RANKL-Targeted Combination Therapy with Osteoprotegerin Variant Devoid of TRAIL Binding Exerts Biphasic Effects on Skeletal Remodeling and Antitumor Immunity

Osteolytic bone metastasis is a major event in patients with breast cancer, accounting for about 70% with skeletal pathology (1). A key signaling event that triggers aggressive osteoclast activity involves receptor activator of NF-κB ligand (RANKL; TNFSF11), produced by osteoblast, bone marrow stromal cells, monocytes, Th cells, and disseminated cancer cells, and its cognate receptor, RANK (TFNRSF11A), on osteoclast precursors (2). In normal skeletal homeostasis, a balance between osteoclast function and bone remodeling is controlled by osteoprotegerin (OPG; TNFRSF11B), a soluble decoy receptor for RANK with high-affinity binding to RANKL thereby blocking osteoclast activation (2). However, OPG also binds to TNF-related apoptosis-inducing ligand (TRAIL; TNFSF10) of the TNF superfamily, enabling tumor cells to evade apoptosis (3). As skeletal dissemination of breast cancer progresses, the checkpoint effect of OPG is compromised, which leads to aggressive osteoclast activity (4). This effect is further compounded by the production of an inhibitor of Wnt signaling pathway, Dickkopf-related protein 1 (DKK1), that blocks osteoblast differentiation (5). Studies have also shown that RANKL–RANK axis promotes tumor cell development and mammary tumorigenesis. RANK expression leads to the spontaneous development of mammary tumor and MMTV-RANK mice showed higher incidence of adenocarcinomas (6), but a lack of RANK showed reduced mammary tumorigenesis (7). Importantly, treatment with soluble RANK-Fc attenuated mammary tumorigenesis in MMTV-RANK mice (6). Recent studies have provided direct proof that RANKL–RANK pathway plays a vital role in the progression of BRAC1 mutation–associated breast cancer (8). Moreover, high RANK expression was found to be significantly associated with poor disease-free survival in patients with breast cancer, especially the triple-negative breast cancer (TNBC; ref. 9). Taken together, RANKL targeting is promising for breast cancer with bone metastasis, beyond its effects on skeletal pathology.

In addition to its role in bone remodeling, RANKL–RANK signaling has been implicated in the immune regulation (10). In the context of bone metastasis, tumor-promoting immunosuppressive M2 macrophages have been identified to express high surface RANK (10–12), prompting the importance of RANKL–RANK signaling in macrophage polarization (13, 14). Furthermore, expression of RANK has also been reported in breast cancer cells (15, 16). Recent clinical and preclinical studies have revealed a potential role for blockade of RANKL–RANK interaction in patients with cancer and in animal models to potentiate anticancer immunotherapy (17–20). These studies showed that the combination of anti-RANKL antibody with anti-CTLA and/or anti-PD-1 antibodies significantly reduced human melanoma and experimental mouse melanoma lung metastases via CD8⁺ T cells and natural killer cells, compared with individual treatment (18, 19). Thus, targeting RANKL–RANK signaling is emerging as a promising approach to treat pleomorphic tumor-associated pathologies.

In this context, this study investigated the potential of individual and combination cell therapy, using a novel OPG variant (OPG^Y49R), devoid of TRAIL binding with an agonistic antibody (MD5-1) targeting the TRAIL receptor, DR5 (21), in an immunocompetent mouse model of osteolytic breast cancer. Results of the study indicated targeting excess RANKL with OPG^Y49R without affecting endogenous TRAIL function not only restored bone remodeling against breast cancer–induced osteolytic damage, but also improved antitumor immune response by directly affecting M2 macrophages polarization and production of immunosuppressive cytokines and chemokines. Furthermore, in combination with MD5-1 antibody, OPG cell therapy approach significantly increased long-term antitumor immunity, leading to long-term survival and protection against tumor relapse.

The human embryonic kidney cell line, HEK 293T, human breast cancer cell lines, MDA-MB-231 (TNBC) and MCF-7 [estrogen receptor (ER)-positive], and the murine macrophage cell line, RAW 264.7 (a kind gift from Dr. Xu Feng, The University of Alabama at Birmingham, UAB, Birmingham, AL), were maintained in DMEM (Gibco, 11965-092), supplemented with 10% FBS (Omega Scientific, FB-02) and 1% Penicillin–Streptomycin (Gibco, 15140-122). BALB/c mouse mesenchymal stem cells (MSC) were provided by Dr. Donald G. Phinney (Scripps Research, Jupiter, FL), and maintained in alpha-MEM (Gibco, 32561-037) supplemented with 10% FBS and 1% penicillin–streptomycin. The human prostate cancer cell line, PC3 (a kind gift from Dr. Kenneth J. Pienta, University of Michigan, Ann Arbor, MI), and murine TNBC cell line, 4T1.2 (kindly provided by Dr. Robin L. Anderson, Olivia Newton-John Cancer Research Institute, Melbourne, Victoria, Australia), were maintained in RPMI1640 Medium (Gibco, 11875-093), supplemented with 10% FBS and 1% penicillin–streptomycin.

