Inhibition of TWIST1 Leads to Activation of Oncogene-Induced Senescence in Oncogene-Driven Non–Small Cell Lung Cancer

Lung cancer is the leading cause of cancer death in the United States and worldwide. The 5-year survival rate for all patients with lung cancer is a dismal 15% (1). The recent advances in the treatment of non–small cell lung cancers (NSCLC) have come from recognition that NSCLCs that is not a single disease entity but rather a collection of distinct molecularly driven neoplasms. This paradigm is typified by the recent progress made in the treatment of patients with EGFR-mutant and EML4-ALK translocation–driven adenocarcinomas of the lung with tyrosine kinase inhibitors targeting these oncogenes (2). Unfortunately, acquired resistance to the currently targetable driver mutations arises in nearly all cases, typically within the first 2 years of therapy (3, 4). Furthermore, little if any progress has been made in the treatment of patients with the most frequently observed driver oncogene, mutant KRAS. It has been over 2 decades since the first mutation in the KRAS oncogene was identified in NSCLCs. KRAS is mutated in one third of all malignancies and approximately 20% of all NSCLCs (5). KRAS mutation, which results in a constitutively active oncoprotein, has been shown to predict lack of benefit from chemotherapy in the adjuvant setting, to portend worse prognosis in metastatic NSCLCs, and possibly to increase risk of recurrence in early-stage disease (5). Although testing for KRAS mutation status is routinely done, there are no currently approved therapies targeting this critical pathway in NSCLCs.

The principle of “oncogene addiction,” the absolute dependency on an oncogene for a tumor's survival, has been well-supported by several preclinical models and in multiple tumor types (6). An emerging concept of “non-oncogene addiction,” the dependency of tumor cells on factors which are not inherently oncogenic, but which are required for survival due to the oncogenic state of the tumor, is now gaining acceptance. To that end, several investigators have conducted large-scale synthetic lethal screens looking for genes,. which are essential for survival of KRAS-mutant tumor cells and have suggested therapeutic strategies that may specifically target KRAS-mutant tumors (7–11). To date, it is unclear whether the inhibition of candidate genes identified in these screens can reproducibly produce synthetic lethality in KRAS-mutant cancer specifically or are more generally required for tumor proliferation (12).

Another potential therapeutic strategy in KRAS-mutant tumors is activation of latent oncogene–induced senescence (OIS). OIS is an irreversible cell-cycle arrest that is characterized by cells displaying an enlarged, flattened cytoplasm, increased senescence-associated β-galactosidase (SA-β-Gal) activity, increased chromatin condensation and changes in gene expression associated with DNA damage and cell-cycle checkpoint pathways. OIS is thought to be triggered early during tumorigenesis after oncogene activation and serves as a checkpoint to prevent premalignant lesions from progressing to malignancy (13).

The emerging concept of using prosenescence therapies for cancer has generated considerable interest (14, 15). Interestingly, several studies have shown that bypass of senescence in mouse models of mutant Kras--mediated adenocarcinoma of the lung and pancreas is essential for tumorigenesis (16–18). Therefore, activation of OIS in KRAS-mutant NSCLCs may be an effective therapeutic strategy. Disruption of the p53–ARF and p16–Rb pathways was previously thought to be required for bypass of RAS-mediated senescence (18). However, recent data suggest that the requirement of these pathways is context-dependent, as inactivation of Skp2 can lead to activation of OIS independent of the p53–ARF pathway through Atf4, p27, and p21 (19). Activation of suppressed senescence pathways in KRAS-mutant NSCLCs through inhibition of key downstream mediators of OIS would be a valuable treatment in NSCLCs.

