Histological specimens collected from January 2014 to December 2018 were retrieved from our hospital's archives according to the guidelines of the Japanese Society of Pathology. Informed consent was provided by each patient for the use of the clinical images. Two cases of BSCC and 20 cases of conventional SCC were collected from craniocervical lesions, and each case included lymph node metastasis. Specimens were obtained from the Department of Diagnostic Pathology, Wakayama Medical University. Another BSCC specimen was obtained from the Department of Pathology, Tsurumi University School of Dental Medicine. Diagnosis was performed by at least two pathologists. Clinical and pathological information is described below. Case 1 is a female patient, 81 years of age, with a clinically diagnosed mandibular malignant tumor; pathological diagnosis is BSCC with abundant squamous components. Case 2 is a male patient, 57 years of age, with a clinically diagnosed malignant tumor in the palatine tonsil; pathological diagnosis is BSCC with moderate squamous components and lymph node (LN) metastasis. Case 3 is a male patient, 68 years of age, with clinically diagnosed tongue cancer; pathological diagnosis is BSCC without squamous components. Case 4 is a female patient, 87 years of age, with clinically diagnosed gingival cancer; pathological diagnosis is well-differentiated SCC. Case 5 is a male patient, 78 years of age, with clinically diagnosed tongue cancer; pathological diagnosis is well-differentiated SCC and LN metastasis. Case 6 is a female patient, 74 years of age, with clinically diagnosed buccal cancer; pathological diagnosis is poorly differentiated SCC. The clinical information regarding the other SCC cases is documented in Table I. Partial resection and LN dissection specimens were used in cases 2 and 5, and biopsy specimens were used in the other cases.

Claudin-4, occludin, SOX2 and PCNA expression in BSCC and SCC tissues were evaluated from serial deparaffinized sections. Our specimens had been fixed with formalin from 24 to 48 h at room temperature and treated with routine processing as in a previous study (12). Ten 4-mm serial sections were prepared for staining and were incubated with primary antibodies for 2 h. Immunohistochemistry was performed using a Discovery Auto-Stainer with automated protocols (Ventana Medical Systems, Inc.; Roche) as previously described (13). The average number of PCNA-positive cells from case 1 to case 23 were counted in 10 random microscopic fields at 400 magnification. The SOX2 intensity was determined by qualitative assessment of three levels as weak, 1; moderate, 2; and strong, 3. We defined a diffuse staining pattern as that positively detected in almost all cancer cells but a dot-like staining pattern as that positively detected in partial and marginal cancer cells.

HSC2 and HSC3 human oral squamous cell carcinoma cells, which do not have basaloid characteristics were obtained from the Health Science Research Resources Bank (Osaka, Japan). These cells were cultured in Dulbecco's modified Eagle's medium (DMEM) (Sigma Chemical Co.; Merck KGaA) supplemented with 10% fetal bovine serum and 1% penicillin streptomycin antibiotics. The construct expression vector for SOX2 was used for transfection. The human SOX2 plasmid (26817) was purchased from Addgene. Transient plasmid transfection was performed as previously described (14).

HSC2 cells were lysed using M-PER lysis buffer (Thermo Fisher Scientific, Inc.). Protein determination was performed using the bicinchoninic method. A total of 40 µg protein was loaded in each lane. The total cell lysates were run on 14% SDS-polyacrylamide gels followed by western blotting using standard procedures. The proteins were transferred on to PVDF membrane. For blocking, membranes were incubated with 5% skim milk for 60 min at room temperature after protein transfer. A WesternBright Sirius kit (Advansta) was used for antibody detection, and an AE-9300 Ez capture MG (ATTO) was used for image data capture. We repeated the western blot analysis three times and the results were similar.

The following commercial antibodies were purchased: claudin-4 (1:100, mouse monoclonal, sc-376643; Santa Cruz Biotechnology Inc.), occludin (1:100, mouse monoclonal, sc-133255; Santa Cruz Biotechnology), SOX2 (1:100, rabbit polyclonal, Ab97959, Abcam), PCNA (1:1,000, mouse monoclonal, sc-56; Santa Cruz Biotechnology) and actin (1:10,000, mouse monoclonal, A5441; Sigma Chemical Co.; Merck KGaA).

We prepared three independent RNA samples from HSC3 cells. Total RNA was isolated and first-strand cDNA was synthesized as previously described (13). Real-time PCR was performed using SYBR Green Master Mix (Bio-Rad Laboratories, Inc.). The thermocycling conditions were the following: Initial denaturation at 95°C for 3 min, followed by 40 cycles of 95°C for 10 sec and 60°C for 30 sec. The 2^−∆∆Cq method was used for relative quantification (15). Amplification primer sequences were designed as follows: SOX2-F, 5′-GAATGCCTTCATGGTGTGGT-3′ and R, 5′-TTGCTGATCTCCGAGTTGTG-3′; claudin-4-F, 5′-ATGGCCTCCATGGGGCTACA-3′ and R, 5′-AGCGAGTCGTACACCTTGCA-3′; occludin-F, 5′-GCGGATTGGTTTATCTTGGA-3′ and R, 5′-CTGGATGACATGGCTGATTG-3′; PCNA-F, 5′-AGGTGTTGGAGGCACTCAAG-3′ and R, 5′-AGTCCATGCTCTGCAGGTTT-3′; and 18S rRNA-F, 5′-GCGCCGCTAGAGGTGAAAT-3′ and R, 5′-GAAAACATTCTTGGCAAATGCTT-3′. Data were normalized using 18S rRNA. qPCR was repeated three times and the results were similar.

