Targeted extracellular vesicle-photoimmunotherapy remodels stromal-immune microenvironment to boost chemo-immunotherapy in preclinical models Authors: Chengming Qu, Jie Liu, Jinpeng Cen, Ming Tian, Yibo Chai, Peng Xia, Runhua Zhou, Jingbo Ma, Huifang Wang, Haitao Yuan, Felix Rückert, Ji Li, Bo He, Zhiyong Yang, Rui Hou, Zhijie Li, Jigang Wang, Yufeng Yuan Published: 2026 ABSTRACT SummarySolid tumors, especially pancreatic ductal adenocarcinomas (PDACs), activate quiescent fibroblasts into cancer-associated fibroblasts (CAFs) that generate dense desmoplastic stroma. This barrier restricts drug penetration and immune infiltration, promoting tumor progression. Here, we engineer Midkine (MDK)-targeting nanobody-functionalized extracellular vesicles (D4-EV) as a precision photoimmunotherapy platform. These vesicles selectively accumulate in the tumor microenvironment through MDK overexpression. Loaded with chlorin e6 (Ce6), Ce6@D4-EV induces immunogenic cell death upon light irradiation, triggering dsDNA release and cGAS-STING activation in tumor-associated macrophages. Concurrently, it reprograms CAFs, reduces extracellular matrix deposition, improves vascular perfusion, and alleviates hypoxia. This stromal-immune remodeling substantially enhances the therapeutic efficacy of immune checkpoint blockade, adoptive T cell therapy, and chemotherapy, leading to prolonged survival in multiple MDK-positive preclinical tumor models. The platform provides a promising strategy to overcome stromal barriers in desmoplastic tumors. Introduction The tumor microenvironment (TME) of pancreatic ductal adenocarcinoma (PDAC) features dense desmoplastic stroma composed of heterogeneous cancer-associated fibroblasts (CAFs), abundant extracellular matrix (ECM) proteins, immunosuppressive immune cells, and a hypovascular network. This pathological remodeling severely impedes drug penetration and immune cell infiltration, reinforcing resistance to chemotherapy, immune checkpoint inhibitors (ICIs), and adoptive cell transfer (ACT).1 CAFs are the primary producers of ECM components that increase tumor stiffness and foster immunosuppression.2,3 However, strategies that fully deplete CAFs—such as targeting α-smooth muscle actin-positive (α-SMA+) myofibroblasts, inhibiting the Hedgehog pathway, or using lysyl oxidase-like-2 (LOXL2) antibodies—have proven counterproductive. These approaches often accelerate tumor progression by enhancing regulatory T cell infiltration, exacerbating immunosuppression, and shortening survival in preclinical models and clinical settings.4,5,6,7,8 In contrast, reprogramming CAFs from a pro-tumorigenic activated state toward a quiescent phenotype has shown greater promise, particularly when combined with chemotherapy.9,10,11,12 A critical driver of PDAC progression is the reciprocal crosstalk between cancer cells and CAFs, which establishes a self-reinforcing cycle that amplifies ECM deposition, stromal stiffness, and therapeutic resistance.13,14 Hence, strategies that simultaneously inhibit malignant cells and reprogram CAFs hold superior potential for breaking this vicious cycle compared with monotherapies. Photodynamic therapy (PDT) employs light-activated photosensitizers (PSs) to generate reactive oxygen species (ROS) that induce localized tumor cell death. Clinically approved for superficial and accessible tumors including melanoma and gastrointestinal, breast, and lung cancers, PDT also triggers immunogenic cell death (ICD), vascular priming, and stromal modulation, functioning effectively as photoimmunotherapy (PIT).15,16,17,18,19,20 While preclinical and clinical studies of PDT in PDAC using agents such as chlorin e6 (Ce6) have yielded encouraging results, the dense stroma and elevated interstitial fluid pressure (IFP) substantially hinder PS delivery and intratumoral distribution.21,22 Therefore, innovative targeted delivery systems with enhanced TME selectivity are essential to unlock the full therapeutic potential of PDT for PDAC therapy. Extracellular vesicles (EVs), particularly exosomes (30–150 nm), offer distinct advantages as drug carriers owing to their natural lipid bilayer, excellent biocompatibility, low immunogenicity and toxicity, high loading capacity, and ability to cross biological barriers.23,24 Genetic surface engineering further enables tumor-specific targeting. Midkine (MDK) is a multifunctional heparin-binding cytokine that is highly overexpressed and extracellularly enriched in the TME of numerous cancers, rather than being predominantly membrane bound.25,26,27 This secretion-based distribution parallels clinically successful targets such as vascular endothelial growth factor A (VEGFA) (targeted by bevacizumab) and acid phosphatase 3 (ACP3), supporting selective TME accumulation and therapeutic delivery.28,29,30,31,32,33,34,35 MDK promotes tumor growth, metastasis, therapeutic resistance, and immunosuppression across a broad spectrum of malignancies, with pan-cancer analyses showing upregulation in far more tumor types than VEGFA. Single-cell transcriptomics and immunohistochemistry consistently localize MDK predominantly to malignant cells, with additional contributions from stromal fibroblasts and endothelial cells. Our prior work demonstrated that MDK nanobody (Nb)-engineered nanoparticles achieve selective PDAC accumulation and enhance PIT.36 In this study, we developed MDK-targeting nanobody (D4)-functionalized EVs (MDK D4-EVs) loaded with the second-generation clinically approved PS-Ce6, generating the platform Ce6@D4-EV. Near-infrared (NIR) imaging in subcutaneous and orthotopic PDAC models with high MDK expression and dense desmoplastic stroma confirmed superior tumor-targeting specificity. Upon irradiation, Ce6@D4-EV-mediated targeted PDT induced ICD in cancer cells, releasing dsDNA that activates the cGAS-STING pathway in tumor-associated macrophages (TAMs). Simultaneously, it reprogrammed CAFs, reducing ECM deposition, improving vascular perfusion, alleviating hypoxia, and promoting immune cell infiltration and therapeutic agent penetration. These coordinated stromal-immune remodeling effects markedly potentiated the efficacy of ICIs (anti-PD-1 and anti-CD47), adoptive T cell therapy, and chemotherapy (gemcitabine [Gem] and paclitaxel [PTX]), yielding significant survival benefits across multiple MDK-positive preclinical models. This EV-based PIT platform represents a powerful approach to surmount stromal barriers in desmoplastic tumors and provides a foundation for clinical translation of targeted PIT. Result Recapitulation of MDK expression in tumors MDK is a multifunctional secretory protein of approximately 13 kDa that binds to sulfated mucopolysaccharides and interacts with multiple receptors, including syndecans, low-density lipoprotein receptor-related protein (LRP), integrins, tyrosine phosphatase-ζ (PTP-ζ), and CD109.37,38,39 It plays crucial roles in tumorigenesis, progression, and the establishment of an immunosuppressive TME in various cancers.40,41,42,43 Unlike many other cytokines, MDK shows minimal or undetectable expression in most normal adult tissues, while it is significantly elevated in a broad spectrum of tumors, including PDAC and hepatocellular carcinoma (HCC).37 To systematically evaluate its potential as a therapeutic target, we first analyzed MDK gene expression across tumor types using The Cancer Genime Atlas database and compared it with VEGFA, a classical cytokine involved in neo-angiogenesis and tumor development (Figure S1). MDK was significantly upregulated in 22 tumor types relative to corresponding normal tissues. In striking contrast, VEGFA elevation was limited to only four tumor types (cholangiocarcinoma, glioblastoma multiforme, head and neck squamous cell carcinoma, and kidney renal clear cell carcinoma). Moreover, VEGFA was downregulated compared with normal tissues in several other cancers, including kidney renal papillary cell carcinoma, liver HCC, skin cutaneous melanoma, thyroid carcinoma, and uterine corpus endometrial carcinoma. These findings highlight that MDK overexpression is far more extensive across tumor types than that of VEGFA. We next examined MDK expression at single-cell resolution using publicly available clinical single-cell RNA sequencing (RNA-seq) datasets from 11 human tumor types, including PDAC,44 HCC,45 breast cancer,46 papillary thyroid carcinoma,47 gastric cancer,48 esophageal cancer,49 renal cell carcinoma,50 lung adenocarcinoma,51 lung squamous cell carcinoma,51 colon adenocarcinoma, and rectal adenocarcinoma.48 These analyses were complemented by immunohistochemical (IHC) staining of a tissue microarray containing the same 11 tumor entities. Across all examined tumors, MDK transcripts were predominantly expressed in malignant cell populations, with additional partial expression detected in stromal compartments at the mRNA level. At the protein level, however, MDK showed consistent and predominant enrichment within the TME (Figure S2). We further confirmed MDK protein expression in PDAC cell lines (KPC and AsPC-1) and the liver cancer cell line Hepa1-6, as well as in xenograft tumor tissues generated from KPC and Hepa1-6 cells (Figures S3A and S3B). Taken together, these comprehensive analyses indicate that MDK represents a valuable “pan-cancer” biomarker and therapeutic target. Its broad overexpression pattern suggests that MDK-targeted strategies may have wider clinical applicability than many existing TME-directed approaches. Preparation and characterization of MDK-Nb-engineered EVs We previously identified two nanobodies (D4 and C3) with high affinity for both human and murine MDK proteins (Figure 1A).36 To achieve stable and efficient surface modification of EVs with these nanobodies, HEK-293 cells were genetically engineered using the pDisplay vector system, which incorporates the platelet-derived growth factor receptor β (PDGFRβ) transmembrane domain and a hemagglutinin (HA) tag for surface display (Figure 1B).52 Stable cell lines expressing D4-Nb, C3-Nb, or control (Ctrl)-Nb were established through antibiotic selection. Successful surface expression of the nanobodies on HEK-293 cells was confirmed by western blotting and flow cytometry (Figures 1C and 1D).Figure 1Preparation and characterization of MDK-Nb-engineered EVs(A) Purification and confirmation of MDK-C3-Nb, MDK-D4-Nb, and Ctrl-Nb by western blotting using HA-tag, His-tag, and VHH antibodies.(B) Schematic workflow of HEK-293 cell transfection with pDisplay vectors encoding Nb sequences and isolation of Nb-engineered EVs.(C) Western blot analysis of Nb expression in stable HEK-293 cells after antibiotic selection.(D) Flow cytometric detection of surface Nb expression on engineered HEK-293 cells (n = 3).(E) Nano-flow cytometric detection of surface Nb on engineered EVs (n = 3).(F) Nanoparticle tracking analysis (NTA) of size distribution and concentration of engineered EVs.(G) Western blot assessment of EV biomarkers in engineered parental cells and isolated EVs.(H) TEM images of engineered EVs showing typical cup-shaped morphology. Scale bars, 40 nm.(I) Zeta potential measurement of engineered EVs (n = 3).(J) SDS-PAGE and western blot analysis of recombinant human and mouse MDK proteins.(K and L) ELISA evaluation of binding of engineered EVs to human (K) and murine (L) MDK proteins (n = 3).(M and N) Stability of particle size (M) and concentration (N) of engineered EVs monitored over 4 weeks (n = 3).(O) Total protein content of Blank-EV, Ctrl-EV, C3-EV, and D4-EV across three independent batches measured by BCA assay (n = 3). One-way ANOVA.(P) Particle-to-protein ratios of Blank-EV, Ctrl-EV, C3-EV, and D4-EV across three batches, showing high purity and low batch-to-batch variability (CV < 20%) (n = 3). One-way ANOVA.