The NSCLC cell line PC9 was purchased from the American Type Culture Collection (ATCC; Manassas, VA, USA) and was cultured in RPMI-1640 medium containing 10% fetal bovine serum (FBS; Gibco, Waltham, MA, USA). For isolating CSCs of PC9 cell line, FACSVantage (BD FACSCalibur; BD Biosciences, San Diego, CA, USA) was used to identify and collect the CD133-positive PC9 cells (CD133-FITC antibody was purchased from Miltenyi Biotec GmbH, Bergisch Gladbach, Germany). To estimate the purity of CSC population in PC9, flow cytometry was performed.

Negative control oligonucleotides (miR-NC), microRNA-23a mimics (miR-23a), miR-23a inhibitors (anti-miR-23a) and PTEN siRNA were purchased from Guangzhou RiboBio, Co., Ltd. (Guangzhou, China). For transfection, 50 pmol/ml miR-NC, miR-23a, anti-miR-23a or PTEN siRNA were packaged with Lipofectamine 2000 (Invitrogen, Waltham, MA, USA). Then, PC9 CSCs were incubated with RNAs in serum-free medium (Gibco). Six hours later, the culture medium was changed to 10% FBS RPMI-1640 medium for 24 h.

PC9 CSCs (5×10³) or PC9 non-CSCs (5×10³) were plated on 96-well plates with 200 µl RPMI-1640 medium. Then, cells were transfected with RNAs followed by treating with EGFR-TKIs for 48 h. After incubation, 20 µl 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) (5 mg/ml; Sigma-Aldrich, St. Louis, MO, USA) was added into the medium for 4 h. Subsequently, we replaced the culture medium with 150 µl dimethyl sulfoxide (DMSO) for 30 min. Cell viability of PC9 CSCs and PC9 non-CSCs was evaluated according to the absorbance at 570 nm using a microplate reader. Half maximal inhibitory concentration (IC₅₀) of EGFR-TKIs to PC9 CSCs and PC9 non-CSCs was calculated according to the results of cell viability assays.

PC9 CSCs (5×10⁵) or PC9 non-CSCs (5×10⁵) were plated on 6-well plates with 2 ml RPMI-1640 medium. Then, cells were transfected with RNAs followed by treating with 0.04 µM erlotinib for 48 h. After incubation, ³H- thymidine was added into the culture medium for 6 h. Then, ³H-thymidine incorporation assays were performed to determine the cell proliferation of PC9 CSCs and PC9 non-CSCs.

PC9 CSCs and PC9 non-CSCs were seeded at a density of 50 cells/cm² (185 cells/well) in a 12-well plate format. Subsequently, cells were treated with 0.04 µM erlotinib for 10 days. After treatment, cell colonies were stained with crystal violet/ethanol mixture and then were counted under a microscope. The proportion of colonies formed is shown as a percentage of its untreated control (18).

Total RNA was extracted from the PC9 CSCs and PC9 non-CSCs using TRIzol reagent (Invitrogen). Expression of miR-23a, Oct4 and Sox2 was measured by One-Step RT-qPCR with SYBR Premix Ex Taq (Takara Bio Inc., Otsu, Japan) according to the manufacturers instructions. U6 snRNA was used as the internal control to calculate the relative expression of miR-23a. GAPDH was used as the internal control to calculate the relative expression of Oct4 and Sox2. 2^−∆∆CT method was used to analyze the results of RT-qPCR (19).

Total proteins were extracted from the PC9 CSCs and PC9 non-CSCs using RIPA buffer (Cell Signaling Technology, Beverly, MA, USA) and then quantified by bicinchoninic acid assay kit (Beyotime Institute of Biotechnology, Haimen, China). Proteins were then separated by 12% sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) and were transferred onto a PVDF membrane (Millipore, Billerica, MA, USA). Membranes were then blocked in 5% skim milk for 1 h at room temperature and then incubated overnight at 4°C with the following antibodies: anti-PTEN, anti-EGFR, anti-phosphorylated EGFR, anti-PI3K, anti-phosphorylated PI3K, anti-AKT, anti-phosphorylated AKT, anti-pro-caspase-9, anti-cleaved caspase-9, anti-pro-caspase-3, anti-cleaved caspase-3 and anti-β-actin (Cell Signaling Technology). After washing the membranes with Tris-buffered saline, they were incubated with goat anti-rabbit secondary antibody (Cell Signaling Technology) for 1 h at room temperature. The blots were visualized using an enhanced chemiluminescence detection kit (Pierce, Rockford, IL, USA).

