Icotinib was gifted from Zhejiang Beta Pharma, Co. Ltd. and was prepared in 5% dimethyl sulfoxide to obtain a stock solution of 10 mM at −20°C. Stattic (cat. no. ab120952) was obtained from Abcam, used in western blot as an inhibitor of STAT3. Anti-AKT (1:1,000; cat. no. #9272S), anti-phosphorylated (p)-AKT (1:500; cat. no. #9271L, Ser473), anti-p-ERK (1:1,000; cat. no. #4370S, Thr202/Tyr204), anti-SRC (1:1,000; cat. no. #2110S), anti-p-SRC (1:500; cat. no. #6943S), anti-FAK (1:1,000; cat. no. #3285S), anti-p-FAK (1:250; cat. no. #3284, Y925), anti-STAT3 (1:1,000; cat. no. #132L), anti-p-STAT3 (1:1,000; cat. no. #9131L, Tyr705) anti-EGFR (1:1,000; cat. no. #2646S), anti-p-EGFR (1:500; cat. no. #2234S, Tyr1068) and anti-integrin α5 (1:500; cat. no. #4705) were purchased from Cell Signaling Technology, Inc. Anti-integrin β4 (1:1,000; cat. no. MAB4060) was purchased from Novus Biologicals, Ltd. Anti-ERK (1:2,000; cat. no. sc292838), anti-actin (1:1,000; cat. no. sc1616), horseradish peroxidase (HRP)-conjugated secondary goat anti-rabbit (cat. no. sc-sc-2004, targeting anti-AKT, p-AKT, p-ERK, p-SRC, FAK, p-FAK, STAT3, p-STAT3, EGFR, p-EGFR, integrin α5 and ERK) and goat anti-mouse antibodies (cat. no. sc-sc-2005, targeting anti-SRC, integrin β4 and actin) were purchased from Santa Cruz Biotechnology, Inc.

Human lung adenocarcinoma cells, PC9 and HCC827, were obtained from The Cell Bank of Type Culture Collection of The Chinese Academy of Sciences (Shanghai, China). The cells were cultured in RPMI-1640 medium purchased from Thermo Fisher Scientific, Inc. containing 10% heat-inactivated foetal bovine serum (FBS; Biological Industries), penicillin (100 U/ml) and streptomycin (100 mg/ml) at 37°C in an humidified atmosphere with 95% air and 5%CO₂. Resistant cells were established by exposing to 0.05 µM icotinib at the beginning and the icotinib concentration was gradually increased in the totally same incubator condition as previously mentioned. Finally, the PC9/IcoR and 827/IcoR cells were maintained at 10 and 5 µM icotinib, respectively.

The effect of icotinib on the cell viability was evaluated using an MTT assay. PC9 and HCC827 cells (3,000 or 5,000 cells/well) in 96-well plate were treated with different concentration of icotinib for 24 or 48 h. A total of 20 µl of the MTT reagent (5 mg/l) was added per well and incubated for another 4 h. The supernatant was removed and 200 µl of dimethyl sulfoxide was added. The absorbance was measured at 570 nm using an iMark Absorbance Microplate (Bio-Rad Laboratories, Inc.).

A colony formation assay was used to estimate the ability of cell proliferation. PC9, PC9/IcoR, HCC827 and 827/IcoR cells at a density of 400 or 600/well in six-well plate were treated with different concentrations of icotinib (0, 0.1, 1 and 10 µM) for 72 h at 37°C. Then, the supernatant of the medium was discarded and the cells were cultured under the aforementioned routine conditions for another 7 days. Subsequently, the colonies were fixed with 95% ethanol for 5 min at room temperature, stained with Wright-Giemsa (Wright staining for 1 min and Giemsa staining for 25 min at room temperature). Only colonies containing >50 cells were counted were counted under an optical microscope (BX41; Olympus Corporation) with white light by eye.

To knockdown the expression of integrin α5, siRNAs targeting integrin α5 (si-integrin α5) and a scrambled negative control (NC) were transfected into PC9/IcoR and 827/IcoR cells were transfected with 0.1 mM siRNAs using Lipofectamine^® 2000 (Thermo Fisher Scientific, Inc.) reagent according to manufacturer's instruction at 37°C. After 48 h, subsequent experiment was performed. The coding strand of ITGA5 si1 was 5′-GUUUCACAGUGGAACUUCA-3′, and ITGA5 si2 was 5′-GCAGUGCUAUUCCCAGUAA-3′. The coding strand of NC was 5′-UUCUCCGAACGUGUCACGU-3′.

