RHOA-FAK Is a Required Signaling Axis for the Maintenance of KRAS-Driven Lung Adenocarcinomas

Activating mutations of the proto-oncogene KRAS (mutant KRAS) promote tumorigenesis in several common human cancers such as non–small cell lung cancer (NSCLC; ref. 1). Mutant KRAS exerts its oncogenic activity through the regulation of several signaling networks. Among these, the most extensively characterized are the RAF/MAP-ERK kinase (MEK)/extracellular signal-related kinase (ERK) and the phosphoinositide 3-kinase (PI3K)/AKT/mTOR signaling pathways (1, 2).

In NSCLC, KRAS mutations occur frequently in combination with inactivating mutations or epigenetic silencing of the CDKN2A locus, which encodes 2 distinct but overlapping tumor suppressors: p19/ARF (p14 in humans, ARF hereafter) and p16/INK4a (INK4a hereafter). Both p19/ARF and p16/INK4a restrain inappropriate cellular proliferation induced by mutant KRAS by positively regulating p53 and retinoblastoma (Rb) tumor suppressors, respectively (3). Indeed, mutant KRAS in association with CDKN2A deficiency results in high-grade lung and pancreatic cancer in mouse models and has been associated with the development of aggressive NSCLC in humans (4–7).

The genotype of cancer cells not only determines their phenotype, but also defines specific vulnerabilities that can be exploited in cancer therapy. Certain cancers are critically dependent on a single oncogenic activity, a phenomenon defined as oncogene addiction (8). For instance, continuous expression of mutant KRAS is required for the survival of NSCLC in both mouse cancer models and in human-derived cells (5, 9). However, attempts to develop direct inhibitors of mutant KRAS have been unsuccessful (10). Therefore, mutant KRAS is still a high-priority therapeutic target.

There has been a tremendous interest in identifying molecular targets that are required for the maintenance of mutant KRAS-dependent cancers (11–13). Pharmacologic inhibitors of MEK1/2, PI3K and/or mTORC1/2 led to promising antitumor effects in preclinical lung cancer models (14, 15). In addition, several compounds targeting RAF/MEK/ERK and PI3K/AKT/mTOR signaling pathways are currently under clinical investigation and hold promise for the treatment of RAS-mutant tumors (16). On the other hand, it is still unknown whether PI3K and MEK1/2 inhibitors are effective therapies in lung cancer. Thus, it is of interest to develop alternative therapeutic strategies that target mutant KRAS tumors.

The goal of this work was the identification of the vulnerabilities of mutant KRAS that can be harnessed for cancer therapy. For this purpose, we dissected the signaling pathways downstream of mutant KRAS in NSCLC developed in a genetically defined mouse model and in cellular systems. With this analysis, we determined that the RHOA-focal adhesion kinase (FAK) signaling axis is a critical vulnerability for high-grade lung tumors.

To identify the cellular networks required for the maintenance of high-grade lung cancer, we crossed tetracycline operator-regulated Kras^G12D (tetO-Kras^G12D) mice with Clara cell secretory protein-reverse tetracycline transactivator (CCSP-rtTA) mice (5) in a Cdkn2a-null background (Ink4a/Arf^−/−) (17). These mice express Kras^G12D in the respiratory epithelium when exposed to doxycycline.

In agreement with previous findings (5), the induction of Kras^G12D combined with Ink4a/Arf deficiency resulted in increased tumor burden as shown by histologic examination and tumor volume quantification of the lungs among Kras^G12D;Ink4a/Arf^+/+, Kras^G12D;Ink4a/Arf^+/− and Kras^G12D;Ink4a/Arf^−/− mice (Supplementary Fig. S1A and S1B).

After 12 weeks of doxycycline exposure, about 50% of the lungs of Kras^G12D;Ink4a/Arf^−/− mice were occupied by adenocarcinomas (high-grade tumors) consisting of cancer cells with atypical nuclei and cytoplasmic blebbing arranged in papillary structures. We did not detect any adenocarcinomas in the Kras^G12D;Ink4a/Arf^+/+ lungs, which contained only well-differentiated adenomas with some areas occupied by cells with atypical nuclei (Fig. 1A and B). Furthermore, the Kras^G12D;Ink4a/Arf^−/− tumors exhibited increased cell proliferation, as shown by the percentage of Ki-67–positive cells (Supplementary Fig. S1C and S1D) as well as upregulation of cyclin D1, a regulator of G₁–S phase transition (Supplementary Fig. S1E). Accordingly, mice lacking Ink4a/Arf displayed a remarkable reduction in median survival compared with Kras^G12D;Ink4a/Arf^+/+ and Kras^G12D;Ink4a/Arf^+/− mice (Supplementary Fig. S1F).