HSC2 and HSC3 cells were seeded into 96-well plates. After transfection, 50 ml of XTT kit reaction solution was added to each well (XTT-based) (Biological Industries) and the cells were incubated at 37°C for an additional 2 h according to the manufacturer's instructions. Absorbance at optical density at 480 nm (OD₄₈₀) and at (OD₆₅₀) was measured using a 96-well microplate reader (SH-9000; Hitachi).

Statistical analysis with Dunnett's test was performed using JMP Pro software version 13.0 (SAS Institute Japan, Tokyo, Japan). The data are expressed as the mean value ± SE (bars) of three independent samples. *P<0.05 was considered to indicate a statistically significant result, and ANOVA was used followed by Dunnett's test.

Representative macroscopic and radioscopic aspects in case 1 of BSCC are shown in Fig. 1. Fig. 1A shows flat appearance of gingival mucosa in a right anterior lesion. Fig. 1B shows the huge mandibular swelling that was found. Diffuse swelling was found in the palatine tonsil or tongue in case 2 and 3, respectively (data not shown). An orthopantomograph of the BSCC at first medical examination showed a multilocular cyst surrounded by well-defined radiolucent borders (Fig. 1C, left panel). The patient was treated with radiotherapy and the orthopantomograph at 8 months after the first examination showed a decreased multilocular cyst with calcification (Fig. 1C, right panel). This indicated that radiotherapy was effective. The other two patients with BSCC were not treated with radiotherapy.

Fig. 2A shows representative histological hematoxylin and eosin (H&E)-stained images from case 1, 2 and 3 of BSCC. Fig. 2B shows representative histological images from case 4, 5 and 6 of SCC.

Representative images of claudin-4, occludin, SOX2 and PCNA immunoreactivities for case 1, 2 and 3 of BSCC are shown in Fig. 3. Claudin-4 immunoreactivity in case 1 was strongly detected in the cell membrane of squamous components, whereas no obvious detection was observed in cancer cells. Claudin-4 immunoreactivity in case 2 of the primary tumor was weakly detected in the membrane of squamous components, whereas no obvious detection was observed in the cancer cells. No obvious detection of claudin-4 immunoreactivity in case 2 was observed in cancer cells of LN metastasis. Claudin-4 immunoreactivity in case 3 was not detected in the cancer cells. Occludin immunoreactivity was similar to claudin-4 as above, but in case 1 it was weakly detected when compared to claudin-4. We defined diffuse staining patterns as positively detected in almost all cancer cells but dot-like staining patterns as positively detected in partial and marginal cancer cells. SOX2 in case 1 had a diffuse staining pattern in the nucleus, cytoplasm and in the membrane of squamous components. SOX2 immunoreactivity in cases 2 and 3 of the primary tumor and LN metastasis exhibited a diffuse staining pattern in the nucleus and some positive staining in squamous components was observed. The PCNA immunoreactivity of the primary tumor and LN metastasis in case 1, 2 and 3 also had diffuse staining in the nucleus, but there was no obvious detection observed in squamous components.

Representative images of claudin-4, occludin, SOX2 and PCNA immunoreactivities in SCC are shown in Fig. 4A. No positive immunoreactivities for claudin-4 and occludin in the primary tumor and LN metastasis from cases 4, 5 and 6 were detected in either cancer cells or squamous components. SOX2 immunoreactivities from case 4, 5 and 6 were dot-like staining patterns in the nucleus of cancer cells. No obvious detection was observed in squamous components. PCNA immunoreactivity from case 4, 5 and 6 of the primary tumor was similar to that of SOX2. There appeared to be weaker detection of PCNA immunoreactivity in case 6 compared with case 4 and 5. SOX2 intensities and PCNA-positive cells of BSCC and SCC cases are provided in Table I and Fig. 4B. The numbers of PCNA-positive cells were higher in BSCC compared to SCC. SOX2 intensities in BSCC were all strong, but they exhibited a variation in SCC.

SOX2 expression was strongly detected in all cases of BSCC. Therefore, it was ascertained whether SOX2 overexpression affects expression of claudin-4, occludin and PCNA using oral SCC cancer HSC2 cells. SOX2 overexpression had little effect on the protein expression of occludin and PCNA (Fig. 5A). Endogenous protein expression of claudin-4 was not detected in HSC2 cells, thus expression levels were further examined by qPCR in HSC3 cells. SOX2 overexpression had little effect on the mRNA expression of claudin-4, occludin and PCNA in HSC3 cells (Fig. 5B).

Abnormalities in SOX2 expression affect proliferation, migration and apoptosis in various types of cancer cells. It was thus ascertained whether SOX2 overexpression affects cell proliferation of oral cancer HSC2 and HSC3 cells. SOX2 overexpression had little effect on the cell proliferation of HSC2 and HSC3 cells at 24 and 48 h (Fig. 5C).