Cells (HCC827, PC9, 827/IcoR and PC9/IcoR) in six-well plate were scratched with 200-µl pipet tips. After being washed with PBS two or three times, 0 and 5 µmol/ml icotinib (for HCC827 and 827/IcoR) or 10 µmol/ml icotinib (for PC9 and PC9/IcoR) was added with RPMI-1640 medium containing 2% FBS onto the plate. The scratches were observed using the CK40 inverted light microscope (Olympus Corporation) and images were captured at 0 and 24 h and the wound width was measured using ImageJ version 1.52a (National Institutes of Health). The healing rate was evaluated relative to the starting wound width (0 h time point).

The migration assay was performed using 8-µm Transwell chambers (Corning, Inc.). PC9 or PC9/IcoR cells with a density of 3×10⁴ cells/200 µl and HCC827or 827/IcoR cells with a density of 8×10⁴ cells/200 µl were placed into the upper chamber, and 500 µl of RPMI-1640 medium containing 2.5% FBS was added to the lower chamber with 37°C. After 24 h, the remaining cells on the upper membrane were removed with a cotton swab and cells that migrated to the bottom of the membranes were stained with Wright-Giemsa staining (Wright staining for 1 min and Giemsa staining for 25 min at room temperature). Next, four randomly selected fields were counted using the BX41 microscope with white lights and the average mean number of cells was presented. Images were also captured. The invasion assay used the same steps as the migration assay except for inserting 3% Matrigel into the upper chamber before seeding 3×10⁴ cells into the culture system. The Matrigel was diluted to a 1:30 ratio at 4°C and solidified at 37°C for 4 h, then the cells were seeded into the upper chamber at 37°C for 24 h.

HCC827, 827/IcoR, PC9 and PC9/IcoR cells with or without icotinib or Stattic or siRNA treatments were lysed in lysis buffer, prepared with 1% Triton X-100,50 mM Tris-HCl pH 7.4, 150 mM NaCl, 10 mM EDTA, 100 mM NaF, 1 mM Na3VO4, 1 mM PMSF, 2 µg/ml aprotinin. After quantification by Bradford assay, the protein samples were mixed with 3×loading buffer. Equal amounts of protein (20–40 µg/lane) were separated using 8% SDS-PAGE and transferred to a PVDF membrane (EMD Millipore). The membranes were blocked using 5% skimmed milk for 1 h at room temperature and incubated with primary antibodies overnight at 4°C. After exposure to appropriate secondary antibodies for 30 min at room temperature, the proteins were detected with chemiluminescence reagent (SuperSignal™ Western Pico Chemiluminescent Substrate; Thermo Fisher Scientific, Inc.) and analyzed using the Electrophoresis Gel Imaging Analysis system (version 6.12, DNR Bio-Imaging Systems, Ltd.).

Total RNA of PC9 and PC9/IcoR cells were extracted for gene expression microarray analysis using Illumina HumanHT12 v3 BeadChip (Shanghai Oui Biomedical Technology Co., Ltd.). The DEseq2 method was used to analyze the different gene expression levels between the two groups with P<0.05 and a fold-change >2.0 as the significance cut-off. Microarray data set GSE62504 were downloaded from the Gene Expression Omnibus (GEO) database (https://www.ncbi.nlm.nih.gov/gds/) (32). GSE62504 contains the expression profiling data of NSCLC sensitive EGFR-mutant cells HCC827 and EGFR-TKIs resistant HCC827-BR cells, which were established based on individual clones by maintenance of HCC827 cells in the presence of escalating concentrations of BIBW2992 of up to 2 µM. GEO2R was used to screen differentially expressed genes using a 2-fold-change cut-off. After downloading all the differentially expressed genes, Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis was performed using the Database for Annotation, Visualization and Integrated Discovery (33). The pathways of ‘focal adhesion’ and ‘extracellular matrix (ECM)-receptor interaction’ enriched from expression profiling of PC9 and PC9/IcoR, and the pathway of ‘focal adhesion’ from microarray data of HCC827 and HCC827/BR were selected to screen potential target genes.

The experimental results were expressed as mean ± standard deviation from at least three independent experiments and analyzed using SPSS v16.0 software. The graphs were constructed using GraphPad v6.0 Software. Unpaired Student's t-test was used to analyze the differences between two independent groups. One-way ANOVA was used to analyze the differences among multiple groups, followed by Dunnett's and Sidak's multiple comparisons tests. P<0.05 was considered to indicate a statistically significant difference.

The PC9/IcoR and 827/IcoR resistant cell lines were established by gradually exposing PC9 and HCC827 parental cells to icotinib up to 10 and 5 µM, respectively. As shown in the MTT assays of 48 h, the resistant cell lines showed increased resistance to icotinib compared with parental cells, which showed dose-dependent inhibition of proliferation. Colony formation also showed that icotinib failed to inhibit the formation of colonies derived from PC9/IcoR and 827/IcoR compared with parental cells (Fig. 1B). However, when treated with icotinib for 24 h, there were no remarkable differences in the inhibition of cell viability between parental and resistant cells (Fig. 1A). Similar results were also obtained when cells were exposed to gefitinib for 48 h (Fig. S1A). Together, these results confirmed that the resistant cell lines are stable in the presence of icotinib for long period (48 h), while there was no difference between sensitive and resistant cell lines on the cell viability inhibited by icotinib in a short period (24 h).

