NSCLC cells, A549, H1299, SPC-A1, SK-MES-1 and a human bronchial epithelial cell line (HBE), were cultured separately in RPMI-1640 medium (Gibco, Invitrogen Life Technologies, Carlsbad, CA, USA) supplemented with 10% fetal bovine serum (FBS), 100 IU/ml penicillin and 100 mg/ml streptomycin in a 5% CO₂ humidified atmosphere at 37°C. For isolation of CD133-positive cells, the parental H1299 cells were labeled with 1 ml CD133 per liter micromagnetic beads per million cells, using a CD133 cell isolation kit (CD133 MicroBead kit; Miltenyi Biotec, Auburn, CA, USA). The isolated CD133-positive cells were cultured in medium consisting of serum-free Dulbecco’s modified Eagle’s medium (DMEM)/F-12 medium (Gibco, Invitrogen Life Technologies) supplemented with 20 ng/ml human epidermal growth factor (EGF) (PeproTech, Rocky Hill, NJ, USA), 10 ng/ml human basic fibroblast growth factor (bFGF; PeproTech) and 2% B-27 serum-free supplements (Invitrogen Life Technologies). The medium was replaced or supplemented with fresh growth factors twice weekly until floating aggregates formed.

To determine the percentage of single cells capable of regenerating novel spheres, cells were plated at a density of 1,000 cells/ml in ultralow-attachment 6-well plates (Corning Life Sciences, Union City, CA, USA) to obtain the novel spheres. The total number of tumor spheres was counted following 10 days of culture. Efficiency of sphere formation was calculated by the following equation: (Total number of spheres formed / total number of living cells seeded) × 100.

NSCLC and matched adjacent normal biopsies were obtained from 30 patients (see supplementary Table I) undergoing pulmonary resection at the Zhoushan Hospital of Zhejiang province (Zhoushan, Zhejiang, China) between January 2009 and May 2010. All samples were collected with informed consent obtained at an internal review. The ethics board committees of the Zhoushan Hospital of Zhejiang province approved the current study. Patients recruited into the study had not undergone chemotherapy or radiotherapy prior to surgery. Surgical specimens of the resected tumors were collected following the confirmation of diagnosis by pathological examination. Tumor sections and matched adjacent noncancerous tissues, were placed in separated cryovials and snap-frozen in liquid nitrogen until analysis. All cases were reviewed by two pathologists and diagnosis was confirmed according to criteria previously established by the National Comprehensive Cancer Network (23).

For flow cytometry analysis, cells (10⁶/100 μl) were dissociated using non-enzymatic solution Cellstripper (Mediatech, Herndon, VA, USA) and incubated with the appropriate dilution of control or specific Phycoerythrin (PE)-conjugated anti-CD133/2 antibody (Miltenyi Biotec) at 4°C for 10 min. All samples were measured using a FACSCalibur flow cytometer (BD Biosciences, Franklin Lakes, NJ, USA) and analyzed with CellQuest software (BD Biosciences).

Total RNA from specimens or cells was extracted using TRIzol (Invitrogen Life Technologies), according to the manufacturer’s instructions and the concentration was determined using a Bioanalyzer UV spectrophotometer Q3000 (Quawell Technology, Inc., San Jose, CA, USA). Complementary DNA (cDNA) was reverse transcribed by 3 μg total RNA in a 20 μl reaction using 0.5 μg Oligo(dT)₁₈ primer and 200 units of RevertAid™ M-MuLV Reverse Transcriptase (Fermentas, Burlington, ON, Canada) and treated with RNase-Free DNase (Qiagen, Valencia, CA, USA). The primers of the housekeeping gene glyceraldehyde-3-phosphate dehydrogenase (GAPDH) and target genes CD133, octamer-binding transcription factor 4 (OCT4A), sex determining region Y-box 2 (SOX2) and Nanog homeobox (NANOG), multidrug resistance protein 1 (MDR1) and ATP-binding cassette subfamily G member 2 (ABCG2) are shown in Table I. The cDNA amplification was performed in a final reaction volume of 20 μl containing 0.5 μM concentrations of each primer, 8 μl 2.5X Real Master mix, 1 μl 20X SYBR solution (Tiangen Biotech, Shanghai, China) and 2 μl 1:10 diluted cDNA. PCR was prepared in triplicate and heated to 95°C for 10 min, followed by 40 cycles of the following sequences: Denaturation at 95°C for 15 sec, annealing at 60°C for 20 sec and extension at 68°C for 35 sec. The expression of genes was detected by qPCR using Applied Biosystems 7500 Real-Time RT-PCR system (Applied Biosystems, Carlsbad, CA, USA). The cycle threshold (Ct) values were calculated with the SDS 2.0.1 software (Applied Biosystems). All target gene Ct values were determined in reference to GAPDH using the 2^−ΔΔCt method (24).

