrifuged at 2,500 rpm for 10 min to remove cell debris and then filtered (0.2‐μm filter, Millipore) before use in experiments.

2.6. In vitro osteoclast induction of Raw264.7 cells

Osteoclast induction of Raw264.7 cells was performed as described (Sun et al., 2016), with modifications. Briefly, Raw264.7 cells were seeded into 96‐well plates at a density of 1 × 10⁴ cells/well, allowed to attach for 24 h, and then cultured in the presence or absence of recombinant murine sRANKL (100 ng/ml, PeproTech, USA) in combination with exosomes or miRNAs for 6 days. The medium was changed every 2 days, and after 6 days of culture, osteoclast differentiation was assessed by means of Trap staining and Trap‐activity detection (Beyotime).

2.7. In vitro osteoclast induction of bone marrow macrophages (BMMs)

We prepared bone marrow cells and cultured BMMs according to previously reported protocols (Li et al., 2016), with minor modifications. Briefly, bone marrow cells were isolated from the tibia and femur of 8–12‐week‐old male mice and cultured in α‐MEM (Invitrogen) containing 10% FBS for 24 h to generate BMMs. To produce osteoclasts, BMMs were cultured in α‐MEM supplemented with recombinant murine M‐CSF (25 ng/ml, PeproTech) and recombinant murine sRANKL (50 ng/ml, PeproTech) with or without treatment with exosomes or miRNAs for 6 days. The culture medium was changed every 2 days until Day 6, and then Trap staining was performed.

2.8. Trap staining

Trap staining for osteoclasts was performed according to the manufacturer's instructions (Sigma‐Aldrich), with minor modifications. Cells were fixed with Fixative Solution (a mixture of 4.05 ml of acetone, 1.55 ml of citrate solution, and 0.5 ml of formaldehyde) for 3 min at room temperature and then washed with deionized water once. Next, 0.05 ml of Fast Garnet GBC base solution and 0.05 ml of sodium nitrite solution were mixed for 2 min and then mixed with another solution containing 4.5 ml of 37°C deionized water, 0.05 ml of naphthol AS‐BI phosphate solution, 0.2 ml of acetate solution, and 0.1 ml of tartrate solution. Lastly, the fixed cells were incubated in the final mixed solution for 1 h at 37°C.

2.9. In vitro osteogenic induction of MC3T3‐E1 cells

MC3T3‐E1 cells were seeded at 1 × 10⁴ cells/well into 24‐well plates on Day 0, cultured in α‐MEM until 80%–90% confluence, and then changed into an osteogenic differentiation medium (Cyagen Biosciences, China). After induction for 21 days, Alizarin Red S staining was performed as previously reported (Gao et al., 2019). Briefly, on Day 21, cells were fixed with 4% PFA for 15 min, incubated with Alizarin Red S (Cyagen Biosciences) for 3 min, and washed with deionized water. To quantify mineralization, 2% hexadecyl pyridinium chloride monohydrate solution was added and the 490‐nm absorbance was measured. For alkaline phosphatase (ALP) staining, cells were stained using a BCIP/NBT Alkaline Phosphatase Color Development Kit (Beyotime) after 7 days of osteogenic induction, and ALP activity was measured using an Alkaline Phosphatase Assay Kit (Beyotime), according to the manufacturer's instructions.

2.10. Library preparation and sequencing

Small‐RNA sequencing was conducted by a commercial service (RiboBio, China). Briefly, total RNAs were extracted from MDA PCa 2b exosomes and RWPE‐1 exosomes, and the quality and concentration of the RNAs were checked using an Agilent 2200. After reverse‐transcription‐PCR amplification, small‐RNA libraries were prepared and sequenced, with raw reads being collected using an Illumina HiSeq 2500 platform.

2.11. Differential expression analysis of miRNAs

To validate miRNAs identified through RNA sequencing, RNA was extracted using a Total Exosome RNA and Protein Isolation Kit (Invitrogen) and then a Mir‐X™ miRNA First Strand Synthesis kit (Takara, Japan) and SYBR® Premix Ex Taq™ II (Takara) were used for qPCR‐based measurement of the expression levels of miR‐92a‐1‐5p, miR‐148a‐3p, and miR‐375 in exosomes. U6 and cel‐miR‐39 were used as controls. The miRNA qPCR primer sets were purchased from RiboBio and mRNA qPCR primers were synthesized by Sangon (Shanghai, China). Table S1 lists the qPCR primers used in this study.

2.12. Target‐gene prediction and KEGG/GO pathway enrichment

Targetscan, miRanda, miRWalk, and miRDB were used for miRNA target‐gene prediction. We used the tool DIANA to discover the pathways enriched by the miRNA targets (http://www.microrna.gr/miRPathv3) (Vlachos et al., 2015), and then used David 6.8 Bioinformatics Resources (https://david.ncifcrf.gov/) to analyse the functional annotations (Huang et al., 2009).

2.13. Lentivirus and oligonucleotide transfection

Lentiviral plasmids encoding miR‐92a‐1‐5p and a negative control were constructed by HANBIO (Shanghai, China). Raw264.7 and MC3T3‐E1 cells were transfected with the lentivirus (pHB‐U6‐mir‐92a‐1‐5p‐EF1‐LUC‐PURO) at a multiplicity of infection of 10, after which selection was performed using 4 μg/ml puromycin for 6 days. Synthetic miRNA mimics, miRNA inhibitors, and control miRNAs (RiboBio) were synthesized and the oligonucleotides were transfected using a RiboFECT CP Transfection Kit (RiboBio).

