Nullifying the CDKN2AB Locus Promotes Mutant K-ras Lung Tumorigenesis

Lung adenocarcinoma, a leading source of cancer-related mortality worldwide (1), harbors recurrent genomic abnormalities that provide a framework for the selection of targeted therapeutic agents (2–4). For these reasons, intense research efforts are under way to characterize the biologic and therapeutic significance of recurrent genomic abnormalities in non–small cell lung cancer (NSCLC). Activating mutations of KRAS occur in approximately 30% of lung adenocarcinomas. KRAS is a small GTPase that regulates several oncogenic networks (5). Mutant KRAS plays a causative role in lung tumorigenesis; however, it is not sufficient for the induction of high-grade lung adenocarcinomas in the absence of cooperating mutations that often involve the CDKN2AB locus (6–9).

The CDKN2AB locus encodes for several tumor suppressors. CDKN2A contains p16^INK4A and p14^ARF (p19^Arf in mice), while CDKN2B contains p15^INK4B. Both p16^INK4A and p15^INK4B promote G₁ cell-cycle arrest by inhibiting the CDK4/CDK6 retinoblastoma family of tumor suppressors, whereas p14^ARF upregulates TP53 by inactivating its negative regulator MDM2 (10, 11). Both p16^INK4A and p15^INK4B are highly similar and appear to have originated from gene duplication (12–14), whereas p14^ARF gene expression is initiated from an exon intercalated between p16^INK4A and p15^INK4B and an alternative reading frame of exon 2 and 3 of p16^INK4A (refs. 14, 15; Fig. 1A). Several studies established that p16^INK4A and p14^ARF are bona fide tumor suppressors in vivo, including in mutant KRAS lung adenocarcinomas (6, 9, 15–17).

Despite convincing biochemical evidence that p15^INK4B is part of the TGF-β signaling pathway, much less is known regarding its tumor suppressor function in vivo (18). For example, p15^INK4B is rarely mutated independently of the other CDKN2AB genes (14). Furthermore, in the absence of other mutations, p15^INK4B null mice have only a mild tumor predisposition (19). However, Krimpenfort and colleagues demonstrated that Cdkn2ab null mice are tumor-prone and develop an expanded tumor spectrum as compared with Cdkn2a null mice (20). This study led to the conclusion that p15^INK4B provides a tumor-suppressive function that is critical in the absence of p16^INK4A and p19^ARF (20). It is unknown whether p15^INK4B has a tumor-suppressive role in other tumor models or whether its status influences tumorigenesis driven by activating mutations of proto-oncogenes commonly occurring in human cancer. For instance, even though p15^INK4B loss has been reported to occur in lung adenocarcinoma, its biologic significance has yet to be established in this context (2, 21, 22).

In this article, we study the significance of Cdkn2ab deficiency in the biology of mutant Kras lung adenocarcinoma with two genetically engineered mouse models that express mutant KRAS in the respiratory epithelium and human lung cancer cell lines. Furthermore, we determined the mutation frequency and pattern of the CDKN2A and CDKN2B loci by analyzing The Tumor Cancer Genome Atlas database. Our data indicate that p15^INK4B provides a tumor-suppressive function in mutant KRAS lung tumorigenesis.

The cDNA of murine Cdkn2b (clone ID 3495097) was obtained from Open Biosystems (Thermo Scientific) and cloned into pLVX-tight-puro (Clontech Laboratories). Recombinant lentiviral particles were generated in 293T cells according to manufacturer's procedures.

