miR-138-5p exhibits downregulated expression in pancreatic CAFs and can reprogram the phenotype of pancreatic CAFs

Analysis using the EPIC algorithm revealed that the CAF proportion in The Cancer Genome Atlas (TCGA) pancreatic cancer dataset was increased relative to that in the Gene-Tissue Expression (GTEx) healthy pancreas dataset (Fig. 1a). Primary CAFs and healthy fibroblasts (NFs) were cultured using the outgrowth method. CAFs and NFs exhibited a spindle shape. Immunofluorescence showed that CAFs had a stronger expression of ACTA2, FAP, and FSP than NFs (Fig. 1b, Supplementary Fig. 1a). The differentially expressed miRNAs were identified using the miRNA array of pancreatic CAFs and NFs (Fig. 1c). Among the differentially expressed miRNAs, miR-138-5p was found to have a great significant difference in expression between CAFs and NFs (Fig. 1d). Quantitative real-time polymerase chain reaction (qRT-PCR) proved that miR-138-5p in CAFs was downregulated when compared with that in paired NFs (Fig. 1e). Fluorescence in situ hybridization (FISH) analysis further illustrated that miR-138 expression in CAFs was lower than that in NFs (Fig. 1f). To facilitate subsequent in vitro and in vivo studies, an immortalized CAF cell line was established (Supplementary Fig. 1b–c). The function of miR-138-5p was investigated. Notably, the results demonstrated that miR-138-5p mimic transfection markedly upregulated miR-138-5p expression (Fig. 1g). In addition, the 5-bromo-2’-deoxyuridine (EdU) and CCK-8 assays demonstrated that miR-138-5p mimic decreased the proliferative rate of CAFs (Fig. 1h, Supplementary Fig. 1d). The findings from the wound-healing and transwell assays indicated that the migratory rate of CAFs was reduced by miR-138-5p mimic (Fig. 1i, Supplementary Fig. 1e). Further, the ACTA2, FAP, FSP, and collagen1 protein expression was downregulated by transfection with miR-138-5p, as determined by Western blotting analysis (Fig. 1j).

Fig. 1

Effect of miR-138-5p on pancreatic cancer-associated fibroblasts (CAFs). a Analysis of CAF proportion in The Cancer Genome Atlas (TCGA) pancreatic cancer dataset and Genotype-Tissue Expression (GTEx) healthy pancreas dataset using the EPIC algorithm. b Morphology of CAFs and healthy fibroblasts (NFs) and immunofluorescence analysis of ACTA2, FAP, and FSP (scale bar = 100 μm). c Identification of differentially expressed microRNAs (miRNAs) between CAFs and NFs using the miRNA array. d miR-138-5p had the highest significant difference in expression between CAFs and NFs. e Quantitative real-time polymerase chain reaction (qRT-PCR) validation of miR-138-5p downregulation in CAFs relative to paired NFs. Data are presented as mean (± SD); n = 3 per group. f miR-138 expression in CAFs and NFs as determined by Fluorescent in situ hybridization (FISH) analysis (scale bar = 100 μm). g qRT-PCR validation of miR-138-5p mimic transfection efficiency. Data are presented as mean (± SD); n = 3 per group. h EdU assay of CAFs transfected with miR-138-5p mimic (scale bar = 100 μm). Data indicate the mean (± SD); n = 3 per group. i The wound-healing assay results of CAFs transfected with miR-138-5p mimic (scale bar = 100 μm). Data are presented as mean (± SD); n = 3 per group. j Western blotting analysis of the impact of miR-138-5p mimic transfection on the ACTA2, FAP, FSP, and collagen1 protein expression levels (scale bar = 100 μm). Data represent the mean (± SD); n = 3 per group

Preparation and characterization of IEVs-PFD/138

We constructed engineered BMSCs-derived EVs for pancreatic CAFs-targeted co-delivery of the gene-drug miR-138-5p mimic and anti-fibrosis drug pirfenidone (PFD) (Fig. 2a). The primary BMSCs were infected with human telomerase reverse transcriptase (hTERT)-encoding lentivirus to construct immortalized BMSCs. The immortalized BMSCs exhibited a spindle-shaped arrangement of radial concentric circles with broom-like growth (Supplementary Fig. 2a). Immortalized BMSCs were identified based on osteogenic, adipogenic, and chondrogenic induction (Supplementary Fig. 2a). Flow cytometry showed that immortalized BMSCs negatively expressed CD45, CD14, CD11B, CD34, and HLA, but positively expressed CD29 and CD73 (Supplementary Fig. 2b). Ultracentrifugation was used to separate the EVs, and Western blotting, transmission electron microscopy (TEM), and nanoparticle tracking analysis were used to characterize them. A characteristic round morphology was observed in EVs, as shown by TEM (Fig. 2b). In addition, EVs and IEVs were positive for CD9, CD63, Alix, and Tsg101, as determined by the Western blotting analysis (Fig. 2c). To confirm the EVs were loaded with integrin α5 peptide, we labeled DSPE-PEG-CRYYRITY with FAM fluorescence and EVs with DID fluorescence. Under confocal microscopy, images were captured, revealing co-localization between the FAM-labeled DSPE-PEG-CRYYRITY and DID-labeled EVs (Fig. 2d). NTA demonstrated that the average sizes of EVs and engineered IEVs were 148.6 ± 12.9 and 163.2 ± 27 nm (Fig. 2e), respectively. The increased size of IEVs can be attributed to the peptide modification on the surface of EVs. miR-138-5p was observed to be significantly upregulated in EV-138 compared to its expression levels in EV-negative control (NC) as determined by qRT-PCR, indicating the successful loading of miR-138-5p mimic (Fig. 2f). Ultraviolet (UV) spectroscopy analysis revealed that the PFD loading efficiency, which was calculated using the standard curve, was 28.61% ± 1.2% (Supplementary Fig. 2c).

