The growth of cancers in the skeleton is associated with devastating effects on bone and results in considerable morbidity and mortality. It is estimated that over 350,000 people in the United States alone suffer with cancers that spread to bone.1 Approximately 75% of patients who die of breast cancer have metastatic bone disease, and more than 80% of patients with myeloma develop osteolytic bone disease.2 Tumor-related osteolytic disease is caused by a disruption to normal bone homeostasis and is mediated by increased osteoclastic bone resorption and suppressed bone formation.3–7 The uncoupling of resorption and formation leads to rapid bone loss and the development of destructive and painful bone lesions. Despite antiresorptive therapy, repair of existing lesions is not seen, and progressive disease is associated with new skeletal events.8–13 New strategies to reduce skeletal-related events and improve patient outcome are urgently required.
The ability to stimulate new bone formation by promoting osteoblastogenesis offers the potential to retain the coupling of bone resorption with formation and to prevent the development of osteolytic disease. We sought to determine whether a novel bone-forming therapy that antagonizes activin-A signaling could block bone metastases and prevent the development of osteolytic disease in two classic examples of osteolytic disease, myeloma and breast cancer metastases.
Activins are members of the transforming growth factor β (TGF-β) superfamily, closely related to their natural antagonists, the inhibins. Inhibins are comprised of a common α subunit coupled to one of two β subunits leading to inhibin-A (αβA) or inhibin-B (αβB). Activins are homo- or heterodimers of the β subunits. The most commonly occurring β subunits are βA and βB, leading to the homodimers activin-A (βAβA), activin-B (βBβB), and the heterodimer activin-AB (βAβB). Activins and inhibins were first identified as important regulators of pituitary follicle-stimulating hormone release.14–17 Subsequent studies have demonstrated that activins are expressed in diverse tissues and play important roles in embryology, wound healing, and tissue homeostasis.18–22 Activins signal by binding either the activin receptor type IIA (ActRIIA) or the activin receptor type IIB, which recruits the type I receptor, activin-like kinase 4 (Alk4), leading to phosphorylation of cytoplasmic Smads 2/3. This complex associates with Smad 4 and translocates to the nucleus to regulate gene transcription.
Activin-A is expressed abundantly in bone.23 Neither activin-B nor activin-AB has been detected in bone. However, conflicting evidence exists concerning the role of activin-A in bone. It has been reported to both inhibit and stimulate osteoblastogenesis in vitro and to promote osteoclast formation in vitro.24–27 In vivo, direct administration of activin-A increases bone mineral density (BMD),28 whereas overexpression of inhibin-A, which blocks activin-A and other TGF-β family members, increases bone mass.29, 30 We have demonstrated recently that activin-A signaling can be blocked in vivo with a soluble ActRIIA.muFc fusion protein, leading to increased bone mass and strength.31
Levels of activin-A have been shown to be increased in the sera of patients with breast and prostate cancer and to be increased further in the presence of metastases.32 However, the effect of targeting activin-A in tumors that grow in bone or metastasize to bone is not yet known. Equally, the effect of restoring osteoblast function on tumor-induced osteolytic disease is unclear. The aim of this study was to determine whether targeting activin-A specifically could prevent the uncoupling of resorption and formation and the development of osteolytic disease and bone metastasis in models of myeloma and breast cancer metastasis.
Materials and Methods
Expression of recombinant ActRIIA.muFc protein
The extracellular domain of the high-affinity activin type IIA receptor was cloned into a pAID4 expression vector generated in-house using a cytomegalovirus promoter and a dihydrofolate reductase selection/amplification marker and transfected into HEK293 cells. ActRIIA.muFc was purified by sequential column chromatography from the supernatant.31 Receptor-ligand binding affinities were determined using a Biacore 3000 instrument (GE Healthcare Life Sciences, Piscataway, NJ, USA).31
Osteoblast mineralization assays
Human osteoblasts (NHOst, Lonza, Allendale, NJ, USA) cultured in α-MEM containing penicillin-streptomycin, amphotericin B, and 10% fetal bovine serum (FBS) were seeded at 2.5 × 104/well in 12-well plates in medium supplemented with FBS (2%), 20 mM HEPES, 1.8 mM CaCl2, 100 nM dexamethasone, and 10 nM β-glycerol phosphate. Cells were incubated with vehicle (PBS), activin-A (20 µg/m; R&D Systems, Minneapolis, MN, USA), ActRIIA.muFc (20 µg/mL), or a monoclonal antibody against activin-A (mab3381, 10 µg/mL; R&D Systems). Concentration of alizarin red after 14 days was measured by determining absorption at 450 nm using a VERSAmax microplate reader (Molecular Devices, Tokyo, Kanto, Japan). The control receptor.muFc was a construct combining the extracellular domain of the activin-like kinase receptor 1 (ALK-1) fused to the same IgG (Alk1-Fc). In-house studies have shown that Alk1-Fc fails to bind to any of the activins. Assays were performed in triplicate.
