MaterialsCell lines

Mouse normal mammary epithelial cell line HC11 (CRL-3062), and mouse triple-negative breast cancer cell lines 4 T1 (CRL-2539) and EMT6 (CRL-2755) were obtained from ATCC (USA). Reagents and chemicals: Dulbecco’s Modified Eagle Medium (DMEM, D0822), RPMI-1640 medium (R4130), fetal bovine serum (FBS, F8687), antibiotic–antimycotic solution (HY-K1058), red blood cell lysis buffer (11814389001), and disodium ethylenediaminetetraacetate (EDTA-Na₂, E5134) were purchased from Sigma (USA). Antibodies and separation kits: CD8+ T cell magnetic sorting kit (130-049-401, Miltenyi, Germany), CD3 (ab34722, Abcam, 1:10), CD8 (ab313760, Abcam, 1:2500). Chemical reagents: Nanocarrier components: CHEMS-PEG2000-COOH (Seebio, China), dicyclohexylcarbodiimide (DCC, D80002), dimethylformamide (DMF, 227056), GNGRG peptide (ChinaPeptides, Shanghai), cholesteryl hemisuccinate (CHEMS, 850524P), phosphatidylethanolamine (PE, 840021 C), cholesterol (C8667), Tween 80 (655206), Coumarin-6 (442631), and IR-780 iodide (425311) were provided by Sigma (USA). Experimental reagents: Trypsin (T4799), Polybrene (TR-1003), G418 (A1720), 4% Paraformaldehyde (441244), SA-β-gal Staining Kit (CS0030), Protein Extraction Kit (BB3101, Bestbio), BCA Kit (P0012, Beyotime), Polyvinylidene fluoride (PVDF) (IPVH85R, Millipore), Trizol reagent (15596026, Invitrogen), MGIEasy rRNA Removal Kit (1000005953, MGI), PrimeScript RT Kit (RR047 A, Takara), SYBR Green qPCR Master Mix (RR420 A, Takara), CCK-8 (96992), Xylene (247642), Hematoxylin Solution (H3136), TUNEL-POD Apoptosis Detection Kit (11772465001, Roche). Other materials: Dialysis membrane (MWCO 3.5 kDa, F132592-0001, Sangon Biotech, Shanghai), TLR9 inhibitor AT791 (MedChemExpress, USA), Cur (C1386, Sigma, USA).

Data retrieval and compilation

Transcriptional gene expression data and corresponding clinical information of breast cancer patients (TCGA-BRCA) were retrieved from The Cancer Genome Atlas (TCGA) database (https://www.cancer.gov). Transcriptomic data were obtained in FPKM (fragments per kilobase of transcript per million mapped reads) format. As the dataset is publicly available, no ethical approval or patient consent was required.

Prediction and retrieval of cur target sites

Canonical SMILES identification for Cur was obtained from the PubChem database and used as input in the SwissTargetPrediction database (http://swisstargetprediction.ch), with “Homo sapiens” selected as the reference species. Predicted targets with probability values greater than zero, indicating a significant likelihood between the compound and the targets. These selected target sites were further converted to corresponding genes via the Uniprot database (http://www.uniprot.org/).

Selection of potential therapeutic targets for Cur

The GeneCards database (https://www.genecards.org/) (Stelzer et al. 2016) was queried using the keywords “triple-negative breast cancer” and “immune.” The top 1000 genes ranked by Relevance score were retrieved as TNBC or immune-related genes, and Cur candidate target genes were visualized through a Venn diagram created using the Venny 2.1 online tool. The genes in the intersection were considered as potential therapeutic targets for reversing immunotherapy resistance in TNBC with Cur.

Transcriptome data analysis of TCGA database

Differential expression analysis was carried out on the TCGA-BRCA utilizing the “DESeq2” package in R. Genes with |log2 FC| > 0.5 and a P-value below 0.001 were considered as significantly differentially expressed, and a volcano plot was developed utilizing the “pheatmap” package. The intersection of differentially expressed genes (DEGs) with candidate targets was defined as key targets for Cur treatment of TNBC. Gene expression data from matched tumor and adjacent normal tissues of the same patients in the TCGA dataset were aligned utilizing patient IDs and analyzed via the “limma” package to validate differential expression.

