miR-200 deficiency promotes lung cancer metastasis by activating Notch signaling in cancer-associated fibroblasts.

Lung adenocarcinoma, the most prevalent lung cancer subtype, is characterized by its high propensity to metastasize. Despite the importance of metastasis in lung cancer mortality, its underlying cellular and molecular mechanisms remain largely elusive. Here, we identified miR-200 miRNAs as potent suppressors for lung adenocarcinoma metastasis. miR-200 expression is specifically repressed in mouse metastatic lung adenocarcinomas, and miR-200 decrease strongly correlates with poor patient survival. Consistently, deletion of mir-200c/141 in the Kras LSL-G12D/+ ; Trp53 flox/flox lung adenocarcinoma mouse model significantly promoted metastasis, generating a desmoplastic tumor stroma highly reminiscent of metastatic human lung cancer. miR-200 deficiency in lung cancer cells promotes the proliferation and activation of adjacent cancer-associated fibroblasts (CAFs), which in turn elevates the metastatic potential of cancer cells. miR-200 regulates the functional interaction between cancer cells and CAFs, at least in part, by targeting Notch ligand Jagged1 and Jagged2 in cancer cells and inducing Notch activation in adjacent CAFs. Hence, the interaction between cancer cells and CAFs constitutes an essential mechanism to promote metastatic potential.

Lung adenocarcinoma, the most prevalent lung cancer subtype, is characterized by its high propensity to metastasize. Metastatic lung adenocarcinoma cells escape primary tumors, disseminate via blood and lymphatic circulation, and ultimately colonize distant organs, particularly the liver, bones, and brain (Budczies et al. 2015; Obenauf and Massagué 2015). Metastasis leads to rapid disease progression and organ failure, accounting for the majority of patient mortality (Chaffer and Weinberg 2011). Despite its immense clinical significance, the cellular and molecular basis for lung adenocarcinoma metastasis remains largely unknown.

Both cancer cell intrinsic mechanisms and tumor-stroma interactions regulate metastatic progression (Quail and Joyce 2013). Previous studies have mostly focused on intrinsic metastasis mechanisms that increase chromosome instability and accessibility, promote developmental plasticity, alter the cancer secretome, or enhance epithelial–mesenchymal transition (EMT) (Winslow et al. 2011; Celià-Terrassa and Kang 2016). However, emerging evidence has implicated interactions between cancer cells and the tumor microenvironment, as lung adenocarcinoma metastases are frequently characterized by desmoplasia and substantial remodeling of the tumor microenvironment (Bremnes et al. 2011; Altorki et al. 2019). One major obstacle in studying lung adenocarcinoma metastasis is the limited number of faithful in vivo models that recapitulate the entire metastatic processes. Widely used transplantation models largely depend on cancer cell lines, often bypassing important in vivo processes to achieve metastatic growth (Gómez-Cuadrado et al. 2017); popular genetically engineered mouse models of lung adenocarcinoma often require a long latency to develop metastasis (Jackson et al. 2005; DuPage et al. 2009).

One of the most widely used models for metastatic lung adenocarcinoma is the Kras; Trp53 (KP) model, in which inducible Kras expression and p53 deletion initiate the growth of lung adenocarcinomas (Jackson et al. 2005). While this model can eventually manifest metastatic tumors that highly resemble human pathology, its usefulness is hampered by long latency and incomplete penetrance (Jackson et al. 2005; DuPage et al. 2009). Here, we identified miR-200 miRNAs as key suppressors for metastasis in KP lung adenocarcinomas. Engineered miR-200 deficiency promoted rapid metastasis to lymph nodes and distant organs, faithfully recapitulating the desmoplastic tumor microenvironment in metastatic human disease. miR-200 deficiency in neoplastic cells increases the expression of Notch ligand Jag1 and Jag2. Enhanced Notch signaling in neighboring stromal fibroblasts induced their proliferation and their remodeling into cancer-associated fibroblasts (CAFs), which in turn promoted metastases. Our findings uncover key miRNAs that suppress metastasis by restricting CAF activation, and highlight the role of tumor microenvironment remodeling in lung cancer metastasis.

Results

Down-regulation of miR-200 miRNAs in a mouse model for metastatic lung cancer

To compare the miRNA expression profiles between primary and metastatic lung adenocarcinomas, we used the Kras;Trp53;R26 (KPT) lung adenocarcinoma mouse model, in which inducible Kras activation and p53 loss in lung epithelial cells yield malignant lung adenocarcinomas with a long latency for metastasis (Jackson et al. 2005; DuPage et al. 2009). Primary tumors and spontaneous metastases from peritoneum, liver, adrenal gland, and soft tissue were collected 5–9 mo after tumor initiation induced by lenti-Cre administration (Winslow et al. 2011). These KPT primary tumors may or may not have metastatic potential. KPT lung cancer cells were marked with the tdTomato Cre-reporter, which enabled isolation of cancer cells (Fig. 1A). Among 641 miRNAs analyzed, 20 highly expressed miRNAs were differentially expressed between primary tumors and metastases (Fig. 1B; Supplemental Fig. S1A; Supplemental Table S1). Strikingly, nearly all miR-200 miRNAs were repressed in metastases (Supplemental Fig. S1A). In particular, miR-141 and miR-200c were among the most down-regulated miRNAs in KPT metastases (Fig. 1B; Supplemental Fig. S1A).

Figure 1.

