Analysis of molecular subtypes and prognostic signature of senescence-associated secretory phenotype in pancreatic cancer

Introduction

Pancreatic cancer (PC), predominantly pancreatic ductal adenocarcinoma (PDAC), is the most aggressive malignancy of the digestive system, with a five-year survival rate (SR) of ∼12% (Siegel et al., 2023). The pancreas is anatomically concealed; hence, most PC patients are often diagnosed at advanced stages—a phenomenon that is projected to make PC the second deadliest malignancy by 2030 (Nevala-Plagemann, Hidalgo & Garrido-Laguna, 2020; Park, Chawla & O’Reilly, 2021; Rahib et al., 2021). While surgical resection remains the standard intervention for PC, adjuvant therapies, neoadjuvant treatments, immunotherapy, targeted therapy, and radiotherapy (RT) have demonstrated transient efficacy in select populations, although often followed by drug resistance (Mizrahi et al., 2020). Additionally, various factors, including cellular senescence—an emerging focus in cancer research—modulate the tumor microenvironment (TME), highlighting its complex nature, which confers limited response to cytotoxic therapies and immunotherapy in PC patients (Park, Chawla & O’Reilly, 2021).

Cellular senescence is a stress-triggered, irreversible proliferative arrest featuring structural and functional degeneration (Dong et al., 2024; McHugh, Durán & Gil, 2025). Senescence, which was initially recognized as replicative senescence in human fibroblasts, presently encompasses oncogene-induced, therapy-induced, virus-induced, mitochondrial dysfunction-associated, and immune-mediated forms (Wang et al., 2024). Besides growth arrest, senescent cells also exhibit Senescence-Associated Secretory Phenotype (SASP) (Wang et al., 2024), a hyper-secretory profile featuring the secretion of diverse proteins, cytokines, proteases, and chemokines (Dong et al., 2024). According to reports, SASP performs dual roles in tumorigenesis. In addition to inducing cell cycle arrest and immune-mediated damaged cell clearance to suppress tumor initiation, it also fosters immunosuppressive microenvironments conducive to tumor progression and metastasis (Dong et al., 2024; Eggert et al., 2016; Liu & Hornsby, 2007; Orjalo et al., 2009). Furthermore, SASP exerts context-dependent effects in PC. For instance, MEK and CDK4/6 inhibition-induced senescence promoted T-cell infiltration and tumor suppression in murine models (Ohuchida et al., 2004; Ruscetti et al., 2020), whereas radiation-induced fibroblast SASP enhanced PC invasiveness (Ohuchida et al., 2004). Nonetheless, the prognostic implications of specific SASP components in PC remain unclear, necessitating bioinformatics-driven identification of SASP-related prognostic biomarkers.

Herein, we stratified PC patients from the TCGA-ICGC cohort into two SASP subtypes based on the consensus cluster analysis results and established a 7-gene SASP-based prognostic signature. We also conducted functional experiments, validating the role of ANGPTL4 in promoting PC proliferation and metastasis. Our proposed risk stratification model could provide computational evidence to guide prognostic prediction and personalized therapy in PC.

Materials & Methods

Results

Genetic alterations of SASP genes

Figure 1 summarizes the workflow, and Table S1 lists the 77 SASP genes. The merged TCGA-ICGC dataset comprised a cohort of 269 PC patients. Somatic CNV analysis revealed frequent alterations: CCL5, CCL7, and CCL2 exhibited copy number gains, while BMP6, VEGFC, and EDN1 showed losses (Fig. 2A).

