Intercellular Horizontal Transfer of TXNDC5 mRNA via Extracellular Vesicles Contributes to Tumor‐Associated Macrophage‐Mediated Prostate Cancer Metastasis

ABSTRACT

As one of the predominant male malignancies globally, prostate cancer (PCa) transitions to a treatment‐refractory phase upon metastasis, for which no curative modalities currently exist. Tumor‐associated macrophages (TAMs), a crucial component of the tumor microenvironment (TME), primarily adopt a metastasis‐promoting M2 phenotype. However, the mechanisms underlying the TAM‐cancer cell crosstalk and resultant PCa metastasis remain elusive. In this study, primary lesions of metastatic PCa (mPCa) exhibit both greater infiltration of M2 macrophages and a higher proportion of M2 macrophage‐derived extracellular vesicles (M2 EVs) compared to those of non‐metastatic PCa (nmPCa). Furthermore, M2 EVs can be internalized by PCa cells, promoting a mesenchymal‐like state (MLS) in PCa and affecting tumor metastasis. Mechanistically, thioredoxin domain‐containing 5 (TXNDC5) mRNA encapsulated in M2 EVs contributes to MLS of DU145 and PC3 cells, enhancing migration and invasion. Single‐vesicle particle analysis confirms that TXNDC5 mRNA encapsulated within M2 EVs can be horizontally transferred to target cells, where it is translated to produce functional proteins. In conclusion, our study demonstrates that M2 macrophages can promote MLS and metastasis of PCa through EV‐mediated horizontal mRNA transfer. A novel role of EVs in the communication between the TME and tumor cells is discovered, offering new insights into tumor metastasis.

This study investigates the role of M2 macrophage‐derived extracellular vesicles (M2 EVs) in prostate cancer (PCa) metastasis. Mechanistically, M2 EVs horizontally transfer TXNDC5 mRNA to PCa cells, where the molecule induces the mesenchymal‐like state (MLS), enhancing PCa migration and invasion. Interfering with specific aspects of this cellular interaction process may serve as novel strategies for controlling metastasis.

1. Introduction

Prostate cancer (PCa) accounts for 29% of the estimated new cancer cases among men in the United States [1]. Tumor metastasis is the leading cause of death in PCa patients. The five‐year survival rate for patients with non‐metastatic PCa (nmPCa) exceeds 98%, whereas it drops to less than 30% once metastasis occurs [2, 3]. Tumor metastasis is a complex process involving multiple factors and stages, with the tumor microenvironment (TME) playing a crucial role in PCa [4, 5].

Tumor‐associated macrophages (TAMs) are integral constituents of the TME and are broadly classified as tumor‐suppressive M1 and tumor‐promoting M2 macrophages [6, 7, 8]. TAMs predominantly exhibit an M2 phenotype, characterized by the elevated expression of CD163 and CD206 [9, 10]. These cells secrete a variety of extracellular signals that play critical roles in tumor invasion, angiogenesis, drug resistance, and immune suppression, thereby driving disease progression [9, 11]. Numerous studies have identified the mesenchymal‐like state (MLS) as a particularly significant outcome of phenotypic plasticity capable of driving tumor progression in various cancer contexts [12, 13, 14, 15]. MLS refers to a state in which cells exhibit characteristics or properties resembling those of mesenchymal cells. Mesenchymal‐like cancer cells not only undergo changes in cell morphology but also exhibit altered gene expression profiles and rewired signaling pathways. Studies have shown that crosstalk between mesenchymal‐like cells and immune cells enhances the mesenchymal‐like phenotype in cancer cells, thereby promoting tumor growth and metastasis [12, 16, 17, 18]. TAMs are closely associated with the induction of MLS [12]. However, the mechanisms by which TAMs promote tumor metastasis through MLS in PCa remain incompletely understood. Extracellular vesicles (EVs), which possess a lipid bilayer nanostructure, serve as crucial mediators of intercellular communication [19]. EVs carry bioactive molecules, including proteins, nucleic acids, and metabolites, and facilitate a wide range of biological effects in recipient cells [20]. Many studies have highlighted the pivotal role of EVs in tumor initiation and progression [21, 22]. In PCa, tumor‐derived exosomal CXCL14 facilitates M2 polarization via the NF‐κB signaling pathway, thereby driving PCa progression [23]. Another study found that EV‐mediated transfer of RNF157 from PCa cells resulted in M2 polarization via destabilization of HDAC1 [24]. However, it is unclear whether M2 macrophage‐derived EVs (M2 EVs) can facilitate MLS and metastasis in the context of PCa.

In our study, we demonstrated that primary lesions of metastatic PCa (mPCa) exhibit increased M2 macrophage infiltration and a higher M2 EVs/tissue‐derived EVs (Ti‐EVs) ratio than those of nmPCa. M2 EVs promote MLS and facilitate PCa metastasis through horizontal transfer of thioredoxin domain‐containing 5 (TXNDC5) mRNA. In summary, our study elucidated a novel mechanism of intercellular communication between tumors and TAMs via EVs in the TME of PCa. Additionally, our study expanded the understanding of EVs and their roles in tumor progression, providing insights into the development of novel therapeutic strategies.

2. Result

2.1. Primary Lesions of mPCa Harbor More M2 Macrophages and Exhibit a Higher Proportion of M2 Macrophage‐derived EVs Compared to Those of nmPCa

Comparisons between metastatic and primary lesions, or between tumor and adjacent tissues, are insufficient for identifying the intrinsic factors that drive the initial stages of cancer metastasis. To elucidate the differences in M2 macrophage infiltration between nmPCa and mPCa primary sites, we integrated large‐scale single‐cell RNA sequencing data, constructing a comprehensive cohort for analysis. This study ultimately incorporated 77 PCa primary samples (46 mPCa and 31 nmPCa), successfully capturing a total of 3 49 850 cells after cell filtration for in‐depth analysis. This resource encompasses samples from public databases and our own sequencing data (which included 3 mPCa and 4 nmPCa samples) (Figure 1A). Based on the detection of marker genes in the integrated single‐cell transcriptomes, all acquired cells were subsequently clustered into 10 main cell types using the UMAP algorithm, including luminal epithelial cells, basal epithelial cells, T cells, endothelial cells, smooth muscle cells, macrophages, fibroblasts, and several other cell types (Figure S1A,B). Pronounced differences between mPCa and nmPCa were observed in the composition of cells in the primary TME (Figure S1C). Notably, the primary lesions of mPCa patients contained a considerably larger proportion of macrophages than those of nmPCa patients (Figure S1D). Following the characterization of overall cell composition, we performed a focused analysis of the macrophage compartment. Based on the expression of canonical M1 and M2 marker genes, we resolved this population into three transcriptionally defined subsets: M1, M2, and a Mixed group (Figure S1E). UMAP visualization demonstrated the distribution of each subset (Figure 1B). Further analysis revealed significant heterogeneity in the macrophage composition between the two groups of patients (Figure S1F,G). Specifically, primary lesions of mPCa patients exhibited a lower proportion of M1 macrophages and a higher proportion of M2 macrophages than those of nmPCa patients (Figure 1C). Furthermore, a higher M2 signature score was associated with worse patient prognosis (Figure 1D). We next employed IHC and IF to spatially resolve the expression of key M2‐associated proteins (represented by CD68, CD163, and CD206) in primary mPCa and nmPCa sites. IHC revealed a higher proportion of CD163⁺ cells in mPCa primary lesions in both high‐risk and low‐risk PCa groups (Figure 1E,F). IF further demonstrated more CD68⁺CD206⁺ macrophage infiltration in mPCa than nmPCa (Figure 1G). Collectively, our findings revealed increased M2 macrophage infiltration in mPCa primary lesions and indicated the underlying prognostic significance.

FIGURE 1.

Primary lesions of mPCa exhibit increased M2 macrophage infiltration and a higher M2‐EVs/Ti‐EVs ratio compared to those of nmPCa. (A) Schematic diagram of the technical workflow to characterize M2 macrophages and M2 EVs in the primary lesions of PCa. (B) UMAP projection of all macrophages from the integrated scRNA‐seq dataset, colored by annotated cell subtypes: M1 macrophages, M2 macrophages, and Mixed macrophages. Cell numbers for each subset were indicated. (C) Analysis of the relative proportion of macrophage groups based on the integrated scRNA‐seq dataset in nmPCa and mPCa. (D) Association between the M2 phenotype and clinical outcome. M2 phenotype was defined by the expression of CD68, CD163, CD206, IL10, ARG1, TGFB1, VEGFA, and CCL22. (E–G) Expression of M2 macrophage‐associated markers (represented by CD68, CD163, and CD206) in primary mPCa and nmPCa sites was shown by IHC (E, F) and IF(G). LR: low risk (patients with PSA value below 10 ng/mL, Gleason score below or equal to 7, and cT1‐cT2a disease); HR: high risk (patients with PSA value above 20 ng/mL, Gleason score above 7, cT2c‐cT4 disease, or a node‐positive disease). Scale bar, 50 µm. (H) Representative TEM images of Ti‐EVs from primary mPCa and nmPCa sites. Scale bar, 200 nm. (I) Size distribution of Ti‐EVs from primary mPCa and nmPCa sites showed by nFCM. (J) Western blotting of EV marker proteins (CD63, ALIX, and CD9) and contaminating protein (GM130). CL: cell lysate of tissue. (K, L) The proportions of CD68⁺ CD206⁺ EVs measured with nFCM in total Ti‐EVs from mPCa (n = 5) compared to those from nmPCa (n = 5). All experiments were repeated three times. Data presented as the mean ± SD. ^*, p < 0.05; ^**, p < 0.01; ^***, p < 0.001; ^****, p < 0.0001; and ns for non‐significant data. Φ: macrophages.

