Lung cancer is the leading cause of cancer death worldwide, with lung adenocarcinoma being the most common subtype. Many oncogenes and tumor suppressor genes are altered in this cancer type, and the discovery of oncogene mutations has led to the development of targeted therapies that have improved clinical outcomes. However, a large fraction of lung adenocarcinomas lacks mutations in known oncogenes, and the genesis and treatment of these oncogene-negative tumors remain enigmatic. Here, we perform iterative in vivo functional screens using quantitative autochthonous mouse model systems to uncover the genetic and biochemical changes that enable efficient lung tumor initiation in the absence of oncogene alterations. Generation of hundreds of diverse combinations of tumor suppressor alterations demonstrates that inactivation of suppressors of the RAS and PI3K pathways drives the development of oncogene-negative lung adenocarcinoma. Human genomic data and histology identified RAS/MAPK and PI3K pathway activation as a common feature of an event in oncogene-negative human lung adenocarcinomas. These Onc-negativeRAS/PI3K tumors and related cell lines are vulnerable to pharmacologic inhibition of these signaling axes. These results transform our understanding of this prevalent yet understudied subtype of lung adenocarcinoma.

Significance:

To address the large fraction of lung adenocarcinomas lacking mutations in proto-oncogenes for which targeted therapies are unavailable, this work uncovers driver pathways of oncogene-negative lung adenocarcinomas and demonstrates their therapeutic vulnerabilities.

Lung cancer is the leading cause of cancer death (1). Lung adenocarcinoma, the most prevalent subtype of lung cancer, has frequent alterations in receptor tyrosine kinase and RAS/RAF pathway oncogenes, including mutations in EGFR and KRAS (2). The identification of driver oncogenes has enabled a shift from toxic chemotherapies to less toxic and more effective therapies that often target the oncogenes (3). However, approximately 30% of lung adenocarcinomas are thought to lack a driving oncogene (4–6). Consequently, developing targeted therapies for these tumors remains a major unmet challenge for precision thoracic oncology.

Extensive genomic and transcriptomic studies suggest that neither technical reasons nor the presence of novel oncogenes likely explains this large and clinically significant population of patients with lung cancer (1, 2, 4–10). Thus, despite the diagnosis of more than 150,000 patients per year with oncogene-negative lung adenocarcinomas worldwide, the genetic events and biochemical pathway changes that drive the initiation and growth of these tumors remain almost entirely unknown.

Oncogenes and tumor suppressor genes are parts of signaling networks that generate and sustain the biochemical changes that drive tumor initiation and growth (11–13). Combinatorial alterations in tumor suppressor genes could cooperate to activate pathways driving oncogene-negative lung tumors. However, human lung adenocarcinomas have complex patterns of mutations across many putative tumor suppressor genes (4), and predicting which combinations of genomic alterations drive cancer in the absence of oncogene activation based on human genomic data alone remains challenging. Although human genomic data can predict combinations of genomic mutations as likely cancer drivers when the mutations co-occur at very high frequencies (14–17), identifying pathogenic combinations of less frequently mutated genes poses a nearly insurmountable statistical challenge. The large numbers of mutations in lung cancers, nongenomic mechanisms that often inactivate tumor suppressor genes, and generation of similar biochemical effects through inactivation of different genes further reduce the ability of human cancer genomic studies to identify combinatorial alterations that activate driver pathways in lung cancer (18–21).

Functional genomic studies within autochthonous cancer models can help identify the pathways involved in tumorigenesis in vivo (22). Here, we leveraged quantitative mouse model systems to assess the ability of hundreds of combinatorial alterations of tumor suppressor genes, acting across many different signaling pathways, to generate oncogene-negative lung adenocarcinomas in vivo. We uncover pathway-level changes that drive lung cancer in the absence of oncogene mutations, translate these findings to human oncogene-negative lung adenocarcinoma, and leverage these results to identify therapeutic vulnerabilities.

Analysis of human lung adenocarcinoma datasets

Somatic mutation data (SNPs and indels, including silent mutations), The Cancer Genome lung adenocarcinoma (TCGA-LUAD) clinical and exposure data, and GISTIC2 thresholded copy-number variation for 513 TCGA LUAD tumors were downloaded from https://tcga.xenahubs.net/download/mc3/LUAD_mc3.txt.gz, https://portal.gdc.cancer.gov/projects/TCGA-LUAD, https://tcga.xenahubs.net/download/TCGA.LUAD.sampleMap/LUAD_clinicalMatrix, https://tcga.xenahubs.net/download/TCGA.LUAD.sampleMap/Gistic2_CopyNumber_Gistic2_all_thresholded.by_genes.gz.

Amplifications were defined as “2” and deletions as “−2”. Genes with conflicting CNV values within a single tumor were ignored. Fusion data were obtained from ref. 23. Fusion and CNV data were filtered to include only data from the 513 samples within the somatic mutation set. Duplicate fusions were collapsed into single fusions. MET-exon skipping data were taken from (24). Curated survival data from (25) were downloaded from https://tcga.xenahubs.net/download/survival/LUAD_survival.txt.gz.

Somatic mutations, copy-number alteration (CNA) data, fusion data, panel information (genomic_information.txt), and clinical data (both sample- and patient-level) from AACR Project GENIE v8 were downloaded from https://www.synapse.org/#!Synapse:syn22228642 (25). Data were filtered to only include LUAD tumors. A single tumor was kept for patients with multiple different tumor samples, with priority for earlier sequenced samples and those from primary tumors. If tumor samples appeared identical within the clinical metadata, the related patient data were excluded. Criteria for determining the fraction of lung adenocarcinomas without known oncogenic drivers, classification of mutations and tumors, and gene and pathway alteration co-occurrences are described in Supplementary Materials and Methods.

Animal studies

The use of mice for this study has been approved by Institutional Animal Care and Use Committee at Stanford University, protocol number 26696. KrasLSL-G12D/+ [Jax no. 008179 (K)], R26LSL-tdTomato(ai9) [Jax no. 007909 (T)], and H11LSL-Cas9 [Jax no. 026816 (C)], Keap1flox, Pten flox (Jax no. 006440), Lkb1 flox (Jax no. 014143), Nf1 flox (Jax no. 017640), and Trp53flox (Jax no. 008462) mice have been described previously (26 33). All mice were on a C57BL/6:129 mixed background except the mice used for generation of oncogene-negative cell lines and some of the Trp53flox/flox;TC mice used for metastasis analysis, which were on a pure C57BL/6 background.

Tumor initiation

Tumors were initiated by intratracheal delivery of pooled or individual Lenti-sgRNA/Cre vectors. Tumors were initiated with the indicated titers and allowed to develop tumors for 3 to 12 months after viral delivery, as indicated in each figure (see Supplementary Materials and Methods).

Tumor barcode sequencing and analysis

For DNA extraction from single dissected tumors to generate libraries for Tuba-seq, targeted sequencing of selected oncogenes, and whole-exome sequencing, we used Qiagen AllPrep DNA/RNA Micro Kit. For Tuba-seq on bulk lungs, genomic DNA was isolated from bulk tumor-bearing lung tissue from each mouse as described previously (34). Q5 High-Fidelity 2x Master Mix (New England Biolabs, M0494X) was used to amplify the sgID-BC region from 50 ng of DNA from dissected tumors or 32 μg of bulk lung genomic DNA. Unique dual-indexed primers were used for each sample (35). The PCR products were purified with Agencourt AMPure XP beads (Beckman Coulter, A63881) using a double size selection protocol. The libraries were pooled based on lung weights to ensure even reading depth, and sequenced (read length 2 × 150bp) on the Illumina HiSeq 2500 or NextSeq 500 platform (Admera Health Biopharma Services). Tuba-seq analysis of tumor barcode reads was performed as described previously (Supplementary Materials and Methods; refs. 34, 36). We used several metrics of tumor number, burden, and size (see Supplementary Fig. S4 in ref. 35; Supplementary Materials and Methods).

Multiple transductions

A fraction of lung tumors initiated with Lenti-sgRNA/Cre vectors contained multiple barcoded Lenti-sgRNA/Cre vectors. If multiple barcodes (sgID-BC) have unexpectedly similar read counts, we suspect transduction of the initial cell with multiple Lenti-sgRNA/Cre vectors. To capitalize on these multiple transductions as a way to find higher-order interactions between tumor suppressor genes, we identified the combinations of sgRNA that appear to cooperate as potent drivers of tumor growth. We developed methods to identify tumors with likely multiple transductions (i.e., those tumors with complex genotypes with multiple tumor suppressor genes inactivated). For each sgID-BC, we consider sgID-BCs from the same sample with read counts within 10% as possible multiple transduction events. Multiple transductions that lead to synergistic combinatorial tumor suppressor alterations would confer a growth advantage. Thus, synergistic combinatorial alterations of tumor suppressor genes would be expected to be overrepresented among the largest tumors.

To have a dataset with a higher signal-to-noise ratio, we analyzed the largest tumors that were co-infected with up to six Lenti-sgRNA/Cre vectors. With this method, for each tumor, we assembled a list of genes that were possibly co-mutated. We then ranked all possible combinations of genes by their frequency in the largest tumors. An inherent problem with this analysis is that the genotypes that increase tumor growth will be overrepresented among the largest tumors even without multiple transductions and specific synergistic interactions. To account for the different number of tumors with different sgIDs, we performed a permutation test, where we control for the number of tumors of each genotype but randomize the sizes of tumors by randomly matching the genotypes with tumor sizes (10,000 repetitions). Synergistic tumor suppressor combinations would occur at significantly higher than expected frequencies based on this permutation test. Reassuringly, although our analysis resulted in significant enrichment of complex genotypes based on the permutation test, a control analysis performed on smaller tumors within the same mice with high noise-to-signal ratio resulted in a loss of statistical significance. Thus, our permutation test controls for the bias of different frequency of sgIDs among the tumors.

Histology and IHC

Lung lobes were inflated with 4% formalin and fixed for 24 hours, stored in 70% ethanol, paraffin-embedded, and 4 µm thick sections were used for hematoxylin and eosin staining and IHC. Primary antibodies were anti-RFP (Rockland, 600–401–379), anti-TTF1(Abcam, ab76013), anti-UCHL1(Sigma, HPA005993), anti-TP63 (Cell Signaling Technology, 13109), anti-phospho-S6 (Cell Signaling Technology, 4858), anti-PTEN (Cell Signaling Technology, 9559), anti-phospho-ERK (Cell Signaling Technology, 4370), anti-phospho-AKT (Thermo Fisher Scientific, 44–621G), and anti-HMGA2 (Biocheck, 59170AP). IHC was performed using Avidin/Biotin Blocking Kit (Vector Laboratories, SP-2001), Avidin-Biotin Complex Kit (Vector Laboratories, PK-4001), and DAB Peroxidase Substrate Kit (Vector Laboratories, SK-4100) following standard protocols.

To quantify the positivity of phospho-ERK and phospho-AKT stained slides, H scores were calculated using Qupath. The H score is determined by adding the results of multiplication of the percentage of cells with staining intensity ordinal value (scored from 0 for “no signal” to 3 for “strong signal”) with possible values ranging from 0 to 300 (37). To normalize potential variations between different rounds of IHC, one patient sample was included and stained for both pERK and pAKT in all rounds of staining as a control.

