Activation of Wnt signalling reduces the population of cancer stem cells in ameloblastoma

Abstract Objectives The treatment of ameloblastoma, an odontogenic epithelial tumour destroying jawbone, mainly depends on radical destructive resections. Other therapeutic options are limited by the characteristics of ameloblastoma, such as high recurrence rates and resistance to radiation and chemotherapy, which implies possible existence of cancer stem cells (CSCs) in ameloblastoma. Here, we identified a putative CSC population in immortalized and primary human ameloblastoma cells and examined possible therapeutic reagents to reduce the CSC population. Methods We investigated subpopulations of AM‐1 cell line and human ameloblastoma cells using immunocytochemistry and flow cytometry and the effects of Wnt signalling activators on the 2‐ and 3‐dimensional cultured ameloblastoma cells using molecular biological analyses. Result Among heterogenous ameloblastoma cells, small‐sized and round‐shaped cells were found to be proliferative and expressed a marker of dental epithelial stem cells, SRY‐box 2 (Sox2). Exogenous activation of Wnt signalling using glycogen synthase kinase 3β inhibitors, lithium chloride (LiCl) and valproic acid (VPA), increased the cell size and decreased proliferation of cells and expression of Sox2 in 2 dimensionally cultured AM‐1 and human primary ameloblastoma cells. Furthermore, the growth of 3 dimensionally cultured AM‐1 cells as suspended or embedded in gel was suppressed by treatment with Wnt signalling activators, VPA and CHIR99021, or antibodies to sclerostin, an antagonist of Wnt signalling. Conclusion We suggest that Wnt signalling activators are potential drug candidates to suppress CSCs in ameloblastoma.

responsible for metastasis and recurrence in cancer patients, which can lead to chronic and incurable cancers. Therefore, elimination of CSCs is an important goal for treatment of cancer. 2 Ameloblastoma is the most common odontogenic tumour, accounting for 1% of all tumours of the head and neck region and around 11% of all odontogenic tumours. Classification of ameloblastomas has currently been established, which includes three types-conventional, unicystic and extraosseous/peripheral. 3 Its biological behaviour is considered to be more aggressive due to its higher incidence of recurrence 4 ; consequently, the current standard treatment is wide resection with appropriate margins and immediate reconstruction, which is associated with significant patient morbidity. 5 Although radiotherapy has been attempted for decreasing the recurrence rate, its efficacy is not clear. 5 A systemic genomic analysis showed that over 80% of ameloblastomas harbour oncogenic mutations in the sonic hedgehog (SHH) and mitogen-activated protein kinase (MAPK) pathways. 6 AM-1 is an ameloblastoma cell line that is immortalized using human papillomavirus type-16. 7 The AM-1 cell line exhibits characteristics that are similar to in situ ameloblastoma cells related to marker expression and invasive properties, which shows that this cell line is an appropriate model system to study ameloblastoma. [7][8][9][10] Studies using this cell line revealed that Akt, MAPK and SHH signalling pathways are related to proliferation and apoptosis of AM-1. [11][12][13][14][15] Interestingly, the Wnt pathway, an osteogenic signalling pathway, is suppressed in this cell line. 16,17 Early progeny of Sox2-positive dental epithelial stem cells (DESCs) are known to transiently express a Wnt inhibitor, secreted frizzled-related protein (Sfrp) 5. 18 AM-1 cells also express the Wnt antagonist Sfrp2, 16 and osteogenic genes related to Wnt signalling are suppressed in this cell line. 17 A recent study showed that Wnt signalling is important for enamel formation by facilitating ameloblast differentiation and movement. 19 In this study, we demonstrated the presence of a putative CSC population in AM-1, a well-established human ameloblastoma cell line.
Immunocytochemistry and flow cytometry of AM-1 cells showed that small-sized and round-shaped cells were proliferative and expressed a marker of DESCs, SRY-related HMG box 2 (Sox2). 18,20,21 Interestingly, Sox2 expression in the cells was negatively correlated with activation of Wnt signalling. We examined the effect of various exogenous Wnt activators in 2 dimensionally (2D) or 3 dimensionally (3D) cultured AM-

cells and human primary ameloblastoma cells. These Wnt activators
showed an inhibitory effect in growth of ameloblastoma cells.