Gene expression profiles were integrated with survival data utilizing a custom Perl script. The “limma” package in R was utilized to retrieve transcript abundance, while survival outcomes were analyzed via the “survival” package, using overall survival (OS) as the clinical endpoint. Based on gene expression levels, patients were stratified into high- and low-expression cohorts. Kaplan–Meier survival curves were plotted. A P-value of less than 0.05 indicated a significant difference, implicating a prognostic association between gene expression and patient outcomes.

Functional enrichment analysis

GO and KEGG functional enrichment analyses were applied to the core targets in protein–protein interaction (PPI) network associated with the potential therapeutic efficacy of Cur to clarify the principal biological functions modulated by these targets. The enrichment analysis was performed utilizing the “ClusterProfiler” package in R (Stelzer et al. 2016). GO analysis included three components: Biological Process (BP), Cellular Component (CC), and Molecular Function (MF), and a P-value less than 0.05 was considered indicative of statistical significance. KEGG enrichment was conducted under the same threshold. Pathway maps reflecting gene expression trends were developed utilizing the “pathview” package in R. Based on the key pathways enriched by the core targets in the PPI network, in conjunction with relevant literature, potential mechanisms of Cur in reversing TNBC immunotherapy resistance were hypothesized.

Construction of pharmacophore model

The two-dimensional and three-dimensional chemical structure diagrams of Cur (Compound CID: 969516) were retrieved from the PubChem database and saved in “SDF” format. The three-dimensional structure was imported to the PharmMapper database (http://lilab-ecust.cn/pharmmapper/) to simulate molecular-target protein docking. In PharmMapper, the target range was set to “Druggable Pharmacophore Models (v2017, 16159)”, with the number of matched targets limited to 1000, while other parameters were kept at default settings.

Molecular simulation docking

Crystal structures of candidate Cur targets—TLR9 (PDB ID: 8 AR3), EGFR (PDB ID: 2EB2), AKT1 (PDB ID: 3O96), and SERPINE1 (PDB ID: 3 CVM)— were retrieved from the Protein Data Bank (https://www.rcsb.org). The molecular structure of Cur was converted into a three-dimensional format utilizing Chem3D Ultra 14.0 and energy-minimized with the MM2 algorithm. Receptor proteins were dehydrated, and the organic compounds were removed using PyMOL software. Hydrogen atoms were added, and partial charges were calculated utilizing AutoDockTools 1.5.6. Both ligands and receptors were converted into PDBQT format, and appropriate binding sites and grid parameters were defined. Molecular docking was carried out using AutoDock Vina 1.5.6 to calculate binding free energies. Docking conformations were visualized for structural analysis.

Immunoinfiltration analysis

Immunoinfiltration analysis was performed utilizing the TCGA-BRCA dataset with the CIBERSORT algorithm in R. This gene expression-based deconvolution algorithm estimated the relative proportions of immune cell types within tumor tissues. Only samples with P-values < 0.05 were included to ensure result reliability. Immune cell composition was visualized utilizing the “Vioplot” and “Pheatmap” while comparing the differences in immune cell abundance between normal and tumor samples to validate the significance of key immune cells in tumor development. The association between CD8+ T cell infiltration and TLR9 expression levels was assessed across all samples through correlation analysis.

Metabolomics data downloading and preprocessing

The breast cancer-related metabolomics dataset ST000356 was obtained from the Metabolomics Workbench. The metabolite composition of the samples was visualized and analyzed using the “built-in statistical” module (Waghu et al. 2015). Data were then submitted to MetaboAnalyst (https://www.metaboanalyst.ca/). After data verification, missing value handling, and selection, the metabolomics data underwent calibration and normalization processes. Finally, the preprocessed results were visualized using the “View results” module.

Multivariate statistical analysis and metabolic pathway analysis of metabolomics data

The preprocessed data was further subjected to multivariate analysis using the MetaboAnalyst platform. Orthogonal partial least squares discriminant analysis (OPLS-DA) was applied, and 100 permutation tests were conducted to prevent overfitting. Metabolites with variable importance in projection (VIP) scores > 1 were identified as differential metabolites. Based on univariate analysis, metabolites with |log₂ fold change| > 2 and a false discovery rate (FDR) < 0.05 were selected as significant. These differential metabolites were further analyzed using KEGG pathway enrichment via the functional analysis module. Integrating the aforementioned analysis results with the metabolic pathway analysis and relevant literature, the role of glycolysis in overcoming immunotherapy resistance in Cur was conclusively established.