Tet-op-Kras;CCSP-rtTA were obtained from H.E. Varmus (6), HIST1H2BJ-GFP mice were from the Jackson Laboratory (23), Ink4ab^−/− mice (hereafter named Cdkn2ab^−/−) were provided by Dr. Anton Berns (The Netherlands Cancer Institute, Amsterdam, the Netherlands), LSL-Kras^G12D mice were from the National Cancer Institute (NCI) mouse repository (24) and Ink4a/Arf^flox/flox mice were from the Jackson Laboratory (25). Mice were maintained in a mixed background (FVB/N/CD-1). Experiments were performed with F3 generation progeny or later progenies, and comparisons were made with littermates. We obtained lung-specific KRAS expression at 4 weeks of age either by feeding mice with doxycycline-implemented food pellets (Harlan Laboratories) or by intratracheal administration of Adenovirus-Cre at 8 weeks of age (University of Iowa, Gene transfer Vector Core; refs. 6, 26). We administered doxycycline at 4 weeks of age to attempt to develop lung tumors before the development of tumors involving other organs, a common occurrence in Cdkn2ab null mice. All animal studies were completed according to the policies of the University of Texas Southwestern Institutional Animal Care and Use Committee. We used digital quantification of the area occupied by tumors to the area of total lung using the NIH ImageJ (v1.42q) software.

Single-cell suspensions of primary respiratory cells were sorted with a MoFlo cell sorter to select and plate GFP fluorescent cells on either irradiated or mitomycin C–treated mouse embryonic fibroblasts (MEF).

Human NSCLC cell lines H460 and A549 were from the Hamon Center cell line repository (UT Southwestern Medical Center, Dallas, TX).

We extracted RNA using the RNeasy Kit (Qiagen) and generated cDNA using iScript cDNA kit (Bio-Rad). For quantitative real-time PCR (qRT-PCR), we used iTaq SYBR green supermix with ROX (Bio-Rad).

We used the following antibodies: anti-PARP (Cell Signaling Technology), anti-HSP90 (BD Biosciences), anti-p15 (Sigma), and anti-phospho-H3 (Millipore).

We used the following antibodies: GFP (Chemicon), phospho-S6 ribosomal protein, HMG2A and phospho-Erk (Cell Signaling Technology), ALDH1A1 (Abcam), phospho-H3 (Millipore), NKX2-1 (Santa Cruz Biotechnology), SP-C (Chemicon), and CCSP (Santa Cruz Biotechnology). An HMG2A staining scoring system was used as follows: tumors were considered negative if no stain was visible, positive in few cells, positive in clusters of cells, and positive if >50% of tumor cells scored HMG2a positive.

Information regarding mutation, segmented copy number, methylation, and clinical data of primary lung adenocarcinomas were obtained from TCGA data portal (https://tcga-data.nci.nih.gov/tcga) on June 8, 2013. Only nonsilent mutations were considered as described (27). Kaplan–Meier was used to estimate the survival curves and comparisons were performed using the log-rank test using SPSS Statistics 17.0. One-way ANOVA followed by Tukey post hoc test was used to compare the pathologic stage among groups.

Most of the KRAS and CDKN2A mutations data were obtained from the COSMIC database (Sanger Institute, United Kingdom), the literature, or from our unpublished whole-exome sequencing data (28). mRNA expression microarrays were performed by us with Illumina HumanWG-6 V3 as previously described (29). They were deposited at Gene Expression Omnibus (http://www.ncbi.nlm.nih.gov/geo) under the accession GSE32036. Whole-genome Single-nucleotide polymorphism (SNP) profiling was done as previously described (30).

To characterize the effect of deficiency of genes encoded by Cdkn2ab on lung tumorigenesis (Fig. 1A), we generated compound mutant mice expressing mutant KRAS in the respiratory epithelium. We crossed LSL-Kras^G12D mice (LSL-Kras mice), which encode a latent mutant Kras gene, with mice harboring conditional alleles of Cdkn2a or Cdk2ab (20, 24, 25). We achieved lung-specific tumor development by intratracheal delivery of adenovirus encoding for the Cre recombinase.

