The TGFβ–miR200–MIG6 Pathway Orchestrates the EMT-Associated Kinase Switch That Induces Resistance to EGFR Inhibitors

The sensitivity of some tumors to epidermal growth factor receptor (EGFR) inhibitors can be explained by the presence of mutations in the EGFR tyrosine kinase domain (1, 2). However, such mutations are rare in tumors other than non–small cell lung carcinoma (NSCLC; refs. 3–6). There is a need to elucidate the mechanisms underlying the differential drug response of cancer cells with wild-type EGFR to identify those patients who could respond and clinically benefit from tyrosine kinase inhibitors (TKI), and to develop new therapeutic strategies to circumvent the de novo or acquired resistance of tumors to EGFR inhibitors.

The response to EGFR-targeted agents is inversely correlated with epithelial–mesenchymal transition (EMT) in multiple types of tumors without known EGFR mutations, including NSCLC, head and neck (H&N), bladder, colorectal, pancreas, and breast carcinomas (7–11). Notably, epithelial tumor cells have been shown to be significantly more sensitive to EGFR inhibitors than tumor cells, which have undergone an EMT-like transition and acquired mesenchymal characteristics (11). These data suggest that EMT is a common denominator of tumors that are resistant to EGFR inhibitors. However, the precise molecular mechanisms underlying this association have not been defined and no specific EMT-associated biomarker of clinical benefit has been identified.

EMT is driven by a network of transcriptional repressors, which include SNAIL1, SNAIL2 (SLUG), ZEB1 (zinc-finger E-box binding factor), ZEB2, and TWIST (12). Transforming growth factor β (TGFβ)–activated SMAD3/4 stimulates the expression of SNAIL1 and TWIST1, which cooperate with SMAD proteins to repress the expression of epithelial genes such as CDH1 (which encodes E-cadherin; refs. 12, 13). These transcriptional effects of TGFβ cooperate with TGFBR2-mediated phosphorylation of partitioning defective 6 (PAR6) to trigger EMT (12, 14). Whereas TGFβ stimulates EMT, bone morphogenetic protein signaling through SMAD1/4 induces expression of proepithelial microRNAs (miRNA; miR200 and miR205) that oppose EMT (12, 15). The miR200 family consists of five members localized on two genomic clusters that can be further divided into two subgroups according to their seed sequences—subgroup I: miR141 and miR200a; subgroup II: miR200b, miR200c, and miR429 (16). During TGFβ-induced EMT, miR200 family and miR205, but not the other miRNAs, are greatly downregulated to facilitate this transition (10, 16, 17).

Members of the miR200 family not only inhibit EMT, but also influence sensitivity to EGFR inhibitors (10, 17–19). miR200c may directly inhibit the expression of the mitogen-inducible gene 6 (MIG6; also known as RALT, ERRFI1, or Gene 33; ref. 10), a negative regulator of EGFR, which plays an important role in signal attenuation of the EGFR network by blocking the formation of the activating dimer interface through interaction with the kinase domains of EGFR and ERBB2 (20–23). We recently reported that EGFR activity was markedly decreased during acquired resistance to the EGFR TKI erlotinib, with a concomitant increase of MIG6 through the activation of the PI3K–AKT pathway. A low MIG6/EGFR ratio was highly correlated with erlotinib sensitivity in a panel of cancer cell lines and early passage xenografts of human tumors with wild-type EGFR (24).

In this study, we report that in response to tumor cell–autonomous expression of TGFβ, erlotinib-sensitive tumor cells undergo EMT-associated suppression of the miR200 family and subsequent upregulation of MIG6 expression. We show that the MIG6-mediated reduction of EGFR occurs concomitantly with a TGFβ-induced EMT-associated kinase switch of tumor cells to an AKT-activated state, thereby leading to an EGFR-independent phenotype that is refractory to EGFR TKI. In a panel of 25 cancer cell lines of different tissue origins, we find that the ratio of the expression levels of MIG6 and miR200c is highly correlated with EMT and resistance to erlotinib. Moreover, analyses of primary tumor xenografts of patient-derived lung and pancreatic cancers carrying wild-type EGFR showed that the tumor MIG6(mRNA)/miR200 ratio is inversely correlated with response to erlotinib in vivo. Our data demonstrate that the TGFβ–miR200–MIG6 network orchestrates the EMT-associated kinase switch that induces resistance to EGFR inhibitors, and identify the ratio of MIG6 to miR200 as a promising predictive biomarker of the response of tumors to EGFR TKIs.

