Senescent fibroblasts in the tumor stroma rewire lung cancer metabolism and plasticity

Abstract

Senescence has been demonstrated to either inhibit or promote tumorigenesis. Resolving this paradox requires spatial mapping and functional characterization of senescent cells in the native tumor niche. Here, we identified senescent p16 ^Ink4a + cancer-associated fibroblasts with a secretory phenotype that promotes fatty acid uptake and utilization by aggressive lung adenocarcinoma driven by Kras and p53 mutations. Furthermore, rewiring of lung cancer metabolism by p16 ^Ink4a + cancer-associated fibroblasts also altered tumor cell identity to a highly plastic/dedifferentiated state associated with progression in murine and human LUAD. Our ex vivo senolytic screening platform identified XL888, a HSP90 inhibitor, that cleared p16 ^Ink4a + cancer-associated fibroblasts in vivo. XL888 administration after establishment of advanced lung adenocarcinoma significantly reduced tumor burden concurrent with the loss of plastic tumor cells. Our study identified a druggable component of the tumor stroma that fulfills the metabolic requirement of tumor cells to acquire a more aggressive phenotype.

Figure 1.. Contribution of senescent p16^Ink4a+ fibroblasts to CAFs in mouse LUAD.

(A) Schematic of experimental design to investigate p16^Ink4a+ fibroblasts in mouse LUAD and demonstration of capacity to sort for p16^Ink4a+ fibroblasts from LUAD tissues. (B) Representative immunofluorescence images showing nuclear GFP+ cells (green) within the stroma (Laminin+, red) in the lungs of KPTI mice at 8–10 weeks post-tumor induction. Scale bars, 200 μm. (C) Representative image of immunostaining of GFP, ACTA2, PI16, and tdTomato in the lungs of KPTI mice at 10 weeks post-tumor induction. Scale bars, 50 μm. (D) Top: UMAP plot of scRNA-seq data from fibroblasts isolated from KPTI and normal mouse lungs. Bottom: Proportion of fibroblast subtype relative to the total fibroblast population within each condition. (E) Left: Representative image of immunostaining for ɣH2AX and F-Actin in p16^Ink4a− and p16^Ink4a+ fibroblasts isolated from KPTI mouse lungs. Right: Quantitative analysis of ɣH2AX+ cells (n=18 per group). Scale bars, 50 μm. (F) Spatial profiling of mouse LUAD section using Xenium In Situ to elucidate the distribution of p16^Ink4a+ fibroblasts expressing CAF markers within LUAD tissue. Each colored dots represents transcript detection overlaid on segmented cell borders. Tumor/stroma defined by presence/absence of tumor and stromal-specific transcripts. (G) Gene expression density map of iCAF/adventitial and myCAF markers relative to the tumor margins. Unpaired t-test was used in (E) to test statistical significance. Data are represented as mean ± SD.; ***P < 0.001

Figure 2.. p16^Ink4a+ CAFs promote LUAD progression by supporting adjacent HPCS cells.

(A) UMAP derived from spatially analyzed transcripts from KPTI mouse lung section. (B) Gene Density of HPCS and AT2 markers within the tumor. (C) ImageDimPlot of KPTI mouse lung with cell positions annotated by cluster labels, with localization of AT2, HPCS, and myCAF clusters in the region of interest. Scale bars, 500 μm. (D) Transcript mapping with cell segmentation highlighting the proximity between p16^Ink4a+ myCAFs and S100a14+ HPCS cells. Scale bars, 50μm. (E) S100a14 transcript localization aligned with H&E of KPTI lungs with tumor histologic grading for aggressive features. Scale bars, 1000 μm. (F) Immunofluorescence identification of GFP+ACTA2+ fibroblasts (indicated by green arrowheads) and S100A14+tdTomato+ HPCS cells (indicated by yellow arrowheads) in KPTI mouse LUAD. Scale bars, 100 μm. (G) Quantification of the distance between GFP+ACTA2+ fibroblasts and S100A14+ or S100A14− tumor cells, with individual measurements represented as data points (n=103 for S100A14−, n=115 for S100A14+). Unpaired t-test was used in (G) to test statistical significance. Data are represented as mean ± SD.; *P < 0.05, ***P < 0.001, ****P < 0.0001

Figure 3.. p16^Ink4a+ fibroblasts support LUAD growth by increasing HPCS cells in vitro and in vivo.

