GMPS drives pancreatic cancer progression via suppressing DNA damage, STING activation and cellular senescence

Highlights

• •GMPS is a prognostic biomarker and therapeutic target in PDAC.
• •GMPS knockdown suppresses PDAC proliferation in organoids, and PDX models.
• •GMPS deficiency triggers DNA damage, G1 arrest and cellular senescence.
• •GMPS knockdown activates the STING‑TBK1 innate immune pathway.

Abstract

Background

Pancreatic ductal adenocarcinoma (PDAC) is an extremely aggressive malignancy with scarce viable treatment targets. Metabolic rewiring is a typical characteristic of PDAC, but the roles of many metabolic enzymes remain poorly defined.

Methods

We integrated DepMap and TCGA data to screen metabolic genes essential for PDAC proliferation and poor prognosis. Expression patterns were validated by immunohistochemistry and western blotting in 97 PDAC tissues and multiple cell lines. Genetic manipulation experiments in PDAC cell lines, organoids, and a patient‑derived xenograft model were performed to assess proliferation and tumor growth. RNA sequencing, proteomics, flow cytometry, SA-β-gal staining, and immunofluorescence were used to investigate underlying mechanisms.

Results

Among 2752 metabolic genes, GMPS had the highest hazard ratio and its high expression correlated with worse prognosis. GMPS-high tumors showed increased mutation frequencies of KRAS and TP53, while GMPS-low tumors had higher immune infiltration and drug sensitivity. GMPS protein was elevated in PDAC tissues and cell lines (MIA PaCa-2, Capan-1). GMPS knockdown reduced colony formation, proliferation, organoid growth, and PDX tumor growth, whereas overexpression promoted organoid expansion. Mechanistically, GMPS knockdown induced DNA damage (γH2AX⁺ cells), G1 phase arrest, cellular senescence (upregulation of IL6, MMP2, p21, and increased SA-β-gal staining), and activated the STING-TBK1 pathway.

Conclusions

GMPS drives PDAC proliferation by suppressing senescence and STING signaling. GMPS is a promising therapeutic target and prognostic biomarker in PDAC.

Graphical abstract

Introduction

Pancreatic ductal adenocarcinoma (PDAC) is still among the most lethal malignancies, characterized by its 13% five-year overall survival rate [1]. Treatment of PDAC relies on multimodal approaches. Surgical resection combined with adjuvant chemotherapy represents standard care for resectable PDAC, whereas neoadjuvant therapy is applied to borderline resectable tumors and locally advanced tumors to facilitate downstaging [[2], [3], [4]]. First-line chemotherapy remains the mainstay of treatment for individuals with advanced or metastatic disease [5]. Immunotherapy demonstrates limited clinical activity in PDAC, conferring therapeutic benefits exclusively to the rare MSI-H/dMMR molecular subtypes [6]. Although targeted therapies including PARP inhibitors [7] and KRAS inhibitors [8] have exhibited encouraging preliminary efficacy, they have not yet been incorporated into routine clinical practice. Despite these options, overall prognosis for PDAC remains poor, emphasizing the urgent need for novel therapeutic targets and strategies [9].

Metabolic rewiring is acknowledged as a key hallmark of pancreatic cancer [10]. Aberrant metabolic activity is closely linked to the uncontrolled proliferation inherent in cancer, conferring resistance to pharmaceutical interventions [11,12]. Numerous metabolic enzymes are substantially upregulated in pancreatic cancer cells, facilitating cellular proliferation through canonical metabolic pathways or modulating the transcription of pro-proliferative genes via non-canonical pathways [13,14]. Despite documented metabolic dysregulation in various cancer cells, effective clinically available drugs targeting these pathways remain lacking [15]. Currently, chemotherapy remains the primary treatment for unresectable pancreatic cancer [16]. Consequently, continued exploration and screening of metabolic therapeutic targets are imperative to support clinical interventions aimed at prolonging the survival of PDAC patients.

Guanine monophosphate synthase (GMPS), a key rate-limiting enzyme in the de novo purine biosynthesis pathway, mainly participates in the biosynthesis of guanine nucleotides. It also exerts vital functions in sustaining intracellular nucleic acid metabolic homeostasis and regulating proliferation signaling [17]. Existing studies have confirmed that GMPS is aberrantly overexpressed in multiple malignancies, such as melanoma [18], lung cancer [19], prostate cancer [20] and cervical cancer [21]. Its upregulation is not only closely associated with rapid tumor cell proliferation but also serves as an important biomarker indicating aggressive disease progression and unfavorable clinical prognosis.

Apart from its canonical enzymatic activity, recent studies have revealed that GMPS participates in multiple biological processes, such as transcriptional regulation and signaling pathway activation, through non-enzymatic mechanisms, thereby exerting multifaceted roles in tumorigenesis and progression [22,23]. GMPS binds to the deubiquitinating enzyme USP7 and allosterically enhances its activity, thereby participating in histone deubiquitination and the regulation of p53 stability [24]. Under normal conditions, GMPS is ubiquitinated by the E3 ligase TRIM21 and retained in the cytoplasm [25]; upon DNA damage or nucleotide deprivation, GMPS translocates into the nucleus, where it transfers p53 from MDM2 to the USP7 complex, removing ubiquitin chains to stabilize p53 and suppress tumorigenesis [26]. In breast cancer, the small nucleolar RNA SNORD50A/B regulates this process by bridging TRIM21 and GMPS [22]. Furthermore, in nasopharyngeal carcinoma, SERPINB5 acts as an adaptor protein that prevents GMPS from entering the nucleus and directs TRIM21 to ubiquitinate and degrade GMPS, thereby repressing p53 and promoting radioresistance [27]. Its non‑canonical functions, as an allosteric activator of USP7, a coordinator of p53 stabilization, and a cytoplasm‑to‑nucleus shuttling protein, provide novel intervention targets for cancer therapy. Despite these advances, the potential involvement of GMPS in safeguarding genomic integrity, particularly through the regulation of replication stress and DNA damage repair, remains largely unexplored. Moreover, GMPS expression profile, clinical significance, and underlying biological mechanisms in PDAC remain poorly understood.