Fig. 2

Preparation and characterization of extracellular vesicles (EVs) loaded with miR-138-5p and pirfenidone (PFD) and subjected to surface modification with integrin α5-targeting peptides (IEVs-PFD/138) and their enhanced cancer-associated fibroblast (CAF)-targeting ability. a Schematic representation of engineered bone marrow mesenchymal stem cell (BMSC)-derived EVs for the co-delivery of miR-138-5p mimic and PFD to pancreatic CAFs. b Transmission electron microscopy images of EVs (scale bar = 100 nm). c The EV markers (CD9, CD63, Alix, and Tsg101) in both EVs and IEVs were analyzed by Western blotting. d The fluorescence co-localization staining of FAM-peptide (green) and DID-labeled EVs (red). e Nanoparticle tracking analysis (NTA) of EVs and IEVs. f Quantitative real-time polymerase chain reaction (qRT-PCR) of miR-138-5p expression in EVs. Data represent the mean (±SD); n = 3 per group. g, h Cellular uptake of DID-labeled EVs and fluorescein amidite (FAM)-miR-138-5p-loaded EVs in CAFs. Laser scanning confocal microscopy (LSCM) (scale bar = 100 μm) and flow cytometric analyses of the uptake of IEVs by CAFs. i–l Quantification of LSCM and flow cytometric analysis results. Data are presented as mean (±SD); n = 3 per group. m qRT-PCR analysis of miR-138-5p expression in CAFs. Data represent the mean (±SD); n = 3 per group. n In vivo circulation time of IEVs and undecorated EVs. o Analysis of the pancreatic cancer model with in vivo Imaging System (IVIS). Data are presented as mean (±SD); n = 3 per group. p Analysis of the main organs and tumor with in vivo Imaging System (IVIS). Data are presented as mean (±SD); n = 3 per group. q Analysis of the intra-tumoral distribution of IEVs in the tumor section (scale bar = 100 μm). Data are presented as mean (±SD); n = 3 per group

IEVs exhibit improved CAF-targeting ability

To test the CAF-targeting ability of engineered IEVs, the cellular uptake of DID-labeled EVs by laser scanning confocal microscope (LSCM) and flow cytometry was assessed. The uptake of IEVs by CAFs was higher than that of non-targeted EVs (Fig. 2g, i, k). Similarly, we investigated the delivery efficiency of miR-138-5p (FAM) to CAFs. The FAM signal in CAFs incubated with IEVs-138 was higher than that in CAFs incubated with EVs-138, confirming that the integrin α5 peptide increased the CAF-targeting ability (Fig. 2h, j, l). qRT-PCR analysis of miR-138-3p revealed that the efficiency of IEVs to deliver the miRNA to CAFs was higher than that of non-targeted EVs (Fig. 2m). To investigate the targeting ability of IEVs in vivo, a subcutaneous fibrotic pancreatic cancer model was developed by co-implanting CAFs and PANC-1 cells into mice. The circulation time of IEVs in the peripheral blood was higher than that of undecorated EVs (Fig. 2n). Analysis with an in vivo imaging system (IVIS) revealed that the targeting ability of IEVs was higher than that of undecorated EVs (Fig. 2o). The fluorescence signal emitted by IEVs was significantly higher than that by undecorated EVs, indicating the enhanced ability of IEVs to reach pancreatic cancer site. The fluorescence intensity peaked at 4 h following the administration of EV. In addition, the kidneys, livers, lungs, spleens, hearts, and tumors were excised from mice 4 h following EV administration and subjected to ex vivo imaging. The results revealed a higher accumulation of IEVs in the tumor site compared to EVs (Fig. 2p). The tumors were sectioned to visualize the presence and spatial distribution of EVs within the tumor tissue. IEVs infiltrated the tumor tissue, which manifested as an expanded fluorescence area and enhanced fluorescence intensity when compared with undecorated EVs (Fig. 2q, Supplementary Fig. 2d).

miR-138-5p reprograms CAF phenotype by downregulating FERMT2

Functional experiments demonstrated that the proliferation and migration of CAFs were significantly suppressed by IEVs-PFD and IEVs-138, which produced strong inhibitory effects. Compared to other EVs, IEVs-PFD/138 had a stronger inhibition effect on CAF proliferation and migration (Fig. 3a–d, Supplementary Fig. 3a, c). These results suggest the potent anti-proliferative and anti-migratory properties of IEVs-PFD/138. Thus, IEVs-PFD/138 can modulate CAF behavior and promote TME remodeling.

