EML4-ALK Variant-Specific Genetic Interactions Shape Lung Tumorigenesis

Abstract

Diverse fusions of echinoderm microtubule-associated protein-like 4 (EML4) and anaplastic lymphoma kinase (ALK) are oncogenic drivers in lung adenocarcinoma. EML4-ALK variants have distinct breakpoints within EML4, but their functional differences remain poorly understood. In this study, we use somatic genome editing to generate autochthonous mouse models of EML4-ALK-driven lung tumors and show that variant 3 (V3) is more oncogenic than variant 1 (V1). By using multiplexed genome editing and quantifying the effects of 29 putative tumor-suppressor genes on V1- and V3-driven lung cancer growth, we show that many tumor-suppressor genes have variant-specific effects on tumorigenesis. Pharmacogenomic analyses further suggest that tumor genotype can influence therapeutic responses. Analysis of human EML4-ALK-positive lung cancers also identified variant-specific differences in their genomic landscapes. These findings suggest that EML4-ALK variants behave more like distinct oncogenes than a uniform entity and highlight the dramatic impact of oncogenic fusion partner proteins and coincident tumor-suppressor gene alterations on the biology of oncogenic fusion-driven cancers.

Significance: EML4-ALK-driven lung cancer is treated as a uniform disease despite the presence of distinct fusion variants in patients. Our findings show that EML4-ALK variants are functionally distinct, which may have implications for the treatment of this cancer type and highlights the need to consider differences among variants of other oncogenic fusions.

Figure 1.. Lentiviral CRISPR/Cas9-mediated initiation of Eml4-Alk V1 and V3 lung tumors and efficient coincident gene inactivation.

a. Schematic of CRISPR/Cas9-mediated initiation of Eml4-Alk variant 1 (V1)- and variant 3 (V3)-driven lung tumors in R26^{LSL-Cas9-EGFP/+} (Cas9-EGFP) mice by intratracheal injection of Lenti-sgV1/Cre and Lenti-sgV3/Cre lentiviral vectors. Viral titer per mouse, mouse number, and genotype are indicated. b. The overall survival of Cas9-EGFP mice with V1 or V3 tumors is presented, with a median survival of 46 weeks for V1 (n=9) and 20.57 weeks for V3 (n=9). A Log-rank (Mantel-Cox) test shows a p-value of <0.0001. c. Individual Eml4-Alk V1 and V3 tumor volumes. The marginal density plots show the distribution of V1 and V3 tumor volumes along the Y-axis. An unpaired t-test was performed, and the nominal p-value is displayed. d. Tumor burden (Tumor area/Total area, %) from end-point mice with V1 or V3 tumors. Each dot represents a mouse. Unpaired t-test, nominal p-value is shown. e. Representative H&E staining and immunohistochemistry for NKX2-1 and proSP-C of lung sections from mice with V1 and V3 lung tumors. Higher magnification of V1 and V3 tumors with high-grade adenocarcinoma features. Scale bars are 1 mm, 100 μm, and 10 μm. f. Lenti-sgTSG¹⁹-sgV1/Cre and Lenti-sgTSG¹⁹-sgV3/Cre. Pools include 2-3 barcoded vectors targeting 6 putative tumor suppressor genes of interest, 5 vectors encoding inert sgRNAs, and a vector with a sgRNA targeting an essential gene (Pcna). g. Initiation of lung tumors in R26^{LSL-Cas9-EGFP/+} and R26^{LSL-Cas9-EGFP/LSL-Cas9-EGFP+} (Cas9-EGFP) mice with Lenti-sgTSG¹⁹-sgV1/Cre (“V1 cohort”) and Lenti-sgTSG¹⁹-sgV3/Cre (“V3 cohort”). Lentiviral titer, median time-point of analysis, and mouse number are indicated. h. Cumulative distribution functions of tumor size for adaptively sampled sgInert tumors in the V1 and V3 cohorts. Each translucent line represents an sgRNA; the solid lines are the cumulative density across sgRNAs. Note that the plot is truncated at 10⁶ cells to improve resolution at smaller tumor sizes. i-j. Adaptively sampled mean size (ASM) of each tumor genotype normalized to the ASM of sgInert tumors in the V1 (h) and V3 cohorts (i). ASM is a summary metric of tumor fitness that integrates the impact of inactivating each gene on tumor size and number. Each bar is a sgRNA. sgRNAs are grouped by gene. Baseline tumor growth (Y=1): no impact relative to sgInert. sgRNAs that significantly increase or decrease ASM (two-sided FDR-adjusted P-value) are in color. P-values and confidence intervals were calculated using nested bootstrap resampling.

