Osteonecrosis of the Jaw Developed in Mice

Background: The pathological mechanism of osteonecrosis of the jaw (ONJ) is unknown. Results: Mouse ONJ-like lesions exhibited epithelial hyperplasia associated with γδ T cells of mouse or human origin. Conclusion: γδ T cells may modify the oral disease phenotypes of ONJ. Significance: ONJ pathogenesis may involve multiple mechanisms separately leading to the development of osteonecrosis or oral epithelial abnormality. Osteonecrosis of the jaw (ONJ), an uncommon co-morbidity in patients treated with bisphosphonates (BP), occurs in the segment of jawbone interfacing oral mucosa. This study aimed to investigate a role of oral mucosal barrier γδ T cells in the pathogenesis of ONJ. Female C57Bl/6J (B6) mice received a bolus zoledronate intravenous injection (ZOL, 540 μg/kg), and their maxillary left first molars were extracted 1 week later. ZOL-treated mice (WT ZOL) delayed oral wound healing with patent open wounds 4 weeks after tooth extraction with characteristic oral epithelial hyperplasia. γδ T cells appeared within the tooth extraction site and hyperplastic epithelium in WT ZOL mice. In ZOL-treated γδ T cell null (Tcrd−/− ZOL) mice, the tooth extraction open wound progressively closed; however, histological ONJ-like lesions were identified in 75 and 60% of WT ZOL and Tcrd−/− ZOL mice, respectively. Although the bone exposure phenotype of ONJ was predominantly observed in WT ZOL mice, Tcrd−/− ZOL mice developed the pustule/fistula disease phenotype. We further addressed the role of γδ T cells from human peripheral blood (h-γδ T cells). When co-cultured with ZOL-pretreated human osteoclasts in vitro, h-γδ T cells exhibited rapid expansion and robust IFN-γ secretion. When h-γδ T cells were injected into ZOL-treated immunodeficient (Rag2−/− ZOL) mice, the oral epithelial hyperplasia developed. However, Rag2−/− ZOL mice did not develop osteonecrosis. The results indicate that γδ T cells are unlikely to influence the core osteonecrosis mechanism; however, they may serve as a critical modifier contributing to the different oral mucosal disease variations of ONJ.


Osteonecrosis of the jaw (ONJ), an uncommon co-morbidity in patients treated with bisphosphonates (BP), occurs in the segment of jawbone interfacing oral mucosa. This study aimed to investigate a role of oral mucosal barrier ␥␦ T cells in the pathogenesis of ONJ. Female C57Bl/6J (B6) mice received a bolus zoledronate intravenous injection (ZOL, 540 g/kg), and their maxillary left first molars were extracted 1 week later. ZOLtreated mice (WT ZOL) delayed oral wound healing with patent open wounds 4 weeks after tooth extraction with characteristic oral epithelial hyperplasia. ␥␦ T cells appeared within the tooth extraction site and hyperplastic epithelium in WT ZOL mice.
In ZOL-treated ␥␦ T cell null (Tcrd ؊/؊ ZOL) mice, the tooth extraction open wound progressively closed; however, histological ONJ-like lesions were identified in 75 and 60% of WT ZOL and Tcrd ؊/؊ ZOL mice, respectively. Although the bone exposure phenotype of ONJ was predominantly observed in WT ZOL mice, Tcrd ؊/؊ ZOL mice developed the pustule/fistula disease phenotype. We further addressed the role of ␥␦ T

cells from human peripheral blood (h-␥␦ T cells). When co-cultured with ZOL-pretreated human osteoclasts in vitro, h-␥␦ T cells exhibited rapid expansion and robust IFN-␥ secretion. When h-␥␦ T cells were injected into ZOL-treated immunodeficient (Rag2
؊/؊ ZOL) mice, the oral epithelial hyperplasia developed. However, Rag2 ؊/؊ ZOL mice did not develop osteonecrosis. The results indicate that ␥␦ T cells are unlikely to influence the core osteonecrosis mechanism; however, they may serve as a critical modifier contributing to the different oral mucosal disease variations of ONJ.
Osteonecrosis of the jaw (ONJ) 2 in the maxilla and mandible has emerged as an uncommon and occasionally severe co-morbidity among patients treated with a group of anti-resorptive agents such as amino-bisphosphonates (BP) (1) and denosumab, a humanized anti-RANKL monoclonal antibody (2). By contrast, ONJ has not been reported in patients treated with hormone replacement therapy or selective estrogen receptor modulators. ONJ is initially defined as an exposed jawbone in the oral cavity for more than 8 weeks even after appropriate intervention therapies in patients with current or past BP treatment (3,4). An increasing number of case reports suggest that considerable variations exist in the clinical manifestations of ONJ, ranging from radiographic bone pathology without ulcerative oral mucosal lesions to localized swelling with persistent fistula formation (5)(6)(7).
ONJ is most frequently reported in zoledronate (ZOL)treated patients with a primary diagnosis of multiple myeloma or bone metastatic tumors such as breast cancer (8,9). These high risk patients are generally middle-aged to elderly, and they are often treated with immunosuppressive and chemotherapeutic agents, glucocorticoids, or both (10,11). Therefore, other studies have addressed the combinatory effect of BP with multiple medications on the development of ONJ-like lesions in mice (Table 1). Although additional systemic and poly-pharmaceutical conditions might act as important cofactors that increase the prevalence of ONJ and contribute to its clinical variations (12,13), the complex conditions have made it difficult to determine its core pathological mechanism.
ONJ primarily occurs in the oral segment of the jawbone that is tightly covered by keratinized oral mucosa. The oral mucosa, i.e. gingiva or palatal tissue, is composed of stratified epithelium and thin connective tissue; furthermore, it is considered to be one of the most protective tissue barriers against physical and chemical damage, bacterial infection, and environmental stress (14). Case control studies of patients with ONJ have indicated an increased risk of developing this condition with tooth extraction or the use of ill-fitting removable dental prostheses (15,16). These "event-related" oral conditions among BPtreated patients can lead to inflammation in the oral mucosa tissue that likely activates oral barrier immunity. Thus, we hypothesized that the close proximity of the jawbone to the oral mucosa enables the involvement of abnormally stimulated oral barrier immunity during ONJ pathogenesis.
T cells expressing canonical ␥␦ T cell receptors represent a small subset of circulating immune cells and account for 2-5% of peripheral blood T cells in humans. A deficiency in circulating ␥␦ T cells has been reported in patients with long term and repeated BP administrations (17,18), and BP-induced ␥␦ T cell deficiency was postulated to promote an underlying susceptibility to the development of ONJ (17). Because ␥␦ T cells are preferentially involved in barrier immunity (19,20), we hypothesized that the ␥␦ T cells in the oral barrier tissue play an important role in the development of ONJ. This study developed a mouse model exhibiting ONJ-like lesions. The role of ␥␦ T cells was addressed in the ␥␦ T cell-deficient Tcrd Ϫ/Ϫ mice and human ␥␦ T cell-injected Rag2 Ϫ/Ϫ mice. This study suggests that ␥␦ T cells may not be centrally involved in its core osteonecrosis mechanism; however, their presence or absence in oral barrier immunity appears to contribute to the epithelial disease variations of ONJ.

Materials and Methods
Ethics Statement-The UCLA Animal Research Committee reviewed and approved all experimental protocols involving animals (ARC 1997-136). The UCLA Institutional Review Board reviewed and approved all protocols involving human subjects (IRB 12-001176).