The overall median survival of LSL-Kras;Cdkn2a^−/− was significantly decreased as compared with their littermates with a wild-type Cdkn2a locus, whereas LSL-Kras;Cdkn2a^+/− had a trend toward an intermediate survival between Cdkn2a wild-type and null genotypes (Fig. 1B). The survival of LSL-Kras;Cdkn2ab^−/− mice was significantly decreased as compared with LSL-Kras, LSL-Kras;Cdkn2ab^+/− or LSL-Kras;Cdkn2a^−/− mice (Fig. 1C). These findings demonstrate that complete loss of Cdkn2ab significantly promotes mutant KRAS lung tumorigenesis.

At the time of death, the majority of lung parenchyma was occupied by lung tumors in every genotype (Supplementary Fig. S1A–S1C). This finding suggests that death is caused by respiratory compromise.

We determined genotype-specific histologic features: lungs of LSL-Kras and LSL-Kras;Cdkn2a^−/− mice contained predominantly lung adenomas (Fig. 1D and E, panel 1 and 2) and occasional carcinomas with nuclear atypia (not shown). In contrast, Cdkn2ab null mice carried primarily high-grade lung adenocarcinomas consisting of atypical nuclei arranged in papillary structures (Fig. 1E, panel 3). Moreover, we detected tumors arising from the bronchial epithelia in all three genotypes (Fig. 1D and E, panel 4).

We observed that 44% of LSL-Kras;Cdkn2a^−/− mice carried metastatic lesions in mediastinal lymph nodes, or the thoracic wall (Fig. 1F; Supplementary Table S1). This metastatic behavior was even more pronounced in LSL-Kras;Cdkn2ab^−/− mice, where 89% of the mice carried lung tumors that metastasized to mediastinal lymph nodes and the myocardium (Fig. 1F; Supplemental Table S1). These finding indicate that p15^Ink4b deficiency leads to high-grade lung adenocarcinomas with aggressive behavior, including the property to generate locoregional metastasis.

We determined the level of activity of targets of KRAS, which have been implicated in tumorigenesis, and of the mitotic marker phospho-Histone 3 (p-H3). LSL-Kras;Cdkn2a null and LSL-Kras;Cdkn2ab null tumors were diffusely positive for p-S6 and p-Erk1/2, with more intense staining in tumor segments with atypical nuclei and papillary morphology, whereas LSL-Kras;Cdkn2ab^+/+ lung tumors were faintly positive for p-S6 and p-Erk1/2 (Fig. 2A).

The staining pattern of p-H3 was genotype dependent: LSL-Kras;Cdkn2a^−/− tumors stained positive for p-H3 mostly at the tumor edge. In contrast, LSL-Kras;Cdkn2ab^−/− tumors stained positive throughout the tumor mass. Furthermore, the number of mitotic cancer cells increases significantly in LSL-Kras;Cdkn2ab^−/− tumors as compared with Cdkn2a^−/− and Cdkn2ab^+/+ lung tumors (Fig. 2A and B). We conclude that the loss of Cdkn2ab promotes cellular proliferation and progression of KRAS-driven lung tumors.

To further characterize the lung tumors that develop in LSL-Kras;Cdkn2ab null mice, we evaluated the lung epithelial marker NK2.1 homeobox 1 (NKX2-1) and chromosomal high motility group protein HMGA2, which define the status of differentiation of lung adenocarcinoma driven by mutant KRAS in the mouse (31, 32). NKX2-1 is a master regulator of lung differentiation, which is expressed in the majority of human lung adenocarcinomas (33). Human lung tumors that are negative for NKX2-1 are poorly differentiated, display high-grade histologic features, and an aggressive behavior (33–35). NKX2-1 restrains lung cancer progression by enforcing a lung epithelial differentiation program by repressing the chromatin regulator HMGA2 (31, 32).

