TGFβ1/integrin β3 positive feedback loop contributes to acquired EGFR TKI resistance in EGFR-mutant lung cancer

Abstract Inevitable emergence of acquired resistance to EGFR TKIs including third-generation TKI osimertinib limits their long-term efficacy in treating EGFR-mutant lung cancer. A fuller investigation of novel molecular mechanisms underlying acquired resistance is essential to develop efficacious therapeutic strategies. Consequently, we have identified a novel TGFβ1/integrin β3 loop that contributes to the occurrence of EGFR TKI-acquired resistance. EGFR TKIs dramatically and sustainably increased the expression of both TGFβ1 and integrin β3 in in vitro and in vivo EGFR-mutant lung cancer models with acquired resistance to EGFR TKIs. Previously, we reported that integrin β3 expression was partially induced by TGFβ1 in these models. Moreover, elevated TGFβ1 in these models was secreted mostly from lung cancer cells. Mechanistically, TGFβ1 was induced and activated by overexpressed integrin β3, forming a positive feedback loop. More importantly, the interruption of TGFβ1/integrin β3 positive feedback loop was shown to dramatically delay the occurrence of acquired resistance and greatly improve the efficacy of EGFR TKI in treating EGFR-mutant lung cancer. Taken together, our study first demonstrated the TGFβ1/integrin β3 loop a new mechanism and target for acquired EGFR TKI resistance in EGFR-mutant lung cancer.


Introduction
Lung cancer is the leading cause of death worldwide with a dismal five-year survival rate of less than 20%. Although EGFR targeted therapy has improved the clinical outcomes of lung cancer patients carrying EGFR mutation, the inevitable development of acquired resistance to EGFR TKI, including third-generation osimertinib, leads to tumour progression. The tumour microenvironment (TME) not only plays a pivotal role during tumour initiation, progression and metastasis but also has profound effects on therapeutic efficacy [1]. The survival signals from soluble factors or cell adhesion molecules within the TME could affect drug response and mediate resistance, such as integrins and TGFb [2][3][4].
Integrin is a major family of cell surface receptors which are heterodimers of a and b subunit. Integrin is expressed in a cellspecific and context-dependent manner. Integrin has been demonstrated as regulators of cancer progression, such as tumour growth, metastasis, treatment resistance and cancer stemness [5]. Integrin has also been implicated in acquired resistance to EGFR TKI in lung cancer [6,7]. Our previous work has reported that integrin b3 expression was dramatically and consistently upregulated in in vitro and in vivo EGFR-mutant NSCLC models with acquired resistance to EGFR TKIs [8]. Mechanistically, integrin b3 was induced partially by elevated TGFb1 in acquired TKI-resistant lung cancer as we reported previously [8].
TGFb1 is a ubiquitous and pleiotropic cytokine that plays a dual role in carcinogenesis. In the early stage of the disease, it acts as a tumour suppressor, while in later stages it acts as a tumour promoter by activation of cancer cell proliferation, epithelial-to-mesenchymal transition (EMT), metastasis and immune escape [3,4]. The upregulation of TGFb1 is often correlated with poor patient prognosis, early recurrence after surgery and therapeutic resistance [9,10]. Shen et al. reported that TGFb1 overexpression was correlated with EGFR TKI resistance and poor prognosis in non-small cell lung cancer (NSCLC) patients [11]. However, the underlying mechanisms and whether TGFb1 inhibition could improve the outcome of cancer patients are not clear yet.
The aim of our investigation was to study how TGFb1 was elevated in acquired TKI-resistant lung cancer and whether TGFb1 inhibition could improve the efficacy of EGFR TKIs. We have established in vitro and in vivo EGFR-mutant NSCLC models with acquired resistance to EGFR TKIs, including third-generation TKI osimertinib. In this study, we reported that elevated TGFb1 in these models was secreted mostly from lung cancer cells. Mechanistically, TGFb1 was induced by overexpressed integrin b3 in acquired TKI-resistant lung cancer, forming a positive feedback loop. Moreover, the interruption of TGFb1/integrin b3 loop was shown to dramatically delay the occurrence of acquired resistance and greatly improve the efficacy of EGFR TKI in treating EGFRmutant lung cancer. Taken together, our study first demonstrated the TGFb1/integrin b3 loop a new mechanism and target for acquired EGFR TKI resistance in EGFR-mutant lung cancer.

