Research ArticleLUNG CANCER

Frequent and Focal FGFR1 Amplification Associates with Therapeutically Tractable FGFR1 Dependency in Squamous Cell Lung Cancer

See allHide authors and affiliations

Science Translational Medicine  15 Dec 2010:
Vol. 2, Issue 62, pp. 62ra93
DOI: 10.1126/scitranslmed.3001451


Lung cancer remains one of the leading causes of cancer-related death in developed countries. Although lung adenocarcinomas with EGFR mutations or EML4-ALK fusions respond to treatment by epidermal growth factor receptor (EGFR) and anaplastic lymphoma kinase (ALK) inhibition, respectively, squamous cell lung cancer currently lacks therapeutically exploitable genetic alterations. We conducted a systematic search in a set of 232 lung cancer specimens for genetic alterations that were therapeutically amenable and then performed high-resolution gene copy number analyses. We identified frequent and focal fibroblast growth factor receptor 1 (FGFR1) amplification in squamous cell lung cancer (n = 155), but not in other lung cancer subtypes, and, by fluorescence in situ hybridization, confirmed the presence of FGFR1 amplifications in an independent cohort of squamous cell lung cancer samples (22% of cases). Using cell-based screening with the FGFR inhibitor PD173074 in a large (n = 83) panel of lung cancer cell lines, we demonstrated that this compound inhibited growth and induced apoptosis specifically in those lung cancer cells carrying amplified FGFR1. We validated the FGFR1 dependence of FGFR1-amplified cell lines by FGFR1 knockdown and by ectopic expression of an FGFR1-resistant allele (FGFR1V561M), which rescued FGFR1-amplified cells from PD173074-mediated cytotoxicity. Finally, we showed that inhibition of FGFR1 with a small molecule led to significant tumor shrinkage in vivo. Thus, focal FGFR1 amplification is common in squamous cell lung cancer and associated with tumor growth and survival, suggesting that FGFR inhibitors may be a viable therapeutic option in this cohort of patients.


Oncogenic protein kinases are frequently implicated as potential targets for cancer treatment. For examples, the ERBB2 amplification in breast cancer is associated with clinical response to antibodies targeting ERBB2 (1), and KIT or PDGFRA (platelet-derived growth factor receptor A) mutations in gastrointestinal stromal tumors lead to sensitivity to the KIT/ABL/PDGFR inhibitor imatinib (2). In lung adenocarcinoma, patients with EGFR-mutant tumors (35) experience tumor shrinkage and prolonged progression-free survival when treated with epidermal growth factor receptor (EGFR) inhibitors (6). Furthermore, EML4-ALK gene fusion–positive lung cancers can be effectively treated with anaplastic lymphoma kinase (ALK) inhibitors (7, 8).

However, these alterations almost exclusively occur in the rare adenocarcinomas of patients who never smoked, but are uncommon in squamous cell lung cancer, which is almost invariably associated with smoking (9). Although previous studies have reported recurrent genetic alterations in squamous cell lung cancer (10), no therapeutically tractable targets have so far been identified. Thus, therapeutic options for squamous cell lung cancer patients remain scarce, because molecularly targeted drugs such as erlotinib, gefitinib, pemetrexed, and cetuximab are either poorly active (6, 11) or contraindicated (for example, bevacizumab) (12). These observations emphasize the need for new “druggable” targets in squamous cell lung cancer patients.