Given the observed enrichment of M2 macrophages in mPCa primary lesions, we sought to elucidate the mechanism underlying their pro‐tumor effects. Since intercellular communication is critical in the tumor microenvironment, we hypothesized that M2 macrophages might mediate these functions, at least in part, through the release of EVs. Compared with EVs derived from urine or blood, which contain a mixture of EVs originating from multiple cellular or tissue sources, Ti‐EVs are considered to better reflect the tissue‐intrinsic characteristics of pathological or physiological events [25]. Analysis of Ti‐EVs may provide deeper insights into the molecular mechanisms underlying disease onset, progression, and therapeutic responses [26, 27]. We separated Ti‐EVs and specifically assessed the proportion of M2 EVs (carrying M2‐associated markers) within the total EV pool from mPCa and nmPCa specimens (Figure 1A, bottom right). Using transmission electron microscope (TEM), we observed a typical cup‐shaped structure of Ti‐EVs separated from the primary lesions of nmPCa and mPCa (Figure 1H). Nano‐flow cytometry (nFCM) revealed a size distribution of 50–200 nm for Ti‐EVs from both groups (Figure 1I). Western blotting confirmed the expression of positive EV markers, such as CD63, ALIX, and CD9, as well as the absence of the contaminating protein GM130 for Ti‐EVs, indicating successful EV separation with high purity (Figure 1J). To resolve the limitations of bulk analysis, which averages signals across heterogeneous EV subpopulations, we performed single‐vesicle characterization using nFCM. This high‐resolution strategy was critical for probing the heterogeneity of the EV pool and specifically quantifying the proportion of M2 EVs in our patient samples, revealing differences that would otherwise be obscured. Immunophenotyping with CD68 and CD206 antibodies by nFCM allowed us to dissect the EV pool into four distinct subpopulations (Figure 1K). Critically, we observed a significantly higher proportion of M2 EVs in mPCa primary foci compared with nmPCa. This enrichment was not only evident in the CD68⁺CD206⁺ EVs (P1) but also in the broader CD206⁺ (P1+P2) and CD68⁺ (P1+P4) fractions, indicating a pervasive increase in macrophage/M2‐related vesicles (Figure 1K,L). This finding directly linked the abundance of M2 macrophages in the tumor microenvironment to a corresponding increase in their EV output, providing a missing link between cellular infiltration and potential distal signaling via vesicles. Collectively, our integrated analysis demonstrated the elevated proportion of M2 EVs in mPCa positions them as a potential key mediator of the pro‐metastatic functions exerted by the M2‐enriched microenvironment.

2.2. M2 EVs Promote the Invasion and Migration of PCa

To transition from our clinical observations to a functional investigation of causality, we established a controlled in vitro model system. Given the pivotal role of M2 macrophages implicated by our human tissue data, we utilized the human monocytic THP‐1 cell line to generate a consistent source of macrophages for mechanistic studies. Specifically, we differentiated THP‐1 cells into macrophages and subsequently polarized a subset toward an M2 phenotype using a cytokine cocktail (IL‐4/IL‐13). This model enabled us to directly probe the functional capacity of M2 macrophages, and their derived M2 EVs, in regulating key aspects of PCa cell aggressiveness, including migration and invasion (Figure 2A). The successful polarization of THP‐1 cells into an M2 phenotype was confirmed through a multi‐modal assessment. Morphologically, induced M2 macrophages adopted the characteristic spindle‐shaped or stellate appearance, distinct from the adherent, rounded morphology of M0 macrophages (Figure 2B). Functionally, the induced M2 macrophages secreted elevated levels of pro‐tumorigenic factors (TGF‐β, CCL22, VEGFA), as quantified by ELISA (Figure 2C). At the molecular level, this morphological and functional shift was concomitant with a significant upregulation of canonical M2 surface markers (CD163, CD206), as validated by flow cytometry (Figure 2D). Furthermore, qRT‐PCR analysis confirmed the transcriptional activation of key M2‐associated genes (MRC1, CD163, VEGFA, TGFB1, and CCL22) (Figure 2E). Critically, IF analysis demonstrated a substantial increase in CD68⁺CD163⁺ cells upon induction, providing direct visual evidence of a definitive M2 macrophage population (Figure 2F). Collectively, these data from morphological, molecular, and functional assays comprehensively verified the successful establishment of a reliable M2 macrophage model in vitro. To assess the direct impact of M2 macrophage‐secreted factors on the metastatic potential of PCa cells, we treated DU145 and PC3 cell lines with cell‐conditioned medium (CCM) collected from M2‐polarized macrophages. Functional analyses, including wound healing and Transwell assays, demonstrated that M2 CCM significantly potentiated the migratory and invasive capacities of both PCa cell lines in vitro (Figure 2G–J). To delineate the specific contribution of EVs within the M2 secretome to this pro‐metastatic effect, we employed two complementary approaches: physical removal of EVs from M2 CCM via ultracentrifugation and pharmacological inhibition of EV secretion using GW4869. The efficacy of EV depletion was confirmed by nano‐flow cytometry, which showed a substantial reduction in EV concentration following both treatments (Figure S2A,B). Strikingly, the pro‐migratory and pro‐invasive effects conferred by M2 CCM were significantly attenuated when PCa cells were treated with EV‐depleted CCM (M2 CCM^EV−) or CCM from GW4869‐treated M2 macrophages (M2 CCM^GW4869), as evidenced by both wound healing and Transwell assays (Figure 3A–C). This result strongly indicates that EVs are crucial functional mediators in the M2 macrophage‐driven enhancement of PCa cell migration and invasion.

FIGURE 2.

CCM of M2 macrophages promotes migration and invasion of PCa cells. (A) Illustration of the strategy used to induce M0 and M2 macrophages in human leukemia monocytic THP‐1 cells. THP‐1 cells were differentiated into M0 macrophages by incubation with 100 ng/mL phorbol‐12‐myristate‐13‐acetate (PMA) for 48 h. M0 macrophages were polarized into M2 macrophages by culturing in 20 ng/mL IL‐4 and IL‐10 for 48 h. (B) Characterization of morphological changes in the course of differentiation from THP‐1 cells to M2 macrophages under a light microscope. Scale bars, 200 µm (100×), 100 µm (200×), 50 µm (400×). (C) ELISA revealed elevated levels of secretory TGF‐β, CCL22, and VEGFA in the CCM of M2 macrophages compared with those of M0 macrophages. (D) Evaluation of M2 macrophage‐associated protein markers by flow cytometry before and after differentiation. (E) Verification of classical M2‐associated genes by qRT‐PCR in M0 and M2 macrophages. Gene expression normalized to GAPDH. (F) The proportional change of CD68⁺CD163⁺ cells upon induction was shown by IF. Scale bar, 100 µm. (G, H) Migration and invasion assays in M2 CCM‐treated versus M0 CCM‐treated DU145 (G) and PC3 (H) cells. (I, J) The wound healing assay showed different migration rates of DU145 (I) and PC3 (J) cells upon M2 CCM treatment. All experiments were repeated three times. Data presented as the mean ± SD. ^*, p < 0.05; ^**, p < 0.01; ^***, p < 0.001; ^****, p < 0.0001; and ns for non‐significant data.

FIGURE 3.

EVs are a crucial contributor to M2 macrophage‐mediated migration and invasion of PCa cells. (A,B) Migration, invasion (A), and wound healing assay (B) of DU145 and PC3 cells treated with M2 CCM, EV‐depleted M2 CCM, or M2 CCM plus GW4869. (C) Quantitative statistics for the above experiment. (D‐E) Migration, invasion (D), and wound healing assay (E) of DU145 and PC3 cells treated with 50 µg/mL M0 or M2 EVs. (F) Quantitative statistics for the above (D‐E) experiment. All experiments were repeated three times. Data presented as the mean ± SD. ^*, p < 0.05; ^**, p < 0.01; ^***, p < 0.001; ^****, p < 0.0001; and ns for non‐significant data.

