EML4-ALK fusions drive lung adeno-to-squamous transition through JAK-STAT activation

Abstract

Human lung adenosquamous cell carcinoma (LUAS), containing both adenomatous and squamous pathologies, exhibits strong cancer plasticity. We find that ALK rearrangement is detectable in 5.1-7.5% of human LUAS, and transgenic expression of EML4-ALK drives lung adenocarcinoma (LUAD) formation initially and squamous transition at late stage. We identify club cells as the main cell-of-origin for squamous transition. Through recapitulating lineage transition in organoid system, we identify JAK-STAT signaling, activated by EML4-ALK phase separation, significantly promotes squamous transition. Integrative study with scRNA-seq and immunostaining identify a plastic cell subpopulation in ALK-rearranged human LUAD showing squamous biomarker expression. Moreover, those relapsed ALK-rearranged LUAD show notable upregulation of squamous biomarkers. Consistently, mouse squamous tumors or LUAD with squamous signature display certain resistance to ALK inhibitor, which can be overcome by combined JAK1/2 inhibitor treatment. This study uncovers strong plasticity of ALK-rearranged tumors in orchestrating phenotypic transition and drug resistance and proposes a potentially effective therapeutic strategy.

Figure 1.

ALK fusion lung tumor heterogeneity in human LUAS and mouse models. (A) Left panel shows experimental design for WGS and RNA-seq of treatment-naïve human LUAS surgical samples (93 samples were RNA-sequenced and 81 samples were analyzed by WGS). The right panel shows the plot between somatic mutation status of significantly mutated genes from TCGA LUSC database and ALK-rearrangement events. Somatic mutations were identified through WGS analyses and gene fusion events were identified through RNA-seq data. (B) Frequency of ALK rearrangements in three different LUAS cohorts. (C) Representative ALK, p40, and TTF1 immunostaining of one human adenosquamous carcinoma tissue sample. Scale bar, 50 μm. (D) Schematic illustration of Rosa26-Loxp-Stop-Loxp-EML4-ALK and Rosa26-Loxp-Stop-Loxp-EML4-ALK L1196M mouse models. (E) Histological characteristics of ALK-rearranged tumors from EML4-ALK L1196M mice. Representative micrographs showing papillary and acinar tumors, lesions originating in intrabronchial hyperplasia, tumors with signet-ring cell pattern, adenosquamous cell carcinoma, and squamous cell carcinoma. Scale bar, 50 μm. (F) Quantification of average numbers for LUAD, LUAS, and LUSC in the EML4-ALK L1196M model. n = 5 for each time point. (G) Statistical analysis of indicated tumors in the EML4-ALK L1196M model. n = 5 for each time point. (H) Multicolor IHC staining of LUAS from the EML4-ALK L1196M model. TTF1 in green, p63 in red, nucleus in blue (DAPI staining). Scale bar, 50 μm. (I) Multicolor IHC staining of LUAD and LUSC from the EML4-ALK L1196M model. Scale bar, 50 μm. (J) Left panel shows statistical analysis of TTF1⁺p63⁻/TTF1⁺p63⁺LUAD ratio in the EML4-ALK L1196M model. Right panel shows statistical analysis of TTF1⁺p63⁺/TTF1⁻p63⁺ LUSC ratio in the EML4-ALK L1196M model. n = 5 for each time point. (K) Schematic illustration of AST process: at LUAD stage, some LUAD cells become TTF1⁺/p63⁺; at LUAS stage, LUAD cells (TTF1⁺/p63⁻ or TTF1⁺/p63⁺) are mixed with LUSC cells (TTF1⁺/p63⁺); at LUSC stage, TTF1^high/p63⁺, TTF1^low/p63⁺, and TTF1⁻/p63⁺ LUSC cells become dominant. LUAD, LUAS, and LUSC were defined pathologically (see Materials and methods for details). Data in F, G, and J were collected from three independent experiments. *P < 0.05, **P < 0.01 by two-tailed unpaired Student’s t test (F). Data are represented as mean ± SEM. W, week.

Figure S1.