PC9 CSCs (5×10⁵) or PC9 non-CSCs (5×10⁵) were plated on 6-well plates with 2 ml RPMI-1640 medium. Then, cells were transfected with RNAs followed by treating with 0.04 µM erlotinib for 48 h. After incubation, cells were collected and were stained with PI and Annexin V (Sigma-Aldrich) for 20 min at room temperature. Subsequently, percentage of apoptotic cells was quantified using flow cyto-metry.

3′UTR of PTEN mRNA was amplified and inserted into pGL3 Luciferase reporter vectors (Promega, Madison, WI, USA) downstream of the luciferase gene. PC9 CSCs (1×10⁵) were plated on 24-well plates with 500 µl RPMI-1640 medium. Then, cells were transfected with 50 pmol/ml miR-23a mimics, 2 µg/ml firefly luciferase reporters and 100 ng/ml Renilla luciferase pRL-TK vector (Promega) for 24 h. After transfection, cells were collected and washed with phosphate-buffered saline (PBS). Luciferase reporter assays were then performed using the Dual-Luciferase reporter assay system (Promega) according to the manufacturers protocol.

Quantitative data are presented as the mean ± SD. All of the experiments were repeated in triplicate. For comparison analysis, the statistical analysis was performed by the Students t-test using SPSS 15.0 software (SPSS; Chicago, IL, USA). A value of P<0.05 was considered to indicate a statistically significant difference.

To study the effect of EGFR-TKIs on lung cancer stem cells, we isolated the CSCs population of PC9 using CD133 antibody. Our separated PC9 CSCs expressed high level of CD133, and CD133 negative PC9 cell population was considered as the non-CSCs (Fig. 1A). As Oct4 and Sox2 were identified as the markers of CSCs (20), we detected their expression by using RT-qPCR analysis. As shown in Fig. 1B, the CD133 positive PC9 cell population expressed significantly higher level of Oct4 and Sox2 compared to the PC9 non-CSCs. Notably, we found that PC9 CSCs showed significant resistance to erlotinib compared to the PC9 non-CSCs. IC₅₀ of erlotinib to PC9 CSCs is 12.6-fold higher than the PC9 non-CSCs (Fig. 1C). The results suggested that lung cancer stem cells are resistant to EGFR-TKIs-induced cell death.

Suppression of EGFR/PI3K/AKT pathway is the mechanism of EGFR-TKIs in killing the NSCLC cells (21). Consistent with this, phosphorylation of EGFR, PI3K and AKT was significantly inhibited in PC9 non-CSCs treated with erlotinib. However, in PC9 CSCs, we observed that erlotinib failed to suppress the activation of PI3K and AKT obviously, despite inhibiting the EGFR phosphorylation. High activation of PI3K and AKT promotes cell proliferation and inhibits apoptosis in cancer cells (22,23). Since erlotinib failed to suppress the PI3K/AKT pathway in PC9 CSCs, the results of ³H-thymidine incorporation assays showed that PC9 CSCs still exhibited high proliferative activity even when they were treated with erlotinib (Fig. 2B). Similarly, the effect of erlotinib on suppressing the colony formation of PC9 CSCs was significantly less than the PC9 non-CSCs (Fig. 2C). Furthermore, PC9 CSCs showed obvious resistance to erlotinib-induced apoptosis rather than the PC9 non-CSCs (Fig. 2D). Taken together, we demonstrated that erlotinib failed to suppress the activation of PI3K and AKT in lung cancer stem cells.

