Abstract
Platinum-based chemotherapy is commonly used for non-small cell lung cancer (NSCLC) and high-grade serous ovarian cancer (HGSOC) treatments, yet clinical outcomes remain poor. Cellular senescence and its associated secretory phenotype (SASP) can have multiple tumor-promoting activities, but both are largely unexplored in these cancers. In this study, using xenograft, orthotopic and KrasG12V-driven murine NSCLC models, we demonstrate that cisplatin-induced senescence strongly promotes malignant phenotypes and tumor progression, which is stimulated by aging. Mechanistically, we found that a transforming growth factor-beta (TGFβ)-enriched SASP drives pro-proliferative effects through TGFBR1 and AKT/mTOR. TGFBR1 inhibition with galunisertib or senolytic treatment reduces tumor progression driven by cisplatin-induced senescence, and concomitant use of TGFBR1 inhibitors with platinum-based chemotherapy reduces tumor burden and improves survival. Finally, we validate the translational relevance of tumor-promoting TGFβ-enriched SASP using clinical NSCLC and HGSOC samples from patients who received neoadjuvant platinum-based chemotherapy. Together, our findings identify a potential cancer therapy resistance mechanism and provide preclinical proof of concept for future trials.
Subject terms: Senescence, Cancer models, Ageing
González-Gualda, Reinius et al. demonstrate that platinum-based chemotherapy-induced senescence promotes malignancy in ovarian and lung cancer via TGFβ ligands, with evidence in mouse models validated in clinical samples. Concomitantly blocking TGFβ signaling with chemotherapy reduces tumor burden and increases survival in mice.
Main
Lung cancer accounts for the highest proportion of cancer-related deaths worldwide, with more than 1.8 million deaths annually. Although therapy with immune checkpoint and EGFR inhibitors has substantially improved outcomes in selected patients, platinum-based therapy, either alone or in other combinations, remains an essential standard-of-care treatment for most patients with lung cancer1. Platinum-based therapy also has high efficacy in cancers with high chromosomal instability, particularly HGSOC2–4. Recent studies on cytotoxic resistance have focused on genetic and non-genetic evolution in the cell-autonomous tumor compartment that can generate complex cellular resistance traits5–7. However, platinum-based therapy may induce adverse tumor microenvironment phenotypes that drive resistance by crosstalk with the cell-autonomous tumor compartment.
Platinum-based chemotherapy results in increased levels of intratumoral and stromal senescence-associated β-galactosidase (SA-β-gal) activity in NSCLC, indicating cellular senescence8. Additional studies have demonstrated the accumulation of senescent cells in different human tumor types after neoadjuvant chemotherapy (NACT), including breast, mesothelioma, prostate, renal cell and rectal cancers9, suggesting that senescence is a common response to anticancer therapies10. A key hallmark of senescence is strong paracrine secretion of cytokines, growth factors, proteases and other mediators, jointly named the senescence-associated secretory phenotype (SASP)11. Senescence is considered a tumor-suppressive mechanism, but increasing evidence indicates that persistent senescent cells can also drive potent tumor-promoting effects in a paracrine manner12,13. However, the specific SASP factors responsible for these effects remain unknown.
Multiple components of the SASP have been identified (including IL-6, IL-1α, CXCL2, MMP-3 and VEGF), but, in different clinical states, the SASP is heterogeneous and dynamic and highly dependent on the inducer, tissue of origin or cell type and even the duration of the senescent burden12. TGFβ is a pleiotropic cytokine commonly secreted by senescent cells14–16 and is capable of exerting both tumor-suppressive and tumor-promoting effects depending on the extracellular context and the recipient cell17. Although TGFβ can induce proliferative cell arrest and senescence16,18,19, it also activates non-canonical pathways, including interplay with AKT/mTOR, MAPK and p38/JNK signaling cascades20, leading to enhanced cancer cell proliferation, epithelial−mesenchymal transitions (EMTs) and metastasis21,22. The contextual nature of TGFβ in the SASP and its potential connection with chemotherapy-induced senescence remains incompletely understood.
Here we show that platinum-based chemotherapy induces senescent cells in human and murine lung adenocarcinomas and in HGSOC clinical samples. The SASP from cisplatin-induced senescent lung and ovarian cancer cells promotes tumor-supportive traits in neighboring cells, a phenomenon recapitulated in lung cancer xenograft, orthotopic and KrasG12V-driven mouse models. Mechanistically, the cisplatin-derived SASP is enriched in TGFβ ligands and activates TGFBR1–AKT/mTOR signaling, driving increased proliferation and malignant behavior that can be reversed by pharmacological TGFBR1 inhibition with galunisertib in vitro and in vivo.
Results
Cisplatin-induced senescence promotes malignant traits in recipient cancer cells through the secretion of SASP factors
Therapy-induced senescence can drive a chronic inflammatory niche in the tumor microenvironment, which has the potential to promote a variety of tumor-promoting activities12,23. To investigate the effects of chemotherapy-derived SASP in NSCLC, human lung adenocarcinoma A549 cells were subjected to cisplatin (15 μM), docetaxel (200 nM) and palbociclib (15 μM) treatment. Drug-induced senescence was confirmed by increased SA-β-gal staining, stable proliferation arrest and expression markers of senescence at RNA and protein (Fig. 1a,b and Supplementary Fig. 1a,b). Next, untreated A549 cells were exposed to conditioned medium (CM) from each type of therapy-induced senescent cells (Fig. 1c). Exposure to cisplatin-derived SASP resulted in increased cell proliferation and cell migration, compared to control CM and docetaxel-derived and palbociclib-derived SASPs, with exposed cells achieving an average 85.7% confluency at 42 hours in the presence of cisplatin-induced senescent CM versus 49.4% in control conditions (Fig. 1d, left). Cell tracking during a wound healing scratch assay showed that exposure to cisplatin-SASP also significantly increased A549 cell migration compared to all other conditions (Fig. 1d, right). The potential confounding effect of acidification or lower nutrient levels in CM was excluded by pH and glucose measurements (Supplementary Fig. 1c,d).
Fig. 1. Cisplatin-induced senescent A549 and L1475(luc) cells and their secretomes promote increased tumor growth in xenograft and orthotopic models of NSCLC.
a, Representative images of control and chemotherapy-treated A549 cells fixed and stained for SA-β-gal activity after 7 days of treatment. Scale bar, 100 μm. Right, proportion of SA-β-gal+ A549 cells for each treatment. b, Cell confluency over time of control and senescent A549 cells after 7 days of treatment, respectively. c, Schematic representation of experimental layout to test the effect of chemotherapy-derived SASP on non-senescent cells. A549 cells were treated with 15 μM CDDP, 150 nM docetaxel or 15 μM palbociclib for 7 days for senescence induction. Senescent cells were subsequently washed, and fresh medium was added and conditioned for 48 hours. CM was collected, centrifuged and added onto non-senescent A549 cells, which were subsequently assayed for phenotypic assessment. d, Left, A549 cell confluency (%) over a 72-hour incubation period exposed to CM from control or chemotherapy-induced senescent cells. Right, relative wound density of migrating A549 cells after scratching over a 66-hour period exposed to CM from control or chemotherapy-induced senescent cells. e, Left, representative images of A549 colonies formed after 10 days upon the exposure to the different CM from non-senescent and chemotherapy-induced senescent cells. Right, quantification of the number of colonies for each experimental condition. f, Left, representative images of A549 three-dimensional tumor spheres formed after 7 days of two-phase co-culture with control or chemotherapy-induced senescent cells. Right, quantification of sphere size in each experimental condition. g, Left, representative images of A549 spheres formed after 10 days upon the exposure to the different CM from non-senescent and chemotherapy-induced senescent cells. Right, quantification of the number of spheres for each experimental condition. Scale bar, 400 μm, top panels; 200 μm, bottom panels. h, OCR and SRC of non-senescent A549 cells at basal conditions and upon injection of oligomycin, FCCP, rotenone and antimycin A, measured after 72-hour incubation with CM from control or cisplatin-induced senescent A549 cells. i, L1475(luc) cell count relative to control condition of cells exposed to control or chemotherapy-induced senescent CM for 48 hours. j, Schematic representation of experimental layout using the subcutaneous xenograft model. In brief, animals were transplanted with either 4 × 106 untreated A549−mCherry+ cells, 4 ×106 untreated A549−mCherry+ cells and 1 × 106 cisplatin-induced senescent A549−GFP+ cells or 4 × 106 cisplatin-induced senescent A549−GFP+ cells alone in the flank. k, Tumor volume of xenografts over time. Data are shown as mean ± s.e.m. (n = 18 tumors per group). l, Fluorescent images of resected tumors at the end of the experiment. m, Schematic representation of experimental layout. In brief, untreated L1475(luc) cells were continuously exposed to the CM from control or cisplatin-induced senescent cells for 10 days prior to lung transplantation via tail vein injection. n, Representative images of L1475(luc) cells exposed to control and cisplatin-induced senescent CM prior to transplantation. Scale bar, 400 μm. o, Representative images of histological sections from lungs resected at 17 days after transplantation of lung cancer cells in each experimental condition. Scale bar, 1 mm. p, Representative images of luciferase activity in mice 17 days after transplantation with lung cancer cells exposed to control or cisplatin-induced senescent cell CM. q, Quantification of luciferase activity at day 17 after transplantation, relative to activity recorded on day 1 after cell transplantation (n = 14 per group). r, Survival curve of mice in each experimental group (n = 14 per group). All data for in vitro experiments are plotted as mean ± s.d. (n = 3), except for data in h, which are shown as mean ± s.d. from one representative experiment (out of three independent experiments performed). Statistical significance for in vitro assays was assessed using unpaired t-test or two-tailed one-way or two-way ANOVA, followed by Tukey’s multiple comparisons test. In vivo data in k and q are shown as mean ± s.e.m. Statistical significance for tumor volume was assessed using two-way ANOVA, followed by Tukey’s multiple comparisons test, and by two-tailed Student’s t-test for relative luciferase activity. Survival analysis was performed using the Kaplan−Meier method, and a two-sided log-rank test was conducted to determine statistical significance. Orthotopic model data represent three independent experiments. Images were created with BioRender.com. CDDP, cisplatin; ctrl, control; SEN, senescent; Untr., untreated.
We next showed that exposure to cisplatin-induced senescent CM for 10 days led to a 3.4-fold increase in a colony formation assay compared to control CM, whereas palbociclib-induced senescent CM had no effect, and docetaxel-derived SASP resulted in significantly fewer colonies (Fig. 1e). The number of spheres formed in low-attachment conditions was also significantly higher in the presence of cisplatin-derived SASP compared to all other conditions (Fig. 1f). Notably, these spheres also exhibited irregular shapes as well as the formation of protrusions, which, together with the enhanced migration properties observed after wound healing, suggested the acquisition of an invasive phenotype consistent with an EMT. To determine the effect of the direct interaction between senescent and untreated tumor cells, we cultured A549 cells in a three-dimensional spheroid system alone or with chemotherapy-induced senescent cells. Notably, three-dimensional co-culture with cisplatin-induced senescent A549 cells resulted in the development of larger spheres compared to all other conditions (Fig. 1g). Mitochondrial stress metabolic assays on A549 cells after incubation with CM with measurement of oxygen consumption rate (OCR) and spare respiratory capacity (SRC) showed increased maximal respiration in cells exposed to cisplatin-derived SASP compared to control (Fig. 1h). These data suggest metabolic rewiring with an increased ability to respond to high ATP demand and, therefore, improved cellular endurance during experimentally induced stress. Incubation of A549 cells with docetaxel-induced or palbociclib-induced senescent CM did not change respiratory ability (Supplementary Fig. 1e). Interestingly, bioenergetic profiling of control and chemotherapy-induced senescent A549 cells revealed that senescent cells also harbor an increased maximal respiratory capacity compared to control cells, with cisplatin-induced senescent cells showing the highest OCR levels among the three anticancer therapies (Supplementary Fig. 1f).
We next aimed to validate our findings using the primary KrasG12D/WT;p53−/− murine L1475(luc) cell line, induced to cellular senescence by cisplatin (3 μM), docetaxel (100 nM) or palbociclib (20 μM) treatments (Supplementary Fig. 2a–d). Proliferation of untreated L1475(luc) cells was increased by cisplatin-induced senescent CM compared to all other conditions (Fig. 1i and Supplementary Fig. 2e). Consistent with our findings in A549 cells, culture of L1475(luc) with cisplatin-induced senescent CM for 10 days resulted in a higher colony formation and significantly enhanced maximal respiration (Fig. 1e,h and Supplementary Fig. 2f,g). In contrast to A549 cells, L1475(luc) cells exposed to docetaxel-induced and palbociclib-induced senescent CM also showed increased OCR levels, albeit to a lower extent compared to cisplatin-derived SASP exposure (Supplementary Fig. 2h). Bioenergetic analysis of control and senescent L1475(luc) cells showed trends similar to those observed in A549 cells, except for docetaxel-induced senescent cells, which presented a respiratory profile similar to control cells (Supplementary Fig. 2i). To exclude confounding effects from apoptosis, we performed Annexin V assays during media conditioning of both control and cisplatin-induced senescent A549 and L1475(luc) cells, which demonstrated only minimal levels of apoptosis from CM (Supplementary Fig. 3a–c). Annexin V assays were also tested during the initial phase of chemotherapy exposure (Supplementary Fig. 3d–i).
To explore whether the cisplatin analog carboplatin also resulted in similar phenotypes, we first determined the induction of senescence in A549 cells upon 7-day treatment with 7.5 μM carboplatin (Supplementary Fig. 4a–c). Carboplatin-derived SASP induced a significant increase in colony formation in untreated A549 cells compared to control CM (Supplementary Fig. 4d). The generality of these observations was confirmed in an HGSOC cancer model by exposing PEO4 cells to cisplatin and carboplatin treatments (2.5 μM and 10 μM, respectively) (Supplementary Fig. 4e–g). Notably, both cisplatin-derived and carboplatin-derived SASPs also increased proliferation rates of untreated PEO4 cells versus control (Supplementary Fig. 4h). Similar proliferation effects of platinum-induced SASP were demonstrated in lung squamous cell carcinoma LK-2 and NCI-H226 cells (Supplementary Fig. 4i–p).
Altogether, our in vitro assessments show that platinum-based therapy results in treatment-specific senescence SASP that promoted tumor cell proliferation, migration, sphere formation and enhanced respiratory endurance in multiple lung and ovarian cancer cells.
Cisplatin-induced senescent A549 and L1475(luc) cells support increased tumor growth in xenograft and orthotopic models of NSCLC
To test in vivo effects of cisplatin-induced senescence, we subcutaneously co-transplanted cisplatin-induced senescent A549−GFP+ cells with untreated A549−mCherry+ cells and analyzed the growth of the xenografts formed over time (Fig. 1j). As shown in Fig. 1k,l, relative tumor volume was significantly higher when tumor cells were co-transplanted with senescent cells as compared to untreated cells transplanted alone. Notably, by day 21, co-transplanted tumors (untreated + senescent cells) were more than twice the volume of control tumors (average tumor volume of co-transplanted: 325.44 ± 116.82 mm3; control: 149.00 ± 60.44 mm3; N = 18 tumors per group) (Fig. 1k,l). Transplantation of senescent cells alone resulted in xenograft recession, confirming that the effect in co-transplanted tumors is likely driven through paracrine support from senescent cells and not because of senescence escape. Mean tumor weight of co-transplanted xenografts (251.11 ± 87.30 mg, N = 18) was also significantly higher than untreated tumors (156.67 ± 78.74 mg, N = 18) and senescent tumors (37.22 ± 17.08 mg, N = 18) (Extended Data Fig. 1a). Histological analyses confirmed decreased levels of the proliferative marker Ki67 and increased expression of p21 in senescent xenografts compared to untreated and co-transplanted tumors (Extended Data Fig. 1b). Representative pictures of the resected tumors at experiment completion are shown in Fig. 1l.
Extended Data Fig. 1. Cisplatin-induced senescence results in lung tumour promoting effects that are exacerbated during aging.
a. Mean tumour weight at experimental endpoint resected from mice transplanted with untreated A549 cells, a combination of untreated and cisplatin-induced senescent A549 cells or senescent-cells only (n = 18 tumours per group). b. Representative histological images of resected xenograft from each experimental group, stained for H&E, Ki67 and p21 expression. Scale bars: 1 mm and 250 μm for magnifications. c. Relative cell count of L1475(luc) cells exposed to control or cisplatin-induced senescent CM for 10 days. Assay was performed in parallel to orthotopic transplantation and cells were grown in normal media (not CM) for 48 h before cell quantification by flow cytometry (n = 3, biological repeats). d. Schematic representation of the experimental layout. e. Quantification of luciferase activity at day 14 post-transplantation, relative to activity recorded on day 1 after cell transplantation (n = 7 mice in control, n = 8 cisplatin, n = 9 palbociclib group). f. Survival curve of mice for the indicated experimental groups. g. Schematic representation of experimental layout. Young (8-10-week old) and aged (42-45-week old) animals were subjected to two cycles of 1.5 mg/kg body weight CDDP treatment. Mice were then transplanted with L1475(luc) in the lungs via tail-vein injection; tumour burden was assessed twice a week by bioluminescence imaging and survival was determined as the time from tumour transplantation until the onset of moderate signs of disease. h. Fold change gene expression of Cdkn1a, Il6, Tgfb1 and Tgfb2 in samples from each experimental condition (n = 3 samples from independent mice per group). i. Quantification of luciferase activity at day 7 and 14 post-transplantation, relative to activity recorded on day 1 after cell transplantation (n = 4 mice per group). j. Representative images of luciferase activity in mice 7 and 14 days after transplantation with lung cancer cells. k. Survival curve of mice for the indicated experimental groups. l. Weight relative to baseline (before first CDDP dose) in each experimental group (young, n = 6; middle-aged, n = 4) over time. Data are shown as mean ± SEM (a, e, i) and as mean ± SD (c, h, l). One-way ANOVA followed by Tukey’s multiple comparisons test was used to assess statistical significance for luciferase activity (e), tumour weight (a) and relative weight (l) over time in mice. Two-tailed Student’s t test was applied to assess significance for relative cell count (c) and relative luciferase activity (i). Survival analysis was performed using the Kaplan-Meier method and a two-sided log-rank test was conducted to determine statistical significance. Images created with BioRender.com.
