Intelligent plant exosomes synergize miRNAs and cisplatin for spatiotemporally precise multimodal treatment for TNBC with high safety

Abstract

Triple-negative breast cancer (TNBC) lacks common receptors and exhibits aggressive behavior, limiting treatment options due to drug resistance and systemic toxicity. TNBC chemotherapy is hindered by poor tumor targeting, drug resistance, and systemic toxicity. Herein, this study presented a cascade targeting exosomal-cisplatin synergistic microneedle nanoplatform (CDDP@RKTExo-MN) as an intelligent wearable therapeutic device for TNBC treatment. Medicinal plant Taxus chinensis derived exosomes (TExo), carrying therapeutic miRNA, was synergized with cisplatin that induced ER stress to trigger a multimodal anti-tumor effects. The cisplatin-loaded TExo was further modified with αvβ3 integrin peptides and an ER-targeting motif for tumor homing and precise subcellular delivery. Leveraging the superficial localization of TNBC, the engineered TExo was integrated into a 3D-printed microneedle patch to construct a closed-loop transdermal delivery system (CDDP@RKTExo-MN). This bioactive architecture ensures precise drug delivery at the tumor site, effectively maximizing therapeutic efficacy while circumventing the systemic off-target toxicity inherent to conventional delivery strategies. CDDP@RKTExo-MN was shown for the cascade targeting capabilities with both cancer cells and their endoplasmic reticulums. By coordinated regulation of MAPK and TNF pathways, the system generated synergistic effects in both significantly amplifying apoptotic signaling and activating immunological protection. In vivo studies conclusively validated its superior tumor suppression efficacy alongside a favorable biosafety.

Graphical abstract

Highlights

• •Engineered medicinal plant exosomes synergize with chemotherapy via transdermal cascade-targeted delivery for safe therapy.
• •Local microneedle delivery of tumor-ER-targeted TExo enhances anti-tumor effects with low systemic toxicity.
• •Spatiotemporal drug release improves targeting specificity and reduces off-target effects in cancer therapy.

1. Introduction

Triple-negative breast cancer (TNBC) is a highly aggressive subtype of breast cancer that lacks estrogen, progesterone and human epidermal growth factor receptors [1,2].

Cisplatin, a chemotherapy drug working by forming covalent interactions with the nitrogenous bases of DNA, which prevents transcription and replication of tumor cells and induces ER stress, ultimately leading to apoptosis, has been widely used as a cost-effective standard therapy for TNBC [[3], [4], [5]]. However, a variety of adverse reactions, including nephrotoxicity, hematological toxicity, and immunotoxicity of cisplatin, have been described to seriously limit its clinical applications [6,7]. To address these issues, the delivery of cisplatin with tumor targeting carriers have been widely explored to replace the traditional vehicles to enhance transfer accuracy and decrease toxicity [[8], [9], [10], [11], [12]]. Recently, plant exosome-like extracellular vesicles (PLEVs) as the new natural platforms with high biocompatibility have been shown for the obvious potential in external payloads delivery [13,14]. Furthermore, our previous studies demonstrated that, exosomes derived from medicinal plants can furthermore act as effective nanotherapeutics to interference the difficult-to-treat diseases by transferring its incorporated endogenous plant therapeutic miRNAs [[15], [16], [17], [18], [19]]. Recent advances in miRNA-based cancer therapy highlight its potential through tumor pathway regulation, innovative delivery systems, and promising preclinical and clinical applications [20,21].

Taxus chinensis is a medicinal plant that has been widely used in the treatment of cancer in clinic [[22], [23], [24]]. It is reasonably to hypothesize that Taxus chinensis exosomes (TExo) can not only act as vehicle to deliver external molecules, but also can transfer their intrinsic anti-cancer biocargoes simultaneously for synergistic anti-cancer potency. However, to ensure the therapeutic index of exosome therapy in vivo, the further engineering on it is advantageous.

The key to the functionality of exosomes is their uptake by cells. The selection of the targeted fractions on exosome membranes such as ligands that bind to proteins overexpressed on the surface of cancer cells, modulate exosome uptake [25]. αvβ3-integrin plays a key role in angiogenesis, the formation of new vessels in tissues that lack them. Besides being involved in angiogenesis, the αvβ3-integrin is also presented on tumor cells of various origin, where it is involved in the processes that govern metastasis [26,27]. Therefore, the αvβ3 integrin-targeting ligand (RGP) represents an excellent option for enhancing the cell membrane targeting capabilities of TExos, thereby facilitating cascade targeting.

The destruction of specific subcellular structures via activating apoptotic signals has recently evolved as a burgeoning therapeutic strategy [28,29]. Due to the heightened sensitivity of cancer cells to the endoplasmic reticulum (ER) stress response, ER-targeting has garnered significant attention for the potential to enhance therapeutic payload accumulation while minimizing off-target effects for cancer therapy [30,31]. Cisplatin that can induce endoplasmic reticulum stress and promote tumor cells apoptosis, was therefore proposed to be loaded into the dual targeting engineered TExo. However, small-molecule drugs show no organ selectivity, thus typically failing to accumulate in specific organelles [32]. Hence, selecting modifiers that can synergistically target organelles such as the endoplasmic reticulum-resident protein signal peptide (KDEL) to achieve cascade targeting is also very important [33]. Anti-cancer drugs are frequently challenged by the toxicity and side effects because of the absorption and accumulation of anti-cancer components in organs and blood upon oral administration or intravenous injection [34,35]. By contrast, skin meditated transdermal drug delivery can minimize the circulation and accumulation of drug molecules in the blood and organs, causing wide attention in cancer drug treatment [36,37].

Herein, we developed a novel intelligent cascade targeting exosomal-cisplatin synergistic nanoplatform with spatiotemporally precise control, CDDP@RKTExo-MN. Specifically, TExo were modified with both the synthetic α_vβ₃ integrin-targeting ligand (RGP) and the endoplasmic reticulum-resident protein signal peptide (KDEL) to achieve a cascade targeting to cell membrane and endoplasmic reticulum in sequence. The smart exosomes were further incorporated into wearable porous microneedle patches to initiate a skin meditated sustained drug delivery at the precise localization of breast cancer site, to both enhance the anti-tumor efficacy and reduce the systemic toxicity induced by oral administration or intravenous injection. Through the multi-omics analysis, the molecular regulatory mechanisms of CDDP@RKTExo-MN in TNBC were fully investigated (see Fig. 1).

Fig. 1.

Schematic illustration of the construction, transdermal delivery, antitumor performance and molecular mechanism of CDDP@RKTExo-MN in TNBC. (1) The construction of CDDP@RKTExo-MN; (2) The CDDP@RKTExo-MN demonstrated exceptional anti-tumor efficacy in a subcutaneous (heterotopic) TNBC model through 4T1 membrane-directed tumor homing and transdermal microneedle-mediated precision delivery; (3) TE-miRs triggers apoptosis via MAPK/TNF axis while cooperatively targeting ER and upregulating ROS, yielding synergistic antitumor effects.

2. Results

2.1. Identification and characterization of TExo

TExos were isolated from Taxus chinensis (Fig. 2a). Exosomes were primarily composed of various membrane phospholipids, small molecules, nucleic acids and proteins [38]. Lipidomic analysis demonstrated that the lipids present in TExo mainly consisted of triglycerides (TG, 28%), diglycerides (DG, 11.8%), phosphatidyl ethanolamine (PE, 7.9%), and phosphatidylcholine (PC, 5.5%) (Fig. 2b). These natural lipids contributed to maintaining the spherical configuration of TExo, protecting the encapsulated bioactive cargos, and endowing them with good bio-membrane penetration capacity. Small molecules metabolomics data showed that each of the top 10 small molecules was directly or indirectly involved in the regulation of tumor physiology, among which the largest content of lipid and lipophilic molecules (55% relative abundance) were involved in the regulation of cell membranes fluidity and energy metabolism [39], suggesting the anti-tumor potential of the exosomes (Fig. 2c). The protein gel electrophoresis result revealed that the contained proteins in TExos were concentrated in the range of 35-55 kDa (Fig. S1). These TExo proteins were mainly concentrated in the chloroplast, cytoplasm, and nucleus, suggesting the functional versatility of the exosomes in intercellular communication through the integration of organelle-specific protein networks (Fig. 2d) [40]. MiRNAs have been widely demonstrated to be enclosed within exosomes as active substances contributing to the bioactivities of exosomes [41]. Consequently, miRNAs were identified within TExos (referred as TExo-miRNAs), and the sequencing results revealed 264 reliable miRNA candidates in TExo (Fig. 2e₁). The predominance of rRNA within the small RNA (sRNA) population indicated a high level of rRNA expression in the cell. Moreover, the majority of sRNAs are concentrated in the 16–24 nt, suggesting that RNAs of this length are particularly abundant within the cellular environment (Fig. 2e2, e3). 115 known miRNA completely matched the Taxus chinensis plant database, and 27 novel miRNA were found (Fig. 2e4).