miR-200 deficiency induces rapid metastasis in KP lung cancer model. (A) Cartoon illustrating isolation of tdTomato+ cancer cells from primary and metastatic KPT lung adenocarcinomas for miRNA profiling. (B) miR-200 miRNAs are strongly down-regulated in distant metastases in the KPT mouse lung adenocarcinoma model. A heat map is shown for the most differentially expressed miRNAs between seven KPT primary lung tumors and four distant metastases (P < 0.05; expression level ≥−20). (Red text) miR-200 miRNAs. (C) The five miR-200 miRNAs reside in two genomic loci and segregate into two subfamilies based on seed sequences. Red box, seed sequences. (D) miR-200 miRNAs are strongly down-regulated in metastatic KPT tumors. Real-time PCR detected miR-200 expression in KP early stage primary tumors (N = 2) and late stage primary tumors (n = 10), but not in KP metastases (n = 7). Error bars indicate SEM. Late stage primary tumors versus metastases. miR-200a: (**) P = 0.0052, t = 3.268, df = 15; miR-429: (*) P = 0.0113, t = 2.887, df = 15; miR-141: (**) P = 0.01, t = 2.047, df = 15; miR-200c: (**) P = 0.0068, t = 3.395, df = 10. All statistical analyses were performed using unpaired, two tailed, Student's t-test. (E) Decreased total MIR-200 expression human lung adenocarcinomas is associated with poor patient survival in the TCGA data (LUAD). A Kaplan-Meier plot compares patient survival between two cohort of patients, with either high (N = 410) or low (N = 22) expression of all five MIR-200 miRNAs. (**) P = 0.00188, log rank test. (F) miR-200c/141 deficiency in the KP200c model significantly reduces overall survival. A Kaplan-Meier plot compares survival of KP and KP200c mice after Adeno-Cre administration (5 × 106 Ad-Cre PFU/mouse), demonstrating a significant acceleration of tumor progression in KP200c mice. (***) P = 0.0004, log rank test. (G) miR-200c/141 deficiency in KP200c mice induces a significant increase in tumor progression. Representative H&E staining of lung sections are shown for three pairs of KP and KP200c mice collected at 90 d after tumor initiation. (Red arrows) Lymph node metastases. (H) KP200c primary tumors are characterized by desmoplastic stroma (blue arrows) and pleomorphic nuclei in cancer cells (red arrows). Representative images are shown for a pair of late stage KP and KP200c primary tumors collected at terminal stage. Scale bars, 100 μm. (I) miR-200 deficiency leads to a significant increase in tumor burden, tumor size, and tumor grade in KP200c mice. (Left) Tumor burden is quantified as the percentage of tumor area versus total lung area. Error bars indicate SEM. (***) P < 0.001. (N) Animal number, unpaired two-tailed Student's t-test, t = 8.199, df = 21. (Middle) Tumor size was measured by tumor area. Error bars indicate SEM. (*) P = 0.0245. (n) Tumor number, unpaired two-tailed Student's t-test, t = 2.278, df = 118. (Right) Tumor grade was determined by histopathological scores. (n) Tumor number. (***) P < 0.001, χ2 = 70.33, df = 2. All KP and KP200c tumors were collected at 12 wk after tumor initiation. (J) KP200c mice exhibit a greater metastasis frequency compared with KP mice. Twenty-six KP and 28 KP200c mice were compared upon tumor initiation with high-dose Cre virus (5 × 106 Ad-Cre PFU/mouse, [***] P < 0.001, χ2 = 156.3, df = 2), with KP200c mice developing more LN metastases. Four KP and six KP200c mice were compared upon tumor initiation with low-dose Cre virus (1 × 105 Lenti-Cre PFU/mouse, [***] P < 0.001, χ2 = 2428, df = 2), with KP200c mice developing more distant metastases. (K) KP200c mice develop metastases in multiple sites. Representative images of H&E staining of a lymph node metastasis, a pleural metastasis, and a distant metastasis in adrenal gland from KP200c mice are shown. Dotted line shows the boundary of metastasis and normal tissue. Scale bar, 100 μm.

The miR-200 family contains five evolutionarily conserved miRNAs in two genomic loci, mir-200b/200a/429 (chromosome 4) and mir-200c/141 (chromosome 6). They are classified into two subfamilies that differ by one nucleotide in seed sequences (Fig. 1C). miR-200 miRNAs were expressed in both early and late stage primary lung adenocarcinomas, exhibiting only a slight down-regulation during tumor progression (Fig. 1D). However, miR-200s were strongly repressed in metastatic lung adenocarcinomas in KPT mice (Fig. 1D). Consistently, in the 432 human lung adenocarcinomas in TCGA, a collective, low miR-200 level was correlated with poor patient survival ([***] P = 0.0018) (Fig. 1E). Hence, down-regulation of miR-200 is associated with tumor metastasis in lung adenocarcinomas.

miR-200 deficiency promotes the development of lung cancer metastases

We next investigated the role of miR-200 miRNAs in lung cancer metastasis in the Kras;Trp53 (KP) mouse model. We engineered individual deletion of mir-200c/141 or mir-200b/200a/429 in mice (Supplemental Fig. S1B; Prosser et al. 2011). mir-200c/141 mice were phenotypically normal, deficiency of mir-200b/200a/429 impaired fertility (Hasuwa et al. 2013), and mir-200c/141; mir-200b/200a/429 mice are lethal (data not shown). While further deletion of mir-200c/141 in KP mice (KP;mir-200c/141, designated as KP200c) exhibits only partial miR-200 deficiency, it nonetheless provides essential insights into the role of miR-200 in lung cancer metastases.

Upon tumor initiation with Adeno-Cre virus (5 × 106 PFU/mouse), KP200c mice quickly developed highly malignant lung adenocarcinomas. Advanced lesions with severe pleomorphic nuclei were present as early as 1 mo after tumor initiation (Supplemental Fig. S1C). At the same time point, KP mice only had adenomas with mild nuclei atypia (Supplemental Fig. S1C). At 12 wk, Cre-transduced cells developed into significantly larger and more aggressive lung adenocarcinomas in KP200c mice (Supplemental Fig. S1D), with a tumor burden five times as high as that of KP mice (Fig. 1G–I). Nearly all KP200c tumors were grade 2 or above, with >63% between grades 3 and 5. In comparison, only 27% KP tumors were grade 3 or above (Fig. 1I). High-grade KP200c tumors were characterized by a large size, a solid histological pattern, large and pleomorphic nuclei, and high degree of desmoplastic stroma (Fig. 1H). Consistently, KP200c mice exhibited a significant decrease in overall survival compared with KP mice ([***] P = 0.0004) (Fig. 1F; Supplemental Table S2).

miR-200c/141 deficiency not only accelerated tumor growth and progression, but also increased the metastasis frequency. More than 60% of KP200c mice (n = 28) developed lymph node (LN) metastases within 120 d, mostly affecting mediastinal and thoracic lymph nodes (Fig. 1G,J; Supplemental Table S2). Within 120 d, only 4% KP mice (n = 26) developed LN metastasis (Fig 1J). When a lower dose of Adeno-Cre virus initiates tumors in KP200c and KP mice (1 × 105 Lenti-Cre PFU/mouse), 50% of KP200c mice developed full-blown distant metastases to liver, pleura, pericardium, and adrenal gland by 150 d (Fig. 1J,K; Supplemental Fig. S1F; Supplemental Table S2), and KP mice exhibited much lower primary tumor burden with no distant metastasis (Fig. 1J; Supplemental Table S2). Interestingly, all KP200c metastases exhibited a complete silencing of mir-200b/200a/429 (Supplemental Fig. S1E), indicating a strong selective pressure for a complete miR-200 loss. Hence, KP200c mice provide a powerful experimental system to probe the underlying mechanisms of adenocarcinoma metastasis.

