KCOT tissues derived from BCNS patient NS11 or from a sporadic KCOT from a 25-year-old Japanese woman were kindly donated to our department. Patient NS11 was diagnosed as BCNS and a point mutation in intron 3 (c.584+2T>G) of PTCH1 was detected in this patient (10). The parents of the patients gave their written informed consent for participation in this study. This study was approved by the Medical Ethics Board of Hyogo College of Medicine (No. 108), and is in accordance with the principles expressed in the Declaration of Helsinki.

KCOT cells (iKCOT1 and sKCOT1) were maintained in Epilife medium supplemented with HKGS (both from Cascade Biologics, Portland, OR, USA) or F-medium (16). Lentiviral vectors, CSII-CMV-cyclin D1, CSII-CMV-CDK4R24C and CSII-CMV-TERT were used to immortalize KCOT cells. Production and infection of recombinant lentiviruses with the vesicular stomatitis virus G glycoprotein (VSV-G) were described previously (17). The lentivirus backbone vector, CSII-MV-RfA, was a gift from Dr Miyoshi (Riken, BRC). After several passages, established cell lines were designated as iKCOT1 and sKCOT1.

To analyze the ability to form 3-D structures, 3-D life hydrogel (Cellendes GmbH, Reutlingen, Germany) was used. Cells were incubated in the Hydrogel in the presence of RGD-peptides for 12 days according to the manufacturer's MDCK cyst formation protocol. Cells were fixed in 4% formaldehyde in phosphate-buffered saline [PBS(−)]. After permeabilization, cells were stained with rhodamine phalloidin (Cytoskeleton, Denver, CO, USA) and mounted using Vectashield containing DAPI (Vector Laboratories, Burlingame, CA, USA). Imaging was carried out using a Zeiss LS780 confocal microscope.

Genomic DNAs from KCOT cell lines were extracted using the DNeasy Blood and Tissue kit (Qiagen, Hilden, Germany). All coding exons and intron-exon boundaries were amplified by PCR and sequenced as described previously (10).

For analysis of mRNA, total RNA was extracted from a sKCOT1 cell line using the QIAamp RNA Blood mini kit (Qiagen) and subjected to RT-PCR using a Cells-to-cDNA^® II kit (Applied Biosystems, Foster City, CA, USA) with random primers. The forward primer for exon 7 was 5′-GAATGGTGGATGTCATGGCTTATCC-3′, and the reverse primer for exon 19 was 5′-ACGGCACTGAGCTTGATTCCGATGA-3′. Amplified products were purified and cloned into pCRII (Thermo Fisher Scientific, Waltham, MA, USA). Eleven randomly selected clones were sequenced using M13 forward or reverse primers.

For immunofluorescence staining, cells were plated in 8-well culture slides (BD Bioscience, Bedford, MA, USA) and half of the medium in each well was changed to a high calcium medium (1.23 mM; 1:1 mix of F medium and DMEM supplemented with 5% FCS). Medium was changed every 3 days. Ten to twelve days later, cells were fixed in 4% paraformaldehyde in PBS(−) and permeabilized with 0.1% Tween-20 in PBS(−). After blocking in PBS(−) with 2% BSA, cells were subjected to immunofluorescence staining. Antibodies were visualized with Alexa-488 or Alexa-546 secondary antibodies (Thermo Fisher Scientific). Rhodamine phalloidin (Cytoskeleton) was used to detect actin filaments. DNA was stained with DAPI (Vector Laboratories). Imaging was carried out using a Zeiss LS780 confocal microscope.

For western blotting, iKCOT1 and sKCOT1 were harvested or incubated with low (0.65 mM) or high (1.23 mM) calcium medium for 10–12 days. Gel electrophoresis and western blotting were performed as described previously (18). Signals were detected by chemiluminescence using a Pierce SuperSignal western blotting kit (Thermo Fisher Scientific). All antibodies used are listed in Table I.

We initially tried to culture cells from the KCOT of the patient with BCNS using standard culture methods for normal keratinocytes of the oral mucosa. In primary culture, KCOT cell colonies grew slowly, and predominantly consisted of closely packed polygonal cells with epithelial morphology. KCOT cells could be subcultured 3 to 4 times, but the polygonal cells were replaced by larger flattened cells and subsequently became senescent and could no longer be cultured. Therefore, we attempted to immortalize primary human KCOT cells by introducing CDK4R24C (inhibitor-resistant form of CDK4), cyclinD1 (which can inactivate Rb by phosphorylation) and hTERT genes using lentiviral vectors. The combination of CDK4R24C, cyclin D1 and hTERT resulted in efficient prolonged proliferation of the cells and they were designated as iKCOT1, and they were cultured for >50 passages. To our knowledge, this is the first study of the successful establishment of immortalized KCOT cells derived from a patient with BCNS. In addition, we also immortalized sporadic KCOT cells by the same method and designated this line as sKCOT1.