Fig. 3

Functional and mechanistic analyses of miR-138-5p in pancreatic cancer-associated fibroblasts (CAFs). a–d EdU (Red: EdU-positive cells, blue: cell nuclei) and wound-healing assay demonstrated the inhibitory effects of extracellular vesicles (EVs) loaded with miR-138-5p and pirfenidone (PFD) and subjected to surface modification with integrin α5-targeting peptides (IEVs-PFD/138) on CAF cell proliferation and migration (scale bar = 100 μm). Data are presented as mean (± SD); n = 3 per group. e Six databases were used for predicting the potential target genes of miR-138-5p. f Predicted miR-138-5p-FERMT2 binding sites. g The results of the dual-luciferase reporter assay with miR-138-5p and FERMT2. Data are presented as mean (± SD); n = 3 per group. h, i Quantitative real-time polymerase chain reaction (qRT-PCR) and western blotting analyses of the impact of miR-138-5p mimic or inhibitor on the mRNA and protein levels of FERMT2 in CAFs. Data are presented as mean (± SD); n = 3 per group. j, k Gene set enrichment analysis of FERMT2 in pancreatic cancer. l Analysis of FERMT2 in pancreatic cancer with the Tumor Immune Single-cell Hub (TISCH) database (http://tisch.comp-genomics.org/home/). m Tissue protein microarray analysis of FERMT2 in pancreatic cancer tissues (scale bar = 100 μm). n Correlation of pathological scoring of FERMT2 with clinical stage in patients with pancreatic cancer. N = 50 for stage I/II; n = 7 for stage III/IV. o, p qRT-PCR and western blotting analyses of FERMT2 in primary CAFs and healthy fibroblasts (NFs). Data are presented as mean (± SD); n = 3 per group. q qRT-PCR and western blotting analyses of FERMT2 and the effect of miR-138-5p inhibitor and small interfering RNA (siRNA) against FERMT2. Data are presented as mean (± SD); n = 3 per group. r, s EdU (Red: EdU-positive cells, blue: cell nuclei) and wound-healing assay verified that miR-138-5p inhibitor mitigated the inhibitory effect of FERMT2 knockdown on cell proliferation and migration (scale bar = 100 μm). Data are presented as mean (± SD); n = 3 per group

To elucidate the underlying mechanism of miR-138-5p in CAFs, miR-138-5p’s potential target genes were identified by mining six databases (Fig. 3e). Among the predicted target genes, FERMT2 exhibited the strongest correlation with CAFs according to the TIMER dataset analysis (Supplementary Fig. 3e). Thus, the miR-138-5p-FERMT2 interaction was experimentally examined. Subsequently, the binding sites between miR-138-5p and FERMT2 were predicted (Fig. 3f). The results of the dual-luciferase reporter assay illustrated the capability of miR-138-5p to bind to the wild-type FERMT2, as indicated by the notable decrease in luciferase activity (Fig. 3g). This indicates the direct miR-138-5p-FERMT2 interaction. Furthermore, transfection with the miR-138-5p mimic led to the downregulation of both mRNA (Fig. 3h) and protein (Fig. 3i) levels of FERMT2 in CAFs. Conversely, transfection with the miR-138-5p inhibitor resulted in the upregulation of both mRNA and protein levels of FERMT2 in CAFs.

The functional role of FERMT2 in pancreatic cancer was examined using gene set enrichment analysis (GSEA) with the TCGA dataset. FERMT2 expression was found to positively correlate with the proliferation and migration of CAFs (Fig. 3j, k). After analyzing the Tumor Immune Single-cell Hub (TISCH) database, it was discovered that among the cell populations in pancreatic cancer, CAFs exhibited the highest FERMT2 expression, followed by cancer cells (Figs. 3l and Supplementary Fig. 3f). This provided valuable insights into the specific cellular contexts of FERMT2 expression, indicating its potential roles in tumor-stroma interactions and the TME. The expression of FERMT2 was observed in pancreatic cancer tissues but not in healthy pancreatic tissues, as determined by tissue protein microarray analysis. Notably, the expression levels of FERMT2 in cancer tissue can be categorized into three distinct patterns (Fig. 3m). The pathological scoring of FERMT2 was positively related to the clinical stage of pancreatic cancer patients (Fig. 3n). Protein (Fig. 3p) and mRNA (Fig. 3o) expression levels of FERMT2 were significantly higher in primary CAFs compared to NFs, as determined by qRT-PCR and western blot analysis.