Then, we conclusively demonstrate that M2 EVs are not just necessary but are sufficient to drive malignant phenotypes. EVs were separated from the CCM of both M0 and M2 macrophages. The separated EVs were rigorously characterized to confirm their identity and integrity. TEM revealed the expected cup‐shaped morphology (Figure S2C), western blotting analysis confirmed the enrichment of canonical EV markers (e.g., CD63, CD9, ALIX) with the absence of GM130 (Figure S2D), and nFCM confirmed a particle size distribution typical of small EVs (Figure S2E). DU145 and PC3 cells were treated with purified EVs derived from M0 or M2 macrophages. Strikingly, exposure to M2 EVs alone was sufficient to recapitulate the pro‐metastatic effects observed with complete M2 CCM, significantly enhancing the migratory and invasive capabilities of the PCa cells compared to those treated with M0 EVs (Figure 3D–F).

Collectively, this complementary experimental evidence solidifies the conclusion that M2 EVs function as potent mediators of PCa cell motility and invasion.

2.3. M2 EVs Are Efficiently Internalized by PCa Cells Inducing MLS and Motility

To investigate how macrophage‐derived EVs interact with PCa cells, M0 and M2 EVs were labeled with the green fluorescent lipophilic dye PKH67 and co‐incubated with PCa cells. Confocal microscopy revealed robust internalization of both M0 and M2 EVs by DU145 and PC3 cells, as evidenced by the presence of green fluorescent dot within the cell cytoplasm (Figure 4A,B). Crucially, Z‐stack analysis confirmed the intracellular localization of these signals, verifying genuine cellular uptake rather than mere surface adsorption (Figure S3A). Notably, the internalization efficiency of M0 and M2 EVs by PCa cells did not differ significantly, suggesting that the phenotypic changes induced by M2 EVs were likely attributable to the intrinsic characteristics defined by their source or content rather than differential uptake by the recipient cells (Figure S3B). Potential mechanisms involved in the uptake of macrophage‐derived EVs by PCa cells were explored using inhibitors targeting different EV internalization pathways. These inhibitors include cytochalasin D (an actin polymerization inhibitor), wortmannin (a phagocytosis inhibitor), methyl‐β‐cyclodextrin (MβCD, an inhibitor of lipid raft‐mediated endocytosis), and omeprazole (an inhibitor of direct membrane fusion). Inhibition with omeprazole showed minimal impact on EV internalization, whereas cytochalasin D, wortmannin, and MβCD significantly reduced EV uptake in both DU145 and PC3 cells as shown by FCM (Figure S3C) [28, 29].

FIGURE 4.

M2 EVs are internalized by PCa cells and mediate MLS transition in recipient cells. (A, B) The internalization of both M0 and M2 EVs by DU145 (A) and PC3 (B) cells was shown by confocal microscopy. Scale bar, 10 µm. (C) GO analysis in PC3 cells based on the difference between treatment with 50 µg/mL M0 EVs and M2 EVs. (D) Regulation of MLS‐related markers in DU145 and PC3 after treatment with EVs based on qRT‐PCR analysis. (E) Characterization of MLS‐related protein markers with western blotting following treatment with 50 µg/mL M0 or M2 EVs. (F, G) Migration and invasion of DU145 and PC3 cells treated with M0 EVs (Con), MLSi, M2 EVs, and M2 EVs + MLSi (F). Quantitative statistics for the above experiment (G). MLSi: an MLS inhibitor Apigenin, 10 µM. All experiments were repeated three times. Data presented as the mean ± SD. ^*, p < 0.05; ^**, p < 0.01; ^***, p < 0.001; ^****, p < 0.0001; and ns for non‐significant data.

RNA sequencing was performed to explore the downstream transcriptional programs underlying the enhancement in migratory and invasive abilities of PCa cells treated with M2 EVs compared with M0 EVs. Gene ontology (GO) analysis revealed significant enrichment of genes related to Biological Processes (BP) such as “cell adhesion”, Cellular Components (CC) including “cell junctions and the cell surface”, as well as Molecular Function (MF) associated with “integrin binding” (Figure 4C). Complementarily, Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis showed that differentially regulated genes were prominently enriched in pathways including “focal adhesion and cell adhesion molecules” (Figure S3D). Based on the above findings that M2 EVs significantly enhance the migratory and invasive capabilities of PCa cells, and considering the transcriptomic analyses implicating their mechanism of action in cell adhesion and cytoskeletal reorganization, we hypothesized that M2 EVs might drive metastatic progression by inducing a phenotypic switch in tumor cells. The acquisition of migratory and invasive properties is often accompanied by a cellular phenotype known as the MLS. MLS is a key phenotype enabling tumor cell migration, invasion, and drug resistance. It can be driven by the epithelial‐mesenchymal transition (EMT) process, during which tumor cells gradually lose epithelial characteristics and acquire features of mesenchymal cells [12, 13, 30, 31]. Further analysis of MLS‐related gene expression using qRT‐PCR and western blotting demonstrated that treatment with M2 EVs significantly upregulated the expression of MLS‐related genes CDH2 (N‐Cadherin), SLUG, and ZEB1 in both DU145 and PC3 cells, while the expression of the canonical epithelial marker CDH1 (E‐Cadherin) and a classic mesenchymal marker VIM showed no significant change (Figure 4D,E). This gene expression pattern, characterized by the acquisition of mesenchymal traits without full loss of epithelial characteristics, was consistent with a hybrid or partial EMT state, which has been associated with enhanced metastatic potential in cancer cells by conferring high plasticity and motility. To definitively validate the necessity of MLS in EV‐mediated PCa metastasis, we utilized apigenin, a flavonoid known to suppress MLS‐associated phenotypes (MLS inhibitor, MLSi) [32]. Under this MLS‐suppressed condition, the potent pro‐metastatic effect of M2 EVs was markedly attenuated. This result demonstrates that a functional MLS program in recipient cells is a requisite downstream event for the metastasis driven by M2 EVs. Importantly, treatment with MLSi alone did not significantly alter the basal migratory or invasive capacity of the cells compared to the control, indicating that the inhibitor itself at this concentration is not cytotoxic and its effect is specific to the pathway activated by M2 EVs (Figure 4F,G). This critical finding demonstrates that the MLS program is influential for M2 EVs‐mediated promotion of PCa metastasis.

In summary, our data establish a coherent mechanistic pathway: M2 macrophages release EVs that are efficiently internalized by PCa cells, subsequently activating gene expression programs related to focal adhesion and cell adhesion. This rewiring of transcriptional networks promotes MLS, which is functionally essential for the enhanced migratory and invasive capabilities driven by M2 EVs, ultimately facilitating PCa metastasis.

2.4. TXNDC5 Upregulation Induced by M2 EVs Promotes PCa Migration and Invasion

Building upon the finding that M2 EVs enhance the migratory and invasive capacities of PCa cells, we sought to identify the key molecular effectors responsible for these phenotypic changes. Transcriptomic profiling of DU145 and PC3 cells treated with M2 EVs versus M0 EVs revealed substantial gene expression alterations. In DU145 cells, 115 genes were upregulated, and 91 were downregulated, while PC3 cells exhibited more extensive remodeling with 739 upregulated and 587 downregulated genes (Figure S4A). Cross‐analysis across different cellular contexts identified five genes consistently upregulated in both cell lines: TXNDC5, CC2D2B, HMCN1, and two unprocessed pseudogenes (Figure 5A). Following a standard analytical pipeline focusing on protein‐coding potential, we prioritized TXNDC5, CC2D2B, and HMCN1 for subsequent investigation.

FIGURE 5.

TXNDC5 is the key mediator of M2 EVs‐induced PCa cell migration and invasion. (A) Venn diagram illustrating the cross‐analysis of upregulated genes in DU145 and PC3 cells following M2 EVs treatment based on RNA sequencing. The analysis integrated two independent DU145 datasets (M2 EV vs. M0 EV; 6 vs. 3 and 3 vs. 3) and one PC3 dataset (M2 EV vs. M0 EV; 4 vs. 3). Significantly upregulated genes (Foldchange > 1.5, p < 0.05) from the two DU145 datasets were combined (union) and subsequently intersected with upregulated genes from the PC3 dataset to define the final core set of conserved targets. (B) Pan‐cancer analysis of TXNDC5 mRNA expression across multiple cancer types from the TCGA database. (C) TXNDC5 expression in PCa tissues and normal prostate tissues based on public transcriptome data using the Xiantao Academic cloud platform. (D) Kaplan‐Meier survival analysis showing the association between TXNDC5 expression and PCa patient prognosis. (E, F) Representative IHC images (E) and quantitative analysis (F) of TXNDC5 protein expression in primary lesions of mPCa and nmPCa. Scale bar, 50 µm and 25 µm (for magnification). (G) Tumor Immune Estimation Resource (TIMER) analysis of the correlation between TXNDC5 expression and the infiltration of all macrophages or M2 macrophages. (H‐N) Changes of migration, invasion (H, I) and wound healing abilities (J, K) of DU145 and PC3 cells with TXNDC5 knockdown. Quantitative statistics for the above experiment (L‐N). All experiments were repeated three times. Data presented as the mean ± SD. ^*, p < 0.05; ^**, p < 0.01; ^***, p < 0.001; ^****, p < 0.0001; and ns for non‐significant data.