To gain insights into the cellular networks that regulate high-grade lung cancer, we assessed the activation status of the main Kras^G12D-regulated signaling pathways in microdissected tumors. We determined that Erk1/2 phosphorylation (p-Erk1/2^T202/T204) declined in Kras^G12D;Ink4a/Arf^+/+ lung tumors after 8 and 12 weeks of Kras^G12D induction, whereas it was sustained in Kras^G12D;Ink4a/Arf^−/− lung adenocarcinomas (Fig. 1C). Moreover, immunohistochemistry revealed that Erk1/2^T202/T204 staining was intense throughout the adenocarcinomas of Kras^G12D;Ink4a/Arf^−/− mice (Fig. 1D, red arrowhead) compared with adjacent adenomas (Fig. 1D, black arrowhead) and to Kras^G12D;Ink4a/Arf^+/+ tumors (Fig. 1D, bottom). Interestingly, other investigators reported that p-Erk1/2 is also deregulated in Kras^G12D-induced high-grade lung tumors in Trp53-null mice (18, 19). However, the functional significance of this event is still unknown.

The RHO family of small GTPases (GTPs) (which comprises RAC, CDC42, and RHO) has been implicated in mutant KRAS-induced tumorigenesis. These proteins regulate the cytoskeleton, cell migration, proliferation, and survival (20). RAC1 is required for mutant KRAS-induced transformation in fibroblasts in vitro and for initiating tumorigenesis in a mouse model of lung cancer (21, 22). In addition, ERK1/2 and RHOA regulate common pathways such as cell migration and chemotaxis (23). Indeed, RHOA is also required for mutant KRAS-induced transformation in vitro (24). Finally, deregulation of RHOA occurs in a variety of cancer types (25, 26). Therefore, we interrogated the functional status of Rac1 and RhoA during induction of Kras^G12D.

With glutathione S-transferase (GST) pull-down experiments, we found that the active form of Rac1 (Rac1-GTP) declined over time in Kras^G12D;Ink4a/Arf^−/− lung adenocarcinomas compared with Kras^G12D;Ink4a/Arf^+/+ tumors (Fig. 1E and Supplementary Fig. S2A). In contrast, as in the case of p-Erk1/2, the active form of RhoA (RhoA-GTP) was elevated in Kras^G12D;Ink4a/Arf^−/− adenocarcinomas compared with Kras^G12D;Ink4a/Arf^+/+ adenomas at 12 weeks after Kras^G12D induction (Fig. 1E and Supplementary Fig. S2B). We did not detect differences in Ras activity (Ras-GTP) between Kras^G12D;Ink4a/Arf^+/+ adenomas and Kras^G12D;Ink4a/Arf^−/− adenocarcinomas (Fig. 1E). This finding suggests that changes in total amount of Ras-GTP are not responsible for p-Erk1/2 deregulation.

In concordance with the GST pull-down experiments, immunohistochemical (IHC) stainings confirmed that RhoA-GTP was present solely in high-grade lung tumors of Kras^G12D;Ink4a/Arf^−/− mice (Fig. 1F and Supplementary Fig. S2C–S2E). Furthermore, we noted no differences in Rac1-GTP or RhoA-GTP in the respiratory epithelium of Ink4a/Arf^+/+ or Ink4a/Arf^−/− mice in the absence of Kras^G12D (Supplementary Fig. S2F).

Our observation that Rac1 and RhoA activation are mutually exclusive in vivo confirms long-standing observations obtained in tissue culture systems (27, 28) and suggests that antagonistic regulation of Rac1-RhoA signaling is of biologic significance. Consistent with these results, p27/Kip1, a cyclin-dependent kinase inhibitor whose degradation is promoted by RhoA-GTP (29, 30), was undetectable in high-grade Kras^G12D;Ink4a/Arf^−/− tumors (Supplementary Fig. S2G).

