Ticagrelor regulates osteoblast and osteoclast function and promotes bone formation in vivo via an adenosine-dependent mechanism

Abstract

As many as 10% of bone fractures heal poorly, and large bone defects resulting from trauma, tumor, or infection may not heal without surgical intervention. Activation of adenosine A_2A receptors (A_2ARs) stimulates bone formation. Ticagrelor and dipyridamole inhibit platelet function by inhibiting P2Y₁₂ receptors and platelet phosphodiesterase, respectively, but share the capacity to inhibit cellular uptake of adenosine and thereby increase extracellular adenosine levels. Because dipyridamole promotes bone regeneration by an A_2AR-mediated mechanism we determined whether ticagrelor could regulate the cells involved in bone homeostasis and regeneration in a murine model and whether inhibition of P2Y₁₂ or indirect A_2AR activation via adenosine was involved. Ticagrelor, dipyridamole and the active metabolite of clopidogrel (CAM), an alternative P2Y₁₂ antagonist, inhibited osteoclast differentiation and promoted osteoblast differentiation in vitro. A_2AR blockade abrogated the effects of ticagrelor and dipyridamole on osteoclast and osteoblast differentiation whereas A_2BR blockade abrogated the effects of CAM. Ticagrelor and CAM, when applied to a 3-dimentional printed resorbable calcium-triphosphate/hydroxyapatite scaffold implanted in a calvarial bone defect, promoted significantly more bone regeneration than the scaffold alone and as much bone regeneration as BMP-2, a growth factor currently used to promote bone regeneration. These results suggest novel approaches to targeting adenosine receptors in the promotion of bone regeneration.—Mediero, A., Wilder, T., Reddy, V. S. R., Cheng, Q., Tovar, N., Coelho, P. G., Witek, L., Whatling, C., Cronstein, B. N. Ticagrelor regulates osteoblast and osteoclast function and promotes bone formation in vivo via an adenosine-dependent mechanism.

Treatment of bone defects caused by trauma, infection, or tumors often requires surgical repair with some means of promoting local bone regeneration. Indeed, 5–10% of fractures may result in nonunion or delayed closure leading to major treatment challenges in orthopedic reconstructive surgery (1). In addition, spinal fusion and defects caused by osteolysis adjacent to implants also require regeneration of bone for adequate rehabilitation (2–4). Once regenerated at a site, new bone tissue is continuously remodeled throughout the individual’s lifetime to maintain homeostasis through the balancing activities of bone-forming osteoblasts and of bone-resorbing osteoclasts (5).

Although substantial advances have been made in enhancing bone regenerative therapeutic methods, the innate capacity of bone for regeneration and healing significantly reduces as the size of the bone defect increases (6, 7). Among bone regenerative therapy methods, application of growth factors and autologous or cadaveric bone grafts are currently in use (8). Bone morphogenetic proteins (BMPs), members of the transforming growth factor-β superfamily, are involved in committing multipotent stromal cells toward an osteogenic lineage and induction of new bone formation (9, 10). Several studies have demonstrated that BMP-2 promotes large bone defect healing and this agent has been approved by the U.S. Food and Drug Administration since 2002 for bone regeneration (11). Although BMP-2 is commercially available for clinical use, its application is restricted to its supraphysiological dosage (1000 times the normal physiologic concentration) usually required to achieve bone formation with somewhat unpredictable results (12, 13). This remains a problem as clinical studies show that rhBMP-2 and BMP-7 have critical side effects that include vertebral osteolysis, ectopic bone formation, radiculitis, and potential stimulation of neoplasias (14–16). Thus, new approaches to enhancement of bone regeneration are warranted.

Antiplatelet drugs are widely used in the treatment and prevention of coronary artery disease, especially in the setting of acute coronary syndrome (ACS) and percutaneous coronary intervention (17). Acetylsalicylic acid and clopidogrel are the most widely used platelet function inhibitors, and dual antiplatelet treatment with these drugs is the most common treatment for ACS after percutaneous coronary intervention (17). Clopidogrel is a thienopyridine that undergoes metabolic transformation in the liver (18, 19) to generate an active metabolite that binds irreversibly to P2Y₁₂ and thereby inhibits ADP-induced platelet activation. Recently, the ADP receptor P2Y₁₂ has been reported to play a critical role in the differentiation and function of osteoclasts, particularly in preclinical models of pathologic bone loss associated with postmenopausal osteoporosis, rheumatoid arthritis, and bone metastases (20). P2Y₁₂ antagonism with clopidogrel decreased bone loss in murine models of osteoporosis and tumor-associated bone loss. Others have reported that treatment with clopidogrel slows both osteoclast and osteoblast proliferation and reduces cell viability in vitro, thereby diminishing both bone formation and resorptive activity (21). Ticagrelor is a more recently developed antiplatelet agent that, in addition to P2Y₁₂ inhibition, prevents cellular adenosine uptake by the equilibrative nucleoside transporter (ENT)-1 and thereby increases extracellular adenosine levels. Dipyridamole inhibits platelet function via inhibition of phosphodiesterase and, like ticagrelor, also blocks ENT1 and increases extracellular adenosine levels.

Previous work from our laboratory has established that ligation of adenosine A_2A receptors (A_2AR) inhibits osteoclast differentiation (22–24) both in vitro and in vivo, and application of an A_2AR agonist reduces osteoclast-mediated bone resorption in a murine calvaria model of wear-particle–induced bone resorption (23). A_2AR activation reduces particle-induced bone pitting and porosity, diminishes inflammation, and reduces osteoclast number in this model. Moreover, treatment of critical bone defects with a collagen sponge injected with an A_2AR agonist markedly increases bone regeneration. Application of dipyridamole to the collagen sponge also promotes bone regeneration by an adenosine A_2A receptor–mediated mechanism (25). Adenosine A_2B receptor ligation also inhibits osteoclast differentiation (26, 27). We report that both the antiplatelet agents ticagrelor and CAM (the active metabolite of clopidogrel) regulate osteoclast differentiation by different mechanisms and that ticagrelor, like dipyridamole, inhibits osteoclast differentiation by an A_2AR-mediated mechanism.