Long bones were isolated from 6- to 8-week-old MF1 mice. Epiphyses were removed, and marrow was flushed out with PBS. Cells were centrifuged at 1000g, washed in PBS, and resuspended at 5 × 106 cells/mL in MEM (Gibco, Paisley, Renfrewshire, Scotland, UK) containing 10% FBS, 2 mM L-glutamine, 100 U/mL penicillin, and 100 µg/mL amphotericin B, supplemented with 10−7 M prostaglandin E2 (PGE2) (Sigma, St Louis, MO, USA), and 50 ng/mL of macrophage-colony stimulating factor (M-CSF; R&D Systems). Cells were placed into tissue culture flasks and incubated for 24 hours at 37°C in 5% CO2. Nonadherent cells were collected by centrifugation and resuspended in MEM containing 10−7 M PGE2, 25 ng/mL of M-CSF, and 3 ng/mL of receptor activator nuclear factor-κ ligand (RANKL; R&D Systems). Then 250-µm-thick dentine slices in 96-well trays were incubated with 200 µL of cell suspension (106 cells) overnight at 37°C and 5% CO2. Disks were transferred to 12-well trays containing 2 mL of medium and 10 to 100 ng/mL of activin-A (R&D Systems). Trays were incubated for 5 days at 37°C in 5% CO2, 95% air, with half medium changes every 2 days. The medium was maintained at approximately pH 7.4 using 5 mM NaOH. the medium was acidified to approximately pH 6.9 using 10 mEq/l H+ (as HCl) on day 7 to activate osteoclastic resorption.33 Cultures were fixed in 2% glutaraldehyde and stained to demonstrate tartrate-resistant acid phosphatase (TRACP). Osteoclasts were defined as TRACP+ cells with two or more nuclei and/or clear evidence of resorption pit excavation. Total numbers of osteoclasts were counted and surface area of resorption pits measured using dot-counting morphometry. Assays were repeated at least three times.
Measuring serum concentrations of activin-A
Concentration of activin-A in serum was measured by ELISA. Briefly, microtiter plates (Immulon, Thermo Labsystems, VWR, Poole, Dorset, UK) were coated overnight with anti-activin-A antibody (1 µg/mL; R&D Systems) diluted in PBS at room temperature. Plates then were incubated for 1 hour at room temperature with PBS, 1% BSA, and 5% sucrose. After three cycles of washing with PBS and 0.05% Tween 20, recombinant activin-A used as standard (R&D Systems) and serum were incubated for 2 hours at RT. After washing, the biotinylated form of the anti-activin-A antibody (0.25 µg/ml, R&D Systems) was added for 2 hours at room temperature. Streptavidin–horseradish peroxidase (R&D Systems) then was added for 20 minutes. After washing, the substrate solution (R&D Systems) was incubated for 20 minutes. Finally, the reaction was stopped by the addition of 50 µL of 1 M H2SO4. The optical density (OD), which is proportional to the concentration of activin-A, was measured at a wavelength of 450 nm and corrected for absorbance at 540 nm using a SPECTRAmax M5e microplate reader (Molecular Devices).
Determining the effect of ActRIIA.muFc on bone mass in normal mice
Eight-week-old female C57/BL/6 mice were treated with vehicle (PBS) or ActRIIA.muFc (0.1, 0.3, 3.0, and 10 mg/kg, i.p. 2× per week) for 6 weeks (n = 10 per group). The effect of ActRIIA.muFc on BMD and bone structure was measured as described below (Supplemental Fig. 1A).
Determining the effect of ActRIIA.muFc on osteolytic disease in 5T2MM-bearing mice
The 5T2MM murine model of myeloma originated in C57BL/KaLwRij mice and is propagated by sequential transfer of diseased bone marrow into young syngeneic mice.34, 35 In a preliminary longitudinal study, C57BL/KaLwRiJ mice were injected with 2 × 106 5T2MM cells via the tail vein and did not receive any other interventions in order to study the development of disease features. Mice were euthanized (n = 3) at intervals, serum paraprotein was determined, and long bones were processed for radiogaphic and histologic analysis. Remaining mice were euthanized at 13 weeks (Supplemental Fig. 1B).
