Nordihydroguaiaretic acid inhibits bladder cancer metastasis through suppression of α1,3-mannosyltransferase expression and LRFN4 N-glycosylation

Abstract

Background

Metastasis remains a predominant contributor to cancer-related mortality worldwide. Elucidating the molecular mechanisms underlying cancer metastasis is crucial for developing strategies to inhibit tumor progression and improve clinical outcomes. Protein glycosylation, a hallmark of cancer pathogenesis mediated by specific glycosyltransferases, has emerged as a critical regulatory mechanism. This study investigates the functional role of glycosylation in cancer progression and explores its therapeutic potential.

Methods

We employed an integrated approach combining bioinformatics analysis of datasets, in vitro biological assays, transcriptional profiling, immunoprecipitation, and lectin pull-down assays to characterize α1,3-mannosyltransferase (ALG3)-dependent glycosylation mechanism in bladder cancer metastasis. Therapeutic targeting was investigated through virtual screening, molecular docking, molecular dynamics simulation, cellular thermal shift assay (CETSA), and validated in nude mice model using the Tranditional Chinese Medicine (TCM) monomer (NDGA).

Results

Our analysis revealed significantly elevated ALG3 expression in bladder cancer patients. Combined measurement of ALG3 in urine and serum samples demonstrated strong diagnostic potential, with higher AUC, sensitivity and specificity. Mechanistically, ALG3 promoted oncogenic cell behaviors through Ras signaling pathway activation. Immunoprecipitation and lectin pull-down assays identified LRFN4 as a novel ALG3 target, with ALG3 mediated N-glycosylation of LRFN4 being essential for its oncogenic function. Virtual screening identified nordihydroguaiaretic acid (NDGA) as a potent ALG3 inhibitor, which was validated through molecular docking, molecular dynamics simulation and CETSA. NDGA exhibited significant anti-tumor effects in both in vitro and in vivo models.

Conclusions

Our findings established ALG3 as a promising detection glycobiomarker for bladder cancer, and a key regulator for metastasis through LRFN4 N-glycosylation and Ras signaling pathway activation. The identified ALG3 inhibitor-NDGA demonstrated significant therapeutic potential, offering a foundation for developing personalized treatment strategies against bladder cancer metastasis.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12967-025-06571-7.

Background

Bladder cancer (BCa), a malignancy arising from the urothelial lining of the urinary bladder, ranks as the sixth most common cancer in males globally and the ninth leading cause of cancer-related mortality [1]. Clinically, BCa is classified into two major subtypes based on invasion depth: non-muscle invasive bladder cancer (NMIBC) and muscle invasive bladder cancer (MIBC). While NMIBC exhibits a risk of progression to MIBC, MIBC itself is associated with metastatic recurrence, frequently to lymph nodes, lungs or liver, resulting in significantly worse outcomes than NMIBC [2, 3]. This is reflected in a five-year survival rate below 60% and a substantially reduced quality of life for MIBC patients [4]. Current diagnostic approaches for BCa primarily rely on urine cytology and cystoscopy [5]. However, these methods present critical limitations: Urine cytology, while non-invasive and highly specific, suffers from suboptimal sensitivity, particularly for low-grade tumors [6]. Cystoscopy, despite its high sensitivity, is invasive, costly, and often uncomfortable for patients, limiting its utility for routine monitoring [7]. The high mortality rate of BCa is further compounded by missed or delayed diagnosis, underscoring the urgent need for reliable, non-invasive biomarkers. Additionally, the lack of effective therapies for metastatic BCa highlights the necessity for novel therapeutic targets. Thus, identifying alternative diagnostic biomarkers and metastatic-associated treatment strategies remains crucial to improving clinical detection, therapeutic efficacy, and patient survival.

Metastasis is a major driver of mortality in bladder cancer patients [8], primarily mediated by two key biological processes: epithelial to mesenchymal transition (EMT) and angiogenesis. During EMT, epithelial cells lose cell–cell adhesion and acquire mesenchymal traits, enhancing their migratory and invasive potential. In bladder cancer, EMT has been strongly linked to tumor progression [9], metastatic dissemination [10] and altered therapeutic sensitivity [11]. Tumor angiogenesis, the formation of aberrant blood vessels, supports metastasis by facilitating nutrient and oxygen supply, waste removal, and metastatic spread [12]. Vasculogenic mimicry (VM), a hallmark of aggressive tumors, involves cancer cells forming vascular-like channels that mimic endothelial-lined vessels, further promoting tumor survival and dissemination [13]. Both EMT and VM are associated with increased malignancy, therapeutic resistance, and poorer clinical outcomes in BCa. Therefore, deciphering the molecular mechanisms underlying cancer metastasis is critical for controlling cancer progression, improving patient prognosis, and developing targeted therapies against metastatic BCa.

Mounting evidence establishes protein glycosylation as a critical regulator of cancer metastasis [14, 15]. This dynamic post-translational modification, mediated by glycosyltransferases, involves the enzymatic transfer of oligosaccharide moieties to specific amino acid residues, forming structurally and functionally diverse glycoproteins. Protein glycosylation participates in all stages of oncogenesis, from tumor initiation to metastatic dissemination. When dysregulated, it drives aberrant cell signaling, enhances metastatic potential, and immune evasion mechanisms [16, 17]. Presently, two subtypes of glycosylation, N-glycosylation and O-glycosylation, are the subject of extensive research. Aberrant N-glycosylation, occurring at conserved Asn-X-Ser/Thr (where X represents any amino acid except proline), drives malignant transformation through multiple oncogenic mechanisms. This post-translational modification facilitates tissue invasion [18], induces a pro-inflammatory tumor microenvironment [19], modulates immune checkpoint molecules [20], and promotes metabolic reprogramming [21], collectively contributing to cancer progression and metastasis. The specific inhibition of aberrant glycosylation represents a promising therapeutic strategy, potentially targeting multiple hallmarks of cancer simultaneously.

Glycosyltransferases serves as the primary enzymatic mediators of protein glycosylation. Among hundreds of identified glycosyltransferases, α1,3-mannosyltransferase (ALG3) has emerged as a critical player in cancer pathogenesis. Emerging evidence demonstrated that ALG3 overexpression contributes to tumor progression through distinct mechanisms across multiple cancer types. In triple-negative breast cancer, elevated ALG3-mediated glycosylation of PD-L1 modulated the tumor immune microenvironment and compromised immunotherapy efficacy [22]. In ovarian cancer, high expression of ALG3 promoted peritoneal metastasis by enhancing uPA/uPAR signaling and facilitating uPAR-ADAM8 interactions, establishing its potential as a therapeutic target [23]. Our preliminary findings indicated that ALG3 was significantly upregulated in bladder cancer, and contributed to malignant transformation. However, the precise molecular mechanisms underlying ALG3’s role in bladder cancer progression remains to be fully elucidated.

Building upon the extensive historical applications of Traditional Chinese Medicine (TCM) in disease treatment [24, 25], particularly its emerging role in oncology [26], our study provided compelling evidence for the therapeutic targeting of glycosylation in bladder cancer. We identified ALG3 as both a diagnostic glycobiomarker and a functional mediator of bladder cancer progression. Mechanistically, ALG3 drove tumor aggressiveness through activation of Ras signaling pathway and N-glycosylation-dependent modification of LRFN4. Notably, we discovered that nordihydroguaiaretic acid (NDGA), a bioactive TCM monomer, specifically inhibited ALG3 expression and functions. The results demonstrated NDGA’s potent anti-metastatic effects through ALG3 downregulation, suggesting its potential as a novel therapeutic agent. These findings not only elucidated the molecular mechanisms underlying ALG3-mediated glycosylation in bladder cancer progression, but also validated the scientific rationale for exploring TCM-derived compounds in modern cancer therapy.

Methods and materials

Clinical specimens

Urine and serum samples (n = 44) were collected from bladder cancer patients and healthy controls from the First Affiliated Hospital of Dalian Medical University during the period from 2022 to 2024 (Table S1), with approval from the ethics committee (grant no. PJ-KS-KY-2022-439). The inclusion criteria for bladder cancer were as follows: 1) histologically confirmed bladder cancer by cystoscopy-guided biopsy; 2) newly diagnosed or recurrent cases with documented prior treatment history. The exclusion criteria included: 1) concurrent malignancies of other genitourinary organs (prostate/kidney cancer); 2) active urinary tract infection or inflammatory conditions (interstitial cystitis); 3) patients who had received adjuvant therapy (chemotherapy or radiotherapy) before. The inclusion criteria for healthy controls included: 1) age and gender matched with bladder cancer patients; 2) normal urinary cytology and negative cystoscopy/ultrasound findings; 3) no history of malignancy or chronic systemic diseases (diabetes, autoimmune disorders). The collected samples were immediately frozen at − 80 ℃ for subsequent analysis.

