SGLT2 inhibition induces autophagic flux blockade and sensitizes pancreatic cancer to EGFR-targeted therapy

Abstract

Background

Pancreatic ductal adenocarcinoma (PDAC) is a lethal malignancy with profound metabolic rewiring and resistance to therapy. Sodium-glucose cotransporter 2 (SGLT2) regulates glucose uptake, but its role in PDAC remains unclear.

Methods

SGLT2 expression was analyzed in clinical samples and public datasets. PDAC cell lines were subjected to genetic knockdown or canagliflozin (CANA) treatment to assess proliferation, migration, apoptosis, and glucose metabolism. Mechanistic studies investigated AMPK-ULK1 signaling, autophagy dynamics, oxidative stress, and EGFR signaling. Xenograft models were used to assess in vivo efficacy.

Results

SGLT2 was upregulated in PDAC and associated with poor prognosis. SGLT2 inhibition suppressed proliferation and migration while promoting apoptosis. Mechanistically, CANA induced ATP deficiency and initiated autophagy, but concurrently impaired autophagosome-lysosome fusion. This dual effect led to autophagic flux blockade, resulting in excessive ROS accumulation, mitochondrial dysfunction, and apoptosis. Inhibition of AMPK reduced ROS levels, while ROS scavenging partially rescued mitochondrial damage and cell death. Notably, SGLT2 inhibition enhanced sensitivity to EGFR-targeted therapy, producing synergistic anti-tumor effects in vitro and in vivo.

Conclusions

SGLT2 maintains metabolic and autophagic homeostasis in PDAC. Its inhibition induces metabolic stress, autophagic flux blockade, and ROS-driven mitochondrial apoptosis. In addition, targeting SGLT2 sensitizes tumors to EGFR-targeted therapy, offering a novel combinatorial strategy.

Graphical Abstract

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1007/s13402-026-01256-9.

Background

Pancreatic ductal adenocarcinoma (PDAC) remains one of the most lethal malignancies worldwide, with a 5-year survival rate around 13% [1]. Surgical resection remains the only potentially curative option, but fewer than 20% of patients are eligible at diagnosis [2]. Chemotherapy and targeted therapy provide limited benefits due to both intrinsic and acquired resistance [3]. Its aggressive biology, dense desmoplastic stroma, and profound metabolic rewiring contribute to both intrinsic and acquired resistance to standard treatments [3]. As such, identifying tumor-specific vulnerabilities that integrate metabolic and signaling dependencies is a critical unmet need in PDAC research.

One hallmark of PDAC is its reliance on glucose metabolism to sustain rapid proliferation in a nutrient-deprived microenvironment [4]. Although PDAC cells can utilize glutamine, fatty acids, and amino acids as alternative fuels, glucose remains the dominant substrate that drives energy production and biosynthesis [5]. Despite the preference for aerobic glycolysis (the Warburg effect), PDAC cells also retain the capacity to enhance oxidative phosphorylation (OXPHOS) under stress [6]. Metastatic cells may activate the pentose phosphate pathway to produce nucleotides and maintain redox balance, thereby promoting tumor progression [6]. Moreover, the high glycolytic flux contributes to lactate accumulation, acidifying the tumor microenvironment and modulating immune cell function [5]. To meet the elevated metabolic demands, PDAC cells activate autophagy as a survival strategy [7]. Within the nutrient-scarce tumor microenvironment, autophagy recycles cellular components to generate alternative nutrients, thereby sustaining glycolytic flux and supporting continued proliferation under metabolic stress [7].

Glucose uptake and utilization are regulated by specific transporters, including the sodium-glucose cotransporter 2 (SGLT2, gene name SLC5A2) [8]. SGLT2 is a transmembrane protein primarily responsible for active glucose reabsorption in the renal proximal tubule [9]. Emerging evidence indicates that SGLT2 is also expressed in several tumors, potentially supporting glucose uptake under metabolic stress [10, 11]. Pharmacological inhibitors of SGLT2, such as canagliflozin (CANA), are widely used for type 2 diabetes management by inhibiting SGLT2-mediated glucose reabsorption in the renal proximal tubules [8]. Preclinical studies suggest that these agents may have anti-tumor properties in certain cancer types [12–14]. CANA attenuates the proliferation of cancer cells by inhibiting glucose uptake [15] and mitochondrial complex I-supported respiration [16]. In vitro experiments showed that CANA promotes mitochondrial dysfunction and reticulophagy in colorectal cancer cells [17]. Moreover, several studies have reported that CANA can enhance the efficacy of chemotherapy [18, 19], radiotherapy [20], or immunotherapy [13] in certain tumors. Clinical evidence further indicates that SGLT2 inhibitor use is associated with a 23% reduced risk of prostate cancer in diabetic men [21]. However, studies on the expression and function of SGLT2 in PDAC remain inconclusive [22, 23], and the underlying mechanisms remain largely unexplored. Addressing these gaps may reveal novel therapeutic opportunities.

In parallel, aberrant epidermal growth factor receptor (EGFR) signaling is frequently observed in PDAC and contributes to tumor progression [24, 25]. EGFR activates downstream pathways such as PI3K/AKT and MAPK, driving tumor growth and survival [26]. Despite widespread EGFR expression in PDAC, clinical responses to EGFR-targeted therapies are limited [27, 28]. Resistance mechanisms include compensatory signaling, metabolic plasticity, and stromal-mediated protection [27–29]. Previous studies have shown that EGFR activity is closely intertwined with glucose metabolism, where it promotes glycolysis and enhances glucose uptake in cancer cells [30, 31]. This metabolic-signaling crosstalk raises the possibility that metabolic disruption may sensitize PDAC cells to EGFR inhibition, providing a rationale for combinatorial strategies.

In this study, we investigated SGLT2 expression and its functional role in PDAC. We aimed to examine how SGLT2 influences cellular metabolism and stress responses and to explore potential interactions with EGFR-targeted therapy, with the goal of identifying metabolic vulnerabilities that may inform combinatorial treatment strategies.

