Anti-TNF-α, activin A and IL-8 antibodies, as well as recombinant RANKL, OPG, activin A and nonspecific mouse immunoglobulin (Ig)G were purchased from R&D Systems, Inc., (Minneapolis, MN, USA). Anti-RANKL, cathepsin K, and acid phosphatase 5, tartrate resistant (TRAP) antibodies were purchased from Abcam (Cambridge, UK). Anti-JUN N-terminal kinase (JNK), anti-phosphorylated (p)-JNK and anti-nuclear factors of activated T-cells (NFATc-1) antibodies were purchased from Cell Signaling Technology, Inc., (CST; Danvers, MA, USA). Recombinant human IL-8 and recombinant murine M-CSF were purchased from Sino Biological (Beijing, China). The JNK pathway inhibitor SP600125 was purchased from Sigma-Aldrich (Merck KGaA, Darmstadt, Germany).

The AM tissues were obtained from 2 male patients (27 and 29 years old) and 2 female patients (23 and 24 years old) treated at the Department of Oral and Maxillofacial Surgery (Hospital of Stomatology, Sun Yat-sen University, Guangzhou, China) from June 2017 to February 2018. Informed consent was obtained according to a protocol approved by the Ethical Committee of the Guanghua School of Stomatology, Hospital of Stomatology and Sun Yat-sen University [Guangzhou, China; ERC-(2017)-5]. Principles outlined in the Declaration of Helsinki were followed. All AM tissues were resected from the mandible, two were from plexiform and two were from follicular AM (Table SI). Normal bone tissue was obtained from a 24-year-old female patient with dento-maxillofacial deformities during the orthognathic surgery.

Primary culture of AM cells was performed as previously described (31). Briefly, the specimen was diced into pieces at an approximate size of 1 mm³ following removing the soft connective tissue, placed into plates coated with collagen I (Invitrogen; Thermo Fisher Scientific, Inc., Waltham, MA, USA) and incubated at 37°C in a 5% (v/v) CO₂ atmosphere for 5 h. Dulbecco's modified Eagle's medium (DMEM; Gibco; Thermo Fisher Scientific, Inc.) containing 15% (v/v) fetal bovine serum (FBS; Gibco; Thermo Fisher Scientific, Inc.) was added and used for subculture of the proliferative cells. The epithelial cells were purified and collected with a differential adhesion method.

Mouse bone marrow-derived monocyte/macrophage precursor cells (BMMs) were isolated as previously described with slight modifications (32). Briefly, bone marrow cells were collected from the femur and tibiae of 12 6-week-old female C57BL/6J mice (mean weight, 17.2 g). The mice were purchased from the Laboratory Animal Center of Sun Yat-sen University. Following washing, the cells were resuspended in α-minimum essential medium (Gibco; Thermo Fisher Scientific, Inc.) containing 10% FBS. Following 16 h of culture, the nonadherent cells were collected and incubated in M-CSF (25 ng/ml) at a density of 3×10⁵ cells/ml in flasks. The cells were used as BMMs following 3 days of culture. The study was performed in accordance with the Guidelines laid down by the National Institute of Health (Bethesda, MD, USA) in the USA regarding the care and use of animals for experimental procedures, and in accordance with local laws and regulations. Adequate measures were taken to minimize the pain or discomfort of the mice. The experiment was approved by the Institutional Animal Care and Use Committee of Sun Yat-sen University (IACUC-DB-2017-0605).

The AM cell line, hTERT-AM, was previously established by the authors' group (31). HS-5, a human bone marrow stromal cell line, was purchased from the American Type Culture Collection (Manassas, VA, USA). The cells were cultured in DMEM containing 10% FBS at 37°C in a 5% (v/v) CO₂ atmosphere.

The hTERT-AM cells and HS-5 cells were directly cocultured in 1:1 ratio or independently cultured at a density of 2.5×10⁵ cells/ml for 24 h. For indirect cocultures, HS-5 cells (1.5×10⁵ cells) were seeded in the lower chamber of a 24-tran-swell plate (0.4-mm pore size; Corning-Costar; Conig, Inc., Corning, NY, USA) and 100 µl medium containing 5×10⁴ AM cells was added to the upper chamber. Cells were cultured for 24 h. The culture media (CM) was collected and stored at -80°C until use. Cells were used for RNA extraction.