MATERIALS AND METHODS

Reagents

Ticagrelor and CAM were provided by AstraZeneca (Mölndal, Sweden). ZM241385 and GS6021 were purchased from Tocris (Bristol, United Kingdom). Dipyridamole, sodium acetate, glacial acetic acid, Naphtol AS-MX phosphate disodium salt, Fast Red Violet LB, sodium tartrate, Alizarin Red, dexamethasone, β-glycerophosphate, l-ascorbic acid, and goat anti-rabbit-FITC were from Sigma-Aldrich (St. Louis, MO, USA). α-MEM, FBS, penicillin/streptomycin were from Thermo Fisher Life Sciences (New York, NY, USA). Sodium tartrate was from Thermo Fisher Scientific (Pittsburgh, PA, USA). Rabbit polyclonal anti-cathepsin K was from Santa Cruz Biotechnology (Dallas, Texas, USA). Rabbit polyclonal anti-alkaline phosphatase (ALP) was from Abcam (Cambridge, MA, USA). Murine M-CSF and receptor activator of NFκB ligand (RANKL) were from R&D (Minneapolis, MN, USA).

Animals

C57Bl/6 wild-type (WT) mice (n = 90) and adenosine A_2AR knockout (A_2AKO) mice (n = 30) age 6–8 wk were used. A_2AKO mice were provided by Dr. J. F. Chen (Boston University School of Medicine, Boston, MA, USA). Female A_2AKO mice were bred onto a C57Bl/6 background (≥10 backcrosses) in the New York University School of Medicine (NYUSoM) Animal Facility. A_2AKO animals were derived from 4 original heterozygous breeding pairs for each mouse strain. Mice described as WT were all maintained on the C57Bl/6 background by the breeder (Taconic Laboratories, Hudson, NY, USA). Genotyping was performed by PCR, as reported previously (28). All protocols were approved by the NYUSoM Institutional Animal Care and Use Committee.

Osteoclast differentiation

The marrow cavity was flushed out with α-MEM from aseptically removed femora and tibiae from 6–8 wk C57Bl/6 female mice. Bone marrow cells (BMCs) were incubated overnight in α-MEM containing 10% FBS and 1% penicillin/streptomycin to obtain a single-cell suspension. Nonadherent cells (∼200,000) were collected and seeded in α-MEM with 30 ng/ml macrophase colony-stimulating factor (MCSF) for 2 d. At d 3 (d 0 of differentiation), 30 ng/ml RANKL was added to cultures in the presence/absence of ticagrelor, CAM, or dipyridamole in doses ranging from 1 nM to 100 µM (n = 6). In some experiments, ZM241385 (A_2AR antagonist) 1 µM, GS6021 (A_2BR antagonist) 1 µM, or ADP 1 µM were added to the culture (n = 6). Medium and reagents were replaced every third day. After incubation for 7 d, cells were stained for tartrate-resistant acid phosphatase (TRAP) for osteoclast quantification (22, 23). The number of TRAP-positive mononuclear cells containing ≥3 nuclei/cell was scored (29). Data are expressed as percentage of control osteoclast differentiation because of the variability in number of osteoclasts formed on different days and by cells from different mice. Because all drugs were dissolved in DMSO, this solvent was added to control medium (diluted 1:10,000) to control for the effects of the highest concentration of DMSO present in the medium with drugs.

Osteogenesis assay

Osteogenesis assays were performed as previously described (26). BMCs were isolated by flushing out the bone marrow cavity from 6-8-wk-old C57Bl/6 female mice. BMCs were cultured for 3 d, nonadherent cells were discarded, and the adherent cells were cultured until confluent. Stromal cells were washed and reseeded in culture dishes at 1 × 10⁵cell/cm² density with osteogenic medium (αMEM containing 1 µM dexamethasone, 50 µg/ml ascorbic acid, and 10 mM β-glycerophosphate) in the presence/absence of ticagrelor, CAM, or dipyridamole in doses ranging from 1 nM to 100 µM. All culture conditions were performed in quadruplicate on marrow from 6 different mice on 6 different occasions (n = 6). In some experiments, ZM241385 1 µM (A_2AR antagonist), CG6021 1 µM (A2BR antagonist), or ADP 1 µM were added to the culture (n = 6). Because all agents were dissolved in DMSO, we added DMSO (1:10,000) to medium control cultures. Alizarin Red staining was performed 10 d after culture. The cells were fixed in 4% paraformaldehyde and stained for 45 min with 2% Alizarin Red. Staining intensity (measured as intensity of red color) was quantified with SigmaScan Pro5 software (Systat, Inc., San Jose, CA, USA). Data were normalized to control, untreated cells for each experiment because osteogenesis varied from culture to culture and mouse to mouse. The results are expressed as the percentage of control osteogenesis.

Real-time quantitative RT-PCR

To validate the effect of ticagrelor and CAM in osteoclast and osteoblast differentiation, we measured the activation of the, cathepsin K and nuclear factor of activated T cells NFATc1 (osteoclast differentiation markers), osteopontin (an extracellular structural protein that initiates the development of osteoclast ruffle borders), RANKL, osteoprotegerin (OPG) and Runx2 (osteoblast differentiation marker). Briefly, bone marrow cells (BMCs) from 8-wk C57BL/6 female mice were isolated by flushing out the bone marrow cavity. non-adherent cells were seeded at a density of 5 × 10⁴ cells/cm² in the presence of MCSF and RANKL 30 ng/ml each and adherent BMCs were seeded at a density of 5 × 10⁴ cells/cm² with osteogenic medium (αMEM containing 1µM dexamethasone, 50 µg/ml ascorbic acid, 10 mM β-glycerophosphate) in the presence/absence of ticagrelor and CAM 10 µM alone or in the presence of ZM241385 or CG6021 1 µM each (n = 3) for 7 (osteoclast differentiation) and 10 (osteoblast differentiation) days. Total RNA was extracted with an RNeasy Mini Kit (Qiagen, Carpinteria, CA, USA). RNA was retrotranscribed using MuLV reverse transcriptase PCR kit (Thermo Fisher Scientific, Carlsbad, CA, USA). Real-time RT-PCR was performed with Brilliant Fast SYBR Green Kit QPCR Master Mix (Agilent Technologies, Santa Clara, CA, USA). The following primers were used: cathepsin K forward: 5′-GCTGAACTCAGGACCTCTGG-3′, reverse: 5′-GAAAAGGGAGGCATGAATGA-3′; NFATc1 forward: 5′-TCATCCTGTCCAACACCAAA-3′, NFATc1 reverse: 5′-TCACCCTGGTGTTCTTCCTC -3′; osteopontin forward: 5′-TCTGATGAGACCGTCACTGC-3′, reverse: 5′-TCTCCTGGCTCTCTTTGGAA-3′; RANKL forward 5′-AGCCGAGACTACGGCAAGTA-3′ and reverse 5′-GCGCTCGAAAGTACAGGAAC-3′; OPG forward 5′-CTGCCTGGGAAGAAGATCAG-3′ and reverse 5′-TTGTGAAGCTGTGCAGGAAC-3′; Runx2 forward: 5′-CCCAGCCACCTTTACCTACA-3′ and reverser: 5′-TATGGAGTGCTGCTGGT-3′; and GAPDH forward: 5′-CTACACTGAGGACCAGGTTGTCT-3′ and reverse: 5′-GGTCTGGGATGGAAATTGTG-3′. The Pfaffl method (30) was used for relative quantification.