Figure 2.. EML4-ALK V1 and V3-driven lung tumor growth is differentially constrained by diverse tumor suppressor genes.

a. Lenti-sgTSG⁷⁵-sgV1/Cre and Lenti-sgTSG⁷⁵-sgV3/Cre pools include 2-4 barcoded vectors targeting each of the 29 putative tumor suppressor genes of interest, 6 vectors encoding inert sgRNAs, and a vector with a sgRNA targeting an essential gene (Pcna). b. Initiation of lung tumors in R26^{LSL-Cas9-EGFP/LSL-Cas9-EGFP} (Cas9-EGFP) mice with Lenti-sgTSG⁷⁵-sgV1/Cre (“V1 cohort”) and Lenti-sgTSG⁷⁵-sgV3/Cre (“V3 cohort”). Lentiviral titer, median time point of analysis, and mouse number are indicated. c. Cumulative distribution functions of tumor size for adaptively sampled sgInert tumors in the V1 and V3 cohorts. Each translucent line represents one of the inert sgRNAs; the solid lines are the cumulative density across all inert sgRNAs. d. Principal components analysis of tumor suppressive effects performed within each mouse reveals consistent separation of mice with Lenti-sgTSG⁷⁵-sgV1/Cre and Lenti-sgTSG⁷⁵-sgV3/Cre initiate tumors. Each dot is a mouse. e-f. Adaptively sampled mean size (ASM) of each tumor genotype normalized to the ASM of sgInert tumors in the V1 (e) and V3 cohorts (f). ASM is a summary metric of tumor fitness that integrates the impact of inactivating each gene on tumor size and number. Baseline tumor growth (Y=1): no significant impact. Genes that significantly increase or decrease ASM (two-sided FDR-adjusted p-value) are in color. P-values and confidence intervals were calculated using nested bootstrap resampling. g. Summary of the impact of all genes assayed in the Lenti-sgTSG⁷⁵ pools on V1- and V3-driven tumorigenesis. Left, comparison of ASM for all tumor genotypes in the V1 and V3 cohorts. The impact of Pcna inactivation is not shown to improve visualization of smaller magnitudes of effect. Genes are colored as indicated on the right. Right, categorization of genes as tumor suppressor genes (TSG) or tumor-promoting genes in V1- and/or V3-driven tumorigenesis. TSG: ASM > 1 and two-sided FDR-adjusted p-value < 0.05. Tumor-promoting gene: ASM < 1 and two-sided FDR-adjusted p-value < 0.05.

Figure 3.. Validation experiments and molecular analyses of EML4-ALK V1 and V3-driven tumors confirm variant-specific effects of tumor suppressor gene inactivation.

a. Schematic of tumor initiation in R^{26LSL-Cas9-EGFP/+} (Cas9-EGFP) mice with individual Lenti-sgTSG-sgV1/Cre and Lenti-sgTSG-sgV3/Cre vectors. Titers were 1.5x10⁶ IFU/mouse for all Lenti-sgTSG-sgV1/Cre vectors and 1x10⁶ IFU/mouse for all Lenti-sgTSG-sgV3/Cre vectors. b. Overall survival of Cas9-EGFP mice transduced with Lenti-sgTSG-sgV1/Cre and Lenti-sgTSG-sgV3/Cre vectors. Grey line indicates sgNeo-V1 or sgNeo-V3 mice; colored lines indicate sgTSG-V1 or sgTSG-V3 mice. Median survival for each group is indicated. Log-rank (Mantel-Cox) was performed and nominal p-values are indicated. c. Surface tumor area (mm²) from indicated tumor genotypes. Each point represents the median tumor size in one mouse. Horizontal lines indicate the mean within each group. Nominal p-values for significant comparisons versus sgNeo are indicated. d. Volcano plots showing changes in gene expression for indicated comparisons. The number of up- or downregulated genes is indicated. Triangles indicate gene with Log2 fold change >|10|. e. Gene-set enrichment analysis of Hallmark pathways from MSigDB. Top enriched pathways between V1 sgNeo and V3 sgNeo. Normalized enrichment score (NES) is shown. f. Heatmaps showing Z-scored normalized counts for differentially expressed genes from indicated comparisons across all Setd2-deficient and Setd2-proficient V1- and V3-driven tumors. g. Inferred pathway activity scores comparing Setd2-deficient V3 (V3 sgSetd2) to Setd2-deficient V1 (V1 sgSetd2) tumors. The inferred activity is based on the t-values of the differentially expressed genes for each comparison. Positive scores (red) indicate that a pathway is more active in V3 sgSetd2 tumors; negative scores (blue) indicate that a pathway is more active in V1 sgSetd2. Stars denote that a pathway is differentially activated (p<0.05, PROGENy multivariate linear model).

Figure 4.. EML4-ALK V1 and V3 resemble distinct oncogenes with respect to their tumor-suppressive landscapes.