A CRISPR interference‑based high‑throughput screening platform allows for systematic, accurate evaluation of gene effects on tumor cell proliferation. We applied this approach to 2752 metabolic genes in 44 pancreatic cancer cell lines and found that GMPS is the metabolic gene most critical for proliferation. Additionally, GMPS expression correlated with the poorest prognosis in PDAC. Despite the significant upregulation of GMPS in PDAC and its potential pivotal role in disease progression, no study has systematically elucidated the function of GMPS in pancreatic cancer. Here, we employed computational prediction, cellular experiments and animal model assays to demonstrate that GMPS serves as an oncogene promoting PDAC progression. Furthermore, we focused on elucidating the molecular mechanisms by which GMPS knockdown suppresses PDAC proliferation.

Materials and methods

Bioinformatics analysis

CERES scores of 2752 metabolism‑related genes across 44 PDAC cell lines were obtained from the DepMap database to screen for metabolic genes associated with proliferation and prognosis. Transcriptomic, mutational, and clinical follow‑up information of PDAC patients were sourced from the TCGA database. All enrolled patients were categorized into high and low GMPS level subgroups using the median expression value as the cutoff. Differentially expressed genes (DEGs) were screened out via the limma package with thresholds of |log2 fold change| > 1 and false discovery rate < 0.05. The clusterProfiler package was used to perform KEGG pathway enrichment analysis. Oncoplots were generated using the maftools package to compare mutation frequencies of core driver oncogenes, including KRAS, TP53, SMAD4 and CDKN2A. Immune cell abundance and immune scores were calculated using the xCell algorithm. Drug IC50 values were sourced from the GDSC database, and differences between the two GMPS level subgroups were further evaluated with the oncoPredict package (Table S1).

Clinical tissue specimens

Surgical samples were collected from patients who received radical pancreatectomy at Fudan University Shanghai Cancer Center between 2018 and 2022, with pathologically confirmed pancreatic ductal adenocarcinoma. Follow‑up information was also collected (Table S2 and S3). These specimens were processed for IHC staining. IHC score = intensity score (scored 0–3) × proportion of positive cells (scored 0–4). Based on GMPS IHC scores, patients were classified into high and low groups. Kaplan–Meier analysis with log‑rank test was used for OS.

Cell culture

The following cell lines were acquired from the American Type Culture Collection (ATCC): HPDE6-C7 (normal pancreatic ductal epithelium), HEK293T (human embryonic kidney), and the pancreatic cancer cell lines PANC-1, AsPC-1, BxPC-3, SW1990, MIA PaCa-2, and Capan-1. All cell lines were cultured as recommended by the manufacturer in RPMI-1640, IMDM or DMEM medium containing 10% fetal bovine serum and 1% antibiotic mixture at 37 °C under a humidified atmosphere with 5% CO₂.

Organoid culture

Freshly resected pancreatic cancer specimens were dissected into 1–3 mm³ pieces and incubated in 10 mL tissue digestion buffer in a 15 mL centrifuge tube at 37 °C on an orbital shaker for 40–60 min. Digestion was terminated by adding FBS. The mixture was strained (100 μm), treated with red blood cell lysis solution, and rinsed. The resulting cell pellet was gently resuspended in Matrigel (Corning, NY, USA), plated in 24-well plates, and solidified at 37 °C for 20 min. After solidification, 500 μL of pre‑warmed culture medium was supplemented, and the plates were returned to the 37 °C, 5% CO₂ incubator. Organoid viability was detected using Organoid viability ATP assay kit (bioGenous, Suzhou, Jiangsu, China).

Virus packaging and transfection

Lentivirus was produced by transfecting HEK293T cells with packaging plasmids and shGMPS or GMPS overexpression plasmid using Lipofectamine 3000 (Thermo Fisher Scientific, Waltham, MA, USA). Supernatant was collected after 72 h and concentrated. PDAC cells were transduced with the virus plus Polybrene (Sigma‑Aldrich, St. Louis, MO, USA), followed by puromycin selection for 7 days. Knockdown efficacy was confirmed by western blot. For organoid transduction, organoids were dissociated into single cells, suspended in the viral stock, and mixed with Matrigel. Stable knockdown and overexpression were verified.

Colony formation assay

PDAC cells (1000 per well) were plated into 6‑well plates and cultured in complete medium for 7–10 days. After visible colonies had formed, the cells were rinsed twice with PBS, immobilized and stained by crystal violet for 20 min. After rinsing and natural air-drying, colonies containing ≥50 cells were quantified. All assays were carried out in triple replicates and independently repeated three times.

