Integrin β3 Inhibition Enhances the Antitumor Activity of ALK Inhibitor in ALK-Rearranged NSCLC

Anaplastic lymphoma kinase (ALK)-rearranged non–small cell lung cancer (NSCLC) behaves more aggressively than other NSCLCs, showing early metastasis and an advanced stage at the time of diagnosis (1–3). Similarly, patients with ALK-rearranged NSCLC and high expression of ALK are more likely to be in advanced stages and have worse 2-year progression-free survival (PFS) rates (4). Nevertheless, patients who have advanced ALK-positive NSCLC are highly responsive to the ALK inhibitor crizotinib (Xalkori, Pfizer), with an objective response rate (ORR) of approximately 60% and a median PFS of 8 to 10 months (5). Because of their benefits over other therapies, ALK tyrosine kinase inhibitors (ALK TKI) are used as the chemotherapeutic agents of choice in NSCLC with ALK rearrangement (6). However, most cases receiving ALK inhibitors relapse within a few years after starting therapy (7–10). Ongoing research mainly attempts to develop new small-molecule inhibitors with greater potency against ALK and activity against acquired mutations (11). Unfortunately, the second- and third-generation ALK inhibitors also show failure (8, 12). Several studies have suggested that the combination of two or more therapeutic treatments is more effective than treatment with ALK TKI alone. Treatments that add therapeutic targets, such as IGF-1R or PD-L1 along with ALK, was shown to improve responses through a synergistic or additive effect in patients with ALK-rearranged NSCLC (12–14).

The integrin family has been studied extensively in cancer cases because of their crucial role in tumor cell signal transduction and cellular communication with the microenvironment (15). With at least 80 clinical trials on the application of integrins ongoing, these proteins have long been the focus of the biotechnology and pharmaceutical industries as potential therapeutic targets, given their central roles in almost all aspects of human biology, as well as in the pathobiology of many diseases (16). In NSCLC, integrin expression has been reported in tumor cells and the cellular matrix and has been found to accompany angiogenesis (17). Integrins have also been reported to work in conjunction with tyrosine kinases, leading to angiogenesis, cell proliferation, and cell survival (18). There have been studies targeting integrins in NSCLC, such as the CERTO phase II trial, in which integrin β3 and integrin β5 were combined with cetuximab and platinum-based chemotherapy (19). Even though a good response was not elicited, the trial showed that cilengitide, a dual β3 and β5 antagonist, had no additional toxicity or safety concerns (19, 20).

In this study, we conducted a multiplex gene expression analysis of gene expression profiles in a large cohort of ALK-rearranged NSCLC cases. We also conducted clinicopathologic analyses, assessed the correlation between integrin β3 and ALK expression levels, and tested responses to a combination treatment in in vitro assays and in vivo models. Our study provides the rationale for cotargeting ALK with noncanonical targets that become available through ALK rearrangement in NSCLC and indicates a new potential therapeutic target in ALK-fusion–driven NSCLC.

Three thousand primary lung cancer patients who visited Samsung Medical Center (SMC) from 2001 to 2014 and received curative resection or biopsy for lung cancer were screened. Baseline clinical characteristics and mutational status of epidermal growth factor receptor (EGFR) and KRAS were extracted from in-hospital charts and electronic medical records. As shown in Fig. 1A, squamous cell carcinoma, small cell carcinoma, recurrent cancer, and EGFR or KRAS mutation-positive cases were excluded. A cohort of patients with ALK-rearranged NSCLC was established, based on ALK-positive immunohistochemistry (IHC) and fluorescence in situ hybridization (FISH), which was performed as described previously (21). Excluding RNAs that could not pass the quality control, a total of 100 ALK-rearranged NSCLC samples were included in the gene expression panel, along with samples from 14 cases of ALK-negative NSCLC. In the ALK-negative NSCLC cohort, four had EGFR mutations, four had KRAS mutations, and six exhibited EGFR wild-type, KRAS wild-type, and ALK-negative genotypes (Supplementary Table S1). The samples used for β3 IHC in Fig. 1E derived from the initial 3,000 patients in the NSCLC cohort; 87 ALK-rearranged NSCLC cases were drawn from the 158 patients in the ALK-rearranged NSCLC cohort, and 135 ALK-negative NSCLC cases were drawn from those that did not show ALK positivity by ALK IHC or ALK rearrangement by FISH.