Animal experimentation and ethical statement

A total of 36 female BALB/c-nu nude mice (6–8 weeks old, 18–22 g) and 3 healthy female BALB/c mice of similar age and weight were purchased from Changzhou Cavens Experimental Animals Co., Ltd. All animals were maintained under specific pathogen-free (SPF) conditions, housed at 23 ± 1 °C with 55 ± 5% humidity, and exposed to a 12-hour light/dark cycle. All animal experiments were granted by the Institutional Animal Ethics Committee and conducted in accordance with institutional protocols and internationally recognized standards for animal welfare. Measures were taken to alleviate discomfort and minimize suffering, while ensuring a reduction in the number of animals used whenever possible. All animals received appropriate care throughout the study, and humane endpoints were implemented to ensure ethical treatment after the experiments.

Isolation and identification of CD8+ T cells

Single-cell suspensions were generated from spleens of three BALB/c mice by mechanical dissociation through a 70 μm mesh in RPMI-1640 medium (R4130, Sigma, USA) containing 10% FBS (F8687, Sigma, USA). After centrifugation, red blood cells were lysed with 5 ml lysis buffer (11814389001, Sigma, USA) for 5 minutes and neutralized with an equal volume of RPMI-1640 containing 10% FBS, followed by centrifugation at 500 g for 5 minutes at 4 °C. Cell pellets were washed with PBS-EDTA (PBE) buffer (200 ml PBS, 0.148896 g EDTA-Na₂ [E5134, Sigma, USA], 2 ml FBS), centrifuged again, and resuspended in 2 ml PBE to obtain the final single-cell suspension. For tumor-derived samples in the orthotopic model, digestion buffer (abs9522, Absin, China) was added prior to tissue dissociation, followed by 1-hour incubation at 37 °C, subsequent steps remained consistent.

CD8+ T cells were isolated utilizing a magnetic sorting kit (130-049-401, Miltenyi, Germany). Following cell concentration adjustment, immunomagnetic beads were added and incubated at 4 °C for 24 minutes with intermittent mixing. After washing with 10 ml PBE, cells were resuspended in 500 μl PBE. LS columns were equilibrated on a VarioMACS separator and washed with PBE before and after sample loading. Flow-through was collected, and columns were flushed to recover the isolated CD8+ T cells. Cells were maintained in RPMI-1640 medium at 37 °C with 5% CO₂.

For purity assessment, approximately 1×106 CD8+ T cells were stained with anti-CD3 (ab34722, 1:10, Abcam) and anti-CD8 (ab313760, 1:2500, Abcam) antibodies at 4 °C for 30 minutes, washed with PBS, and analyzed by flow cytometry (FCM, Accuri C6, BD).

Cell culture

Normal mouse mammary epithelial cells HC11 (CRL-3062, ATCC, USA), mouse TNBC cells 4 T1 (CRL-2539, ATCC, USA), and EMT6 (CRL-2755, ATCC, USA) were procured from the American Type Culture Collection (ATCC). Cells were cultured in DMEM (D0822, Sigma, USA) enriched with 10% FBS and 1% antibiotic-antimycotic solution (HY-K1058, MCE, USA) and maintained in a 37 °C, 5% CO2 humidified incubator.

Synthesis of NGR-NVs@Cur

The synthetic route of CHEMS-PEG2000-NGR is shown in Figure S1. Briefly, 10 mmol of CHEMS-PEG2000-COOH (Seebio, Shanghai, China), 10 mmol of DCC (D80002, Sigma, USA), and a catalytic amount of DMAP (107700, Sigma, USA) were dissolved in DMF (227056, Sigma, USA). Then, 12 mmol of GNGRG peptide (ChinaPeptides, Shanghai, China) was added to the solution and stirred under nitrogen at ambient temperature for 24 hours. The reaction mixture was dialyzed (MWCO 3.5 kDa; F132592-0001, Sangon Biotech, China) against deionized water for 48 hours to remove unreacted peptides and by-products. The dialyzed sample was freeze-dried and stored at –20 °C. The conjugation product CHEMS-PEG2000-NGR was characterized by 1H NMR (Figure S2). In the spectrum, δ = 2.5 ppm corresponds to the DMSO solvent peak; δ = 3.50–3.88 ppm represents the methylene (-CH₂-) protons of PEG; δ = 5.3, 4.5, and 0.65–2.30 ppm are characteristic peaks of CHEMS; and δ = 3.5–4.5 and 1.4–2.7 ppm correspond to the side-chain protons of amino acids in the NGR sequence.

$$Coupling\ rate\%=\left(\frac}}/\frac}}\right)\times 100\%$$

The coupling efficiency of NGR was calculated by comparing the integrated peak areas of NGR side-chain protons (δ = 1.4–2.7 ppm) and PEG methylene protons (δ = 3.50–3.88 ppm), yielding a conjugation rate of 86.45%.