Previous murine studies have suggested a role for Twist1 as a factor in preventing OIS through inhibition of the Rb and p53 pathways (20, 21). TWIST1 is a basic helix-loop-helix (bHLH) transcription factor which plays critical roles throughout development influencing mesoderm formation, neurogenesis, myogenesis, and neural crest cell migration and differentiation (22). These roles are highlighted by the findings that Twist1 deletion is embryonic lethal in mice and that germ line haploinsufficiency of TWIST1 in humans leads to Saethre–Chotzen syndrome, which is characterized by craniosynostosis and limb abnormalities (23, 24). A number of reports have implicated TWIST1 in oncogenesis through its ability to inhibit DNA damage–induced apoptosis, prevent OIS and promote metastasis through induction of the epithelial–mesenchymal transition (EMT; ref. 25). Increased TWIST1 expression is correlated with increased risk of metastasis and poor prognosis in a number of solid tumor types including breast, prostate, ovarian, and lung (25). We have recently shown that Twist1 cooperates with mutant Kras to induce adenocarcinoma of the lung in transgenic mouse models and inhibition of Twist1 in these models led to activation of Kras-induced senescence (26). However, significant differences exist between transgenic mouse models of cancer and human disease. In the current study, we explore whether TWIST1 plays a critical role in human NSCLCs characterized by multiple defined oncogenic drivers.

All human NSCLC cell lines, (A549, H460, H358, H727, H23, Calu-1, Calu-6, H1975, H3255, H1993) and embryonic kidney cell line HEK 293T were obtained from the American Type Culture Collection and grown in media as recommended.

At 12 ± 2 days after infection with short hairpin RNA (shRNA) lentiviruses, the cells were washed twice with PBS and then fixed with PBS containing 2% formaldehyde and 0.25% glutaraldehyde for 5 minutes. The cells were then incubated at 37ºC for 20 hours with staining solution [40 mmol/L citric acid sodium phosphate, pH 6.0, 1 mg/mL 5-bromo-4-chloro-3-isolyl-b-d-galactoside (X-gal), 5 mmol/L potassium ferricyanide, 5 mmol/L potassium ferrocyanide, 150 mmol/L NaCl, 2 mmol/L MgCl₂]. After incubation, cells were washed twice with PBS and viewed with bright-field microscopy.

Cells were lysed on ice for 60 minutes in radioimmunoprecipitation assay buffer supplemented with protease and phosphatase inhibitors (Sigma-Aldrich) and clarified by centrifugation. Protein concentrations were determined by Bradford protein assay (Bio-Rad Laboratories). Equal protein concentrations of each sample were run on NuPAGE bis-Tris gels (Invitrogen) and electrophoretically transferred to polyvinylidene difluoride membranes. After being blocked with 5% dried milk in TBS containing 0.2% Tween 20, the filters were incubated with primary antibodies. The following primary antibodies were used: goat anti-Actin (C-11, Santa Cruz), rabbit anti-GAPDH (FL-335, Santa Cruz), anti-p21 (Ab-1, Calbiochem), mouse monoclonal anti-p27 (F-8, Santa Cruz), rabbit anti-Parp-1/2 (H-250, Santa Cruz) rabbit anti-cleaved PARP (Cell Signaling), and anti-p53 (DO-1, Santa Cruz). After washing and incubation with horseradish peroxidase (HRP)-conjugated anti-goat or anti-mouse IgG (Amersham), the antigen–antibody complexes were visualized by chemiluminescence (ECL detection system; Perkin Elmer or SuperSignal West Femto Maximum Sensitivity Substrate; Thermo Fisher Scientific).

On days 4 or 6 after infection with the indicated shRNA lentiviruses, cells were plated in 12-well plates at a density of 5,000 to 10,000 cells/well. On day 12, the cells were stained with crystal violet (0.5% in 95% ethanol).

Cells were harvested at indicated times, fixed with 70% cold ethanol and stained with stained with propidium iodide, and analyzed by flow cytometry as previously described (27).

293T cells were seeded (2.5 × 10⁶ cells) in T25 flasks. Lentiviral particles were generated using a three-plasmid system and infected as per the TRC Library Production and Performance Protocols, RNAi Consortium, Broad Institute (28) shRNA constructs were obtained from the Broad RNAi Consortium and clone IDs as are follows: TWIST1: TRCN0000020539 (shTWIST1-1) and TRCN0000020543 (shTWIST1-2). P53: TRCN0000003756 (shp53-1) and TRCN0000010814 (sh53-2). The pLKO.1-shRNA scramble vector was obtained from Dr. David M. Sabatini through Addgene (Addgene plasmid 1864) as previously described (29). The Tet-pLKO-puro vector was from Dmitri Wiederschain through Addgene (Addgene plasmid 21915) as previously described (30). The pLKO.1–Neo vector was obtained from Sheila Stewart through Addgene (Addgene plasmid 13425), and oligonucleotides for TWIST1-1, TWIST1-2, and scrambled control were cloned into this vector and used to for experiments with shRNAs targeting p53. All constructs were sequence-verified. Twenty-four hours after infection, cells were treated with 1 μg/mL puromycin or 500 μg/mL G418 and passaged once 80% confluent.