Materials and methods

Public database analysis

Pan-cancer expression analysis of SGLT2 was performed using Gene Expression Profiling Interactive Analysis 2 (GEPIA2, http://gepia2.cancer-pku.cn/#index) [32] based on The Cancer Genome Atlas (TCGA) and Genotype-Tissue Expression (GTEx) datasets.

Single-cell RNA sequencing data analysis

Single-cell RNA sequencing (scRNA-seq) data for PDAC were obtained from the Gene Expression Omnibus (GEO) database (GSE205013 [33]). Untreated PDAC samples were included for downstream analysis. Quality control was performed using Scanpy (v1.10.4) in Python 3.12.2. Cells with fewer than 200 detected genes or with > 15% mitochondrial gene content were excluded, and genes expressed in fewer than 3 cells were removed. High-quality cells with ≥ 500 detected genes and ≤ 15% mitochondrial proportion were retained, resulting in 80,642 cells for subsequent analyses. Gene expression matrices were normalized and log-transformed, followed by scaling to unit variance.

Principal component analysis (PCA) was applied for dimensionality reduction, and batch effects across samples were corrected using the Harmony algorithm. Uniform Manifold Approximation and Projection (UMAP) was then used for visualization. Cell clustering was performed using the Leiden algorithm, and cell types were manually annotated based on canonical marker gene expression and established literature.

Immunohistochemistry (IHC)

The tissue microarray (TMA) used in this study contained 49 PDAC tissues and paired adjacent normal paraffin-embedded specimens, provided by the Department of Hepatobiliary and Pancreatic Surgery, Peking University First Hospital, in accordance with the guidelines of the Ethics Committee of Peking University First Hospital (approval number: 2024 − 194). Written informed consent was obtained from all participants prior to inclusion in the study. Paraffin sections were baked at 62 °C for 1 h, dewaxed, and subjected to heat-mediated antigen retrieval. After washing, the sections were incubated with the primary anti-SGLT2 antibody (Invitrogen, PA5-101893; Thermo Fisher Scientific, Waltham, MA, USA; 1:100) overnight at 4 °C. The following day, sections were incubated with a horseradish peroxidase (HRP)-conjugated secondary antibody for 30 min at 37 °C. Finally, slides were mounted with neutral gum and cover slipped. Following immunohistochemical staining, two independent pathologists evaluated the expression levels of the target protein. Quantitative analysis of staining intensity and the percentage of positive tumor cells was performed using ImageJ IHC Profiler.

Cell culture

Normal pancreatic ductal epithelial cells (HPNE) and pancreatic cancer cell lines (PANC-1, Capan-1, MIA PaCa-2, AsPC-1, BxPC-3, PA-TU-8988T, and T3M-4) were purchased from the American Type Culture Collection (ATCC, Manassas, VA, USA), the German Collection of Microorganisms and Cell Cultures (DSMZ, Braunschweig, Germany), or AcceGen (Fairfield, NJ, USA). Cells were cultured in DMEM or RPMI-1640 medium (Gibco, Thermo Fisher Scientific, Waltham, MA, USA) supplemented with 10% fetal bovine serum (FBS; Procell Life Science & Technology, Wuhan, China) at 37 °C in a humidified atmosphere containing 5% CO₂.

Cell proliferation and colony formation

Cell viability was measured using the Cell Counting Kit-8 (CCK-8; Dojindo Laboratories, Kumamoto, Japan; CK04). For colony formation, 500 PANC-1 cells or 700 Capan-1 cells per well were seeded into six-well plates and cultured for 10–14 days until visible colonies formed. Colonies were fixed with 4% paraformaldehyde and stained with 0.1% crystal violet.

Wound-healing and transwell assays

Wound-healing and Transwell assays were performed to evaluate cell migration. For wound healing, confluent cells were scratched with a pipette tip, washed, and incubated in serum-free medium. Images were captured at 0, 24, and 48 h. For Transwell assays, 3 × 10⁴ Capan-1 cells or 1 × 10⁴ PANC-1 cells in serum-free medium were seeded into the upper chamber (6.5 mm, 8.0 μm pore; Corning Inc., Corning, NY, USA; Cat. No. 3422) without Matrigel. The lower chamber was filled with medium containing 10% fetal bovine serum (FBS) as a chemoattractant. After 48 h, migrated cells were fixed with 4% paraformaldehyde, stained with 0.1% crystal violet, and counted. For knockdown experiments, cells were analyzed 48 h after seeding, while for drug treatment experiments, drugs were added 24 h after seeding and cells were analyzed 48 h later. Images were acquired using a DP74 Microscope Digital Camera (Olympus, Tokyo, Japan).

Flow cytometry

Apoptosis was assessed using an Annexin V-APC/7-AAD apoptosis detection kit (KeyGEN BioTECH, Jiangsu, China; KGA1106-50) according to the manufacturer’s protocol. Samples were analyzed on a CytoFLEX flow cytometer (Beckman Coulter, USA).

Metabolic assays

Glucose concentration in culture supernatants was determined using the Glucose Assay Kit-WST (Dojindo; G264). Lactate, Adenosine triphosphate (ATP), Nicotinamide adenine dinucleotide (oxidized form, NAD⁺), and Nicotinamide adenine dinucleotide (reduced form, NADH) levels were measured using the L-Lactate Assay Kit with WST-8 (Beyotime Biotechnology, Shanghai, China; S0208S), ATP Assay Kit (Beyotime Biotechnology; S0026), and NAD⁺/NADH Assay Kit with WST-8 (Beyotime Biotechnology; S0175), respectively, according to the manufacturers’ protocols and normalized to protein content. Mitochondrial function was assessed using the Seahorse XF Cell Mito Stress Test Kit (Agilent Technologies, Santa Clara, CA, USA) with a Seahorse XFe24 Analyzer. Final drug concentrations were 1.5 µM oligomycin, 0.5 µM rotenone/antimycin A (Rot/AA), and Carbonyl cyanide p-trifluoromethoxyphenylhydrazone (FCCP) at 1.5 µM for Capan-1 cells and 2.5 µM for PANC-1 cells.