Next, small interfering RNAs (siRNAs) against FERMT2 (si-FERMT2) were used to knock down the expression of FERMT2 and determine how FERMT2 knockdown affected the proliferative rate and migration of CAFs. In addition, the impact of FERMT2 knockdown and related phenotypes was abolished by transfecting CAFs with miR-138-5p inhibitor (Fig. 3q). The knockdown of FERMT2 considerably reduced CAF proliferation and migration, demonstrated by the EdU, CCK-8, wound-healing, and transwell assays. In addition, the miR-138-5p inhibitor effectively abrogated the FERMT2 knockdown-mediated downregulation of the proliferation and migration of CAFs (Fig. 3r, s and Fig. s3b, d). These findings revealed the effect of FERMT2 knockdown on CAFs and the pivotal function of miR-138-5p in mediating the inhibitory effects of FERMT2 knockdown on the proliferation and migration of CAFs.

FERMT2 interacts with TGFBR1 and PYCR1 in CAFs

Analysis of the TCGA pancreatic cancer dataset revealed that FERMT2 was positively correlated with the TGF-β signaling pathway (Fig. 4a). In addition, single-sample GSEA (ssGSEA) confirmed the positive correlation between FERMT2 and the TGF-β signaling pathway (Fig. 4b). Furthermore, analysis using the STRING database indicated that FERMT2 can potentially interact with TGFBR1 (Fig. 4c). This interaction was experimentally validated using co-IP analysis (Fig. 4d). Western blotting demonstrated that the protein expression levels of TGFBR1 and phosphorylated SMAD2/3 (p-SMAD2/3) were downregulated upon FERMT2 knockdown and restored upon miR-138-5p inhibition (Fig. 4e). Besides, the mRNA expression of FERMT2 positively correlated with ACTA2, COL1A1, FAP, and FSP based on TCGA pancreatic cancer data. (Fig. 4f). Western blotting results revealed that the protein expression levels of ACTA2, collagen 1, FAP, and FSP were decreased upon FERMT2 knockdown and restored upon the inhibition of miR-138-5p (Fig. 4g). These findings demonstrated that the inhibitory impact of miR-138-5p on the TGF-β signaling pathway could counteract the inhibitory effects of FERMT2 knockdown, indicating a regulatory function of miR-138-5p in the FERMT2-TGFBR1-TGF-β signaling pathway, which can further regulate CAF phenotype and collagen formation.

Fig. 4

FERMT2 interacts with TGFBR1 and PYCR1 in cancer-associated fibroblasts (CAFs). a Correlation of FERMT2 expression with the TGF-β pathway in The Cancer Genome Atlas (TCGA) pancreatic cancer cohort. b Single-sample gene set enrichment analysis (ssGSEA) of FERMT2 in TCGA pancreatic cancer cohort. c Analysis of the interaction of FERMT2 with TGF-β signaling-related proteins using STRING database (https://cn.string-db.org/). d Co-immunoprecipitation (Co-IP) analysis of FERMT2 and TGFBR1. e Western blotting analysis of the effect of FERMT2 knockdown and miR-138-5p inhibitor on the TGFBR1 and p-SMAD2/3 expression levels. f Correlation of FERMT2 with ACTA2, COL1A1, FAP, and FSP in TCGA pancreatic cancer dataset. g Western blotting analysis of the effect of FERMT2 knockdown and miR-138-5p inhibitor on the ACTA2, collagen1, FAP, and FSP expression levels. h Correlation of FERMT2 expression with the collagen production pathway in TCGA pancreatic cancer dataset. i Co-IP analysis of FERMT2 and TGFBR1. j Western blotting analysis of the impact of FERMT2 knockdown and miR-138-5p inhibitor on PYCR1 expression. k Effect of FERMT2 knockdown and PYCR1 overexpression on the proline content. Data are presented as mean (±SD); n = 3 per group. l Western blotting analysis of the effect of FERMT2 knockdown and PYCR1 overexpression on the PYCR1 and collagen1 expression levels

FERMT2 was positively correlated with the collagen formation pathway (Fig. 4h). PYCR1, a key enzyme involved in proline metabolism, regulates collagen deposition and remodeling. Co-IP experiments revealed the interaction between FERMT2 and PYCR1, indicating that these two proteins can form a complex (Fig. 4i). In addition, miR-138-5p inhibitor mitigated the FERMT2 knockdown-mediated downregulation of PYCR1 levels, as shown by Western blotting analysis (Fig. 4j). FERMT2 knockdown downregulated the proline levels. Conversely, PYCR1 overexpression upregulated the proline levels (Fig. 4k). Western blotting analysis revealed that PYCR1 overexpression mitigated the FERMT2 knockdown-mediated downregulation of PYCR1 and COL1A1 (Fig. 4l). Based on these data, the inhibition effect of miR-138-5p on PYCR1 expression can counteract the inhibitory effects of FERMT2 knockdown, indicating a regulatory function of miR-138-5p in modulating the FERMT2-PYCR1 axis, which can further regulate proline and collagen formation.