Recombinant TRAIL (GF092) and chemiluminescent horseradish peroxidase substrate (WBKLS0500) were purchased from MilliporeSigma, and bacterially purified, recombinant RANKL was a kind gift from Dr. Xu Feng (UAB, Birmingham, AL; ref. 22). Recombinant OPG (459-MO) and M-CSF (216-MC) proteins were purchased from R&D Systems. Recombinant murine IL4 (214-14) and IL13 (210-13) were purchased from PeproTech. Mouse IL2 (130-120-662) was purchased from Miltenyi Biotec. Details of all antibodies used in this study are listed in the Supplementary Table S1.

4T1.2 cells were transduced with a recombinant lentiviral vector expressing luciferase and stable clones were selected by puromycin. Luciferase expression was confirmed by in vitro luciferase assay and in vivo noninvasive luciferase imaging. Identical in vivo growth kinetics of 4T1.2^Luc to the parental 4T1.2 cell line were verified in female BALB/c mice. For further details, see Supplementary Materials and Methods.

MSCs were plated in 60-mm dishes at a density of 4 × 10⁵. Next day, the cells were infected with rAAV-OPG^WT or rAAV-OPG^Y49R variant at a multiplicity of infection of 1 × 10⁵ vector particles/cell and cultured for 2 days. On day 3, OPG-expressing MSCs were harvested for in vivo injection. Prior to in vivo application, lysates and culture supernatants from rAAV-OPG^WT/Y49R–transduced MSCs were confirmed by Western blotting for expression and extracellular secretion of the transgenic protein. For cloning, packaging, and purification of rAAV-OPG^WT and rAAV-OPG^Y49R variant, see Supplementary Materials and Methods.

TRAIL-induced apoptosis assay was performed using PC3 cells by culturing them in media containing recombinant TRAIL alone or in combination with conditioned media from wild-type OPG (OPG^WT) or OPG^Y49R plasmid–transfected HEK 293T cells. Osteoclast differentiation assay was performed using RAW264.7 cells by culturing cells in medium containing recombinant RANKL alone or in combination with conditioned media containing OPG^WT or OPG^Y49R. 4T1.2^Luc cell proliferation was performed by culturing cells in medium containing recombinant RANKL ± recombinant OPG. Further details are provided in Supplementary Materials and Methods.

Femurs and tibiae were obtained from 8-week-old female BALB/c mice and bone marrow cells (BMC) were flushed out with a syringe filled with RPMI1640 medium. Next, cells were treated with ACK Buffer (Lonza, 10-548E) to lyse red blood cells, then resuspended in FACS buffer, and stained with CD11b-VioBlue antibody. CD11b⁺ BMCs were flow sorted and plated in 6-well plate (1 × 10⁶ cells/well). To generate bone marrow–derived macrophages, cells were cultured in RPMI1640 medium, supplemented with 10% FBS, 1% penicillin–streptomycin, and 20 ng/mL M-CSF for 7 days, and then differentiated to M2 macrophages with IL4 (50 ng/mL) and IL13 (20 ng/mL) in the presence of RANKL ± OPG. M2 differentiation was confirmed by flow cytometry using M2 macrophage markers and by qRT-PCR for M2 macrophage–specific gene transcripts.

4T1.2^Luc cells (1 × 10⁵) in PBS were injected into the fourth mammary fat pad of 8-week-old female BALB/c mice. Mice were divided into four groups and treated with OPG^Y49R, MD5-1, OPG^Y49R and MD5-1 combination; and untreated group was used as the control. MD5-1 antibody (100 μg/mouse) was injected intraperitoneally three times: on days 2, 5, and 9. Approximately, 2 × 10⁵ of MSC-OPG^Y49R was injected by tail vein into mice on days 3 and 10. Tumor progression was monitored using a digital caliper every other day, starting 1 week postimplantation, and tumor volume was calculated as: volume = (length × width²)/2. For the metastasis model, primary tumors developed as above were surgically removed when tumors reached a volume of approximately 100 mm³, following which OPG^Y49R and MD5-1 combination treatment was given. One cohort was maintained without treatment as control. Metastasis development was monitored using noninvasive IVIS imaging. All animal studies were conducted in accordance with an approved protocol of the Institutional Animal Care and Use Committee of UAB (Birmingham, AL).