Cell culture and establishment of EGFR TKI-resistant lung cancer cell lines in vitro
Human NSCLC HCC827 cells, human normal lung epithelial 16HBE cells, human lung fibroblast MRC5 cells and human umbilical vein endothelial cells (HUVEC) were purchased from the Cell Bank of Type Culture Collection of the Chinese Academy of Sciences (Shanghai, China). Cells were cultured following standard protocol as previously described [8,12] and tested by certified third-party laboratories for authenticity using short tandem repeat analysis and examined for mycoplasma regularly. Gefitinib-or osimertinibresistant HCC827 cells (HCC827GR, HCC827OR) were established by the stepwise escalation method and maintained as previously described [8,12].

Mouse xenograft models and establishment of EGFR TKIresistant lung cancer tumours in vivo
Male athymic BALB/c nude mice were purchased from the Shanghai Laboratory Animal Centre (Chinese Academy of Sciences, Shanghai, China) and housed in environmentally controlled, specific pathogen-free conditions for one week before the study. All experimental procedures were reviewed and approved in accordance with the guidelines for the care and use of laboratory animals and obtained informed written consent at Shanghai Jiao Tong University.
To establish mouse xenograft models, the same amount of the indicated tumour cells (HCC827, HCC827GR or HCC827OR) was injected subcutaneously into both flanks of each mouse. The tumour volume was measured one week after injection and then every other day. Tumour volumes (mm 3 ) were calculated as length Â width 2 /2.

Isolation of primary cancer cells and cancer-associated fibroblasts from xenograft tumours
Xenograft-bearing athymic mice were euthanized by cervical dislocation. Xenograft tumours were collected, minced and enzymatically digested with a mixture of collagenase I (1 mg/ml) and hyaluronidase (1 mg/ml) in DMEM supplemented with 10% FBS, 2 mM glutamine and 1% Penicillin/Streptomycin in 37 C for 1 h with gentle mixing.
After digestion, the suspension was centrifuged at 40 Â g for 2 min to separate the epithelial and fibroblast cells. Fibroblast cells in the supernatant were pelleted by centrifugation at 100 Â g for 5 min at 4 C followed by two washes with DMEM and then resuspended and cultured in fibroblast medium (Cell Biologics) at 37 C with 5% CO2. All the fibroblasts used in the experiments were at early passage (between three and seven).
The pellet containing tumour epithelial cells was re-suspended in DMEM, filtered through a 40 mM cell strainer and then centrifuged at 100 Â g for 5 min at 4 C followed by two washes with DMEM and then re-suspended and cultured in DMEM medium supplemented with 10% FBS, 2 mM glutamine and 1% Penicillin/Streptomycin at 37 C with 5% CO2.

Cell characterisation by immunohistochemistry staining
General cell morphology was viewed under an inverted microscope. Immunohistochemistry staining were performed using standard protocols. The antibodies were listed in Supplementary  Table S1.

TGFb1 ELISA
Protein levels of active and total TGFb1 were determined using Legend MAX ELISA kits (Biolegend) following the manufacturer's instructions.

Western blot and qRT-PCR assays
Protein and mRNA expression levels were measured by Western blot and qRT-PCR assays, respectively. b-actin was used as a loading control for Western blots, and GAPDH was used as a control for qRT-PCR. The lists of antibodies used are available in Supplementary Tables S1. The primers used for qRT-PCR analysis of TGFb1 and GAPDH were previously reported [8].
Reporter constructs and dual-luciferase assay The 1.3 kb promoter region of human ITGB3 gene (NM_008332) or 0.8 kb promoter region of human TGFb1 gene (J04431) was cloned and inserted into multiple cloning sites of psiCheck2 vector (Promega) to construct psiCheck2-ITGB3 and psiCheck2-TGFb1, respectively, which were verified by sequencing. A 1.0 kb random sequence was inserted into the same site of psiCheck2 to construct psiCheck2-NC which was used as a negative control. Dual-Luciferase Reporter Assay System (Promega) was used to measure luciferase activity according to manufacturer's manual.

Statistical analysis
All data are presented as the mean ± SEM. Statistical analysis was conducted using GraphPad Prism 7.0 software (La Jolla, CA, United States). Differences between groups were examined using Student's t test and two-way ANOVA. Differences were considered significant if the p value was less than .05.

