Engineered Osteoclasts Resorb Necrotic Alveolar Bone in Anti-RANKL Antibody-Treated Mice

Abstract

Medication-related osteonecrosis of the jaw (MRONJ) is a serious side effect of antiresorptive medications such as denosumab (humanized anti-RANKL antibody), yet its pathophysiology remains elusive. It has been posited that inhibition of osteoclastic bone resorption leads to the pathological sequelae of dead bone accumulation, impaired new bone formation, and poor wound healing in MRONJ, but this hypothesis has not been definitively tested. We previously engineered myeloid precursors with a conditional receptor activator of nuclear factor kappa-Β intracellular domain (iRANK cells), which differentiate into osteoclasts in response to a chemical inducer of dimerization (CID) independently of RANKL. In this study, we showed that CID-treated iRANK cells differentiated into osteoclasts and robustly resorbed mineralized surfaces even in the presence of anti-RANKL antibody in vitro. We then developed a tooth extraction-triggered MRONJ model in nude mice using anti-RANKL antibody to deplete osteoclasts. This model was used to determine whether reconstitution of engineered osteoclasts within sockets could prevent specific pathological features of MRONJ. Locally delivered iRANK cells successfully differentiated into multinucleated osteoclasts in response to CID treatment in vivo as measured by green fluorescent protein (GFP), tartrate-resistant acid phosphatase (TRAP), carbonic anhydrase II, matrix metallopeptidase 9 (MMP-9), and cathepsin K staining. Sockets treated with iRANK cells + CID had significantly more osteoclasts and less necrotic bone than those receiving iRANK cells alone. These data support the hypothesis that osteoclast deficiency leads to accumulation of necrotic bone in MRONJ.

Introduction

Medication-related osteonecrosis of the jaw (MRONJ), which was first described in 2003, is a potentially serious side effect of treatment with antiresorptive and antiangiogenic medications that often occurs in the oral cavity after traumatic dentoalveolar procedures such as tooth extraction.(1, 2) These antiresorptive medications, including bisphosphonates (BPs) and denosumab, are used to decrease the risk of skeletal-related events in patients with severe osteoporosis, multiple myeloma and bone metastasis of other cancers.(1) MRONJ is clinically defined as accumulation of an area of exposed and necrotic bone that has persisted for more than eight weeks with no history of radiation therapy.(2, 3) Initially, lesions are asymptomatic but may become symptomatic when the surrounding tissues become inflamed. MRONJ incidence in patients with cancer receiving intravenous BPs is reported to range from 2% to 15%(4, 5), whereas MRONJ development in osteoporotic patients still remains controversial. The incidences of ONJ from BPs and denosumab were comparable in cancer patients, suggesting that ONJ induced by both drugs adversely affects patients’ quality of life and produces significant morbidity.(4, 5)

Denosumab is a human monoclonal antibody that binds to receptor activator of nuclear factor kappa-Β ligand (RANKL) and prevents RANKL from activating its receptor, RANK, on the surface of osteoclasts and their precursors.(6) Prevention of the RANKL/RANK interaction inhibits osteoclast formation, function, and survival, thereby decreasing bone resorption and increasing bone mass and strength in both cortical and trabecular bone.(7–9) In contrast to denosumab, many potential mechanisms by which BPs inhibit osteoclastic bone resorption have been proposed. BPs, attached to hydroxyapatite on bony surfaces, are impregnated into osteoclasts when they begin to resorb bone and impair their ability to form the ruffled border, to adhere to the bony surface, and to produce protons necessary for bone resorption.(10, 11) Several studies also showed that BPs reduce osteoclast activity by decreasing osteoclast progenitor development and recruitment, and by inducing osteoclast apoptosis via the activation of caspases.(12–14) In addition to their inhibitory effect on osteoclasts, BPs appear to have an effect on osteoblasts. An in vitro study showed that zoledronic acid (ZOL) can increase production of osteoprotegerin (OPG) in a dose-dependent manner from primary human osteoblasts.(15) OPG then binds to RANKL and blocks the interaction with RANK, thereby inhibiting osteoclastogenesis. Another study also demonstrated that ZOL markedly increased OPG protein secretion and reduced transmembrane RANKL protein expression in osteoblast-like cells.(16)

Despite the fact that the first case of MRONJ was described over a decade ago, the mechanism and pathophysiology of MRONJ still remain elusive. Proposed hypotheses that attempt to explain the unique localization of necrotic bone exclusively to the jaw include altered bone remodeling or oversuppression of bone resorption, angiogenesis inhibition, constant microtrauma, suppression of innate or acquired immunity, vitamin D deficiency, soft tissue toxicity, and inflammation or infection.(1) In addition, the management of MRONJ remains as a significant clinical challenge, with little progress having been made on treatment.(17) Patients with MRONJ are treated conservatively with mouth rinses, antibiotics and oral analgesics. Surgical debridement is typically not indicated unless it is absolutely necessary.(1, 18) Even with drug holiday before traumatic dentoalveolar procedures, many patients still experience a lot of pain and other side effects in later stages of MRONJ which tremendously affect their quality of life.(1) Therefore, it is critical to understand the pathophysiology of MRONJ and develop proper regimen for prevention and treatment of this disease.