Enzyme-linked immunosorbent assay

ALG3 concentrations in urine and serum samples were quantified using commercial human ALG3 enzyme-linked immunosorbent assay (ELISA) kits (Westang, Shanghai, China), following the manufacturer’s protocol. Briefly, 100 μl of urine (no dilution), serum (1:10 dilution) or standard solutions (0, 125, 250, 500, 1000, 2000, 4000, 8000 pg/ml) were added to anti-ALG3 pre-coated 96-well plates and incubated at 37 ℃ for 40 min in a humidified chamber (Thermo Fisher Scientific, Waltham, MA, USA). After washing 5 times with 300 μl/well of washing buffer, 50 μl of biotin-labeled primary antibody was added, followed by a second incubation at 37 ℃ for 20 min. After repeat washing, 100 μl of streptavidin-horseradish peroxide (HRP)-conjugated antibody was applied and incubated for 10 min at 37 ℃. A colorimetric reaction was initiated by adding 100 μl of tetramethylbenzidine substrate solution to each well, with signals allowed to develop at 37 ℃ for 15 min in the dark. Absorbance was then measured at 450 nm (OD450nm) using a microplate reader (Thermo Fisher Scientific, Waltham, MA, USA), and ALG3 concentration in urine and serum was calculated using standard curve generated with 4-parameter logistic regression.

Immunohistochemical staining

Formalin-fixed, paraffin-embedded tissue sections (4 μm) were obtained from the First Affiliated Hospital of Dalian Medical University, and baked at 60 °C for 2 h prior to processing. After deparaffinization with xylene and rehydration in gradual ethanol (100%, 95%, 80%, 70%), tissue slides underwent antigen retrieval using EDTA buffer (1 mM, pH8.0) for 20 min in a decloaking chamber. Endogenous peroxidase was blocked with 3% hydrogen peroxide for 10 min at room temperature, and non-specific binding was blocked using 10% normal goat serum (ZSGB-BIO, Beijing, China). The slides were then incubated overnight at 4 ℃ in a humidified chamber with primary antibodies: ALG3 (DF14333, 1:100, Affinity Biosciences, Jiangsu, China), E-cadherin (A22850, 1:100, ABclonal Technology, Wuhan, China), N-cadherin (A19083, 1:100, ABclonal Technology, Wuhan, China), VE-cadherin (A25003, 1:200, ABclonal Technology, Wuhan, China), and MMP14 (A2549, 1:100, ABclonal Technology, Wuhan, China). Following incubation, horseradish peroxidase-conjugated secondary antibodies were applied for 1 h at room temperature. Diaminobenzidine and hematoxylin were used for staining, and dehydrated through graded ethanol (70%, 80%, 95%, 100%) and xylene. Finally, the slides were visualized under an inverted microscope (Olympus, Tokyo, Japan), and images were taken. The staining intensity score was determined based on the percentage of positive cells (< 10%: 0; 10–25%: 1, 26%-50%: 2, 51–75%: 3, 76%-100%: 4) and the staining intensity (negative: 0; weak: 1; medium: 2; strong: 3).

Data selection

The TCGA-BLCA dataset, comprising 411 bladder cancer samples and 19 normal tissue samples, was obtained from The Cancer Genome Atlas (TCGA) database (https://portal.gdc.cancer.gov/) via the Genomic Data Commons (GDC) Data portal. Only samples with complete clinical annotations (including age, gender, TNM stage, and survival data) and high-quality RNA-seq data (FPKM ≥ 0.1 in ≥ 50% of samples) were included. Additionally, the GSE222315 dataset, which contained 9 bladder cancer tissues and 4 normal adjacent tissues, was retrieved from the Gene Expression Omnibus (GEO) database (https://www.ncbi.nlm.nih.gov/geo/).

Identification of core cell type expressing ALG3 and associated biological processes using scRNA-seq

The single-cell RNA sequencing (scRNA-seq) dataset-GSE222315 from bladder cancer patients was analyzed using R package “Seurat”. Cells with less than 300 detected genes or more than 5% mitochondrial gene content were excluded to remove low-quality or dying cells, and genes expressed in less than 3 cells were discarded. Counts were normalized using the "LogNormalize" method, and top 2000 highly variable genes (HVGs) were identified via the FindVariableFeatures function. Scaling was performed using ScaleData to regress out UMI counts and mitochondrial gene percentage. Then, principal component analysis was run using RunPCA on scaled HVGs, and significant PCs were selected based on the ElbowPlot. Cell clusters were generated via FindNeighbors and FindClusters, and Uniform Manifold Approximation and Projection (UMAP) was applied for visualization. Clusters were annotated using marker genes from the CellMarker 2.0 database (http://117.50.127.228/CellMarker/) and literature-curated bladder cancer signatures. ALG3 expression across cell clusters was quantified using “FindAllMarkers” function in “Seurat”. Additionally, gene set variation analysis (GSVA) was performed on Hallmark gene sets (MSigDB) using "gsva" function, and differential pathway activity between ALG3-high and ALG3-low clusters was assessed via R package "limma".

Cell culture and transfection

Two human bladder cancer cell lines, T24 (JNO-H0484) and UMUC3 (JNO-H0486), were obtained from Jennio Biotechnology (Guangzhou, China). T24 cells were maintained in RPMI-1640 medium (Gibco, California, USA) supplemented with 10% fetal bovine serum (FBS, Gibco) and 1% penicillin/streptomycin (Gibco). UMUC3 cells were cultured in DMEM-H medium (Gibco) supplemented with 10% FBS and 1% penicillin/streptomycin. All cells were maintained in a humidified incubator with 5% CO₂ at 37 ℃, and the medium was renewed every 2–3 days depending on cell conditions.

For genetic modulation of ALG3, three specific siRNA sequences targeting ALG3 (siRNA-1, siRNA-2, siRNA-3) and a pcDNA3.1-ALG3 overexpression plasmid were designed and synthesized by GenePharma (Shanghai, China). A non-targeting scramble siRNA and empty pcDNA3.1 vector served as negative controls. Transient transfection was performed using Lipofectamine 3000 reagent (Invitrogen, Carlsbad, CA, USA) according to the manufacturer’s protocol. Cells were seeded in six-well plates (3 × 10⁵ cells/well), and siRNA (50 pmol/well) or plasmid DNA (4 μg/well) was mixed with Lipofectamine 3000 at a 1:2 ratio before adding to cells. After 6 h incubation, the medium was replaced with complete culture medium. RNA and protein samples from the transfected cells were harvested 48 h post-transfection for subsequent analysis. The siRNA sequences of ALG3 were as follows: ALG3 siRNA-1: 5′-GCGUCAUCAAUGGUACCUATT-3′ (sense), 5′-UAGGUACCAUUGAUGACGCTT-3′ (anti-sense); ALG3 siRNA-2: 5′-GCUGUGCUCUACCUGGCUATT-3′ (sense), 5′-UAGCCAGGUAGAGCACAGCTT-3′ (anti-sense); ALG3 siRNA-3: 5′-CCAGUUCUACGUCUGGUAUTT-3′ (sense), 5′-AUACCAGACGUAGAACUGGTT-3′ (anti-sense).

Quantitative real-time polymerase chain reaction

Total RNA was isolated from cells using RNAiso Plus reagent (TAKARA, Japan) according to the manufacurer’s instructions. RNA concentration and purity were assessed using NanoDrop 2000 (Thermo Fisher Scientific, Waltham, MA, USA) with acceptable A₂₆₀/A₂₈₀ ratios (1.8–2.0). For cDNA synthesis, 1 μg of total RNA was reverse-transcribed using the PrimeScript™ RT reagent kit (TAKARA). The resulting complementary DNA was subjected to quantitative real-time polymerase chain reaction (qRT-PCR) using TB Green^® Premix Ex Taq™ II (TAKARA) on an ABI7500 real-time PCR system (Applied Biosystems, Life Technologies, Carlsbad, USA). The amplification program was set as follows: initial denaturation at 94 ℃ for 30 s, followed by 40 cycles of amplification at 94 ℃ for 5 s and 60 ℃ for 30 s. Relative gene expression was calculated using the 2^−△△CT method with normalization to GAPDH as the endogenous control. Primer sequences were provided in Table S2.

Western blot

Cells were lysed in ice-cold RIPA buffer (Beyotime Biotechnology, Beijing, China) supplemented with 1 × protease inhibitor Cocktail and 1 mM PMSF (Beyotime Biotechnology) for 30 min on ice. Lysates were centrifuged at 12,000 g for 10 min at 4 °C, and the supernatants were collected. Protein concentration was determined using a BCA protein assay kit (Beyotime Biotechnology). Equal amounts of protein (20 μg) were separated by 12% SDS-PAGE gels and transferred onto nitrocellulose membrances (Millipore, Burlington, MA, USA). Membranes were blocked with 5% non-fat milk in TBST for 2 h at room temperature, followed by incubation with primary antibodies at 4 ℃ overnight: ALG3 (DF14333, 1:1000, Affinity Biosciences, Jiangsu, China), E-cadheirn (A22850, 1:1000, ABclonal Technology, Wuhan, China), N-cadherin (A19083, 1:2000, ABclonal Technology, Wuhan, China), VE-cadherin (A22659, 1:1000, ABclonal Technology, Wuhan, China), MMP14 (A2549, 1:500, ABclonal Technology, Wuhan, China), MMP2 (1:1000, ABclonal Technology, Wuhan, China), MMP9 (1:500, ABclonal Technology, Wuhan, China), Ras (A23382, 1:1000, ABclonal Technology, Wuhan, China), Ras-GAP (A4876, 1:500, ABclonal Technology, Wuhan, China), Raf1 (A0223, ABclonal Technology, Wuhan, China), p-Raf1 (AP0498, 1:500, ABclonal Technology, Wuhan, China), MEK1/2 (A4868, 1:2000, ABclonal Technology, Wuhan, China), p-MEK1/2 (AP1349, 1:2000, ABclonal Technology, Wuhan, China), ERK1/2 (A4782, 1:1000, ABclonal Technology, Wuhan, China), p-ERK1/2 (AP0974, 1:1000, ABclonal Technology, Wuhan, China) and GAPDH (AC001, 1:5000, ABclonal Technology, Wuhan, China). After three 10 min washes with TBST, membranes were incubated with HRP-conjugated goat anti-rabbit/mouse secondary antibodies (1:5000, ABclonal Technology, Wuhan, China) at room temperature for 1 h. Protein bands were visualized using enhanced chemiluminescence reagent and imaged with ChemiDoc™ System (Bio-Rad, Marne la Coquette, France). Quantification was performed using ImageJ software (NIH, Bethesda, MD, USA) with background subtraction.