ROS, mitochondrial membrane potential, and lysosomal detection

Intracellular ROS was detected using the Reactive Oxygen Species Assay Kit (DHE; Beijing Solarbio Science & Technology Co., Ltd., Beijing, China; CA1420). Mitochondrial membrane potential (ΔΨm) was measured using the Mitochondrial Membrane Potential Assay Kit with TMRE (Beyotime Biotechnology; C2001S). Lysosomes were stained with Lyso-Tracker Red (Beyotime Biotechnology; C1046). Nuclei were counterstained with Hoechst 33,342 Staining Solution for Live Cells (100X; Beyotime Biotechnology; C1028). Images were captured using the Olympus DP74 camera.

Dual-fluorescence LC3B reporter assay

Cells were infected with pLenti-CMV-mCherry-GFP-LC3B-IRES-Puro-WPRE lentivirus (OBiO Technology). Images were acquired using a Confocal Laser Scanning Microscope (Leica Microsystems, Wetzlar, Germany).

In vivo tumor xenograft assays

Male BALB/c-nu mice were obtained at 3 weeks of age (average 13 g; Charles River Laboratories, Viatonglihua, CAnN.Cg-Foxn1nu/Crl, code 401) and acclimated for 1 week. At 4 weeks of age (average 16 g), mice were randomly assigned to different groups. PANC-1 cells were resuspended in a 1:1 mixture of DMEM and Corning^® Matrigel^® Basement Membrane Matrix High Concentration (HC; Corning, 354248) at 4 × 10⁶ cells per 100 µL and injected subcutaneously into the axilla of each mouse. One week after injection, mice received daily oral gavage of 100 µL vehicle (10% DMSO, 40% PEG300, 10% Tween 80, 40% H₂O), Canagliflozin (50 mg/kg), Erlotinib (25 mg/kg), or the combination. Tumor volume and body weight were measured three times per week. Tumor volume was calculated as V = length × width² / 2. Experiments were planned to terminate when tumors reached 1000 mm³, after 4 weeks of treatment, or if body weight decreased by more than 15%. In practice, tumors did not exceed 800 mm³, and experiments were stopped accordingly. All procedures were approved by the Ethics Committee for Animal Research at Peking University First Hospital (approval number J2025049) and followed institutional guidelines.

Drug treatment and gene silencing

Canagliflozin (S2760; Selleck Chemicals, Houston, TX, USA), Erlotinib (S7786; Selleck Chemicals), and Compound C (HY-13418 A; MedChemExpress, MCE, Monmouth Junction, NJ, USA) were dissolved in DMSO and diluted in culture medium. N-acetylcysteine (HY-B0215; MCE) was dissolved in water and diluted in culture medium. SGLT2 knockdown was achieved using a lentiviral shRNA vector (pSLenti-U6-shRNA(SLC5A2)-CMV-EGFP-F2A-Puro-WPRE; OBiO Technology, Shanghai, China). Stable cells were selected with puromycin (HY-K1057; MCE), and knockdown efficiency was confirmed by Western blot.

Western blot and qRT-PCR

Proteins were extracted using RIPA buffer supplemented with protease and phosphatase inhibitors (Beijing LABLEAD, Beijing, China). Equal amounts of protein were separated by SDS-PAGE and transferred to 0.22 μm PVDF membranes. Membranes were blocked with 5% nonfat milk for 1 h at room temperature, incubated with primary antibodies overnight at 4 °C, and then washed three times with TBST. After incubation with HRP-conjugated secondary antibodies (Abclonal) for 1 h at room temperature, signals were detected using the enhanced chemiluminescence (ECL) reagent. The following primary antibodies were used: SGLT2 (Invitrogen, PA5-101893, 1:1000; #14120, CST, 1:1000), Phospho-AMPKα (Thr172; #2535, Cell Signaling Technology, CST, Danvers, MA, USA; 1:1000), AMPKα1/α2 (A27099, Abclonal, Wuhan, China; 1:2000), Phospho-ULK1 (Ser555, #5869, CST; 1:1000), ULK1 (#8054, CST; 1:1000), SQSTM1/p62 (#5114, CST; 1:1000), LC3B (#3868, CST; 1:1000), Bax (#2772, CST; 1:1000), Bcl-2 (#3498, CST; 1:1000), Cleaved Caspase-3 (#9661, CST; 1:1000), Caspase-3 (#14220, CST; 1:1000), EGFR (A11351, Abclonal; 1:2000), and Phospho-EGFR (Tyr1068, AP0994, Abclonal; 1:1000; #3777, CST, 1:1000). β-Actin (AC026, Abclonal; 1:80000) was used as the loading control. HRP-conjugated secondary antibodies were from Abclonal (AS014, goat anti-rabbit IgG (H + L), 1:10000; AS003, goat anti-mouse IgG (H + L), 1:10000).

Total RNA was extracted using TRIzol reagent (Invitrogen, Thermo Fisher Scientific). Reverse transcription was performed with Hifair^® AdvanceFast 1st Strand cDNA Synthesis SuperMix for qPCR (Yeasen Biotechnology, Shanghai, China). qRT-PCR was carried out using Hieff^® qPCR SYBR Green II Master Mix (Yeasen Biotechnology). Primers were synthesized by Generay Biotech (Shanghai, China). The qRT-PCR primer sequences were as follows:

SLC5A2, forward 5′-CTGTTTGCACCCGTGTACCT-3′, reverse 5′-CCTGTCACCGTGTAAATCATGG-3′; ACTB, forward 5′-CCTGGACTTCGAGCAAGAGATGG-3′, reverse 5′-CAGGAAGGAAGGCTGGAAGAGTG-3′.

Statistical analysis

Unless otherwise stated, all in vitro experiments were repeated independently at least three times, with three biological replicates included in each experiment. For in vivo experiments, n = 5 mice per group. Data are presented as mean ± standard error of the mean (SEM). Statistical analyses were conducted using GraphPad Prism (version 10.1.2; GraphPad Software, San Diego, CA, USA). Parametric tests were applied when the data were considered approximately normally distributed with comparable variance among groups. Paired IHC data were analyzed using the Wilcoxon signed-rank test. Comparisons between two groups were performed using two-tailed Student’s t-test, and multiple group comparisons were conducted using one-way ANOVA followed by Tukey’s or Dunnett’s post hoc test, as appropriate. Kaplan-Meier survival curves were analyzed using the log-rank test.