In addition, we established a consecutive serial cohort of four patients with ALK-rearranged NSCLC who had been treated with crizotinib and for whom pre-crizotinib and post-crizotinib formalin-fixed paraffin-embedded (FFPE) blocks were available. Post-crizotinib refers to samples obtained at the time of acquired resistance. Treatment information is shown in Supplementary Table S2. All patients provided informed consent; this study was approved by the SMC Institutional Review Board (IRB No. 2015-01-108-002) and in accord with the principles listed in the Helsinki declaration.

From the collected ALK-rearranged NSCLC FFPE samples with sufficient tissue remaining, RNA was extracted using an RNeasy FFPE kit according to the manufacturer's instructions (Qiagen). Samples were further prepared as previously described (22) and processed using an nCounter PanCancer Pathways Panel (XT-CSO-PATH1-12), which allowed testing for 730 genes related to cancer (www.nanostring.com). Briefly, 200 ng of RNA was used, and the results were normalized using NanoString nSolver software. After performing image quality control using a predefined cutoff value, we excluded outlier samples using a normalization factor defined as the sum of positive control counts greater than threefold. Therefore, of the 100 ALK-rearranged NSCLC samples, seven were excluded. A two-sample t test was used to determine the significance of differential gene expression in ALK-rearranged and ALK-negative NSCLC samples. An adjusted P value was calculated using a single-step procedure (23) to control the family-wise error rate (FWER). Power was calculated using a single-step procedure (24) using the following parameters: FWER alpha = 0.05, number of genes = 730, ALK-rearranged samples = 93, ALK-negative samples = 14, number of predictive genes selected = 4, delta = effect size of the four predictive genes selected, and block size = 10, with a common correlation coefficient of rho = 0.1 and K = 10,000 replications. The distribution of integrin β3 mRNA expression levels was determined based on the cutoff value that showed the greatest difference in clinical parameters (Supplementary Table S3).

Calu3, H23, H2228, and BEAS-2B were purchased from American Type Culture Collection. H3122 was a generous gift from Dr. Christine M. Lovly (Vanderbilt University). Cell lines H23, H2228, and H3122 were cultured in RPMI 1640 medium, and Calu3 and BEAS-2B were maintained in DMEM. All cultures were supplied with 10% fetal bovine serum and a 1% antibiotic-antimycotic solution (both from Thermo Fisher Scientific). All cell lines used were authenticated by short tandem repeat (STR) analysis and determined mycoplasma free (both from Applied Biosystems) every 6 months. In addition, all cell lines were used and collected within 20 passages from the initial thaw.

ALK (1:30, clone 5A4; Novocastra) and integrin β3 (1:100, clone EPR2417Y; Millipore) antibodies were applied in the IHC analysis using previously described procedure (25). Immunoblotting was performed using pALK (Tyr1604), ALK (D5F3), pSTAT3 (Tyr705), STAT3 (79D7), pAKT (Ser473), AKT (40D4), pERK (Thr202/Tyr204), ERK, and vimentin (D21H3) (all from Cell Signaling Technology), and integrin β3, β-actin, lamin B, and α-tubulin (all from Santa Cruz Biotechnology), all under standardized procedures. ImageJ software was used for densitometry quantification (NIH). For the FACS analysis and migration assay, anti-αvβ3 (clone LM609; Millipore) was used to stain cells as previously described (26).