ZOL Injection to Mice-This study sought to establish a mouse model exhibiting ONJ-like lesions. Seven-week-old female C57Bl/6J mice (the Jackson Laboratory, Bar Harbor, ME) were subjected to a bolus injection of 540 g/kg ZOL (Reclast, Novartis, East Hanover, NJ) through the retro-orbital venous plexus (WT ZOL) (21). In the control group, 0.9% NaCl vehicle solution was injected through the retro-orbital venous plexus (WT NaCl). The ZOL dose, which is pharmacologically relevant to oncological doses in humans, was estimated via allometric scaling (22). Each mouse was administered only a single ZOL injection. In the initial study, mouse renal toxicity was evaluated using a creatinine clearance test for serum or urine samples obtained 1 day after ZOL (n ϭ 6) or NaCl (n ϭ 6) injection.
Maxillary First Molar Extraction-One week after the ZOL or NaCl injection, the maxillary left first molar was extracted (23). Mice were anesthetized via isoflurane inhalation and placed on a custom-made surgical table in a supine position using the fixed positioner on the maxillary incisors. A nasal tube was used for the continuous inhalation of 2-4% isoflurane mixed with oxygen during the surgical manipulations in the oral cavity. After the suprabony circumferential periodontal ligament of the attached gingiva was dissected with a dental explorer, the maxillary left first molar was laterally luxated by inserting the tip of a dental explorer between the first and second molars. The luxated molar was then gently removed using surgical forceps. Surgical complications such as tooth fracture occurred and appeared to cause confounding problems. As such, those mice were eliminated from further evaluation. Immediately prior to tooth extraction, 5.0 mg/kg carprofen was subcutaneously injected, and this injection was repeated every 24 h for 48 h.
Maxillary Tissue, Femur, and Whole Blood Collection-Euthanasia by 100% CO 2 inhalation was performed on day 4 (WT NaCl, n ϭ 6; WT ZOL, n ϭ 7), week 1 (WT NaCl, n ϭ 8; WT ZOL, n ϭ 9), week 2 (WT NaCl, n ϭ 11; WT ZOL, n ϭ 11), or week 4 (WT NaCl, n ϭ 8; WT ZOL, n ϭ 12) after tooth extraction. The maxilla containing the tooth extraction wound and femur were harvested. The maxillary tissue was subjected to standardized digital photo recording. The clinical photograph was enlarged and examined for tooth extraction wound healing. The harvested maxillary tissue and femurs were fixed in 10% buffered formalin and used for imaging by micro-computed tomography (micro-CT: CT40, Scanco Medical, Bassersdorf, Switzerland) at an x-ray energy level of 55 peak kV with an intensity of 145 A. The voxel size was 20 m with a slice increment of 20 m. The fixed maxillary tissues were further treated with a formic acid-based decalcifying solution (Immunocal, Ummunotec, Swanton, VT) or 10% EDTA for 7 days for histological section preparation as described below.
Separately, whole blood samples were obtained at the time of euthanasia via cardiac puncture using a 23-gauge needle. Serum chemistry was determined for alkaline phosphatase, calcium, and phosphorus (24).
Characterization of ␥␦ T Cells in Mouse Oral Mucosal Tissue-To evaluate ␥␦ T cells in the oral mucosa barrier tissue, a cell dissociation study was performed. Two weeks after molar extraction, the entire gingival/palatal oral mucosa tissue, including the wound area over the tooth extraction socket, was harvested from WT ZOL (n ϭ 3) and WT NaCl (n ϭ 3) mice. The gingival/palatal tissue was cut into small pieces, incubated with the premixed enzymes of a commercially available cell dissociation kit (Tumor Dissociation Kit, Miltenyi Biotec, Auburn, CA), and subjected to repeated mechanical agitations at room temperature and incubation at 37°C. Dissociated gingival/palatal tissue cells were washed and incubated with FITCconjugated monoclonal antibody against CD45 and PE-conjugated monoclonal antibodies against CD3, ␥␦TCR (GL3), or DX5 (BioLegend, San Diego). IgG2b was used as the isotype control. After 15 min of incubation on ice, cells were analyzed by flow cytometry (EPICS XL-MCL, Coulter, Miami, FL) (25,26). The data were presented using the lymphocyte gate.
To further investigate the presence of ␥␦ T cells in the tooth extraction socket histologically, Tcrd-H3BEGFP mice were used. Because of a knock-in mutation of an internal ribosome entry site-controlled histone 2B-enhanced GFP inserted into the 3Ј end of the T cell receptor ␦ constant gene, ␥␦ T cells of Tcrd-H2BEGFP mice were highlighted by GFP fluorescence (Tcrd-GFP ϩ ) (25). Tcrd-H2BEGFP mice were subjected to ZOL or NaCl injection followed by maxillary first molar extraction as described above. At day 4 (n ϭ 3 in each group) and week 2 (n ϭ 3 in each group) of tooth extraction, the maxillary tissue was harvested. After being fixed in 10% buffered formalin and decalcified, maxillary tissues were embedded in paraffin. Deparaffinized frontal sections, including the tooth extraction site, were subjected to antigen retrieval via microwave irradiation and incubated with an anti-GFP antibody (Santa Cruz Biotechnology, Dallas). GFP ϩ cells were visualized using the diaminobenzidine substrate, and the sections were counterstained with hematoxylin. Representative tooth extraction sockets were identified, and the number of GFP ϩ cells within the socket was determined. The data are presented as the number of GFP ϩ cells per extraction socket area (mm 2 ).
ONJ Development in ␥␦ T Cell Null (Tcrd Ϫ/Ϫ ) Mice-␥␦ T cell null (Tcrd Ϫ/Ϫ ) mice carry T cell receptor C␦ that was dis-rupted by gene targeting, resulting in the loss of T cells with ␥␦TCR expression in the lymphoid and epithelial tissues (27). To address the role of ␥␦ T cells in the pathological development of ONJ, female 7-week-old Tcrd Ϫ/Ϫ mice (B6.129P2-Tcrd tm1Mom /J, The Jackson Laboratory) were treated with a ZOL injection (Tcrd Ϫ/Ϫ ZOL) followed by maxillary first molar extraction as described above.
Because WT ZOL mice exhibited delayed tooth extraction healing and developed ONJ-like lesions at week 4, we primarily investigated the wound healing phenotype of Tcrd Ϫ/Ϫ ZOL mice at week 4 (n ϭ 15). In addition, this study examined the early wound healing process of Tcrd Ϫ/Ϫ ZOL mice at day 4 (n ϭ 3), week 1 (n ϭ 3), and week 2 (n ϭ 6). The harvested maxillary tissues were photo-recorded, fixed with 10% buffered formalin, and subjected to micro-CT imaging. The specimens were further subjected to decalcification and paraffin embedding for histological specimen preparation. In a parallel experiment, Tcrd Ϫ/Ϫ mice injected with 0.9% NaCl vehicle solution via the retro-orbital venous plexus were subject to left maxillary first molar extraction (n ϭ 6). Maxillary tissue was harvested on week 2.
Evaluation of Oral Mucosa Inflammation-The digital photo-records from WT NaCl, WT ZOL, and Tcrd Ϫ/Ϫ ZOL mice were used to identify the presence or absence of clinical alveolar bone exposure of WT and Tcrd Ϫ/Ϫ mice. The swelling area of gingival/palatal tissue was measured using a Java-based imageprocessing program (ImageJ, National Institutes of Health, Bethesda), which was standardized using the circumferential crown area of the remaining first molar.