We first analyzed the status of PTCH1 in both cell lines. Sequence analysis of iKCOT1 cells showed a germline c.584+2T>G mutation (10) and a wild-type PTCH1 sequence indicating no evidence of loss of heterozygosity (LOH) at least in the coding region. RT-PCR analysis indicated that mutated and wild-type alleles were equally expressed in this cell line (10). These data suggested that the wild-type allele is expressed in iKCOT1 cells. sKCOT1 cells showed two mutations in the PTCH1 gene: the c1120G>T mutation in exon 8 is a nonsense mutation and the c3064_3065delAT mutation in exon 18 is a frameshift mutation (Fig. 1A). Both mutations were designated as loss of function mutations; therefore, we analyzed mRNA from sKCOT1 cells to address whether these mutations were on the same allele. After purification of mRNA from sKCOT1 cells, RT-PCR was performed using a primer set that covers exon 8 to 18 of the PTCH1 gene. Obtained fragments were cloned into pCRII-TOPO (Thermo Fisher Scientific) and 11 randomly selected independent clones were sequenced. Three out of 11 clones contained only the exon 8 c1120G>T mutation and eight out of 11 clones had exon 18 c3064_3065delAT mutation only (Fig. 1A). These results indicated that each PTCH1 allele carried a single, different lethal mutation in sKCOT1 cells. Sequence analysis of the blood sample from the patient from which sKCOT1 cells were derived showed no mutations in the PTCH1 gene indicating that these mutations were sporadic and that the patient was not BCNS. In addition, we sequenced the coding regions of the PTCH2, SMO and SUFU genes of both KCOT cells, but did not detect any mutation (data not shown).

To characterize the nature of the KCOT cell lines, immunostaining was performed. In both cell lines in F medium, the stem cell markers BMI1, SOX2 and CD44 and the mesenchymal cell markers SNAI2, CDH2 and VIM, were observed. Furthermore, both cell lines expressed NEFL, which is characteristic of neuronal cells (Fig. 1B).

KCOT cells retain several characteristics of keratinocytes; therefore, we tested the response of the cell lines to calcium treatment by increasing calcium concentration of the culture media from 0.65 to 1.23 mM. Calcium-treated sKCOT1 cells stratified and parakeratinized, and iKCOT1 cells piled up and formed nodule-like structures by day 10 (Fig. 2A). Fig. 2B shows CDH1 and actin staining at 10 days of KCOT cells in low or high calcium medium. High expression of CDH1 was observed especially in nodule-like structure of iKCOT1 cells (Fig. 2B). When cells were cultured in dishes with high calcium medium for 12 days, sKCOT1 cells were stratified with the surface cells being of bigger cell size and with bigger nuclei compared with the bottom cells. Fig. 2C shows sequential images of CDH1 (green) and DAPI (magenta) stained sKCOT1 cells. The upper cells, predicted to be differentiated cells, expressed CDH1 and parakeratinized. Fig. 3 shows sequential images of sKCOT1 and iKCOT1 cells in 3-D hydrogels in low calcium concentration. Under low calcium conditions, sKCOT1 cells formed spheroids with nuclei at the center that were small and condensed, and lacked cell membranes (Fig. 3A). Conversely, iKCOT1 cells formed spheroids with normal nuclei and cell membranes (Fig. 3B). However, under high calcium conditions, some nuclei were condensed and lacked cell membranes, and were located at the center of the iKCOT1 spheroids similar to sKCOT1. These data indicated that both cell lines showed cyst-like structures in 3-D hydrogels with different calcium concentrations.

Because calcium treatment induced dramatic morphological changes to the KCOT cells, we examined protein levels by western blotting (Fig. 4). In both cell lines, CDH1, CLDN1, KRT10 and IVL were induced by calcium. However, levels of CDH2, VIM, TJP1, SNAI2, CD44 and BMI1 were reduced. CTNNB1 expression was reduced in sKCOT1 cells but was not changed in iKCOT1 cells. Although an increase of calcium concentration induced these changes in protein expression, the expression of some proteins changed with respect to time. Therefore, other factors, such as cell-to-cell contact or cell proliferation, may also be involved in changes in protein expression. NEFL expression was not influenced by calcium.

iKCOT1 from a BCNS patient and sKCOT1 from a non-BCNS sporadic patient have mutations in PTCH1. iKCOT1 cells have a germline mutation of c.584+2T>G (10) in one allele but we found no additional mutations on the other allele. RT-PCR analysis showed the existence of normal PTCH1 transcripts indicating that one of the PTCH1 alleles appears to be wild-type.