We found that lung tumors of LSL-Kras;Cdkn2ab^+/+ and LSL-Kras;Cdkn2a^−/− mice are predominately NKX2-1–positive and HMGA2-negative (Fig. 2C and D and Supplementary Fig. S2A). In contrast, about 21% of LSL-Kras;Cdkn2ab^−/− tumors lose NKX2-1 completely (Supplementary Fig. S2A). In addition, about 73% of LSL-Kras;Cdkn2ab^−/− tumors increase their number of HMGA2-positive cells (Fig. 2C–E and Supplementary Fig. S2B). Finally, we also determined that the metastases that developed in Cdkn2a^−/− and in Cdkn2ab^−/− lung tumors stain positive for HMGA2 (Supplementary Fig. S2C). This observation strongly suggests that HMGA2-positive lung tumors have metastatic potential.

These data indicate that complete loss of the Cdkn2ab locus significantly promotes lung cancer tumor progression in the mouse. These observations imply that p15^INK4B provides an important tumor suppressor function in opposing mutant KRAS tumorigenesis.

To characterize in better detail the biologic consequence of Cdkn2ab deficiency in oncogenic Kras lung adenocarcinoma, we generated mice that allow the tagging and isolation of lung cancer cells. We took advantage of CCSP-rtTA;Tet-op-Kras mice, which carry a transgene encoding Kras^G12D under the control of the tetracycline operator (Tet-op-Kras), and a transgene expressing the reverse tetracycline transactivator in the respiratory epithelium under the control of the Clara cell secretory protein promoter (CCSP-rtTA). Tet-op-Kras;CCSP-rtTA mice develop lung cancer with 100% penetrance when continuously exposed to doxycycline (6).

To visualize mutant KRAS–expressing lung epithelial cells, we crossed the Tet-op-Kras;CCSP-rtTA strain with mice that carry a transgene encoding histone H2B fused to GFP (H2B-GFP) under the control of the tetracycline operator (tetO-HIST1H2BJ-GFP; ref. 23). We determined that upon exposure to food pellets impregnated with doxycycline, Tet-op-Kras,CCSP-rtTA;tetO-HIST1H2BJ-GFP (Kras-GFP mice) transgenic mice express H2B-GFP in lung epithelial cells (Supplementary Fig. S3A–S3C) with the characteristics of type II pneumocytes (Fig. 3A).

When exposed continuously to doxycycline, Kras-GFP mice develop mutant KRAS lung tumors that are also positive for H2B-GFP (Fig. 3B). Lung tumor cells showed predominantly nuclear NKX2-1 staining and cytoplasmic SP-C staining (Fig. 3C). Tumor cells were predominantly negative for CCSP but we detected faint CCSP staining in a portion of tumor cells as compared with the strong staining in the bronchiolar epithelial lining (Fig. 3C). We conclude that this mouse model allows successful H2B-GFP tagging of lung cancer cells.

We determined the effect of Cdkn2a or Cdkn2ab deficiency on lung tumorigenesis in Kras-GFP mice. Significantly, Kras-GFP;Cdkn2ab^−/− mice developed an increased number of tumors as well as an increased tumor burden 4.5 months after induction of mutant KRAS (Fig. 4A–C). In this mouse model, loss of Cdkn2a resulted in an intermediate number of tumors and tumor burden at 4.5 months (Supplementary Fig. S4A and S4B). We also observed an increase in tumor burden in Kras-GFP;Cdkn2ab^−/− mice at time of death (Supplementary Fig. S4C). These results indicate that Cdkn2ab loss promotes both mutant KRAS tumor initiation and progression.

Tumors arose within the lung parenchyma and alveoli in all three genotypes (Fig. 4D); however, in this model only Cdkn2ab null mice generated tumors that arose from the bronchial epithelia (Fig. 4D, panels 4 and 9 and Supplementary Fig. S4D). This observation suggests that Cdkn2ab loss reduces the threshold to develop oncogenic KRAS-induced tumors in distal bronchioles. In addition, Cdkn2a or Cdkn2ab null mice developed lung tumors characterized by nuclear atypia and papillary structures (Fig. 4D, panels 2, 3, 7 and, 8; Supplementary Fig. S4D) that closely resembles aggressive human NSCLC (36, 37). Others and we described similar morphologic changes in lung adenocarcinomas of Cdkn2a null mice induced by the doxycycline-regulated mutant Kras transgene (6, 9, 38).