Our findings suggest that Erk1/2 is a positive regulator of RhoA in Kras^G12D;Ink4a/Arf^−/−-induced adenocarcinomas. To test this hypothesis, we conducted gain-of-function and loss-of-function experiments. Quantitative RHOA activity assays (G-LISA assays) revealed that ablation of KRAS resulted in the downregulation of RHOA-GTP in H460 and A549 human NSCLC cells (mutant KRAS, INK4a/ARF-deficient; Fig. 2A and Supplementary Fig. S3A), whereas its ectopic expression in H838 NSCLC cells (wild-type KRAS, INK4a/ARF-deficient) led to the upregulation of RHOA-GTP (Fig. 2B and Supplementary Fig. S3B). Moreover, ectopic expression of constitutively active MEK1 (MEK1^Q56P), which is the main regulator of ERK1/2 in H292 NSCLC cells (wild-type KRAS, INK4a/ARF-deficient), upregulated RHOA-GTP (Fig. 2C). Conversely, pharmacologic inhibition of MEK1/2 with UO126 and ARRY-142886 led to the downregulation of RHOA-GTP and the consequent upregulation of RAC1-GTP in H460 and A549 NSCLC cells (Fig. 2D and Supplementary Fig. S3C). In agreement with these findings, treatment of Kras^G12D;Ink4a/Arf^−/− mice with ARRY-142886 also resulted in the downregulation of RhoA-GTP in mouse lung tumors (Fig. 2E and F).

Next, we determined the mechanisms underlying p-ERK1/2–dependent upregulation of RHOA-GTP. It has been reported that p-ERK1/2 inhibits p190RHO GTPase-activating protein (p190RHOGAP), a specific GAP for RHOA, in fibronectin-stimulated cells in vitro (31). We determined whether this mechanism is functionally relevant in NSCLC induced by mutant KRAS. With pull-down assays, we detected an increased association of endogenous RHOA-GTP with p190RHOGAP after inhibition of MEK1/2 (Fig. 2G). Remarkably, knockdown of p190RHOGAP completely abolished the effect of MEK1/2 inhibition on RHOA (Fig. 2H). These findings indicate that p-ERK1/2 blocks the ability of p190RHOGAP to interact with RHOA and strongly supports the conclusion that p-ERK1/2 is the major activator of RHOA in this context.

To assess the functional significance of sustained RHOA-GTP in lung adenocarcinomas, we determined the effect of RHOA silencing in cell lines derived from human NSCLC specimens. Remarkably, RHOA silencing with 2 unrelated siRNAs resulted in apoptosis as shown by increased levels of cleaved PARP and cleaved caspase-3 in A549 cells (mutant KRAS, INK4a/ARF-deficient). This effect was virtually absent in HCC95 cells (wild-type KRAS, INK4a/ARF-deficient) suggesting that the apoptotic effect seen upon RNA interference (RNAi)-mediated RHOA ablation reflects a specific requirement for mutant KRAS (Fig. 3A and B). To confirm this result, we silenced RHOA in a panel of lung cancer cell lines expressing either mutant or wild-type KRAS in addition to INK4a/ARF deficiency. Indeed, we found that the presence of mutant KRAS in association with INK4a/ARF deficiency conferred vulnerability to RHOA silencing (72 hours after transfection; Fig. 3C and Supplementary Fig. S4A). RHOA silencing also resulted in loss of cell viability in NSCLC cells expressing mutant KRAS in association with mutations of TP53, although the effect was less pronounced compared with mutant KRAS, INK4a/ARF-deficient NSCLC cells (Fig. 3C and Supplementary Fig. S4A). Importantly, RHOA silencing produced similar results 120 hours after transfection (Supplementary Fig. S4B), suggesting that these effects are not due to differences in doubling time between cell lines. Taken together, these results indicate that mutant KRAS in combination with INK4a/ARF deficiency triggers a requirement for RHOA-GTP for tumor cell viability.

Next, we used human bronchial epithelial cells (HBEC3KT cells immortalized by introducing hTERT and cyclin-dependent kinase 4 (CDK4), which partially overcome the inhibitory effect of INK4a on cell-cycle progression) to test the impact of mutant KRAS expression in this context. The constitutive expression of mutant KRAS resulted in increased RHOA-GTP that became significantly higher upon additional p53 knockdown (Supplementary Fig. S4C and S4D). In both cases, the expression of mutant KRAS resulted in a considerable induction of cell death upon RHOA silencing (Supplementary Fig. S4E). Thus, HBEC3KT cells confirm our in vivo observations with transgenic mice and suggest that their susceptibility is specifically dependent on a genotype-induced activation of the ERK1/2-RHOA pathway.