2.14. Luciferase reporter assay

Cells were co‐transfected with luciferase vectors containing wild‐type or mutant 3ʹ‐UTR of COL1A1 and miR‐92a‐1‐5p mimics or miR‐Control by using Lipofectamine 3000 (Invitrogen). Luciferase activity was measured using a Dual‐Luciferase Reporter Assay System (Beyotime) at 48 h after transfection.

2.15. COL1A1 knockdown by siRNA

The siRNA oligonucleotides for mouse COL1A1 and siRNA control were purchased from Ribobio. MC3T3‐E1 and Raw264.7 cells were transfected using a RiboFECT CP Transfection Kit (RiboBio), followed by validation of mRNA change at 36 h after transfection using qPCR and protein change at 48 h using western blotting.

2.16. Animal studies

All experiments were performed following the animal welfare guidelines approved by the Intramural Animal Use and Care Committee at Air Force Medical University. All efforts were made to reduce animal suffering to the extent possible. Male BALB/C nude mice (6 weeks old; 18–20 g) were purchased from SJA Laboratory Animal Co. (Changsha, China) and randomly assigned to different treatment groups. The mice were acclimatized to the specific‐pathogen‐free animal housing facility for one week before the experiments were started.

2.17. Exosomes labelling and tracking in vivo

To visualize the target organ and the metabolism process of exosomes in vivo, exosomes were labelled according to the instructions of the manufacturer, Life Technologies (USA), with modifications. Briefly, 100 μg of exosomes were incubated with Vybrant DID (1:1000 in PBS) in the dark for 20 min, and then the labelled exosomes were washed in 35 ml of PBS, with centrifugation at 120,000 × g for 90 min, to remove the excess dye. Next, the Vybrant DID‐labelled exosomes were injected into the tail vein of BALB/C nude mice (7 weeks old, n = 3 per group/time point, dosage per mouse: 100 μg of exosomes, in 150 μl of PBS), and PBS with/without Vybrant DID was used as the control in all experiments. At 4, 24, and 48 h after injection, the mice and the harvested tissues were subject to in vivo and ex vivo imaging. Fluorescence intensity was determined using an IVIS Spectrum system and Living Image Software (PerkinElmer, USA).

2.18. Enrichment of bone marrow cells and flow cytometry

To verify that the detected fluorescence was from bone marrow cells that had internalized Vybrant DID‐labelled exosomes, mice were euthanized at 24 h post‐injection and the femur and tibia were flushed with PBS by using a syringe attached with a 27‐G needle. Cell suspensions were passed through a 70‐μm cell strainer (BD Biosciences, USA) and red blood cells were lysed using a Red Blood Cell Lysis Buffer (Beyotime), and then bone marrow cells were assessed using flow cytometry.

2.19. Immunofluorescence staining of tibia

To visualize the colocalization of Vybrant DID‐labelled exosomes and bone stromal cells, immunofluorescence analysis of the tibia was performed as reported (Sui et al., 2016), with modifications. The left tibia was isolated at sacrifice 24 h after injection of labelled exosomes, fixed in 4% PFA, decalcified rapidly, and embedded in paraffin wax; subsequently, specimens were sectioned at 3‐μm thickness by using a tissue slicer (Leica, Germany) away from light. Without quenching the spontaneous fluorescence (red) from Vybrant DID‐labelled exosomes, nuclei and bone stromal cells in the bone were stained with DAPI (blue) and anti‐osteocalcin (OCN)/RANKL/cathepsin K (CTSK)/TRAP (green), respectively. Immunofluorescence labelling for COL1A1 and type I collagen was performed using a routine staining method. Samples were examined using a fluorescence microscope (Nikon) or scanned using a case viewer system (3DHISTEC, Hungary).

2.20. Micro‐computed tomography (micro‐CT) analysis

To evaluate bone mass, a micro‐CT system (GE Healthcare, USA) was used as reported (Bouxsein et al., 2010; Sui et al., 2016), with modifications. Mice were sacrificed and the right tibia was removed and fixed in 4% PFA for 24 h, and then 7‐mm blocks of the specimen were prepared. Subsequently, the specimens were imaged using a 120‐min micro‐CT scan, at a resolution of 8 μm, a current of 80 μA and a voltage of 80 kV. Micview V2.1.2 software was used for data analysis, and the following parameters were quantified: bone mineral density (BMD), bone volume per tissue volume (BV/TV), bone surface per bone volume (BS/BV), trabecular number (Tb.N), trabecular thickness (Tb.Th), and trabecular separation (Tb.Sp) (Bouxsein et al., 2010).

2.21. Mouse bone marrow education

Mice (7 weeks old, n = 3–6 per group) were injected with a small dose of exosomes, as previously reported (Costa‐Silva et al., 2015; Peinado et al., 2012; Wen et al., 2016); 10 μg of total exosomes (in 100 μl of PBS) were injected into the tail vein thrice weekly for 4 weeks, and after this initial treatment, 50 μg of exosomes (in 150 μl of PBS) were injected once a week. PBS was used as the control. On Days 28, 42, and 56, mice were sacrificed and specimens were collected for different experiments.