Although osteoclast suppression has been observed in most MRONJ animal models, there is still no evidence to date whether restoring osteoclast and its bone resorptive function can alleviate MRONJ in animals that have received antiresorptive drugs. We previously engineered RAW264.7 murine myeloid precursor cells with an inducible intracellular RANK (iRANK) construct to allow differentiation into osteoclasts under the control of a chemical inducer of dimerization (CID) drug. This differentiation is independent of RANKL and macrophage colony-stimulating factor (M-CSF), and also resistant to OPG.(19) In this study, we hypothesize that osteoclast formation, differentiation and function are impaired at the extraction site and normal bone healing is disturbed due to the inhibition of osteoclastic activities by antiresorptive drugs. Thus, we further hypothesize that locally-delivered engineered osteoclasts which are resistant to this inhibition may promote bone healing processes after tooth extraction in a MRONJ mouse model. This study determines the effects of restoring osteoclast functions on MRONJ development and also provides a new and innovative strategy to investigate mechanisms leading to different stages of MRONJ.

Materials and Methods

In vitro osteoclast differentiation and mineral resorption

RAW264.7 cells were purchased from American Type Culture Collection (ATCC). RAW264.7 cells containing a CID-inducible, intracellular RANK signaling domain (iRANK cells) were created as previously described.(19) RAW264.7 and iRANK cells were cultured in α-MEM containing 10% fetal bovine serum and 100 U/mL of penicillin/streptomycin and incubated at 37°C with 5% CO₂. RAW264.7 and iRANK cells were plated at 20,000 cells/well in 4-well Nunc^™ Lab-Tek^™ chamber slides. Four hours after plating, 2 nM recombinant mouse RANKL (R&D Systems) was applied to RAW264.7 cells for osteoclast differentiation concurrently with the treatment of 0 nM, 2 nM, 5 nM or 20 nM Ultra-LEAF^™ purified rat anti-mouse CD254 (TRANCE, RANKL) IgG2a, κ monoclonal antibody (BioLegend) or 20 nM Ultra-LEAF^™ purified rat IgG2a, κ isotype control antibody (BioLegend). Similarly, 10 nM AP20187 (CID; Clontech Laboratories) was applied to iRANK cells for osteoclast differentiation concurrently with the treatment of 0 nM, 10 nM, 25 nM or 100 nM of the same rat anti-mouse RANKL monoclonal antibody listed above or 100 nM of the same rat IgG isotype control antibody also listed above. Cells were cultured for 5 days with media and treatments replaced once at day 3. Cells were washed twice with PBS, fixed with 10% neutral buffered formalin for 10 minutes and subjected to TRAP staining following the manufacturer’s instructions (387A, Sigma-Aldrich). Slides were mounted with Aqua-Mount (Thermo Fisher Scientific) and images were obtained using an upright microscope (Nikon E800). The numbers of multinucleated TRAP-positive osteoclasts (with ≥ 3 nuclei) were quantified.

To determine the resorptive function of osteoclasts in the presence of anti-RANKL antibody in vitro, RAW264.7 and iRANK cells were cultured on an Osteo Assay (Corning Inc.) and treated with the inducers; 2 nM recombinant mouse RANKL (R&D Systems) and 10 nM CID (Clontech Laboratories) respectively. Concurrently, cells were treated with either 20 nM rat anti-mouse RANKL monoclonal antibody (BioLegend) or 20 nM rat IgG isotype control antibody (BioLegend). At day 10, all cells on the Osteo Assay plate were removed by 10% bleach and resorption pits were visualized using von Kossa staining. Images were taken by Nikon D5100 camera and the resorption area was quantified using ImageJ software version 1.52a (NIH, Bethesda, MD).

MRONJ model in nude mice

Animals and surgical procedures were handled in accordance with the University of Washington’s Institutional Animal Care and Use Committee. A MRONJ model in athymic nude mice was utilized to allow delivery of allogeneic iRANK cells. Briefly, eight- to ten-week-old female nude mice were randomly divided into two groups (10 mice per group); IgG isotype control (Ct) and anti-RANKL antibody (Ab); same antibodies as in the in vitro experiments were used. Mice in Ct and Ab group were injected with 10 mg/kg rat IgG isotype control and 10 mg/kg of rat anti-mouse RANKL monoclonal antibody intraperitoneally respectively three times a week until termination. One week after the first antibody injection, left maxillary first molar was extracted in both groups. Under general anesthesia using ketamine/xylazine, a rodent molar luxator (iM3) was used to separate the gingival attachment and a 25-gauge needle was used to gently luxate the tooth. Then, a curved hemostat was used to pull the tooth out. Approximately three weeks after tooth extraction, mice were sacrificed to harvest serum and maxilla.