Tgfb1 level was elevated in acquired EGFR TKI-resistant lung cancer cells and xenograft tumours
First, we examined the TGFb1 level in acquired EGFR TKI-resistant lung cancer cell lines HCC827GR and HCC827OR established by stepwise escalation method [8,12]. Both active and total concentrations of TGFb1 were dramatically elevated in the culture media of HCC827GR and HCC827OR cells compared to those of their parental HCC827 cells as showed in Figure 1(A) upper panel detected by ELISA assay. We then treated HCC827 cells with gefitinib or osimertinib for 72 h and found that active and total concentrations of TGFb1 were also increased significantly in the culture media, suggesting that EGFR TKI treatment induced TGFb1 secretion and activity ( Figure 1(B) upper panel). Moreover, we examined the protein levels of receptors of TGFb1 (TGFbR1 and TGFbR2) in the above experiments. As shown in Figure 1(A) lower panel, compared to parental HCC827 cells, there were no changes in TGFbR1 levels and slight increases in TGFbR2 levels in both HCC827GR and HCC827OR cells. As shown in Figure 1(B) lower panel, gefitinib or osimertinib treatment of 72 h decreased the expression of TGFbR1 and TGFbR2, the latter with larger extent, in HCC827 cells. These results suggested that short-term EGFR TKI treatment decreased the expression of both TGFb1 receptors, while longterm treatment had no effects on TGFbR1 levels and slight increased TGFbR2 levels. Therefore, long-term EGFR TKI treatment mainly increased TGFb1 level to activate downstream signalling.
Next, in vivo models were established by treating HCC827 xenograft tumour-bearing nude mice with daily gefitinib or osimertinib for two to three months as we previously reported [8,13,14]. Similarly, both active and total concentrations of TGFb1 were also dramatically elevated in EGFR TKI-resistant xenograft tumours (xGR-HCC827 and xOR-HCC827, Figure 1(C)) compared to parental/sensitive xenograft tumours (xHCC827).

Tgfb1 was mainly secreted by acquired EGFR TKI-resistant lung cancer cells
To establish whether EGFR TKI-induced TGFb1 was secreted by lung cancer cells, first human normal lung epithelial 16HBE cells, human lung fibroblast MRC5 cells and human umbilical vein endothelial cells (HUVEC) were treated with gefitinib for 72 h and then total concentrations of TGFb1 were examined by ELISA assay. As shown in Figure 1(D), there were no significant changes in total concentration of TGFb1 in all these three cell lines after gefitinib treatment, suggesting EGFR TKI-induced TGFb1 was mostly secreted by lung cancer cells. To extend our findings in vivo, we isolated cancer-associated fibroblast (CAF) and osimertinib-resistant HCC827 cells (xOR-HCC827) from same HCC827 xenograft tumours which were treated with osimertinib daily for two to three months. An epithelial marker cytokeratin 7 (CK7) or fibroblast marker a-SMA was used to identify xOR-HCC827 and CAF cells, respectively. As shown in Figure S1, xOR-HCC827 cells were stained strongly positive for CK7 and negative for a-SMA, while CAF cells were stained strongly positive for a-SMA and negative for CK7. We treated xOR-HCC827 cells with osimertinib for 72 h in the presence or absence of CAF and found that TGFb1 levels were elevated only in the xOR-HCC827 cells alone or with CAF group treated with osimertinib ( Figure 1(E)). These results demonstrated that EGFR TKI-induced TGFb1 was mostly secreted by acquired resistant lung cancer cells.