To assess the inflammation degree of the oral mucosal tissue, cells dissociated from gingival/palatal tissues harvested at day 4 and week 2 (n ϭ 3 in each group) were incubated with a PEconjugated CD45 monoclonal antibody (BioLegend) and analyzed using flow cytometry. The number of CD45 ϩ lymphocytes among the total cells dissociated from gingival/palatal tissues was determined.
Histological Examination of ONJ-like Lesions-The fixed and decalcified maxillary tissues of WT NaCl, WT ZOL, and Tcrd Ϫ/Ϫ ZOL mice were subjected to a conventional paraffinembedded histological preparation. A series of frontal sections (8 m thick) of maxilla were prepared by bisecting the tooth extraction site and the contralateral remaining first molar. Histological sections were stained with hematoxylin and eosin. Histological images were digitized using a high throughput imaging system (Ariol SL-50, Applied Imaging, Grand Rapids, MI) and archived. ONJ-like lesions were determined based on the following criteria following the updated ONJ definition by the American Association of Oral and Maxillofacial Surgeons (28): 1) jawbone exposure associated with abnormal epithelial hyperplasia communicating to the palatal/alveolar bone, or 2) the development of a pustule directly on the surface of the alveolar bone associated with epithelial fistula. The jawbone exposure exhibited food and debris impaction, whereas the pustule only contained inflammatory cells. The prevalence was expressed as the number of animals with ONJ-like oral mucosal abnormalities over the number of animals at week 4.
For alveolar bone characterization, the histological images of palatal/alveolar bone containing the first molar extraction site were horizontally bisected to divide the oral and nasal sides. The primary focus was placed on the oral side of palatal/alveolar bone, which encompassed the buccal border of the alveolar bone to the mid-palatine suture. The degree of osteonecrosis was defined as the ratio between the number of osteocytes and the number of nonvital osteocytes determined using osteocytic lacunae without a cellular component (i.e. empty lacunae) or abnormally condensed small nuclei (pyknotic osteocytes). An operator blind to condition performed the histological evaluation.
The number of osteoclasts (OCs) within the oral half of the palatal/alveolar bone containing the first molar extraction site was counted in the histological specimens under ϫ20 magnification. OCs were defined as large cells with multiple nuclei (Ն2 nuclei) on the bone surface. Specific histological sections were stained with tartrate-resistant acid phosphatase using a commercially available kit (Sigma). The area of palatal/alveolar bone was measured using an image-processing program (ImageJ, National Institutes of Health). The OC number was normalized to the palatal/alveolar bone area.
Micro-CT Examination-Micro-CT data were reconstructed to generate three-dimensional images at a threshold of 220 using the software provided by the manufacturer of the micro-CT scanner.
Maxillary micro-CT three-dimensional images of WT NaCl, WT ZOL, and Tcrd Ϫ/Ϫ ZOL mice were used to identify alveolar bone abnormalities such as periosteal reaction, bone sequestra associated with ONJ, and bone remodeling. A blinded operator reviewed the micro-CT three-dimensional images, and the degrees of bone formation and bone resorption at the tooth extraction site were separately rated from 0 to 2. The bone remodeling index was calculated by combining the bone formation and resorption indices. Various micro-CT images exhibited severe periosteal reactions at the external surface of the alveolar bone, which were not considered in the quantitative evaluation.
Bone morphometry of femurs harvested 2 weeks after tooth extraction or 3 weeks after ZOL injection was characterized in WT NaCl, WT ZOL, and Tcrd Ϫ/Ϫ ZOL mice (n ϭ 6 in each group) by micro-CT. Volume of interest was set 600 m below the end point of the growth plate, which was used as an anatomical landmark. One hundred slices (1.2 mm) were evaluated in the distal femur metaphysis at a threshold of 220. The volume of interest only included secondary spongiosa. Trabecular bone morphometric measurements in three-dimensions included the following: bone volume normalized to tissue volume, trabecular number, trabecular thickness, and trabecular separation.
Human ␥␦ T Cells and CD3 ϩ T Cells Co-cultured with ZOLpretreated Human Osteoclasts-For preparing human osteoclasts (h-OCs), peripheral blood was collected from healthy donors, and peripheral blood mononuclear cells were separated by Ficoll-Hypaque gradient centrifugation. Peripheral blood mononuclear cells were cultured onto the tissue culture plate for 1 h, after which the adherent subpopulation of peripheral blood mononuclear cells was detached from the tissue culture plates, and the CD14 ϩ monocytes were purified (EasyStep Human Monocyte Isolation Kit, Stem Cell Technologies, Vancouver, Canada). Approximately 94% purity was achieved based on flow cytometric analysis of CD14. Monocytes (10.3 ϫ 10 6 cells) were cultured in ␣-minimal essential medium containing 25 ng/ml macrophage colony-stimulating factor (M-CSF) and receptor activator of nuclear factor-B ligand (RANKL) (25 ng/ml) (29). Medium was refreshed every 3 days. Development of h-OCs was monitored by tartrate-resistant acid phosphatase staining (Primary Cell Co., Sapporo, Japan) and resorption pit formation on synthetic carbonate apatite pre-coated plate (Cosmo Bio USA, Carlsbad, CA) following the manufacturer's protocols. Mature h-OCs were generated after 14 -21 days of incubation. Once developed, h-OCs were treated with 100 nM ZOL for 6 days.
The number of h-␥␦ T and h-CD3 ϩ T cells in each plate was counted at co-culture days 1, 3, and 6. The co-culture medium of h-OC/h-␥␦ T cells and h-OC/h-CD3 ϩ T cells was harvested at day 6, and IFN-␥ concentration was determined by ELISA. h-␥␦ T cells were then collected, washed with PBS, and prepared for injection to the Rag2 Ϫ/Ϫ mouse as described below. On the day of mouse injection, surface markers of h-␥␦ T cells were analyzed by flow cytometry using PE-conjugated monoclonal anti-human CD3, CD4, CD19, and CD69 antibodies (BioLegend) (25,26).
Maxillary tissue containing the tooth extraction site was photographed and harvested. After being fixed and decalcified, the conventional histological preparation as described above was stained with H&E.
Using histological sections stained with H&E, osteonecrosis was assessed by the cluster area of Ն5 necrotic osteocytes over the alveolar bone area. Epithelial abnormality was assessed for the presence or absence of hyperplastic epithelium.

Standardized Osteoclast Number of WT and Rag2
Ϫ/Ϫ Mice-The histological sections through the distal root area of the first molar from the above experiment were stained by tartrate-resistant acid phosphatase, and the number of OCs on the surface of palatal bone and in the tooth extraction socket was separately counted. The number of OC was standardized by the bone surface linear length.

Effect of m-OC on h-␥␦ T Cell and Mouse (m)-␥␦ T Cells in
Vitro-m-OCs were generated from mouse femur flowthrough cells with supplementation with 25 ng/ml M-CSF and 25 ng/ml RANKL for 7 days. h-␥␦ T cells were isolated from peripheral blood of healthy subjects as described above. m-␥␦ T cells were isolated from mouse spleen (TCR␥/␦ T cell isolation kit, mouse, Miltenyi Biotec). After the activation with rh-IL2 (100 units/ml) and lipopolysaccharide (LPS) (100 ng/ml), h-␥␦ T cells and m-␥␦ T cells were co-cultured with m-OCs that were exposed to 100 nM ZOL for overnight. h-␥␦ T cells and m-␥␦ T cells were also co-cultured with m-OCs without ZOL pretreatment. At 1, 4, and 8 days of co-culture, the medium was collected and assayed for IFN-␥ by ELISA.