To examine whether RHOA is required for the establishment of NSCLC, we conducted xenograft experiments using A549 cells, which are representative of the NSCLC cells we used in vitro. We transduced the A549 cells with a retrovirus expressing RHOA^T19N, a dominant negative mutant of RHOA. Indeed, we found that RHOA^T19N significantly decreased the amount of RHOA-GTP in A549 cells before implantation in mice and in tumors excised at the study end point (Supplementary Fig. S4F). We detected a more than 4-fold decrease in tumor formation of xenografts expressing RHOA^T19N (Fig. 3D and E), which correlated with a dramatic difference in survival (Fig. 3F). We conclude that the activation of RHOA is critical in promoting the growth of NSCLC.

To date, there are no pharmacologic drugs that target RHOA-GTP available for use in preclinical trials. Thus, we silenced the main direct and indirect downstream targets of RHOA, such as ROCK1, LIMK2, FAK, Villin 1, Cortactin, Kinectin, and DIAPH1 (30, 32, 33), to identify “druggable” therapeutic targets. We found that only the silencing of FAK caused significant loss of cell viability that at least in part recapitulated the effects on cell viability of RHOA silencing (Supplementary Fig. S5A and S5B). Indeed, siRNA-mediated FAK knockdown led to significant apoptosis (72 hours after siRNA transfection) in mutant KRAS;INK4A/ARF-deficient NSCLC cells (Fig. 4A and B). Moreover, FAK silencing also triggered apoptosis in mutant KRAS and p53-deficient cells (Fig. 4A and B). Finally, assessment of cell viability at 120 hours after transfection with siRNAs against FAK revealed a more dramatic and selective cell loss (Supplementary Fig. S5C). As predicted by our previous findings regarding RHOA activation (Supplementary Fig. S4C–S4E), ectopic expression of mutant KRAS in HBEC3KT cells increased FAK activation and cell death upon FAK silencing, an effect that was amplified by the knockdown of p53 (Supplementary Fig. S5D and S5E).

FAK plays a pivotal role in focal adhesion regulation and turnover in addition to the control of cell adhesion and cancer cell survival (34–37). Moreover, it has been shown that in some cell types, FAK also contributes to actin remodeling by regulating RHOA (38). Indeed, ablation of FAK in NSCLC cells was accompanied by disruption of focal adhesions, reduced actin stress fiber formation, and remarkable upregulation of the p27/KIP1 tumor suppressor (Fig. 4C and D).

Next, we confirmed that FAK is a physiologic target of RHOA in NSCLC cells. Silencing of RHOA induced a dramatic downregulation of p-FAK^Y397 in A549, H460, and H2009 cells (Fig. 4E). Conversely, forced expression of a constitutively active RHOA mutant (RHOA Q63L) led to increased p-FAK^Y397 (Supplementary Fig. S5F and S5G).

We predicted that, similar to RHOA-GTP, FAK activation would also be sustained in high-grade lung cancer in vivo. Indeed, immunoblot analysis conducted on microdissected lung tumors showed that p-Fak^Y397 declined to almost undetectable levels after 12 weeks of doxycycline exposure in Kras^G12D;Ink4a/Arf^+/+ lung adenomas, whereas it remained easily detectable in Kras^G12D;Ink4a/Arf^−/− adenocarcinomas (Fig. 5A). We found no differences in Fak activation between Ink4a/Arf^+/+ or Ink4a/Arf^−/− mouse lungs in the absence of Kras^G12D (Supplementary Fig. S6A).

IHC staining confirmed that p-Fak^Y397 was present solely in high-grade lung adenocarcinomas of Kras^G12D;Ink4a/Arf^−/− mice (Fig. 5B and Supplementary Fig. S6B). Finally, RhoA-GTP strongly colocalized with p-Fak^Y397 in mouse lung adenocarcinomas (Fig. 5C and Supplementary Fig. S6C).