Calcination and attrition milling of β-TCP ink

The ink for the scaffolds was fabricated through a series of powder-processing steps, beginning with calcination (800°C for 11 h), attrition milling (3 mm zirconia milling medium; Union Process, Akron, OH, USA) in DI-H2O for ∼30 min, and drying (∼65–75°C) of the β-tricalcium phosphate (β-TCP) (Sigma-Aldrich). The dried ceramic was then transferred into a polyethylene bottle and dry milled for ∼10 min using a paint shaker with a charge of ∼20 pieces of zirconia milling medium of 10 mm diameter (31).

Ink formulation

Previously calcined and milled ceramic powders were used for the colloidal gel formulation. Concentrated β-TCP suspensions, where the volume fraction (ϕ_ceramics) of ceramic was ∼0.46, were produced by mixing a precalculated amount of ceramic powder and ammonium polyacrylate (Darvan 821A; RT Vanderbilt, Norwalk, CT, USA) solution to disperse particles into deionized (DI)-H₂O. The dispersant proportion per gram of ceramics was ∼15 mg. First, ∼25 g of milling medium was added to the DI-H₂O, then the dispersant, and then the ceramic powder in 3 parts (∼33% per step). After each addition of powder, the suspension was mixed in a planetary mixer (AR-250; Thinky, Tokyo, Japan) for 1 min. Next, hydroxypropyl methylcellulose, also referred to as F4M, (Methocel F4M; Dow Chemical Company, Midland, MI, USA) was added as the thickening agent. The F4M is used in a 5% wt/wt aqueous solution with a proportion of 7 mg/ml ceramic. The suspension was then mixed for 1 min followed by a defoaming step for 1 min in the planetary mixer. Finally, the suspension was gelled by adding ∼150 mg/30 ml of ink of polyethylenimine (Sigma-Aldrich, St. Louis, MO, USA) in a 10% wt/wt solution. Mixing and defoaming (1 min and 30 s, respectively) after the final addition completed the ink preparation procedure (31, 32).

Scaffold design and assembly

The β-TCP scaffolds were fabricated via robocasting, with a 3-dimensional (3-D) direct-write microprinter gantry robot system used to extrude the colloidal ink (Aerotech Inc., Pittsburgh, PA, USA). The 3-D circular plug-shaped scaffolds with cap (cap layer diameter, 4.4-mm; plug diameter, 3.3-mm; 250-μm struts, and 300-μm pore spacing) were designed with a computer-aided design system (RoboCAD 4.1; 3D Inks LLC, Tulsa, OK, USA) (Supplemental Fig. 1). The colloidal ink was loaded into a 3 ml syringe (Nordson Corp., Westlake, OH, USA) and subsequently equipped with 250-μm-diameter extrusion nozzle (Nordson Corp.). The 3-D circular plug-shaped scaffolds were printed in layer-by-layer fashion at 8 mm/s print speed. The entire deposition process occurred in a low-viscosity paraffin oil tray to prevent drying of the structure during fabrication. After the scaffold was complete, it was removed from the oil reservoir and allowed to partially dry, before being sintered in a multistep process which included a ramp to 400°C for 2 h, followed by 900°C for 2 h, and ultimately 1100°C for 4 h; the dwell time at the 2 lower temperatures allowed for the burnout of the organics, whereas the higher temperature allowed for the densification.

In vivo murine model surgical procedure

WT (n = 90) and A_2AKO (n = 30) mice were anesthetized by intraperitoneal injection of 100 mg/kg ketamine and 10 mg/kg xylazine. The hair over the skull was shaved, and the underlying skin was aseptically prepared. A full-thickness midline incision, extending from the nasofrontal to occipital region, was made under sterile conditions, and the underlying periosteum was sharply incised on the midline and subsequently elevated off the skull to obtain sufficient exposure for the trephine. A 3 mm defect to remove bone from the middle of the dorsal calvarium was applied with a trephine drill and irrigated with saline to avoid overheating the bone, with caution to prevent damage to the underlying sagittal sinus and dura mater. In one set of experiments, the defect in WT mice was covered with a collagen sponge (DuraGen Plus; Integra LifeScienes Corp., Plainsboro, NJ, USA) soaked in 20 µl of 0.9% saline (control; n = 10), ticagrelor 1 or 10 µM (n = 10 each) and CAM 1 or 10 µM (n = 10 each). As control for bone formation, some animals were treated with BMP-2 200 ng (n = 10). Treatment was applied by a daily injection of ticagrelor or CAM vs. a single dose for BMP-2, beginning immediately after incision closure and continuing every day until death. In another set of experiments, the defect in both WT and A_2AKO mice was filled with the bioresorbable 3-D printed collagen-coated HA/β-TCP scaffold (31,33,34). In this case, the 3 mm defects were created using a biopsy punch (Acu Punch; Acuderm, Inc., Fort Lauderdale, FL, USA). The bioresorbable digitally printed collagen-coated scaffolds were soaked in saline (control), ticagrelor 1 mM, or CAM 1 mM (n = 5 each treatment). No more treatments were applied until death. Water and food were given ad libitum until death. Animals were euthanized after 4 wk of defect formation in a CO₂ chamber and the calvaria were removed, fixed, and prepared for micro–computed tomography (microCT) and histologic staining.