Erlotinib was purchased from the Johns Hopkins Pharmacy. LY294002 and U0126 were obtained from the Cell Signaling Technology. TGFβ and TGFβ RII/Fc were purchased from R&D Systems. All other chemicals were purchased from Sigma. All reagents were dissolved according to the manufacturer's recommendations.

Human NSCLC cell lines (H226, H292, H358, H1838, A549, Calu6, H460, H1703, H1915, H1299, Calu3, H1437, and H23), human bladder cancer cell lines (5637, SCaBER, UMUC-3, T24, HT-1376, BFTC-905, and J82), and the human H&N cell line FaDu were obtained from the American Type Culture Collection. The cell lines were freshly ordered and used within 6 months of order date.

Drug-resistant cell lines were generated via a process of slowly escalating exposure to erlotinib, as reported previously (24). SCC-S is used to designate the parental UM-SCC1 cells exposed to DMSO, and SCC-R refers to the erlotinib-resistant clone.

Pelleted cells were lysed on ice with RIPA lysis buffer (Thermo Scientific) supplemented with protease/phosphatase inhibitors (Roche). Protein concentrations were determined by the bicinchoninic acid method and lysates diluted in SDS sample buffer (Bio-Rad) before SDS–PAGE. Anti-MIG6 antibody was a gift from Dr. Ingvar Ferby of the Ludwig Institute for Cancer Research, Uppsala, Sweden (25). β-Actin was obtained from Abcam. All other antibodies were obtained from Cell Signaling Technology. Secondary horseradish peroxidase–conjugated antibodies were from KPL and signals developed using West-Pico chemiluminescence substrate (Thermo Scientific). ImageJ software was used to quantify immunoblot signals on exposed films.

RNA was extracted using TRizol (Invitrogen) followed by RNeasy kit cleanup (Qiagen). RNA was reverse transcribed to cDNA using Superscript III (Invitrogen), which was then used as a template for real-time PCR. Gene products were amplified using iTaq SYBR green Supermix with Rox dye (Bio-Rad Laboratories). All reactions were performed in triplicate and relative quantity was calculated after normalizing to GAPDH expression.

RNA from cultured cells was extracted using the mirVana Kit (Ambion). Total RNA from fresh-frozen tumors was isolated using the TRizol reagent (Invitrogen). Specific quantitative real-time PCR was carried out using TaqMan MicroRNA Assays for miR200a, miR200b, miR200c, miR205, and control RNU6b (Applied Biosystems) on a 7900HT detector (Applied Biosystems).

Relative cell viability was determined using an Alamar Blue assay as outlined by the manufacturer (AbDSerotec). New media containing 1/10 volume of Alamar Blue reagent were added to the wells and cells were incubated at 37°C for 1 hour. Fluorescence (545-nm excitation, 590-nm emission wavelengths) was measured using a SpectraMax-Plus384 fluorometer. Cell viability was calculated relative to an untreated culture of cells incubated in parallel.

A total of 1 × 10⁶ cells were plated in media containing 0.1% FBS. Tumor cell supernatants were evaluated by ELISA (R&D Systems) to determine the amount of TGFβ expressed by 1 × 10⁶ cells per 24 hours.

The xenografts were generated and erlotinib treatment was performed as published previously (26, 27). Relative tumor growth inhibition (TGI) in response to erlotinib (35 mg/kg) was calculated as the relative tumor growth of treated mice divided by relative tumor growth of control mice (T/C). The animals were maintained in accordance to guidelines of the American Association of Laboratory Animal Care and the research protocol was approved by the Johns Hopkins University Animal Use and Care Committee.

