The human lung cancer cell line A549 was obtained from the American Type Culture Collection (ATCC) and was cultured in DMEM (Hyclone, USA) with 10% fetal bovine serum at 37°C, 5% CO₂. The induced CD133⁺/CD326⁺ subpopulation cells were suspended in serum-free medium supplemented with 0.4% BSA (Sigma, USA), insulin (5 ng/ml, Sigma), bFGF (10 ng/ml, PeproTech, USA), EGF (20 ng/ml, PeproTech), and B27 (20 ng/ml, Invitrogen, USA) at a density of 10³ cells/3 ml in ultralow attachment plates (Corning, USA).

To induce the EMT process, adherent A549 cells and CD133⁺/CD326⁺ spheroids were treated with 5 ng/ml TGF-β1 (Sigma) for 72 h. Adherent A549 cells and suspended CD133⁺/CD326⁺ cells were transfected with agomiR-181b-5p and antagomiR-181b-5p, which were purchased from RiboBio Co. Ltd. (China). Cells were plated in a 24-well plate at 1×10⁵ cells/plate. Agomirs or antagomirs of miR-181b-5p were appropriately diluted according to the manufacturer's protocol and added to the culture medium to transfect the cells. The concentration of agomiR-181b-5p and antagomiR-181b-5p were 50 nM and 100 nM, respectively. The expression of 181b-5p was determined 48 h after transfection.

Peripheral blood samples were obtained from NSCLC patients prior to treatment at the Xinqiao Hospital of the Third Military Medical University between 2014 and 2015 and were stored at −80°C. This project was approved by the ethics committee of the Xinqiao Hospital of the Third Military Medical University, and informed consent was obtained from all the patients.

Spheres were dissociated into single cells, washed and incubated with monoclonal antibodies specific for human CD133/1-PE and CD326-FITC (Miltenyi, Germany). The appropriate dilution and procedures were carried out according to the manufacturer's instructions. After incubation, the samples were washed with PBS and analyzed by FACSAria II (BD, USA). All CD133⁺/CD326⁺ cells were collected for subsequent experiments.

Total RNA was isolated using RNAiso Plus (Takara, Japan). Circulating microRNA in peripheral blood was isolated using the mirVana PARIS Kit (Ambion, USA). Reverse transcription reactions were performed using the PrimeScript™ RT reagent kit with gDNA Eraser (Takara) for mRNA and Bulge-Loop™ miRNA qRT-PCR Starter kit (RiboBio) for miRNA to make cDNA from total RNA in a MyCycler PCR system (Bio-Rad, USA). Subsequently, quantitative real-time PCR was performed using SYBR^® Premix Ex Taq II (Takara). Primer pairs for miR-181b-5p, cel-miR-39 and U6 were purchased from RiboBio Co. Ltd. Primer pairs for GAPDH and genes associated with stemness and EMT were designed by Sangon Biotech Co. Ltd. (China). Each sample was performed in triplicate, and the reaction products were analyzed using the ABI 7500 Prism Sequence Detection system (Applied Biosystems, USA). Data analysis was based on the Ct method (∆∆Ct according to Applied Biosystems). All operations followed the manufacturer's protocol.

Spheres were centrifuged (800 rpm, 5 min) on slides by cytospin and fixed with 4% paraformaldehyde and 0.1% Triton for 30 min, washed with PBS, blocked with BSA for 30 min at room temperature, and then incubated with primary antibodies at 4°C overnight. Primary antibodies were rabbit monoclonal anti-CD133 (Abcam, UK) and goat polyclonal anti-CD326 (Santa Cruz, USA) at a dilution of 1:300. After washing, the spheroids were incubated with goat anti-rabbit IgG-FITC (Beyotime, China) and donkey anti-goat IgG-Cy3 (BioLegend, USA) fluorescent antibodies at a dilution of 1:400 for 30 min and protected from light. After DAPI staining for the nucleus, the spheres were observed under an Olympus confocal microscope.

