Tankyrase and the Canonical Wnt Pathway Protect Lung Cancer Cells from EGFR Inhibition

Lung cancer is the leading cause of cancer death in the United States (1). NSCLC account for ∼82% of lung cancers (2). Since most NSCLC are diagnosed at late stages, prognosis is quite poor. Even when combining newer agents, such as bevacizumab with traditional chemotherapy, the median overall survival of patients with metastatic NSCLC remains less than 1 year, and only 3.5% of NSCLC patients survive 5 years after diagnosis (3, 4).

Adenocarcinomas, the most common histologic type of NSCLC, are frequently (40–80%) associated with mutations and/or amplifications in the epidermal growth factor receptor (EGFR) gene (5). For patients with activating mutations in EGFR, the EGFR tyrosine kinase inhibitors (TKI) erlotinib and gefitinib provide better treatment than conventional chemotherapy, with impressive response rates (>70%) and prolonged progression-free survival (2, 6–8). However, the duration of responses is short-lived, and all patients eventually relapse.

The dysregulation of the Wnt/β-catenin pathway has been observed in various forms of cancer. Approximately 90% of sporadic colon cancers exhibit aberrant Wnt signaling activity, usually as a result of adenomatous polyposis coli (APC) gene mutations, but also because of mutations in genes encoding β-catenin or Axin (9). Importantly, it has been recently shown that altered β-catenin expression, or WNT1, WNT3A, and WNT5A overexpression, are associated with poor prognosis in NSCLC patients (10). Thus, the Wnt/β-catenin pathway may play an important role in NSCLC pathogenesis and resistance to therapy.

Secreted Wnt family proteins bind to specific Frizzled (Fzd) receptor complexes on cells and activate intracellular pathways (11, 12). The canonical Wnt pathway regulates the activity of β-catenin and its ability to drive activation of specific gene targets. In the absence of a Wnt signal, β-catenin is targeted for degradation. APC and Axin compose a scaffold that permits β-catenin phosphorylation by casein kinase 1α (CK1α) and glycogen synthase kinase 3β (GSK3β), resulting in β-catenin ubiquitinylation and proteosomal degradation. Conversely, upon binding of a Wnt ligand to the Fzd receptor, Axin is shuttled to the cell membrane, preventing β-catenin degradation. Accumulating β-catenin can then enter the nucleus and on interaction with members of the Tcf/Lef transcriptional activators, facilitates the expression of various gene targets. Interestingly, the poly-ADP-ribosylating enzymes tankyrases 1 and 2, which play various roles in the cell including in telomere maintenance (13), have recently been shown to positively regulate canonical Wnt signaling (14).

Although targeted therapies, such as using EGFR TKI, have shown clinical promise, these therapies do not produce durable remissions for NSCLC. Although in some cases secondary mutations in the kinase may prevent inhibitor binding, additional escape mechanisms are mostly poorly defined. As lung cancers are typically diagnosed at advanced phases and seems to possess inherent or acquired survival mechanisms that can protect cells from EGFR inhibition, the discovery of pathways that mediate compensatory survival mechanisms could reveal novel therapeutic targets that render pharmacological EGFR inhibitors more effective for the treatment of lung cancer. For this reason, we carried out a genome-wide short hairpin RNA (shRNA) screen to identify gene products whose inhibition synergizes with the EGFR inhibitor gefitinib to eliminate NSCLC cells. Our screen identified a number of druggable gene products within the Wnt/tankyrase/β-catenin pathway as mediators of NSCLC maintenance during EGFR inhibition.

A total of 1 × 10⁷ cells from each cell line were transduced with the SBI shRNA library (multiplicity of infection of ∼0.3) and 72 hours later subjected to puromycin selection (1 μg/mL) for 10 days. Cells (1 × 10⁷) were treated in triplicate with gefitinib at IC₇₀ (0.6 μmol/L for H322C; 0.05 μmol/L for HCC4006) or vehicle (DMSO) for 48 hours followed by culture for 96 hours without drugs. Total RNA was isolated from each group using Trizol (Invitrogen) and reverse transcribed (using vector-specific primer) with M-MLV reverse transcriptase (Epicenter). The cDNA was amplified by nested PCR, with addition of Illumina-specific adapter sequences. After sequencing on an Illumina GenomeAnalyzer (Illumina), shRNAs were identified and the number of clusters for each shRNA sequence quantified.