Fig. 2.

Preparation and characterization of the compositions of TExo. (a) Diagram of TExo (b) Category and quantities of the lipids in TExo. (c) Compositional distribution and classification of small molecules. (d) Subcellular localization in TExo. (e₁, e₂, e₃, e₄) Analysis of the number, type, length distribution and Heatmap of miRNAs identified in TExo.

2.2. TExo exhibited efficient cellular uptake and multimodal anti-TNBC efficacy

Exosome uptake by target cells was a prerequisite for their functions. To explore the potential of TExo as the nanodrug with anti-cancer efficacy, the cellular uptake of TExos by 4T1 cells was investigated (Fig. 3a). The maximum uptake of TExo was 88.4% (Fig. 3b), demonstrating TExo could be recognized by mammalian cell surface and the uptake of TExo by 4T1 cells within 24 h was relatively considerable (Fig. S2). Meanwhile, the TExos exhibited significant inhibitory effects on cellular proliferation (Fig. 3c) and demonstrated a pronounced capacity to induce apoptosis in 4T1 cells (Fig. 3d–S3). TExo apparently reduced the migration ability of 4T1 cells (Fig. 3e₁, e₂). Subsequently, transwell tests demonstrated TExo significantly inhibited the invasion ability of 4T1 cells (Fig. 3f₁, f₂), showing a 1.35-fold increase compared to CON group. Within 6 h, the ROS levels in 4T1 cells were elevated by TExo (Fig. 3g–S4, S5). These findings demonstrated the intrinsic anti-tumor therapeutic properties of TExo.

Fig. 3.

Cellular uptake and antitumor effects of TExos. (a) Diagram of the mechanism of the cellular uptake of TExos by 4T1 cells. (b) Quantified images of cellular uptake of 4T1 co-cultured with the DiO-labeled TExos. (c) Viability of 4T1 cells treated by TExos with different concentrations. (d) Quantitative results of apoptosis of 4T1 cells with blank medium or TExos for 24 h (e₁,e₂) Light microscope images and quantified migration of 4T1 cells treated with blank medium or TExos at 0, 6, 12, and 24 h after co-incubation, respectively. Scale bar: 400 μm. (f₁,f₂) Invasion of 4T1 cells with blank medium or TExos for 24 h and quantitative results. Scale bar: 200 μm. (g) The ROS quantitative results of 4T1 co-cultured with TExos for 24 h, analyzed by flow cytometry.∗∗∗∗p < 0.0001, ∗∗∗p < 0.001, ∗∗p < 0.01, ∗p < 0.05.

2.3. Transcriptional interference of TExo to 4T1 cells by transferring miRNAs

MiRNAs are important direct gene regulators, which bind to the target genes and act as the post-transcriptional regulators to regulate proteins expression by preventing translation or triggering mRNAs degradation. The transfer of miRNAs contained in PLEVs to the target cells has proven to be one of the main principles for the execution of exosomal therapeutic efficacy as nanodrug [42]. To systematically decode this cross-kingdom regulatory mechanism in TNBC, we first profiled the miRNA cargo of TExos and their transfer into 4T1 cells.

Four types of miRNAs derived from TExos were identified to be transferred into 4T1 cells and influenced the expression of cancer cell miRNAs (Fig. 4a₁). It was further disclosed that, among the 4866 differentially expressed genes in 4T1 cells upon TExo treatment, 974 of them were predicted to be the targets of these 4 TExo miRNAs, implying their significant regulatory roles (Fig. 4a₂). Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses were next performed on the differentially expressed mRNAs in the cells, revealing that the genetic modification of TExo on cells was mainly concentrated in cancer-related pathways. GO enrichment analysis showed that these mRNAs were primarily enriched in processes such as “Intracellular anatomical structure”, “Intracellular organelle” and “Intracellular membrane-bounded organelle" among others, which suggested that these mRNAs played a role in various cellular biological processes, including regulation of cell morphology maintenance, material transport, and structural stability (Fig. 4b). KEGG enrichment analyses of the delivered miRNAs revealed that they are significantly enriched in cancer-related signaling pathway, especially in "MicroRNAs in cancer", “Protein processing in endoplasmic reticulum” and “Phosphatidylinositol signaling system” which influence numerous cancer-relevant processes such as proliferation, cell cycle control, apoptosis, differentiation, migration and metabolism. Notably, its strong association with the endoplasmic reticulum pathway suggested potential ER-targeting properties (Fig. 4c). Followingly, enrichment analysis of the target genes of 4 miRNAs was similar with the enrichment analysis of the differentially expressed miRNAs.These findings mainly related to signal transduction, cancer, and cellular processes, including MAPK signaling pathway, endometrial cancer, TGF-beta signaling pathway, AMPK signaling pathway, and colorectal cancer (Fig. S6). These results demonstrated the potential of TExos to function as delivery vehicles for active miRNA in cancer cells and ER-targeting.

Fig. 4.

Transcriptional interference of TExo to 4T1 cells by transferring miRNAs. (a₁) Venn diagram of miRNAs contained in both TExo and 4T1 cells. (a₂) Venn diagram of mRNAs contained in both TExos target gene prediction and up-regulated in 4T1 cells. (b)Top 20 of GO and c) KEGG enrichment analysis. (d) The heatmap of mRNA changes regulated by TExos (e₁, e₂) WB analysis of MAPK signaling pathway and quantitative. (f₁, f₂) WB analysis of TNF signaling pathway and quantitative. (g) The diagram of the mechanism of tumor growth inhibition and apoptosis.∗∗∗∗p < 0.0001, ∗∗∗p < 0.001, ∗∗p < 0.01, ∗p < 0.05.

MiRNAs serve as pivotal regulators by influencing protein expression through their binding to specific gene targets. After TExos treatment, the expression of relevant genes was significantly modulated, exhibiting both upregulation and downregulation (Fig. 4d). Gene Set Enrichment Analysis (GSEA) and western blot experiment were further performed and results revealed two distinct signaling pathways closely associated with TNBC development and progression. There was an overall downregulation of MAPK pathway-related proteins, which helps attenuate tumorigenic inflammation and immune evasion (Fig. 4e₁, e₂, S7). Conversely, Tumor Necrosis Factor (TNF) signaling was upregulated following TExo intervention (Fig. 4f₁, f₂, S8). TNF was a cytokine that can directly kill tumor cells without significant cytotoxicity to normal cells [43]. This dual modulation which suppressing survival-associated MAPK pathways while promoting TNF-induced cell death, provided a comprehensive therapeutic mechanism, leading to significant tumor growth inhibition (Fig. 4g).

2.4. Metabolic alteration and transcriptomic–metabolomic co-analysis of TExo intervention to 4T1 cells

The structural formulas of differential metabolites associated with cancer-related immunomodulation and energy metabolism were characterized in the metabolic profiling of 4T1 cells following TExos intervention (Fig. 5a₁-b₂). In the amino acid and infection-related metabolism, TExos triggered marked alterations in phosphocreatine, carnitine, and carnosine levels (Fig. 5a₁, a₂), suggesting disrupted energy shuttling through the creatine kinase system and modified β-oxidation dynamics. The concomitant decrease in 4-pyridoxic acid (vitamin B6 catabolite) and lumichrome (riboflavin derivative), coupled with elevated dehydrobiotin implies compromised vitamin B6/B2-dependent enzymatic activities and potential oxidative stress adaptation. Creatine Kinase B (CKB), the source of phosphocreatine, highly expressed in TNBC, supports proliferation by boosting glycolysis and lactate secretion [44,45]. The invasiveness of TNBC also increases through β-oxidation pathway modifications. TExos reduced CKB and β-oxidation, thereby inhibiting glycolysis and enhancing mitochondrial oxidative phosphorylation to reverse tumor metabolism. Nucleotide, carbohydrate, and lipid metabolism exhibited pathway-specific modulations, suggesting that the coordinated metabolic shifts induced by the treatment disrupted cellular homeostasis. These alterations in nucleotide and lipid pathways indicate a state of severe metabolic stress and energetic exhaustion in 4T1 cells, thereby potentiating the therapeutic efficacy. (Fig. 5b₁, b₂). This phenotype, characterized by enhanced redox balancing, improved membrane plasticity, and nucleotide conservation, may facilitate adaptation to therapeutic stress. Compared to the blank control group, the T-4T1 group showed an upregulation of 49 significant differential metabolites and a downregulation of 140 differential metabolites after intervention (Fig. S9a and b). Notably, many metabolites significantly shifted, indicating potential reprogramming of cellular metabolism upon TExo exposure and deepening the understanding of tumor microenvironment remodeling (Figs. S10 and S11). Collectively, these multiple mechanisms results underscored the profound metabolic reprogramming induced by TExo in 4T1 cells, with the identified metabolites served as potential biomarkers or therapeutic targets for further mechanistic studies.