Tgfb1 and integrin b3 formed a positive feedback loop
We previously reported that overexpressed integrin b3 in EGFR TKI-acquired resistant lung cancer was partially induced by elevated TGFb1 [8]. To further investigate the interaction between integrin b3 and TGFb1, HCC827 cells were transfected with plasmid psiCHECK-ITGB3, which was previously established by inserting ITGB3 promoter into psiCHECK2, and then treated with TGFb1 [8]. Consistent with our previous findings, as shown in Figure 2(A) left panel, TGFb1 significantly increased luciferase activity in psiCHECK-ITGB3 group compared to psiCHECK-NC. Furthermore, HCC827OR cells were transfected with psiCHECK-ITGB3 and then treated with TGFbRI inhibitor SB-431542. As expected, SB-431542 significantly decreased luciferase activity in psiCHECK-ITGB3 group compared to psiCHECK-NC (Figure 2(A), right panel). Those results demonstrated that TGFb1 induced integrin b3 expression in EGFRmutant lung cancer. Next, to investigate the effect of integrin b3 on TGFb1 expression, we inserted TGFb1 promoter into psiCHECK2 to construct psiCHECK-TGFb1 and then transfected psiCHECK-TGFb1 or psiCHECK-NC into HCC827 or HCC827GR cells. And then HCC827 cells were treated with ITGB3 overexpressing lentivirus (LV-ITGB3) to overexpress integrin b3 and HCC827GR cells were treated with specific integrin b3 siRNA (siITGB3) to knockdown integrin b3 expression or selective integrin avb3 inhibitor c(RGDfK) to inhibit integrin b3 downstream signalling. As shown in Figure 2(B), overexpression of integrin b3 in parental HCC827 cells increased luciferase activity, while antagonising integrin b3 pharmacologically by c(RGDfK) or genetically by siITGB3 decreased luciferase activity, suggesting that integrin b3 regulated the expression of TGFb1 in EGFR-mutant lung cancer cells. Moreover, we treated EGFR TKIsensitive parental HCC827 cells with LV-ITGB3 to overexpress integrin b3, Western blot and RT-qPCR showed that TGFb1 activity and expression including protein and mRNA were both increased (Figure 2(C,D)). On the contrary, EGFR TKI-resistant cells HCC827GR or HCC827OR were treated with c(RGDfK) or siITGB3, both TGFb1 activity and expression were both decreased (Figure 2(E-H), Figure  S2). Consistently, as shown in Figure 2(I), correlation analysis revealed a significant positive correlation in terms of mRNA expression between TGFb1 and integrin b3 using online web tool GEPIA (http://gepia.cancer-pku.cn/) [13]. Taken together, in EGFRmutant lung cancer cells, TGFb1 and integrin b3 formed a positive feedback loop by reciprocally inducing each other expression.
Blocking TGFb1 or integrin b3 individually inhibited reciprocal downstream signalling through integrin b3/FAK/Src/ERK and TGFb1/Smad pathways To further demonstrate that TGFb1 and integrin b3 could form a positive feedback loop, we blocked TGFb1 or integrin b3 individually to examine the effects on reciprocal downstream signalling. As shown in Figure 3(A,B), Figure S3A and Figure S3B, EGFR TKIacquired resistant HCC827GR and HCC827OR cells were treated with integrin avb3 inhibitor c(RGDfK) or siITGB3 and Western blot showed that p-Smad 2 and p-Smad 3 were dramatically downregulated compared to Ctrl or siNC, respectively. Then, HCC827GR and HCC827OR cells were treated with SB-431542 and Western blot showed that the levels of p-FAK, p-Src and p-ERK were significantly decreased (Figure 3(C)) and Figure S3C. Moreover, integrin b3 was overexpressed using LV-ITGB3 in EGFR TKI-sensitive HCC827 cells and p-Smad 2 and p-Smad 3 levels were dramatically induced in LV-ITGB3 group compared to LV-NC group (Figure 3(D)). In all above experiments, the protein levels of TGFbR1 and TGFbR2 were not changed (Figure 3(A-D) and Figure S3).
Next, HCC827 cells were treated with exogenous TGFb1 in the presence or absence of TGFbRI inhibitor SB-431542. As shown in Figure 3(E), TGFb1 dramatically induced the expression levels of integrin b3 and downstream p-FAK, p-Src and p-ERK, which were abrogated by pre-treatment with SB-431542. Then, HCC827 cells were treated with avb3 inhibitor c(RGDfK) in the presence or absence of exogenous TGFb1. As shown in Figure 3(F), c(RGDfK) significantly inhibited the TGFb1 downstream Smad2/Smad3 signalling by dramatically decreased the phosphorylation of Smad2 and Smad3, which were abrogated by pre-treatment with exogenous TGFb1. As expected, in these two recovery experiments, the protein levels of TGFbR1 and TGFbR2 were not changed ( Figure  3(E,F)). These results demonstrated that blocking TGFb1 or integrin b3 individually inhibited reciprocal downstream signalling.
Furthermore, to explore the effect of blocking TGFb1 or integrin b3 on EGFR TKI-induced TGFb1/integrin b3 positive loop activation, we treated HCC827 cells with osimertinib in the presence or absence of SB-431542 or c(RGDfK). As shown in Figure  3(G), osimertinib dramatically induced the expression levels of integrin b3 and activation of downstream FAK/Src/ERK which were abrogated by pre-treatment with SB-431542. Similarly, as shown in Figure 3(H), osimertinib dramatically induced the expression levels of integrin b3 and activation of TGFb1 downstream Smad2/3 which was abrogated by pre-treatment with c(RGDfK).