cDNAs were synthesized from patient samples or plasmids and then subcloned into a pLenti6.3 backbone. pLenti6.3-LacZ-V5, pLenti6.3-EML4-ALK(EA)-V5, and pLenti6.3-Luc2 vectors were transfected into BEAS-2B and Calu3 cell lines in a lentiviral form (Thermo Fisher Scientific) for stable expression. The EML4–ALK fusion used in this study was derived from a patient sample. It harbors EML4 exon 14 with 142 bases of intron 14 and ALK exon 20 with 56 bases inserted into intron 19 (E14ins142;ins56A20). pCDNA3.1–EML4–ALK (variant 1) was a generous gift from Dr. Richard Bayliss (University of Leeds). The ITGB3 promoter region [−939 bp from transcription start site (TSS)] was PCR amplified using genomic DNA from BEAS-2B cells. The PCR product was then cloned into the pGL4.10-Luc2 vector (Promega). The pRL-SV40 plasmid was a generous gift from Dr. Jeong Hoon Kim (Sungkyunkwan University).

siRNAs were custom made or purchased from Bioneer and were transfected using Lipofectamine RNAiMAX reagent (Thermo Fisher Scientific). siRNA sequences are listed in Supplementary Table S5 (27). We tested two or three siRNA sequences for ALK, integrin β3, and STAT3 and conducted further studies using ALK-1, integrin β3-1, and STAT3-1 siRNAs, since they exhibited the most specific knockdown and greatest knockdown efficiencies (Fig. 2C and D; Supplementary Fig. S2A and S2B).

For the in vitro inhibitory assays, BEZ235, PD0325901, S3I-201, crizotinib, ceritinib, alectinib, brigatinib, and cilengitide were all purchased from Selleck Chemicals and diluted in dimethyl sulfoxide (DMSO).

Crizotinib and ceritinib (both from LC laboratories), as well as cilengitide (Sigma-Aldrich), were used in in vivo experiments and diluted in 0.1 N HCl, 0.5% Tween-80 and 0.5% methyl cellulose, and in physiologic saline buffer, respectively.

RNA was extracted using an RNeasy kit (Qiagen), and cDNA synthesis was performed using a SuperScript III First-Strand Synthesis kit (Thermo Fisher Scientific) according to the manufacturer's instructions. The PCR was conducted as previously described (28). The sequences of the primers used are listed in Supplementary Table S5. Because of the homology to HPRT1, the same primers were used for hprt1 amplification in Ba/F3 samples.

BEAS-2B and H3122 cells were seeded at 1 × 10⁵ cells/well into 12-well plates a day prior to transfection. Cells were transfected with the following plasmids and harvested for use in luciferase assays (Dual-Glo Luciferase assay system, Promega): reporter plasmids (100 ng), pGL4.10-Luc2 and pGL4.10-ITGB3-Luc2; EML4-ALK expression vectors (100 ng), pCDNA3.1-V5 and pCDNA3.1-EML4-ALK; Renilla expression vector (10 ng), pRL-SV40.

Transcription Factor BIND (http://tfbind.hgc.jp/) was used to search for STAT3-binding sites within the ITGB3 promoter region (up to −3,000 bp from TSS). BEAS-2B/EA and H3122 cells, 1 × 10⁷ cells from each line, were lysed, and only the nuclear extract (NE) was used for chromatin immunoprecipitation (ChIP). Rabbit IgG and STAT3 antibodies (Cell Signaling Technology) were used for immunoprecipitation, which was performed manually as previously described (29). For the sequences of qPCR primers used, refer to Supplementary Table S5. We included an input sample, nonspecific antibody sample (IgG), and STAT3 IP-sample. STAT3 is known to bind to the c-fos region; thus, c-fos was used as a positive control in the ChIP assay. Additionally, myogenic differentiation 1 (MYOD1) was used as a negative control. The final qPCR data were normalized for differences in the amount of input chromatin.

Cells expressing the EML4–ALK fusion form (E14ins142;ins56A20) were used for the Transwell migration assay. BEAS-2B cells transfected with integrin β3 siRNA (100 nmol/L) and/or pretreated with crizotinib (0.5 μmol/L) and/or prerolled with anti-αvβ3 antibody (10 μg/mL) were seeded at a concentration of 5 × 10⁴ cells/well in the upper chamber of a conventional 8-μm pore Transwell plate (Corning). For Calu3 cells, 2 × 10⁶ cells/well were seeded into plates. BEAS-2B cells were incubated for 3 hours to assess their migratory properties under chemotactic conditions; Calu3/EA cells were incubated for 24 hours. Cells were fixed in ice-cold methanol for 30 minutes and then stained with hematoxylin and eosin (H&E) to visualize the number of cells that migrated (Sigma-Aldrich). Cells from five regions were counted and averaged to obtain this number.