Cell lines

Mouse oncogene-negative cell lines were generated from tumors initiated in Trp53flox/flox;TC BL6 mice 4 months after transduction with Lenti-sgNf1-sgRasa1-sgPten/Cre. After dissociation of tumors, cells were cultured in DMEM supplemented with 10% FBS, 1% penicillin/streptomycin (Gibco), and 0.1% Amphotericin (Life Technologies). HC494 and MW389T2 (KrasG12D and Trp53 mutant) lung adenocarcinoma cells were generated previously. Human oncogene-negative cell lines (NCI-H1838, NCI-H1623) and oncogene-positive cell lines (A549, H2009, NCI-H2009, SW1573, HOP62, NCI-H358, NCI-H1792) were purchased from ATCC and cultured in RPMI supplemented with 5% FBS, 1% penicillin/streptomycin (Gibco), and 0.1% amphotericin (Life Technologies). We performed Mycoplasma testing using MycoAlert Mycoplasma Detection Kit (Lonza). Cells were maintained at 37°C in a humidified incubator at 5% CO2. NCI-H1838 and NCI-H1623 cell lines do not have any genomic mutation in components of PI3K pathway. Mutations of these two cell lines in components of RAS pathway are indicated below (extracted from DepMap):

  • NCI-H1838 mutations in RAS pathway: NF1(p.N184fs) and IQGAP2 (p.P780L)

  • NCI-H1623 mutations in RAS pathway: RASA1 (p.A47fs), FGFR2 (p.A355S), and ERF (p.G255C)

Clonogenic, apoptosis, and proliferation assays

For clonogenic assays, mouse cells were seeded in triplicate into 24-well plates (4,000 cells per well) and allowed to adhere overnight in regular growth media. Cells were then cultured in the absence or presence of the drug as indicated on each figure panel in complete media for 4 days. Growth media with or without drugs was replaced every 2 days. The remaining cells were stained with 0.5% crystal violet in 20% methanol and photographed using a digital scanner. Relative growth was quantified by densitometry after extracting crystal violet from the stained cells using 100% methanol (38).

Clonogenic assay of human oncogene-negative lung adenocarcinoma cell lines was done in spheroids (39). 400 to 5,000 cells/well were seeded in round bottom ultra-low attachment 96-well plates (Corning) in growth media and incubated for 72 hours at 37°C in 5% CO2. Spheroid formation was confirmed visually, and spheroids were treated in triplicate with dilutions of RMC-4550 and capivasertib in complete growth media. Following drug exposure for 5 days, cell viability in spheroids was determined using the CellTiter-Glo 3D Assay Kit (Promega), following the manufacturer's instructions. Data were normalized to DMSO values.

Drug synergy was analyzed using SynergyFinder (https://synergyfinder.fimm.fi) web application (40). The degree of combination synergy, or antagonism, was quantified by comparing the observed drug combination response against the expected response, calculated using Loewe's model that assumes no interaction between drugs (41).

For apoptosis and proliferation assays, 3 × 105 cells were seeded into 6-well plates, allowed to adhere overnight in regular growth media, and cultured in the presence or absence of 10 µmol/L of capivasertib, RMC-4550, or a combination of both drugs. After 24 hours, apoptosis and cell proliferation were determined through staining with Fixable Viability Dye eFluor 450 (Thermo Fisher Scientific, 65–0863–14), cleaved caspase-3 antibody (Cell Signaling Technology, 9669), and Click-iT EdU Alexa Fluor 647 Flow Cytometry Assay Kit (Thermo Fisher Scientific, C-10424) according to the manufacturers' instructions. Data were acquired using a BD LSR II Flow Cytometer. All experiments were performed independently two times on three different cell lines.

In vivo drug response studies

For drug efficacy studies in autochthonous mouse models, TC mice (8–12 weeks old) were divided into four groups 3.5 months after tumor initiation. They received the vehicle, capivasertib (100 mg/kg; MedChemExpress), RMC-4550 (30 mg/kg; MedChemExpress), or a combination of both dissolved in 10% DMSO, 40% PEG, 5% Tween 80, and 45% PBS through oral gavage. Mice were treated daily for 8 days, and the treatment was stopped for 2 days for recovery, and continued for 2 more days before the tissue harvest. The last two doses of combination therapy were half of the initial doses.

Cell line-derived allografts were generated through subcutaneous injection of 300,000 of MY-C3 (Nf1, Rasa1, Pten, and Trp53 mutant) oncogene-negative mouse cell line in 200 μL of PBS into male (6- to 8-week-old) BL6 mice (two tumors per mouse). Once tumors reached an average size of ∼100 mm3, they were administered RMC-4550 (30 mg/kg; MedChemExpress) and capivasertib (100 mg/kg, MedChemExpress) was administered (5 days on, 2 days off) for 17 days.

Tumor dissociation, cell sorting, and RNA sequencing

Primary tumors were dissociated using collagenase IV, dispase, and trypsin at 37°C for 30 minutes. After dissociation, the samples remained continually on ice, were in contact with ice-cold solutions, and were in the presence of 2 mmol/L EDTA and 1 U/ml DNase to prevent aggregation. Cells were stained with antibodies to CD45 (BioLegend, 103112), CD31 (BioLegend, 303116), F4/80 (BioLegend, 123116), and Ter119 (BioLegend, 116212) to exclude hematopoietic and endothelial cells [lineage-positive (Lin+) cells]. DAPI was used to exclude dead cells. FACS Aria sorters (BD Biosciences) were used for cell sorting.

RNA was purified using RNA/DNA All Prep Kit (Qiagen, 80284). RNA quality of each tumor sample was assessed using the RNA6000 PicoAssay for the Agilent 2100 Bioanalyzer as per the manufacturer's recommendation. 4.4 ng total RNA per sample was used for cDNA synthesis and library preparation using Trio RNA-Seq, Mouse rRNA Kit (Tecan, 0507–32), according to the manufacturer's instructions. The purified cDNA library products were evaluated using the Agilent bioanalyzer and sequenced on NextSeq High Output 1×75 (Admera Health Biopharma Services).

Analysis of mouse model-derived RNA-seq datasets

Paired-end RNA-seq reads were aligned to the mm10 mouse genome using STAR (v2.6.1d) 2-pass mapping and estimates of transcript abundance were obtained using RSEM (v1.2.30; refs. 42, 43). The differentially expressed genes between different tumor genotypes and treatment groups were called by DESeq2 using transcript abundance estimates via tximport (44, 45). The DESeq2-calculated fold changes were used to generate ranked gene lists for input into GSEA (46).

The upregulated genes with absolute log2 fold change greater than 1 and a FDR less than 0.05 in the comparison of Nf1, Rasa1, and Pten mutant oncogene-negative tumors with KrasG12D-driven tumors (KTC+sgInert and KTC+sgPten) were compiled into a signature reflecting the oncogene-negative adenocarcinoma state. This gene signature was utilized in the analysis of human oncogene-positive and oncogene-negative tumors. Scaled estimates of transcript abundance for TCGA LUAD samples were obtained from the GDC data portal (gdc-portal.nci.nih.gov). Each expression profile was then scored on the basis of the mouse-derived gene signature using single-sample GSEA within the Gene Set Variation Analysis (GSVA) package (47).

Statistical analysis

The statistical analyses were performed using R, Python, and Prism software environments. For all bar plots showing relative tumor size, tumor number, tumor burden and frequency, P values and 95% confidence intervals (represented by whiskers) were calculated using bootstrap resampling (10,000 repetitions, see Supplementary Materials and Methods). For all box plots, center lines represent the median, box limits represent the interquartile range. For all strip plots, lines represent the mean. For all box plots and strip plots, statistical significance was calculated by the Wilcoxon rank sum test. For “frequency in large tumors,” a permutation test was used to calculate P values (see Supplementary Materials and Methods). Two-sided Fisher exact test was used to determine statistical significance of alteration frequencies in tumor suppressor genes and oncogenic pathways. Survival curves were compared with the log-rank test. For H scores, center lines represent the mean, and the P values are calculated using Mann–Whitney test in Prism.

Data availability statement

Tuba-seq barcode sequencing and RNA-seq data have been deposited in NCBI's Gene Expression Omnibus (https://www.ncbi.nlm.nih.gov/geo/; GSE174393). Whole exome sequencing data generated are publicly available in SRA-NCBI (www.ncbi.nlm.nih.gov/sra; PRJNA769722).

A large fraction of human lung adenocarcinomas lack oncogene mutations

To better understand the genomics of lung adenocarcinomas that lack oncogene mutations, we analyzed data from TCGA and AACR Genomics Evidence Neoplasia Information Exchange (GENIE; refs. 48, 49). We classified tumors as oncogene-positive if they had high-confidence oncogenic alterations in previously described proto-oncogenes, oncogene-indeterminate if they had alterations of unknown significance in known proto-oncogenes, and oncogene-negative if they had no alterations in known proto-oncogenes (Materials and Methods). Consistent with previous publications, we found that 17% to 18% of lung adenocarcinomas were oncogene-negative (Fig. 1A; Supplementary Fig. S1A; refs. 50–52). In addition, 15% to 27% of lung adenocarcinomas were oncogene-indeterminate and thus 32% to 45% of lung adenocarcinomas lack known oncogene mutations. Patients with oncogene-negative, oncogene-indeterminate, and oncogene-positive lung adenocarcinomas have broadly similar mutational burden and clinical characteristics (Supplementary Figs. S1B–S1E).

Figure 1.

Combinatorial tumor suppressor inactivation enables lung tumor development in the absence of engineered oncogenes. A, Frequency of oncogene-positive, oncogene-indeterminate, and oncogene-negative human lung adenocarcinomas. Data from TCGA. PM, point mutation; indel, insertion and deletion; amp, amplification; multiple, multiple alterations in the same gene. B, Combined Cre/lox and CRISPR/Cas9-mediated tumor suppressor gene inactivation to generate lung epithelial cells with diverse genotypes. The number of mice in each group is indicated. C, Representative light and fluorescence images of lung lobes from the indicated genotypes of mice 1 year after transduction with the Lenti-sgTS102/Cre pool. Lung lobes are outlined with white dotted lines. Scale bar, 4 mm. D, The number of surface tumors (defined as Tomato-positive expansions greater than 0.5 mm in diameter) quantified by direct counting. Each dot represents a mouse, and the bar is the mean. E, Representative hematoxylin and eosin (H&E)–, TTF1-, and TP63-stained sections of tumors from the indicated genotypes of mice. Scale bar, 100 µm. F, Heatmap showing two measures of tumor suppressor strengths in each genotype detected using Tuba-seq analysis: “Tumors” is the occurrence of tumor suppressor gene targeting vectors in dissected tumors. P < 0.001 (red); P < 0.1 (pink; see Supplementary Fig. S4). “Lung” represents the increase in median sizes of clonal expansions in bulk lung lobe samples. Significant increases (P < 0.05) in sizes of clonal expansions with all sgRNAs (red) and those with only one significant sgRNA (pink; see Supplementary Fig. S5). Gray boxes indicate redundant targeting of genes by both Cre/loxP and CRISPR/Cas9.

Figure 1.