| Cell culture, 3-dimensional culture and cell sheet generation
AM-1 cells were cultured in keratinocyte growth medium supplemented with pituitary extract (Gibco, Grand Island, NY, 17005-042) at 37°C and 5% CO 2 in a humidified incubator. For spheroid formation, 1.0 × 10 5 or 2 × 10 5 cells were plated onto Ultra-Low Attachment Surface Costar 6 Well Plates (Corning Inc, Corning, NY, 4371) with keratinocyte growth medium or Dulbecco's Modified Eagle's Medium (DMEM, Gibco, 11995-065) supplemented with 10% foetal bovine serum (FBS, Gibco, 12484-020) and 1% penicillin/streptomycin (Gibco, 15140-112) solution at 37°C for 1 week. For 3-dimensional culture in gels, 1.0 × 10 3 cells were suspended in 40 μl of gels and spot into one well of the pre-warmed 24 well plate. The gels were solidified in 37°C incubator and then keratinocyte growth media supplemented with 0, 0.6 or 1.2 mmol/L calcium chloride. For cell sheet generation, AM-1 cells were plated onto temperature-responsive dishes (Nunc UpCell 3.5 cm dish, ThermoFisher Scientific, Somerset, NJ, NUN-174904) and cultured with keratinocyte growth medium until fully confluent. The confluent cells were detached in the form of a cell sheet as described in the manufacturer's instructions.

| Realtime PCR analysis
Total RNA of AM-1 cells cultured with conventional cell culture methods or cultured as spheroids were extracted using Trizol™ rea-

| Flow cytometry
AM-1 cells detached using trypsin-EDTA were centrifuged and resuspended in keratinocyte growth medium. Light scattering characteristics of cells were analysed using forward scattered light (FSC) and side-scattered light (SSC), and based on the characteristics, cells were sorted into two subpopulations using a BD FACSAria III cell sorter (BD Biosciences).
Mice had access to food and water ad libitum. For orthotopic grafts, the upper first molars of 8-week-old nude mice were extracted, and a hole was prepared using a portable drill with a 0.75 mm tip in the extraction site under deep anaesthesia. Subsequently, a properly sized AM-1 cell sheet was grafted into the hole using forceps. AM-1 cell sheet-grafted mice were housed for 1 week for healing and subsequently sacrificed with CO 2 for histological and immunohistological analyses.

| A putative CSC population in AM-1 cells
As previously described, 7 AM-1 cells exhibited a heterogeneous morphology-from small round cells to large flattened cells. The average cell size was 60.6 ± 32.4 μm 2 , and the largest cell was 12 times larger than the smallest one. To detect a possible CSC population inside these heterogeneous cells, Sox2, a marker of DESCs, 18,20,21 was stained and visualized ( Figure 1A, red). Sox2 expression significantly differed among the cells and was especially strong in small round cells ( Figure 1A, arrows). Quantification of the results revealed a negative correlation between Sox2 expression and the cell size of AM-1 cells ( Figure 1B, Pearson's correlation coefficient r = −0.616). Additionally, the expression of Ki67, a marker of cell proliferation, showed a wide variation among the cells ( Figure 1C) and was also negatively correlated with cell size in AM-1 cells ( Figure 1D r = −0.691).
Wnt signalling, an essential signalling pathway for ameloblast differentiation, is known to be suppressed in DESCs and AM-1 cells. [17][18][19] Staining of β-catenin, an effector molecule of Wnt signalling, revealed that it was localized differently in AM-1 cells; some cells displayed the presence of nuclear β-catenin, which indicates activation of Wnt signalling, but cytosolic β-catenin was observed in other cells ( Figure 1E). Quantitative analysis showed that the nu-  Figure 2D). We classified AM-1 cells into four types ( Figure 2E) according to size (small or large), shape (round or flat), Sox2 expression (Sox2 high or Sox2 low ) and nuclear accumulation of β-catenin (nuclear β-cat high or nuclear β-cat low ). Upon treatment with LiCl, type I (small, round, Sox2 high and nuclear β-cat low ) and type II (small, dented, round, Sox2 high and nuclear β-cat low ) cells increased, whereas type III (small, flat, Sox2 low and nuclear β-cat low ) and type IV cells (large, flat, Sox2 low and nuclear β-cat high ) decreased  Figure 3A,B). The transcription of Axin2, a target gene of Wnt signalling, also increased in dose-dependent manner by VPA treatment ( Figure S2). Expression of Bcl11b and Sox2 decreased upon VPA treatment ( Figure 3C,D). An opposite effect was observed in cells treated with basic fibroblast growth factor (bFGF), a mitogenic factor that stimulates ameloblastoma proliferation 15 ; the number of small, round, Sox2 high cells increased upon bFGF treatment ( Figure 3E, arrows). These results show that small, round, Sox2 high cells respond to their surrounding microenvironment, which is one of the basic features of CSCs. 2