We next investigated the paracrine effects of cisplatin-induced senescence using a KrasG12D/WT;p53−/− orthotopic murine model. L1475(luc) cells were exposed to control or cisplatin-induced senescent CM for 10 days and were subsequently transplanted in the lungs of C57BL/6 mice via tail vein injection (Fig. 1m). Tumor burden was measured over time using D-luciferin administration and bioluminescence imaging. As expected, cells exposed to cisplatin-derived SASP in culture proliferated faster than those exposed to control CM (Fig. 1n and Extended Data Fig. 1c). Transplanted cells that had been exposed to cisplatin-derived SASP resulted in a significantly greater luciferase activity in the lungs, relative to day 1 after injection (Fig. 1o–q), a higher number and greater size of tumor foci and a decrease in median survival by 30% (23 days compared to 32.5 days in mice transplanted with tumors conditioned with control CM; Fig. 1r). By contrast, parallel transplantation experiments using exposure to the palbociclib-induced senescent secretome did not alter tumor burden and resulted in slightly increased survival (Extended Data Fig. 1d–f).
Chemotherapy is known to be gerontogenic and can induce senescence affecting both tumor and non-cancer tissues24, whereas senescence promotes adverse effects of chemotherapy and cancer relapse25. To investigate the potential interplay between age-related senescence burden in normal tissues26 and genotoxic stress, young and middle-aged mice were subjected to two cycles of cisplatin treatment, and then L1475(luc) cells were transplanted orthotopically in the lungs (Extended Data Fig. 1g). Reverse transcription quantitative polymerase chain reaction (RT−qPCR) analysis of RNA extracted from whole lungs at day 14 after treatment showed increased gene expression levels of the senescence markers Cdkn1a, Il6 and Tgfb2 in cisplatin-treated animals compared to the vehicle group (Extended Data Fig. 1h). This observation is consistent with our previous experiments showing that cisplatin treatment after orthotopic transplantation of L1475(luc) cancer cells also results in the onset of senescence in the lung27. Tumor burden analysis by bioluminescence imaging was significantly higher in middle-aged animals (Extended Data Figs. 1i,j). Of note, middle-aged mice had a 31% decrease in median survival (14.5 days compared to 21 days in young individuals; Extended Data Fig. 1k). Indirect measures of chemotherapy-related adverse effects, including change in body weight, did not show major differences between the groups during cisplatin administration. However, middle-aged mice rapidly underwent marked weight loss after tumor transplantation compared to young mice (Extended Data Fig. 1l). These results suggest that young individuals present higher tolerability toward cisplatin treatment and that aging exacerbates tumor progression.
Taken together, our in vivo findings confirm that the induction of senescence in response to cisplatin treatment in mice exerts detrimental effects in a paracrine fashion, boosting lung cancer progression and shortening lifespan.
Transcriptomic and proteomic analyses reveal chemotherapy context-dependent differences in SASP signatures and that TGFβ ligands are overrepresented in cisplatin-derived SASP
We next performed RNA sequencing (RNA-seq) analyses of control, cisplatin-induced, docetaxel-induced and palbociclib-induced senescent A549 cells that confirmed increased expression of known senescence mediators such as TP53 (p53) and CDKN1A (p21WAF1/Cip1) and reduced expression of genes promoting cell cycle progression (including E2F, CDC and CDK genes) and DNA replication (PCNA, POL and MCM), consistent with induction of stable cell cycle arrest (see scaled expression profile in Fig. 2a and Extended Data Fig. 2a). Gene set enrichment analysis (GSEA) also showed that cisplatin-treated, docetaxel-treated and palbociclib-treated A549 cells were all positively enriched for signatures of senescence and negatively enriched for cell cycle and DNA replication pathways compared to vehicle-treated cells (Extended Data Fig. 2b). As expected, cisplatin-treated cells displayed much greater transcription of mismatch repair and DNA repair-related genes. Analysis of treatment-specific senescence expression profiles showed that docetaxel-induced senescent cells had the highest number of uniquely upregulated genes (N = 1,961) versus 751 genes in cisplatin-induced and 363 genes in palbociclib-induced senescent cells (Fig. 2b).
Fig. 2. Transcriptomic and proteomics analyses reveal different transcriptional and SASP signatures of chemotherapy-induced senescent cells and highlight TGFβ ligands as potential candidate drivers of malignant traits in cisplatin-derived SASP.
a, Heatmap displaying expression z-scores of most significantly altered gene expression changes in cellular senescence pathway (hsa04218) and their hierarchical clustering in control and cisplatin-induced, docetaxel-induced and palbociclib-induced senescent A549 cells. b, Venn diagram of the number of genes significantly upregulated in cisplatin-induced, docetaxel-induced and palbociclib-induced senescent A549 cells versus control. c, Heatmap of expression z-scores of selected SASP genes in control and cisplatin-induced, docetaxel-induced and palbociclib-induced senescent A549 cells. d, Top differentially enriched pathways in the MSigDB Hallmark 2020 collection, sorted by combined score ranking, in cisplatin-induced, docetaxel-induced and palbociclib-induced A549 senescent cells versus control. e, Left, proportion of EdU high-intensity A549 cells upon 72 hours of exposure to different SASP ligands using MEMA technology, ranked left to right from lowest induced proportion to highest. Right, representative immunofluorescence images of MEMA wells containing A549 cells upon exposure to PBS, TGFβ1 and TGFβ2 recombinant ligands, depicting DAPI and KRT5 staining and EdU incorporation. The experiment was performed with 12–15 technical replicates per ligand combination. The central line within each box represents the median of the nuclei-gated EdU-positive proportion for the indicated ligand combinations. The box spans from the first quartile (Q1) to the third quartile (Q3) encompassing the interquartile range (IQR), representing the middle 50% of the data. The whiskers extend from the box to the minimum and maximum data points within 1.5 times the IQR from the box edges. Scale bar, 150 μm. f, Fragments per kilobase of transcript per million mapped reads (FPKM) of TGFB1, TGFB2 and TGFB3 ligands in control and cisplatin-induced, docetaxel-induced and palbociclib-induced senescent A549 cells. g, Western blot analysis of whole-cell extracts of control and cisplatin-induced, docetaxel-induced and palbociclib-induced senescent A549 cells. h, Average concentration in CM of TGFβ1, TGFβ2 and TGFβ3 free ligands (activated) in each experimental condition measured by ELISA. Data are shown as mean ± s.d. (n = 3 independent biological repeats). Statistical significance was assessed by two-tailed one-way ANOVA, followed by Tukey’s multiple comparisons test. Ctl, control; Veh, vehicle.
Extended Data Fig. 2. Transcriptomic profiles of A549 and L1475(luc) cells upon distinct chemotherapy treatments.
a. Heatmap displaying expression z-scores of most significantly altered gene expression changes and their hierarchical clustering in cell cycle (top) and DNA replication pathways (bottom) of control and cisplatin-, docetaxel- and palbociclib-induced senescent A549 cells. b. GSEA of KEGG Senescence, KEGG Cell Cycle and KEGG DNA replication sets in cisplatin-, docetaxel- and palbociclib-induced senescent A549 cells versus untreated A549 cells. c. Heatmap displaying expression z-scores of most significantly altered gene expression changes and their hierarchical clustering in the cellular senescence pathway (mmu04218) of control and cisplatin-, docetaxel- and palbociclib-induced senescent L1475(luc) cells. d. Venn diagram of the number of genes significantly upregulated in cisplatin-, docetaxel- and palbociclib-induced senescent L1475(luc) cells vs control. e. Heatmap of expression z-scores of selected SASP genes in control and cisplatin-, docetaxel- and palbociclib-induced senescent L1475(luc) cells. f. Fold change TGFB1/Tgfb1, TGFB2/Tgfb2 and TGFB3/Tgfb3 gene expression levels in control and chemotherapy-treated A549 and L1475(luc) cells.
To further dissect gene expression profiles, we next investigated the expression of known SASP factors in vehicle-treated and chemotherapy-treated A549 cells (Fig. 2c). As expected, we observed a general increase in the transcription of inflammatory ligands and proteases in senescent A549 cells in most of the clusters, such as genes encoding for IL-6, IL-1B, CCL2 and MMPs. Clusters 3 and 8 appeared particularly enriched in cisplatin-induced senescent cells, which included genes encoding for TGFB2, CSF2RB, PDGFA and LIF. We performed pathway enrichment analyses of significantly differentially expressed genes in each type of induced senescence against predefined sets from the Molecular Signatures Database (MSigDB). This revealed a significant enrichment in metabolic pathways across the three types of senescence (Fig. 2d). Notably, cisplatin-induced senescent cells were uniquely enriched for IL-2/STAT5 and TGFβ pathways.
We also observed distinct transcriptomic profiles for murine L1475 cells with cisplatin-induced and docetaxel-induced senescence (Extended Data Fig. 2c,d). Similar to human A549 cells, cisplatin-treated mouse L1475 cells showed increased expression of senescence mediators such as Cdkn1a (p21WAF1/Cip1), with reduced expression of genes promoting cell cycle progression (such as E2f, Cdc and Cdk genes) and DNA replication (such as Pol, Dna2 helicase and Ssbp1 genes), and had the highest number of uniquely upregulated genes as compared to docetaxel-treated cells (Extended Data Fig. 2d). Characterization of the expression of known SASP factors confirmed a general increase in the transcription of inflammatory ligands and proteases in senescent murine cells, such as Il1a, Il6, Tgfb2, Tgfb3 and Mmp3 genes (Extended Data Fig. 2e).
To further identify the specific SASP factors driving increased proliferation in A549 cells, we used a microenvironment microarray (MEMA) platform to probe the functional effects of multiple combinations of individual factors and cytokines in an unbiased high-throughput manner28,29. Notably, analysis of 64 human SASP-related ligands showed that TGFβ2 and TGFβ1 led to the highest A549 proliferation rate of all factors tested in the array (Fig. 2e). Together with the gene expression data, this suggested that increased secretion of TGFβ ligands could be mediating the tumor-promoting effects observed in cells that are exposed to cisplatin-derived SASP. Thus, we next confirmed that cisplatin-induced senescent cells present an increased expression of TGFB1, TGFB2 and TGFB3 ligands in our RNA-seq data (Fig. 2f), which was further validated by RT−qPCR analysis in both A549 and L1475(luc) cells (Extended Data Fig. 2f). Western blot examination of control and senescent A549 cell extracts revealed increased protein levels of inactive and active TGFβ1 in cisplatin-induced senescent cells versus control A549 cells, whereas free TGFβ2 and TGFβ3 expression was decreased (Fig. 2g). Biologically active TGFβ ligands are tightly regulated in producing cells, where they are normally stored in a latent form, whereas the efficient secretion, folding and extracellular deposition requires the complex regulation of multiple steps30. To determine whether the lower intracellular levels of the ligands observed by western blot were due to an increased secretion compared to control cells, we analyzed the levels of the active form in the SASP by ELISA. Notably, this confirmed that the CM of cisplatin-induced senescent cells is significantly enriched in active TGFβ1, TGFβ2 and TGFβ3 compared to control and docetaxel-induced and palbociclib-induced senescent cells (Fig. 2h).
Together, these data indicate that chemotherapeutic treatment results in markedly distinct senescent transcriptional phenotypes in human and murine lung adenocarcinoma cells and suggest the enriched secretion of TGFβ ligands in cisplatin-induced senescent CM as potential SASP candidates promoting increased proliferation rates of recipient lung cancer cells.
TGFBR1-driven activation of AKT/mTOR pathway orchestrates the induction of malignant NSCLC traits upon exposure to cisplatin-derived SASP
Bioactive TGFβ cytokines secreted to the extracellular space activate downstream signaling responses in recipient cells by driving the dimerization of TGFBR1 and TGFBR2, receptor serine and threonine kinases, respectively17,31. To test whether TGFBR1 was involved in the observed tumor-promoting effects derived from the exposure to cisplatin-induced SASP, A549 and L1475(luc) cells were grown in the presence of control and cisplatin-derived CM and treated with galunisertib, a TGFBR1 inhibitor. Cell confluency and relative cell count analysis at 72 hours revealed that TGFBR1 inhibition significantly reduced the proliferation of cells after exposure to cisplatin-derived SASP (Fig. 3a and Extended Data Fig. 3a). Galunisertib effects were not mediated by altered cell survival or mitochondrial function, as shown by measuring apoptosis levels by Annexin V assays (Fig. 3b) and OCR by Seahorse assays (Extended Data Fig. 3b), respectively. We next observed that galunisertib treatment also reduced the number of colonies formed after exposure to cisplatin-induced senescent A549 and L1475(luc) CM for 10 days compared to cisplatin-derived SASP exposure alone (Fig. 3c and Extended Data Fig. 3c). In addition, TGFBR1 inhibition reduced the number of spheres and sphere size observed in three-dimensional co-cultures of A549 and L1475(luc) cells with cisplatin-induced senescent cells (Fig. 3d). To determine if the effects observed with galunisertib treatment were specific to the inhibition of TGFBR1, we silenced the expression of TGFBR1 and Tgfbr1 by generating knockdown A549 and L1475(luc) cell lines (Extended Data Fig. 3d). In agreement with the effects observed with galunisertib, reduced expression of TGFBR1 and Tgfbr1 resulted in a decreased impact on enhanced proliferation when A549 and L1475(luc) cells were exposed to cisplatin-derived CM compared to parental and scrambled transduced cells (Extended Data Fig. 3e). Of note, this effect is partially rescued when using a doxycycline-dependent inducible and reversible short hairpin RNA (shRNA) interference assay for TGFBR1 in A549 cells exposed to cisplatin-derived CM (Fig. 3e,f). In addition, we also observed that treatment of A549 cells with recombinant human TGFβ1 (rhTGFβ1) ligand significantly increased cell growth (Extended Data Fig. 3f), further validating our observations obtained with the high-throughput MEMA platform. These results, therefore, confirm that the activation of TGFBR1 in recipient cells, likely through TGFβ ligands secreted by cisplatin-induced senescent cells, is specifically responsible for the increased tumor growth driven by this SASP.
Fig. 3. TGFBR1-driven activation of AKT/mTOR pathway orchestrates the induction of malignant traits upon exposure to cisplatin-derived SASP.
a, Cell confluency fold change of A549 and L1475(luc) cells exposed to control and cisplatin-induced senescent CM with and without 50 μM galunisertib or 0.5 μM staurosporine treatment for 72 hours. b, Annexin V measurements per cell confluence of A549 and L1475(luc) cells exposed to control and cisplatin-induced senescent CM with and without 50 μM galunisertib or 0.5 μM staurosporine treatment for 72 hours. c, Quantification of A549 and L1475(luc) colonies formed relative to control upon exposure to control and cisplatin-induced senescent CM for 10 days with and without 50 μM galunisertib treatment. d, Left, representative images of A549 and L1475(luc) tumor spheres formed after 7 days of two-phase co-culture with control or chemotherapy-induced senescent cells with and without 50 μM galunisertib treatment. Middle, number of spheres formed in each condition, relative to co-culture with control cells treated with vehicle. Right, average sphere size in each condition. For average sphere size, a total of 150 spheres from three independent biological repeats were measured. e, Western blot analysis of whole extracts of A549 cells upon lentiviral transduction of shRNAs for (1 μg ml−1) doxycycline-dependent TGFBR1 knockdown, as indicated. f, Cell counts relative to scrambled shRNA control (GEPIR) for A549 cells exposed to control and cisplatin-induced senescent CM for 7 days. Bars are represented at the indicated conditions of lentiviral transduction of shRNAs for doxycycline-dependent TGFBR1 knockdown. g, Heatmap of pixel intensity z-score of each kinase phosphorylation listed upon 30-minute exposure to control and cisplatin-induced senescent CM in A549 cells. h, Western blot analysis of whole extracts of A549 cells exposed to control and cisplatin-induced senescent CM for 30 minutes with a range of concentrations of galunisertib, as indicated. i, Western blot analysis of whole extracts of A549 and L1475(luc) cells exposed to control and cisplatin-induced senescent CM for 30 minutes with and without 50 μM galunisertib and 1 μM rapamycin. j, Cell confluency and cell count of A549 and L1475(luc) cells exposed to control and cisplatin-induced senescent CM for 48 hours upon 1 μM rapamycin treatment, relative to cells exposed to control CM with vehicle. k, Heatmap displaying expression z-scores of gene expression changes in cell cycle, PI3K and mTOR pathways and their hierarchical clustering in L1475(luc) cells exposed for 12 hours to control and cisplatin-induced senescent cells CM. All data represent mean ± s.d. (n = 3 independent biological repeats). Heatmaps represent average z-score of three independent biological repeats. Statistical significance in a−d, f and j was assessed by two-tailed one-way ANOVA or two-way ANOVA, followed by Tukey’s multiple comparisons test. ctl, control; NS, not significant.