MicroCT

After death, WT and A_2AKO (n = 5 calvaria per treatment group) were fixed in 70% ethanol and prepared for high-resolution microCT. This 3-D imaging technology was used to perform qualitative and quantitative analyses of new bone formation areas in murine calvarial bone. Analyses were performed in Skyscan 1172 microCT (Bruker, Fitchburg, WI, USA) using the following imaging parameters: 60 kV, 167 μA, 9.7 μm pixel size, 2000 × 1332 matrix, 0.3° rotation steps, 6 averages, movement correction of 10, 0.5 mm Al filter, 2 segments scanned per sample (56 min/segment). Images were reconstructed with NRecon software (Skyscan; Micro Photonics, Allentown, PA, USA) [histogram range 0–0.065, beam hardening correction of 35, Gaussian smoothing (factor 1), ring artifact correction of 7]. For qualitative analysis, 3-D images of the mice heads were then reconstructed from the cross-sectional slices using CTNa software provided by Skyscan, and processing was done to get direct morphometric measurements in 3-D. For quantitative analysis of new bone formation, the area of interest was segmented manually by marking the volume of interest (VOI) a round region across the defective bone∼3 mm in diameter. Data were calculated as a percentage to avoid intrinsic differences among animals. The percentage of bone regenerated was calculated by subtracting the remaining defect area from the total defect area. For qualitative and quantitative 3-D analysis, all data were exported in DICOM format and imported into Amira software (Visage Imaging GmbH, Berlin, Germany). The software extracted volumetric and densitometric numbers from the gray value distribution of the segmented image: the VOI as a circular region across the defective bone, ∼3 mm in diameter. Data were calculated in bone volume as a function of mouse type and biologic group.

Histologic studies

WT and A_2AKO calvaria (n = 5 per treatment group) were removed and fixed in 4% paraformaldehyde for 48 h, followed by decalcification in 10% EDTA for 4 weeks and paraffin embedding. Sections (5 µm) were cut and stained with hematoxylin and eosin. Photomicrographs were taken at an original magnification of ×800.

WT and A_2AKO calvaria (n = 5 per treatment group) were removed and fixed in 70% ethanol and coarsely dehydrated in a series of graded alcohols. The specimens were then embedded in polymethyl methacrylate according to standard hard tissue histology protocols. In brief, after dehydration, specimens were immersed in methyl salicylate for 30 h, with a change in solution after 6 h. Tissue was infiltrated with methyl methacrylate (MMA), with the solution changed every 24 h 3 times, to ensure complete infiltration (first infiltration at room temperature and the other 2 at 4°C). After this, the solution was changed again and the samples were kept under ultraviolet light until polymerization. MMA blocks were cut in multiple sagittal sections and placed onto a slide with thin film of industrial adhesive (408; Loctite, Rocky Hill, CT, USA). Sections were then ground down manually to ∼80 μm thickness with a series of graded emery papers up to 1200 grit (Buehler, Lake Bluff, IL, USA) and a 1 μm alumina powder (MicroPolish II; Buehler, Lake Bluff, IL, USA). Sections were then stained with Van Gieson’s picrofuchsin and Stevenel Blue to determine tissue response. In this procedure, calcified bone stains bright red with an intensity that varies according to bone maturity. Cells, fibers, and unmineralized matrix stain blue, and scaffold is unstained, appearing dark on histologic sections.

TRAP staining was performed in paraffin sections with a homemade TRAP buffer (0.1 M acetate buffer, 0.3 M sodium tartrate, 10 mg/ml Naphtol AS-MX phosphate, 0.1% Triton X-100, and 0.3 mg/ml Fast Red Violet LB; Sigma-Aldrich). After they were deparaffinized and washed in acetate buffer, the samples were incubated in TRAP buffer for 30 min and counterstained with Fast Green.

Immunohistochemistry analyses for cathepsin K and ALP were performed (23). In brief, deparaffinized and hydrated sections were incubated with proteinase K solution (20 μg/ml in TE buffer; pH 8.0) for 15 min in a water bath at 37°C for antigen retrieval. After blocking of nonspecific binding with PBS 3% BSA and 0.1% Triton X-100 for 1 h, primary antibodies (rabbit polyclonal anti-ALP 1:100, rabbit polyclonal anti-cathepsin K 1:25) were incubated overnight at 4°C in a humidifying chamber. Secondary antibody goat anti-rabbit-FITC (1:200) was incubated for 1 h in the dark. Slides were mounted with DAPI antifade mounting medium (Fluoroshield; Sigma-Aldrich).

Statistical analysis

Statistical significance for differences between groups was determined by use of 1-way ANOVA and Bonferroni post hoc test. All statistics were calculated using Prism software (GraphPad, La Jolla, CA, USA).

RESULTS

Ticagrelor inhibits osteoclast differentiation and promotes osteoblast formation by an adenosine A_2AR mechanism

Because prior studies have suggested that dipyridamole diminishes bone resorption by an adenosine receptor-mediated mechanism (35) and adenosine A_2A and A_2B receptors inhibit osteoclast differentiation (26), we asked whether ticagrelor would regulate osteoclast differentiation from primary murine bone marrow–derived precursors. Ticagrelor inhibited osteoclastogenesis in a concentration-dependent manner (IC₅₀ = 4 µM, decrease of 61 ± 2% inhibition with 10 µM; P < 0.001; n = 6). Both dipyridamole (IC₅₀ = 0.5 µM, decrease of 62 ± 3% inhibition with 1 µM; P < 0.001; n = 6) and the P2Y₁₂ antagonist CAM (IC₅₀ = 7 µM, decrease of 58 ± 4% inhibition with 10 µM; P < 0.001; n = 6) also inhibited osteoclastogenesis (Fig. 1). Dipyridamole was 10-fold more potent than ticagrelor or CAM at inhibiting osteoclastogenesis, consistent with the known potency of dipyridamole for inhibition of ENT1 (36).