Combination effects of canagliflozin (CANA) and erlotinib (Erlo) were evaluated using the Chou-Talalay method. Cells were treated with increasing concentrations of CANA and Erlo alone or in combination at a fixed ratio of 2:1 (CANA: Erlo). After 48 h of treatment, cell viability was assessed using the CCK-8 assay. The combination index (CI) was calculated using CompuSyn software (ComboSyn Inc., USA), where CI < 1 indicates synergism, CI = 1 indicates an additive effect, and CI > 1 indicates antagonism.

P < 0.05 was considered statistically significant.

Results

SGLT2 is overexpressed in pancreatic cancer and correlates with poor prognosis

We first evaluated the expression pattern of SGLT2 across various cancer types using transcriptomic data from the GEPIA2 platform. SGLT2 expression was found to be upregulated in multiple cancer types, including PDAC (Fig. 1A). Subsequently, single-cell RNA sequencing data were obtained from a publicly available pancreatic cancer dataset (GSE205013). The expression pattern of SGLT2 was further examined using single-cell RNA sequencing data. After rigorous quality control and filtering, 11 PDAC samples were integrated, resulting in a total of 80,642 high-quality cells retained for downstream analyses (Fig. 1B). Unsupervised clustering identified 23 distinct clusters, which were subsequently annotated into 11 major cell types based on canonical marker gene expression profiles. Specifically, NK/T cells (CD3D, CD3E, CD2, CD7, IL7R), ductal cells (EPCAM, KRT19, KRT7, MUC1, MUC5AC, CEACAM5, CEACAM6), acinar cells (PRSS1, CPA1, CELA3A, CELA3B), endocrine cells (INS, IAPP), fibroblasts (DCN, COL1A1, COL1A2, COL3A1, ACTA2), endothelial cells (VWF, PECAM1, PLVAP), macrophages/monocytes (CD14, S100A8, S100A9, C1QA, C1QB, FCGR3A), pericytes (RGS5, PDGFRB), Schwann cells (S100B), mast cells (TPSAB1, TPSB2), and B cells (CD19, MS4A1, CD79A, CD79B) were identified. These results provide a comprehensive cellular landscape of PDAC, forming the basis for subsequent analyses of cell-type-specific SGLT2 expression (Fig. 1C-E). Overall, SGLT2 expression was relatively low across the PDAC single-cell landscape, with SGLT2-positive cells accounting for a small proportion of the total cell population. Nevertheless, SGLT2 expression was preferentially observed in subsets of ductal epithelial cells and acinar cells, as illustrated by UMAP feature plots and violin plots (Fig. 1F, G). We further verified and analyzed the expression of SGLT2 in pancreatic cancer tissues at the histological level. IHC was performed on a tissue microarray containing paired pancreatic tumor and adjacent normal samples. Representative images demonstrated stronger SGLT2 staining localized predominantly to tumor epithelial regions compared to adjacent normal tissues (Fig. 1H). Quantitative analysis of percentage of SGLT2⁺ cells revealed significantly elevated SGLT2 expression in tumor samples (Fig. 1I; p = 0.0002). To investigate the prognostic relevance of SGLT2, Kaplan-Meier survival analysis was conducted. Patients with high SGLT2 expression (percentage of SGLT2⁺ cells > median) exhibited significantly shorter overall survival than those with low expression (percentage of SGLT2 + cells ≤ median) (Fig. 1J).

Fig. 1.

(A) Pan-cancer SGLT2 mRNA expression based on TCGA and GTEx datasets, showing upregulation in pancreatic adenocarcinoma (PAAD). (B-E) UMAP visualization of 11 cell clusters from single-cell RNA sequencing of untreated pancreatic cancer samples. (F, G) Feature and violin plots showing SGLT2 expression in ductal epithelial and acinar cell subsets. (H, I) Representative immunohistochemical images and quantification of SGLT2 expression in paired pancreatic tumor and adjacent normal tissues. (J) Kaplan-Meier curves for overall survival of pancreatic cancer patients stratified by SGLT2 expression. Hazard ratio (HR) = 4.522, 95% confidence interval (CI): 1.098–18.62; p = 0.0316. ns, not significant; *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001

SGLT2 promotes proliferation, migration, and tumorigenicity of pancreatic cancer cells

To evaluate the biological significance of SGLT2 in pancreatic cancer, its mRNA expression levels were examined across a panel of pancreatic cancer cell lines. Compared with the non-tumorigenic pancreatic ductal epithelial cell line HPNE, all tested cancer cell lines showed markedly elevated SGLT2 expression, with Capan-1 and PANC-1 exhibiting the highest levels (Fig. 2A). Consistently, Western blot analysis confirmed increased protein expression of SGLT2, with the highest expression observed in PA-TU-8988T, followed by PANC-1 and Capan-1 (Fig. 2B). Given their robust expression and extensive use as in vitro models, PANC-1 and Capan-1 were selected for subsequent functional studies.

Fig. 2.

(A) Relative SGLT2 mRNA expression and (B) SGLT2 protein expression in HPNE and pancreatic cancer cell lines. (C) Western blot confirming SGLT2 knockdown in Capan-1 and PANC-1. (D, E) Cell proliferation curves post-knockdown. (F) Colony formation assays. (G) Transwell assays. (H, I) Wound healing assays. (J, K) Apoptotic cell analysis. (L) Representative tumor images from nude mice injected with shNC or shSGLT2 PANC-1 cells. (M) Tumor weights at endpoint. shNC vs. shSGLT2: 0.3251 g vs. 0.1841 g, p = 0.0012 (N) Tumor growth curves. ns, not significant; *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001

Stable knockdown of SGLT2 was achieved in both Capan-1 and PANC-1 cells using two independent shRNAs. Knockdown efficiency was confirmed by Western blot (Fig. 2C). CCK-8 assays revealed that SGLT2 knockdown significantly inhibited cell proliferation in both cell lines. Compared with control, both shRNA groups displayed consistently lower viability over time (Fig. 2D, E). The effect on suppressing long-term growth was further supported by colony formation assays, which showed a significant decrease in colony numbers in knockdown groups (Fig. 2F). SGLT2 knockdown also attenuated cell motility, as evidenced by reduced cell migration in Transwell assays (Fig. 2G; p < 0.0001 for Capan-1, p < 0.01 for PANC-1), and corroborated by wound healing assays (Fig. 2H, I). To assess the impact of SGLT2 on pancreatic cancer cell survival, Annexin V-APC/7-AAD staining combined with flow cytometric analysis demonstrated a marked increase in apoptosis in SGLT2-deficient cells (Fig. 2J, K), indicating that SGLT2 may contribute to the survival of pancreatic cancer cells.