IEVs-PFD/138 treatment reverses CAF activation in vitro

The immunoprecipitation (IP) assay results demonstrated that IEVs-PFD and IEVs-138 decreased the abundance of the FERMT2-TGFBR1 and FERMT2-PYCR1 complexes. IEVs-PFD/138 exerted the highest inhibitory effects on the formation of these complexes (Fig. 5a). Western blotting analysis demonstrated that IEVs-PFD and IEVs-138 effectively downregulated the expression levels of FERMT2, PYCR1, and TGFBR1 (Fig. 5b), as well as the expression levels of p-SMAD2/3, indicating the suppression of TGF-β signaling pathway activity (Fig. 5b). In addition, the levels of ACTA2, FAP, FSP, and collagen1 were considerably downregulated by IEVs-PFD and IEVs-138 (Fig. 5b). Compared to other EVs, IEVs-PFD/138 had greater inhibitory impacts on the expression of these proteins. Both IEVs-PFD and IEVs-138 effectively suppressed CAF secretion of IL-6, TGFB1, and CXCL12, according to the results of the enzyme-linked immunosorbent assay (ELISA) (Fig. 5c). In particular, IEVs-PFD/138 exerted the highest inhibitory effects on the secretion of IL6, TGFB1, and CXCL12. These findings suggest that IEVs-PFD/138 suppressed the production of pro-inflammatory and pro-tumorigenic factors in CAFs. Analysis of the proline levels revealed that IEVs-PFD and IEVs-138 effectively downregulated the intracellular proline content in CAFs (Fig. 5d). In particular, IEVs-PFD/138 exerted the highest inhibitory effects on the proline levels. These findings suggest that IEVs-PFD/138 modulated proline metabolism in CAFs and consequently downregulated intracellular proline content. Thus, the application of IEVs-PFD/138 is a potential therapeutic strategy for modulating the activation and secretory profile of CAFs and the formation of collagen.

Fig. 5

Role of extracellular vesicles (EVs) loaded with miR-138-5p and pirfenidone (PFD) and subjected to surface modification with integrin α5-targeting peptides (IEVs-PFD/138) in cancer-associated fibroblast (CAF) phenotype reversal, cancer cell inhibition, and enhanced drug penetration. a Immunoprecipitation (IP) analysis of the FERMT2-TGFBR1 and FERMT2-PYCR1 complexes. b Western blotting analysis of the expression levels of FERMT2, PYCR1, TGFBR1, p-SMAD2/3, ACTA2, FAP, FSP, and collagen1. c Enzyme-linked immunosorbent assay (ELISA) for analyzing the secretion of IL6, TGFB1, and CXCL12 by CAFs. Data are presented as mean (± SD); n = 3 per group. d Proline levels in CAFs. Data represent the mean (± SD); n = 3 per group. e Scheme of the effect of conditioned medium (CM) derived from EV-treated CAFs on PANC-1 cells. f Western blotting analysis of PCNA, VIM, and N-cad in PANC-1 cells treated with CM derived from EV-treated CAFs. Data are presented as mean (±SD); n = 3 per group. g Evaluation of the proliferation, migration, and invasion of PANC-1 cells treated with CM derived from CAFs treated with EVs (Red: EdU-positive cells, blue: cell nuclei), (scale bar = 100 μm). Data are presented as mean (±SD); n = 3 per group. h–j Enhanced fluorescence intensity and penetration depth of Hoechst 33258 in stroma-rich three-dimensional (3D) multicellular tumor spheroids treated with IEVs-PFD/138 (scale bar = 100 μm). k–m Comparison of 3D multicellular tumor spheroid volume after treatment with the combination of IEVs-PFD/138 and gemcitabine (GEM) (scale bar = 100 μm). Data represent the mean (± SD); n = 3 per group

IEVs-PFD/138 treatment suppresses the tumor-promoting effect of CAFs in vitro

To elucidate the effect of EV-treated CAFs on the biological behaviors (proliferation, migration, and invasion) of pancreatic cancer cells, PANC-1 cells were incubated with the conditioned medium (CM) of EV-treated CAFs (Fig. 5e). After subjecting PANC-1 cells to CM treatment, Western blotting analysis showed an increase in the levels of PCNA, VIM, and N-cadherin. However, the application of CM derived from EV-treated CAFs, specifically IEVs-PFD and IEVs-138, attenuated this effect. Notably, the inhibitory effect was most pronounced with the application of IEVs-PFD/138 (Fig. 5f). The results of the EdU, transwell migration, and invasion assays revealed that the CM of CAFs the capacity of PANC-1 cells to proliferate, migrate, and invade. In contrast, the CM of CAFs treated with EVs, especially IEVs-PFD and IEVs-138, suppressed the capacity of PANC-1 cells to proliferate, migrate, and invade. Treatment with the CM of CAFs treated with IEVs-PFD/138 exerted the highest inhibitory effect on the proliferation, migration, and invasion of PANC-1 cells (Fig. 5g). These findings revealed the regulatory effects of CAF-derived CM on the proliferation, migration, and invasion of PANC-1 cells. Overall, the potent inhibitory impact of IEVs-PFD/138 improved our understanding of the reciprocal interactions between pancreatic cancer cells and CAFs and can aid in developing novel strategies targeting cancer progression.