Combinatorial tumor suppressor inactivation enables lung tumor development in the absence of engineered oncogenes. A, Frequency of oncogene-positive, oncogene-indeterminate, and oncogene-negative human lung adenocarcinomas. Data from TCGA. PM, point mutation; indel, insertion and deletion; amp, amplification; multiple, multiple alterations in the same gene. B, Combined Cre/lox and CRISPR/Cas9-mediated tumor suppressor gene inactivation to generate lung epithelial cells with diverse genotypes. The number of mice in each group is indicated. C, Representative light and fluorescence images of lung lobes from the indicated genotypes of mice 1 year after transduction with the Lenti-sgTS102/Cre pool. Lung lobes are outlined with white dotted lines. Scale bar, 4 mm. D, The number of surface tumors (defined as Tomato-positive expansions greater than 0.5 mm in diameter) quantified by direct counting. Each dot represents a mouse, and the bar is the mean. E, Representative hematoxylin and eosin (H&E)–, TTF1-, and TP63-stained sections of tumors from the indicated genotypes of mice. Scale bar, 100 µm. F, Heatmap showing two measures of tumor suppressor strengths in each genotype detected using Tuba-seq analysis: “Tumors” is the occurrence of tumor suppressor gene targeting vectors in dissected tumors. P < 0.001 (red); P < 0.1 (pink; see Supplementary Fig. S4). “Lung” represents the increase in median sizes of clonal expansions in bulk lung lobe samples. Significant increases (P < 0.05) in sizes of clonal expansions with all sgRNAs (red) and those with only one significant sgRNA (pink; see Supplementary Fig. S5). Gray boxes indicate redundant targeting of genes by both Cre/loxP and CRISPR/Cas9.

Close modal

Combinatorial tumor suppressor gene inactivation enables lung tumor development

To determine whether combinatorial tumor suppressor gene inactivation can drive lung tumor initiation in the absence of oncogene activation, we coupled Cre/loxP-based genetically engineered mouse models and somatic CRISPR/Cas9-based genome editing with tumor barcoding and high-throughput barcode sequencing (Tuba-seq; refs. 29, 30, 34–36, 53, 54). We used Cre/loxP to inactivate each of five “core” tumor suppressor genes (Trp53, Lkb1/Stk11, Keap1, Nf1, and Pten). These genes are within diverse pathways and are frequently inactivated in human lung cancers, including oncogene-negative lung adenocarcinomas (Supplementary Figs. S2A and S2B). We used CRISPR/Cas9 to coincidentally inactivate panels of additional tumor suppressor genes in lung epithelial cells in mice with floxed alleles of each of the “core” tumor suppressors, a Cre-reporter allele [R26LSL-Tom(T); ref. 29], and a Cre-regulated Cas9 allele [H11LSL-Cas9(C); ref. 30].

We transduced Nf1f/f;TC, Ptenf/f;TC, Trp53f/f;TC, Lkb1f/f;TC, Keap1f/f;TC, TC, and T mice with two pools of barcoded Lenti-sgRNA/Cre vectors that target ∼50 putative tumor suppressor genes that we previously investigated in KRASG12D-driven lung tumors (Lenti-sgTS15/Cre and Lenti-sgTS102/Cre; Fig. 1B; Supplementary Figs. S2C, S2D and S3A; Supplementary Table S1; refs. 34–36). The mutation frequency of these genes varied, and mutations in some were enriched in oncogene-negative human lung adenocarcinomas (Supplementary Table S1; Supplementary Figs. S2C and S2D). The combination of Cre/LoxP and CRISPR/Cas9-based genome editing should generate hundreds of combinations of genomic alterations in lung epithelial cells. We previously found that a small percent of lung tumors initiated with Lenti-sgRNA/Cre vectors in other lung cancer models contained multiple sgRNAs, consistent with the transduction of the initial cell with multiple Lenti-sgRNA/Cre vectors (34, 36). Thus, we used a high titer of the Lenti-sgRNA/Cre pools in these experiments to increases the likelihood of finding higher-order genetic interactions that drive tumorigenesis.

One year after transduction with the Lenti-sgRNA/Cre pools, Nf1f/f;TC, Ptenf/f;TC, and Trp53f/f;TC mice developed a modest number of tumors (defined as Tomatopositive expansion >0.5 mm in diameter; Fig. 1C and D; Supplementary Figs. S3B and S3C). Interestingly, Nf1f/f;TC, Ptenf/f;TC, and Trp53f/f;TC, and TC mice transduced with the larger Lenti-sgTS102/Cre pool developed many more tumors than those transduced with the Lenti-sgTS15/Cre pool. These tumors were positive for TTF1/NKX2–1, a marker for lung adenocarcinoma, and negative for P63 and UCHL1, markers for squamous cell and small cell lung cancer, respectively (Fig. 1E).

To determine whether these tumors contained spontaneous oncogene mutations, we sequenced 10 genomic regions in Kras, Braf, Nras, and Egfr (Supplementary Fig. S3D, Supplementary Table S2, and Materials and Methods; refs. 28, 53, 55–62). Across 29 samples, we detected only one oncogene mutation (a KrasG12V mutation in a tumor from a Ptenf/f;TC mouse). Thus, the majority of these tumors arose in the absence of hotspot mutations in these proto-oncogenes. These results suggest that the inactivation of combinations of specific tumor suppressor genes in Nf1f/f;TC, Ptenf/f;TC, and Trp53f/f;TC mice drives the development of lung cancer in vivo. Notably, the overall low number of tumors indicates that inactivation of the “core” tumor suppressor genes alone, and most combinations of tumor suppressor genes tested, are insufficient to generate lung tumors.

Identification of top candidate tumor suppressor genes involved in oncogene-negative lung tumor formation

The Lenti-sgRNA/Cre vectors contain two-component barcodes in which an sgID identifies the sgRNA and a random barcode (BC) uniquely tags each clonal tumor. Thus, high throughput sequencing of the sgID-BC region can identify the sgRNA(s) present in each tumor and quantify the number of cancer cells in each tumor (Fig. 1B). To determine which sgRNAs were present in the largest tumors, we PCR-amplified the sgID-BC region from genomic DNA from dissected tumors and performed high-throughput sgID-BC sequencing. Most large tumors contained multiple Lenti-sgRNA/Cre vectors therefore, we calculated the statistical enrichment of each sgRNA based on their relative representation in the dissected tumors (Fig. 1F; Supplementary Fig. S4; see Materials and Methods).

To further quantify the impact of inactivating each tumor suppressor gene on clonal expansion of lung epithelial cells, we performed tumor barcode sequencing (Tuba-seq) on bulk DNA from one lung lobe from each Nf1f/f;TC, Ptenf/f;TC, Trp53f/f;TC, and TC mouse (Fig. 1C). Analysis of the number of cells in clonal expansions further nominated tumor suppressor genes that may contribute to tumor initiation and growth (Fig. 1F; Supplementary Fig. S5). On the basis of these two analyses, we selected 13 genes for further analysis (Fig. 1F). The potential importance of these tumor suppressor genes was often supported by both sgRNAs targeting each gene, consistent with on-target effects (Fig. 1F; Supplementary Figs. S4 and S5).

Inactivation of candidate tumor suppressors efficiently generates lung tumors

To determine whether inactivation of candidate tumor suppressor genes can initiate oncogene-negative tumors, we generated a pool of Lenti-sgRNA/Cre vectors targeting each of these tumor suppressor genes (Lenti-sgTS14/Cre pool; Fig. 2A). We targeted each gene with the sgRNA that had the most significant effect on tumor growth and used five times higher titer of each lentiviral vector per mouse than we used in Lenti-sgTS102/Cre pool, thus increasing the potential for the transduction of the initial cell with multiple Lenti-sgRNA/Cre vectors.

Figure 2.

Nf1, Rasa1, and Pten emerge as key drivers of oncogene-negative lung adenocarcinoma. A, Combined Cre/loxP and CRISPR/Cas9-mediated tumor suppressor gene inactivation to generate lung epithelial cells with diverse genotypes. The number of mice in each group is indicated. B, Representative light and fluorescence images of lung lobes from the indicated genotypes of mice. Lung lobes are outlined with white dotted lines. Scale bar, 4 mm. C, The number of tumors (defined as Tomato-positive expansions greater than 0.5 mm in diameter) quantified by direct counting. Each dot represents a mouse and the bar is the mean. D and E, The number of tumors with a minimum size of 1,000 neoplastic cells, relative to inert sgRNA containing expansions are shown as blue bars. Ninetieth percentile of tumor sizes relative to inert sgRNAs is shown as red bars. sgRNAs that significantly impact tumor number or size (P < 0.05) are in darker colors. Whiskers show 95% confidence intervals. F and G, Barcodes with the highest counts in each mouse were investigated for coinfection with multiple Lenti-sgTS/Cre vectors (i.e., tumors initiated from cells transduced with multiple viruses, which result in complex tumor suppressor genotypes, see Materials and Methods). The most frequently co-mutated pairs of tumor suppressor genes are shown. Combinations of sgRNAs that lead to the generation of Nf1, Rasa1, and Pten mutant cancer cells are in bold. *, P < 0.05; **, P < 0.01; ***, P < 0.001 based on a permutation test. H, Total number of neoplastic cells in clonal expansions with more than 200 cells in the indicated genotypes of mice after receiving Lenti-sgTS14/Cre or Lenti-sgTS11/Cre (which lacks lentiviral vectors containing sgNf1, sgRasa1, and sgPten). The magnitude of reduction in neoplastic cell number is indicated.

Figure 2.

Nf1, Rasa1, and Pten emerge as key drivers of oncogene-negative lung adenocarcinoma. A, Combined Cre/loxP and CRISPR/Cas9-mediated tumor suppressor gene inactivation to generate lung epithelial cells with diverse genotypes. The number of mice in each group is indicated. B, Representative light and fluorescence images of lung lobes from the indicated genotypes of mice. Lung lobes are outlined with white dotted lines. Scale bar, 4 mm. C, The number of tumors (defined as Tomato-positive expansions greater than 0.5 mm in diameter) quantified by direct counting. Each dot represents a mouse and the bar is the mean. D and E, The number of tumors with a minimum size of 1,000 neoplastic cells, relative to inert sgRNA containing expansions are shown as blue bars. Ninetieth percentile of tumor sizes relative to inert sgRNAs is shown as red bars. sgRNAs that significantly impact tumor number or size (P < 0.05) are in darker colors. Whiskers show 95% confidence intervals. F and G, Barcodes with the highest counts in each mouse were investigated for coinfection with multiple Lenti-sgTS/Cre vectors (i.e., tumors initiated from cells transduced with multiple viruses, which result in complex tumor suppressor genotypes, see Materials and Methods). The most frequently co-mutated pairs of tumor suppressor genes are shown. Combinations of sgRNAs that lead to the generation of Nf1, Rasa1, and Pten mutant cancer cells are in bold. *, P < 0.05; **, P < 0.01; ***, P < 0.001 based on a permutation test. H, Total number of neoplastic cells in clonal expansions with more than 200 cells in the indicated genotypes of mice after receiving Lenti-sgTS14/Cre or Lenti-sgTS11/Cre (which lacks lentiviral vectors containing sgNf1, sgRasa1, and sgPten). The magnitude of reduction in neoplastic cell number is indicated.

Close modal

We initiated tumors with Lenti-sgTS14/Cre in Nf1f/f;TC, Ptenf/f;TC, Trp53f/f;TC, TC, and KrasLSL-G12D;T (KT) mice. Less than 4 months after tumor initiation, several Nf1f/f;TC and Ptenf/f;TC mice showed signs of extensive tumor burden. These mice developed many more tumors than mice of the same genotypes 1 year after transduction with the Lenti-sgTS102/Cre (compare Fig. 2B and C with Fig. 1C and D). We performed Tuba-seq on DNA from tumor-bearing lungs to determine the number and size of tumors with each barcoded Lenti-sgRNA/Cre vector. Inactivation of Nf1, Rasa1, and Pten most dramatically increased tumor size and/or tumor number across all mouse genotypes (Fig. 2D and E; Supplementary Figs. S6A and S6B, and Materials and Methods). Inactivation of several other tumor suppressor genes less dramatically but significantly increased tumor size and/or tumor number in a genotype-specific manner, suggesting that additional molecular pathways may also lead to early epithelial expansions.