To investigate the mechanism by which erlotinib failed to inhibit the PI3K/AKT pathway in PC9 CSCs, we detected the miRNA expression in PC9 cell line. We found that the expression of miR-23a was significantly upregulated in PC9 CSCs rather than PC9 non-CSCs (Fig. 3A). Furthermore, PTEN, which is an effective suppressor of PI3K and AKT (22), was downregulated in PC9 CSCs compared to the corresponding non-CSCs (Fig. 3B). Negative correlation between miR-23a and PTEN was observed. As the public miRNA TargetScan database (http://www.targetscan.org) as well as miRanda database (http://www.microrna.org/) predicted that PTEN mRNA 3′UTR contained putative binding sites to miR-23a (Fig. 3C), we inferred that PTEN is the target of miR-23a, and overexpression of miR-23a in PC9 CSCs was responsible for their erlotinib resistance. To confirm the relationship between miR-23a and PTEN, we performed luciferase reporter assays. The results of these assays showed that transfection with miR-23a in PC9 CSCs significantly decreased the activities of luciferase in pGL3-PTEN plasmid. In contrast, introduction with anti-miR-23a was able to increase the activities of luciferase in pGL3-PTEN vector (Fig. 3D). These results implied the effect of miR-23a on regulating the PTEN gene. In addition, we observed that transfection with miR-23a decreased the protein level of PTEN, whereas the miR-23a antisense oligonucleotides (anti-miR-23a) significantly increased the expression level of PTEN in PC9 CSCs (Fig. 3E). Taken together, we proved that miR-23a regulated the expression of PTEN in lung cancer stem cells.

The preceding results demonstrated that miR-23a which was overexpressed in PC9 CSCs decreased the PTEN expression. We therefore investigated whether knockdown of miR-23a enhances the antitumor effect of erlotinib on PC9 CSCs. Results of western blot analysis illustrated that transfection with miR-23a antisense oligonucleotides (anti-miR-23a) significantly increased the expression of PTEN, whereas the PTEN siRNA abolished the effect of anti-miR-23a (Fig. 4A). Next, we found that miR-23a antisense oligonucleotides significantly enhanced the effect of erlotinib on killing PC9 CSCs. However, co-transfection with PETN siRNA obviously impaired the effect of anti-miR-23a (Fig. 4B). These results proved the synergistic effects of miR-23a antisense oligonucleotides on EGFR-TKIs. Mechanically, although erlotinib alone treatment inhibited the EGFR signaling, absence of PTEN in PC9 CSCs still induced the activation of PI3K and AKT. By contrast, combination with erlotinib and anti-miR-23a inhibited the phosphorylation of EGFR, and upregulated the expression of PTEN. Phosphorylation of PI3K and AKT was significantly suppressed (Fig. 4C). As cell proliferation and apoptosis were regulated by PI3K/AKT pathway (22–24), we observed that combination with miR-23a antisense oligonucleotides and erlotinib induced significant inhibition of cell proliferation in PC9 CSCs (Fig. 4D). Moreover, anti-miR-23a promoted erlotinib-induced apoptosis and cleavage of caspase-9 and caspase-3 by upregulating the expression of PTEN in PC9 CSCs (Fig. 4E and F). Taken together, we demonstrated that miR-23a antisense oligonucleotides resensitize lung cancer stem cells to erlotinib by increasing the expression of PTEN.

To investigate the effect of anti-miR-23a on other EGFR-TKIs, we treated the PC9 CSCs with gefitinib and afatinib after transfection with anti-miR-23a and PTEN siRNA. The results of MTT demonstrated that knockdown of miR-23a also enhanced the antitumor effect of gefitinib and afatinib by upregulating the expression of PTEN (Fig. 5A). Intuitively, transfection with miR-23a antisense oligonucleotides significantly decreased the IC₅₀ of all these EGFR-TKIs (gefitinib, erlotinib and afatinib) to PC9 CSCs (Fig. 5B). We therefore demonstrated the effect of miR-23a antisense oligonucleotides on resensitizing lung cancer stem cells to EGFR-TKIs.