Prognostic signature construction and validation

Univariate Cox regression identified 33 prognostic SASP genes (Fig. 3A), of which FGF1 (HR = 0.813), IGFBP2 (HR = 0.862), and VGF (HR = 0.865) emerged as protective factors. Protein-Protein Interaction (PPI) networks highlighted gene connectivity (Fig. 3B), with LASSO regression yielding a 7-gene signature: Risk score = (0.0081 × ANGPTL4) + (0.0086 × DKK1) + (0.0177 × EREG) + (0.0425 × IGFBP3) + (0.1902 × IL18) + (0.1702 × MMP13) + (0.0567 × PLAU) (Range: 0.58–3.22, Fig. 3C). Furthermore, high-risk patients exhibited shorter survival (Figs. 3D, 3E). High-risk patients also exhibited signature gene upregulation (Fig. 3F). The AUC values for the training cohort were 0.722 (1 year), 0.693 (2 years), and 0.677 (3 years); for the test cohort, they were 0.661 (1 year), 0.641 (2 years), and 0.626 (3 years). The overall cohort AUC values were 0.695 (1 year), 0.669 (2 years), and 0.644 (3 years). Compared to traditional clinical characteristics, the risk score demonstrated superior predictive performance in forecasting prognosis for patients with PC (Fig. 3G). The PCA plot demonstrates a clear separation between the two risk groups of patients (Fig. 3H).

Validation of signature genes in single-cell RNA-seq data

Quality control processes yielded 23 cell clusters (Figs. 4A, 4B). Furthermore, annotation revealed 11 cell types, with elevated ductal epithelial cells and fibroblasts in metastatic samples and primary tumors, respectively (Figs. 4C–4E). The signature genes were predominantly expressed in ductal epithelial and myeloid cells (Fig. 4F).

Clinical relevance of the prognostic signature

Higher risk scores correlated with advanced T-stage (Fig. 5A) and SASPcluster B (Fig. 5B), highlighting the significance of SASP in tumor development or risk. Furthermore, most SASP genes showed differential expression between groups (Fig. 5C). Multivariate Cox regression confirmed risk score as an independent prognostic factor (Figs. 5D, 5E), and GSVA and GSEA highlighted the enrichment of the glycolysis, mTORC1, and EMT pathways in high-risk patients (Figs. 5F, 5G). Moreover, the resulting nomogram effectively predicted survival (Figs. 5H, 5I).

Association of prognostic signature with TME

According to CIBERSORT analysis, high-risk patients exhibited naive B cell, memory B cell, plasma cell, and CD8+ T cell downregulation, alongside M0 macrophage and activated Dendritic Cell (DC) upregulation (Figs. 6A, 6B). The risk score exhibited a negative correlation with naïve B cells, plasma cells, and CD8 T cells, whereas it showed a positive correlation with Macrophage M0, Macrophage M2, and activated dendritic cells (Fig. 6C). Furthermore, in high-risk patients, immune functions (e.g., Type I IFN response, checkpoint activity) were suppressed (Fig. 6D), whereas checkpoints CD44, PD-L1, and PDCD1LG2 were upregulated (Fig. 6E). On the other hand, low-risk patients showed enhanced sensitivity to cisplatin, irinotecan, olaparib, and oxaliplatin (Figs. 6F–6I).

Functional validation of ANGPTL4 in PC

Centrosome amplification-induced senescence-associated secretory phenotype variation activated the Hypoxia-Inducible Factor-1α (HIF-1α) and related gene pathways, including migration-promoting factors such as ANGPTL4, highlighting the significance of ANGPTL4 in SASP (Wu et al., 2023). Herein, immunohistochemistry (IHC) analysis revealed ANGPTL4, IL-18, MMP13, and PLAU upregulation in PC tissues (Figs. 7A–7D). Furthermore, pan-cancer analysis revealed ANGPTL4 overexpression in PC, as well as breast, colorectal, and liver cancers (Fig. 7E). High ANGPTL4 expression correlated with a poor prognosis, poor differentiation, and lymph node metastasis (LNM) (Figs. 7F–7J). Additionally, GeneMANIA identified the interactions of ANGPTL4 with LPL, PPARD, and PPARG (Fig. 7K), with ANGPTL4 expression correlating positively with M0 macrophages, and negatively with NK cells activated and T cells CD4+ memory resting (Fig. 7L). We also detected the cellular senescence markers p21 and p16 to validate the construction of the oxidative stress (OS)-induced cellular senescence model (Fig. 8A). According to the ELISA results, senescent cells secreted significantly higher levels of ANGPTL4 compared to the control group, confirming ANGPTL4’s involvement in SASP (Fig. 8B). Functional assays further demonstrated rANGPTL4-driven cell proliferation (Fig. 8C), migration, and invasion (Fig. 8D).