The largest tumors in Nf1f/f;TC, Ptenf/f;TC, Trp53f/f;TC, and TC mice were frequently generated through the inactivation of multiple tumor suppressor genes. Vectors targeting Nf1, Rasa1, and/or Pten were often present in the largest tumors, and the coincident targeting of Nf1, Rasa1, and Pten was the most frequent combination (Fig. 2F and G; Supplementary Figs. S6C–S6H). To gain greater insight into the contribution of Nf1, Rasa1, and Pten inactivation to the generation of oncogene-negative tumors, we transduced Nf1f/f;TC, Ptenf/f;TC, Trp53f/f;TC, TC, and KT mice with a pool of Lenti-sgRNA/Cre vectors that lacked the vectors targeting Nf1, Rasa1, and Pten (Lenti-sgTS11/Cre; Supplementary Fig. S7A). Approximately 4 months after transduction, these mice had many fewer tumors than mice transduced with Lenti-sgTS14/Cre pool (Supplementary Figs. S7B and S7C). Tuba-seq analysis confirmed a dramatic decrease in tumor burden relative to mice that received the Lenti-sgTS14/Cre pool (Fig. 2H). Thus, the inactivation of Nf1, Rasa1, and Pten emerged as the most important contributors to the generation of oncogene-negative lung tumors.

Extensive experiments generating single and pairwise inactivation of tumor suppressor genes in individual mice led to the development of very few tumors even after long periods of time (Supplementary Figs. S8 and S9). Thus, single and pairwise tumor suppressor gene inactivation is rarely sufficient to generate lung tumors and combinatorial inactivation of three or more tumor suppressor genes increases the efficiency of tumor development and/or accelerates the growth of oncogene-negative lung tumors.

Combinatorial inactivation of Nf1, Rasa1, and Pten drives lung adenocarcinoma development comparably with oncogenic KRAS

To investigate the contribution of Nf1, Rasa1, and Pten, we transduced TC and Trp53f/f;TC mice with a pool of eight lentiviral vectors that would inactivate Nf1, Rasa1, and Pten individually, in pairwise combinations, and all three simultaneously (Lenti-sgTSTriple-pool/Cre, Fig. 3A). Three months after tumor initiation, TC mice had hundreds of large adenomas and adenocarcinomas (Fig. 3BD; Supplementary Figs. S10A–S10E). Tuba-seq analysis showed that most of the tumor burden arose as a consequence of concomitant inactivation of all three tumor suppressors (Fig. 3E; Supplementary Fig. S8). Additional inactivation of Trp53 in Trp53f/f;TC mice did not increase tumor initiation, suggesting that Trp53 is not a major suppressor of oncogene-negative lung adenocarcinoma development at these early stages (Fig. 3BF; Supplementary Figs. S10A–S10F). Finally, to compare the tumor initiation potential of combinatorial Nf1, Rasa1, and Pten inactivation with that of a known oncogene, we transduced KrasLSL-G12D;T mice (which lack Cas9) with Lenti-sgTSTriple-pool/Cre (Fig. 3A). Strikingly, coincident inactivation of Nf1, Rasa1, and Pten in TC and Trp53f/f;TC mice was nearly as potent as oncogenic KRASG12D in driving lung tumor initiation in vivo (Fig. 3F; Supplementary Fig. S10G and Materials and Methods).

Figure 3.

Inactivation of Nf1, Rasa1, and Pten allows the acquisition of growth advantage during lung adenocarcinoma development. A, Barcoded triple sgRNA vectors for CRISPR/Cas9-mediated inactivation of all combinations of Nf1, Rasa1, and Pten in TC and Trp53flox/flox;TC mice. sgNeo1 and sgNeo2 are active cutting, but inert sgRNAs. sgNT is a nontargeting inert sgRNA. Mouse genotype, mouse number, and titers of virus are indicated. Tuba-seq was performed on tumor-bearing lungs 3 months after tumor initiation. B, Brightfield and fluorescence images of lungs from the indicated mouse genotypes. Lung lobes are outlined with a dashed white line. Scale bar, 4 mm. C, The number of surface tumors (defined as Tomato-positive expansions greater than 0.5 mm in diameter) quantified by direct counting. Each dot represents a mouse and the bar is the mean. D, Representative hematoxylin and eosin (H&E)– and Tomato-stained sections of lungs from TC and Trp53flox/flox;TC mice 3 months after transduction with Lenti-sgTSTriple-pool/Cre. Scale bar, 500 µm. E, Numbers of tumors (with >1,000 neoplastic cells) relative to the inert sgRNA-containing expansions. sgRNAs resulting in a significantly higher number of tumors than the inert vector (P < 0.05) are shown in a darker color. Mean ± 95% confidence interval is shown. F, Quantification of the ability of combined Nf1/Rasa1/Pten inactivation in TC mice and oncogenic KrasG12D in KT mice to initiate tumors. The number of tumors (with >1,000 neoplastic cells) per infectious unit (ifu) is shown. The bar is the median, the box represents the interquartile range, and the whiskers show minimum and maximum values. ns, nonsignificant.

Figure 3.

Inactivation of Nf1, Rasa1, and Pten allows the acquisition of growth advantage during lung adenocarcinoma development. A, Barcoded triple sgRNA vectors for CRISPR/Cas9-mediated inactivation of all combinations of Nf1, Rasa1, and Pten in TC and Trp53flox/flox;TC mice. sgNeo1 and sgNeo2 are active cutting, but inert sgRNAs. sgNT is a nontargeting inert sgRNA. Mouse genotype, mouse number, and titers of virus are indicated. Tuba-seq was performed on tumor-bearing lungs 3 months after tumor initiation. B, Brightfield and fluorescence images of lungs from the indicated mouse genotypes. Lung lobes are outlined with a dashed white line. Scale bar, 4 mm. C, The number of surface tumors (defined as Tomato-positive expansions greater than 0.5 mm in diameter) quantified by direct counting. Each dot represents a mouse and the bar is the mean. D, Representative hematoxylin and eosin (H&E)– and Tomato-stained sections of lungs from TC and Trp53flox/flox;TC mice 3 months after transduction with Lenti-sgTSTriple-pool/Cre. Scale bar, 500 µm. E, Numbers of tumors (with >1,000 neoplastic cells) relative to the inert sgRNA-containing expansions. sgRNAs resulting in a significantly higher number of tumors than the inert vector (P < 0.05) are shown in a darker color. Mean ± 95% confidence interval is shown. F, Quantification of the ability of combined Nf1/Rasa1/Pten inactivation in TC mice and oncogenic KrasG12D in KT mice to initiate tumors. The number of tumors (with >1,000 neoplastic cells) per infectious unit (ifu) is shown. The bar is the median, the box represents the interquartile range, and the whiskers show minimum and maximum values. ns, nonsignificant.

Close modal

Finally, we initiated tumors in TC and Trp53f/f;TC mice using only the lentiviral vector that targets all three genes (Lenti-sgNf1-sgRasa1-sgPten/Cre; Supplementary Fig. S11A). After only 3 months, these mice developed very large numbers of lung adenomas and adenocarcinomas (Supplementary Figs. S11B–S11E). We confirmed the inactivation of Nf1, Rasa1, and Pten in these tumors and whole-exome sequencing uncovered no putative oncogene mutations and only a few putative tumor suppressor mutations (Supplementary Fig. S11F; Supplementary Table S3). Interestingly, at later time points after initiation, tumors in Trp53f/f;TC mice progressed to an invasive NKX2–1negative HMGA2positive state and metastasized to other organs (Supplementary Fig. S12; ref. 63).

Oncogene-negative murine lung adenocarcinomas have activated RAS and PI3K pathways

NF1 and RASA1 are negative regulators of RAS, whereas PTEN is a negative regulator of the PI3K–AKT pathway. Therefore, we investigated the impact of inactivating these tumor suppressor genes on RAS and PI3K pathway activation by IHC and RNA-seq on FACS-isolated Tomatopositive cancer cells. We generated autochthonous tumors in which Nf1, Rasa1, and Pten were inactivated (TC mice with Lenti-sgNf1-sgRasa1-sgPten/Cre; Nf1/Rasa1/Pten tumors), KRASG12D was expressed (KT;H11LSL-Cas9 mice with Lenti-sgInert/Cre; Kras tumors), or KRASG12D was expressed and Pten was inactivated (KT;H11LSL-Cas9 mice with Lenti-sgPten/Cre; Kras/Pten tumors; Supplementary Fig. S13A). Nf1/Rasa1/Pten tumors had positive staining for pERK (indicative of RAS pathway activation) and pAKT (indicative of PI3K pathway activation; Fig. 4A). Compared with Kras/Pten tumors, the average pERK staining in Nf1/Rasa1/Pten tumors was less intense and pAKT staining was similar (Fig. 4B and C). Single-sample gene set variation analysis (ssGSVA) on our RNA-seq data confirmed that Nf1/Rasa1/Pten tumors had lower RAS pathway gene signature scores than Kras/Pten tumors (Supplementary Fig. S13B; refs. 64, 65). PI3K–AKT pathway gene signature scores were similar in Nf1/Rasa1/Pten and Kras tumors (Supplementary Fig. S13C). The rare tumors that eventually developed after pairwise inactivation of Nf1, Rasa1, and Pten also had strong activation of RAS and PI3K pathways (Supplementary Fig. S8; Supplementary Fig. S13D). On the basis of these analyses, we propose that these tumors represent a subtype of oncogene-negative lung adenocarcinomas with activated RAS and PI3K pathways (Onc-negativeRAS/PI3K subtype).

Figure 4.

Oncogene-negative mouse and human lung adenocarcinomas have frequent activation of RAS and PI3K pathways. A–C, Representative IHC for pERK and pAKT on tumors with the indicated genotypes and quantification of these stainings. The bar is the mean. ns, nonsignificant; ****, P < 0.0001 using Mann–Whitney U test. Scale bar, 50 μm. D and E, Representative IHC for pAKT and pERK on oncogene-negative human tumors. H scores for the whole section are indicated for each representative image. Scale bar, 200 µm (left) and 50 µm (right). F and G, Quantification of pAKT and pERK staining on 35 oncogene-negative and 18 oncogene-positive human lung adenocarcinomas. Genotypes of oncogene-positive tumors, with the lowest pERK and pAKT staining intensities in red. Significance between groups, Mann–Whitney U test; ns, nonsignificant; ****, P < 0.0001. H, pERK and pAKT H-scores for oncogene-negative human tumors replotted from F and G. Red dotted lines, thresholds for low versus medium pERK and pAKT based on the lowest pERK and pATK staining intensity of oncogene-positive lung adenocarcinomas. Black dotted lines, thresholds for medium versus high pERK and pAKT staining based on the mean pERK and pAKT H-scores in oncogene-positive tumors. The number of tumors in each staining intensity group (low, medium, high) is indicated. I, Alteration frequency of established components of RAS and PI3K pathways (see Supplementary Table S6) and their co-occurrences in TCGA data sets. P value calculated by two-sided Fisher exact test. J, Cumulative distribution function (CDF) plot of the signature scores for human tumors stratified by genes upregulated in mouse oncogene-negative tumors generated by inactivation of Nf1, Rasa1, and Pten (Supplementary Fig. S14A; see Supplementary Table S4). Cohort size and P value calculated by Kolmogorov–Smirnov test are indicated. *, P < 0.05; **, P < 0.01.