For interventional studies, in which treatment with ActRIIA.muFc commenced from week 1 after tumor cell inoculation with 2 × 106 5T2MM cells via the tail vein, female mice were grouped and left uninjected (naive, n = 10) or injected with 5T2MM cells (tail vein) and treated with ActRIIA.muFc (n = 10) or vehicle (n = 8) (10 mg/kg, 2× per week, i.p.). Six and two days prior to euthanization, mice were injected with the fluorochrome calcein (30 mg/kg, i.v.; Sigma-Aldrich) to enable analysis of bone-formation rate (Supplemental Fig. 1C). For interventional 5T2MM studies, in which mice were treated with ActRIIA.muFc from the time of serum paraprotein detection (week 8), female mice were grouped and left uninjected (naive, n = 9) or injected with 2 × 106 5T2MM cells (tail vein) and treated with ActRIIA.muFc (n = 9) or vehicle (n = 10) (10 mg/kg, 2× per week, i.p.) (Supplemental Fig. 1D). Mice were euthanized at 12 weeks.
Determining the effect of ActRIIA.muFc on osteolytic disease in the MDA-MB-231 model of metastatic breast cancer
In this experiment, 1 × 105 MDA-MB-231-luc-D3H2LN cells (Caliper Life Sciences, Hopkinton, MA, USA) were injected into the left ventricle of athymic nude mice (Nude-Foxn1nu/Foxn1+; Harlan, Indianapolis, IN, USA). Mice were treated with ActRIIA.muFc (n = 12) or vehicle (n = 10) (10 mg/kg, 2× per week, s.c.) commencing 2 weeks prior to injection of tumor cells. A cohort of naive mice (n = 10) also was treated with ActRIIA.muFc (10 mg/kg, 2× per week, s.c.). BMD was measured at intervals, and bone structure was examined at death (8 weeks) (Supplemental Fig. 1E).
Radiographic and micro–computed tomographic (µCT) analysis of BMD, bone structure, and osteolytic bone disease
In studies of healthy, non-tumor bearing mice, BMD was measured at the beginning of the study and following 6 weeks of ActRIIA.muFC treatment by in vivo peripheral quantitative computed tomographic (pQCT) analysis using an XCT Research SA+ pQCT scanner (Stratec Medizintechnik, Pforzheim, Baden-Württemberg, Germany). Tibias were scanned at 2.0 and 2.5 mm below the tibial tuberosity. BMD was determined using Stratec software. Following euthanization, tibias and femurs were scanned on a Scanco µCT40 (Scanco Medical AG, Wayne, PA, USA) using a 12-µm voxel size. Trabecular volume (BV/TV, %) and number (Tb.N, /mm) were determined using Scanco software in a region 360 µm proximal to the growth plate extending 1.8 mm proximally.36
In the preliminary longitudinal study of 5T2MM-bearing mice, tibas were radiographed using a Faxitron (Hewlett-Packard, Lincolnshire, IL, USA), and osteolytic lesion number was counted. For interventional 5T2MM studies, in which mice were treated with ActRIIA.muFc from week 1 after tumor cell inoculation, long bones were scanned on a Skyscan µCT 1172 (Skyscan, Kontich, Belgium) at 50 kV and 200 µA using a 0.5-mm aluminium filter and a detection pixel size of 4.3 µm2. Images were captured every 0.7 degrees through 180 degrees of rotation and analyzed using Skyscan software. A volume of 1 mm3 of trabecular bone positioned 0.2 mm from the growth plate was analyzed. BV/TV (%), Tb.Th (µm), and Tb.N (mm−1) were assessed. Osteolytic lesions were analyzed along a 4-mm region of cortical bone proximal to the growth plate using Image J software (National Institute of Health, Bethesda, MD, USA). For interventional 5T2MM studies, in which mice were treated with ActRIIA.muFc from time of serum paraprotein detection (week 8), long bones were radiographed as described earlier, and osteolytic lesion number was counted prior to scanning using the Skyscan µCT.
For interventional studies in MDA-MB-231-bearing mice, BMD was measured weekly by pQCT, as described earlier. Following euthanization, tibias were scanned and trabecular bone volume (BV/TV%) determined.