ALK fusion lung tumor heterogeneity in human LUAS and mouse models. (A) Statistical analyses of ALK-rearranged samples in three different LUAS cohorts. (B) Representative ALK, TTF1, and p40 immunostaining of the adenomatous components and squamous components within one human LUAS sample in Cohort 2. Red box indicates adenomatous components; blue box indicates squamous components. Scale bar, 1 mm. (C) Statistical analysis of ALK immunostaining of adenomatous components and squamous components from the available LUAS samples (n = 7) in different LUAS cohorts. (D) Left panel shows schematic illustration of Rosa26-Loxp-Stop-Loxp-EML4-ALK and Rosa26-Loxp-Stop-Loxp-EML4-ALK L1196M mouse models. Right panel shows H&E staining of lung tumors from EML4-ALK and EML4-ALK L1196M mice at 4 wk after Ad-Cre infection. WT, the wildtype EML4-ALK mice; L1196M, the EML4-ALK L1196M mice. (E) Representative H&E staining, TTF1, NapsinA, p63, p40, and CK5 immunostaining of LUAD, LUAS, and LUSC from wt EML4-ALK model. Scale bar, 50 μm. (F) Representative H&E staining, TTF1, NapsinA, p63, p40, and CK5 immunostaining of LUAD, LUAS, and LUSC from EML4-ALK L1196M model. Scale bar, 50 μm. (G) Representative H&E staining, TTF1, p63, and CK5 immunostaining from lung tumors in the wt EML4-ALK and EML4-ALK L1196M mice treated with Ad-Cre for 4 wk (W). Scale bar, 50 μm. (H–J) Statistical analyses of tumor burden (H), average tumor size (I), and tumor size for LUAD and LUAS/LUSC (J) in EML4-ALK L1196M mice at 4 wk (n = 5), 6 wk (n = 5), and 8 wk (n = 5) after Ad-Cre treatment. (K) Quantification of individual tumor size for LUAD (n = 209) and LUSC (n = 33) in the EML4-ALK L1196M mice at 8 wk after Ad-Cre treatment. (L) Multicolor IHC staining of LUAS from the same region illustrated in Fig. 1 H. TTF1 in green, p63 in red, CK5 in white, and nucleus in blue (DAPI staining). Scale bar, 50 μm. (M) Multicolor IHC staining of LUAD and LUSC from the same region illustrated in Fig. 1 I, respectively. TTF1 in green, p63 in red, CK5 in white, nucleus in blue (DAPI staining). Scale bar, 50 μm. (N) Representative TTF1 and p40 immunostaining of one human LUAS sample. Scale bar, 50 μm. Data in H–K were collected from three independent experiments. *P < 0.05, ***P < 0.001 by multiple t test (J), two-tailed unpaired Student’s t test (K). Data are represented as mean ± SEM.

Figure 2.

Single-cell analyses of mouse and human ALK fusion tumors reveal the evolution route of AST. (A) Scheme depicting experimental design for scRNA-seq of mouse LUAD and LUSC. (B) UMAP visualization of epithelial cells (cancer cells) derived from LUAD and LUSC. (C) UMAP visualization of LUAD and LUSC cells labeled with Seurat clusters. (D) Marker gene expression across various clusters. Dot diameter indicates the proportion of cells expressing a given gene; color indicates the expression level. (E) Feature plots of LUAD-related genes (Nkx2-1, Sftpc) and LUSC-related genes (p63, Krt5). (F) Dot plot showing expression of LUAD and LUSC-related genes in each cluster. Dot diameter indicates the proportion of cells expressing a given gene; color indicates the expression level. (G) Pseudotime ordering of cancer cells using Monocle3. (H) Biomarker gene expressional dynamics over pseudotime trajectory. The curves were fitted with the Lowess method. (I) Comparison of squamous expression signature from bulk RNA-seq in 41 human (h) EML4-ALK LUAD samples and 1 human EML4-ALK LUSC sample. The bulk RNA-seq data of 10 human LUSC were downloaded from Zhang et al. (2019b). (J) Statistical analyses of p63, p40, and CK5/6 immunostaining of ALK-positive LUAD samples. (K) Scheme depicting experimental design for scRNA-seq of ALK-rearranged human LUAD biopsies. (L) UMAP visualization of cells from five patients (two LUSC samples and three ALK-rearranged LUAD samples) with cells colored based on the cell types (upper row) and squamous score using the well-established squamous biomarker genes (lower row). The minimum score is indicated by light gray and the maximum score is indicated by yellow. (M) Heat maps showing expression of LUAD-related genes (FOXA2, KRT8, SFTPC, NKX2-1, NAPSA, and SFTPA1) and LUSC-related genes (SOX2, p63, KRT5, KRT14, CLCA2, KRT6A, and DSG3) in human cancer cell clusters from scRNA-seq data.