2.22. Intratibial (i.t.) injection

We performed i.t. injections as reported (Berlin et al., 1993; Wu et al., 1998), with minor modifications. Briefly, after the 4‐week bone marrow education, BALB/C nude mice (n = 6 per group) were anesthetized using sodium pentobarbital and the right legs were cleaned with iodine solution. PCa cells infected with pHB‐U6‐mir‐92a‐1‐5p‐EF1‐LUC‐PURO lentivirus (HANBIO) were aspirated into a 25‐μl Hamilton syringe (Alltech Associates, Switzerland) mounted with a 27‐G needle, and the needle was inserted into the cortex of the anterior tuberosity with a rotating “drill‐like” movement and inserted down 3–5 mm once the bone cortex was traversed (Berlin et al., 1993). Subsequently, 25 μl of the cell suspension (2.5 × 10⁶ cells/inoculum) was slowly injected into the bone marrow space, after which the needle was removed from the bone, and bone wax was applied to the injection area.

2.23. Orthotopic injection

Male BALB/C nude mice (7 weeks old, n = 6 per group) were anesthetized using sodium pentobarbital, the abdomen was surgically prepared with iodine solution, and a surgical incision was made in the lower abdomen to expose the prostate. Subsequently, miR‐92a‐1‐5p‐overexpressing MDA‐PCa‐2b‐luc cells (10⁶ cells in 10 μl) were injected into each dorsal prostate lobe by using a Hamilton syringe attached with a 27‐G needle.

2.24. In vivo imaging

Tumor growth was detected using an IVIS Spectrum system (Perkin Elmer), as previously reported (Dai et al., 2019). Briefly, D‐luciferin at a working dose of 150 mg/kg (in 100 μl of PBS) was intraperitoneally injected, and 1.5% isoflurane was used for anesthetization before imaging. Tumor growth and tumor burden were measured based on bioluminescence imaging by using Living Image Software (Perkin Elmer).

2.25. Statistical analysis

All data are presented as means ± standard deviation. Comparisons were performed using t test and one‐way ANOVA as appropriate, by using GraphPad Prism version 7.0. P < 0.05 was considered statistically significant (*, P < 0.05; **, P < 0.01; ***, P < 0.001).

3.1. Characteristics of exosomes derived from osteoblastic, osteoclastic, and mixed PCa cells

PCa cell lines are frequently divided into three groups: osteoblastic, osteolytic, and mixed cell lines. According to previous studies, MDA PCa 2b cells exhibit the osteoblastic phenotype of PCa cells and induce osteogenesis when implanted into the mouse tibia (Figure 1a) (Navone et al., 1997). By contrast, two other PCa cell lines, PC3 and DU145, are known to induce osteolytic lesions in bone metastasis (Figure 1a) (Fisher et al., 2002; Russell & Kingsley, 2003). PCa cell lines featuring a mixed phenotype, LNCap and its derivative C4‐2/C4‐2B, produce mixed lesions (Thalmann et al., 1994).

We isolated exosomes derived from osteoblastic, osteoclastic, and mixed PCa cells and then characterized single vesicles within each mixture by using TEM. The exosomes isolated from MDA PCa 2b, C4‐2, PC3, and LNCap AI+F cells appeared homogenous and displayed an ovoid morphology (Figure 1b). Moreover, examination of the size distribution of exosomes by using NTA revealed that the majority of the population was in the ∼90–130‐nm range (Figure 1c), which is within the characteristic diameter range of 30–150 nm of these vesicles. As recommended (Lötvall et al., 2014), the values acquired using NTA were compared with the TEM results to verify that the isolated vesicles were not non‐membranous particles of a similar size. To further confirm that exosomes had been isolated, we used western blotting to test for the presence of several “exosome markers”: a transmembrane protein (CD9), three cytosolic proteins (Annexin V, Alix, HSP70), and an intracellular protein (GM130) (Figure 1d). All exosomes isolated from the four types of PCa cells expressed CD9 and Alix, although the exosomes showed clear differences in the enrichment of the proteins; furthermore, the exosomes from MDA PCa 2b, PC3, and LNCap AI+F cells were positive for Annexin V, whereas C4‐2 exosomes were negative for this marker. Our results also agree with the report that GM130 is not expected to be enriched in exosomes (Lötvall et al., 2014). HSP70 was not expressed in PCa exosomes, which is consistent with our previous results (data not shown). Collectively, these results on the morphology, size, and marker expression of the isolated vesicles confirmed that the vesicles were exosomes. Lastly, quantification of the isolated exosomes by using the BCA method revealed that the exosome yield differed markedly between PCa cells and normal prostate cells (RWPE‐1), with PCa cells producing ∼3‐fold more exosomes (Figure 1e).