To assess whether our results are relevant to human NSCLC, we used 120 primary human NSCLC specimens and assessed the mutation status of KRAS together with FAK activation. This analysis revealed an association between mutant KRAS and p-FAK^Y397 (Fig. 5D). In addition, we evaluated the activation status of RHOA/FAK and the mutation status of INK4a/ARF and p53 in an independent cohort of 20 consecutive primary human NSCLC specimens expressing mutant KRAS (Table 1 and Supplementary Fig. S6D and S6E). We found that tumors carrying either INK4a/ARF, TP53 tumor-associated mutations or INK4a/ARF hemizygous deletions evidenced high levels of activated RHOA and FAK (11/20 tumors; Table 1, ID number 1–11), whereas tumors that did not show strong coactivation of RHOA/FAK carried wild-type INK4a/ARF or TP53 (8/20 tumors; Table 1, ID number 13–20). We found only one specimen that did not conform to this pattern (Table 1, ID number 12). Although there was a significant association between the activation of RHOA/FAK and deficiency of the INK4a/ARF or TP53 locus, the small sample number does not allow the establishment of firm correlations. The observation that hemizygous deletions of INK4a/ARF occur mainly in specimens that show coactivation of RHOA and FAK suggests that these samples may harbor additional mutations inactivating the INK4a/ARF or p53 tumor suppressive networks. Moreover, it seems that FAK is active regardless of INK4a/ARF or p53 status in a minority of NSCLCs. This observation suggests that in these cases FAK activation may occur through RHOA-independent mechanisms. This could be explained by the fact that several signals, other than RHOA, positively regulate FAK (36).

Our findings suggest that Kras^G12D promotes the progression of lung adenomas into adenocarcinomas through a mechanism that involves sustained activation of the Erk1/2-RhoA-FAK signaling axis in Ink4a/Arf-deficient cancer cells. Thus, we predicted that inhibition of FAK would lead to significant antitumor effects specifically in lung adenocarcinomas and not in adenomas. We evaluated the preclinical efficacy of PF562271, an ATP-competitive inhibitor, in Kras^G12D;Ink4a/Arf^−/− and Kras^G12D;Ink4a/Arf^+/+-induced lung tumors. This inhibitor is well tolerated and displayed antitumor activity in a phase I clinical trial (39). We determined that PF562271 inhibits Fak phosphorylation in vivo in a dose-dependent manner. We selected 40 mg/kg daily as a nontoxic dose that effectively inhibits p-Fak^Y397 (Supplementary Fig. S7A).

Remarkably, treatment with PF562271 resulted in significant tumor regression in Kras^G12D;Ink4a/Arf^−/− mice, whereas it had antiproliferative effects in Kras^G12D;Ink4a/Arf^+/+ mice (Fig. 6A). Detailed histologic evaluation determined that the treatment with PF562271 significantly suppressed the number and size of Kras^G12D;Ink4a/Arf^−/− lung adenocarcinomas compared with the placebo-treated mice (Fig. 6B and C). Lung adenocarcinomas treated with PF562271 were distinguished by the presence of foamy macrophages and enrichment of reparative connective tissue (Fig. 6D, top). In addition, FAK inhibition caused localized induction of apoptosis and reduced cellular proliferation [Fig. 6D–F (bottom)].

FAK is a positive regulator of AKT (33). Indeed, we found that PF562271 not only efficiently inhibited p-FAK^Y397in vivo but also resulted in striking downregulation of p-Akt^S473 (Fig. 6G). The antitumor effects observed with PF562271 were confirmed with the FAK14 inhibitor (Supplementary Fig. S7B–S7G). FAK14 is structurally unrelated to PF562271 and inhibits the autophosphorylation of FAK at Y397 (40).

Finally, we conducted studies with A427 and A549 NSCLC cells to determine the impact of the PF562271 inhibitor on the survival of athymic nude mice carrying xenografts of human cells. Notably, we detected a significant increase in the survival rate of PF562271-treated mice compared with placebo-treated mice (Fig. 6H and I). We obtained similar results with the FAK14 inhibitor (Supplementary Fig. S7H and S7I). Thus, we conclude that pharmacologic inhibition of FAK is a potent therapeutic strategy for advanced KRAS-induced lung tumors.

The identification of patients that may benefit from treatment with targeted cancer therapies is still a significant challenge. In this study, we show that NSCLCs characterized by mutant KRAS and loss of either INK4a/ARF or p53 are highly sensitive to inhibition of the RHOA-FAK signaling axis. The data imply that pharmacologic inhibitors of FAK are effective, genotype-specific anticancer agents. Our findings are of clinical significance because these genotypes are associated with aggressive cancers, which are refractory to conventional therapy.