Figure 1.

In vitro characterization of the effect of ticagrelor, dipyridamole, and CAM on bone marrow–derived osteoclasts. Primary bone marrow–derived osteoclast precursors (from WT animals) were stimulated to undergo osteoclast differentiation, fixed, and stained for TRAP after being cultured for 7 d in the presence of ticagrelor, CAM, and dipyridamole (1 nM–100 µM). TRAP⁺ cells containing 3 or more nuclei were counted as osteoclasts. Data are expressed as a percentage of control osteoclast differentiation and results represent means ± sem of 6 independent cultures. **P < 0.005, ***P < 0.001 compared to control (ANOVA).

At low concentrations, both ticagrelor, at concentrations lower than those previously shown to inhibit ent1-mediated adenosine uptake (36, 37), and CAM inhibited osteoblast differentiation (Fig. 2), although at higher concentrations, neither agent altered osteoblast differentiation. In previous studies P2Y12 antagonism has been shown to diminish osteoblast differentiation in vivo and in vitro (21) and this effect most likely explains the effects of low doses of these antagonists on osteoblast differentiation. We further speculate that higher concentrations of CAM and ticagrelor reverse the effect of P₂Y₁₂ inhibition of osteoblast differentiation by directly or indirectly (by increasing extracellular adenosine concentrations) stimulating A2BR, respectively.

Figure 2.

In vitro characterization of the effect of ticagrelor, dipyridamole, and CAM on bone marrow–derived osteoblasts. Mesenchymal stem cells were isolated from primary bone marrow of WT animals and induced to undergo differentiation into osteoblasts. After 10 d of culture in the presence of medium, ticagrelor, CAM, and dipyridamole cells were fixed and matrix stained with Alizarin Red. The intensity of the staining was determined for each well, and results are expressed as the mean ± sem percentage of control osteoblast differentiation of 6 independent cultures. *P < 0.05 compared to control (ANOVA).

We have previously reported that dipyridamole regulates bone metabolism by increasing extracellular adenosine levels which activate adenosine A_2AR (35). To determine whether both ticagrelor and CAM inhibited osteoclastogenesis via A_2AR stimulation, we determined the effect of an A_2AR antagonist (ZM241385, 1 μM) and A_2BR antagonist (GS6021, 1 μM) on the capacity of these agents to regulate osteoclastogenesis. Blockade of A_2AR with ZM241385 abrogated the capacity of both dipyridamole and ticagrelor to inhibit osteoclastogenesis (5 ± 1 and 1 ± 2% inhibition vs. 61 ± 2 and 62 ± 3% inhibition, in the presence or absence of ZM241385, respectively; P < 0.001; n = 6; Fig. 3A). In contrast, the effect of CAM was reversed by the A_2BR antagonist GS6021 1 µM (11 ± 2% inhibition vs. 58 ± 4 and 62 ± 3% inhibition; P < 0.001; n = 6). The A_2AR agonist ZM241385 1 µM inhibited osteoblast differentiation and function, even in the presence of either ticagrelor or dipyridamole (58 ± 3 and 49 ± 4% inhibition vs. 100 ± 2 and 95 ± 5% osteoblast differentiation, respectively; P < 0.001; n = 6). As with osteoclast differentiation GS6021 reversed the effect of CAM on osteoblast differentiation (24 ± 2% inhibition; P < 0.05; n = 6; Fig. 3B).

Figure 3.

Ticagrelor inhibits osteoclast differentiation while promoting osteoblast formation and differentiation by an adenosine A_2AR mechanism. A) Primary bone marrow–derived osteoclast precursors (from WT animals) were stimulated to undergo osteoclast differentiation, fixed, and stained for TRAP after being cultured for 7 d in the presence of ticagrelor (10 µM), CAM (10 µM), and dipyridamole (1µM), alone or in combination with ZM241385, GS6021 (1 µM each). TRAP⁺ cells containing 3 or more nuclei were counted as osteoclasts. Data are expressed as the percentage of control osteoclast differentiation. B) Mesenchymal stem cells were isolated from primary bone marrow from WT animals and induced to undergo differentiation into osteoblasts. After 10 d of culture in the presence of ticagrelor (10 µM), CAM (10 µM), and dipyridamole (1µM) alone or in combination with ZM241385 and GS6021 (1 µM each), cells were fixed and matrix stained with Alizarin Red. The intensity of staining was determined for each well, and results are expressed as the percentage of control osteoblast differentiation. C) Changes in cathepsin K, NFATc1 and osteopontin mRNA in MCSF/RANKL precursors during the 7 d of osteoclast differentiation in the presence of ticagrelor and CAM alone or with ZM241385/GS6021 compared to control. D) Changes in RANKL, OPG and Runx2 mRNA in mesenchymal precursors and osteoblasts during 10 d of osteoblast differentiation in the presence of ticagrelor and CAM alone or with ZM241385/GS6021. Results are expressed as means ± sem of 6 independent cultures. *P < 0.05, **P > 0.005, ***P < 0.001 compared to control (ANOVA).

To confirm the effect of the agents on osteoclast differentiation, we analyzed cathepsin K, NFATc1, and osteopontin gene expression. Ticagrelor and CAM inhibited cathepsin K gene expression when compared with the control (maximum increase, 2.5 ± 0.8 and 2.6 ± 0.03, respectively, vs. 5.14 ± 1.44 control; P < 0.01; n = 3; Fig. 3C), and treatment with ZM241385 reversed the effect of ticagrelor (maximum increase, 4.6 ± 1 vs. 2.5 ± 0.8; P < 0.05; n = 3). Similarly treatment with GS6021 reversed the effect of CAM (maximum increase, 4.9 ± 1.3 vs. 2.6 ± 0.03; P < 0.05; n = 3; Fig. 3C). Similar changes were observed in NFATc1 (maximum increase, 27.8 ± 7.8 for ticagrelor and 35.2 ± 10.3 for CAM vs. 51 ± 3.3; P < 0.01 and P < 0.05, respectively; n = 3) whereas osteopontin expression was decreased only in the presence of ticagrelor (maximum increase, 150 ± 128 for ticagrelor and 249 ± 26 for CAM vs. 197.4 ± 2.7; P < 0.05 and P = ns, respectively; n = 3).