Interrogation of public databases revealed TXNDC5 as the most promising candidate. It demonstrated significant overexpression in multiple cancer types (Figure 5B), including PCa where it showed elevated expression in tumor tissues compared to normal controls (Figure 5C). Critically, high TXNDC5 expression correlated with unfavorable patient prognosis in PCa (Figure 5D). In contrast, CC2D2B and HMCN1 displayed no significant differential expression between tumor and normal tissues, nor any association with clinical outcomes (Figure S4B–E). Furthermore, IHC analysis of our patient cohort confirmed significantly enhanced TXNDC5 expression in primary lesions of mPCa compared to nmPCa counterparts (Figure 5E,F), reinforcing its clinical relevance with disease progression. To functionally connect M2 macrophages with MLS promotion, we analyzed our scRNA‐seq data. We found that M2 macrophages exhibited an enrichment for genes driving cellular plasticity, a hallmark of the MLS (derived from EMT‐associated gene set enrichment) (Figure S4F). Simultaneously, TXNDC5 expression was intrinsically linked to the M2 phenotype, as evidenced by a significant co‐expression pattern with MRC1 (Figure S4G). When compared to their M1 counterparts, M2 macrophages also expressed higher levels of TXNDC5 based on integrated scRNA‐seq data (Figure S4H), suggesting a potential synergy within M2 macrophages to facilitate metastatic potential. These findings establish an intrinsic link between TXNDC5 and the M2 phenotype. Furthermore, we performed an independent analysis using Tumor Immune Estimation Resource (TIMER), confirming significant positive correlations between TXNDC5 expression and infiltration levels of both total macrophages and specifically M2 macrophages in PCa samples (Figure 5G).

TXNDC5 encodes a protein of the disulfide isomerase family. The protein is located in the endoplasmic reticulum (ER) and contributes to the regulation of oxidative stress and protein folding [33, 34, 35]. Increased expression of TXNDC5 in the tumor microenvironment promotes tumor progression by fostering a microenvironment conducive to migration, invasion, and metastasis [35]. However, its role in PCa metastasis has not been fully elucidated [36]. To further investigate the function of TXNDC5, we designed two short hairpin RNA (shRNAs) for TXNDC5 knockdown in DU145 and PC3 cells. Both shTXNDC5#1 and shTXNDC5#2 significantly reduced TXNDC5 expression compared to that in the control group at both the transcriptional and protein levels (Figure S4I,J). We observed that TXNDC5 knockdown significantly attenuated PCa cell migration and invasion (Figure 5H–N). Collectively, these functional data demonstrated that TXNDC5 was necessary for the migration and invasion of PCa cells, positioning it as an important driver of the metastatic process.

2.5. M2 EVs Horizontally Transfer Functional TXNDC5 mRNA to PCa Cells

Considering that EVs serve as crucial mediators of molecular cargo transport, elevated TXNDC5 expression in PCa cells might result from EV‐mediated horizontal transfer. As mentioned above, we observed a consistent elevation of TXNDC5 mRNA levels in the two PCa cell lines following M2 EVs treatment. Then we validated the result with qRT‐PCR (Figure S5A). We systematically characterized the TXNDC5 content in the donor‐recipient system. We found that M2 macrophages expressed higher levels of TXNDC5 mRNA than M0 macrophages did (Figure S5B). Crucially, this differential expression was reflected in their respective EV cargo. Analysis using qRT‐PCR confirmed higher enrichment of TXNDC5 transcripts in M2 EVs compared to M0 EVs (Figure S5C). To address whether TXNDC5 enrichment in M2 EVs represents a ubiquitous phenomenon beyond the model, we extended our validation to both human peripheral blood mononuclear cell (PBMC)‐derived macrophages and murine RAW264.7 macrophages. Successful M2 polarization in these systems was confirmed by elevated expression of the characteristic markers (Figure S5D). We subsequently isolated EVs from conditioned media of M0 and M2 macrophages through ultracentrifugation, with successful isolation validated by standard characterization methods including nFCM, TEM, and western blotting for EV markers (Figure S5E,F). Crucially, we found consistently higher levels of TXNDC5 mRNA in M2 EVs compared to M0 EVs across both human PBMC‐derived and murine macrophage models, confirming the broader relevance of this finding (Figure S5G). This multi‐system validation substantially strengthens our conclusion that the horizontal transfer of TXNDC5 mRNA represents a conserved mechanism in M2 EVs‐mediated intercellular communication. After confirming the enrichment of TXNDC5 mRNA in M2 EVs, we sought to determine whether these EVs could correspondingly upregulate TXNDC5 protein expression in recipient PCa cells. To this end, we treated DU145 and PC3 cells with EVs of varying types and concentrations. Western blotting analysis demonstrated that, compared to M0 EVs, M2 EVs significantly upregulated TXNDC5 protein levels in recipient cells, and this effect was observed in a dose‐dependent manner (Figure S5H). Importantly, western blotting analysis demonstrated no corresponding package of TXNDC5 protein in both M0 and M2 EVs (Figure S5I), effectively ruling out substantial protein transfer as the primary mechanism. These findings prompted us to investigate whether TXNDC5 mRNA undergoes horizontal transfer from macrophages to PCa cells via EVs. To definitively distinguish EV‐mediated mRNA transfer from potential transcriptional upregulation in recipient cells, we employed a transcriptional inhibition approach. Pretreatment of DU145 and PC3 cells with actinomycin D effectively suppressed endogenous RNA synthesis, maintaining consistently low basal TXNDC5 mRNA levels throughout the 6–24 h period (Figure 6A). Under transcriptional inhibition, treatment with M2 EVs resulted in a significant recovery of TXNDC5 mRNA in recipient DU145 and PC3 cells (Figure 6B). This specific restoration under complete transcriptional blockade provided definitive evidence for the horizontal transfer of TXNDC5 mRNA via M2 EVs.

FIGURE 6.

M2 macrophages promote PCa metastasis via horizontal TXNDC5 mRNA transfer. (A) Demonstration of efficient inhibition of de novo TXNDC5 mRNA synthesis in DU145 and PC3 cells with 1 µg/mL actinomycin D at 6 and 24 h. (B) TXNDC5 mRNA levels in PCa cells under transcriptional inhibition. TXNDC5 mRNA levels were measured at 0 h (Con, 0 h) as a baseline, and at 6 h post‐inhibition in cells receiving no EVs (Con, 6 h), M0 EVs (50 µg/mL), or M2 EVs (50 µg/mL). (C) Schematic depiction of the generation of PKH67‐labeled M2 EVs containing Cy5‐tagged TXNDC5‐Flag mRNA. (D) Single‐particle analysis by nFCM validating the successful generation of positive (PKH67⁺/Cy5⁺) EVs groups. (E, F) Validation of the protective role of EV lipid bilayer via RNase and Triton treatment (E). Quantitative statistics of P1 (PKH67⁺/Cy5⁺), P2 (PKH67−/Cy5⁺ group), P4 (PKH67⁺/Cy5− group) subgroups for the above experiment (F). (G) Confocal microscopy images showing internalization of dual‐labeled M2 EVs (PKH67, green; Cy5‐TXNDC5 mRNA, red) by DU145 and PC3 cells. Nuclei were stained with DAPI (blue). Scale bar, 20 µm. (H, I) Quantitative comparison of mRNA loading efficiency between M0 and M2 EVs, shown by (H) the percentage of total mRNA‐carrying EVs (P1+P2) and (I) the ratio of mRNA‐carrying to total macrophage‐derived EVs (P1/[P1+P4]). (J) Western blotting analysis of Flag‐tagged protein expression in PCa cells after treatment with M0 or M2 EVs loaded with identical amounts of TXNDC5‐Flag mRNA. All experiments were repeated three times. Data presented as the mean ± SD. ^*, p < 0.05; ^**, p < 0.01; ^***, p < 0.001; ^****, p < 0.0001; and ns for non‐significant data.