A schematic diagram of the synthetic route for NGR-NVs@Cur is shown in Figure S3. Briefly, Curcumin (Cur, C1386, Sigma, St. Louis, MO, USA), L-α-phosphatidylethanolamine (PE, 840021 C, Sigma, St. Louis, MO, USA), cholesterol (C8667, Sigma, St. Louis, MO, USA), and CHEMS (850524P, Sigma, St. Louis, MO, USA) were combined in chloroform in a molar proportion of 6:35:15:15, together with CHEMS-PEG2000-NGR (at 8% of the PE molar content) (Gu et al. 2017). The mixture was dried using a rotary evaporator at 20 °C and then placed in a vacuum desiccator overnight. The resulting lipid film was hydrated in 20 mL of PBS buffer at 60 °C for 1 hour to obtain the NGR-NVs@Cur suspension.

During synthesis, coumarin-6 (C6, 442631, Sigma, St. Louis, MO, USA) was used in place of Cur as a fluorescent probe to obtain NGR-NVs@C6. When Cur was omitted (for mechanistic in vitro experiments) or replaced with IR-780 (425311, Sigma, St. Louis, MO, USA) (for in vivo targeting evaluation of NGR-NVs), yielding NGR-NVs. When both IR-780 was used in place of Cur and CHEMS-PEG2000 was substituted for CHEMS-PEG2000-NGR in the formulation, the resulting product was termed NVs.

Characterization and performance evaluation of NGR-NVs@Cur

Vesicle morphology was visualized utilizing transmission electron microscopy (TEM, JEM-2100 F, JEOL, Japan) after negative staining with phosphotungstic acid (Cat# 76960, Sigma, St. Louis, MO, USA). Particle size and zeta potential were analyzed at 25 °C via dynamic light scattering (DLS, Zetasizer Nano, Malvern Panalytical, Malvern, UK).

Encapsulation efficiency (EE) was evaluated by diluting 200 μL of vesicle suspension with 1.8 mL PBS (pH 7.4) containing 0.5% Tween-80. The mixture was centrifuged at 14,000 rpm for 10 minutes. The amount of encapsulated Cur was quantified by high-performance liquid chromatography (HPLC), and EE was calculated as: EE% = (W_total − W_free)/W_total × 100% The drug loading efficiency of the liposomes was determined to be 46.23% (Gu et al. 2017).

For in vitro release studies, NGR-NVs@Cur were suspended in PBS and sealed into pre-treated dialysis bags (molecular weight cut-off, MWCO = 3.5 kDa) and immersed in PBS (pH 5.0 and 7.4) containing 1% Tween-80. Samples were incubated at 37 °C in a shaking water bath at 100 rpm. At predetermined time points (0, 2, 4, 8, 12, 24, 48, and 72 h), aliquots were collected from the release medium and replaced with fresh buffer. Cumulative release was measured by HPLC (detection conditions provided separately), and release profiles were analyzed.

4 T1 cells (1×105/well) were plated in 6-well plates and cultured for 24 hours. Cells were treated with 100 ng/mL free coumarin-6 or NGR-NVs@C6 for 4 hours. Intracellular fluorescence was observed by confocal microscopy. Cells were harvested using trypsin (Cat# T4799, Sigma, St. Louis, MO, USA), and fluorescence intensity was measured by FCM (Accuri C6, BD) using a 467 nm argon laser (Gu et al. 2017).

Cell treatment and co-culture

To explore the role of TNBC cells in inducing CD8+ T cell senescence, co-cultures were established with HC11 group (untreated HC11 cells), 4 T1 group (untreated 4 T1 cells), and EMT6 group (untreated EMT6 cells). To study the impact of Cur on the TLR9 and mTOR pathways in TNBC cells, TNBC cells were co-cultured with CD8+ T cells in three groups: CON (treated with PBS), Cur (treated with Cur), and Cur + AT791 (treated with Cur and TLR9 inhibitor AT791). For assessing the efficacy of different drug treatments, co-cultures were assigned to four groups: CON (PBS), NGR-NVs (TNBC cells treated with NGR-NVs), Cur (TNBC cells treated with Cur), and NGR-NVs@Cur (TNBC cells treated with NGR-NVs@Cur). Experimental groupings are summarized in Table 1.