Student t tests were used for statistical analysis between two groups. The significance level was defined as 0.05. All statistical analyses were performed using SPSS. The IC₅₀ value was generated using GraphPad Prism software.

To identify the molecular mechanisms underlying the resistance of tumor cells to EGFR TKI, we examined tumor cell expression and activity of EGFR and alternative receptor tyrosine kinases (RTK) that lead to EGFR-independent AKT activation. We evaluated pairs of cancer cell lines with wild-type EGFR that were either sensitive or resistant to the EGFR TKI, erlotinib; lung carcinoma (H358/H1703 and Calu3/Calu6) and H&N cancer (SCC-S/SCC-R and JHU011/JHU028). Erlotinib-resistant (SCC-R) and erlotinib-sensitive (SCC-S) isogenic cell lines were generated by chronic exposure of human H&N squamous cell carcinoma UM-SCC1 cells to either erlotinib or DMSO (vehicle control; ref. 24). The other three pairs of cell lines (JHU011/JHU028, H358/H1703, and Calu3/Calu6) are intrinsically erlotinib-sensitive or erlotinib-resistant. For every sensitive/resistant cell line pair tested, the IC₅₀ value of the resistant cells was at least 10 times more than that of their sensitive counterparts (Fig. 1A). Comparison of the expression and activity of EGFR family members in resistant and sensitive cell lines revealed that the levels of phosphorylated EGFR, HER2, and HER3 were markedly decreased in resistant cells (Fig. 1B). In resistant cells, low activity of EGFR family kinases was associated with a significantly higher expression of the endogenous EGFR family negative regulator, MIG6. Consistent with the observed upregulation of MIG6 expression by PI3K-dependent pathways (24, 28), the resistant cell lines exhibited higher AKT phosphorylation levels compared with their sensitive counterparts (Fig. 1B). In accordance with their increased AKT phosphorylation despite low activity of the EGFR family members, erlotinib-resistant cells exhibited a switch from EGFR to activation of an alternative tumor cell–specific RTKs (PDGFR, FGFR, VEGFR, and/or IGFR; Fig. 1C).

The nonreceptor focal adhesion kinase (FAK) plays an important role in TGFβ-induced EMT progression (29) and upregulation of mesenchymal markers (30). We tested FAK phosphorylation and total expression level in our four pairs of erlotinib-resistant and -sensitive cell lines and found that FAK activity is significantly higher in erlotinib-resistant cells from lung and H&N origin (Supplementary Fig. S1). To evaluate whether resistance to erlotinib is associated with features of EMT, we tested levels of E-cadherin and vimentin in the panel of 25 erlotinib-sensitive or erlotinib-resistant cell lines with wild-type EGFR from lung, H&N, and bladder cancer origin (24). Although erlotinib-sensitive cells displayed characteristics of typical epithelial cells, including expression of E-cadherin and absence of vimentin, the majority of resistant cells displayed a mesenchymal phenotype manifested by loss of E-cadherin and acquisition of vimentin (Fig. 2A). To determine whether erlotinib sensitivity correlates with levels of tumor cell expression of TGFβ (31), we measured the amount of TGFβ produced in cell supernatants of each of the 25 tumor cell lines. Erlotinib-resistant, mesenchymal-like tumor cell lines produced higher levels of TGFβ compared with the erlotinib-sensitive, epithelial-like tumor cells (Fig. 2B). To examine whether TGFβ induces the EMT-associated kinase switch responsible for resistance to erlotinib, erlotinib-sensitive epithelial cell lines were exposed to TGFβ1 or TGFβ3. These cell lines included one H&N (SCC-S) and two lung (H358 and H292) cancer cell lines. Serial examination of EMT markers (loss of E-cadherin and upregulation of vimentin) in a time course (1–21 days) showed that TGFβ treatment resulted in complete EMT by day 14 (Fig. 2C and Supplementary Fig. S2A). Strikingly, both total EGFR and phospho-EGFR were reduced with this transition and were accompanied by elevated expression of MIG6 in cells with a mesenchymal phenotype (Fig. 2C and Supplementary Fig. S2A). Concomitant with these molecular alterations, the mesenchymal-like cells acquired a relative resistance to erlotinib (Fig. 2D and Supplementary Fig. S2B). The acquisition of an erlotinib-resistant EMT phenotype in response to TGFβ was associated with a significant increase in AKT activity (Fig. 2C and Supplementary Fig. S2A). To confirm the causal role of AKT in upregulating MIG6 in tumor cells that have acquired resistance to erlotinib, we treated H358, H358/TGFβ1-day 21, and H358/TGFβ3-day 21 cells with LY294002 (PI3K inhibitor), U0126 (MEK inhibitor), or erlotinib (Fig. 2E). Whereas all three inhibitors reduced basal expression of MIG6 in H358 cells, only LY294002 resulted in significant inhibition of MIG6 in the erlotinib-resistant H358/TGFβ1-day 21 and H358/TGFβ3-day 21 cells. These data indicate that basal EGFR activity induces an autoregulatory expression of MIG6 in epithelial cells, and that TGFβ-induced activation of AKT coopts this activity in mesenchymal cells (Fig. 2E). Together with the data shown in Fig. 1C, these data suggested that MIG6 elevation in EMT cells is due to activation of AKT by EGFR-independent tyrosine kinases. To test whether TGFβ can promote this kinase switch, levels of phospho IGFR, PDGFR, FGFR, and FAK kinases were assessed in response to treatment of erlotinib-sensitive cells (H358, H292, and SCC-S) with TGFβ1 for 21 days. These kinases showed significantly greater activity in TGFβ1-treated cells when compared with the untreated counterparts (Fig. 2F and Supplementary Fig. S2C). These data indicate that TGFβ-mediated activation of AKT via alternative kinases may substitute for the loss of EGFR activity in a cell-specific manner and contribute to the acquisition of an erlotinib-resistant phenotype.