When we analyzed the expression of RANKL and OPG by osteoblasts we observed a decrease in RANKL expression 7 d after incubation with ticagrelor and CAM (maximum increase: 4.4 ± 2.04 for ticagrelor and 1 ± 0.13 for CAM vs. 16.4 ± 2.7 for the control; P < 0.05; n = 3) with an increase in OPG (maximum increase of 3.9 ± 0.9 for ticagrelor and 2.9 ± 0.5 for CAM vs. 1.63 ± 0.2; P < 0.01; n = 3); A_2AR and A_2BR antagonists reversed the effects of ticagrelor and CAM, respectively (Fig. 3D). Finally, Runx2 expression was slightly increased by both ticagrelor and CAM (maximum increase of 3.1 ± 0.2 for ticagrelor and 2.5 ± 0.2 for CAM vs. 1.58 ± 0.8; P < 0.05 and P = ns; n = 3). These results further indicate that the effects of CAM and ticagrelor on osteoblast differentiation are complex because of opposing effects of P2Y₁₂ antagonism and A_2BR stimulation on osteoblast differentiation and that the complexity results from actions of these 2 receptors on differing stages of osteoblast differentiation.

Because these results suggest that CAM inhibits osteoclast differentiation in an A_2BR-dependent fashion, we confirmed these results in A_2BKO mice (Supplemental Fig. 2). CAM did not inhibit osteoclast differentiation or affect osteoblast differentiation by A_2BR-deficient cells. To confirm the differential adenosine receptor involved in the effect of these P2Y₁₂ antagonist, ticagrelor was also tested. As observed, when A_2BKO cells were exposed to ticagrelor, osteoclast differentiation was inhibited (Supplemental Fig. 2A) and osteoblast formation was promoted (Supplemental Fig. 2B).

Ticagrelor promotes bone formation in vivo

To determine whether the effects observed in vitro were relevant to the effects of the agents tested in vivo we determined whether ticagrelor or clopidogrel affected bone regeneration in the calvarium when applied locally to a collagen sponge placed in a critical bone defect. Both ticagrelor and CAM, 1 and 10 µM, markedly enhanced the area of bone regenerated after 4 wk (21 ± 3% ticagrelor 1 µM, 28 ± 1% ticagrelor 10 µM, 21 ± 3% CAM 1 µM, and 28 ± 1% CAM 10 µM vs. 15 ± 1% in saline-treated sponges; P < 0.005; n = 5 mice per condition; Fig. 4A and Table 1). The increment in bone regeneration induced by ticagrelor and CAM was similar to that induced by BMP-2 treatment (28 ± 1% bone regeneration, P < 0.005, n = 5). The bone volume, total volume, ratio of bone volume to total volume, and bone mineral density (BMD) were also increased in treated animals when compared to control mice, and all values were similar to those observed in BMP-2-treated mice (Table 1).

Figure 4.

Ticagrelor promotes bone formation in vivo. A) The figures show representative microCT images of calvaria of mice following trephination and treatment with ticagrelor 10 and 1 µM and CAM 10 and 1 µM incorporated into a collagen gel (n = 5 mice per group). B) Calvaria were stained with hematoxylin & eosin to determine the new bone formation. Shown are representative images for TRAP staining (arrow) for osteoclasts in mouse calvaria and the means ± sem (n = 5 mice per group) number of osteoclasts/hpf. Original magnification: ×40 (left), ×400 (right). Data were expressed as mean ± SEM (n = 5 per group). ***P < 0.001 compared to control (ANOVA).

TABLE 1.

MicroCT analysis of calvaria after treatment

Parameter	Control	Ticagrelor, 10 µM	Ticagrelor, 1 µM	CAM, 10 µM	CAM, 1 µM	BMP-2
Bone formation (%)	15.1 ± 1.29	27.93 ± 1.07**	21.37 ± 2.8	27.79 ± 1.33**	21.31 ± 3	28.13 ± 0.65**
TV (mm3)	0.51 ± 0.04	0.72 ± 0.03*	0.52 ± 0.02	0.73 ± 0.08*	0.61 ± 0.02	0.71 ± 0.04*
BV (mm3)	0.42 ± 0.04	0.57 ± 0.02*	0.42 ± 0.02	0.58 ± 0.05*	0.049 ± 0.02	0.53 ± 0.03*
BV/TV	82.61 ± 2.22	79.71 ± 1.8	80.28 ± 1.09	79.66 ± 2.27	84.96 ± 3.05	80.77 ± 1.62
BMD	0.91 ± 0.004	0.93 ± 0.012*	0.91 ± 0.016	0.94 ± 0.01*	0.90 ± 0.009	0.93 ± 0.01*

Digital morphometric analysis of microCT images from the calvaria treated with ultrahigh-molecular-weight polyethylene wear particles combined with ticagrelor, 10 and 1 µM, and CAM, 10 and 1 µM (n = 5 mice per group). Data are expressed as means ± sem. TV, total volume; BV, bone volume. *P < 0.05, **P < 0.01 compared with control (ANOVA).