Adherent A549 cells and CD133⁺/CD326⁺ cells were left untreated or were treated with TGF-β1 as described above. Cells were lysed using TRIzol (Life Technologies, USA) according to the manufacturer's instructions. First, E. coli poly(A) polymerase was used to generate polyadenylated tails at the 3′-end of all RNA molecules. Second, after annealing oligo-dT primers, cDNA was synthesized using the qScript Flex cDNA synthesis kit (Quanta Biosciences, USA) according to the manufacturer's instructions for gene-specific priming (a universal tag that would extend from the 3′-end of cDNA molecules was added during reverse transcription). With the addition of this universal tag, individual miRNAs were detected with miRNA-specific forward primers and a reverse universal primer mix. A SYBR Green-based quantitative real-time PCR method was used to quantify the relative expression of mature miRNAs. A total of 362 mature miRNAs were evaluated in the microRNA expression profiling array. miRNA expression was normalized by geometric mean-based global normalization using the Real-Time StatMiner (Integromics, Spain) analysis software. The array data were processed by bioinformatics analysis.

For western blot analysis, 40 µg of total protein was resolved by SDS polyacrylamide gel (Boster, China) electrophoresis, and the proteins were then transferred onto polyvinylidene difluoride (PVDF) membranes (Millipore, USA). The membranes were incubated with the Blocking kit (Boster) for 1 h and then incubated overnight at 4°C with antibodies from the EMT antibody sampler kit (1:1,000 dilution, Cell Signaling Technology, USA). The protein levels were normalized against GAPDH from the same sample (1:3,000 dilution, ZSGB-Bio, China). After three washes, the membranes were incubated for 1 h at 37°C with species-specific horseradish peroxidase-conjugated secondary antibodies. The membranes were developed using an enhanced chemiluminescence (ECL) detection system followed by exposure to Hyperfilm ECL (Beyotime).

A Transwell invasion assay was performed using 24-well Transwell permeable supports with 8-µm pores (Corning). Cells (2×10⁵) were suspended in 200 µl serum-free DMEM/F12 medium and seeded into the Transwell inserts coated with Matrigel (BD Biosciences, USA). The bottoms of the wells were filled with 500 µl complete medium containing 20% fetal bovine serum. All cells were incubated at 37°C, 5% CO₂. After 24 h, the cells in the upper chamber were removed with cotton swabs and the filter membrane was fixed with 4% paraformaldehyde for 20 min. Subsequently, the remaining cells of the filter membrane in the lower chamber were stained with 0.1% crystal violet stain solution. From five randomly selected fields, the cells that had invaded through the membrane to the lower surface were counted under a light microscope.

Five-week-old male nude mice were purchased from the Chinese Academy of Medical Sciences (Beijing, China). To generate xenografts, serial dilutions (1×10⁶, 1×10⁵, 1×10, and 1×10³) of A549 cells and CD133⁺/CD326⁺ cells were injected into the mice (n=5). Mice were monitored twice a week and sacrificed after 6 weeks.

A549 cancer cells at 2×10⁶/100 µl were injected into the right armpit of the forelimb in all 10 mice. The mice were observed twice per week. When xenografts reached ~5×5 mm (after nearly 2 weeks), we randomly chose 5 mice and injected 1 nmol micrON™ agomiR-181b-5p every 3 days for 4 weeks. The other 5 mice injected with micrON agomiR-181b-5p negative control were the control group. After 6 weeks, all mice were sacrificed and metastases were observed. All mouse experiments were performed in accordance with local ethics guidelines.

The data are expressed as the means ± SD. Statistical analysis of data was assessed using the Student's t-test and one-way ANOVA with SPSS 18.0 software. Differences with p<0.05 were considered statistically significant.