Histologic and histomorphometric analysis
Right tibias were fixed in formalin, decalcified in EDTA, and embedded in paraffin, and 3-µm sections were cut on a microtome (Leica Microsystems, Milton Keynes, Northamptonshire, UK). Sections were stained with hematoxylin and eosin (H&E) or TRACP and counterstained with Gills hematoxylin. Osteoblast number (N.OB/BS, /mm) and proportion of surface occupied by osteoblasts (OB.S/BS, %) were determined in trabecular bone in a 0.75-mm2 area 0.25 mm from the growth plate and on corticoendosteal surface 3 mm in length starting 0.25 mm from the growth plate using Osteomeasure histomorphometry software (Osteometrics, Decatur, GA, USA). Osteoclast numbers (N.OC/BS, /mm) and osteoclast surface (OC.S/BS, %) were measured in the same region.
Left tibias were fixed in 70% ethanol and embedded in LRWhite resin (Taab Laboratories, Reading, Berkshire, UK), and undecalcified sections prepared and examined under ultraviolet (UV) illumination (DMRB microscope; Leica Microsystems). Mineralized surface and separation between two fluorochrome labels were measured using Osteomeasure software. Mineralized surface (MS, %) was calculated as (dL + 0.5sL/BS) × 100, where dL = double-label surface, sL = single-label surface, and BS = total bone surface. Mineral apposition rate (MAR, µm/day) was calculated as Il.W/time between injections, where Il.W = interlabel width. BFR/BS (mm3/mm2/day) was calculated as [MAR × (dL + 0.5sL)]/BS.37
Assessing tumor burden and survival in studies in the 5T2MM and 5T33MM models of myeloma
In the preliminary longitudinal study, 5T2MM tumor burden was determined by measuring serum paraprotein by electrophoresis, by isolating the bone marrow from the femurs and staining tumor cells with an anti-idiotype antibody, by determining the proportion of tumor cells by flow cytometry, and by measuring the proportion of bone marrow (%) occupied by 5T2MM cells on H&E-stained sections of the tibia in 7 areas, each measuring 0.25 mm2, beginning 0.25 mm from the growth plate. In interventional studies, tumor burden was determined using the techniques described earlier and by histologic assessment of right and left tibias.
In separate studies, 5T33MM bearing mice were treated with ActRIIA.muFc or vehicle (10 mg/kg, 2× per week, i.p.) commencing 1 week before injection of 5T33MM cells. Mice were euthanized when they showed signs of morbidity (Supplemental Fig. 1F). The effect of ActRIIA.muFc on time to morbidity was determined by Kaplan-Meier analysis. The 5T33MM murine model of myeloma exhibits a more rapidly aggressive phenotype than 5T2MM cells, allowing rapid accumulation of survival data.38
Determining the effect of ActRIIA.muFc on the metastasis potential of MDA-MB-231-bearing mice and survival
MDA-MB-231-luc-D3H2LN cells (purchased from Caliper Life Sciences) transfected with luciferase to facilitate in vivo imaging were injected into the left ventricle (1 × 105 cells) of athymic nude mice (Nude-Foxn1nu/Foxn1+; Harlan). Mice were treated with vehicle (n = 16) or ActRIIA.muFc (10 mg/kg, 1× per week, s.c.) (n = 9) commencing 2 weeks prior to injection of tumor cells. At week 6, mice were injected with 150 mg/kg of luciferin (Caliper Life Sciences), and the presence of metastasis was determined by luciferase imaging using the IVIS Spectrum detector (Caliper Life Sciences). Number of mice with bone metastasis and number of bone metastasis per mouse were determined. Mice were euthanized on exhibiting signs of morbidity.
Comparisons of the effect of ActRIIA.muFc on osteoblast mineralization, osteoclastogenesis, and bone parameters between naive mice and 5T2MM-bearing mice treated with vehicle or ActRIIA.muFc were undertaken by one-way ANOVA with a Bonferroni post-hoc test. The effect of ActRIIA.muFc on BMD and bone structure in normal mice was determined by ANOVA with a post hoc test (Holm-Sidak for pQCT, Fisher's PLSD for µCT data). Time-dependent changes in indices of tumor burden and bone turnover were analyzed by ANOVA with a Tukey post hoc test. Lytic lesion data were analyzed using Kruskal Wallis ANOVA on ranks with Dunn's method for multiple comparisons.