3.2. PCa exosomes target bone and are internalized by bone stromal cells in vivo

To elucidate the roles of PCa exosomes, it is critical to first determine the distribution and metabolism of PCa exosomes in vivo. We determined the tissue distribution of PCa exosomes by labelling the exosomes with Vybrant DID, a lipid‐based fluorescent dye, injecting the labelled exosomes into the tail vein of BALB/C nude mice (7 weeks old, n = 3 per group/time point), and assessing the biodistribution of the exosomes in vivo and ex vivo at 4, 24, and 48 h post‐injection. The in vivo fluorescence signal was only observed in the bones of the PCa‐Exos groups (MDA PCa 2b Exos group and LNCap AI+F Exos group), but not the RWPE‐1 Exos group which did not localize to the bone, indicating this bone target is PCa specific (Figure 2a). Next, bones were harvested and their fluorescence intensity was measured ex vivo; at both 24 and 48 h, the average fluorescence intensity of the PCa‐Exos groups was elevated substantially (Figure 2b,c). Examination of the dynamics of bone fluorescence intensity in the PCa‐Exos groups (Figure 2d) revealed that the intensity peaked at 24 h and then decreased drastically by 48 h. Therefore, we selected the 24‐h time point for further analysis; this time point is the same as that used in most studies (Inder et al., 2014; Wen et al., 2016), although (to our knowledge) the peaking of the signal at this time point has not been previously verified. We also evaluated the fluorescence signal from various soft organs (heart, liver, spleen, kidney, stomach, and intestine) and found no differences between the groups (Figure S1a). At 24 h post‐injection, no pathological changes were detected in these soft organs in the PCa‐Exos groups, which suggested that the exosomes are appropriately targeted and show almost no systemic toxicity (Figure S1b). Overall, these results suggested that PCa‐derived exosomes, relative to normal exosomes, exhibit a high tendency for bone‐specific targeting.

Next, we determined whether the observed fluorescence was derived specifically from cells that had internalized the fluorescently labelled exosomes. Flow cytometry was used to measure the fluorescence signal of bone marrow cells flushed from long bones; the fluorescence level was higher in the MDA PCa 2b‐Exos group than in the Vybrant DID group (Figure 2e), which indicated that the red fluorescence was from bone marrow cells that had internalized PCa exosomes. We also sought to identity the types of bone marrow cells that had taken up the exosomes. Without quenching the spontaneous fluorescence from the labelled MDA PCa 2b exosomes (red), bones were subject to immunofluorescence labelling by using anti‐OCN/RANKL/CTSK/TRAP (green), and nuclei were stained with DAPI (blue). The fluorescence of the labelled exosomes colocalized with the signal from osteoblasts and osteoclasts (white arrows indicated), which confirmed that PCa exosomes were internalized by bone stromal cells following tail‐vein injection (Figure 2f‐i).

3.3. PCa exosomes are internalized rapidly by Raw264.7 cells and facilitate osteoclast differentiation in vitro

Based on the PCa‐exosome organotropism detected here with bone and bone stromal cells, we investigated whether PCa exosomes are involved in osteoclast differentiation. First of all, we prepared MDA PCa 2b CM from cells cultured with or without the exosome inhibitor GW4869. The decrease of exosome yield by GW4869 was confirmed using BCA Protein Assay and NTA (Figure S2a and S2b). After priming with 100 ng/ml RANKL for 2 days, osteoclast differentiation of Raw264.7 cells was increased following culture with MDA PCa 2b CM for 48 h, but this effect was reversed by GW4869 (Figure 3a). We further observed that MDA PCa 2b exosomes were rapidly internalized by Raw264.7 cells (Figure 3b, upper); the uptake occurred within only 90 min of culture, which was a shorter time than that reported for other recipient cells (Kalra et al., 2019; Raimondi et al., 2015). The results of ImageJ analysis showed that the uptake peaked at 24 h and then decreased markedly by 48 h (Figure 3b, lower). Next, we sought to examine whether cell fusion occurred during osteoclast differentiation induced by PCa exosomes. After culture for 24 or 48 h with exosomes alone, Raw264.7 cells were fixed with 4% PFA and stained for F‐actin and nuclei (with DAPI) and examined under a confocal microscope; our result showed that MDA PCa 2b exosomes induced the fusion of Raw264.7 cells at both 24 and 48 h (Figure 3c), whereas exosomes derived from normal RWPE‐1 cells did not induce cell fusion (Figure 3d).