(Q) Endotoxin levels of Blank-EV, Ctrl-EV, C3-EV, and D4-EV measured by LAL assay (n = 3). One-way ANOVA.Data are presented as mean ± SEM. ns, no significance. Preparation and characterization of MDK-Nb-engineered EVs (A) Purification and confirmation of MDK-C3-Nb, MDK-D4-Nb, and Ctrl-Nb by western blotting using HA-tag, His-tag, and VHH antibodies. (B) Schematic workflow of HEK-293 cell transfection with pDisplay vectors encoding Nb sequences and isolation of Nb-engineered EVs. (C) Western blot analysis of Nb expression in stable HEK-293 cells after antibiotic selection. (D) Flow cytometric detection of surface Nb expression on engineered HEK-293 cells (n = 3). (E) Nano-flow cytometric detection of surface Nb on engineered EVs (n = 3). (F) Nanoparticle tracking analysis (NTA) of size distribution and concentration of engineered EVs. (G) Western blot assessment of EV biomarkers in engineered parental cells and isolated EVs. (H) TEM images of engineered EVs showing typical cup-shaped morphology. Scale bars, 40 nm. (I) Zeta potential measurement of engineered EVs (n = 3). (J) SDS-PAGE and western blot analysis of recombinant human and mouse MDK proteins. (K and L) ELISA evaluation of binding of engineered EVs to human (K) and murine (L) MDK proteins (n = 3). (M and N) Stability of particle size (M) and concentration (N) of engineered EVs monitored over 4 weeks (n = 3). (O) Total protein content of Blank-EV, Ctrl-EV, C3-EV, and D4-EV across three independent batches measured by BCA assay (n = 3). One-way ANOVA. (P) Particle-to-protein ratios of Blank-EV, Ctrl-EV, C3-EV, and D4-EV across three batches, showing high purity and low batch-to-batch variability (CV < 20%) (n = 3). One-way ANOVA. (Q) Endotoxin levels of Blank-EV, Ctrl-EV, C3-EV, and D4-EV measured by LAL assay (n = 3). One-way ANOVA. Data are presented as mean ± SEM. ns, no significance. Nb-engineered EVs were subsequently isolated from the culture supernatant of these stable cells by ultracentrifugation. Surface display of the nanobodies on the isolated EVs was verified using nano-flow cytometry with anti-HA antibodies (Figure 1E). Nanoparticle tracking analysis (NTA) demonstrated that all EV preparations exhibited an average particle size of approximately 120 nm, and surface modification with nanobodies did not significantly alter this size distribution (Figure 1F). Western blotting confirmed the presence of canonical EV markers (CD9, CD81, and HSP70) in both parental cells and derived EVs, while the endoplasmic reticulum marker calnexin was absent in EV fractions, consistent with high purity. Importantly, only EVs and cells engineered with Nb constructs showed positive signals for the HA tag and nanobody framework when probed with anti-VHH antibodies (Figure 1G). Transmission electron microscopy (TEM) revealed the characteristic cup-shaped morphology of EVs in all groups, regardless of Nb engineering (Figure 1H). Zeta potential measurements indicated that all EV preparations carried a slightly negative surface charge, with no significant differences among groups following Nb modification (Figure 1I). Binding functionality was assessed by ELISA. Both MDK D4-EV and MDK C3-EV demonstrated specific binding to recombinant human and murine MDK proteins. Notably, D4-EV exhibited significantly stronger binding activity than C3-EV (Figures 1J–1L), leading to its selection as the primary candidate for subsequent experiments. To evaluate the multivalent display, we quantified D4-Nb copy number per EV, revealing approximately 78 copies displayed on the surface of each particle (Figures S3C–S3E). To support potential clinical translation, we performed rigorous quality control across three independently produced batches. NTA profiling confirmed highly consistent modal diameters centered around 100–120 nm at day 0, with excellent physical stability maintained after 28 days of storage at 4°C and no significant changes in particle concentration (Figures 1M and 1N). Total protein content measured by bicinchoninic acid (BCA) assay was comparable across Blank-EV, Ctrl-EV, C3-EV, and D4-EV preparations (Figure 1O), while particle-to-protein ratios showed low batch-to-batch variability (coefficient of variation [CV] < 20%) (Figure 1P), indicating high purity and reproducible manufacturing. Endotoxin levels, determined by the Limulus amebocyte lysate (LAL) assay,53 remained well below the FDA-recommended threshold of 0.25 EU/mL for injectable products in all batches (Figure 1Q). These comprehensive characterizations confirm that functional MDK-targeted EVs were successfully engineered with excellent stability, purity, and safety profiles suitable for further preclinical development. MDK-Nb-engineered EVs loaded with PS repress cancer cell proliferation and fibroblast activation in vitro Efficient internalization by target cells is a prerequisite for effective EV-mediated drug delivery. We, therefore, evaluated the uptake of PKH67-labeled Blank-EV, C3-EV, and D4-EV by pancreatic cancer cell lines KPC and AsPC-1. All three formulations showed time-dependent internalization, with significantly higher uptake at 8 h compared with 2 h (Figure 2A). Importantly, functionalization with MDK-targeting nanobodies (C3 or D4) did not impair cellular uptake relative to unmodified Blank-EV. This observation aligns with the intended TME-targeting design of the platform, which relies on extracellular MDK enrichment rather than direct cell-surface receptor-mediated endocytosis. Given the superior binding affinity of D4-EV to MDK (Figures 1K and 1L), it was selected for all subsequent studies.Figure 2Biocompatibility of MDK-D4-Nb engineered EVs and suppression of TGF-β/SMAD-dependent CAF activation by Ce6@D4-EV in vitro(A) Internalization and quantification of Blank-EV, C3-EV, and D4-EV in KPC and AsPC-1 cells (n = 3). Scale bars, 10 μm. Student’s t test.(B and C) Internalization and quantification of Ce6@D4-EV in MRC5-CAF and 10T1/2-CAF cells. Scale bars, 25 μm.(D) Volcano plot of DEGs between two groups (n = 3).(E) KEGG pathway enrichment analysis of downregulated genes.(F) Immunofluorescence staining of CAF biomarkers (α-SMA, Col-IV, and FN) in 10T1/2-CAF cells after Ce6@D4-EV plus irradiation (n = 3). Scale bars, 500 μm.(G) Quantification of α-SMA, Col-IV, and FN expression in 10T1/2-CAF cells after indicated treatments (n = 3). One-way ANOVA.(H) Representative immunofluorescence images showing p-SMAD2/3 nuclear translocation and expression of α-SMA, Col-IV, and FN in 10T1/2-CAF cells under indicated treatments. Scale bars, 10 μm.(I) Quantification of immunofluorescence signals for p-SMAD2/3, α-SMA, Col-IV, and FN in (H) (n = 3). One-way ANOVA.Data are presented as mean ± SEM. ns, no significance; ∗∗p < 0.01, ∗∗∗p < 0.001. Biocompatibility of MDK-D4-Nb engineered EVs and suppression of TGF-β/SMAD-dependent CAF activation by Ce6@D4-EV in vitro (A) Internalization and quantification of Blank-EV, C3-EV, and D4-EV in KPC and AsPC-1 cells (n = 3). Scale bars, 10 μm. Student’s t test. (B and C) Internalization and quantification of Ce6@D4-EV in MRC5-CAF and 10T1/2-CAF cells. Scale bars, 25 μm. (D) Volcano plot of DEGs between two groups (n = 3). (E) KEGG pathway enrichment analysis of downregulated genes. (F) Immunofluorescence staining of CAF biomarkers (α-SMA, Col-IV, and FN) in 10T1/2-CAF cells after Ce6@D4-EV plus irradiation (n = 3). Scale bars, 500 μm. (G) Quantification of α-SMA, Col-IV, and FN expression in 10T1/2-CAF cells after indicated treatments (n = 3). One-way ANOVA. (H) Representative immunofluorescence images showing p-SMAD2/3 nuclear translocation and expression of α-SMA, Col-IV, and FN in 10T1/2-CAF cells under indicated treatments. Scale bars, 10 μm. (I) Quantification of immunofluorescence signals for p-SMAD2/3, α-SMA, Col-IV, and FN in (H) (n = 3). One-way ANOVA. Data are presented as mean ± SEM. ns, no significance; ∗∗p < 0.01, ∗∗∗p < 0.001. The second-generation PS-Ce6 was loaded into D4-EV and unmodified Blank-EV to produce Ce6@D4-EV and Ce6@Blank-EV, respectively. Based on a standard curve (Figure S3F), the encapsulation efficiency reached 72%, with each D4-EV particle carrying approximately 3.05 × 106 Ce6 molecules, a loading capacity higher than many previously reported EV systems.54 These data indicate the excellent drug-loading capability of EVs as delivery vectors. As a PS for PDT, Ce6 generates ROS upon light irradiation. Ce6@D4-EV produced robust, time-dependent ROS generation following irradiation at 30 mW/cm2, compared with the blank (Figure S3G). Importantly, the photochemical activity of Ce6 remained intact after EV encapsulation, as Free-Ce6, Ce6@Blank-EV, and Ce6@D4-EV exhibited comparable ROS production kinetics (Figure S3G). We next assessed the in vitro antitumor efficacy of Ce6@D4-EV. In pancreatic cancer cells, Ce6@D4-EV plus irradiation induced concentration-dependent cytotoxicity, while treatment without light irradiation showed no effect on cell viability (Figure S3H). Colony formation assays and Annexin V/PI apoptosis analysis further confirmed that Ce6@D4-EV plus light exposure significantly inhibited proliferation and induced apoptosis compared with multiple control groups (Figures S3I and S3J). In addition, Ce6@D4-EV-mediated PDT triggered hallmark features of ICD, including surface translocation of calreticulin (CRT) and ERp57, as well as extracellular release of HMGB1 and ATP (Figures S3K–S3M). These ICD markers suggest the potential to convert dying tumor cells into an endogenous vaccine capable of stimulating antitumor immune responses. Because dense desmoplastic stroma driven by CAFs represents a major barrier in PDAC, we also investigated the impact of Ce6@D4-EV-mediated PDT on fibroblasts. CAF-like phenotypes were induced in human MRC5 lung fibroblasts and murine 10T1/2 embryonic fibroblasts by stimulation with TGF-β (5 ng/mL). PKH67-labeled Ce6@D4-EV was efficiently internalized by both activated fibroblast populations, with markedly increased intracellular accumulation at 8 h compared with 2 h (Figures 2B and 2C). Irradiation of Ce6@D4-EV-treated 10T1/2-CAFs also induced partial cell apoptosis (Figure S3N). To elucidate the molecular mechanisms underlying CAF modulation, we performed RNA-seq analysis on 10T1/2-derived CAFs after Ce6@D4-EV plus irradiation. This identified 553 significantly differentially expressed genes (DEGs) (Figure 2D). Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis revealed that downregulated genes were strongly associated with CAF maintenance and ECM remodeling pathways, including ECM-receptor interaction, focal adhesion, PI3K-Akt signaling, and particularly TGF-β signaling (Figure 2E). Several key genes involved in CAF activation and ECM deposition, such as Col12a1, Fbln5, Serpine1, Aspn, Lox, Gas2, and Ccn2, were markedly suppressed. Quantitative PCR (qPCR) validation confirmed that TGF-β stimulation (5 ng/mL) strongly induced Serpine1 and Ccn2 expression, whereas both Ce6@D4-EV plus irradiation and the TGF-β receptor kinase inhibitor (TRKI, SB-431542) significantly downregulated these genes. Increasing TGF-β concentration to 15 ng/mL partially reversed the inhibitory effect (Figure S3O). At the protein level, immunofluorescence staining demonstrated that Free-Ce6 or Ce6@D4-EV plus irradiation significantly reduced expression of the CAF activation marker α-SMA and suppressed deposition of collagen IV (Col-IV) and fibronectin (FN) in both MRC5- and 10T1/2-derived CAF models (Figures 2F and 2G; Figure S3P). Mechanistically, immunofluorescence further showed that Ce6@D4-EV plus irradiation or TRKI treatment markedly decreased nuclear translocation of p-SMAD2/3, accompanied by reduced expression of α-SMA, Col-IV, and FN (Figures 2H and 2I). These effects were partially rescued by elevated TGF-β stimulation (15 ng/mL) (Figures 2H and 2I). Collectively, these in vitro data indicate that Ce6@D4-EV-mediated PDT not only directly kills pancreatic cancer cells and induces ICD but also effectively reprograms activated CAFs and inhibits ECM production, at least in part through suppression of the canonical TGF-β/SMAD signaling axis—a central driver of desmoplasia in PDAC.55,56 MDK D4-EVs efficiently target subcutaneous and orthotopic