A total of 5 × 10⁵ viable cells were resuspended in equal volumes of PBS and Matrigel and injected subcutaneously in adult homozygous NOD/SCID mice (Harlan Laboratories). For doxycycline-inducible cell line experiments, once the tumors reached ≥150 mm³ (average, 14 days), the mice were given 5% dextrose either with or without 2 mg/mL doxycycline in the drinking water. Drinking water was replenished at least once a week. Statistical analyses were conducted for each xenograft separately. Initial tumor volume measurements are summarized using means, SDs, and ranges. Growth patterns were summarized graphically by plotting the mean and SD for each treatment group over time. A conditional t test is used to compare shTWIST1 tumors with the control (shScram). P < 0.05 is considered statistically significant.

To explore the role of TWIST1 in human NSCLCs, we used an shRNA strategy to silence TWIST1 in NSCLC cell lines with distinct mutational backgrounds (KRAS-mutant, EGFR-mutant and c-Met–amplified; Supplementary Table S1). In our previous study and current study, we have shown that the shRNAs used in this report (shTWIST1-1 and shTWIST-2) are able to significantly suppress TWIST1 levels on both the transcript and protein level (ref. 26 and Supplementary Fig. S1A and SB). Upon silencing of TWIST1 with 2 separate shRNAs compared with a control scrambled shRNA, we observed dramatic growth inhibition in colony formation assays for all 9 NSCLC cell lines examined (Fig. 1). To confirm that these effects were not due to off-target effects of the shRNAs used in this study, we stably overexpressed a mouse Twist1 cDNA in A549 cells and showed that this overexpression rescued the A549 cells from the growth-inhibitory effects of silencing human TWIST1 (ref. 26 and Supplementary Fig. S2). In addition to the 9 lines shown in Fig. 1, an additional KRAS-mutant NSCLC cell line, Calu-6 showed extensive apoptosis by day 4 after silencing and could not be plated for this assay. In all, we observed significant growth inhibition in 7 KRAS-mutant NSCLC cell lines, 2 cell lines which harbor an activating mutation in EGFR as well as in a cell line with c-Met amplification. Taken together, these data suggest that TWIST1 may play a critical role in lung cancer tumorigenesis in the context of multiple driver mutations.

We observed a dramatic activation of latent OIS as shown by positive SA-β-Gal staining and characteristic enlarged, flattened cytoplasm morphology (Fig. 2) in 3 KRAS-mutant NSCLC cells lines (H460, A549, and H358). These findings were further supported by profound cell-cycle arrest in these cell lines (Fig. 3A) as well as by induction of the senescence markers, p21 and/or p27, in all cell lines examined (Fig. 3B). Furthermore, a gene expression analysis after silencing of TWIST1 in 3 KRAS-mutant cell lines (H460, H727, and H358) revealed several highly significant gene sets involved with growth arrest (Supplementary Figs. S7 and S8 and data not shown). Collectively, these series of experiments support a role for TWIST1 in the suppression of OIS in KRAS-mutant lung cancer cell lines. Interestingly, in a subset of cell lines (Calu-1, Calu-6), we also observed evidence of apoptosis including PARP cleavage. These findings are consistent with previously known antiapoptotic role of TWIST1 and suggest complementary mechanisms by which inhibition of TWIST1 may block tumor growth (31).