To identify therapeutically relevant genome alterations in squamous cell lung cancer, we analyzed 155 primary squamous cell lung cancer specimens using Affymetrix 6.0 SNP (single-nucleotide polymorphism) arrays, which yielded high-resolution genomic profiles (median intermarker distance <1 kb). To separate driver lesions from random noise, we applied the GISTIC algorithm (13, 14). We identified 25 significant amplification peaks, including the previously described amplification of SOX2 on chromosome 3q26.33 (Fig. 1A and table S1) (10) and 26 significant deletions (fig. S1 and table S1). The second most significant amplification (q = 8.82 × 10−28) peak was identified on 8p12 and included FGFR1 (fibroblast growth factor receptor 1) as well as FLJ43582 in each amplified sample (Fig. 1A). This region spanned 133 kb (table S1) and was amplified at high amplitude (four or more copies) in 15 of 155 (9.7%) squamous cell lung cancer specimens (Fig. 1A). Notably, 11 of the tumors with FGFR1 amplification were from smokers, whereas none of these were from patients who had never smoked (table S2). Ten of the 15 tumors with amplified amounts of FGFR1 also harbored a mutation in TP53 (table S2). Moreover, patients who had tumors with FGFR1 amplification [copy number > 9 in fluorescence in situ hybridization (FISH) analysis] had a nonsignificant trend toward inferior survival compared to patients whose tumors lacked FGFR1 amplifications (copy number = 2 in FISH analysis) (fig. S2). We next analyzed copy number alterations in lung adenocarcinoma specimens (n = 77) and found no significant (q > 0.25) amplification (four or more copies; 1.3%) at 8p12 (Fig. 1B).

Fig. 1

FGFR1 is amplified in squamous cell lung cancer (SQLC). (A) Left panel: Significant (14) [FDR (false discovery rate) value; x axis] amplifications across all chromosomes (y axis) in SQLC (n = 155) as assessed by GISTIC. Right panel: Copy number alterations (blue, deletion; white, copy number–neutral; red, amplification) at chromosome 8 (y axis) across all SQLC samples (x axis). Samples are ordered according to focal amplification of FGFR1. (B) Significant (G score; y axis) copy number changes in adenocarcinoma (AC; n = 77) (black line) and SQLC (red dotted line) at chromosome 8. The q value for the presence of 8p12 amplification is 8.82 × 10−28 for SQLC and greater than 0.25 for adenocarcinoma. The chromosomal positions of FGFR1 (8p12) and MYC are highlighted (black arrows). (C) Frequency of FGFR1 amplification (% of samples ≥ copy number 4; y axis) in non-SQLC from a published data set (14), adenocarcinoma, and SQLC. P values indicate statistical significance. (D) FISH analysis (green, control; red, FGFR1) of 153 SQLC samples (FGFR1-HA: copy number >9; FGFR1-LA: copy number >2 and <9; FGFR1-N: copy number 2). Presented are example images from the three different FGFR1 amplification groups.

Finally, we analyzed a publicly available lung cancer SNP array data set (14) for the presence of FGFR1 amplifications (four or more copies); FGFR1 was amplified in 6 of 581 (1%) nonsquamous cell lung cancers (Fig. 1C). Thus, FGFR1 amplification is significantly enriched in squamous cell lung cancer when compared to our own adenocarcinoma data set (P = 0.03) (table S3) and when compared to a published data set of nonsquamous cell lung cancer (P < 0.0001) (Fig. 1C). FISH using an 8p12-specific probe on an independent set of 153 squamous cell lung cancers confirmed the presence of frequent high-level amplification of FGFR1 in 34 of 153 (22%) patients (Fig. 1D and table S4), 27 of whom were current smokers and none of whom were nonsmokers. We note that FISH is not sensitive to the admixture of nontumoral cells; thus, focal amplification of FGFR1 is likely to be more frequent in squamous cell lung cancer than as estimated by SNP arrays (table S4) (15). We also sequenced the FGFR1 gene in 94 squamous cell lung cancers and 94 adenocarcinomas and found one mutation (FGFR1P578H) in the adenocarcinoma cohort, indicating that FGFR1 mutations might play only a minor role and might not drive alterations in the pathogenesis of lung cancer (16).