Cell Delivery and CID treatment

Eight- to ten-week-old nude mice were randomly divided into two groups (10 mice per group); anti-RANKL antibody with iRANK cells (Abc), and anti-RANKL antibody with iRANK cells and CID (Abcc); same antibodies used throughout study. Mice in both groups were injected with rat anti-mouse RANKL monoclonal antibody and tooth extraction was performed as described above. Immediately after tooth extraction, 100,000 iRANK cells in 0.5 mg/mL neutralized collagen were delivered to the tooth socket and covered with Gelfoam® (Pfizer Pharmaceutical) in both Abc and Abcc groups. Only the mice in Abcc group were injected with 10 mg/kg CID intraperitoneally for 3 consecutive days starting on the same day as cell delivery, then twice a week throughout the experiment to induce osteoclast differentiation. Mice were sacrificed to harvest serum and maxilla 3 weeks after tooth extraction. The tissue specimens from all groups were immediately fixed in 4% paraformaldehyde in PBS, pH 7.4, at 4°C for 48 hours and stored in 70% EtOH solution. Tissues were decalcified with 5% EDTA and 4% sucrose in PBS, pH 7.4 for 4 to 5 weeks at 4°C. Decalcification solution was changed daily. Tissue samples were then processed for paraffin embedding and sectioned for immunohistochemistry. Additionally, terminal blood was collected to measure the serum concentration of TRAP-5b and osteocalcin (OCN), biochemical markers for bone resorption and formation, respectively, using the Mouse TRAP-5b (Immunodiagnostic Systems) and the Mouse Osteocalcin ELISA Kit (Immunotopics).

H&E and TRAP staining

4-μm-thickness tissue sections in sagittal plane were used for immunohistochemistry. For hematoxylin and eosin (H&E) staining, tissue sections were deparaffinized, rehydrated and placed in Harris hematoxylin solution for 3 minutes before rinsing. This was followed by 40 seconds in ammonium water (0.25% v/v). Finally, slides were dipped in eosin solution 10 times before dehydrating. The slides were mounted with Permount (Thermo Fisher Scientific). For TRAP staining, tissue sections were deparaffinized, rehydrated and incubated for 30 minutes with TRAP staining solution at 37°C, according to manufacturer’s protocol (387A, Sigma-Aldrich). Nuclei were counterstained with 300 nM DAPI Dilactate (Life Technologies) for 5 minutes. Slides were mounted with Aqua-Mount (Thermo Fisher Scientific) and images were obtained using an upright microscope (Nikon E800).

Immunofluorescence staining

Engineered osteoclasts express green fluorescent protein (GFP), so we could visualize and distinguish them from endogenous osteoclasts. To detect GFP expression, tissue sections were deparaffinized and rehydrated with TBS-Tween (TBS-T; 0.05 M Tris buffer, 0.15 M NaCl, 0.1% Tween 20, pH 7.6) before permeabilization with 0.1% Triton X-100 in PBS, pH 7.4. for 10 minutes. Blocking was performed with 4% normal donkey serum (NDS) for 30 minutes. Rabbit anti-GFP polyclonal antibody (Thermo Fisher Scientific) was diluted to 10 μg/mL in 2% NDS and incubated at room temperature for 1 hour. Slides were rinsed 3 times in TBS-T for 5 minutes. Cy3 AffiniPure Donkey Anti-Rabbit IgG secondary antibody (Jackson ImmunoResearch Laboratories, Inc.) was used to decrease autofluorescence in the green channel in order to accurately detect GFP expression. The secondary was diluted to 7.5 μg/mL in 2% NDS and incubated for 30 minutes at room temperature. In addition, tissue sections were used for double-staining of additional osteoclast markers including MMP-9, carbonic anhydrase II, and cathepsin K. For double-staining images in Figure 4E, slides were incubated in primary antibodies: Goat Anti-Mouse MMP-9 at 5 μg/mL (Novus Biologicals) and Rabbit Anti-Carbonic Anhydrase II at 8 μg/mL (Abcam) or Goat Anti-Mouse MMP-9 at 5 μg/mL and Rabbit Anti-Cathepsin K at 10 μg/mL (Abcam) for 1 hour. Then, slides were incubated with appropriate secondary antibodies: Cy3 Donkey Anti-Goat at 7.5 μg/mL (Jackson ImmunoResearch) and Alexa Fluor 488 Donkey Anti-Rabbit at 7.5 μg/mL (Jackson ImmunoResearch) for 30 minutes at room temperature. Slides were then rinsed three times in TBS-T for 5 minutes and once in Milli-Q water. Nuclei were counterstained with 300 nM DAPI Dilactate (Life Technologies) for 5 minutes. Slides were rinsed with Milli-Q water and PBS before mounting with ProLong Gold Antifade Mountant (Thermo Fisher Scientific). Images were obtained using an upright microscope (Nikon E800) and overlaid using Adobe Photoshop to show double-staining of osteoclast markers in the same cell.

