A soluble activin Type IIA receptor induces bone formation and improves skeletal integrity

The ActRIIA–mFc fusion protein was purified from conditioned medium by sequential column chromatography as described below. SDS/PAGE analysis showed a single protein band with a molecular mass of ≈106 kDa, under nonreducing conditions and of ≈57 kDa under reducing conditions (data not shown).

The binding affinity of the purified ActRIIA–mFc to activin A was determined by using a Biacore 3000 biosensor. To collect detailed kinetic information, a concentration series of activin A was injected over immobilized ActRIIA–mFc and a reference surface. A simple 1:1 interaction model provided an excellent fit to the data, as shown by the overlay of the simulated binding responses [red lines, supporting information (SI) Fig. S1A]. The K_D of activin A binding to the ActRIIA–mFc was determined to be 3.39 e⁻¹² M (Fig. S1A).

To evaluate the ability of the purified ActRIIA–mFc to antagonize activin-induced Smad signaling, a reporter gene assay was performed by using a human rhabdomyosarcoma cell line (A204) transfected with pGL3(CAGA)12-luc, a Smad 2/3-responsive construct (29), and pRL-CMV-luc as internal control. The cells were treated with activin A in the presence or absence of ActRIIA–mFc and assayed for luciferase activity after 6 h (Fig. S1B). ActRIIA–mFc inhibited activin A activity in a dose-dependent manner with an inhibitory concentration (IC₅₀) of 22 ng/ml (≈275 pM) by using 10 ng/ml (≈400 pM) activin A. These results are similar to those reported in previous studies of ActRIIA-binding affinities (25, 26). The data demonstrate that the purified ActRIIA–mFc fusion protein inhibited activin-induced signaling through the Smad pathway.

To test the function of ActRIIA–mFc in osteoclastogenesis, RAW246.7 cells were induced to differentiate by the addition of receptor activator of nuclear factor κb ligand (RANKL) and macrophage colony-stimulating factor (MCSF). The addition of activin (20 ng/ml) increased tartrate-resistant acid phosphatase (TRAP) expression, indicating an increase in osteoclast differentiation (Fig. S1C). The addition of 20 μg/ml ActRIIA–mFc inhibited activin-dependent osteoclast differentiation and restored TRAP expression to the levels of RANKL/MCSF differentiation.

We also assessed the ability of ActRIIA–mFc to inhibit activin function in osteoblasts by using normal human osteoblast (NHOst) cells. Cells were induced to differentiate by the addition of medium containing 100 nM dexamethasone and 10 nM β-glycerol phosphate. The addition of 50 ng/ml activin A to the differentiation medium at the beginning of the culture inhibited mineralized nodule formation, as reported previously (25). The addition of 20 μg/ml ActRIIA–mFc to the differentiation medium at the same time as activin A restored the ability of the osteoblasts to differentiate and form mineralized nodules (Fig. S1D).

To assess the skeletal response to in vivo activin inhibition, 12-week-old gonadally intact female C57BL/6N female mice received twice weekly i.p. injections of ActRIIA–mFc (10 mg/kg) or vehicle (VEH, PBS). Mice were euthanized after 2, 4, 6, and 12 weeks of treatment, and bones were assessed by static and dynamic histomorphometry, microcomputed tomography (μCT), and biomechanical testing. Static histomorphometry of trabecular bone in the distal femoral metaphysis showed that ActRIIA–mFc increased trabecular bone volume by 45%, 120%, 130%, and 248% versus VEH at 2, 4, 6, and 12 weeks, respectively (Fig. 1A; P ≤ 0.01). The increase in trabecular bone volume was due to an increase in both trabecular number (TbN) and trabecular thickness (TbTh) (Fig. 1 B and C; P ≤ 0.01 for both). The eroded surface per bone surface (ES/BS), osteoblast number per bone perimeter (Nob/Bpm), and osteoclast number per bone perimeter (Noc/Bpm) were decreased by ActRIIA–mFc treatment at two weeks (P ≤ 0.01) but did not differ from VEH thereafter (Fig. 1 D–F). The increased trabecular bone volume was secondary to an increase in bone formation because ActRIIA–mFc-treated mice exhibited a significant increase in mineralizing surface per bone surface (MS/BS) (80–120%), mineral apposition rate (MAR) (14–25%), and bone-formation rate (BFR, 115–155%) relative to VEH-treated mice over the course of the study (Fig. 1 G–I; P ≤ 0.01 for all). Representative images of von Kossa-stained femurs at 6 weeks of treatment are shown in Fig. 1J. Consistent with the histomorphometry, serum osteocalcin, a marker of bone formation, was increased after 2 and 4 weeks of ActRIIA–mFc treatment (Fig. 1K). In comparison, serum TRAP5b levels, which reflect osteoclast cell number, were unchanged after 2 weeks but began increasing after 4 weeks of ActRIIA–mFc treatment (Fig. 1L), indicating an increase of bone formation followed by an increase in osteoclastogenesis.