Next, we performed high-throughput cell line screening (17, 18) to determine the activity of the non–isoform-specific FGFR inhibitor PD173074 (19) in a collection of 83 lung cancer cell lines (table S5) (17, 20). Of all cell lines tested, three had a half-maximal growth-inhibitory concentration (GI50 values) below 1.0 μM (Fig. 2A); remarkably, two of the three sensitive lung cancer cell lines exhibited focal amplification at 8p12 by 6.0 SNP array analysis (Fig. 2B), suggesting that FGFR1 amplifications are significantly (P = 0.005) associated with FGFR inhibitor activity (Fig. 2A). As expected, FGFR1-amplified cells expressed higher amounts of total FGFR1 protein (fig. S3). Previous studies indicated that expression of FGFR ligands might contribute to the sensitivity to FGFR inhibitors in lung cancer (21). We did not observe elevated amounts of FGF2 in the FGFR1-amplified cell lines (fig. S4A), nor did we observe a difference in the expression of FGFR ligands between patients harboring FGFR1 amplification and those without FGFR1 amplification (fig. S4B). However, FGFR1-amplified cells showed robust phosphorylation of FGFR, suggesting ligand-independent activation, which was further enhanced upon addition of exogenous FGF2 or FGF9 (fig. S4C), compatible with paracrine activation of FGFR1 in FGFR1-amplified cells. We next measured induction of apoptosis in FGFR1-amplified cells after treatment with PD173074 (Fig. 2C and table S6). The two PD173074-sensitive cell lines bearing FGFR1-amplifications (DMS114 and H1581) were also the ones with the highest proportion of apoptotic cells after treatment with the FGFR inhibitor. Furthermore, FGFR inhibition led to decreased colony formation of FGFR1-amplified but not of EGFR-mutant cells in soft agar (Fig. 2D), further enforcing the notion that amplification of FGFR1 drives proliferation of these lung cancer cell lines. Treatment with PD173074 reduced the amounts of phosphorylated FGFR1 and of the adaptor molecule FRS2 in a dose-dependent manner only in FGFR1-amplified H1581 cells, but not in the EGFR-mutant cell line HCC827 (Fig. 2E). We also observed inhibition of phosphorylation of extracellular signal–regulated kinase (ERK) but not of AKT and S6, indicating that the mitogen-activated protein kinase (MAPK) pathway, and not the phosphatidylinositol 3-kinase (PI3K) pathway, is the major signaling pathway engaged by amplified FGFR1 (Fig. 2E).

Fig. 2

FGFR1 amplifications are associated with FGFR inhibitor activity. (A) GI50 values (y axis) of PD173074 across 83 lung cancer cell lines (x axis). FGFR1-amplified (copy number ≥4) cell lines are marked with asterisks. (B) Copy number alterations (x axis; blue, deletion; white, copy number 2; red, amplification) on chromosome 8 with a zoom in on 8p12 (FGFR1 locus is highlighted) across all cell lines (y axis). (C) Induction of apoptosis (difference between PD173074 at 1 μM and DMSO control after 72 hours; y axis) across 24 cell lines (x axis; asterisks denote FGFR1 amplification copy number ≥4) as measured by flow cytometry (after annexin V/PI staining). (D) FGFR1-amplified cell lines were plated in soft agar and treated with either DMSO (control) or decreasing concentrations of PD173074. (E) Phosphorylation of FGFR and of downstream molecules in FGFR1-amplified (H1581 and H520) and in FGFR1 wild-type (EGFR-mutant) cells (HCC827) after treatment with PD173074 as assessed by immunoblotting.

To validate FGFR1 as the critical target of PD173074 in FGFR1-amplified lung cancer cells, we ectopically expressed the V561M mutation (22) at the gatekeeper position of FGFR1 (FGFR1V561M), preventing access of the compound to the hinge region of the kinase (23) (fig. S5). Expression of FGFR1V561M in FGFR1-amplified lung cancer cells abolished PD173074-mediated cytotoxicity and dephosphorylation of FGFR (Fig. 3A), consistent with the notion that FGFR1 is the critical target of PD173074 in FGFR1-amplified lung cancer cells. Furthermore, in a panel of 105 biochemically screened kinases, FGFR1 was one of only two kinases strongly inhibited by PD173074 (table S7), recapitulating previous studies (22).

Fig. 3

FGFR1-amplified cells are dependent on FGFR1 in vitro and in vivo. (A) Left panel: Viability (PD173074 treatment as compared to DMSO control) of FGFR1-amplified cells expressing wild-type (wt) or mutant (V561M) FGFR1 treated with PD173074 [0.5 μM (white bars) and 1.0 μM (gray bars)]. Right panel: Phosphorylation of FGFR in the FGFR1V561M and FGFR1wt cells detected by immunoblotting. (B) Left panel: Viability (PD173074 treatment as compared to DMSO control; y axis) of H1581 cells after transduction with control shRNA or shRNA targeting FGFR1. Right panel: Silencing of FGFR1 in H1581 cells was confirmed by immunoblotting. (C) In mice engrafted with H1581 cells treated with either vehicle or PD173074 (dosage as indicated; y axis), tumor volume was measured over time (x axis).