| In vitro spheroid-forming capacity of AM-1 cells
In vitro spheroid-forming assay is a well-established method for demonstrating self-renewal capacity of stem cells from various organs. 22 We plated various numbers of AM-1 cells on low attachment surface cell culture plates with several different culture media to optimize spheroid-forming conditions ( Figure S3A). No spheroid formation was observed when the seeded number of cells was 2 × 10 5 per well in a 6-well plate with keratinocyte growth medium ( Figure S3A). However, the same number of cells grown in DMEM displayed spheroid formation ( Figure S3A). Interestingly, the dissected spheroids showed a similar structure to that of ameloblastoma; the hyperchromatic outer shell implied the presence of peripheral palisading cells at the basal layer of ameloblastoma, and the eosinophilic spots inside the spheroids were similar to keratin pearls, typical structures found in acanthomatous ameloblastoma ( Figure S3B). Real-time PCR showed that the expression of stem cell markers (OCT3/4, Sox2 and CD49f) and an anti-apoptotic marker (Bcl11b) increased in cells that were three dimensionally (3D) cultured as spheroids compared with those cultured in a conventional cell culture system ( Figure S3C-F).

| In vivo tumour-forming capacity of AM-1 cells
In vivo tumour-forming capacity of AM-1 cells was assessed by orthotopic grafts of ameloblastoma cells ( Figure S4A-D). A mass of AM-1 cell sheet was implanted into a hole drilled at the extraction site of the maxillary first molar of 8 week-old BALB/c nude mice ( Figure S4A,B). After a week, complete closure of the extraction site was observed ( Figure S4B

