Two OSCC cell lines, SAS and OSC20, both derived from primary lesions of a patient with OSCC, were used in the experiment. SAS was purchased from the Health Science Research Resources Bank. OSC20 was donated by the Research Center for Innovative Cancer Therapy, Molecular Targeting Therapeutics Division, Kurume University, School of Medicine. SAS was grown in Dulbecco's modified Eagle's medium (DMEM; Nissui Seiyaku Co., Ltd., Tokyo, Japan) and Ham's F12 medium supplemented with heat-inactivated (56°C, 30 min) 5% fetal bovine serum (FBS; Bioserum, Victoria, Australia), 100 U/ml, penicillin and 100 μg streptomycin (Gibco-BRL/Life Technologies Inc., Gaitherburg, MD, USA). OSC20 was grown in Eagle's minimum essential medium (EMEM; Gibco, BRL/Life Technologies Inc.) with 5% FBS. Cells were cultured in an atmosphere of 5% CO₂ in air at 37°C.

SAS and OSC20 cells with 80% confluence were washed once with phosphate-buffered saline (PBS), detached with accutase (Innovative Cell Technologies, Inc., San Diego, CA, USA), suspended at 1×10⁶ cells/ml in PBS supplemented with 2% FBS, incubated with 10 μg/ml human IgG (R&D Systems, Inc., Minneapolis, MN, USA) for 15 min at room temperature, and then incubated with allophycocyanin (APC)-conjugated mouse anti-human CD44v3 (cat. no. FAB5088A; R&D Systems) combined with phycoerythrin (PE)-conjugated mouse anti-human CD24 (cat. no. 555428; BD Biosciences, San Jose, CA, USA) at 4°C for 45 min. Samples were washed, centrifuged at 500 × g for 3 min, resuspended in 2 ml cold PBS supplemented with 2% FBS, then 1 μg/ml propidium iodide (PI; BD Biosciences) was added and the cells were filtered through a 40-μm cell strainer (BD Biosciences). Analysis and separation were carried out with a FACSAria II (BD Biosciences).

A total of 2,500 cells of each of the four cell fractions, i.e., CD44v3⁺/CD24⁻, CD44v3⁺/CD24⁺, CD44v3⁻/CD24⁻ and CD44v3⁻/CD24⁺ cells isolated from two cell lines were plated in 96-well plates and cultured in a CO₂ incubator. The cells were harvested at 24, 48, 72 or 96 h and the proliferation was examined in colorimetric assays using 3-(4,5-dimethylthiazol-2yl-yl-)-2, 5-dimethyl tetrazolium bromide (MTT) cell growth assay kits (Chemicon, Temecula, CA, USA) as described elsewhere (24).

Four cell fractions isolated from two cell lines (5,000 cells/dish) were cultured in serum-free medium including 10 ng/ml epidermal growth factor (EGF; Sanko Junyaku Co., Ltd., Tokyo, Japan) and 20 ng/ml basic fibroblast growth factor (bFGF) (PeproTech, Rocky Hill, NJ, USA) using ultra-low attachment 6-well plates (Corning Inc., Corning, NY, USA) for 1 week. Sphere formation was assessed by counting the number of spheres (>3 cells) under a microscope (x200).

The four isolated cell fractions sorted from two cell lines were plated at 2,500 cells/well in 96-well plates, and the effect of CDDP (1 or 5 μM) (Nihon Kayaku, Tokyo, Japan), 5-fluorouracil (5-FU) (10 or 100 μM) (Kyowa Hakko, Tokyo, Japan), and cetuximab (100 or 1,000 nM) (Merck Serono Co., Ltd., Tokyo, Japan) was examined. Drug resistance was determined after treatment for 96 h by MTT assay.

Various numbers of cells (1×10³, 1×10⁴, or 1×10⁵) were injected subcutaneously into 4-week-old female non-obese diabetic/severe combined immunodeficiency (NOD/SCID) mice (n=5 in each group). Tumorigenic capacity was judged 8 weeks after injection. The animal procedures were approved by the Ethics Review Committee for Animal Experimentation of Kurume University School of Medicine.