We developed Bioinformatics for Next Generation Sequencing! (BiNGS!) for analyzing and interpreting synthetic lethal screen data (15). A preprocessing step filtered out erroneous and low-quality reads. Filtered reads were mapped against the shRNA reference library using Bowtie (16). Output from this step is a P × N matrix, where P and N represent shRNA counts and samples, respectively. A secondary filtering step removes shRNA reads mapping to sequences without gene annotations. We also filtered out shRNAs where the median raw count in the control group is greater than the maximum raw count in the treatment group if the shRNA is enriched in the control group, and vice versa. We then employed negative binomial to model the count distribution in the sequencing data using edgeR (17). We computed the q-value of false discovery rate for multiple comparisons for these shRNAs, and carried out meta-analysis by combining adjusted P-values for all shRNAs representing the same gene using weighted Z-transformation (18). We used the associated P-value, P(wZ), to sort lists of genes with differentially represented shRNAs.

Calu-3 cells were obtained from American Type Culture Collection (ATCC) in 2002. Drs. John Minna and Adi Gazdar (University of Texas Southwestern Medical School, Dallas, TX) provided HCC4006 cells in 2006. H322C were obtained from Dr. Al Moustada (Biotechnology Research Institute, Montreal, Quebec, Canada) in 2003. H3255 and H3122 were obtained from Dr. Bruce Johnson and Dr. Pasi Janne (Dana-Farber Cancer Institute, Boston, MA) in 2004 and 2008, respectively. Calu-3, HCC4006, H322C, H3255, and H3122 were last authenticated by STR DNA profiling in February 2012 by our Molecular Biology Core. The STR profiles for these lines matched the profiles on file at the ATCC (Calu-3) and profiling data provided by Drs. Minna and Gazdar (HCC4006, H322C, H3255, and H3122).

Athymic nude mice (4- to -6-week-old females) were obtained from the National Cancer Institute. Cultured cells (2 × 10⁶) were injected into the flanks of mice at day 0. Vehicle or gefitinib was injected i.p. daily (5/7 days). Tumor volumes were evaluated by caliper measurement and calculated by the formula: π × (short diameter)² × (long diameter)/6. Hematoxylin and eosin (H&E) and Ki67 staining were carried out using standard protocols.

Graphing and statistical analysis were carried out using PRISM-5 software. Unless otherwise indicated, 2-tailed Student t test was used.

See Supplementary Methods for descriptions of methods for lentivirus preparation, cell culture, colony forming assays, cell viability/cell cycle/apoptosis/senescence assays, Western blotting, RT-PCR, and chemicals used.

To identify gene targets whose inhibition cooperates with gefitinib to more effectively eliminate NSCLC cells, we designed a genome-wide RNAi-based loss-of-function screen (Fig. 1A). We generated EGFR TKI-sensitive NSCLC cell lines H322C and HCC4006 expressing lentiviral-encoded shRNAs targeting all known human genes, with most cells expressing a single shRNA. H322C, a bronchioalveolar carcinoma with wild-type EGFR, exhibits intermediate sensitivity to gefinitib (IC₅₀ 0.25 μmol/L). HCC4006, an adenocarcinoma that possesses an amplified EGFR gene with exon 19 deletion (an activating mutation), exhibits high sensitivity to gefitinib (IC₅₀ 0.02 μmol/L; ref. 19).

Each cell line was divided into 6 populations: 3 were treated with DMSO and 3 were treated with gefitinib for 48 hours at doses that inhibit expansion by ∼70%, followed by 96 hours of culture without drug. shRNA sequences were amplified, sequenced, identified, and statistically analyzed as described in Materials and Methods. Over 5 million shRNAs were read per sample, representing more than 60,000 unique shRNAs. Unsupervised hierarchical clustering of mapped shRNA sequences from the individual samples revealed that DMSO samples clustered together and gefitinib-treated samples clustered together (Fig. 1B and Supplementary Fig. S1). Thus, gefitinib treatment is reproducibly affecting the representation of shRNAs; therefore, this screen should reveal shRNA targeted gene products whose knockdown either increases or decreases gefitinib-mediated inhibition of NSCLC cells.

For synthetic lethality, we considered genes targeted by shRNAs that were depleted in the treatment group with P-values, P(wZ), of <0.05 as synthetic lethal hits (1237 and 748 genes each for H322C and HCC4006; Supplementary Table S1). Importantly, we identified 104 genes with significant underrepresentation in both NSCLC cell lines (Supplementary Table S1). As the inhibition of shRNA targeted genes sensitizes NSCLC cells to gefitinib, we have dubbed these genes as “SLuG”s, for Synthetic Lethal upon Gefitinib. The 104 synthetic lethal genes found in both cell lines are denoted as SLuG², and those identified in a single cell line will be labeled SLuG^H322C and SLuG^HCC4006. Note that SLuG²s are enriched in cancer-related genes (see legend of Supplementary Table S1).