PDAC To evaluate the tumor-targeting capability of MDK-Nb-engineered EVs, we first performed NIR fluorescence imaging. IR-783, a dye suitable for both NIR-I and NIR-II imaging due to its emission tail,57 was loaded into EVs. In subcutaneous KPC models, IR783@D4-EV exhibited significantly enhanced tumor accumulation compared with IR783@Blank-EV at 24 h post-injection, enabling clearer visualization of small tumor lesions by NIR-II imaging (Figures S4A–S4D). Ex vivo organ imaging confirmed higher tumor-associated signals in the D4-EV group. Although some fluorescence was detectable in control organs due to the high sensitivity of NIR-II detection, no significant differences were observed between groups, consistent with physiological EV clearance patterns.58,59 Immunofluorescence and IHC staining using anti-HA and anti-VHH antibodies further localized Nb-engineered EVs predominantly to tumor tissues, with only weak signals in the liver and negligible presence in other major organs (Figures S4E and S4F). For more stringent assessment of MDK-dependent specificity, we labeled D4-EV with the clinically approved NIR dye indocyanine green (ICG) and compared biodistribution in KPC-wild type (WT) versus KPC-MDK-knockout (KO) subcutaneous tumor models. ICG@D4-EV showed markedly greater intratumoral accumulation in KPC-WT mice than in MDK-KO mice across multiple time points (4, 8, 12, and 24 h) (Figures 3A and 3C). Ex vivo imaging revealed minimal off-target signals in most non-tumor organs (heart, spleen, lung, kidney, and pancreas) in WT mice, with liver fluorescence comparable to tumor levels. In contrast, MDK-KO mice displayed predominant liver accumulation (Figures 3B–3E).Figure 3MDK-D4-Nb-engineered EVs enhance MDK-dependent tumor targeting in subcutaneous and orthotopic PDAC models(A) In vivo NIR-I imaging of subcutaneous WT and MDK-KO KPC tumors after ICG@D4-EV injection.(B) Ex vivo NIR-I imaging of major organs from subcutaneous WT and MDK-KO KPC tumor-bearing mice 24 h post-ICG@D4-EV injection.(C) Quantitative fluorescence analysis of in vivo NIR-I tumor imaging in (A) (n = 3). Two-way ANOVA.(D) Quantitative fluorescence analysis of ex vivo organ imaging in (B) (n = 3). Student’s t test.(E) Quantitative fluorescence analysis of excised livers and tumors from WT versus MDK-KO KPC tumor-bearing mice. Student’s t test.(F) Ex vivo NIR-II imaging of major organs from orthotopic KPC tumor-bearing mice 24 h after injection of IR783@Blank-EV or IR783@D4-EV (n = 3). H, heart; Li, liver; S, spleen; Lu, lungs; K, kidneys; P/T, pancreas/tumor.(G) Quantitative analysis of ex vivo NIR-II organ imaging in (F) (n = 3). Student’s t test.(H) Immunofluorescence staining of orthotopic KPC tumor tissues with anti-HA and anti-VHH antibodies after IR783@D4-EV treatment (IR783@Blank-EV as control). Scale bars, 50 μm.(I) Immunofluorescence staining of major organs and tumors from IR783@D4-EV-treated mice. Scale bars, 50 μm.Data are presented as mean ± SEM. ns, no significance; ∗∗∗p < 0.001. MDK-D4-Nb-engineered EVs enhance MDK-dependent tumor targeting in subcutaneous and orthotopic PDAC models (A) In vivo NIR-I imaging of subcutaneous WT and MDK-KO KPC tumors after ICG@D4-EV injection. (B) Ex vivo NIR-I imaging of major organs from subcutaneous WT and MDK-KO KPC tumor-bearing mice 24 h post-ICG@D4-EV injection. (C) Quantitative fluorescence analysis of in vivo NIR-I tumor imaging in (A) (n = 3). Two-way ANOVA. (D) Quantitative fluorescence analysis of ex vivo organ imaging in (B) (n = 3). Student’s t test. (E) Quantitative fluorescence analysis of excised livers and tumors from WT versus MDK-KO KPC tumor-bearing mice. Student’s t test. (F) Ex vivo NIR-II imaging of major organs from orthotopic KPC tumor-bearing mice 24 h after injection of IR783@Blank-EV or IR783@D4-EV (n = 3). H, heart; Li, liver; S, spleen; Lu, lungs; K, kidneys; P/T, pancreas/tumor. (G) Quantitative analysis of ex vivo NIR-II organ imaging in (F) (n = 3). Student’s t test. (H) Immunofluorescence staining of orthotopic KPC tumor tissues with anti-HA and anti-VHH antibodies after IR783@D4-EV treatment (IR783@Blank-EV as control). Scale bars, 50 μm. (I) Immunofluorescence staining of major organs and tumors from IR783@D4-EV-treated mice. Scale bars, 50 μm. Data are presented as mean ± SEM. ns, no significance; ∗∗∗p < 0.001. In the more clinically relevant orthotopic KPC model, ex vivo NIR-II imaging at 24 h post-injection demonstrated significantly higher tumor enrichment with IR783@D4-EV compared with IR783@Blank-EV (Figures 3F and 3G). D4 functionalization notably improved the spatial resolution and accuracy of PDAC lesion detection within the pancreas. Immunofluorescence staining confirmed the presence of Nb-engineered EVs in orthotopic tumor tissues, alongside detectable liver signals consistent with normal EV clearance (Figures 3H and 3I). Collectively, these multi-modal imaging and histological data demonstrate that MDK-Nb surface engineering confers selective and efficient accumulation of EVs in MDK-enriched PDAC TME in both subcutaneous and orthotopic settings, while reducing non-specific distribution. This targeting strategy markedly enhances delivery efficiency for both imaging probes and therapeutic payloads. Ce6@D4-EV-mediated targeted PDT suppresses tumor growth in vivo by inducing ICD and remodeling CAFs Prior to in vivo antitumor efficacy studies, dark toxicity was excluded by administering Free-Ce6, Ce6@Blank-EV, or Ce6@D4-EV without light irradiation in subcutaneous KPC tumor-bearing mice (Figure S5A). No significant differences in tumor growth or body weight were observed compared with PBS control (Figures S5B and S5C), confirming that the observed effects require light activation. Therapeutic efficacy was then evaluated using a schedule of two treatment cycles combining intravenous injection and localized light irradiation (Figure 4A). In subcutaneous KPC tumors, Free-Ce6 or Ce6@Blank-EV plus irradiation produced no meaningful tumor growth inhibition or survival benefit compared with PBS control. This limited efficacy is likely attributable to insufficient intratumoral accumulation via the enhanced permeability and retention effect and consequent inadequate ROS generation. In sharp contrast, Ce6@D4-EV plus irradiation resulted in pronounced tumor growth suppression and significantly prolonged animal survival (Figures 4B and 4C), highlighting the critical contribution of MDK-Nb-mediated active targeting.Figure 4Ce6@D4-EV-mediated PDT inhibits tumor progression and reprograms stromal and vascular features in MDK-high tumors in vivo(A) Schematic of subcutaneous KPC tumor model and treatment schedule (two cycles of injection and irradiation).(B) Tumor growth curves of subcutaneous KPC tumors across treatment groups (n = 5). Two-way ANOVA.(C) Kaplan-Meier survival curves of KPC tumor-bearing mice (n = 5). Log rank (Mantel-Cox) test.(D) Immunofluorescence detection of immunogenic cell death markers (surface CRT and ERp57) in tumor tissues. Scale bars, 50 μm.(E) Immunofluorescence staining of α-SMA to assess CAF abundance in tumor tissues. Scale bars, 50 μm.(F) Co-immunofluorescence staining of p-SMAD2/3 (red) and α-SMA (green) in KPC tumor tissues. Scale bars, 10 μm.(G) Immunofluorescence staining of FN (red) and Masson’s trichrome staining of collagen (blue) in tumor tissues. Scale bars: 25 μm (top) and 50 μm (bottom).(H) Immunofluorescence staining of CDH5 (green) in CD31+ vessels (red). Scale bars, 50 μm.(I) Vascular perfusion assessed by FITC-lectin (green) and CD31 (red) co-staining. Scale bars: 25 μm (top) and 50 μm (bottom).(J) Hypoxia detection in tumor tissues by pimonidazole staining. Scale bars, 50 μm.(K and L) Immunofluorescence staining of ICAM-1 (K) and VCAM-1 (L) (green) on CD31+ vessels (red). Scale bars: 25/50 μm (top) and 50/100 μm (bottom).(M) Tumor growth curves of subcutaneous Panc02-MDK tumors (n = 5). Two-way ANOVA.(N) Survival curves of Panc02-MDK tumor-bearing mice (n = 5). Log rank (Mantel-Cox) test.(O) Tumor growth curves of subcutaneous 4T1 tumors (n = 5). Two-way ANOVA.(P) Survival curves of 4T1 tumor-bearing mice (n = 5). Log rank (Mantel-Cox) test.(Q) Tumor growth curves of subcutaneous Hepa1-6 tumors (n = 5). Two-way ANOVA.(R) Survival curves of Hepa1-6 tumor-bearing mice (n = 5). Log rank (Mantel-Cox) test.Red arrows in (B, M, and O) indicate tumor recurrence after treatment cessation. Data are presented as mean ± SEM. ns, no significance; ∗∗p < 0.01, ∗∗∗p < 0.001. Ce6@D4-EV-mediated PDT inhibits tumor progression and reprograms stromal and vascular features in MDK-high tumors in vivo (A) Schematic of subcutaneous KPC tumor model and treatment schedule (two cycles of injection and irradiation). (B) Tumor growth curves of subcutaneous KPC tumors across treatment groups (n = 5). Two-way ANOVA. (C) Kaplan-Meier survival curves of KPC tumor-bearing mice (n = 5). Log rank (Mantel-Cox) test. (D) Immunofluorescence detection of immunogenic cell death markers (surface CRT and ERp57) in tumor tissues. Scale bars, 50 μm. (E) Immunofluorescence staining of α-SMA to assess CAF abundance in tumor tissues. Scale bars, 50 μm. (F) Co-immunofluorescence staining of p-SMAD2/3 (red) and α-SMA (green) in KPC tumor tissues. Scale bars, 10 μm. (G) Immunofluorescence staining of FN (red) and Masson’s trichrome staining of collagen (blue) in tumor tissues. Scale bars: 25 μm (top) and 50 μm (bottom). (H) Immunofluorescence staining of CDH5 (green) in CD31+ vessels (red). Scale bars, 50 μm. (I) Vascular perfusion assessed by FITC-lectin (green) and CD31 (red) co-staining. Scale bars: 25 μm (top) and 50 μm (bottom). (J) Hypoxia detection in tumor tissues by pimonidazole staining. Scale bars, 50 μm. (K and L) Immunofluorescence staining of ICAM-1 (K) and VCAM-1 (L) (green) on CD31+ vessels (red). Scale bars: 25/50 μm (top) and 50/100 μm (bottom). (M) Tumor growth curves of subcutaneous Panc02-MDK tumors (n = 5). Two-way ANOVA. (N) Survival curves of Panc02-MDK tumor-bearing mice (n = 5). Log rank (Mantel-Cox) test. (O) Tumor growth curves of subcutaneous 4T1 tumors (n = 5). Two-way ANOVA. (P) Survival curves of 4T1 tumor-bearing mice (n = 5). Log rank (Mantel-Cox) test. (Q) Tumor growth curves of subcutaneous Hepa1-6 tumors (n = 5). Two-way ANOVA. (R) Survival curves of Hepa1-6 tumor-bearing mice (n = 5). Log rank (Mantel-Cox) test. Red arrows in (B, M, and O) indicate tumor recurrence after treatment cessation. Data are presented as mean ± SEM. ns, no significance; ∗∗p < 0.01, ∗∗∗p < 0.001. To elucidate the underlying mechanisms, tumor tissues were comprehensively analyzed. Ce6@D4-EV plus irradiation induced prominent surface exposure of ICD biomarkers CRT and ERp57 on tumor cells (Figure 4D; Figure S5D), consistent with in vitro observations and supporting the potential for immune activation. Given the central role of CAFs in PDAC desmoplasia, we examined stromal changes. Expression of the activated CAF marker α-SMA was significantly decreased in the Ce6@D4-EV plus irradiation group compared with controls (Figure 4E; Figure S5E). Co-immunofluorescence staining further revealed reduced nuclear localization of p-SMAD2/3 within α-SMA+ CAFs (Figure 4F; Figure S5F), indicating suppression of TGF-β/SMAD signaling at the tissue level. Concurrently, deposition of ECM proteins including FN, collagen I (Col-I), and laminin was markedly reduced, and Masson’s trichrome staining confirmed decreased collagen fiber content (Figure 4G; Figures S5G and S5H). These changes demonstrate effective CAF reprogramming and attenuation of fibrotic ECM production. The reduction in stromal density and CAF activity was accompanied by improvements in vascular architecture and function. Expression of VE-cadherin (CDH5), an important mediator of endothelial adherens junctions,60 was significantly upregulated in CD31+ blood vessels (Figure 4H; Figure S5I). Functional perfusion, assessed by fluorescein isothiocyanate (FITC)-lectin staining, was substantially enhanced, while tumor