IEVs-PFD/138 treatment enhances drug penetration into stroma-rich 3D multicellular tumor spheroids

The dense ECM within the desmoplastic pancreatic cancer TME is a major obstacle to the penetration of anticancer therapeutics. To address this challenge, a stroma-rich 3D multicellular tumor spheroid model was established by co-culturing PANC-1 cells with CAFs. This model closely mimics the desmoplastic characteristics of the pancreatic cancer TME. The effect of IEVs-PFD/138 on the stromal components and the penetration of small-molecule chemotherapy drugs in stroma-rich 3D multicellular tumor spheroids was examined. To simulate the penetration of small-molecule drugs, Hoechst 33258, a commonly used fluorescence probe was applied. IEVs-PFD/138-treated 3D multicellular tumor spheroids exhibited enhanced Hoechst 33258 fluorescence intensity and penetration depth (Fig. 5h–j). To validate the effect of IEVs-PFD/138 on drug penetration, 3D multicellular tumor spheroids were co-treated with IEVs-PFD/138 and GEM. At equivalent concentrations of GEM (10 and 20 μM), the volume of 3D spheroids pre-treated with IEVs-PFD/138 was significantly reduced relative to that of 3D spheroids pre-treated with phosphate-buffered saline (PBS) (Fig. 5k–m). Furthermore, necrotic cells were observed surrounding the 3D multicellular tumor spheroids. These findings suggested that IEVs-PFD/138 potentiate the efficacy of small-molecule anti-tumor drugs in the TME of pancreatic cancer by abrogating the buildup of fibrotic stroma and improving drug delivery.

IEVs-PFD/138 treatment reduces activated CAFs and proliferation in 3D human PDAC tumor explants

EVs were introduced into 3D human PDAC tumor explants derived from surgical resections, followed by seven days of continuous culture (Supplementary Fig. 4a). This model maintains the 3D organization and interactions of PDAC cells and stromal cells in a tissue culture dish. IEVs-PFD/138 exhibited a significant decrease in activated CAFs and cellular proliferative rate compared to the control groups (Supplementary Fig. 4b, c).

IEVs-PFD/138 treatment exerts anti-tumor effects and reverses CAF activation in a subcutaneous model of desmoplastic pancreatic cancer

To examine the therapeutic implications and fundamental mechanisms of IEVs-PFD/138 in vivo, a model of subcutaneous desmoplastic pancreatic cancer was developed in mice via the co-transplantation of PANC-1 cells and CAFs. Subsequently, mice were intravenously injected with EVs or PBS for seven treatment cycles (Fig. 6a). The results demonstrated a significant inhibition of tumor growth following treatment with both IEVs-PFD and IEVs-138 (Fig. 6b–e). Remarkably, IEVs-138/PFD exhibited a pronounced anti-tumor effect, leading to a significant extension in the survival time of nude mice (time required for the tumor volume to reach 1000 mm3) (Fig. 6d).

Fig. 6

Effect of treatment with extracellular vesicles (EVs) loaded with miR-138-5p and pirfenidone (PFD) and subjected to surface modification with integrin α5-targeting peptides (IEVs-PFD/138) in the subcutaneous desmoplastic pancreatic cancer model. a Scheme of IEVs-PFD/138 treatment. b Images of the mouse model after treatment on day 21. c, d Tumor growth curves of the mouse models. Data are presented as mean (± SD); n = 5 per group. e Tumor weight of the mouse models. Data are presented as mean (± SD); n = 5 per group. f Effect of treatment on the survival of the mouse model. g Ki67 immunofluorescence staining, Masson’s staining, ACTA2 immunofluorescence staining, FAP immunofluorescence staining, FSP immunofluorescence staining, PYCR1 immunofluorescence staining, and collagen1 immunohistochemical staining of tumors from mice under different treatments (scale bar = 100 μm). h Western blotting analysis of the FERMT2, PYCR1, TGFBR1, and p-SMAD2/3 expression levels in tumors from mice administered various treatments. i Hematoxylin and eosin (HE) staining of vital organs from mice administered various treatments (scale bar = 100 μm)