Lung tumors of Cdkn2ab-deficient mice show an increase in the number of cells stained positive for targets of activated KRAS, such as p-S6 ribosomal protein and p-ERK1/2 (Supplementary Fig. S4E). These findings suggest that Cdkn2ab deficiency leads to increased proliferative activity. These features reproduced the characteristics of lung tumors developed in LSL-Kras;Cdkn2ab^−/− mice.

Kras-GFP;Cdkn2ab^+/+ mice had the longest survival, Kras-GFP;Cdkn2a^−/− mice had an intermediate survival and Kras-GFP;Cdkn2ab^−/− mice had the shortest survival among the three groups (Supplementary Fig. S4G). These differences, however, were due mainly to sarcomas and lymphomas, which preclude the assessment of lung cancer–specific mortality (Supplementary Fig. S4H and Supplementary Table S2).

To evaluate the properties of lung epithelial cells marked by H2B-GFP, we obtained single-cell suspensions from the lungs of Kras-GFP and CCSP-rtTA;tetO-HIST1H2BJ-GFP control transgenic mice (control-GFP mice). We found that one of 85 H2B-GFP–positive oncogenic KRAS-expressing lung epithelial cells formed colonies, whereas only 1 of 150 GFP-positive, but wild-type Kras lung epithelial cells were able to do so (Supplementary Fig. S5A and S5B). We did not detect any difference in the number of lung epithelial cells present in colonies of either control or mutant KRAS–expressing GFP-positive cells (Supplementary Fig. S5C). This finding suggests that mutant KRAS promotes clonogenicity, but not cell proliferation under these conditions.

We were able to passage these cells up to 5 times onto feeder cultures, suggesting that they can be propagated in vitro and have short-term self-renewal abilities. Of note, similar colony formation ratios were reported for bronchoalveolar stem cells, putative epithelial stem cells present at the bronchoalveolar junction (39).

We were unable to propagate H2B-GFP–positive cells obtained from the respiratory epithelium beyond 5 passages due to loss of viability. In contrast, H2B-GFP–positive primary lung epithelial cells obtained from Kras-GFP;Cdkn2a^−/− or Kras-GFP;Cdkn2ab^−/− mice readily give rise to cell lines when grown as monolayer cultures in conventional tissue cultures plates and retain a subpopulation of H2B-GFP–positive cells for up to 20 passages (Supplementary Fig. S5D and S5E). Moreover, H2B-GFP–positive cells are positive for lung epithelial markers and the cancer stem cell marker ALDH1A1 (Supplementary Fig. S5F). As expected, only H2B-GFP–positive cells express mutant Kras (Supplementary Fig. S5G). Moreover, mutant Kras expression is doxycycline dependent and cannot be reinduced (Supplementary Fig. S5H). Taken together, these data suggest that deficiency of the genes of the Cdkn2ab locus promote immortalization of the respiratory epithelial cells.

To determine the frequency of the loss of genes of the CDKN2AB locus, we determined the mutation status of p15^INK4B, p16^INK4A and p14^ARF in a panel of 23 human NSCLC lines which express mutant KRAS. For this purpose, we used direct sequencing to detect point mutations and SNP analysis the detect gene copy number alteration (Table 1). We established that 6 of the cell lines carried homozygous deletions of CDKN2AB and 15 cell lines carried at least heterozygous deletion of CDKN2AB. Only two cell lines (HCC44 and H1155) did not lose a single copy of the CDKN2AB allele. We validated nullyzygosity of the CDKN2AB locus in A549 and H460 by qRT-PCR (Fig. 5A). These findings indicate that loss-of-function of all three tumor suppressors of CDKN2AB is a common event in human lung adenocarcinoma.