Blocking activin-A signaling stimulates mineralization and increases bone mass
To assess the effect of targeting activin-A on osteoblast differentiation, we first determined the effect of activin-A on osteoblast mineralization in vitro. Activin-A blocked mineralization of normal human osteoblasts (NHOsts, p < .05), whereas ActRIIA.muFc and a monoclonal anti-activin-A antibody (mab3381) prevented this inhibitory activity (p < .01 and p < .05, respectively; Fig. 1A). Both ActRIIA.muFc and mab3381 alone increased mineralization of NHOsts independent of exogenous activin-A, consistent with activin-A being expressed by NHOsts26 and having an autocrine effect (p < .05 and p < .01, respectively; Fig. 1A). A control receptor.muFc fusion protein with no activin-A binding activity had no effect on mineralization, confirming the specificity of the ActRIIA.muFc construct (data not shown). The effect of targeting activin-A on osteoclast number and activity was assessed using murine osteoclast cultures. Activin-A administered at 10 and 100 ng/mL did not increase number or activity of osteoclasts, as evidenced by total osteoclast count and unchanged bone resorption area (Fig. 1B).
We next investigated whether blocking activin-A with ActRIIA.muFc would increase bone mass in vivo. Treatment of C57BL/6N female mice with ActRIIA.muFc for 6 weeks showed dose-dependent increases in trabecular BMD in the tibias (p < .01; Fig. 1C), as measured by µCT. µCT analysis also showed dose-dependent increases in trabecular bone volume (p < .001) and trabecular number (p < .001; Fig. 1D).
Blocking activin-A with ActRIIA.muFc treatment from the time of tumor cell injection prevents development of osteolytic bone disease in 5T2MM myeloma-bearing mice
We then investigated whether targeting activin-A with ActRIIA.muFc could prevent the formation of osteolytic bone lesions in models of cancer-induced bone disease. C57BL/KaLwRiJ mice were injected with 5T2MM murine myeloma cells, and disease was assessed. In untreated mice, idiotype-positive 5T2MM cells were first detected in the bone marrow at week 4 (0.7% ± 0.3%) until the bone marrow was fully infiltrated with myeloma cells (83% ± 17%) at week 13 (data not shown). A serum paraprotein, measured by electrophoresis, was detected at week 8 and increased to week 13 (data not shown). Radiographic and histologic analyses demonstrated the presence of osteolytic bone lesions (p < .001) and a time-dependent reduction in trabecular bone area (p < .05; Fig. 2A, D). Histomorphometric analysis demonstrated that this was caused by an increase in the number of osteoclasts (p < .05), a decrease in the number of osteoblasts (p < .001), and reduction in the bone undergoing mineralization (p < .05; Fig. 2B–D). Serum concentrations of activin-A were assessed in mice bearing 5T2MM cells by ELISA. The concentration of activin-A was 520 ± 302 pg/mL.
Mice injected with 5T2MM cells were treated with ActRIIA.muFc from time of tumor cell injection and were euthanized at 12 weeks. In non-myeloma-bearing mice, osteoblasts were found lining trabecular and corticoendosteal surfaces, whereas in disease-bearing mice, osteoblasts were not present in areas infiltrated with tumor cells. In contrast, mice injected with 5T2MM cells and treated with ActRIIA.muFc retained osteoblasts lining bone surfaces (Fig. 3A). ActRIIA.muFc treatment significantly increased osteoblast number and the surface covered by osteoblasts compared with mice treated with vehicle (p < .001 in each case; Fig. 3A). Furthermore, mineralization also was reduced in tumor-bearing mice (p < .001, Fig. 3C, D) and ActRIIA.muFc treatment increased mineralization (p < .05; Fig. 3C, D). 5T2MM cells also reduced bone-formation rate and mineral apposition rate, which reflects new bone formation at individual bone remodeling sites (p < .001 and p < .01, respectively; Fig. 3C, D). ActRIIA.muFc treatment also prevented this inhibition (p > .05 in each case; Fig. 3C, D). ActRIIA.muFc treatment reduced osteoclast number by 21% and bone surface occupied by osteoclasts by 23%, but this was not significant (Fig. 3B).