Because the miR200 family of miRNAs is downregulated to facilitate EMT (10, 17), we used RT-PCR to assess the level of expression of miR200 in three sensitive cell lines (SCC-S, H358, and H292) in response to exposure to TGFβ for 21 days. In all tested cell lines, expression of the miR200 family members (200a, 200b, 200c, and 205) was significantly reduced upon TGFβ treatment (Fig. 3A). Consistent with the observed ability of miR200c to directly inhibit expression of MIG6 (10), the loss of miR200 family in response to TGFβ was attended with elevation in MIG6 expression during EMT-associated resistance to erlotinib (Fig. 2C and Supplementary Fig. S2A). We next examined changes in miR200 levels in erlotinib-sensitive (SCC-S) and erlotinib-resistant (SCC-R) isogenic H&N cell lines. We found that parental erlotinib-sensitive SCC-S cells displayed significantly higher levels of miR200 family members than the resistant, mesenchymal like, SCC-R cells (Fig. 3B). The same pattern was observed in the other three intrinsically sensitive/resistant cell lines pairs (JHU011/JHU028, H358/H1703, and Calu3/Calu6). Finally, examination of the 25 H&N, bladder, and lung cancer cell lines used in this study demonstrated a clear inverse correlation of miR200 levels and erlotinib sensitivity (Fig. 3C). Notably, the levels of miR200 family members were also inversely correlated with the expression of MIG6. Although erlotinib-sensitive cells demonstrated a high level of miR200 and a low level of MIG6, most of the erlotinib-resistant cells showed decreased levels of miR200 miRNAs and elevated MIG6 expression (Fig. 3C and Supplementary Fig. S3). Taken together, these data indicate that TGFβ-induced repression of miR200 family unleashes the expression of MIG6 in tumor cells during their EMT-associated conversion to an erlotinib-resistant phenotype.