(A) Top: Schematic of the co-culture setup of FACS-sorted fibroblasts and tdTomato+ LUAD cells under air-liquid interface conditions. Bottom: Images of 3D tumor organoids. (B) Quantitative analysis of organoid sizes established in (A). (C) Representative images of LY6A immunofluorescence in organoids formed as described in (A). Scale bars, 100 μm. (D) Flow cytometry analysis of LY6A+ cell populations in 3D tumor organoids. (E) Single-cell RNA sequencing of tdTomato+ cells from organoids established in (A). Left: Unsupervised clustering of scRNA-seq data, annotated based on Marjanovic et al. Right: UMAP plot showing distinct cellular population that emerges in the tumor organoids co-cultured with p16^Ink4a+ fibroblasts. (F) Pie graph depicting the proportion of cells contributing to identified clusters. (G) Top: Outline of the in vivo transplantation of tdTomato+ LUAD cells with either p16^Ink4a+ or p16^Ink4a− fibroblasts into NSG mice. Bottom: microCT images of mouse lungs 4 weeks post-transplantation, Scale bars, 1 cm. (H) Tumor burden quantification in transplanted mice, expressed as tumor volume relative to whole lung volume at 4 weeks post-transplantation. (I) Representative image of S100A14 and LY6A immunofluorescence in the lungs of recipient mice, 4 weeks post-transplantation. Scale bars, 100 μm. (J) Quantitative analysis of HPCS cell prevalence within tdTomato+ tumor lesions. Unpaired t-test was used in (B), (D), (H), (J) to test statistical significance. Data are represented as mean ± SD.; *P < 0.05, ***P < 0.001, ****P < 0.0001

Figure 4.. APOE derived from p16^Ink4a+ fibroblasts promotes LUAD expansion by enriching the HPCS population.

(A) NicheNet analysis of ligand-receptor interactions between p16^Ink4a+ fibroblasts and tdTomato+ LUAD cells from KPTI mouse lungs. (B) Representative image of APOE immunofluorescence in KPTI lung tissue, indicating high expression of APOE in GFP+ACTA2+ fibroblasts. (C) qPCR evaluation of Apoe expression in fibroblasts sorted from KPTI mouse lungs. Each data point represents a separate biological replicate. (D) Xenium in situ visualization of Postn, Apoe, and GFP expression patterns within tumor areas in KPTI mouse lung. (E) Images of 3D tumor organoids treated with recombinant mouse APOE (rmAPOE). (F) Assessment of tumor organoid sizes with rmAPOE treatment. (G) Immunofluorescence detection of LY6A in tumor organoids derived from the experiment in (E). Scale bars, 100 μm. (H) Determination of LY6A+ cell percentages in organoids via flow cytometry. (I) Representative image of 3D tumor organoids treated with COG133, an ApoE mimetic peptide, to observe inhibitory effects on p16^Ink4a+ fibroblasts function. (J) Measurement of sizes of tumor organoids established in (H). (K) Flow cytometry-based quantification of LY6A+ cells in tumor organoids treated with COG133. Unpaired t-test was used in (C), (F), (H), (J), and (K) to test statistical significance. Data are represented as mean ± SD.; *P < 0.05, ***P < 0.001, ****P < 0.0001

Figure 5.. APOE from p16^Ink4a+ fibroblasts modulates lipid metabolism in LUAD.

(A) Left: Oil Red O staining of KPTI mouse lung tissue for lipid deposits. Scale bars, 100 μm. Right: LY6A immunofluorescence coupled with Bodipy 493/503 staining for neutral lipids in KPTI mouse lung. Scale bars, 50 μm. (B) Xenium in situ imaging aligned with H&E staining demonstrates localization of Apoe and Cldn4 transcripts within vacuolized regions in KPTI lung tissue. (C) Top: Schematic of 3D tumor organoid culture with treatment of rmAPOE protein. Bottom: Representative images of immunofluorescence of LY6A and Bodipy 493/503 staining for neutral lipids in 3D tumor organoids. Scale bars, 25 μm. (D) Diagram outlining the Bodipy fatty acid (Bodipy-C12) transfer assay in 3D tumor organoids. (E) Graph illustrating the increase in Bodipy-C12 Mean Fluorescence Intensity (MFI) in tdT+ tumor cells following rmAPOE treatment, suggesting enhanced fatty acid uptake. Right: MFI quantification of Bodipy-C12 in tdT+ tumor cells co-cultured with p16^Ink4a+ fibroblasts (F) MFI quantification of Bodipy-C12 in tdT+ tumor cells co-cultured with p16^Ink4a+ fibroblasts; downregulation of Apoe using shRNA indicates the role of APOE in supporting tumor cell fatty acid uptake. (G) Heatmap showing the log2 fold change in free fatty acid levels in tumor organoids cultured with p16^Ink4a− fibroblasts, comparing the effects of vehicle and rmAPOE treatment. (H) Schematic representing the metabolism of fatty acids, including their conversion into acyl-carnitines via carnitine palmitoyltransferase 1 (CPT1) for entry into fatty acid oxidation pathways. (I) Images of tumor organoids co-cultured with p16Ink4a+ fibroblasts and treated with either vehicle or etomoxir, a CPT1 inhibitor. (J) Organoid size quantification reveals that inhibition of fatty acid oxidation disrupts the tumor-supportive role of p16^Ink4a+ fibroblasts. (K) Flow cytometry analysis of the LY6A+ cell proportion in tumor organoids subjected to fatty acid oxidation inhibition. One-way ANOVA was used in (E) and (F) and unpaired t-test was used in (J) and (K) to test statistical significance. Data are represented as mean ± SD.; *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001