CCK‑8 proliferation assay

Cell proliferative activity was assessed via Cell Counting Kit‑8 (CCK‑8, Selleck, Houston, TX, USA). Briefly, MIA PaCa‑2 and Capan‑1 cells were inoculated into 96‑well plates at 2000 cells per well in complete medium. At predetermined time points (0, 24, 48, 72, and 96 h), 10 µL of CCK‑8 solution was supplemented and incubated at 37 °C for 2 h. Subsequently, the optical density value at 450 nm was detected by a microplate spectrophotometer. All experiments were performed in five replicate wells per condition and repeated three times independently.

Cell cycle analysis

PDAC cells were harvested, rinsed twice with prechilled PBS, and preserved in 70% ethanol for 4 h at 4 °C. After fixation, the cells were rinsed and stained with propidium iodide/RNase staining buffer (Beyotime, Shanghai, China) for 30 min in the dark conditions. Cell cycle distribution was analyzed by Cytoflex S (Beckman, Brea, CA, USA) and ModFit LT software. All assays were conducted three separate times.

SA‑β‑gal staining

A SA‑β‑gal staining kit (Beyotime, Shanghai, China) was employed to detect senescence. PDAC cells were rinsed, fixed, stained with SA‑β‑gal staining mix at 37 °C under CO₂-free conditions overnight, and visualized under an optical microscope. ImageJ software was utilized to quantify the staining intensity of SA-β-gal. Each experiment was performed three times independently.

Quantitative real‑time polymerase chain reaction (qRT‑PCR)

Total RNA was isolated from cells with RNA extraction kits (ACCURATE Biotechnology, Changsha, Hunan, China). Reverse transcription was conducted using a HiScript III RT SuperMix for qPCR (Vazyme, Nanjing, Jiangsu, China) with 1 µg of total RNA. qRT-PCR was implemented with SYBR Green Master Mix (Vazyme, Nanjing, Jiangsu, China) on a QuantStudio 6 System (Thermo Fisher Scientific, Waltham, MA, USA). Primers used for GMPS, CDKN1A, IL6, MMP2, and the internal control ACTB are listed in Table S4. Gene relative expression was quantified by the 2^⁻ΔΔCt approach. All reactions were run in triplicate.

Immunofluorescence staining for γH2AX

To assess DNA damage, cells were subjected to immunofluorescence staining for γH2AX. PDAC cells grown on glass coverslips were fixed with fixation solution for 15 min at ambient temperature, permeabilized with immunostaining permeabilization solution for 10 min, and blocked with 3% bovine serum albumin for 45 min. The cells were then incubated with anti‑γH2AX antibody (Table S5) at 4 °C overnight. After PBS rinsing, the coverslips were treated with AF594‑conjugated secondary antibody (1:200, Invitrogen, Carlsbad, CA, USA) for 1 h in the dark. Cell nuclei were counterstained with 1 μg/mL DAPI for 5 min. Fluorescence images were captured via an Olympus fluorescence microscope (Tokyo, Japan). The percentage of γH2AX‑positive cells was determined by analyzing five randomly selected visual fields. The experiment was repeated three times.

Western blotting

Cells were lysed in RIPA lysis buffer (Beyotime, Shanghai, China) containing protease and phosphatase inhibitor cocktail (Beyotime, Shanghai, China) on ice for 30 min. The cell lysates were centrifuged and the supernatants were harvested. Proteins were resolved by 10% SDS‑polyacrylamide gel electrophoresis and then electrotransferred onto nitrocellulose membranes (Millipore, USA). The membranes were blocked using 5% skim milk and incubated with primary antibodies overnight at 4 °C. After washing, the membranes were incubated with secondary antibodies for 1 h. Protein bands were detected with enhanced chemiluminescence reagents (Epizyme, Shanghai, China) and photographed using a chemiluminescence imaging device (Tanon, Shanghai, China). Detailed information of all antibodies adopted in this research is listed in Table S5.

RNA sequencing

MIA PaCa‑2 cells stably expressing shGMPS or NC were harvested when reaching 80–90% confluence. Total RNA was isolated with TRIzol reagent (Invitrogen, Carlsbad, CA, USA). RNA integrity was verified on an Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA). cDNA libraries were constructed via the TruSeq Stranded mRNA Library Prep Kit (Illumina, San Diego, CA, USA). Subsequent sequencing was performed on the Illumina NovaSeq 6000 platform adopting a 150 bp paired-end sequencing mode. Raw sequencing reads were assessed for quality via FastQC and further processed for sequence trimming using Trimmomatic software. Relative gene expression abundances were calculated with featureCounts, and DEGs were screened out through DESeq2 software under the criteria of ∣log2​fold change∣>1 and FDR <0.05. Additionally, KEGG pathway enrichment analysis was conducted utilizing the clusterProfiler package within R language. The resulting quantitative data are presented in Tables S6.

Proteomics analysis

MIA PaCa‑2 cells with GMPS knockdown or control were lysed, and the extracted proteins were subjected to tryptic digestion. The resultant peptides were further analyzed using an Orbitrap Fusion Lumos mass spectrometer (Thermo Fisher, Waltham, MA, USA). Original mass spectrometry data were matched against the human UniProt database using MaxQuant. Label‑free quantification was used for protein quantification. Differentially expressed proteins were screened with the criteria of |log₂fold change| > 1 and adjusted p‑value < 0.05 (t‑test). GO functional enrichment analyses were carried out using clusterProfiler. The resulting quantitative data are presented in Tables S7.