Figure S2.

scRNA-seq of mouse and human ALK fusion tumors uncovers the evolution route of AST. (A) Top panel shows schematic illustration of tdTomato; EML4-ALK L1196M mouse model. Bottom panel shows flow cytometry sorting of CD45⁺ immune cells and tdTomato⁺ cancer cells for scRNA-seq. (B) t-Distributed Stochastic Neighbor Embedding (t-SNE) visualization of mouse LUAD and LUSC labeled by cell types annotated with LUSCancer package. (C) t-SNE visualization of immune cell types of LUAD and LUSC. Following markers were used for identification of immune subtypes: immune cell (Ptprc), B cell (Cd19, Cd79a), T cell (Cd3d, Cd3e), macrophage (Cd14, Fogr2b), dendritic (Cd22, Fscn1), neutrophil (S100a8, S100a9). (D) Bar plot showing distribution of immune subtypes in mouse LUAD and LUSC. (E) Flow cytometry analysis of T cells (CD45⁺CD3⁺) in LUAD and LUSC. (F) Quantification of T cell populations in LUAD (n = 22) and LUSC (n = 7) by flow cytometry. (G) Flow cytometry analysis of neutrophils (CD45⁺CD11b⁺Ly-6G⁺) in LUAD and LUSC. (H) Quantification of neutrophil populations in LUAD (n = 22) and LUSC (n = 7) by flow cytometry. (I) Heatmap showing mean expression of top differentially expressed genes in each epithelial cell cluster. (J) Feature plots of known LUAD-related genes (Krt8, Foxa2) and LUSC-related genes (Krt14, Krt6a). (K) Expression levels of LUAD- and LUSC-related genes over pseudotime trajectory. (L) Heatmap of IHC staining of p63, p40, and CK5/6 for 206 human LUAD samples with ALK rearrangements. (M) Comparison of squamous expression signatures from bulk RNA-seq in human EML4-ALK LUAD samples, human EGFR LUAD samples, and human LUSC samples. The bulk RNA-seq data were downloaded from the TCGA database and studies (Zhang et al., 2019b; Fang et al., 2021). (N) ssGSEA score of squamous signature among human EML4-ALK LUAD samples, human EGFR LUAD samples, and human LUSC samples. (O) UMAP visualization of scRNA-seq data labeled with Seurat clusters. (P) Heat map displays the scores for all cells across all reference labels annotated with SingleR. DC, dendritic cell; NK cell, natural killer cell. (Q) Expression level for markers of different cell types across various Seurat clusters. Dot diameter indicates the proportion of positive cells. Color indicates the expression level. (R) Representative p40, CK6A, CK14, and CK5 immunostaining of samples from sample hLUAD-1. One LUSC sample was used as a positive control. Scale bar, 50 μm. (S) UMAP visualization of cells of the single-cell sequencing data (Maynard et al., 2020) from four ALK-rearranged LUAD with cells colored based on the cell types and squamous score using the well-established squamous biomarker genes. The minimum score is indicated by blue; the maximum score is indicated by yellow. Data in E–H represent one experiment of three independent experiments. *P < 0.05, **P < 0.01, ***P < 0.001 by two-tailed unpaired Student’s t test (F and H), Wilcoxon rank sum test (N). ns: not significant. Data are represented as mean ± SEM.

Figure S3.