miR-200 deficiency induces the expansion of CAFs in metastatic KP200c tumors

Metastases in KP200c mice were accompanied by an increase in MAPK signaling and a loss of cell differentiation (Fig. 2A). Compared with KP primary tumors, phosphorylated Erk1/2 level was significantly increased in KP200c primary tumors ([*] P = 0.0263), and further elevated in KP200c LN and distant metastases ([***] P < 0.0001) (Fig. 2A). In addition, 64% of KP200c lymph node and distant metastases were negative for surfactant protein C (SPC), a marker for type 2 alveolar epithelial cells (AT2), and club cell antigen 10 (CC10), a marker for bronchiolar club cells (Fig. 2B). KP200c primary tumors, in comparison, were more heterogeneous in SPC and CC10 expression, with only 28% of tumors being SPClowCC10low (Fig. 2B; Supplemental Fig. S2A). Nearly all KP200c metastases lost expression of Nkx2.1 (Fig. 2C), a lung lineage transcription factor that controls cellular differentiation and limits metastatic potential (Winslow et al. 2011). KP200c primary tumors, however, mostly expressed Nkx2.1, with a down-regulated expression in high-grade tumors (Fig. 2C). Hence, KP200c metastases could be derived from poorly differentiated cancer cells with highly elevated MAPK signaling.

Figure 2.

CAFs are enriched in metastatic KP200c lung adenocarcinomas. (A) Primary and metastatic KP200c tumors exhibit a strong increase in MAPK signaling. Immunohistochemistry staining of phosphorylated Erk1/2 (p-Erk) (left) and quantitation of p-Erk-positive cells (right) indicate a strong elevation of MAPK signaling in KP200c tumors, particularly the metastatic KP200c tumors. Twenty-seven KP primary tumors, 40 KP200c primary tumors and 6 KP200c metastases were analyzed. Scale bar, 100 μm. Error bar indicates SEM. KP versus. KP200c primary tumor. (*) P = 0.0263, unpaired two-tailed Student's t-test, t = 2.273, df = 65; KP200c primary versus KP200c lymph node metastases, (***) P < 0.0001, unpaired two-tailed Student's t-test, t = 11.06, df = 44. (B) Metastatic KP200c tumors are characterized by a loss of cell lineage markers. (Left) Quantitation is shown for immunohistochemistry staining of CC10 (a marker for Club cells) and SPC (a marker for alveolar type 2 cells) in KP200c primary and metastatic tumors. Tumor stratification shows a preferential loss of CC10 and SPC expression in KP200c metastases compared with KP200c primary tumors. KP200c primary tumors versus metastases. (***) P < 0.0001, χ2 = 61.67, df = 3. (C) Metastatic KP200c tumors exhibit a down-regulation of Nkx2.1. Representative images (left) and quantitation (right) of Nkx2.1 immunohistochemistry staining are shown for KP200c primary tumors, lymph node, and distant metastases. Scale bar, 100 μm. Error bar indicates SEM. (***) P < 0.0001, unpaired two-tailed Student's t-test, t = 9.474, df = 57. Forty-seven KP200c primary tumors and 12 KP200c metastases (both lymph node and distant metastases) were analyzed. (D,E) α-SMA-expressing CAFs are enriched in KP200c primary tumors and metastases. (D) Representative images of H&E and α-SMA staining are shown for high-grade regions of KP and KP200c primary tumors, as well as KP200c metastases. Scale bars, 100 μm. (E) Abundance of α-SMA+ CAFs is correlated with the size of KP200c metastases. The percentage of α-SMA staining in total tumor area is plotted against the size of metastatic tumors for 10 KP200c lymph node metastases and two KP lymph node metastases, A simple linear regression line against KP200 metastases is superimposed. R = 0.3599. (F) α-SMA positive fibroblasts are absent from KP and KP200c normal lung and early primary tumors. Representative images of α-SMA staining are shown for KP and KP200c normal lung and early primary tumors. Scale bars, 100 μm. (G) Deficiency of miR-200c/141 specifically in cancer cells promotes metastasis progression and desmoplastic stroma. (Left) KP200cCKO lungs collected 90 d after tumor initiation (5 × 106 Adeno-Cre PFU/mouse) exhibit evident lymph node metastases, with metastasis frequency comparable with that of KP200c mice. (Right) H&E and α-SMA staining of high-grade regions of KP200cCKO primary tumors and lymph node metastases indicate strong CAF enrichment in the tumor microenvironment. Scale bars, 100 μm.

A prominent cellular hallmark of metastatic KP200c tumors was the abundance of collagen-rich, stromal desmoplasia within the metastatic tumor microenvironment. In comparison, KP tumors only exhibited widespread desmoplasia in rare lymph node and liver metastases, but not in the primary tumors (Supplemental Fig. S2B–S2D). The desmoplastic tumors are enriched in CAFs, a heterogeneous population of fibroblasts that can promote cancer growth, survival, and invasiveness by secreting cytokines and growth factors (interleukins, HGF, CXCLs, TGF-β, etc.) (Bhowmick et al. 2004; Orimo et al. 2005; Kalluri and Zeisberg 2006; Frantz et al. 2010; Pietras and Östman 2010; Gascard and Tlsty 2016; Ahirwar et al. 2018), as well as collagen to increase tissue stiffness (Pankova et al. 2016). In KP200c late stage primary tumors, lymph node metastases, and distant metastases, we observed a marked increase of α-smooth muscle actin-expressing CAFs (α-SMA+ CAFs) (Fig. 2D), which grew into multiple layers with stress-fiber-like feature intertwined with clusters of KP200c cancer cells (Fig. 2D). This α -SMA+ stromal structure was largely absent in KP primary tumors, including high-grade tumors where α-SMA+ cells are thin single layer (Fig. 2D). Unlike α-SMA+ CAFs, FSP-1-expressing CAFs did not show a significant increase in KP200c tumors (Supplemental Fig. S2E). Importantly, the percentage of α-SMA+ CAFs in lymph node metastasis was proportional to the tumor size in KP200c metastases, but not in KP metastases (Fig. 2E), with 70% of KP200c lymph node metastases containing at least 20% of α-SMA+ CAFs (Fig. 2E). α-SMA staining was absent from the epithelial compartment in both KP and KP200c normal lung and early lesions (Fig. 2F). Moreover, the CAFs in the advanced KP200c tumors were derived from untransduced cells shown by the absence of tdTomato expression (Supplemental Fig. S1D). Our findings suggest that miR-200 deficiency in lung adenocarcinomas lead to the recruitment and expansion of α-SMA+ CAFs in tumor microenvironment, which could be a critical step for developing metastatic lung adenocarcinomas.