Tumors occurring in Japanese BCNS patients are usually KCOTs, and BCCs are sometimes observed in older patients (10). BCC occurs at a low rate in African-American patients, but at a high rate (up to 97%) in Caucasian patients with BCNS (6). Differences in type and onset of symptoms in BCNS patients from different ethnic populations clearly demonstrate the participation of other factor(s) besides PTCH1 in BCNS phenotypes. Data from BCNS patients with BCC support Knudson's two-hit hypothesis (22,23), a theory of oncogenesis stipulating that normal cells require two mutagenic hits (two distinct episodes of DNA damage) to generate cancer. In BCNS, various benign and malignant tumors, such as BCC, KCOT, medulloblastoma and ovarian fibromas, exhibit loss of heterozygosity (LOH) (3,24). Pan et al reported that not all KCOTs in BCNS-derived as well as sporadic KCOT had two PTCH1 mutations (25). Some KCOTs, as well as the sKCOT1 cell line, carry loss of function mutations in both PTCH1 alleles. Previously, we reported PTCH1 mutations in both alleles of a BCNS-derived ovarian fibroma case (20). Therefore, loss of PTCH1 functions itself does not confirm tumor malignancy. Further genome-wide analysis of KCOTs will elucidate additional mutations or alterations that are involved in the malignancy of the tumor. Retinoblastoma is a malignant eye cancer that is common in childhood. It is known that loss of function of both copies of the tumor suppressor RB1 gene induces retinoblastoma. However, why depletion of the RB1 gene does not induce other types of cancers is not well understood. Xu et al showed that only cone precursor cells are sensitive to the Rb defect in mice and that a cell type-specific gene network is critical for the formation of a tumor (26). In BCNS patients or Ptch^+/− mice, PTCH1 function is halved throughout the body; however, only specific cells or tissues show tumor formation. These data indicated that in addition to the depletion of PTCH1 function, cell type-specific gene networks may specify the sensitivity of cells and the development of tumors. Next-generation sequencing will elucidate the genes involved in tumor formation and gene expression networks that are essential for cell type restricted tumor formation.

KCOT cell characteristics were examined by immunofluorescence and western blot analyses (Figs. 1B and 4). Both cell lines expressed stem cell markers, such as CD44, BMI1 and SOX2, and mesenchymal cell markers, including CDH2, VIM and SNAI2. However, levels of these proteins were decreased under high calcium conditions. Conversely, levels of epithelial cell markers, CDH1 and CLDN1, and also the keratinocyte markers, KRT10 and IVL were dramatically increased (Fig. 4). These data suggested that both cell lines have a mesenchymal stem cell-like character, and high calcium conditions induced differentiation of epithelial characteristics. Grachtchouk et al showed that disrupting normal Hh signaling at the epithelial cell rests of Malassez (ERM) induced KCOTs (14). Gao et al induced KCOTs from ERM by disrupting TGF-β signaling (27). ERM are odontogenic epithelial cells in the periodontal ligaments and are normally quiescent and predicted to work during periodontal regeneration (28,29). ERM are known to express several proteins that are usually associated with mesenchymal cells (28). Therefore, ERM are a candidate for the origin of KCOTs. Recent analysis reveals that ERM have stem cell-like properties with capacities to undergo epithelial-mesenchymal transition (EMT) and to differentiate into mesenchymal lineages (29). Cells in which proliferation or differentiation activity is activated or repressed by abnormal Hh or TGF-β signaling in ERM may be an origin of KCOTs. Therefore, ERM-derived KCOT cells may also have stem cell and mesenchymal properties. Alternatively, KCOT may be a result of EMT of ERM cells and a high calcium concentration induces reverse transition of KCOT cells from mesenchymal cells to epithelial-like cells. One interesting feature of these cell lines was the expression of NEFL, which is characteristic for neuronal cells (Fig. 4). This may be due to the mesenchymal stem cell-like characteristics of these cell lines. Recently, genome-wide expression analysis of fresh Ameloblastoma and KCOT samples were carried out, the KCOTs are distinguished by elevated levels of expression of epithelial differentiation markers, and stem cell marker SOX2 (30). These expression profiles partially support our KCOT cell lines as retaining original KCOT characteristics.

In conclusion, we have established immortalized KCOT cell lines, and have developed an in vitro culture model of KCOT. Our experimental models may facilitate further studies to understand the genesis of syndromic and sporadic KCOTs and provide a unique resource to elucidate unsolved questions.