Critically, these studies demonstrated that promoting bone formation with ActRIIA.muFc prevented the formation of osteolytic bone lesions induced by 5T2MM cells (p < .05; Fig. 4A–C). Additionally, 5T2MM cells caused a loss of trabecular bone in the tibias (39%, p < .05), femurs (50%, p < .01), and vertebrae (21%, p < .05) that was completely blocked with ActRIIA.muFc (p < .001 in each case). ActRIIA.muFc treatment also prevented the tumor-induced decrease in trabecular number in the tibia (35%, p < .05), femur (43%, p < .001), and vertebrae (15%, p < .01) and increased trabecular thickness (p < .001 in each case; Fig. 4A, B, D; vertebrae data not shown).
Treatment with ActRIIA.muFc after the emergence of a serum paraprotein also prevents osteolytic bone disease in 5T2MM myeloma-bearing mice
In separate studies, 5T2MM myeloma-bearing mice were treated with ActRIIA.muFc only after the detection of a serum paraprotein and were treated until euthanization. µCT analysis of tibias demonstrates that treatment of 5T2MM myeloma-bearing mice with ActRIIA.muFc from week 8 prevents 5T2MM-induced reductions in trabecular volume and trabecular number (p < .05; Fig. 5). Cortical bone in tibias analyzed in this study were not significantly altered between groups. Analysis of plain radiographs of the tibias also demonstrated that osteolytic lesion number was reduced in the tibias of 5T2MM myeloma-bearing mice treated with ActRIIA.muFc compared with 5T2MM myeloma-bearing mice treated with vehicle only (p < .05; Fig. 5).
Inhibiting activin-A increases survival in 5T33MM myeloma-bearing mice
In addition to determining the effect of ActRIIA.muFc on bone disease, we also sought to establish the effect of treatment on tumor burden and survival. Myeloma cells were shown to express the ActRIIA receptor directly; however, activin-A had no effect on cell proliferation (data not shown). ActRIIA.muFc treatment caused a 41% reduction in serum paraprotein, a marker of whole-body tumor burden, and a 37% reduction in tumor burden in bone (Fig. 6A, B), although these reductions were not significant. In contrast, the spleen, a site of extramedullary tumor growth, did not show a change in weight with ActRIIA.muFc treatment (Fig. 6C). Importantly, ActRIIA.muFc treatment of 5T33MM myeloma-bearing mice caused an increase in time to morbidity (p < .05; Fig. 6D).
Blocking activin-A prevents the development of osteolytic disease and reduces bone metastases in MDA-MB-231-bearing mice
We have shown previously that bone formation is also suppressed in a model of osteolytic disease induced by breast cancer cells.39 This suggests that similar cellular mechanisms may contribute to the development of bone lesions in both myeloma and breast cancer. Since ActRIIA.muFc had such a profound effect on preventing myeloma bone disease, we also sought to establish whether blocking activin-A would prevent metastatic bone disease using the MDA-MB-231 model of breast cancer–induced bone destruction. Intracardiac injection of MDA-MB-231-luc-D3H2LN cells resulted in a decrease in BMD and the development of bone lesions, both of which were blocked by treatment with ActRIIA.muFc (p < .001; Fig. 7A). µCT analysis confirmed that ActRIIA.muFc prevented the reduction in bone volume induced by the tumor (p < .005; Fig. 7A). To investigate the effect of ActRIIA.muFc on metastasis and survival, mice were injected with MDA-MB-231-luc-D3H2LN cells and their growth was monitored by in vivo bioluminescence imaging (Fig. 7B). ActRIIA.muFc reduced the numbers of mice with evidence of bone metastasis (4 of 10 treated with ActRIIA.muFc compared with 9 of 10 treated with vehicle, p < .05) and the numbers of metastases per mouse (p < .05; Fig. 7B, C). Furthermore, MDA-MB-231-bearing mice treated with ActRIIA.muFc survived for 11 days longer than those treated with vehicle (p = .06; Fig. 7D). As with myeloma cells, MDA-MB-231 cells expressed the ActRIIA receptor, although neither activin-A, nor ActRIIA.muFc had any effect on cell proliferation. Furthermore, treatment of mice bearing MDA-MB-231 cells grown as subcutaneous xenografts with ActRIIA.muFc had no effect on tumor burden (data not shown).