Recently, several groups reported strategies to induce antitumor responses in high-grade mouse lung cancer (11, 13, 41). To the best of our knowledge, our study is the first example of an effective genotype-specific monotherapy for high-grade mutant KRAS tumors.

Consistent with the known roles of FAK in the regulation of the cytoskeleton, we determined that its inhibition resulted in the reduction of F-actin stress fibers, disruption of focal adhesions, induction of the p27/Kip1 tumor suppressor, and decreased p-AKT^S473. These events occurred in conjunction with the induction of apoptosis; therefore, we propose that multiple cooperative functions of FAK contribute to its requirement for the maintenance of high-grade lung cancer. In addition, our studies show that FAK is the main effector of RHOA. However, it is still possible that other downstream targets of RHOA may contribute to its tumor-promoting ability. Future studies will be necessary to determine the mechanisms of cell death that contribute to this antitumor response.

Pharmacologic inhibition of MEK1/2 leads to compensatory upregulation of the PI3K/AKT signaling pathway (42), which in turn promotes cancer cell survival. On the contrary, we have shown that pharmacologic inhibition of FAK in vivo downregulates p-Akt. Thus, inhibition of FAK does not trigger the emergence of PI3K/AKT-dependent compensatory mechanisms. Collectively, these data show that the inhibition of an ultimate effector-arm of mutant KRAS, in this case RHOA/FAK, has detrimental antitumor effects.

It has been reported that p19/ARF and p53 restrain the progression of lung adenomas into adenocarcinomas and that their loss leads to the upregulation of MEK1/2 signaling through multiple mechanisms, including genomic amplification of mutant KRAS, inactivation of negative feedback mechanisms, or emergence of cooperative oncogenes (18, 19). We did not detect differences in the overall Ras activity (Ras-GTP) between Kras^G12D;Ink4a/Arf^+/+ adenomas and Kras^G12D;Ink4a/Arf^−/− adenocarcinomas. Thus, we propose that in this mouse model, mechanisms other than increased Ras-GTP signaling are responsible for the deregulation of p-Erk1/2.

Several questions remain to be answered regarding the role of FAK in lung cancer. As shown by our IHC data, a subset of mutant KRAS NSCLCs displayed upregulation of p-FAK in the absence of INK4a/ARF or p53 mutations/deletions. Furthermore, a subset of wild-type KRAS NSCLCs displayed either strong or moderate p-FAK staining. It will be of interest in the future to determine the significance of FAK activation in these settings. Furthermore, larger cohorts of patients will be needed to firmly establish that a correlation exists between mutant KRAS, INK4a/ARF, and/or p53 deficiency and activation of RHOA-FAK in human primary NSCLCs.

In view of the fact that NSCLCs are often comprised of heterogeneous populations of neoplastic cells, a possible mechanism of emergence of resistance to FAK inhibitors could be fueled by the persistence of neoplastic clones primarily driven by low-level oncogenic signals that are still able to develop high-grade tumors. Although this targeted therapy will have a significant benefit in cancer treatment, the elimination of less-advanced tumors is still an unmet need that must be resolved.

We conclude that the RHOA-FAK signaling axis is a novel, critical synthetic lethal partner of mutant KRAS in NSCLCs that are INK4a/ARF- or TP53-deficient. We propose that this information would serve as a biomarker for the selection of patients undergoing personalized cancer treatment protocols involving FAK inhibitors.

The PF562271 inhibitor (Pfizer) was reconstituted in 50% dimethyl sulfoxide and 50% PEG300 and additionally diluted with saline to a final concentration of 40 mg/kg before administration to the mice twice daily for a total of 12 days by oral gavage. The FAK14 inhibitor (Tocris) was reconstituted in H₂O and administered to the mice by intraperitoneal injection, once daily for a total of 10 days at the dosage of 20 mg/kg for the transgenic mice and 30 mg/kg for athymic nude mice. ARRY-142886 (AZD6244) (Selleck) was reconstituted in 0.5% methyl cellulose (Fluka) and 0.4% polysorbate (Tween 80; Fluka) and administered at 35 mg/kg by oral gavage. Xenograft experiments using A549 or A427 NSCLC cells were carried out by subcutaneous inoculation of 2 × 10⁶ cells into 6-week-old female athymic nude mice (nu/nu). For all in vivo experiments, we used age-matched littermates. The body weight of the mice remained stable during treatment with PF562271 and decreased by less than 5% during treatment with FAK14. Tumor burden was assessed by digital quantification of the area occupied by tumors compared with unaffected tissue using NIH ImageJ (v1.42q) software. All studies were done according to the guidelines of the University of Texas Southwestern Institutional Animal Care and Use Committee (Dallas, TX). See also Supplementary Methods.