Club cells serve as the major cell-of-origin of AST. (A) Schematic illustration of the Sftpc-CreERT2; EML4-ALK L1196M or Scgb1a1-CreERT2; EML4-ALK L1196M model. (B) Representative H&E staining of Sftpc-CreERT2; EML4-ALK L1196M or Scgb1a1-CreERT2; EML4-ALK L1196M mice tumors without tamoxifen induction. (C) Representative H&E staining, TTF1, and p63 immunostaining from lung tumors in the Sftpc-CreERT2; EML4-ALK L1196M mice without tamoxifen treatment. Scale bar, 50 μm. (D) Representative H&E staining, TTF1, and p63 immunostaining from lung tumors in the Scgb1a1-CreERT2; EML4-ALK L1196M mice without tamoxifen treatment. Scale bar, 50 μm. (E) Statistical analyses of tumor burden, tumor number, and average tumor size of EML4-ALK L1196M mice at 6 wk (W; n = 10), 8 wk (n = 6), and 10 wk (n = 13) after Ad-CC10-Cre treatment. (F) Statistical analyses of tumor burden, tumor number, and average tumor size of EML4-ALK L1196M mice at 6 wk (n = 6), 10 wk (n = 6), 14 wk (n = 5), and 18 wk (n = 9) after Ad-SPC-Cre treatment. (G) Representative H&E staining, TTF1, p63, and CK5 immunostaining of LUAD, LUAS, and LUSC from EML4-ALK L1196M mice after Ad-SPC-Cre treatment for 18 wk. Scale bar, 50 μm. (H) Expression (FPKM) comparison of LUAD and LUSC signature genes among ATII cell–derived LUAD, club cell–derived LUAD, LUAS, and LUSC samples. (I and J) Representative TTF1, p63, and CK5 immunostaining of tumoroids derived from LUAD (I) and LUSC (J) in the EML4-ALK L1196M model. Scale bar, 50 μm. (K) Representative H&E staining, TTF1, p63, and CK5 immunostaining from parental tumors and club cell–derived plastic LUAD tumoroids at early (P0/P1) and late (P6–P9) passages. Scale bar, 50 μm. (L and M) Real-time PCR detection of mRNA levels for p63 (L) and Krt5 (M) of club cell–derived plastic LUAD tumoroids at P6-P9. Tumor, the parental tumor (LUAD) used for organoid culture. (N) Representative photos for bright field (BF), p63, and CK5 immunostaining of four tumoroids (SPC-1, SPC-4, SPC-5, and SPC-6) derived from ATII cells at indicated passages. Scale bar, 50 μm. (O) Real-time PCR detection of p63 and Krt5 mRNA levels of four tumoroids derived from ATII cells. CC10-2P7, club cell–derived LUAD tumoroid #2 at passage 7. (P) Representative H&E staining, p63, SOX2, and CK5 immunostaining of primary tumor and allograft tumor of the ATII cell–derived non-plastic tumoroids (SPC-4). Scale bar, 50 μm. Data in E and F were collected from four independent experiments. Data in K and N were collected from three independent experiments. Data in L, M, and O represent one experiment of three independent experiments. ***P < 0.001 by DEseq2 (H). ns: not significant. Data are represented as mean ± SEM.

Figure 3.

Club cells serve as the main cell-of-origin for AST in EML4-ALK model. (A) Kaplan–Meier curve shows the overall survival of the EML4-ALK L1196M mice receiving Ad-SPC-Cre (n = 9) or Ad-CC10-Cre (n = 13). (B) Statistical analysis of LUAS/LUSC incidence in the EML4-ALK L1196M mice after Ad-SPC-Cre (n = 9) or Ad-CC10-Cre (n = 13) treatment for 18 or 10 wk, respectively. (C) Scheme depicting experimental design for analysis of Ad-SPC-Cre or Ad-CC10-Cre treated mice at serial time points. W, week. (D) Representative H&E staining and immunostaining for TTF1, p63, and CK5 of LUAD, LUSC, and LUAS from the EML4-ALK L1196M mice at 10 wk after Ad-CC10-Cre treatment. Scale bar, 50 μm. (E) Quantification of average numbers of indicated tumors in the EML4-ALK L1196M mice at 6 wk (W; n = 10), 8 wk (n = 6), and 10 wk (n = 13) after Ad-CC10-Cre treatment. (F) Statistical analysis of indicated tumors from the EML4-ALK L1196M mice at 8 wk (n = 6) and 10 wk (n = 13) after Ad-CC10-Cre treatment. (G) Quantification of average numbers indicated tumors in the EML4-ALK L1196M mice at 6 wk (n = 6), 10 wk (n = 6), 14 wk (n = 5), and 18 wk (n = 9) after Ad-SPC-Cre treatment. (H) Statistical analysis of indicated tumors from the EML4-ALK L1196M mice at 14 wk (n = 5) and 18 wk (n = 9) after Ad-SPC-Cre treatment. (I) Heat maps showing gene signatures across various tumors from multiple mouse models. (J) PCA analysis of various tumors from multiple mouse models. (K) Expression (FPKM) comparison of Dsg3, Krt14, and p63 across ATII cell–derived LUAD, club cell–derived LUAD, LUAS, and LUSC from EML4-ALK L1196M mice. Data in A, B, and E–H were collected from four independent experiments. ***P < 0.001 by DEseq2 (K). Data are represented as mean ± SEM.

Figure 4.