Fig. 4. iRANK cells were retained and differentiated into osteoclasts in anti-RANKL antibody and CID-treated mice.

(A) Experiment timeline (B) TRAP and immunofluorescence staining; GFP+ cells (arrowheads) indicating delivered iRANK cells were found in the sockets of both Abc and Abcc groups. However, only Abcc group exhibited osteoclasts (arrows) which are TRAP+GFP+ with three or more nuclei. (Dot line indicates outline of alveolar bone; M2, second molar; scale bar = 100μm) (C) Quantification of TRAP+GFP+ osteoclasts (***p < 0.001; n ≥ 7 per group) (D) High magnification of iRANK osteoclasts; In Abcc group, engineered iRANK osteoclasts are positive for TRAP as indicated by red granules (black arrow). Un-fused iRANK cells are TRAP-negative as depicted by the absence of red staining (black arrowhead). Engineered iRANK cells also express GFP showing that these cells are the implanted iRANK cells which have differentiated into osteoclasts indicated by double-labeling of the same cell with TRAP enzyme and GFP antibody (black and white arrows). Osteoclast at center has 13 nuclei. Undifferentiated implanted iRANK cells also express GFP as expected (white arrowhead). 60X oil immersion images; scale bar = 50μm (E) Immunofluorescence of iRANK osteoclasts in the extraction site of adjacent sections double-labeled with antibodies to either MMP-9 and carbonic anhydrase II, or MMP-9 and cathepsin K. (MMP-9 = matrix metalloprotease 9; CAII = carbonic anhydrase II; CatK = cathepsin K. 60X oil immersion images; scale bar = 50μm)

Quantitation of epithelial closure, necrotic and new bone, and osteoclast number

Several characteristics were used to evaluate the degree of ONJ-like lesions. After termination, the images of epithelial closure at the extraction site were taken before all the tissues were decalcified and processed for histochemistry. To Investigate necrotic bone area, newly formed bone, and number of osteoclasts at the extraction site, all the quantifications were done using ImageJ software version 1.52a (NIH, Bethesda, MD). To quantify necrotic bone area, four uniformly spaced (approximately 50 μm) H&E-stained sections were selected per sample. The area of interest covered the alveolar bone adjacent to the tooth sockets. The percentage of necrotic area was calculated by the area containing 5 or more empty osteocytic lacunae divided by the total alveolar bone area. To quantify newly formed bone, four uniformly spaced (approximately 50 μm) H&E-stained sections were selected per sample. The newly formed bone was quantified only in the sockets without root fragments as the percentage of bone area in the socket/total socket area (%BA/TA). The number of osteoclasts was also quantified by using two uniformly spaced (approximately 100 μm) TRAP-stained sections per sample for Ct and Ab groups, and TRAP & GFP-stained sections for Abc and Abcc groups. Endogenous and engineered osteoclasts were defined as TRAP+ and TRAP+GFP+ cells with three or more nuclei respectively. Eight samples were excluded from the histological analysis due to tissue damage at the extraction site during embedding and sectioning process, except one sample from Abcc group which was because the CID system did not work, and osteoclast differentiation could not be induced.

Blood chemistry

Serum samples collected from each mouse were stored in −20°C and thawed once. The presence of TRAP-5b was detected by sandwich ELISA using the Mouse TRAP-5b ELISA Kit (Immunodiagnostic Systems). Serum levels of OCN were detected by sandwich ELISA using the Mouse Osteocalcin ELISA Kit (Immunotopics). The serum samples were assayed in triplicate and the procedure was performed according to the manufacturer’s protocol.

Statistical analysis

Results are expressed as mean ± SD. GraphPad Prism version 8.4.1 was used to perform Kolmogorov-Smirnov test to test normality of data, student’s t-test to compare means of two individual groups with equal variances, student’s t-test with Welch’s correction to compare means of two individual groups with unequal variances, one-way ANOVA with post-hoc Tukey test to compare means of three or more individual groups, and Fisher exact’s test to compare epithelial closure among groups. A value of p < 0.05 was considered statistically significant.