In KP200c mice, both neoplastic cells and stromal cells were deficient for mir-200c/141. Hence, it is unclear whether mir-200c/141 deficiency in cancer cells and/or stromal cells drove tumor growth and metastasis. To address the role of mir-200c/141 specifically in neoplastic cells, we generated KP mice with a mir-200c/141 conditional allele (Kras; p53; mir-200c/141, designated as KP200cCKO mice) (Park et al. 2012). This model generates mir-200c/141-deficient KP cancer cells in an otherwise mir-200c/141-wild type tumor microenvironment (Supplemental Fig. S2F). Similar to KP200c mice, KP200cCKO mice exhibited an increase in metastasis frequency, with 80% of KP200cCKO mice developing lymph node or distant metastases by 120 d (Fig. 2G; Supplemental Table S2; Supplemental Fig. S2G). Metastases in KP200cCKO mice were also characterized by strong desmoplasia with extensive infiltration of α-SMA+ CAFs (Fig. 2G). Hence, miR-200 deficiency in lung cancer cells promotes local and distant metastases, while inducing the expansion of α-SMA+ CAFs in the tumor microenvironment.

Enrichment of SMA+ CAFs occurs in human metastatic lung adenocarcinomas

Consistent with α-SMA+ CAF enrichment in metastatic KP200c mouse lung tumors, α-SMA+ CAFs are also enriched in metastatic human lung adenocarcinomas (Fig. 3A). We compared α-SMA staining in paired primary lung adenocarcinomas and lymph node metastases (n = 8) and observed a significant increase of α-SMA+ CAFs in lymph node metastases (Fig. 3A). Two patients exhibited a similar enrichment of α-SMA+ CAFs in liver metastases relative to their paired primary tumors (Fig. 3B).

Figure 3.

Enrichment of SMA+ CAFs occurs in human metastatic lung adenocarcinomas. (A,B) Human metastatic lung adenocarcinomas exhibit an increase of α-SMA+ CAFs compared with paired primary tumors. (A) Quantitation of α-SMA+ area is shown for eight pairs of human primary and metastatic lung adenocarcinomas; representative immunohistochemistry staining images are shown for three pairs. Scale bars, 100 μm. Pair 1: (ns) P = 0.0661, t = 2.509, df = 4.000; pair 2: (*) P = 0.0148, t = 4.101, df = 4.000; pair 3: (***) P = 0.0004, t = 11.22, df = 4.000; pair 4: (*) P = 0.0467, t = 2.883, df = 3.869; pair 5: (***) P = 0.0010, t = 6.955, df = 4.951; pair 6: (*) P = 0.0250, t = 3.530, df = 3.920; pair 7: (***) P = 0.0001, t = 27.80, df = 3.000; pair 8: (***) P < 0.0001, t = 19.42, df = 4.000. All statistical analyses were performed using unpaired two-tailed Student's t-test with Welch's correction. (B) α-SMA+ CAFs are enriched in liver metastases from relapsed human lung adenocarcinoma patients. H&E staining are shown for two independent liver metastases samples and α-SMA staining are shown for the boxed metastasis region. Dotted line indicates the tumor/liver boundary. Scale bars, 100 μm. (C) Increased cancer-CAF interaction is associated with tumor progression in human lung adenocarcinomas. A stage IA lung adenocarcinoma was stained for α-SMA, and regions representing a progressive histopathological pattern including normal lung, atypical adenomatous hyperplasia (AAH), and adenocarcinoma (ADC) are marked and shown magnified. Quantitation of α-SMA+ area was shown. Scale bar, 100 μm.

Interestingly, we captured a continuum of CAF activation patterns in a stage IA human lung adenocarcinoma (Fig. 3C). In normal lung, the thin alveoli septa consist of a monolayer of α-SMA+ pericytes embedded in the basement membrane of type 1 alveolar cells (AT1) (Fig. 3C, panel a). When the lesion progressed to atypical adenomatous hyperplasia (AAH), the transformed epithelial cells grew into the alveolar space with the α-SMA+ CAFs starting to expand within the alveolar septa (Fig. 3C, panel b). In high-grade adenocarcinomas where the alveolar lumen is filled with cancer cells (Fig. 3C, panel c), the α-SMA+ CAFs formed multilayered structures intermingled with cancer cells. Hence, the expansion of the α-SMA+ CAFs in the tumor microenvironment, as well as the extent of tumor-CAF interaction, both correlate with histological progression in human adenocarcinomas.

Fibroblast coculture promotes the metastatic features in KP200c tumor organoids

We next established a 3D tumor organoid system by culturing one KP and three KP200c lung cancer cell lines (Supplemental Fig. S3A) on top of a matrix containing 50% Matrigel + 0.5 mg/mL Collagen I to mimic a collagen-rich lung tumor microenvironment (Fig. 4A). When cultured alone, KP and KP200c cancer cells both formed well-polarized spherical tumor organoids with hollow lumina (Fig. 4B), with KP200c organoids slightly larger in size (Fig. 4B), but with neither exhibiting invasive features (Fig. 4B).

Figure 4.