Statistical Analyses-Student's t tests were used to compare means. A one-way analysis of variance followed by a Bonferroni correction was used when more than two groups were compared. Fisher's exact test was used to describe the prevalence of ONJ-like lesions in contingency tables. The correlation between osteonecrosis and OC numbers was assessed by Pearson correlation and Spearman's rank correlation tests.

Results
ONJ-like Lesions in WT Mice-In control WT mice injected with 0.9% NaCl vehicle solution (WT NaCl), the maxillary first molar extraction created an open wound the size of the maxillary first molar crown, which remained open after surgery. Four days after tooth extraction, all WT NaCl mice continued to exhibit an open wound in the gingival tissue, which was surrounded by localized swelling (Fig. 1A). The open wound of the maxillary gingival tissue progressively closed and was completely covered by epithelial tissue between weeks 2 and 4. In the ZOL-treated WT mice (WT ZOL), the closure of gingival open wounds was slow. In total, 50% (6/12) of the animals exhibited clinically open wounds at week 4, and the soft tissue swelling continued to surround the wound opening (Fig. 1A). The open wound was more common at the large mesial root, which was observed in all WT ZOL mice.
Histological sections revealed that the tooth extraction wounds of WT NaCl mice healed with new bone formation in the bony socket, and the oral epithelial integrity was re-established between weeks 2 and 4 (Fig. 1B). Mild inflammatory cell infiltrates were observed in the palatal oral mucosa. By contrast, WT ZOL mice exhibited alveolar bone exposure at the mesial root extraction site with food and debris impaction. At the small distal root areas, where the oral mucosa wound was closed based on gross observation, histological evaluations revealed definitive bone exposure facilitated by the abnormal migration of oral epithelial cells. The abnormal proliferation and migration of epithelial cells resembling pseudoepitheliomatous hyperplasia (24,31) resulted in direct epithelial contact with the partially necrotic alveolar bone. Reduced new bone formation was noted in the extraction socket, and densely localized inflammatory cell infiltration was observed on the surface of the palatal bone adjacent to the tooth extraction site (Fig. 1B).
Oral ␥␦ T Cells in the Tooth Extraction Wound-The cellular component was dissociated from gingival/palatal oral mucosa tissues harvested from WT NaCl and WT ZOL mice 2 weeks after the tooth extraction. Cells in the lymphocyte gate in the flow cytometry were predominantly composed of CD45 ϩ cells ( Fig. 2A). Within the CD45 ϩ population, the WT NaCl and WT ZOL samples contained 7.1 and 5.5% CD3 ϩ T cells, respectively. GL3 ϩ ␥␦ T cells were 0.4 and 1.3%, and DX5 ϩ natural killer cells were 1.1 and 2.3% in the WT NaCl and WT ZOL samples, respectively. It was estimated that the fraction of ␥␦ T cells in the infiltrated T cell population was 5.6 and 23.6% in the WT NaCl and WT ZOL oral wound samples at week 2 of tooth extraction.
Using Tcrd-H2BEGFP mice, GFP ϩ ␥␦ T cells were examined in the extraction socket (Fig. 2B). At day 4 after tooth extraction, the number of GFP ϩ ␥␦ T cells did not differ between Tcrd-H2BEGFP NaCl and ZOL mice (Fig. 2C). However, at week 2 of wound healing, the number of GFP ϩ ␥␦ T cells significantly decreased in the control mice. By contrast, Tcrd-H2BEGFP ZOL mice exhibited an increase of GFP ϩ ␥␦ T cells in the tooth extraction socket (Fig. 2C). A few GFP ϩ ␥␦ T cells were identified in the unwounded junctional epithelium of the free gingiva adjacent to the tooth (Fig. 2D). However, in the oral epithelial hyperplasia in Tcrd-H2BEGFP ZOL mice, an increased density of GFP ϩ ␥␦ T cells was observed (Fig. 2D). Combining the data of GL3 ϩ and GFP ϩ ␥␦ T cells, ZOL treatment appeared to associate with the sustained appearance of ␥␦ T cells in the oral mucosa at the tooth extraction wound site.
Different Oral Mucosa Wound Healing in Tcrd Ϫ/Ϫ ZOL Mice-We addressed the possible role of ␥␦ T cells in the development of ONJ using ␥␦ T cell-deficient (Tcrd Ϫ/Ϫ ) mice. The healing patterns of the tooth extraction wound were indistinguishable between the WT ZOL and Tcrd Ϫ/Ϫ ZOL mice until week 1. At week 2, the tooth extraction wound area was visibly reduced in Tcrd Ϫ/Ϫ ZOL mice compared with WT ZOL mice. At week 4, the tooth extraction wound was closed in 86.7% of Tcrd Ϫ/Ϫ ZOL mice (13/15); however, gingival/palatal tissue swelling remained (Fig. 3A). The standardized oral mucosal swelling area was significantly larger in the Tcrd Ϫ/Ϫ ZOL mice compared with the WT ZOL mice at weeks 2 and 4 (Fig. 3B).
Pustule/Fistula Development on the Necrotic Bone Surface of Tcrd Ϫ/Ϫ ZOL Mice-Histological cross-sections of Tcrd Ϫ/Ϫ ZOL mice revealed well healed extraction sockets and oral mucosa wounds at week 4 (Fig. 3C). In 26.7% of the Tcrd Ϫ/Ϫ ZOL mice, small sequestrations of necrotic bone were observed in the connective tissue of the oral mucosa. Unlike the ONJ-like lesions of the WT ZOL mice, the development of large pustules on the surface of necrotic palatal bones was observed in Tcrd Ϫ/Ϫ ZOL mice (Fig. 3C). The associated epithelium at the pustule was consistent with fistula formation.

FIGURE 2. Appearance of ␥␦ T cells in the oral mucosa after tooth extraction.
A, palatal/gingival tissue of WT NaCl or WT ZOL mice was harvested 2 weeks after tooth extraction, and cells were dissociated by repeated enzymic digestions. Dissociated cells were incubated with CD45, CD3, ␥␦TCR (GL3), or DX5 antibodies and examined by flow cytometry. In a representative analysis, CD45 ϩ population contained CD3 ϩ T cells at similar rates; however, the WT ZOL sample showed higher levels of GL3 ϩ ␥␦ T cells and DX5 ϩ NK cells than the WT NaCl sample. B, to further identify ␥␦ T cells, maxillary histological specimens were prepared from NaCl or ZOL-injected Tcrd-H2BEGFP mice 4 days or 2 weeks after tooth extraction (Ext.). Anti-GFP antibody was used to highlight GFP ϩ ␥␦ T cells by immunohistology. A cluster of GFP ϩ ␥␦ T cells was found in the connective tissue of the tooth extraction wound. C, number of GFP ϩ ␥␦ T cells in the tooth extraction socket was found to be similar 4 days after tooth extraction in WT NaCl (n ϭ 4) and WT ZOL (n ϭ 4) mice. However, although in the control WT NaCl mice (n ϭ 4), GFP ϩ ␥␦ T cells decreased 2 weeks after tooth extraction, the GFP ϩ ␥␦ T cell number increased in WT ZOL mice (n ϭ 4). *, p Ͻ 0.05 D, in some WT ZOL mice, PEH-like oral epithelial hyperplasia (Epi Hyp) was observed at the tooth extraction site. GFP ϩ ␥␦ T cells were abundant in the hyperplastic epithelium but not in normal oral epithelium. JULY 10, 2015 • VOLUME 290 • NUMBER 28

JOURNAL OF BIOLOGICAL CHEMISTRY 17355
The effect of ZOL on the femurs was characterized in WT ZOL and Tcrd Ϫ/Ϫ ZOL mice by micro-CT, which demonstrated the similar increases in bone volume normalized to tissue volume and trabecular number and decreases in trabecular separation as compared with the WT NaCl control group (Fig.   3E). No significant difference was noted between WT ZOL and Tcrd Ϫ/Ϫ ZOL femurs.