Discussion

The global annual PC diagnosis has doubled over the past two decades, posing substantial challenges to public health systems and society as a whole (Klein, 2021). Furthermore, effective treatments for PC remain limited. Although immunotherapy and targeted therapies demonstrated great promise, the poor response attributed to the unique immunosuppressive TME of PC has introduced new challenges.

Cellular senescence, a biological concept first proposed in 1961, has extended beyond replicative exhaustion to include stress-induced senescence (Dong et al., 2024). Tumor heterogeneity dictates that senescence adopts the “double-edged sword” adage in cancer progression. While senescence restricts uncontrolled proliferation via cell cycle arrest or immune cell recruitment, SASP secretes factors that shape an immunosuppressive TME, thus promoting tumor progression (Dong et al., 2024; McHugh, Durán & Gil, 2025). Despite extensive research on senescence and SASP, their roles in PC remain unclear, forming the basis of this study, which aimed to explore the associations of SASP-related genes with PC molecular subtypes, prognosis, genetic mutations, TME infiltration, and chemosensitivity.

Our findings revealed two SASP-associated clusters (SASPclusters) based on SASP gene expression. Compared to Cluster A, Cluster B exhibited worse prognosis and higher SASP gene expression. Furthermore, a 7-gene prognostic model was constructed using LASSO regression, effectively stratifying the patients into high- and low-risk groups. High-risk patients exhibited signature gene overexpression, predominantly localized to epithelial and myeloid cells in single-cell RNA-seq analyses. Additionally, cancer-associated pathways, including glycolysis, angiogenesis, and EMT, were enriched in Cluster B and high-risk patients. Nacarelli et al. (2019) reported that the HMGA-NAMPT-NAD+ axis could enhance glycolysis and mitochondrial respiration, driving pro-inflammatory SASP. Furthermore, in preclinical PDAC models, therapy-induced senescence promoted angiogenesis via SASP-secreted factors, improving gemcitabine efficacy (Ruscetti et al., 2020). Similarly, senescent fibroblast-secreted SASP promoted EMT in premalignant epithelia (Yang et al., 2018), while myofibroblast-derived SASP facilitated EMT in oral submucous fibrosis (OSF) (Sharma et al., 2021), implying shared mechanisms between Cluster B and high-risk patients.

Chemotherapy and radiotherapy (RT) eliminate cancer cells, while making them, along with normal cells, senescent. Although cellular senescence could inhibit tumor progression, residual senescent cells post-treatment may pose new risks, such as promoting cancer stem cell (CSC) activation and re-proliferation, ultimately causing tumor recurrence (Nardella et al., 2011; Milanovic et al., 2018). Besides alleviating treatment-related side effects, combining senescent cell clearance with existing cancer therapies might also significantly reduce the risk of tumor progression (Xiao et al., 2023). Navitoclax (ABT263) could inhibit the activity of anti-apoptotic proteins, clearing senescent cells. It could also effectively eliminate PARP inhibitor olaparib-induced senescent ovarian and breast cancer cells (Fleury et al., 2019). Notably, navitoclax-galactose (Nav-Gal) could be a potent senescent cell clearance intervention, selectively inducing senescent cell apoptosis, with a senescent cell clearance index higher than that of navitoclax alone. It enhanced the cytotoxic effect of cisplatin in A549 lung cancer (LC) cells. Additionally, Cisplatin+Nav-Gal effectively eliminated senescent LC cells and significantly inhibited tumor growth (González-Gualda et al., 2020). Furthermore, artemisinin could inhibit irinotecan treatment-activated mTOR activity, alleviating irinotecan treatment-induced intestinal senescence. Moreover, drug combinations significantly inhibited colorectal tumor growth in mice (Jia et al., 2023). A combination formulation comprising ch282-5 and chloroquine, rapamycin, oxaliplatin, or navitoclax effectively inhibited colon cancer growth and liver metastasis (Wang et al., 2016). Overall, in addition to effectively inhibiting tumor progression, the combined use of chemotherapy drugs and the senescence-inducing agent navitoclax may improve high-risk patients’ low sensitivity to drugs such as olaparib, cisplatin, irinotecan, and oxaliplatin, providing novel insights for clinical treatment.