For determination of the three-dimensional (3D) architecture of the trabecular bones, mice were sacrificed, tibiae were harvested, and scanned using a SCANCO μCT40 Instrument (Scanco Medical). For histology, bone tissues were decalcified and sections of the bones were prepared. Hematoxylin and eosin (H&E) staining, Goldner trichrome staining, and tartrate-resistant acid phosphatase (TRAP) staining were performed as detailed in Supplementary Materials and Methods.

Soluble proteins were extracted from primary tumors of untreated and OPG^Y49R-treated mice and analyzed using Proteome Profiler Mouse Cytokine Array Kit (R&D Systems, ARY006) following the manufacturer's instructions. For further details, see Supplementary Materials and Methods.

For adoptive T-cell transfer, CD8⁺ T cells were flow sorted from treated mice on day 20. Normal BALB/c mice were used as control. Sorted CD8⁺ T cells were plated in anti-CD3–coated (10 μg/mL) 12-well tissue culture plates and cultured in RPMI1640 medium plus IL2 (50 U/mL) and anti-CD28 antibody (2 μg/mL) for 3 days. Later, 1 × 10⁶ CD8⁺ T cells were harvested, and injected through tail vein into recipient mice with preestablished orthotopic tumors. Tumor progression was monitored using digital caliper every other day, starting 5 days after tumor challenge. For tumor cell rechallenge studies, 1 × 10⁵ 4T1.2^Luc cells were inoculated at the opposite fourth mammary fat pad of the BALB/c mice that were originally implanted with tumor cells and treated with MD5-1 or a combination of OPG^Y49R and MD5-1 antibody. A new cohort of 4T1.2^Luc-challenged female BALB/c mice was used as control for tumor growth kinetics. Tumor progression was monitored using a digital caliper every other day, starting 1 week postimplantation.

Results consisting of three groups or more were analyzed using two-way ANOVA with Tukey multiple comparisons test by GraphPad Prism 8. Analysis of results containing two groups was performed using Student t test. All data are presented as mean ± SD. P < 0.05 was considered statistically significant.

Additional details on Materials and Methods are provided in Supplementary Materials and Methods.

We recently identified a TRAIL-binding aromatic residue on human OPG and established that removal of this side chain abolished interaction with TRAIL, without affecting RANKL binding (23, 24). Interestingly, comparison of OPG orthologs revealed that this motif is highly conserved across species (Supplementary Fig. S1). To utilize this OPG variant (OPG^Y49R) for therapy effects on bone remodeling without affecting immune cell–produced TRAIL effects on breast cancer cells in an immunocompetent mouse model, we first created a mouse OPG^Y49R variant by site-directed mutagenesis and tested its effects, first in a TRAIL assay. Conditioned media from mouse OPG^WT or OPG^Y49R variant plasmid–transfected HEK 293T cells were tested on PC3 cells in the presence of recombinant TRAIL protein. TRAIL-induced apoptosis was analyzed by cell viability and Western blot analysis, using anti-caspase-3 antibody. We observed that TRAIL treatment induced significant apoptosis in both recombinant TRAIL–treated group and in combination with conditioned medium containing OPG^Y49R (Fig. 1A). However, when conditioned medium from OPG^WT-transfected HEK 293T cells was added with TRAIL, there was a significant protection from apoptosis. Furthermore, Western blot analyses corroborated the above results with increased levels of cleaved caspase-3 and a concomitant decrease in total caspase-3 when conditioned medium from OPG^Y49R-transfected cells was used, but not with medium from OPG^WT-transfected cells (Fig. 1B). This result confirmed that OPG^Y49R does not bind to TRAIL. In parallel, we also confirmed by osteoclast assay that, substitution of an aromatic side chain in OPG^Y49R did not compromise RANKL binding. In an osteoclast assay using mouse macrophage cell line, RAW 264.7, addition of conditioned medium containing OPG^Y49R in combination with RANKL significantly prevented osteoclast differentiation. This effect was found to be comparable with the OPG^WT-treated group (Fig. 1C). These data confirmed that murine OPG^Y49R abolished TRAIL binding, but retained RANKL binding.