Establishment of an organoid model recapitulating the AST process. (A) Schematic illustration of the organoid culture system. Primary tumors from the EML4-ALK L1196M model were dissected for organoid culture. (B) Representative photos for the LUAD or LUSC tumoroids. Scale bar, 50 μm. (C) Representative photos, H&E staining, and immunostaining for TTF1, SOX2, p63, and CK5 in club cell–derived tumoroid (CC10-3). Scale bar, 50 μm. P, passage; BF, bright field. (D) Real-time PCR detection of Nkx2-1, Sox2, p63, and Krt5 mRNA levels in club cell–derived LUAD tumoroid (CC10-3) at indicated passages. (E) Representative H&E staining and p63 immunostaining for primary tumors from Ad-CC10-Cre–treated group used for organoid culture (CC10-2 and CC10-3). Scale bar, 50 μm. (F) Representative H&E staining and immunostaining for TTF1, p63, p40, SOX2, and CK5 of allograft tumors of club cell–derived LUAD tumoroids (CC10-2 and CC10-3). Scale bar, 50 μm. (G) Statistical analysis of squamous transition incidence of LUAD tumoroids from Ad-SPC-Cre (n = 11) or Ad-CC10-Cre (n = 11) treatment. Data in G were collected from three independent experiments. Data are represented as mean ± SEM.

Figure S4.

Inhibition of the JAK-STAT pathway blocks the AST process. (A) Representative immunostaining and statistical analyses of 8-oxo-dGuo⁺ and NQO1⁺ cells in LUAD and LUSC from EML4-ALK L1196M mice. Scale bar, 50 μm. 40 representative images for each group were counted. (B–D) Western blot analyses of p-ERK1/2 (B), p-STAT3 (C), and p-S6 (D). Club cell–derived EML4-ALK L1196M tumoroids were treated with indicated doses of trametinib (Tra), ruxolitinib (Rux), and capivasertib (Cap) for 12 h and then subjected to western blot analyses. (E–G) Cell viability of club cell–derived LUAD tumoroids treated with indicated doses of trametinib (E), ruxolitinib (F), and capivasertib (G) for 72 h. (H) Representative morphology photos of club cell–derived LUAD tumoroids at early passages (p0/1) treated with indicated doses of capivasertib, trametinib, and ruxolitinib for 4 days. Scale bar, 50 μm. (I) Representative p63 immunofluorescence staining for club cell–derived LUAD tumoroids treated with capivasertib (100 nM), trametinib (10 nM), ruxolitinib (5 μM), or DMSO for six passages. p63 in red; nucleus in blue (DAPI staining). Scale bar, 50 μm. (J) Statistical analysis of p63⁺ cell ratio per tumoroid. Tumoroids analyzed for DMSO, capivasertib, trametinib, and ruxolitinib groups were 8, 44, 12, and 21, respectively. (K) Representative TTF1 immunofluorescence staining for club cell–derived LUAD tumoroids treated with capivasertib (100 nM), trametinib (10 nM), ruxolitinib (5 μM), or DMSO for six passages. TTF1 in red; nucleus in blue (DAPI staining). Scale bar, 50 μm. (L) Statistical analysis of TTF1⁺ cell ratio per tumoroid. Tumoroids analyzed for DMSO, capivasertib, trametinib, and ruxolitinib groups were 9, 3, 6, and 6, respectively. (M) Representative morphology photos of club cell–derived LUAD tumoroids treated with 5 μM AZD-1480 for six passages. Scale bar, 50 μm. (N) Western blot analyses of p63, CK5, and SOX2. Club cell–derived EML4-ALK L1196M tumoroids were treated with AZD-1480 (5 μM) or DMSO for six passages and then subjected to western blot analyses. (O) Representative chromogram sequences showing knockout efficiency of sgStat1, sgStat3, sgStat5b, and sgStat6 in club cell–derived plastic EML4-ALK L1196M tumoroids. (P) Relative mRNA levels of p63 in club cell–derived EML4-ALK L1196M plastic tumoroids with or without Stat1, Stat3, Stat5b, or Stat6 knockout. (Q) Protein levels of p63 in club cell–derived plastic EML4-ALK L1196M tumoroids with or without Stat1, Stat3, Stat5b, or Stat6 knockout. (R) Quantification of LUAS/LUSC incidence in the EML4-ALK L1196M model with or without Stat3 knockout. n = 6 for control group (sgTomato, sgTom). n = 5 for sgStat3 group. (S) Representative whole lungs of Stat3 knockout (sgStat3) and control (sgTom) in the EML4-ALK L1196M model. Scale bar, 1 mm. (T) Quantification of LUSC number. (U) Statistical analysis of LUAS and LUSC in the EML4-ALK L1196M model with or without Stat3 knockout. (V) Heat map showing the expression dynamics of multiple STAT3 target genes from TRRUST database (https://www.grnpedia.org/trrust/) in LUAD-like, intermediate, and LUSC-like states (FDR<0.01) according to scRNA-seq data analyses. (W) Representative p-STAT3, TTF1, and p40 immunostaining of the adenomatous components and squamous components within human LUAS samples. Scale bar, 50 μm. Data in A–Q represent one experiment of two independent experiments. Data in R–U were collected from two independent experiments. *P < 0.05, ***P < 0.001 by two-tailed unpaired Student’s t test (A), one-way ANOVA (J and L), one-tailed unpaired Student’s t test (T). ns: not significant. Data are represented as mean ± SEM. Source data are available for this figure: SourceData FS4.