Among identified SLuGs were multiple genes with known roles in protecting NSCLC or other cancer cells from EGFR inhibition, which provides confidence in the screening results. Examples include: (i) MET (c-Met; SLuG^H332C): amplification of which has been shown to contribute to preexisting and acquired clinical resistance to EGFR inhibitors (20–22). (ii) PIK3CA (SLuG^H332C and phosphoinositide-3-kinase α subunit): pharmacological inhibition of the PI3K/Akt/mTOR pathway restores sensitivity in gefitinib-resistant tumor cell lines, forming the basis of a current phase I trial combining gefitinib with everolimus (PI3K/Akt/mTOR pathway inhibitor; ref. 23). (iii) Fibroblast growth factor 2 (FGF2) and the FGF receptor FGFR2 (both SLuG^H332C): FGFR signaling has been shown to confer intrinsic resistance to EGFR inhibition (24–26). (iv) BIRC5 (SLuG²): overexpression of BIRC5/survivin has been shown to protect lung cancer cells from gefitinib-induced apoptosis, and has been implicated in conferring resistance to gefitinib (27). (v) PTPN11 (SHP2; SLuG²): inhibition of SHP2 has been shown to sensitize to gefitinib (28), and (vi) SOS2 (SLUG^HCC4006): SOS2 was recently discovered in a synthetic lethal siRNA screen with EGFR inhibition in a cervical adenocarcinoma cell line (29).

Importantly, multiple SLuG gene products mapped throughout the Wnt/tankyrase/β-catenin pathway (Fig. 1C). In particular, TNKS1 was present as a SLuG² and TNKS2 was identified as a SLuG^HCC4006. These poly-ADP-ribosylating enzymes have recently been shown to be key positive regulators of β-catenin–dependent transcription by promoting the degradation of Axin (14). The SLuG²CSNK2A1 (casein kinase 2α; CK2α) has been shown to promote canonical Wnt signaling by phosphorylation of β-catenin, leading to decreased degradation and increased transcriptional activity (30). Moreover, high expression of CSNK2A1 is predictive of poor prognosis in NSCLC patients (31). Additional SLuGs in this Wnt pathway include SLuG^H322C genes TCF3, DVL3, SMAD3, CCND1, WNT2B, PP2A, and DVL3, as well as SLuG^HCC4006 genes LRP8, FZD3, SMAD2, JUN, CSNK1A1 (CK1α), and APC2. Note that pathway genes include several β-catenin/TCF target genes, including the SLuG²BIRC5. Thus, the inhibition of numerous components of the canonical Wnt signaling pathway seems to potentiate gefitinib-mediated elimination of NSCLC cells.

To validate screening results, we generated H322C sublines expressing either negative control shRNAs or shRNAs targeting TNKS1. Vectors from the TRC collection were used for validation of TNKS1 and other SLuGs (in each case, pools of retroviral integrants), with shRNA sequences distinct from those used for the screen. As expected, knockdown of TNKS1 decreased mRNA levels of AXIN2 (Supplementary Fig. S2B), a well-characterized target of Wnt/β-catenin-dependent transcription (32). Although TNKS1 knockdown had a mild effect on cell expansion by itself, when coupled with gefitinib-mediated EGFR inhibition, NSCLC cell expansion was greatly inhibited (Fig. 2A and Supplementary Fig. S2A). We next asked whether pharmacological inhibition of tankyrase activity, using the small molecule inhibitor XAV939 (14), could also sensitize NSCLC cells to EGFR inhibition. As shown in Fig. 2B and Supplementary Fig. S2C, although exposure of H322C and HCC4006 cells to XAV939 or gefitinib alone moderately inhibited proliferation, combination of the 2 drugs led to more potent inhibition of cell expansion. In addition, pharmacological stabilization of Axin using endo-IWR1, which also reduces β-catenin levels by inhibiting tankyrase (14, 33), also potentiated gefitinib-mediated elimination of H322C and HCC4006 cells (Fig. 2C). We validated the synergistic interaction of gefitinib and XAV939 or endo-IWR1 using 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assays in multiple NSCLC lines with varying sensitivities to gefitinib (Fig. 2D, E and Supplementary Fig. S2D and S2E), including H3255 (with activating EGFR mutation L858R) and Calu3 (with increased wild-type EGFR gene copy number; ref. 19). Note that synergism is strongest in the EGFR mutant lines. Importantly, we further showed that treatment with the FDA approved EGFR inhibitor erlotinib also synergized with XAV939 to reduce NSCLC cell expansion (Fig. 2F and Supplementary Fig. S2E). Interestingly, cell lines that are not driven by EGFR mutation or amplification, such as H3122 (ALK fusion) and human foreskin fibroblasts are not affected by the combinatorial treatment (Supplementary Fig. S2F and S2G) and tankyrase inhibition does not synergize with the first line NSCLC chemotherapeutic paclitaxel (Supplementary Fig. S2H), thus suggesting a unique reliance on Wnt signaling in EGFR-dependent cancers after EGFR inhibition. Finally, we showed that combined treatment of H322C or HCC4006 cells with gefitinib and either XAV939 or endo-IWR1 resulted in enhanced elimination of colony forming ability (Fig. 3A).