Severe Periosteal Reactions Occurred in WT ZOL Mice but Not in Tcrd Ϫ/Ϫ ZOL Mice-Radiographic and CT image examinations of patients with ONJ have revealed that periosteal reac-tion predominantly occurs during advanced disease stages (32,33). This study used micro-CT images to determine the presence of abnormal alveolar bone in mouse maxilla. WT NaCl mice revealed evidence of bone resorption in the palatal alveolar bone at week 2 of tooth extraction as well as bone formation in the extraction socket at week 4 (Fig. 4, A and B). WT NaCl mice did not exhibit abnormal alveolar bone wound healing. ZOL treatment of WT and Tcrd Ϫ/Ϫ mice appeared to decrease osteolytic signs at the extraction site, whereas bone formation in the extraction socket was less affected (Fig. 4, A and B). A total of 53% of WT ZOL mouse specimens from weeks 2 and 4 exhibited periosteal reactions, including severe forms, at the tooth extraction wound healing site or the contralateral periodontal disease area (Fig. 4A). Tcrd Ϫ/Ϫ ZOL mice demonstrated a small "sun-ray" type of calcified spikes on the surface of the alveolar bone; however, the severity of the periosteal reaction was attenuated considerably (Fig. 4A).
Chronic Oral Mucosa Inflammation in WT ZOL and Tcrd Ϫ/Ϫ ZOL Mice-Oral epithelial hyperplasia in WT ZOL mice migrated toward the inflammatory lesion and established direct contact between the basal cells and the surface of partially necrotic alveolar bone (Fig. 5A). These characteristic features were less apparent in Tcrd Ϫ/Ϫ ZOL mice, and the epithelial tissue did not reach the bone surface except at the periphery of the pustule (Fig. 5A). Despite the different oral mucosal disease phenotypes of ONJ-like lesions in WT ZOL and Tcrd Ϫ/Ϫ ZOL mice, both models exhibited sustained inflammation. WT ZOL mice exhibited localized and dense inflammatory cell infiltrates near the bone surface, whereas inflammation in the gingival/palatal mucosa tissues was more diffuse and often exhibited cellulitis-like inflammatory cell infiltration in Tcrd Ϫ/Ϫ ZOL mice (Fig. 5A).
Dissociated cells from the gingival/palatal mucosa tissues contained over 50% CD45 ϩ leukocytes in both the WT NaCl and WT ZOL mice at day 4 (Fig. 5B). Control WT NaCl mice exhibited a significant reduction in CD45 ϩ cells at week 2, whereas WT ZOL mice maintained high levels of CD45 ϩ cells at week 2 (Fig. 5B). Similarly, Tcrd Ϫ/Ϫ ZOL mice exhibited an increase in CD45 ϩ cells at week 2 (Fig. 5B). Two weeks after the maxillary first molar extraction, Tcrd Ϫ/Ϫ NaCl mice also sustained oral mucosa swelling and CD45 ϩ cells infiltration (data not shown).
Osteonecrosis in WT ZOL and Tcrd Ϫ/Ϫ ZOL Mice-The degree of osteonecrosis in the palatal/alveolar bone was defined as the proportion of nonvital osteocyte lacunae in the oral and nasal sides. The nasal side of palatal bone contained 10 -15% nonvital osteocyte lacunae in all groups at all of the tested time points (data not shown). The oral side of the palatal bone exhib-ited a cluster of ϳ20% nonvital osteocyte lacunae immediately after tooth extraction, which gradually decreased over time in control WT NaCl mice (Fig. 5C). In WT ZOL mice, the area of osteonecrosis was observed in the bone exposure site as well as the palatal bone that interfaced the oral mucosa and dense inflammatory infiltrates (Fig. 5A). The area of osteonecrosis was not reduced over time but rather increased at week 4, which resulted in persistent osteonecrosis (Fig. 5C). The osteonecrosis areas of the Tcrd Ϫ/Ϫ ZOL mice were associated with bone sequestra and pustules (Fig. 5A). The Tcrd Ϫ/Ϫ ZOL group exhibited increased osteonecrosis development at week 2, followed by a mild decrease at week 4 (Fig. 5C). The reference Tcrd Ϫ/Ϫ NaCl group developed a significantly smaller necrotic bone area at week 2, despite the remaining open wound (data not shown).
The number of OCs in the palatal/alveolar bone of WT NaCl control mice demonstrated an early increase at day 4 and week 1 after tooth extraction, followed by a gradual decrease (Fig.  5D). By contrast, OCs in the palatal/alveolar bone of WT ZOL and Tcrd Ϫ/Ϫ ZOL mice remained relatively unchanged throughout the experiment, resulting in a significantly increased number compared with the WT NaCl mice at week 4 ( Fig. 5D). Active bone resorption was observed at the pustule periphery (Fig. 5A). The removal of necrotic bone sequestra might contribute to the late decrease of osteonecrosis area in Tcrd Ϫ/Ϫ ZOL mice.
The correlation between the area of osteonecrosis and the number of osteoclasts in the combined data of weeks 1, 2, and 4 was evaluated by Pearson correlation and Spearman's rank correlation tests. The WT NaCl group revealed the significant correlation in both tests, whereas the WT ZOL and Tcrd Ϫ/Ϫ ZOL groups showed a similar correlation pattern with no statistical significance (Fig. 5E).
In summary, ZOL-treated Tcrd Ϫ/Ϫ mice developed the different oral mucosa disease phenotype characterized by the pustule and fistula formation; however, the osteonecrosis occurred similarly to WT ZOL mice.

Activation of Human Peripheral Blood ␥␦ T Cells by ZOLpretreated h-OC-h-␥␦ T cells contain V␥9V␦2 T cells. BPtreated macrophages accumulate phosphorylated metabolites
such as isopentenyl pyrophosphate (IPP) (34), which have been shown to selectively activate h-␥␦ T cells but not mouse ␥␦ T cells. In the context of ONJ pathogenesis, we investigated the role of ZOL-treated h-OCs, rather than peripheral blood macrophages, on the activation of h-␥␦ T cells in this project.