hypoxia (detected by pimonidazole) was markedly alleviated compared with control groups (Figures 4I and 4J; Figures S5J and S5K). In addition, the endothelial adhesion molecules ICAM-1 and VCAM-1, which facilitate immune cell recruitment,61 were significantly increased on tumor vasculature (Figures 4K and 4L; Figures S5L and S5M). These vascular improvements likely result from decreased physical compression by remodeled stroma, restored vessel integrity, and reduced hypoxia. Safety evaluation at the end of treatment showed no significant changes in body weight, hematological parameters, or histopathological appearance of major organs (Figures S5N–S5P), supporting a favorable therapeutic index. Given the pronounced responses in both tumor cells and CAFs, we next examined the intratumoral cellular distribution of D4-EV in vivo. To directly visualize cellular distribution, HA-FITC-labeled D4-EV was injected and single-cell flow cytometry performed on dissociated tumors and major organs (Figures S5Q and S5R). In KPC-WT tumors, strong enrichment occurred in both EpCAM+ tumor cells and FAP+ CAFs with comparable fluorescence intensities, while liver uptake was significantly lower; negligible signals were detected in non-tumor organs (heart, spleen, lung, and kidney). In KPC-MDK-KO tumors, no enrichment was seen in tumor cells or CAFs, with predominant accumulation restricted to liver cells. These data confirm that D4-EV preferentially accumulates in MDK-enriched tumor tissues and is efficiently internalized by both cancer cells and CAFs in vivo. To assess generalizability, the platform was tested in additional MDK-high models. In an independent Panc02-MDK subcutaneous model in immunocompetent mice, Ce6@D4-EV plus irradiation significantly inhibited tumor growth and extended survival (Figures 4M and 4N). Similar potent efficacy was observed in the desmoplastic 4T1 breast cancer model (Figures 4O and 4P) and in Hepa1-6 HCC tumors, which also express high MDK levels (Figures 4Q and 4R; Figures S3A and S3B). In summary, MDK-targeted Ce6@D4-EV-mediated PDT achieves robust antitumor efficacy across multiple MDK-high solid tumor models through simultaneous induction of ICD in cancer cells and stromal remodeling via CAF reprogramming, ECM reduction, vascular normalization, and hypoxia relief. MDK-targeted Ce6@D4-EV-mediated PDT reprograms the TME Given the induction of ICD, improved vascular function, and reduced hypoxia, we hypothesized that Ce6@D4-EV-mediated PDT could reprogram the immunosuppressive TME. Immune profiling was performed in the subcutaneous Panc02-MDK model in immunocompetent C57BL/6 mice. Flow cytometry revealed that Ce6@D4-EV plus irradiation significantly increased the proportions of mature dendritic cells (CD45+ CD11c+ MHC II+ CD80+ CD86+), CD4+ T cells, CD8+ T cells, and M1-polarized macrophages (CD45+ F4/80+ CD86+) compared with control groups (Figures 5A–5D). M2 macrophage populations (CD45+ F4/80+ CD206+) remained unchanged (Figure 5E), resulting in a net shift toward an immunostimulatory TME. Comparable immune activation was observed in KPC tumors, including increased infiltration of dendritic cells, CD4+ T cells, and CD8+ T cells, as well as an elevated M1/M2 macrophage ratio (Figures S6A–S6G). Expression of Iba-1, a marker associated with phagocytic macrophages, was significantly upregulated in F4/80+ cells after treatment (Figure S6H), suggesting enhanced phagocytic activity.Figure 5Ce6@D4-EV-mediated targeted PDT enhances immune infiltration and activates STING-associated TME reprogramming in subcutaneous PDAC tumors(A) Flow cytometric quantification of mature dendritic cells in subcutaneous Panc02-MDK tumors after indicated treatments (n = 5). One-way ANOVA.(B–E) Quantification of CD4+ T cells (B), CD8+ T cells (C), M1 macrophages (D), and M2 macrophages (E) in Panc02-MDK tumors (n = 5). One-way ANOVA.(F) KEGG pathway enrichment analysis of up-regulated genes.(G) Co-immunofluorescence staining and quantification of p-IRF3 (red) in F4/80+ TAMs (green). Scale bars: 10 μm (inset) and 20 μm. One-way ANOVA.(H) qPCR analysis of ISGs in tumor tissues (n = 5). Student’s t test.Data are presented as mean ± SEM. ns, no significance; ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001. Ce6@D4-EV-mediated targeted PDT enhances immune infiltration and activates STING-associated TME reprogramming in subcutaneous PDAC tumors (A) Flow cytometric quantification of mature dendritic cells in subcutaneous Panc02-MDK tumors after indicated treatments (n = 5). One-way ANOVA. (B–E) Quantification of CD4+ T cells (B), CD8+ T cells (C), M1 macrophages (D), and M2 macrophages (E) in Panc02-MDK tumors (n = 5). One-way ANOVA. (F) KEGG pathway enrichment analysis of up-regulated genes. (G) Co-immunofluorescence staining and quantification of p-IRF3 (red) in F4/80+ TAMs (green). Scale bars: 10 μm (inset) and 20 μm. One-way ANOVA. (H) qPCR analysis of ISGs in tumor tissues (n = 5). Student’s t test. Data are presented as mean ± SEM. ns, no significance; ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001. RNA-seq analysis of KPC tumor tissues post-treatment showed clear separation of gene expression profiles (Figure S6I). KEGG pathway enrichment of upregulated genes highlighted activation of chemokine signaling, cytokine-cytokine receptor interaction, oxidative phosphorylation, and innate immune pathways, including the cytosolic DNA-sensing and NOD-like receptor signaling pathways (Figure 5F). Downregulated pathways included MAPK signaling (Figure S6J), consistent with reduced cancer cell proliferation and CAF activity. The enrichment of the cytosolic DNA-sensing pathway prompted investigation of cGAS-STING signaling. Nuclear p-IRF3 staining, a key indicator of STING activation, was significantly increased in F4/80+ TAMs in both KPC and Panc02-MDK tumors after Ce6@D4-EV plus irradiation (Figure 5G; Figure S6K). qPCR confirmed upregulation of multiple interferon-stimulated genes (ISGs), including IFNB1, CCL5, CXCL9, CXCL10, and ISG15 (Figure 5H). Collectively, these findings demonstrate that Ce6@D4-EV-mediated PDT effectively reprograms the immunosuppressive TME of PDAC by enhancing immune cell infiltration and activation, promoting M1 macrophage polarization, and triggering cGAS-STING signaling in TAMs, likely in response to dsDNA released from PDT-induced dying tumor cells. STING activation in the TAMs is essential for antitumor effects of Ce6@D4-EV-mediated targeted PDT To investigate whether dsDNA released from ICD activates the cGAS-STING pathway in TAMs, we isolated primary peritoneal macrophages and stimulated them with supernatants from variously treated KPC cells. Direct incubation with D4-EV alone did not activate STING signaling. In contrast, supernatants from Ce6@D4-EV plus irradiation-treated KPC cells strongly induced cGAS, p-TBK1, and p-IRF3 expression while decreasing total STING protein levels (Figures 6A and 6B). Downstream ISGs (Ccl5, Cxcl10, Ifnb1, and Tnf) were correspondingly upregulated (Figure 6C; Figure S7A). Pretreatment of PDT supernatants with DNase I almost completely abolished STING pathway activation and ISG induction (Figures 6D and 6E; Figure S7B), confirming that tumor-derived dsDNA is the critical mediator. Pharmacological inhibition with the STING antagonist H151 similarly reversed both pathway activation and ISG expression (Figures 6F and 6G; Figure S7C). Nuclear translocation of p-IRF3 was markedly enhanced by PDT supernatants and blocked by H151 (Figure 6H).Figure 6Ce6@D4-EV-mediated targeted PDT activates macrophage cGAS-STING signaling to drive antitumor efficacy in PDAC(A) Western blot of STING pathway components in macrophages treated with D4-EV or supernatants from Ce6@D4-EV plus irradiation-treated KPC cells.(B) Western blot of STING pathway activation in macrophages incubated with supernatants from KPC cells under indicated conditions.(C) qPCR analysis of cGAS-STING downstream genes (Ccl5 and Cxcl10) in primary peritoneal macrophages (n = 3). One-way ANOVA.(D) Western blot showing abolition of STING activation by DNase I pretreatment of PDT-treated KPC supernatants.(E) qPCR analysis of downstream genes after DNase I treatment (n = 3). One-way ANOVA.(F) Western blot showing reversal of STING activation by H151 in macrophages exposed to PDT-treated KPC supernatants.(G) qPCR analysis of downstream genes after H151 treatment (n = 3). One-way ANOVA.(H) Immunofluorescence staining and quantification of p-IRF3 in primary peritoneal macrophages exposed to KPC supernatants with indicated treatments (n = 3). Scale bars, 45 μm.(I) Western blot of cGAS-STING signaling in WT and STING−/− BMDMs incubated with PDT-conditioned media ± C176 collected from KPC cells.(J) qPCR quantification of downstream genes (Ccl5, Cxcl10, Ifnb1, and Tnf) in WT and STING−/− BMDMs in (I) (n = 3). One-way ANOVA.(K) Schematic of KPC subcutaneous tumor model and treatment schedule.(L–N) Tumor growth curves of subcutaneous KPC tumors treated with Ce6@D4-EV plus irradiation, with or without macrophage depletion (CLO) (L), H151 (L), C176 (N), or in STING−/− mice (M) (n = 5). Two-way ANOVA. Red arrows indicate tumor recurrence after treatment cessation.Data are presented as mean ± SEM. ns, no significance; ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001. Ce6@D4-EV-mediated targeted PDT activates macrophage cGAS-STING signaling to drive antitumor efficacy in PDAC (A) Western blot of STING pathway components in macrophages treated with D4-EV or supernatants from Ce6@D4-EV plus irradiation-treated KPC cells. (B) Western blot of STING pathway activation in macrophages incubated with supernatants from KPC cells under indicated conditions. (C) qPCR analysis of cGAS-STING downstream genes (Ccl5 and Cxcl10) in primary peritoneal macrophages (n = 3). One-way ANOVA. (D) Western blot showing abolition of STING activation by DNase I pretreatment of PDT-treated KPC supernatants. (E) qPCR analysis of downstream genes after DNase I treatment (n = 3). One-way ANOVA. (F) Western blot showing reversal of STING activation by H151 in macrophages exposed to PDT-treated KPC supernatants. (G) qPCR analysis of downstream genes after H151 treatment (n = 3). One-way ANOVA. (H) Immunofluorescence staining and quantification of p-IRF3 in primary peritoneal macrophages exposed to KPC supernatants with indicated treatments (n = 3). Scale bars, 45 μm. (I) Western blot of cGAS-STING signaling in WT and STING−/− BMDMs incubated with PDT-conditioned media ± C176 collected from KPC cells. (J) qPCR quantification of downstream genes (Ccl5, Cxcl10, Ifnb1, and Tnf) in WT and STING−/− BMDMs in (I) (n = 3). One-way ANOVA. (K) Schematic of KPC subcutaneous tumor model and treatment schedule. (L–N) Tumor growth curves of subcutaneous KPC tumors treated with Ce6@D4-EV plus irradiation, with or without macrophage depletion (CLO) (L), H151 (L), C176 (N), or in STING−/− mice (M) (n = 5). Two-way ANOVA. Red arrows indicate tumor recurrence after treatment cessation. Data are presented as mean ± SEM. ns, no significance; ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001. These observations were extended to additional MDK-high cell lines (human PANC-1, MDK-overexpressing Panc02, and 4T1). Conditioned media from PDT-treated cells activated macrophage STING signaling and ISG expression in a strictly STING-dependent manner, with effects fully abrogated by H151 (Figures S7D–S7I). Genetic validation using bone marrow-derived macrophages (BMDMs) from WT and STING−/− mice showed robust p-IRF3 activation and ISG induction only in WT cells; the response was absent in STING−/− BMDMs and strongly suppressed by the covalent STING inhibitor C176 (Figures 6I and 6J). In vivo functional studies further confirmed the essential role of TAM STING signaling. Macrophage depletion using clodronate liposomes (CLOs) or systemic STING inhibition with H151 significantly diminished the tumor growth suppression achieved by Ce6@D4-EV plus irradiation in KPC tumors (Figures 6K and 6L). Therapeutic efficacy was also markedly reduced in STING−/− mice compared with WT hosts (Figure 6M). Similarly, pharmacological blockade with C176 abrogated antitumor activity without affecting tumor growth when used alone (Figure 6N). Collectively, these in vitro mechanistic and in vivo genetic/pharmacological data provide strong evidence that activation of the cGAS-STING pathway in TAMs is a critical mediator of the antitumor efficacy of Ce6@D4-EV-mediated targeted PDT. Ce6@D4-EV-mediated targeted PDT enhances the antitumor efficacy of immunotherapy and chemotherapy in preclinical PDAC The immunosuppressive TME of PDAC, characterized by poor T cell infiltration and dominant M2-like macrophages, limits responses to ICIs such as anti-PD-1 and anti-CD47. Given that Ce6@D4-EV-mediated PDT enhances immune cell infiltration and promotes M1 polarization, we evaluated its potential to sensitize tumors to ICI therapy. In subcutaneous KPC tumors, anti-PD-1 antibody or anti-CD47 nanobody monotherapy showed negligible antitumor activity, consistent with the known resistance of PDAC to single-agent immunotherapy.62,63 Ce6@D4-EV plus irradiation alone produced initial tumor control but allowed gradual regrowth after treatment cessation. In contrast, the combination of Ce6@D4-EV plus irradiation with either PD-1 or CD47 blockade achieved sustained tumor suppression, increased tumor cell apoptosis, and significantly prolonged survival (Figures 7A–7C; Figures S8A and S8B). No additional systemic toxicity was observed across groups (Figures S8C–S8E).Figure 7MDK-targeted Ce6@D4-EV-mediated PDT enhances the antitumor efficacy of immunotherapy and chemotherapy across multiple preclinical tumor models(A) Schematic of KPC subcutaneous tumor model and combinatorial treatment schedule.(B) Tumor growth curves of KPC tumors treated with Ce6@D4-EV plus irradiation ± anti-PD-1 or anti-CD47 (n = 5). Two-way ANOVA.(C) Survival curves of KPC tumor-bearing mice receiving PDT plus ICI combinations (n = 5). Log rank (Mantel-Cox) test.(D) Tumor growth curves of KPC tumors treated with Ce6@D4-EV plus irradiation ± Gem (n = 5). Two-way ANOVA.(E) Survival curves for PDT plus Gem combination (n = 5). Log rank (Mantel-Cox) test.(F) Tumor growth curves of KPC-OVA tumors treated with Ce6@D4-EV plus irradiation ± OT-1 CD8+ T cell transfer (n = 5). Two-way ANOVA.(G) Survival curves for PDT plus adoptive T cell therapy (n = 5). Log rank (Mantel-Cox) test.(H) Quantification of infiltrated CD45.1+ OT-1 cells in KPC-OVA tumors ± PDT (n = 5). Student’s t test.(I) Tumor growth curves of Panc02-MDK tumors treated with Ce6@D4-EV plus irradiation ± anti-PD-1 or Gem (n = 5). Two-way ANOVA.(J) Survival curves for Panc02-MDK tumors receiving PDT plus PD-1 or Gem (n = 5). Log rank (Mantel-Cox) test.(K) Tumor growth curves of orthotopic 4T1 tumors treated with Ce6@D4-EV plus irradiation ± anti-PD-1 or PTX (n = 5). Two-way ANOVA.(L) Survival curves for orthotopic 4T1 tumors receiving PDT plus PD-1 or PTX (n = 5). Log rank (Mantel-Cox) test.(M) Tumor growth curves of subcutaneous Hepa1-6 tumors treated with Ce6@D4-EV plus irradiation ± anti-PD-1 (n = 5). Two-way ANOVA.(N) Survival curves for Hepa1-6 tumors receiving PDT plus PD-1 blockade (n = 5). Log rank (Mantel-Cox) test.Red arrows indicate tumor recurrence after cessation of Ce6@D4-EV plus light treatment. Data are presented as mean ± SEM. ns, no significance; ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001. MDK-targeted Ce6@D4-EV-mediated PDT enhances the antitumor efficacy of immunotherapy and chemotherapy across multiple preclinical tumor models (A) Schematic of KPC subcutaneous tumor model and combinatorial treatment schedule. (B) Tumor growth curves of KPC tumors treated with Ce6@D4-EV plus irradiation ± anti-PD-1 or anti-CD47 (n = 5). Two-way ANOVA. (C) Survival curves of KPC tumor-bearing mice receiving PDT plus ICI combinations (n = 5). Log rank (Mantel-Cox) test. (D) Tumor growth curves of KPC tumors treated with Ce6@D4-EV plus irradiation ± Gem (n = 5). Two-way ANOVA. (E) Survival curves for PDT plus Gem combination (n = 5). Log rank (Mantel-Cox) test. (F) Tumor growth curves of KPC-OVA tumors treated with Ce6@D4-EV plus irradiation ± OT-1 CD8+ T cell transfer (n = 5). Two-way ANOVA. (G) Survival curves for PDT plus adoptive T cell therapy (n = 5). Log rank (Mantel-Cox) test. (H) Quantification of infiltrated CD45.1+ OT-1 cells in KPC-OVA tumors ± PDT (n = 5). Student’s t test. (I) Tumor growth curves of Panc02-MDK tumors treated with Ce6@D4-EV plus irradiation ± anti-PD-1 or Gem (n = 5). Two-way ANOVA. (J) Survival curves for Panc02-MDK tumors receiving PDT plus PD-1 or Gem (n = 5). Log rank (Mantel-Cox) test. (K) Tumor growth curves of orthotopic 4T1 tumors treated with Ce6@D4-EV plus irradiation ± anti-PD-1 or PTX (n = 5). Two-way ANOVA. (L) Survival curves for orthotopic 4T1 tumors receiving PDT plus PD-1 or PTX (n = 5). Log rank (Mantel-Cox) test. (M) Tumor growth curves of subcutaneous Hepa1-6 tumors treated with Ce6@D4-EV plus irradiation ± anti-PD-1 (n = 5). Two-way ANOVA. (N) Survival curves for Hepa1-6 tumors receiving PDT plus PD-1 blockade (n = 5). Log rank (Mantel-Cox) test. Red arrows indicate tumor recurrence after cessation of Ce6@D4-EV plus light treatment. Data are presented as mean ± SEM. ns, no significance; ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001. Chemotherapy with Gem remains a cornerstone of PDAC treatment, yet its efficacy is often limited by poor penetration through dense stroma. We, therefore, tested combination with low-dose Gem (25 mg/kg), a dose that alone produced no tumor growth inhibition. While Ce6@D4-EV plus irradiation delayed growth, tumors eventually regrew. The combination regimen induced durable tumor control, enhanced apoptosis and proliferation inhibition, and significantly extended survival compared with PDT alone (Figures 7D and 7E; Figure S8F). Importantly, this improved efficacy occurred without observable increases in body weight loss, hematological abnormalities, or organ toxicity (Figures S8G–S8I). ACT has shown limited success in solid tumors such as PDAC due to poor infiltration and immunosuppressive signals. To test whether stromal and immune remodeling by PDT could enhance ACT, we engineered KPC-OVA cells and used OVA-specific OT-1 CD8+ T cells (Figure S8J). OT-1 transfer alone (1 × 106 cells per injection) had no antitumor effect. However, when combined with Ce6@D4-EV plus irradiation, the regimen produced durable tumor suppression, increased apoptosis, and prolonged survival (Figures 7F and 7G; Figure S8K). Flow cytometric analysis confirmed significantly higher infiltration of CD45.1+ OT-1 cells into tumors pretreated with PDT compared with OT-1 transfer alone (Figure 7H). No systemic toxicity was detected (Figures S8L–S8N). These improvements likely stem from PDT-induced cytokine/chemokine production, enhanced vascular perfusion, and upregulated endothelial ICAM-1/VCAM-1 expression, which collectively facilitate recruitment and retention of transferred T cells. To validate the broader applicability of these combination strategies, we extended testing to additional models. In the Panc02-MDK model, combination of Ce6@D4-EV plus irradiation with either Gem or anti-PD-1 resulted in sustained tumor control and prolonged survival (Figures 7I and 7J). In orthotopic 4T1 breast tumors with pronounced desmoplasia, the platform significantly enhanced the efficacy of PTX or anti-PD-1 blockade (Figures 7K and 7L). Similar synergistic effects were observed in the Hepa1-6 HCC model, where anti-PD-1 monotherapy was ineffective but produced long-term tumor control when combined with Ce6@D4-EV-mediated PDT (Figures 7M and 7N). Across four independent preclinical models, these results demonstrate that Ce6@D4-EV-mediated targeted PDT can effectively sensitize MDK-high solid tumors to multiple standard therapies, including ICIs, chemotherapy, and adoptive T cell transfer, by overcoming stromal barriers and reprogramming the TME into a more immunopermissive state. Long-term safety and immunogenicity of Ce6@D4-EV under repeated dosing To evaluate long-term safety and potential immunogenicity, healthy C57BL/6 mice received six intravenous administrations of PBS, Free-Ce6, D4-EV, or Ce6@D4-EV over 19 days, followed by observation until day 28 (Figure S9A). Body weights remained stable across all groups with no significant differences (Figure S9B). At study endpoint, hematological and serum biochemical parameters were within normal ranges, and histopathological examination of major organs showed no detectable abnormalities (Figures S9C and S9D). Serial ELISA measurements revealed that anti-EV IgG levels in D4-EV and Ce6@D4-EV groups were comparable to PBS controls or only mildly elevated, with signals returning to baseline upon sample dilution (Figures S9E and S9F). Anti-HA tag IgG responses remained negligible in all groups (Figure S9G). These data indicate that repeated dosing of Ce6@D4-EV exhibits a favorable safety profile and low immunogenicity, supporting its potential for further translational development. Discussion The dense fibrotic stroma and high IFP characteristic of PDAC cause vessel collapse, poor perfusion, hypoxia, inadequate drug delivery, and limited immune cell infiltration, thereby generating an immunosuppressive TME that fuels tumor progression and resistance to chemotherapy and immunotherapy. This study constructed MDK-targeted EV nanoplatforms to execute PDT in preclinical PDAC models, achieving extensive TME reprogramming and substantially improved antitumor efficacy when combined with standard therapies. MDK acts as a multifunctional cytokine that engages multiple receptors to drive tumor growth, metastasis, therapeutic resistance, and immunosuppression.40,41,42,43,64 Pan-cancer transcriptomic data indicate MDK upregulation across more than 20 tumor types, considerably broader than VEGFA, while integrated single-cell analyses and IHC staining reveal consistent TME enrichment, with predominant expression by malignant cells and contributions from fibroblasts and endothelial cells. These features establish MDK as a valuable pan-cancer biomarker for imaging and targeted therapy. Similar to the clinical success of VEGFA-targeted agents such as bevacizumab in colorectal cancer, non-small cell lung cancer, glioblastoma, and HCC,65 MDK-directed targeting is feasible for enhancing drug homing to MDK-positive tumors, including PDAC and HCC. Prior MDK-Nb-modified nanoparticles demonstrated selective PDAC accumulation and tumor suppression.36 Here, surface engineering of MDK-D4 nanobodies onto EVs conferred high tumor selectivity, as evidenced by NIR-I and NIR-II imaging in both subcutaneous and orthotopic PDAC models featuring high MDK levels and dense stroma. Unlike ligands targeting cell-surface antigens (e.g., HER2, c-Met, or cadherin 17 [CDH17]), MDK is a secreted protein distributed in the extracellular tumor microenvironment; therefore, MDK-targeted agents can efficiently home to the TME and be primarily taken up by tumor cells and CAFs.52,66,67,68,69,70 EVs possess inherent advantages over synthetic nanoparticles, including superior biocompatibility, low immunogenicity, high stability, robust drug-loading capacity, and the ability to penetrate tumors and cross biological barriers.24,66 However, unmodified EVs exhibit non-specific accumulation in organs such as the liver, lung, spleen, and pancreas.71 Genetic display of targeting moieties overcomes this limitation, enhancing tumor selectivity and therapeutic efficiency. In this work, both MDK D4-EV and Ce6@D4-EV demonstrated pronounced TME-targeting profiles in MDK-positive KPC tumors. Light irradiation of Ce6@D4-EV triggered robust antitumor effects through direct cancer cell killing via PDT, functional inactivation of CAFs (evidenced by reduced α-SMA and p-SMAD expression), and decreased synthesis of ECM proteins (FN, Col-I, and laminin). These changes