Figure 4.

Oncogene-negative mouse and human lung adenocarcinomas have frequent activation of RAS and PI3K pathways. A–C, Representative IHC for pERK and pAKT on tumors with the indicated genotypes and quantification of these stainings. The bar is the mean. ns, nonsignificant; ****, P < 0.0001 using Mann–Whitney U test. Scale bar, 50 μm. D and E, Representative IHC for pAKT and pERK on oncogene-negative human tumors. H scores for the whole section are indicated for each representative image. Scale bar, 200 µm (left) and 50 µm (right). F and G, Quantification of pAKT and pERK staining on 35 oncogene-negative and 18 oncogene-positive human lung adenocarcinomas. Genotypes of oncogene-positive tumors, with the lowest pERK and pAKT staining intensities in red. Significance between groups, Mann–Whitney U test; ns, nonsignificant; ****, P < 0.0001. H, pERK and pAKT H-scores for oncogene-negative human tumors replotted from F and G. Red dotted lines, thresholds for low versus medium pERK and pAKT based on the lowest pERK and pATK staining intensity of oncogene-positive lung adenocarcinomas. Black dotted lines, thresholds for medium versus high pERK and pAKT staining based on the mean pERK and pAKT H-scores in oncogene-positive tumors. The number of tumors in each staining intensity group (low, medium, high) is indicated. I, Alteration frequency of established components of RAS and PI3K pathways (see Supplementary Table S6) and their co-occurrences in TCGA data sets. P value calculated by two-sided Fisher exact test. J, Cumulative distribution function (CDF) plot of the signature scores for human tumors stratified by genes upregulated in mouse oncogene-negative tumors generated by inactivation of Nf1, Rasa1, and Pten (Supplementary Fig. S14A; see Supplementary Table S4). Cohort size and P value calculated by Kolmogorov–Smirnov test are indicated. *, P < 0.05; **, P < 0.01.

Close modal

Oncogene-negative human lung adenocarcinomas frequently have activation of RAS and PI3K pathways

To investigate the activation of RAS and PI3K pathways in human oncogene-negative lung adenocarcinomas, we analyzed oncogene-negative (N = 35) and oncogene-positive (N = 18) lung adenocarcinomas. IHC for pERK and pAKT showed that ∼45% of oncogene-negative human tumors had moderate to strong activation of both RAS and PI3K pathways and thus represent the Onc-negativeRAS/PI3K subtype (Fig. 4DH; Supplementary Figs. S13E–S13J). Activation of the RAS and PI3K pathways were rarely explained by mutations in NF1, PTEN, or other genes profiled by Stanford's Solid Tumor Actionable Mutation Panel (STAMP; Supplementary Tables S5 and S6), likely due to the noncomprehensiveness of this gene panel, as well as epigenetic mechanisms of RAS and PI3K pathway activation (66). Epigenetic silencing and other nongenomic mechanisms have been well documented to inhibit tumor suppressor genes including PTEN (19, 20, 67, 68). Therefore, we performed IHC for PTEN on 20 oncogene-negative lung adenocarcinomas that did not have genomic PTEN mutations. Consistent with previous reports, we observed low PTEN protein levels in 13 of 20 of these tumors (Supplementary Figs. S14A–S14F; ref. 19).

To assess a larger set of oncogene-negative lung adenocarcinomas for alterations that could lead to the activation of RAS and PI3K pathways, we analyzed oncogene-negative tumors in TCGA and GENIE datasets. We queried a set of well-established negative regulators of the RAS and PI3K pathways for alterations in oncogene-negative tumors (Supplementary Table S6). Consistent with previous reports, NF1 and RASA1 alterations were enriched in oncogene-negative tumors; however, coincident genomic alterations in NF1, RASA1, and PTEN were rare (Supplementary Figs. S14G and S14H; refs. 69, 70). Importantly, over 60% of oncogene-negative lung adenocarcinomas in TCGA had alterations in either the RAS or PI3K pathways, and 22% of these tumors had alterations in components of both pathways, likely representing Onc-negativeRAS/PI3K lung adenocarcinomas (Fig. 4I). These frequencies were lower in the GENIE dataset, possibly because only a fraction of the known genes in these pathways were analyzed (Supplementary Fig. S14I). These histologic and genomic analyses support a model in which activation of the RAS and PI3K pathways in Onc-negativeRAS/PI3K tumors can be generated by diverse genomic and/or epigenetic alterations.

Finally, we assessed whether Onc-negativeRAS/PI3K tumors in our mouse model more broadly exhibit transcriptional features that are consistent oncogene-negative human lung adenocarcinoma. We generated a gene expression signature of genes that are higher in Nf1/Rasa1/Pten tumors relative to Kras tumors in mice. We then calculated gene signature activity scores for each TCGA lung adenocarcinoma for this Onc-negativeRAS/PI3K gene expression signature using single-sample GSEA (Supplementary Table S4). Interestingly, the Onc-negativeRAS/PI3K signature was highest in oncogene-negative human lung adenocarcinomas relative to lung adenocarcinomas driven by oncogenic KRAS or other known oncogenes (Fig. 4J). Collectively, these data indicate that the molecular and biochemical state of mouse Onc-negativeRAS/PI3K tumors recapitulates that of a substantial fraction of oncogene-negative human lung adenocarcinomas.

Onc-negativeRAS/PI3K tumors are vulnerable to inhibition of RAS and PI3K–AKT pathways

Understanding the biochemical changes that drive tumor development can nominate potential therapeutic strategies (55). To investigate the therapeutic impact of targeting key nodes in Onc-negativeRAS/PI3K lung cancer, we initiated tumors in TC mice with a pool of barcoded sgRNA viral vectors targeting Nf1, Rasa1, and Pten. We treated these mice with the SHP2 inhibitor RMC-4550 (39), AKT1/2 inhibitor capivasertib (71, 72), or a combination of the two (Fig. 5A; Supplementary Figs. S15A and S15B). These drugs were chosen based on their ongoing clinical development and ability to reduce activation of the RAS and PI3K pathways (39, 72).

Figure 5.

SHP2 and AKT inhibition reduces the growth of autochthonous oncogene-negative lung tumors. A, Barcoded triple sgRNA vectors for CRISPR/Cas9-mediated inactivation of combinations of Nf1, Rasa1, and Pten in TC mice. Indicated numbers of mice were treated with RMC-4550 (SHP2 inhibitor), capivasertib (AKT inhibitor), or a combination of these two drugs for 2 weeks 3.5 months after tumor initiation. Tuba-seq and histologic analysis were performed on tumor-bearing lungs. B, Brightfield and fluorescence images of lungs from the indicated mice. Lung lobes are outlined with a dashed white line. Scale bar, 4 mm. C, Representative hematoxylin and eosin (H&E)– and Tomato-stained sections of tumors from TC mice 3.5 months after transduction with Lenti-sgTripleTS6/Cre and 2 weeks after treatment with the indicated drugs. Scale bar, 100 µm. D, Relative tumor burden after treatment with capivasertib, RMC-4550, and combination of these drugs compared with tumor burden in vehicle-treated mice. ns, nonsignificant; ***, P < 0.001. Drug response is shown for all the tumors.

Figure 5.

SHP2 and AKT inhibition reduces the growth of autochthonous oncogene-negative lung tumors. A, Barcoded triple sgRNA vectors for CRISPR/Cas9-mediated inactivation of combinations of Nf1, Rasa1, and Pten in TC mice. Indicated numbers of mice were treated with RMC-4550 (SHP2 inhibitor), capivasertib (AKT inhibitor), or a combination of these two drugs for 2 weeks 3.5 months after tumor initiation. Tuba-seq and histologic analysis were performed on tumor-bearing lungs. B, Brightfield and fluorescence images of lungs from the indicated mice. Lung lobes are outlined with a dashed white line. Scale bar, 4 mm. C, Representative hematoxylin and eosin (H&E)– and Tomato-stained sections of tumors from TC mice 3.5 months after transduction with Lenti-sgTripleTS6/Cre and 2 weeks after treatment with the indicated drugs. Scale bar, 100 µm. D, Relative tumor burden after treatment with capivasertib, RMC-4550, and combination of these drugs compared with tumor burden in vehicle-treated mice. ns, nonsignificant; ***, P < 0.001. Drug response is shown for all the tumors.

Close modal

Direct fluorescence imaging and histology indicated that SHP2 inhibition and combined SHP2 and AKT1/2 inhibition greatly reduced tumor burden (Fig. 5B and C; Supplementary Fig. S15C). Tuba-seq analysis provided insight into the overall and genotype-specific responses of tumors to the therapeutic interventions. Capivasertib monotherapy was ineffective in vivo whereas RMC-4550 reduced the total tumor burden. The combination of RMC-4550 and capivasertib trended towards being the most efficient therapeutic approach reducing tumor burden by ∼30% compared with RMC-4550 alone (Fig. 5D; Supplementary Figs. S15D–S15G).

We confirmed the inhibition of RAS and PI3K pathways in Onc-negativeRAS/PI3K tumors in mice treated with RMC-4550 and capivasertib (Supplementary Fig. S15H). Furthermore, global gene expression analysis confirmed the downregulation of RAS and PI3K-AKT gene expression signatures after coincident SHP2 and AKT1/2 inhibition (Supplementary Figs. S16A–S16D). Treated tumors tended to have higher expression of an apoptosis gene expression signature and lower expression of a G2–M gene expression signature, suggesting that this combination induces broad cellular changes (Supplementary Figs. S16E and S16F).

Inhibition of SHP2 and AKT synergizes to reduce the growth of Onc-negativeRAS/PI3K lung adenocarcinoma cell lines

To more extensively characterize the responses to SHP2 and AKT inhibition, we generated Nf1/Rasa1/Pten-deficient Onc-negativeRAS/PI3K cell lines from tumors initiated with Lenti-sgNf1-sgRasa1-sgPten/Cre in Trp53flox/flox;TC mice (Fig. 6A; Supplementary Figs. S17A and S17B). As anticipated, RAS and PI3K signaling was reduced in response to treatment with RMC-4550 and capivasertib, respectively (Supplementary Fig. S17C). RMC-4550 and capivasertib each decreased the overall growth of Onc-negativeRAS/PI3K cell lines in a dose-dependent manner (Fig. 6B; Supplementary Figs. S17D and S17E). Consistent with our in vivo observations, RMC-4550 and capivasertib synergized to inhibit the growth of these cell lines (Fig. 6C; Supplementary Figs. S17F and S17G). RMC-4550 and capivasertib inhibited proliferation and induced apoptosis to a greater extent than RMC-4550 or capivasertib alone (Fig. 6D and E). RMC-4550 and capivasertib treatment also led regression of subcutaneous allografts generated from these Onc-negativeRAS/PI3K cell lines (Fig. 6F and G; Supplementary Figs. S17H–S17J).

Figure 6.