To utilize OPG^Y49R for therapy effects on bone remodeling in breast cancer–induced osteolytic damage, without compromising immune cell–produced TRAIL effects in vivo, we adopted a cell therapy approach using mouse MSCs. MSCs were genetically engineered by infection of recombinant adeno-associated virus expressing OPG^WT (rAAV-OPG^WT) or OPG^Y49R (rAAV-OPG^Y49R). Prior to producing rAAV-expressing OPG, we tested different AAV serotypes for their ability to transduce MSCs from BALB/c mouse using rAAV-GFP and established that AAV serotype-4 capsid possesses highest transduction efficiency in mouse MSCs (Supplementary Fig. S2A). On the basis of this observation, we cloned, packaged, and purified rAAV4, encoding OPG^WT or OPG^Y49R. Prior to tumor challenge studies in vivo, high level expression and extracellular secretion of OPG were confirmed by transducing mouse MSCs with rAAV-OPG^WT or rAAV-OPG^Y49R and analyzing both cell lysates and supernatants by Western blotting (Supplementary Fig. S2B). Next, we examined the potential of mouse OPG^Y49R in protecting tumor-induced bone loss in vivo. To monitor the bone metastasis noninvasively, we established a clonal derivative of 4T1.2 cells by stable transduction with lentiviral vector expressing luciferase (4T1.2^Luc) and confirmed the identical in vivo growth kinetics to that of parental 4T1.2 cells (Supplementary Fig. S3A). Using the 4T1.2^Luc cell line, an osteolytic breast cancer model was established by inoculating this syngeneic mouse breast cancer cell line into fourth mammary fat pad of female BALB/c mice. It was interesting to note that the 4T1.2/BALB/c displayed bone damage by systemic effects without visible bone metastasis (Supplementary Fig. S3B). To first verify the potential of OPG^Y49R against tumor-induced bone loss, cohorts of female BALB/c mice with preestablished orthotopic breast tumors were untreated or treated with MSCs producing OPG^WT or OPG^Y49R, or MSC infected with rAAV-GFP. Normal BALB/c mice were used as control. Bone damage was examined on day 18 by micro-CT. Tumor challenge caused significant bone damage in untreated and MSC-GFP–treated groups, as shown by decreased trabeculae and bone volume compared with normal BALB/c mice (Fig. 1D). However, treatment with MSCs producing either OPG^WT or OPG^Y49R displayed a significant decrease in osteolytic bone damage compared with untreated and MSC-GFP–treated groups (Fig. 1D), suggesting therapy effects with OPG^Y49R were effective in restoring skeletal homeostasis against cancer-induced osteolysis.

Prior to testing the effects of blocking RANKL–RANK signaling directly on tumor growth in vivo, we evaluated the significance of RANKL in 4T1.2^Luc cell growth and downstream signaling on tumor cell proliferation. We identified by flow cytometry (Fig. 2A) and Western blotting (Fig. 2B) using anti-RANK antibody (R12-31) (25) that, RANK expression was explicit in 4T1.2^Luc cells as in human breast cancer cell lines, MBA-MB-231 and MCF 7. Next, we tested direct effects of RANKL on 4T1.2^Luc proliferation and activation of downstream signaling pathways, and how OPG influences RANKL-induced effects in these cellular processes. To determine this, 4T1.2^Luc cells were cultured in the presence or absence of RANKL. Recombinant OPG was added to the cultures, either alone or with RANKL. Results indicated a significant increase in 4T1.2^Luc cell proliferation when RANKL was added to the culture. Interestingly, when OPG was added to the culture in the presence of RANKL, the growth rate of 4T1.2^Luc cells decreased to the level without RANKL (Fig. 2C). Addition of OPG alone to 4T1.2^Luc cells did not influence rate of cell proliferation (Fig. 2C). These data provide key evidence that RANKL–RANK axis promotes breast cancer proliferation. A major activation pathway following RANKL–RANK ligation involves the recruitment of an adaptor protein, TNF receptor associated factor 6 (TRAF6), that relays downstream signaling via IκB to activate transcription factors, NF-κB and NFATc1, or via the MAPK cascade involving p38, ERK, and JNK. RANKL–RANK interaction also relays downstream signaling via the PI3K pathway through c-Src, which leads to AKT signaling (26). To investigate which signaling pathway is involved in enhanced breast cancer cell proliferation upon RANKL stimulation, we examined the changes in phosphorylation status of NF-κB p65, ERK1/2, and AKT. We observed a significant increase in NF-κB and MAPK activation, whereas there was no change in AKT phosphorylation (Fig. 2D). Interestingly, when OPG was added to 4T1.2^Luc cells in the presence of RANKL, both NF-κB and ERK1/2 phosphorylation restored to levels in cultures without RANKL (Fig. 2D). These data indicated that OPG inhibited RANKL-induced 4T1.2^Luc cell proliferation by blocking RANKL–RANK signaling pathway through NF-κB and ERK1/2 MAPK.