Immunofluorescence analysis revealed that tumor cell proliferation was significantly downregulated in the IEVs-PFD/138-treated group as evidenced by the decreased immunofluorescence intensity of the proliferation marker Ki67 (Fig. 6g). The immunofluorescence intensities of ACTA2, FAP, and FSP, which serve as activation markers for CAFs, were markedly downregulated in the IEVs-PFD/138-treated group, indicating the reversal of CAF activation (Fig. 6g). In addition, the collagen content of the tumors decreased considerably in the group that received IEVs-PFD/138, as evidenced by Masson’s trichrome staining. Moreover, the PYCR1 and collagen1 levels were downregulated in the IEVs-PFD/138-treated group, suggesting that treatment with IEVs-PFD/138 decreases the ability of CAFs to produce collagen (Fig. 6g). Western blotting analysis revealed the downregulation of FERMT2 and PYCR1 in the IEVs-PFD/138-treated group (Fig. 6h). In addition, the TGFBR1 expression level was significantly downregulated in the IEVs-PFD/138-treated group. Furthermore, the p-SMAD2/3 levels were downregulated in the IEVs-PFD/138-treated group, indicating the suppression of the TGF-β signaling pathway (Fig. 6h). Furthermore, histological examination of vital organs using hematoxylin and eosin (HE) staining demonstrated no significant damage in important organs of nude mice treated with EVs (Fig. 6i). This indicates that EVs do not exert toxic effects on vital organs. Thus, IEVs-PFD/138 treatment exerts inhibitory effects on pancreatic cell proliferation and the TGF-β signaling pathway, reverses CAF activation, and downregulates collagen deposition. These findings provide valuable insights into the mechanisms governing the anti-tumor and anti-fibrotic properties of IEVs-PFD/138 and suggest that IEVs-PFD/138 treatment is a promising treatment strategy for pancreatic cancer.

IEVs-PFD/138 treatment enhances the therapeutic efficacy of GEM and reverses desmoplastic TME phenotype in a pancreatic cancer PDX model

The PDX model is considered a clinically relevant model as it utilizes patient tumor samples for transplantation into immunodeficient mice, preserving the tumor heterogeneity and stromal microenvironment. In this study, PDX at the F3 passage was used as the experimental model. EVs and GEM were sequentially administered via the tail vein for eight treatment cycles (Fig. 7a). On day 25, tumors were collected from five mice in each group for further analysis. Notably, the tumor growth rate was found to be considerably reduced in the GEM-treated group compared to the PBS-treated group, as indicated by the tumor volume growth curve analysis. The tumor growth rate was the slowest in the group treated with the combination of IEVs-PFD/138 and GEM (Fig. 7b–d). In addition, tumors treated with GEM had a reduced weight relative to those treated with PBS. The tumor weight was the lowest in the group treated with the combination of IEVs-PFD/138 and GEM (Fig. 7e).

Fig. 7

Effect of treatment with extracellular vesicles (EVs) loaded with miR-138-5p and pirfenidone (PFD) and subjected to surface modification with integrin α5-targeting peptides (IEVs-PFD/138) in the patient-derived xenograft (PDX) pancreatic cancer models. a Scheme of PDX model establishment and treatment. b Images of mouse models and tumors after treatment on day 24. c, d Tumor growth curves of the mouse models. Data are presented as mean (± SD); n = 5 per group. e Tumor weight of the mouse models. Data are presented as mean (±SD); n = 5 per group. f Relative gemcitabine (GEM) concentrations in tumors. Data are presented as mean (±SD); n = 3 per group. g Elastic modulus of tumors. Data are presented as mean (± SD); n = 3 per group. h Solid stress of tumors (three tumors with each tumor having three parts). i Effect of treatment on the survival of the mouse model. j Hematoxylin and eosin (HE) staining, Masson’s staining, collagen1 immunohistochemical staining, ACTA2 immunohistochemical staining, Ki67 immunohistochemical staining, and Tunel staining of tumors from mice administered various treatments (scale bar = 200 μm). k Western blotting analysis of the FERMT2, PYCR1, TGFBR1, P-SMAD2/3, HIF1A, and ENT1 expression levels in tumors from mice subjected to different treatments

IEVs-PDF/138 suppressed tumor stroma generation, enhancing drug penetration. Thus, this study quantified GEM content in tumor tissues using UV spectroscopy. In comparison to the GEM-treated group, the IEVs-PDF/138-treated group exhibited reduced tumor concentrations of GEM (Fig. 7f). The tumor elastic modulus was significantly downregulated in the IEVs-PFD/138-treated group (Fig. 7g). Furthermore, the effect of IEVs-PFD/138 on the solid stress of tumors was examined by measuring solid stress levels according to an established protocol. Significantly reduced solid stress was observed within tumors subjected to IEVs-PFD/138 as compared to the groups treated with PBS and GEM (Fig. 7h). These findings suggest that IEVs-PFD/138 treatment effectively decreases tumor stroma generation. The observed reduction in elastic modulus and solid stress indicated a modulation of the mechanical forces exerted by the stromal components within the TME. The desmoplastic stroma in pancreatic cancer is reported to generate physical barriers that limit drug penetration and hinder therapeutic efficacy. The observed reduction in elastic modulus and solid stress within the tumors induced by IEVs-PFD/138 treatment can explain the enhanced efficiency of GEM. The reduction in solid stress alleviated the mechanical compression exerted on blood vessels and enhanced vascular perfusion within the tumor, facilitating the delivery of GEM to tumor cells. In addition, the decreased stromal density and interstitial fluid pressure resulting from stroma reduction may contribute to enhanced drug diffusion and distribution throughout the tumor. Survival analysis indicated that mice in the EV-treated group exhibited the longest survival time (time required for the tumor volume to reach 1000 mm3) (Fig. 7i). These findings suggest the potential of IEVs-PFD/138 in synergistically enhancing the therapeutic efficacy of GEM in the PDX model.