RMC-4550 and capivasertib synergize to inhibit the growth of Onc-negativeRAS/PI3K lung adenocarcinoma cell lines. A, Cell-line generation from Onc-negativeRAS/PI3K tumors from Trp53flox/flox;TC mice. B, Drug dose–response matrix depicting percent growth inhibition of a murine Onc-negativeRAS/PI3K cell line after 4 days of treatment with the indicated doses of RMC-4550 and capivasertib. The average responses of three to four replicates are shown for each treatment. C, Loewe synergy score calculated based on drug responses in B. Synergy scores indicate the percentage of response beyond expectation. D and E, Cell proliferation and apoptosis analysis using EdU incorporation, cleaved caspase-3 staining, and flow cytometry analysis. Three independent Onc-negativeRAS/PI3K murine cell lines were treated with 10 µmol/L of the indicated drug(s) for 2 days before the analysis. F, Volumes of cell-line derived allografts from a murine Onc-negativeRAS/PI3K cell line (MY-C3). Seventeen days after subcutaneous transplantation, three mice were treated with vehicle and three mice received RMC-4550 and capivasertib. **, P < 0.001; ***, P < 0.0001. G, Weights of subcutaneous tumors at the endpoint of the experiment. *, P < 0.05. H, Drug dose–response matrix depicting percent growth inhibition of H1838, a human Onc-negativeRAS/PI3K lung adenocarcinoma cell line. I, Loewe synergy score calculated based on drug responses in H. J, Model of biochemical progression and molecular drivers of Onc-negativeRAS/PI3K tumors.

Figure 6.

RMC-4550 and capivasertib synergize to inhibit the growth of Onc-negativeRAS/PI3K lung adenocarcinoma cell lines. A, Cell-line generation from Onc-negativeRAS/PI3K tumors from Trp53flox/flox;TC mice. B, Drug dose–response matrix depicting percent growth inhibition of a murine Onc-negativeRAS/PI3K cell line after 4 days of treatment with the indicated doses of RMC-4550 and capivasertib. The average responses of three to four replicates are shown for each treatment. C, Loewe synergy score calculated based on drug responses in B. Synergy scores indicate the percentage of response beyond expectation. D and E, Cell proliferation and apoptosis analysis using EdU incorporation, cleaved caspase-3 staining, and flow cytometry analysis. Three independent Onc-negativeRAS/PI3K murine cell lines were treated with 10 µmol/L of the indicated drug(s) for 2 days before the analysis. F, Volumes of cell-line derived allografts from a murine Onc-negativeRAS/PI3K cell line (MY-C3). Seventeen days after subcutaneous transplantation, three mice were treated with vehicle and three mice received RMC-4550 and capivasertib. **, P < 0.001; ***, P < 0.0001. G, Weights of subcutaneous tumors at the endpoint of the experiment. *, P < 0.05. H, Drug dose–response matrix depicting percent growth inhibition of H1838, a human Onc-negativeRAS/PI3K lung adenocarcinoma cell line. I, Loewe synergy score calculated based on drug responses in H. J, Model of biochemical progression and molecular drivers of Onc-negativeRAS/PI3K tumors.

Close modal

Building on these findings, we assessed activation of RAS and PI3K pathways and driver pathway vulnerabilities in two oncogene-negative human lung adenocarcinoma cell lines, NCI-H1838 (NF1LOF) and NCI-H1623 (RASA1LOF). H1838 and H1623 had activation of RAS and PI3K pathways (Supplementary Fig. S17K). Consistent with our findings in mouse Onc-negativeRAS/PI3K cell lines, RMC-4550 synergized with capivasertib to inhibit the growth of these human Onc-negativeRAS/PI3K lung adenocarcinoma cell lines (Fig. 6H and I; Supplementary Figs. S17L and S17M). These in vivo and cell culture analyses indicate that Onc-negativeRAS/PI3K tumors are vulnerable to therapeutic inhibition of these pathways.

It is often overlooked that lung adenocarcinomas without genomic alterations in oncogenes afflict as many patients as those driven by either oncogenic KRAS or EGFR. By querying an extensive set of tumor suppressor gene alterations, we uncovered combinatorial tumor suppressor inactivation as a driver of oncogene-negative lung adenocarcinomas. Importantly, combinatorial inactivation of negative regulators of RAS and PI3K pathways are as potent as oncogenic KRASG12D in initiating lung tumors in vivo.

Furthermore, although NF1 inactivation is sometimes suggested to be an “oncogenic driver” in lung adenocarcinoma (4, 51, 73), Nf1 inactivation alone is insufficient to initiate lung tumors (Supplementary Fig. S8). Even pairwise inactivation of Nf1 and Rasa1, as well as many other tumor suppressor genes, generated very few tumors even after long time periods (Supplementary Fig. S8). These data suggest that genomic and/or epigenetic alterations in multiple genes within and across pathways may be required to surpass the thresholds necessary for Onc-negativeRAS/PI3K lung adenocarcinoma initiation and growth.

Although cancers harbor diverse genomic and epigenomic alterations, these alterations often converge on key pathways and generate similar biochemical changes (12, 74). Pathway activation through genomic and epigenomic inactivation of tumor suppressors can be very diverse, precluding the identification of non-oncogene drivers from gene-centric analysis of human cancer genomic data. Notably, our pathway analysis in oncogene-negative lung adenocarcinomas indicated that mutations in different genes that converge on the RAS and PI3K pathways frequently co-occur (Fig. 4I; Supplementary Fig. S14I). Furthermore, previous reports and our observations suggest frequent nongenomic mechanisms of downregulation of RAS GAPs and PTEN (Fig. 4FH; Supplementary Figs. S14A–S14F; refs. 4, 19–21, 67, 68). Thus, genomic alterations should be viewed as a floor, not a ceiling, in estimating the frequency of pathway alteration.

We assessed the ability of hundreds of complex tumor suppressor genotypes to generate lung tumors. While activation of RAS and PI3K pathway emerged as the most potent driver of oncogene-negative lung adenocarcinomas, our data also suggest that combinatorial inactivation of tumor suppressor genes outside these two pathways can likely initiate tumorigenesis (Fig. 2; Supplementary Fig. S6). Given the mutational diversity and complexity of oncogene-negative human lung adenocarcinomas (75), there remain many other mutational combinations to be investigated. We anticipate that additional studies will uncover other oncogene negative tumor subtypes beyond Onc-negativeRAS/PI3K lung adenocarcinomas.

Knowledge of the genes underlying human cancer is a pillar of cancer diagnostics, personalized medicine, and the selection of rational combination therapies. Our data demonstrate RAS and PI3K pathway activation in the absence of oncogene mutations in a sizable fraction of human lung adenocarcinomas, which predicted therapeutic vulnerability to SHP2 and AKT inhibitors. Beyond SHP2 and AKT, extensive efforts have generated inhibitors for many other components of the RAS and PI3K pathways. Thus, further investigation of the therapeutic targeting of key nodes within the RAS pathway (e.g., SOS, MEK, ERK) and PI3K pathway (e.g., PI3K, mTOR), could contribute to the development of the most effective therapies for Onc-negativeRAS/PI3K lung adenocarcinomas.

Our findings uncover tumorigenic mechanisms and clinical features of oncogene-negative lung adenocarcinomas. This work identifies biomarkers and new therapeutic targets for Onc-negativeRAS/PI3K tumors. The generation of comprehensive molecular and pharmacogenomic maps of oncogene-negative lung adenocarcinomas will transform our understanding of these heretofore poorly characterized lung cancer subtypes.

M. Yousefi reports grants from NIH, American Lung Association, and Stanford Dean's fellowship during the conduct of the study; also has a patent pending. G. Boross reports grants from Tobacco-Related Disease Research Program (TRDRP) during the conduct of the study and also has a patent pending. C. Weiss reports grants from Stanford University during the conduct of the study. S. Chew reports grants from Ono Pharmaceuticals outside the submitted work. C. Swanton reports grants from Pfizer, Boehringer-Ingelheim, Ono Pharmaceutical, Archer Dx (Invitae); grants and personal fees from AstraZeneca, Bristol Myers Squibb, and Roche-Ventana; other support from AstraZeneca, Apogen Biotech, GRAIL, Epic Biosciences, Achilles Therapeuics, and GRAIL; personal fees from Amgen, Novartis, GlaxoSmithKline, MSD, Illumina, Genentech, GRAIL, Medicxi, Bicycle Therapeutics, Metabomed, Roche Innovation Centre Shanghai, Sarah Canon Research Institute outside the submitted work; also has a patent for assay technology to detect tumor recurrence (PCT/GB2017/053289) issued, a patent for to targeting neoantigens (PCT/EP2016/059401) issued, a patent for identifying patent response to immune checkpoint blockade (PCT/EP2016/071471) issued, a patent for determining HLA LOH (PCT/GB2018/052004) issued, a patent for predicting survival rates of cancer patients (PCT/GB2020/050221) issued, a patent for to treating cancer by targeting insertion/deletion mutations (PCT/GB2018/051893) issued, a patent for identifying insertion/deletion mutation targets (PCT/GB2018/051892) issued, a patent for tumor mutations (PCT/US2017/28013) issued, and a patent for identifying responders to cancer treatment (PCT/GB2018/051912) issued; and reports employment with Royal Society Napier Research (RSRP\R\210001). Work in Swanton laboratory was supported by the Francis Crick Institute that receives its core funding from Cancer Research UK (FC001169), the UK Medical Research Council (FC001169), and the Wellcome Trust (FC001169). This research was funded in whole, or in part, by the Wellcome Trust (FC001169). For the purpose of Open Access, the author has applied a CC BY public copyright licence to any author accepted manuscript version arising from this submission. C. Swanton is funded by Cancer Research UK (TRACERx, PEACE and CRUK Cancer Immunotherapy Catalyst Network), Cancer Research UK Lung Cancer Centre of Excellence (C11496/A30025), the Rosetrees Trust, Butterfield and Stoneygate Trusts, NovoNordisk Foundation (ID16584), Royal Society Professorship Enhancement Award (RP/EA/180007), the National Institute for Health Research (NIHR) Biomedical Research Centre at University College London Hospitals, the Cancer Research UK-University College London Centre, Experimental Cancer Medicine Centre, and the Breast Cancer Research Foundation (BCRF 20-157). This work was supported by a Stand Up To Cancer‐LUNGevity-American Lung Association Lung Cancer Interception Dream Team Translational Research Grant (Grant No.: SU2C-AACR-DT23-17 to S.M. Dubinett and A.E. Spira). Stand Up To Cancer is a division of the Entertainment Industry Foundation. Research grants are administered by the American Association for Cancer Research, the Scientific Partner of SU2C. C. Swanton is in receipt of an ERC Advanced Grant (PROTEUS) from the European Research Council under the European Union's Horizon 2020 research and innovation programme (grant agreement no. 835297). D.A. Petrov reports grants from NIH during the conduct of the study, other support from D2G Oncology outside the submitted work and also has a patent for vulnerability of oncogene-negative adenocarcinoma to combined repression of the PI3K and RAS/MAPK pathways pending. M.M. Winslow reports grants from NIH during the conduct of the study; personal fees and other support from D2G Oncology, Inc. outside the submitted work; also has a patent for patent pending. No disclosures were reported by the other authors.

