Identification and characterization of cancer stem cells in human head and neck squamous cell carcinoma

HNSCC cell line A253 (ATCC^®HTB-41) was obtained from American Type Culture Collection (ATCC, Manassas, VA). HNSCC cell line KCCT873 was obtained from Yokohama City University Hospital [22]. A253 cells were established from tumor originated from submaxillary salivary gland. KCCT873 cells were originated from tongue tumor. A253 cells were grown in McCoy’s Modified Medium, and KCCT873 cells in RPMI 1640 medium. Cell culture media were supplemented with 10% fetal bovine serum and 1% penicillin/streptomycin (Lonza, Walkersville, MD). The cells were maintained at 37°C in a humidified atmosphere containing 5% CO₂.

Cell sorting by flow cytometry was performed by Mr. Howard Mostowski at the Flow Cytometry Core facility, Center for Biologics Evaluation and Research, FDA. Cells were labeled with mouse anti-human CD44-PE (Millipore, Temecula, CA) and mouse anti-human CD24-FITC (Santa Cruz Biotech, Santa Cruz, CA) antibodies. The top or bottom cells in the 0.5 to 1 percentile fluorescence intensity of each CD24+/CD44+ and CD24-/CD44+ subpopulations were sorted and collected separately for further experiments.

For flow cytometric analysis of other markers, cells (10⁶ cells/ml) were resuspended and incubated with various antibodies, CD29-APC, CD73-APC, and CD90-PerCP-Cy5.5 (eBioscience - http://​www.​ebioscience.​com), CD24-FITC (Santa Cruz Biotech), and CD44-PE (Millipore), according to the manufacturer’s instructions for 30 min on ice, washed with PBS three times, and fixed with 1% paraformaldehyde for later analysis. For controls, relevant isotype control antibody (eBioscience) and no antibody was used in parallel. Data were analyzed using FlowJo software (Tree Star Inc., Ashland, OR).

For qRT-PCR, total RNAs was extracted by Trizol reagent according to the manufacturer’s instructions (Invitrogen, Carlsbad, CA). The 1^st strand cDNA was synthesized from 1 μg of total RNA using Superscript II Reverse Transcriptase (Invitrogen) according to manufactures specifications. The resulting cDNA was amplified by using gene-specific primers. The primer sequences for each tested gene are listed in Additional file 1: Table S1. For amplification, samples were prepared with SsoAdvanced^TM SYBR^® Green Supermix (Bio-Rad) following the manufacture’s protocol, and run on a Bio-Rad CFX96 Touch^TM Real-Time Detection System. Buffer only and no template were included in each assay run as controls. All samples and controls were run in triplicate. Gene-specific amplification was normalized to β-Actin and relative fold change was calculated following the manufacture’s protocol (Bio-Rad).

One thousand sorted cells per well were cultured in quadruplicate in 96-well plates for the indicated period of time. Cell proliferation was detected by using CellTiter-Glo® Luminescent Cell Viability Assay kit (Promega, Madison, WI). Cell viability was quantified by measuring the absorbance using a microplate reader (Molecular Devices, Sunnyvale, CA) with 500 ms integration. Experimental background was determined by using empty wells with medium.

Collagen type I gels were prepared with cell culture medium to make final collagen concentration of 2 mg/mL (pH = 7.0) [23]. For cell cultures within collagen gels, 1.5 mL cell suspension (500 cells/mL) was mixed with 1.5 mL of collagen solution. The mixture was plated in six-well plates, and placed in 37°C incubator for gelation. After gelation, the collagen gels were overlaid with 3 mL of complete medium and incubated in a humidified atmosphere containing 95% air and 5% CO₂. Cells were cultured for six days. Cell colonies were visualized with Coomassie Blue solution staining (0.5% Coomassie Brilliant Blue G250, Bio-Rad), and visible colonies were counted. Assays were performed in triplicate.

Cell invasion was studied by using BD BioCoat Matrigel invasion chambers (BD Biosciences; 24-well, 8 μm pore size) with 10% fetal bovine serum as a chemo attractant, and following the manufacture’s protocol. Briefly, one thousand cells were loaded into the chamber and incubated for 24 to 72 hrs at 37°C. Noninvasive cells were removed from the upper surface of the membrane with a cotton swab, and cells on the bottom surface of the membrane were fixed and stained with H&E. Cells in five random fields per well were counted. The experiments were performed in duplicate.