Extended Data Fig. 3. TGFBR1-driven activation of AKT/mTOR pathway orchestrates the induction of malignant traits upon exposure to cisplatin-derived SASP.
a. Cell confluency and relative cell count of A549 and L1475(luc) cells exposed to control and cisplatin-induced senescent CM with and without 50 μM galunisertib treatment for 48 h. b. OCR of A549 cells exposed or not exposed to 50 μM galunisertib in basal conditions and upon the injection of mitochondria-targeting drugs. c. Representative images of A549 and L1475(luc) colonies formed upon exposure to control and cisplatin-induced senescent CM for 10 days with and without 50 μM galunisertib treatment. d. Fold change TGFβR1/Tgfbr1 gene expression of scrambled- and shTGFβR1-A549 cells (left) and scrambled- and shTgfbr1-L1475(luc) cells (right). e. Cell confluency and relative cell count of A549 (left) and L1475(luc) (right) original, scrambled and shTGFBR1/shTgfbr1 cells upon exposure to control and cisplatin-induced senescent CM for 48 h. f. Normalised cell confluency over time of A549 cells treated with human recombinant TGFβ1 ligand. Data in panels a, d, e, f represent mean ± SD (n = 3 independent biological repeats). Statistical significance was determined by two-tailed, one- or two-way ANOVA followed by Tukey’s multiple comparisons test, and by two-tailed Student’s t-test. g. Representative pictures of human phospho-kinase array panels of whole-protein extracts from A549 cells exposed for 30 min to either control- or cisplatin-induced senescent CM with and without 50 μM galunisertib treatment. h. Representative images depicting A549 cell confluency from cells exposed to control- or cisplatin-induced senescent CM for 48 h treated with either vehicle, 50 μM galunisertib or 1 μM rapamycin. Images are representative of 1 out of 3 independent experiments carried out with similar results. Scale bar = 300 μm as depicted in the image.
Intriguingly, TGFβ cytokines can signal through the activation of several different pathways in recipient cells, which, paradoxically, can result in both tumor-suppressive and tumor-promoting effects17,32. To uncover the mechanism involved in the transduction of TGFBR1 activation upon exposure to cisplatin-induced senescent CM, we explored protein phosphorylation changes in A549 recipient cells exposed to control and cisplatin-induced senescent CM with and without galunisertib treatment. Phospho-kinase array analysis revealed an increased phosphorylation of AKT1/AKT2/ATK3 at residue S473, one of the activating sites of this kinase, after exposure of cells to cisplatin-induced senescent CM, whereas TGFBR1 inhibition through galunisertib treatment reduced levels of this phosphorylation (Fig. 3g and Extended Data Fig. 3g). This suggests that binding to TGFBR1 upon exposure to cisplatin-derived SASP mediates the downstream activation of AKT and cell proliferation via the AKT/mTOR/p70S6K signaling cascade. As observed in our phospho-kinase arrays, we detected an increase in phospho-AKT S473 in both A549 and L1475(luc) cells exposed to cisplatin-induced CM by western blotting, which was diminished upon galunisertib treatment in a dose-dependent manner (Fig. 3h,i). Notably, we observed the same trend in the phosphorylation of p70S6K, suggesting that phosphorylation of AKT results in the activation of p70S6K via mTOR signaling (Fig. 3h,i). We then treated CM recipient cells with the mTOR inhibitor rapamycin33 and observed that increased phosphorylation of AKT at S473 remained unchanged in cells after exposure to cisplatin-derived SASP and rapamycin but with reduced phosphorylation of p70S6K (Fig. 3i). To further validate the effect of this pathway on the phenotypic response observed upon exposure to the SASP, we exposed A549 and L1475(luc) cells to control and cisplatin-induced senescent CM and observed that rapamycin also hampered the effect on proliferation driven by cisplatin-derived SASP (Fig. 3j and Extended Data Fig. 3h).
We next performed RNA-seq analyses of L1475(luc) cells exposed to control and cisplatin-derived SASP for 12 hours. This confirmed the upregulation of genes involved in the cell cycle in cisplatin CM-exposed cells (such as E2f, Cdc and Cdk genes) (Fig. 3k), consistent with the effects observed in our in vitro and in vivo analyses. In addition, we explored changes at the transcriptional level in PI3K/Akt and mTOR pathways and observed a marked increase in the expression of genes involved in both pathways in cells exposed to cisplatin-derived SASP compared to control (Fig. 3k), validating the activation of these pathways upon exposure to this specific SASP in murine lung adenocarcinoma cells.
Together, these results demonstrate that the exposure of lung cancer A549 and L1475(luc) cells to the SASP derived from cisplatin-induced senescent cells results in the TGFBR1-mediated activation of the AKT/mTOR/P70S6K pathway, leading to increased cell proliferation.
TGFBR1 pharmacologic inhibition effectively blocks pro-tumorigenic effects derived from exposure to cisplatin-induced senescence in vivo
To test these mechanistic hypotheses in vivo, we subcutaneously transplanted cisplatin-induced senescent A549 cells together with untreated A549 cells and subjected mice to galunisertib and senolytic treatment with ABT-737 during tumor development (Fig. 4a). As expected, the ablation of senescent cells in the tumors through senolytic treatment prevented the increase in tumor growth in co-transplanted xenografts compared to co-transplanted tumors treated with vehicle only (Fig. 4b,c). Of note, galunisertib treatment in mice also significantly blocked the tumor-promoting effect derived from co-transplantation with senescent cells in the xenografts (Fig. 4b,c), confirming the efficiency of inhibiting TGFBR1 to prevent the deleterious effects derived from cisplatin-induced senescent SASP in vivo. No impact on the progression of control A549 tumor xenografts was observed with either ABT-737 or galunisertib treatment, suggesting that the effect is driven by transplanted cisplatin-induced senescent A549 cells.
Fig. 4. TGFBR1 inhibition effectively prevents enhanced proliferation derived from the exposure of cisplatin-induced senescent SASP in xenograft and orthotopic models of lung adenocarcinoma.
a, Schematic representation of experimental layout. In brief, animals were transplanted subcutaneously with either 4 × 106 untreated A549 cells or 4 × 106 untreated A549 combined with 1 × 106 cisplatin-induced senescent A549 cells. At 5 days after transplantation, animals were subjected to 150 mg kg−1 body weight galunisertib treatment, 25 mg kg−1 body weight ABT-737 or vehicle at timings depicted. Tumor volume was measured twice a week, and tumors were resected at day 30 after initiation of treatment. b, Tumor volume over time for each experimental condition (n = 12 independent tumors per group from six mice transplanted with one xenograft per flank). c, Tumor weight of each experimental group upon resection of tumors at day 30. d, Schematic representation of experimental layout. Animals were transplanted with L1475(luc) in the lungs via tail vein injection, and, after 3 days, they were subjected to 1.5 mg kg−1 body weight CDDP, 150 mg kg−1 body weight galunisertib, a combination of both or vehicle at the frequency shown in the timeline. Tumor burden was assessed twice a week by bioluminescence imaging, and survival was determined as the time from tumor transplantation until the onset of advanced signs of disease. e, Representative images of luciferase activity in mice at days 2 and 10 after initiation of treatment. f, Absolute radiance and relative luciferase activity recorded in each experimental group at days 0, 8 and 15 after initiation of treatment (n = 9 per group). g, Survival curve of mice in each experimental condition. h, Weight of mice (n = 10 per group) subjected to CDDP and CDDP + galunisertib treatment over time. Data are shown as mean ± s.d. Statistical significance was assessed by ANOVA test using Welch’s correction for unequal variances in luciferase activity measurements and by two-way ANOVA followed by Bonferroni’s multiple comparison tests in fold change tumor volume (b) and weight analyses (c). Survival analysis (g) was performed using the Kaplan−Meier method, and a two-sided log-rank test was conducted to determine statistical significance. Statistical significance in weight differences (h) was assessed by two-sided one-way ANOVA, followed by Tukey’s comparisons test. Images were created with BioRender.com. CDDP, cisplatin; ctl, control.
We next investigated the effects of TGFBR1 inhibition of cisplatin-induced senescence CM in the KrasG12D/WT;p53−/− orthotopic murine model. L1475(luc) cells were exposed to cisplatin-induced senescent CM with or without galunisertib for 10 days and subsequently transplanted in the lungs of C57BL/6 mice via tail vein injection (Extended Data Fig. 4a). Note that, at the tested concentrations, galunisertib alone had no effect, or a negligible effect, on the proliferation of both L1475(luc) and A549 cells (Fig. 3a,c,d). Analysis of relative luciferase signal in mice at day 16 after transplantation showed a significantly decreased tumor burden in animals transplanted with cells simultaneously exposed to cisplatin-derived SASP and galunisertib (Extended Data Fig. 4b,c). Notably, we observed an increased survival in mice transplanted with tumors that had been simultaneously pretreated with the Tgbfr1 inhibitor during the exposure to cisplatin-induced senescent CM (Extended Data Fig. 4d).
Extended Data Fig. 4. Cisplatin treatment induces tumour senescence in orthotopically-transplanted lung cancer mouse models and galunisertib co-treatment reduces tumour burden.
a. Schematic representation of experimental layout. Briefly, L1475(luc) cells were exposed to cisplatin-induced senescent CM continuously for 10 days and treated with either vehicle or 50 μM galunisertib. Pre-conditioned cells were subsequently transplanted in the lungs of mice via tail-vein injection. Tumour burden was then assessed twice a week by bioluminescence imaging and survival was determined as the time from tumour transplantation until the onset of moderate signs of disease. b. Relative luciferase activity at day 16 of mice transplanted with cells pre-conditioned to cisplatin-derived CM with and without galuniseritb (n = 6). c. Representative images of luciferase activity at day 13 post-transplantation from each experimental group. d. Survival curve of mice from each experimental group. e. Schematic representation of experimental layout. Animals were transplanted with L1475(luc) cells in the lung via tail-vein injection, and after 3 days, they were subjected to 1.5 mg/kg body weight CDDP, 150 mg/kg body weight galunisertib, a combination of both or vehicle as shown in the timeline. Lungs were resected at day 7 post-transplantation and RNA was extracted from whole lungs. f. Representative histological images for SA-β-gal activity (top) and SenTraGor (GL13) staining (bottom) in the indicated experimental groups (n = 10 representative images of lung tumours from three independent mice per experimental group). Red arrows point to examples of GL13-positive cells. g. Quantification of SenTraGor (GL13)+ cells per total cells in lung tissues from vehicle, cisplatin, galunisertib and co-treated mice. Scale bars: 250 μm and 50 μm, as depicted in the Figure. h. Fold change gene expression of Cdkn1a, Il6, Tgfb1 and Tgfb2 in samples from each experimental condition (n = 3 samples from independent mice). Statistical significance was assessed by one-way ANOVA followed by Tukey’s multiple comparisons test (g, h), and by two-tailed Student’s t-test (b). Survival analysis was performed using the Kaplan-Meier method and a two-sided log-rank test was conducted to determine statistical significance. Images created with BioRender.com.
Altogether, the above results provide in vivo evidence that TGFBR1 signaling drives tumor progression through cisplatin-induced senescent paracrine effects in human and murine lung cancer models.
Cisplatin and galunisertib concomitant treatment reduces tumor burden and significantly enhances survival in murine models of lung cancer
We next aimed to determine whether combination treatments to prevent the deleterious effects of senescence could be an efficient strategy to improve lung cancer treatment. Using the KrasG12D/WT;p53−/− orthotopic murine model, lung tumor-bearing mice received either single treatment with cisplatin or galunisertib or a regimen combining the two pharmacologic drugs (Fig. 4d). Cisplatin treatment of L1475(luc)-transplanted mice resulted in increased SA-β-gal activity and SenTraGor (GL13)-positive staining (alternative to SA-β-gal activity but suitable for formalin-fixed material) as well as expression of other biomarkers of senescence (for example, p21 and p16) that correlated with increased phospho-AKT levels in close vicinity in the same or adjacent lung tumor areas (Extended Data Fig. 4e–g and Supplementary Figs. 5 and 6a). In addition, RT−qPCR analysis of RNA extracted from whole lungs after 7 days of treatment revealed increased gene expression levels of the senescence marker Cdkn1a as well as Tgfb1 and Tgfb2 in cisplatin-treated animals compared to the vehicle group (Extended Data Fig. 4h).
Single-drug treatments did not decrease luciferase activity signal compared to vehicle treatment, but the combined cisplatin and galunisertib treatment had decreased tumor burden as compared to the other experimental groups, although this was not statistically significant (Fig. 4e,f). Notably, cisplatin and galunisertib co-treatment significantly improved median survival by 33% (20 days compared to 15 days in vehicle-treated mice; Fig. 4g). We also observed that animals receiving the cisplatin and galunisertib combination displayed improved tolerability and had decreased weight loss and fluctuation compared to cisplatin-only-treated animals (Fig. 4h). No galunisertib-specific toxicity was seen at the doses tested, including no evidence of cardiotoxicity (Supplementary Fig. 6b,c).
Finally, we used a more physiologically relevant model of KRAS-driven lung adenocarcinoma, driven by endogenous expression of oncogenic KrasG12V (ref. 34). Tumor-bearing KrasFSFG12V/+;p53frt/frt mice34,35 were treated with four cycles of cisplatin-based chemotherapy, alone or in combination with galunisertib treatment (Fig. 5a). The combined administration of cisplatin with the TGFBR1 inhibitor resulted in a significantly decreased fold change in tumor volume based on computed tomography imaging of murine lungs compared to animals treated with vehicle or single treatments (Fig. 5b,c and Extended Data Fig. 5a). Histological analyses showed accumulation of p16INK4A and SenTraGor (GL13)-positive staining (as a surrogate of lipofuscin content) in cisplatin-treated tumors and in combination with galunisertib, whereas both biomarkers of senescence were negligible in tumors of either vehicle-treated or galunisertib-treated mice (Fig. 5e,f and Extended Data Fig. 5b,c). The assessment of biomarkers of senescence in the cisplatin-treated group correlated with high positivity for phospho-AKT and phospho-p70S6K kinases, indicating the activation of the AKT/mTOR pathway (Fig. 5e,f). Double and triple staining co-localization experiments showed double-positive Ki67/pAKT cells in close vicinity to p16-positive cells in tumor cells after platinum treatment (Fig. 5g, Extended Data Fig. 6a–c and Supplementary Fig. 7a–d). Of note, the levels of phospho-AKT and phospho-p70S6K were significantly reduced upon TGFBR1 inhibitor co-treatment, which is compatible with the increased proliferation levels (measured by Ki67 staining) in the cisplatin-treated group (Fig. 5e,f and Extended Data Fig. 6c). Immunophenotypic profiles, including myeloid and lymphoid lineages, measured by flow cytometry did not show significant changes across the four experimental groups, except for the proportion of alveolar macrophages. Therefore, these results are in broad agreement with our in vitro functional/mechanistic assays, showing that cisplatin-induced senescence/SASP can stimulate tumor growth and proliferation of nearby lung cancer cells by the activation of the AKT/mTOR pathway and that there was no strong evidence of potential immunosuppressive activities (Extended Data Fig. 5d). Notably, cisplatin and galunisertib co-treatment substantially improved median survival of mice by 42.7% (from 48 days to 68.5 days) when compared to the vehicle-treated and monotherapy-treated groups (Fig. 5d).