Confirming the histomorphometric results of increased trabecular bone volume, μCT of the fifth lumbar (L5) vertebrae revealed that mice treated with ActRIIA–mFc had greater trabecular bone volume when compared with VEH-treated mice (8%, 29%, 39%, and 51% after 2, 4, 6, and 12 weeks, respectively, Fig. 2A). The increase was mainly due to increased trabecular number (18%, 25%, and 48% at weeks 4, 6, and 12; P ≤ 0.01), and trabecular thickness (Fig. 2 B and C; P ≤ 0.05). Representative images from the μCT analysis are shown for VEH and ActRIIA–mFc after 6 weeks of treatment (Fig. 2D). To examine the impact of increased trabecular bone volume on bone strength, compression testing of the L5 vertebrae was performed. Mice treated with ActRIIA–mFc had greater compression strength (25–66%, Fig. 2E) and absorbed more energy (66–167%, Fig. 3F) when compared with VEH. Taken together, these data indicate that in normal adult mice, ActRIIA–mFc stimulates bone formation, resulting in increased trabecular bone mass and strength.

To further explore the effects of activin inhibition in a disease-state model, we determined the skeletal effects of ActRIIA–mFc in estrogen-deficient mice with established bone loss. Four-week-old C57BL/6 mice underwent ovariectomy (OVX) or sham (SHAM) surgery. After an 8-week period for bone loss to occur, mice were treated twice per week for 12 weeks with ActRIIA–mFc (10 mg/kg, i.p.) or VEH. In vivo pQCT measurements at the proximal tibia before treatment showed that trabecular bone density was ≈20% lower in OVX mice when compared with SHAM mice, indicating that ovariectomy had induced osteopenia (Fig. 3A, P ≤ 0.01). Four weeks after ActRIIA–mFc treatment, trabecular bone density (TbBMD) was increased in the proximal tibia in both OVX and SHAM mice relative to VEH (Fig. 3 A and B; P ≤ 0.01). At the end of 12 weeks of treatment, TbBMD in ActRIIA–mFc-OVX mice had increased 12% versus baseline (P ≤ 0.01), whereas TbBMD in OVX-VEH mice showed a 15% decrease (P ≤ 0.01) from baseline, for a net difference in TbBMD of 27% in between VEH control and ActRIIA–mFc-treated mice. In SHAM mice, ActRIIA–mFc treatment increased TbBMD by 27% relative to baseline (P ≤ 0.01), consistent with its anabolic activity. At the end of 12 weeks of treatment with ActRIIA–mFc, OVX mice had TbBMD levels comparable with SHAM-VEH mice (P = 0.1), indicating a reversal of the osteopenic phenotype.

Similarly, μCT of the L5 vertebrae revealed that OVX-VEH mice had reduced trabecular bone volume (Tb BV/TV) 20% when compared with SHAM-VEH mice (P ≤ 0.01), indicating that OVX had successfully induced osteopenia (Fig. 3C). Both OVX (+58%) and SHAM (+51%) ActRIIA–mFc-treated mice had significantly higher vertebral Tb BV/TV than their respective VEH-treated controls (Fig. 3C, P ≤ 0.01 for both). Furthermore, trabecular bone volume was higher in OVX mice treated with ActRIIA–mFc than in age-matched VEH-SHAM controls (P ≤ 0.01). Representative μCT images of the vertebrae from each treatment group are shown in Fig. 3D. A similar effect on trabecular bone was seen in the distal femur, where Tb BV/TV was significantly increased in ActRIIA–mFc-treated mice when compared with VEH controls (295% in OVX, 180% in SHAM, P ≤ 0.01 for both). Compression testing of the L5 vertebrae confirmed that ActRIIA–mFc treatment improved both the strength and the energy-absorbed-to-failure in OVX and SHAM mice (Fig. 3 E and F, P ≤ 0.01). Vertebral compressive strength of OVX mice treated with ActRIIA–mFc did not differ from age-matched SHAM-VEH controls (P = 0.1), whereas the OVX-VEH mice were significantly weaker than SHAM-VEH (P ≤ 0.01).