Pre-chilled 8-μm pore Transwell plates (Corning) were coated with 10 μL of 1 mg/mL growth factor–reduced mouse sarcoma extracellular matrix (Matrigel; BD Biosciences). BEAS-2B cells were transfected with 100 nmol/L integrin β3 siRNA 48 hours prior to the experiment and/or pretreated with 0.5 μmol/L crizotinib for 24 hours. Then, 5 × 10⁴ cells were seeded in the upper chamber and incubated for 24 hours to assess their invasive properties under chemotactic conditions. Cells were H&E stained, and five regions were selected and averaged to determine the number of invaded cells.

A pre-chilled 48-well plate was first coated with 30 μL of growth factor–reduced mouse sarcoma extracellular matrix (Matrigel; BD Biosciences). After 15 minutes of incubation at 37°C, 5 × 10⁴ BEAS-2B/EA cells were mixed with 150 μL of growth factor–reduced Matrigel and then seeded on top of the original layer. Thirty minutes after incubation, 200 μL of culture medium containing the appropriate drug concentration [DMSO (control), 1 μmol/L crizotinib (CZT), 1 μmol/L crizotinib + 5 μmol/L cilengitide (CZT + CIL), 1 μmol/L ceritinib (CER), or 1 μmol/L ceritinib + 5 μmol/L cilengitide (CER + CIL)] was added. Media were changed every 2 days for a week. Cells were extracted and analyzed for protein expression, using PBS-EDTA to break down the Matrigel. Cells were lysed and processed using standard procedures (30).

All animal experiments were performed under a protocol approved by the SMC ethics review committee (20160923001). Male 8-week-old Balb/c-nude mice were anaesthetized using 2% isoflurane and placed in the right lateral decubitus position. Inferior to the left scapula, a 2-cm skin incision was made to expose the pleura. BEAS-2B/EA/Luc2 cells (3 × 10⁶) in 70 μL of culture medium and mouse sarcoma extracellular matrix (BD Biosciences), mixed at a ratio of 1:1, was injected directly into the left lung with a 29 G insulin syringe (BD Biosciences; ref. 31). A 4-0 polypropylene suture (Prolene, Ethicon Inc.) was used to close up the skin incision, and then the mice were laid in the right lateral decubitus position under a UV lamp for recovery.

Two weeks after injection, mice were randomly distributed into six treatment groups: vehicle, 4 mg/kg cilengitide, 100 mg/kg crizotinib, 100 mg/kg crizotinib + 4 mg/kg cilengitide, 50 mg/kg ceritinib, and 50 mg/kg ceritinib + 4 mg/kg cilengitide. Mice that did not bear tumors by the end of the 2-week period were exempted from the experiment. Each treatment group was composed of 6 to 8 mice, and drug administrators were blinded. Crizotinib and ceritinib were administered orally every day, whereas cilengitide was intraperitoneally injected daily (vehicle treated with physiologic saline buffer). Mice were monitored and treated daily, imaged weekly for evidence of engraftment and progression of disease, and drug efficacy was determined based on bioluminescence intensity. Mice were only treated for 3 weeks, as mortality was observed in the vehicle group at that time.

After luciferase-expressing tumor cells were injected, mice were examined by bioluminescent imaging on a weekly basis. d-luciferin (Perkin Elmer) was diluted to a working concentration of 150 mg/kg in physiologic saline buffer prior to intraperitoneal injection. Images were taken 10 minutes after injection. After anesthetizing each animal with 2% isoflurane, bioluminescence (photons/s/cm²/steradian) was assessed at an exposure time of 1 second using an IVISw Spectrum system. Mice that did not show bioluminescence at an exposure time of 1 second or that preexpired during the study were disregarded from the final analysis. The region of interest (ROI) in images was analyzed, and bioluminescence was quantified in photon/s using Living Image version 3.2.