Figure 5.

EML4-ALK phase separation regulates AST through activating JAK-STAT signaling. (A) Immunofluorescence staining in club cell–derived EML4-ALK L1196M tumors. ALK in green, nucleus in blue (DAPI staining). Scale bar, 25 μm. (B) Immunofluorescence staining of club cell–derived EML4-ALK L1196M tumoroids. ALK in green, nucleus in blue (DAPI staining). Scale bar, 25 μm. (C) Human bronchial epithelial BEAS-2B cells were transfected with GFP-EML4-ALK L1196M for 12 h and GFP fluorescence was monitored through live imaging. Snapshots at indicated time points showed the fusion event. Scale bar, 10 μm. (D) Top panel shows representative FRAP images of GFP-EML4-ALK L1196M condensates in BEAS-2B cells. The images were taken before and after photobleaching. Scale bar, 2 μm. Bottom panel shows FRAP recovery curve of GFP-EML4-ALK L1196M condensates in BEAS-2B cells. n = 13. (E) BEAS-2B cells were transfected with GFP-EML4-ALK L1196M or GFP-EML4-21S-ALK L1196M and analyzed via confocal microscopy. Scale bar, 25 μm. (F) Western blot analyses of AKT, ERK1/2, and STAT3 phosphorylation. BEAS-2B cells stably expressing EML4-ALK L1196M or EML4-21S-ALK L1196M were deprived of serum and glucose for 2 h and then subjected to western blot analyses. Ctrl, control; WT, GFP-EML4-ALK L1196M; 21S, GFP-EML4-21S-ALK L1196M. (G) Schematic illustration of EML4-ALK downstream signaling pathways and their corresponding inhibitors. (H) Bright field photos of club cell–derived EML4-ALK L1196M tumoroids treated with DMSO, capivasertib (100 nM), trametinib (10 nM), and ruxolitinib (5 μM) for six passages. Scale bar, 50 μm. (I) Representative CK5 immunofluorescence staining in club cell–derived EML4-ALK L1196M tumoroids. CK5 in red, nucleus in blue (DAPI staining). Scale bar, 50 μm. (J) Statistical analysis of CK5⁺ cell ratio. The numbers of tumoroids analyzed over DMSO, capivasertib, trametinib, and ruxolitinib groups were 20, 18, 21, and 22, respectively. (K) Western blot analyses of p63, CK5, and SOX2. Club cell–derived EML4-ALK L1196M tumoroids were treated with DMSO, 100 nM capivasertib (Cap), 10 nM trametinib (Tra), and 5 μM ruxolitinib (Rux) for six passages and then subjected to western blot analyses. (L) Relative mRNA levels of LUSC signature genes. (M) Experimental design for in vivo treatment of ruxolitinib. W, week. (N) Representative CK5 immunostaining. Scale bar, 1 mm. (O–Q) Statistical analyses of tumor number (O), LUAD number (P), and LUAS/LUSC number (Q) in mice receiving ruxolitinib (n = 5) or vehicle (n = 4). Data in A–F represent one experiment of three independent experiments. Data in H–L and O–Q represent one experiment of two independent experiments. *P < 0.05, **P < 0.01, ***P < 0.001 by one-way ANOVA (J and L), two-tailed unpaired Student’s t test (O–Q). ns: not significant. Data are represented as mean ± SEM. Source data are available for this figure: SourceData F5.

Figure S5.