Results

iRANK engineered osteoclast differentiation and resorptive function were resistant to inhibition by anti-RANKL antibody

RAW264.7 cells containing a CID-inducible, intracellular RANK signaling domain (iRANK cells) were created as previously described.(19) The main feature of CID technology is that induction of signaling is triggered only by the presence of the small molecule CID and is not affected by the presence of endogenous RANK/RANKL inhibitors, such as OPG. We previously showed that iRANK cell differentiation to osteoclasts in the presence of CID was OPG-resistant(19), therefore, we hypothesized that this differentiation would not be inhibited by anti-RANKL antibody which binds to RANKL in a similar manner as OPG. As expected, iRANK cells formed multinucleated TRAP-positive osteoclasts following CID treatment even in the presence of the highest concentration of anti-RANKL antibody (100 nM). On the other hand, RANKL-mediated osteoclastogenesis of RAW264.7 cells was almost completely inhibited by anti-RANKL antibody even at the lowest concentration (2 nM) (Figure 1A–1D).

Fig. 1. CID-induced osteoclastogenesis in iRANK cells was anti-RANKL antibody-independent.

(A), (B) TRAP staining of osteoclasts differentiated from RAW264.7 cells and iRANK cells treated with different concentration of anti-RANKL antibody (C), (D) Quantification of osteoclasts (n=3 per group; ****p < 0.0001; scale bar = 200 μM)

Next, we determined the effect of anti-RANKL antibody on engineered osteoclast function in vitro by differentiating RAW264.7 and iRANK cells into osteoclasts directly on a 2D-mineralized well-plate using RANKL and CID respectively. Concurrently, both cell types were treated with 20 nM anti-RANKL antibody or control IgG antibody for 10 days. Visualized with von Kossa staining, there was no resorption in RANKL-treated RAW264.7 cells in the presence of anti-RANKL antibody. On the contrary, the resorption area in CID-treated iRANK cells in the presence of anti-RANKL antibody was comparable to the group without treatment (Figure 2A and 2B). Interestingly, the resorption area in the IgG-treated iRANK group was significantly higher than the other iRANK groups (Figure 2B) as IgG can interact with the Fc gamma receptors (FcγRs) on hematopoietic cells and facilitate osteoclastogenesis as well as phagocytosis/resorption.(20, 21) These in vitro data suggest that osteoclast differentiation of iRANK cells via CID system and their resorptive function are not inhibited by anti-RANKL antibody.

Fig. 2. CID-induced osteoclasts resorbed 2D mineral substrate in the presence of anti-RANKL antibody.

(A) von Kossa staining of 2D mineral substrate (Osteo Assay) surfaces which osteoclasts differentiated from RAW264.7 cells and iRANK cells were cultured on (B) Quantification of resorption area (n=3 per group; *p < 0.05, **p < 0.01, and ***p < 0.001)

Absence of osteoclasts and accumulation of necrotic bone were observed following tooth extraction in anti-RANKL antibody-treated nude mice

As the osteoclast differentiation of iRANK cells and their resorptive function were resistant to anti-RANKL antibody in vitro, our next step was to develop the MRONJ animal model and locally deliver these cells hypothesizing that they would improve the MRONJ lesions.

Since iRANK cells are allogeneic, nude mice were required to develop the MRONJ model in order to avoid unfavorable host responses following cell delivery. Prior to establishing the MRONJ model, we first demonstrated that nude mice had normal wound healing compared to wild-type mice following tooth extraction (Supplementary Figure 1). Clinical presentation of maxilla at 3 weeks after tooth extraction showed that 100% of wild type and nude mice had complete closure of epithelium at the extraction site. Furthermore, connective tissue was observed directly above the bone-filled sockets in all mice, indicating normal healing process at the extraction site.

To establish the MRONJ model, nude mice were randomly divided into two groups (10 mice per group); Control IgG (Ct) and anti-RANKL antibody (Ab) as shown in Figure 3A. Ct group received isotype control rat IgG, while Ab group received anti-RANKL antibody. Several features were studied at approximately three weeks following tooth extraction. In Ct group, we observed osteoclasts, identified as TRAP-positive cells with three or more nuclei, within the extraction sites (Figure 3B and 3E). Osteoclasts were found within and around the tooth sockets demonstrating typical resorptive morphology near damaged bones. On the other hand, there were no osteoclasts detected in the tooth sockets of the Ab group (Figure 3B and 3E). These data indicated that the injections of anti-RANKL antibody inhibited osteoclast formation in nude mice.

Fig. 3. Absence of osteoclasts and presence of necrotic bone were observed following tooth extraction in anti-RANKL antibody-treated mice.