Fig. 5. Cisplatin and galunisertib concomitant treatment significantly reduces tumor burden and enhances survival in a KrasG12V-oncogene-driven model of lung cancer.
a, Schematic representation of experimental layout. In brief, lung tumors were induced in 6−10-week-old KrasFSFG12V/+;p53frt/frt animals. After 8 weeks, lungs were imaged by computed tomography scan, and the number of tumors/tumor volume was analyzed. Animals were subjected to 1.5 mg kg−1 body weight CDDP, 150 mg kg−1 body weight galunisertib, a combination of both or vehicle at the frequency shown in the timeline. At 4 weeks upon treatment completion, animals were subjected to computed tomography scanning for lung tumor burden assessment (vehicle and CDDP groups, n = 3 mice; galunisertib and CDDP + galunisertib groups, n = 4 mice). b, Fold change in tumor volume at week 12 post-AdFLP (4 weeks after completion of treatment) relative to initial volume measured by computed tomography scan prior to treatment (vehicle, n = 15 tumors; CDDP only, n = 17 tumors; galunisertib only, n = 14 tumors; CDDP + galunisertib, n = 11 tumors). c, Three-dimensional rendering reconstructions of scanned lungs from each experimental group at week 8 (pre-treatment) and week 12 (post-treatment) post-AdFLP. Tumors are shown in red color. d, Survival curve of mice in each experimental condition. Representative images of IHC (e) and quantification (f) for p16INK4a, Ki67, pAKT and p70S6K in vehicle-treated (n = 3–6), cisplatin-treated (n = 7), galunisertib-treated (n = 9–10) and co-treated (n = 5–6) KrasFSFG12V/+;p53frt/frt mice. N in f represents the number of tumors analyzed (indicated in the figure) from three independent mice per experimental group. Data are shown as mean ± s.d. Statistical significance was assessed by one-way ANOVA, followed by Dunnett’s multiple test in histological analyses and Kruskal−Wallis test in fold change tumor volume analyses. Survival analysis was performed using the Kaplan−Meier method and a two-sided log-rank test. g, Top left panels, representative images for p16INK4a (magenta) and pAKT (teal, cytoplasmic); bottom left panels, Ki67 (teal) and pAKT (magenta, cytoplasmic) double IHC staining in the indicated lung cancer experimental groups (vehicle-treated, cisplatin-treated, galunisertib-treated and co-treated KrasFSFG12V/+;p53frt/frt mice). Magenta arrowheads point to examples of p16-positive cells. Teal arrowheads point to examples of pAKT-positive cells. Blue arrows point to examples of Ki67-positive cells. Blue/magenta arrows point to examples of Ki67/pAKT double-positive cells. Top right panels, representative images for p16INK4a (brown) and Ki67 (teal); bottom right panels, Ki67 (teal) and pAKT (brown, cytoplasmic) double IHC staining in a cisplatin-treated lung cancer experimental group. Blue arrowheads point to examples of Ki67-positive cells (teal). Brown arrowheads point to examples of p16-positive cells (brown). Blue/brown arrows point to examples of Ki67/pAKT double-positive cells. Scale bar in e, 50 μm; scale bar in g, 10 μm. AdFLP, adenovirus vector to deliver FLP recombinase enzyme. Images were created with BioRender.com. CDDP, cisplatin.
Extended Data Fig. 5. Concomitant treatment of cisplatin and galunisertib reduces tumour burden in KrasG12V oncogene-driven models of lung cancer.
a. micro-CT scans of lungs from each experimental group at week 8 (pre-treatment) and week 11 (post-treatment) post-AdFLP. b. Quantification of SenTraGor (GL13)+ cells per total cells in lung tissues from vehicle, cisplatin, galunisertib and co-treated mice (n = 3 independent mice samples per group). c. Representative histological images for SenTraGor (GL13) staining in the indicated experimental groups. Scale bar = 25 μm as depicted in the image. d. Immunophenotypic profiles by flow cytometry in lung tissues from vehicle, cisplatin, galunisertib and co-treated mice. Data are shown as mean ± SD (n = 3 independent mice samples in cisplatin group; n = 4 in the rest of the groups). Statistical significance was assessed by one-way ANOVA, followed by Tukey’s multiple comparisons test (b) or Kruskal-Wallis test (d). AdFLP, adenovirus vector to deliver FLP recombinase enzyme.
Extended Data Fig. 6. Cisplatin and galunisertib concomitant treatment reduces phospho-AKT and Ki67 levels in a KrasG12V oncogene-driven model of lung cancer.
a. Schematic representation of experimental layout. Briefly, lung tumour was induced in 6-10 week-old KrasFSFG12V/+;p53frt/frt animals. After 8 weeks, lungs were imaged by CT scan and number of tumours/tumour volume was analysed. Animals were subjected to 1.5 mg/kg body weight CDDP, 150 mg/kg body weight galunisertib, a combination of both or vehicle at the frequency shown in the timeline. At 4 weeks upon treatment completion, animals were subjected to CT scanning for lung tumour burden assessment (vehicle and CDDP groups, n = 3 mice; galunisertib and CDDP + galunisertib groups, n = 4). b. Representative images for p16INK4a (magenta) and Ki67 (teal) double immunohistochemistry staining in the indicated lung cancer experimental groups (vehicle-, cisplatin-, galunisertib- and co-treated KrasFSFG12V/+;p53frt/frt mice). c. Representative images for p16INK4a (magenta) and phospho-AKT (pAKT) (teal, cytoplasmic) or Ki67 (teal) and pAKT (magenta, cytoplasmic) double immunohistochemistry staining in the indicated lung cancer experimental groups (vehicle-, cisplatin-, galunisertib- and co-treated KrasFSFG12V/+;p53frt/frt mice). Black arrowheads point to examples of p16-positive cells (magenta). White arrowheads point to examples of pAKT-positive cells (teal). Orange arrowheads point to examples of Ki67-positive cells (teal). Grey arrowheads point to examples of Ki67 (teal)/pAKT (magenta) double-positive cells. Scale bars: 20 μm and 10 μm for magnifications. Images created with BioRender.com.
To further demonstrate the therapeutic benefit of targeting platinum-induced senescence, we conducted an experiment where KrasFSFG12V/+;p53frt/frt mice were subjected to three cycles of cisplatin-based chemotherapy, alone or in combination with navitoclax (ABT-263, a Bcl-2 inhibitor) that has senolytic activities36 (Extended Data Fig. 7a). Assessment of tumor burden by computed tomography imaging showed the best therapeutic effect in the co-treatment group compared to vehicle and monotherapies and also showed reduced levels of therapy-induced SA-β-gal activity and proliferation markers (Ki67 and pRb) (Extended Data Fig. 7a–c). Additionally, by using KrasFSFG12V/+ mice in a wild-type p53 background, we also found that concomitant treatment of cisplatin and senolytic ABT-737 (an alternative Bcl-2 inhibitor) significantly reduces the lung tumor burden when compared to platinum-based monotherapy (Supplementary Fig. 8a–c), and this correlates with decreased levels of biomarkers of therapy-induced senescence and proliferation as well as increased levels of tumor-associated apoptosis (Supplementary Fig. 8d–f). Taken together, our in vivo results emphasize the importance of modulating tumor-promoting activities of therapy-induced senescence and reveal the clinical potential of combining platinum chemotherapy and TGFBR1 inhibitors to improve therapeutic outcomes in lung cancer management.
Extended Data Fig. 7. Concomitant treatment of cisplatin and navitoclax reduces tumour burden in KrasG12V oncogene-driven models of lung cancer.
a. Schematic representation of experimental layout. Briefly, lung tumour was induced in 6-10 week-old KrasFSFG12V/+;p53frt/frt animals. After 8 weeks, lungs were imaged by CT scan and number of tumours/tumour volume was analysed. Animals were subjected to 1.5 mg/kg body weight CDDP, 85 mg/kg body weight navitoclax, a combination of both or vehicle at the frequency shown in the timeline. At 3 weeks upon treatment completion, animals were subjected to CT scanning for lung tumour burden assessment (vehicle, CDDP and navitoclax groups, n = 3 mice; CDDP + navitoclax groups, n = 2). b. Fold change in tumour volume after treatment completion relative to initial volume measured by CT scan prior to treatment (vehicle, n = 12 tumours; CDDP only, n = 10; galunisertib only, n = 10; CDDP + galunisertib, n = 6). c. Representative images of histological sections for SA-β-gal activity and immunohistochemistry for Ki67 and phospho-Rb (pRb) in vehicle-, cisplatin-, galunisertib- and co-treated KrasFSFG12V/+;p53frt/frt mice. Scale bar = 50 μm as depicted in the image. Data are shown as mean ± SD. Statistical significance was assessed by one-way ANOVA, followed by Kruskal-Wallis test in fold change tumour volume analyses. Images created with BioRender.com.
Platinum-based NACT induces tumor senescence in patients with NSCLC and results in the increased activation of the AKT/mTOR pathway
Evidence of therapy-induced senescence in patients with NSCLC, other than imperfect studies based solely on the detection of SA-β-gal in tumor specimens8, remains very limited. To convincingly demonstrate whether senescent response is seen in humans after exposure to platinum-based chemotherapy, we used human samples from patients who received preoperative cisplatin-based chemotherapy followed by surgery (Fig. 6a). We analyzed the expression of different proliferation and senescence-related markers in stage II−III human lung adenocarcinoma samples resected from individuals within 1 month of chemotherapeutic regimen completion as well as stage II−III treatment-naive lung adenocarcinomas (see Supplementary Table 1 for detailed clinical and pathological information). Histological analysis showed that Ki67 detection is heterogeneous and scattered but predominantly expressed throughout the tumors of treatment-naive patients compared to chemotherapy-treated specimens (Fig. 6b,d and Extended Data Fig. 8a). When compared to treatment-naive tumor samples, cisplatin-treated tumors presented a significantly higher proportion of p21-positive cells, generally accumulating in the outer regions of the lesions and presenting a more dispersed pattern in central tumor areas (Fig. 6b,d). Notably, consecutive tissue sections showed a lack of overlap between positive signal for the Ki67 proliferative marker and expression of the senescent marker p21, potentially indicating the implementation of a diverse senescent response in the tumors (Fig. 6b and Extended Data Fig. 8a). Moreover, p21-positive areas overlapped with SenTraGor (GL13)-positive staining, indicating senescence (Fig. 6b,d)37,38. The regional association between GL13 and p21 is, therefore, compatible with the accumulation of a cisplatin-induced senescent response across different tumor areas (see also Supplementary Fig. 9a,b).
Fig. 6. Neoadjuvant platinum-based therapy induces senescence in patients with NSCLC, and this correlates with increased phospho-AKT and phospho-P70S6K signaling in surrounding cells.
a, Schematic representation of human NSCLC samples analyzed in this study. b, Representative histological images of NSCLC specimens resected from treatment-naive (TN) or platinum-treated patients and subjected to Ki67, p21 and SenTraGor (GL13) staining. Scale bar is 50 μm or 20 μm as depicted. Black arrowheads show SenTraGor+ or p21+ cells; white arrowheads show SenTragor− or p21− cells. c, Quantification of phospho-AKT+ and phospho-P70S6K+ cells per total cells in TN or platinum-treated specimens. For quantification, five representative images were quantified per sample. d, Quantification of Ki67+, p21+ and SenTraGor (GL13)+ cells per total cells in TN or platinum-treated specimens. Bottom, percentage of p21+ and SenTraGor (GL13)+ cells in consecutive sections of platinum-based treated patients. For quantification, five representative images were quantified per sample. e, Representative histological images of NSCLC specimens resected from platinum-treated patients and subjected to staining for p21 and TGFβ1 in consecutive tissue sections. f, Representative histological images of NSCLC specimens resected from platinum-treated patients and subjected to co-staining for p21 (magenta) and phospho-AKT or phospho-P70S6K (brown). Scale bar is 50 μm as depicted in the image. g, Quantification of phospho-AKT+ and phospho-P70S6K+ cells per total cells in p21-low and p21-high areas. For quantification, three representative images for each p21-low/high area from each patient were analyzed. P21-low areas were defined as those with <10% p21+ cells/total cells; p21-high areas were defined as those with >10% p21+ cells/total cells. Data are shown as mean ± s.d. Statistical significance was determined by two-tailed Student’s t-test. Images were created with BioRender.com.
Extended Data Fig. 8. Neoadjuvant platinum-based therapy induces biomarkers of senescence in tumours of NSCLC patients and this correlates with increased phospho-AKT and phospho-P70S6K signalling in surrounding cells.
a. Representative histological images of NSCLC specimens resected from treatment-naïve (n = 2) or platinum-treated (n = 4) patients and subjected to Ki67, p21, phospho-AKT and phospho-P70S6K staining. Scale bar = 100 μm. b, c. Representative histological images of NSCLC specimens resected from platinum-treated patients and subjected to co-staining for p21 (magenta) and phospho-AKT (brown, cytoplasmic) or phospho-P70S6K (brown) double immunohistochemistry. Scale bar = 100 μm or 20 μm as depicted in the image. d. Representative histological images of NSCLC specimens resected from platinum-treated patients and subjected to co-staining for p21 (magenta) and phospho-P70S6K (brown) double immunohistochemistry, or Ki67 (brown), as indicated. Scale bar = 50 μm as depicted in the image. Images from all available patient samples for this study are shown in the figure, and micrographs represent similar staining patterns across sample and subgroups (treatment-naïve and platinum-treated). Images created with BioRender.com.
Given previous evidence generated in this study, we also explored the expression of TGFβ ligands, phospho-AKT and phospho-P70S6K in the human tumor samples. Representative sections show the accumulation of TGFβ1 in areas of p21-positive cells (Fig. 6e). Notably, these analyses showed a considerably increased reactivity against AKT and P70S6K phosphorylation within the cisplatin-treated tumors, which intriguingly correlated with the regions of accumulated p21-positive expression within the lesions (Extended Data Fig. 8a). Quantification of cells positive for phospho-AKT and phospho-P70S6K revealed significant increased phosphorylation of these kinases compared to treatment-naive specimens (Fig. 6c), indicating an association between senescence induction and the activation of this pathway. To further determine the expression patterns and the potential link between the expression of the senescent marker p21 and the activation of these kinases, we analyzed co-staining for the levels of each phosphorylation state in p21-low and p21-high expression regions in chemotherapy-treated tumors. Notably, these analyses revealed a significantly higher expression of phospho-AKT and phospho-P70S6K in regions with a marked accumulation of p21-positive cells (Fig. 6f,g and Extended Data Fig. 8b–d), indicating an association between the induction of senescence in the tumors and the proximal activation of the AKT/mTOR pathway in nearby human lung adenocarcinoma cells after platinum-based chemotherapy.
Taken together, these analyses provide supportive evidence in human NSCLC tumors of the induction of cellular senescence after cisplatin treatment and show that accumulated senescent cells are tightly linked to the activation of the AKT/mTOR signaling in the tumor microenvironment, providing clinical validation of the mechanistic insights from preclinical models.
Tumor senescence is induced after platinum-based NACT in ovarian cancer and is associated with AKT/mTOR pathway activation
To test whether senescence responses to platinum chemotherapy occur beyond NSCLC, we first used gene expression profile datasets from HGSOC patient samples treated with neoadjuvant carboplatin and paclitaxel from the CTCR-OV01 study (Fig. 7a)39. GSEA with normalized enrichment scores (NESs) of post-chemotherapy samples (carboplatin, paclitaxel and combination) relative to pretreatment samples shows downregulation of cell-cycle-related signatures, including G2/M checkpoint and E2F-mediated regulation of DNA replication (Fig. 7b,c), which is consistent with the induction of therapy-induced senescence. Of note, platinum-based therapy was associated with significant upregulation of the TGFβ signaling pathway, TGFβ regulation of the extracellular matrix, TGFBR in EMT and activation of SMAD signatures (including non-canonical TGFβ signaling), and these findings correlated with upregulation of inflammatory response and AKT/mTOR signaling pathways (Fig. 7b,c and Extended Data Fig. 9a,b).
Fig. 7. GSEA of microarray datasets from patients with ovarian cancer after chemotherapy treatment.
a, Schematic overview of the CTCR-OV01 study. b, Heatmap showing positive and negative NESs indicating positive and negative enrichment, respectively, for 34 evaluable objects. c, Heatmap showing normalized gene expression values (z-score) associated with the indicated signaling pathways for each patient sample before and after treatment. The genes selected were differentially expressed between pretreatment and posttreatment samples (Benjamini−Hochberg-adjusted P < 0.05). The microarray dataset was downloaded from the GEO (GSE15622), and the genes selected were differentially expressed between pretreatment and posttreatment samples (statistical testing implemented in the limma R Bioconductor package with the Benjamini−Hochberg adjustment for multiple testing). Images were created with BioRender.com. ECM, extracellular matrix; Expr, expression; PDAC, pancreatic ductal adenocarcinoma; sign., signaling.
Extended Data Fig. 9. Gene Set Enrichment Analyses of ovarian cancer patients microarray datasets after chemotherapy treatment.
a. Heat map showing normalised gene expression values (z-score) genes from TGF-β and PI3K, and MAPK pathways for each patient sample before and after treatment. The genes selected were differentially expressed between pre- and post-treatment samples (statistical testing implemented in the limma R Bioconductor package with the Benjamini-Hochberg adjustment for multiple testing). b. Network (cnetplot) showing the links between significantly enriched pathways (Fisher’s exact test implemented in the EnrichR tool, with the Benjamini-Hochberg adjustment for multiple testing) in the genes DE post-treatment and the top 10 most frequently co-occurring DE genes in these pathways. Pathway node size reflects the number of differentially expressed genes in the pathway. Gene nodes are coloured by the (log-)fold change before and after treatment. Links are coloured by adjacency to the same pathway. The microarray dataset was downloaded from GEO (GSE15622) and processed using R Bioconductor packages.
Next, we analyzed unmatched treatment-naive and platinum-treated tissue microarray (TMA) cores from patients with HGSOC from the NeOv cohort40. Treatment-naive samples were obtained from 26 patients with International Federation of Gynecology and Obstetrics (FIGO) stage I–IV HGSOC treated with immediate primary surgery (IPS). Tumors exposed to carboplatin-based NACT were obtained from 17 patients with FIGO stage III–IV disease who subsequently underwent delayed primary surgery (DPS) (Fig. 8a; see Supplementary Table 2 for clinical details).