M. Yousefi: Conceptualization, resources, data curation, formal analysis, supervision, funding acquisition, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. G. Boross: Conceptualization, data curation, software, formal analysis, supervision, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. C. Weiss: Data curation, formal analysis, visualization, methodology, project administration, writing–review and editing. C.W. Murray: Data curation, software, formal analysis, investigation, visualization, methodology, project administration, writing–review and editing. J.D. Hebert: Investigation, writing–review and editing. H. Cai: Data curation, formal analysis, investigation, visualization, methodology. E.L. Ashkin: Investigation. S. Karmakar: Investigation. L. Andrejka: Investigation. L.C. Chen: Investigation, writing–review and editing. M. Wang: Investigation. M.K. Tsai: Investigation, writing–review and editing. W.-Y. Lin: Investigation, writing–review and editing. C. Li: Investigation. P. Yakhchalian: Investigation. C.I. Colón: Investigation. S.-K. Chew: Investigation, writing–review and editing. P. Chu: Investigation, writing–review and editing. C. Swanton: Funding acquisition, investigation, writing–review and editing. C.A. Kunder: Resources, funding acquisition, investigation, writing–review and editing. D.A. Petrov: Resources, funding acquisition, investigation, methodology, writing–review and editing. M.M. Winslow: Conceptualization, resources, data curation, supervision, funding acquisition, visualization, methodology, writing–review and editing.

The authors thank the Stanford Shared FACS Facility, Stanford Veterinary Animal Care staff, and Stanford Protein and Nucleic Acid Facility; A. Orantes and S. Mello for administrative support; Stanford's Molecular Genetic Pathology Laboratory and Henning Stehr for help in providing genetically profiled tumor tissues. David Feldser, Joseph Lipsick, Eric Collisson, Christopher McFarland, and members of the Winslow and Petrov laboratories for helpful discussions and reviewing the manuscript. They thank Florent Elefteriou and Alejandro Sweet-Cordero for providing mouse strains. M. Yousefi was supported by a Stanford University School of Medicine Dean's fellowship, an American Lung Association grant, and an NIH Ruth L. Kirschstein National Research Service Award (F32-CA236311). G. Boross, H. Cai, and J.D. Hebert were supported by Tobacco-Related Disease Research Program Postdoctoral Fellowships (T31FT-1772, 28FT-0019, and T31FT-1619). C.W. Murray was supported by the NSF Graduate Research Fellowship Program and an Anne T. and Robert M. Bass Stanford Graduate Fellowship. W-Y. Lin was supported by an AACR Postdoctoral fellowship (17-40-18-LIN). C. Li was the Connie and Bob Lurie Fellow of the Damon Runyon Cancer Research Foundation (DRG-2331). E.L. Ashkin and C.I. Colón were supported by PHS Grant Number CA09302. E.L. Ashkin was also supported by HHMI Gilliam Fellowship for Advanced Study (GT14928). This work was supported by NIH R01-CA231253 (to M.M. Winslow and D.A. Petrov), NIH R01-CA230919 (to M.M. Winslow), and NIH R01-CA234349 (to M.M. Winslow and D.A. Petrov), as well as by the Stanford Cancer Institute, an NCI-designated Comprehensive Cancer Center.

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.