From recent studies it is becoming increasingly clear that RANKL–RANK pathway is involved in promoting mammary tumorigenesis and protumorigenic immune cell functions. Together with our observations on the potential of OPG^Y49R in skeletal remodeling in osteolytic malignancy, we hypothesized that systemically stable, genetically engineered cell therapy to produce OPG^Y49R in physiologically required concentration, without interfering in TRAIL-induced apoptosis of breast cancer cells, would provide therapeutic benefits for osteolytic bone damage by minimizing tumor growth directly and by improving antitumor immunity through dampening tumor-associated M2 macrophage function. We tested this possibility in female BALB/c mice with preestablished orthotopic 4T1.2^Luc tumors. To further examine whether antibody targeting of TRAIL receptor, DR5, as in clinically tested agonistic mAb approach (27), would enhance antitumor effects, we included cohorts of mice that received only MD5-1 antibody (mouse DR5-targeting agonistic mAb) or in combination with cell-based OPG^Y49R therapy. MD5-1 antibody was applied intraperitoneally and MSC-OPG^Y49R was delivered by tail vein injection. Overall experimental schema is shown in Supplementary Fig. S4A. Results of the study demonstrated a significant decrease in tumor growth in both MD5-1 and MSC-OPG^Y49R individual treatment groups. However, the combination of MD5-1 and OPG^Y49R resulted in a significantly greater decrease in tumor growth kinetics (Fig. 3A). Tibiae from cohorts of mice from control, MD5-1, and/or MSC-OPG^Y49R treatment groups were further analyzed by micro-CT and histology to distinguish therapy effects on bone remodeling. Data obtained from micro-CT were analyzed for bone volume, trabecular thickness, trabecular number, and trabecular spacing. Results from micro-CT data, H&E staining, and trichrome staining showed a significant destruction of the trabecular bone in the untreated tumor group, compared with age-matched control (Fig. 3B and C; Supplementary Fig. S4B). Increased osteoclast activity was also observed in the untreated tumor group by TRAP staining (Fig. 3D). Interestingly, although MD5-1 standalone treatment decreased tumor growth, there was significant bone damage and osteoclast activity in this group (Fig. 3B,–D; Supplementary Fig. S4). However, treatment with MSC-OPG^Y49R alone or in combination with MD5-1 restored bone integrity and decreased osteoclast activity, similar to the control group (Fig. 3B,–D; Supplementary Fig. S4), confirming superior effects of this combination in simultaneously restoring skeletal integrity and decreasing tumor growth.

Recent studies suggest that RANKL–RANK signaling promotes cancer-associated inflammation through M2 macrophage activation, and RANKL induces M2 macrophage polarization (10, 14). On the basis of significant therapy effects of MSC-OPG^Y49R on both bone remodeling and tumor growth, indicating benefits of targeting RANK expression on tumor cells, we further established higher RANK expression in M2 macrophages and monocytic myeloid-derived suppressor cells (MDSC), compared with M1 macrophages and polymorphonuclear MDSCs, respectively. Furthermore, there was no significant difference in RANK expression between myeloid dendritic cells (DC) and plasmacytoid DCs (Fig. 4A; Supplementary Fig. S5A and S5B). Because M2 macrophages displayed highest RANK expression, next, we examined the impact of dampening RANKL signaling systemically with MSC-OPG^Y49R on M2 macrophages. Spleen and tumors were isolated from control and 4T1.2^Luc tumor–challenged mice with/without OPG^Y49R treatment, and macrophages were analyzed by flow cytometry. Results of this study showed tumor challenge significantly increased total macrophages (Fig. 4B). Interestingly, there was neither a significant decrease in macrophage numbers nor in M2/M1 ratio based on M1 and M2 macrophage surface markers following OPG^Y49R treatment (Fig. 4B and C). On the basis of this, we sought to understand possible effect of OPG on M2 macrophage polarization and/or function. CD11b⁺ monocytes from bone marrow of BALB/c mice were differentiated with IL4 and IL13 for M2 macrophage in the presence of RANKL ± OPG, following which cells were analyzed for M2 polarization markers by flow cytometry and for gene expression signatures by qRT-PCR. Results indicated that addition of RANKL did not significantly increase M2 macrophage numbers (Fig. 4D, left). However, both arginase 1 (28, 29) gene expression and enzyme levels were significantly increased following exposure to RANKL, but addition of OPG blocked RANKL-induced arginase 1 production (Fig. 4D, right and E, left). We also observed that RANKL increased CCL17 (30, 31) expression and addition of OPG significantly blocked it (Fig. 4E, right).