miR-200 deficiency in cancer cells promotes CAF activation to drive tumor metastasis. (A) A diagram illustrating the tumor-fibroblast organoid coculture model. GFP-labeled KP or KP200c lung cancer cells are cocultured with tdTomato labeled normal lung fibroblasts to form tumor organoids. (B–D) miR-200 deficiency in KP cancer cells induces metastatic features upon coculture with lung fibroblasts. (B) Fluorescence images (left) and quantitation (right) of tumor-fibroblast organoids indicate that miR-200 deficiency in KP cancer cells, combined with fibroblast coculture, induced tumor clustering and promoted metastatic cellular features. Scale bar, 100 μm. KP200c alone versus KP200c + fibroblasts, (***) P < 0.0001, t = 7.404, df = 568; KP200c + fibroblasts versus KP + fibroblasts, (***) P < 0.0001, t = 7.894, df = 488; unpaired two-tailed Student's t-test. (C,D) Representative images and quantification are shown for clustering and invasiveness in KP and KP200c tumor organoids with fibroblast coculture. Scale bars, 100 μm. (C) KP200c organoids exhibit invasive features in cancer cells (arrows) and elongated morphology in cocultured fibroblasts (arrowheads). (D) Analysis of 87 KP and 112 KP200c tumor organoids indicate a strong increase in invasive features in KP200c tumor organoids. (E) Ki67 immunostaining in KP200c organoid with fibroblast coculture indicates strong cell proliferation in both cancer cells and fibroblasts (tdTomato+). Scale bar, 100 μm. (White arrowheads) Ki67+ cells. (F) miR-200 re-expression in KP200c cancer cells decreases cell proliferation of tumor-associated lung fibroblasts in organoid coculture, but fails to affect non-tumor-associated fibroblasts. The tdTomato signal coverage was measured. Error bar indicates SD. Non-tumor-associated fibroblasts versus tumor-associated fibroblasts: KP: (ns) P = 0.3823, t = 1.021, df = 3; KP200c-MSCV: (***) P = 0.0006, t = 6.484, df = 6; KP200c-MSCV-miR-200c/141: (ns) P = 0.2456, t = 1.440, df = 3, unpaired two-tailed Student's t-test. (G) Coculture of KP200c cancer cells and lung fibroblasts in organoids promotes α-SMA induction in fibroblasts. Scale bars, 100 μm. (H) Increased α-SMA+ CAF proliferation is observed in KP200c high-grade primary tumors and metastatic tumors, but not in KP200c MSCV low-grade primary tumors. Low-grade primary tumors, high-grade primary tumors, lymph node metastases, and adrenal gland metastases from KP200c mice were immunostained for Ki67 and α-SMA. White arrowheads, proliferating CAFs (Ki-67+, α-SMA+). Scale bars, 20 μm. (I) miR-200 expression in KP200c cancer cells regulates the induction of specific cytokine and growth factors by CAFs. Control KP200c cancer cells enhanced fibroblast production of prometastatic cytokines and growth factors including Hgf, Tgfb2, Fgf1 and Ccl5, while miR-200 re-expression in KP200c cancer cells reversed this phenotype. Error bar indicates SD. All statistical analyses are performed using unpaired, two-tailed Student's t-test. (J,K) KP200c cancer cells primed with lung fibroblasts yield highly metastatic lung tumors in an orthotropic allograft tumor model 4 wk after tumor transplantation. (J) A cartoon illustrating the orthotropic allograft model to study lung cancer metastasis using cocultured KP200c cancer cells and lung fibroblasts. (K) H&E staining reveals a highly metastatic tumor phenotype in primary tumors, lymph node metastases, and distant metastases generated from KP200c cancer cells primed by lung fibroblast coculture. Scale bar, 500 μm, black arrowheads, metastases. (L) Transplanted exogenous fibroblasts and endogenous fibroblasts from recipient mice both contribute to CAFs in metastatic tumor microenvironment. Disseminating tumor clusters from pleural cavity and pleural metastasis both contain transplanted tdTomato+ CAFs and endogenous tdTomato− CAFs. Scale bar, 100 μm. (Red arrowheads) tdTomato+ CAFs, (black arrowheads) tdTomato- CAFs.

Interestingly, when cocultured with primary lung fibroblasts, the miR-200 deficient KP200c organoids, but not the miR-200-expressing KP organoids, clustered to form large multi-acinar structures with cellular protrusions, indicative of invasive behavior (Fig. 4C,D). This clustering behavior substantially increased the size of KP200c tumor organoids (Fig. 4B; Supplemental Fig. S3B). The extent of organoid clustering, along with the degree of cell proliferation and invasiveness in KP200c organoids, were proportional to the ratio of fibroblasts to cancer cells (Supplemental Fig. S3C). Consistently, re-expression of mir-200c/141 in KP200c cancer cells reverted this invasive phenotype when cocultured with lung fibroblasts, restoring a spherical organoid morphology (Supplemental Fig. S3D,F). Hence, miR-200 miRNAs act in lung cancer cells to restrict the metastatic phenotype, at least in part, by regulating the functional interaction between cancer cells and fibroblasts.

miR-200 deficiency in tumors promotes the proliferation and activation of CAFs

The interaction between KP200c cancer cells and stromal fibroblasts induced invasive features in neoplastic cells, while promoting their proliferation in 3D organoid culture (Fig. 4E), but not in 2D culture (Supplemental Fig. S3E). Increased fibroblast proliferation was only observed in the vicinity of KP200c tumor organoids (Fig. 4B), as tumor-associated fibroblasts exhibited a 3.8-fold increase in cell density compared with non-tumor-associated fibroblasts (Fig. 4F). No proliferation difference was observed between KP tumor-associated and non-tumor-associated fibroblasts (Fig. 4F). KP200c tumor-associated fibroblasts exhibited elongated morphology, induction of α-SMA, and infiltration into tumor organoids to form direct cell-cell contact with cancer cells (Fig. 4G). Consistently, highly proliferative α-SMA+ CAFs with an elongated morphology were prevalent in KP200c high-grade primary tumors, lymph nodes, and distant metastases, but not in low-grade primary tumors (Fig. 4H; Supplemental Fig. S3G).

miR-200 deficiency in lung cancer cells not only promotes the proliferation of surrounding lung fibroblasts, but also regulates their activity. A major function of CAFs is to secrete cytokines, chemokines, and growth factors to promote tumor growth, survival, and invasion (Gascard and Tlsty 2016). We isolated tdTomato+ lung fibroblasts that were cultured alone or cocultured with KP200c tumor organoids, and compared their expression of CAF-derived secreted factors. Fibroblasts cocultured with KP200c tumor organoids exhibited a strong induction of hepatocyte growth factor (HGF), TGF-β, fibroblast growth factor-1 (Fgf1) and Ccl5 (Fig. 4I), all of which have been reported to contribute to metastatic potential in different cancer models including lung adenocarcinoma (Ao et al. 2007; Karnoub et al. 2007; Borczuk et al. 2008; Kojima et al. 2010; Straussman et al. 2012; Yu et al. 2015; Sun et al. 2017). Interestingly, mir-200c/141 re-expression in KP200c tumor organoids reduced the induction of these growth factors and cytokines in cocultured fibroblasts (Fig. 4I), suggesting that miR-200 miRNAs in cancer cells regulate CAF gene expression via a cell-nonautonomous mechanism. Moreover, treating KP200c cancer cells with HGF, TGF-β, or a combination of both is sufficient to induce the invasive behavior, resulting in enlarged invasive tumor organoid clusters (Supplemental Fig. S3H).

Interplay between KP200c cancer cells and fibroblasts promotes metastatic potential in vivo

Our findings implicate a complex interplay between cancer cells and CAFs regulated by miR-200 miRNAs. To investigate the importance of cancer–fibroblast interactions in promoting metastases, we established an orthotropic allograft lung tumor model in which we intratracheally transplanted lung cancer cells with or without primary lung fibroblasts into immune deficient recipient mice (Fig. 4J). Among two KP and three KP200c lung cancer lines tested, KP lines rarely grafted to form tumors, yet all KP200c lines successfully developed allograft lung tumors by 7 wk (Supplemental Fig. S3I). When KP200c cancer cells were injected with cocultured, tdTomato-labeled, primary lung fibroblasts, tumor growth was significantly accelerated with an increased tumor burden (Fig. 4K; Supplemental Fig. S3J). More importantly, within a short latency of 40 d, all mice exhibited early onset distant metastases affecting the pleural membrane in rib cage, pericardium, adrenal gland, kidney, and subcutaneous soft tissue (Fig. 4K; Supplemental Fig. S3K). In comparison, mice injected with KP200c cancer cells alone, while developing lung tumors, rarely developed metastases at this time point (Fig. 4K; Supplemental Fig. S3K). These allograft KP200c tumors likely gain metastatic potential eventually, as they promoted the proliferation and activation of endogenous stromal fibroblasts in recipient mice. Altogether, our data demonstrate that boosting cancer-fibroblast interactions in KP200c cancer cells significantly accelerated tumor outgrowth and metastases in vivo.