Autocrine or paracrine TGFβ signaling is required for the maintenance of the mesenchymal state. Blockage of this signaling can inhibit or reverse EMT by upregulating miR200 and subsequently downregulating ZEB1/2 (31–34). As a corollary to this observation, overexpression of miR200c restores the sensitivity of resistant NSCLC cells to the anti-EGFR antibody cetuximab (35). To determine whether inhibition of TGFβ can restore miR200 expression and reverse the erlotinib-resistant phenotype, we blocked TGFβ signaling in two erlotinib-resistant cell lines of lung (H1703) and H&N (JHU028) origin with SB-431542, a potent inhibitor of the activin receptor–like kinase (ALK) receptors family. Tumor cells were cultured with TGFβ alone or in combination with TGFβ inhibitor for 7 days, and then treated with 1 μmol/L erlotinib for an additional 72 hours. In both cell lines, exposure to TGFβ inhibitor resulted in a significant increase in miR200c and miR205 levels, and concurrent downregulation of AKT phosphorylation and MIG6 expression (Fig. 4A and B). Treatment with SB-431542 increased the sensitivity of tumor cells to erlotinib (Fig. 4C). Likewise, cells incubated with TGFβ RII/Fc (recombinant TGFβ receptor II, which binds to and inhibits TGFβ1, TGFβ3, and TGFβ5), displayed a similar increase in erlotinib sensitivity (Fig. 4C).

We observed a strong correlation between MIG6 mRNA and protein levels in 25 tumor cell lines (Fig. 5A). Akin to the MIG6 protein (24), MIG6 mRNA expression was considerably lower in erlotinib-sensitive cell lines. Next, we tested whether the ratio between MIG6 mRNA and miR200 levels is a reliable predictor of tumor cells response to erlotinib. We found that across the panel of 25 cancer cell lines, the ratio of MIG6 mRNA to each one of the miR200 family members tested seemed to be a reliable predictor of tumor cell responsiveness to erlotinib (Fig. 5B and Supplementary Fig. S4A). Given the strong correlation between MIG6 mRNA and protein levels (Fig. 5A), the ability of MIG6(mRNA)/miR200 ratios to predict erlotinib sensitivity was equal to the predictive value of the MIG6/miR200 protein expression ratio (Fig. 5C). Interestingly, the ability of the MIG6(mRNA)/miR200 ratio to predict erlotinib sensitivity in cancer cell lines was even better than the predictive value of the MIG6/EGFR protein expression ratio (Fig. 5D and Supplementary Fig. S4B).

We obtained 18 human NSCLCs, and 27 pancreatic tumors that were directly xenografted into nude mice (27). Tumor characteristics, including KRAS, NRAS, and p53 mutation status, are summarized in Supplementary Table S1. No erlotinib-sensitizing mutations in EGFR were detected in any of these tumors and there was no correlation of KRAS mutation with erlotinib response. For all models tested, miR200 levels were measured by quantitative RT-PCR and mRNA levels of MIG6 and EGFR were determined by Affymetrix expression array. Relative TGI in response to erlotinib (35 mg/kg) was calculated as the relative tumor growth of treated mice divided by relative tumor growth of control mice. We next plotted the MIG6(mRNA)/miR200 ratio against erlotinib responsiveness, with the more resistant tumors clustered to the left and the more sensitive models clustered on the right. Lung and pancreatic tumors that display a high MIG6(mRNA)/miR200 ratio tended to cluster on the left side of the chart, indicating that they were more resistant to erlotinib (Fig. 6A and B). Lung models with a TGI more than 40% and pancreatic models with TGI greater than 50%, were associated with significantly lower MIG6(mRNA)/miR200a, MIG6(mRNA)/miR200b, or MIG6(mRNA)/miR200c ratios and greater miR200 expression (Fig. 6). Our data showed that expression of miR200c (Supplementary Fig. S5A) and subsequently the MIG6(mRNA)/miR200c ratio (Fig. 6) showed the strongest correlation with erlotinib response compared with miR200b and miR200a, suggesting that miR200c might play a more dominant role in regulating MIG6. Supporting this observation, an inverse correlation between miR200c and MIG6 expression levels was noted across the pancreatic models (Supplementary Fig. S5B). In lung models, tumors with higher erlotinib sensitivity displayed a similar pattern of the low MIG6(mRNA)/miR200c ratio. Of note, four erlotinib-resistant lung tumors with low EGFR and MIG6 expression (CTG-0167, CTG-0502, CTG-0199, and CTG-0157; Supplementary Fig. S5C) exhibited even lower levels of miR200c (Supplementary Fig. S5B). Unlike the limited predictive ability of the MIG6/EGFR ratio in such tumors with low EGFR expression (24, 27), the MIG6(mRNA)/miR200c ratio was still able to correctly identify three of four of these lung tumors with low EGFR mRNA as erlotinib resistant. Therefore, the ratio of MIG6 to miR200 was a reliable predictive biomarker of the primary tumors response to EGFR TKIs regardless of their EGFR status.