RANKL has been reported to play a crucial role in osteoclast differentiation (Boyle et al., 2003). BMMs (bone marrow macrophages) and Raw264.7 cells (murine macrophage cell line) are pre‐osteoclasts, often used for osteoclastogenesis. Here, we first checked for Trap⁺ osteoclasts after culturing with various concentrations of RANKL (0, 20, 50, 100, 150, 200 ng/ml) and then selected 100 ng/ml RANKL for further studies (Figure S2c). To observe the mRNA expression of osteoclast markers, Raw264.7 cells were incubated with 100 ng/ml RANKL and (or) 20 μg/ml MDA PCa exosomes for 6 days. Next, qPCR analysis revealed that MDA PCa 2b exosomes markedly induced the mRNA expression of Cstk and Trap in Raw264.7 cells after a 6‐days incubation (Figure 3e,f). The synergistic effect by RANKL and exosomes was observed by Trap mRNA but not observed significantly by Ctsk mRNA, indicating that PCa exosomes may promote osteoclastogenesis in a RANKL independent manner (Figure 3e,f). Furthermore, our results demonstrated that MDA PCa 2b exosomes induce osteoclast differentiation of Raw264.7 cells in the presence of RANKL in a dose‐dependent manner (Figure 3g). We also verified that MDA PCa 2b exosomes induced the osteoclast differentiation of BMMs in the presence of RANKL and M‐CSF (Figure 3h). We further confirmed that exosomes from mixed C4‐2/LNCap AI+F cells and osteoclastic PC3 cells also induced the osteoclast differentiation of BMMs (Figure S2d, Figure 3h) and Raw264.7 cells (Figure S2e). To determine whether PCa exosomes alone can induce osteoclast differentiation, we cultured Raw264.7 cells with exosomes for 6 days and then performed Trap staining and quantified Trap⁺ osteoclasts (Figure 3i). Unexpectedly, osteoclastogenesis was also induced by PCa exosomes in the absence of RANKL (Figure 3j). Next, our results inspired us to determine whether RANKL is transferred by PCa exosomes. As examined by western blotting, RANKL was expressed at low level in all three types of PCa cells and no expression was found in PCa exosomes (Figure S2f). qPCR analysis also demonstrated that RANKL was lowly expressed in PCa cells with an ∼28 ct value in 0.2 μg total RNA (Figure S2g). This result contradicts the accepted notion that RANKL is necessary for osteoclast differentiation, but is in accord with the results shown in Figure 3c. In summary, we used Raw264.7 cells and BMMs, two commonly used models for osteoclast differentiation (Song et al., 2019), to verify that exosomes derived from osteoblastic, osteoclastic, and mixed PCa cells can induce osteoclast differentiation both in the absence and presence of RANKL.

3.4. PCa exosomes inhibit osteoblastogenesis of MC3T3‐E1 cells in vitro

To further analyse the observed exosome internalization by bone stromal cells, we investigated the potential effects of the exosomes on osteoblasts. We examined the differentiation of MC3T3‐E1 cells, which represent a murine pre‐osteoblast cell line and can differentiate into mature osteoblasts in osteogenic induction medium containing β‐glycerophosphate and ascorbic acid (Wang et al., 1999).

PCa cell‐derived exosomes labelled with PKH67 were internalized by MC3T3‐E1 cells after 4‐h incubation at 37°C (Figure 4a); the internalization was inhibited when cells were incubated at 4°C, which indicated that exosome uptake by the recipient cells was a biologically active process. Next, we cultured the cells with the exosomes in normal medium for 6 days and then measured the mRNA expression of the osteoblastic markers Alp, Ocn, Runx2, and Osx (Figure 4b–e); the results confirmed that all four markers were drastically downregulated, which indicated that the osteoblastogenesis of MC3T3‐E1 cells was inhibited by MDA PCa 2b exosomes. Moreover, in agreement with the qPCR data, the MDA PCa 2b exosomes decreased ALP activity at high or low concentration (Figure 4f). As a classical and recommended experiment of osteogenic differentiation, Alizarin Red S staining was also performed: When MC3T3‐E1 cells were cultured with MDA PCa 2b exosomes in osteogenic induction medium for 21 days, the cells showed markedly reduced capacity for mineralized nodule formation (Figure 4g). Furthermore, in agreement with the ALP mRNA expression, ALP staining was decreased after the cells were cultured with MDA PCa 2b exosomes (Figure 4h). We also examined the ALP activity and ALP staining of MC3T3‐E1 cells after treatment with exosomes derived from other PCa cells (C4‐2, PC3, and LNCap AI+F) (Figure S3). Overall, our results presented in the preceding subsection demonstrated that PCa exosomes accelerated osteoclast differentiation, and our results in this subsection further showed that PCa exosomes inhibited osteoblastogenesis.

To further strengthen our results on the role of PCa cell‐derived exosomes on bone homeostasis, we assessed the effect of exosomes isolated from the serum of six PCa patients (three localized PCa and three PCa with bone metastasis) on the osteoclast differentiation of BMMs in three independent experiments. The exosome yield was assessed by BCA protein assay; the amounts of exosomes from serum of PCa with bone metastasis were significantly higher (Figure S6a). Exosomes from localized PCa and PCa with bone metastasis both promoted osteoclast differentiation of BMMs (Figure S6b). Moreover, we discovered that exosomes from PCa with bone metastasis had a stronger ability to promote osteoclast differentiation in all three independent experiments. We also investigated the potential of bone metastatic PCa patient‐derived exosomes to inhibit osteoblastogenesis; ALP staining assay revealed a trend toward reduction (Figure S6c).

3.5. Osteogenic tumor‐derived exosomes induce osteolysis in vivo

Considering the key role of PCa exosomes in bone homeostasis in vitro, we sought to test whether PCa‐derived exosomes influence bone formation and bone lysis in vivo. We treated male BALB/C nude mice (7 weeks old, n = 3 per group/time point) with PBS (Control group), PCa exosomes, or normal exosomes through tail‐vein injection and then harvested the tibia for experiments (Figure 5a); we harvested the tibia after 4 weeks of injection of exosomes (10 μg in 100 μl of PBS, thrice weekly) or 100 μl of PBS alone. No differences between the groups was observed in morphological examinations of the mice and of bone or in X‐ray scanning (Figure S4a–c). The results of micro‐CT analyses showed that four measured parameters, BMD, BV/TV, Tb.N, and Tb.Th of diaphysis, from the right tibia were markedly lower in the MDA PCa 2b Exos group than in the control groups, whereas BS/BV and Tb.Sp of diaphysis were notably elevated in the MDA PCa 2b Exos group (Figure 5b‐g). These results indicated that osteolysis occurred in bone after 4‐weeks MDA PCa 2b exosomes education via tail‐vein. Representative 3D‐reconstruction models of the separated trabecular bones are shown in Figure 5h. At sacrifice, the left tibia was collected for H&E and TRAP staining and OCN immunofluorescence labelling. Increased severity of bone destruction in the MDA PCa 2b Exos group was indicated by H&E staining results (Figure S4d), and increased numbers of osteoclasts with fewer osteoblasts were indicated by the results of TRAP staining (Figure 5i) and OCN immunolabeling (Figure 5j). By contrast, H&E staining of soft tissues revealed no overt tissue damage, which confirmed the bone‐specific functional targeting of the PCa exosomes (Figure S4e).