Fig. 5.

Metabolic alteration and co-analysis of transcriptomic–metabolomic after TExo intervention in 4T1 cells. (a) Structural formulas of differential metabolites related to cancer-associated immune regulation in 4T1 cells; (b) Structural formulas of differential metabolites related to energy metabolism in 4T1 cells; (c) Differential heatmap of metabolites associated with immune regulation in 4T1 cells under Tc intervention; (d) Differential heatmap of metabolites associated with energy metabolism in 4T1 cells under Tc intervention; (e) Correlation network diagram of differential genes and metabolites.

Through Metabolite Set Enrichment Analysis (MESA), the top five pathways enriched were primarily associated with fatty acid metabolism (Fig. S12). TNBC tumor cells heavily relied on fatty acid oxidation for energy, particularly in microenvironments characterized by hypoxia or nutrient deficiency [46]. Concurrently, enhanced lipid synthesis constitutes a crucial mechanism of chemoresistance in TNBC. Consequently, TExos inhibit tumor cells to a certain extent by modulating fatty acid metabolism.

Transcriptomic-metabolomic integrative analysis demonstrated that the significant pathway in the metabolite-gene were metabolic pathways, central carbon metabolism in cancer, AMPK signaling pathway, carbohydrate digestion and absorption and inositol phosphate metabolism (Fig. 5c). The top 10 differentially expressed genes and differential metabolites were found to be intimately linked to the previously validated pathways, including the MAPK signaling pathway, Central carbon metabolism in cancer, and Arginine and proline metabolism (Fig. S13). The joint analysis of differential genes and metabolites indicated that the strongest associations were observed with four differential metabolites, specifically Vitamin C, Uric acid, 4-pyridoxic acid and Phosphocreatine, which were significantly linked to the metabolic pathways of Arginine and proline metabolism, as well as Microbial metabolism in diverse environments (Fig. 5d–S14). Therefore, transcriptomic–metabolomic confirmed that TExos modulate these arginine metabolic products through gene regulation. There was one differential gene, Map4k2, which was associated with the MAPK signaling pathway.

Through the above research, we have demonstrated for the first time that TExo not simply as carriers, but also exert the significant antitumor effects on 4T1 cells. By influencing transcriptional metabolism, the intervention of TExo on 4T1 tumors affects multiple pathways, such as MAPK, as well as metabolic pathways involving arginine and proline, thereby inhibiting tumor cell growth and modulating the tumor microenvironment, ultimately exerting an anti-TNBC effect.

2.5. Engineering of TExo into intelligent CDDP@RKTExo-MN

Plant-derived nanomedicines for the treatment of TNBC still face the challenge of non-targeting and limited penetration efficacy towards mammalian cells in vivo [47]. Furthermore, endoplasmic reticulum (ER)-targeted therapy induces the selective death of tumor cells or enhances their sensitivity to treatment [48]. ER-targeting therapeutics can increase the cytotoxicity of cisplatin to tumor cells and enhance antitumor immune responses. To enhance the active binding, penetration and endoplasmic Reticulum targeting of TExo in vivo, TExo were conjugated with both a novel α_vβ₃ integrin-targeting ligand (RGP) and a endoplasmic reticulum-resident protein signal peptide (KDEL), forming the cascade levels of targeting system, CDDP@RKTExo (Fig. 6a,b,c). Integrin α_vβ₃ is usually expressed at a low or undetectable level in most adult epithelial cells but highly upregulated in many kinds of cancer. Integrin αvβ3 participates in every step of tumor progression, including tumorigenesis, epithelial-mesenchymal transition (EMT), angiogenesis, tumor stemness, metabolic reprogramming, immune escape, bone metastasis and drug resistance. CDDP@RKTExo employed RGP which targeting αvβ3-integrin highly expressed exclusively in tumor sites and neovasculature, and KDEL modifications to achieve stepwise targeting of TNBC. Integrin αvβ3 is typically expressed at low levels in most quiescent adult tissues but is significantly upregulated in various malignancies and angiogenic vessels [49]. CDDP@RKTExo incorporates the RGP peptide to selectively target αvβ3 integrin by exploiting its preferential enrichment in the tumor microenvironment, combined with KDEL modifications to achieve precise, stepwise targeting of TNBC. Through the transmission electron microscope (TEM) analysis, CDDP@TExo and CDDP@RKTExo maintained the TExo morphology, and no significant rupture was observed to the membrane structure (Fig. 6d,e,f). After the targeting engineering, the particle sizes were increased from 152.63 ± 8.55 nm to 163.50 ± 11.55 nm for CDDP@TExo and 181.07 ± 10.57 nm for CDDP@RKTExo, respectively. Meanwhile, the changes in zeta potential for the PLEVs were not significant (Fig. 6g,h,i). The highest encapsulation efficiency and drug loading percentage CDDP onto TExo amount was determined as 0.25 μg/μg and 26.3%, respectively. Compared to the most reported results, this drug loading capacity of CDDP into the PLEV was increased with around four-fold.

Fig. 6.

Engineering and characterization of CDDP@RKTExo-MN (a) The diagram of drug-carrying microneedles loading CDDP@RKTExo. (b)Grafting rate of RGP (c) Grafting rate of KDEL (d, e, f) TEM image of TExo, CDDP@TExo, and CDDP@RKTExo. Scale bar: 150 nm. (g,h,i) Size distribution of TExo, CDDP@TExo and CDDP@RKTExo. (j) MNs and (l) CDDP@RKTExo-MN. Scale bar: 500 μm. (k₁-k₃) SEM images of the overall morphology and surface microstructure of MNs and (m₁-m₃) CDDP@RKTExo-MN. Scale bar: b1/d1: 1 mm, b₂/d₂: 20 μm, b3/d3: 1 μm (n₁, n₂) The internal elemental scans and energy level diagrams of CDDP@RKTExo-MN, C: red; N: green; O: blue; Si: yellow; P: orange; S: purple. (o) The overall drug release efficiency of CDDP@RKTExo-MN.∗∗∗∗p < 0.0001, ∗∗∗p < 0.001, ∗∗p < 0.01, ∗p < 0.05.

Previous studies have demonstrated that both cisplatin has obvious side effects to heart, kidney and muscles, along with the conventional oral administration [50,51]. Conventional Taxus chinensis extract has also been reported for the toxicity. To enhance the precise targeting and specific accumulation of CDDP@RKTExo at tumor sites, decreasing the circulation and accumulation of cisplatin and TExos in blood circulation and organs, CDDP@RKTExo was integrated into a novel wearable porous microneedles patch, which facilitates drug delivery through the skin, significantly reducing the accumulation of drugs in the bloodstream compared to oral or injectable administration (Fig. 6j–l).

The versatility, simplicity, high reproducibility and accuracy in the microscale associated by default with 3D printing have encouraged this research on its applicability on the fabrication of transdermal microneedle systems [52]. Mechanical strength characterization of porous MNs was measured through optimized condition. Preparation at 40 °C yielded a compression force of 374 mN/needle. This value was well above the threshold required for effective skin insertion, ensuring practical reliability (Fig. S15). SEM was performed to identify the drug-loaded in the microneedles (Fig. 6k₁-k₃, m₁-m₃). The surface of the blank microneedles exhibited significant pores, ranging in size from 100 to 150 nm, whereas the surface of the drug-loaded microneedles showed distinct accumulation of flaky and particulate matter. Elemental scanning of C, O, N, P, S, and Si within the microneedles indicated that nitrogen was a marker for TExo loading. The spectrum revealed that C, O, and Si were the predominant elements, while N was present in trace amounts, reflecting successful drug loading (Fig. 6n₁,n₂). The loading amount of CDDP@RKTExo within the microneedles patch is 380.95 ± 25.01 μg, resulting in an encapsulation efficiency of 20.51 ± 2.17%. At 37 °C, the release of CDDP from the CDDP@RKTExo-MN system demonstrated an initial rapid release of up to 28% within the first hour, likely due to a minor fraction of CDDP adhering to the surface. This was followed by a gradual release over the subsequent 24 h, stabilizing after 32 h and achieving a maximum release rate of 60%. These findings indicate the sustained-release efficacy of CDDP from the CDDP@RKTExo-MN system (Fig. 6o).