We next analyzed patient data from The Cancer Genome Atlas (TCGA) database. Data with mutation, copy number, and methylation was available for 323 lung adenocarcinoma patients and showed that 30% (n = 96) had oncogenic KRAS mutations. In addition, 51 of the 96 patients (53%) with KRAS mutations also had two-hit CDKN2A inactivation by mutation, chromosomal deletion, and/or promoter methylation. Of these patients, 17 (18%) also showed inactivation in the CDKN2B locus by homozygous deletions. None of the patients in this database showed inactivation of CDKN2B but not of CDKN2A. These data indicate that the loss of the CDKN2AB locus is a common occurrence in human NSCLC (Fig. 5B). Notably, deficiency of CDKN2A is associated with higher pathologic stage (Supplementary Table S3), poorer overall survival (Fig. 5C), and progression-free survival (Fig. 5D) in patients with mutant KRAS lung adenocarcinoma. While there is a tendency to poorer progression-free survival with inactivation of CDKN2B in addition to CDKN2A (P = 0.19), this difference was not significant. Anyway, inactivation of either CDKN2A or CDKN2AB is associated with poorer progression-free survival and overall survival (Supplementary Fig. S6A and S6B) independently of the KRAS mutation status.

To determine the functional effect of p15^INK4B expression in vitro, we ectopically expressed 15^INK4B in mutant KRAS human lung cancer cell lines that also carry homozygous deletions of the CDKN2AB locus. We determined that reintroduction of 15^INK4B leads to an increase in the percentage of A549 and H460 cells in the G₁ phase of the cell cycle (Fig. 6A). Furthermore, we noticed a significant increase of the percentage of subdiploid cells in A549 cells, which indicate the presence of apoptosis. Western blot analysis confirmed that expression of p15^INK4B downregulates the phosphorylation of mitotic marker p-H3 (Ser 10; Fig. 6B). These results suggest that the expression of p15^INK4B promotes a tumor suppressor program that counteracts the proliferative effects of mutant KRAS in NSCLC cells.

Recent technical advancements have allowed a systematic evaluation of statistically recurrent somatic mutations in NSCLC (2, 3). This data set holds the promise to better understand lung tumorigenesis and to identify novel treatments for patients with lung cancer.

In this study, we show that genetic inactivation of the CDKN2A and CDKN2B loci is a common event in mutant KRAS lung adenocarcinoma. By analyzing the TCGA database, we determined that inactivation of CDKN2A by either somatic mutation, chromosomal deletion, and/or promoter hypermethylation, and inactivation of CDKN2B by homozygous deletion occur in about 50% and 20% of human lung adenocarcinomas, respectively. Previous reports did not have sufficient power to determine whether inactivation of the genes of CDKN2AB occurs with significant frequency in mutant KRAS lung cancer (2, 40). By analyzing the large and comprehensive TCGA sample cohort, it is apparent that inactivation of CDKN2AB occurs at a significantly higher frequency than previously noted. Furthermore, we noticed that loss of CDKN2B rarely, if ever, occurs in the absence of CDKN2A somatic inactivation. Our analysis of the TCGA data set also indicates that loss of CDKN2A is associated with higher stage, higher incidence of metastasis, and consequently, decreased overall survival in patients with mutant KRAS lung cancer. These findings have significant clinical implications since they suggest that mutant KRAS lung cancers with two-hit inactivation of CDKN2A identify a subset of patients with high-risk disease. In this regard, it is of interest that these patients could benefit from treatment with focal adhesion kinase inhibitors (9).

We recently reported that expression of oncogenic KRAS in the respiratory epithelium significantly reduced the longevity of p16^Ink4a/p19^Arf null mice (9). We observed a further increase in lung cancer–mediated mortality in mutant Kras mice conditionally deficient in the entire Cdkn2ab locus. This observation suggests a model whereby p15^INK4B loss confers a selective advantage to tumor cells that have lost p16^INK4A and p14^ARF. However, we did not detect a worse clinical outcome in patients with mutant KRAS lung cancers with loss of CDKN2AB as compared with patients with lung cancers mutant for KRAS and CDKN2A. We reason that this observation could be due to several variables, including insufficient sample size of the TCGA data set, or the presence of concurrent mutations that affect the phenotype of human cancer that are not present in genetically engineered mice.