A commercial antibody-based protein microarray assay was performed using G-Series Human Bone Metabolism Array 1 (RayBiotech Life, Norcross, GA, USA). In brief, array glass slides were incubated in serum-free media from HS-5 cultures, hTERT-AM/HS-5 cocultures and hTERT-AM cultures (500 µg/ml) for 2 h. Following washing, the glass slides were incubated with a cocktail of 31 biotinylated antibodies and the remaining experimental procedure was carried out following the manufacturer's protocol. The intensity of the signal was normal-ized to that of the internal positive control.

The serum-free medium from cell cultures were collected and the level of TNF-α (cat. no. CSB-E04740h; Cusabio, Wuhan, China), IL-8 (cat. no. D8000C; R&D Systems, Inc.), and activin A (cat. no. DAC00B; R&D Systems, Inc.) were analyzed in accordance with the manufacturer's protocol.

Total RNA was extracted using TRIzol (Invitrogen; Thermo Fisher Scientific, Inc.) and reverse synthesized to cDNA using a PrimeScript RT Reagent kit (Takara Bio Inc., Otsu, Japan). The following thermal cycling steps were used: 37°C for 15 min, 85°C for 5 sec, and store at 4°C. GAPDH was used as internal control. The primers used in the present study are presented in Table SII. Amplification of the cDNA template was conducted using LightCycler 480 SYBR-Green I Master (Roche Diagnostics, Basel, Switzerland). PCR cycles were run as follows: 95°C for 5 min, 40 cycles of 95°C for 10 sec, 60°C for 20 sec and 72°C for 20 sec. Finally followed by 95°C for 5 sec and 65°C for 1 min. Following RT-PCR, the PCR products were separated on a 2% agarose gel and stained with a new type of DNA dye GoldView (cat. no. G8140; Beijing Solarbio Science & Technology Co., Ltd., Beijing, China), which is an ideal substitute for ethidium bromide. Expression of each gene was normalized to that of GAPDH as a loading control.

Total cellular proteins were extracted with Radioimmunoprecipitation assay lysis buffer (cat. no. P0013B; Beyotime Institute of Biotechnology, Shanghai, China) containing 10 mg/ml aprotinin, 10 mg/ml leupeptin and 100 mM PMSF. The bicinchoninic acid assay was used for protein quantification. The aliquots (containing 40 µg protein) were separated by 8% SDS-PAGE, transferred to polyvinyli-dene difluoride membrane. The membranes were incubated at 4°C overnight with antibodies against TRAP (cat. no. ab191406; Abcam; 1:1,000), NFATc-1 (cat. no. 8032; CST; 1:1,000), Cathepsin K (cat. no. ab187647; Abcam; 1:1,000), RANKL (cat. no. ab97864; Abcam; 1:1,000), JNK (cat. no. 9252; CST; 1:1,000) and p-JNK (cat. no. 4668; CST; 1:1,000). Anti-β-actin (cat. no. 4970; CST; 1:1,000) was used as a loading control. After washing, the membranes were incubated with IRDye^® 800CW goat anti-Rabbit IgG (cat. no. 926-32211; LI-COR Biosciences, Lincoln, NE, USA; 1:10,000) or Alexa Fluor^® 680 goat anti-rabbit IgG (cat. no. ab175773; Abcam; 1:10,000) for 1 h at room temperature. The target proteins were visual-ized and quantification was performed with the Odyssey Clx Infrared Imaging System (LI-COR Biosciences) and Image Studio 5.0 software (LI-COR Biosciences).

Isolated BMMs (3×10⁴ cells/well) were cultured in a 96-well plate in medium containing M-CSF (25 ng/ml), RANKL (50 ng/ml), IL-8 (100 ng/ml), activin A (50 ng/ml), the culture mediums of hTERT-AM cells, HS-5 cells and the coculture medium with or without corresponding antibodies for 5 days. The cultures were fed every 2 days with fresh medium. To detect osteoclast formation, the cells were subjected to TRAP staining using an Acid Phosphatase Leukocyte kit (Sigma-Aldrich; Merck KGaA). In addition, multinucleated TRAP-positive cells (≥3 nuclei) were counted as osteoclasts.

BMMs (3×10⁴ cells/well) were seeded onto 6 mm-diameter dentin slices in 96-well plates and cultured in medium containing M-CSF (25 ng/ml), RANKL (50 ng/ml) with or without activin A (50 ng/ml), or the CM from HS-5 cells, hTERT-AM cells and HS-5/hTERT-AM cocultures for 15 days. CM was changed every 2 days. The cells were lysed with 0.25% ammonia water and the slices were stained with toluidine blue for 4 min at room temperature. To evaluate bone resorption, the areas covering the pit were calculated by counting the number of mesh squares (100×100 µm) with a light microscope (magnification, ×100; Olympus Corporation, Tokyo, Japan).