TGFβ, a multifunctional cytokine that regulates cell growth and differentiation, is frequently upregulated in many human cancers (31, 36–38). Although TGFβ exerts a suppressive effect on normal epithelial cells, tumor cells frequently become refractory to the growth-inhibitory effect of TGFβ and acquire an ability to increase expression and secretion of TGFβ (31–34). This switch enables tumor cells to leverage the tumor-promoting effects of TGFβ in the tumor microenvironment to facilitate tumor progression, invasion, and metastasis (31–39). Previous studies have demonstrated that TGFβ plays a key role in promoting EMT, a switch of epithelial cells into a mesenchymal migratory phenotype that is driven by a network of transcriptional repressors that include SNAIL1, ZEB1, ZEB2, and TWIST (12, 39, 40). TGFβ uses both, canonical and noncanonical signaling pathways to engineer this kinase switch. TGFβ-activated SMAD3/4 stimulates the expression of SNAIL1 and TWIST1, which repress the expression of epithelial genes, such as CDH1 (which encodes E-cadherin; ref. 13). TGFβ also inhibits the expression of proepithelial miRNAs (miR200 and miR205) that inhibit ZEB1/2 and oppose EMT (17, 18, 32, 41). Besides promoting EMT, TGFβ engages SMAD-independent pathways to activate PI3K–AKT, such as tumor necrosis factor alpha converting enzyme (TACE)-mediated secretion of EGFR ligands (42).

In this study, we report that TGFβ induces tumor cells to undergo an EMT-associated kinase switch that renders them resistant to EGFR inhibitors. TGFβ-mediated suppression of the miR200 family not only facilitates EMT, but also enables upregulation of MIG6, a negative regulator of EGFR whose expression is held in check by miR200c. In addition to curtailing EGFR activity via upregulation of MIG6, TGFβ promotes EGFR-independent activation of alternative RTKs and PI3K–AKT signaling. We find that the net effect of TGFβ signaling is the loss of EGFR activity with a concomitant EMT-associated kinase switch of tumor cells to an AKT-activated state, thereby leading to an EGFR-independent mesenchymal phenotype that is refractory to EGFR TKI. Our study demonstrates that the molecular signature of this resistant tumor phenotype is an elevated MIG6/miR200 ratio.

Our data demonstrate that the TGFβ–miR200–MIG6 network orchestrates the EMT-associated kinase switch that induces resistance to EGFR inhibitors (Fig. 7). As such, the autonomous production of TGFβ by tumor cells may be a frequent mechanism by which cancers induce an erlotinib-resistant phenotype. Our study provides the following lines of evidence to support this conclusion. In a panel of 25 cancer cell lines of different tissue origins (H&N, bladder, and lung), erlotinib-resistant, mesenchymal-like cells produced higher levels of TGFβ than the epithelial-like, erlotinib-sensitive cells, suggesting that increased autocrine exposure to TGFβ may be a driving force behind the erlotinib-resistant phenotype. In the same panel, resistance to erlotinib was highly correlated with EMT and an elevated MIG6/miR200c ratio. Besides the high TGFβ expression and elevated MIG6/miR200 ratio exhibited by de novo erlotinib-resistant cell lines, this phenotype was also exhibited by SCC-R tumor cells that had acquired erlotinib resistance by culturing erlotinib-sensitive SCC-S cells in the presence of escalating concentrations of erlotinib. SCC-R cells expressed more than 10-fold higher levels of TGFβ compared with SCC-S cells (31), and this was associated with reduction of miR200 family members (200a, 200b, 200c, and 205) and concomitant increase in MIG6 expression. Furthermore, these cells showed evidence of EMT and manifested a kinase switch involving reduced activity of the EGFR kinase family and activation of alternative RTKs (pPDGFR, pFGFR, pVEGFR, and pIGFR) and AKT. In support of the causal association of tumor cell expression of TGFβ with an elevated MIG6/miR200 ratio and erlotinib resistance, exposure of various erlotinib-sensitive epithelial tumor cells to exogenous TGFβ resulted in their EMT-associated conversion to an erlotinib-resistant phenotype with an attendant reduction of miR200, increase in MIG6 expression, decrease in EGFR activity, and activation of AKT. Conversely, blockade of TGFβ signaling in erlotinib-resistant, mesenchymal-like cell lines resulted in a concurrent increase of miR200c and miR205 transcripts, downregulation of AKT activity and MIG6 levels, and a significant increase in erlotinib sensitivity.