| Effect of a Wnt signalling activator on suspension-cultured AM-1 cells
To investigate the effect of Wnt signalling on the formation of AM-1 spheroids, we activated Wnt signalling in the spheroids by treatment with VPA. The sectioned spheroids showed a dosedependent decrease in size ( Figure 5A,B). Immunostaining and immunoblotting confirmed an increase in β-catenin by VPA treatment of the spheroids ( Figure 5C,D). With an increase in β-catenin, the expression of a proliferation marker, Ki67, and a stem cell marker, Sox2, was decreased ( Figure 5C-H). However, cleaved Caspase 3, a marker of apoptosis, was not changed by VPA treatment ( Figure S6).
The SOST antibody also reduced the size of AM-1 spheroids ( Figure 6A,B). The β-catenin level was increased, and the number of Ki67-positive cells decreased in the spheroids after treatment with the antibody (Figure 6C-F). Notably, the expression of ameloblastin, a marker of ameloblast differentiation, was increased by the antibody ( Figure 6G,H). However, no significant effect of the antibody was observed on the apoptosis of AM-1 spheroids ( Figure S7). F I G U R E 1 AM-1 harbours a cancer stem cell-like population. A-F, AM-1 cells were subjected to immunocytochemistry (ICC) using antibodies to Sox2 (A, E), Ki67 (B), or β-catenin (E). Small round cells strongly expressing Sox2 are indicated by white arrows (A). The cell size (B, D), Sox2 intensity (B, F), or ratio of nuclear to cytosolic β-catenin (F) of each cell were measured and are displayed as a dot plot. A trend line (dotted line) shows the relationship between the measured factors (B, D and F). Cytoskeletons (A and C) and nuclei (A, C and E) were visualized using Alexa 488-conjugated phalloidin (green) and DAPI (blue), respectively. Scale bar = 20 μm. G-J, Scattered light characteristics of AM-1 cells based on forward scattered light (FSC) and side scattered light (SSC). AM-1 cells were sorted into two groups (P1 and P2) based on FSC and SSC (G) and cultured for 1 (H, I) or 3 d (J). Three dimensionally reconstructed image (H, I) and conventional confocal image (J). The cells were subjected to ICC using an antibody to Cytokeratin 14 and Sox2 (H-J). Cytoskeletons and nuclei (H-J) were visualized using fluorophore-conjugated phalloidin and DAPI (blue), respectively. Scale bar = 100 μm. K-L, Total RNA of AM-1 cells in P1 and P2 were extracted and subjected to real-time PCR analyses using primer sets of CD49f (K) and Sox2 (L). RQ, relative quantity. n = 3 F I G U R E 2 Treatment with lithium chloride increased the cell size and decreased proliferation and expression of stem cell markers of AM-1 cells. A-F, AM-1 cells were cultured for 24 h with the indicated dose of lithium chloride (LiCl). The cells were subjected to ICC using antibodies to Ki67 (A), β-catenin, or Sox2 (D). Cytoskeletons (A) and nuclei (A, D) were visualized using Alexa 488-conjugated phalloidin (green) and DAPI (blue), respectively. Distribution of cell size (B) and ratio of cell populations classified by Ki67 expression pattern (C; none, partially, or fully covered nucleus with Ki67) or by size, shape, Sox2 expression and nuclear accumulation of β-catenin (E and F, class I-IV) were analysed and are displayed as a graph (B, C and F). The average cell size in each group is indicated by a red bar (B). Scale bar = 20 μm F I G U R E 3 Treatment with valproic acid increased the cell size and decreased proliferation and expression of stem cell markers of AM-1 cells. A-E, AM-1 cells were cultured for 24 h with 1 mmol/L valproic acid or 100 mg/ml basic fibroblast growth factor (bFGF). The cells were subjected to ICC using antibodies to β-catenin (A) and Sox2 (E), and immunoblotting using antibodies to β-catenin or α-tubulin (B). Cytoskeletons and nuclei were visualized using Alexa 488-conjugated phalloidin (green) and DAPI (blue), respectively (A, E). Scale bar = 20 μm. Total RNA of the cells was extracted and subjected to real-time PCR analyses using primer sets of Bcl11b (C) and Sox2 (D). n = 3. VPA, valproic acid. RQ, relative quantity

| Effect of a Wnt signalling activator on 3Dcultured AM-1 cells in gels
Recent advances in 3D cell culture have enabled the development of more physiological in vitro models using cancer cells. 24 First, we tried 3D culture of AM-1 cells in Matrigel, a well-defined soluble basement membrane used widely. 25 Initially, we cultured AM-1 cells in 3D Matrigel using a keratinocyte medium. The cells were observed to proliferate in Matrigel; however, they formed loosely connected cell masses ( Figure 7A). When Ca 2+ was supplied to the medium, AM-1 cells adhered to each other to form tightly compacted cell spheroids ( Figure 7B). The spheroids showed a complex structure -a basal cell-like hematoxylinophilic outer region and keratinizing eosinophilic inner region were observed ( Figure 7B). In both cases, exogenous activation of Wnt signalling using CHIR99021, a GSK3β inhibitor, suppressed proliferation and F I G U R E 4 Treatment with VPA or α-SOST increased the cell size and decreased proliferation and expression of stem cell markers of AM-1 cells. A-F, AM-1 cells were cultured for 24 h with the indicated dose of VPA or α-SOST. The cells were subjected to ICC using antibodies to cytokeratin 14 (K14, green) and Ki67 (red). Nuclei were visualized using DAPI (blue), respectively. Cell size (B and E) and ratio of Ki67-positive cells per total cells (C and F) were analysed and are displayed as a graph. Scale bar = 100 μm, *p-value < 0.05, **p-value < 0.01. mass formation of the cells ( Figure 7C). Next, to provide an environment more similar to an in vivo microenvironment, we cultured the cells in collagen gel. Interestingly, a network-like growth pattern of AM-1 cells was observed in the collagen gel ( Figure 7D).
In addition, the cells formed more prominent and thicker networks at higher concentrations of Ca 2+ and collagen ( Figure 7E). Wnt  Figure 7F and Figure S8).