Dual blocking of TGFb1 and integrin b3 overcame acquired EGFR TKI resistance in vitro and in vivo
Next, we explored the effects of dual-blocking TGFb1 and integrin b3 using SB-431542 and c(RGDfK) on cell proliferation/apoptosis and acquired resistance to EGFR TKI in EGFR-mutant lung cancer. As shown in Figure 4(A) and Figure S4A, single agent SB-431542 or c(RGDfK) slightly or moderately promoted apoptosis of EGFR TKI-resistant HCC827GR and HCC827OR cells, while combination of SB-431542 and c(RGDfK) (Combo) significantly and dramatically increased the percentage of apoptotic cells by flow cytometry analysis. Western blot also showed that cleaved PARP and cleaved caspase-3 expression (Figure 4(B) and Figure S4B) were slightly upregulated in single agent SB-431542 or c(RGDfK) group, while dramatically increased in combo group. These data demonstrated that dual-blocking TGFb1 and integrin b3 dramatically promoted apoptosis in EGFR TKI-resistant cells. To examine the potential toxic effects of combination of gefitinib, SB-431542 and c(RGDfK), we treated 16HBE, MRC5, HUVEC and HCC827GR cells with triple combo and examined the levels of cleaved PARP and cleaved caspase-3. As shown in Figure S4C, triple combo dramatically increased the levels of cleaved PARP and cleaved caspase-3 in HCC827GR cells but had no such effects in other cells, suggesting that triple combo had minimal toxic effects on normal cells.
Then, HCC827GR or HCC827OR cells were treated with SB-431542 or c(RGDfK) or combo in the absence or presence of EGFR TKI. IncuCyte growth curves showed that SB-431542 had no effect on cell proliferation and c(RGDfK) or EGFR TKI alone slightly inhibited cell proliferation (Figure 4(C) and Figure S4D) compared to control, while EGFR TKI þ SB moderately inhibited proliferation and EGFR TKI þ c(RGDfK), SB þ c(RGDfK) or EGFR TKI þ SB þ c(RGDfK) almost completely inhibited proliferation of EGFR TKI-resistant cells. Those results suggested that dual blocking could significantly inhibit resistant cell proliferation and reverse acquired resistance to EGFR TKI. Furthermore, we sought to determine whether dual-blocking TGFb1 and integrin b3 prevents or delays the emergence of acquired resistance. Low confluence HCC827 cells (200-500/well) were seeded and treated in a 96-well plate, and wells of 50% or greater confluence were scored as positive weekly [15]. We found that an EGFR TKI in combination with c(RGDfK) or SB significantly reduced the percentage of positive wells compared with single agents while an EGFR TKI in combination with both c(RGDfK) and SB completely prevent the occurrence of positive wells up to six weeks (Figure 4(D)). Those results indicated that single agent SB-431542 or c(RGDfK) could effectively delay while combination of SB-431542 and c(RGDfK) prevent the emergence of acquired resistance to EGFR TKI.
To extend our findings in vivo, subcutaneous xenograft tumours were established using HCC827GR cells. Then xenograft tumour-bearing nude mice were treated with gefitinib daily (12.5 mg/kg) with or without combination of SB-431542 (10 mg/ kg) and c(RGDfK) (10 mg/kg). As shown in Figure 5(A), SB-431542 or c(RGDfK) alone had no effect on tumour growth compared to control, while gefitinib had some inhibitory effects on HCC827GR xenograft tumours for the first week which did not last long and tumours gradually grew back. Importantly, triple combination of gefitinib, SB-431542 and c(RGDfK) had dramatically shrunk tumour burden and inhibited tumour growth up to seven weeks. Moreover, Western blot showed that gefitinib treatment slightly promoted resistant tumour apoptosis compared to control group, while triple combination treatment dramatically promoted apoptosis by greatly inducing c-PARP and c-caspase 3 expression ( Figure 5(B)). Consistently, gefitinib induced integrin b3 expression and downstream FAK/Src activation which was inhibited not only by c(RGDfK) but also by TGFb1 signal inhibitor SB-431542. Similarly, gefitinib induced TGFb1 levels and downstream Smad activation which was inhibited not only by c(RGDfK) but also by integrin b3 signal inhibitor SB-431542 treatment ( Figure 5(B,C)). Taken together, dual blocking of TGFb1 and integrin b3 significantly delayed the progression of acquired resistance to EGFR TKI by interrupting the positive TGFb1/integrin b3 feedback loop.