We next assessed the impact of SGLT2 silencing on tumorigenicity in vivo. PANC-1 cells transduced with shNC or shSGLT2 were injected subcutaneously into nude mice, and tumor formation was monitored. Tumors derived from shSGLT2 group exhibited markedly smaller size and reduced mass compared with control (Fig. 2L, M). Consistently, growth curves showed that SGLT2 knockdown significantly suppressed tumor progression over time (Fig. 2N). Together, these findings indicate that SGLT2 promotes pancreatic cancer cell survival and growth.

Pharmacological inhibition of SGLT2 suppresses proliferation and induces apoptosis in pancreatic cancer cells

To further evaluate the therapeutic potential of SGLT2 inhibition in pancreatic cancer, we examined the effect of selective SGLT2 inhibitor CANA in vitro. A dose-dependent cytotoxicity was observed via CCK-8 assays following 48-hour treatment, with IC₅₀ values of 132.4 µM for Capan-1 and 99.11 µM for PANC-1 cells (Fig. 3A, B). In addition, CANA suppressed cell proliferation in a time-dependent manner (Fig. 3C, D). Colony formation assays revealed a significant reduction in long-term proliferative potential following CANA treatment (Fig. 3E). CANA treatment also impaired migration capacity. Transwell assays showed a marked decrease in the number of migrated cells upon treatment (Fig. 3F). Consistently, wound-healing assays confirmed reduced migration in both Capan-1 and PANC-1 cells (Fig. 3G, H). After 48 h of exposure to 50 µM CANA, the migration area rate was reduced significantly. Furthermore, flow cytometric analysis of Annexin V-APC/7-AAD staining demonstrated that CANA induced an increased proportion of apoptotic cells compared to control (Fig. 3I, J). Collectively, these findings indicate that CANA-mediated inhibition of SGLT2 suppresses proliferation and induces apoptosis in pancreatic cancer cells.

Fig. 3.

(A, B) CCK-8 assays in Capan-1 and PANC-1 cells after 48-hour CANA treatment; IC₅₀ values were calculated. (C, D) Effect of CANA on cell proliferation. (E) Colony formation assays. (F) Transwell assays. (G, H) Wound healing assays. (I, J) Apoptotic cell analysis. ns, not significant; *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001

SGLT2 inhibition disrupts glucose metabolism and mitochondrial function, triggering energy crisis and oxidative stress in pancreatic cancer cells

To elucidate the mechanisms underlying the cytotoxic effects induced by SGLT2 inhibition, we next examined its impact on cellular energy metabolism and mitochondrial function. Following 48-hour treatment with CANA, the glucose uptake, lactate generation, and ATP production were significantly decreased (Fig. 4A-C). To further characterize oxidative metabolism, we performed Seahorse XF Cell Mito Stress Test. CANA exposure resulted in a pronounced reduction in ATP production and maximal respiration (Fig. 4D-F). These observations suggest that SGLT2 activity is essential for sustaining energy homeostasis in pancreatic cancer cells, and its inhibition impairs both glycolytic activity and mitochondrial oxidative phosphorylation. In parallel, the NAD⁺/NADH ratio declined after treatment (Fig. 4G), suggesting NADH accumulation and disruption of redox homeostasis (Fig. 4H). To assess downstream consequences of energy dysregulation, we evaluated intracellular oxidative stress and mitochondrial integrity. DHE staining showed a marked increase in reactive oxygen species (ROS) levels following CANA treatment (Fig. 4I, J), suggesting increased oxidative stress. In parallel, measurement of mitochondrial membrane potential using TMRE staining revealed a significant hyperpolarization (Fig. 4K, L), a hallmark of early mitochondrial dysfunction and apoptotic priming. To determine whether these alterations culminate in mitochondria-dependent apoptosis, we assessed apoptosis-related protein expression by Western blot. While BCL-2-associated X protein (Bax) expression remained relatively unchanged, BCL-2 levels were significantly reduced and cleaved cysteine-aspartic acid protease 3 (Caspase-3) was markedly upregulated, indicating activation of mitochondria-dependent apoptosis (Fig. 4M).

Fig. 4.

(A, B) Glucose and lactate concentration in the culture medium. (C) ATP generation normalized to total protein content. (D-F) Seahorse XF Cell Mito Stress Test: oxygen consumption rate (OCR) curves, ATP production, maximal respiration. (G, H) NAD⁺/NADH ratio and NADH levels. (I, J) Intracellular ROS by DHE staining. (K, L) Mitochondrial membrane potential by TMRE staining. (M) Western blot of Bax, Bcl-2, cleaved caspase-3, and total caspase-3. (N) KEGG pathway enrichment of significantly altered metabolites following CANA treatment (top 15 pathways shown). ns, not significant; *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001

Finally, metabolomic profiling and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis revealed that CANA treatment led to significant alterations in metabolic pathways, particularly those involved in nucleotide biosynthesis, redox regulation, energy metabolism, and stress signaling (Fig. 4N). Top enriched pathways included purine metabolism, nicotinate and nicotinamide metabolism, oxidative phosphorylation, and the adenosine monophosphate-activated protein kinase (AMPK) signaling pathway, consistent with an energy crisis and metabolic stress. Similar metabolic and mitochondrial alterations were observed in SGLT2-knockdown cells (Supplementary Fig. 1), further supporting the role of SGLT2 in maintaining energy homeostasis in pancreatic cancer cells. Given that CANA may exert SGLT2-independent effects at relatively high concentrations, we further examined whether CANA-induced metabolic and apoptotic changes were influenced by SGLT2 expression levels. SGLT2 knockdown reduced the additional ATP-depleting and pro-apoptotic effects of CANA, whereas SGLT2 overexpression alleviated CANA-induced ATP depletion and apoptosis only to a modest extent (Supplementary Fig. 2). These findings suggest that SGLT2 contributes to the cellular response to CANA, although the incomplete rescue indicates that SGLT2-independent mechanisms may also be involved.