After the four fractions were isolated, total RNA was extracted using an RNAqueous Micro kit (Life Technologies, Carisbad, CA, USA), and complementary DNA (cDNA) was synthesized using the Reverse Transcription System (Promega, Madison, WI, USA) according to the manufacturer's instructions. Quantitative real-time RT-PCR (RT-qPCR) was performed to examine the expression of CSC-LC property-related genes [e.g., ABC transporter genes (ABCB1 and ABCG2), anti-apoptosis genes (BCL2 and CFLAR), self-replication genes (Oct4 and Nanog) and hypoxia-related genes [hypoxia inducible factor 1α (HIF1α)] with ABI PRISM 7500 (Applied Biosystems, Foster City, CA, USA). Gene expression assays and primer and probe mixes were used for ABCB1, ABCG2, ALDH1A1, BCL2, CFLAR, Oct4, Nanog, HIF1α, and β-actin [assay IDs (Hs 00184500_m1, Hs01053790_m1, Hs00946916_m1, Hs00608023_m1, Hs00153439_m1, Hs03666771, Hs04260366, Hs00153153_m1, and Hs99999903_m1, respectively; Applied Biosystems)], and thermal conditions were as follows: initial incubation at 95°C for 10 min, then 40 cycles alternating in turn with 95°C for 10 sec, 60°C for 20 sec, and 72°C for 15 sec, and then maintained at 72°C for 10 min. Comparative gene expression analysis was performed using the 2^(−ΔΔCq) method with normalization to the level of internal control gene, β-actin.

OSCC tissue samples for immunohistochemistry were obtained from 50 patients who underwent surgical resection at Kurume University Hospital between 2007 and 2008. These specimens were fixed in 10% buffered formalin followed by paraffin embedment. None of the patients had previously received any treatments, including chemotherapy or radiotherapy. This cohort was composed of 33 men and 17 women aged from 48 to 87 years (median age, 70 years). The average observation time for overall survival was 52 months for patients still alive at the time of analysis, and ranged from 1 to 95 months.

All tissues were immunohistochemically examined for expression of CD44v3 and CD24 using mouse anti-human CD44v3 monoclonal antibody (VFF-327v3, 1:50 dilution; Novocastra, New Castle upon Tyne, UK) and mouse anti-human CD24 monoclonal antibody (528807, 1:100 dilution; R&D Systems). BenchMark XT (Ventana Medical Systems, Inc., Tucson, AZ, USA) was used for immunostaining.

All slides were evaluated by two of the authors (K.T. and J.A.). CD44v3 and CD24 expressions were evaluated according to staining intensity and staining area in the non-invasive portion and invasive portion within the tumor. The staining intensity was graded as 0, negative; 1, weakly positive; 2, moderately positive; or 3, strongly positive. The staining intensity in the normal epithelium was used as an internal control. The staining area was graded as 1, 0–20%; 2, 20–40%; 3, 40–60%; 4, 60–80%; and 5, 80–100%. The scores of staining intensity and area were calculated. The expression scores of CD44v3 and CD24 were compared between the non-invasive portion and invasive portion. The expression scores of CD44v3 and CD24 were categorized as either negative or positive. The median value of the expression score was used to separate the negative and positive groups. The relationship between isolated CD44v3 and isolated CD24 immunoexpression or CD44v3/CD24 immunophenotype and each clinicopathological feature of OSCC, such as T stage, nodal status, mode of invasion (25) and histological grade, was analyzed.

JMP software version 11.0 was used for all statistical analysis. Comparisons of cell growth assay, sphere forming assay, drug resistance assay, and quantitative real-time RT-PCR assay were normally distributed and assessed by the Shapiro-Wilk test, and then, a comparison of each in vitro assay except for the sphere forming assay was performed using one-way ANOVA with Dunnet post hoc comparisons. A comparison of the sphere forming assay was performed using one-way ANOVA with the Tukey post hoc comparisons. The expression score data of the invasive portion were compared with that of the non-invasive portion using a Mann-Whitney U test. Correlation of isolated CD44v3 and CD24 immunoexpression or the CD44v3/CD24 immunophenotypes with the clinicopathological parameters was assessed by standard Chi square tests or the Fisher's exact test. The overall survival rate was defined as the interval between the diagnosis and the date of death (uncensored data) or the date of the last available clinical information (censored data). Comparison and estimation of cumulative survival rates were performed using the Kaplan-Meier curves and the log rank test. All tests were two-sided and a value of P<0.05 was considered significant.

In SAS, the proportion of CD44v3⁺/CD24⁻, CD44v3⁺/CD24⁺, CD44v3⁻/CD24⁻ and CD44v3⁻/CD24⁺ cells was 10.7, 48.2, 7.8 and 33.3%, respectively. Whereas, the proportion of CD44v3⁺/CD24⁻, CD44v3⁺/CD24⁺, CD44v3⁻/CD24⁻ and CD44v3⁻/CD24⁺ cells was 24.1, 52.0, 11.2 and 12.7%, respectively (Fig. 1).

CD44v3⁺/CD24⁻ cells in SAS showed a significantly higher proliferative ability than that of CD44v3⁻/CD24⁻ and CD44v3⁻/CD24⁺ cells after culturing for 96 h (P<0.05), and, CD44v3⁺/CD24⁻ cells in OSC20 showed a significantly higher proliferative ability than that of CD44v3⁻/CD24⁻ cells (P<0.05) (Fig. 2A).