STING inhibition and rescue experiments

To investigate the role of STING signaling in GMPS-knockdown-induced cellular senescence, Capan-1 cells stably expressing shGMPS or a negative-control shRNA were treated with H-151, a selective STING inhibitor (MedChemExpress, NJ, USA, Catalog No HY-112,693), at a concentration of 0.5 μM for 48 h. H-151 was dissolved in dimethyl sulfoxide (DMSO), and an equivalent concentration of DMSO was used as the vehicle control. Four groups were included: NC, H-151, sh1, and sh1+H-151. After treatment, cells were harvested to evaluate STING pathway activation and senescence-associated phenotypes.

In vivo PDX model experiments

Male NOD-Prkdc^scidIl2rg^null (NPSG) mice aged 5 to 6 weeks with body weights ranging from 18 to 20 g were purchased from Shanghai Jihui Laboratory Animal Co., Ltd. (China) and raised under specific pathogen-free environments. All experimental animals were maintained under specific pathogen‑free conditions in accordance with relevant ethical standards.

Fresh human PDAC tumor tissues were collected from patients undergoing curative pancreatectomy at Fudan University Shanghai Cancer Center. The tissues were immediately placed in ice‑cold RPMI‑1640 medium. Under sterile conditions, the tumor tissue was minced into small fragments. These fragments were implanted into the right dorsal of NPSG mice. Once the tumor grew to roughly 1000 mm³, they were collected, cut into tiny pieces, and retransplanted into recipient mice for propagation. Third‑generation (F3) PDX tumors were used for the intervention experiment. The validity of the PDX models was confirmed by hematoxylin and eosin (H&E) staining, immunohistochemistry for the pancreatic cancer marker CK19, and detection of human DNA content.

When the F3 PDX tumors reached a volume of approximately 200 mm³, mice were divided into two groups with six animals in each group, namely the negative control (NC) group and siGMPS intervention group. In vivo siRNA targeting GMPS (RiboBio, Guangzhou, Guangdong, China) or non‑targeting negative control siRNA was delivered via intratumoral injection at a dosage of 10 nmol per injection, once every three days (10 injections). Tumor dimensions (length and width) were measured using a vernier caliper, and tumor volume was calculated using the formula: 0.5 × length × width². At the end of the experiment (day 28), mice were sacrificed via CO₂ inhalation. Isolated tumor tissues were used for measurement, photography and subsequent IHC.

Immunohistochemistry

Tissue sections were dewaxed in xylene and rehydrated in gradient ethanol solutions and finally rinsed with distilled water. Antigen retrieval was achieved by heating tissue slices in citrate buffer within a pressure cooker for 3 min. Subsequently, endogenous peroxidase activity was inhibited via incubation of the slices in 3% hydrogen peroxide solution for 15 min at ambient temperature. Then, tissue slices were blocked with 5% normal goat serum in PBS for 1 h at ambient temperature to reduce non‑specific binding. The sections were then incubated overnight at 4 °C with primary antibodies. After rinsing with PBS, the slices were incubated with secondary antibody for 1 h at ambient temperature. Antibody binding was stained with 3,3′‑diaminobenzidine (DAB) substrate for 2–5 min. Afterwards, hematoxylin was applied for nuclear counterstaining. The prepared slices were further dehydrated, cleared with xylene and sealed using neutral mounting medium. For PDX xenograft tissues, the DAB H‑score was determined using the APTime 3.0 software, which automatically calculates the score based on the standard formula: H‑score = [1 × (% weakly stained cells)] + [2 × (% moderately stained cells)] + [3 × (% strongly stained cells)], with a theoretical range of 0–300.

Statistical analysis

Two-group comparisons used two-tailed Student’s t-test. Multiple-group comparisons were analyzed using one- or two-way ANOVA, with Tukey’s post-hoc test. Survival curve analysis was implemented by the Kaplan-Meier approach combined with the log-rank test. P < 0.05 was considered significant (* p < 0.05, ** p < 0.01, *** p < 0.001). Data analyses were completed with GraphPad Prism v8.0.2 or R software v4.4.1.

Results

GMPS associates with proliferation and poor prognosis in PDAC

To systematically identify metabolic genes critical for PDAC proliferation, we first leveraged data from the DepMap database. We ranked 2752 metabolism-related genes based on their CERES scores across 44 PDAC cell lines, where a lower score indicates greater gene essentiality for cell growth. This initial screening yielded 41 genes with CERES scores less than −1 (Fig. 1A, B). To explore the clinical significance of these candidates, we analyzed their associations with patient survival. GMPS exhibited the highest hazard ratio (HR) among the 41 genes, suggesting its strong association with increased mortality risk in PDAC patients (Fig. 1C). Subsequent Kaplan-Meier analysis confirmed that high GMPS expression was correlated with inferior disease-free survival and overall survival (OS), establishing it as a robust prognostic factor for adverse outcomes (Fig. 1D). Furthermore, transcriptomic data analysis showed that GMPS mRNA expression was elevated in PDAC tissues relative to normal pancreatic counterparts (Fig. 1E). Collectively, these integrated bioinformatics analyses pinpointed GMPS as a key metabolic gene linked to PDAC progression and poor prognosis, prompting its selection for further functional investigation.

Fig. 1.