Although XAV939 alone did not affect cell-cycle profiles, H322C cells treated with gefitinib exhibited an increase of cells in G₁ and a decrease in S-phase (Fig. 3B), as expected. Notably, cells treated with both gefitinib and XAV939 exhibited an even greater number of cells in G₁ phase with a corresponding decrease in S and G₂–M phases relative to gefitinib alone. Although gefitinib treatment alone does increase the proportion of H322C cells undergoing apoptosis (Supplementary Fig. S3A), combined treatment with XAV939 only modestly increased the presence of apoptotic cells detected by Annexin-V staining. But notably, the combination treatment resulted in more cells with senescence-associated β-galactosidase staining than gefitinib alone (Fig. 3C and Supplementary Fig. S3B), indicating that the combinatorial treatment causes increased senescence to H332C cells. Moreover, known Wnt/β-catenin targets (AXIN2, MYC, and BIRC5) are differentially modulated by the inhibition of EGFR and TNKS (Fig. 3D). As expected, AXIN2 mRNA levels are decreased by XAV939 treatment, but surprisingly increased 2-fold by gefitinib. Interestingly, the combination of gefitinib and XAV939 induce a synergistic surge of AXIN2 mRNA levels. Whereas MYC mRNA levels are reduced by either gefitinib or XAV939 alone, the combinatorial treatment results in further reductions in MYC message levels. Finally, BIRC5 mRNA levels are reduced by gefitinib treatment, but seem unaffected by XAV939 alone. In summary, our results indicate that inhibiting both EGFR and TNKS synergistically impairs the proliferation of NSCLC cells, correlating with a marked increase in senescence, and altered mRNA levels of various Wnt target genes (model in Supplementary Fig. S3C).

We next sought to validate other SLuGs that also mapped to the canonical Wnt pathway. Wnt signaling is tightly controlled with multiple levels of feedback. For example, whereas CK1α can phosphorylate β-catenin and thus target it for destruction (34), it also can destabilize the destruction complex, thus promoting canonical Wnt signaling (35). As shown in Fig. 4A and B and Supplementary Fig. S4A and S4E, knockdown of either CSNK1A1 (CK1α) or CSNK2A1 (CK2α) impaired the proliferation and colony forming ability of H322C cells, and sensitized these cells to EGFR inhibition with gefitinib. Interestingly, pyrvinium, a small-molecule CK1α activator recently showed to inhibit the Wnt/β-catenin pathway (ref. 36; EC₅₀ ∼ 10 nmol/L), synergistically increases gefitinib-mediated inhibition of H322C and HCC4006 cells (Fig. 4C and Supplementary S4F and S4G). Pyrvinium has been an FDA approved an anti-helminth drug for >50 years, and is a potent, specific, and direct activator of CK1α (36). Although it may seem counterintuitive that both activation (with pyrvinium) and inhibition (with shRNAs) of CK1α would sensitize NSCLC cells to gefitinib, as mentioned earlier, previous studies have also suggested dual roles for CK1α in both positive and negative modulation of Wnt signaling (37). Notably, pyrvinium does not synergize with paclitaxel (Supplementary Fig. S4H) in NSCLC cells or with gefitinib in human foreskin fibroblasts (Supplementary Fig. S4I). These data reinforce the importance of appropriate regulation of Wnt signaling for NSCLC maintenance during EGFR inhibition (Fig. 4D).