To directly mimic packaging and transfer of TXNDC5 mRNA via M2 EVs into recipient cells with visual and quantitative evidence, we established a dual‐fluorescence mRNA tracking system in vitro (Figure 6C). PKH67‐labeled THP‐1 cells were differentiated into M2 macrophages (Figure S6A, top panel), and Cy5‐tagged TXNDC5‐Flag mRNA was transfected into the PKH67‐labeled macrophages (Figure S6A, bottom panel). CCM was subsequently collected for EV separation. Using nFCM for high‐resolution single‐particle analysis, representative real‐time nFCM plots illustrated the simultaneous detection of three parameters for each individual particle of M0 EV (Figure S6B) and M2 EVs (Figure 6D): side scatter (SS, indicating particle size and internal complexity), PKH67 green fluorescence (confirming the particle's origin from the labeled macrophage membrane), and Cy5 red fluorescence (reporting the presence of the transfected TXNDC5 mRNA). Then we resolved the EV population into four distinct subsets based on their fluorescence profiles. These included PKH67−/Cy5− group (background, P3), PKH67⁺/Cy5− group (EVs without the target mRNA, P4), PKH67−/Cy5⁺ group (mostly EVs with inefficient membrane labeling containing target mRNA, P2), and the definitive PKH67⁺/Cy5⁺ population (EVs of macrophage origin with the Cy5‐tagged TXNDC5 mRNA package, P1). The clear identification of this PKH67⁺/Cy5⁺ EV subpopulation provides direct evidence that EVs of macrophage origin could package TXNDC5 mRNA during their biogenesis. To definitively confirm that the Cy5 signal originated from TXNDC5 mRNA enclosed within the EV lumen rather than adhering to the surface, we performed a critical enzymatic protection assay. Treatment with RNase A alone did not significantly alter the proportion of fluorescent signals (P1, P2, and P4). In stark contrast, concurrent treatment with RNase A and the membrane‐disrupting Triton X‐100 drastically reduced the positive population (Figure 6E,F; Figure S6C,D). This result conclusively demonstrates that the TXNDC5 mRNA is protected within the EV lumen. To visually track the delivery process via M2 EVs, we co‐incubated PKH67‐labeled M2 EVs containing Cy5‐tagged TXNDC5 mRNA with DU145 and PC3 cells. Confocal microscopy revealed the successful internalization of these double‐positive EVs into the recipient PCa cells (Figure 6G). Having established the intravesicular localization of TXNDC5 mRNA, we quantitatively compared the mRNA loading and transfer efficiency between M0 and M2 EVs. Quantitative analysis of the nFCM subpopulations revealed that M2 EVs possessed a significantly higher TXNDC5 mRNA loading ratio. This was evidenced by both a greater proportion of total mRNA‐carrying EVs (P1 + P2) and a higher ratio of mRNA‐carrying EVs to all EVs of macrophage origin (P1/[P1+P4]), when compared to M0 EVs (Figure 6H,I). This finding indicated that M2 polarization not only enhances TXNDC5 expression in macrophages but also specifically promotes its selective packaging into EVs.

To further confirm that EV‐encapsulated TXNDC5 mRNA was translated following internalization by recipient cells, we treated DU145 and PC3 cells with EVs containing TXNDC5‐Flag mRNA and validated Flag expression by western blotting. To definitively decouple the effects of EV cargo loading from intrinsic functional delivery efficiency, we performed a critical controlled experiment. We isolated M0 and M2 EVs and loaded them with an identical amount of in vitro‐transcribed TXNDC5‐Flag mRNA via electroporation. Quantitative assessment using nFCM confirmed that the two EV groups exhibited comparable Cy5 mean fluorescence intensity (MFI) (Figure S6E). Subsequent co‐culture of these mRNA‐loaded EVs with DU145 and PC3 cells revealed M2 EVs mediated significantly higher expression of the Flag‐tagged protein in recipient cells compared to M0 EVs loaded with the same mRNA volume (Figure 6J). This result provides direct and compelling evidence that M2 EVs possess a superior intrinsic capacity to functionally deliver their mRNA cargo, a property that is independent of and synergistic with their enhanced TXNDC5 mRNA packaging.

In summary, M2 EVs can horizontally transfer TXNDC5 mRNA to PCa cells, where it is successfully translated and potentially exerts functional effects.

2.6. TXNDC5 mRNA Transferred by M2 EVs Affects PCa MLS and Metastasis

To definitively establish the functional significance of TXNDC5 mRNA within M2 EVs, we performed loss‐of‐function studies in the donor macrophages. We knocked down TXNDC5 in M2‐polarized THP‐1 cells, with knockdown efficiency confirmed at the protein level by western blotting (Figure S7A), and isolated EVs (M2 EVs^shTXNDC5) from their conditioned medium. The knockdown was specific, as it did not alter the expression of characteristic M2‐related markers (Figure S7B). Knockdown of TXNDC5 mRNA in M2 EVs was validated using qRT‐PCR (Figure S7C). Functionally, M2 EVs^shTXNDC5 exhibited a significantly attenuated ability to promote the migration and invasion of DU145 and PC3 cells compared to control M2 EVs (Figure 7A–C). At the molecular level, this impaired pro‐metastatic capacity was associated with a marked downregulation of MLS drivers N‐Cadherin, SLUG, and ZEB1 in PCa cells treated with M2 EVs^shTXNDC5 (Figure 7D). To further investigate the role of M2 EVs‐derived TXNDC5 in PCa in vivo, we pretreated DU145‐Luc cells with different EVs (M0 EVs, M2 EVs, and M2 EVs^shTXNDC5) and injected the cells into the left ventricle of nude mice. EVs were administered every three days via the tail vein until day nine to mimic sustained systemic effects of circulating EVs under physiological conditions (Figure 7E). Then bioluminescence imaging was performed to assess the systemic metastasis of PCa cells and the metastasis burden of main organs in the three groups. The in vivo metastasis assay yielded results fully consistent with our in vitro findings. Bioluminescence imaging demonstrated that systemic administration of M2 EVs dramatically enhanced the metastatic burden in mice compared to the M0 EV control group. This was evidenced by quantified total photon flux throughout the mouse body. This pro‐metastatic effect was critically dependent on TXNDC5, as mice treated with M2 EVs^shTXNDC5 showed a significant reduction in systemic metastasis (Figure 7F,G). The systemic pro‐metastatic effect of M2 EVs quantified by whole‐body bioluminescence imaging was consistently reflected in ex vivo bioluminescence imaging of major organs, with particularly prominent metastatic loads observed in the lung and the liver. To rigorously evaluate the associated phenotypic and molecular changes, we focused our analysis on lung tissues, which consistently developed macroscopic metastases across all experimental groups, thereby enabling a standardized comparison (Figure S7D). Histological examination of lung sections revealed that the M2 EVs group developed significantly more numerous and larger metastatic nodules compared to both the control and M2 EVs^shTXNDC5 groups, demonstrating the potent pro‐metastatic effect of TXNDC5 in M2 EVs (Figure 7H). Additionally, molecular analysis of these lung metastases further demonstrated that the elevated expression of MLS‐associated proteins (N‐Cadherin, SLUG, and ZEB1) in the M2 EVs group was markedly attenuated in M2 EV^shTXNDC5 group (Figure 7I,J).

FIGURE 7.

TXNDC5 in M2 EVs drives MLS and metastatic progression in vitro and in vivo. (A–C) Migration, invasion (A), and wound healing assay (B) of DU145 and PC3 cells treated with 50 µg/mL M2 EVs and M2 EVs^shTXNDC5. Quantitative statistics for the above experiment (C). (D) Western blotting of MLS‐related markers of PCa cells treated with 50 µg/mL M2 EVs or M2 EVs^shTXNDC5. (E) Schematic diagram of the in vivo experimental design: DU145‐Luc cells pretreated with EVs were intracardially injected, followed by repeated EV administration via the tail vein. EV‐treated PCa cells are injected into the left ventricle of nude mice (day 0). 5 mg/kg EVs of three types were routinely injected into the tail vein of the mice at day 0, day 3, and day 6. The animals were subjected to downstream analyses at day 9 (n = 6). (F,G) In vivo metastasis assessment by bioluminescence imaging. Representative whole‐body bioluminescence images (F). and quantification of total flux (G) showed that M2 EVs enhanced metastatic burden, which was attenuated by M2 EVs^shTXNDC5. (H) Representative H&E‐stained lung sections harvested from mice treated with control, M2 EVs or M2 EVs^shTXNDC5. Scale bar, 2 mm. (I, J) IHC analysis of MLS markers in lung metastases. Representative images (I) and quantitative scoring (J) showing expression of N‐Cadherin, SLUG, and ZEB1 in the control (M0 EVs), M2 EVs, and M2 EVs^shTXNDC5 group. Data presented as the mean ± SD. ^*, p < 0.05; ^**, p < 0.01; ^***, p < 0.001; ^****, p < 0.0001; and ns for non‐significant data.

Collectively, these in vivo findings established that TXNDC5 mRNA encapsulated in M2 EVs serves as a pivotal molecular driver of systemic metastasis, with its functional role being mechanistically linked with the induction and maintenance of a metastasis‐prone MLS phenotype in disseminated tumor cells.

In conclusion, TXNDC5 mRNA‐containing M2 EVs potentially influence PCa metastatic ability by affecting MLS both in vitro and in vivo.

3. Discussion

Metastatic PCa remains incurable, and understanding the key steps involved in PCa metastasis is critical for improving patient survival. The process of PCa metastasis is complex and involves a series of events that depend on both the intrinsic characteristics of tumor cells and their interactions with various other cell types within the TME. Our study demonstrates that EV‐mediated mRNA transfer participates in interactions between tumor cells and other components of the TME to promote PCa metastasis from a unique perspective.