Fig. 8. Senescence induction in patients with HGSOC after neoadjuvant treatment with platinum-based chemotherapy.
a, Schematic representation of human HGSOC samples analyzed in this study. b, Heatmap of positive staining by IHC stain and compartment (S, stroma; T, tumor). Values represent scaled percentages of positive IHC staining by marker and compartment. Individual cases (columns) are annotated according to subcohort. Hierarchical clustering of columns by Ward’s method, specifying row order to keep tumor and stromal compartments adjacent by marker. c, Representative histopathological images of HGSOC TMA cores from the IPS and NACT cohorts, subjected to IHC staining against p21, p16, pS6RP, pAKT1 and Ki67. Scale bar, 50 µm. d, Quantification of positive IHC staining of cells detected in the tumor compartments of IPS and NACT cores (significance test: Wilcoxon rank-sum test/Mann−Whitney U-test) as a percentage of all tumor cells detected per TMA core (IPS: n = 24 for p21, p16, pS6RP and pAKT1 and n = 26 for Ki67; NACT: n = 17 with evaluable TMA cores for all markers). Each point represents an individual patient (mean value of available primary and metastatic samples per case). e, Percentages of pS6RP-positive cells by surgical debulking outcome, at the total compartment (tumor and stroma) and tumor and stromal compartment levels, in the entire HGSOC cohort and by IPS/NACT subcohort (IPS optimal debulking n = 16, IPS residual disease n = 7, NACT optimal debulking n = 9, IPS residual disease n = 8). Optimal debulking, no residual disease; residual disease, minimal (<1 cm) or gross (>1 cm) residual disease. Significance test: two-sided unpaired Wilcoxon rank-sum test/Mann−Whitney U-test. Each point represents a patient-level mean value.
We analyzed 899 digitized tissue core images (IPS N = 574, NACT N = 325; Supplementary Fig. 10, REMARK diagram) from 100 distinct formalin-fixed paraffin-embedded (FFPE) blocks (IPS N = 72, NACT N = 28). A total of 7,187,809 cell detections were included in the analysis (IPS N = 3,418,690 tumor, N = 1,414,179 stromal; NACT N = 905,527 tumor, N = 1,449,413 stromal). As expected, the proportions of tumor cells per core were significantly higher in the treatment-naive IPS samples compared to the platinum-exposed samples (Extended Data Fig. 10a,b), and subsequent analyses were, therefore, conducted by tissue compartment (tumor, stroma and tumor + stroma).
Extended Data Fig. 10. Proportions of tumour cells of all cell detections (using binary labels: tumour or stroma) per TMA core and quantification of immunohistochemical staining.
a. All TMA cores, unmatched comparison between IPS (n = 574) and NACT (n = 325) samples (two-sided unpaired Mann Whitney U, p = 1.1 × 10-49). b. All TMA cores, unmatched comparison between IPS and NACT by binary anatomical site category (IPS metastatic: n = 286; IPS ovary/fallopian tube: n = 288, NACT metastatic: n = 137; NACT ovary/fallopian tube: n = 188). Two-sided unpaired Mann Whitney U, p = 8.1 × 10-15 (metastatic), p = 5.7 × 10-39 (ovary/fallopian tube). c. Quantification of immunohistochemical staining of HGSOC TMA cores by sub-cohort and anatomical site in the tumour compartment, each point representing a site-level mean value per patient. P values (metastatic): ki67 (p = 6.1 × 10-3), p16 (p = 1.7 × 10-2), p21 (p = 2.3 × 10-2), pAKT1 (p = 4.8 × 10-1), pS6RP (p = 6.1 × 10-3). P values (ovary/fallopian tube): ki67 (p = 1.7 × 10-4), p16 (p = 1.2 × 10-1), p21 (p = 4.3 × 10-4), pAKT1 (p = 5.7 × 10-7), pS6RP (p = 4.8 × 10-6). d. Quantification of immunohistochemical staining of HGSOC TMA cores by sub-cohort and anatomical site in the stromal compartment, each point representing a site-level mean value per patient. P values (metastatic): ki67 (p = 1 × 10-4), p16 (p = 0.9), p21 (p = 2.8 × 10-3), pAKT1 (p = 1.8 × 10-1), pS6RP (p = 6.1 × 10-3). P values (ovary/fallopian tube): ki67 (p = 5.7 × 10-7), p16 (p = 0.9), p21 (p = 1.2 × 10-4), pAKT1 (p = 2.5 × 10-2), pS6RP (p = 7.2 × 10-5). Ovarian case representation with evaluable TMA cores in c. and d. as follows: IPS ovary/fallopian tube n = 21 for p16/p21/pAKT1/pS6RP and n = 22 for Ki67; IPS metastatic n = 16 for all 5 markers; NACT ovary/fallopian tube n = 15 for p16/p21/pAKT1/Ki67 and n = 14 for pS6RP; NACT metastatic n = 10 for all 5 markers. Significance test: Wilcoxon rank sum test (two-sided unpaired Mann Whitney U test). Boxplot median and lower/upper hinges show 50th, 25th and 75th centiles, with hinges extending 1.5 × interquartile range below and above the 25th and 75th centiles respectively. P values: * p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001.
We quantified expression levels of proliferation and senescence-related immunohistochemistry (IHC) markers and compared these between unmatched treatment-naive IPS and platinum-treated NACT samples. Cellular proliferation was significantly lower in post-NACT tissues compared to treatment-naive specimens as measured by positive Ki67 staining in the tumor compartment (Fig. 8b–d and Supplementary Table 3), in both primary ovarian and metastatic tissues in tumor (Extended Data Fig. 10c and Supplementary Table 4) and stroma (Extended Data Fig. 10d and Supplementary Table 5). Conversely, the proportions of cells expressing the senescence marker p21 were significantly higher in the platinum-treated samples (Fig. 8b–d and Supplementary Table 3), in primary and metastatic tumor (Extended Data Fig. 10c and Supplementary Table 4) and in stroma (Extended Data Fig. 10d and Supplementary Table 5). The senescence-related marker p16 was also expressed at higher levels in the tumor compartments of NACT samples compared to unmatched IPS tissues (Fig. 8b–d and Supplementary Table 3) but was only statistically significant in tumor cells at metastatic tissue sites and not in primary tumors or in stroma (Extended Data Fig. 10c,d and Supplementary Tables 4 and 5).
We tested whether senescence induction is associated with activation of the AKT/mTOR pathway in platinum-treated HGSOC. pAKT1 was expressed in tumor at significantly higher levels in NACT compared to IPS cases overall (Fig. 8b–d and Supplementary Table 3) but only in primary pelvic tissues and not in metastatic sites (Extended Data Fig. 10c,d and Supplementary Tables 4 and 5). NACT tissues also expressed the mTOR target phospho-S6 ribosomal protein (pS6RP) in significantly higher proportions of tumor cells compared to IPS tissues overall (Fig. 8b–d and Supplementary Table 3), in both primary and secondary sites (Extended Data Fig. 10c,d and Supplementary Tables 4 and 5).
After dichotomizing p21 staining into high and low expression groups using the median values across the cohort, pS6RP expression was significantly higher in p21-high cases compared to p21-low cases in all compartments (Supplementary Fig. 11a–c). At the subcohort level, pS6RP staining was significantly higher or showed a trend toward being elevated in p21-high samples (Supplementary Fig. 11d–f). High-p21 tumors also displayed significantly higher pAKT1 expression levels than low-p21 tumors at the cohort level (Supplementary Fig. 11h); however, this was only significant in IPS cases at the subcohort level (Supplementary Fig. 11k).
The association between p21 and pS6RP expression was also significantly correlated between the two markers overall (Supplementary Fig. 12a–c). These correlations remained significant at the subcohort level in the total and stromal compartments and in IPS tumors (Supplementary Fig. 12d–f). Expression levels of p21 and pAKT1 were also significantly correlated overall (Supplementary Fig. 13a–c). At the subcohort level, these were statistically significant in the total and stromal compartments and in IPS tumors (Supplementary Fig. 13d–f). Some correlations between p16 and pS6RP (Supplementary Fig. 12h,k) and between p16 and pAKT1 were identified in tumor (Supplementary Fig. 13h,k) but not in the stromal or total compartments.
To capture the global relationships between expression levels of senescence and mTOR pathway-related markers on a patient and cohort level, we performed hierarchical clustering on normalized percentage values of positive IHC staining for the five markers by case. This resulted in two primary patient clusters, which reflected subcohort and, thereby, platinum exposure status in most cases (predominantly platinum-exposed cluster: 92% (12/13) NACT cases; predominantly platinum-naive cluster: 83% (25/30) IPS cases; Fig. 8b). Unsupervised hierarchical clustering on cases with unspecified stain and compartment order resulted in a minor cluster with Ki67 and p16 (stroma) and a major cluster with pAKT1, pS6RP, p21 and p16 (tumor) (Supplementary Fig. 14a,b). Within the latter cluster, p21 and pS6RP expression in both tumor and stromal clustered closest together (Supplementary Fig. 14a,b).
Regarding clinical outcome, higher pS6RP staining was associated with residual disease and suboptimal surgical debulking at the whole-cohort level (Fig. 8e). On subcohort analysis, this was statistically significant only in platinum-exposed NACT samples and not in treatment-naive samples (Fig. 8e).
Collectively, these analyses in HGSOC indicate p21 induction associated with platinum chemotherapy in a tumor type that is genomically and clinically distinct from NSCLC and further support the potential role of AKT/mTOR pathway activation in platinum-induced senescence. We identified a potential link between pS6RP expression and suboptimal tumor resectability, which, in turn, is a known poor prognostic marker. These results suggest that first-line platinum-induced senescence and associated AKT/mTOR pathway activation could have an impact on clinical outcome in HGSOC.
Discussion
Multiple preclinical studies have shown that therapy-induced senescence can enhance tumor persistence via diverse SASP-induced phenotypes and tumor-promoting activities12,22,37–40. Accordingly, early-phase clinical trials have been developed to target therapy-induced senescence41–44; however, progress has been limited by toxicity and a lack of specific biomarkers for patient selection. Based on preliminary observations of increased SA-β-gal activity in lung cancer after neoadjuvant treatment8, we reasoned that targeting platinum-induced senescence might enable precision medicine approaches to reduce platinum therapy resistance. Here we systematically characterize clinically relevant mouse models and patient-derived samples to demonstrate that cisplatin selectively induces an SASP enriched in TGFβ ligands, which, in turn, strongly promotes cancer cell fitness and fuels tumor progression and therapy resistance in lung and ovarian cancer.
Our findings confirm that cisplatin-induced senescence—but not senescence induced by docetaxel or palbociclib—is a driver of cancer cell proliferation, increased tumor growth and reduced survival in both xenograft and orthotopic lung cancer models. We developed a functional screening platform to characterize platinum-induced SASP using MEMAs to measure individual and combinatorial factors in an integrative multiomic analysis. We identified that platinum-based treatment specifically and strongly induces TGFβ ligands as critical drivers of cancer cell proliferation and tumor promotion during platinum-induced senescence. These deleterious effects were mitigated by galunisertib or the senolytic agent ABT-737, supporting a causal role for platinum-induced senescence in driving cancer progression. We also showed similar benefits for navitoclax in experiments modeling clinical trial designs. Notably, our preclinical findings are consistent with clinical observations showing that patients with NSCLC who exhibit in situ tumor-associated SA-β-gal activity after platinum-based treatment have significantly worse survival compared to those without detectable senescence-associated activity45. It is likely that platinum-induced DNA damage responses (DDRs) drive TGFβ secretomes46 as palbociclib and docetaxel have predominantly non-DDR mechanisms of action and weaker induction of TGFβ ligands. These SASP analyses should now be applied to other clinically relevant therapies, such as PARP inhibitors or radiotherapy, to determine if there is common involvement of DDR in TGFβ secretomes.
Increasing age is a negative prognostic factor in patients with lung and ovarian carcinoma and is associated with reduced response. We show that cisplatin pretreatment in middle-aged mice creates a permissive lung microenvironment that significantly enhances tumor progression of transplanted lung cancer KP cells compared to younger recipients. Of note, this leads to significant weight loss and reduced survival rates. In these animals, the cisplatin-induced tumor-promoting niche is characterized by senescence-associated biomarkers, consistent with our previous findings27. Defining the precise relationships between senescence and contributions from other factors arising from broader tissue remodeling responses will require larger treatment cohorts in young and aged mice and spatial single-cell profiling.
Recent studies using endogenous KRAS-driven transgenic mouse models have shown that aging can limit lung tumorigenesis by suppressing cancer stemness47 and reshaping the tumor suppressor gene landscape48. Both studies conclude that aged alveolar cancer-initiating cells exhibit diminished transformation and proliferative potential. Nonetheless, as shown in our data, transplantation of KP cells from young donors into aged recipient mice resulted in increased numbers of tumors than in young recipients47, supporting the notion that aged and/or damaged lung microenvironments—when decoupled from the intrinsic properties of the transformed cells—can adopt a pro-tumorigenic role. These observations are also consistent with reports showing that cellular senescence can facilitate cancer relapse by generating tumor-permissive niches25 and that aging promotes metastasis in KRAS-driven lung cancer49. Cellular senescence can also exacerbate adverse effects of chemotherapy25, and older individuals have higher grades of chemotherapy-associated toxicities45. Notably, dysregulated TGFβ signaling has been implicated in the pathogenesis of age-related cancers50, and TGFβ1 has been identified as a key mediator of both senescence and skin aging through PI3K/AKT pathway activation51. Studying larger cohorts of neoadjuvantly treated tumor specimens from different age groups could determine whether they respond differently to SASP-induced TGFβ signaling and activation of AKT/mTOR pathways.
Our study now establishes a mechanistic link between chemotherapy-induced senescence and TGFβ-mediated activation of TGFBR1 signaling, leading to non-canonical downstream stimulation of the AKT/mTOR pathway in both NSCLC and HGSOC. We infer this mechanism from multiple lines of evidence presented here. Additionally, we extended these findings through transcriptomic and gene set enrichment analyses of datasets from our previous CTCR-OV01 study of 34 patients with HGSOC who were treated with carboplatin or paclitaxel NACT before interval debulking surgery39. Interestingly, in HGSOC TMAs collected after NACT, we show that phosphorylation of kinases within the AKT/mTOR pathway was associated with suboptimal surgical debulking—a major poor prognostic factor in ovarian cancer52. Collectively, these findings highlight a potentially clinically relevant connection among platinum-induced senescence, TGFβ signaling and mTOR pathway activation, which warrants further investigation in larger, independent cohorts to evaluate the broader impact of these mechanisms on treatment response and overall survival in both NSCLC and HGSOC.
Of translational relevance, our data show that platinum−anti-TGFBR1 doublet combination treatment with cisplatin and galunisertib reduces tumor burden and significantly enhances survival in orthotopic and KrasFSFG12V-driven lung cancer mouse models. Although none has achieved regulatory approval to date, multiple therapeutic strategies targeting TGFβ signaling in cancer clinical trials have been explored, including direct TGFβ sequestration (for example, neutralizing antibodies or ligand traps), suppression of expression (for example, antisense oligonucleotides or peptide vaccines) and inhibition of downstream signaling, most commonly with small-molecule inhibitors of TGFBR1 (ref. 53). Oral small-molecule anti-TGFBR1 agents such as galunisertib and vactosertib have been under extensive clinical investigation in early-phase trials both as monotherapies and in combination with standard anticancer regimens for the treatment of various cancer types54. Our work supports the value of testing drug combinations with a senolytic agent as an alternative strategy to using TGFBR1 inhibitors to target platinum-induced senescence in future trials.
This work has several limitations. Mechanistically, understanding of how specific combinations of TGFβ ligands and signaling inputs dictate activation of either canonical (SMAD dependent) or non-canonical TGFBR1 signaling is incomplete. Likewise, the precise regulatory network linking non-canonical TGFBR1 signaling to the AKT/mTOR pathway needs to be fully elucidated. Our findings demonstrate that SASP-responsive lung cancer cells exhibit enhanced malignant traits upon activation of these pathways; however, activation of the AKT/mTOR pathway may also influence other cellular processes, such as metabolism and mitochondrial function, and we cannot exclude the involvement of additional upstream regulators. Furthermore, TGFβ is known to exert various pro-tumorigenic effects—including immunosuppressive, angiogenic and metastatic activities—that warrant further investigation53. Finally, despite extensive preclinical progress, the clinical translation of current anti-TGFβ and senolytic therapies remains limited by on-target and off-target toxicities in some patient groups, including cardiotoxicity, skin reactions and thrombocytopenia27,43,53.
Despite notable therapeutic advances, treatment resistance in NSCLC and HGSOC still represents a major barrier to improving clinical outcomes. Collectively, our data demonstrate how the beneficial effects of a clinically critical cytotoxic therapy could be pharmacologically enhanced for patient benefit. Notably, we demonstrate that the detrimental effects of cisplatin-induced senescence can be ameliorated through concomitant TGFBR1 inhibition and senolytic agents (Supplementary Fig. 15). We propose that combining platinum chemotherapy with TGFBR1 inhibition represents an important therapeutic strategy to be investigated in future clinical trials in NSCLC and HGSOC. This approach may help prevent cancer treatment resistance and limit the harmful impact of unresolved bystander senescence within the tumor microenvironment.
Methods
Ethics declaration
The use of human NSCLC and HGSOC tissues was approved by ethics review committees at Royal Papworth Hospital Research Tissue Bank (RPHRTB) (T02722) and Cambridge University Hospitals NHS Foundation Trust (08/H0306/61), respectively. Written informed consents were obtained from all patients.