After 4 weeks of injection, it became challenging to further inject the mouse tails with exosomes. Therefore, we altered our injection strategy to a single injection per week of a larger dose of exosomes (50 μg of exosomes in 150 μl of PBS, with 150 μl of PBS alone used as a control), and we harvested the tibia on Days 42 and 56 and performed micro‐CT analysis; representative longitudinal‐section images and 3D maps of the tibia are shown in Figure 5k,l, respectively. Bone loss was detected in the MDA PCa 2b Exos group at all three indicated time points as compared with both the Control group and the RWPE‐1 Exos group.

Next, we educated the bone marrow of BALB/C nude mice (7 weeks old, n = 6 per group) with C4‐2 and PC3 exosomes for 4 weeks and then harvested the right tibia and performed micro‐CT analysis. Osteolysis was verified based on the bone parameters BMD, BV/TV, BS/BV, and Tb.Th (Figure S5a–d), although Tb.N and Tb.Sp did not differ between the groups (Figure S5e,f). Representative 3D‐reconstruction models are shown in Figure S5g. Increase in osteoclasts was indicated by Trap staining images and quantification (Figure S5h‐j).

Collectively, these results confirmed that a small‐dose injection of PCa exosomes derived from osteogenic, osteoclastic, or mixed PCa cells (without the help of tumor cells) can markedly induce osteolysis and decrease bone formation in model mice; this agrees with our in vitro findings and is the opposite of what was previously accepted.

3.6. PCa exosomes accelerate tumor growth in bone

After successful bone marrow education and induction of osteoclastic lesions in BALB/C nude mice, we aimed to determine, based on our previous results, whether PCa exosomes can accelerate tumor growth in bone. After 4 weeks of bone marrow education, 2.5 × 10⁶ MDA PCa 2b cells infected with a lentivirus vector (GFP‐Puro‐Luc) were injected into the right tibia of BALB/C nude mice (n = 6 per group, start time point) (Figure 6a), and then in vivo imaging was performed from Day 1 to Day 56 post‐injection; the imaging on Day 1 was performed to ensure that the i.t. injection was successful and that the bioluminescence intensity did not differ between the different groups (Figure 6b). During Days 14–42, no bioluminescence was detectable in any of the groups, probably due to immune elimination, but then on Days 49 and 56 (n = 3 per group, end time point, three mice in the M‐Exos+M‐Cells group were dead before Day 49), bioluminescence was detectable only in the MDA PCa 2b group, which indicates that pre‐education with tumor exosomes accelerated tumor growth in bone (Figure 6b). On Days 49 and 56, the tumor burden in the different groups was quantified through in vivo imaging. Bioluminescence was detected both on Day 49 (Figure 6c) and Day 56 (Figure 6d), and the bioluminescence intensity was increased drastically (Figure 6e) in the M‐Exos+M‐Cells group, which indicated that the growth of the tumor cells was accelerated. When the mice of the MDA Exos group were sacrificed on Day 56, bone tumor was detected in 3D‐reconstruction images from micro‐CT analysis (Figure 6f), and H&E staining revealed the presence of prostatic tumor in bone (Figure 6g).

3.7. miRNAs transferred by MDA PCa 2b exosomes influence bone homeostasis

The results obtained thus far led us to investigate whether exosome cargos influence bone homeostasis and tumor development. Among the exosome cargos, miRNA have been studied most thoroughly and are highly conserved among species; they are small but powerful RNA molecules with multiple target genes that have been demonstrated to be critical regulator of multiple diseases. Therefore, we used next‐generation sequencing to profile small RNAs from MDA PCa 2b exosomes and RWPE‐1 exosomes, and we screened for differentially expressed miRNAs by using these criteria: |log2(FC)| > 2.5 (FC: fold‐change) and P < 0.001. Among the miRNA identified to be differentially expressed, 18 were upregulated and 30 were downregulated in MDA PCa 2b exosomes relative to controls (Figure 7a). Next, we focused on three specific miRNAs, miR‐92a‐1‐5p, miR‐375, and miR‐148a‐3p. miR‐92a‐1‐5, a newly identified and seldom reported member of the miR‐17‐92‐cluster miRNAs, has been shown to be associated with cancer (Liu et al., 2019); it was reported to be markedly downregulated in exosomes isolated from the urine of PCa patients and could serve as a potential noninvasive biomarker of PCa (Rodríguez et al., 2017). Our previous work demonstrated that urinary exosomal miR‐375‐3p is a potential biomarker that exhibits strong association with PCa clinical stage and bone metastasis (Li et al., 2019). miR‐148a‐3p was reported to be upregulated in the serum of PCa patients (Dybos et al., 2018). The results of qPCR analyses confirmed that miR‐92a‐1‐5p, miR‐375, and miR‐148a‐3p were expressed at higher levels in MDA PCa 2b exosomes than in RWPE‐1 exosomes (Figure 7b), and miR‐92a‐1‐5p was the most abundant miRNA among eight miRNAs verified to be upregulated by using qPCR (data not shown). To examine the potential roles of the three miRNAs in bone homeostasis, we first transfected MC3T3‐E1 cells with the miRNAs for 7 days; after the induction, ALP activity was measured, which revealed that miR‐92a‐1‐5p overexpression inhibited the osteogenic differentiation of MC3T3‐E1 cells, whereas overexpression of miR‐148a‐3p or miR‐375 enhanced osteogenic differentiation (Figure 7c).