Surprisingly, when KP200c cancer cells were cotransplanted with primary lung fibroblasts into recipient mice, tdTomato+ fibroblasts could be detected within the disseminating cancer cell clusters in the pleural cavity and in distant metastases (Fig. 4L). Similarly, in KP200c mice, we observed cancer cell clusters that contained α-SMA+ CAFs inside blood/lymphatic vessels in the lung (Supplemental Fig. S3L). Consistently, in a human patient with minimally invasive lung adenocarcinomas and lymph node micrometastases, we detected clusters of cells containing α-SMA+ CAFs adjacent to CK19+ cancer cells in blood (Supplemental Fig. S3M), and observed lymph node micrometastases encased by a single layer of α-SMA+ CAFs in the vicinity of lymphatic vessels (Supplemental Fig. S3N). These findings suggest that α-SMA+ CAFs can adhere to metastatic cancer cells during dissemination and extravasation, and prolonged tumor-CAF interaction could play a role throughout different stages of tumor metastasis.

miR-200 miRNAs repress Jag1/Jag2 and inhibit Notch signaling in CAFs to repress metastasis

Cancer cells can convert lung fibroblasts into tumor-promoting CAFs via paracrine signaling or direct cell-cell interactions. Interestingly, KP200c cancer cells induced lung fibroblast proliferation and activation in a tumor organoid coculture system that permitted cell-cell contact, but not in trans-well assays that separated cancer cells and fibroblasts in different compartments to prohibit cell-cell contacts (Fig. 5A). Hence, miR-200 miRNAs likely regulate pathways for direct cell-cell interactions between cancer cells and CAFs.

Figure 5.

miR-200 represses Jag1/Jag2 in cancer cells and inhibits Notch signaling in CAFs. (A) Direct cell–cell contact mediates the activation of cocultured fibroblasts by KP200c cancer cells. KP200c cancer cells promote proliferation of fibroblasts only in tumor organoid culture, where direct cell-cell contact was permitted, but not in transwell assays. Error bars indicate SD. Fibroblasts + KP tumor organoids versus fibroblasts + KP200c tumor organoids, (*) P = 0.0104, t = 3.185, df = 2, unpaired two-tailed Student's t-test. (B) Candidate miR-141 and miR-200c binding sites in the 3′ UTR of Jag1 and Jag2, respectively, as predicted by TargetScan. (C) Metastatic KP200c tumors exhibit an induction of Jag1 and Jag2 compared with KP200c primary tumors. Ten KP200c primary tumors and 2 lymph node metastases were subjected to real-time PCR for Jag1 and Jag2. Error bars indicate standard error. Jag1 KP200c primary tumor versus LN met: (***) P < 0.0001, t = 12.03, df = 10; Jag2 KP200c primary tumor versus LN met: (*) P = 0.0290, t = 2.548, df = 10, unpaired Student's t-test. (D) Overexpression of mir-200c/141 in KP200c cancer cells suppresses the mRNA (left) and protein (right) expression of both Jag1 and Jag2. Error bars indicate SD. (E) Coculture of lung fibroblasts and KP200c cancer cells induces Notch signaling in fibroblasts, but represses Notch signaling in cancer cells. tdTomato+ fibroblasts cocultured with KP200c cancer cells in tumor organoid assay exhibit an increase in Notch target genes (Hes1 and Nrarp) compared with those cocultured with KP cancer cells. When cocultured with fibroblasts, KP200c cancer cells exhibit an increase in Jag1 and Jag2 expression, but decrease in Notch target genes compared with KP cancer cells. Error bar indicates SD. (F) miR-200 re-expression in KP200c cancer cells represses Notch targets Hes1 and Nrarp in cocultured fibroblasts. Error bars indicate SD. All statistical analyses were performed using unpaired, two-tailed Student's t-test. (G) Activated Notch signaling in fibroblasts promotes metastatic behavior of cocultured, miR-200 expressing cancer cells. Lung fibroblasts expressing mouse NICD1, NICD3, or control pcDNA3 vector were cocultured with mir-200c/141 expressing KP200c cancer cells, and invasive features of cancer cells were examined by light microscopy. Fibroblasts with activated Notch signaling, but not control fibroblasts, conferred a metastatic phenotype in miR-200 expressing KP200c cancer cells in organoid culture. Scale bars, 100 μm. (Red arrows) Invasive cellular structures in cancer cells cocultured with fibroblasts with activated Notch signaling. Morphology of tumor organoids was quantified as described in Figure 3D. (H) Metastatic KP200c mouse tumors and metastatic human lung adenocarcinomas both contain CAFs with strong expression of nuclear Notch1, compared with paired primary tumors. Scale bars, 100 μm. Red arrows point to the CAF stained positive for activated Notch1. (I) Knockdown of both Jag1 and Jag2 in KP200c cancer cells inhibits invasiveness in cancer cells and proliferation in fibroblasts when cultured as tumor organoids with lung fibroblasts. Jag1 and Jag2 double knockdown in KP200c cancer cells, when cocultured with lung fibroblasts, restores a spherical morphology of tumor organoids (left), reduces the size of tumor organoids (middle), and decreases the invasive features of tumor organoids (right). Scale bar, 100 μm.

Many studies suggest that a key function of miR-200 miRNAs is to suppress EMT by targeting Zeb1 and Zeb2 (Korpal et al. 2008; Park et al. 2008; Gibbons et al. 2009; Gregory et al. 2011). However, primary and metastatic KP200c tumors remained largely intact in epithelial morphology (Supplemental Fig. S4A), and failed to exhibit miR-200-dependent regulation of Zeb1, Zeb2, vimentin, and E-cadherin (Supplemental Fig. S4B,C). Furthermore, Zeb1 and Zeb2 knockdown in KP200c lung cancer cells did not alter the invasive morphology of KP200c tumor organoids (Supplemental Fig. S4D,E). We cannot exclude the possibility that transient EMT still occurs during lung cancer metastasis in vivo, since miR-200-deficient lung cancer cells are more susceptible to TGF-β-induced EMT (Gibbons et al. 2009). However, our data strongly suggest that the miR-200-EMT pathway is not required for cancer-CAF interactions that play an essential role in promoting metastasis.