Statistical significances for mutation status, sex, smoking, T stage, N stage, metastasis, stage, recurrence, and death were estimated using Pearson χ² test. An independent t test was used to analyze age, tumor size, qRT-PCR, qPCR, migration, invasion, cell viability, and results of in vivo experiments (32). All tests were carried out using SPSS, with significance defined as P < 0.05.

This study examined a total of 3,000 patients with NSCLC and 158 samples from ALK IHC-positive and FISH-confirmed ALK-rearranged FFPE cases. A total of 100 ALK-rearranged and 14 ALK-negative NSCLC RNAs were included in the gene expression panel. The remaining RNAs were excluded because of unavailability, insufficiency of remaining tissue, or poor RNA quality. Basic demographic data, along with clinical and pathologic characteristics of patients, are summarized in Table 1 and Supplementary Table S1. The major clinical differences between the two groups are age, smoking history, and tumor size (all P < 0.05). The median age of ALK-negative patients with NSCLC was 64 years, whereas those with ALK-rearranged NSCLC was 53 years of age. The ALK-negative cohort contained more ever (both current and former) smokers (77.8%), and the ALK-rearranged cohort had more never smokers (61%) than ever smokers. Eleven stage 4 patients in the ALK-rearranged NSCLC cohort received the following treatments: wedge resection (N = 7), lobectomy (N = 2), and biopsy (N = 2). The ALK-rearranged cases had smaller tumor sizes at the time of resection, with a median tumor size of 2.5 cm and tumors ranging from 0.7 to 8 cm. In total, 26% of patients with ALK rearrangement showed NSCLC recurrence; the rest of these patients were reported recurrence-free at the last follow-up visit. The follow-up period varied from 2 to 150 months, with a median follow-up duration of 39 months. Death was reported for 18% of ALK-rearranged patients until the last follow-up.

To explore gene expression patterns in ALK-rearranged NSCLC cases, we assessed their gene expression levels (Fig. 1A). Interestingly, the expression patterns for 730 cancer-related genes in patients with ALK-rearranged NSCLC were relatively homogenous (Fig. 1B). Notably, five patients (PA-01, -52, 56, -69, and -100) showed decreased expression in genes associated with DNA repair, RAS, and MAPK pathways (data not shown). To investigate genes associated with ALK expression, we scored and ranked the genes that were highly expressed in the ALK-rearranged NSCLC samples compared with those from the ALK-negative cohort. As a result, we discovered four genes that were highly correlated with ALK expression levels (Fig. 1C): integrin beta-3 (ITGB3), ETS translocation variant 1 (ETV1), tumor necrosis factor receptor superfamily member 10C (TNFRSF10C), and human ephrin-A5 (EFNA5) (adjusted P = 0.0003, 0.0088, 0.0096, and 0.0302, respectively). The power of these four genes was estimated at 99.75%. Of the four genes, ITGB3 was the gene with the most statistically significant overexpression (Fig. 1C and D). This was confirmed by the higher integrin β3 IHC scores in ALK-rearranged NSCLC cases compared with those in ALK-negative NSCLC cases, based on the four-tier scoring system of tumor cells shown in Supplementary Fig. S1A (P < 0.005; Fig. 1E). We further investigated the mRNA expression levels to validate whether other integins, which were not included in the panel, were also impacted by the expression of ALK. Similar to the panel result, integrin β3 showed the most significant increased levels of expression compared with other integrins (P < 0.005; Supplementary Fig. S1B). Nevertheless, the correlation between ALK and integrin β3 levels could not be confirmed in The Cancer Genome Atlas, because only three ALK-rearranged NSCLC samples were included (Supplementary Fig. S1C; ref. 33).