Cell counting kit-8 assay

Cell proliferation was assessed using the Cell Counting Kit-8 (CCK-8, Beyotime Biotechnology, Beijing, China) according to the manufacturer’s protocol. Briefly, cells were seeded into 96-well plates at a density of 3 × 10³ cells/well in 100 μl complete medium and allowed to adhere overnight. Three to four biological replicates were conducted for each group. After treatment, 10 μl of CCK-8 reagent was added to each well containing 90 μl of serum-free medium and incubated at 37 ℃ for 2 h in a humidified 5% CO₂ incubator. The OD_450nm was measured using a microplate reader, where blank wells (medium + CCK-8 without cells) were included for background subtraction. Data were collected from three independent experiments.

Colony formation assay

A total of 1 × 10³ cells were plated into six-well plates and cultured in complete medium for 10–14 days at 37 ℃ with 5% CO₂, with medium refreshed every 2–3 days. Only colonies containing more than 50 cells were counted as valid. Cell colonies were fixed with 100% methanol at room temperature for 20 min, and stained with 0.5% crystal violet (Beyotime Biotechnology) for 15 min. Stained colonies were imaged and quantified using ImageJ software.

Migration and invasion assays

For migration assays, 1 × 10⁵ cells suspended in 200 μl of FBS-free medium were placed in the upper chamber of the transwell insert (8 μm, Corning Life Sciences, MA, USA), while the lower chamber was filled with 600 μl complete medium as a chemoattractant. For invasion assays, the transwell chambers were pre-coated with Matrigel (1:8 dilution, BD Biosciences, San Jose, CA, USA) and polymerized at 37 ℃ for 2–4 h. 1 × 10⁵ cells suspended in 200 μl of FBS-free medium were added to the upper chamber, while 600 μl complete medium was added to the lower chamber. After incubation for the indicated time at 37 ℃/5% CO₂, non-migrated/invaded cells on the upper membrane surface were removed with cotton swabs, while migrated or invaded cells on the lower surface were fixed with 100% methanol for 20 min and stained with 0.5% crystal violet for 15 min. Following PBS (pH7.4) washes, stained cells were imaged at 100× magnification from 10 random fields per membrane, and counted using ImageJ software.

Tube formation assay

To assess angiogenic potential, a total of 50 μl of Matrigel per well was polymerized in 96-well plates at 37 ℃ for 2–4 h. Subsequently, 5 × 10⁴ treated cells resuspended in 200 μl of serum-free medium were seeded into the gel matrix. Plates were incubated at 37 °C/5% CO₂ for 6–8 h until tubular networks formed. Images were captured under an inverted microscope (Olympus) and quantitative analysis was performed by capturing 3 random fields per well using the Angiogenesis Analyser plugin tool in ImageJ software. The total tube length, number of meshes and branching junctions were measured to evaluate the tube formation ability of bladder cancer cells.

RNA-seq analysis

ALG3-knockdown UMUC3 cells were generated by transfecting cells in six-well plates with 50 pmol ALG3 siRNA for 48 h using Lipofectamine 3000 reagent. Total RNA was isolated using RNAiso Plus reagent (TARAKA), and quantified using a Nanodrop 2000 spectrophotometer (Thermo Fisher). After confirming the quality of the RNA by A₂₆₀/A₂₈₀ ratio, samples were sent to Genedenovo Biotechnology Co., Ltd (Beijing, China) for RNA sequencing (RNA-seq). Differentially expressed genes (DEGs) were identified based on the following criteria: |log(fold change)|≥ 0.59, P-value < 0.05. Gene Ontoloy (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis were conducted based on the DEGs.

Weight Gene Co-expression Network Analysis (WGCNA)

TCGA-BLCA RNA-seq data (n = 411 patient samples) was downloaded from the TCGA database. After removing low-expressed genes and filtering for the top 50% most variable genes based on median absolute deviation, we constructed co-expression networks using the R package “WGCNA”. The soft-thresholding power was determined by scale-free topology criterion (R² > 0.85), and gene modules were identified through topological overlap matrix transformation followed by dynamic tree-cutting. Module-trait associations with ALG3 expression were assessed by correlating module eigengenes using Pearson correlation, with significance determined by Benjamini–Hochberg false discovery rate correction (FDR < 0.05). The most significantly associated module was selected for further analysis, with hub genes defined as those showing high module membership.

Lectin blot

Protein lysates (20 μg per group) were separated by 12% SDS-PAGE gel and electrophoretically transferred to nitrocellulose membranes (Millipore). Membranes were blocked with 5% bovine serum albumin at room temperature for 2 h to prevent nonspecific binding. For glycan detection, membranes were incubated overnight at 4 ℃ with biotinylated Galanthus Nivalis Agglutinin (GNA, B-1245-2, 1:5000, Vector Labs, California, USA) which specifically recognized terminal α1,3-linked mannose residues. Following three 10 min TBST washes, membranes were probed with HRP-conjugated streptavidin secondary antibody (ZB-2404, 1:5000, ZSGB-BIO, Beijing, China) for 1 h at room temperature. Immunoreactive protein bands were visualized using an enhanced chemiluminescent system.

Immunoprecipitation

Total cell proteins were extracted using ice-cold RIPA buffer (Beyotime, P0013B) supplemented with 1 × protease inhibitor cocktail and 1 mM PMSF. 500 μg of total proteins from each group were incubated with 50 μl of anti-LRFN4 antibody-conjugated magnetic beads (Beyotime, Beijing, China) at room temperature for 2 h with gental rotation. Beads were then washed five times with PBS containing 0.1% Tween-20 (PBST) to remove nonspecifically bound proteins. The immunoprecipitated complexes were eluted by boiling in 1 × loading buffer (Beyotime) at 95 ℃ for 10 min. The eluted proteins were separated by 10% SDS-PAGE and transferred to nitrocellulose membranes. Glycosylation status of LRFN4 was evaluated using biotinylated GNA (1:5000) followed by HRP-streptavidin detection, with parallel Western blotting using anti-LRFN4 antibody (1:1000) serving as loading control.

Lectin pull-down assay

Total cell lysates were extracted using ice-cold lysis buffer containing protease inhibitors. A total of 1 mg proteins were incubated with 50 μl of GNA-conjugated agarose beads (AL-1243-5, Vector Laboratories, California, USA) at 4 ℃ overnight with costant rotation. Beads were washed six times with washing buffer, and bound glycoproteins were eluted by boiling in 1 × loading buffer at 95 ℃ for 10 min. Eluted proteins were resolved by 10% SDS-PAGE gel and analyzed by Western blot using anti-LRFN4 antibody (PA5-20,699, 1:1000, Invitrogen, Carlsbad, CA, USA) followed by HRP-conjugated secondary antibody (1:5000). Protein bands were visualized using an enhanced chemiluminescent system.

Virtual screening of the natural products targeting ALG3

The three-dimensional structure of human ALG3 (PDB ID: 6XY7) was prepared using the Protein Preparation Wizard in Schrödinger Suite. The L6810 natural product library (TopScience, 2688 compounds) was prepared for docking using LigPrep with OPLS4 force field, generating possible tautomers and stereoisomers. Molecular docking was performed using Glide SP and XP protocols with the following parameters: (1) grid generation centered on the ALG3 catalytic domain; (2) flexible ligand sampling; (3) enhanced precision scoring. Compounds were ranked by Glide docking score (kcal/mol) and subjected to binding pose analysis using Maestro's interaction fingerprint tool. Based on the binding energy and interactions between the compounds and the ALG3 protein, six candidates were identified as potential therapeutic agents targeting ALG3.