Parental H1299 cells and sorted cells (5×10³) were plated in 96-well plates. Cisplatin and paclitaxel (Sigma-Aldrich, St. Louis, MO, USA) were added to the culture medium at a final concentration of 25 μg/ml and 20 μM, respectively, for 24 h. Subsequently, cell viability was analyzed by an MTT assay.

Female nude mice (strain, BALB/c; age, 4 weeks) were obtained from the Shanghai Laboratory Animal Center (Shanghai, China). Specific numbers of parental and sorted cells (10³, 10⁴ or 10⁶ cells) were subcutaneously injected into the flank of the nude mice. Tumor growth was monitored and volume was measured weekly using a caliper (volume = width × length² × π/6) for a maximum of 8 weeks. Subsequently, mice were euthanized by cervical vertebra dislocation and the tumors were collected, fixed in formalin and embedded in paraffin for further use.

Statistical analyses were performed with the GraphPad Prism 5.0 statistical software (GraphPad Software, Inc., La Jolla, CA, USA). The paired t-test, unpaired t-test and Mann-Whitney U test were used to analyze the correlation between mRNA expression levels and the clinicopathological features. The paired sample t-test was used to compare the differences of mRNA expression between lung tumors and the surrounding normal tissue. Survival analysis was estimated by the Kaplan-Meier method and the log-rank test was used to compare the survival between groups. The Cox proportional hazard regression model was used to analyze the risk factors for lung cancer. P<0.05 was considered to indicate a statistically significant difference.

To investigate the involvement of CD133 in the link between CSCs and lung cancer biology, CD133-positive cells were isolated from lung cancer cell lines using the magnetic bead method to determine the expression levels of CD133. CD133 was shown to be highly expressed in, 95C, SPC-A1, A549 and H1299 NSCLC lines. The highest expression levels were observed in the H1299 cells (Fig. 1A). Thus, H1299 cells were used for all subsequent experiments.

The isolated CD133-positive cells identified by flow cytometry formed floating sphere-like bodies in DMEM/F-12 serum-free medium with bFGF and EGF. When CD133-positive cells were grown under adherent conditions in culture medium supplemented with 10% FBS, the cells acquired the typical morphological features of parental H1299 cells with the loss of CD133 expression (decrease in expression from 94.9 to 4.9%; Fig. 1B). qPCR results demonstrated that the expression of stemness genes (OCT4A, SOX2 and NANOG) and drug resistant genes (MDR1 and ABCG2) in CD133-positive cells were higher compared with those in the CD133-negative cells (Fig. 1C). The ability to form spheroid-like bodies was significantly higher in the CD133-positive cells compared with the parental or CD133-negative cells (P<0.05; Fig. 1D). The multidrug chemotherapy resistant abilities of the CD133-positive cells and CD133-negative cells were then investigated. Compared with the CD133-negative and parental cells, CD133-positive cells sorted from H1299 cells were more resistant to cisplatin and paclitaxel (P<0.05; Fig. 1E).