Discussion
The roles of integrin b3 in acquired EGFR TKI resistance in EGFRmutant lung cancer have been reported [6]. Seguin et al. found that in erlotinib-resistant lung cancer xenograft tumours and patients, integrin b3 expression was significantly upregulated [6]. Consistently, in EGFR TKI-acquired resistant lung cancers, we previously found that integrin b3 was dramatically and consistently overexpressed which was induced by elevated TGFb1 [8]. Surprisingly, ITGAV, ITGA5 and ITGA4 were not overexpressed in EGFR TKI-resistant lung cancers and not induced by EGFR TKI treatment even though they could be induced by exogenous TGFb1. The underlying mechanisms were not clear yet [8]. In this study, we found that in both EGFR-mutant lung cancer cells and xenograft tumours with acquired resistance to EGFR TKI, TGFb1 production, secretion and activation were all increased. In sensitive EGFR-mutant lung cancer cells, TGFb1 production, secretion and activation could also be induced by EGFR TKI treatment.
TGFb1 is secreted in a latent form and activated via various mechanisms [3]. Total and active forms of TGFb1 were measured with a sandwich ELISA Kit. The sources of TGFb1 in tumours vary and include the cancer cells themselves as well as various cells of the tumour stroma [14]. To determine the sources of EGFR TKIinduced TGFb1, we first treated normal lung epithelial 16HBE cells, human lung fibroblast MRC5 cells and human umbilical vein endothelial cells (HUVEC) with gefitinib and found that TGFb1 secreted by those cells were not increased. And then we isolated CAF from osimertinib-resistant HCC827 xenograft tumours and found that elevated TGFb1 was from resistant tumour cells but not from CAF. These in vitro and in vivo results established that EGFR TKI-induced TGFb1 was mostly produced and secreted by lung cancer cells.
In our previous study, we found that TGFb1 could dramatically and consistently induce integrin b3 expression on transcriptional level and activate downstream signalling [8]. As expected, in this study, we used TGFbRI inhibitor SB-431542 to block TGFb1 signalling and found that integrin b3 expression was reduced on transcriptional level and downstream FAK, Src and Erk signalling was inhibited, further demonstrating that TGFb1 positively regulate integrin b3 signalling. Furthermore, integrin av has been reported to locally activate TGFb1, while other integrins remained to be confirmed [15]. Therefore, to investigate the effects of integrin b3 on TGFb1 activation and expression in EGFR-mutant lung cancer, we used lentivirus to overexpress and siRNA to knockdown integrin b3 or integrin avb3 inhibitor c(RGDfK) to inhibit integrin b3 downstream signalling and found that integrin b3 could positively regulate TGFb1 production on transcriptional, secretion and activation of downstream SMAD signalling. Moreover, exogenous TGFb1-induced or osimertinib-induced integrin b3 expression and downstream FAK, Src and Erk signalling activation was abrogated by SB-431542, while osimertinib-induced integrin b3 expression and downstream Smad signalling activation was abrogated by c(RGDfK). Also, the inhibition of SMAD signalling by c(RGDfK) was recovered by exogenous TGFb1. Put together, our study first showed that TGFb1 and integrin b3 formed a positive feedback loop by reciprocally inducing the expression of each other on transcriptional level and activating reciprocal downstream signalling.
Therefore, we hypothesised that dual blocking of TGFb1 and integrin b3 could improve the efficacy of EGFR TKI against EGFRmutant lung cancer. Indeed, we found that combination of SB-431542 and c(RGDfK) promoted apoptosis, inhibited growth and increased sensitivity of acquired resistant cells to EGFR TKI treatment as well as greatly delayed the occurrence of acquired resistance. Furthermore, in in vivo models, triple combination of gefitinib, SB-431542 and c(RGDfK) dramatically and significantly repressed the growth of resistant xenograft tumours and induced resistant tumour apoptosis as well as inhibited FAK/Src/ERK and Smad signalling. Also, triple combination had minimal toxic effects on normal cells. Taken together, our study first demonstrated the TGFb1/integrin b3 loop a new mechanism and target for acquired EGFR TKI resistance in EGFR-mutant lung cancer.

Author contributions
YZ, LL and LX designed all the experiments. YZ, HC and TW conducted the experiments, analysed and interpreted the results. LX supervised the project. YZ, LX and LL wrote the draft manuscript. LL and LX reviewed and edited the manuscript. All authors read and approved the final manuscript.

Disclosure statement
No potential conflict of interest was reported by the author(s).

Funding
The work was supported by National Natural Science Foundation of China (81773747 and 81372522).