(A) Experiment timeline (B) TRAP staining; Osteoclasts, defined as TRAP+ cells with three or more nuclei, were observed in Ct group, but not in Ab group. (C) H&E staining; Clusters of empty lacunae locating at unresorbed alveolar ridge (red arrow) were observed with no sign of osteoclast in Ab group. On the contrary, osteoclasts (black arrow) were observed in Ct group. (Black dot line indicates the tooth socket outline. Red dot line indicates the level of newly formed woven bone. Black arrowheads indicate reversal line.) (D) Clinical presentation at the extraction site (Black arrowheads indicate the outline of the extraction site.) (E) Quantification of osteoclasts, (F) % necrotic bone area, (G) % bone area/total socket area (%BA/TA), and (H) epithelial closure (M2, second molar; M3, third molar; LB, lamellar bone; WB, woven bone; scale bar = 100 μm; ***p < 0.001, ****p < 0.0001; n ≥ 7 per group for osteoclast and % necrotic bone area; n ≥ 4 per group for %BA/TA)

Although tooth extraction has widely been used to induce ONJ-like lesions in rodents, root fracture is one of the most common complications following tooth extraction especially in mice with age of 8 weeks and older due to the accumulation of cementum at the root apices.(22) We excluded the samples with retained root fragments from the analysis of new bone formation since the bone formation process was complicated in those sockets.(data not shown and 23, 24) However, there has been no evidence showing that retained root fragments are associated with accumulation of necrotic bone around tooth sockets following tooth extraction in mice. To confirm this, we determined that no correlation existed between the retained root fragments and necrotic bone area in all groups (Supplementary Figure 2). Thus, we included all animals in the analysis of necrotic bone area.

As expected, Ct group showed significantly less % necrotic bone area at the extraction site compared to Ab group. H&E-stained sections displayed an osteoclast resorbing an alveolar bone next to the socket in Ct group, while the alveolar bone containing clusters of empty lacunae remained unresorbed in Ab group (Figure 3C and 3F). On the other hand, % newly formed bone area in the tooth socket (%BA/TA) showed no significant difference between Ct and Ab groups (Figure 3C and 3G). In addition, all of the samples in Ct and Ab groups, except one sample from Ab group which had a granulation tissue mass above the socket, exhibited normal soft tissue healing and complete epithelial closure at the extraction site (Figure 3D and 3H).

Lastly, we measured serum TRAP-5b and osteocalcin levels which are markers of bone resorption and bone formation, respectively. We found that Ab group showed significantly less serum TRAP-5b and osteocalcin levels compared to Ct group (Supplementary Figure 3). According to all these findings, our MRONJ model developed in nude mice exhibited absence of osteoclasts and accumulation of necrotic bone following tooth extraction.

iRANK cells were retained and differentiated into osteoclasts in anti-RANKL antibody and CID-treated mice

To determine whether iRANK cells were retained following local injection, in vivo bioluminescence imaging was performed (Supplementary Figure 4). Bioluminescence was detected in all animals that received cells with or without Gelfoam®. As expected, no signal was detected in the animals that did not receive cells but received luciferin alone. Luciferase transfected iRANK cells were retained at the tooth socket up to 10 days following delivery.

Next, we locally delivered iRANK cells to the extraction sites of mice treated with anti-RANKL antibody to determine whether CID treatment would induce iRANK osteoclast differentiation and improve the healing of MRONJ-like lesions. Nude mice were randomly divided into two groups (10 mice per group); anti-RANKL antibody with iRANK cells (Abc), and anti-RANKL antibody with iRANK cells and CID (Abcc) as shown in Figure 4A. Anti-RANKL antibody injections and tooth extraction were performed in all mice as described above. In this experiment, iRANK cells were delivered to the tooth sockets immediately after tooth extraction in both groups. However, only the mice in Abcc group were injected with CID drug intraperitoneally to induce osteoclast differentiation.

As shown in Figure 4B and 4C, GFP-positive iRANK cells were observed in both Abc and Abcc groups. However, osteoclasts were only observed in Abcc group. These osteoclasts were not only TRAP-positive, but also GFP-positive, indicating that they differentiated from the delivered iRANK cells under the control of CID (Figure 4D). In addition to TRAP, the engineered osteoclasts expressed other osteoclast markers including matrix metallopeptidase 9 (MMP-9), carbonic anhydrase II (CAII), and cathepsin K (CatK) (Figure 4E). Taking both in vitro and in vivo findings together, we demonstrated that iRANK cells delivered to the extraction sites were retained and successfully differentiated into osteoclasts via CID system despite the treatment of anti-RANKL antibody in nude mice.

Necrotic bone resorption, but not new bone formation or epithelial closure, was restored by the iRANK engineered osteoclasts following tooth extraction in anti-RANKL antibody and CID-treated mice

Next, we investigated how local replenishment of osteoclasts affected necrotic bone area, a main feature of MRONJ. Histologically, we observed engineered osteoclasts resorbing alveolar bone with empty lacunae in the Abcc group, while the alveolar bone containing clusters of empty lacunae remained unresorbed in the Abc group (Figure 5A). Furthermore, quantitative data indicated a statistically significant decrease in the necrotic bone area in Abcc group compared to Abc group, (p = 0.045) (Figure 5C). Thus, these data suggest that the osteoclasts’ resorptive function could be restored by delivery of engineered osteoclasts.