Molecular dynamic simulation

The binding dynamics of ALG3 with NDGA were investigated through all-atom molecular dynamics simulations using Schrödinger software. The initial ALG3-NDGA complex structure, obtained from molecular docking, was processed using the Protein Preparation Wizard, with protonation states of ionizable residues determined at pH 7.4 using PROPKA3.0 and histidine tautomers assigned based on local hydrogen-bonding networks. The system was solvated in an orthorhombic water box with a 10 Å buffer using TIP3P explicit water molecules, neutralized with chloride ions, and parameterized with the OPLS4 force field, where partial atomic charges for NDGA were derived from quantum mechanical calculations at the B3LYP/6-31G level. The energy-minimized system underwent a multi-stage equilibration process, including initial minimization, NVT ensemble equilibration, and NPT ensemble equilibration. Production MD was then performed for 100 ns under NPT conditions using a RESPA integrator with 2 fs time step, LINCS constraints on hydrogen bonds, Particle Mesh Ewald method for long-range electrostatics, and isotropic pressure coupling, with trajectories saved every 10 ps. The stabilized production phase was analyzed for root-mean-square deviation of protein backbone and ligand, root-mean-square fluctuation of residue positions, secondary structure evolution using DSSP algorithm, interaction fingerprints, binding free energy through MM-GBSA, and principal component analysis of collective motions. Visualization and trajectory analysis were performed using Schrödinger's Simulation Interaction Diagram tools, with simulation quality validated by monitoring potential energy convergence, density equilibration, and thermodynamic stability throughout the production phase.

Cellular thermal shift assay

Cells are treated with NDGA or DMSO for 2 h, at 37 ℃ in a 5% CO₂ atmosphere, and havested by gental scraping which were divided into aliquots (100 μl) for heating. The aliquots were subjected to a temperature gradient (43–61 °C in 3 °C increments) using a Bio-Rad T100 thermal cycler, with each temperature maintained for exactly 3 min to allow for protein denaturation. Following heat treatment, samples were immediately cooled on ice for 5 min, and lysed by three freeze–thaw cycles in liquid nitrogen. The lysates were centrifuged (12000g, 10 min, 4 ℃) to separate the soluble (non-denatured) proteins from insoluble (denatured) aggregates. The thermal stability of ALG3 was assessed by quantifying its presence in the soluble fraction through Western blot analysis using a specific anti-ALG3 primary antibody (1:1000). Band intensities were quantified using ImageJ software and normalized to the 43 ℃ control sample. Thermal denaturation curves were generated by plotting the relative soluble ALG3 levels against temperature and fitted to a sigmoidal dose–response model using GraphPad Prism 10.0. A rightward shift in the thermal denaturation curve accompanied by increased protein remaining in the soluble fraction at elevated temperatures in NDGA-treated samples compared to controls was interpreted as evidence of direct target engagement and thermal stabilization of ALG3 by NDGA.

Cell viability assay

To evaluate the cytotoxic effects of NDGA, cells were seeded in 96-well plates at a density of 5 × 10³ cells/well in 100 μl complete medium. Cells were treated with various concentrations of NDGA (0, 1, 3, 6, 9, 12, 15, 18 μM) (CAS 500–38-9, TJS2190, purity 99.91%, TOPSCIENCE, Shanghai, China) for 48 h under standard culture conditions. Subsequently, 10 μl of CCK-8 reagent was added to each well containing 90 μl FBS-free medium, and incubated at 37 ℃ for 2 h. The metabolic activity of viable cells was quantified by measuring absorbance at 450 nm using a microplate reader. Blank wells containing medium plus CCK-8 reagent without cells were included for background subtraction. Dose–response curves were generated using non-linear regression analysis (log(inhibitor) vs. response-variable slope) in GraphPad Prism 10.0, from which the half-maximal inhibitory concentration (IC₅₀) was calculated.

Animal experiments

All animal procedures were approved by the Animal Ethics Committee of Dalian Medical University (No. AEE23158), and conducted in compliance with relevant ethical regulations for animal research.

For xenograft studies, 4 week-old male BALB/c nude mice were acclimatized for one week in SPF conditions (22 ± 1 ℃, 55 ± 5% humidity, 12 h light/dark cycle) with ad libitum access to autoclaved food and water. T24 cells stably overexpressing ALG3 (oeALG3, 3 × 10⁶ cells in 100 μl PBS) mixed 1:1 with Matrigel were injected subcutaneously into the flank of the nude mice. One week post-inoculation when palpable tumors formed, mice were randomly allocated into two groups (n = 10/group): (1) oeALG3, (2) oeALG3 + NDGA (5 mg/kg/day oral gavage). NDGA was freshly prepared in vehicle daily. Tumor dimensions were measured every 48 h by digital caliper, with volume calculated using (length × width²)/2. After 14 days of treatment, mice were euthanized, and tumors were excised, weighed, and divided for either snap-freezing in liquid nitrogen or 4% paraformaldehyde fixation.

For the metastasis model, 4–6 week-old male BALB/c nude mice received intravenous injection of 1 × 10⁶ oeNC and oeALG3 cells in 100 μl PBS via tail vein. After one week, mice were randomized into three groups (n = 10/group): (1) oeNC, (2) oeALG3, (3) oeALG3 + NDGA (5 mg/kg/day, 14 days). After 7 weeks, lungs were harvested, inflated with 4% paraformaldehyde, and metastatic nodules were quantified microscopically after processing for H&E staining. Histological analysis was performed by a blinded pathologist using ImageScope software.

Statistical analysis

All data were derived from a minimum of three biologically independent replicates and were presented as mean ± standard deviation (SD). Graphical representations and preliminary analysis were generated using GraphPad Prism 10.0 (GraphPad, San Diego, CA, USA), with advanced statistical analysis conducted in SPSS 26.0 (SPSS Inc., Chicago, IL, USA). For comparisons between two groups, Student’s t test was performed after verifying normality and homogeneity of variance. Multiple group comparisons were analyzed by one-way analysis of variance (ANOVA) with Tukey's post-hoc test for normally distributed data or Kruskal–Wallis test with Dunn's correction for non-parametric distributions. All statistical tests were performed at a 95% confidence interval, with significance thresholds denoted as follows: *P < 0.05, **P < 0.01, ***P < 0.001 and ns indicated no significance.

Results

Elevated ALG3 expression as a potential screening and diagnostic glycobiomarker in bladder cancer

ALG3 demonstrates widespread expression across multiple human tissues, including the gastrointestinal tract, hepatobiliary system, genitourinary tract, and reproductive organs (Fig. S1). Building upon our previous findings of pan-cancer ALG3 upregulation, particularly in bladder cancer tissues and serum, we conducted a comprehensive evaluation of its diagnostic potential. Immunohistochemical analysis confirmed significant ALG3 overexpression in bladder cancer tissues compared to adjacent normal tissues (Fig. 1A). Quantitative ELISA assays revealed markedly elevated ALG3 levels in both urine (Fig. 1B) and serum (Fig. 1C) samples from bladder cancer patients versus age- and gender-matched healthy controls. Diagnostic performance assessment through ROC analysis yielded distinct profiles for different sample types: serum ALG3 detection showed superior sensitivity (79.5%, AUC: 0.816), while urine analysis demonstrated higher specificity (88.6%, AUC: 0.750). Notably, the combined evaluation of both urine and serum samples achieved optimal diagnostic accuracy with an AUC of 0.866, specificity of 90.9%, and maintained sensitivity of 75.0% (Fig. 1D). These findings supported a stratified diagnostic approach utilizing serum ALG3 for initial screening (maximizing sensitivity) followed by urine confirmation (ensuring specificity), with the combined assay representing the most robust diagnostic indicator.

Fig. 1.

Diagnostic performance of ALG3 in urine and serum samples for bladder cancer detection. A Representative immunohistochemical staining of ALG3 in bladder cancer tissues and adjacent normal tissues from five patients. Brown staining indicated positive ALG3 expression. Scale bar = 20 μm. B, C ALG3 concentrations measured by ELISA in urine (n = 44) (B) and serum (n = 44) (C) samples from bladder cancer patients versus age- and gender-matched healthy controls. D ROC curves demonstrating the diagnostic accuracy of urinary ALG3 (blue line, AUC: 0.750, sensitivity: 56.8%, specificity: 88.6%) and serum ALG3 (purple line, AUC: 0.816, sensitivity: 79.5%, specificity: 77.3%) based on ELISA data for bladder cancer detection. Optimal cutoffs were determined by Youden's index. ***P < 0.001

ALG3 was predominantly expressed in anti-FcgRIIB monocytes and tissue stem cells, closely associated with cancer metastasis

To investigate the role of ALG3 in cancer initiation and progression, we analyzed scRNA-seq data from the GSE222315 dataset. Following quality control, principle component analysis, and dimensionality reduction (t-SNE/UMAP), cell clusters were identified based on specific cell markers (Fig. 2A–C). The proportional distribution of cell populations in adjacent normal tissues versus bladder cancer tissues was subsequently quantified (Fig. 2D). Gene Set Variation Analysis (GSVA) revealed functional heterogeneity among cell subtypes: gamma_delta T cells were enriched in proliferation-related pathways, whereas epithelial cells, monocyte, and tissue stem cells exhibited signatures linked to cancer metastasis. In contrast, anti-FcgRIIB monocyte and NK cells were implicated in signal transduction (Fig. 2E). Notably, ALG3 expression was restricted to six cell types (Fig. 2F–K), with highest levels detected in anti-FcgRIIB monocytes (Fig. 2J) and tissue stem cells (Fig. 2K). GSVA analysis further demonstrated that ALG3-high cell populations were positively correlated with Hallmark pathways critical for metastasis, including angiogenesis and epithelial to mesenchymal transition. These findings suggested that ALG3 might serve as a key regulator in bladder cancer progression, particularly by promoting metastatic mechanisms.