Only 0.1% of adherent A549 cells co-expressed CD133 and CD326 (Fig. 1A), which are markers of stem cells. To enrich the CD133⁺/CD326⁺ cells, A549 cells were resuspended in serum-free medium and grown as spheroids. We confirmed by flow cytometry that 44.2% cells in the spheroids co-expressed CD133 and CD326 (Fig. 1B). We collected CD133⁺/CD326⁺ cells for the subsequent experiments. Next, we studied the expression of stemness-associated genes, including CD133, CD326, OCT-4 and Nanog, using quantitative real-time PCR. We demonstrated that the expression of these genes was higher in CD133⁺/CD326⁺ cells compared to the adherent A549 cells (p<0.05, Fig. 1C). Immunofluorescence assay confirmed that most of the cells in spheroids co-expressed CD133/CD326 (Fig. 1D). We demonstrated that as few as 1×10⁴ CD133⁺/CD326⁺ cells generated tumor xenografts in nude mice, whereas 1×10⁵ unselected A549 cells failed to do so. All xenografts were confirmed as adenocarcinoma by histologic examination (Fig. 1E and Table I). According to these results and our previous study (13), we concluded that the CD133⁺/CD326⁺ cell subpopulation of A549 cells has the traits of CSLCs.

TGF-β1 has been shown to induce EMT in cancer cells (24). We treated adherent A549 cells and CD133⁺/CD326⁺ cells with 5 ng/ml of TGF-β1 for 72 h. As expected, adherent A549 cells treated with TGF-β1 lost their epithelial morphology. Their shape had changed drastically, the junction between cells became loose and cells acquired a fibroblast-like appearance. In contrast, TGF-β1 treatment did not trigger significant morphological changes in the CD133⁺/CD326⁺ subpopulation (Fig. 2A). We used quantitative real-time PCR to assess the expression of E-cadherin, vimentin, Snail and Slug. We demonstrated that TGF-β1 treatment induced the upregulation of vimentin, Snail and Slug and the downregulation of E-cadherin in both adherent A549 cells and CD133⁺/CD326⁺ cells (Fig. 2B). This result indicates that although the distinctive morphological changes of cells undergoing EMT were not observed in CD133⁺/CD326⁺ cells, the expression of EMT-associated genes indicative of the epithelial to mesenchymal transition was observed in parental A549 cells and in CD133⁺/CD326⁺ cells. Cancer cells undergoing EMT are widely known to gain the ability to invade and metastasize. We used Transwell invasion assay to show that compared to control A549 cells, A549 cells treated with TGF-β1 have increased invasion potential through the Matrigel membrane. CD133⁺/CD326⁺ have an increased basal ability to invade compared to unselected A549 cells and TGF-β1 further increased their invasive capacity (Fig. 2C and D). Based on these results, we concluded that TGF-β1 promotes EMT both in CSLCs and non-CSLCs.

To identify microRNAs involved in EMT and stem cell phenotype, we performed microRNA expression profiles in A549 cells, A549 cells treated with TGF-β1, CD133⁺/CD326⁺ cells and in CD133⁺/CD326⁺ cells treated with TGF-β1 using a quantitative real-time PCR miRNA expression profiling array. We identified common microRNAs between two groups (A549 vs A549/TGF-β1 and CD133⁺/CD326⁺ vs CD133⁺/CD326⁺/TGF-β1), their predicted target genes (n=1180) and their binding sites using the TargetScan and miRanda databases. We performed GO analysis and pathway analysis based on the Gene Ontology database and the KEGG database. Finally, according to the interactions between these genes and their corresponding microRNAs, we constructed a microRNA-gene network. The microRNAs in the network were evaluated by the degree, which is the number of target genes regulated by one microRNA. A high degree indicates that a large number of target genes is regulated by this particular miRNA. Similarly, the degree of the gene indicates the number of microRNAs targeting this gene. If a gene has a high degree, multiple microRNAs may target it. The key miRNA and gene in the network always have the highest degrees (Tables II and III). Several miRNAs in Table II have an established role in cancer; however, the role of miR-181b-5p in EMT of CSLCs has not been described. We confirmed by quantitative real-time PCR that miR-181b-5p was significantly upregulated both in A549 and CD133⁺/CD326⁺ cells treated with TGF-β1 (Fig. 2E). We hypothesize that this may contribute to the invasion and metastasis of CSLCs and non-CSLCs.