Table 1 Cell Experimental Groups

For drug treatment, 3×105 4 T1 or EMT6 cells were seeded in 12-well plates and exposed to one of the following: Cur (60 μmol/L) (Calibasi-Kocal et al. 2019; Guneydas and Topcul 2022); addition of an equivalent dose of NGR-NVs@Cur; simultaneous addition of Cur and TLR9 inhibitor AT791 at a final concentration of 0.04 μmol/L (sourced from MedChemExpress, USA); or addition of an equal amount of NGR-NVs or PBS. Subsequently, the cells were incubated for another 12 hours.

For co-culture, 12-well plates were pre-coated with 500 μL of anti-CD3 solution (2 μg/mL) and incubated overnight at 4 °C. The next day, the solution was removed, and 3×106 isolated CD8+ T cells were added for 24-hour activation at 37 °C. Separately, 3×105 4 T1 or EMT6 cells were seeded in fresh plates with 1 mL RPMI-1640 medium. After adherence, activated CD8+ T cells were added at a 1:1 ratio along with fresh medium, and co-culture was maintained for 24 hours. Supernatants were collected and centrifuged to remove residual medium, and CD8+ T cells were re-cultured for 3 additional days prior to SA-β-gal staining and FCM.

Cell transfection

To investigate the impact of Cur activation of TLR9 on the mTOR pathway, glycolysis, and T cell senescence, 4 T1 cells were treated with sh-TLR9, sh-AKT, oe-AKT, and sh-GLUT1. This treatment resulted in either knockdown or overexpression of TLR9, the upstream mTOR factor AKT, or the crucial sugar transporter protein GLUT1 in the treated cells. For the experiments, RFP-tagged sh-TLR9-1, sh-TLR9-2, sh-AKT-1, sh-AKT-2, sh-GLUT1-1, sh-GLUT1-2, and oe-AKT lentivirus (virus titer 108 TU) were constructed based on the pGL3-Basic plasmid. Additionally, negative controls were RFP-tagged sh-NC and oe-NC lentivirus (virus titer 108 TU). The lentiviruses and sequences were synthesized by Suzhou Jikai Gene Co., Ltd., with the shRNA sequences detailed in Table S1.

To verify transfection efficiency, 4 T1 cells were transduced with lentiviral vectors targeting sh-TLR9-1, sh-TLR9-2, sh-AKT-1, sh-AKT-2, sh-GLUT1-1, sh-GLUT1-2, oe-AKT, sh-NC, and oe-NC to establish gene knockdown or overexpression models. The more effective shRNA sequences were selected for subsequent experiments. Based on the experimental design, the transfected 4 T1 cells, alone or co-cultured with CD8+ T cells, were divided into groups receiving either PBS or Curcumin (Cur) treatment (e.g., Cur + sh-TLR9 group, Cur + sh-TLR9 + sh-AKT group, Cur + oe-AKT + sh-GLUT1 group, etc.). Details of all experimental groups are listed in Table 1. PBS and Cur treatment protocols were consistent with previous descriptions.

Briefly, 4 T1 cells were plated at 1×105 cells/well in 96-well plates. When confluence reached approximately 50%, cells were infected with lentivirus at a multiplicity of infection (MOI) of 5 in the presence of 8 μg/mL Polybrene (TR-1003, Sigma, USA). Selection was performed using 5 μg/mL G418 (A1720, Sigma, USA), and after 24 hours, the medium was replaced. Transfection efficiency was assessed 3–4 days post-infection by observing RFP fluorescence under a fluorescence microscope. The expression levels of the corresponding mRNA and protein were measured by reverse transcription quantitative polymerase chain reaction (RT-qPCR) and Western blot. The transfected 4 T1 cells were subsequently co-cultured with CD8+ T cells treated with Cur or PBS according to the respective protocols.