Herein, the established model comprised several signature genes: IL18, MMP13, PLAU, IGFBP3, EREG, DKK1, and ANGPTL4. Interleukin-18 (IL-18) could downregulate the anti-aging protein Klotho, promoting fibroblast senescence in pulmonary fibrosis (Zhang et al., 2019a). In tumors, IL-18/IL-12-stimulated γδ T cells directly kill cancer cells and induce senescence in the absence of TCR signaling, exerting potent anti-tumor activity (Schilbach et al., 2020). However, IL-18 is also recognized as a tumor-promoting factor (Zhang et al., 2024). This duality mirrors SASP’s context-dependent roles, where IL-18 exerts contrary effects under various conditions. During EMT, Matrix Metallopeptidase 13 (MMP13), a key regulator of chondrocyte senescence in Osteoarthritis (OA) (Zhang et al., 2019b), is upregulated, contributing to PC metastasis (Rath et al., 2017; Wei et al., 2017). However, its role in PC cell senescence remains unclear. According to reports, hyperglycemia might promote endothelial senescence via the AQR/PLAU axis (Wan et al., 2022). Senescent cancer-associated fibroblasts (CAFs) secrete plasminogen activator urokinase (PLAU), driving PC progression and immunosuppression (Liu et al., 2025). Additionally, PLAU induces neuronal senescence (Sun et al., 2024), highlighting its dual role as an outcome and inducer of SASP. Although senescent Endometrial Mesenchymal Stromal Cell (EMSC)-secreted Insulin-like Growth Factor-Binding Protein 3 (IGFBP3) (Ushakov et al., 2021) could induce senescence via telomerase inhibition, thus suppressing BC proliferation (Kwon et al., 2023), its role in PC senescence remains unclear. On the other hand, Epiregulin (EREG), a senescent stromal cell-secreted soluble factor, was identified as a potential therapeutic target for SASP and could be leveraged to overcome treatment resistance (Wang et al., 2022a; Wang et al., 2022b). While EREG upregulation in PC tissues could stimulate tumor growth, its silencing inactivates ERK/p38 MAPK pathways, thus inhibiting PDAC (Liu et al., 2024; Zhu et al., 2000). Nonetheless, the relationship between EREG and PC senescence remains undefined. Regarding DKK1, its knockout in adipose-derived stem cells was reported to activate Wnt signaling, enhancing cell viability, reducing senescence, and suppressing NF-κB-mediated inflammation (Choi et al., 2024). Exogenous DKK1 partially rescued FEZ1 inhibition-induced growth arrest and senescence in limbal stem/progenitor cells (Zhu et al., 2023). Furthermore, YAP prevented dermal fibroblast senescence and inhibited melanogenesis via DKK1 paracrine signaling (Li et al., 2024). Additionally, ZEB1 activated the DKK1/mutant p53/Mdm2/CtBP axis and suppressed macroH2A1-mediated senescence, promoting tumorigenesis (De Barrios et al., 2017). Finally, ANGPTL4, a key mediator of senescence in diabetic cardiomyopathy, exhibits dual roles. Centrosome amplification-induced SASP promoted tumorigenesis via ANGPTL4 secretion (Wen et al., 2024). Furthermore, ANGPTL4 accelerated KRAS(G12D)-driven acinar-to-ductal metaplasia and PC initiation (Yan et al., 2021). Its overexpression also induced pro-invasive, proliferative, and anti-apoptotic transcriptional programs, conferring gemcitabine resistance and shortening survival in PC (Gordon et al., 2023). Herein, bioinformatics analyses revealed ANGPTL4 expression in epithelial, EMT-like, and perivascular cells, correlating with PC prognosis. Functional experiments further demonstrated that recombinant ANGPTL4 could promote PC cell proliferation, invasion, and migration, although its link to senescence remains unclear, necessitating further investigation.