softened the stromal matrix, reperfused collapsed tumor vessels, and alleviated hypoxia. Hypoxia is a key driver of CAF activation, immunosuppression, low T cell infiltration, and therapeutic resistance in PDAC.72,73 Improved oxygenation upregulated vascular adhesion molecules (ICAM-1 and VCAM-1), promoted M1 macrophage polarization, increased cytokine and chemokine production for immune recruitment, and enhanced infiltration of dendritic cells and T lymphocytes. Consequently, Ce6@D4-EV-mediated targeted PDT sensitized PDAC to ICIs, augmented the efficacy of Gem and PTX, and facilitated recruitment of adoptively transferred T cells. A central mechanistic insight is the activation of the cGAS-STING pathway in TAMs following PDT. Dying cancer cells undergoing ICD released endogenous dsDNA, which, after phagocytosis by TAMs, stimulated cGAS-STING signaling and upregulated a panel of ISGs. This response enhanced antigen presentation, type I IFN production, and antitumor immunity while reversing TME immunosuppression.74,75 Supernatants from PDT-treated cancer cells activated cGAS-STING in isolated macrophages; pretreatment with DNase I abolished the effect, confirming dsDNA dependence. Macrophage depletion or systemic STING inactivation significantly attenuated the therapeutic benefits, underscoring the pathway’s importance. In contrast to exogenous STING agonists, which have shown limited clinical success due to poor tumor targeting, instability, potential toxicity to T cells, and unintended activation of immunosuppressive B cells,75,76 this endogenous, localized dsDNA-driven mechanism provides a more physiological route to invigorate innate and adaptive immunity within the TME.77,78,79,80 By simultaneously targeting cancer cells and activated CAFs, Ce6@D4-EV disrupts their mutual crosstalk, normalizes the desmoplastic stroma, reprograms the immune landscape, and creates a more permissive environment for conventional therapies. Overall, this MDK-Nb-engineered EV platform enables precise delivery of Ce6 for PDT in desmoplastic tumors, reduces ECM deposition, alleviates hypoxia and stiffness, activates TAM cGAS-STING signaling via dsDNA release, and markedly potentiates chemotherapy, ICIs, and adoptive T cell therapy. It offers an alternative strategy for engaging the STING pathway through localized, physiological signals and highlights the translational potential of EV-based PIT for MDK-positive solid tumors. Limitations of the study Several limitations of this study should be acknowledged. First, although efficient uptake of MDK-engineered EVs by tumor cells and CAFs was observed, the precise mechanisms underlying EV internalization remain to be elucidated. Second, this work primarily focused on STING pathway activation in TAMs due to their abundance and strong p-IRF3 induction; the contribution of other immune populations, particularly dendritic cells, warrants further investigation. Third, the EVs used in this study were derived from HEK-293 cells, which may pose challenges related to immunogenicity and translational scalability. Future studies should consider clinically suitable EV sources, such as human mesenchymal stem cells,81,82,83,84 together with good manufacturing practice-compliant production systems.85 In addition, while Ce6@D4-EV-mediated PDT was effective in orthotopic breast cancer models, technical barriers limit its implementation in murine orthotopic PDAC models; advances in minimally invasive, image-guided light-delivery technologies may enable future evaluation in this setting. Finally, given the broad upregulation of MDK across cancer types, whether the mechanisms identified here extend to other MDK-positive tumors or the clinical context remains to be determined. Resource availability Lead contact Requests for further information and resources should be directed to and will be fulfilled by the lead contact, Yufeng Yuan (yuanyf1971@whu.edu.cn). Materials availability This study did not generate new unique reagents. Data and code availability •High-throughput RNA-seq data generated in this study have been deposited in the Gene Expression Omnibus (GEO) database and are publicly available under accession numbers GEO: GSE318327 and GSE318328.•This paper does not report original code.•Any additional information required to reanalyze the data reported in this work paper is available from the lead contact upon request. •High-throughput RNA-seq data generated in this study have been deposited in the Gene Expression Omnibus (GEO) database and are publicly available under accession numbers GEO: GSE318327 and GSE318328. High-throughput RNA-seq data generated in this study have been deposited in the Gene Expression Omnibus (GEO) database and are publicly available under accession numbers GEO: GSE318327 and GSE318328. •This paper does not report original code. This paper does not report original code. •Any additional information required to reanalyze the data reported in this work paper is available from the lead contact upon request. Any additional information required to reanalyze the data reported in this work paper is available from the lead contact upon request. Acknowledgments This work was supported by grants from the 10.13039/501100001809National Natural Science Foundation of China (82373775, 82304518, and 82400727), the Shenzhen Medical Research Fund (D2403013 and B2302051), the International Science and Technology Cooperation Program of Shenzhen (GJHZ20240218114508015), the 10.13039/501100003819Natural Science Foundation of Hubei Province (2026AFB181), the Medical Sci-Tech Innovation Platform of Zhongnan Hospital of Wuhan University (PTXM2025021), the Program of Excellent Doctor of Zhongnan Hospital of 10.13039/501100007046Wuhan University (ZNYB20240015), the 10.13039/501100002858China Postdoctoral Science Foundation (GZC20241274 and 2025M772195), and the Postdoctoral Project of Hubei Province (2004HBBHJD058, 2025HBBSHCXB047). We are grateful to Professor Ruth Ganss from the University of Western Australia for kindly providing the OT-1 mice. Author contributions C.Q., J. Liu, J.C., M.T., and Y.C. performed the majority of experiments, analyzed the data, and contributed equally to this work. Y.Y., J.W., Z.L., and R.H. conceived and designed the experiments. P.X., R.Z., and J.M. conducted the database searches and bioinformatics analysis. H.W., H.Y., F.R., J. Li, B.H., and Z.Y. provided technical advice and analyzed the data. J. Liu, J.C., M.T., and Y.C. assisted with in vitro experiments and animal experiments. C.Q., Z.L., and Y.Y. wrote the manuscript and designed the figures. All authors read and approved the final paper. Declaration of interests The authors declare no competing interests. STAR★Methods Key resources table REAGENT or RESOURCESOURCEIDENTIFIERAntibodiesAnti-CD9AbcamCat# ab236630; RRID: AB_2922400Anti-CD81AbcamCat# ab79559; RRID: AB_1603682Anti-HSP70AbcamCat# ab181606; RRID: AB_2910093Anti-TSG101AbcamCat# ab125011; RRID: AB_10974262Anti-CalnexinAbcamCat# ab133615; RRID: AB_2864299Anti-CalretininAbcamCat# ab92341; RRID: AB_2049245Anti-ERp57AbcamCat# ab10287; RRID: AB_297026Anti-Camelid VHHGenScriptCat# A01861; RRID: AB_3083750Anti-His-TagCell Signaling TechnologyCat# 2365; RRID: AB_2115720Anti-HA-tagAbmartCat# M20003; RRID: AB_2864345Anti-cGASABclonalCat# A8335; RRID: AB_2770305Anti-STINGABclonalCat# A21051; RRID: AB_3083450Anti-TBK1ABclonalCat# A3458; RRID: AB_2863061Anti-Phospho-TBK1ABclonalCat# AP0847; RRID: AB_2771611Anti-IRF3ABclonalCat# A2172; RRID: AB_2764190Anti-Phospho-IRF3ABclonalCat# AP0623; RRID: AB_2771210Anti-F4/80Bio-RADCat# MCA497R; RRID: AB_323279Anti-FITC anti-mouse CD3BiolegendCat# 100203; RRID: AB_312660Anti-PE anti-mouse CD45BiolegendCat# 147711; RRID: AB_2563597Anti-APC/Cyanine7 anti-mouse CD8b.2BiolegendCat# 140421; RRID: AB_2860690Anti-APC anti-mouse CD4BiolegendCat# 100411; RRID: AB_312696Anti-APC/Cyanine7 anti-mouse CD86BiolegendCat# 159217; RRID: AB_3106042Anti-PE/Cyanine7 anti-mouse CD206 (MMR)BiolegendCat# 141719; RRID: AB_2562247Anti-PE anti-mouse CD11c AntibodyBiolegendCat# 117307; RRID: AB_313776Anti-APC/Cyanine7 anti-mouse CD11bBiolegendCat# 101225; RRID: AB_830641Anti-PE anti-mouse CD8b.2BiolegendCat# 140408; RRID: AB_10644002Anti-Brilliant Violet 650™ anti-mouse CD45.1BiolegendCat# 110735; RRID: AB_11124743Anti-PE/Cyanine7 anti-mouse CD45.2BiolegendCat# 109829; RRID: AB_1186103Anti-Phospho-SMAD2/SMAD3ThermoFisherPA5-110155; RRID: AB_2855566Chemicals, peptides, and recombinant proteinsM-CSFMedChemExpressHY-P7085IR783MedChemExpressHY-139109ICGMedChemExpressHY-W088089PKH67 Green Fluorescent Cell Linker Mini KitSigma-AldrichCat#MINI67-1KTIRDye 680LI-CORCat#D01202-12Critical commercial assaysBCA Protein Assay KitBeyotimeCat#P0012CCK-8 KitEpizymeCat#CX001MAnnexin V-FITC/PI KitThermoFisherCat#A13199Deposited dataRNA-seqThis paperGEO: GSE318327 and GSE318328Experimental models: Cell linesHEK-293ATCCCRL-1573Huh-1FuHengFH1331HepG2FuHengFH0076Hepa1-6ATCCCRL-1830HuH-7FuHengFH0075AsPC-1ATCCCRL-1682Panc02FuHengFH1102PANC-1ATCCCRL-14694T1ATCCCRL-2539Experimental models: Organisms/strainsMouse: BALB/c.GempharmatechN/AMouse: C57BL/6JGempharmatechN/AMouse: STING1−/− C57BL/6JShulaibao (Wuhan)N/ARecombinant DNApDisplay plasmidAddgeneN/ASoftware and algorithmsGraphPad Prism version 10.1GraphPadhttps://www.graphpad.com/BiorenderBiorenderhttps://www.biorender.com/FlowJo 10.6.5FlowJo, LLChttps://www.flowjo.com/Living Image softwarePerkinElmerN/A REAGENT or RESOURCESOURCEIDENTIFIERAntibodiesAnti-CD9AbcamCat# ab236630; RRID: AB_2922400Anti-CD81AbcamCat# ab79559; RRID: AB_1603682Anti-HSP70AbcamCat# ab181606; RRID: AB_2910093Anti-TSG101AbcamCat# ab125011; RRID: AB_10974262Anti-CalnexinAbcamCat# ab133615; RRID: AB_2864299Anti-CalretininAbcamCat# ab92341; RRID: AB_2049245Anti-ERp57AbcamCat# ab10287; RRID: AB_297026Anti-Camelid VHHGenScriptCat# A01861; RRID: AB_3083750Anti-His-TagCell Signaling TechnologyCat# 2365; RRID: AB_2115720Anti-HA-tagAbmartCat# M20003; RRID: AB_2864345Anti-cGASABclonalCat# A8335; RRID: AB_2770305Anti-STINGABclonalCat# A21051; RRID: AB_3083450Anti-TBK1ABclonalCat# A3458; RRID: AB_2863061Anti-Phospho-TBK1ABclonalCat# AP0847; RRID: AB_2771611Anti-IRF3ABclonalCat# A2172; RRID: AB_2764190Anti-Phospho-IRF3ABclonalCat# AP0623; RRID: AB_2771210Anti-F4/80Bio-RADCat# MCA497R; RRID: AB_323279Anti-FITC anti-mouse CD3BiolegendCat# 100203; RRID: AB_312660Anti-PE anti-mouse CD45BiolegendCat# 147711; RRID: AB_2563597Anti-APC/Cyanine7 anti-mouse CD8b.2BiolegendCat# 140421; RRID: AB_2860690Anti-APC anti-mouse CD4BiolegendCat# 100411; RRID: AB_312696Anti-APC/Cyanine7 anti-mouse CD86BiolegendCat# 159217; RRID: AB_3106042Anti-PE/Cyanine7 anti-mouse CD206 (MMR)BiolegendCat# 141719; RRID: AB_2562247Anti-PE anti-mouse CD11c AntibodyBiolegendCat# 117307; RRID: AB_313776Anti-APC/Cyanine7 anti-mouse CD11bBiolegendCat# 101225; RRID: AB_830641Anti-PE anti-mouse CD8b.2BiolegendCat# 140408; RRID: AB_10644002Anti-Brilliant Violet 650™ anti-mouse CD45.1BiolegendCat# 110735; RRID: AB_11124743Anti-PE/Cyanine7 anti-mouse CD45.2BiolegendCat# 109829; RRID: AB_1186103Anti-Phospho-SMAD2/SMAD3ThermoFisherPA5-110155; RRID: AB_2855566Chemicals, peptides, and recombinant proteinsM-CSFMedChemExpressHY-P7085IR783MedChemExpressHY-139109ICGMedChemExpressHY-W088089PKH67 Green Fluorescent Cell Linker Mini KitSigma-AldrichCat#MINI67-1KTIRDye 680LI-CORCat#D01202-12Critical commercial assaysBCA Protein Assay KitBeyotimeCat#P0012CCK-8 KitEpizymeCat#CX001MAnnexin V-FITC/PI KitThermoFisherCat#A13199Deposited dataRNA-seqThis paperGEO: GSE318327 and GSE318328Experimental models: Cell linesHEK-293ATCCCRL-1573Huh-1FuHengFH1331HepG2FuHengFH0076Hepa1-6ATCCCRL-1830HuH-7FuHengFH0075AsPC-1ATCCCRL-1682Panc02FuHengFH1102PANC-1ATCCCRL-14694T1ATCCCRL-2539Experimental models: Organisms/strainsMouse: BALB/c.GempharmatechN/AMouse: C57BL/6JGempharmatechN/AMouse: STING1−/− C57BL/6JShulaibao (Wuhan)N/ARecombinant DNApDisplay plasmidAddgeneN/ASoftware and algorithmsGraphPad Prism version 