Statistical analysis was conducted using SPSS version 23.0 software (IBM, Corps., Armonk, NY, USA). Data are presented as the mean ± standard deviation. Student's t-tests were used in two-group comparisons. One-way analysis of variance with multiple comparisons followed by Tukey's test was used for multiple groups comparisons. All tests were two-tailed and a value of P<0.05 was considered to indicate a statistically significant difference. Each experiment was performed in triplicate.

A total of two subtypes of AM were first tested for their ability to form osteolytic lesions and induce osteoclast formation. The histological examination results indicated a follicular and a plexiform AM, respectively (Fig. 1A). The sections were then stained with TRAP, a specific enzyme of osteoclasts. TRAP-positive osteoclasts were located adjacent to tumor cells, but were more commonly demonstrated at more distant sites from tumor cells, forming resorption pits. However, the normal bone demonstrated no TRAP-positive cells and a smooth cortical bone surface (Fig. 1B). Additionally, The AM tissues and primary AM cells as well as hTERT-AM cells were screened for the expression of osteoclast-activating factors including RANKL, M-CSF, IL-1α, IL-1β, TNF-α, IL-6 and IL-8 by RT-PCR. Due to different sources or conditions of the tissues and cells, these factors demonstrated various expression levels among the specimens; however, RANKL expression was only detected in AM tissues rather than in AM cells (Fig. 1C).

Mouse BMMs were treated with culture media from hTERT-AM. A slightly increased number of TRAP-positive cells was observed [Fig. 2A(c)], suggesting that the soluble factors released from AM cells have a limited potential to stimulate osteoclast formation. It was observed that osteoclasts were mostly located in the bone marrow far from the tumor cells, which prompted the investigation of whether cells in the bone microenvironment, particularly BMSCs, serve a role in osteoclastogenesis in AM. Therefore a cell-to-cell coculture of hTERT-AM cells, HS-5 cells and a human BMSC line was established and the coculture medium was added to the culture of mouse BMMs. Notably, the average number of TRAP-positive cells was identified to be significantly increased when mouse BMMs were treated with culture medium from the coculture [P<0.05; Fig. 2A(d) and B]. However, BMSCs plated in the absence of hTERT-AM cells did not induce osteoclast formation [Fig. 2A(b)]. A pit formation assay was further performed to investigate the potential of the multi-nucleated cells to resorb calcified tissue. The results demonstrated that the formation of resorption pits was seen on the dentin slices treated with hTERT-AM CM or HS-5/ hTERT-AM CM and the pit area of wells containing HS-5/hTERT-AM CM was ~4-fold greater than those seeded with hTERT-AM CM (Fig. 2C). Furthermore, mRNA expression of regulatory genes of osteoclast proliferation (PU.1), differentiation (NFATc-1), fusion (DC-STAMP and MITF) and function (Cathepsin K, integrin β3, and TRAP) were identified to be elevated in BMMs following treatment with HS-5/AM CM (Fig. 2D). These, in combination with an elevated protein level of NFATc-1, Cathepsin K and TRAP (Fig. 2E), which suggested that interactions between AM cells and BMSCs are essential for osteoclast differentiation and function.

A total of 31 cytokines and growth factors potentially involved in bone metabolism were screened for in the culture media of monoculture of hTERT-AM cells, HS-5 cells and coculture of the two cell lines using a cytokine array. The analysis indicated that certain cytokines were relatively abundant in the HS-5/ hTERT-AM cell coculture medium. Notably, the expression of IL-8 and TNF-α in the CM of hTERT-AM cells was significantly increased compared with HS-5 cells and a markedly elevated level of activin A was observed in the coculture medium of the two cell lines (Fig. 3A and Table SIII). These three factors were previously observed to be involved in the process of bone resorption (13,33,34).

To validate the findings of the cytokine array, IL-8 levels were quantified by ELISA in the 24-h serum-free culture media from the HS-5 cells, AM cells, or coculture of the two cell types. The results revealed that AM cells secreted increased amounts of IL-8 compared with HS-5 cells and a dramatic increase in the production of IL-8 protein was observed following direct or indirect co-culture of the two cell types (Fig. 3B). To define the source of IL-8 in the coculture, RT-qPCR was performed to compare IL-8 expression in AM cells and HS-5 cells cultured separately using a Transwell plate. The results revealed a significant increase in IL-8 expression in HS-5 cells following indirect coculture with hTERT-AM cells, whereas IL-8 expression in hTERT-AM cells remained nearly unchanged (P<0.05; Fig. 3C). These data indicate that AM cells stimulated the IL-8 secretion of BMSCs either though cell-cell contact or soluble factors.