All animal experiments were approved for ethical conduct by the Home Office England and Central Biomedical Services, performed under PPL holder numbers P7EC604EE and PP7061972 and regulated under the Animals (Scientific Procedures) Act 1986, as stated in the International Guiding Principles for Biomedical Research involving Animals.
NSCLC human biopsies
Human lung adenocarcinoma samples were obtained from the RPHRTB after review by the RPHRTB project review committee (project number T02722). RPHRTB has a derogation under the UK Human Tissue Authority (HTA) to supply samples (HTA number 12212), surplus to clinical need, that have been collected using Research Ethics Committee (REC)-approved RPHRTB consent. Patients signed the RPHRTB general consent and accepted the use of their biopsy for research purposes. Information about the human samples used can be found in Supplementary Table 1.
HGSOC clinical cohort
Patients with histopathological confirmation of HGSOC who provided written, informed consent to the CTCR-OV04 observational study (Molecular Analysis of Response to Treatment in Ovarian Cancer, Cambridge University Hospitals NHS Foundation Trust, East of England – Cambridgeshire and Hertfordshire, REC approval number 08/H0306/61) were included in the NeOv cohort studied in this work.
Forty-three cases diagnosed between May 2010 and August 2017 with archival FFPE tumor tissue obtained at the time of debulking surgery were included (Supplementary Table 2: summary of clinical characteristics). Twenty-six patients were treated with IPS followed by adjuvant chemotherapy (NeOv-IPS cohort). Seventeen patients were treated with NACT followed by DPS and adjuvant chemotherapy (NeOv-NACT cohort).
TMA construction
Tissue areas containing tumor on hematoxylin and eosin (H&E) sections from FFPE blocks were marked for core retrieval by a consultant pathologist specializing in gynecological cancers (M.J.L.). TMA construction was carried out at the Cancer Research UK (CRUK) Cambridge Institute Histopathology Core Facility. Paraffin TMAs of 1-mm-diameter cores were constructed with a manual Arrayer (Beecher Instruments, MTA-1) according to predetermined maps.
For the NeOv-NACT cohort, TMAs were constructed according to anatomical site (primary ovarian and metastatic tumor tissue), and each donor FFPE block was represented in triplicate. For the NeOv-IPS cohort, a previously constructed set of primary and metastatic TMAs was used where each donor FFPE block was represented in duplicate.
Three-micrometer-thick sections were cut from each TMA block with a Leica RM2255 microtome, floated on a 45 °C water bath for 30 seconds, mounted on positively charged glass slides and dried at 60 °C for 60 minutes. Automated pre-IHC dewaxing, rehydration and clearing and post-IHC dehydration and clearing were carried out using a Leica ST5020 Multistainer. Sections were mounted using a Leica CV5030 Coverslipper.
Image processing and analysis
Stained TMA slides were scanned at ×20 magnification into Aperio eSlide Manager software (Leica Biosystems) using an Aperio AT2 (automated digital scanner; Leica Biosystems). Scanned images were imported into QuPath version 0.4.1 for processing. For core-wise analysis, each slide image was segmented using a de-arraying function. Cores were categorized by quality criteria to inform inclusion/exclusion during downstream analysis (Supplementary Fig. 10, REMARK diagram).
After positive cell detection, minor focal areas of poor quality, including folded areas and artifacts, were removed. Smooth features (10 µm/25 µm/50 µm) were applied to the cell detections followed by tumor/stroma random forest classification. Manual quality control and corrections were performed on each core, and core annotations and cell classifications were subjected to final review by a consultant pathologist (M.J.L.) prior to automated batch export of core-wise detection measurements.
Downstream computational analyses were performed in RStudio (R version 4.2.1). After removal of cores with substantial tissue damage, loss or folding, cores were retained for analysis if a minimum 20 or more tumor cells were detected. Absolute counts and proportions of DAB-positive and DAB-negative tumor and stromal cells were computed per core, and mean values were computed per parent FFPE block (referred to as sample-level means), per binary anatomical site (primary or metastatic) and per patient. Percentages of DAB-positive staining cell detections were compared between the NACT and IPS cohorts using the two-sided unpaired Wilcoxon rank-sum test (Mann−Whitney U-test) using the ‘wilcoxon.test’ function from the ‘stats’ R package (version 4.4.1). Adjustments for multiple comparisons were performed using the Benjamini−Hochberg method using the ‘stats’ R package (version 4.4.1). Heatmaps were generated using the ‘ComplexHeatmap’ R package (version 2.14.0).
Animal work and in vivo treatment studies
For senescence induction upon chemotherapeutic treatment and evaluation of chemotherapy and galunisertib treatments in vivo, lung tumors were generated in C57BL/6 KrasFSFG12V/+ (ref. 34) and KrasFSFG12V/+;p53frt/frt (refs. 34,35) mice through intranasal or intratracheal administration with FLP-expressing adenovirus. In brief, 6−10-week-old mice (including male and female mice) were treated once with Adeno-FLP particles (2.5 × 107 plaque-forming units (PFU) per mouse of virus) after anesthesia (ketamine/xylazene by intraperitoneal injection followed by antisedan injection). For virus preparation, viral particles were diluted in DMEM, precipitated with 10 mM CaCl2 and incubated for 20 minutes prior to nasal inhalation or intratracheal administration. In KrasFSFG12V/+;p53frt/frt mice, mice were subjected to either vehicle or 1.5 mg kg−1 body weight cisplatin (by intraperitoneal injection) once weekly for four subsequent weeks, followed by six daily administrations of galunisertib (LY2157299; 150 mg kg−1 body weight, by oral gavage) after each cisplatin dose. In senolytic treatment of KrasFSFG12V/+;p53frt/frt mice, mice were subjected to either vehicle or 1.5 mg kg−1 body weight cisplatin (by intraperitoneal injection) once weekly for three subsequent weeks, followed by six daily administrations of navitoclax (ABT-263; 85 mg kg−1 body weight, by oral gavage) after each cisplatin dose. In KrasFSFG12V/+ mice, at 9 months after tumor induction, mice were treated with four doses of 1.5 mg kg−1 body weight cisplatin (by intraperitoneal injection) once weekly and ABT-737 (25 mg kg−1 body weight, twice per week on weeks 2 and 4 after cisplatin treatment, by intraperitoneal injection) (MedChemExpress, HY-50907) or vehicle (solution of 30% propylene glycol, 5% Tween 80 and 3.3% dextrose).
For xenograft experiments, 8–12-week-old female SCID mice were used. Cell suspension was prepared by mixing Matrigel (Corning) 1:1 with 1 × 106 control A549 cells, 1 × 106 control and 1 × 106 cisplatin-induced senescent A549 cells or 4 × 106 cisplatin-induced senescent A549 cells alone. Cells were injected subcutaneously, and xenograft growth was monitored using a digital caliper twice a week. Five days after transplantation, animals were treated with 150 mg kg−1 body weight galunisertib (LY2157299), 25 mg kg−1 body weight ABT-737 or vehicle following the experimental timing detailed in the respective figures. Tumor volume was calculated as volume = (D × d2) / 2, where D and d refer to the long and short tumor diameters, respectively. Experiments were terminated either at preestablished time endpoints or when tumors reached an average diameter of 1.5 cm.
For orthotopic transplantation of lung cancer cells, C57BL/6J female mice (8–10 weeks for young cohorts and 42−45 weeks for middle-aged cohorts) were sublethally irradiated at 4 Gy using a cesium source irradiator 6 hours before cell injection. Luciferase-expressing KrasG12D/WT;p53−/− murine lung tumor L1475(luc) cells were injected intravenously (2 × 105 cells per mouse). Baseline luminescence was recorded 24 hours after transplantation, and tumor growth was monitored twice a week by bioluminescence imaging after intraperitoneal injection with D-luciferin (150 mg kg−1 body weight; PerkinElmer) using an IVIS Spectrum Xenogen imaging system (Caliper Life Sciences) and Living Image software (version 4.7.3). Relative luciferase activity was calculated as the change from baseline at indicated timepoints, normalized to a blank control (luciferase-negative animal). Tumor survival represents the onset of moderate signs of disease and/or 15% weight loss.
Mice in this study were matched with age and starting tumor volume. Due to the variability of in vivo tumor growth, animals were randomly allocated to experimental groups once initial tumor burden was determined, such that each experimental group was normalized by similar tumor volumes before treatment (<25 mm3 per mouse of total tumor burden for KrasFSFG12V/+;p53frt/frt mouse survival). Maximal tumor size/burden was not exceeded. Investigators were blinded to group allocation during data analysis only when relevant. For example, the investigators who performed and analyzed the micro-computed tomography (µCT) acquisitions were not informed of the experimental groups in advance.
µCT imaging and analysis
Mice were anesthetised by isoflurane inhalation and scanned using a micro-PET/CT scanner (Mediso Medical) and Nucline nanoScan software (version 2.0) with an X-ray energy of 35 kVp, an exposure time of 450 ms, a total number of projections of 720 and one projection per step following a semicircular single field of view scan method. Scans were reconstructed through a Butterworth filter and a small voxel size. Three-dimensional reconstructions and analyses were performed using Slicer 4.10.2 software (3D Slicer). To determine tumor size, the length (longest diameter (L)) and width (diameter perpendicular to the length (W)) were measured, and the ellipsoid formula was applied (tumor volume = (4 / 3) × PI × (W / 2)2 × (L / 2)).
Cells and reagents
For in vitro experiments, the human lung adenocarcinoma A549 cell line was obtained from the American Type Culture Collection and maintained in DMEM growth medium supplemented with 10% FBS. The heterozygous KrasG12D/WT;p53−/− murine L1475(luc) cell line was generated from KrasLSLG12D/WT;p53Fx/Fx mice55,56 and maintained in F-12/DMEM growth medium supplemented with 10% FBS. The ovarian cancer cell line PEO4 was maintained in RPMI growth medium supplemented with 10% FBS. All cells were grown at 37 °C and 5% CO2 and were routinely authenticated and tested for mycoplasma contamination.
For the induction of cellular senescence in vitro, A549 cells were treated with 15 µM cisplatin (Sigma-Aldrich), 7.5 µM carboplatin (Sigma-Aldrich), 100 nM docetaxel (Sigma-Aldrich) and 15 µM palbociclib (Cambridge Bioscience) for 7 days. L1475(luc) cells were treated with 3 µM cisplatin, 100 nM docetaxel and 30 µM palbociclib for 5 days. PEO4 cells were treated with 2.5 µM cisplatin and 10 µM carboplatin. LPK-2 and NCI-H226 cells were treated with 15 µM cisplatin. For proliferation and colony formation assays, A549 and L1475(luc) cells were treated with 50 µM galunisertib (Stratech) and 1 µM rapamycin (Stratech). A549 cells were also exposed to rhTGFβ1 ligand (Bio-Techne) at a concentration of 10 ng ml−1.
Apoptosis assays
To assess the induction of apoptosis A549 or L1475(luc), cells cultured with control CM or cisplatin-treated CM were seeded in 96-well plates. Twenty-four hours later, Annexin V Fluorescent Reagent (Essen BioScience) was added to the media, and, after a first scan, the cells were treated with vehicle, 50 µM galunisertib or 0.5 µM staurosporine for 48 hours. Images were collected every 2 hours with an IncuCyte S3 Live Cell Analysis System microscope (Essen BioScience) over time. A total of five pictures per well were analyzed using the IncuCyte ZOOM software analyzer, and total and Annexin V-positive cells were counted using ImageJ software.
In vitro assessment of malignant traits
CM preparation and analysis
To evaluate the effect of the different SASPs on cell behavior in a paracrine manner, cells that had been under chemotherapeutic treatment as specified above were thoroughly washed with PBS twice, and then either FBS-free or 10% FBS medium was added. After 48−72 hours of incubation, CM from senescent cells was collected and centrifuged for 10 minutes at 1,000 r.p.m. before adding onto non-senescent cells for further assessment. CM used on all assays was freshly conditioned and was not subjected to storage/freezing.
Proliferation assays
For proliferation assessment, 20,000 A549 cells were seeded in a 24-well plate and let to attach overnight. The next morning, media were removed, cells were washed twice with PBS, and fresh CM was added onto cells. Cells were let to grow over a period of 72 hours, and pictures were taken every 2−3 hours with an IncuCyte S3 Live Cell Analysis System microscope (Essen BioScience). Cell confluency was analyzed for each timepoint using the IncuCyte ZOOM software analyzer (v2016B). A total of 50,000−100,000 L1475(luc) cells were seeded in six-well plates, and, the next morning, cells were washed with PBS, and CM from senescent plates was added onto the plates. At specified timepoints, cells were trypsinised, and 50 µl of Precision Count Beads (BioLegend) was added to each tube for cell count by flow cytometry using an LSRFortessa Cell Analyzer (BD Biosciences). Data were recorded using FACSDiva software (version 8.0.1) and analyzed with FlowJo software (version 10.8.1) (Tree Star).
Scratch wound cell migration assay
A total of 50,000 cells were seeded in a 96-well plate, and, when cells reached 100% confluency, the WoundMaker tool (Essen BioScience) was applied to create a horizontal wound on each well. Cells were washed once with PBS, and then CM was added onto each well. Cells were allowed to migrate over a period of 42 hours, and pictures were taken every 2 hours with an IncuCyte S3 Live Cell Analysis System microscope (Essen BioScience). Relative wound confluency was analyzed for each timepoint using the IncuCyte ZOOM software analyzer.
Colony formation assay
A total of 500 cells per well were seeded in six-well plates and let to attach overnight. The next morning, cells were carefully washed, and CM from senescent and control cells was added to the cells. Fresh CM was replaced every 24−48 hours, and colonies were let to grow for 10−14 days. At the end of the experiment, colonies were washed with PBS and fixed with 4% paraformaldehyde (PFA) for 10 minutes. They were permeabilized in ice-cold methanol for 20 minutes, dried and stained with 0.2% crystal violet (Acros Organics) in 20% methanol for 30 minutes. Plates were washed with distilled water, colony plates were scanned and the number of colonies was analyzed using Fiji software (version 2.1.0/1.53c).
Low attachment sphere-forming assay
Cover glasses were placed inside six-well plates, and 0.01% polylysine was added and allowed to set for 30 minutes. A total of 2,000 cells per well were resuspended in collected CM, and 10 ng ml−1 rhFGF (R&D Systems), 10 ng ml−1 hEGF (Invitrogen) and 1% N-2 Supplement (Life Technologies) were added. Cells were seeded onto coverslips and allowed to form spheres for up to 14 days. Supplemented CM was replaced every 48 hours, and, at the end of the experiment, spheres were fixed with 4% PFA for 20 minutes and imaged with an AxioScan.Z1 microscope and ZEN Blue software (version 2.6) at ×10 magnification. Sphere number and shape were analyzed using Fiji software.
Three-dimensional Matrigel co-culture assay
For three-dimensional direct co-culture experiments, a total of 5,000 non-senescent cells were plated in the middle of a 24-well plate immersed in Matrigel (Corning). After matrix solidification for 20 minutes at 37 °C, 15,000 senescent cells were carefully seeded surrounding the solidified three-dimensional droplet. Spheres were allowed to grow for 5−7 days, and representative pictures were taken with an EVOS Cell Imaging System microscope (Thermo Fisher Scientific) and analyzed using Fiji software.
Cell mitochondrial stress test
To assess oxygen consumption rate (OCR) and extracellular acidification rate (ECAR) of senescent, non-senescent and SASP recipient cells, a total of 40,000 cells were seeded in an XFe24 cell culture microplate. Once attached, fresh media or CM were added (supplemented with 25 mM glucose, 1 mM pyruvate and 4 mM glutamine). Cells were incubated for 2 hours at 37 °C with atmospheric CO2 in a non-humidified incubator. Previously calibrated Seahorse cartridges were loaded with 2 μM oligomycin, 1 μM carbonyl cyanide-p-trifluoromethoxyphenylhydrazone (FCCP), 1 μM rotenone and 1 μM antimycin A. The cartridge was placed onto a microplate containing cells and inserted into a Seahorse XF-24 Flux Analyzer (Agilent Technologies). Three measurements of 2 min-mix, 2 min-wait and 4 min-measure were carried out at basal conditions and after each drug injection into the wells. After measurements, cells were washed with PBS, and protein was extracted with RIPA lysis medium for quantification. OCR and ECAR values were normalized to total micrograms of protein per well.
Gene expression analysis (RNA extraction and qPCR)
For gene expression analyses, RNA was extracted using the RNeasy Mini Kit (Qiagen), and, where applicable, cDNA was synthesized with the High-Capacity RNA-to-cDNA Kit (Thermo Fisher Scientific) following the manufacturer’s instructions.
Gene expression was measured by quantitative real-time PCR performed on a QuantStudio Thermocycler (Applied Biosystems) using QuantStudio Design & Analysis Software (version 1.5.1) following Luna Universal qPCR Master Mix (New England Biolabs) protocol and amplification parameters and using predesigned KiCqStart SYBR Green Primers (see details in Supplementary Table 6). Relative quantification was carried out using 2−ΔΔCt methodology.