To further examine whether dampening RANKL influences macrophage- and protumorigenic immune cell–produced cytokines and chemokines in vivo, we investigated cytokine/chemokine signals in the tumor microenvironment (TME). Primary tumors from untreated and OPG^Y49R-treated 4T1.2^Luc tumor–challenged mice were explanted and soluble proteins were extracted and analyzed using proteome profiler mouse cytokine array kit. Result showed that treatment with MSC-OPG^Y49R induced remarkable changes in cytokine/chemokine profiles (Fig. 5A and B). Following OPG^Y49R treatment, there was upregulation of Th1 cytokines: IL1α, IL1β, IL2, IL12, IL16, IL17, IL23, INFγ, and TNFα (Fig. 5C). In addition, OPG^Y49R treatment also resulted in an increase in IL1ra and CCL5 and a decrease in chemokines, CXCL1 and CXCL2 (Fig. 5A,–C), indicating the potential of OPG^Y49R in blocking RANKL-mediated protumorigenic functions.

Upon establishing the dual therapeutic effects of OPG^Y49R in restoring skeletal homeostasis and reducing M2 macrophage functional mediators, we reasoned that simultaneous targeting of tumor cells and the TRAIL signaling pathway would maximize antitumor effects and effector/memory T-cell functions. To determine the effect of OPG^Y49R and MD5-1 combination therapy on long-term T-cell function, we performed adoptive transfer studies, wherein splenetic CD8⁺ T cells from control mice, tumor-challenged untreated mice, and tumor-challenged MSC-OPG^Y49R, MD5-1, and MSC-OPG^Y49R + MD5-1–treated mice were sorted by flow cytometry and adoptively transferred to female BALB/c mice with preestablished tumors in mammary fat pad and the tumor growth kinetics were monitored. Results of this study indicated a significant decrease in tumor growth following administration of CD8⁺ T cells from both MD5-1 and MD5-1 and OPG^Y49R combination treatment received mice, but the effect was more pronounced in the group of mice that received CD8⁺ T cells from OPG^Y49R and MD5-1 combination treatment (Fig. 6A). However, there was no protection against the growth of tumors following adoptive transfer of T cells from control, untreated, or OPG^Y49R alone treated mice. In addition, cohorts of tumor-free mice following MD5-1 ± OPD^Y49R treatment were rechallenged with 4T1.2^Luc cells to determine antitumor memory cell reactivation. Results of this study showed a significant protection against regrowth of secondary tumors in mice that were tumor free following OPG^Y49R and MD5-1 combination therapy. Mice that were treated initially with MD5-1, showed a significant decrease in the growth of rechallenged tumors, but in mice that received combination therapy, tumor regrowth was completely blocked (Fig. 6B). These data indicate that OPG^Y49R and MD5-1 combination therapy improves tumor-specific T-cell memory functions.

Finally, we tested whether application of the combination therapy would protect against visceral metastasis, extending its therapeutic benefit when given during resection of the primary tumors. Breast cancer disseminates to both skeletal and nonskeletal tissues. In addition to bone, common sites of breast cancer metastasis are liver, lung, and brain. To determine this, we developed 4T1.2^luc tumors orthotopically in fourth mammary fat pad of female BALB/c mice. When the primary tumors reached a volume of approximately 100 mm³, they were surgically resected, followed by either no treatment (naïve) or application of MSC-OPG^Y49R and MD5-1 combination therapy. Recurrent tumor growth at the primary site and/or visceral metastasis was monitored by noninvasive luciferase imaging. Results of this study demonstrated a significant decrease, both in the incidence of recurrent tumors at the primary site and in visceral metastasis, compared with naïve group without treatment (Fig. 6C). Metastasis-free survival rate was significantly increased in mice that received the combination therapy (Fig. 6D). These data suggest that OPG^Y49R and MD5-1 combination therapy can also be used in conjunction with debulking surgery to prevent/reduce metastasis.

Over the past decade, it has become clear that RANKL–RANK signaling axis is not only involved in bone homeostasis and remodeling, but also in mammary gland development, immune function, and tumorigenesis (32, 33). Especially in breast cancer pathophysiology, RANKL–RANK pathway seems to play a vital role in aggressive growth and metastasis, as supported by clinical data correlating RANK and RANKL expression with poor disease-free survival in patients with TNBC (9), higher RANKL levels in ER-positive breast cancers correlating with higher risk (34), and RANKL as a regulator of BRCA1 mutation–driven breast cancer (8). Therefore, targeting overt RANKL–RANK signaling is emerging as a viable option for antitumor functions. Using an immunocompetent mouse TNBC model, recapitulating breast cancer–associated pathologies as in the human disease, we established the potential of a variant OPG, devoid of TRAIL inhibitory function, both as a standalone therapy and in combination with agonistic activation of the death receptor pathway in tumor cells.