SGLT2 inhibition causes autophagic flux blockade, leading to oxidative stress and mitochondrial apoptosis

In response to this energy deficit and redox imbalance, we next investigated whether key metabolic signaling pathways were activated. AMPK, a well-established energy sensor, is rapidly activated under conditions of metabolic stress. Western blot showed increased phosphorylation of AMPK and unc-51-like kinase 1 (ULK1), accompanied by accumulation of microtubule-associated protein 1 light chain 3 beta (LC3B)-II and sequestosome 1 (p62) after CANA treatment (Fig. 5A). Inhibition of AMPK with Compound C (CC) reduced p-AMPK and reversed the increase of LC3B-II/I and p62 (Fig. 5B), suggesting that CANA-induced autophagy initiation is AMPK dependent. Time-course Western blot analysis showed that, compared with DMSO, CANA treatment led to a progressive increase in LC3B-II/I and p62 levels from 6 to 48 h (Fig. 5C, D), suggesting altered autophagic turnover. To further evaluate autophagic flux, cells were treated with the lysosomal inhibitor bafilomycin A1 (BafA1). As expected, BafA1 treatment resulted in the accumulation of LC3B-II and p62. Notably, co-treatment with CANA and BafA1 caused no marked additional increase in LC3B-II and p62 compared with CANA treatment alone (Fig. 5E), supporting that CANA impairs autophagic flux. Consistent with this, confocal imaging of cells expressing mCherry-GFP-LC3B revealed increased puncta formation and a higher yellow/red ratio after CANA treatment (Fig. 5F, G), further supporting impaired autophagosome-lysosome fusion. LysoTracker staining showed that lysosomal acidification appeared to be preserved (Fig. 5H, I), suggesting that the blockade was not due to primarily lysosomal dysfunction.

Fig. 5.

(A) Western blot analysis of p-AMPK, p-ULK1, LC3B-II/I and p62 after CANA treatment. (B) Western blot showing the effects of CC on p-AMPK, LC3B-II/I and p62 in the presence of CANA. (C, D) Time-course Western blots of LC3B-II/I and p62 at 6, 12, 24 and 48 h in DMSO- and CANA-treated cells. (E) Western blot analysis of LC3B-II/I and p62 in cells treated with CANA and/or the lysosomal inhibitor BafA1. (F, G) Representative confocal images and quantification of mCherry-GFP-LC3B expressed in cells; puncta counts and yellow/red ratio are shown. (H, I) Representative images and quantification of lysosomal acidification (LysoTracker). (J, K) Intracellular ROS levels in cells treated with CANA ± CC. (L, M) Mitochondrial membrane potential in cells treated with CANA ± NAC. (N, O) Apoptosis analysis in cells treated with CANA ± NAC. ns, not significant; *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001

We next examined whether the AMPK-dependent autophagy contributed to ROS accumulation and mitochondrial changes. CC treatment partially reversed CANA-induced ROS increase (Fig. 5J, K). Similarly, the ROS scavenger N-Acetylcysteine (NAC) partially reduced mitochondrial hyperpolarization and apoptosis caused by CANA (Fig. 5L-O). These results indicate that CANA activates AMPK-ULK1-dependent autophagy initiation but impairs autophagosome-lysosome fusion. The autophagy blockade results in excessive ROS generation, mitochondrial dysfunction, and apoptotic cell death.

SGLT2 inhibition enhances the sensitivity of pancreatic cancer to EGFR-targeted therapy

To further investigate essential pathways involved in the cellular response to SGLT2 inhibition, we performed transcriptomic analysis of PANC-1 cells following CANA treatment. KEGG pathway enrichment analysis of upregulated genes identified several significantly enriched pathways, including the ErbB signaling pathway (Fig. 6A). As EGFR is a central effector in the ErbB family, we next evaluated EGFR expression and phosphorylation. Western blot analysis showed that CANA treatment increased total EGFR and phosphorylated EGFR (p-EGFR, Tyr1068) levels compared with DMSO controls (Fig. 6B), while SGLT2 knockdown reduced EGFR and p-EGFR levels (Fig. 6C). To further examine the divergent effects of pharmacological treatment and genetic knockdown on EGFR signaling, we generated SGLT2 knockdown, SGLT2 overexpression, and SGLT2 re-expression cell models. SGLT2 knockdown reduced EGFR phosphorylation, whereas SGLT2 re-expression restored the decrease in p-EGFR induced by SGLT2 knockdown. In contrast, SGLT2 overexpression alone did not further increase p-EGFR levels (Fig. 6D). We further performed a time-course analysis of EGFR activation following CANA treatment in PANC-1 cells. CANA induced a time-dependent increase in p-EGFR, which reached a relatively high level at approximately 24 h and remained elevated at 48 h (Fig. 6E). These results indicate that CANA treatment and SGLT2 knockdown affect EGFR signaling differently, and that SGLT2 re-expression can rescue the reduction in p-EGFR caused by SGLT2 knockdown.

Fig. 6.