A previous study had shown a requirement for p16 in mediating OIS in pancreatic ductal epithelial cells after silencing of Twist1 (21). Furthermore, prior work had suggested a possible role for the p53/p21 pathway in mediating OIS after suppression of TWIST1 in human cancer cell lines (20). In the present study, we examined whether these pathways played a key role in mediating the growth inhibition observed after TWIST1 silencing in human lung cancer cell lines. As the p16 pathway was disrupted either through deletion, methylation, or mutation in every cell line examined (Supplementary Table S1), it appeared that p16 was not required for the growth inhibition observed. In addition, only 2 (H460, A549) of the 10 lines had wild-type p53 which suggested that inhibition of p53 could not be the only mediator of growth inhibition after silencing of TWIST1. To test whether p53 was required for growth inhibition in p53 wild-type cell lines, we silenced p53 using an shRNA strategy and confirmed knockdown of p53 at baseline and under conditions that resulted in p53 stabilization (TWIST1 knockdown and DNA damage; Fig. 4A, Supplementary Fig. S3 and data not shown). After knockdown of TWIST1, stabilization of p53 was observed in both the shScrambled control H460 and A549 cells and was accompanied by induction of p21. However, after silencing of p53, p53 expression and p21 induction were lost (Fig. 4A and Supplementary Fig. S3A). Despite the loss of p53 in these cell lines, growth inhibition induced by suppression of TWIST1 was unaltered (Fig. 4B and Supplementary Fig. S3B). It is possible that through disabling a critical member of the OIS pathway such as p53, the cellular response after suppression of TWIST1 may shift from a senescence program to an apoptotic program. To assess whether a shift to an apoptotic program is occurring with depletion of p53, we have checked for evidence of PARP cleavage which is a well-validated marker of apoptosis. In H460 and A549 cells, we observed only a modest amount of apoptosis in our control shScram line after silencing of TWIST1. Interestingly, after depletion of p53, we observed a decrease in apoptosis compared with the control cells after silencing TWIST1 (Fig. 4). This would argue against apoptosis as significant mechanism for the observed growth inhibition as the modest decrease in apoptosis after p53 depletion did not result in even a partial rescue of the observed decrease in clonogenic survival following silencing of TWIST1.

As we had observed p53-independent p21 induction in several of our p53-disrupted cell lines (Fig. 3B), we examined whether p21 was required for the growth inhibition observed. We silenced p21 using shRNAs in 3 cell lines (H460, A549, and H727) and observed a significant reduction in p21 expression but not in the cyclin-dependent kinase inhibitor, p27 (Supplementary Fig. S4A). Silencing of p21 did not evidently affect the inhibition of colony formation induced by knockdown of TWIST1 (Supplementary Fig. S4B).

Recent studies had suggested that the SKP2/p27 pathway was important for the induction of senescence (19). In addition, after silencing of TWIST1, we observed induction of p27 in several of these cell lines (Fig. 3B). We therefore attempted to knockdown p27 in H460 cells. Despite screening 8 shRNAs targeting p27, we were only able to derive a partial knockdown of this gene; complete knockdown prevented the establishment of a puromycin-resistant cell line even at early time points (Supplementary Fig. S5 and data not shown). It is possible that these cells lines require cytoplasmic p27, which has been shown in several studies to be oncogenic and can cooperate with KRAS in lung tumorigenesis (32) and reviewed in ref. 33). Partial knockdown of p27 markedly suppressed p27 induction after loss of TWIST1, however, growth inhibition was unperturbed (Supplementary Fig. S5B). To further evaluate the requirement of the p27 pathway in mediating growth suppression after loss of TWIST1, we overexpressed the key regulator of p27 stability, SKP2 (34) in the A549 and H460 cell lines. Overexpression of SKP2 largely blocked p27 induction after silencing of TWIST1 but did not rescue loss of TWIST1 (Supplementary Fig. S6). Interestingly, loss of TWIST1 expression in the setting of SKP2 overexpression did result in modest increase in apoptosis as shown by the appearance of cleaved PARP.