In corroboration with published reports that human breast cancer cells express RANK (15, 16), we found that 4T1.2^Luc cells express RANK and addition of RANKL to the culture significantly increases cell proliferation (Fig. 2A–C). Binding of RANKL to membrane-bound RANK triggers intracellular signaling through recruitment of adaptor proteins: TRAFs, Gab2, Cbl, and c-Src (26). In a cell-specific context, this pleomorphic signaling leads to activation of the NF-κB, PI3K-AKT, and MAPK cascades (26). Although a key activation pathway of RANK signaling during osteoclastogenesis involves NF-κB, NFATC1, and downstream osteoclast-specific genes, including cathepsin K, we found RANKL stimulation of 4T1.2^Luc cells resulting in the activation of both the NF-κB and MAPK pathways, involved in aggressive tumor growth (35, 36). However, treatment with OPG in the presence of RANKL reverted to basal proliferation rate and protumorigenic signaling molecules in breast cancer cells (Fig. 2C and D). Furthermore, application of systemic MSC-OPG^Y49R as a standalone therapy not only reverted osteolytic bone damage, but also decreased tumor growth (Fig. 3A–C), indicating the potential of OPG^Y49R in directly targeting RANKL-mediated tumor growth. Recent reports also identified that RANK-positive tumor cells of luminal progenitors are highly proliferative, compared with RANK-negative tumors (8). Furthermore, in three distinct datasets, including 684 breast cancer patient samples, RANK mRNA expression was higher in primary tumor tissues of patients with TNBC, compared with other molecular subtypes, and more interestingly, patients with TNBC who exhibited both RANK and RANKL had significantly worse relapse-free survival (9), signifying the use of OPG^Y49R in coadjuvant setting for TNBC. Considering the lack of receptors for therapeutic targeting in TNBC and this molecular subtype disease associated with poor prognosis with least 5-year survival rate among all breast cancer subtypes, the use of OPG^Y49R as a biological molecule for dual therapy implications on skeletal lesions and against tumor proliferation without interfering in TRAIL-mediated apoptosis of tumor cells further augments the relevance of its potential application.

Although cell-based OPG^Y49R therapy significantly reversed bone damage and established bone architecture to that seen in normal bone anatomy, and decreased tumor growth in vivo, the latter effect was significantly improved when combined with agonistic activation of DR5, using MD5-1 antibody (Fig. 3A,–C). This superior combinatorial effect was found to be due to induction of tumor apoptosis and activation of antitumor immune functions. In particular, following combination treatment with OPG^Y49R and MD5-1, there was a significant increase in tumor-specific T-cell memory functions, the potency of which was further verified by adoptive transfer and tumor rechallenge studies (Fig. 6A and B). Although there was no effect on total M2 macrophage numbers (Fig. 4B,–D), arginase 1 and CCL17 production induced by RANKL was blocked by OPG (Fig. 4D and E). Both arginase 1 and CCL17 are related to immunosuppression of tumor-associated macrophages (TAM; refs. 28–31). Superimposing our findings with published studies, which indicated immunosuppressive effects of RANKL on M2 macrophages (11, 12) and on the existence of two different populations of M2 macrophages (Arg⁺ and Arg⁻; ref. 29), our data suggest that RANKL plays a major role in converting Arg⁻ M2 macrophages to Arg⁺ highly immunosuppressive cells. These data support a direct therapeutic potential of blocking RANKL–RANK signaling in M2 macrophages by OPG^Y49R to decrease immunosuppressive effects. Differential analysis of soluble factors in the TME by immune array in mice treated with OPG^Y49R alone identified upregulation of key Th1 cytokines IL2, IL12, and INFγ, and downregulation of chemokines, CXCL1 and CXCL2 (Fig. 5A,–C). Although a precise source of these cytokines and chemokines is not established in this study, most of these soluble immune mediators are known to be produced by cells of both innate and adaptive immune system (37). Interestingly, we also found a stark upregulation of IL1ra, a biological antagonist of IL1 signaling (38). IL1ra has been reported to inhibit tumor growth and metastasis by decreasing MDSCs and TAMs accumulation (39). Thus, blocking the RANKL–RANK axis with OPG^Y49R at the TME would decrease TAM function and consequently metastasis.