2.6. Active targeting and interference of CDDP@RKTExo-MN to 4T1 cells in vitro

Cell imaging observation indicated that, within the 6h period, CDDP@RKTExo could more effectively enter the 4T1 tumor cells (Fig. 7a₁,a_2, S16). Furthermore, compared to that of TExo, the fluorescence intensity of CDDP@RKTExo in the endoplasmic reticulum was significantly higher than that in TExo and CDDP@TExo, indicating the obvious targeting efficacy of CDDP@RKTExo to the endoplasmic reticulum organelle in 4T1 tumor cells (Fig. 7b₁, b_2, S17a,b).

Fig. 7.

Cellular uptake, active targeting and anti-tumor effects of CDDP@RKTExo and CDDP@RKTExo-MNin 4T1 cells. (a₁-a₂) CLSM and quantified images of 4T1 cells co-treated 6h with Dio, Scale bar: 50 μm and (b₁-b₂) endoplasmic reticulum ER-Tracker dyes. ER-Tracker: λ_ex = 587 nm, λ_em = 615 nm. Scale bar: 100 μm (c) 4T1 uptake of CDDP@TExo and CDDP@RKTExo for 0, 3, 6, 9,18 and 24 h (d) Viability of 4T1 cells treated by CDDP@TExo and CDDP@RKTExo with different concentrations. (e) Quantitative results of apoptosis of 4T1 cells with blank medium or CDDP@TExo and CDDP@RKTExo for 24h. (f) The ROS quantitative results of 4T1 co-cultured with CDDP, CDDP@TExo and CDDP@RKTExo for 24 h, analyzed by flow cytometry. (g₁, g₂) Light microscope images and quantified migration of 4T1 cells treated with blank medium or TExo, CDDP@TExo and CDDP@RKTExo at 0, 6, 12, and 24 h after co-incubation, respectively. Scale bar: 400 μm. (h₁,h₂) Invasion of 4T1 cells with blank medium or TExo, CDDP@TExo and CDDP@RKTExo for 24 h and quantitative results. Scale bar: 200 μm∗∗∗∗p < 0.0001, ∗∗∗p < 0.001, ∗∗p < 0.01, ∗p < 0.05.

The uptake of CDDP@RKTExo by 4T1 cells within 24 h was significantly higher than those of TExo and CDDP@TExo (Fig. S18). The maximum uptake of CDDP@RKTExo was 97.7%, which was increased with 1.1 and 1.07 folds higher than those of TExo (88.4%) and CDDP@TExo (90.9%), respectively (Fig. 7c), demonstrating enhancement of cell surface recognition performance due to ligand-receptor affinity effect. The results from cellular uptake indicate that the modified system exhibited a significant enhancement in its targeted recognition and uptake capabilities. Compared to TExo, the resistance to 4T1 cells of CDDP@TExo and CDDP@RKTExo has been significantly decreased, with CDDP@RKTExo demonstrating superior resistance to 4T1 cells (Fig. 7d). Meanwhile, CDDP@RKTExo also exhibited a significant pro-apoptotic effect on 4T1 cells within 24 h compared to the other groups (Fig. 7e–S19). The underlying mechanism might be contributed to that, first, natural lipids composed bilayer nanovesicles with particle size of approximately 100 nm were favorable for penetration. Second, RGP and KDEL reduced the structural integrity of tight junction and increased the paracellular permeability. Within 6 h, the ROS levels in 4T1 cells were significantly elevated by the modified TExo up to fivefolds (Fig. 7f). Compared to the control group, the ROS fluorescence intensity in the CDDP group was significantly increased (p < 0.01), and the ROS fluorescence intensity in the TExo group was further increased compared to the CDDP group (p < 0.0001), further supporting the aforementioned conclusions (Fig. S5). The drug-loaded modified CDDP@RKTExo significantly reduced the migration ability of 4T1 cells (Fig. 7g₁, g₂). CDDP@RKTExo significantly attenuated 4T1 cell invasion. Quantitatively, the invasion rate was reduced by 3.02-, 2.24-, and 1.31-fold compared to the CON, TExo, and CDDP@TExo groups, respectively (Fig. 7h₁, h₂).

2.7. Transdermal and targeting programmed delivery of CDDP@RKTExo-MN retards TNBC in vivo

The in vivo anti-tumor efficacy of CDDP@RKTExo was further evaluated in 4T1 tumor-bearing mice, which were randomly divided into 6 groups and treated with different formulations via MN (PBS, CDDP-IV, CDDP-MN, TExo-MN, CDDP@TExo-MN, and CDDP@RKTExo-MN). Treatments were initiated when tumors reached 100 mm³, with drug administration every two days and tumor volume and body weight monitored throughout the study (Fig. 8a). Body weight remained stable across all groups except CDDP-IV-treated and CDDP-MN-treated mice, highlighting the low systemic toxicity of our strategy (Fig. 8b). The sizes and weights of the tumors after 14 days of treatment and resection showed that the tumor in the CDDP@RKTExo-MN group was significantly smaller than those in the other groups, with a volume reduction of 42.69% compared to the CON group (Fig. 8c–S20, S21). Meanwhile, the tumor growth trends over 14 days were statistically analyzed, displaying that tumors in the control group exhibited explosive growth, while all drug treatment groups showed varying degrees of tumor growth inhibition.

Fig. 8.

Anti-tumor and angiogenesis inhibition efficacy of TExo, CDDP@TExo and CDDP@RKTExo in vivo. (a) Diagram of the establishment of TNBC mice model and treatment regimen. (b,c) Mouse weight and tumor volume variations over 14 days after the treatment of different treatments. (d) Histological analysis of resected tumor sections from different treatments. Scale bar: 100 μm. (e₁-e₄) TUNEL assay, Ki67 assay, VEGF and CD31staining quantification. Data are presented as mean ± SD (n = 6 biologically independent animals).

To further evaluate pathological changes in tumor tissues, we performed hematoxylin-eosin (H&E) and immunofluorescence staining. H&E staining of the control-MN group revealed typical malignant features, including high cell density, disorganized structure, irregular nuclei, and high mitotic activity (Fig. 8d). In contrast, drug-treated groups exhibited reduced cell density, lighter nuclear staining, smaller nucleoli, and decreased mitotic activity, with CDDP@RKTExo-MN showing the most pronounced effects, including extensive nuclear fragmentation and necrotic areas. Meanwhile, Ki67 staining revealed that the control group exhibited the deepest brown spots, indicating the highest proliferative activity. By contrast, the CDDP@RKTExo-MN group showed minimal brown spots, demonstrating significant inhibition of tumor cell proliferation (Fig. 8e₁). To assess the extent of apoptosis in tumor tissues following drug intervention, we performed TUNEL staining. Results demonstrated that the control tumors exhibited only sparse green fluorescence, indicating minimal apoptotic activity (Fig. 8e₂). In contrast, the CDDP@RKTExo-MN treated group displayed widespread green fluorescence across the observed area, suggesting robust induction of tumor cell apoptosis and potent anti-proliferative effects. Furthermore, CD31 is a recommended marker in the clinical diagnosis and treatment guidelines for breast cancer, and has high sensitivity and specificity in reflecting breast cancer metastasis. The percentage of VEGF and CD31 positive tumor cells were much higher in control groups, confirming the strong angiogenesis inhibition capacities of CDDP@RKTExo-MN (Fig. 8e₃, e₄).