Gene ontology analysis on predicted miR-200 targets revealed multiple candidates involved in cell-cell communication, including Jagged1 (Jag1), Jagged2 (Jag2), Postn, Gjc1, and Pcdh9, as predicted by RNA22 and TargetScan algorithms (Miranda et al. 2006; Agarwal et al. 2015). Notch ligands Jag1 and Jag2 emerged as strong candidate miR-200 targets, each containing a predicted miR-141 or miR-200c binding site, respectively (Fig. 5B). Furthermore, Jag1 and Jag2 mRNA expression was elevated in miR-200-deficient metastatic cancer cells (Fig. 5C; Supplemental Fig. S1D), but reduced upon miR-200 re-expression (Fig. 5D). Hence, the di-cistronic miRNA cluster mir-200c/141 acts coordinately to down-regulate both Jag1 and Jag2 expression in cancer cells.

Notch signaling operates in diverse developmental and pathological processes. Notch ligands, including Jag1, Jag2, Dll1, and Dll4, bind to membrane-bound Notch receptors in neighboring cells, inducing a cascade of proteolytic events that ultimately release the Notch intracellular domain (NICD), which translocates into the nucleus (Artavanis-Tsakonas 1999; Kopan and Ilagan 2009; Bray 2016). In the nucleus, NICD interacts with DNA binding protein RBPJ and coactivator MAML to stimulate the transcription of Notch target genes, such as Hes, Hey, and Nrarp (Lamar et al. 2001; Iso et al. 2003). Notch expressing cells are often activated by cell surface ligands in adjacent cells, and Notch ligand producing cells are known to dampen their own Notch activity through cis-inhibition (del Álamo et al. 2011). Consistently, miR-200 deficiency resulted in elevated expression of Notch ligand Jag1 and Jag2 in cancer cells and elevated expression of Notch target genes Hes1 and Nrarp in cocultured fibroblasts, but a decrease in Notch target gene expression in KP200c cancer cells (Fig. 5E). This induction of Notch activity in cocultured fibroblasts was reversed by re-expressing miR-200c/141 in KP200c cancer cells (Fig. 5F).

To investigate the sufficiency of elevated fibroblast Notch signaling to promote the metastatic potential of cancer cells, we manipulated Notch signaling in lung fibroblasts and examined their effects on cocultured tumor organoids. Since Notch1 and Notch3 are the main Notch receptors expressed in lung fibroblasts (data not shown), we activated Notch signaling in lung fibroblasts by overexpressing Notch1 NICD (NICD1) or Notch3 NICD (NICD3), and cocultured them with miR-200c/141-re-expressing KP200c tumor organoids. miR-200 expression abolished metastatic features of KP200c tumor organoids cocultured with normal lung fibroblasts (Supplemental Fig. S3F), yet coculture with NICD1- and NICD3-expressing fibroblasts reversed this miR-200 effect, promoting cancer cell clustering and invasive cellular features (Fig. 5G). This is likely due to the activation of Notch directly inducing the expression of CAF-related cytokines including Ccl2, Ccl5, Hgf and Tgfb2 (Supplemental Fig. S4H). Consistently, α-SMA+ CAFs in KP200c metastases exhibited specific activated Notch1 nuclear staining (Fig. 5H, top), indicative of a strong Notch activation in these cells. Similar observations were made in two human lung adenocarcinoma samples, where activated Notch1 nuclear staining was evident in CAFs associated with metastases (Fig.5H; Supplemental Fig. S4G).

Consistent with Jag1 and Jag2 being the key miR-200 targets that mediate tumor-CAF interactions, knocking down Jag1 and Jag2 in KP200c cancer cells restored a spherical morphology in tumor organoids, abolished tumor clustering and invasive features, and reduced fibroblasts elongation and expansion (Fig. 5I). Hence, Jag1 and Jag2 derepression caused by miR-200 loss in cancer cells triggered Notch activation in adjacent CAFs, which promoted fibroblast proliferation and activation, ultimately enhancing the metastatic potential of cancer cells (Fig. 6).

Figure 6.

A model illustrating the role of miR-200 miRNAs in lung cancer metastasis by regulating the interaction between cancer cells and fibroblasts. In miR-200-expressing cancer cells, miR-200 targets Jag1 and Jag2; hence, the cancer cells fail to activate Notch signaling in adjacent fibroblasts and lack metastatic potential. miR-200 deficiency in cancer cells de-represses Jag1 and Jag2, triggering Notch activation in neighboring CAFs, which promotes fibroblast proliferation and activation, ultimately enhancing the metastatic potential of the cancer cells.

Discussion

miR-200 miRNAs act as potent suppressors of lung cancer metastasis, and complete miR-200 inactivation was observed in metastatic lung adenocarcinomas in both KP and KP200c models. The KP200c model is among the best for studying lung adenocarcinoma metastasis, owing to its short latency and faithful recapitulation of patient pathology. Metastatic KP200c lung adenocarcinomas faithfully recapitulate the desmoplastic stroma found in human metastatic tumors (Bremnes et al. 2011), characterized by a prometastatic niche filled with SMA+ CAFs. These SMA+ CAFs, derived from normal lung fibroblasts, strongly foster the metastatic potential of cancer cells, at least in part, by secreting cytokines and growth factors (Fig. 3G,I). A recent study showed that human metastatic lung adenocarcinomas possess transcriptomes encompassing a developmental continuum ranging from stem cell-like cells to regenerative progenitors in lung epithelia (Laughney et al. 2020). It is likely that CAFs in the desmoplastic microenvironment not only promote tumor cell proliferation, but also create a prometastatic niche to induce developmental plasticity in cancer cells. This is consistent with the reports that Wnt-expressing fibroblasts establish a stem cell niche for AT2 cells in normal development and for cancer stem cells in lung adenocarcinoma (Tammela et al. 2017; Nabhan et al. 2018). Given the heterogeneity of the tumor microenvironment, CAF-dependent prometastatic niches are likely to require interactions between CAFs and other stromal cell types in lung adenocarcinomas.

In many miR-200 studies, miR-200 miRNAs are characterized as key inhibitors of EMT by repressing Zeb1 and Zeb2 (Korpal et al. 2008; Park et al. 2008; Gibbons et al. 2009; Gregory et al. 2011), acting through a cell intrinsic mechanism to enhance metastasis. However, EMT may not be essential for metastasis in KP200c lung adenocarcinomas, as overt EMT was not observed in metastatic KP200c tumors, and Zeb1 and Zeb2 knockdown failed to reverse invasiveness in tumor organoids (Supplemental Fig. S4A,D,E). Rather, miR-200 deficiency in KP200c cancer cells mediates a cell extrinsic mechanism to alter the interaction between cancer cells and CAFs. miR-200 deficiency in cancer cells increases Jag1 and Jag2, elevating Notch signaling in neighboring fibroblasts to promote their conversion into highly proliferative, α-SMA+ CAFs, thereby establishing a prometastatic niche. This Notch activation in CAFs is strongly associated with metastatic lung adenocarcinomas in both mouse models and human patients (Fig. 4H), contributing to a prometastatic microenvironment. Notch signaling regulates fibroblast dependent, cell-cell interactions in multiple lung pathological processes, at least in part, by promoting the secretory factors to remodel microenvironment (Liu et al. 2009). Hence, the miR-200/Jag/Notch axis that orchestrates cancer–CAF interactions are likely to constitutes an important regulation to suppress metastasis, additional in vivo studies are needed to verify the Jag/Notch interaction in driving CAF recruitment and cancer metastases.