TAMs are important components of the PCa tumor microenvironment, and their role in the initial stages of tumor metastasis remains poorly understood. Previous studies have demonstrated higher infiltration of M2 TAMs in PCa metastases than in primary lesions [37]. However, it may not be feasible to analyze the primary site‐intrinsic factors driving tumor metastasis through a metastasis‐primary lesion comparison because the TME of the metastasis site may have been complicated by the characteristics of the target organ. In our study, we investigated the infiltration of M2 macrophages in the primary tumors of patients with nmPCa and mPCa. We found that the mPCa primary lesions exhibited greater infiltration of M2 macrophages. To further explore whether M2 macrophages potentially influence PCa metastasis, we focused on the role of EVs, which have gained considerable attention in intercellular communications. Previous studies have concentrated on the effects of tumor cell‐derived EVs on the polarization of M2 macrophages; however, the potential role of M2 EVs in PCa has not been sufficiently studied [38, 39, 40, 41, 42, 43, 44]. Ti‐EVs are more tissue‐specific than EVs from other sources and better reflect the physiological and pathological conditions of the source organ. We were the first to separate and analyze clinical sample‐derived PCa Ti‐EVs and demonstrate that mPCa contained a higher proportion of M2 macrophage‐derived EVs compared to nmPCa [26, 45, 46, 47]. Our findings provide strong evidence for the role of M2 EVs in promoting PCa progression. Furthermore, previous research has focused on non‐coding RNAs (e.g., microRNAs and circRNAs) or proteins in EVs, and there has been limited attention on the role of EV‐encapsulated mRNA. We discovered that M2 EVs could horizontally transfer TXNDC5 mRNA to promote the MLS phenotype in PCa. Studies have shown that TXNDC5 can promote the proliferation of castration‐resistant PCa, but its role in tumor metastasis has not been systematically explored [36]. In our study, we found that TXNDC5 was highly expressed in multiple cancers. We further demonstrated that mPCa primary tumors have greater TXNDC5 expression than nmPCa, and showed that TXNDC5 mRNA‐containing M2 EVs promote PCa metastasis. Beyond broadly inhibiting EV biogenesis or uptake, future therapeutic strategies could aim for precision inhibition by targeting specific EV subpopulations from specific cellular sources. Among these, EVs derived from M2 macrophages represent a highly promising target. Developing strategies to specifically neutralize M2 EVs—for instance, by exploiting their unique surface markers—holds significant research value and clinical potential. Such a strategy potentially offers improved efficacy and reduced off‐target side effects compared to the inhibition approaches targeting broader EV populations. [48, 49, 50, 51]

Despite the important insights gained from our study, there are several limitations that should be acknowledged. M2 EVs are a complex biological system containing various biomolecules, and the potential pro‐tumor roles of other molecules, including microRNAs and proteins, which may collectively influence the phenotype of target cells, have not been fully explored. Additionally, the mechanisms by which mRNA is packaged into EVs remain unclear and warrant further investigation. Furthermore, it is challenging to evaluate the precise efficiency of EVs in mediating metastasis under pathophysiological conditions. Intrinsic efficiency of EVs‐mediated delivery remains to be addressed, which is influenced by dynamic variables such as the absolute EVs concentration in vivo, the duration of EVs‐cell contact, and the recipient cells’ translational capacity. Finally, the mechanism by which the TXNDC5 protein drives MLS and ultimately metastatic progression warrants deeper elucidation. Based on previously reported functions of TXNDC5, we propose the following model: Multiple stressors in the TME perturb the protein‐folding capacity of the endoplasmic reticulum in malignant and stromal cells. Moderate yet persistent ER stress responses restore ER homeostasis, promote cell adaptation to stress, and thus bestow the malignant cells with the ability to survive under harsh conditions such as those experienced in early metastatic disseminations. By contrast, extreme ER stress caused by the uncontrolled accumulation of misfolded proteins in this organelle can lead to a terminal unfolded protein response (UPR) that induces cell death [52, 53, 54, 55]. TXNDC5 belongs to the protein disulfide isomerase family. It is located in the ER and is known to facilitate protein folding and thiol‐disulfide interchange reactions, mitigating ER stress [56, 57]. Thus, we hypothesize that high TXNDC5 levels, replenished by its mRNA, could enhance the capacity for protein folding, thereby promoting cancer cell survival under extremely adverse conditions, enabling the accumulation of survival advantages that would eventually facilitate metastatic dissemination. Future work will focus on the exact downstream signaling axis responsible for the pro‐metastatic function of TXNDC5.

In conclusion, our study is the first to validate the horizontal transfer of TXNDC5 mRNA in M2 EVs using single‐vesicle analysis. Our research provides new insights into intercellular communication within the tumor microenvironment and highlights potential avenues for the inhibition of EV‐mediated tumor metastasis.

4. Methods

4.1. Study Approval

All animal experiments and collection of human specimens with relevant experiments were approved by the Ethics Committee of Ren Ji Hospital, Shanghai Jiao Tong University School of Medicine (KY2023‐112‐C). Human samples and clinical information were collected from patients who provided written, informed consent.

4.2. Human Specimens

Tissue samples from the primary lesions of PCa were collected from patients undergoing transrectal ultrasound‐guided prostate biopsy or radical prostatectomy at the Department of Urology, Ren Ji Hospital. The cancerous nature of the collected tissue samples was pathologically confirmed.

4.3. Cell Culture

Human leukemia monocytic THP‐1 cells (TIB‐202, ATCC, RRID: CVCL_0006) and human PCa cell lines PC3 (CRL‐1435, ATCC, RRID: CVCL_0035) were cultured in RPMI‐1640 medium (72400047, Gibco) containing 10% fetal bovine serum (FBS) (BS1612‐109, Bioexplorer) and 1% penicillin‐streptomycin solution (100X, BL505A, Biosharp). For THP‐1, medium was additionally supplied with β‐mercaptoethanol (0.05 mm, M8211, Solarbio). Human PCa cell lines DU145 (HTB‐81, ATCC, RRID: CVCL_0105) cells and murine RAW264.7 (TIB‐71, ATCC, RRID: CVCL_0493) were maintained in Dulbecco's modified eagle medium (11965092, Gibco) supplemented with 10% FBS and 1% penicillin–streptomycin. All cell lines have been tested for mycoplasma after their purchase in 2021 and STR analysis was performed every month. All cells were maintained at 37°C with 5% CO₂ in a humidified incubator and were routinely tested for mycoplasma contamination. THP‐1 cells were differentiated into M0 macrophages by incubation with PMA (100 ng/mL, 19–144, Sigma‐Aldrich) for 48 h. M0 macrophages were polarized into M2 macrophages by culturing in IL‐4 (20 ng/mL, #200‐04‐1MG, PeproTech) and IL‐13 (20 ng/mL, #200‐13‐50UG, PeproTech) for 48 h. RAW264.7 cells were seeded in appropriate culture plates and allowed to adhere overnight. For M2 polarization, cells were stimulated with IL‐4 (20 ng/mL, #214‐14‐20UG, PeproTech) and IL‐13 (20 ng/mL, #210‐13‐10UG, PeproTech) for 48 h. Human peripheral blood mononuclear cells (hPBMCs) were isolated from healthy donor peripheral blood via density gradient centrifugation. Isolated hPBMCs were cultured in complete medium containing 10% FBS and antibiotics. To induce differentiation into macrophages, hPBMCs were cultured with M‐CSF (25 ng/mL, #300‐25‐10UG, PeproTech) for approximately 7 days until stable differentiation was achieved. Then IL‐4 and IL‐13 were used as primary M2 polarization inducing agents like THP‐1. Polarized cells were changed to culture medium supplemented with 10% EV‐depleted FBS (EXO‐FBS‐50A‐1, System Biosciences) before EV separation.

4.4. Separation of EVs From Primary PCa Tissue and Cell Culture Medium

Pretreatment of human PCa tissues was performed as previously described [58]. Briefly, PCa tissue (1–2 g per patient) was dissected into 2 × 2 × 2 mm blocks with a sterile scalpel immediately after prostatectomy and subsequently dissociated with collagenase D (2 mg/mL, 11088882001, Roche) and DNase I (40 U/mL, 11284932001, Roche) at 37°C for 30 min. The mixture was then filtered through a 70 µm cell strainer (352350‐50, Falcon). Cell debris, organelles, and large EVs were removed by sequential centrifugation at 1000 g for 10 min and 10 000 g for 20 min. Tissue‐derived EVs were isolated from the supernatants by size exclusion chromatography (SEC), according to the manufacturer's instructions (Izon). EVs were separated from cell conditioned medium (CCM) via differential centrifugation, as described in our previous study [59]. Briefly, CCM was centrifuged at 1000 g for 10 min and 10 000 g for 20 min. The supernatant was then ultracentrifuged at 12 00 00 g for 2 h at 4°C. After the first round of ultracentrifugation, the pellet was resuspended in PBS (P1020, Solarbio) and spun in the same tube for the second round. The supernatant was discarded and the pellet was dissolved in PBS for storage or downstream analysis.