For bulk RNA-seq, RNA samples were assessed for optimal quality using the Agilent RNA ScreenTape System (Agilent Technologies) and submitted to BGI Sequencing Services for library preparation and sequencing (75 × 2 reads). Reads were aligned to the reference genome (either hg38 for human data or mm10 for mice data) using STAR (version_2.6.1d) in two-pass mode following STAR best practices and recommendations57. The quality of the data was evaluated using STAR (version_2.6.1d)57 and SAMtools (version 1.9)58,59. PCR duplicates were removed from aligned BAM files using SAMtools (version 1.9)58. Read counts were extracted from the aligned BAM files using subread’s FeatureCounts (version 1.6.4)60. Normalization of read counts for analysis was done according to edgeR recommendations using the ratio of the variance method, which accounts for intersample variance, and differential expression analysis of the normalized read counts between the sample groups was performed following best practices and recommendations of edgeR61–63 and limma64 on the R environment (version 3.0.6). Pathway overrepresentation analyses were performed as implemented in Kyoto Encyclopedia of Genes and Genomes (KEGGREST version 1.24.1)65, Gene Ontology (GOSEQ version 1.36.0)66 and GSEA (FGSEA version 1.10.1)67–69 R packages. RNA-seq raw data are available in the European Molecular Biology Laboratory’s European Bioinformatics Institute database: European Nucleotide Archive repository accession number PRJEB52271 (https://www.ebi.ac.uk/ena/browser/home), and code used for data analyses is available upon reasonable request.
Microarray dataset analysis
The microarray dataset was downloaded from the Gene Expression Omnibus (GEO) (GSE15622) and processed using R Bioconductor packages (affy version 1.76.0, affycoretools version 1.76.0, limma version 3.54.2 and hgu133a.db version 3.13.0). Differentially expressed genes were defined based on a 0.05 threshold for the adjusted P values (Benjamini−Hochberg multiple testing correction). The GSEA was performed using the R Bioconductor package fgsea version 1.22.0 using the pretreatment samples as a baseline and selecting gene sets with adjusted P values less than 0.05 (Benjamini−Hochberg) from the WikiPathways (2024) and BioPlanet (2019) databases downloaded from the EnrichR website (https://maayanlab.cloud/Enrichr/#libraries) as well as the SenMayo senescence (human) gene set70. NESs were calculated as part of GSEA (https://www.gsea-msigdb.org/gsea/doc/GSEAUserGuideTEXT.htm#_Normalized_Enrichment_Score). Heatmaps were plotted using the R package ComplexHeatmap (version 2.14.0). Network plots were designed using the R packages graph (version 2.2.1) and graphlayouts (version 1.2.2).
shRNA knockdown
For TGFBR1 and Tgfbr1 knockdowns, viral particles were produced in 293T cells upon transfection with pCMVAR8.91, pMD2.G and shRNA plasmids mixed at a 1:2:3 proportion, respectively, with Lipofectamine 2000 (Thermo Fisher Scientific) in Opti-MEM medium (Thermo Fisher Scientific). Cells were incubated with the mix overnight, and fresh medium was added the next morning. After a 48-hour incubation, supernatant was collected and filtered. Viral particles were added onto A549 and L1475(luc) cells, and positive selection was performed with puromycin treatment for 5 days. RNA was then extracted as described above, and lines with the highest knockdown were used for further assessment and experiments. TGFBR1 knockdown in A549 cells also included generation of viral particles produced in 293T cells upon PEI-mediated co-transfection of plasmids psPAX2 and pMD2.G. Backbone LT3-GEPIR (Addgene, 111177) was used to make constructs to knock down TGFBR1. Viral supernatants were harvested 48 hours after transfection and filtered through a 0.45-µm PVPD sterile filter. A549 cells were seeded on a six-well plate 1 day before transduction, and virus supernatant was added to each well with medium containing 8 μg ml−1 polybrene (Millipore). Twenty-four hours after transduction, the media were changed to fresh complete DMEM. Puromycin selection (4 μg ml−1) was started 48 hours after transduction for 2 days at least. Clone IDs, target sequences and further details of the shRNA constructs used can be found in Supplementary Table 7.
Western blotting
Protein was extracted using radioimmunoprecipitation assay (RIPA) buffer (Sigma-Aldrich) supplemented with 1 mM EDTA, cOmplete EDTA-free EASYpak protease inhibitor cocktail (Roche) and PhosSTOP EASYpak phosphatase inhibitor cocktail (Roche). Lysates were incubated on ice for 15 minutes and centrifuged at 14,000g for 15 minutes. Protein concentration in supernatant was quantified using the Pierce BCA Protein Assay Kit (Thermo Fisher Scientific). A total of 30 μg of protein per sample was diluted in Laemmli Sample Buffer (Bio-Rad) and run for 50 minutes at 120 V into a Mini-PROTEAN TGX Precast Gel. Proteins were electrotransferred from the gel onto a PDVF membrane by wet tank transfer overnight at 4 °C. Membrane was washed with Tris-buffered saline with 1% Tween 20 (TBS-T) and blocked in 5% milk solution. Membranes were incubated with primary antibodies overnight at 4 °C, washed three times with TBS-T buffer and incubated with HRP-conjugated secondary antibodies for 1 hour at room temperature. Finally, membranes were incubated with Enhanced Chemiluminescence (ECL) Detection Solution (Amersham) and imaged using a Xograph Compact X4 automatic processor or a ChemiDoc Imager (Bio-Rad) using Image Lab (version 6.1). A full list of antibodies used for western blotting can be found in Supplementary Table 8.
ELISA
Senescent and control cells were washed, and FBS-free fresh media were conditioned for 72 hours. CM was then collected, centrifuged for 10 minutes at 1,000g and immediately stored at −80 °C for ELISA analysis of TGFβ ligands. Prior to the assay, ligands were activated by incubating the samples with 1 N HCl for 10 minutes and neutralizing them with 1.2 N NaOH/0.5 M HEPES solution. Samples were assayed immediately after using the Quantikine Human TGF-β1 Immunoassay Kit (R&D Systems), the Human TGF-beta 2 Quantikine ELISA Kit (R&D Systems) and the Human TGF-beta 3 DuoSet ELISA Kit (R&D Systems).
Phospho-kinase array
Changes in kinase phosphorylation upon exposure to the SASP was assayed using the Proteome Profiler Human Phospho-Kinase Array (R&D Systems). In brief, A549 cells were exposed to FBS-free CM from control and senescent cells (with either vehicle or galunisertib) for 30 minutes at 37 °C. Protein was extracted immediately after as described above, and protein concentration was calculated following the Pierce BCA Protein Assay Kit (Thermo Fisher Scientific). A total of 600 μg was immediately used for the assay following the manufacturer’s indications. Membranes from the kit were incubated with ECL solution (Amersham), and X-ray film was exposed to 1 minute and 10 minutes. Pixel intensity was calculated using Fiji software (version 2.1.0/1.53c) and normalized against loading controls.
MEMA
The MEMA platform28,29 was used as a high-throughput technology to determine the impact of different SASP ligands on A549 cell proliferation. In brief, MEMA plates were blocked for 20 minutes with 1% non-fouling blocking agent and rinsed three times with PBS. A total of 3 × 104 A549 cells per well were seeded in DMEM medium containing 0.1% FBS. After 18 hours of adhesion, each well was supplemented with an experimental SASP ligand. Cells were allowed to grow for up to 72 hours and were then incubated with 10 µM 5-ethynyl-2′-deoxyuridine (EdU) for 1 hour prior to fixation. Next, cells were fixed with 2% PFA for 15 minutes at room temperature and stored at 4 °C in PBS. Cells were then permeabilized with 0.1% Triton X-100 for 15 minutes, washed with PBS, washed with 0.05% Tween 20 PBS (PBS-T) and incubated with Click-iT EdU detection reaction reagents (Click-iT EdU Cell Proliferation Kit (Thermo Fisher Scientific)) for 1 hour while protected from light. After quenching, cells were rinsed with PBS-T and subjected to immunofluorescence by using anti-KRT5 and anti-KRT19 antibodies (Supplementary Table 8). Cells were then washed again with PBS-T, stained with secondary antibodies, washed with PBS-T, washed with and stored in PBS and imaged on an automated imaging system. Segment cells and intensity were calculated using CellProfiler. The R environment with custom code was used to normalize, correct variations and summarize the raw Cell Profiler-derived data for each condition. Custom code and raw data are accessible on GitHub (https://github.com/markdane/A549_low_serum/tree/main/R) and Synapse (ID: syn27665118).
Full lists of ligands used for MEMA experiments, UniProt IDs and concentrations are found in Supplementary Table 9.
SA-β-gal staining
Upon senescence induction, cells were washed with PBS and fixed and stained for SA-β-gal using the Senescence β-Galactosidase Staining Kit (Cell Signaling Technology) following the manufacturer’s instructions. Stained cells were imaged using an Olympus Compact Brightfield Modular Microscope (Life Technologies) and ZEN Blue (version 2.6). Resected lungs from experimental mice were also stained whole-mount for SA-β-gal following the same procedure and kit.
SenTraGor staining
SenTraGor staining was performed as described previously37,38,71. Whole slides were digitized using an AxioScan Microscope Slide Scanner (Zeiss) using polarized light, and ZEN Blue (version 2.6.) was used for image acquisition. The mean percentage of GL13-positive cells in at least 5−10 high-power fields (×400) per sample was quantified. For each power field, at least 100 cells were measured.
Lung processing for fluorescence-activated cell sorting and immune system phenotyping
For the mechanical dissociation of tissue, lungs were transferred to a conical tube containing dissociation solution (3 mg ml−1 Collagenase IV, 1% 1 M HEPES and 1% FBS) and incubated for 30 minutes at 37 °C in a rocking platform with regular vortexing. Washing solution (consisting of DMEM + 10% FBS) was added, and cell pellets were filtered using 100/70-μm filters. Then, 5 ml of DMEM was employed to wash through any remaining cells in the filter and spun down for 5 minutes at 4 °C. Supernatant was discarded, and the cell pellet was resuspended in 1 ml of RBC lysis buffer (ACK buffer, 0.15 M NH4Cl, 10 mM KHCO3 and 0.1 mM EDTA) and lysed for 3–5 minutes on ice. One milliliter of fluorescence-activated cell sorting (FACS) buffer (1% BSA and 5 mM EDTA in PBS/Cell Staining Buffer; BioLegend) was added and passed through a 40-μm filter. Samples were then centrifuged at 300g for 5 minutes at 4 °C. The supernatant was removed, and the cell pellet was resuspended in 3× FACS buffer for further staining. The antibodies used are as stated in Supplementary Table 8. Data were acquired on an LSRFortessa cell analyzer running FACSDiva software (BD Biosciences) and then analyzed with FlowJo software.
Tissue sectioning and IHC
Lungs and subcutaneous tumors were collected and fixed in 10% formalin overnight. Samples were then embedded in paraffin and cut in 3−7-μm-thick sections. Slides were deparaffinized in xylene and re-hydrated through a series of graded ethanol until water. Antigen retrieval and IHC were performed by the Spanish National Cancer Research Centre (CNIO) Histopathology Service. Whole slides were digitalized using an AxioScan Microscope Slide Scanner (Zeiss) using polarized light, and ZEN Blue (version 2.6.) was used for image acquisition. Analysis of IHC staining was carried out using HALO (version 3.3.2541) (Indica Labs). An automated analysis method for each staining was established based on Indica Labs’ CytoNuclear version 2.0.9 settings. A positive control for each stain was used to establish optimal settings that were then used for all samples, and the percent positive stain was determined across the whole lesion of interest. A full list of antibodies used for IHC can be found in Supplementary Table 8.
For HGSOC TMAs, automated staining was carried out using the Leica Polymer Refine Detection System (DS9800), using a modified standard template on the BOND-III platform. Post-primary antibody and polymer (anti-rabbit Poly-HRP-IgG) were obtained from the associated kit. Antibodies are detailed in Supplementary Table 8. Retrieval pretreatments were run at 100 °C, and DAB enhancer (Leica, AR9432) was used for all antibodies.
Statistics and reproducibility
GraphPad Prism software (version 9) was used for statistical analysis. All data are displayed as mean ± s.d. unless stated otherwise. Group sizes were determined based on the results of preliminary experiments and discussions with a senior statistician. Group allocation was performed with simple randomization. Normality was tested using the Shapiro−Wilk test, and equality of variances was assessed using the F-test. For normally distributed data with equal variance, statistical significance was determined using two-tailed unpaired Student’s t-test. Welch’s correction was performed for samples with unequal variance. The two-sided Mann–Whitney U-test was performed for datasets without normal distribution; one-way ANOVA was used for comparison of more than two samples; and two-way ANOVA was used to analyze data with two variables, such as cell confluency or tumor growth over time. Tukey’s multiple comparisons test was performed to compare the mean of each group with the mean of every other group after ANOVA testing. A two-sided log-rank test was performed to analyze survival for categorical variables. Investigators were blinded to group allocation during data analysis for µCT scans. The ROUT coefficient, Q = 1, as recommended by GraphPad Prism, was set up to identify and remove outliers for immunostaining quantifications, but, otherwise, no data or samples were excluded from analyses under the experimental designs. P values less than 0.05 were considered significant.
Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.
Supplementary information
Supplementary Figs. 1−15, Supplementary Uncropped Immunoblots 1 and 2 and Supplementary Tables 1−9.
Source data
Single Excel file containing all source data for Figs. 1–8 and Extended Data Figs. 1−10.
PDF file containing original immunoblots for Fig. 3.
Acknowledgements
We acknowledge the support of the University of Cambridge, Cambridge University Hospitals NHS Foundation Trust and CRUK and thank all patients who participated and donated tissue samples to this study. The Muñoz-Espín laboratory was supported by the CRUK Cambridge Centre Early Detection Programme (RG86786), by a CRUK Programme Foundation Award (C62187/A29760), by a CRUK Early Detection OHSU Project Award (C62187/A26989) and by a Medical Research Council New Investigator Research Grant (NIRG) (MR/R000530/1). E.G.-G. was holder of a ‘La Caixa’ Foundation Scholarship for postgraduate studies at European universities and funded by the CRUK Cambridge Centre Early Detection Programme. D.M. was funded by an NIRG (MR/R000530/1) and a CRUK Programme Foundation Award (C62187/A29760). S.M. was funded by a CRUK Programme Foundation Award (C62187/A29760). J.G. was funded by a Darley/Sands Downing College Fellowship (G109261). H-L.O. was funded by a CRUK Early Detection OHSU Project Award (C62187/A26989). M.D. was funded by an Evelyn Trust Fellowship (G113971) and CRUK Cambridge Centre Fellowship (G118236). R.K. was funded by a CRUK Cambridge Centre Fellowship (G125208). V.G. was financially supported by the European Regional Development Fund of the European Union and Greek national funds through the Operational Program Competitiveness, Entrepreneurship and Innovation, under the call RESEARCH − CREATE – INNOVATE (project code: T1EDK02939), and NKUA-SARG grant 70/3/8916. This research was supported by the NIHR Cambridge Biomedical Research Centre (BRC-1215-20014) through funding to R.C.R. and D.M.R. G.J.D. was supported by funding through the CRUK Cambridge Centre, Thoracic Cancer Programme. M.N. and I.O. were supported by the CRUK Cambridge Institute’s Core Grant (C9545/A29580) and a CRUK Pioneer Award (C63389/A30462). J.D.B. was supported by CRUK and CRUK Cambridge Centre (A25177, A22905, CTRQQR-2021-100012 and -100005) and the National Institute for Health and Care Research (NIHR203312) and Cambridge Biomedical Research Centre (BRC-1215-20014). M.A.V.R. was supported by a CRUK Cambridge Centre Clinical Research Training Fellowship (A25177). We thank the Addenbrooke’s Hospital Gynaecological Oncology clinical team and the OV04 study team for facilitating clinical sample collection, the Addenbrooke’s Human Research Tissue Bank and the CRUK Cambridge Institute Compliance and Biobanking Team for sample management and the CRUK Cambridge Institute Histopathology Core Facility for ovarian tissue processing. We thank M. Vias, J. Miller and J. Jones for discussions around ovarian assay development. We thank the Anne McLaren Building Animal Facility and its Imaging Facility for acquisition of µCT scans and the CNIO Core Imaging Unit for µCT scan analyses. We thank D. Kirsch (Duke University, UK) for the provision of the p53frt/frt mouse allele. We also thank M. Serrano (Altos Labs, Granta Park, Cambridge, UK) for critical reading and insightful comments. The views expressed are those of the authors and not necessarily those of the NIHR or the Department of Health and Social Care. We also thank BioRender.com.