Screening of proliferation- and prognosis-related metabolic genes in PDAC. (A) Schematic diagram of screening GMPS from 2752 metabolism-related genes in 44 PDAC cell lines using Depmap database, sorted by CERES scores, selected 41 metabolic genes associated with proliferation phenotype, then further filtered by correlation with proliferation phenotype and prognosis. (B) Source information of 44 PDAC cell lines and distribution of CERES scores of metabolism-related genes in PDAC cell lines. (C) Heatmap of metabolism-related genes with CERES score < −1, ranked by hazard ratio (HR). (D) Correlation of GMPS expression with disease-free survival (DFS) and overall survival (OS) in PDAC patients based on The Cancer Genome Atlas (TCGA) database. (E) GMPS expression levels were compared between PDAC tumor tissues and paired normal tissues in TCGA database. **** p < 0.0001.

GMPS expression defines distinct genomic, immune, and drug response traits in PDAC

To investigate the molecular correlates of GMPS expression in PDAC, we performed subgroup analysis on The Cancer Genome Atlas (TCGA) cohort stratified by GMPS expression levels. Differential expression analysis uncovered significant transcriptomic alterations between GMPS‑high and GMPS‑low groups (Fig. S1A). KEGG pathway enrichment analysis using a ridge plot showed that pathways including pancreatic cancer, pathways in cancer, cell cycle, DNA replication, and p53 signaling exhibited distinct enrichment patterns between the two groups (Fig. S1B). Oncoplot analysis of well-established driver genes (KRAS, TP53, SMAD4, CDKN2A) [16] demonstrated that GMPS‑high tumors exhibited significantly higher mutation frequencies of KRAS and TP53 compared with GMPS‑low tumors (Fig. S1C, D). Immune cell abundance estimated by xCell revealed that the GMPS-low group had higher infiltration of immune cells, including CD8⁺T cells, macrophages (both M1 and M2), NKT cells, and an increased overall immune score (Fig. S1E). Drug sensitivity analysis indicated that GMPS‑low tumors had lower predicted IC50 values for several agents, including Axitinib, GSK269962A, PRIMA.1MET, Niraparib, Sorafenib, and Irinotecan, suggesting increased sensitivity to these drugs. Moreover, GMPS-low tumors also exhibited lower predicted IC50 values for first-line chemotherapeutic agents for pancreatic cancer, including gemcitabine, paclitaxel, 5-fluorouracil, and oxaliplatin (Fig. S1F). Collectively, these data indicate that low GMPS expression defines a distinct PDAC subtype characterized by enhanced immune infiltration and greater susceptibility to multiple targeted agents and chemotherapies.

Elevated GMPS expression in PDAC tissues predicts poor survival

To verify the protein-level expression pattern of GMPS, we conducted immunohistochemistry (IHC) on cancerous and adjacent non‑cancerous specimens from 97 PDAC patients having follow‑up information. GMPS protein levels were markedly increased in PDAC cancerous tissues relative to adjacent non‑cancerous tissues (Fig. 2A, B). Western blot (WB) detection of five matched cancerous and adjacent non‑cancerous tissues further confirmed GMPS upregulation in tumors (Fig. 2C, D). Representative IHC images of GMPS‑high and GMPS‑low tumors are displayed in Fig. 2E. Kaplan‑Meier survival analysis showed that increased GMPS level was strongly linked to shortened OS (Fig. 2F). In vitro, GMPS mRNA and protein levels were examined across multiple pancreatic cancer cell lines as well as the pancreatic ductal epithelial cell HPDE6-C7. Among the tested cell lines, MIA PaCa-2 and Capan-1 exhibited the highest GMPS protein expression levels (Fig. 2G, H). Based on these results, MIA PaCa‑2 and Capan‑1 were selected for subsequent loss‑of‑function experiments. Collectively, these data confirm that GMPS is elevated in PDAC, and its high abundance correlates with poor survival outcomes.

Fig. 2.

GMPS is upregulated in PDAC tissues and cell lines and correlates with poor patient survival. (A) Typical immunohistochemistry (IHC) staining views of GMPS in PDAC matched cancerous and non-cancerous samples (scale bar = 400 μm). (B) Statistical graph of GMPS IHC expression scores in matched tumor (T) versus normal specimens (N). (C) Western blot (WB) results of GMPS protein in matched tumor (T) versus normal specimens (N). (D) Statistical graph of GMPS protein expression level in matched tumor (T) versus normal specimens (N). (E) Typical IHC staining patterns showing differential GMPS staining intensity in tumor specimens (scale bar = 400 μm). (F) OS of PDAC patients stratified by GMPS expression levels based on our clinical cohort. (G) GMPS mRNA expression levels in HPDE6-C7 and various pancreatic cancer cell lines. (H) GMPS protein expression levels and corresponding densitometric values in the same cell lines. *** p < 0.001.

GMPS promotes the proliferative capacity of PDAC cell lines and organoids

To examine the function of GMPS in PDAC proliferation, we first knocked down GMPS in MIA PaCa‑2 and Capan‑1 cells. Colony formation assays indicated that suppression of GMPS markedly diminished the number and size of colonies in both cell lines (Fig. 3A, B). Consistently, CCK‑8 proliferation curves showed that GMPS knockdown impaired cell proliferation over time (Fig. 3C). We next extended these findings to patient‑derived organoid models. WB confirmed efficient GMPS knockdown and overexpression in organoids (Fig. 3D). Organoid growth was monitored on days 1, 4, and 7. Knockdown of GMPS markedly suppressed organoid expansion (Fig. 3E), whereas GMPS overexpression promoted organoid growth (Fig. 3F). Quantification of viable cells by ATP‑based luminescence on day 10 further demonstrated that GMPS knockdown reduced organoid viability, while overexpression increased it (Fig. 3G). Collectively, these loss‑ and gain‑of‑function experiments establish GMPS as a positive regulator of PDAC cell and organoid proliferation.