Fig. 5. Bone resorptive function, but not new bone formation and epithelial closure, was restored by the engineered osteoclasts in anti-RANKL antibody-treated mice.

(A) H&E staining; Left panels: low magnification images of the extraction site. Black dotted line indicates the tooth socket outline. Red dotted line indicates the level of newly formed woven bone. The boxed areas represent alveolar bone adjacent to the tooth sockets. Scale bar = 100 μm. Right panels: Higher magnification images of the boxed areas in left panels. Large clusters of empty lacunae located at unresorbed alveolar ridge (red arrow) were observed with no sign of osteoclasts in Abc group. In contrast, iRANK engineered osteoclasts (black arrows) located next to smaller clusters of empty lacunae were observed in Abcc group. Black dotted line indicates the necrotic bone area. Black arrowheads indicate reversal line. LB, lamellar bone; WB, woven bone. Scale bar = 100 μm. (B) Clinical presentation at the extraction site. Black arrowheads indicate the outline of the extraction site. (C) Quantification of necrotic bone area, (D) % bone area/total socket area (%BA/TA), and (E) epithelial closure (M2, second molar; M3, third molar; *p < 0.05; n ≥ 7 per group for % necrotic bone area; n ≥ 3 per group for %BA/TA)”

In addition to necrotic bone area, we measured whether new bone formation and epithelial healing were altered in mice receiving anti-RANKL antibody and iRANK engineered osteoclasts. Interestingly, there was no significant difference between Abc and Abcc groups in %BA/TA (Figure 5A and 5D) or epithelial closure (Figure 5B and 5E). However, in contrast to the mice that did not receive iRANK cells, the majority of the samples from Abc and Abcc groups exhibited partial or incomplete epithelial closure at the extraction site (Figure 5B and 5E) as well as less newly formed bone in the sockets.

Finally, serum TRAP-5b and osteocalcin levels were reduced equally in Abc and Abcc groups, compared to Ct group, indicating that addition of engineered cells did not alter peripheral bone remodeling in general (Supplementary Figure 3).

Discussion

Deficiency of osteoclastic bone resorption has been hypothesized as a main mechanism leading to MRONJ. In this study, we developed a tooth extraction-triggered MRONJ model in nude mice using anti-RANKL antibody to deplete osteoclasts. Replenishment of osteoclasts within sockets was achieved using engineered, inducible myeloid precursor cells. Lesions treated with iRANK cells + CID showed significantly decreased accumulation of necrotic bone in the sockets compared to those treated with iRANK cells alone. These data support the hypothesis that osteoclast deficiency leads to accumulation of necrotic bone, a major feature of MRONJ.

Throughout the past decade, many studies have shown that osteoclast inhibition is one of the potential mechanisms underlying MRONJ pathophysiology. In the normal healing processes following tooth extraction, osteoclasts play a role in resorption of damaged or necrotic bone caused by traumatic force in early stages, as well as, bone formation and remodeling in later stages.(25, 26) However, no study has shown whether replenishment of osteoclasts can restore these functions in MRONJ models. In this study, we are the first group to demonstrate that the bone resorptive function could be restored by delivered iRANK osteoclasts as decreased necrotic bone area was observed in Abcc group compared to Abc group.

Interestingly, the absence of osteoclasts caused by anti-RANKL antibody did not impair new bone formation (Ct vs. Ab group) and the local repletion of osteoclasts did not affect this process either (Abc vs. Abcc group). These findings suggest that new bone formation may not only rely on the osteoclast-osteoblast interaction. Woven bone formation occurs in two ways: appositional formation from the existing bone surfaces primarily mediated by the coupling mechanism of osteoclast and osteoblasts, and de novo formation which is a new bone formation by osteoblasts without osteoclasts or any existing bone surfaces.(27) It is most likely that the appositional formation was inhibited in our model, leaving the de novo formation as the only source of new bone formation in the Ab group. These findings also raise another crucial point that osteoclast suppression alone may not be enough to induce severe-stage ONJ-like lesions. Other factors or mechanisms should also be considered toward the pathophysiology and different stages of MRONJ. Several hypotheses apart from osteoclast suppression, for instance, unbalanced M1 and M2 macrophages(28, 29) and angiogenesis inhibition(30–32), have been studied in BP-treated animal models. Hence, further studies of those possibilities need to be investigated in anti-RANKL antibody-treated animal models in order to clearly understand the pathophysiology of MRONJ.