4.5. TEM Analysis

For EV visualization, 10 µL of fresh EV suspension was loaded onto a 200‐mesh carbon‐coated copper grid (BZ11032b, Zhongjingkeyi Technology) for 1 min. Excess liquid was then removed from the edge of the copper mesh by aspiration with filter paper before negative staining with one drop of 1% uranyl acetate (U25690‐1g, Shanghai Acmec Biochemical Technology) for 10 s. Excess fluid was removed, and the grid was washed twice with distilled water and dried in the dark. Images were obtained using TEM (HT7700, HITACHI).

4.6. Flow Cytometry (FCM)

Briefly, the cells were harvested and washed twice with PBS to remove residual culture medium. The cell concentration was adjusted to 1 × 10⁶ cells/mL. To block non‐specific binding, the cells were incubated with Fc receptor blocking reagent for 10 min at 4°C. The cells were then stained with PE anti‐human CD163 (333605‐25T, Biolegend) and APC anti‐human CD206 (321109‐25T, Biolegend) at the recommended dilution rate in staining buffer (100 µL). The staining was performed in the dark for 30 min at 4°C. After staining, the cells were washed twice with PBS containing 2% FBS and resuspended in staining buffer (200 µL). Flow cytometry was performed using a flow cytometer (CytoFLEX, Beckman), and the data were analyzed using FlowJo software (BD Biosciences).

4.7. EV Uptake Experiments

The green fluorescent dye PKH67 Fluorescent Cell Linker Kit (MINI67, Sigma‐Aldrich) was used for EV labeling. EVs were resuspended in Diluent C (0.5 mL), and PKH67 (4 µL) was diluted in Diluent C (0.5 mL). The two solutions were thoroughly mixed and incubated in the dark at room temperature for 4 min. To stop excessive staining, 1% BSA (5 mL) was added, and the labeled EVs were re‐separated by ultracentrifugation at 4°C to remove residual dye. The EV pellet was then resuspended in PBS (100 µL) and stored in the dark at 4°C for later use. DU145 and PC3 cells were cultured in a confocal dish (No81156, Ibidi), and the PKH67‐labeled EVs were added to the recipient cells. The cells were incubated at 37°C in the dark for 12 h. After incubation, the cell supernatant was removed and the cells were washed once with PBS. Next, 4% paraformaldehyde (1 mL, bry‐0026, Runnerbio) was added to fix the cells in the dark for 15 min. The fixative was removed and the cells were washed once with PBS with gentle shaking to ensure thorough coverage. The cells were permeabilized with 0.5% Triton X‐100 for 5 min at 25°C. After PBS wash, TRITC‐labeled phalloidin (200 µL, 40734ES80, Yeasen) was incubated with the cells in the dark at 25°C for 30 min. After washing with PBS, DAPI (500 µL, C0065, Solarbio) was added and incubated in the dark for 5 min to stain the nuclei. After removing excess DAPI, the cells were washed three times with PBS. Finally, PBS (1 mL) was added to the confocal dish for storage in the dark, and the cells were imaged using a laser confocal microscope (CSU‐W1 SoRa, Nikon). To investigate the pathways of EV uptake, DU145 and PC3 cells were pretreated with a panel of pharmacological inhibitors targeting distinct internalization mechanisms. Following pretreatment, cells were incubated with PKH67‐labeled EVs. After incubation, cells were thoroughly washed to remove non‐internalized EVs, and the extent of EV uptake was quantified by measuring the mean fluorescence intensity (MFI) using FCM.

4.8. Plasma Membrane Labeling With PKH67

THP‐1 cells were labeled with the green fluorescent dye PKH67 fluorescent cell linker kit following the manufacturer's instructions. A suspension of 2 × 10⁷ single cells was washed once in serum‐free medium and centrifuged at 400 × g for 5 min. The supernatant was carefully aspirated, leaving ≤ 25 µL of residual medium. To prepare a 2 × Cell Suspension, the pellet was resuspended in Diluent C (1 mL) by gentle pipetting. A 2 × Dye Solution (4 × 10⁻⁶ m) was prepared by mixing PKH67 dye (4 µL) with Diluent C (1 mL). A 2 × Cell Suspension (1 mL) was added to 2 × Dye Solution (1 mL), mixed immediately, and incubated for 1–5 min with periodic mixing to achieve final concentrations of 1 × 10⁷ cells/mL and 2 × 10⁻⁶ m PKH67. Staining was stopped with serum or 1% BSA (2 mL, B2064, Sigma) for 1 min. Cells were centrifuged at 400 × g for 10 min at 20°C–25°C, resuspended in complete medium (10 mL), and washed twice to remove unbound dye. The final pellet was resuspended in complete medium (10 mL) and tested for cell viability and labeling efficiency. PKH67‐labeled THP‐1 cells were then induced to differentiate into M2 macrophages, as described above.

4.9. RNA Isolation and Quantitative Reverse Transcription PCR (qRT‐PCR)

Total RNA was extracted from the cells and EVs using TRIzol reagent (99940001, Invitrogen). RNA concentrations were quantified using a NANODROP 2000c spectrophotometer (Thermo Fisher Scientific). RNA was reverse‐transcribed using the Reverse Transcription Kit (R323‐01, Vazyme). qRT‐PCR was performed using the SYBR green method (Q711‐02, Vazyme) on a LightCycler 480 qPCR machine (Roche). The primers used in this study were synthesized by Sangon Biotech. The GAPDH gene was chosen as reference gene for all the qRT‐PCR experiments.

4.10. Western Blotting Assay

Cells and EVs were lysed using RIPA lysis buffer (P0013B, Beyotime Biotechnology) supplemented with a protease inhibitor cocktail (P1005, Beyotime Biotechnology). The protein concentration was determined using the BCA Protein Assay kit (P0009, Beyotime Biotechnology). Proteins were incubated with 5 × loading buffer (P0286, Beyotime Biotechnology) at 95°C for 10 min and resolved using SDS‐PAGE (Tanon). Proteins were then transferred to 0.45 µm PVDF membranes, and the membranes were blocked with non‐fat milk and incubated with primary antibodies at 4°C overnight. The membranes were then washed and incubated with secondary antibodies (1:1000) for 1 h at 25°C. Protein bands were visualized using ECL solution (P0018S, Beyotime Biotechnology) in an ECL detection system (Tanon). The antibodies used in this experiment are listed in supplementary table.

4.11. Enzyme‐Linked Immunosorbent Assay (ELISA)

For the measurement of secretory protein concentrations, conditioned medium was collected and centrifuged for 20 min at 10 000 × g, and the supernatant was used to test the concentrations of TGF‐β1 (E‐EL‐0162, Elabscience), CCL22 (E‐EL‐H0029, Elabscience), and VEGFA (E‐EL‐H0111, Elabscience) with ELISA kits according to the manufacturer's instructions. The absorbance of the plates was measured at 450 nm wavelength using a microplate reader (Thermo Fisher Scientific).

4.12. Immunohistochemistry (IHC) Analysis

All tissue samples were fixed with 10% formalin, embedded in paraffin, and cut into 4 µm sections. The tissue sections were dewaxed and rehydrated. Slides were boiled in Tris‐EDTA buffer (Sigma) in a microwave oven for 15 min to retrieve the antigen. Endogenous peroxidase activity was blocked for 25 min with hydrogen peroxide solution (3%). Slides were subsequently blocked with 3% BSA and incubated with primary antibodies at 4°C overnight. Slides were then washed and incubated with horseradish peroxidase‐labeled secondary antibodies for 60 min at 25°C. The slides were subsequently incubated with DAB solution, counterstained, dehydrated, and mounted for analysis under a Nikon light microscope, where representative images were taken, and cells of interest per field of view were determined. The histochemistry score (H‐score) was calculated as follows: H‐score = (percentage of weak intensity area × 1) + (percentage of moderate intensity area × 2) + (percentage of strong intensity area × 3). The antibodies used in the experiment are listed in the Supplementary Table.

4.13. Immunofluorescence (IF) Staining

Tyramide signal amplification (TSA) technique was used to detect the expression of target antigens in PCa tissue paraffin sections. First, PCa tissue samples were routinely embedded in paraffin and sectioned into 4–5 µm thick slices. The slices were deparaffinized with xylene and rehydrated using a series of alcohol solutions with a concentration gradient, followed by washing with PBS buffer three times for 5 min each to remove residual alcohol. The sections were then treated with a 0.3% hydrogen peroxide solution for 20 min to eliminate endogenous peroxidase activity, followed by washing with PBS to remove hydrogen peroxide. To block non‐specific binding, the sections were incubated with 5% BSA for 1 h. Subsequently, the sections were incubated overnight at 4°C with primary antibodies specific to the target antigen. The following day, the sections were washed with PBS to remove unbound primary antibodies. The sections were then incubated with the secondary antibody from the TSA amplification system. The TSA system amplifies the fluorescence signal using hydrogen peroxide and substrate reactions, significantly increasing the sensitivity of immunolabeling. After incubation, the sections were washed again with PBS and mounted using an anti‐fade reagent. Finally, images of the stained tissue sections were captured, observed, and interpreted using a fluorescence imaging equipment. For cell samples, the IF protocol was generally unchanged.