1.
Barta
JA
,
Powell
CA
,
Wisnivesky
JP
.
Global epidemiology of lung cancer
.
Ann Glob Health
2019
;
85
:
8
.
2.
Devarakonda
S
,
Morgensztern
D
,
Govindan
R
.
Genomic alterations in lung adenocarcinoma
.
Lancet Oncol
2015
;
16
:
e342
51
.
3.
McDermott
U
,
Downing
JR
,
Stratton
MR
.
Genomics and the continuum of cancer care
.
N Engl J Med
2011
;
364
:
340
50
.
4.
Cancer Genome Atlas Research Network
.
Comprehensive molecular profiling of lung adenocarcinoma
.
Nature
2014
;
511
:
543
50
.
5.
Carrot-Zhang
J
,
Yao
X
,
Devarakonda
S
,
Deshpande
A
,
Damrauer
JS
,
Silva
TC
, et al
.
Whole-genome characterization of lung adenocarcinomas lacking the RTK/RAS/RAF pathway
.
Cell Rep
2021
;
34
:
108707
.
6.
Campbell
JD
,
Alexandrov
A
,
Kim
J
,
Wala
J
,
Berger
AH
,
Pedamallu
CS
, et al
.
Distinct patterns of somatic genome alterations in lung adenocarcinomas and squamous cell carcinomas
.
Nat Genet
2016
;
48
:
607
16
.
7.
Vaishnavi
A
,
Capelletti
M
,
Le
AT
,
Kako
S
,
Butaney
M
,
Ercan
D
, et al
.
Oncogenic and drug-sensitive NTRK1 rearrangements in lung cancer
.
Nat Med
2013
;
19
:
1469
72
.
8.
Jonna
S
,
Feldman
RA
,
Swensen
J
,
Gatalica
Z
,
Korn
WM
,
Borghaei
H
, et al
.
Detection of NRG1 gene fusions in solid tumors
.
Clin Cancer Res
2019
;
25
:
4966
72
.
9.
Takeuchi
K
,
Soda
M
,
Togashi
Y
,
Suzuki
R
,
Sakata
S
,
Hatano
S
, et al
.
RET, ROS1 and ALK fusions in lung cancer
.
Nat Med
2012
;
18
:
378
81
.
10.
Izumi
H
,
Matsumoto
S
,
Liu
J
,
Tanaka
K
,
Mori
S
,
Hayashi
K
, et al
.
The CLIP1-LTK fusion is an oncogenic driver in non-small-cell lung cancer
.
Nature
2021
;
600
:
319
23
.
11.
Vogelstein
B
,
Papadopoulos
N
,
Velculescu
VE
,
Zhou
S
,
Diaz
LA
,
Kinzler
KW
.
Cancer genome landscapes
.
Science
2013
;
339
:
1546
58
.
12.
Sanchez-Vega
F
,
Mina
M
,
Armenia
J
,
Chatila
WK
,
Luna
A
,
La
KC
, et al
.
Oncogenic signaling pathways in The Cancer Genome Atlas
.
Cell
2018
;
173
:
321
37
.
13.
Krogan
NJ
,
Lippman
S
,
Agard
DA
,
Ashworth
A
,
Ideker
T
.
The cancer cell map initiative: defining the hallmark networks of cancer
.
Mol Cell
2015
;
58
:
690
8
.
14.
George
J
,
Lim
JS
,
Jang
SJ
,
Cun
Y
,
Ozretić
L
,
Kong
G
, et al
.
Comprehensive genomic profiles of small cell lung cancer
.
Nature
2015
;
524
:
47
53
.
15.
Gouyer
V
,
Gazzéri
S
,
Bolon
I
,
Drevet
C
,
Brambilla
C
,
Brambilla
E
.
Mechanism of retinoblastoma gene inactivation in the spectrum of neuroendocrine lung tumors
.
Am J Respir Cell Mol Biol
1998
;
18
:
188
96
.
16.
Sekido
Y
,
Fong
KM
,
Minna
JD
.
Molecular genetics of lung cancer
.
Annu Rev Med
2003
;
54
:
73
87
.
17.
Meuwissen
R
,
Linn
SC
,
Linnoila
RI
,
Zevenhoven
J
,
Mooi
WJ
,
Berns
A
.
Induction of small cell lung cancer by somatic inactivation of both Trp53 and Rb1 in a conditional mouse model
.
Cancer Cell
2003
;
4
:
181
9
.
18.
Govindan
R
,
Ding
L
,
Griffith
M
,
Subramanian
J
,
Dees
ND
,
Kanchi
KL
, et al
.
Genomic landscape of non-small cell lung cancer in smokers and never-smokers
.
Cell
2012
;
150
:
1121
34
.
19.
Soria
JC
,
Lee
HY
,
Lee
JI
,
Wang
L
,
Issa
JP
,
Kemp
BL
, et al
.
Lack of PTEN expression in non-small cell lung cancer could be related to promoter methylation
.
Clin Cancer Res
2002
;
8
:
1178
84
.
20.
Kazanets
A
,
Shorstova
T
,
Hilmi
K
,
Marques
M
,
Witcher
M
.
Epigenetic silencing of tumor suppressor genes: paradigms, puzzles, and potential
.
Biochim Biophys Acta
2016
;
1865
:
275
88
.
21.
Ding
L
,
Getz
G
,
Wheeler
DA
,
Mardis
ER
,
McLellan
MD
,
Cibulskis
K
, et al
.
Somatic mutations affect key pathways in lung adenocarcinoma
.
Nature
2008
;
455
:
1069
75
.
22.
Lee
JS
,
Grisham
JW
,
Thorgeirsson
SS
.
Comparative functional genomics for identifying models of human cancer
.
Carcinogenesis
2005
;
26
:
1013
20
.
23.
Gao
Q
,
Liang
W-W
,
Foltz
SM
,
Mutharasu
G
,
Jayasinghe
RG
,
Cao
S
, et al
.
Driver fusions and their implications in the development and treatment of human cancers
.
Cell Rep
2018
;
23
:
227
38
.
24.
Lu
X
,
Peled
N
,
Greer
J
,
Wu
W
,
Choi
P
,
Berger
AH
, et al
.
MET Exon 14 mutation encodes an actionable therapeutic target in lung adenocarcinoma
.
Cancer Res
2017
;
77
:
4498
505
.
25.
Liu
J
,
Lichtenberg
T
,
Hoadley
KA
,
Poisson
LM
,
Lazar
AJ
,
Cherniack
AD
, et al
.
An integrated TCGA pan-cancer clinical data resource to drive high-quality survival outcome analytics
.
Cell
2018
;
173
:
400
16
.
26.
Liu
C
,
Sage
JC
,
Miller
MR
,
Verhaak
RGW
,
Hippenmeyer
S
,
Vogel
H
, et al
.
Mosaic analysis with double markers reveals tumor cell of origin in glioma
.
Cell
2011
;
146
:
209
21
.
27.
Zhu
Y
,
Romero
MI
,
Ghosh
P
,
Ye
Z
,
Charnay
P
,
Rushing
EJ
, et al
.
Ablation of NF1 function in neurons induces abnormal development of cerebral cortex and reactive gliosis in the brain
.
Genes Dev
2001
;
15
:
859
76
.
28.
Jackson
EL
,
Willis
N
,
Mercer
K
,
Bronson
RT
,
Crowley
D
,
Montoya
R
, et al
.
Analysis of lung tumor initiation and progression using conditional expression of oncogenic K-ras
.
Genes Dev
2001
;
15
:
3243
8
.
29.
Madisen
L
,
Zwingman
TA
,
Sunkin
SM
,
Oh
SW
,
Zariwala
HA
,
Gu
H
, et al
.
A robust and high-throughput Cre reporting and characterization system for the whole mouse brain
.
Nat Neurosci
2010
;
13
:
133
40
.
30.
Chiou
S-H
,
Winters
IP
,
Wang
J
,
Naranjo
S
,
Dudgeon
C
,
Tamburini
FB
, et al
.
Pancreatic cancer modeling using retrograde viral vector delivery and in vivo CRISPR/Cas9-mediated somatic genome editing
.
Genes Dev
2015
;
29
:
1576
85
.
31.
Okawa
H
,
Motohashi
H
,
Kobayashi
A
,
Aburatani
H
,
Kensler
TW
,
Yamamoto
M
.
Hepatocyte-specific deletion of the keap1 gene activates Nrf2 and confers potent resistance against acute drug toxicity
.
Biochem Biophys Res Commun
2006
;
339
:
79
88
.
32.
Bardeesy
N
,
Sinha
M
,
Hezel
AF
,
Signoretti
S
,
Hathaway
NA
,
Sharpless
NE
, et al
.
Loss of the Lkb1 tumour suppressor provokes intestinal polyposis but resistance to transformation
.
Nature
2002
;
419
:
162
7
.
33.
Jonkers
J
,
Meuwissen
R
,
van der Gulden
H
,
Peterse
H
,
van der Valk
M
,
Berns
A
.
Synergistic tumor suppressor activity of BRCA2 and p53 in a conditional mouse model for breast cancer
.
Nat Genet
2001
;
29
:
418
25
.
34.
Rogers
ZN
,
McFarland
CD
,
Winters
IP
,
Naranjo
S
,
Chuang
C-H
,
Petrov
D
, et al
.
A quantitative and multiplexed approach to uncover the fitness landscape of tumor suppression in vivo
.
Nat Methods
2017
;
14
:
737
42
.
35.
Cai
H
,
Chew
SK
,
Li
C
,
Tsai
MK
,
Andrejka
L
,
Murray
CW
, et al
.
A Functional taxonomy of tumor suppression in oncogenic KRAS-driven lung cancer
.
Cancer Discov
2021
;
11
:
1754
73
.
36.
Rogers
ZN
,
McFarland
CD
,
Winters
IP
,
Seoane
JA
,
Brady
JJ
,
Yoon
S
, et al
.
Mapping the in vivo fitness landscape of lung adenocarcinoma tumor suppression in mice
.
Nat Genet
2018
;
50
:
483
6
.
37.
Fedchenko
N
,
Reifenrath
J
.
Different approaches for interpretation and reporting of immunohistochemistry analysis results in the bone tissue - a review
.
Diagn Pathol
2014
;
9
:
221
.
38.
Feoktistova
M
,
Geserick
P
,
Leverkus
M
.
Crystal violet assay for determining viability of cultured cells
.
Cold Spring Harb Protoc
2016
;
2016
:
pdb prot087379
.
39.
Nichols
RJ
,
Haderk
F
,
Stahlhut
C
,
Schulze
CJ
,
Hemmati
G
,
Wildes
D
, et al
.
RAS nucleotide cycling underlies the SHP2 phosphatase dependence of mutant BRAF-, NF1- and RAS-driven cancers
.
Nat Cell Biol
2018
;
20
:
1064
73
.
40.
Ianevski
A
,
Giri
AK
,
Aittokallio
T
.
SynergyFinder 2.0: visual analytics of multi-drug combination synergies
.
Nucleic Acids Res
2020
;
48
:
W488
93
.
41.
Loewe
S
.
The problem of synergism and antagonism of combined drugs
.
Arzneimittelforschung
1953
;
3
:
285
90
.
42.
Dobin
A
,
Davis
CA
,
Schlesinger
F
,
Drenkow
J
,
Zaleski
C
,
Jha
S
, et al
.
STAR: ultrafast universal RNA-seq aligner
.
Bioinformatics
2013
;
29
:
15
21
.
43.
Li
B
,
Dewey
CN
.
RSEM: accurate transcript quantification from RNA-Seq data with or without a reference genome
.
BMC Bioinf
2011
;
12
:
323
.
44.
Soneson
C
,
Love
MI
,
Robinson
MD
.
Differential analyses for RNA-seq: transcript-level estimates improve gene-level inferences
.
F1000Res
2015
;
4
:
1521
.
45.
Love
MI
,
Huber
W
,
Anders
S
.
Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2
.
Genome Biol
2014
;
15
:
550
.
46.
Subramanian
A
,
Tamayo
P
,
Mootha
VK
,
Mukherjee
S
,
Ebert
BL
,
Gillette
MA
, et al
.
Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles
.
Proc Natl Acad Sci U S A
2005
;
102
:
15545
50
.
47.
Hanzelmann
S
,
Castelo
R
,
Guinney
J
.
GSVA: gene set variation analysis for microarray and RNA-seq data
.
BMC Bioinf
2013
;
14
:
7
.
48.
Hutter
C
,
Zenklusen
JC
.
The Cancer Genome Atlas: creating lasting value beyond its data
.
Cell
2018
;
173
:
283
5
.
49.
Consortium
APG
.
AACR Project GENIE: Powering precision medicine through an international consortium
.
Cancer Discov
2017
;
7
:
818
31
.
50.
Jorge
SE
,
Kobayashi
SS
,
Costa
DB
.
Epidermal growth factor receptor (EGFR) mutations in lung cancer: preclinical and clinical data
.
Braz J Med Biol Res
2014
;
47
:
929
39
.
51.
Skoulidis
F
,
Heymach
JV
.
Co-occurring genomic alterations in non-small-cell lung cancer biology and therapy
.
Nat Rev Cancer
2019
;
19
:
495
509
.
52.
Saito
M
,
Shiraishi
K
,
Kunitoh
H
,
Takenoshita
S
,
Yokota
J
,
Kohno
T
.
Gene aberrations for precision medicine against lung adenocarcinoma
.
Cancer Sci
2016
;
107
:
713
20
.
53.
Winters
IP
,
Chiou
S-H
,
Paulk
NK
,
McFarland
CD
,
Lalgudi
PV
,
Ma
RK
, et al
.
Multiplexed in vivo homology-directed repair and tumor barcoding enables parallel quantification of Kras variant oncogenicity
.
Nat Commun
2017
;
8
:
2053
.
54.
Winters
IP
,
Murray
CW
,
Winslow
MM
.
Towards quantitative and multiplexed in vivo functional cancer genomics
.
Nat Rev Genet
2018
;
19
:
741
55
.
55.
Lynch
TJ
,
Bell
DW
,
Sordella
R
,
Gurubhagavatula
S
,
Okimoto
RA
,
Brannigan
BW
, et al
.
Activating mutations in the epidermal growth factor receptor underlying responsiveness of non-small-cell lung cancer to gefitinib
.
N Engl J Med
2004
;
350
:
2129
39
.
56.
Ohashi
K
,
Sequist
LV
,
Arcila
ME
,
Lovly
CM
,
Chen
X
,
Rudin
CM
, et al
.
Characteristics of lung cancers harboring NRAS mutations
.
Clin Cancer Res
2013
;
19
:
2584
91
.
57.
Lin
Q
,
Zhang
H
,
Ding
H
,
Qian
J
,
Lizaso
A
,
Lin
J
, et al
.
The association between BRAF mutation class and clinical features in BRAF-mutant Chinese non-small cell lung cancer patients
.
J Transl Med
2019
;
17
:
298
.
58.
Paez
JG
,
Jänne
PA
,
Lee
JC
,
Tracy
S
,
Greulich
H
,
Gabriel
S
, et al
.
EGFR mutations in lung cancer: correlation with clinical response to gefitinib therapy
.
Science
2004
;
304
:
1497
500
.
59.
Politi
K
,
Zakowski
MF
,
Fan
P-D
,
Schonfeld
EA
,
Pao
W
,
Varmus
HE
.
Lung adenocarcinomas induced in mice by mutant EGF receptors found in human lung cancers respond to a tyrosine kinase inhibitor or to down-regulation of the receptors
.
Genes Dev
2006
;
20
:
1496
510
.
60.
Li
D
,
Shimamura
T
,
Ji
H
,
Chen
L
,
Haringsma
HJ
,
McNamara
K
, et al
.
Bronchial and peripheral murine lung carcinomas induced by T790M-L858R mutant EGFR respond to HKI-272 and rapamycin combination therapy
.
Cancer Cell
2007
;
12
:
81
93
.
61.
van Veen
JE
,
Scherzer
M
,
Boshuizen
J
,
Chu
M
,
Liu
A
,
Landman
A
, et al
.
Mutationally-activated PI3'-kinase-alpha promotes de-differentiation of lung tumors initiated by the BRAF(V600E) oncoprotein kinase
.
Elife
2019
;
8
:
e43668
.
62.
Dankort
D
,
Filenova
E
,
Collado
M
,
Serrano
M
,
Jones
K
,
McMahon
M
.
A new mouse model to explore the initiation, progression, and therapy of BRAFV600E-induced lung tumors
.
Genes Dev
2007
;
21
:
379
84
.
63.
Winslow
MM
,
Dayton
TL
,
Verhaak
RGW
,
Kim-Kiselak
C
,
Snyder
EL
,
Feldser
DM
, et al
.
Suppression of lung adenocarcinoma progression by Nkx2–1
.
Nature
2011
;
473
:
101
4
.
64.
Sweet-Cordero
A
,
Mukherjee
S
,
Subramanian
A
,
You
H
,
Roix
JJ
,
Ladd-Acosta
C
, et al
.
An oncogenic KRAS2 expression signature identified by cross-species gene-expression analysis
.
Nat Genet
2005
;
37
:
48
55
.
65.
Agarwal
A
,
Das
K
,
Lerner
N
,
Sathe
S
,
Cicek
M
,
Casey
G
, et al
.
The AKT/I kappa B kinase pathway promotes angiogenic/metastatic gene expression in colorectal cancer by activating nuclear factor-kappa B and beta-catenin
.
Oncogene
2005
;
24
:
1021
31
.
66.
Yang
S-R
,
Lin
C-Y
,
Stehr
H
,
Long
SR
,
Kong
CS
,
Berry
GJ
, et al
.
Comprehensive genomic profiling of malignant effusions in patients with metastatic lung adenocarcinoma
.
J Mol Diagn
2018
;
20
:
184
94
.
67.
Maertens
O
,
Cichowski
K
.
An expanding role for RAS GTPase activating proteins (RAS GAPs) in cancer
.
Adv Biol Regul
2014
;
55
:
1
14
.
68.
Song
MS
,
Salmena
L
,
Pandolfi
PP
.
The functions and regulation of the PTEN tumour suppressor
.
Nat Rev Mol Cell Biol
2012
;
13
:
283
96
.
69.
Hayashi
T
,
Desmeules
P
,
Smith
RS
,
Drilon
A
,
Somwar
R
,
Ladanyi
M
.
RASA1 and NF1 are preferentially co-mutated and define a distinct genetic subset of smoking-associated non-small cell lung carcinomas sensitive to MEK inhibition
.
Clin Cancer Res
2018
;
24
:
1436
47
.
70.
Kitajima
S
,
Barbie
DA
.
RASA1/NF1-mutant lung cancer: racing to the clinic?
Clin Cancer Res
2018
;
24
:
1243
5
.
71.
Middleton
G
,
Fletcher
P
,
Popat
S
,
Savage
J
,
Summers
Y
,
Greystoke
A
, et al
.
The National Lung Matrix Trial of personalized therapy in lung cancer
.
Nature
2020
;
583
:
807
12
.
72.
Davies
BR
,
Greenwood
H
,
Dudley
P
,
Crafter
C
,
Yu
D-H
,
Zhang
J
, et al
.
Preclinical pharmacology of AZD5363, an inhibitor of AKT: pharmacodynamics, antitumor activity, and correlation of monotherapy activity with genetic background
.
Mol Cancer Ther
2012
;
11
:
873
87
.
73.
O'Neill
AC
,
Jagannathan
JP
,
Ramaiya
NH
.
Evolving cancer classification in the era of personalized medicine: a primer for radiologists
.
Korean J Radiol
2017
;
18
:
6
17
.
74.
Hanahan
D
,
Weinberg
RA
.
Hallmarks of cancer: the next generation
.
Cell
2011
;
144
:
646
74
.
75.
Lawrence
MS
,
Stojanov
P
,
Polak
P
,
Kryukov
GV
,
Cibulskis
K
,
Sivachenko
A
, et al
.
Mutational heterogeneity in cancer and the search for new cancer-associated genes
.
Nature
2013
;
499
:
214
8
.

Supplementary data