Three million h-␥␦ T cells and 51 million h-CD3 ϩ T cells were isolated from approximately equal volumes of peripheral blood of different healthy donors. The estimated purity of h-␥␦  (black dotted line). B, area of swelling normalized by the circumferential crown size of the contralateral first molar was significantly increased in Tcrd Ϫ/Ϫ ZOL mice compared with WT ZOL mice at weeks 2 and 4; *, p Ͻ 0.05. C, histological examination revealed that Tcrd Ϫ/Ϫ ZOL mice did not exhibit the bone exposure as observed in WT ZOL mice. The development of pustule (arrows; Pus) associated with epithelial fistulation (arrowheads) was demonstrated. The tooth extraction socket (black dotted line) healed uneventfully and was filled with woven bone. Osteonecrosis (red dotted line; Ost.Nec) was observed at the pustule areas and bone sequestration (red arrows). D, prevalence of necrotic jawbone exposure was significantly higher in WT ZOL mice than Tcrd Ϫ/Ϫ ZOL or WT NaCl control mice. However, Tcrd Ϫ/Ϫ mice exhibited more pustule/fistula formation than WT ZOL and WT NaCl mice. Both ONJ phenotypic variants were combined, and the prevalence of ONJ-like lesion development was found indistinguishable between WT ZOL and Tcrd Ϫ/Ϫ ZOL mice. *, p Ͻ 0.05; †, p ϭ 0.06. E, micro-CT characterization of femur trabecular bone morphology of WT NaCl, WT ZOL, and Tcrd Ϫ/Ϫ ZOL mice. The effect of ZOL was shown as the increased bone mass and structure of trabecular bone of WT ZOL and Tcrd Ϫ/Ϫ ZOL mice. *, p Ͻ 0.05; **, p Ͻ 0.01.
T cells and h-CD3 ϩ T cells was 76.9 and 96.4%, respectively (Fig. 6A). After overnight activation by anti-CD3 antibody, anti-CD28 antibody, and rh-IL-2, T cells were co-cultured with ZOL-pretreated h-OC. The number of h-␥␦ T cells robustly increased during the co-culture period, whereas the number of h-CD3 ϩ T cells decreased (Fig. 6B).
After 6 days of co-culture, the secretion of IFN-␥ was significantly higher by h-␥␦ T cells than h-CD3 ϩ T cells (Fig. 6C). The cell surface antigen analysis suggested that h-␥␦ T cells maintained the T cell profile with a small population of CD4 ϩ T cells. The large representation of CD69-positive staining indicated the lymphocyte activation (Fig. 6D). As such, this study suggested that h-␥␦ T cells were activated by ZOL-pretreated h-OCs.
Rag2 Ϫ/Ϫ Mice Engrafted with h-␥␦ T Cells-One day after injection of h-␥␦ T cells to ZOL-treated Rag2 Ϫ/Ϫ mice, the maxillary first molar was extracted. At the time of euthanasia 2 weeks after tooth extraction, spleen and bone marrow cells were harvested. The profile of endogenous mouse immune effectors showed the lack of CD3 ϩ T cells and CD19 ϩ B cells FIGURE 4. Alveolar bone healing assessment by micro-CT imaging. A, WT NaCl control mice exhibited signs of osteolysis at week 2 as roughened alveolar and palatal bone surfaces (white arrowheads). Bone formation within the extraction socket (black arrowhead) appeared to complete the wound healing process at week 4. We noted that one WT NaCl mouse exhibited periodontal disease at the nonextraction side (white arrowhead). In WT ZOL and Tcrd Ϫ/Ϫ ZOL mice, the evidence of bone resorption was not clearly observed, and the extraction socket was not completely filled by new bone formation (black arrowheads). Severe periosteal reaction (black arrows) was observed at the molar extraction site in 53% of the WT ZOL mice. A periosteal reaction was also noted in one WT ZOL mouse at the contralateral nonextraction side with periodontal disease (confirmed with histology). Although the severity of periosteal reaction in Tcrd Ϫ/Ϫ ZOL mice was much attenuated (arrows), the sun-ray type of calcified spikes was observed both in WT ZOL and Tcrd Ϫ/Ϫ ZOL mice (SR with black arrow). B, micro-CT images were used to rate bone formation (0 -2) and bone resorption (0 -2) appearance. The bone-remodeling rate was expressed by combining the rates of bone formation and resorption in each animal. Active bone remodeling was suggested in WT NaCl control mice. Both WT ZOL and Tcrd Ϫ/Ϫ ZOL mice showed significantly decreased bone resorption. *, p Ͻ 0.05. FIGURE 5. Characterization of osteonecrosis in WT NaCl, WT ZOL, and Tcrd ؊/؊ ZOL mice. A, localized inflammation (Inf) was evident near the surface of alveolar bone in WT NaCl mice. In WT ZOL mice, the intense inflammatory reaction was demonstrated in the oral mucosa (double-headed arrow; Inf). Abnormal epithelial tissue adhesion (Epi) was evident on the surface of partially necrotic alveolar bone (Ost. Nec.). In Tcrd Ϫ/Ϫ ZOL mice, inflammation was more diffused, and epithelial tissue did not reach the necrotic bone surface except at the pustule periphery. In both lesions, clusters of osteoclasts (arrowheads) were observed at and adjacent to the osteonecrosis site. B, CD45 ϩ lymphocytes accounted for 30 -45% of the dissociated cells from gingival/palatal tissue at day 4. The fraction of CD45 ϩ cells decreased in WT NaCl mice at week 2. However, WT ZOL and Tcrd Ϫ/Ϫ ZOL mice exhibited similar fractions of CD45 ϩ lymphocytes in the oral mucosa. C, area of osteonecrosis is presented as the percent of nonvital osteocytes over the total number of osteocytes on the oral side of palatal/alveolar bone. At week 4, the area of osteonecrosis was significantly larger in WT ZOL and Tcrd Ϫ/Ϫ ZOL mice than WT NaCl control mice. D, number of osteoclasts in the palatal/alveolar bone area suggested that the early increase in the tooth extraction wound of WT NaCl control mice followed by progressive decrease. Both WT ZOL and Tcrd Ϫ/Ϫ ZOL mice indicated a consistent appearance of osteoclasts throughout the experimental period, resulting in the elevated osteoclast number at week 4. *, p Ͻ 0.05; †, p ϭ 0.06 compared with the WT NaCl control group. E, correlation between the osteonecrosis area and the number of OCs in week 1, week 2, and week 4 specimens. The WT NaCl group demonstrated the significant correlation between the osteonecrosis area and osteoclast number. By contrast, correlation pattern was similar in the WT ZOL and Tcrd Ϫ/Ϫ ZOL groups, which did not show statistical significance.
but the presence of DX5 ϩ natural killer cells in spleen (data not shown) and F4/80 ϩ monocytes in bone marrow (Fig. 6E). The profile of human immune effectors indicated the presence of CD45 ϩ ␥␦TCR ϩ and CD3 ϩ ␥␦TCR ϩ cells (Fig. 6E), suggesting the successful engrafting of h-␥␦ T cells in Rag2 Ϫ/Ϫ mice, which survived for the experimental period of 2 weeks.
Lack of ONJ-like Legions in ZOL-treated Rag2 Ϫ/Ϫ Mice-Rag2 Ϫ/Ϫ mice with NaCl or ZOL injection showed accelerated tooth extraction wound healing as compared with WT mice.
All Rag2 Ϫ/Ϫ NaCl mice showed the closure of the tooth extraction wound at week 2 (Fig. 7A). Although Rag2 Ϫ/Ϫ ZOL mice showed small food and debris impaction at the wound site in three out of five mice in the same healing period, histological examination revealed well healed extraction sockets (Fig. 7A).