LUSC or LUAD with squamous signature show poor TKI therapeutic responses. (A) Schematic illustration of lorlatinib treatments in tumoroids. (B) Drug response curves of the non-plastic and plastic tumoroids. Cell viability was measured after 72 h lorlatinib treatment. Non-plastic LUAD tumoroids: LUAD-1, LUAD-2. Plastic tumoroids: LUSC-1, LUSC-2, LUSC-3, and LUSC-4. (C) Representative CC3 immunostaining for LUAD and LUSC in 0.5 mg/kg, 2 mg/kg lorlatinib, and control (vehicle) groups. Scale bar, 50 μm. (D) Statistical analysis of CC3 staining. (E) UMAP visualization of epithelial cells from mouse LUAD, LUSC, and LOR (remaining tumors after 2 wk of 10 mg/kg lorlatinib treatment). Batch effect was removed by fastMNN. (F) Heat map showing mean expression of top differentially expressed genes in each epithelial cell cluster for mouse LUAD, LUSC, and LOR samples. (G) Feature plots of known LUAD-related genes including Foxa2, Krt8, Sftpa1, and Sftpc. (H) Feature plots of known LUSC-related genes including Sox2, p63, Krt14, and Krt6a. (I) Real-time PCR detection confirmed the ectopic expression of p63 and Sox2 in the non-plastic tumoroids. (J) Cell viability detection of Sox2-expressing or control tumoroids treated with lorlatinib. Cell viability was measured after 72 h lorlatinib treatment. (K) Representative immunofluorescence staining of the GFP⁺ (expressing p63) and GFP⁻ (without p63 expression) mixed tumoroids. p63 in red; nucleus in blue (DAPI staining). Scale bar, 50 μm. (L) Flow cytometry analysis of the GFP⁺ cells. The GFP⁺ cells (expressing p63) and GFP⁻ cells (without p63 expression) were mixed at 1:1 ratio and cultured with 100, 300, and 500 nM lorlatinib for 24 h before being subjected to flow cytometry analyses. (M) Representative H&E staining and p63 immunostaining of remaining LUAD in 0.5, 2, 10 mg/kg lorlatinib group. Scale bar, 50 μm. (N) Quantification of average tumor size of the p63⁻ or p63⁺ LUAD in mice with vehicle, 0.5, 2, and 10 mg/kg lorlatinib treatments. (O) Top panel shows TTF1 immunostaining photos in two ALK-rearranged human LUAD biopsies with p40 positivity. Scale bar, 50 μm. Bottom panel shows the PFS of LUAD patients with or without p40 positivity. PFS was calculated from the start date of TKI treatments to the date of tumor progression. (P) Heat map of IHC score (scale 0–300) of p40, CK5, CK6A, and CK14 for seven paired treatment-naïve and drug-resistant biopsies with ALK rearrangements. (Q) The p40, CK5, CK6A, and CK14 immunostaining of treatment-naïve and drug-resistant biopsy samples from patient #6. Scale bar, 50 μm. (R) The H&E staining, p40, and CK5/6 immunostaining of treatment-naïve and drug-resistant biopsy samples from patients #8 and #9. Scale bar, 50 μm. Data in B and I–L represent one experiment of three independent experiments. Data in C, D, M, and N represent one experiment of two independent experiments. *P < 0.05, **P < 0.01, ***P < 0.001 by one-way ANOVA (D), multiple t test (J), two-tailed unpaired Student’s t test (N), Mantel–Cox test (O). ns: not significant. Data are represented as mean ± SEM.

Figure 6.