2.8. The molecular interference of CDDP@RKTExo-MN to TNBC in vivo

Building on the potent in vivo antitumor efficacy, Western blot analysis further elucidated a coordinated dual-mechanism. Consistent with in vitro findings, treatment with CDDP@RKTExo-MN significantly suppressed the MAPK-driven proliferative signaling, which indicated by downregulated p-ERK/ERK, p-P38/P38, and p-JNK/JNK. Conversely, it activated the pro-apoptotic and immunogenic signaling pathways, evidenced by the upregulation of TNF-α, p-NF-ĸB, and p-AKT. This distinct regulatory pattern suggests that the therapeutic system acts by simultaneously inhibiting MAPK proliferation signals while triggering TNF-α-inducting extrinsic apoptotic cascades, thereby synergistically driving tumor regression (Fig. 9a_1-a₂, b_1-b₂). Furthermore, the expression levels of key endoplasmic-reticulum-related proteins were examined in tumor tissues following treatment with different groups. In vivo WB analysis also confirmed that the treatment of CDDP@RKTExo towards TNBC mice dependent on cholesterol metabolism/MAPK pathway (Fig. 9c₁-c₂, d₁-d₂) with higher expression of Caspase3/9/12 and lower expression of MMP2/9. Consistent with the in vitro results, the CDDP@RKTExo treatment group exhibited a 2.01-fold downregulation of p-ERK1/2/ERK1/2, a 1.81-fold downregulation of p-p38/p38 and a 2.00-fold downregulation of p-JNK/JNK. Typically, mild stress activates MAPK for survival. However, the observed global suppression of MAPK phosphorylation suggests a collapse of the cellular homeostatic machinery, rendering the cells unable to counteract the oxidative damage induced by the CDDP@TExo system. ELISA experiments indicated CDDP@RKTExo-MN had a significant decrease in pro-inflammatory factors level such as IL-1β, IL-6, and TNF-α, while the multifunctional negative regulator IL-10 showed a notable increase in secretion (Fig. 9e–h). These findings further validate the dual-pathway antitumor mechanism of CDDP@RKTExo in that, on the one hand, TExo-mediated activation of the MAPK/TNF-α axis induces oxidative stress and apoptosis; on the other hand, targeting the MMP2/9 and Caspase3/9/12 axis enhances tumor cell anti-endoplasmic-reticulum, thereby amplifying the antitumor efficacy.

Fig. 9.

The anti-cancer molecular mechanisms of CDDP@RKTExo in vivo. (a₁, a₂) WB analysis and Quantitation of MAPK signaling pathway,and (b₁, b₂) TNF signaling pathway. WB analysis and Quantitation of (c₁, c₂) WB analysis and Quantitation of tumour apoptosis and (d₁, d₂) metastasis. (e-h) Expression levels of inflammatory factors in mouse serum. (i) Diagram of the in vivo mechanism of action.∗∗∗∗p < 0.0001, ∗∗∗p < 0.001, ∗∗p < 0.01, ∗p < 0.05.

2.9. The biosafety of CDDP@RKTExo-MN in vivo

Previous studies have demonstrated with the exception of the aril, all parts of the yew plant contain taxine alkaloids and are poisonous, and Nephrotoxicity of Cisplatin is the most well-known and potentially most clinically significant toxicity. To verify the in vivo biosafety of CDDP@RKTExo-MN, bio-tissues were sectioned and stained by H&E. H&E staining images reveal that all groups showed no significant organ lesions (Fig. 10a). The treated mice exhibited no significant abnormalities in body weight, major organ H&E staining, blood routine tests, or liver function indicators (Fig. 10b₁-b₇, c₁-c₄). This indicated that the treatment had good biosafety. The aforementioned biosafety analysis experiments illustrated that, on one hand, TExo exhibited no obvious toxicity in short term; on the other hand, the research system employed an active targeting strategy combined with transdermal administration, which not only enhanced the targeted delivery of Taxus and cisplatin, reducing the systemic toxicity compared to free cisplatin., but also significantly decreases the circulation and accumulation concentration of the drug in the bloodstream of organs through transdermal-mediated local delivery, thereby reducing systemic toxicity.

Fig. 10.

Biosafety evaluation of CDDP@RKTExo in vivo. (a) H&E staining images of major organs after being treated with TExo, CDDP@TExo and CDDP@RKTExo. Scale bar = 100 μm (b1-b7) Blood routine analysis (WBC, white blood cell; RBC, red blood cell; HGB, hemoglobin; PLT, blood platelet; Lymph, lymphocyte; Mon, monocytes; Gran, neutrophilicgranulocyte. (c₁-c₅) Blood biochemistry analysis (ALT, alanine aminotransferase; AST, aspartate aminotransferase; ALP, alkaline phosphatase; UREA, urea nitrogen; CREA, creatinine; n = 3).∗∗∗∗p < 0.0001, ∗∗∗p < 0.001, ∗∗p < 0.01, ∗p < 0.05.

3. Discussion

TNBC remains a significant clinical challenge due to its aggressive nature, absence of targetable receptors, and resistance to conventional therapies. Current treatment strategies are limited by poor drug delivery efficiency, inadequate tumor specificity, and significant systemic toxicity, particularly with chemotherapeutics like CDDP [53]. Despite that exosomes have emerged as both nanodrugs and promising nanocarriers for cancer therapy [54], contributing to their intrinsic biocompatibility and capacity for therapeutic miRNA delivery, their therapeutic potential is still hindered by limited active-targeting capabilities to mammalian cells and efficacy in overcoming stromal barriers fluently [55]. These limitations underscore the urgent need for innovative strategies that integrate precise drug delivery with multi-targeted therapeutic efficacy. To address these issues, herein, we developed a novel intelligent transdermal-cascade targeting programmed exosomal-cisplatin synergistic nanoplatform with spatiotemporally precise control to TNBC.

The proposed intelligent platform, CDDP@RKTExo-MN achieves precise multidimensional targeting by integrating three synergistic innovations that collectively overcome stromal, tumor, and subcellular barriers in the management of TNBC. Exploiting the pathological overexpression of αvβ₃ integrins in TNBC vasculature and tumor cells, RGP-modified plant-derived exosome-like vesicles (PLEVs) demonstrate enhanced tumor specificity via ligand-receptor interactions, effectively bypassing the stromal-rich microenvironment that typically impedes nanoparticle penetration through aberrant extracellular matrix architecture and elevated interstitial fluid pressure. In parallel, the incorporation of KDEL trafficking signals capitalizes on the tumor's susceptibility to ER stress, directing cisplatin-loaded PLEVs to the ER to amplify drug potency. Additionally, the microneedle array facilitates spatiotemporal control over drug release kinetics through its dissolvable polymeric matrix, achieving sustained delivery of the therapeutic payload that maintains tumoricidal concentrations while mitigating systemic toxicities. These coordinated modifications address the combined challenges of stromal exclusion, inadequate intracellular drug accumulation, and systemic toxicity via physiologically informed targeting strategies that exploit both tumor pathophysiology and subcellular vulnerabilities.

The platform further leverages the synergistic interactions between the intrinsic miRNAs of TExo and cisplatin to achieve multi-modal therapeutic efficacy. The PLEVs deliver a repertoire of conserved miRNAs that regulate key signaling pathways, such as MAPK and TNF, to induce apoptosis and modulate the tumor microenvironment. When combined with cisplatin-induced ER stress, these miRNAs enhance apoptotic signaling, generating a synthetic lethality effect that potentiates tumor cell death. Moreover, the platform reprograms the immunosuppressive TNBC microenvironment by reducing pro-angiogenic factors and inflammatory cytokines while promoting anti-tumor immune activation. However, whether this kind of sustained and localized release microneedle platform will cause a long-term chronic toxicity still need to be studied. In conclusion, This cascade targeting mechanism-targeting both tumor cells and their supportive stroma-marks a significant advancement over conventional single-target therapies.

4. Conclusion

This study introduces a novel intelligent plant-derived exosome-like vesicles delivery system for TNBC treatment, combining cascade-targeting treatment with chemotherapy into a transdermal delivery platform. The system effectively treated the tumor with all the potential side effects and toxicity reduced. These results offer new solutions for TNBC treatment and exemplify the successful integration of natural bioactive miRNAs with chemical drug for diseases therapy. The presented design and platform provide the flexible framework that can be extended in treating various challenging malign.

5. Methods

5.1. Cell source and culture

4T1 cell line (RRID:CVCL_0125) was obtained from the Institute of Biochemistry and Cell Biology of Chinese Academy of Science on Sept.25,2022. STR authentication was performed confirming the cell line was contamination free. The cells were cultured in RPMI 1640 medium containing 10% FBS, L-glutamine, penicillin (50 U/mL), streptomycin (50 U/mL), and were maintained at 37 °C with 5% CO₂.

5.2. Isolation and characterization of TExo

For the isolation and purification of plant-derived exosome-like vesicles (PLEVs), TExos were extracted from the juice of Taxus chinensis (Qianfeng South Medicine Garden in Guangzhou). The procedure involved initially grinding the plant material in a blender to obtain the juice, which was then subjected to a series of centrifugation steps at 1000×g for 5-20 min, 4000×g for 20-60 min, and finally 10,000×g for 60-120 min to eliminate larger debris. The resulting supernatant was subsequently ultra-centrifuged at 150,000 g for 60-120 min using an XPN-100 centrifuge (Beckman Coulter), and the pellet was resuspended in PBS and then filtered through a 0.22-μm membrane.