Figure 6.. Senolytic compound XL888 clears p16Ink4a+ fibroblasts and reduces tumor burden in KPTI mice.

(A) Generation of PCLS cultures from KPTI mouse lungs. (B) Brightfield and fluorescence images of PCLS demonstrating the preservation of tdTomato+ LUAD and GFP+ cells. Scale bars, 1000 m. (C) Flow cytometry strategy for evaluating GFP+ cells within sorted (CD45-EpCAM-CD31-) fibroblasts post-XL888 treatment. (D) Flow cytometry analysis of GFP+ fibroblasts in PCLS treated with XL888 (n=3 slices for each group). (E) Representative image of GFP and ACTA2 immunofluorescence in PCLS treated with vehicle or XL888. Scale bars, 100 m. (F) Quantitative analysis of GFP+ cells among ACTA2+ fibroblasts (n=10 for each group). (G) Experimental design to test tumor suppressive effect of XL888 in KPTI mouse (top), alongside representative macroscopic lung images from vehicle- and XL888-treated groups (bottom). (H) MicroCT images of lungs from KPTI mice following treatment. Scale bars, 1 cm. (I) Assessment of tumor burden in KPTI mice by microCT, expressed as the tumor volume to whole lung volume ratio (n=14–15 mice for each group). (J) Representative image of GFP and ACTA2 immunofluorescence of KPTI mouse lung tissue. Scale bars, 200 m. (K) Quantification of p16Ink4a+ myCAFs in KPTI moues lung (n=6–7 mice for each group). (L) Representative image of S100A14, LY6A, and tdTomato immunofluorescence of KPTI lung tissue. Scale bars, 200 m. (M) Quantification of the proportion of HPCS cells, identified by S100A14 (left) and LY6A (right), within the tdTomato+ LUAD cell population in KPTI mouse lungs (n=6–7 mice for each group). Unpaired t-test was used in (d), (f), (i), (k), and (m) to test statistical significance. Data are represented as mean ± SD.; **P < 0.01, ***P < 0.001, ****P < 0.0001

Figure 7.. Contribution of p16^INK4a+ fibroblasts to human LUAD progression.

(A) UMAP derived from spatially analyzed transcripts from human LUAD section. (B) ImageDimPlot of human LUAD section with cell positions annotated by cluster labels (left), localization of KRT8+/CLDN4+ cells (blue), and myCAFs (yellow) with transcripts of CDKN2A, APOE, and CLDN4 in the region of interest (right). Scale bars, 200 μm. (C) Survival analysis of LUAD patients in the TCGA based on expression level of gene signature in the p16^INK4A+ CAF cluster from a human CAF single cell data set (Grout et al.). (D) Experimental design to test the effects of p16^INK4A-overexpression (p16OE) in human lung fibroblasts on human LUAD (hLUAD) organoids. (E) Images of hLUAD organoids cultured with either control or p16OE fibroblasts. Scale bars, 2000 μm. (F) Quantificaiton of hLUAD organoids in (e) (n=5 for each group). (G) qPCR analysis of p16^INK4A and p21 in normal human lung fibroblasts after overexpression of p16^INK4A (n=3 for each group). (H) qPCR analysis of control or p16OE lung fibroblasts with or without co-culture with tumor (n=3 for each group). (I) Left: Flow cytometry image of tumor organoids from (E). Right: Quantification of Integrin α2 high plastic cell population within tumor organoids (n=4 for each group). (J) Analysis of fatty acid transfer within tumor organoids established in (E) (n=3 for each group). Unpaired t-test was used in (F), (G), (I), (J) and and one-way ANOVA was used in (H) to test statistical significance. Data are represented as mean ± SD.; *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001