We have shown that blocking activin-A signaling stimulates osteoblast differentiation and mineralization in vitro and increases bone mass in vivo. This is consistent with reports showing that activin-A inhibits osteoblast differentiation in cultures of fetal rat calvarial cells24 but contrasts with other studies that report that activin-A stimulates osteoblastogenesis.25 However, we also have demonstrated that activin-A inhibits osteoblast differentiation and that this inhibition may be reversed with the addition of ActRIIA.muFc. Interestingly, we also show that both ActRIIA.muFc and mab3381 alone increased mineralization in vitro. One explanation for this observation is that these agents bind activin-A, which is produced endogenously by osteoblasts. This is consistent with previous reports showing activin-A release from osteoblasts.26, 40 However, we cannot exclude the fact that these both ActRIIA.muFc and mab3381 bind other ligands and stimulate osteoblast activity.
Importantly, we showed that ActRIIA.muFc increased BMD in C57BL/6N female mice. This is consistent with our previous data and demonstrates that inhibiting activin-A with ActRIIA.muFc can increase bone mass under physiologic conditions.31 This is also consistent with the observation that inhibin-A, which also blocks activin-A, is a stimulator of bone mass and strength.29
In addition to investigating the effects of activin-A and ActRIIA.muFc on osteoblastogenesis, we also examined their effects on osteoclast formation. We found that activin-A did not significantly alter osteoclast number or activity in vitro. This contrasts with other studies.25, 27 The reason for the discrepancy between these studies is unclear but may reflect differences between culture systems and/or other technical differences.
Despite this, our observation that activin-A had no effect on osteoclasts in vitro is consistent with our observations in vivo, in which we found that ActRIIA.muFc had no significant effect on osteoclasts numbers. The absence of any effect of treatment on osteoclasts in vivo in our tumor model may reflect the choice of time point. Mice were examined after 12 weeks of treatment, and it is possible that early effects on osteoclastic bone resorption may have been missed. However, our studies in normal mice have examined the effect of ActRIIA.muFc at 2, 4, 6, and 12 weeks of treatment.31 Two weeks of treatment was associated with a small but significant decrease in osteoclast number and eroded surface, whereas treatment for 4, 6, or 12 weeks had no effect. This suggests that there may be short-term effects but that these are not sustained and may not be sufficient to account for the prevention of development of bone lesions seen in this study. Furthermore, in a recent report, osteoclast numbers in SCID mice bearing INA6 myeloma cells and treated with ActRIIA.muFc also were not significantly different from those in vehicle-treated mice.40 It is also worthy of note that the effect of human ActRIIA.IgG1.Fc (ACE-011) on bone structure and cells has been studied in cynomolgus monkeys. As with the murine construct, ActRIIA was able to increase bone mass and stimulate bone formation. Osteoclast resorption was reduced at some but not all sites.41 Whether the discrepancies between studies reflect differences between the human and mouse constructs, differences between rodent and primate systems, or differences between anatomic sites is unclear.
We then assessed the effect of treatment with ActRIIA.muFc in models of cancer-induced bone disease. In mice bearing 5T2MM myeloma cells treated with ActRIIA.muFc, bone volume was increased, osteolytic lesions were reduced, and osteoblast numbers were increased. However, no effects were observed on osteoclast numbers. This is consistent with data reporting that overexpression of inhibin-A has no effect on osteoclast number.28 The absence of significant effect on osteoclasts suggests that osteoblasts also play a pivotal role in regulating the development of osteolytic bone disease in multiple myeloma.
Myeloma in humans presents only once disease is well established. In order to model the clinical situation, separate studies withheld treatment with ActRIIA.muFc until the emergence of a serum paraprotein. This strategy also conferred significant protection against the development of osteolytic bone disease. This raises the possibility that targeting activin-A with an appropriate soluble ActRIIA construct could offer thereapeutic potential in the clinical setting.
Increased survival also was observed in 5T33MM myeloma-bearing mice treated with ActRIIA.muFc, suggesting antitumor effects. In support of this, tumor burden in bone, measured histologically and by determining serum paraprotein, a marker of overall tumor burden, was reduced in mice bearing 5T2MM cells. In contrast, spleen weight, a surrogate marker of extramedullary tumor load, was unchanged between mice bearing 5T2MM cells treated with either ActRIIA or vehicle. This suggests that ActRIIA.muFc treatment reduces tumor growth in bone but not at extramedullary sites. Furthermore, treatment of myeloma cells with activin-A also had no effect on cell proliferation. These data therefore suggest that any change in myeloma burden is likely to be mediated indirectly by changes in the local bone microenvironment rather than any direct antimyeloma effect.