The high analytical resolution of the 6.0 SNP arrays, together with the large size of our data set, limited the number of candidate genes in the 8p12 amplicon to only two genes, FGFR1 and FLJ43582. A previous study analyzing the 8p12 locus in lung cancer applying lower-resolution techniques suggested WHSC1L1 to be the relevant oncogene in the 8p12 amplicon (24). To test whether genes other than FGFR1 drive tumorigenesis in the 8p12-amplified tumors, we silenced the genes WHSC1L1 (24) and FLJ43582 using five different short hairpin RNA (shRNA) constructs in the 8p12-amplified lung cancer cell line H1581. Although silencing of either one of these genes did not inhibit cellular viability (fig. S6), silencing of FGFR1 strongly reduced the viability of the FGFR1-amplified lung cancer cells (Fig. 3B). In light of the focality of the 8p12 amplicon (including FGFR1 and FLJ43582) and the lack of effect of shRNA-mediated knockdown of either FLJ43582 or WHSC1L1 in FGFR1-amplified cells, our data suggest that FGFR1 is the relevant target in these cells. Notably, the cell line H1703, which bears a copy number gain at 8p12 and that had been reported to depend on WHSC1L1 (24), was not sensitive to FGFR inhibition (fig. S7). By contrast, H1703 cells depend on PDGFRA for their survival (25) because of amplification (copy number >2.8) of the gene encoding this kinase (26, 27). Thus, our data suggest that the gene targeted by the 8p12 amplicon is primarily FGFR1 and its amplification induces FGFR1 dependency.

Finally, treatment with PD173074 (100 mg/kg, twice a day) resulted in tumor shrinkage in mice engrafted with FGFR1-amplified cells (Fig. 3C). This reduction in tumor size was paralleled by reduction in the amounts of phospho-ERK but not of phospho-AKT in immunohistochemical analyses of explanted tumors, validating our in vitro findings that MAPK signaling is the key pathway engaged by amplified FGFR1 (fig. S8A). Treatment at 50 mg/kg twice a day resulted in only a minimal exposure when compared to the gavage of 100 mg/kg twice a day because of the short half-life of the compound in vivo (fig. S8B). Thus, although we cannot formally exclude inhibitory effects on VEGFR2 (vascular endothelial growth factor receptor 2), the observed tumor regression is likely to be mediated by inhibition of FGFR1. In contrast, xenografted EGFR-mutant H1975 cells did not show signs of regression upon PD173074 treatment (fig. S8C). Thus, FGFR1 amplification leads to FGFR1 dependency in vivo.


Here, we have identified frequent high-level amplification of FGFR1 in squamous cell lung cancer of smokers; this amplification sensitizes the tumors to FGFR1 inhibition. Previous studies in lung cancer cohorts of mixed subtypes and low technological resolution (24, 28) or small size (10) have reported occasional amplification of the 8p locus in lung cancer. However, the large size of our sample set was necessary to reveal the high prevalence of this amplicon in squamous cell lung cancer (~10%) in comparison to other lung cancer subtypes (1%). Given the insensitivity of FISH analyses to admixture of nontumoral cells, the true prevalence of this amplification is likely to still be substantively underestimated by SNP arrays and to be up to 20%. We conclude that FGFR1 amplification is one of the hallmark alterations in squamous cell lung cancer, similar to amplification of SOX2. These two alterations were almost completely mutually exclusive (table S8), suggesting an epistatic relationship. Furthermore, FGFR1 amplification induced a strong FGFR1 dependency that could be exploited therapeutically, resulting in induction of apoptosis. Thus, FGFR1 amplification represents an opportunity for targeted therapy in squamous cell lung cancer. We therefore suggest that FGFR1 inhibitors, which are currently in clinical testing in tumor types bearing genetic alterations in FGFR genes (2931), should be evaluated in patients with FGFR1-amplified squamous cell lung cancer.