Fig. 3.

GMPS promotes PDAC cell and organoid proliferation. (A) Representative images of colony formation assays after GMPS knockdown in PDAC cells (one image per 6‑well plate well). (B) Statistical quantification of colony formation assays. (C) Proliferation curves from CCK-8 assays after GMPS knockdown in PDAC cells. (D) WB validation of GMPS protein levels upon knockdown and overexpression in organoids. (E) Example images of organoid growth after GMPS knockdown (scale bar = 500 μm). (F) Example images of organoid growth after GMPS overexpression (scale bar = 500 μm). (G) Quantification of viable cells by ATP-based luciferase reaction (Day 10). ** p < 0.01, *** p < 0.001.

GMPS silencing regulates cell cycle and induces cellular senescence

To explore the molecular mechanisms that underlie GMPS‑mediated proliferation, we performed RNA sequencing on MIA PaCa‑2 cells following GMPS knockdown. KEGG enrichment analysis showed that cell cycle, p53 signaling pathway, and DNA damage repair‑related pathways were significantly enriched (Fig. 4A). Flow cytometry analysis showed that GMPS knockdown led to an altered cell cycle distribution, characterized by an increased percentage of cells in the G1 phase and concomitant decreases in both the S and G2 phases in MIA PaCa‑2 and Capan‑1 cells (Fig. 4B, C). Given that DNA damage and cell cycle arrest are characteristic features of cellular senescence [28], we next examined senescence‑related phenotypes. qRT‑PCR confirmed that GMPS knockdown upregulated mRNA levels of senescence‑associated genes, including CDKN1A, IL6 and MMP2 (Fig. 4D). WB confirmed that GMPS knockdown increased protein levels of p21 (Fig. 4E). Furthermore, Senescence‑associated β‑galactosidase (SA‑β‑gal) staining, a characteristic feature of cellular senescence, was markedly increased upon GMPS knockdown in both cell lines (Fig. 4F). Together, these data suggest that GMPS knockdown induces G1 phase cell cycle arrest and triggers cellular senescence, thereby attenuating PDAC cell proliferation, likely through DNA damage repair pathways.

Fig. 4.

GMPS regulates cell cycle and induces senescence in PDAC cells. (A) KEGG pathway enrichment bubble plot from RNA sequencing of MIA PaCa-2 cells with GMPS knockdown versus control. (B) Cell cycle distribution (Modfit) of PDAC cells with GMPS knockdown versus control. (C) Statistical analysis of G1, G2, and S phase percentages in PDAC cells. (D) mRNA levels of GMPS and senescence-related genes (CDKN1A, IL6, MMP2) in MIA PaCa-2 cells with GMPS knockdown versus control. (E) WB analysis of GMPS and p21 protein levels in PDAC cells with GMPS knockdown versus control. (F) SA-β-gal staining and statistical quantification in PDAC cells with GMPS knockdown versus control (scale bar = 100 μm). * p < 0.05, ** p < 0.01, *** p < 0.001.

GMPS knockdown induces DNA damage and triggers the STING‑TBK1 pathway

Having observed that GMPS knockdown leads to cell cycle arrest and senescence, we next examined whether DNA damage was involved. Immunofluorescence staining revealed that GMPS knockdown significantly increased the percentage of γH2AX‑positive cells, indicating DNA damage in PDAC cells (Fig. 5A, B). To gain further mechanistic insight, we performed proteomic analysis followed by Gene Ontology (GO) enrichment. The results showed that pathways related to DNA damage repair, positive regulation of canonical NF‑κB signaling and type I interferon production, and cellular response to double‑stranded RNA were significantly enriched (Fig. 5C). Given that the STING signaling pathway is a well‑established bridge connecting DNA damage with the activation of NF‑κB and type I interferon responses, we investigated whether STING was activated upon GMPS knockdown. WB results showed that GMPS knockdown increased the phosphorylation levels of STING and TBK1, indicating the STING‑TBK1 signaling pathway activation (Fig. 5D).

Fig. 5.

GMPS knockdown induces DNA damage and activates STING signaling. (A) Fluorescence images of γH2AX-positive cells in PDAC cells with GMPS knockdown versus control (scale bar = 100 μm). (B) Statistical analysis of γH2AX-positive cells in (A). (C) GO enrichment analysis of proteomics data from MIA PaCa-2 cells with GMPS knockdown versus control. (D) WB showing activation of the STING-TBK1 signaling pathway in PDAC cells with GMPS knockdown versus control. (E) Western blot analysis of GMPS, p-STING, STING, p-TBK1, TBK1, and p21 in control and GMPS-knockdown Capan-1 cells treated with DMSO or H-151. (F) Representative images and quantification of SA-β-gal staining in the indicated groups (scale bar = 100 μm). (G) qRT-PCR analysis of GMPS, CDKN1A, IL6, and MMP2 mRNA expression in the indicated groups. ** p < 0.01, *** p < 0.001.