SA-β-gal staining

CD8+ T cells were harvested, rinsed twice with PBS, and transferred to glass slides and air-dried, followed by fixation with 50 μL of 4% paraformaldehyde (441244, Sigma, USA) for 3–5 minutes. After removing the paraformaldehyde, cells were rinsed twice with PBS and dried again. Subsequently, 100 μL of pre-prepared SA-β-gal staining solution (CS0030, Sigma, USA) was added, and slides were incubated at 37 °C in the dark for 24 hours. After gentle washing with water, slides were mounted, and SA-β-gal-positive cells were visualized under a microscope. The proportion of positive cells was quantified. Each experiment was performed in triplicate.

Flow cytometry (FCM)

After counting, 1–2×106 cells were aliquoted into flow tubes and washed once with staining buffer (PBS enriched with 1% FBS and 1 mM EDTA). After discarding the supernatant, cells were gently resuspended in the residual volume (~20–25 μL) by tapping the tube. Fluorescent antibody cocktails were prepared according to the number of tubes (see Table S2 for details), with 5 μL staining buffer added to each to facilitate mixing. The appropriate volume of antibody cocktail was introduced to each tube, gently agitated, and maintained on ice in light-protected conditions for 30 minutes. After incubation, 2 mL staining buffer was added, and cells were vortexed and centrifuged at 350 g for 6 minutes. The supernatant was discarded, and cells were gently resuspended. The final volume was adjusted to 250–300 μL, filtered through a 300-mesh strainer, and subjected to FCM analysis.

Western blot (WB)

Total protein was extracted from target cells utilizing a commercial lysis kit (Bestbio, BB3101, China), and quantified via BCA assay (P0012, Biyuntian, Shanghai, China). Equal protein amounts were resolved by sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE) and transferred to PVDF membranes (IPVH85R, Millipore, Germany). Membranes were blocked in 5% BSA at ambient temperature for 1 h, followed by immunoblotting with the corresponding primary and secondary antibodies. After washing three times with TBST (5 minutes each), proteins were detected utilizing a chemiluminescence imaging system. Band intensities were quantified utilizing ImageJ 1.48u (NIH, USA), and normalized to Glyceraldehyde-3-phosphate dehydrogenase (GAPDH). Antibody sources are listed in Table S2.

RT-qPCR

Total RNA was isolated utilizing a Trizol reagent (15596026, Invitrogen, Car, USA). Ribosomal RNA was removed utilizing the MGIEasy rRNA Removal Kit (1000005953, MGI, China). RNA concentration and purity were measured with the NanoDrop 2000 UV-Vis spectrophotometer (ThermoFisher, USA). cDNA was synthesized from purified RNA utilizing the PrimeScript RT Reagent Kit (RR047 A, Takara, Japan).

Quantitative PCR was conducted with SYBR Green Master Mix (RR420 A, Takara, Japan), with GAPDH as the internal control. Each sample was analyzed in triplicate, and relative gene expression was calculated utilizing the 2-ΔΔCt method, where ΔΔCT = ΔCt (treatment) - ΔCt (control), and ΔCt = Ct (target) − Ct (reference). The cycle threshold (Ct) represents the number of amplification cycles required for the fluorescence signal to reach the set threshold, corresponding to the exponential phase of amplification. Primer sequences were designed utilizing Primer-BLAST and are listed in Table S3.

Detection of cell proliferation using CCK-8 assay

Cells in the logarithmic growth phase were plated at 8×103 cells per well in a 96-well plate and incubated under standard conditions. Following 24, 48, and 72 hours of incubation, 10 μL of CCK-8 solution (96992, Sigma, USA) was introduced to each well. After 1 hour of incubation at 37 °C, absorbance was recorded at 450 nm utilizing a microplate reader (Epoch, Bio-Tek, USA). Cell viability was determined as: (OD experimental group - OD blank group)/(OD control group - OD blank group). The experiment was repeated thrice.

Wound healing assay

Cells were seeded at 5×105 cells per well in a 24-well plate for cultivation. After reaching confluence, a scratch was made utilizing a 200 μL pipette tip. Cells were washed with PBS and cultured in serum-free RPMI 1640 medium for 24 hours. Images were captured at 0 and 24 hours utilizing an inverted microscope (5× magnification), and wound closure was quantified utilizing ImageJ software. Experiments were conducted in triplicate.

Construction of subcutaneous tumor and adoptive T cell transfer mouse model

Thirty-six female BALB/c-nu nude mice (6–8 weeks old, 18–22 g) were housed under SPF)conditions. A single-cell suspension of 4 T1 cells (2×106 cells in 100 μL PBS) was prepared and injected subcutaneously into the right dorsal region of each mouse under anesthesia. On day 6, tumor-bearing mice were randomly assigned into six groups (n = 6/group): NVs, NGR-NVs, CON, J43, NGR-NVs@Cur, and J43 + NGR-NVs@Cur.