Human NSCLC cell lines H125, H441, H2087, H522, EKVX, H2030, H1264, HCC95, HCC44, H2009, H358, H460, H1563, A549, and HBEC3KT cells (HBEC3 cells immortalized by introducing CDK4 and hTERT), together with the variants HBEC3KT-shp53, HBEC3KT-KRAS, and HBEC3KT-shp53+KRAS, were kindly provided by Dr. John Minna (UT Southwestern Medical Center; refs. 43–45). All NSCLC cell lines have been tested and authenticated by DNA fingerprinting using the PowerPlex 1.2 kit (Promega) and confirmed to be the same as the DNA fingerprint library maintained by American Type Culture Collection and the Minna/Gazdar lab (the primary source of the lines). See also Supplementary Methods.

siRNA (siGenome) against RHOA, FAK, p190RHOGAP, or nontargeting siRNA control were purchased from Dharmacon (Thermo Scientific). See also Supplementary Methods.

Cells were lysed in cold MLB buffer (125 mmol/L Hepes pH 7.5, 750 mmol/L NaCl, 5% NP-40, 50 mmol/L MgCl₂, 5 mmol/L EDTA, protease inhibitors cocktail tablet, phenylmethylsulfonylfluoride). See also Supplementary Methods.

The exonic regions of interest (NCBI Human Genome Build 36.1) were broken into amplicons of 350 bp or less, and specific primers were designed using Primer 3 to cover the exonic regions plus at least 50 bp of intronic sequences on both sides of intron–exon junctions. See also Supplementary Methods.

Mutations were detected using an automated detection pipeline at the Memorial Sloan-Kettering Cancer Bioinformatics Core (New York, NY). Bidirectional reads and mapping tables (to link read names to sample identifiers, gene names, read direction, and amplicon) were subjected to a quality control filter which excludes reads that have an average phred score of <10 for bases 100 to 200. See also Supplementary Methods.

No potential conflicts of interest were disclosed.

Conception and design: G. Konstantinidou, G. Ramadori, P.P. Scaglioni

Development of methodology: G. Konstantinidou, P.P. Scaglioni

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): G. Konstantinidou, K. Kangasniemi, R.E. Ramirez, Y. Cai, C. Behrens, M.T. Dellinger, R.A. Brekken, I.I. Wistuba, A. Heguy, J. Teruya-Feldstein

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): G. Konstantinidou, G. Ramadori, A. Heguy

Writing, review, and/or revision of the manuscript: G. Konstantinidou, G. Ramadori, R.E. Ramirez, Y. Cai, I.I. Wistuba, P.P. Scaglioni

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): R.E. Ramirez, Y. Cai

Study supervision: P.P. Scaglioni

The authors thank Michael A. White for critically reading the manuscript, John D. Minna, Michael Peyton, and Jill Larsen for providing reagents (UT Southwestern Medical Center), Marina Asher (Memorial Sloan-Kettering Cancer Center) for technical assistance in paraffin staining, and Masaya Takahashi (UT Southwestern Medical Center) for advice on MRI imaging.

This study was supported by grants from NIH K08 CA 112325, R01CA137195, American Cancer Society Institutional Research Grant #02-196, the Concern Foundation, the Gibson Foundation, Leukemia Texas Inc. (to P.P. Scaglioni), Cancer Prevention and Research Institute of Texas RP101496 (to G. Konstantinidou), Department of Defense (to M.T. Dellinger), American Heart Association (to G. Ramadori), Department of Defense W81XWH-07-1-0306, and Specialized Program of Research Excellence in Lung Cancer P50CA70907 (to I.I. Wistuba). In vivo imaging studies were made possible by the UT Southwestern small animal imaging resource, supported in part by grants NCI U24 CA126608, NIH BTRP P41-RR02584, and the Harold C. Simmons Cancer Center through NCI Cancer Center support grant 1P30 CA 142543-01.