Following cell sorting, both CD24+/CD44+ and CD24-/CD44+ cells were cultured for 2 days to eliminate damaged cells caused by the sorting process. Cells were then plated at a density of 1 × 10³/well in 96-well plates. Chemotherapeutic reagents, Gemcitabine or Cisplatin, were added to the cells at gradually increasing concentrations. The cells were cultured for 72 hrs, and the cell viability was determined by CellTiter-Glo^® Assay (Promega, Madison, WI) according to the manufacturer’s protocol.

Immunohistochemical (IHC) studies of tumor sections were performed on formalin-fixed, paraffin-embedded tumors isolated from tumor xenografts in the study. Tissue sections were deparaffinized by xylene, and re-hydrated with sequential washes of 100%, 75%, and 50% ethanol, and PBS. For antigen retrieval, slides were placed in 50 mM citrate buffer pH 6.0 (Vector Lab, CA), boiled for 5 min, and stayed in the buffer for 15 min. Endogenous peroxidase activity was inhibited with 3% hydrogen peroxidase in PBS. Non-specific binding was blocked with 2.5% normal serum and 1% bovine serum albumin (BSA) for 1 hr. Tissue sections were incubated with various antibodies, CD24 and CD44 (Millipore), or isotype control (IgG) (Sigma) overnight at 4°C. Immunodetection was performed using ABC staining systems according to manufacturer’s instructions (Santa Cruz Biotech). All sections were counterstained with haematoxylin. After dehydration with washes of 95% and 100% ethanol and xylene, tissue sections with permanent mounting medium were covered with glass coverslips, and viewed by light microscope. H&E staining was also performed on the section from each tumor tissue sample.

Statistical analyses were performed by paired Student’s t-test between two groups. Data were presented as mean ± SD. P value of < 0.05 was considered statistically significant. Each experiment was repeated at least twice including animal experiments.

To determine the percentage of the putative cancer stem-like cells in the HNSCC cell population, cell suspensions from cell lines A253 and KCCT873 were analyzed and sorted for cell surface markers CD24 and CD44 by flow cytometry. Two phenotypic subpopulations were separated. CD24+/CD44+ cells were only ~5-8% in whole cell population. In contrast, CD24-/CD44+ cells were >90% in whole cell population of both HNSCC cell lines (Figure 1A).

We next investigated the expression of known “stemness” genes in the isolated CD24+/CD44+ and CD24-/CD44+ subpopulations by real-time RT-PCR technology. We tested expression of six genes including ALDH1, BMI1, CD133, Nanog, Oct3/4, and Sox2. BMI1 and Nanog genes showed a significantly higher expression in CD24+/CD44+ compared to CD24-/CD44+ subpopulations from both HNSCC cell lines. However, there was no significant difference in ALDH1 expression between CD24+/CD44+ and CD24-/CD44+ subpopulations from both cell lines (Figure 1B and C). CD133 was only expressed in one cell line (KCCT873) at a very low level and did not show a clear difference between two subpopulations of cells (data not shown). A253 cells did not show any expression of CD133 gene. The expression of Oct3/4 and Sox2 was absent in both cell subpopulations in both cell lines (data not shown).

To explore the self-renewal and differentiation capacity of CD24+/CD44+ cells, the purified CD24+/CD44+ cells were cultured in vitro for 3 weeks, and variations in CD24 expression were examined by flow cytometry. We found that the proportion of CD24+/CD44+ cells dramatically declined in a time dependent manner in the CD24+/CD44+ sorted population of cells. CD24+ cells in CD24+/CD44+ population decreased to ~62% one week after culture and continued to decrease to 28% two weeks after cell culture. The proportion of the CD24+/CD44+ cells returned to similar presorting level (< 10%) after three weeks culture. In contrast, the proportion of CD24-/CD44+ cells in the cell population gradually increased from ~30% at the first week to ~86% after three weeks, indicating that the CD24+/CD44+ cells give rise to CD24-/CD44+ cells (Figure 2A and B).

Cell proliferation assays indicated that the growth rate of CD24+/CD44+ cells was slightly lower compared to CD24-/CD44+ cells for up to 5 days after cell sorting (Figure 3A and B). These results indicate that CD24+/CD44+ cells show asymmetric division-like proliferation pattern, indicating the self-renewal and differentiation potential to produce heterologous descendent CD24-/CD44+ cells in culture.

We next investigated the invasion ability of CD24+/CD44+ and CD24-/CD44+ subpopulations by matrigel invasion assays. We observed that the number of invading cells in the CD24+/CD44+ cells was significantly higher compared to CD24-/CD44+ cells, indicating that CD24+/CD44+ cells have higher invasion ability compared to CD24-/CD44+ cells (p < 0.02 for A253 and p < 0.01 for KCCT873 compared to CD24-/CD44+ cells) (Figure 4A).