The sphere forming ability of CD44v3⁺/CD24⁻ cells in both SAS and OSC20 was significantly higher than the other fractions (P<0.05 or P<0.001). There was also a significant difference in sphere forming ability between the two fractions, for example, between CD44v3⁺/CD24⁺ and CD44v3⁻/CD24⁻, and between CD44v3⁺/CD24⁺ and CD44v3⁻/CD24⁺ cells (P<0.001) (Fig. 2B).

After 96-h treatment with CDDP, 5-FU, or cetuximab, the sensitivity to each drug was assessed with the MTT assay in SAS and OSC20. Cell growth was significantly suppressed in cells treated with CDDP, 5-FU and cetuximab, as compared with control cells in both SAS and OSC20 (Fig. 2C).

In SAS, CD44v3⁺/CD24⁻ cells had a significantly higher resistance in CDDP (1 μM) or 5-FU (10 or 100 μM) treatment than the other fractions (P<0.05 or P<0.001), and had a significantly higher resistance in CDDP (5 μM) or cetuximab (100 or 1,000 nM) treatment than CD44v3⁺/CD24⁺ and CD44v3⁻/CD24⁺ cells (P<0.001) (Fig. 2C).

In OSC20, CD44v3⁺/CD24⁻ cells had a significantly higher resistance in CDDP (1 or 5 μM) or 5-FU (100 μM) or cetuximab (100 nM) treatment than CD44v3⁺/CD24⁺ cells and/or CD44v3⁻/CD24⁺ cells (P<0.05, P<0.01 or P<0.001). There was no significant difference in sensitivity in 5-FU (10 μM) or cetuximab (1,000 nM) treatment among the cell fractions (Fig. 2C).

We performed RT-qPCR analysis to compare CSC-LC property related gene expression in CD44v3⁺/CD24⁻ cells and the other cell fractions in SAS and OSC20 cells. In SAS, CD44v3⁺/CD24⁻ cells showed significantly higher mRNA expression of transporter-related genes (ABCB1 and ABCG2), ALDH1A1, anti-apoptotic gene (BCL2), and self-replication genes (Oct-4, Nanog), than the other fractions (Fig. 3).

Injection of 1×10³ cells from all of the cell fractions produced no tumors in NOD/SCID mice. In contrast, two mice that received 1×10⁴ CD44v3⁺/CD24⁻ cells or CD44v3⁺/CD24⁺ cells developed tumors 8 weeks after the inoculation. In addition, injection of 1×10⁵ cells from all of the cell fractions produced tumors in NOD/SCID mice. The ratio of tumorigenicity among CD44v3⁺/CD24⁻ cell fractions and the other fractions in SAS was not significantly different (Table I).

CD44v3 and CD24 were predominantly expressed on the cell membrane and in the cytoplasm, respectively. A representative immunostaining photomicrograph of each grade is shown in Fig. 4. CD44v3 and CD24 immunoscoring of the non-invasive portion and invasive portion is shown in Fig. 5. A significant difference was observed between the expression scores of CD24 in the non-invasive portion and those in the invasive portion (P<0.01) (Fig. 5).

There was no significant correlation between isolated CD44v3 and isolated CD24 immunoexpression or CD44v3/CD24 immunophenotypes and each clinicopathological parameter in the non-invasive portion (data not shown). The association between CD44v3 and CD24 expression in the invasive portion and clinicopathological parameters is summarized in Table II. CD44v3 expression was significantly correlated with lymph node metastasis (P=0.039), but not with age, gender, T stage, mode of invasion and tumor differentiation. CD24 expression was significantly correlated with gender (P=0.01), but not with the other clinicopathological parameter. The association between CD44v3/CD24 immunophenotypes and clinicopathological parameters in 50 OSCC patients is summarized in Table III. When the 50 cases were subdivided into four groups i.e., CD44v3⁺/CD24⁻, CD44v3⁺/CD24⁺, CD44v3⁻/CD24⁻ and CD44v3⁻/CD24⁺ cases, the four immunophenotypes were not significantly correlated with each clinicopathological parameter.

Kaplan Meier analysis established the relationship between CD44v3 expression in the invasive portion and OS (Fig. 6A). CD44v3⁺ cases tended to show poor OS compared with CD44v3⁻ cases (P=0.055). The relationship between CD24 expression in the invasive portion and OS is shown in Fig. 6B. There was no significant difference in OS between CD24⁺ cases and CD24⁻ cases. The Kaplan Meier curves for OS in OSCC patients that were subdivided into four groups according to CD44v3/CD24 immunophenotypes are shown in Fig. 6C. CD44v3⁺/CD24⁻ cases showed significantly worse OS than CD44v3⁺/CD24⁻ cases (P=0.029).