(A) KEGG pathway enrichment analysis of upregulated genes in PANC-1 cells following CANA treatment, highlighting significantly enriched pathways including the ErbB signaling pathway. (B, C) Western blots of EGFR and p-EGFR in CANA-treated or SGLT2-knockdown cells, respectively. (D) Western blot analysis of SGLT2, EGFR, and p-EGFR in SGLT2 knockdown (KD), SGLT2 overexpression (OE), and SGLT2 re-expression cell models. (E) Time-course Western blot analysis of p-EGFR in PANC-1 cells treated with 0.1% DMSO or 100 µM CANA for 0, 12, 24, and 48 h. (F, G) CI plots for CANA and Erlo co-treatment. (H, I) CCK-8 assays. (J) Colony formation assays. (K) Transwell assays. (L, M) Wound healing assays. (N, O) Flow cytometry analysis of apoptosis. (P-S) Xenograft tumor images, growth curves, and endpoint tumor weights; body weight monitored during treatment. ns, not significant; *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001

To assess the therapeutic potential of combined SGLT2 and EGFR inhibition, Capan-1 and PANC-1 cells were co-treated with CANA and the EGFR tyrosine kinase inhibitor (EGFR-TKI) erlotinib (Erlo). Dose-response curves for CANA, Erlo, and their combination are shown in Supplementary Fig. 3. Combination index (CI) analysis demonstrated synergistic effects (CI < 1) at moderate to high fraction affected in both cell lines (Fig. 6F, G). Consistently, combination treatment more effectively suppressed short-term proliferation (Fig. 6H, I) and long-term colony formation (Fig. 6J) than either agent alone. Cell migration was also more strongly impaired by the combination, as shown in Transwell (Fig. 6K) and wound healing assays (Fig. 6L, M). Furthermore, flow cytometry analysis revealed a significant increase in apoptotic cells in the combination group relative to monotherapies (Fig. 6N, O).

To validate these findings in vivo, a subcutaneous xenograft model was established using PANC-1 cells. Tumor growth was significantly suppressed by the combination of CANA and erlotinib compared to either monotherapy or vehicle control (Fig. 6P-R). At the endpoint, the mean tumor volume in the combination group was markedly lower than that in the control, CANA, and erlotinib groups. No significant body weight loss was observed during treatment, indicating acceptable tolerability (Fig. 6S). Collectively, these results demonstrate a synergistic effect between SGLT2 inhibition and EGFR-targeted therapy in pancreatic cancer models, and support the potential of combination therapeutic strategies.

Discussion

In this study, we found that SGLT2 is significantly upregulated in pancreatic cancer tissues and correlates with poor overall survival. Functional experiments demonstrated that both genetic knockdown and pharmacologic inhibition of SGLT2 suppressed tumor cell proliferation, clonogenic capacity, and migration, while promoting apoptosis. Mechanistically, our data suggest that SGLT2 targeting disrupted energy homeostasis, impaired autophagic flux, induced oxidative stress, and triggered mitochondrial apoptosis. Notably, co-treatment with an EGFR inhibitor exerted a synergistic anti-tumor effect.

SGLT2 has been reported to be overexpressed in several types of malignancies, including lung [34], gastric [35], and breast cancers [36]. However, SGLT2 expression in PDAC has not been fully characterized. In this study, we re-evaluated SGLT2 expression using both bulk and single-cell transcriptomic data, as well as clinical samples. In our cohort, although the proportion of SGLT2-positive tumor cells was relatively low, its expression was significantly higher in tumor tissues compared with paired adjacent normal tissues (n = 39). More importantly, high SGLT2 expression was associated with significantly shorter overall survival (OS) among PDAC patients (n = 47), suggesting its potential as a negative prognostic marker. Despite limitations in sample volume and follow-up, our data provide clinically relevant evidence linking SGLT2 to PDAC progression. Recent studies indicated that SGLT2 inhibitors may reduce pancreatic cancer risk in patients with type 2 diabetes compared to other glucose-lowering medications [37, 38]. Together, these findings support a functional and prognostic role for SGLT2 in PDAC.

Given the metabolic function of SGLT2 and the dependence of PDAC cells on glucose-derived energy, we investigated how SGLT2 supports tumor survival. SGLT2 inhibition induces acute ATP depletion, leading to activation of the AMPK/ULK1 axis, which promotes autophagy initiation under metabolic stress [39]. Meanwhile, our findings suggest that autophagic degradation is impaired under these conditions, resulting in altered autophagic flux. Autophagic flux blockade impairs mitochondrial clearance and leads to the accumulation of dysfunctional mitochondria [40]. Moreover, under downstream flux blockade, the enhanced formation of autophagosomes promotes the accumulation of membranous structures [41], potentially exacerbating metabolic stress. In this context, autophagy may shift from a protective to a stress-promoting process, thereby contributing to oxidative stress [42]. Consistently, our experimental results showed that inhibition of AMPK with CC partially attenuated CANA-induced ROS accumulation. Excessive ROS can damage mitochondrial membranes and compromise mitochondrial quality control [43]. ROS disrupts mitochondrial homeostasis, as indicated by increased mitochondrial membrane potential, ultimately promoting mitochondria-dependent apoptosis. These effects were partially reversed by the ROS scavenger NAC. While previous studies reported that SGLT2 inhibition by CANA induces autophagy in benign and malignant diseases [17, 44, 45], our findings reveal a distinct context in PDAC, where SGLT2 inhibition triggers autophagy initiation but is associated with impaired autophagic degradation. Together, these results support a role for SGLT2 in maintaining metabolic and autophagic homeostasis in PDAC cells under energy stress. Disruption of this balance may create a metabolic vulnerability that can be therapeutically exploited.

Another important observation of this study is that SGLT2 inhibition is associated with alterations in EGFR signaling and enhances the efficacy of EGFR-targeted therapy. In the present study, pharmacological CANA treatment increased EGFR phosphorylation, whereas genetic SGLT2 knockdown reduced EGFR and p-EGFR levels. Additional rescue experiments showed that SGLT2 re-expression restored the decrease in p-EGFR caused by SGLT2 knockdown, while SGLT2 overexpression alone did not further enhance p-EGFR. These findings suggest that SGLT2 contributes to the maintenance of basal EGFR signaling, although increased SGLT2 expression alone may not be sufficient to further activate EGFR. By contrast, CANA treatment induced a time-dependent increase in p-EGFR, suggesting that EGFR activation after pharmacological SGLT2 inhibition may reflect an adaptive response to acute metabolic stress. Therefore, the divergent effects of CANA treatment and SGLT2 knockdown on EGFR phosphorylation may be related to different modes and durations of SGLT2 perturbation rather than a simple linear relationship between SGLT2 expression and EGFR activation.