Fig. 2.

Association between ALG3 expression and metastatic potential in bladder cancer revealed by scRNA-seq. A UMAP visualization of scRNA-seq data depicting the global transcriptomic landscape of cells derived from bladder cancer tissues (blue) and normal bladder tissues (orange). B UMAP plot annotated with major cell types identified by canonical marker genes. C Dot plot illustrating the expression levels and percentage of cells expressing cell type-specific markers. D Stacked bar plot comparing the relative proportions of major cell populations between normal tissues and bladder cancer tissues. E GSVA enrichment scores for 50 HALLMARK pathways across cell types, with red indicating activation and blue indicating suppression in different cell populations. F–K Violin plots showing ALG3 expression in specific subsets: B cells (F), Epithelial cells (G), CD16+ Monocyte (H), gamma delta T cells (I), anti-FcgRIIB Monocyte (J) and Tissue stem cells (K). Dots represented individual cells; horizontal lines marked median expression. ***P < 0.001

ALG3 overexpression promoted tumorigenesis in bladder cancer

To elucidate the functional role of ALG3 in bladder cancer pathogenesis, we designed three ALG3 targeting siRNAs (ALG3 siRNA-1, ALG3 siRNA-2, ALG3 siRNA-3) and an ALG3 overexpression plasmid to modulate endogenous ALG3 expression. Quantitative analysis confirmed significant downregulation or upregulation of ALG3 at both mRNA (Fig. 3A–D) and protein levels (Fig. 3E, F). Among the siRNAs, ALG3 siRNA-1 demonstrated the most efficient knockdown and was selected for subsequent experiments alongside the overexpression construct. Functional assays revealed that ALG3 knockdown substantially suppressed cell proliferation (Fig. 3G, H) and colony formation (Fig. 3K), whereas ALG3 overexpression enhanced both proliferative capacity (Fig. 3I, J) and clonogenicity (Fig. 3L). Collectively, these findings demonstrated that ALG3 acted as a pro-tumorigenic driver in bladder cancer, fostering malignant growth and survival.

Fig. 3.

ALG3 promoted bladder cancer cell proliferation and tumorigenesis in vitro. A–D qRT-PCR quantification of ALG3 mRNA levels in T24 and UMUC3 cells transfected with three independent ALG3-targeting siRNA sequences (ALG3 siRNA) or negative control siRNA (scramble) (A, B), and cells overexpressing ALG3 cDNA (ALG3 cDNA) versus empty vector control (mock) (C, D). Data normalized to GAPDH and expressed as fold-change relative to scramble or mock controls. E, F Western blot analysis confirmed ALG3 knockdown by ALG3 siRNA (E) and overexpression by ALG3 cDNA (F) transfection in T24/UMUC3 cells. GAPDH served as the loading control. Representative blots from three replicates shown. G–J Cell proliferation measured by CCK-8 assays at 0, 24, 48, 72, and 96 h post-transfection in ALG3-knockdown (G-H) and ALG3-overexpressing cells (I, J). K–L Colony formation assays of ALG3 siRNA-treated (K) and ALG3 cDNA-treated cells (L) after 14 days culture. Left: Representative crystal violet-stained wells. Right: Quantified colonies (> 50 cells/colony). *P < 0.05, **P < 0.01, ***P < 0.001

ALG3 promoted epithelial to mesenchymal transition and vasculogenic mimicry in bladder cancer

EMT is a critical process driving cancer metastasis [27], characterized by the lose of epithelial nature and gain of mesenchymal phenotype, enabling cell migration and invasion. ALG3 knockdown significantly attenuated the migratory (Fig. 4A) and invasive (Fig. 4C) capabilities of bladder cancer cells, concomitant with upregulated E-cadherin and downregulated N-cadherin expression both at mRNA (Fig. 4E, F) and protein levels (Fig. 4I). Conversely, ALG3 overexpression potentiated EMT, enhancing cell migration (Fig. 4B), invasion (Fig. 4D), and EMT markers expression (Fig. 4G–I). Angiogenesis, particularly vasculogenic mimicry (VM), is essential for tumor metastasis, wherein cancer cells form vessel-like networks to sustain nutrient supply and hematogenous dissemination [28]. Strikingly, ALG3 knockdown impaired VM, reducing tube formation (length, mesh and junctions), whereas ALG3 overexpression augmented these parameters (Fig. 4J). Consistent with this, VM-associated markers (VE-cadherin, MMP14, MMP2 and MMP9) were downregualted upon ALG3 silencing (Fig. 4K, L, 4O), and elevated upon ALG3 upregulation (Fig. 4M–O). Together, these data highlighted that ALG3 orchestrated bladder cancer metastasis by coordinately regulating EMT and VM, suggesting its potential as a therapeutic target.

Fig. 4.

ALG3 promoted metastatic potential in bladder cancer through EMT and VM. A–D Transwell assays evaluating metastatic potential of T24/UMUC3 cells following ALG3 knockdown or ALG3 overexpression. Left: Representative images (Scale bars = 50 μm). Right: Quantified cell counts (10 random fields/well). E–H qRT-PCR analysis of EMT markers (E-cadherin, N-cadherin) showing ALG3-mediated transcriptional regulation. Data normalized to GAPDH and expressed as fold-change vs. controls. I Western blot confirmed ALG3-dependent protein-level changes in EMT markers after ALG3 siRNA or cDNA transfection. GAPDH served as the loading control. J Tube formation assays demonstrating ALG3’s role in promoting VM, as measured by total tube length, mesh and junction number. Scale bar = 50 μm. K–N qRT-PCR of VM markers (VE-cadherin, MMP14, MMP2 and MMP9) revealed ALG3-dependent regulation. O Western blot validated VM marker protein expression changes. GAPDH served as the loading control. *P<0.05, **P < 0.01, ***P < 0.001

ALG3 activated Ras signaling pathway to drive bladder cancer progression

After observing the roles of ALG3 in the tumorigenesis and metastasis of bladder cancer, we hypothesized how cells might respond to signals from ALG3 and alter their biological behaviors. To elucidate the molecular mechanisms underlying ALG3-mediated oncogenic effects, we performed RNA-seq analysis of ALG3 knockdown UMUC3 cells. Differential expression analysis revealed 716 upregulated and 772 downregulated genes (Fig. 5A). Functional enrichment analysis (GO) of these 1488 ALG3-associated DEGs identified significant associations with extracellular matrix remodeling, cell growth, wound healing, and apoptosis, processes critical for tumorigenesis and cancer metastasis (Fig. 5B). Notably, the Ras, Rap1, and TGF-beta signaling pathways emerged as top-related pathways (Fig. 5C), prompting further investigation into Ras due to its established oncogenic role. Mechanistically, ALG3 knockdown inhibited Ras activation, reducing the conversion of GDP-bound (inactive) to GTP-bound (active) form. This suppression propagated downstream, decreasing phosphorylation of Raf1, MEK1/2 and ERK1/2 (Fig. 5D). Conversely, ALG3 overexpression potentiated Ras signaling, enhancing GTP-Ras levels and subsequent phosphorylation of Raf1, MEK1/2 and ERK1/2 (Fig. 5E). These findings demonstrated that ALG3 acted as a critical upstream regulator of the Ras-Raf-MEK-ERK cascade, thereby promoting proliferation and metastasis in bladder cancer.

Fig. 5.

ALG3 activated the Ras signaling pathway. A RNA-seq analysis identified differentially expressed genes between ALG3-knockdown and scramble control groups in UMUC3 cells. Thresholds: |log2 fold change|≧0.59 and P < 0.05. Red/green dots represented significantly up-/down-regulated genes. B GO enrichment analysis of ALG3-regulated DEGs, showing the top 20 significantly enriched biological processes (BP), molecular functions (MF), and cellular components (CC). Bubble size indicated gene count; color represented P-value. C KEGG pathway analysis of ALG3-associated DEGs to identify the altered signaling pathways, revealed Ras signaling pathway as one of the most significantly enriched pathways. Bubble size indicated gene count; color represented P-value. D, E Western blot validation of Ras signaling pathway activation (Ras-GTP, p-Raf1, p-MEK1/2, p-ERK1/2) following transfection with ALG3 siRNA or cDNA. GAPDH served as the loading control. Band intensities were quantified using ImageJ and normalized to GAPDH. **P < 0.01, ***P < 0.001

ALG3 catalyzed aberrant N-glycosylation of LRFN4

WGCNA is a widely used tool for constructing co-expression networks that are highly associated with specific phenotypes. By applying stringent thresholds (gene significance > 0.2 and module membership > 0.3), adjusting correlation coefficients, and clustering various gene modules (Fig. 6A, B), the MEyellow module showed strongest correlation with ALG3 (cor = 0.37, P-value = 7e−14) was selected (Fig. 6C). In total, we identified 835 ALG3-associated genes (Fig. 6D). Intersection of these genes with 716 ALG3-downregulated genes from RNA-seq yielded 30 overlapping candidates (Fig. 6E). After excluding non-coding RNAs, GeneMANIA network analysis prioritized three ALG3-interacting candidates: LRFN4, MOGS and SLC37A4 (Fig. 6F). Molecular docking confirmed the interaction between ALG3 and LRFN4 (Fig. 6G). In silico prediction (NetNGlyc 1.0) revealed that only LRFN4 and MOGS had potential glycosites as glycoproteins. LRFN4, a single-pass type I membrane protein, harbored six conserved N-glycosites (Asn25, 70, 324, 333, 376, 449) (Fig. 6H, I). Functional assays demonstrated that ALG3 critically regulated LRFN4 glycosylation: Lectin blot showed reduced α-1,3-mannosylation upon ALG3 knockdown and enhanced α-1,3-mannosylation of LRFN4 after ALG3 overexpression (Fig. 6J). Immunoprecipitation with LRFN4 antibody, followed by GNA lectin detection, confirmed ALG3-dependent N-glycosylation changes on LRFN4 (Fig. 6K). Lectin pull-down assays using GNA-coated beads validated that ALG3 modulated LRFN4 α-1,3-mannosylation levels (Fig. 6L). Collectively, these results established LRFN4 as a functional α-1,3 mannosylated substrate of ALG3, implicating this modification in ALG3-driven oncogenesis.