a. Correlation in the rank order of tumor suppressive effects in Cas9-EGFP mice transduced with the Lenti-sgTSG⁷⁵-sgV1/Cre (V1) and Lenti-sgTSG⁷⁵-sgV3/Cre (V3). Genes were ranked by fold change in ASM relative to sgInert vectors. For panels a-c: Confidence intervals were generated through a nested bootstrap resampling procedure where tumor suppressive effects and ranks were repeatedly recalculated to quantify uncertainty in ρ. Points are median values across 10,000 bootstrap resamplings. “Random” reflects the simulated distribution of ρ observed when there is no relationship between two groups given the number of items being ranked. b. Correlation in the rank order of tumor suppressive effects between KRAS G12D, KRAS G12C, BRAF, and EGFR mice from Blair et al. Genes were ranked by fold change in 95th percentile tumor size relative to sgInert. c. Correlation in the rank order of tumor suppressive effects for KRAS G12D mice from Blair et al (“G12D_Blair”) and an independent KRAS G12D dataset (“G12D_Shuldiner”). Only the subset of genes assayed in both datasets were included (N=18). For both datasets, genes were ranked by fold change in 95^th percentile tumor size relative to sgInert vectors. d-e. Comparisons of the rank order of tumor suppressor effects between EML4-ALK V1 and V3 lung tumors (d) and between KRAS G12D and KRAS G12C lung tumors (data from Blair et al., e). Genes assayed in both the Lenti-[sgTSG-sgV1/V3]PoolBC/Cre pools and in Blair et al. are shown.

Figure 5.. Variant-specific and tumor suppressor genotype-specific responses to ALK inhibition.

a. Bright field images of EML4-ALK V1 and V3 mouse derived tumor organoids (mDTOs) treated with DMSO or the ALK inhibitor Lorlatinib over 5 days. Scale bars are 500 μm. Representative of at least two independent experiments. b. Relative area of EML4-ALK V1 and V3 mDTOs in response to DMSO, 500 nM Lorlatinib, or 1000 nM Lorlatinib across two independent experiments. One-way ANOVA followed by a Dunnetťs multiple comparison test was performed and nominal p-values are shown. c. Lenti-U6BCsgRNAPool-sgEA/Cre includes Lenti-U6BCsgRNA-sgEml4V1-sgAlk/Cre vectors with Inert- and Essential gene-targeting sgRNAs and Lenti-U6BCRNA-sgEml4V3-sgAlk/Cre vectors with Inert-, Essential gene- and Tumor suppressor gene (TSG)-targeting sgRNAs in a 1:2 ratio. Each gene is targeted with 2 sgRNAs. d. Tumors were initiated in Cas9-EGFP mice with Lenti-U6BCsgRNAPool-sgEA/Cre. Mice were assigned to Vehicle (n=24) or Lorlatinib (n=27) treatment groups. Mice were treated as indicated and lungs were collected for Tuba-seq^ultra analyses. e. Cumulative density plots of the V3 sgInert tumors in the Vehicle (dashed line) and Lorlatinib (solid green line) Cas9-EGFP groups, sgInert tumors in the vehicle group were shrunk to match the Lorlatnib-treated sgInert tumors (solid blue line) to quantify the treatment response. f. Volcano plot of genotype-by-treatment response estimated by ScoreRTN. Each dot represents a TSG-targeting or inert sgRNA in V3 tumors. Y-axis shows −log10(raw bootstrapped p-values). V3 tumors with vector with sgRNAs Pten#1 or Pten#2 were resistant to Lorlatinib (FDR < 0.25). Analysis was restricted to tumors with >500 cells in the Lorlatinib group. g. ScoreRTN of V3 tumors initiated with each V3 vector in Lenti-U6BCsgRNAPool-sgEA/Cre. 500 cell cutoff. 95% confidence intervals are shown.

Figure 6.. The genomic landscape of EML4-ALK-driven lung cancer.

a. Number of samples per fusion variant in the EML4-ALK non-small cell lung cancer (NSCLC) patient cohort. Variants are annotated per short or long fusions. The fused exons and protein product are shown for each variant. Relevant EML4 and ALK domains are indicated (trimerisation domain (TD); hydrophobic motif in EML proteins (HELP); tandem atypical propeller domain (TAPE); tyrosine kinase domain (TK)). b. Alteration frequencies across all samples in the EML4-ALK NSCLC patient cohort for the 50 most commonly altered genes. Annotations indicate genes evaluated as tumor suppressors (rather than as oncogenes) in analysis of patient samples (top) and included in the sgTSG⁷⁵ pool (bottom), c-f. Alterations in NOTCH genes (c,d) and genes encoding members of the Fanconi anemia complex (d,f) occur significantly more frequently in EML4-ALK V1-driven tumors relative to EML4-ALK V3-driven tumors. Two-sided Fisher’s Exact test p-values are reported. g-i. Alterations in PTEN (g) and canonical oncogenic alterations of PIK3CA (PIK3CAE^542K and PIK3CAE^545K) (h) occur significantly more frequent in EML4-ALK V1-driven tumors relative to EML4-ALK V3-driven tumors. Two-sided Fisher’s Exact test p-value is reported. j. The mutational spectrum of EML4-ALK V2-driven tumors more closely resembles that of EML4-ALK V1-driven tumors than EML4-ALK V3-driven tumors. Shading reflects the Pearson correlation coefficient between alteration frequencies in EML4-ALK V2-driven tumors and alteration frequencies in EML4-ALK V1- and V3-driven tumors for the indicated gene sets. Genes that best differentiate V1 from V3 are genes that have different effects (p<0.1) in the mouse model. *indicates a significant difference (two-sided Fisher’s z-test p<0.05) between correlation of V2-driven tumors with V1- and V3-driven tumors.