In addition, activin A production was observed to be relatively high in HS-5 cells, but was nearly undetectable in AM cells. Only a direct interaction between AM and HS-5 cells could increase activin A production in the coculture medium (Fig. 3D). To investigate the source of activin A, RT-qPCR was used to examine the expression levels of inhibin-βA subunits, in which activin A is a dimer. The results indicated that the upregulated activin A was derived from BMSCs (Fig. 3E). These findings suggest that cell-to-cell contact was essential for triggering activin A secretion in BMSCs.

Furthermore, ELISA analysis verified increased production of TNF-α in AM cells compared with in HS-5 cells and that coculture of AM and HS-5 cells did not lead to alteration in TNF-α protein level (Fig. S1).

To investigate the role of AM/BMSCs CM-derived IL-8 and activin A in osteoclastogenesis, CM from hTERT-AM/HS-5 cocultures, with or without IL-8 or activin A neutralizing antibodies, were added to the culture of mouse BMMs in the presence of M-CSF. It was demonstrated that osteoclast formation provoked by the coculture medium was decreased either by neutralization of IL-8 or activin A (Fig. 4), which ascertained that the two proteins are involved in osteoclastogenesis induced by AM/BMSCs interactions and that IL-8 seems to serve a more important role in osteoclastogenesis in AM.

Studies have demonstrated that IL-8 production can be induced in response to multiple signals including IL-1α, IL-1β and TNF-α (30,35). In the present study, TNF-α was observed to be highly expressed in hTERT-AM cells (Fig. 3A), therefore anti-TNF-α antibody was added to the culture medium of hTERT-AM cells to examine the effect of TNF-α on IL-8 expression in HS-5 cells. The results demonstrated that neutralizing TNF-α significantly downregulated IL-8 expression in HS-5 cells induced by AM cells (P<0.05; Fig. 5A). ELISA analysis further indicated that the protein level of IL-8 in the coculture medium of the two cell lines was significantly inhibited by addition of the anti-TNF-α antibody (P<0.05; Fig. 5B). Additionally, it was also identified that IL-8 directly induced osteoclast formation in vitro (Fig. 5C and D), which was consistent with previous studies (36,37).

RT-qPCR was performed to examine whether AM-derived CM could induce RANKL expression in BMSCs. The results revealed that CM derived from hTERT-AM cells significantly increased RANKL expression in BMSCs (P<0.05). The expression of OPG, the decoy receptor of RANKL, was not altered (Fig. 6A). Additionally, adding OPG to the coculture medium slightly decreased osteoclast formation in mouse BMMs (Fig. 6B and C). Furthermore, it was identified that recombinant IL-8 induced RANKL level in BMSCs (Fig. 6D and E) and addition of anti-IL-8 antibody to AM-derived CM significantly decreased RANKL expression in BMSCs (P<0.05; Fig. 6E). These results indicate that hTERT-AM cells upregulated the expression of RANKL in HS-5 cells in an IL-8-dependent manner and RANKL further influenced osteoclast formation.

Previous studies demonstrated that activin A secretion was regulated by p38-dependent and JNK-dependent pathways, and that a highly conserved c-Jun-binding sequence was present in the INHβA promoter (38). Whether activation of the JNK pathway was associated with activin A production was therefore investigated by coculture of HS-5 cells and hTERT-AM cells. The results indicated that direct coculture of the two cell lines activated the JNK pathway, as phosphorylation of JNK was observed in HS-5 cells in the presence of fixed AM cells (Fig. 7A). It was also demonstrated that adding the specific JNK inhibitor SP600125 significantly reduced activin A secretion in the coculture medium (P<0.05; Fig. 7B). Furthermore, although activin A itself did not stimulate osteoclast formation, it did slightly support RANKL to induce osteoclastogenesis (Fig. 7C and D). Furthermore, the pit formation assay revealed that the area of resorption pits on the dentin slices treated with RANKL and activin A was about 1.3-fold greater than those seeded with RANKL alone (Fig. 7E). This suggests that activin A acts as a cofactor of RANKL to induce osteoclast formation and function and has a more effective triggering function of osteoclasts.