We then attempted to confirm this correlation in expression levels of ALK and integrin β3 in clinical samples and ALK-rearranged cell lines. Within patients in the ALK-rearranged NSCLC cohort, integrin β3 IHC scores were significantly higher for samples with ALK IHC scores of 2 and 3 than in those with scores of 0 and 1 (P < 0.05; Fig. 1F). This analysis could be conducted only on protein expression levels, not on gene expression levels, as the available panel did not contain the probe designed for ALK-rearranged region. Both integrin β3 mRNA and protein levels were more highly expressed in the H3122 and H2228 (ALK rearranged) than in Calu3 (EGFR/KRAS wild-type and ALK negative) and H23 (KRAS mutant) cell lines (Fig. 2A and B). To ascertain the specificity of this correlation, siRNAs and ALK TKIs were administered to the ALK-rearranged cells. When ALK expression was downregulated using siRNA (Fig. 2C and D) or crizotinib and ceritinib (Fig. 2E and F), integrin β3 levels also decreased in H3122 and H2228 cells. In addition, these results were confirmed in tissue samples from patients with ALK rearrangements. Out of four, three cases showed decreased expression levels in integrin β3 after crizotinib treatment (Fig. 3A). Consistent with the above results, integrin β3 levels increased when the normal lung epithelial, BEAS-2B, Calu3, and murine pro-B (Ba/F3) cell lines expressed the EML4–ALK fusion form (E14ins142;ins56A20; Fig. 3B–E). The opposite was observed when EML4–ALK-expressing BEAS-2B cells (BEAS-2B/EA) were treated with ALK inhibitors (Fig. 3F and G). Taken together, these results indicate that ALK and integrin β3 are strongly correlated with one another, consistent with the results obtained from the multiplex gene expression analysis, clinical analysis, and in vitro experiments.

We have previously reported that in patients with ALK-rearranged NSCLC, differences in ALK expression levels exist, and that high expression levels of ALK correlate with worse prognosis, such as advanced stages and a worse 2-year PFS (4). Because these patients showed various levels of integrin β3 expression (Fig. 1D and F), we decided to assess whether their integrin β3 gene expression levels were associated with clinical characteristics (Supplementary Table S3). When the cohort was divided based on integrin β3 mRNA levels determined in the multiplex gene expression assay, patients with high integrin β3 expression levels showed more advanced stages of NSCLC (P = 0.04) and more distant metastases at the time of diagnosis or during follow-up (P < 0.005). However, higher integrin β3 IHC scores did not show significant association with tumor progression (Supplementary Table S4). Moreover, the patients expressing higher gene expression levels of integrin β3 were more likely to be current or former smokers (P = 0.05). Because higher integrin β3 expression is associated with tumor progression in clinical samples, we next assessed the therapeutic potential of this protein in vitro and in vivo.

Integrins are known to participate in mobility, adhesion, and progression of cancer cells (10, 34). To explore the antimigratory and anti-invasive properties of integrin β3 in ALK-rearranged NSCLC, we used integrin β3 siRNA, blocking antibody, or cilengitide, a specific inhibitor of integrins αvβ3 and αvβ5. Results showed that, compared with a single treatment of 0.5 μmol/L crizotinib, combination therapy with crizotinib and integrin β3 inhibitors, including siRNA (P < 0.05), anti-αvβ3 treatment (P < 0.005), and cilengitide (P < 0.005), substantially decreased migration and invasion (Fig. 4A; Supplementary Fig. S2C and S2D).

In addition, we assessed whether the combination of ALK and integrin β3 activity inhibition would result in a synergistic cell proliferation inhibition in EML4–ALK-expressing cells. In 3D-cultured BEAS-2B/EA cells, the combination of crizotinib plus cilengitide or ceritinib plus cilengitide showed significant tumor growth inhibition, compared with that with crizotinib or ceritinib alone (Fig. 4B; Supplementary Fig. S3A and S3B). In addition, treatment with alectinib plus cilengitide and brigatinib plus cilengitide resulted in increased tumor sensitivity compared with that with single alectinib or brigatinib treatment (Supplementary Fig. S3C and S3D). A decrease in survival signals, integrin β3, and vimentin was observed in the combination treatment group (Fig. 4C). Taken together, these results demonstrate the efficacy of treatment with ALK TKIs and cilengitide on ALK-rearranged lung cancer cell lines and suggest a synergistic response in cell migration, invasion, and growth.