Extended data
Author contributions
Project supervision and funding acquisition: D.M.-E. Conceptualization and design of the project: D.M.-E. and E.G.-G. Experimental designs: D.M.-E., E.G.-G. and D.M. Methodology: E.G.-G., D.M., S.M. and J.G. (in vivo experiments); E.G.-G. (proliferation, colony formation, wound healing, sphere formation and co-culture assays; cell mitochondrial stress test; and molecular biology and mechanistic assays); J.G. (gene knockdown and mechanistic experiments); I.O. (bioinformatic RNA-seq analyses from the OV01 dataset); J.E.M. (RNA-seq analyses); H-L.O. (MEMA analyses and experiments with carboplatin and PEO4 cells); M.H. and L.F. (histology analyses of orthotopic transplantation model); M.P.P.-C. (apoptosis assays); M.D. (RT−qPCR experiments); R.K. (toxicity assays); R.H., M.D. and J.E.K. (MEMA analyses); G.M. and F.M. (CT scan analyses and three-dimensional rendering); C.P.M. (generation of KP reagents and guidance with their in vivo transplantation); M.B. (oversight and generation of KrasFSFG12V/+-driven lung cancer murine model); V.G. (SenTraGor analysis of human tissue sections); D.R., G.J.D. and R.C.R. (provision of human samples from the RPHRTB and clinical oversight); M.N. (omics oversight and resources); M.A.V.R. (HGSOC funding acquisition, project design, TMA dataset generation, clinical data curation, image quality control and processing, cell segmentation, tumor stroma classification, computational analyses and clinical oversight); M.J.L. (TMA dataset generation, tumor stroma classification and clinical oversight); and J.D.B. (HGSOC resources, funding acquisition, project design and clinical oversight). Data analysis and curation: D.M.-E., E.G.-G., M.A.V.R., I.O., M.J.L., M.N. and J.D.B. Writing of the manuscript: D.M.-E., E.G.-G., M.A.V.R. and J.D.B., with input from all authors. All authors commented on the article text. E.G.-G. and M.A.V.R. share co-first authorship and led the NSCLC and HGSOC experiments, respectively. D.M., S.M., J.G. and I.O. made important inputs to this work and contributed equally as co-second authors. D.M.-E., J.D.B. and M.N. are senior authors.
Peer review
Peer review information
Nature Aging thanks Gerardo Ferbeyre and the other, anonymous, reviewer(s) for their contribution to the peer review of this work.
Data availability
All the datasets used in this study are appropriately accessible via the following accession codes. The bulk RNA-seq dataset generated to support the findings of this study is available in the European Molecular Biology Laboratory’s European Bioinformatics Institute database: European Nucleotide Archive repository accession number PRJEB52271 (https://www.ebi.ac.uk/ena/browser/home). A549 MEMA data are available on Synapse (https://www.synapse.org) (ID: syn27665118). Source Data are provided with this paper. All other data needed to evaluate the conclusions of this study are present in the article or are available from the corresponding author upon reasonable request.
Code availability
Published software used for data analysis is provided in the Methods. The A549 MEMA custom code is available on GitHub (https://github.com/markdane/A549_low_serum) and is also available from the corresponding author upon reasonable request.
Competing interests
E.G-G. is now an employee of Eli Lilly and Company, Spain. G.J.D. and C.P.M. are now employees of AstraZeneca, UK. J.D.B. is a cofounder and shareholder of Tailor Bio. The rest of the authors declare no competing interests.
Footnotes
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
These authors contributed equally: Estela González-Gualda, Marika A. V. Reinius.
These authors jointly supervised this work: Masashi Narita, James D. Brenton, Daniel Muñoz-Espín.
Extended data
is available for this paper at 10.1038/s43587-025-01054-2.
Supplementary information
The online version contains supplementary material available at 10.1038/s43587-025-01054-2.
References
- 1.Herbst, R. S., Morgensztern, D. & Boshoff, C. The biology and management of non-small cell lung cancer. Nature553, 446–454 (2018). [DOI] [PubMed] [Google Scholar]
- 2.Drews, R. M. et al. A pan-cancer compendium of chromosomal instability. Nature606, 976–983 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Macintyre, G. et al. Copy number signatures and mutational processes in ovarian carcinoma. Nat. Genet.50, 1262–1270 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Vaughan, S. et al. Rethinking ovarian cancer: recommendations for improving outcomes. Nat. Rev. Cancer11, 719–725 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.França, G. S. et al. Cellular adaptation to cancer therapy along a resistance continuum. Nature631, 876–883 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Marine, J.-C., Dawson, S.-J. & Dawson, M. A. Non-genetic mechanisms of therapeutic resistance in cancer. Nat. Rev. Cancer20, 743–756 (2020). [DOI] [PubMed] [Google Scholar]
- 7.McGranahan, N. & Swanton, C. Clonal heterogeneity and tumor evolution: past, present, and the future. Cell168, 613–628 (2017). [DOI] [PubMed] [Google Scholar]
- 8.Roberson, R. S., Kussick, S. J., Vallieres, E., Chen, S.-Y. J. & Wu, D. Y. Escape from therapy-induced accelerated cellular senescence in p53-null lung cancer cells and in human lung cancers. Cancer Res.65, 2795–2803 (2005). [DOI] [PubMed] [Google Scholar]
- 9.Ou, H.-L. et al. Cellular senescence in cancer: from mechanisms to detection. Mol. Oncol.15, 2634–2671 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Di Micco, R., Krizhanovsky, V., Baker, D. & d’Adda di Fagagna, F. Cellular senescence in ageing: from mechanisms to therapeutic opportunities. Nat. Rev. Mol. Cell Biol.22, 75–95 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Gorgoulis, V. et al. Cellular senescence: defining a path forward. Cell179, 813–827 (2019). [DOI] [PubMed] [Google Scholar]
- 12.Faget, D. V., Ren, Q. & Stewart, S. A. Unmasking senescence: context-dependent effects of SASP in cancer. Nat. Rev. Cancer19, 439–453 (2019). [DOI] [PubMed] [Google Scholar]
- 13.Gonzalez-Meljem, J. M., Apps, J. R., Fraser, H. C. & Martinez-Barbera, J. P. Paracrine roles of cellular senescence in promoting tumourigenesis. Br. J. Cancer118, 1283–1288 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Chen, H. et al. TGF-β1/IL-11/MEK/ERK signaling mediates senescence-associated pulmonary fibrosis in a stress-induced premature senescence model of Bmi-1 deficiency. Exp. Mol. Med.52, 130–151 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Li, Z.-Y., Chen, Z.-L., Zhang, T., Wei, C. & Shi, W.-Y. TGF-β and NF-κB signaling pathway crosstalk potentiates corneal epithelial senescence through an RNA stress response. Aging8, 2337–2354 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Teo, Y. V. et al. Notch signaling mediates secondary senescence. Cell Rep.27, 997–1007 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.David, C. J. & Massagué, J. Contextual determinants of TGFβ action in development, immunity and cancer. Nat. Rev. Mol. Cell Biol.19, 419–435 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Senturk, S. et al. Transforming growth factor-beta induces senescence in hepatocellular carcinoma cells and inhibits tumor growth. Hepatology52, 966–974 (2010). [DOI] [PubMed] [Google Scholar]
- 19.Wu, J. et al. TGF-β1 induces senescence of bone marrow mesenchymal stem cells via increase of mitochondrial ROS production. BMC Dev. Biol.14, 21 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Pickup, M., Novitskiy, S. & Moses, H. L. The roles of TGFβ in the tumour microenvironment. Nat. Rev. Cancer13, 788–799 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Lamouille, S. & Derynck, R. Cell size and invasion in TGF-β−induced epithelial to mesenchymal transition is regulated by activation of the mTOR pathway. J. Cell Biol.178, 437–451 (2007). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Zhang, L., Zhou, F. & ten Dijke, P. Signaling interplay between transforming growth factor-β receptor and PI3K/AKT pathways in cancer. Trends Biochem. Sci.38, 612–620 (2013). [DOI] [PubMed] [Google Scholar]
- 23.Calcinotto, A. et al. Cellular senescence: aging, cancer, and injury. Physiol. Rev.99, 1047–1078 (2019). [DOI] [PubMed] [Google Scholar]
- 24.Sanoff, H. K. et al. Effect of cytotoxic chemotherapy on markers of molecular age in patients with breast cancer. J. Natl Cancer Inst.106, dju057 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Demaria, M. et al. Cellular senescence promotes adverse effects of chemotherapy and cancer relapse. Cancer Discov.7, 165–176 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.van Deursen, J. M. The role of senescent cells in ageing. Nature509, 439–446 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.González-Gualda, E. et al. Galacto-conjugation of navitoclax as an efficient strategy to increase senolytic specificity and reduce platelet toxicity. Aging Cell19, e13142 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Smith, R. et al. Using microarrays to interrogate microenvironmental impact on cellular phenotypes in cancer. J. Vis. Exp.10.3791/58957 (2019). [DOI] [PMC free article] [PubMed]
- 29.Watson, S. S. et al. Microenvironment-mediated mechanisms of resistance to HER2 inhibitors differ between HER2+ breast cancer subtypes. Cell Syst.6, 329–342 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Batlle, E. & Massagué, J. Transforming growth factor-β signaling in immunity and cancer. Immunity50, 924–940 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Massague, J. TGFβ signalling in context. Nat. Rev. Mol. Cell Biol.13, 616–630 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Massagué, J. TGFβ in cancer. Cell134, 215–230 (2008). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Benjamin, D., Colombi, M., Moroni, C. & Hall, M. N. Rapamycin passes the torch: a new generation of mTOR inhibitors. Nat. Rev. Drug Discov.10, 868–880 (2011). [DOI] [PubMed] [Google Scholar]
- 34.Sanclemente, M. et al. c-RAF ablation induces regression of advanced Kras/Trp53 mutant lung adenocarcinomas by a mechanism independent of MAPK signaling. Cancer Cell33, 217–228 (2018). [DOI] [PubMed] [Google Scholar]
- 35.Lee, C.-L. et al. Generation of primary tumors with Flp recombinase in FRT-flanked p53 mice. Dis. Model. Mech.5, 397–402 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Zhu, Y. et al. Identification of a novel senolytic agent, navitoclax, targeting the Bcl-2 family of anti-apoptotic factors. Aging Cell15, 428–435 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Evangelou, K. et al. Robust, universal biomarker assay to detect senescent cells in biological specimens. Aging Cell16, 192–197 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Kohli, J. et al. Algorithmic assessment of cellular senescence in experimental and clinical specimens. Nat. Protoc.16, 2471–2498 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Ahmed, A. A. et al. The extracellular matrix protein TGFBI induces microtubule stabilization and sensitizes ovarian cancers to paclitaxel. Cancer Cell12, 514–527 (2007). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40.Crispin-Ortuzar, M. et al. Integrated radiogenomics models predict response to neoadjuvant chemotherapy in high grade serous ovarian cancer. Nat. Commun.14, 6756 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41.Chaib, S., Tchkonia, T. & Kirkland, J. L. Cellular senescence and senolytics: the path to the clinic. Nat. Med.28, 1556–1568 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42.Hanahan, D. Hallmarks of cancer: new dimensions. Cancer Discov.12, 31–46 (2022). [DOI] [PubMed] [Google Scholar]
- 43.Prasanna, P. G. et al. Therapy-induced senescence: opportunities to improve anticancer therapy. J. Natl Cancer Inst.113, 1285–1298 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44.Wang, L., Lankhorst, L. & Bernards, R. Exploiting senescence for the treatment of cancer. Nat. Rev. Cancer22, 340–355 (2022). [DOI] [PubMed] [Google Scholar]
- 45.Versteeg, K. S., Konings, I. R., Lagaay, A. M., van de Loosdrecht, A. A. & Verheul, H. M. W. Prediction of treatment-related toxicity and outcome with geriatric assessment in elderly patients with solid malignancies treated with chemotherapy: a systematic review. Ann. Oncol.25, 1914–1918 (2014). [DOI] [PubMed] [Google Scholar]
- 46.Rodier, F. et al. Persistent DNA damage signalling triggers senescence-associated inflammatory cytokine secretion. Nat. Cell Biol.11, 973–979 (2009). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 47.Zhuang, X. et al. Ageing limits stemness and tumorigenesis by reprogramming iron homeostasis. Nature637, 184–194 (2025). [DOI] [PubMed] [Google Scholar]
- 48.Shuldiner, E. G. et al. Aging represses oncogenic KRAS-driven lung tumorigenesis and alters tumor suppression. Nat. Aging5, 2263–2278 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 49.Patel, A. A. H. et al. Aging promotes lung cancer metastasis through epigenetic ATF4 induction. Preprint at bioRxiv10.1101/2024.07.03.601209 (2024).
- 50.Tominaga, K. & Suzuki, H. I. TGF-β signaling in cellular senescence and aging-related pathology. Int. J. Mol. Sci.20, 5002 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 51.Haga, M., Iida, K. & Okada, M. Positive and negative feedback regulation of the TGF-β1 explains two equilibrium states in skin aging. iScience27, 109708 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 52.du Bois, A. et al. Role of surgical outcome as prognostic factor in advanced epithelial ovarian cancer: a combined exploratory analysis of 3 prospectively randomized phase 3 multicenter trials: by the Arbeitsgemeinschaft Gynaekologische Onkologie Studiengruppe Ovarialkarzinom (AGO-OVAR) and the Groupe d’Investigateurs Nationaux Pour les Etudes des Cancers de l’Ovaire (GINECO). Cancer115, 1234–1244 (2009). [DOI] [PubMed] [Google Scholar]
- 53.Ciardiello, D., Elez, E., Tabernero, J. & Seoane, J. Clinical development of therapies targeting TGFβ: current knowledge and future perspectives. Ann. Oncol.31, 1336–1349 (2020). [DOI] [PubMed] [Google Scholar]
- 54.Malek, E. et al. The TGFβ type I receptor kinase inhibitor vactosertib in combination with pomalidomide in relapsed/refractory multiple myeloma: a phase 1b trial. Nat. Commun.15, 7388 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 55.Jackson, E. L. et al. The differential effects of mutant p53 alleles on advanced murine lung cancer. Cancer Res.65, 10280–10288 (2005). [DOI] [PubMed] [Google Scholar]
- 56.Turrell, F. K. et al. Lung tumors with distinct p53 mutations respond similarly to p53 targeted therapy but exhibit genotype-specific statin sensitivity. Genes Dev.31, 1339–1353 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 57.Dobin, A. et al. STAR: ultrafast universal RNA-seq aligner. Bioinformatics29, 15–21 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 58.Bonfield, J. K. et al. HTSlib: C library for reading/writing high-throughput sequencing data. Gigascience10, giab007 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 59.Danecek, P. et al. Twelve years of SAMtools and BCFtools. Gigascience10, giab008 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 60.Liao, Y., Smyth, G. K. & Shi, W. The R package Rsubread is easier, faster, cheaper and better for alignment and quantification of RNA sequencing reads. Nucleic Acids Res.47, e47 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 61.Chen, Y., Lun, A. T. L. & Smyth, G. K. From reads to genes to pathways: differential expression analysis of RNA-Seq experiments using Rsubread and the edgeR quasi-likelihood pipeline. F1000Res.5, 1438 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 62.McCarthy, D. J., Chen, Y. & Smyth, G. K. Differential expression analysis of multifactor RNA-Seq experiments with respect to biological variation. Nucleic Acids Res.40, 4288–4297 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 63.Robinson, M. D., McCarthy, D. J. & Smyth, G. K. edgeR: a Bioconductor package for differential expression analysis of digital gene expression data. Bioinformatics26, 139–140 (2010). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 64.Ritchie, M. E. et al. limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Res.43, e47 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 65.Kanehisa, M., Furumichi, M., Tanabe, M., Sato, Y. & Morishima, K. KEGG: new perspectives on genomes, pathways, diseases and drugs. Nucleic Acids Res.45, D353–D361 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 66.Young, M. D., Wakefield, M. J., Smyth, G. K. & Oshlack, A. Gene ontology analysis for RNA-seq: accounting for selection bias. Genome Biol.11, R14 (2010). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 67.Korotkevich, G. et al. Fast gene set enrichment analysis. Preprint at bioRxiv10.1101/060012 (2016).
- 68.Mootha, V. K. et al. PGC-1α-responsive genes involved in oxidative phosphorylation are coordinately downregulated in human diabetes. Nat. Genet.34, 267–273 (2003). [DOI] [PubMed] [Google Scholar]
- 69.Subramanian, A. et al. Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proc. Natl Acad. Sci. USA102, 15545–15550 (2005). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 70.Saul, D. et al. A new gene set identifies senescent cells and predicts senescence-associated pathways across tissues. Nat. Commun.13, 4827 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 71.Evangelou, K. et al. Pulmonary infection by SARS-CoV-2 induces senescence accompanied by an inflammatory phenotype in severe COVID-19: possible implications for viral mutagenesis. Eur. Respir. J.60, 2102951 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Supplementary Figs. 1−15, Supplementary Uncropped Immunoblots 1 and 2 and Supplementary Tables 1−9.
Single Excel file containing all source data for Figs. 1–8 and Extended Data Figs. 1−10.
PDF file containing original immunoblots for Fig. 3.
Data Availability Statement
All the datasets used in this study are appropriately accessible via the following accession codes. The bulk RNA-seq dataset generated to support the findings of this study is available in the European Molecular Biology Laboratory’s European Bioinformatics Institute database: European Nucleotide Archive repository accession number PRJEB52271 (https://www.ebi.ac.uk/ena/browser/home). A549 MEMA data are available on Synapse (https://www.synapse.org) (ID: syn27665118). Source Data are provided with this paper. All other data needed to evaluate the conclusions of this study are present in the article or are available from the corresponding author upon reasonable request.
Published software used for data analysis is provided in the Methods. The A549 MEMA custom code is available on GitHub (https://github.com/markdane/A549_low_serum) and is also available from the corresponding author upon reasonable request.