Materials and Methods

Genomic analyses

The tumor specimens analyzed in this study have been collected under local Institutional Review Board approval. All patients gave written informed consent. Genomic DNA was hybridized to Affymetrix 6.0 SNP arrays following the manufacturer’s instructions. Raw signal intensities were normalized and modeled with a Gaussian mixture model. Background-corrected intensities were normalized across all arrays of one batch by quantile normalization. Raw copy numbers were calculated by dividing the normalized tumor-derived signal intensities by the mean signal intensities derived from the normal samples hybridized in the same batch. Raw copy number data were segmented by circular binary segmentation and visualized in the integrated genome viewer (IGV) (32). GISTIC was performed as described previously (13, 14). The human genome build hg18 was used. Dideoxy sequencing was performed on whole-genome amplified DNA of primary tumors. Cell lines were sequenced with complementary DNA (cDNA). All raw data are publicly available [Gene Expression Omnibus (GEO); GSE25016].

Tissue microarray construction

Tissue microarray slides were obtained from formalin-fixed, paraffin-embedded lung squamous cell carcinoma samples. The tissue microarrays contained samples of a total of 172 patients from the University Hospital Zurich; each of these samples was present in duplicate cores, each core 0.6 mm in diameter (33). A second tissue microarray of 22 patients from Weill Cornell Medical Center was obtained, with each sample present in triplicate cores, each core 0.6 mm in diameter. Subsequently, 153 samples were used for FISH analysis.

Gene expression

After RNA isolation, biotin-labeled complementary RNA (cRNA) preparation was performed with Epicentre TargetAmp Kit (Epicentre Biotechnologies) and Biotin-16-UTP (10 mM; Roche Molecular Biochemicals) or Illumina TotalPrep RNA Amplification Kit (Ambion). Biotin-labeled cRNA (1.5 μg) was hybridized to Sentrix whole-genome bead chips WG6 version 2 (Illumina) and scanned on the Illumina BeadStation 500X. For data collection, we used Illumina BeadStudio software. Gene pattern analysis platform (34) was used to visualize the normalized data.

FGFR1 amplification FISH assay

A FISH assay was used to detect the FGFR1 amplification at the chromosomal level on the tissue microarrays. We performed fluorescence signal detection with two probes on chromosome 8. The reference probe is located on a stable region of chromosome 8p23.2 and selected on the basis of SNP array analysis. Only samples where the control bacterial artificial chromosome (BAC) was detectable were used for the determination of the copy number of FGFR1. The target probe is located on the FGFR1 locus spanning 8p11.23 to 8p11.22. We used the digoxigenin-labeled BAC clones CTD 2523O9, which produces a green signal, as reference probe. The target probe was labeled with biotin to produce a red signal with RP11-148D21 BAC clones (Invitrogen). Deparaffinized sections were pretreated with a 100 mM tris and 50 mM EDTA solution at 92.8°C for 15 min and digested with Digest-All III (dilution, 1:2) at 37°C for 14 min; FGFR1 FISH probes were denatured at 73°C for 5 min and immediately placed on ice. Subsequently, the tissue sections and FGFR1 FISH probes were co-denatured at 94°C for 3 min and hybridized overnight at 37°C. After hybridization, washing was done with 2× SSC at 75°C for 5 min, and the fluorescence detection was performed with streptavidin–Alexa 594 conjugates (dilution 1:200) and antibodies to digoxigenin–fluorescein isothiocyanate (FITC) (dilution, 1:200). Slides were then counterstained with 4′,6-diamidino-2-phenylindole (DAPI) and mounted. The samples were analyzed under a 63× oil immersion objective with a fluorescence microscope (Zeiss) equipped with appropriate filters, a charge-coupled device camera, and the FISH imaging and capturing software Metafer 4 (Metasystems). The evaluation of the tests was done independently by three experienced evaluators (R.M., S.M., and S.P.). At least 100 nuclei per case were evaluated. The thresholds for assigning a sample to the FGFR1 “high-amplification” group were a copy number of nine. All samples that had a copy number below nine and above two were assigned to the group of “low-amplification” cohort. All the remaining samples were assigned “normal.”