LUSC or squamous signature–enriched LUAD show unfavorable response to ALK TKIs. (A) Top panel shows experimental design for in vivo lorlatinib treatments. Bottom panel shows representative lung H&E staining from EML4-ALK L1196M mice treated with vehicle or lorlatinib (0.5, 2, or 10 mg/kg). Scale bar, 1 mm. W, week. (B and C) Quantification of tumor burden (B) and average tumor number (C) in mice with various lorlatinib treatments. Veh, vehicle. (D and E) Quantification of average tumor size of LUAD (D) and LUSC (E) in mice with various lorlatinib treatments. (F) Representative immunostaining for CC3 of LUAD and LUSC from mice receiving vehicle or 10 mg/kg lorlatinib. Scale bar, 50 μm. (G) Statistical analysis of CC3 staining. 18 representative images for each group were counted. (H) Representative immunostaining for Ki-67 of LUAD and LUSC from mice receiving vehicle or lorlatinib. Scale bar, 50 μm. (I and J) Statistical analysis of Ki-67 staining of LUAD (I) and LUSC (J). 18 representative images for each group were counted. (K) Kaplan–Meier curves show the PFS of LUAD (n = 60) and LUSC (n = 8) patients with ALK-rearranged tumors. PFS was calculated from the start date of alectinib treatment to the date of tumor progression. (L) UMAP visualization of cancer cells labeled with Seurat clusters for mouse LUAD, LUSC, and LOR samples. LOR: after 2 wk of 10 mg/kg lorlatinib treatment. (M) Feature plots of known LUAD-related genes (Nkx2-1, Napsa) and LUSC-related genes (Krt5, Dsg3) for mouse LUAD, LUSC, and LOR samples. (N) UMAP visualization of epithelial cells labeled with LUAD-like, intermediate, and LUSC-like state. 1: LUAD-like state; 2: intermediate state; 3: LUSC-like state. (O) Dot plot showing expression of known LUAD- and LUSC-related genes in LUAD-like, intermediate, and LUSC-like state. Dot diameter indicates the proportion of cells expressing a given gene; color indicates the expression level. (P) Bar plot showing distribution of LUAD, LUSC, and LOR cells in LUAD-like, intermediate, and LUSC-like state. (Q) Dot plot showing expression of significantly upregulated transcription factors in intermediate state compared to LUAD-like state (false discovery rate [FDR] < 0.001). (R) Scheme experimental design for ectopic expression of p63 and Sox2 in non-plastic tumoroids. (S) Cell viability detection of p63-expressing or control (Ctrl) tumoroids treated with lorlatinib. Cell viability was measured after 72 h of lorlatinib treatment. (T) Top panel shows the experimental design for cell competition experiments. Bottom panel shows the flow cytometry analysis of GFP⁺ cells. The GFP⁺ cells (expressing p63) and GFP⁻ cells (without p63 expression) were mixed at 1:1 ratio and cultured with 100 nM lorlatinib for 24 h before being subjected to flow cytometry analyses. (U) Statistical analyses of the p63⁻/p63⁺ LUAD ratio in vehicle or lorlatinib (0.5, 2, or 10 mg/kg) treatments group. (V) ssGSEA score of squamous signature of LUAD patients with ALK-rearranged tumors. (W) Kaplan–Meier curves show the PFS of LUAD patients with ALK-rearranged tumors (n = 36) according to squamous score status. According to ssGSEA score of squamous signature, patients were subjected into two groups: squamous signature high (n = 8) and squamous signature low (n = 28). PFS was calculated from the start date of TKI treatments to the date of tumor progression. Data in B–J and U represent one experiment of two independent experiments. Data in S and T represent one experiment of three independent experiments. *P < 0.05, **P < 0.01, ***P < 0.001 by one-way ANOVA (B–E, I, and J), multiple t test (G and S), Mantel–Cox test (K and W). ns: not significant. Data are represented as mean ± SEM.

Figure 7.

Combined lorlatinib and ruxolitinib treatment eradicates ALK-driven tumors and significantly improves therapeutic efficacy. (A) Top panel shows experimental design for in vivo combinational treatments using lorlatinib and ruxolitinib. Bottom panel shows lung H&E staining from four groups. Rux: ruxolitinib; Lor: lorlatinib; Lor+Rux: lorlatinib and ruxolitinib. Scale bar, 1 mm. W, week. (B and C) Quantification of tumor burden (B) and average tumor number (C) in mice with vehicle (n = 4), lorlatinib (n = 4), ruxolitinib (n = 4), or combined (n = 4) treatment. (D) Representative H&E staining of tumors from mice with vehicle and combined treatment. Scale bar, 50 μm. (E) Representative immunostaining for CC3 of tumors from vehicle, ruxolitinib, lorlatinib, or combined treatments group. Scale bar, 50 μm. (F) Statistical analysis of CC3 staining. 30 representative images for each group were counted. (G) Representative immunostaining for Ki-67 of tumors from vehicle, ruxolitinib, lorlatinib, or combined treatments group. Scale bar, 50 μm. (H) Statistical analysis of Ki-67 staining. 30 representative images for each group were counted. (I) Proposed model for AST and histological transition-associated TKI resistance. EML4-ALK LUAD can progressively transition into LUSC, which results in increased TKI resistance. The JAK-STAT signaling is important for driving squamous transition. Combined JAK1/2 inhibitor and TKI treatment significantly inhibits the AST process and regain the high efficacy of molecular targeted therapy. Data in B–H represent one experiment of two independent experiments. **P < 0.01, ***P < 0.001 by one-way ANOVA (B, C, F, and H). ns: not significant. Data are represented as mean ± SEM.