The 25 H&N, bladder, and lung cancer cell lines used in this study showed an inverse correlation between the expression levels of MIG6 and miR200. Whereas erlotinib-sensitive cells displayed a low MIG6/miR200 ratio, erlotinib-resistant cells exhibited a high MIG6/miR200 ratio. A similar pattern was noted during TGFβ-induced EMT, wherein downregulation of miR200 family members was paralleled by upregulation of MIG6. By performing PicTar, TargetScan, miRanda, and miRBase searches to predict miRNA–mRNA interactions on the MIG6 3′ untranslated region (UTR), we found that the 3′UTR of MIG6 contains conserved potential binding sites for miR200 family members. In addition, recent work indicates that miR200c can directly bind to the 3′UTR region of MIG6 mRNA and downregulate its expression (10). In line with these data, the MIG6(mRNA)/miR200c ratio showed the strongest association with erlotinib sensitivity in cancer cell lines as well as primary human tumor xenografts in vivo. These data suggest that TGFβ-mediated suppression of the miR200 family unleashes expression of MIG6, which in turn quenches EGFR activity. The elevation of MIG6 following TGFβ-induced EMT is sustained by EGFR-independent activation of AKT because this is reduced by PI3K inhibitors, but not by erlotinib. Therefore, a high MIG6/miR200c ratio is a sequel of TGFβ-induced EMT and a signature of the EMT-associated kinase switch responsible for resistance to EGFR TKI. Consistent with these observations, our analyses of patient-derived primary tumor xenografts of 18 NSCLC and 27 pancreatic cancers carrying wild-type EGFR showed that the tumor MIG6/miR200 ratio is inversely correlated with response to erlotinib in vivo.

Our study demonstrates that an elevated MIG6/miR200 ratio is a molecular signature that characterizes the erlotinib-resistant tumor phenotype. These observations have important clinical implications for the treatment of patients with EGFR inhibitors. The tumor MIG6/miR200 ratio may have clinical value as a predictive biomarker to discern patients who are likely to benefit from EGFR inhibitors from those who are unlikely to respond to such therapy. Our findings further suggest that inhibition of the molecular determinants of the EMT-associated kinase switch, such as TGFβ, may prevent or reverse the de novo or acquired resistance of cancers to EGFR inhibitors.

No potential conflicts of interest were disclosed

Conception and design: E. Izumchenko, X. Chang, A. Bedi, D. Sidransky

Development of methodology: E. Izumchenko, C. Michailidi, L. Kagohara, R. Ravi, A. Bedi

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): E. Izumchenko, X. Chang, C. Michailidi, S. Ling, A. Bedi

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): E. Izumchenko, X. Chang, C. Michailidi, K. Paz, A. Bedi, D. Sidransky

Writing, review, and/or revision of the manuscript: E. Izumchenko, X. Chang, C. Michailidi, L. Kagohara, M. Brait, M. Hoque, A. Bedi

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): E. Izumchenko, M. Brait, S. Ling, A. Bedi

Study supervision: E. Izumchenko, R. Ravi, A. Bedi, D. Sidransky

This work was supported by NIH grants SPORE P50 DE019032, EDRN U01 CA084986, R37DE012588, and FAMRI-funded 072017_YCSA.

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.