The harvested TExos were stored at −80 °C until further analysis. The particle size and zeta potential of the PLEVs were assessed using a Zeta-sizer Nano Z90 instrument (Malvern Instruments). The particle concentration of the PLEVs was measured via a nanoparticle tracking analyzer (Nanosight NS500, Malvern Instruments), following the manufacturer's instructions. The morphology of the PLEVs was characterized by transmission electron microscopy using a Talos L120C microscope (Thermo Fisher).

5.3. Construction and characterization of CDDP@RKTExo and CDDP@RKTExo-MN

Engineered TExo modifications were prepared through sequential ultrasonication and incubation protocols. TExos were combined with cisplatin (CDDP) at a 1:1 mass ratio (w/w) in sterile PBS, followed by pulsed ultrasound treatment (20% amplitude, 4 s on/2 s off per cycle) administered over 6 cycles (total duration: 3 min) using a probe sonicator maintained at 4 °C via ice bath cooling between cycles. Post-sonication, samples were incubated at 37 °C for 60 min to restore membrane integrity, then subjected to ultracentrifugation at 150,000×g for 90 min at 4 °C to isolate CDDP@TExo. Residual free drug was removed through three wash cycles with PBS under identical centrifugation parameters, with final product purity verified via nanoparticle tracking analysis and drug encapsulation efficiency. Then, CDDP@TExo was mixed with RGP and KDEL at a weight ratio of 1:1 (W:W), respectively. The mixtures were then agitated at 37 °C with 100 rpm for 60 min. Subsequently, the mixtures were transferred to 100KD ultrafiltration tubes and centrifuged at 3000 rpm for 15 min to collect the precipitate of the engineered PLEVs loaded with CDDP (CDDP@RKTExo). Finally, 100 μg of CDDP@RKTExo was mixed with porous microneedles, shaked at room temperature (100 rpm) for 60 min, and removed excess PLEVs to obtain CDDP@RKTExo-MN.

To determine the grafting efficiency of ligand on the surface of the PLEV, a fluorescence-based quantification assay was performed. Briefly, the RGP and KDEL was respectively labeled with FITC and TRAMA according to the manufacturer's protocol. The mixture was purified by centrifugation to remove unbound ligands. The fluorescence intensity of the supernatant was measured. The grafting efficiency was calculated as:

$$\mathrm{G}\text{rafting}\mathrm{E}\text{fficiency}(\%)=\frac{\text{Total}\text{Input}\text{Fluorescence}-\text{Unbound}\text{Fluorescence}}{\text{Total}\text{Input}\text{Fluorescence}}\times 100\%$$

The precipitate was resuspended with a small amount of PBS, and the particle size and zeta potential of the PLEVs were measured using a Malvern Zetasizer NanoSizer (Malvern Instruments). The morphology of the CDDP@RKTExo was characterized using transmission electron microscopy (Talos L120C, Thermo Fisher). For the final preparation of the CDDP@RKTExo-MN system, morphological observations was conducted using optical microscopy and scanning electron microscopy (SEM), followed by elemental scanning to evaluate the drug loading capacity of the TExos. Finally, drug release profiles were simulated at different time points on the complete skin of mice using a small transdermal device for detailed characterization.

5.4. MiRNAs in TExos analysis

rRNA was depleted using Trizol reagent, followed by fragmentation of purified mRNA into short segments (150–300 bp) via divalent cation-mediated hydrolysis. First-strand cDNA synthesis was performed using random hexamer primers, with subsequent end repair, dA-tailing, and adapter ligation steps completed using a library preparation kit. Size-selected fragments (200–500 bp) were recovered via agarose gel electrophoresis, and library quality was validated through RT-qPCR quantification. Sequencing was conducted on an Illumina NovaSeq 6000 platform (150 bp paired-end reads), after which raw data underwent quality trimming (Phred score ≥30), adapter removal, and low-complexity sequence filtering. High-quality reads were aligned to reference transcripts using HISAT2 (v2.2.1) and Bowtie2 (v2.4.2), with gene expression quantified as FPKM (fragments per kilobase of transcript per million mapped reads). Differential expression analysis was performed in R (v4.2.0) using DESeq2 (v1.36.0) under a negative binomial model, identifying genes meeting thresholds of |log2 (fold change)| ≥1 and Benjamini-Hochberg-adjusted FDR <0.05. Enriched Gene Ontology (GO) terms and KEGG pathways were identified through hypergeometric testing with Bonferroni-corrected p-values <0.05 considered statistically significant.

5.5. Lipid analysis

Separation and analysis of lipids contained in TExos: Lipids were extracted using the MTBE method. An aliquot of TExos solution was combined with 200 μL of deionized water and vortex-mixed thoroughly. Subsequently, 800 μL of methyl tert-butyl ether (MTBE) was added, followed by additional vortex mixing. The mixture was treated with 240 μL of pre-cooled methanol (4 °C), vigorously vortexed, and subjected to ultrasonic extraction in a temperature-controlled water bath (4 °C) for 20 min. After equilibration at room temperature (25 °C) for 30 min, the sample was centrifuged at 14,000×g for 15 min at 10 °C. The upper organic phase was isolated and evaporated to dryness under a nitrogen stream.

For mass spectrometry analysis, the dried residue was reconstituted in 200 μL of a 90% isopropanol/acetonitrile (v/v) solution, vortexed for complete dissolution, and centrifuged at 14,000×g for 15 min at 10 °C. The supernatant was filtered through a 0.22 μm membrane prior to LC-MS/MS injection. Chromatographic separation was performed on a C18 column using mobile phase A (acetonitrile/water, 6:4 v/v, supplemented with 0.1% formic acid and 0.1 mM ammonium formate) and mobile phase B (acetonitrile/isopropanol, 1:9 v/v, containing 0.1% formic acid and 0.1 mM ammonium formate). A gradient elution program was implemented as follows: 0–2 min (30% B), 2–25 min (30%→100% B), 25–35 min (30% B), with a constant flow rate of 300 μL/min and column temperature maintained at 45 °C. Analytes were ionized via electrospray ionization (ESI) in alternating positive and negative ion modes following UHPLC separation. Mass spectral data were acquired in full-scan mode (m/z 150–2000) with parallel monitoring of lipid molecular ions and fragment ions. LipidSearch software (version 4.2) facilitated automated peak alignment, feature extraction, and lipid identification through spectral matching against the LIPID MAPS database, incorporating retention time correction and intensity normalization workflows.

5.6. Protein analysis

Protein quantification was conducted using a BCA assay kit, followed by TMT labeling where reagents were dissolved in acetonitrile, vortex-mixed, and centrifuged, with one vial of TMT reagent added per 100 μg of peptides. Post-labeling, first-dimensional separation was achieved via reversed-phase liquid chromatography (RPLC) coupled with tandem mass spectrometry (LC-MS/MS), while second-dimensional analysis employed nanoflow liquid chromatography-tandem mass spectrometry (nanoLC-MS/MS). Raw spectral counts were processed using the DESeq2 package under a negative binomial distribution model, with differentially expressed proteins (DEPs) identified using thresholds of |log2 (fold change)| ≥1 and false discovery rate (FDR) < 0.05, validated through Benjamini-Hochberg correction. SDT (4% SDS, 100 mM Tris-HCl, pH 7.6) buffer was used for sample lysis and protein extraction. 20 μg of protein for each sample were mixed with loading buffer respectively and boiled for 5 min. The proteins were separated on 4%–20% SDS-PAGE gel (constant voltage 180 V, 45 min). Protein bands were visualized by Coomassie Blue R-250 staining.

5.7. Small molecules in TExos analysis

Samples were slowly thawed at 4 °C, followed by the addition of pre-chilled methanol/acetonitrile/water solution (2:2:1, v/v) to appropriate aliquots. The mixture was vortex-mixed, subjected to low-temperature sonication for 30 min, incubated at −20 °C for 10 min, and centrifuged at 14,000×g for 20 min at 4 °C. The supernatant was vacuum-dried and reconstituted in 100 μL acetonitrile/water solution (1:1, v/v) for mass spectrometry analysis, then vortexed and centrifuged at 14,000×g for 15 min at 4 °C prior to injecting the supernatant. Chromatographic separation was performed on an Agilent 1290 Infinity LC UHPLC system equipped with a HILIC column maintained at 25 °C, with a flow rate of 0.5 mL/min and injection volume of 2 μL. Mobile phases consisted of (A) water containing 25 mM ammonium acetate and 25 mM ammonium hydroxide, and (B) acetonitrile. The gradient program was: 0–0.5 min (95% B), 0.5–7 min (95%→65% B, linear gradient), 7–8 min (65%→40% B, linear gradient), 8–9 min (40% B isocratic), 9–9.1 min (40%→95% B, linear gradient), and 9.1–12 min (95% B isocratic). MS1 and MS2 spectra were acquired using an AB Sciex TripleTOF 6600 mass spectrometer. Raw data were converted to. mzML format via ProteoWizard and processed using XCMS for peak alignment, retention time correction, and peak area extraction with parameter settings: centWave (m/z tolerance = 10 ppm, peakwidth = [10, 60], prefilter = [10, 100]); peak grouping (bw = 5, mzwid = 0.025, minfrac = 0.5). Metabolites with >50% missing values within groups were excluded, followed by KNN-imputation for missing values, outlier removal, and total peak area normalization to enable cross-sample comparability.