We also observed upregulation of IL16 following treatment with OPG^Y49R. IL16, a ligand for CD4 and functions as chemoattractant for T cells, has been reported to be expressed in high concentrations in M1 macrophages compared with M2 macrophages (40). Data from cytokine array, identifying increased IL1ra and IL16, together with a decrease in arginase production by M2 macrophages following OPG treatment, are suggestive of beneficial effects of blocking RANKL–RANK signaling to promote effector T-cell infiltration at the TME for antitumor effects. CXCL1 and CXCL2 have been reported to enhance MDSCs and promote breast cancer metastasis through NF-κB/SOX4 signaling (41, 42). Decrease in chemokine ligands, CXCL1 and CXCL2, and increase in Th1 cytokines corroborate the significance of OPG^Y49R combination therapy, possibly by decreasing immunosuppression and promoting infiltration of Th1 cells to the TME, as we found an increase in CD8⁺ T cells following OPG^Y49R and MD5-1 combination therapy (Supplementary Fig. S6). Immunotherapy approaches using immune checkpoint inhibitors (ICI) are emerging as more powerful alternatives for many cancers, including carcinomas of the breast (43). However, only a minority of patients respond effectively to ICIs, including mAbs to CTLA-4, PD1, and PD-L1. Multiple reasons account for this insufficiency, including mutation burden, bulk and clonality of neoantigens, levels of immune checkpoints, IFNγ responsiveness, and composition of the TME itself, which enables evasion of immune surveillance through production of additional suppressive mediators such as VEGF and IKK2 (44, 45). Hence, combining OPG^Y49R with other immune modulating agents would be beneficial for patients refractory to ICIs. In this context, targeting the RANKL–RANK axis with OPG^Y49R, capable of reverting and preventing bone damage, to simultaneously exert antitumor effects directly through arresting RANK-mediated signaling cascades, without interfering in endogenous and/or exogenous TRAIL-induced apoptosis of tumor cells, will be a promising new alternate treatment modality. RANK signaling is also known to play an important role in generating regulatory T cells that are immunosuppressive (46, 47). Hence, blocking augmented RANKL–RANK signaling in both innate and adaptive protumorigenic immune cells will be highly beneficial.

The most prevalent adjuvant therapy for breast cancer with bone dissemination is denosumab, a monoclonal RANKL antibody, used commonly in combination with aromatase inhibitors, especially in patients with malignancy-related hypercalcemia after bisphosphonate therapy (48). From a recently concluded, large ABCSG-18 trial, although denosumab treatment reduced clinical features, there was only a modest disease-free survival (49, 50). As encountered with use of mAbs in clinical regimen, common side effects with denosumab include infection of abdomen and urinary tract, pancreatitis, musculoskeletal pain, and hypocalcemia (51). Furthermore, controversy exists regarding osteonecrosis of the jaw (52), prompting a careful follow-up for patients prescribed denosumab. It was interesting to note in our studies that use of MD5-1 alone although resulted in a significant reduction in tumor growth kinetics in vivo, there was no therapy effect on bone damage. Hence, OPG^Y49R combination with tumor-targeted therapies should extend dual benefits on both skeletal damage and tumor growth. The cell-based therapy approach we have adopted using MSCs offers additional advantages such as an immune-privileged cell source as genetically engineered MSCs can be used in allogenic context to produce systemically stable levels of OPG^Y49R in physiologically relevant concentrations. Further characterization of this approach from pharmacokinetics and long-term standpoints should result in clinical application of this dual targeting not only for patients with osteolytic malignancy in carcinoma of the breast, but also other osteolytic cancers including multiple myeloma, kidney cancer, and prostate cancer with apparent skeletal pathology.

No potential conflicts of interest were disclosed.

H. Wang: Conceptualization, validation, investigation, methodology, writing-original draft. R. Ashton: Methodology. J.A. Hensel: Methodology. J.H. Lee: Methodology. V. Khattar: Methodology. Y. Wang: Methodology. J.S. Deshane: Funding acquisition, methodology. S. Ponnazhagan: Conceptualization, resources, data curation, formal analysis, supervision, funding acquisition, validation, investigation, visualization, project administration, writing-review and editing.

This work was supported by NIH grants R01 AR060948 and R01CA184770 (to S. Ponnazhagan) and R01HL128502 (to J.S. Deshane), and grants from the Breast Cancer Research Foundation of Alabama (to S. Ponnazhagan). We thank the Heflin Center Genomics Core, the Pathology Research Core, the Small Animal Bone Phenotyping Core, the Preclinical Shared Imaging Facility, and the Comprehensive Flow Cytometry Core at UAB for support with DNA sequencing, histology, μCT scan, IVIS imaging, and flow cytometry analyses, respectively.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.