Given the apparent importance of Wnt signaling in residual cell maintenance during gefitinib treatment, we directly assessed the contribution of β-catenin, even though it was not identified as a SLuG. Even modest knockdown of β-catenin had an inhibitory effect on cell expansion by itself, but when coupled with gefitinib, cell expansion was greatly inhibited (Fig. 5A and Supplementary Fig. S5A and S5B). Combined treatment with gefitinib and a recently described small molecule, PNU74654, that can disrupt the interaction between β-catenin and TCF4 (38), similarly led to enhanced inhibition of NSCLC expansion (Fig. 5B). Moreover, knockdown of TCF4 also potentiated gefitinib efficacy in both H322C and HCC4006 lines (Fig. 5C and Supplementary S5C). In summary, our results indicate that blocking both EGFR and Wnt/β-Catenin pathways leads to enhanced inhibition of NSCLC cells.

We next asked whether pharmacological inhibition of the Wnt/tankyrase/β-catenin pathway could cooperate with gefitinib by further inhibiting EGFR phosphorylation. As shown in Supplementary Fig. S6A, the inhibition of activating phosphorylation of EGFR and ERK, which is downstream of EGFR, after gefitinib treatment in HCC4006 and H322C cells was not affected by the addition of XAV939, pyrvinium, or PNU74654. Thus, the ability of Wnt/tankyrase/β-catenin pathway inhibitors to sensitize NSCLC cells to gefitinib does not seems to be because of augmented EGFR pathway inhibition. Moreover, XAV939 and endo-IWR1 treatments resulted in the expected increase in AXIN1 protein levels (Fig. 6A and B). Finally, all small molecules used to disrupt Wnt signaling (XAV939, pyrvinium, and PNU74654) decrease AXIN2 mRNA levels (Supplementary Fig. S6B).

Although pharmacological manipulation of the Wnt/tankyrase/β-catenin pathway has clear relevance for potential clinical applications, a caveat with the use of small-molecule inhibitors is their inevitable off-target effects. Moreover, the inhibition of enzymes such as tankyrase and CK1α will impact cellular processes beyond Wnt signaling (37, 39). To determine the extent to which the efficacy of these inhibitors in cooperating with gefitinib to impair NSCLC cell proliferation was through inhibition of the canonical Wnt pathway, we generated HCC4006 cells that stably express an activated form of β-catenin (S33A/S37A/T41A/S45A; ref. 40; Supplementary Fig. S6C). As shown in Fig. 6C, expression of activated β-catenin not only partially protects NSCLC cells from gefitinib treatment, but provides substantial protection from combined treatments with either XAV939 + gefitinib or pyrvinium + gefitinib. This β-catenin rescue is evident over a range of concentrations for gefitinib, XAV939, and pyrvinium (Fig. 6D and Supplementary Fig. S6D). Thus, the expression of activated β-catenin obviated the effects of pyrvinium, and to a partial extent XAV939, in terms of increasing HCC4006 inhibition after gefitinib treatment. Nonetheless, it is notable that activation of β-catenin cannot fully rescue growth inhibition by high-dose pyrvinium or XAV939 in the absence of gefitinib (Fig. 6C and D), which could represent off-target effects of these drugs or impacts of tankyrase inhibition beyond the β-catenin pathway. Still, the ability of activated β-catenin expression to largely bypass effects of the tankyrase inhibitor XAV939 and the CK1α activator pyrvinium on enhancing gefitinib-mediated NSCLC inhibition indicates that these small molecules, at least in good measure, act by inhibiting the Wnt/β-catenin pathway.

Finally, we asked whether the inhibition of tankyrase could sensitize NSCLC cells to EGFR inhibition in vivo. We injected nude mice subcutaneously with H322C cells expressing either negative control shRNAs (Scrbl) or shRNAs targeting TNKS1. As shown in Fig. 7A and Supplementary Fig. S7, gefitinib treatment delayed, but did not prevent, the outgrowth of the Scrbl shRNA expressing tumor cells. In contrast, tumors expressing shRNAs targeting TNKS1 showed a much more substantial inhibition of tumor growth with gefitinib. H&E and Ki67 staining revealed that tumors from gefitinib-treated mice exhibited reduced cell density and proliferation, with greater reductions in cell density for TNKS1 shRNA expressing tumors (Fig. 7B).