To modulate the expression of miR-181b-5p, we transfected agomiR-181b-5p into A549 cells and CD133⁺/CD326⁺ cells. In addition, we transfected antagomiR-181b-5p into A549 cells and CD133⁺/CD326⁺ cells that were treated with TGF-β1 for 72 h. We confirmed that the expression of miR-181b-5p in A549 cells and CD133⁺/CD326⁺ cells transfected with agomiR-181b-5p was increased 8.75- and 25.3-fold (Fig. 3A). Accordingly, the transfection of antagomiR-181b-5p into A549 cells and CD133⁺/CD326⁺ cells treated with TGF-β1 significantly reduced the expression of miR-181b-5p (Fig. 3B). This result confirmed that agomiR-181b-5p was the agonist of miR-181b-5p and that antagomiR-181b-5p was the antagonist of miR-181b-5p.

Next, we sought to determine whether miR-181b-5p regulates EMT in vitro and in vivo. Transwell invasion assay confirmed that agomiR-181b-5p, similar to TGF-β1 treatment, significantly enhanced the invasion of both A549 cells and CD133⁺/CD326⁺ cells. In contrast, antagomiR-181b-5p reduced the TGF-β1-mediated invasion in A549 cells and CD133⁺/CD326⁺ cells but did not completely eliminate the effect of TGF-β1 (Fig. 3C–E). We treated adherent A549 cells with 50 nM of agomiR-181b-5p for 7 days and observed that a part of adherent A549 cells lost their epithelial morphology and acquired a fibroblast-like appearance (Fig. 5D). These results indicated that the expression of miR-181b-5p correlates with the ability of cells to invade in both CSLCs and non-CSLCs and that miR-181b-5p contributes to TGF-β1-induced EMT. Based on the in vitro studies, we hypothesized that the overexpression of miR-181b-5p may promote metastasis in vivo. To confirm this critical question, we performed a tumor metastasis assay and observed that the group of mice injected with agomiR-181b-5p had more metastasis than the negative control (Fig. 4 and Table IV). Thus, our findings suggest that miR-181b-5p promotes invasion and metastasis both in CSLCs and non-CSLCs in vitro and in vivo.

To identify the target gene of miR-181b-5p, we evaluated the expression of EMT markers (such as E-cadherin, vimentin, Snail and Slug) after the treatment of miR-181b-5p agomir and antagomir in untreated or TGF-β1 treated A549 cells and CD133⁺/CD326⁺ cells. After treatment with agomiR-181b-5p, decreased E-cadherin expression was observed only in A549 cells and CD133⁺/CD326⁺ cells (Fig. 5A). In contrast, antagomiR-181b-5p increased the expression of E-cadherin expression in TGF-β1-treated A549 cells and CD133⁺/CD326⁺ cells (Fig. 5B). Western blot analysis was used to detect the protein levels of E-cadherin and to confirm our results (Fig. 5C). These findings suggested that miR-181b-5p promotes EMT at least in part by targeting E-cadherin in CSLCs and non-CSLCs.

Finally, we compared the expression of miR-181b-5p in peripheral blood of 15 healthy people and 20 patients with NSCLC (patient characteristics are indicated in Table V). We found that the levels of miR-181b-5p in peripheral blood were increased in NSCLC patients (Fig. 6A) compared to healthy individuals (p<0.05). miR-181b-5p was upregulated in stage II/III/IV NSCLC patients (Fig. 6B, p<0.05). According to this result, miR-181b-5p may be a potential therapeutic target and a prognostic marker in NSCLC patients.