Fig. 6.

Identification of LRFN4 as a functional target of ALG3-mediated N-glycosylation in bladder cancer. A WGCNA parameter optimization. Left: Scale independence analysis (R² > 0.9) versus soft threshold power (β). Right: Mean connectivity analysis confirming β = 3 as optimal power for network construction. B Hierarchical clustering dendrogram of co-expression modules from 5000 most varible genes, with each branch representing a gene and each color indicating a module. C Module-trait correlation heatmap showing MEyellow module (Pearson's r = 0.37, P = 7e−14) as most strongly associated with ALG3 expression. Color scale: − 1 (blue) to + 1 (red). D Hub gene identification in MEyellow module. Scatter plot of gene significance (GS) for ALG3 expression versus module membership (MM). Thresholds: MM = 0.3 and GS = 0.2. E Venn diagram intersecting ALG3-downregulated DEGs (n = 716, RNA-seq, |log2FC|≧0.59, P < 0.05) with MEyellow module genes (n = 835), yielding 30 candidate targets. F GeneMANIA functional network of 30 overlapping genes from Venn analysis. Edge thickness reflected interaction confidence (co-expression: purple; physical interactions: pink). ALG3 showed direct predicted binding with LRFN4, MOGS and SLC37A4. G Molecular docking of ALG3-LRFN4 complex (ALG3: green; LRFN4: blue). H, I N-glycosites prediction in LRFN4 using NetNGlyc-1.0 database (H) and structural visualization (I), identifying six potential glycosites (Asn25, 70, 324, 333, 376, 440). J Lectin blot showing global α1,3-mannosylation reduction with ALG3 siRNA and increased with ALG3 cDNA transfection. K Immunoprecipitation followed by Western blot confirmed ALG3-dependent α1,3-mannosylation changes specifically on LRFN4. L Lectin pull-down assays showing decreased α1,3-mannosylation LRFN4 after ALG3 knockdown and increased α1,3-mannosylation after ALG3 overexpression

N-glycosylation of LRFN4 drove bladder cancer progression via Ras signaling activation

To investigate the functional significance of LRFN4 glycosylation in bladder cancer, we generated glycosylation-deficient LRFN4 mutants (LRFN4 MUT) by substituting all six glycosites (Asn 25, 70, 324, 333, 376, 440) and compared their effects with wild-type LRFN4 (LRFN4 WT). As illustrated, deglycosylation of LRFN4 profoundly suppressed malignant phenotypes: reduced cell viability (Fig. 7A), impaired clonogenic capacity with reduced number of colonies (Fig. 7B), attenuated migration and invasion (Fig. 7C, D), and disrupted vasculogenic mimicry with decreased tube length, number of meshes and junctions (Fig. 7E). These demonstrated the crucial role of LRFN4 N-glycosylation in promoting tumorigenesis and metastasis in bladder cancer.

Fig. 7.

N-glycosylation of LRFN4 drove malignant progression in bladder cancer cells. A Cell proliferation assessed by CCK-8 assays in T24 and UMUC3 cells transfected with wild-type LRFN4 (LRFN4 WT) or N-glycosites-deficient mutant-type LRFN4 (LRFN4 MUT). Absorbance at 450 nm was measured at 0, 24, 48, 72, and 96 h post-transfection. B Colony formation capability after 14 days culture. Left: Representative crystal violet-stained wells. Right: Quantification of colonies (> 50 cells/colony). C, D Transwell assays assessing metastatic potential of bladder cancer cells after LRFN4 WT and LRFN4 MUT transfection. Left: Representative images (scale bars = 50 μm). Right: Quantified cell counts (10 random fields/well). E Tube formation assays evaluating the vasculogenic mimicry formation ability in LRFN4 WT or LRFN4 MUT transfected cells. Left: Representative images (scale bars = 50 μm). Right: Quantified parameters (total tube length, number of meshes and junctions). F–H Western blot analysis of Ras signaling pathway activation (Ras-GTP, p-Raf1, p-MEK1/2, p-ERK1/2) in T24 (F) and UMUC3 (G) cells transfected with LRFN4 WT or LRFN4 MUT. Statistical analysis were shown (H). *P<0.05, **P < 0.01, ***P < 0.001

To further explore the role of LRFN4 glycosylation in Ras signaling activation, we compared the effects of wild-type LRFN4 (LRFN4 WT) with its glycosylation-deficient mutant (LRFN4 MUT) in bladder cancer cells. Western blot results revealed that while LRFN4 WT effectively activated the Ras signaling pathway, LRFN4 MUT lost this capacity. Specifically, the conversion of Ras to its active GTP-bound form (Ras-GTP) was significantly impaired in LRFN4 MUT-transfected cells compared to LRFN4 WT controls. This defect in Ras activation was accompanied by reduced phosphorylation of downstream effectors, including Raf1, MEK1/2, and ERK1/2 (Fig. 7F–H). These findings demonstrated that N-glycosylation was essential for LRFN4-mediated activation of the Ras signaling pathway in bladder cancer cells.

Identification and validation of NDGA as a potential ALG3 inhibitor from Traditional Chinese Medicine

Our findings establishing ALG3’s critical role in bladder cancer progression prompted us to explore its therapeutic potential. Using a machine learning-assisted drug discovery approach, we performed structure-based hierarchical virtual screening of 2688 TCM monomers to identify potential ALG3 inhibitors (Fig. 8A). The screening identified six candidate compounds with favorable binding characteristics to ALG3 (Table S3). Notably, nordihydroguaiaretic acid (NDGA) emerged as the top candidate, demonstrating the strongest binding affinity with a docking score of − 9.34 kcal/mol. Molecular docking analysis revealed that NDGA formed six hydrogen bonds and three hydrophobic interactions with ALG3 (Fig. 8B). Additional promising candidates included: Diosbulbin B (docking score: − 9.039 kcal/mol), which interacted with ALG3 via one hydrogen bond, three hydrophobic interactions and one salt bridge (Fig. 8C). Four other compounds (1-piperoylpiperidine, (E,E)-, rel-(8R,8'R)-dimethyl-(7S,7'R)-bis(3,4-m, corylin and dihydrolycorine) with docking scores between − 8.08 and − 9.0 kcal/mol indicated a weaker affinity for ALG3 (Fig. 8D–G). Based on its superior binding profile, we selected NDGA for further investigation as a potential ALG3-targeted therapeutic agent.

Fig. 8.

Virtual screening identified potential Traditional Chinese Medicine targeting ALG3. A Systematic screening workflow: Initial screening of 2688 compounds from the Traditional Chinese Medicine Database; ADME property filtering; Molecular docking against the ALG3 catalytic domain using Schrödinger Glide; Final selection of top candidates based on Lipinski’s rule of five and docking scores. B–G Predicted binding modes and molecular interactions between the ALG3 catalytic domain and top six candidate compounds: Nordihydroguaiaretic acid (docking score: − 9.34 kcal/mol) (B), Diosbulbin B (docking score: − 9.039 kcal/mol) (C), 1-piperoylpiperidine, (E,E)- (docking score: − 8.382 kcal/mol) (D), rel-(8R,8'R)-dimethyl-(7S,7'R)-bis(3,4-m (docking score: − 8.213 kcal/mol) (E), Corylin (docking score: − 8.14 kcal/mol) (F) and Dihydrolycorine (docking score: − 8.08 kcal/mol) (G). All compounds were visualized using PyMOL

To characterize the binding stability of the ALG3-NDGA complex, we conducted 100-ns molecular dynamics (MD) simulation. The root-mean-square deviation (RMSD) analysis demonstrated system equilibration, with both ALG3 and NDGA achieving stable conformations after 70 ns of simulation time (Fig. 9A). Throughout the trajectory, ALG3 maintained structural integrity with modest fluctuations (less than 4 Å), while NDGA exhibited even greater positional stability (less than 3 Å) within the binding pocket (Fig. 9B, C). Secondary structure analysis confirmed that ALG3 maintained its native conformation without significant structural deformation (Fig. 9D). Key stabilizing interactions involved four critical residues: ARG 354, HIS 297, PRO 216 and LEU 212, contributed to the stable binding between ALG3 and NDGA (Fig. 9E–G). The binding free energy calculated by MM-GBSA yielded a thermodynamically favorable value of − 45.7812 kcal/mol (Fig. 9H), strongly supporting NDGA as a high-affinity ALG3 binder. These comprehensive simulations provided robust structural and energetic evidence for NDGA as a promising ALG3-targeting compound derived from Traditional Chinese Medicine.