Although EGFR is widely expressed in pancreatic cancer [24], the clinical benefit of EGFR inhibitors has been limited [46, 47]. EGFR has been shown to promote aerobic glycolysis in triple-negative breast cancer and lung adenocarcinoma cells [30, 31], and ligand binding to EGFR and other ErbB family receptors can enhance glucose uptake [48]. In this context, metabolic perturbation may in turn reshape receptor signaling networks, potentially creating new therapeutic vulnerabilities. Functionally, we observed that SGLT2 inhibition significantly enhanced sensitivity to erlotinib in both Capan-1 and the relatively EGFR-insensitive PANC-1 cell lines, demonstrating synergistic effects in vitro (CI < 1). This synergistic effect was further validated in vivo, where combination treatment produced greater tumor suppression than either monotherapy. These findings suggest that metabolic stress induced by SGLT2 inhibition may increase the functional relevance of EGFR signaling under metabolic stress conditions, thereby rendering tumor cells more susceptible to EGFR-targeted therapy.

Consistently, prior studies in NSCLC have demonstrated that glucose deprivation synergizes with autophagy or AKT inhibitors to overcome acquired resistance to EGFR-targeted therapy [49], and blocking glucose metabolism increases EGFR-TKI sensitivity in resistant tumors [50]. The underlying mechanism remains unclear, but may involve metabolic stress-induced receptor activation, altered trafficking, or adaptive responses to disrupted glucose homeostasis.

Despite the insights provided by this study, several limitations should be acknowledged. First, the tissue microarray cohort used for survival analysis was relatively small, which may have contributed to the wide confidence interval of the hazard ratio. In addition, multivariate analysis was not performed because of the limited sample size and incomplete clinicopathological information. Larger well-annotated cohorts are therefore needed to further evaluate the independent prognostic value of SGLT2 in PDAC. Second, the mechanisms underlying autophagic disruption and EGFR signaling modulation by SGLT2 inhibition remain to be fully elucidated. Although CANA is clinically approved, potential off-target effects at experimental doses cannot be excluded. In our rescue experiments, SGLT2 knockdown and overexpression only partially modified CANA-induced ATP depletion and apoptosis, suggesting that both SGLT2-dependent and SGLT2-independent mechanisms may contribute to the observed anti-tumor effects. Further studies using additional genetic rescue systems and more selective approaches are needed to clarify the specificity of CANA-mediated effects. Additionally, while EGFR inhibitors have shown modest efficacy in K-Ras wild-type PDAC [47], both cell lines used in this study (Capan-1 and PANC-1) harbor K-Ras mutations, and in vivo validation was limited to PANC-1 xenografts. Therefore, whether SGLT2 inhibition broadly enhances EGFR-targeted therapy requires further validation across metabolically heterogeneous PDAC models.

In conclusion, our study identifies SGLT2 as a key metabolic regulator in PDAC, with high expression linked to poor prognosis. SGLT2 inhibition disrupts energy and redox homeostasis. Mechanistically, SGLT2 inhibition blocks autophagic flux, leading to ROS accumulation and apoptosis. Importantly, SGLT2 inhibition sensitizes PDAC cells to EGFR-targeted therapy, revealing a metabolic vulnerability with translational potential. Further studies in genetically diverse and clinically relevant models are warranted to support clinical translation.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

PDAC

Pancreatic ductal adenocarcinoma

SGLT2

Sodium-glucose cotransporter 2

CANA

Canagliflozin

OXPHOS

Oxidative phosphorylation

EGFR

Epidermal growth factor receptor

GEPIA2

Gene Expression Profiling Interactive Analysis 2

TCGA

The Cancer Genome Atlas

GTEx

Genotype-Tissue Expression

GEO

Gene Expression Omnibus

PCA

Principal component analysis

UMAP

Uniform manifold approximation and projection

IHC

Immunohistochemistry

CCK-8

Cell Counting Kit-8

OCR

Oxygen consumption rate

ATP

Adenosine triphosphate

NAD+

Nicotinamide adenine dinucleotide, oxidized form

NADH

Nicotinamide adenine dinucleotide, reduced form

Rot/AA

Rotenone/antimycin A

FCCP

Carbonyl cyanide p-trifluoromethoxyphenylhydrazone

ROS

Reactive oxygen species

Bax

Bcl-2-associated X protein

Bcl-2

B-cell lymphoma 2

KEGG

Kyoto Encyclopedia of Genes and Genomes

AMPK

Adenosine monophosphate-activated protein kinase

ULK1

Unc-51-like kinase 1

p62

Sequestosome 1

LC3B

Microtubule-associated protein 1 light chain 3 beta

CC

Compound C

BafA1

Bafilomycin A1

NAC

N-acetylcysteine

EGFR-TKI

Epidermal growth factor receptor tyrosine kinase inhibitor

Erlo

Erlotinib

CI

Combination index

KD

Knockdown

OE

Overexpression

Author contributions

Y.W. and E.Z. contributed equally to this work. Y.W. performed most of the experiments, analyzed the data, and drafted the manuscript. E.Z. supervised the experimental design and data interpretation. R.M. and W.L. assisted with cell culture and molecular assays. Q.W. and J.Z. contributed to data analysis and figure preparation. Y.Y. provided technical support and critical discussion. Y.M. and X.T. conceived and supervised the study, revised the manuscript, and approved the final version. All authors read and approved the final manuscript.

Funding

This study was supported by National Natural Science Foundation of China (NO. 82171722, 82271764, 82471772 and 82571996), Beijing Natural Science Foundation (L246015).

Data availability

The datasets generated and analyzed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

This study was approved by the Ethics Committee of Peking University First Hospital (approval no. 2024-194-002). Written informed consent was obtained from the patient. Animal experiments were approved by the Institutional Animal Care and Use Committee of Peking University First Hospital (approval no. J2025049) and conducted in accordance with institutional guidelines for animal welfare.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.