The colony-formation capacity of CD24+/CD44+ and CD24-/CD44+ subpopulations was also tested. Our results indicate that CD24+/CD44+ cells form significantly higher number of colonies compared to CD24-/CD44+ cell subpopulation (p < 0.05) (Figure 4B).

Cisplatin (cis-diammine-dichloroplatinum (II)) is used for treatment of a wide range of cancers, including head & neck tumors. Cisplatin often leads to an initial therapeutic success associated with partial response or disease stabilization [24]. Gemcitabine is a nucleoside analog displaying a wide spectrum of antitumor activity [25]. Although both drugs have been used for chemotherapeutic treatment of patients with head & neck tumors, many patients are intrinsically resistant to these drugs [24]. Recent studies have indicated that cancer stem cell phenotypes are associated with drug resistance to chemotherapeutic drugs [26, 27]. To evaluate the drug resistance properties of FACS sorted HNSCC cells, CD24+/CD44+ and CD24-/CD44+ cells were grown and treated with various concentrations of either cisplatin or gemcitabine for 72 hours, and then cell survival was assessed by determining cell viability. CD24+/CD44+ cells seem to show small but significantly higher drug resistance to either chemotherapeutic agent when compared to CD24-/CD44+ cells (Figure 5). For example, CD24+/CD44+ cells showed higher survival rate (53.5%) compared to CD24-/CD44+ cells (40%) when treated with 1000 nM cisplatin (p < 0.01) (Figure 5A). Similarly, CD24+/CD44+ cells showed > 10% higher survival rate (37%) compared to survival rate (26%) of CD24-/CD44+ cells when treated with 10 nM gemcitabine (p < 0.01) (Figure 5B).

We next evaluated whether the two subpopulations (CD24+/CD44+ and CD24-/CD44+) of HNSCC cells were endowed with differential tumorigenic potential. Several independent experiments were performed with two different HNSCC cell lines. The two phenotypic subpopulations of cells, CD24+/CD44+ and CD24-/CD44+, were sorted by flow cytometry, suspended in a Matrigel mixture (1:1), and then S.C. injected into athymic nude mice. The tumor size was measured weekly for 9 weeks, at which time animals were sacrificed. When minimal (1 × 10²) to maximal (1 × 10⁴) numbers of cells per mouse were injected, both CD24+/CD44+ and CD24-/CD44+ cells formed tumors and thus were tumorigenic. However, the size of tumor generated by CD24+/CD44+ cells was significantly larger than the size of the tumors from CD24-/CD44+ or unsorted control cells (Figure 6) indicating CD24+/CD44+ cells are highly aggressive.

Immunohistochemical staining for CD24 and CD44 on tumor tissues isolated from tumor xenografts at the end of the study were performed to determine whether CD24+/CD44+ CSC maintained their phenotype at the end of the experiment. Upon H&E staining, A253 cells showed submaxillary salivary gland features since these cells originated from submaxillary salivary gland tumor. KCCT873 cells showed similar features. By IHC, strong positive staining for CD24 was observed on the surface of salivary gland appearing structure in xenograft tumors derived from both cell lines. In addition, strong positive staining for CD44 was observed not only on the surface of salivary gland appearing structures, but also on the dense carcinoma cells within the tumor mass as well (Figure 7). Since xenograft tumors generated from both CD24+/CD44+ and CD24-/CD44+ cells showed the similar immunohistochemical staining, we hypothesized that the CD24+/CD44+ cells may have been generated during the in vivo tumor growth from CD24-/CD44+ cell subpopulation.

To investigate whether other putative stem cell markers were expressed in HNSCC cells, the mesenchymal stem cell markers, CD29 (β1-integrin), CD73 (5′-nucleotidase), CD90 (Thy-1), and CD105 (Endogin) were selected and analyzed by flow cytometry. CD29 expression showed the strongest correlation with the CD44 expression. Almost all cells (99.6%) were CD29 and CD44 double-positive. Only ~6% of the cells were CD29+ and CD24+, the same percentage found for CD24+/CD44+ cells (Figure 8A). CD73 also showed a strong correlation with CD44 expression. Approximately 92% of the cells were CD73 and CD44 double-positive, while only ~6% of the cells were CD73+/CD24+, similar to CD24+/CD44+ cells (Figure 8B). On the other hand, neither CD90 nor CD105 showed any correlation with either CD24 or CD44 expression (data not shown).