To further investigate the differences between the resistant cells and parental cells, Transwell and wound healing azssays were performed to determine the migration and invasive capability of the cell lines. As shown in Fig. 2A, the healing rate of 827/IcoR and PC9/IcoR cells was faster compared with the parental cells, in which icotinib treatment decreased the healing rate. In addition, results from the Transwell assay showed that the number of 827/IcoR and PC9/IcoR cells that passed through the transwell membranes and the Matrigel-coated membrane was higher compared with that of the parental cells. Similarly, icotinib treatment reduced the migration and invasion abilities of HCC827 and PC9 but had little impact on the 827/IcoR and PC9/IcoR cells (Fig. 2B and C). Collectively, these results showed that the migration and invasion abilities of icotinib-resistant cells were significantly enhanced compared with their parental cells. Furthermore, icotinib could only inhibit the migration and invasion capacities of parental cells but showed no effect on resistant cells.

To determine the causes of enhanced migration and invasion in icotinib-resistant cells, a gene microarray was used to identify differentially expressed genes that were upregulated in PC9/IcoR cells compared with parental cells. The top 20 pathways were identified using KEGG enrichment analysis, among which the ‘focal adhesion’ pathway and the ‘ECM-receptor interaction pathway’ had higher scores, indicating that these significantly enriched pathways were mostly associated with metastasis (Fig. 3A). The gene microarray data of HCC827 and HCC827/BR cell lines were downloaded in the GSE62504 microarray dataset and the ‘focal adhesion’ pathway was identified using GEO2R and KEGG analysis. To narrow the range and to obtain more precise results, the intersection of the three pathways was identified, as shown in Fig. 3B. Finally, four genes were identified, including integrin α5 and integrin β4, as members of the integrin family of proteins.

To verify these results, western blotting analysis was performed to identify the protein expression levels of integrin α5 and integrin β4 in the cell lines (Fig. 3C). The data showed that the protein expression levels of integrin α5 were higher in 827/IcoR and PC9/IcoR cells compared with those in parental cells. The level of integrin β4 expression was enhanced in PC9/IcoR cells but not in 827/IcoR cells compared with their respective parental cell lines. Moreover, knockdown of integrin α5 did not change the expression level of integrin β4, implying no association between them (Fig. S1C). These data suggested that it was integrin α5 instead of integrin β4 that played a key role in icotinib resistance.

Compared with treatment with NC siRNA, treatment with si-ITGA5 (si1 and si2) suppressed the expression of integrin α5 protein in 827/IcoR and PC9/IcoR cells (Fig. 4A). Transwell assays showed that the number of cells that had passed through the Transwell membranes were significantly decreased after transfection with si1 and si2 (Fig. 4B). However, the migration resistant cells decreased when treated with si-ITGA5 and icotinib together (both P<0.001; Fig. 4C). Moreover, the cell viability of 827/IcoR and PC9/IcoR cells were partly decreased when treated with integrin α5 siRNA and icotinib together (Fig. 4D). Therefore, the sensitivity to icotinib of resistant cells was partly restored after transfected with siITGA5.

To further clarify the underlying mechanisms of integrin α5 in icotinib resistance, western blotting was performed to identify its related signaling effects. The data showed that the expression levels of integrin α5 related signaling, such as p-FAK, p-STAT3, p-AKT and p-ERK, were higher in the resistant cell lines compared with the parental cell lines except p-SRC (Fig. 5A). The expression levels of p-FAK, p-STAT3 and p-AKT were decreased in icotinib-treated sensitive cells and did not change in icotinib-treated resistant cells except p-ERK (Fig. 5B), which were consistent with the results indicating that the migration and invasion capacity of resistant cells were not affected by icotinib.

The persistent activation of p-EGFR in PC9/IcoR and 827/IcoR may explain the resistance to gefitinib in icotinib-resistant cells (Fig. S1A and B). Furthermore, after knocking down integrin α5 using siRNAs in the resistant cells, the protein expression levels of p-FAK, p-STAT3 and p-AKT were notably decreased (Fig. 5C). When treated with STAT3 inhibitor, Stattic, the phosphorylation level of AKT was significantly decreased, whereas no change of FAK phosphorylation was observed (Fig. S1D). These results suggested that the integrin α5/FAK-STAT3/AKT signaling pathway played a notable role in icotinib-resistant cells.