Oral Epithelium Hyperplasia in Rag2 Ϫ/Ϫ ZOL Mice Engrafted with h-␥␦TCR-Rag2 Ϫ/Ϫ ZOL mice engrafted with h-␥␦ T cells showed an unusual roughened oral soft tissue healing at week 2 (Fig. 7B). Histological examination revealed vari- ous degrees of oral epithelial hyperplasia in all mice. Oral epithelium exhibited irregular rete pegs and extensions of the basal cells. Where the hyperplastic oral epithelium embedded in the deep connective tissue, small food and debris were enclosed, and the tooth extraction socket showed the delay in bone healing (Fig. 7B).
In this project, we employed the autoclaved cellulose bedding for immunodeficient Rag2 Ϫ/Ϫ mice, which might, in turn, change the wound-healing environment. WT NaCl mice housed in the same bedding environment showed the accelerated tooth extraction wound closure, whereas WT ZOL mice housed in autoclaved cellulose bedding demonstrated the ONJ Oral Mucosal Disease Phenotypes of ONJ in Mice JULY 10, 2015 • VOLUME 290 • NUMBER 28 phenotype and oral wound was open at week 2 in all mice. WT ZOL mice showed the significantly larger osteonecrosis area than WT NaCl mice (Fig. 7C). The osteonecrosis area of Rag2 Ϫ/Ϫ NaCl and Rag2 Ϫ/Ϫ ZOL mice remained small at the level of WT NaCl mice. Despite the development of oral epithelial hyperplasia in h-␥␦ T cell-engrafted Rag2 Ϫ/Ϫ ZOL mice, the osteonecrosis area remained unaffected (Fig. 7C).
Reduced Osteoclastogenesis in Rag2 Ϫ/Ϫ Mice-Rag2 Ϫ/Ϫ ZOL mice with or without h-␥␦ T cell engraftment demonstrated significantly less osteoclasts on the palatal bone surface than WT NaCl and WT ZOL mice (Fig. 8A). Dense inflammatory cell infiltration was observed in the palatal/gingival tissue of WT ZOL mice, which interfaced palatal bone lined by a number of osteoclasts. By contrast, no inflammatory infiltration was observed in Rag2 Ϫ/Ϫ ZOL mice (Fig. 8A). Because the reduced osteoclast number was similarly observed in Rag2 Ϫ/Ϫ NaCl mice (Fig. 8A), it was postulated that osteoclastogenesis on the palatal bone surface required the mediation by T cells, whereas the normalized osteoclast number in the extraction socket of Rag2 Ϫ/Ϫ mice could be associated with the accelerated bone wound healing and remodeling.

m-OC Induced the Secretion of IFN-␥ of h-␥␦ T Cells-
The potential interaction between m-OC and h-␥␦ T cells was addressed by in vitro co-culture study. When h-␥␦ T cells were co-cultured with ZOL pretreated m-OC, IFN-␥ secretion was significantly increased (Fig. 8B). Because m-OC alone did not secrete IFN-␥, activated h-␥␦ T cells were the source of IFN-␥. Surprisingly, IFN-␥ secretion by h-␥␦ T cells was also increased when co-cultured with m-OC without ZOL pretreatment, albeit with a transient fashion (Fig. 8B). In the case of m-␥␦ T cells, nonspecific activation by m-OC with or without ZOL pretreatment was observed resulting in the higher levels of IFN-␥ secretion (Fig. 8B).

Discussion
This study demonstrated the ONJ-like lesions with severe bone exposure (Fig. 1) and the nonexposure variants (Fig. 3) generated in mice that differed in the presence or absence of ␥␦ T cells, respectively. WT ZOL mice exhibited extended exposure of partially necrotic alveolar bone associated with abnormal oral epithelial hyperplasia. By contrast, in the absence of ␥␦ T cells, Tcrd Ϫ/Ϫ ZOL mice developed pustules on the surface of necrotic alveolar bone, which appeared to form fistulae with the oral epithelium. The exposure of necrotic bone has been a clinical hallmark of a fully developed ONJ lesion. However, the current position of the American Association of Oral and Maxillofacial Surgeons now includes intraoral or extraoral fistulae in the maxillofacial region as an additional clinical definition of medication-related ONJ (28). This revised ONJ definition underscores large clinical variations in the development of oral mucosa abnormalities in this disorder.
The area of osteonecrosis in WT ZOL mice progressively increased, whereas the osteonecrosis area in Tcrd Ϫ/Ϫ ZOL mice peaked at week 2 and then declined (Fig. 5C). The number of osteoclasts on the surface of palatal bone was equivalent in the WT ZOL and Tcrd Ϫ/Ϫ ZOL mice (Fig. 5D). However, histologically, osteoclasts appeared to cluster immediately outside the pustule in Tcrd Ϫ/Ϫ ZOL, resulting in bone sequestration (Fig. 3C). The bone sequestra in Tcrd Ϫ/Ϫ ZOL mice appeared to move toward the oral epithelium and was moved out to the oral cavity, which potentially contributed to the late reduction of the osteonecrotic area. In the presence of ␥␦ T cells, osteoclasts in WT ZOL mice did not cluster but distributed on the palatal bone surface interfacing oral mucosa with severe inflammatory reaction (Fig. 8A). It was further noted that Rag2 Ϫ/Ϫ mice lacking T and B cells showed the decreased osteoclast number on the palatal bone surface without inflammatory cell infiltration in the oral mucosa (Fig. 8A). Therefore, we postulate that tooth extraction-induced oral mucosa inflammation containing ␥␦ T cells may be responsible, in part, for the induction of osteoclasts distributed on the palatal bone surface, although the severe inflammatory reaction in the form of pustule formation may be required for the localized osteoclast clustering in Tcrd Ϫ/Ϫ ZOL mice.