To further elucidate possible mechanisms of TWIST1 dependency in NSCLCs, we conducted microarray analysis on 3 KRAS-mutant NSCLC lung cancer cells (H460, H358, and H727) after silencing of TWIST1. A profound cell-cycle arrest signature was observed associated with silencing of TWIST1 (Supplementary Fig. S7). Genes encoding general regulators of cell-cycle progression (cyclin D, E1, A2, B1, CDK2) and specific regulators of progression through S-phase (PCNA, SKP2) and mitosis (BUB1, CDC20, MAD2, AURKA, FOXM1) were significantly decreased after silencing of TWIST1 (Supplementary Fig. S7). Gene set enrichment analysis (GSEA; ref. 35), comparing global gene expression patterns after TWIST1 silencing with patterns from control (shScram) cell lines, revealed a highly significant similarity to multiple gene sets for cell-cycle arrest (Supplementary Fig. S8A). This dataset also showed highly significant similarity to a gene set predicting resistance to doxorubicin (Supplementary Fig. S8B). This is consistent with previous studies showing that TWIST1 overexpression could confer chemoresistance in other tumor types (reviewed in ref. 25). Two gene sets involving the RAS signaling pathway were also highly significantly similar, including a gene signature of NRAS-induced genes (decreased in shTWIST1 compared with control), and a gene signature following exposure of cells to the RAS inhibitor, salirasib (Supplementary Fig. S8C). This is consistent with our findings that inhibition of TWIST1 is critical for KRAS-mutant NSCLC survival. Interestingly, a gene set for downregulation of the target genes of YB-1, a known TWIST1 target gene (36), was also highly similar and suggests a possible role for YB-1 in mediating TWIST1 function (Supplementary Fig. S8D).

To further evaluate TWIST1 as a potential therapeutic target for NSCLCs with a defined oncogenic driver, we examined the effect of TWIST1 targeting on both tumor initiation and maintenance in a xenograft hind flank model. KRAS-mutant (A549, H358, and H727), EGFR-mutant (H1975), and c-Met–amplified (H1993) NSCLC cell lines were infected with either scrambled shRNA or TWIST1-1 shRNA and on day 4 after infection were implanted in the hind flank of NOD/SCID animals. Knockdown of TWIST1 significantly inhibited tumor growth in each of these oncogene-driven NSCLC xenograft models (Fig. 5). Remarkably, silencing of TWIST1 in the c-Met–amplified and -dependent cell line H1993 completely prevented tumor formation, suggesting a potentially critical role for the TWIST1 pathway in this context.

To examine the role of TWIST1 in tumor maintenance of fully established tumors, we constructed syngeneic A549 cells containing either shRNA targeting TWIST1 or a scrambled control under the control of a doxycycline-responsive promoter (30). In absence of doxycycline, these cell lines had similar growth characteristics. However, upon addition of doxycycline, TWIST1 was silenced in the A549 Tet-shTWIST1 cell line, p21 was induced and growth inhibition was observed (Fig. 6A and B). To determine the requirement of TWIST1 in established tumors, the A549 Tet-shScram and A549 Tet-shTWIST1 cell lines were implanted in the hind flank of NOD/SCID animals. After the formation of palpable tumors, doxycycline was added to the drinking water of half the animals for each tumor type. Although the administration of doxycycline to animals with A549 Tet-shScram xenograft tumors did not affect the growth kinetics, silencing of TWIST1 by doxycycline administration to animals with the A549 Tet-shTWIST1 xenografts resulted in significant growth inhibition (Fig. 6C). Interestingly, this growth inhibition was accompanied by induction of p21 specifically in the tumors in which TWIST1 was silenced. This finding supports the role of growth arrest and OIS in mediating the observed growth inhibition in vivo. In summary, these data suggest that inhibition of TWIST1 in established tumors may be an effective therapeutic strategy.

We have shown that TWIST1 is a promising novel therapeutic target for oncogene-driven NSCLCs, through activation of OIS. We found that silencing of TWIST1 resulted in dramatic growth inhibition and activation of latent OIS and/or apoptosis. This marked growth inhibition appeared to be independent of the p53/p21, p16, and SKP2/p27 pathways. Similar effects were seen in KRAS-mutant, EGFR-mutant, and c-MET–amplified cell lines, suggesting that inhibition of TWIST1 may be a potential therapeutic target in a large percentage of patients with oncogene-driven lung cancer.