Next, we conducted in vivo studies with BEAS-2B/EA cells expressing a luciferase gene. BEAS-2B cells were used to better reflect the patients' backgrounds. We conducted studies using lung cancer models to mimic the conditions of primary human disease, but even though the cells were injected directly into the left lung, there is no certainty whether there was a leakage of the cells from the left lung into the intrathoracic cavity through the injection wound. Similar to in vitro results, the combination treatment groups (crizotinib plus cilengitide and ceritinib plus cilengitide) exhibited greater impairment of BEAS-2B/EA_Luc2 cell growth compared with that of the single ALK treatment group (Fig. 4D). Tumor cell growth was impaired the most two weeks after treatment in both combination groups, based on bioluminescence intensity (both P < 0.05; Fig. 4D). Overall, we confirmed the tumor suppression effect of combined ALK TKIs and cilengitide treatment in in vivo lung cancer model.

ALK phosphorylation is known to activate the PI3K/AKT, MAPK/ERK, and JAK/STAT pathways, and these pathways, in turn, stimulate changes in cellular function (35). To investigate which signals mediate expression of integrin β3, we treated ALK-rearranged cells with BEZ235 (a PI3K/mTOR inhibitor), PD0325901 (an MEK inhibitor), and S3I-201 (a STAT3 inhibitor). Of the three, only S3I-201 depleted integrin β3 expression (Fig. 5A) in a dose-dependent manner (Fig. 5B). Crizotinib and ceritinib also inhibited integrin β3 expression, although S3I-201 was more effective. Similar results were observed with STAT3 siRNA (Fig. 5C). Because phosphorylated STAT3 acts as a transcription factor, we hypothesized that this protein directly regulates integrin β3 expression in ALK-rearranged NSCLC. Indeed, in BEAS-2B/EA cells and H3122 cells, phosphorylated STAT3 was more abundant in the NE than in the LacZ control or in the ALK siRNA-treated control, respectively (Supplementary Fig. S2E), and, in the ChIP assay, activated STAT3 was shown to bind to two putative STAT3-binding sites in the ITGB3 promoter region (Fig. 5D). Furthermore, the reporter gene assay confirmed that STAT3 regulates the transcription of ITGB3 in ALK-rearranged cells, because STAT3 inhibition reduced ITGB3 transcriptional activity (Fig. 5E; Supplementary Fig. S2F). Our data suggest that STAT3 activation by oncogenic ALK expression increases integrin β3 transcription and promotes tumor cell migration and invasion in cells expressing ALK.

We examined gene expression patterns of 730 cancer-related genes in 100 patients with ALK-rearranged NSCLC and identified ITGB3 as the most significantly overexpressed gene, as well as a gene with expression levels that positively correlated with those of ALK (Figs. 1–3). Integrins are known to play multiple roles in cancer (10, 34), and integrins α5β1 and αvβ5 participate in tumor cell adhesion and motility in lung cancer (17, 36). We found that high integrin β3 mRNA expression levels in patients with ALK-rearranged NSCLC were associated with more advanced stages of cancer and metastases. However, high integrin β3 IHC scores showed no association with clinical measure other than tumor size (Supplementary Table S4). This difference may be because of the higher sensitivity in assessing gene expression levels by the multiplex gene expression assay than by IHC (Supplementary Fig. S1D) (R² = 0.48). Chen and colleagues also reported superior overall performance of the NanoString-based assay compared with IHC staining (37). In addition, this difference may have been observed because of the lower sensitivity or specificity of the antibody used in the IHC assay.

Another interesting finding is that integrin β3 is regulated by phosphorylated STAT3 in cells expressing ALK. Integrins are regulated by adaptor molecules such as talin-1 and the kindlins (38, 39). Through inhibition of three ALK downstream signaling pathways, ChIP assay, and luciferase reporter assays, we discovered that activated STAT3 is directly involved in integrin β3 expression in ALK-rearranged cells (Fig. 5). This is the first report that STAT3 transcription factor directly regulates integrin β3, indicating that an ALK/STAT3/integrin β3 pathway may exist in ALK-rearranged NSCLC cases.