10.1GraphPadhttps://www.graphpad.com/BiorenderBiorenderhttps://www.biorender.com/FlowJo 10.6.5FlowJo, LLChttps://www.flowjo.com/Living Image softwarePerkinElmerN/A Experimental model and study participant details Human studies A human multi-organ cancer tissue microarray (TMA) (HOrgC120PG04) was obtained from Shanghai OUTDO Biotech Co. Ltd. The TMA contained a total of 120 cores representing 11 human cancer types, including PTC, EC, GC, COAD, READ, HCC, PDAC, LUSC, LUAD, BC, and RCC. For each cancer type, 2–6 patient samples were included, with paired tumor and adjacent paracancerous tissues represented by one core each. IHC staining for MDK protein was performed, and no random allocation was conducted. Detailed clinicopathological information, including diagnosis, pathological classification, and clinical stage, is publicly available from the manufacturer’s website (http://www.superchip.com.cn/biology/tissue.html). All human tissue samples were collected with informed consent and approved by the institutional ethics committee of the sample provider. Ethics approval and animals All animal experiments were approved by the Institutional Animal Care and Use Committee (IACUC) of Shenzhen People’s Hospital (AUP-220516-LZJ-0346-01 and AUP-230220-LZJ-0109-01) and Zhongnan Hospital of Wuhan University (ZN2025149), and were performed in accordance with relevant institutional and national guidelines and regulations. All procedures were conducted in compliance with institutional ethical standards. The study utilized wild-type C57BL/6J mice, wild-type BALB/c mice, and STING1−/− mice on a C57BL/6J genetic background. Female mice aged 6–8 weeks and male mice aged 8–10 weeks were obtained from Gempharmatech (Guangzhou, China) and maintained in the SPF-grade barrier facility at the Animal Center of Shenzhen People’s Hospital under controlled conditions (22 ± 2°C, 50–60% humidity, 12 h light/dark cycle), with ad libitum access to sterile chow and water. All tumor models were established using syngeneic murine cancer cell lines implanted into immunocompetent mice with matched genetic backgrounds, ensuring physiological relevance and reliable evaluation of immune responses. Both female and male mice were included depending on the experimental model, and no obvious sex-dependent differences were observed in the major experimental outcomes. Cell culture HEK-293, Huh-1, Hepa1-6, HuH-7, PANC-1, and KPC (Pdx1-Cre, Lox-stop-lox-KrasG12D/+, Lox-stop-lox-Trp53R172H/+) cells were cultured in DMEM supplemented with 10% FBS and 1% Pen-Strep. HepG2 cells were maintained in MEM with 10% FBS and 1% Pen-Strep. AsPC-1, Panc02, and 4T1 cells were cultured in RPMI-1640 supplemented with 10% FBS and 1% Pen-Strep. All cell lines were maintained at 37°C in a humidified incubator with 5% CO2. Human cell lines were authenticated by STR profiling performed by the providers. Murine cell lines were obtained from authenticated sources. All cell lines were routinely screened for mycoplasma contamination at regular intervals (every 1–2 months) using the AuthentiFiler PCR Amplification Kit (Applied Biosystems), and all tested negative throughout the study. Method details Experimental reagents Cell culture media, including DMEM and RPMI1640, fetal bovine serum (FBS), and penicillin-streptomycin (Pen-Strep), were sourced from Thermo Fisher Scientific (Waltham, MA, USA). Apoptosis detection was performed using an Annexin V-FITC/PI kit from MultiSciences (Hangzhou, China). Fixation was carried out with 4% paraformaldehyde (PFA) from Biyuntian Company (Shanghai, China). Fluorescent cell labeling was achieved using the PKH67 Green Fluorescent Cell Linker Mini Kit, while DNA fragmentation was assessed with a TUNEL staining kit, all obtained from Sigma-Aldrich (St. Louis, MO, USA). Ultracentrifuge tubes were procured from Beckman (Fullerton, California, USA), and Ni NTA Beads were purchased from Smart-Lifesciences (Changzhou, China). Nuclear staining was conducted with DAPI (Beyotime, Shanghai, China). A white light source (wavelength range: 350–750 nm) was supplied by Beijing China Education AU-Light Technology Co., Ltd (Beijing, China). Protein expression and purification Recombinant plasmids pET-14B-MDK (human and mouse) were transformed into BL21(DE3) cells for protein expression. Bacterial cultures were grown at 37°C with shaking at 225 rpm until reaching an OD600 of 0.6, followed by induction with 0.2 mM IPTG at 16°C overnight. Cells were harvested by centrifugation at 8,000×g for 15 min at 4°C. The pellets were resuspended in lysis buffer (300 mM NaCl, 50 mM NaH2PO4, 10 mM imidazole, pH 8.0, 1 mM PMSF) and lysed using high-pressure homogenization at low temperature for three cycles. After centrifugation at 12,000×g for 45 min, the supernatant was applied to a gravity column packed with 1 mL Ni-NTA agarose resin (Qiagen, Germany). The resin was sequentially washed with Wash Buffer I (20 mM imidazole) and Wash Buffer II (40 mM imidazole), followed by elution with Elution Buffer (250 mM imidazole). The eluted protein was further purified by Superdex-150 gel filtration chromatography (AKTA Pure System, GE Healthcare Life Sciences, USA) in 1 × PBS. Purified proteins were verified by SDS-PAGE, flash-frozen in liquid nitrogen, and stored at −80°C. Nanobodies were purified using the same Ni-NTA protocol and validated by Western Blot using anti-6×His, anti-HA, and anti-VHH antibodies. ELISA for Nb binding affinity Human and mouse MDK proteins were immobilized as antigens in 96-well plates overnight at 4°C. Plates were blocked with 3% BSA for 2 h at room temperature, followed by three washes with 0.1% PBST. Serial dilutions of MDK-specific Nbs and control Nb were added and incubated at 37°C for 1 h. After washing, 100 μL/well of HA-HRP antibody (1:3000; Sino Biological, Beijing, China) was added and incubated for 1 h. Plates were washed again, and 100 μL of TMB substrate (Solarbio, Beijing, China) was added to each well, followed by incubation in the dark for 3–5 min. The reaction was stopped with 100 μL/well of 1 M HCl, and absorbance was measured at 450 nm. Loading of EVs with Ce6 Ce6 dye was mixed with EVs and loaded into a 4 mm electroporation cuvette. Electroporation was performed using a Bio-Rad instrument at 400 V with a 1 ms pulse duration. To remove unencapsulated Ce6, the mixture was ultracentrifuged at 120,000×g for 90 min. Unbound Ce6 concentration was quantified using a fluorescence standard curve, and encapsulation efficiency was calculated as follows: [(Initial Ce6 concentration − Free Ce6 concentration)/Initial Ce6 concentration] × 100%. Cellular uptake of EVs EVs (1 × 1010 particles/mL) were labeled with 4 μL of PKH67 dye and incubated at room temperature for 30 min in the dark. The labeled EVs were pelleted by centrifugation at 100,000×g for 60 min at 4°C. Tumor cells were seeded in 96-well plates at 2 × 104 cells/well and cultured for 24 h. Labeled EVs were added to the cells and incubated for 2 h or 8 h. After washing three times with PBS, cells were fixed with 4% PFA for 30 min. Nuclei were stained with DAPI, and images were acquired using a confocal microscope. CCK-8 assay and colony formation assay Cell viability was assessed using a CCK-8 assay kit (HY-K0301, Shanghai, China). For colony formation assays, 1,000 cells were seeded and cultured for 2 weeks. Colonies were fixed with 4% paraformaldehyde for 30 min, stained with 0.5% crystal violet for 1 h, and counted after imaging. Apoptosis analysis Cells were cultured in 6-well plates until reaching 80% confluence and treated with various drugs for 24 h. Apoptosis was assessed using an Annexin V-FITC/PI Apoptosis Kit (Lianke, Hangzhou, China). Stained cells were analyzed by flow cytometry (CytoFLEX; Beckman Coulter, USA), and data were processed using FlowJo software (Version 10.0.7, FlowJo LLC, USA). Primary macrophage isolation and STING activation assays Primary peritoneal macrophages were isolated by peritoneal lavage from C57BL/6 mice stimulated by 4% thioglycolate solution and purified by adherence. Bone marrow-derived macrophages (BMDMs) were generated from femur and tibia bone marrow cells and differentiated for 7 days in RPMI-1640 medium containing 10% FBS and macrophage colony stimulating factor (M-CSF). For macrophage STING activation assays, tumor cells were treated under indicated conditions, and conditioned media were collected after centrifugation to remove debris and applied to macrophage cultures. Where indicated, STING inhibitors (H151 or C176) were added prior to stimulation. STING pathway activation was evaluated by Western Blot for pathway proteins and by qPCR analysis of ISGs (Ccl5, Cxcl10, Ifnb1, and Tnf). Analysis of immune cell infiltration in tumors Immune cell infiltration in tumor tissues from mouse models was assessed using a spectral flow cytometer (Cytek Northern Lights, Shanghai, China). Single-cell suspensions were prepared from tumor tissues via enzymatic digestion (RWD Life Science, Shenzhen, China). After red blood cell lysis with ACK buffer, viable cells were counted using Trypan Blue staining. Approximately 4 × 106 cells per sample were resuspended in 100 μL FACS buffer and incubated with CD16/32 antibody at 4°C for 30 min to block Fc receptors. Fluorophore-conjugated antibodies specific to immune cell markers were added and incubated on ice for 30 min. Macrophages were identified using CD45, F4/80, CD11b, CD86 and CD206; DC cells with CD45, CD11c, MHC II, CD80 and CD86; and T cells with CD45, CD3, CD4, and CD8; OT-1 cells with CD11b (negative), CD3 and CD45.1. Samples were analyzed using a spectral flow cytometer (Cytek, USA), with single-stained controls for compensation. In vivo NIR-I fluorescence imaging KPC-WT or MDK-KO KPC tumor-bearing mice (tumor volume ∼500 mm3) were randomized into groups (n = 3 per group). KPC mice received intravenous injections of ICG@D4-EV (ICG: 60 μg; D4-EVs: 1 × 1010 particles). Mice were imaged at 4, 8, 12, and 24 h post-injection using an IVIS imaging system (PerkinElmer, USA). At 24 h, mice were euthanized, and tumors and major organs (heart, liver, spleen, lung, kidney, colon) were harvested for ex vivo imaging and nanobody staining. In vivo NIR-II fluorescence imaging Subcutaneous and orthotopic KPC tumor-bearing mice (tumor volume ∼500 mm3 or 4 weeks post-implantation) were randomized into two groups (n = 3 per group). Mice received intravenous injections of IR783@Blank-EV (IR783: 100 μg; Blank-EV: 1 × 1010 particles) or IR783@D4-EV (IR783: 100 μg; D4-EV: 1 × 1010 particles). NIR-II fluorescence imaging was performed at 4 and 24 h post-injection using an Artemis Intelligent Imaging System (Artemis Intelligent Imaging Inc., Shanghai, China). At 24 h, mice were euthanized, and tumors along with major organs were harvested for ex vivo imaging and nanobody staining. Preparation and activation of OT-1 T cells from mouse spleen and lymph nodes Spleens and lymph nodes were harvested from 6 to 12-week-old C57BL/6 OT-I (CD45.1) mice, homogenized through a metal strainer, and filtered to obtain single-cell suspensions. Red blood cells were lysed using ACK buffer, and the remaining cells were washed and resuspended in complete RPMI-1640 medium. Cells were activated by culturing with OVA peptide (25 nM) and IL-2 (10 U/mL) for 3 days. Activated T cells were counted, resuspended at 20 × 106 cells/mL, and injected intravenously into mice for further experiments. Tunnel staining Tissue sections were dewaxed and treated with 20 μg/mL DNase-free protease K (Beyotime Biotechnology, China) at 37°C for 30 min. After washing with PBS, 50 μL of Tunnel reaction mixturewas applied and incubated at 37°C for 60 min in the dark. Slides were mounted with anti-fluorescence quenching medium and imaged using a fluorescence confocal microscope (Leica Biosystems, Germany). Immunofluorescence and IHC staining For immunofluorescence staining, tissue sections were fixed with acetone/methanol (4:1) at −20°C for 10 min, blocked with 4% donkey serum in PBS for 30 min at room temperature (RT), and