The clinical manifestations at the extraction site of nude mice treated with anti-RANKL antibody (Ab group) are most similar to stage 0 MRONJ in humans, which presents some histological changes of alveolar bone at the extraction site with no evidence of exposed necrotic bone.(1) Several studies also observed stage 0 ONJ-like lesions in rodents treated with antiresorptive drugs.(4, 33) These lesions can be advanced to more severe stages by the induction of chemotherapy drugs concurrently to antiresorptive drugs.(34, 35) Also, we hypothesize that the mild stage of our model may be due to the extraction of healthy teeth. The major causes of tooth extraction in patients are caries, periodontal and pulpal disease(36), and MRONJ is related to tooth extraction in patients with these diseases.(37) However, we extracted the healthy maxillary first molars in this experiment which may result in less severity of ONJ-like lesions. Even though exposed necrotic bone area as well as other MRONJ characteristics including impaired bone formation and incomplete epithelial closure seen in later stages (stage 1–3) of MRONJ were not observed in Ab group, our model exhibited absence of osteoclasts and accumulation of necrotic bone at the extraction site histologically. These features were adequate for the cell delivery study to investigate whether bone resorptive function could be restored by the iRANK engineered osteoclasts. This is the first loss and gain of function study providing insights on specific osteoclast activities which can potentially contribute to future studies on pathophysiological mechanisms leading to different stages of MRONJ.

Nude mice which lack thymus were utilized in this study. They were necessary to avoid an immune rejection of the delivered cells since they lack normal T cells originating from the thymus. Several studies have shown that alteration of immune cells such as regulatory T cells (Tregs), T helper 17 cells (Th17), and dendritic cells may involve MRONJ pathophysiology(38, 39), so it is possible that utilization of nude mice may have contributed to the mild-stage ONJ-like lesions in this study. However, nude mice still have intact innate immunity (such as natural killer (NK) cells), precursors of antibody forming cells, and T-cell precursors in their bone marrow which can differentiate to γδ T cells without antigen activation.(40–42) Therefore, the majority of immune cells besides thymus-dependent T-cells remain functional in nude mice. Indeed, Park et al. showed that the exposed bone area and open wound at the ONJ-like lesions were smaller in γδ T cell null (Tcrd^−/−) mice treated with zoledronic acid compared to treated wild-type mice.(43) This suggests that at least a few features related to MRONJ could potentially be induced in nude mice consisting of γδ T cells. In addition, we also showed that nude mice had normal wound healing compared to wild-type mice following tooth extraction. According to these data, utilization of nude mice should not hinder the main objective of this study which focuses on the effects of delivered iRANK osteoclasts in the mouse model with the inhibition of endogenous osteoclasts, a common feature observed in all stages of MRONJ.

A limitation of this study is the use of a myeloid precursor cell line, rather than primary cells to create iRANK engineered osteoclasts. We delivered these cells directly to the tooth sockets in neutralized collagen which has been successfully used as a cell carrier to the periodontium without interfering healing processes.(44) However, primary osteoclasts could not be used in this model since they would be inhibited by anti-RANKL antibody treatment. In addition, we have been unable to transduce primary murine bone marrow precursors with iRANK constructs using either adenoviral or lentiviral vectors, most likely due to the size of the construct (unpublished data). In addition, undifferentiated myeloid precursor cells proliferate in vivo and therefore limited the duration of our experiments. These proliferated cells may potentially affect soft tissue healing at the delivery site resulting in the incomplete epithelial closure observed only in the cell-delivery groups (Abc and Abcc), but not in Ab group. Therefore, modifications such as delivery of purified iRANK osteoclasts should be considered for utilizing iRANK cells as a study tool and a potential therapeutic strategy for MRONJ in future studies.

In conclusion, we performed tooth extraction in nude mice treated with anti-RANKL antibody and developed iRANK cell and CID delivery methods to induce osteoclast formation in vivo. Our model captured the absence of osteoclasts and the clusters of empty lacunae representing necrotic bone area similarly to stage 0 MRONJ in human. However, new bone formation and epithelial closure were not affected by the complete inhibition of osteoclasts. We successfully used this model to demonstrate the roles of osteoclasts in healing following tooth extraction and potentially in MRONJ development. Our model suggests that the absence of osteoclasts affects the process of necrotic bone resorption around the tooth sockets after tooth extraction which may contribute to MRONJ. This bone resorptive function can be restored by delivering the engineered osteoclasts to the extraction site. We are the first group to conduct a loss and gain of function study of osteoclasts in vivo, hence, we believe that the optimized engineered osteoclasts along with robust and reliable animal models would be beneficial tools to further study the pathophysiology and mechanisms leading to different stages of MRONJ.

Data and materials availability:

All data associated with this study are available in the main text or the supplementary materials.