As reported by Seguin and colleagues, integrin αvβ3 plays a major role in lung cancer progression in patients with KRAS-mutant and EGFR inhibition–resistant lung cancers, and targeted combination therapy resulted in complete loss of the β3-positive stem cell population (40, 41). In a similar manner, our study showed the potential of dual targeting ALK and integrin β3, which resulted in significant migration and invasion inhibitions in EML4–ALK-expressing cells (Fig. 4). Moreover, we confirmed the benefits of combination therapy of ALK TKIs and cilengitide through the in vitro experiments and the in vivo lung cancer mouse model (Fig. 4). The tumor suppression effect of cilengitide seen in our study can be attributed to the inhibition of the integrin αvβ3 complex because the major partner of αv integrin was assumed to be integrin β3 and functional blocking test using anti-αvβ3 antibody was also shown to play a significant role in cell migration and invasion (Supplementary Fig. S1B and Fig. 4). However, this synergistic effect can be attributed to the inhibitions of ALK and one or more αv integrins since cilengitide targets all five αv integrins (42). Further in-depth analysis needs to be conducted in order to evaluate the role of other integrin beta subunits in tumor regression of ALK-rearranged NSCLC.

Treatment of advanced NSCLC with integrin β3 inhibitor alone or in combination with other inhibitors has failed to elicit a good response in previous clinical trials (19, 20). However, these results were obtained before tumor classification at the molecular level was routinely applied; thus, we believe that results of these studies cannot be directly translated. If the patients were selected based on their molecular subtypes, such as ALK rearrangement, and were treated in combination with tyrosine kinase inhibitors, such as ALK inhibitors, a different result may have been observed. As suggested by our study results, integrin β3 overexpression is one of the downstream effectors of ALK activation that contributes to tumor progression in ALK-rearranged NSCLC. Cilengitide administration alone would not have been sufficient enough to inhibit the growth of tumor cells as other signals related with cancer progression and survival are still active. Therefore, combination inhibition of a major oncogenic driver and additional targets, such as integrin β3, may have been required to impede the tumorigenic effects on patients with ALK-rearranged NSCLC. Furthermore, previous clinical trials showed that cilengitide combined with other chemotherapies was well tolerated and showed no dose-limiting toxicities (19). This shows that an integrin β3 inhibitor is a potential treatment option for NSCLC and supports its exploration in combination with other therapeutic agents.

Our results indicate that patients with ALK-rearranged NSCLC and integrin β3 overexpression have more aggressive tumor progression, and that integrin β3 expression levels can be a factor used to determine the therapeutic approach. Based on our clinical and experimental results, we would like to stress the importance of conducting comprehensive molecular analyses to develop noncanonical therapies that result in better prognoses in patients with ALK-rearranged NSCLC.

No potential conflicts of interest were disclosed.

Conception and design: K.-W. Noh, M.-S. Lee, Y.-L. Choi

Development of methodology: K.-W. Noh, J.-Y. Song, M. Sung, M.-S. Lee, Y.-L. Choi

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): K.-W. Noh, S.-H. Lee

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): K.-W. Noh, I. Sohn, J.-Y. Song, K. Jung, M.-S. Lee, Y.-L. Choi

Writing, review, and/or revision of the manuscript: K.-W. Noh, I. Sohn, Y.-J. Kim, J. Han, S.-H. Lee, M.-S. Lee, Y.-L. Choi

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): K.-W. Noh, H.-T. Shin, K. Jung, M. Sung, M. Kim, S. An, M.-S. Lee

Study supervision: M.-S. Lee, Y.-L. Choi

We would like to thank Dr. Christine Lovly for providing the H2228 cell line, Dr. Richard Bayliss for providing the EML4–ALK plasmid, and Dr. Varun Gupta for providing critical comments on the manuscript. We also thank Areum Han from the Multimedia Department of SMC for providing the graphical illustrations.

This work was supported by National Research Foundation of Korea (NRF) grants funded by the Korean government (Ministry of Science, ICT and Future Planning) (nos. 2015R1C1A1A02037066, 2016R1A2B2012975, 2016R1A5A2945889, and 2018R1C1B6001396).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.