Although it has been more than 2 decades since the first mutation in the KRAS oncogene was identified in NSCLCs, no current therapies target this critical oncogene. In this study, we have shown that TWIST1 is required for proliferative growth of oncogene-driven NSCLCs including multiple KRAS-mutant lines and that silencing of TWIST1 leads to activation of OIS. Previous studies have defined the bypass of senescence as a requirement for mutant KRAS-mediated carcinogenesis in the lung (16–18). In this current study, we elucidated a possible synthetic interaction between mutant KRAS and TWIST1 that in most lines results not in lethality but in the activation of latent OIS. Our findings are consistent with a recent study of KRAS transformation of primary pancreatic duct epithelial cells that showed activated mutant KRAS led to increased expression of TWIST1 by a posttranscriptional mechanism and that TWIST1 expression was essential to bypassing OIS in this model (21) We have recently shown that Twist1 overexpression cooperates with mutant Kras to induce adenocarcinoma of the lung in mouse models, and inhibition of Twist1 in these models led to activation of Kras-induced senescence (26). One possible caveat to this mouse model study was that simultaneous expression of Kras^G12D and Twist1 during tumorigenesis could have resulted in co-dependent tumors and therefore the phenotype of tumor stasis and activation of OIS observed with removal of Twist1 overexpression was due to this induced co-dependency. However, our current work shows the TWIST1 dependency of oncogene-driven NSCLCs generally and KRAS-mutant NSCLCs in particular using both in vitro and xenograft approaches.

Previous work had suggested that bypass of OIS by TWIST1 required p16 and possibly a functional p53/p21 pathway (20, 21). In our multiple lung cancer lines, we observed activation of OIS in both p53 wild-type and mutant backgrounds and in the absence of a functional p16 pathway. Furthermore, our knockdown experiments confirmed that the observed OIS and growth inhibition was p53/p21-independent. A recent study has shown a critical role of TWIST1 in sarcomagenesis through its ability to lead to p53 degradation (37), and previous studies have shown the ability of TWIST1 to interact with and influence the transcriptional function of p53 (38, 39). Furthermore, p53 is required for Ras-induced OIS in mouse embryonic fibroblasts (18). Although the dispensability of p53 for OIS after silencing of TWIST1 was unexpected, it is consistent with the previous work that first established the role of TWIST1 in suppressing OIS (20). Ansieau and colleagues observed induction of p21 and p16 after silencing of TWIST1 and subsequently proposed a role for the p53 and Rb pathways in mediating OIS after suppression of TWIST1. However, in the same study, they clearly showed the induction of OIS after silencing of TWIST1 or TWIST2 in 2 p53-mutant cell lines (T47D and RPMI 7951; ref. 20). The aforementioned study about the role of TWIST1 in sarcoma showed a requirement for p53 in mediating serum starvation senescence after silencing of TWIST1 in a single sarcoma cell line (37). It is not clear whether the requirement for p53 is a sarcoma-specific effect or is required under conditions of serum starvation. Our current study suggests that both the p53 and p16 pathways are dispensable for OIS and growth inhibition observed after suppression of TWIST1.

In addition, using both knockdown of p27 and overexpression of SKP2, we have also shown that activation of OIS is likely independent of the SKP2/p27 pathway. An important caveat is that in neither experiment were we able to completely deplete p27 from the cell and it remains possible that this is a p27-dependent process requiring only a low threshold expression of p27. Alternatively, p21 and p27 may be redundant and knockdown of both CDK inhibitors is required for abrogation of OIS induced by silencing of TWIST1. Furthermore, in our experiments using SKP2 overexpression cell lines, we did observe an increase in apoptosis in the SKP2-overexpressing lines compared with vector control. Therefore, we cannot rule out the possibility that the ability of SKP2 overexpression to rescue the OIS phenotype was masked, in part, by the increase in apoptosis that was observed. However, for all experiments, in which we have silenced TWIST1, we have assessed the cells for SA-β-gal staining and have not seen a rescue of OIS after depletion or overexpression of any genes in this study.