Next, we used the tool DIANA to perform KEGG analyses for the target genes of the three miRNAs, and we identified a total of 22 pathways (P < 0.05). Considering the distinct roles of the three miRNAs in bone homeostasis, for further investigation, we selected these eight potential functional pathways: RNA degradation, Extracellular matrix (ECM)‐receptor interaction, Bladder cancer, p53 signalling pathway, Endocytosis, FOXO signalling pathway, Lysine degradation, and Viral carcinogenesis (Figure 7d). Our results identified 154 genes in the eight pathways as potential target genes of the three miRNAs. Moreover, to reveal the potential functions of these target genes, we performed KEGG and GO enrichment analyses by using the tool DAVID; the target genes were enriched in FOXO signalling pathway, angiogenesis, exosome biosynthesis, cytosis, and cancer‐associated pathways (P < 0.05) (Figure 7e–h). Notably, FoxOs (FoxO1, FoxO3, and FoxO4), which are expressed in numerous types of bone cells, have been widely shown to activate osteoblastogenesis and suppress osteoclastogenesis through several targets and pathways (Ma et al., 2020), and bone tissue regeneration is frequently linked with angiogenesis and cytosis. These results indicated that miRNAs delivered by PCa exosomes strongly and broadly affect bone homeostasis through multiple potential pathways and target genes.

3.8. miR‐92a‐1‐5p transferred by MDA PCa 2b exosomes facilitates osteoclast differentiation and inhibits osteoblast differentiation by targeting COL1A1

Lastly, we investigated the precise role in bone homeostasis and tumor development of miR‐92a‐1‐5p, which was the most abundant miRNA that we detected here, and which has been seldom examined in previous studies. To determine the function of miR‐92a‐1‐5p during osteoclast differentiation, we transfected Raw264.7 cells with miRNA mimics or inhibitors, and after culturing the cells with RANKL for 6 days, we used Trap staining to assess osteoclast formation (Figure 8a). Moreover, we quantified the mRNA expression of the osteoclast markers Ctsk and Rank and measured Trap activity (Figure 8b‐d). Overall, osteoclast differentiation was increased by miR‐92a‐1‐5p and inhibited by anti‐miR‐92a‐1‐5p. Furthermore, to elucidate the role of miR‐92a‐1‐5p during osteoblast differentiation, we transfected MC3T3‐E1 cells with the miRNA mimics and inhibitors, and after induction for 21 days, performed Alizarin Red S staining. Osteoblast differentiation was decreased by the miRNA mimics and increased by the miRNA inhibitors (Figure 8e). Next, we injected cells stably transfected with a miRNA‐expressing lentivirus (miR‐92a‐1‐5p MDA‐PCa‐2b‐luc cells) into the prostate of BALB/C nude mice (7 weeks old, n = 6 per group) (Figure 8f) and then quantified the tumor burden through in vivo imaging and ex vivo imaging on Day 42. Total tumor burden was higher in Lv‐92a‐1‐5p group via observing the number of tumor sites (Figure 8g). Elevated tumor burden was observed in bone (Figure 8i) in Lv‐92a‐1‐5p group, however, tumor burden in prostate was not significantly increased (Figure 8h). Tumor bone metastasis was confirmed via ex vivo imaging (Figure 8j) and PSA immunofluorescence labelling (red, white arrows indicated) (Figure 8k), based on PSA expression in prostate and MDA PCa cells, and no expression in bone.

To uncover the molecular mechanisms underlying miR‐92a‐1‐5p‐regulated bone homeostasis, we first used the tool DIANA to identify enriched KEGG pathways. Among the enriched pathways, the most dominant was ECM‐receptor interaction (Table 1). Next, we predicted target genes by using online miRNA prediction software (Targetscan, miRanda, miRWalk, and miRDB) (Figure 9a). Among the predicted targets, we identified COL1A1, which encodes a type I collagen subunit that is the major organic component of the bone ECM (Kolb & Bussard, 2019). At 48 h after transfection with miR‐92a‐1‐5p, the mRNA level of COL1A1 was not notably altered in MC3T3‐E1 cells but was significantly decreased in Raw264.7 cells (Figure 9b,c). Western blotting results showed that COL1A1 protein levels were decreased in MC3T3‐E1 and Raw264.7 cells stably transfected with the miR‐92a‐1‐5p lentivirus (Figure 9d). Moreover, according to the results of bioinformatics analysis, COL1A1 harboured two 7‐nt seed sites for miR‐92a‐1‐5p within its 3ʹ‐UTR (Figure 9e). Thus, we used a dual‐luciferase reporter‐gene system to determine whether miR‐92a‐1‐5p directly targets COL1A1; luciferase activity in the miR‐92a‐1‐5p and COL1A1‐Wt co‐transfection group was significantly lower than that in the control group (P < 0.001) (Figure 9f), which implied that COL1A1 is a direct target of miR‐92a‐1‐5p. We further confirmed these results by using immunofluorescence labelling to examine the cellular localization and relative expression level of type I collagen in MC3T3‐E1 and Raw264.7 cells stably expressing the miR‐92a‐1‐5p lentivirus. As expectedly, the expression levels of type I collagen in MC3T3‐E1 and Raw264.7 cells were dramatically decreased (Figure 9g,h).