Fig. 9.

Identification and validation of the stability and interaction of the ALG3-NDGA complex. A Root-mean-square deviation (RMSD) trajectories of ALG3 (green) and NDGA (red) over a 100-ns molecular dynamic simulation. The system reached equilibrium after ~ 70 ns, confirming complex stability. B, C Root-mean-square fluctuation (RMSF) profiles revealed regions of ALG3 with high flexibility and stable NDGA-binding sites. D Secondary structure evolution of ALG3 during the simulation process, suggesting structural integrity. E Interaction map identified key ALG3 residues forming hydrogen bonds and hydrophobic contacts with NDGA. F Timeline of interactions and contacts between ALG3 and NDGA demonstrated persistent contacts with specific interaction types throughout simulation. G Summary of interaction types between ALG3 and NDGA, including hydrogen bonds, hydrophobic contacts, ionic interactions, and water bridges. H Binding free energy (MM-GBSA) calculation revealed favorable binding energy, confirming spontaneous complex formation. I, J Cellular thermal shift assays assessing NDGA-ALG3 binding affinity in T24 (I) and UMUC3 cells (J) showed significant thermal stabilization of ALG3 upon NDGA treatment, indicating cellular target engagement. K, L Thermal denaturation curves of ALG3 validated NDGA-ALG3 binding, with NDGA-treated groups showing increased ALG3 stability at higher temperatures compared to DMSO controls in both T24 (K) and UMUC3 (L) cells. Band intensities were quantified by ImageJ and normalized to 43 ℃ control

To validate the NDGA-ALG3 interaction in a cellular context, we performed cellular thermal shift assay in intact cells. Western blot showed that at higher than 46 ℃, ALG3 retained higher percentage of its soluble fraction in NDGA-treated cells than in DMSO-treated cells, confirming that NDGA binding enhanced ALG3 thermostability (Fig. 9I, J). Quantitative analysis of soluble ALG3 levels across a temperature gradient (43–61 ℃) revealed a significant thermal stabilization effect in NDGA-treated cells compared to DMSO-treated controls (Fig. 9K, L). Specifically, NDGA treatment induced a rightward shift in the thermal denaturation curve, corresponding to ΔTm of 50.72 ± 0.18 ℃ (T24) and 53.41 ± 0.23 ℃ (UMUC3). These results provided direct evidence of target engagement and suggested that NDGA stabilized ALG3 in its native cellular environment.

NDGA suppressed bladder cancer tumorigenesis through ALG3 downregulation

Having established NDGA as an ALG3-binding compound, we investigated its anti-tumor effects in bladder cancer models. Dose–response assays revealed NDGA significantly suppressed proliferation of T24 (IC₅₀ = 7.259 ± 0.373 μM) and UMUC3 (IC₅₀ = 15.547 ± 0.256 μM) cells in a concentration-dependent manner (Fig. 10A). Western blot analysis confirmed concomitant dose-dependent reduction of ALG3 protein levels following 48 h NDGA treatment (Fig. 10B), suggesting direct targeting of ALG3 by NDGA. The functional relationship between NDGA and ALG3 was further demonstrated by enhanced cellular sensitivity to NDGA upon ALG3 knockdown, with IC₅₀ values decreasing from 12.23 μM to 6.785 μM in T24 cells (Fig. 10C) and from 12.37 μM to 8.562 μM in UMUC3 cells (Fig. 10D). Rescue experiments showed NDGA effectively counteracted the pro-tumorigenic effects of ALG3 overexpression, as evidenced by CCK8 assays (Fig. 10E, F) and colony formation assays (Fig. 10G). In vivo validation using oeALG3 T24 cells demonstrated NDGA’s therapeutic potential, with significant reductions in both tumor weight (Fig. 10H–J) and volume (Fig. 10K) following 14-days NDGA treatment compared to oeALG3 controls. These consistent in vitro and in vivo findings established NDGA as a promising therapeutic agent that suppressed bladder cancer progression through targeted ALG3 inhibition.

Fig. 10.

NDGA inhibited ALG3-mediated tumorigenesis in bladder cancer. A Dose–response curves of NDGA in T24 and UMUC3 cells were generated after stimulating with various concentrations of NDGA (CCK-8 assays, 48 h treatment). Calculated IC₅₀ values: T24: 7.25 ± 0.373 μM; UMUC3: 15.547 ± 0.256 μM. B Western blot showing NDGA induced dose-dependent reduction of ALG3 protein levels (48 h treatment). GAPDH was loading control. C, D CCK8 assays evaluating alterations of IC₅₀ values after cells transfected with ALG3 siRNA or combined with NDGA blocking. Combined ALG3 siRNA and NDGA treatment synergistically enhanced sensitivity in bladder cancer cells, shifting IC₅₀ values from 12.23 μM to 6.785 μM (T24) and 12.37 μM to 8.562 μM (UMUC3). E, F CCK-8 assays confirmed NDGA resistance reversal in ALG3-overexpressing cells compared to mock controls. G Colony formation assays validated that ALG3 overexpression conferred resistance to NDGA, which was abrogated by combined NDGA treatment. H, I Overview of tumor formation in nude mice after subcutanous injection with oeALG3 cells (3 × 10⁶ cells each mice) and oral administration of NDGA (5 mg/kg/day, 14 days). Representative tumors at endpoint (day 21). J Tumor weight comparison between oeALG3 and oeALG3 plus NDGA groups. Final tumor weight decreased in oeALG3 combined NDGA treatment group. K Tumor volume comparison between oeALG3 and oeALG3 plus NDGA groups at days 5, 7, 9, 11, 13, 15, 17, 19 and 21. Tumor growth curves showing NDGA (5 mg/kg/day, oral gavage) suppressed volume expansion from day 9 onward. *P < 0.05, **P < 0.01, ***P < 0.001

NDGA suppressed bladder cancer metastasis by targeting ALG3-mediated pathways

Our investigation revealed NDGA's potent anti-metastatic effects in bladder cancer models through ALG3 inhibition. Transwell migration and invasion assays demonstrated that NDGA treatment significantly reduced cell motility in ALG3-overexpressing T24 and UMUC3 cells (Fig. 11A, B) compared to controls. The anti-angiogenic potential of NDGA was confirmed through tube formation assays, with reduced tube length, mesh and junction numbers (Fig. 11C). In vivo validation using a tail vain metastasis model showed that while ALG3 overexpression increased lung metastatic nodules, NDGA treatment reduced metastasis at 7 weeks post-injetion (Fig. 11D, E). Immunohistochemical analysis of tumor tissues revealed NDGA’s ability to reverse ALG3-mediated EMT and VM markers alterations: increased E-cadherin and decreased N-cadherin expression (Fig. 11F), along with reduced VE-cadherin and MMP14 levels (Fig. 11G). These evidences suggested that NDGA effectively counteracted ALG3-driven metastatic processes through inhibition of cancer cell migration and invasion, suppression of VM formation, and reversal of EMT and VM markers expression. The consistent anti-metastatic effects across cellular and animal models positioned NDGA as a promising therapeutic candidate for advanced bladder cancer.

Fig. 11.

NDGA suppressed ALG3-drove metastasis in bladder cancer. A, B Transwell assays assessing cell migration and invasion capabilities in T24 and UMUC3 cells after ALG3 overexpression, or combined treatment with NDGA. Scale bars = 50 μm. C Tube formation assays evaluating vasculogenic mimicry formation ability of bladder cancer cells treated with ALG3 cDNA or combined with NDGA, assessing with total tube length, number of meshes and junctions. Scale bars = 50 μm. D Representative lung tissues from nude mice (n = 6/group) injected via tail vein with 1 × 10⁶ oeNC or oeALG3 cells, indicating NDGA treatment (5 mg/kg/day, oral gavage) reduced lung colonization. E H&E staining visualizing metastatic foci in oeNC, oeALG3 and oeALG3 + NDGA groups, showing NDGA reduced metastasis incidence. Scale bars = 20μm. F, G Immunohistochemical staining of EMT markers (E-cadherin, N-cadherin) and VM markers (VE-cadherin, MMP14) in tumor tissues collected from nude mice model, corroborating anti-metastatic mechanism. Scale bars = 20 μm. *P < 0.05, **P < 0.01, ***P < 0.001