To identify effects of ActRIIA–mFc treatment on cortical bone, μCT analysis of the mid-femoral diaphysis was performed (Fig. 4A). Cortical thickness was significantly reduced in OVX-VEH when compared with SHAM-VEH (P ≤ 0.01) and was higher in ActRIIA–mFc-treated OVX mice (13%) and SHAM (19%) mice relative to VEH-treated controls (Fig. 4B; P ≤ 0.01 for both). The total cross-sectional area of the bone was increased in SHAM mice in response to ActRIIA–mFc treatment (P ≤ 0.01), whereas SHAM-VEH-, OVX-VEH-, and OVX-ActRIIA–mFc-treated mice were not significantly different from each other (Fig. 4C). The endosteal circumference in both treatment groups were reduced when compared with the VEH-treated controls (P ≤ 0.05) and mainly contributed to the increased cortical thickness.

To determine whether the cortical bone changes were associated with altered bone strength, three-point bending of the femur was performed. Femoral bending strength tended to be higher (P = 0.10), and energy-absorbed-to-failure was greater (P ≤ 0.01) in OVX mice treated with ActRIIA–mFc relative to controls (Fig. 4 D and E). In SHAM mice treated with ActRIIA–mFc, both femoral bending strength and energy-absorbed-to-failure were higher than VEH controls (Fig. 4 D and E, P ≤ 0.01). The OVX ActRIIA–mFc-treated mice showed similar or improved mechanical properties when compared with SHAM VEH controls, indicating rescue from the osteopenic phenotype.

The ActRIIA extracellular domain (A34-P115) was obtained by PCR amplification. The purified PCR fragment and the pAID4 mIgG2a vector were ligated to create the expression construct, and the sequence was confirmed by double strand dideoxy sequencing.

The pAID4-ActRIIA–mIgG2A construct was transiently transfected into HEK293 cells. After 1 week of growth at 37°C, the conditioned media were harvested by using a two-step column chromatography procedure. The purified ActRIIA–mFc was eluted from the column and then dialyzed into PBS.

A204 cells were plated in a 48-well plate (1 × 10⁵ cell per well) in McCoy's 5A media with 10% FBS and incubated at 37°C. The cells were transiently transfected with pGL3(CAGA)12-luc (a plasmid with the firefly luciferase gene controlled by SMAD 2/3 response element) and pRL-CMV-luc (a plasmid with Renilla luciferase controlled by a constitutively active CMV promoter) at a 10:1 ratio. Cells were incubated for 16 h at 37°C in fresh media with 0.1% BSA containing either activin alone (10 ng/ml) or a complex of activin and varying amount of ActRIIA–mFc preincubated for 45 min at 37°C. After 6 h, cells were lysed and assayed by using a Dual-Luciferase reporter assay kit (Promega). Results are expressed as a ratio of firefly over Renilla luciferase activity in relative light units.

Receptor ligand-binding affinities were determined by using a Biacore 3000 instrument (GE Healthcare Life Sciences). Goat anti-murine Fc antibody (Jackson ImmunoResearch) was immobilized on two flow cells of a CM5 chip. The ActRIIA–mFc was captured on the experimental flow cell by injecting the purified protein (2 μg/ml at 10 μl/min) across the bound antibody, resulting in ≈100 resonance units of captured receptor. The receptor–antibody complex remained stable over each ligand-binding cycle. For kinetic analysis, a concentration series of activin A was injected at a flow rate of 50 μl/ml over experimental and reference antibody surfaces at 20°C. The antibody surface was regenerated between binding cycles with injection of 10 mM glycine (pH 1.7). To obtain kinetic rate constants, corrected data were fit by using BiaEvaluation software. A kinetic analysis of each receptor–ligand interaction was obtained by fitting the response data to a reversible 1:1 biomolecular interaction model. The equilibrium constant K_D was determined by the quotient K_off/K_on. Constants are reported as the average of three or more independent analyses of receptor ligand interactions.

The mouse macrophage cell line RAW264.7 (ATTC TIB-71) was cultured in Dulbecco's modified Eagle's medium high glucose 1× (GIBCO), 10% FBS (SAFC Biosciences), 5 units of penicillin–streptomycin (Cellgro), and 0.25 μg/ml amphotericin B (Sigma-Aldrich). Osteoclast differentiation experiments were performed in 12-well plates seeded at 4 × 10⁴ cells per well in minimal essential medium-α without phenol red (Invitrogen), 10% heat-inactivated charcoal-stripped FBS (Invitrogen) with antibiotic, and antifungal agent. Media were supplemented with 20 ng/ml RANKL (R & D Systems), 30 ng/ml MCSF (R & D Systems), ±20 ng/ml activin A (R & D Systems), and ±20 μg/ml ActRIIA–mFc. Differentiation media were changed every 2–3 days over 12 days and used for the TRAP assay (Immunodiagnostics Systems).