Despite these caveats, these data suggest that alternative p53/p21- and SKP2/p27-independent mechanisms may exist for the activation of OIS. A recent study suggested that the endoplasmic reticulum (ER) stress protein and transcription factor ATF4 was important for p53-independent OIS induced by Skp2 inhibition (19). In addition, there are some data suggesting that unopposed expression of oncogenes such as β-catenin can result in DNA damage resulting in senescence and apoptosis independent of p53 (40). To better characterize the mediators of TWIST dependency, we are conducting overexpression and loss of function genetic screens to elucidate the key mediators of OIS upon inhibition of TWIST1.

The requirement of TWIST1 for bypassing OIS appears to extend across multiple driver oncogenes. Interestingly, a previous study had correlated increase TWIST1 expression with the presence of EGFR-activating mutations (41). A follow-up study found that TWIST1 may play a role in migration and EGF-mediated EMT in a limited subset of EGFR-mutant cell lines in short-term experiments (42). In our present study, we show a critical requirement for TWIST1 in EGFR-mutant NSCLCs in long-term growth experiments in vitro as well as in vivo.

Little is known about the requirement for TWIST1 in c-Met–dependent NSCLCs. Interestingly, a recent report, has suggested that either pharmacologic inhibition of c-Met or silencing of TWIST1 could inhibit metastasis in an in vivo breast cancer model by preventing EMT (43). We report for the first time that TWIST1 expression is essential for c-Met–dependent NSCLCs. The c-Met oncogene is amplified, mutated, and/or overexpressed in significant number of patients with NSCLCs. Furthermore, there is evidence of significant crosstalk and interdependence between the EGFR and c-Met signaling pathways (44). c-Met amplification is a recognized acquired resistance mechanism in patients treated with EGFR-targeted tyrosine kinase inhibitor (TKI; ref. 45). Multiple antibody-based and TKI-based therapies for targeting the c-Met pathway are currently in clinical trials for NSCLCs (44). Therefore, targeting TWIST1 in c-Met–dependent NSCLCs represents a novel and promising therapeutic strategy.

In summary, our data suggest that TWIST1 represents an attractive therapeutic target for oncogene-driven NSCLCs. Inhibition of TWIST1 resulted in OIS and growth inhibition in diverse genetic backgrounds and was effective for 3 major classes of driver mutations in adenocarcinoma of the lung (KRAS-mutant, EGFR-mutant, and c-Met–amplified). While prenatal TWIST1 haploinsufficiency leads to Saethre–Chotzen syndrome (23, 24), the expression of TWIST1 is quite restricted postnatally, and some studies have even suggested that postnatal deletion of this loci may be well-tolerated (46, 47). To this end, we are currently developing a cell-based assay to screen for small molecules that may inhibit TWIST1 transcriptional activity.

Y.-J. Cho has Honoraria from Speakers Bureau of Novartis and is a consultant/advisory board member of Novartis. C.M. Rudin is a consultant/advisory board member of Eli Lilly. No potential conflicts of interest were disclosed by the other authors.

Conception and design: T.F. Burns, P.T. Tran, C.M. Rudin

Development of methodology: T.F. Burns

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): T.F. Burns, I. Dobromilskaya, R.P. Gajula, S. Thiyagarajan, S.N.H. Chatley

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): T.F. Burns, S. Thiyagarajan, S.N.H. Chatley, Y.-J. Cho, P.T. Tran, C.M. Rudin

Writing, review, and/or revision of the manuscript: T.F. Burns, Y.-J. Cho, P.T. Tran, C.M. Rudin

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): T.F. Burns, S.C. Murphy, K. Aziz, P.T. Tran

Study supervision: T.F. Burns, P.T. Tran, C.M. Rudin

The authors thank members of the Rudin, Burns, Tran, and Hann laboratories for discussion and advice regarding this work.

This research was supported by the Flight Attendant Medical Research Institute, the Burroughs Wellcome Fund (C.M. Rudin), the 2010 AACR-AstraZeneca Fellowship for Translational Lung Cancer Research, 2010 ASCO Young Investigator Award and an Institutional T32 grant “Molecular Targets for Cancer Detection and Treatment” (T.F. Burns), and a Parker B. Francis Fellowship, an ACS Scholar award (122688-RSG-12-196-01-TBG), Uniting Against Lung Cancer Award and ASTRO Junior Faculty Research Training Award (to P.T. Tran).

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.