To further determine whether STING signaling functionally contributes to the senescence phenotype induced by GMPS depletion, we performed rescue experiments using H-151, a selective STING inhibitor. GMPS-knockdown Capan-1 cells were treated with H-151 (0.5 μM for 48 h). H-151 treatment reduced the phosphorylation levels of STING and TBK1 and partially reversed the senescence phenotype induced by GMPS knockdown, as evidenced by a decrease in SA-β-gal-positive cells and p21 protein expression (Fig. 5E, F). Consistently, H-151 significantly reduced the mRNA expression levels of the senescence-associated genes CDKN1A, IL6, and MMP2, while GMPS expression remained suppressed (Fig. 5G). Thus, our findings indicate that GMPS knockdown triggers DNA damage, which not only suppresses proliferation via cell cycle arrest and senescence but also activates the STING pathway to reinforce the senescence phenotype.

GMPS knockdown inhibits tumor growth in a patient‑derived xenograft (PDX) model

To assess the in vivo role of GMPS, we generated a PDX model. After three rounds of passaging to ensure stable tumor growth, the third‑generation PDX tumors were used for intervention experiments. In vivo siRNA targeting GMPS (siGMPS) or negative control (NC) was delivered via intratumoral injection. Each injection contained 10 nmol reagents, given once every three days for four consecutive weeks (Fig. 6A). All tumor tissues were collected for subsequent detection after treatment termination. Representative tumor images showed that siGMPS treatment significantly reduced tumor size compared with the NC group (Fig. 6B). Tumor volume measurements further confirmed that GMPS knockdown suppressed tumor growth over time (Fig. 6C). IHC staining of PDX tumor tissues revealed that siGMPS treatment decreased GMPS abundance and the proliferation biomarkers PCNA and Ki67, as quantified by 3,3′-diaminobenzidine (DAB) scores (Fig. 6D-G). Collectively, these in vivo data demonstrate that GMPS knockdown effectively inhibits PDAC tumor progression, supporting its promise as a therapeutic target.

Fig. 6.

GMPS knockdown suppresses tumor growth in a PDX model. (A) Schematic representation of PDX passaging and study strategy. (B) Example tumor images from negative control (NC) and siGMPS groups. (C) Statistical graph of tumor volumes in NC and siGMPS groups. (D) IHC staining of GMPS, PCNA, Ki67 in tumor tissues from NC and siGMPS groups, and corresponding DAB scoring statistics (scale bar = 100 μm). (E–G) Quantification of GMPS, PCNA, and Ki67 staining based on DAB H-scores.** p < 0.01, *** p < 0.001.

Discussion

In this research, we investigated the functional significance of GMPS in the progression of PDAC. Through integrative bioinformatics, cellular and animal functional assays, and multi-omics analyses, we identified GMPS as a critical metabolic oncogene that promotes PDAC proliferation and is associated with unfavorable prognosis. Our results highlight GMPS as a viable therapeutic candidate and elucidate the molecular mechanisms by which GMPS knockdown inhibits PDAC proliferation, including DNA damage, cellular senescence, and STING pathway activation.

Using a CRISPR interference-based high-throughput screening framework applied to 44 PDAC cell lines from the DepMap database, we ranked 2752 metabolic genes and identified GMPS as one of the most critical genes for PDAC proliferation. Survival analysis further revealed that high GMPS expression correlated with the poorest OS and DFS among the 41 candidate genes, underscoring its clinical relevance. These observations are in line with prior studies confirming elevated GMPS levels across various malignancies [[18], [19], [20], [21], [22]], although its functional role in these contexts remains less characterized. This suggests that GMPS may represent a broadly conserved metabolic dependency across multiple tumor types.

Transcriptomic and genomic characterization of the TCGA PDAC cohort stratified by GMPS expression revealed that GMPS-high tumors exhibited higher mutation frequencies of KRAS [29] and TP53 [30], two of the most common driver mutations in PDAC. This suggests that GMPS upregulation may cooperate with these canonical oncogenic pathways to promote tumor aggressiveness. Moreover, immune infiltration analysis implied that GMPS-low tumors may represent a more immunologically “hot” subtype, potentially more responsive to immunotherapies. Consistent with our finding, a previous study demonstrated that GMPS facilitates immune evasion in hepatocellular carcinoma and impairing the functional activity of CD8⁺ T cells [31]. Prediction of drug sensitivity further showed that GMPS-low PDACs are more susceptible to multiple agents, suggesting that GMPS expression levels could guide personalized treatment strategies.

GMPS exhibited elevated levels in PDAC tissues and cell lines, and high GMPS levels correlated with shorter overall survival. Functional studies in cell lines, organoids and PDXs further demonstrated that GMPS promotes PDAC proliferation, confirming it as an oncogenic driver rather than merely a prognostic indicator. Mechanistically, using RNA‑seq, proteomics, and in vitro cellular assays, we demonstrated that GMPS knockdown induces DNA damage, cell cycle arrest, STING pathway activation, and cellular senescence.