The effects of Notch signaling are context-dependent in various tumor types. Activation of Notch3 has been shown to mark cancer stem cells and drive self-renewal in KP lung cancer mouse model (Zheng et al. 2013). Notch activation in fibroblasts promotes CAF activation and tumor progression in lung, breast, and prostate cancer (Su et al. 2017; Strell et al. 2019), yet the loss of CAF Notch signaling promotes an oncogenic tumor microenvironment in squamous cell carcinomas (Hu et al. 2012; Procopio et al. 2015). It is important to recognize that the unique cellular heterogeneity found within each tumor microenvironment could foster a complex crosstalk between cancer cells and stromal cells, and among different stromal cell types. In both mouse and human metastatic lung adenocarcinomas, miR-200 deficient lung cancer cells could also induce Notch signaling in other stromal cell types to promote metastasis.

It has been long recognized that highly metastatic lung adenocarcinomas are frequently associated with a prometastatic and desmoplastic tumor microenvironment, characterized by the expansion of cancer associated fibroblasts (CAFs). Here, we demonstrated that miR-200 is a potent suppressor of tumor metastasis in lung cancer, whose deficiency activates Notch signaling in the tumor microenvironment, particularly CAFs, to remodel tumor microenvironment to favor metastasis.

Materials and methods

Animals

Mice of Kras;Trp53 genotype were crossed with mir-200c/141 mice to generate Kras;Trp53;mir-200c/141 (KP200c) mice (Jackson et al. 2005; Prosser et al. 2011). KP200c mice were crossed with mir-200c/141 mice (Park et al. 2012) to generate Kras;Trp53;mir-200c/141 (KP200cCKO) mice. Lung tumorigenesis was induced by intranasal instillation of 5 × 106 PFU of recombinant Ad-Cre virus (Gene Transfer Vector Core, University of Iowa) or by intratracheal intubation of 1 × 105 PFU Lenti-Cre virus (DuPage et al. 2009).

Human primary and metastatic tumor samples of lung adenocarcinoma

Matched primary and metastatic tumor samples, blood samples liver metastases and early stage lung tumors were collected from patients, approved by the Institutional Review Board of Shanghai Pulmonary Hospital, School of Medicine, Tongji University. These patients were diagnosed with clinical stage Ia/IIa/IIb/IIIa/IV, pathologically confirmed lung adenocarcinoma.

microRNA analysis

For TaqMan microRNA assay, cDNA was synthesized with preamplification from total RNA using Megaplex RT primers, Megaplex PreAmp primers, TaqMan MicroRNA reverse transcription kit and TaqMan PreAmp master mix kit. Subsequently, quantitative-PCR was carried out using TaqMan array rodent microRNA A + B card set v3.0 (Applied Biosystems 4444909). LUAD data were downloaded from TCGA and normalized using upper quantile. Patients were classified into two groups - “lower” and “higher,” according to the expression of the mir-200 family. Survival analysis was performed to compare the two groups, using the package “survival” in R (https://cran.r-project.org/web/packages/survival/index.html). P-value was determined by Log rank test, to quantify the significance of the differences in survival time between the two patient groups. Individual microRNA and mRNA were analyzed by real-time PCR using the primers shown in Supplemental Table S3.

Three-dimensional tumor organoid culture for tumor–stromal interaction

The tumor organoid culture was performed using protocol described previously (Debnath et al. 2003) with modifications. Eight-well chamber slides were coated with 100 µL of Matrigel (Corning 354234)/0.5mǵ/mL Collagen I (Corning 354236) mixture. Tumor cells and fibroblast cells were trypsinized and resuspended in growth media with supplement at twice the desired density and mix with equal volume of 10% Matrigel/Collagen I mixture. The cell/gel mixture were immediately plated on top of the matrix coating and cultured for up to 8 d at 37°C with 5% CO2.

Orthotopic transplantation experiment

Tumor cells and fibroblast were cocultured as organoids in suspension for 5 d prior to transplantation. On the day of orthotopic transplantation, tumor organoids were dissociated into single cells using trypsin/EDTA, and the percentages of tumor cells and tdTomato+ fibroblasts were determined by FACS. Single-cell suspension containing 1 × 105 tumor cells were resuspended in 50 μL of RPMI1640 with 10 mM EDTA and injected into the lung of nu/nu mice (Taconic) by intratracheal intubation. The tumors were collected either at 40 d after the tumor injection or at their terminal stage.

Histopathology and immunohistochemistry

For tumor histology, whole lungs were fixed, dehydrated and embedded in paraffin wax (ACROS Organics). The lung blocks were sectioned at 5-µm thicknesses and stained with hematoxylin and eosin (H&E). Each tumor was given a score of grades 1–5. For IHC, paraffin sections were deparaffinized, dehydrated, and subjected to heat-induced antigen retrieval using Trilogy (Cell Marque 920P). Slides were incubated for 10 min with 3% H2O2, blocked with 10% goat serum in PBS + 0.05% Triton X-100, incubated with primary antibody and then horseradish peroxidase (HRP)-conjugated secondary antibodies.

Jag1/Jag2 knockdown

The Jag1 knockdown lentiviral expression vector were constructed into pRRL-SFFV-GFP-miR-E-Puro (SGEP) vector and the Jag2 knockdown lentiviral expression vector were constructed into pRRL-SFFV-GFP-miR-E-Neo (SGEN) vector (Fellmann et al. 2013). The oligo sequences encoding Jag1 and Jag2 shRNA were designed using splashRNA algorithm (Supplemental Table S4; Pelossof et al. 2017), and cloned into XhoI/EcoRI sites. The shRNA lentiviral vectors were package in HEK293 cells together with pVSV-G and pMD-2 as previously described. Five shRNAs were tested for each gene, and the ones with best knockdown effect were used to generated double knockdown cell line. Cells transduced with both shJag1 and shJag2 were selected 48 h after infection using 2 μg/mL puromycin (Gibco by Thermo Fisher) and 500 μg/mL geneticin (Gibco by Thermo Fisher).