HE staining, Masson’s staining, and immunohistochemical (IHC) analysis of collagen1, revealed that IEVs-PFD/138 markedly downregulated tumor collagen deposition, indicating effective modulation of the ECM components (Fig. 7j). In addition, IHC analysis revealed the downregulation of ACTA2 expression, indicating the suppression of CAF activation. These findings suggest that IEVs-PFD/138 treatment suppresses CAF activation, contributing to the modulation of the TME (Fig. 7j). The expression of KI67 was markedly downregulated in the IEVs-PFD/138-treated group (Fig. 7j), suggesting the decreased proliferative capacity of tumor cells. In addition, TUNEL staining revealed that IEVs-PFD/138 significantly upregulated the number of apoptotic cells (Fig. 7j). This further supports the anti-proliferative and pro-apoptotic effects of IEVs-PFD/138. These results can be attributed to the enhanced efficacy of GEM when combined with IEVs-PFD/138.

The FERMT2 and PYCR1 protein expression were found to be decreased in the group that received IEVs-PFD/138, as determined by Western blotting analysis (Fig. 7k). In addition, IEVs-PFD/138 downregulated the TGFBR1 and p-SMAD2/3 levels, which suggests inhibition of the TGF-β signaling pathway (Fig. 7k). The TGF-β pathway assumes a multifaceted function in cancer as it can promote CAF activation and exert tumorigenic effects. The downregulation of TGFBR1 and p-SMAD2/3 suggests the suppression of TGF-β signaling, which may contribute to the reversal of CAF activation and the inhibition of cancer-promoting effects. HIF1A expression was decreased in the group treated with IEVs-PFD/138 (Fig. 7k). The downregulation of HIF1A, an essential regulator of hypoxia-induced cellular response, indicates the alleviation of tumor hypoxia. The alleviation of hypoxia can be attributed to the IEVs-PFD/138-mediated downregulation of tumor stromal pressure. The downregulation of HIF1A suggests a shift toward a less hypoxic TME. Moreover, the IEVs-PFD/138 treatment resulted in increased expression of ENT1 protein in tumor tissue (Fig. 7k). ENT1 is a nucleoside transporter responsible for the uptake of GEM by cancer cells. The upregulation of ENT1 expression suggests enhanced GEM uptake by pancreatic cancer cells, leading to greater drug sensitivity. The increased ENT1 expression, influenced by the modulation of the TGF-β pathway and HIF1A, facilitates increased intracellular GEM accumulation and improved therapeutic response.

These findings have significant implications for the development of innovative therapeutic approaches for pancreatic cancer. The suppression of CAF activation and collagen deposition induced by IEVs-PFD/138 treatment may effectively alleviate the desmoplastic stromal barrier associated with tumor progression and therapy resistance. Furthermore, the enhanced efficacy of GEM in combination with IEVs-PFD/138 highlights the potential synergistic effects of this treatment strategy for pancreatic cancer.

IEVs-PFD/138 treatment potentiates the therapeutic efficacy of GEM and inhibits metastasis in an orthotopic model of desmoplastic pancreatic cancer

The orthotopic model of desmoplastic pancreatic cancer established using PANC-1 cells and CAFs accurately reflects the native TME and anatomical characteristics of the primary tumor site. This model enables the evaluation of cancer-stroma interactions and metastatic dissemination. In addition, this model provides a comprehensive representation of tumor biology and therapeutic response. IEVs-PFD/138 and GEM were sequentially and intravenously administered for seven treatment cycles. Bioluminescence imaging was performed on days 0, 7, 14, and 21 to monitor the growth of the orthotopic pancreatic tumors (Fig. 8a). The group that received GEM treatment exhibited a reduced growth rate of pancreatic tumors compared to the group that received PBS. The combination of GEM and IEVs-PFD/138 significantly decreased tumor growth (Fig. 8b–d). These findings suggest the potential of IEVs-PFD/138 in enhancing the therapeutic efficacy of GEM against orthotopic pancreatic cancer.

Fig. 8

Effects of treatment with extracellular vesicles (EVs) loaded with miR-138-5p and pirfenidone (PFD) and subjected to surface modification with integrin α5-targeting peptides (IEVs-PFD/138) in the orthotopic desmoplastic pancreatic cancer models. a Scheme of IEVs-PFD/138 and gemcitabine (GEM) treatment. b Illustrative in vivo bioluminescence images of mice. c Bioluminescence intensity quantification of mice with orthotopic pancreatic tumors. Data represent the mean (±SD); n = 3 per group. d Growth-inhibitory effec