Cell lines and reagents

Cell lines were obtained from the American Type Culture Collection (ATCC), the German Resource Centre for Biological Material (DSMZ), or from our own and other cell culture collections and were maintained as described previously. PD173074 was purchased from commercial suppliers, dissolved in dimethyl sulfoxide (DMSO) or vehicle solution, and stored at −20°C.

Cell line screening

Cell line screening was performed as previously described (17) with various concentrations of PD173074. Viability was determined after 96 hours by measuring cellular adenosine triphosphate (ATP) content (CellTiter-Glo, Promega). Half-maximal inhibitory concentrations (GI50) were determined with the statistical data analysis software “R” with the package “ic50.”


For determination of apoptosis, cells were seeded in six-well plates, incubated for 24 hours, treated with either DMSO (control) or 1.0 μM PD173074 for 72 hours, and stained with annexin V and propidium iodide (PI). Finally, the cells were analyzed on a FACSCanto flow cytometer (BD Biosciences). The difference between the relative percentage of annexin V/PI–positive cells treated with DMSO and cells treated with PD173074 was determined (induction of apoptosis rate).

Lentiviral RNA interference and retroviral expression

The V561M mutation was introduced into FGFR1 cloned in pBABE-Puro by site-directed mutagenesis. Replication-incompetent retroviruses were produced by cotransfection with the pCL-ampho plasmid in human embryonic kidney (HEK) 293T cells. Hairpin shRNA targeting the different genes was ordered from Sigma. All sequences are given in table S10. Replication-incompetent lentiviruses were produced from pLKO.1-Puro–based vectors by cotransfection with Δ8.9 and pMGD2 in 293T cells as described previously (35). After transduction, cells were selected with puromycin (1.5 μg/ml), and 5 days after selection, cells were counted with trypan blue.

Western blotting

The following antibodies were used for immunoblotting: β-actin (MP Bioscience); phospho-FGFR (Tyr653, Tyr654), phospho-FRS2 (Tyr196), phospho-AKT (Ser473), phospho-S6, S6, AKT, phospho-ERK, and ERK (Cell Signaling Technology); total FGFR1 (Santa Cruz Biotechnology); and horseradish peroxidase (HRP)–conjugated antibodies to rabbit and mouse (Millipore).

Soft agar assay

Cells were suspended in growth media containing 10% fetal calf serum (FCS) and 0.6% agar and plated in triplicate on 50 μl of solidified growth medium (10% FCS; 1.0% agar). Growth medium containing indicated compound concentrations was added on top. Colonies were analyzed with the Scanalyzer imaging system (LemnaTec).

Xenograft mouse models

All animal procedures were approved by the local animal protection committee and the local authorities. Tumor cells (5 × 106) were injected subcutaneously into male nude mice. After the tumors reached a size of at least 50 mm3, the animals were treated twice daily by oral gavage with PD173074 (15 mg/ml for 50 mg/kg or 30 mg/ml for 100 mg/kg schedule) dissolved in vehicle (sodium lactate) or vehicle detergent alone. Tumor size was monitored by measuring perpendicular diameters as described previously (17). For the determination of tumor growth under treatment with PD173074, each experiment presented in the figures compromises the measurement of five different tumors.

Statistical analyses

Tests for statistical significance were either two-tailed t tests or Fisher’s exact tests. Prediction of compound activity was performed with the KNN algorithm as described previously (17). Multiple hypothesis testing was performed with the statistical data analysis software R using P value adjustment.

Supplementary Material


Fig. S1. Significant deletions are observed in squamous cell lung cancer.

Fig. S2. FGFR1 amplification has no significant impact on overall survival of SQLC patients.

Fig. S3. FGFR1 amplification correlates with FGFR1 protein expression.

Fig. S4. Expression of FGFR ligands does not correlate with FGFR1 amplification status.

Fig. S5. PD173074 binds inside the ATP-binding pocket of FGFR1.

Fig. S6. Knockdown of genes adjacent to FGFR1 on 8p12 does not affect cell viability.

Fig. S7. PD173074 is not active in the PDGFRA- and FGFR1-amplified cell line H1703.

Fig. S8. PD173074 shows antitumor activity in vivo.