5.8. Differentially expressed genes analysis

Small RNA sequencing libraries were prepared using the Small RNA Sample Prep Kit and sequenced on an Illumina platform with single-end 50 bp reads (SE50). Raw sequencing data underwent quality control to remove low-quality reads, adapter contaminants, and polyA/T/G/C sequences, generating clean reads that were subsequently annotated and quantified for small RNA classification using miRDeep2 and SILVA databases. Length distribution profiling and category-specific annotation of small RNAs were performed using sRNAtoolbox. Differential miRNA expression analysis was conducted via DESeq2 under negative binomial distribution, with significantly differentially expressed miRNAs (DEMs) identified using thresholds of |log2 (fold change)| ≥1 and Benjamini-Hochberg-adjusted FDR ≤0.05. Putative targets of DEMs were predicted using TargetScan (v7.2), miRDB (2020 release), and miRWalk (v3.0), with consensus targets retained for functional enrichment analysis via clusterProfiler (v4.0) against Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) databases. Protein-protein interaction (PPI) networks were constructed using STRING (v11.5) with high-confidence interactions (combined score ≥0.9), visualized and analyzed in Cytoscape (v3.9.1) employing CytoHubba to identify top 20 hub genes based on maximal clique centrality (MCC). Integrated miRNA-mRNA regulatory networks were generated using miRNet (v2.0) with experimentally validated interactions from miRTarBase (v9.0).

5.9. Target gene analysis of TExos-miRNAs

Ribosomal RNA was depleted using a commercial rRNA removal kit, followed by fragmentation of purified mRNA into short segments. These fragmented mRNAs served as templates for cDNA synthesis via reverse transcription. Synthesized cDNA underwent end-repair, dA-tailing, and adapter ligation, with library fragments subsequently size-selected via agarose gel electrophoresis. Library quantification was performed using RT-qPCR with SYBR Green chemistry on a QuantStudio 5 system (Thermo Fisher), followed by paired-end sequencing (2 × 150 bp) on an Illumina NovaSeq 6000 platform. Raw sequencing data underwent quality filtering using Trimmomatic (v0.39) to remove adapter sequences and low-quality reads (Phred score <30). High-quality reads were aligned to reference transcriptomes using HISAT2 (v2.2.1) and Bowtie2 (v2.4.5) with default parameters, with gene expression quantified as FPKM (Fragments Per Kilobase Million) values through StringTie (v2.2.1). Differential expression analysis was conducted using DESeq2 (v1.34.0) under negative binomial distribution, identifying significantly differentially expressed genes (DEGs) meeting thresholds of |log2 (fold change)| ≥1 and Benjamini-Hochberg-adjusted FDR <0.05. Functional enrichment analysis of DEGs was performed using clusterProfiler (v4.4.4) against Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) databases, with pathway significance determined at adjusted p-value <0.05 through hypergeometric testing with Bonferroni correction.

5.10. Cellular uptake and internalization of TExos

A 1 mg/mL DiO stock solution was prepared by dissolving the dye in 25 μL DMSO, followed by dilution (100–200 × ) into exosome suspensions to achieve a 5–10 μM working concentration. The labeled exosomes were incubated at 37 °C for 30 min, with unbound dye removed via ultracentrifugation at 150,000×g for 1 h. Logarithmic-phase 4T1 cells were dissociated into single-cell suspensions and seeded into 6-well plates at 2 × 10⁵ cells/well. Following overnight adhesion in a humidified incubator (37 °C, 5% CO₂), cells were treated with DiO-labeled exosomes for 0, 3, 6, 9, 18, and 24 h. Post-incubation, culture medium was aspirated, and cells were washed twice with phosphate-buffered saline (PBS), trypsinized, and pelleted by centrifugation (300×g, 5 min). Cellular uptake efficiency was quantified via flow cytometry (CytoFlex S, Beckman)) by measuring DiO fluorescence intensity (ex/em: 484/501 nm) in ≥10,000 events per sample, normalized to untreated controls.

5.11. The CLSM analysis of the endoplasmic reticulum co-localization of 4T1 with engineered TExos

A confocal petri dish was used to collect 4T1 cells in the logarithmic growth phase, and a single-cell suspension was prepared by gentle pipetting. Cells were then seeded into the dish at a density of 1 × 10^⁵ cells per well and evenly distributed by gentle agitation. The dish was incubated at 37 °C with 5% CO₂ overnight. After attachment, the original medium was removed and replaced with complete medium containing 1411 μg/mL TExos, CDDP@TExo, and CDDP@RKTExo, with a control group established. The plates were further incubated for 12 h. Fluorescence imaging of the samples before and after treatment was conducted using an ER fluorescent probe detection kit, following the manufacturer's instructions, with an excitation wavelength of 587 nm and an emission wavelength of 615 nm.

5.12. Cell viability assay

5 × 10³ cells were seeded in a 96-well plate per cell and cultured for 24 h. TExos of different protein concentrations (0, 1, 10, 50, 100, 200 μg/mL) were added into each well. After incubation for 24 h, the cell proliferation was then evaluated by cell count kit-8 assay (C0038, Beyotime) according to the manufacturer's instructions and determined by measuring absorbance at 450 nm by the microplate reader (Multiskan FC, Thermo Fisher).

5.13. Cell migration and invasion assays

For cell migration, scratch assay was conducted to evaluate the antimigration capacities of TExos, CDDP@TExo, and CDDP@RKTExo against 4T1 cells. Cells were seeded in 6-well plate at 1 × 10⁵ cells/well and culture for 24 h. The confluent cell monolayers were wounded with a sterile 200 μL pipet tip. Thereafter, cells were gently washed, and the wounds were imaged prior to/after the addition of 500 μg/mL Serum-free culture medium of TExos, CDDP@TExo, and CDDP@RKTExo, respectively. Samples were collected at 6, 12, and 24 h post-treatment for photography. The relative covering wound area was analyzed using Image J. For cell invasion, 80 μL Matrigel (dilution 1:8, 0827045, ABW) was coated onto each well of the upper chamber Transwell plate (8 μm, 3422, Corning) incubated at 37 °C for 30 min, and 4T1 cells were seeded into the at 5 × 10⁴ cells/well with serum-free media containing 500 μg/mL TExos, CDDP@TExo, and CDDP@RKTExo or not. The media containing FBS (10%, v/v) were added into the lower chamber. FBS was used as a chemoattractant that attracted cells to cross membrane. After 24 h in-cubation, cells on the upper surface of membranes were scrubbed off with humid cotton buds, while cells on the bottom surface were immobilized with 4% paraformaldehyde and stained with 0.1% crystal violet for 20 min. Images were taken with a microscope (ECLIPSE-TI-S, Nikon).

5.14. Cell apoptosis assay

4T1 cells were seeded in 6-well plates at 1 × 10⁵ cells/well and cultured for 24 h. Cells were exposed to 500 μg/mL TExos, CDDP@TExo, and CDDP@RKTExo for 24 h. Then the cells were collected and stained using Annexin V-FITC Apoptosis Detection Kit (40302ES50, Yeasen). Finally, cells were analyzed immediately by flow cytometry (CytoFlex S, Beckman).

5.15. ROS detection

For confocal microscopy analysis, logarithmic-phase 4T1 cells were dissociated into single-cell suspensions and seeded onto confocal dishes at 1 × 10⁵ cells/dish, followed by gentle agitation to ensure uniform distribution. Cells were incubated overnight in a humidified atmosphere (37 °C, 5% CO₂), after which the medium was replaced with complete medium containing 1411 μg/mL TExos (experimental group) or vehicle control (control group) for 4 h. Intracellular ROS levels were quantified using a commercial ROS detection kit according to the manufacturer's protocol, with imaging performed on confocal microscope (Carl Zeiss AG) using 488 nm excitation/525 nm emission filters. For flow cytometry, parallel-treated cells in 6-well plates (1 × 10⁵ cells/well) were harvested by trypsinization, washed with PBS, and stained with the same ROS probe. Fluorescence intensity was measured using flow cytometer (BD Biosciences).