A recent retrospective analysis of BP-treated patients in Copenhagen concluded that the severe bone exposure and the nonexposure variants of ONJ belonged to the same disease condition. However, although the severe bone exposure variant was observed from the early BP treatment stage, the nonexposure variant was not reported until patients received repeated BP injections (35). ONJ patients who had received multiple injections of BP have been shown to develop deficiency of circulating ␥␦ T cells (17). Human ␥␦ T cells are largely composed of V␥9V␦2 T cells, which can react to phosphoantigens (pAg) such as bacteria or parasite-derived (E)-4-hydroxy-3-methylbut-2-enyl pyrophosphate from the non-mevalonate pathway (36). V␥9V␦2 T cells can also react to mevalonate pathwayrelated pAgs such as IPP and isomer dimethylallyl pyrophosphate, albeit at much lower potencies (37,38). Macrophages FIGURE 7. ONJ lesions of Rag2 ؊/؊ mice and the effect of h-␥␦ T cell engraftment. A, accelerated tooth extraction wound healing was observed in Rag2 Ϫ/Ϫ mice. Two weeks after first molar extraction, both Rag2 Ϫ/Ϫ NaCl and Rag2 Ϫ/Ϫ ZOL mice exhibited uneventful healing (arrows) without oral mucosa swelling. Tooth extraction wound was completely closed in all Rag2 Ϫ/Ϫ NaCl mice (n ϭ 5), whereas Rag2 Ϫ/Ϫ ZOL mice (n ϭ 5) showed nonkeratinized oral mucosa covering the extraction socket or small open wound with food and debris impaction. Histologically, the extraction socket (Soc) was filled with newly formed bone in both Rag2 Ϫ/Ϫ NaCl and Rag2 Ϫ/Ϫ ZOL mice. The area of osteonecrosis (Ost Nec) was small. There were small food and debris impactions (arrow), which appeared to be integrated in the healed tissue. B, Rag2 Ϫ/Ϫ ZOL mice engrafted with h-␥␦ T cells (n ϭ 4) similarly showed the wound closure without swelling; however, there were unusual pits and irregular papilloma-like tissues (arrows) were present at the first molar extraction site. Histologically, all Rag2 Ϫ/Ϫ ZOL h-␥␦ T cell mice exhibited various degrees of oral epithelial hyperplasia (Epi, arrows). In some specimens, epithelial "pouch" was found in the deep connective tissue, which often, but not always, contained small food and debris. The extraction socket (Soc) containing the epithelial pouch showed delayed bone formation. Osteonecrosis was found in small areas, associated with osteoclasts (arrowheads). C, osteonecrosis was determined by the bone area of containing Ն5 empty osteocyte lacunae over the total alveolar bone area. Because all Rag2 Ϫ/Ϫ mice were housed with autoclaved cellulose-based bedding, a separate set of WT NaCl and WT ZOL mice (n ϭ 5 in each group) was housed in the same environment after tooth extraction. Tooth extraction wound healing appeared to be more favorable in this environment; however, WT ZOL mice exhibited open tooth extraction wounds with significantly larger osteonecrosis areas. The osteonecrosis area of Rag2 Ϫ/Ϫ ZOL mice remained small at the levels of WT NaCl and Rag2 Ϫ/Ϫ NaCl mice. The engraftment of h-␥␦ T cells to Rag2 Ϫ/Ϫ ZOL mice did not increase the osteonecrosis area. *, p Ͻ 0.05 compared with the WT NaCl control group.
exposed to BP such as ZOL become an activator of V␥9V␦2 T cells, likely in consequence of intracellular IPP accumulation. The repeated BP injections are postulated to result in the repeated activation of V␥9V␦2 T cells leading to eventual ␥␦ T cell depletion (17,18). It is tempting to speculate that the decreased circulating V␥9V␦2 T cells may causally associate with the development of nonexposure and fistula phenotype of ONJ in humans.
It has been established that the V␥9V␦2 T cell receptor as well as butyrophilin-3A1 critical for IPP-induced activation of ␥␦ T cells are conserved only in humans, primates, and a few other placental animals but not in mice (39,40). Therefore, mouse ␥␦ T cells do not respond to pAgs, and the ZOL injection should not activated the circulating ␥␦ T cells in mice through a similar mechanism as in humans. Thus, this study examined the role of human ␥␦ T cells in the development of oral mucosal phenotypes in ZOL-treated Rag2 Ϫ/Ϫ mice. Like BP-exposed macrophages, this study demonstrated that ZOL-treated h-OCs activated h-␥␦ T cells. By co-culturing with ZOL-pretreated h-OC, h-␥␦ T cells were found to increase the cell counts, secrete IFN-␥ at a much higher level than the similarly activated h-CD3 ϩ T cells, and express the lymphocyte activation marker CD69 (Fig. 6). When h-␥␦ T cells that were preactivated by ZOL-treated h-OCs were engrafted onto Rag2 Ϫ/Ϫ ZOL mice, the tooth extraction wound healed with abnormal oral epithelial hyperplasia (Fig. 7).
The abnormal oral epithelial hyperplasia referred to as pseudoepitheliomatous hyperplasia (PEH) was reported in 60% of human ONJ biopsy specimens (24,31) and was associated with the more aggressive ONJ phenotype with necrotic bone exposure. Akilov et al. (41) reported the development of PEH in mouse ear skin by the injection of TNF-␣ and IFN-␥, suggesting that overproduction of Th1 cytokines in the barrier immunity stimulated the unregulated proliferation and migration of epidermal basal cells. Rag2 Ϫ/Ϫ mice used in this study lacked T and B cells, whereas NK cells and monocytes were present (Fig. 6). Therefore, the engrafted h-␥␦ T cells were likely to be a source of Th1 cytokines such as TNF-␣ and IFN-␥ (42).
This study demonstrated that PEH-like oral epithelial migration was less apparent in Tcrd Ϫ/Ϫ ZOL mice that nonetheless developed the nonexposure variant-like ONJ lesion (Fig. 4). Conversely, the prolonged retention of ␥␦ T cells in the tooth extraction wound of WT ZOL mice may have contributed to PEH-like oral epithelial hyperplasia leading to the gross exposure of necrotic bone and the severe ONJ phenotype.
To better understand the "␥␦ T cell humanized" mice, h-␥␦ T cells and m-␥␦ T cells were co-cultured with mouse OC (m-OC). As expected, the activation of h-␥␦ T cells was observed with ZOL pretreated m-OC. However, co-culturing with m-OC without ZOL pretreatment also activated h-␥␦ T cells. The ZOL treatment of m-OC is expected to accumulate pAg such as IPP, which specifically activate h-␥␦ T cells. However, the surprising observation of h-␥␦ T cell activation by ZOL-untreated m-OC, albeit with a transient nature, should suggest a nonspecific activation of immune effectors by OCs. The activation of m-␥␦ T cells by m-OC with or without ZOL pretreatment further supports the previously unrecognized functions of OCs other than bone resorption, which might include regulatory roles of local immune effectors. Besides the pAg-derived activation, human and mouse ␥␦ T cells are activated by peptide stimulants such as NKG2D ligands (43). It is tempting to speculate that ONJ pathogenesis may involve nonspecific activation of ␥␦ T cells by OCs through the non-pAg activation mechanism in mice as well as in humans.
It is important to report that Rag2 Ϫ/Ϫ ZOL mice with or without h-␥␦ T cell engraftment did not develop osteonecrosis (Fig. 7), whereas WT ZOL and Tcrd Ϫ/Ϫ ZOL developed osteonecrosis (Figs. 1, 3, and 5). It appears that the prolonged or unresolved oral inflammation is commonly observed in our mouse models as well as in ONJ patients. The time course profile of OC numbers at the surface of jawbone was similarly sustained in WT ZOL and Tcrd Ϫ/Ϫ ZOL mice (Fig. 5). By contrast, Rag2 Ϫ/Ϫ mice lacking oral mucosal inflammation significantly reduced the recruitment of osteoclasts (Fig. 8A). Considering the mode of action of BP targeting OCs, it may not be a mere coincidence that osteonecrosis legions were found to localize with or adjacent to a cluster of OCs (Fig. 3C). Therefore, the lack of OC in Rag2 Ϫ/Ϫ ZOL mice may lead to the lack of osteonecrosis.
We previously proposed that ZOL-affected OCs might mediate the pro-inflammatory environment in oral mucosa based on the unique observation of a group of inflammatory cells closely associated with OCs in rat jawbones treated with ZOL (24). This study further suggested the role of OC on the immune regulation (Fig. 8B). Zhang et al. (44) demonstrated an unusually increased number of IL17-expressing helper T cells and M1 macrophages in the oral mucosa of ONJ patients. The ONJ patient demography with co-morbidities such as diabetes type 2 (45) may suggest the presence of systemic pro-inflammatory susceptibility. As such, the dysregulation of systemic as well as local oral mucosa immunity may provide the underlining core mechanism to the prolonged oral inflammation leading to the development of ONJ, in which ␥␦ T cells may modify the oral epithelial disease phenotype.