Histochemical staining of bone defects provided insight into the mechanism by which ticagrelor and CAM increased bone regeneration and confirmed the results obtained by microCT. TRAP staining revealed fewer osteoclasts in ticagrelor- and CAM-treated mice than in control mice [13 ± 1 osteoclasts/high-power field (hpf) for ticagrelor 1 µM, 11 ± 1 osteoclasts/hpf for ticagrelor 10 µM, 13 ± 1 osteoclasts/hpf for CAM 1 µM, and 10 ± 1 osteoclasts/hpf for CAM 10 µM vs. 20 ± 2 osteoclasts/hpf in control; P < 0.001; n = 5 mice per condition], and similar to BMP-2-treated mice (11 ± 2 osteoclast/hpf; P < 0.001; n = 5) (Fig. 4B). These findings were associated with a decrease in immunostaining for cathepsin K (25 ± 1 positive cells/hpf for ticagrelor 1 µM, 22 ± 1 positive cells/hpf for ticagrelor 10 µM, 30 ± 3 positive cells/hpf for CAM 1 µM and 25 ± 1 positive cells/hpf for CAM 10 µM respectively, vs. 46 ± 2 positive cells/hpf in control; P < 0.001; n = 5 mice per condition) (Fig. 5A). Similar reductions in osteoclasts were observed in BMP-2 treated mice (Figs. 4B and 5A). Moreover, we observed an increase in immunostaining for the osteoblast and bone formation marker, ALP, in the bony defects treated with ticagrelor or CAM (38 ± 2 positive cells/hpf for ticagrelor 1 µM, 39 ± 1 positive cells/hpf for ticagrelor 10 µM, 35 ± 1 positive cells/hpf for CAM 1 µM, and 39 ± 1 positive cells/hpf for CAM, 10 µM vs. 26 ± 1 positive cells/hpf in control; P < 0.001; n = 5 mice per condition; Fig. 5B).

Figure 5.

Ticagrelor inhibits osteoclast marker expression and increases osteoblast marker expression after trephination. Calvaria were processed and immunohistologic staining performed. A) Representative sections of calvaria (from n = 5 mice per group) stained for cathepsin K (green). Nuclei are shown in blue (DAPI). B) Representative sections of calvaria (from n = 5 mice per group) stained for alkaline phosphatase (green). Nuclei are shown in blue (DAPI). Quantification of the number of positive cells/hpf was performed in a blinded fashion. Results represent means ± SEM (n = 5 mice per group). Original magnification, ×400. ***P < 0.001 compared with control (ANOVA).

Daily administration of agents into a collagen sponge is not the most efficient approach to promoting bone regeneration. Therefore, we determined whether combining the osteoconductivity of bioresorbable tricalcium phosphate bioceramic scaffolds with the osteoinductivity of the test agents would lead to more effective closure of critical bone defects with ticagrelor. The collagen-coated bioresorbable scaffold was soaked in a 1 mM ticagrelor or CAM solution and implanted in the critical defect. Three-dimensional bone formation was significantly greater in ticagrelor- and CAM-coated scaffolds than control scaffolds by 4 wk after implantation (51 ± 2 µm³ bone volume for ticagrelor and 45 ± 8 µm³ bone volume for CAM vs. 30 ± 3 µm³ bone volume for control; P < 0.05 and P = ns, respectively; n = 5 mice per condition; Fig. 6A). To determine whether ticagrelor exerted its effect via indirect activation of adenosine A_2AR, scaffolds coated with ticagrelor or CAM were implanted in defects made in A_2AKO mice. Increased bone formation mediated by ticagrelor was abrogated in A_2AKO mice, whereas CAM maintained its effect on bone regeneration (26 ± 1 µm³ bone volume for ticagrelor and 49 ± 2 µm³ bone volume for CAM vs. 44 ± 4 µm³ bone volume for control; P < 0.005 and P = not significant, respectively; n = 5 mice per condition). The difference in bone volume between WT controls and A_2AKO controls was not significant.

Figure 6.

3-D analysis of bone formation in a collagen-coated bioresorbable scaffold. A) MicroCT images were analyzed and the bone and scaffold areas were extracted from the DICOM files. Blue: new bone formation; gray: the scaffold. The volume of new bone formation in the scaffolds at the site of trephination are presented as the means ± SEM (n = 5 mice per group). B) Representative histologic sections of scaffolds within trephine defects after 4 wk. Red: new bone; black: scaffold. Sections were embedded in methyl methacrylate and, after cutting and polishing, were stained with Stevenel’s Blue and von Giessen stain. Original magnification, ×20. **P < 0.005, *P < 0.05 compared with control (ANOVA).

Histologic examination of implanted scaffolds demonstrated overall biocompatibility and osteoconductivity for all experimental groups. Bone ingrowth primarily occurred from the defect borders toward the center through the scaffold struts. Qualitative evaluation depicted an increase in bone ingrowth toward the defect center for both ticagrelor and CAM-coated scaffolds vs. control scaffolds in WT mice (Fig. 6B). Moreover, histologic examination in A_2AKO mice was in direct agreement with microCT analysis, where reduced bone ingrowth was observed in scaffolds treated with ticagrelor.

DISCUSSION

Unlike other tissues, bone can regenerate and repair itself, but in pathologic fractures or large and massive bone defects, bone healing and repair fail because of insufficient blood supply, infection of the bone or the surrounding tissues, among other causes, resulting in delayed union or nonunion (38). Many surgical and pharmaceutical options are available, but as many processes are involved during bone healing, treatment of such conditions represents a significant challenge. Combined or not with various bone grafts or scaffolds, growth factors such as BMPs, mesenchymal stem cells, and a variety of other agents have been used to promote repair of bone defects. The results presented here suggest that agents that increase adenosine levels or stimulate adenosine receptors can directly promote bone regeneration.

In the present study, the antiplatelet agents ticagrelor and clopidogrel inhibited osteoclast differentiation in vitro and promote bone regeneration in a murine trephination model using either a collagen sponge or 3-D printed ceramic scaffold as scaffolds. We found that coating these tricalcium-phosphate–based scaffolds with ticagrelor provides sufficient release of the agent to result in enhanced bone growth over the scaffold to the same levels encountered for BMP-2 with the clear advantage of avoiding daily administration of the compound to a collagen sponge.

We had demonstrated that inhibition of osteoclast formation via direct A_2AR stimulation or, indirectly, by increasing local adenosine concentration via dipyridamole-mediated blockade of ENT1-mediated adenosine uptake, stimulates new bone formation as well as BMP-2, a growth factor currently marketed for promotion of bone growth (35). As with dipyridamole, ticagrelor exerts some of its beneficial effects via blockade of adenosine uptake via ENT1 and the resulting increases in extracellular adenosine levels that stimulate adenosine receptors including A_2ARs (36, 37, 39–42). It is interesting to note that dipyridamole was a more potent inhibitor of osteoclastogenesis via blockade of ENT1 than ticagrelor (IC₅₀ of 0.5 vs. 4 µM), consistent with prior reports (36, 43). In suspensions of washed human erythrocytes, ticagrelor dose dependently inhibited adenosine uptake by erythrocytes with a pIC₅₀ of 7.0 ± 0.15 μM, which was 10-fold less potent than dipyridamole (43). This is the first time that a role for ticagrelor in regulation of bone metabolism has been reported.