5.16. Western blot analysis

Total proteins were extracted using a protein extraction reagent (78510, Thermo Scientific™ Pierce™) supplemented with a protease inhibitor cocktail (78445, Thermo Scientific™) and quantified via a BCA assay kit (P0010, Beyotime Biotechnology). SDS-PAGE was performed with 8–12% resolving gels and 5% stacking gels under constant voltage (60 V for stacking, 80 V for resolving), with 60 μg total protein (10–15 μL per lane) loaded and separated for 2 h. For electroblotting, PVDF membranes (IPVH00010, Millipore) were pretreated with methanol and equilibrated in Tris-glycine transfer buffer (25 mM Tris, 192 mM glycine, 5% methanol, pH 8.3). Gels and membranes were assembled in a wet transfer system (Bio-Rad) and transferred at 100 V for 2 h at 4 °C. Post-transfer, membranes were blocked with 5% BSA in TBST (20 mM Tris, 150 mM NaCl, 0.1% Tween-20, pH 7.6) for 1 h at room temperature, followed by three washes (5 min each) in TBST. Primary antibodies were diluted in blocking buffer and incubated overnight at 4 °C, after which membranes were washed three times in TBST and probed with HRP-conjugated secondary antibodies [goat anti-rabbit IgG (31210, Thermo Scientific™ Pierce™, 1:5000) and goat anti-mouse IgG (31160, Thermo Scientific™ Pierce™, 1:5000)] for 1 h at room temperature. Following five washes in TBST (5 min each), protein bands were visualized using SuperSignal™ West Dura Extended Duration Substrate (34075, Thermo Scientific™) prepared as a 1:1 mixture of peroxide solution and luminol/enhancer solution. Membranes were incubated with ECL reagent for 1 min, excess solution removed, and signals captured on X-ray film (Huadong Pharmaceutical Co., Ltd.) with exposure times optimized between 5 and 10 min. Band densitometry was analyzed using ImageJ (v1.53k), with results expressed as mean ± standard deviation from triplicate experiments.

5.17. Animal study

BALB/c mice (6 weeks of age, female) were purchased from Shanghai SLAC Laboratory Animal Co. Ltd. All animal experimental procedures were performed in obedience to the guidelines and protocols of the Animal Experimental Ethics Committee of Zhejiang University. The mice were randomly divided into 6 groups (n = 6 per group) using a randomized block design to minimize baseline weight variations. To reduce bias, tumor volume measurements were performed by an investigator blinded to the treatment grouping. Throughout the treatment period, the general health status of the mice was monitored daily. No mortality or abnormal behaviors were observed in any group. Subcu-taneous transplanted breast tumor was used to evaluate the in vivo antibreast cancer capacity of TExos. Briefly, 1 × 10⁷ 4T1 cells were sus-pended in free-serum medium (100 μL) and subcutaneously injected at right flank BALB/c mice. When the average volumes of tumors reached 50–100 mm³, mice were treated with 30 μL PBS or TExos at the con-centrations of 10 mg protein/kg via intratumoral injection and the tumor volumes and body weights of mice were recorded every 3 days. The tumor tissues were retrieved on days 14 after the treatment. The tumor volume was calculated using the following formula: width² × length × 0.5. Female BALB/c mice (6 weeks old) were randomized into two groups and subcutaneously inoculated with tumor cells to establish xenograft models. For ex vivo imaging, exosomes were fluorescently labeled by incubating with 1 mg/mL DiR dye (dissolved in 25 μL DMSO) diluted 100–200 × into exosome suspensions (final 5–10 μM working concentration) for 30 min at 37 °C, followed by removal of unbound dye via ultracentrifugation at 150,000×g for 1 h. When tumor volumes reached 100 mm³, mice received microneedle-mediated systemic administration of taxus-modified exosomes (DiR-labeled) or unmodified controls.

5.18. Immunofluorescence tissue staining

Paraffin-embedded tissue sections were deparaffinized through sequential immersion in xylene I (15 min), xylene II (15 min), absolute ethanol I (5 min), absolute ethanol II (5 min), 95% ethanol (5 min), 85% ethanol (5 min), and 75% ethanol (5 min), followed by rehydration in distilled water. Antigen retrieval was performed by immersing slides in sodium citrate buffer (pH 6.0) and heating in a pressure cooker until steam generation, maintaining the treatment for 2 min before cooling to room temperature. Non-specific binding was blocked with 10% goat serum in PBS at 37 °C for 30 min. Primary antibodies (VEGF at 1:200 dilution, CD31 at 1:500 dilution) diluted in antibody buffer were applied to sections and incubated overnight at 4 °C in a humidified chamber. After washing with PBST (PBS containing 0.1% Tween-20), HRP-conjugated secondary antibodies (1:100 dilution in PBST) were added and incubated at 37 °C for 1 h. For TUNEL staining, a premixed reagent (1:10 ratio of component A to B) was applied to sections and incubated overnight at 4 °C. Nuclei were counterstained with DAPI (1 μg/mL) for 5 min in the dark, washed three times with PBS, and mounted with anti-fade mounting medium. Imaging was performed using a digital slide scanner with consistent exposure settings across all samples.

5.19. H&E staining

After dewaxing, sections were washed, followed by the addition of hematoxylin (Hangzhou Hulk Biotechnology Co. Ltd) for staining for 5 min, hydrochloric acid aqueous solution for 2 s and ammonia aqueous differentiation solution (10,011,018, Sinopharm Chemical Agents Co. Ltd)1 for 15–30 s, and then washed. The slices were dehydrated with 95% alcohol and then stained with eosin solution (Hangzhou Hulk Biotechnology Co. Ltd) for 5–8 s. After dehydration and sealing, the slices were observed under microscope.

5.20. Biosafety validation of CDDP@RKTExo-MN

The experiment concluded on the 14th day after administration in each group of mice. Major organs and blood samples were collected from the mice (n = 3) for H&E staining and biochemical analysis, respectively, to evaluate systemic toxicity. Whole blood was collected from the mice's ocular cavities. A portion of the samples was placed in vacuum anticoagulant tubes for the detection and analysis of routine blood parameters at the end of the experiment. The other portion was placed in regular centrifuge tubes and allowed to stand at 4 °C. After the samples had separated, they were kept at room temperature for several minutes and then centrifuged at 3000 rpm for 15 min in a refrigerated centrifuge to separate and collect the serum. Subsequently, blood biochemical indicators were analyzed using an automatic biochemical analyzer. Ultimately, the hearts, livers, spleens, lungs, kidneys, and brains of the mice from each group were harvested and placed in 10% neutral formalin for fixation. Subsequently, these organs underwent paraffin embedding, and sectioning was performed using a microtome. The sections were then stained with hematoxylin and eosin to assess the safety of the treatment.

5.21. Statistical analysis

All quantitative data are presented as mean ± standard deviation (SD) derived from at least three biologically independent experiments. Statistical analyses were performed using GraphPad Prism software (Version 9.0). For comparisons between two groups, an unpaired two-tailed Student's t-test was utilized. For comparisons among multiple groups (>2 groups), a one-way analysis of variance (ANOVA) followed by Tukey's post-hoc test was applied to determine statistical significance. A PP-value of <0.05 was considered statistically significant.

CRediT authorship contribution statement

Xi-Yuan Xu: Writing – review & editing, Writing – original draft, Investigation, Data curation. Xuan Zhang: Visualization, Investigation, Data curation. Yi-Yi Wang: Writing – review & editing, Investigation, Data curation. Qi-Yao Xiao: Validation, Investigation. Cheng Yang: Software, Resources, Methodology. De-Sheng Kong: Software, Resources, Methodology. Li-Hua Peng: Writing – review & editing, Supervision, Funding acquisition.

Ethics approval and consent to participate

The animal experiment protocol was approved by the Institutional Animal Experimental Ethics Committee of Zhejiang University (Ethics Approval Number: 25468) and conducted in accordance with the ARRIVE guidelines and the National Institutes of Health (NIH) Guide for the Care and Use of Laboratory Animals. All animals were housed under specific pathogen-free conditions with a 12-h light/dark cycle, ad libitum access to food and water, and daily health monitoring. Every effort was made to minimize animal suffering and reduce the number of animals used.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

The following is the Supplementary data to this article:

Data availability

All data generated or analyzed during this study are included in this published article and its supplementary information files.