We also have shown that blocking activin-A prevents the development of osteolytic bone disease in a murine model of breast cancer. Furthermore, treatment with ActRIIA.muFc reduces bone metastases in mice bearing MDA-MB-231 breast cancer cells. This is consistent with the data in the 5T2MM model and with evidence showing that overexpression of follistatin, a physiologic inhibitor of activin, reduces metastasis in an experimental model of small cell lung cancer metastasis.42 The absence of a direct effect of either activin-A or ActRIIA.muFc on MDA-MB-231 cell proliferation in vitro or subcutaneous tumor growth in vivo supports the suggestion that these effects are mediated indirectly via changes in the bone microenvironment.
Therefore, in the model of myeloma, these data demonstrate that specific blockade of activin-A signaling through the ActRIIA receptor by administration of soluble ActRIIA.muFc prevents suppression of osteoblastogenesis caused by tumor cells and stimulates bone formation. This stimulation of bone formation prevents the development of classic osteolytic bone lesions. This is evidence that stimulating bone formation in this way can prevent what is typically thought to be disease caused by increased osteoclastic bone resorption. Importantly, this effect appears to be predominantly independent of bone resorption because treatment had no significant effect on osteoclast number. Thus the protective effect of targeting activin-A appears to be mediated via preservation of osteoblast differentiation and osteoblastic activity. This suggests that maintaining osteoblast differentiation and osteoblastic activity retains the coupling of bone resorption and bone formation seen in physiologic bone remodeling and perturbed in tumor-induced osteolytic bone disease. This mechanism appears key to the protective effects of ActRIIA.muFc. Since we saw similar structural effects in the breast cancer model and we also have seen osteoblast suppression in this system, it is likely that similar mechanisms maybe responsible in this setting.
Importantly, targeting osteoblasts with ActRIIA.muFc was associated with antitumor effects, decreased metastases, and increased survival. This was common to models of both multiple myeloma and breast cancer metastasis. We and others have shown previously that inhibiting bone resorption is also associated with antitumor effects in these experimental models.43–46 However, osteoblastic bone formation was not typically examined in these previous studies. This report is one of the first to provide evidence that stimulating osteoblast differentiation and bone formation is also associated with antitumor effects and the reduction of metastases. The absence of a significant effect of ActRIIA.muFc treatment on osteoclastic bone resorption but a positive effect on osteoblast number, coupled with the absence of an antitumor effect at extramedullary sites, allows us to conclude that osteoblasts may well regulate tumor growth and survival. Indeed, this is consistent with recent evidence that osteoblasts have inhibitory effects on the growth of myeloma cells mediated by the small leucine-rich proteoglycan decorin.47 Thus we now need to consider that osteoblasts, as well as osteoclasts, play a key role in regulating tumor growth and survival in bone.
We have shown that promoting osteoblast differentiation and bone formation by inhibiting activin-A with ActRIIA.muFc prevents osteolytic bone disease, reduces tumor burden, and increases survival. Therapeutic approaches to maintaining osteoblast function is a critical new approach and represents a new paradigm for the treatment of tumors that develop in bone, such as myeloma and breast cancer bone metastases. These data suggest that one approach to stimulating osteoblast differentiation in this setting is to block activin-A with an ActRIIA.Fc construct. Although it may be argued that such a strategy will decrease follicle-stimulating hormone (FSH) levels and reduce fertility, this is likely to be reversible and to be an acceptable risk in the oncology setting. Thus, blocking activin-A represents a potential new therapeutic approach to stimulating osteoblast differentiation in multiple myeloma– and breast cancer–induced bone disease.
AWM, SP, EDW, and JS are employees of Acceleron Pharma, Cambridge, MA, USA; PC and MLB have received research funding from Acceleron Pharma, Cambridge, MA, USA. ADC, DH, MB, LC, HE, NA, MLK, TRA, and KV state that they have no conflicts of interest.
We would like to acknowledge the following and thank them for their expert technical assistance: Angelo Willems and Orla Gallagher. We also gratefully acknowledge the following sources of funding: AC and PC were supported by Leukaemia and Lymphoma Research; PC is an investigator on the National Institute of Health Research Biomedical Research Unit for Bone; KV was supported by the Fonds voor Wetenschappelijk Onderzoek Vlaanderen; MK and TAt were supported by the Biotechnology and Biological Sciences Research Council and the Arthritis Research Campaign. Acceleron Pharma was supported by a generous grant from the Multiple Myeloma Research Foundation.