Table S1. Significant amplifications and deletions are noted in a subset of 155 SQLC samples.

Table S2. Clinical features and co-occurrent mutations of FGFR1-amplified SQLC samples.

Table S3. Significant amplifications and deletions are noted in a subset of 77 adenocarcinoma samples.

Table S4. FGFR1 amplification is detected using FISH on tumor microarrays.

Table S5. GI50 values are not associated with mutation status across the lung cancer cell line panel.

Table S6. PD173074 induces apoptosis in FGFR1-amplified cell lines.

Table S7. PD173074 has specific activity against two kinases.

Table S8. FGFR1 and SOX2 amplification in squamous cell lung carcinoma.

Table S9. Sequences of all shRNA constructs that were used in the study.


References and Notes

  1. Acknowledgments: We thank M. Meyerson and A. Bass for sharing unpublished data and C. Reinhardt and A. Ullrich for discussion. Funding: This work was supported by the Deutsche Krebshilfe (grant 107954 to R.K.T.); the German Ministry of Science and Education (BMBF) as part of the NGFNplus program (grants 01GS08100 to R.K.T. and 01GS08101 to J. Wolf and P.N.); the Max Planck Society (M.I.F.A.NEUR8061 to R.K.T.); the Deutsche Forschungsgemeinschaft (DFG) through SFB (TP6 to R.K.T. and R.T.U.; TP5 to L.C.H. and R. Buettner); the Ministry for Innovation, Science, Research and Technology of the State of Nordrhein-Westfalen (MIWT, 4000-12 09 to R.K.T. and B.K.); and an anonymous foundation to R.K.T. E.B. and C.B. were supported by the PNES INCA grant 2008. B.S. was supported by the International Association for the Study of Lung Cancer Young Investigator Award. W.P. was supported by a grant from the NIH National Cancer Institute. Z.W. was supported by the Royal Australasian College of Surgeons and a Raelene Boyle Scholarship. G.W. was supported by the Australasian Society of Cardiac and Thoracic Surgeons Foundation Grant, Peter MacCallum Foundation Grant, and a private research donation from family and friends of former patients. Author contributions: J. Weiss and M.L.S. designed and performed the experiments, analyzed the data, and wrote the manuscript. D.S. designed the experiments and analyzed the data. M.P., T.Z., F.L., and D.R. analyzed the data. J.M.H., R.T.U., R.M., S.M., F.F., S.H., M.K., J. Schöttle, F.G., I.D., S.Q., L.C.H., H.B.-W., I.B., and J.A. performed the experiments and discussed the data. A.S., H.M., P.W., S.A., Z.W., M.C., G.W., P.R., B.S., E.B., C.B., P.L., S.S., O.T.B., W.E.-R., C.L., I.P., J. Sänger, J.C., H.G., W.T., H.S., E.T., E. Smit, D.H., F.C., C.L., S.D., M.H., R. Beroukhim, W.P., B.K., M.B., R. Buettner, K.E., E. Stoelben, J. Wolf, and P.N. contributed critical tumor specimens and contributed to the discussion of the data. S.P. reviewed tumor histology, analyzed the FISH data, and wrote the manuscript. R.K.T. conceived the project, designed the experiments, analyzed the data, and wrote the manuscript. Competing interests: R.K.T. receives consulting and lecture fees from Sequenom, Sanofi-Aventis, Merck, Roche, Infinity, Boehringer, AstraZeneca, and ATLAS Biolabs, as well as research support from Novartis and AstraZeneca. J. Wolf is a member of advisory boards of Roche, AstraZeneca, Novartis, Amgen, Bayer, and Merck and has received lecture fees from Roche, AstraZeneca, and Merck. R. Beroukhim receives consulting fees for Novartis Institutes for BioMedical Research. W.P. receives consulting fees from Molecular MD, AstraZeneca, BMS, and Symphony Evolution. H.G. received research support from Eli Lilly, Roche, and Boehringer Ingelheim through the University Medical Center Groningen. The other authors declare that they have no competing interests. Accession numbers: All raw data are publicly available (GEO; GSE25016).

Stay Connected to Science Translational Medicine

Navigate This Article