We determined histologically that Cdkn2a deficiency leads to a higher percentage of high-grade lung tumors in Tet-o-Kras mice than in LSL-Kras mice, while the lung tumor phenotype of mutant Kras;Cdkn2ab null mice is consistent between the two mouse models. It is likely that these differences are either due to the strength of the promoters used to express mutant Kras or to the fact that we exposed mice to doxycycline at 4 weeks of age to attempt to obtain lung tumors before the emergence of tumors in other sites. We reason that LSL-Kras mice, owing to the fact that mutant Kras is expressed from the endogenous promoter, model more closely mutant KRAS tumorigenesis.

Our in vivo cell-tracking experiments are consistent with a recent report indicating that a subset of alveolar type II cell coexpressing CCSP and SP-C give rise to lung adenocarcinomas upon mutant KRAS expression (41). The observation that loss of Cdkn2ab promotes the clonogenic ability of lung epithelial cells and their ability to readily give rise to cell lines supports the notion that Cdkn2ab regulates cell proliferation. This assertion is further supported by our observation that reintroduction of p15^INK4B induces antiproliferative effects in human NSCLC lines. We also noticed that once mutant Kras was extinguished in Cdkn2ab null lung cancer cell lines, it could no longer be induced. This observation leads to the speculation that mutant KRAS facilitates the maintenance of a cell state that is permissive to the activity of the CC10 promoter driving the expression of rtTA or the accessibility of the Tet-op element driving mutant KRAS.

Overall, our data provide new genetic evidence that the loss of the three genes residing in CDKN2AB promote mutant KRAS lung tumorigenesis by fostering cellular proliferation, cancer cell differentiation, and metastatic behavior. We propose that mutant Kras;Cdkn2ab null mice provide a platform to accurately model aggressive mutant Kras lung adenocarcinoma.

No potential conflicts of interest were disclosed.

Conception and design: K. Schuster, N. Venkateswaran, P.P. Scaglioni

Development of methodology: S. Peña-Llopis

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): K. Schuster, N. Venkateswaran, S. Peña-Llopis

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): K. Schuster, N. Venkateswaran, L. Girard, S. Peña-Llopis, P.P. Scaglioni

Writing, review, and/or revision of the manuscript: K. Schuster, N. Venkateswaran, A. Rabellino, L. Girard, S. Peña-Llopis, P.P. Scaglioni

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): K. Schuster, N. Venkateswaran, A. Rabellino, S. Peña-Llopis

Study supervision: P.P. Scaglioni

The authors thank Dr. Diego Castrillon of the Department of Pathology, UTSW for providing the Vasa-Cre transgenic mouse, Dr. Anton Berns (Netherlands Cancer Institute) for the Ink4ab^−/− conditional mouse, Angie Mobley (UTSW Flow Cytometry Core), Rachel Ramirez (Scaglioni's Lab), and John Shelton (Histology Core UTSW) for technical assistance. The authors also thank Drs. John Minna, Adi Gazdar, John Heymach, and Ignacio Wistuba for providing NSCLC lines and access to their expression and genomic profiles (supported by NCI SPORE P50CA70907).

This study was supported by American Cancer Society Scholar Award 13-068-01-TBG, CDMRP LCRP Grant LC110229 NCI R01CA137195, UT Southwestern Friends of the Comprehensive Cancer Center, the Gibson Foundation, (to P.P. Scaglioni). NRSA Award F32CA154237 from the NCI (to K. Schuster). CPRIT-NGS RP 110708, NCI SPORE P50CA70907 (to L. Girard) and the Harold C. Simmons Cancer Center through NCI Cancer Center support grant 1P30 CA 142543-01.

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.