| D ISCUSS I ON
Radical surgery is the current mainstay of ameloblastoma treatment and includes an en bloc resection of 1-2 cm of the bone adjacent to the ameloblastoma, which usually requires immediate bone reconstruction for speech and swallowing. This reconstruction is not necessary for conservative surgery, such as simple enucleation or curettage of the bone margin. However, the rate of recurrence after these operations can be as high as 60%-90%. 5 With such high recurrence rates, the characteristics of ameloblastoma, such as slow-and long-term proliferation, resistance to radiation and chemotherapy, and complex histological structures within the tumours, strongly imply the possible existence of CSCs in ameloblastoma. However, very few pathological studies have reported stem cell marker staining in tumour tissue, 22,26 and the characteristics and roles of CSCs in tumorigenesis remain to be elucidated. In this study, we characterized a putative CSC population in AM-1 cells, a well-established human ameloblastoma cell line. We observed that the small round cell population of AM-1 cells was proliferative and expressed a stem cell marker. We also found that AM-1 cells possessed spheroid-forming capacity, which is an indicator for stemness of cells. In addition, the orthotopic graft of AM-1 cells formed a mass that exhibited keratinization. These findings may serve to pave the way to study CSC populations of ameloblastoma in vitro.
Wnt signalling is an osteogenic signalling pathway that promotes bone generation by activating osteogenesis and inhibiting osteoclastogenesis. 27 In tooth development, Wnt signalling is known to be activated in the late stage of development and facilitates ameloblast differentiation and movement. 19 Expression of Sfrp, an antagonist of Wnt signalling, is observed in Sox2-positive DESCs. 18 Interestingly, AM-1 cell line also expresses the Wnt antagonist Sfrp2, 16 and osteogenic genes related to Wnt signalling are suppressed in this cell line. 17 Here, we revealed that the Sox2-  (Figure 8). Recently, a sclerostin antibody was approved as a therapeutic agent against osteoporosis. 28 The antibody activates Wnt signalling by neutralizing sclerostin, a Wnt antagonist secreted by mature osteocytes, to prevent excessive osteogenesis. 29 A recent study revealed that treatment with the antibody stimulates mandibular bone formation. 23 In our study, treatment with sclerostin antibody activated Wnt signalling and suppressed the formation of AM-1 spheroids ( Figure 8). The effect of Wnt signalling activators was clearly F I G U R E 8 Schematic diagram for the effect of Wnt signalling activators on ameloblastoma. Exogenous activation of Wnt signalling using antibody (α-SOST antibody) or chemicals (LiCl, VPA, or CHIR9902) increases β-catenin (β-cat) in ameloblastoma, leading to an increase in β-cat high and Sox2cells and decrease in β-cat low and Sox2 + cells, the putative cancer stem cells. This change of subpopulation in ameloblastoma results in the suppression of spheroid-forming activity and invasiveness confirmed by human primary ameloblastoma cells. Taken together, we suggest that treatment with Wnt activators is a potential therapeutic option after resection of ameloblastoma, which promotes bone healing along with suppression of tumour recurrence. AM-1 cells showed anchorage-independent growth, which is a property of transformed cells. 30 Proliferation of AM-1 cells was observed in a suspension culture containing keratinocyte growth medium that had a low calcium concentration, 31

| CON CLUS ION
In conclusion, we identified putative CSC population in human immortalized and primary ameloblastoma cells. Furthermore, we showed that exogenous activation of Wnt signalling decreased the CSC population and suppressed their 2-and 3-dimensional growth of ameloblastoma cells. This finding suggests that the Wnt signalling activators could be developed as drugs suppressing CSCs in ameloblastoma, which could provide an alternative option to treat cancer without radical destructive surgery.