To further demonstrate the effect of COL1A1 knockdown in the differentiation of osteoblasts and osteoclasts, we designed and transfected Col1a1 siRNA, and confirmed the decrease of COL1A1 expression in both MC3T3‐E1 cells and Raw264.7 cells using qPCR (Figure 10a,b) and western blotting (Figure 10c). As expected, the mRNA expression of Ctsk and Trap in Raw264.7 cells was upregulated following transfection with Col1a1 siRNA for 36 h (Figure 10d,e); the number of Trap⁺ osteoclasts were confirmed increase in both Raw264.7 cells (Figure 10f) and BMMs (Figure 10g) 6 days after transfection. Moreover, Alp and Runx2 mRNA (Figure 10h,i) expression in MC3T3‐E1 cells was downregulated 36 h post transfection; ALP staining (Figure 10j) and ALP activity (Figure 10k) were also reduced after transfection for 7 days.

To test whether exosomes from MDA PCa 2b cells can diminish COL1A1 expression in vivo, we injected nude mice thrice weekly for 4 weeks with 10 μg of MDA PCa 2b exosomes and then harvested the tibia and examined COL1A1 levels (Figure 11a,b). COL1A1 expression was clearly downregulated. We also measured the level of type I collagen in the tibia harvested after 4 weeks of bone marrow education, and we found that the type I collagen labelling intensity was lower in the MDA PCa 2b treatment group than in the control group (Figure 11c,d). We found that both COL1A1 and type I collagen mainly localize in trabecular bone. However, as mentioned in Figure 5 and Figure S5, 4‐weeks exosomes education decreased the thickness of trabecular bone and increased the trabecular space; thus, these results are consistent with the results we described in the previous part of this study.

In conclusion, these results demonstrated that miR‐92a‐1‐5p derived from PCa exosomes disrupts bone homeostasis, promotes osteoclast differentiation, inhibits osteoblast differentiation, and promotes tumor bone metastasis in PCa; moreover, our findings suggest the following underlying mechanism: miR‐92a‐1‐5p and MDA PCa exosomes containing miR‐92a‐1‐5p target COL1A1 and accelerate bone ECM degradation.

3.9. Reasons for discrepancy of PCa cells and PCa exosomes on bone homeostasis

Notably, unlike PCa cells, exosomes from three cell lines used in the present study (osteoblastic MDA PCa 2b cells, osteoclastic PC3 cells, and mixed C4‐2) all promoted osteoclast differentiation. Although, there may be other factors responsible for this observation, we attempted to investigate this discrepancy from the perspective of this study. Firstly, we performed qPCR to examine the expression of miRNA‐92a‐1‐5p, miRNA‐375, miR‐148a‐3p in different PCa cells. We found all three miRNAs were universally expressed, but at different level, in the three PCa cell lines (Figure S7a‐c). Based on previous data in Figure 7, miRNA‐92a‐1‐5p is osteoclastic and miRNA‐375/miR‐148a‐3p are osteoblastic. We calculated the relative osteogenic index as: relative osteogenic index = relative miRNA expression of (miR‐148a‐3p + miR‐375 ‐ miR‐92a‐1‐5p) (Figure S7d). We found that PC3 cells showed lower osteogenic index, which may partly account for PC3 as an osteoclastic cell line. Furthermore, we compared the exosome yield of the three types of PCa cell lines and found no differences (Figure S7e). We also injected BALB/C nude mice (7 weeks old, n = 3 per group) with 100 μg Vybrant DID labelled exosomes via tail vein and compared the fluorescence intensity in bone 24 h post injection using ex vivo imaging. No distribution difference was discovered among the groups (Figure S7f). Subsequently, we incubated 10 μg Vybrant DID labelled PCa exosomes with Raw264.7 cells and MC3T3‐E1 cells in vitro for 90 min and 180 min, respectively (Figure S7g,h). We discovered that PC3 exosomes were more easily internalized by Raw264.7 and C4‐2 exosomes were more easily internalized by MC3T3‐E1, which indicated that different PCa exosomes may target different recipient cells to play the same role. Thus, we concluded that miRNA expression in the PCa cells may be the determinants of the osteogenic characteristics of PCa cells; PCa exosomes target different recipient cells but play the same role in bone homeostasis. Importantly, the similar role, PCa exosomes derived from different types of PCa cells played in bone homeostasis, has created the basis for PCa‐exosome‐targeted therapeutic strategies for bone diseases.