incubated with primary antibodies overnight at 4°C. Sections were then stained with fluorophore-conjugated secondary antibodies in 0.5% donkey serum/PBS for 1 h at RT in the dark and mounted with DAPI-containing medium. For IHC staining, paraffin-embedded tissues were deparaffinized, rehydrated, and subjected to antigen retrieval in Tris-EDTA buffer (pH 9.0) by steaming for 15 min. Endogenous peroxidase activity was quenched with 0.3% H2O2-methano for 10 min. Sections were blocked with 4% donkey serum in TBST for 1 h, incubated with primary antibodies overnight at 4°C, and then with biotinylated secondary antibodies for 60 min. Signal amplification was performed using VECTASTAIN Elite ABC Reagent (Vectorlabs, USA), followed by development with DAB peroxidase substrate. Slides were dehydrated, air-dried, and mounted with neutral mounting medium (Solaibao, China). TEM EVs morphology was analyzed using TEM (JEOL JEM-1400 Plus, Japan). Samples (30 μL) were applied to carbon-coated copper grids (Xinxing Braim, China) for 5 min, followed by staining with 4% uranyl acetate. After air-drying, grids were imaged under TEM. Western blot analysis Total proteins were extracted using RIPA buffer (Beyotime Biotechnology, China), and protein concentrations were determined using a BCA protein assay kit (Beyotime Biotechnology). Proteins were separated by SDS-PAGE (10%, 12.5%, or 15% gels) and transferred onto PVDF membranes (Millipore, USA). Membranes were blocked with 5% nonfat dry milk (Biosharp, China) in TBST for 1–2 h at room temperature and then incubated with primary antibodies overnight at 4°C. After washing, membranes were probed with HRP-conjugated secondary antibodies. Protein bands were detected using enhanced chemiluminescence (ECL) reagent (GE Healthcare, UK) and visualized using an imaging system. Fluorescent western blotting for Nb copy number determination Nbs of known concentrations and engineered EVs were resolved by 15% SDS-PAGE and transferred onto PVDF membranes (Millipore, USA). Membranes were blocked with 5% nonfat dry milk for 1 h and probed with an HA-tag antibody overnight at 4°C. After washing, membranes were incubated with an IRDye 680-conjugated secondary antibody. Following three washes in TBST and a final wash in TBS, protein bands were visualized using a fully automated electrophoretic fluorescence immunoassay analyzer (Azure Biosystems, USA). Fluorescence intensities were quantified using ImageJ software to generate a standard curve, which was used to calculate the copy numbers of nanobodies displayed on each EV. qPCR The primer information are summarized in Table S1. Total RNA was isolated from tissues or cells using TRIzol reagent (Thermo Fisher Scientific, USA). RNA purity and integrity were verified by agarose gel electrophoresis, assessing 28S and 18S ribosomal RNA bands. Approximately 2 μg of total RNA was reverse transcribed into cDNA. qPCR was performed using SYBR Green RT-PCR Master Mix (BioRad, USA) on a CFX 96 RT-PCR system with gene-specific primers. The cycling conditions were: 95°C for 10 min, followed by 40 cycles of 95°C for 15s and 60°C for 1 min. Gene expression levels were normalized to β-actin using the 2−ΔΔCt method. RNA-seq Total RNA was extracted from PBS-treated and Ce6@D4-EV plus Light-treated KPC tumor samples. RNA concentration and purity were measured using a NanoDrop 2000 spectrophotometer (NanoDrop Technologies, USA), and integrity was assessed via agarose gel electrophoresis and an Agilent 2100 Bioanalyzer (Agilent Technologies, USA). mRNA was isolated using Oligo (dT) magnetic beads, fragmented to ∼300 bp, and used to construct sequencing libraries. Libraries were sequenced on the Illumina NovaSeq 6000 platform (Illumina, USA). Differentially expressed genes (DEGs) were identified with a threshold of p value <0.05 and an absolute log2 (fold change) > 2. Mouse xenograft models and treatments Specific treatment options are summarized in Table S2. For the analysis of dark toxicity, KPC tumor-bearing mice were randomly divided into four treatment groups, which were treated with Blank (200 μL, PBS), Free-Ce6 (100 μg), Ce6@Blank-EV (1 × 1010 particles/injection) and Ce6@D4-EV (1 × 1010 particles/injection) respectively. Mice were given two injections according to the therapeutic schedule. To evaluate the therapeutic efficacy of Ce6@D4-EV, KPC tumor-bearing mice were randomly assigned to four treatment groups (n = 5): (1) Blank (200 μL, PBS); (2) Free-Ce6 (100 μg) plus light (80 mW/cm2, 10 min); (3) Ce6@Blank-EV (1 × 1010 particles/injection) plus light (80 mW/cm2, 10 min); (4) Ce6@D4-EV (1 × 1010 particles/injection) plus light (80 mW/cm2, 10 min). Treatments were administered two times according to the established schedule. To evaluate the effectiveness of Ce6@D4-EV in the treatment of multiple tumors, additional subcutaneous syngeneic tumor models were established in immunocompetent mice. For the MDK-overexpressing Panc02 model, Panc02-MDK cells were implanted subcutaneously into C57BL/6 mice. When tumors became established, mice were randomly assigned into five treatment groups (n = 5): (1) PBS (200 μL); (2) Free-Ce6 (100 μg) plus light (80 mW/cm2, 10 min); (3) D4-EV (1 × 1010 particles); (4) Ce6@Blank-EV (1 × 1010 particles/injection) plus light (80 mW/cm2, 10 min), and (5) Ce6@D4-EV (1 × 1010 particles/injection) plus light (80 mW/cm2, 10 min). Similarly, subcutaneous 4T1 breast tumors were generated in immunocompetent BALB/c mice, and animals were treated using the same five-group regimen (n = 5): (1) PBS (200 μL); (2) Free-Ce6 (100 μg) plus light (80 mW/cm2, 10 min); (3) D4-EV (1 × 1010 particles); (4) Ce6@Blank-EV (1 × 1010 particles/injection) plus light (80 mW/cm2, 10 min); and (5) Ce6@D4-EV (1 × 1010 particles/injection) plus light (80 mW/cm2, 10 min). For the Hepa1-6 HCC model, Hepa1-6 cells were implanted subcutaneously into C57BL/6 mice. Tumor-bearing mice were then randomized into three treatment groups (n = 5): (1) Blank (PBS, 200 μL); (2) Ce6@Blank-EV (1 × 1010 particles/injection) plus light (80 mW/cm2, 10 min); (3) Ce6@D4-EV (1 × 1010 particles/injection) plus light (80 mW/cm2, 10 min). Treatments were administered two times according to the established schedule. To examine the impact of macrophage depletion or STING pathway inhibition on Ce6@D4-EV-mediated photoimmunotherapy, subcutaneous KPC tumor-bearing mice were randomly divided into four groups (n = 5) and treated with: (1) Blank (PBS, 200 μL); (2) Ce6@D4-EV (1 × 1010 particles/injection) plus light (80 mW/cm2, 10 min) plus clodronate liposomes (200 μL, 5 mg/mL, peritumoral injection for 7 days); (3) Ce6@D4-EV (1 × 1010 particles/injection) plus light (80 mW/cm2, 10 min) plus H151 (10 mg/kg, intraperitoneal injection for 7 days); (4) Ce6@D4-EV (1 × 1010 particles/injection) plus light (80 mW/cm2, 10 min). Mice were treated two cycles with PDT. To further address concerns regarding potential off-target anti-inflammatory effects of H151 and to provide more stringent mechanistic validation, STING dependency was additionally confirmed using both genetic and independent pharmacological approaches. Subcutaneous KPC tumors were established in STING1+/+ and STING1−/− C57BL/6 mice (n = 5) and treated with PBS (200 μL) or Ce6@D4-EV (1 × 1010 particles/injection) plus light (80 mW/cm2, 10 min). Mice were treated two cycles with PDT. In parallel, the covalent STING inhibitor C176 was employed as an additional pharmacological control. Subcutaneous KPC tumor-bearing mice were randomly divided into four groups (n = 5) and treated with: (1) Blank (PBS, 200 μL); (2) C176 (13 mg/kg, intraperitoneal injection for 6 days); (3) Ce6@D4-EV (1 × 1010 particles/injection) plus light (80 mW/cm2, 10 min) plus C176 (13 mg/kg, intraperitoneal injection for 6 days); (4) Ce6@D4-EV (1 × 1010 particles/injection) plus light (80 mW/cm2, 10 min). Mice were treated two cycles with PDT. For combination ICB therapy, subcutaneous KPC tumor-bearing mice were randomly assigned to six groups (n = 5 per group): (1) Blank (PBS, 200 μL); (2) PD-1 antibodies (12.5 mg/kg); (3) CD47 Nbs (10 mg/kg); (4) Ce6@D4-EV (1 × 1010 particles/injection) plus light (80 mW/cm2, 10 min); (5) Ce6@D4-EV (1 × 1010 particles/injection) plus light (80 mW/cm2, 10 min) plus PD-1 antibodies (12.5 mg/kg); (6) Ce6@D4-EV (1 × 1010 particles/injection) plus light (80 mW/cm2, 10 min) plus CD47 Nbs (10 mg/kg). Mice were treated two cycles with PDT and then followed two injections of PD-1 antibodies/CD47 Nbs according to the therapeutic schedule. For combination chemotherapy, KPC tumor-bearing mice were randomly assigned to four groups (n = 5): (1) Blank (PBS, 200 μL); (2) Gem (25 mg/kg); (3) Ce6@D4-EV (1 × 1010 particles/injection) plus light (80 mW/cm2, 10 min); (4) Ce6@D4-EV (1 × 1010 particles/injection) plus light (80 mW/cm2, 10 min) plus Gem (25 mg/kg). Mice were treated two cycles with PDT and then followed two injections of Gem according to the therapeutic schedule. For combination adoptive T cell therapy, KPC tumor-bearing mice were randomly assigned to four groups (n = 5): (1) Blank (PBS, 200 μL); (2) OT-1 cells (1 × 106 cells); (3) Ce6@D4-EV (1 × 1010 particles/injection) plus light (80 mW/cm2, 10 min); (4) Ce6@D4-EV (1 × 1010 particles/injection) plus light (80 mW/cm2, 10 min) plus OT-1 cells (1 × 106 cells). Mice were treated two cycles with PDT and then followed two injections of OT-1 cells according to the therapeutic schedule. To further strengthen the robustness and translational relevance of these combination effects beyond the original KPC model, we systematically extended our in vivo validation to multiple non-KPC tumor systems. The subcutaneous MDK-overexpressing Panc02 tumor-bearing mice were randomly assigned to six groups (n = 5): (1) Blank (PBS, 200 μL); (2) Gem (25 mg/kg); (3) PD-1 antibodies (12.5 mg/kg); (4) Ce6@D4-EV (1 × 1010 particles/injection) plus light (80 mW/cm2, 10 min); (5) Ce6@D4-EV (1 × 1010 particles/injection) plus light (80 mW/cm2, 10 min) plus Gem (25 mg/kg). (6) Ce6@D4-EV (1 × 1010 particles/injection) plus light (80 mW/cm2, 10 min) plus PD-1 antibodies (12.5 mg/kg); Mice were treated two cycles with PDT and then followed two injections of Gem/PD-1 antibodies according to the therapeutic schedule. The orthotopic 4T1 tumor-bearing mice were randomly assigned to six groups (n = 5): (1) Blank (PBS, 200 μL); (2) PTX (20 mg/kg); (3) PD-1 antibodies (12.5 mg/kg); (4) Ce6@D4-EV (1 × 1010 particles/injection) plus light (80 mW/cm2, 10 min); (5) Ce6@D4-EV (1 × 1010 particles/injection) plus light (80 mW/cm2, 10 min) plus PTX (20 mg/kg). (6) Ce6@D4-EV (1 × 1010 particles/injection) plus light (80 mW/cm2, 10 min) plus PD-1 antibodies (12.5 mg/kg); Mice were treated two cycles with PDT and then followed two injections of PTX/PD-1 antibodies according to the therapeutic schedule. The subcutaneous Hepa1-6 tumor-bearing mice were randomly assigned to four groups (n = 5): (1) Blank (PBS, 200 μL); (2) PD-1 antibodies (12.5 mg/kg); (3) Ce6@D4-EV (1 × 1010 particles/injection) plus light (80 mW/cm2, 10 min); (4) Ce6@D4-EV (1 × 1010 particles/injection) plus light (80 mW/cm2, 10 min) plus PD-1 antibodies (12.5 mg/kg). Mice were treated two cycles with PDT and then followed two injections of PD-1 antibodies according to the therapeutic schedule. Quantification and statistical analysis Data are presented as mean ± SEM unless otherwise specified. Statistical analyses were conducted using Prism software (GraphPad Prism, version 10.1). To compare datasets from two distinct groups, two-tailed unpaired t test was employed. Assessment of in vitro cell viability and tumor growth curves was performed using one-way ANOVA or two-way ANOVA with Tukey’s post-test. Survival rates were analyzed using the Kaplan-Meier method and Mantel–Cox test. Statistical significance was determined at p values of 0.05 or lower. The notation “ns” indicates a lack of statistical significance, with ∗p < 0.05, ∗∗p < 0.01, and ∗∗∗p < 0.001 denoting various levels of significance.