Herein, high-risk patients exhibited reduced naive B cell, memory B cell, plasma cell, and CD8+ T cell infiltration. While B cells amplify anti-tumor immunity via antibody production, the presence of plasma cells has been associated with favorable immunotherapy responses (Delvecchio et al., 2022; Mirlekar et al., 2022). Additionally, IL-35/STAT3/PAX5/BCL6 signaling-driven transcriptional dysregulation in naïve B cells could impair their differentiation into anti-tumor effector cells (Mirlekar et al., 2022). Fan et al. (2024) reported that DKK1+ tumor cells suppressed CCL19+ fibroblasts and plasma cell infiltration, contributing to immunotherapy resistance in hepatocellular carcinoma (HCC). Elevated DKK1 levels and reduced plasma cell infiltration in high-risk PC patients suggest similar interactions, warranting further validation. Furthermore, PC patients often exhibit suppressed CD8+ T cell infiltration, which is critical for tumor control. Although MEK/CDK4/6 inhibition-induced SASP enhanced CD8+ T cell infiltration and PD-1 inhibitor sensitivity in PDAC models (Ruscetti et al., 2020), our findings of CD8+ T cell downregulation in high-risk patients suggest context-dependent mechanisms. Therefore, given the higher KRAS mutation rates in high-risk patients, combining KRAS pathway inhibitors with PD-1 blockade may enhance CD8+ T cell infiltration, improving outcomes. Primary human senescent CD8+ T cells also exhibit a SASP comprising chemokines, cytokines, and ECM remodeling enzymes (Callender et al., 2018). Furthermore, aging cells could express immune surveillance inhibitors and release ligands for the NKG2D receptor, suppressing their clearance, thus weakening the surveillance of aging cells by NK and cytotoxic CD8+ T cells (Salminen, 2021). In an age-related inflammation model, cells highly expressing PD-L1 were associated with higher levels of SASP, suggesting that senescent cells specifically express PD-L1 to evade monitoring by CD8+ T cells (Wang et al., 2022a; Wang et al., 2022b). Moreover, combinations of KRAS inhibitors could induce RB protein-mediated senescence, inhibiting PDAC proliferation. This senescence-inducing therapy triggers SASP, including pro-angiogenic factors that promote tumor angiogenesis. It is also noteworthy that SASP-mediated Endothelial Cell (EC) activation could stimulate CD8+ T cells to aggregate toward immunologically “cold” tumors, making them sensitive to PD-1 checkpoint blockade (Ruscetti et al., 2020). Consequently, SASP and immune cells might correlate closely. Overall, age-related diseases may evade surveillance by CD8 T cells and others via SASP, while immunotherapy targeting tumors may leverage SASP to reshape the immune IME and achieve antitumor therapeutic effects.

Despite its valuable insights, this study has some limitations. First, we mainly derived SASP genes from published literature, potentially omitting uncharacterized genes. Second, the mechanism linking the seven marker genes in PC with SASP should be verified further in experimental studies. Third, our prognostic model was constructed and validated based solely on public databases. To further validate the reliability of our prognostic signature, additional multicenter clinical cohort studies will be required in the future.

Conclusions

Herein, we developed a novel SASP-associated molecular subtype and prognostic signature, which may aid in predicting PC prognosis, potentially illuminating PC stratification and therapeutic strategies. Although high-risk patients identified with the model might benefit from combined KRAS pathway inhibition and immunotherapy, further rigorous experimental studies will be required to validate this phenomenon.

Supplemental Information