The results reported herein, along with those in previously published work, unequivocally demonstrate that direct and indirect activation of the A_2AR diminishes osteoclast differentiation and function both in vitro and in vivo in a manner consistent with A_2AR-mediated inhibition of NFκB signaling (24, 25, 27, 44). In contrast, A_2AR stimulation does not alter osteoblast differentiation directly although it clearly regulates osteoblast expression of proteins involved in bone formation (24, 25). In contrast, the active metabolite of clopidogrel, a P2Y₁₂ antagonist with no ENT1 activity, inhibits osteoclast differentiation in an A_2BR-dependent fashion. Syberg et al. (21) determined whether the in vitro effects of clopidogrel were caused by direct antagonism of the P2Y₁₂ receptor and found that receptor activation by ADP diminishes cAMP levels, whereas clopidogrel stimulates increased cAMP levels, an observation consistent with both P2Y₁₂ (45) and A_2BR activation. Prior work further indicated that treatment with clopidogrel itself decreases osteoblast growth and reduces cell viability while decreasing bone formation, ALP activity, and collagen production and increasing the number of adipocytes and diminishing osteoclast formation, viability, and resorptive activity (21). The differences in osteoblast differentiation results between this work and ours may arise from the fact that the present study used CAM, the active form of clopidogrel, in contrast to use of clopidogrel, the prodrug. Clopidogrel is rapidly metabolized by cytochrome P450 in the liver to form an active metabolite (18, 19) in vivo, but the degree to which clopidogrel can be metabolized by cultured cells in vitro remains unclear (46). The same work indicated that adult mice treated with clopidogrel for 4 wk had decreased BMD and trabecular bone (21), whereas both the in vitro and in vivo results reported here show that CAM increased bone regeneration. These contrasting results may have been related to the different animal models studied, although our results are consistent with other work performed by Su et al. (20) in which clopidogrel decreased bone loss in murine models of osteoporosis and tumor-associated bone loss. Alternatively, the effects of local application of high concentrations of the active metabolite may differ from the effect of systemic administration as a result of the indirect effect of clopidogrel on other bone catabolic factors.

Bone growth promoted by dipyridamole on bioactive scaffolds is almost completely confined to the scaffold, but BMP-2 promotes as much, if not more, bone regeneration outside of the scaffold as on it (47). We find that bone growth promoted by ticagrelor and CAM is similar. Bone regeneration, which begins at the cut edge of the defect, appears to proceed from the border of the defect toward the center of the defect through the scaffold struts.

Osteoclasts release ATP which may also regulate osteoclast and osteoblast differentiation via interaction with P2 receptors (48). Osteoclasts express multiple P2 purinoceptor subtypes (49). Of note, the P2X7 receptor regulates both osteoblast and osteoclast function. Of note, P2X7 receptor activation in osteoblasts enhances differentiation and bone formation whereas its activation in osteoclasts results in apoptosis (50–52). Polymorphisms in the P2X7 receptor have been suggested to be involved in osteoporosis, aseptic hip loosening, and rheumatoid arthritis (48). P2Y2 activation also has a role in both bone formation and bone resorption in vitro and in vivo, and calcium waves among osteoblasts were partly propagated by the paracrine action of ATP through binding to P2Y2 receptors on neighboring cells, inducing elevation of intracellular calcium. Moreover, activation of the P2Y1 receptor subtype potentiates PTH-induced c-fos gene expression (53).

It was interesting to note that, despite its A_2BR-mediated effects on osteoclasts and osteoblasts in vitro, CAM did not increase bone ingrowth into scaffolds implanted in A_2AKO mouse calvaria. One potential explanation for the failure of CAM to promote bone regeneration in these mice is that A_2BR are only poorly expressed on the cell surface and are therefore minimally functional in the absence of A_2AR (54). Thus, because there was minimal expression and function of A_2BR on bone marrow and bone in A_2AKO mice, CAM did not promote bone regeneration.

In summary, ticagrelor inhibits osteoclast differentiation via blockade of adenosine uptake and the resulting increases in extracellular adenosine levels that stimulate A_2ARs, as we have reported for dipyridamole (25). In contrast, CAM inhibits osteoclast differentiation via stimulation of A_2BRs. Local treatment with both ticagrelor and CAM promoted bone regeneration in a murine trephination model and may be useful for promoting bone regeneration and inhibition of bone destruction.

Glossary

3-D

3 dimentional

β-TCP

β-tricalcium phosphate

ACS

acute coronary syndrome

ALP

alkaline phosphatase

BENT1

equilibrative nucleoside transporter 1

A₂AR

adenosine receptor A2A

A₂BR

adenosine receptor A2B

BMC

bone marrow cell

BMD

bone mineral density

BMP

bone morphogenetic protein

CAM

active metabolite of clopidogrel

DI-H₂O

deionized-H₂O

ENT

equilibrative nucleoside transporter

hpf

high-power field

F4M

hydroxypropyl methylcellulose

MCSF

macrophage colony-stimulating factor

MMA

methyl methacrylate

microCT

micro–computed tomography

OPG

osteoprotegerin

RANKL

receptor activator of NFκB ligand

TRAP

tartrate-resistant acid phosphatase

VOI

volume of interest

WT

wild-type

AUTHOR CONTRIBUTIONS

B. N. Cronstein and C. Whatling designed the experiments and wrote and revised the manuscript; A. Mediero designed the experiments and has been the primary person responsible for carrying out all experimental procedures and writing the manuscript; T. Wilder helped on surgery and animal treatments and revised the manuscript; and V. S. R. Reddy, Q. Cheng, N. Tovar, P. G. Coelho, and L. Witek designed the scaffolds, performed undecalcified bone preparation, sectioning, staining, scanning, and analysis and revised the manuscript.