Mice in the NVs and NGR-NVs groups received 200 μL of the respective formulation via tail vein injection. In vivo fluorescence imaging was conducted utilizing the IVIS® Spectrum system at 0.5 and 12 hours post-injection to assess tumor targeting. Mice were anesthetized for 10 minutes, placed in the imaging chamber, and scanned using filters set to excitation at 745 nm and emission at 800 nm. Signal intensity was recorded to evaluate biodistribution.

In the remaining four groups (CON, J43, NGR-NVs@Cur, and J43 + NGR-NVs@Cur), once the tumors reached 8×8 mm2 (day 6), each mouse received an intravenous injection of 5×105 CD8+ T cells via the tail vein to establish an adoptive T cell transfer mouse model. The extraction method of CD8+ T cells was as previously described, and the cells were pre-activated with CD3 and CD28 antibodies (ab243228, Abcam, UK) before injection. Starting from the 7 th day, every 3 days, each mouse received intravenous injections of 100 μL of saline, J43, NGR-NVs@Cur, or a simultaneous injection of J43 and NGR-NVs@Cur, each containing equivalent doses of 20 μg J43 (Kleinpeter et al. 2016) and 5 μg Cur (Xu et al. 2024). Tumor volume and survival status were recorded regularly. On the 40 th day, the mice were euthanized. Tumors and major organs (heart, liver, spleen, lungs, kidneys) were harvested for further analysis, including CD8+ T cell isolation, FCM, and H&E staining, following established protocols (Hu et al. 2020; Chang et al. 2012).

H&E staining

Mouse tissues were fixed in 4% paraformaldehyde, embedded in paraffin, and sectioned at 5 μm thickness. Sections were mounted on slides and dried at 45 °C. Slides were deparaffinized with xylene (247642, Sigma, USA), followed by graded ethanol washes and a final rinse with distilled water for 2 minutes. Slides were stained with hematoxylin (H3136, Sigma, USA) for 5 minutes, rinsed under running water, subjected to 1% hydrochloric acid ethanol differentiation for 3 seconds, and then stained with 5% eosin (861006, Sigma, USA) for approximately 2 minutes. After dehydration and clearing, slides were coverslipped and examined under a light microscope.

TUNEL staining

Terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) staining was performed utilizing the TUNEL-POD Apoptosis Detection Kit (11772465001, Roche, USA). Tumor sections were incubated with Proteinase K working solution at 37 °C for 15–30 minutes, followed by two PBS washes. The TUNEL reaction mixture (50 μL TdT + 450 μL fluorescent-labeled dUTP) was added, followed by incubation in a dark, moist chamber at 37 °C for 1 hour. After three PBS washes, 50 μL of Converter-POD was added and incubated at 37 °C for 30 minutes in the dark. Sections were then washed with PBS, incubated with 50 μL DAB substrate at ambient temperature for 10 minutes, and washed again. Hematoxylin counterstaining, tap water rinse, dehydration in graded ethanol, xylene clearing, and mounting with neutral gum were subsequently performed. Finally, apoptotic cells and the total cell count (200–500 cells in total) were observed and counted under a light microscope, with photographs taken to calculate the apoptosis index (AI) = apoptotic cells/total cells.

Statistical analysis

Descriptive statistics, including mean, median, standard deviation, and range, were calculated. Comparisons between two groups were conducted utilizing independent samples t-tests, while comparisons among three or more groups were assessed by one-way analysis of variance (ANOVA). When dealing with multiple variables or conditions, a multiple analysis of variance (MANOVA) was employed. Survival time data was analyzed utilizing Kaplan-Meier survival curves, with differences compared through Log-rank tests. Evaluation of relationships between continuous variables were assessed utilizing Pearson correlation coefficients. Linear or logistic regression analyses were employed in predictive or exploratory research scenarios. For non-normally distributed data or variance heterogeneity, non-parametric tests such as the Mann–Whitney U test or Kruskal–Wallis test were employed. All statistical analyses were conducted utilizing SPSS (IBM Corporation) or R. Statistical significance was considered when the p-value was less than 0.05. Graphs were generated utilizing GraphPad Prism software (GraphPad Software, Inc.).