NHOst cells (Lonza) were cultured in minimal essential medium-α without phenol red (Invitrogen), 10% FBS (SAFC Biosciences), 5 units of penicillin–streptomycin (Cellgro), and 0.25 μg/ml amphotericin B. Mineralization experiments were performed in 12-well plates seeded at 5 × 10⁴ cells per well. Mineralization medium was the same as above with the addition of 10% heat-inactivated charcoal-stripped FBS (Invitrogen), 20 mM Hepes (Cellgro), and 1.8 mM CaCl₂. Media were supplemented with 100 nM dexamethasone, 10 nM β-glycerol phosphate, ±20 ng/ml activin A (R & D Systems), and ±20 μg/ml ActRIIA–mFc. Media were changed every 2–3 days over the 14-day assay period. Cells were analyzed for bone mineral deposition by the use of the osteogenesis quantitation kit (Chemicon International). Cells were rinsed with PBS, fixed in formalin-free fixative (Sigma–Aldrich) for 15 min, and stained with alizarin red.

Gonadally intact and 4-week-old OVX- or SHAM-operated C57BL/6NTac female mice were purchased from Taconic. At 12 weeks of age, mice were administered ActRIIA–mFc protein at 10 mg/kg or PBS vehicle, twice per week by i.p. injection. Mice were euthanized after 12 weeks of treatment (n = 8–10). Mice were injected with calcein (20 mg/kg) at 9 and 2 days before euthanasia. Mice were euthanized after 2, 4, 6, or 12 weeks of treatment. At euthanasia, the femurs, tibias, and L4/L5 vertebrae were dissected for further analysis.

All in vivo experiments were performed with the approval of Acceleron Pharma's Institutional Animal Care and Use Committee.

The left tibia of anesthetized mice was analyzed during the course of the study by using an XCT Research SA+ pQCT scanner (Stratec Medizintechnik). Measurements were taken before dosing and then after 4, 6, 8, and 12 weeks of treatment with ActRIIA–mFc. CT slices at the proximal tibia were taken at 2.0 and 2.5 mm from the joint for trabecular bone and at 6.0 mm for cortical bone. Bone densities were calculated by using contour mode 1 and peel mode 2 with the provided software (version 5.50D).

Femurs and vertebrae were evaluated by using a desktop microtomographic imaging system (μCT40, Scanco Medical AG) equipped with a 10-mm focal spot microfocus x-ray tube. Transverse CT slices of femoral midshaft, the distal femoral metaphysis, and vertebral body (L5) were acquired by using 12-μm isotropic voxel size as described (38). For the vertebral body, ≈220 CT slices were obtained from the cranial to caudal growth plates. Images were reconstructed, filtered, and thresholded by using a specimen-specific threshold (39, 40). Morphometric parameters were computed by using a direct three-dimensional approach that does not rely on any assumptions about the underlying structure. For trabecular morphology, volume fraction (BV/TV, %), trabecular thickness (TbTh, μm), trabecular separation (TbSp, μm), and TbN (mm⁻¹) were assessed. For midfemoral cortical bone, the total cross-sectional area (TA, mm²), cortical bone area (mm²), bone area fraction (%), and cortical thickness (μm) were assessed.

Studies of the biomechanical properties of the femur in three-point bending and of the vertebral body in compression assays were performed as described (37–39). Load was applied at a constant rate (0.05 mm/s) until failure. The failure load (N), bending stiffness (N/mm), and work-to-failure (N × mm) were measured from the load-displacement curve.

Static and dynamic histomorphometry was performed after 2, 4, 6, and 12 weeks of treatment. Femurs were processed as described (41) and stained with 0.1% toluidine blue (pH 6.4), or Von Kossa. Static parameters of bone formation and resorption were measured in a defined area between 725 and 1,270 μm from the growth plate by using an OsteoMeasure bone histomorphometry system (Osteometrics). For static parameters, the Tb BV/TV, TbN (mm), TbTh (μm), NOb/BPm (mm⁻¹) and NOc/Bpm (mm⁻¹) per bone perimeter, and ES/BS (%) were measured. For dynamic histomorphometry, MS/BS (%) and mineral apposition rate (μm/day) were measured in unstained sections under UV light and used to calculate BFR (μm³/μm²/day), as described (41). The terminology and units used are those recommended by the Histomorphometry Nomenclature Committee of the American Society for Bone and Mineral Research (42).

For additional details on on materials and methods used, see SI Materials and Methods.