Discussion

Despite significant advancements in cancer research, the efficacy of current therapeutic strategies remains suboptimal, highlighting the urgent need for reliable biomarkers. Aberrant glycosylation patterns, which emerge early in tumorigenesis [29], have emerged as promising diagnostic targets, with glycosyltransferases like ALG3 showing particular potential due to their dysregulation in various malignancies [30, 31]. ALG3 has demonstrated diagnostic and prognostic value across various cancers, serving as a biomarker in lung adenocarcinoma [32], showing negative correlation with survival outcomes in hepatocellular carcinoma [33], and functioning as both biomarker and therapeutic target in ovarian cancer [23]. In cancers of the urinary system, urine offers a non-invasive medium for disease monitoring. Our study revealed consistently elevated ALG3 levels in both urine and serum of bladder cancer patients, with urinary ALG3 showing higher specificity but lower sensitivity than serum, suggesting that reliance on urine alone could lead to missed diagnosis. To overcome this limitation, we implemented a combined urine-serum detection approach, which significantly enhanced both the sensitivity (75.0%) and specificity (90.9%) of ALG3-based diagnosis. These findings established ALG3 as a robust glycobiomarker for bladder cancer and demonstrated how targeting tumor-associated glycosylation anomalies could refine diagnostic accuracy in clinical practice. By combining urine and serum detection, we achieved superior diagnostic performance (75.0% sensitivity, 90.9% specificity) compared to FDA-approved urinary biomarkers like BTA stat (40–72% sensitivity, 29–96% specificity), BTA TRAK (~ 70% sensitivity, ~ 80% specificity), and NMP22 (11–85.7% sensitivity, 77–100% specificity) [34]. Unlike these protein-abundance markers, ALG3 uniquely reflected glycosylation-driven metastatic pathways while also being enzymatically targetable, offering both diagnostic and therapeutic potential. While ALG3 required serum supplementation for optimal sensitivity and awaited multicenter validation, its mechanistic specificity and balanced performance profile positioned it as a clinically valuable glycobiomarker that could significantly improve bladder cancer diagnosis and management.

ALG3, an α-1,3 mannosyltransferase, is implicated in diverse human pathologies, including cancer [35], congenital disorders [36], osteoarthritis [37], and leukemia [38], etc. Our prior studies identified ALG3 as an oncogenic driver in bladder cancer, where its overexpression significantly enhanced cell proliferation, migration and invasion. Notably, its pro-tumorigenic effects have been further documented in breast cancer, lung cancer, hepatocellular carcinoma, and oral squamous cell carcinoma. Beyond distinguishing bladder cancer from healthy individuals, we discovered that ALG3-high tissue stem cells were positively associated with epithelial to mesenchymal transition and angiogenesis, underscoring ALG3’s pivotal role in bladder cancer metastasis. These findings prompted us to investigate the mechanistic link between ALG3 and cancer metastasis. Aberrant N-glycosylation, a recognized hallmark of malignancy [14], involves the dynamic addition, trimming and modification of N-glycans to generate glycoproteins with distinct structures and functions, ultimately driving oncogenesis. For instance, ALG3-mediated glycosylation of TGFBR2 promoted radioresistance and cancer stemness in breast cancer [39], while ALG3-catalyzed α-1,3 mannosylation of uPAR facilitated ovarian cancer metastasis via the ADAM8/Ras/ERK axis [23]. In our study, immunoprecipitation and lectin pull-down assays identified LRFN4 as a novel ALG3 substrate. LRFN4, a highly glycosylated glycoprotein upregulated in gastric cancer [40] and linked to poor prognosis in lung adenocarcinoma [41], has emerging but poorly characterized roles in cancer. Strikingly, genetic ablation of N-glycosylation sites in LRFN4 (via glycosite-mutated plasmids) suppressed its pro-proliferative and pro-metastatic functions, confirming the dependency of LRFN4 activity on glycosylation. This aligns with established paradigms of glycosylation-dependent oncoprotein regulation. For example, SEMA7A’s tumor-promoting effects in head and neck squamous cell carcinoma required FUT8-mediated N-glycosylation [42]. STT3A/B-dependent glycosylation of CD24 sustained breast cancer tumorigenesis, while its inhibition sensitized cells to paclitaxel-induced ferroptosis [43]. N-glycosylation-deficient IL6 exhibited attenuated STAT3 activation, exacerbating EMT and chemoresistance in lung cancer [44]. Collectively, our data demonstrated that ALG3-mediated N-glycosylation of LRFN4 was essential for suppressing metastatic progression in bladder cancer, reinforcing the broader therapeutic relevance of glycosylation pathways in oncology.

Virtual screening is a powerful computational tool for identifying bioactive compounds with favorable drug-like properties from large chemical libraries. Given that ALG3 was upregulated in multiple cancers and served as a potential diagnostic, prognostic and therapeutic target, its inhibition represented a promising anti-cancer strategy. Traditional Chinese Medicine (TCM) monomers, derived from medicinal herbs with unique pharmacological profiles, have garnered increasing interest as perspective anti-tumor agents due to their favorable safety profile, reduced side effects, and sustained therapeutic efficacy [45–47]. Compared to conventional therapies, TCM monomers offer additional advantages, including lower toxicity, potential immune-modulatory effects, and adaptability to personalized treatment regimens [48, 49]. In this study, we employed structure-based virtual screening to identify ALG3-targeting TCM monomers from a natural compound library. Among the candidates, six compounds exhibited promising binding characteristics, with nordihydroguaiaretic acid (NDGA) displaying the highest binding affinity and lowest free energy. NDGA, a polyphenolic compound isolated from Larrea tridentate, features an ortho-dihydroxy structure with four hydroxyl groups. Beyond its well-documented anti-inflammatory and antioxidant properties, utilized in treating rheumatism [50], Parkinson’s disease [51] and SARS-CoV-2 infection [52], NDGA has demonstrated broad-spectrum anti-tumor activity across multiple cancers [53, 54]. Notably, NDGA attenuated cisplatin-induced nephrotoxity [55] while suppressing breast cancer growth through mTORC1 inhibition [56]. NDGA reduced viability in various lung cancer cell lines (H1975, H358, Calu-1, A549, SKLU-1 and H2228), with IC₅₀ values of 15-45 μM and near-complete cytotoxicity at > 100 μM [57]. NDGA induced anoikis-like apoptosis in pancreatic and cervical cancers by disrupting actin cytoskeletal dynamics and activating stress kinases [58]. Expanding upon these findings, our study revealed NDGA’s efficacy in bladder cancer, where it suppressed tumorigenesis and metastasis by downregulating ALG3. These results highlighted NDGA’s potential as a novel therapeutic agent for ALG3-driven malignancies, offering a promising avenue for bladder cancer treatment.

While this study provided valuable insights, several limitations should be acknowledged. The relatively small sample size of serum and urine specimens may limit the statistical power for subgroup analysis across different age groups, genders, tumor differentiation grades, or infiltration patterns, prompting our ongoing efforts to expand sample collection, and ALG3 quantification to enhance its diagnostic utility. Although we have identified ALG3-mediated glycosylation of LRFN4 as a key regulatory mechanism, the combined effects of ALG3 and LRFN4 in bladder cancer initiation and metastasis require further elucidation through comprehensive functional assays, xenograft mouse models, and systems biology approaches.

Conclusion

Recent advances in glycobiology have highlighted glycosylation-targeting strategies as promising therapeutic approaches against cancer progression and metastasis [59–61]. Our study established ALG3 as a critical oncogenic driver in bladder cancer, mechanistically linking its pro-metastatic function to LRFN4 N-glycosylation-dependent activation of the Ras signaling pathway. Furthermore, we identified NDGA as a potent TCM monomers that suppressed bladder cancer progression through ALG3 downregulation (Fig. 12). Collectively, these findings unveiled the NDGA/ALG3/LRFN4-glycosylation axis as a novel therapeutic paradigm, offering both mechanistic insights and translational for treating ALG3-driven malignancies.

Fig. 12.

Mechanistic model of NDGA targeting ALG3-mediated LRFN4 N-glycosylation to suppress bladder cancer metastasis

Abbreviations

TCM

Tranditional Chinese Medicine

NDGA

Nordihydroguaiaretic acid

BCa

Bladder cancer

NMIBC

Non-muscle invasive bladder cancer

MIBC

Muscle invasive bladder cancer

EMT

Epithelial to mesenchymal transition

VM

Vasculogenic mimicry

Asn

Asparagine

ELISA

Enzyme-linked immunosorbent assay

FBS

Fetal bovine serum

qRT-PCR

Quantitative real-time polymerase chain reaction

HRP

Horseradish peroxide

CCK-8

Cell Counting Kit-8

OD

Optical density

DEGs

Differentially expressed genes

GO

Gene Ontology

KEGG

Kyoto Encyclopedia of Genes and Genomes

WGCNA

Weight Gene Co-expression Network Analysis

GNA

Galanthus Nivalis Agglutinin

RMSD

Root mean square deviation

RMSF

Root mean square fluctuation

Author contributions

MLL, QZ and SJL: Conceptualization, project administration, supervision, writing—original draft, writing—review and editing, Funding acquisition. JYZ and TYZ: Methodology, data curation, formal analysis. All authors read and approved the final manuscript.

Funding

All authors are grateful to the support from the Life and Health Guidance Program Project of Dalian (2023).

Availability of data and materials

Data will be made available on request.

Declarations

Ethics approval and consent to participate

This study was approved by the Ethics Committee of the First Affiliated Hospital of Dalian Medical University (No. PJ-KS-KY-2022-439), and the written informed consent was obtained from all individuals. Animal experiments were performed and approval by the Animal Ethics Committee of Dalian Medical University (No. AEE23158).

Consent for publication

Not applicable.

Competing interests

The authors declare no conflict of interests.