To analyze the tumorigenic potential of CD133-positive cells grown under sphere-forming conditions, during which this property of stem cells may be maintained, specific numbers of H1299 CD133-positive, CD133-negative and parental cells were subcutaneously implanted into the flanks of nude mice (BALB/c strain). As shown in Table II, tumor growth was observed in all groups in which mice were inoculated with 10³–10⁶ CD133-positive cells, whereas no tumor growth was observed following inoculation with 10³ H1299 parental or CD133-negative cells (Fig. 2A). Tumor growth curves showed that tumors that had developed from the CD133-positive cells grew at a faster rate compared with those that had developed from the parental or CD133-negative cells (Fig. 2B). These observations indicated that CD133-positive lung cancer cells may be used to enrich tumor-initiating cells under sphere-forming conditions.

Tumor specimens from 30 patients (19 paired samples of lung adenocarcinomas and 11 paired samples of lung squamous-cell carcinomas) with previously untreated NSCLC were recruited into the current study. The patients consisted of 23 males and 7 females. The median age was 61.1 years, with nine cases (30%) aged >65 years. Based on the World Health Organization criteria, 15 patients (50%) presented with stage I disease, six patients (20%) with stage II disease and 9 patients (30%) with stage III/IV disease. Of the 30 patients, 14 (46.67%) had tumors with nodal metastasis and 15 (30%) had a primary tumor >3 cm in diameter. The majority of the patients (22; 73.33%) were smokers.

A primary aim of the current study was to determine whether there was a difference in the frequency of CSC-associated gene expression between paired lung carcinoma and the corresponding noncancerous lung tissue. The results from the qPCR analysis showed that, compared with the adjacent normal lung tissues, the levels of CSC-associated biomarkers, such as OCT4A, CD133, NANOG and MDR1, were significantly increased in lung cancer tissues (P<0.05; Fig. 3). The mRNA levels of CD133 (Fig. 3A) and MDR1 (Fig. 3D) were observed to be significantly higher in the tumor tissue (CD133, P=0.0131 and MDR1, P=0.0159) compared with that in adjacent normal lung tissue. In addition, compared with the normal counterpart, higher NANOG and OCT4A expression was observed in the malignant tissue (P=0.0012 and P=0.0076) in 23 and 19 of 30 pairs (76.67 and 63.33%, respectively; Fig. 3C and B).

The correlation between the expression of CSC-associated genes and clinicopathological features of NSCLC was investigated to gain an improved understanding of the potential involvement of these genes in NSCLC development and progression using the Mann-Whitney U test (Table III). No substantial correlation between the expression of NANOG or MDR1 and the tumor stage, nodal status, tumor size and histological type was observed, with the exception of tumor differentiation. However, overexpression of CD133 in lung tumors was observed to be markedly associated with clinical stage, tumor size and differentiation of NSCLC (P=0.0334, 0.0238 and 0.0376, respectively; Fig. 4A–C). In addition, patients with squamous-cell lung carcinoma or a tumor ≥3 cm in diameter exhibited higher expression levels of CD133. Notably, overexpression of OCT4A in tumors was positively associated with a tumor stage I or II and moderate-well differentiated tumors (P=0.0417 and 0.0396, respectively; Fig. 4D and E). In addition, patients with poor differentiation had higher levels of CD133 and OCT4A.

The expression of CSC-associated genes and its correlation with overall survival of NSCLC patients was investigated (fold change >2 was considered as high expression). The results showed that higher levels of expression of CD133 in NSCLC was independently correlated with shorter overall survival (log-rank test: P=0.0258, Fig. 5). However, no significant correlation between the expression of OCT4A and the overall survival was observed (log-rank test: P=0.1395). In addition, tumor size >3 cm was observed to be associated with decreased overall survival (log-rank test: P=0.015, data not shown).

The Kaplan-Meier survival curve showed a trend for an improved outcome in patients with lower CD133 mRNA signals; however, the Cox hazard regression analysis indicated that CD133 was not considered to be an independent factor in predicting the prognosis of patients with NSCLC (hazard ratio: 0.179; 95% confidence interval: 0.010–3.205; P=0.243).