Our results reveal that the Wnt/β-catenin pathway is critical for NSCLC cell maintenance upon EGFR inhibition. Given that key enzymatic regulators of the Wnt/β-catenin pathway, including tankyrase, can be inhibited by small molecules, these studies reveal potential mechanisms to increase the efficacy of clinically-used EGFR inhibitors such as gefitinib and erlotinib for the treatment of NSCLC. It is important to emphasize that our screen was not designed to discover acquired resistance mechanisms, but to reveal pathways that maintain some EGFR-dependent NSCLC cells despite EGFR inhibition. Clinical studies thus far have shown efficacy for small molecule EGFR inhibitors only for patients with NSCLC bearing activating EGFR mutations (∼15% of cases; refs. 2, 6), and even in these cases relapse is inevitable. Moreover, our experiments indicate that targeting tankyrase and the Wnt/β-catenin pathway may increase the efficacy of EGFR inhibitors for NSCLC that overexpress wild-type EGFR, which could expand the number of patients that benefit from anti-EGFR therapies.

RNAi screens provide powerful tools to identify novel compensatory pathways upon TKI treatment (29, 41–43). Our results highlight the strength of unbiased, genome-wide analyses, which provides unanticipated insight into the pathways that maintain NSCLC cells. Notably, we did not find significant correlations between our SLuG² list with genes that are regulated after gefitinib treatment or genes known to be mutated in lung cancers, and we found only very weak correlations with canonical Wnt target genes and gefitinib resistance of NSCLC lines (Supplementary Fig. S5D and legend). These results emphasize the importance of functional screens, as a pathway does not need to be mutated or deregulated to contribute to therapy resistance or maintenance of the cancer phenotype (44). Importantly, tankyrases have been identified by many pharmaceutical companies as attractive targets for drug development (45), and our studies reveal a potentially valuable application of tankyrase inhibitors for the treatment of lung cancers.

Our results suggest that inhibition of EGFR and tankyrases could each impact β-catenin-dependent transcription through divergent mechanisms. Although XAV939 (but not gefitinib) treated cells exhibited the expected increase in AXIN1 protein levels, gefitinib treatment resulted in increases in AXIN2 mRNA levels, with further increases when combined with XAV939, coinciding with decreased expression of Wnt/β-catenin target genes. Together with previous demonstrations of crosstalk between EGFR and canonical Wnt pathways (46, 47), these data suggest a more complicated impact of EGFR and tankyrase inhibitors on Wnt/β-catenin-dependent transcription than originally anticipated.

Clinical experience thus far has indicated that single-agent, molecularly-targeted chemotherapy will not be curative for NSCLC, or indeed for any advanced cancer. Given the molecular complexity of these cancers, compensatory survival pathways can mediate partial cancer cell maintenance and lead to drug resistance. As shown in a recent report (48), there is considerable heterogeneity in acquired resistance mechanisms in lung cancers, which become refractory to EGFR inhibitors, and one might envision that each patient's cancer will require an individualized therapy. Indeed, activation of the Wnt/tankyrase/β-catenin pathway should be examined as one potential mechanism of acquired resistance to EGFR-targeted therapies. Moreover, we would suggest that there are also pathways, such as the Wnt/tankyrase/β-catenin pathway, which may mediate intrinsic resistance to targeted therapies, such as with EGFR inhibitors, in a wider range of NSCLC. Nononcogene addiction of cancers (44), particularly when specific to a context (such as EGFR inhibition) that the cancer would presumably not have evolved to avoid prior to treatment, may be especially attractive for therapeutic targeting together with the mutationally-activated primary pathway of oncogene addiction.

In conclusion, our studies reveal the effectiveness of genome-wide shRNA screens to discover resistance signaling networks and to help exploit druggable pathways whose inhibition can potentiate current therapeutics for cancer. In particular, our results strongly implicate the Wnt/tankyrase/β-catenin pathway as a promising therapeutic target for NSCLC when combined with EGFR inhibition.

No potential conflicts of interest were disclosed.

The authors thank Mark Gregory, Courtney Fleenor, Joaquin Espinosa, Kelly Sullivan, and Lynn Heasley for their suggestions and critical review of the manuscript, and Cancer Center Functional Genomics, Flow Cytometry, and Biostatistics/Bioinformatics Shared Resources. The authors also thank AstraZeneca for the gift of gefitinib.

Studies were supported by RO1-CA157850, pilot funding from NCI SPORE P50-CA058187 and by Uniting Against Lung Cancer. Cancer Center Shared Resources are supported by NIH grant 2-P30-CA46934.

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.