However, the molecular mechanisms linking GMPS deficiency to DNA damage accumulation remain to be fully elucidated. Given that GMPS functions as a rate-limiting enzyme in the de novo purine biosynthesis pathway, catalyzing the conversion of xanthosine monophosphate (XMP) to guanosine monophosphate (GMP), we speculate that GMPS knockdown may disrupt the intracellular nucleotide pool homeostasis, particularly the guanine nucleotide pools [21]. This imbalance increases replication stress, which may cause fork stalling at guanine‑rich regions and lead to DNA double‑strand breaks [32], as evidenced by the elevated γH2AX observed in our GMPS-knockdown models. Importantly, our findings align with a growing body of evidence highlighting the vulnerability of pancreatic cancer to nucleotide metabolism perturbations. For instance, targeting RRM2, which regulates the dNTP pool by controlling deoxypyrimidine synthesis, has been shown to induce replicative stress and sensitize pancreatic cancer cells to chemotherapy [33]. Similarly, recent studies have identified the de novo guanine nucleotide biosynthesis pathway as a critical dependency in KRAS-mutant PDAC, with IMPDH2 inhibition inducing irreversible guanine nucleotide depletion and augmenting the efficacy of KRAS inhibitors [34]. In parallel, ECM-driven upregulation of IMPDH has been demonstrated to enhance guanylate production, thereby alleviating DNA damage in PDAC [35]. Collectively, these studies suggest that disrupting nucleotide pool homeostasis precipitates replicative stress and genomic instability in pancreatic cancer. While our current data support this mechanistic model, we acknowledge that direct evidence linking GMPS knockdown to replication fork stalling, dNTP pool alterations, or DNA damage pathway activation is currently lacking. Future studies employing DNA fiber assays to directly assess replication fork dynamics, quantitative metabolomics to profile nucleotide pool changes, and exogenous nucleoside supplementation rescue experiments will be essential to definitively establish this mechanism.

Once DNA damage occurs, two downstream signaling axes are triggered. First, the classic p53‑dependent pathway: DNA damage activates p53 signaling, leading to cell cycle arrest and promoting cellular senescence. Second, the innate immune pathway: cytoplasmic DNA generated from DNA damage is sensed by the STING pathway [36,37], and STING activation further reinforces the senescent state and induces antitumor immunity [38,39]. Of note, cellular senescence itself is a recognized cancer hallmark with dual functions in tumor biology [40]. On one hand, senescence exerts tumor-suppressive effects by restraining neoplastic cell growth; on the other hand, senescent cells facilitate tumor development and immune modulation via the secretion of a complex senescence-associated secretory phenotype (SASP) [41,42]. During senescence, STING signaling upregulates cytokines and chemokines, contributing to SASP formation and thereby remodeling the tumor immune microenvironment [43]. Interestingly, in our TCGA analysis, the GMPS-low expression group showed a higher overall immune score, which may be related to STING activation leading to a more immunologically active tumor microenvironment [36]. Thus, our findings suggest that GMPS knockdown triggers DNA damage, which not only directly impairs proliferation but also activates the STING‑TBK1 signaling axis, thereby promoting a SASP and potentially enhancing immune surveillance. Therefore, we position GMPS as a critical regulator of cellular senescence in PDAC.

Finally, intratumoral delivery of GMPS-targeting siRNA in a PDX model suppressed tumor growth, reduced PCNA/Ki67 expression, and lowered GMPS protein levels, validating GMPS as a viable therapeutic target in PDAC. Looking forward, based on our key finding that GMPS knockdown induces cellular senescence, inducing tumor cell senescence with GMPS inhibitors followed by sequential or concurrent senolytics treatment could theoretically eliminate residual senescent tumor cells [42,44], and enhance overall anti‑proliferative efficacy. On the other hand, the STING‑TBK1 pathway activated by GMPS depletion not only promotes senescence but also enhances type I interferon responses, providing a rationale for combination with immunotherapy [45].

In conclusion, this study identifies GMPS as a metabolic oncogene in PDAC that drives proliferation through cell cycle progression and suppression of senescence. GMPS knockdown induces DNA damage, activates the STING‑TBK1 pathway, and triggers cellular senescence, thereby inhibiting tumor growth. Our findings establish GMPS as a prognostic indicator and a promising treatment candidate for PDAC. Future investigations should concentrate on developing specific GMPS inhibitors and exploring combination therapeutic strategies that leverage GMPS‑dependent vulnerabilities.

Declarations

Ethics approval

This study was conducted in accordance with the Declaration of Helsinki, and approved by the Institutional Review Board of Fudan University Shanghai Cancer Center (Approval No 2504-Exp169). The animal study protocol was approved by the Institutional Animal Care and Use Committee of Fudan University (Approval No FUSCC-IACUC-2024,138) and conformed to the relevant guidelines and regulations.

Consent for publication

All authors have read and approved of its submission to this journal.

Funding

This research was financially supported by the Noncommunicable Chronic Diseases-National Science and Technology Major Project(2025ZD0552305) and the National Natural Science Foundation of China (82541017 and 82473106).

Data availability statement

Publicly available datasets analyzed in this study include DepMap (https://depmap.org/), TCGA (https://portal.gdc.cancer.gov/), and GDSC (https://www.cancerrxgene.org/). All other experimental data and methodological details are provided in the main text or Supplementary Materials.

CRediT authorship contribution statement

Yangyi Li: Writing – original draft, Visualization, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Zifeng Zhang: Writing – original draft, Visualization, Validation, Methodology, Formal analysis. Dingru Li: Visualization, Validation, Methodology, Formal analysis. Cong Zhou: Visualization, Validation, Formal analysis. Yueyue Chen: Software, Methodology, Formal analysis. Chunbin Zhu: Visualization, Validation, Formal analysis. Jianhui Yang: Investigation, Data curation. Rong Tang: Software, Formal analysis. Wei Wang: Visualization, Formal analysis. Xianjun Yu: Writing – review & editing, Supervision, Resources, Conceptualization. Jin Xu: Writing – review & editing, Supervision, Resources, Funding acquisition, Conceptualization.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.