4.14. Nano‐Flow Cytometry Analysis

EVs were analyzed using a nano‐flow cytometer (Flow NanoAnalyzer, NanoFCM) following the protocols described in our previous study [59]. Briefly, particles passing through the detector within 1 min were recorded in the range of 2000–8000 particles/min. The particle concentration, size distribution, and intensity of fluorescent signals were calculated using NanoFCM software (NanoFCM Profession V2.0) using calibration curves established with Quality Control Nanospheres (QS 2503, NanoFCM) and Silica Nanospheres Cocktail (S16M‐Exo, NanoFCM).

4.15. Cell Migration and Invasion Assay

Cell migration and invasion assays were performed using Transwell chamber (353097, Falcon). Briefly, after co‐culture of EVs or macrophage CCM with DU145 or PC3 cells for 48 h, cell suspensions (8 × 10⁴ PC3 cells or 5 × 10⁴ DU145 cells resuspended in 200 µL serum‐free culture medium per well) were seeded in the upper chamber of the Transwell system either with inserts coated with Matrigel (354234, Corning) for the cell invasion assay or without the Matrigel for the cell migration assay. Each lower chamber contained 800 µL culture medium (10% FBS for DU145 and 20% FBS for PC3). After incubation (DU145 for 24 h and PC3 for 48 h), the upper chambers were washed and cleaned with a cotton swab, and the cells that crossed the membrane were fixed with 4% paraformaldehyde, stained with 0.1% crystal violet (BL802A, Biosharp) and counted under a microscope in five random fields of view.

4.16. Wound Healing Assay

For the wound healing assay, 2 × 10⁵ DU145 cells or 3 × 10⁵ PC3 cells were seeded in six‐well plates. Cells were starved in serum‐free medium for 24 h, followed by treatment with EVs or CCM. After the cells reached approximately 100% confluence, a 200 µL pipette tip was used to create wounds in the cell monolayer. Representative images were taken at 0 and 24 h under a microscope at 200 × magnification. Wound closure was calculated using ImageJ Software (NIH ImageJ).

4.17. Transcriptomic Profiling of Single Cells

Single‐cell RNA sequencing (scRNA‐seq) was performed using the GEXSCOPE platform (Singleron Biotechnologies) following the manufacturer's protocols. Briefly, fresh tissues were dissociated into single‐cell suspensions using the Singleron PythoN Tissue Dissociation System. After viability assessment, single‐cell suspensions (2 × 10⁵ cells/mL in PBS) were loaded onto microwell chips for library construction. Barcoding beads were subsequently collected from the microwell chips, followed by reverse transcription of mRNA captured by the beads and PCR amplification of cDNA. The amplified cDNA was then fragmented and ligated with sequencing adapters. scRNA‐seq libraries were prepared using the GEXSCOPE Single Cell RNA Library Kit (Singleron), diluted to 4 nm, pooled, and sequenced on an Illumina NovaSeq 6000 platform with 150 bp paired‐end reads.

4.18. Analysis of scRNA‐Seq Data

Raw reads were processed to generate gene expression matrices using CeleScope algorithm (Singleron Biotechnologies) with default parameters. Briefly, Barcodes and UMIs were extracted from R1 reads and corrected, after which R2 reads were trimmed and aligned against the GRCh38 (hg38) transcriptome using STAR (v2.6.1b). Mapped reads were then assigned to exons with FeatureCounts (v2.0.1). Assigned reads, cell barcode, UMI, and genes were grouped into the gene expression matrices for downstream analysis.

Single cell data generated in this study was integrated by the Harmony algorithm with public scRNA‐seq datasets GSE210358, GSE268307, GSE274229, GSE137829, OMIX008930, and HRA002145 [60, 61, 62, 63, 64, 65]. Methods related to downstream single‐cell analysis have been described in our previous work [66].

4.19. Bulk Transcriptomic Sequencing and Analysis

Annoroad's integrated transcriptome sequencing service employs a standardized workflow encompassing rigorous sample QC, MGI‐platform library construction, and 150 bp paired‐end sequencing via cPAS technology. Total RNA was extracted with TRIzol reagent (99940001, Invitrogen). mRNA with ployA tails was enriched by Oligo(dT) magnetic beads, fragmented, reverse‐transcribed, and enriched with PCR to obtain cDNA libraries. The libraries were purified, quantified, and sequenced on the MGI platform. Analytical pipeline processes raw data through quality filtering, HISAT2‐based alignment to reference genomes, and FPKM quantification. Differential expression analysis utilizing DESeq2/DESeq identifies significantly regulated genes, followed by functional enrichment analysis via cluster Profiler to elucidate relevant GO terms and KEGG pathways (p < 0.05), delivering comprehensive transcriptional profiling insights. A core set of genes commonly upregulated by M2 EVs through a comparative analysis of RNA‐Seq data from DU145 and PC3 cell lines was identified. Differential expression analysis for each independent dataset—DU145 (across two sets: M2 EV vs. M0 EV, n = 6 vs. n = 3 and n = 3 vs. n = 3) and PC3 (M2 EV vs. M0 EV, n = 4 vs. n = 3) was performed using DESeq2, with significantly upregulated genes defined as those exhibiting an absolute foldchange > 1.5 and a p < 0.05. To ensure a sensitive and comprehensive representation of the DU145 response, the union of upregulated genes from its two constituent datasets was generated. This combined DU145 gene set was subsequently intersected with the upregulated genes from the PC3 dataset, yielding a final list of high‐confidence targets that are consistently induced by M2 EVs in PC3 cells and in at least one independent DU145 experimental group. Finally, public transcriptome data for TXNDC5, CC2D2B, HMCN1, and other genes were analyzed using the Xiantao Academic cloud platform (https://www.xiantaozi.com).

4.20. Lentiviral Infection

All lentiviruses were generated and purchased from TranSheep Bio, Shanghai. The shRNA target sequences are provided in Supplementary Tables. Viral infections were induced according to the manufacturer's guidelines. Briefly, for DU145 and PC3 cells, a cell suspension at a density of 1 × 10⁵ cells/mL was prepared in complete medium. The cells were cultured at 37°C for 16–24 h until 20%–30% confluence is reached. The following day, the conditioned medium was changed to complete medium containing 1 × Lv‐Enhance, and viruses were added at an appropriate concentration based on the Multiplicity of Infection (MOI) (40 for DU145 cells and 20 for PC3 cells) and virus titer. The cells were cultured at 37°C for 16 h before switching to regular complete medium. The culture medium was routinely replaced to maintain cell viability. After 48–72 h post‐infection, infection efficiency was assessed. Puromycin (1–5 µg/mL) was added 48 h post‐infection for selection and continuous selection was performed for 1–2 weeks to obtain stably transfected cell lines. For THP‐1 cells, a cell suspension at a density of 5 × 10⁵ cells/mL was prepared in complete medium containing 0.5 × Lv‐Enhance. Viruses were added at an appropriate concentration based on the cell MOI (50) and virus titer, and cells were cultured at 37°C for 12–16 h. After infection, the medium was changed, and cells were collected into 1.5 mL EP tubes before centrifugation at 500 g for 2–4 min. The supernatant was discarded, and complete medium was added. Infection efficiency was assessed approximately 48–72 h post‐infection. The subsequent selection strategy was consistent with the previous strategy.

4.21. EV‐Related mRNA Transfection and Electroporation

PKH67‐labeled M2 macrophages were obtained as described previously. Cy5‐tagged TXNDC5‐Flag was generated by Cynbio Technology, Shanghai. Transfection of mRNAs into M2 macrophages was performed as previously described [67]. Briefly, 1 × 10⁶ M2 macrophages were transfected with CALNP mRNA in vitro (DN002, D‐Nano Therapeutics) containing Cy5‐tagged TXNDC5‐Flag mRNA (5 µg/mL) for 6 h. Cells were then washed with PBS and replaced with EV‐depleted medium. The cells were cultured for another 48 h before EV separation from the conditioned medium via differential centrifugation. EVs were then added to the culture medium of DU145 or PC3 cells. Flag expression was analyzed using WB.

TXNDC5 mRNA was introduced into the EVs via electroporation. EV samples were diluted in Gene Pulser Electroporation Buffer (volume ratio of 1:9, Bio‐Rad), followed by the addition of TXNDC5 mRNA. The mixture was then incubated in a 4 mm Gene Pulser/MicroPulser Electroporation Cuvette (Bio‐Rad) on ice for 10 min. The cuvettes were then inserted into the Gene Pulser XcellTM Total System (Bio‐Rad) and electroporated using 200 V square waveforms with a pulse length of 10 ms and five pulses with 1 s intervals. The EVs were then purified with ultrafiltration to eliminate the buffer [68].

4.22. Animal Experiments

Animal experiments were approved by the Committee for Ethical Review of Research Involving Animal Subjects at Ren Ji Hospital (RJ2022‐1210). Nude mice (6–8 weeks, male) were purchased from GemPharmatech. The animals were bred under humidified and pathogen‐free conditions with a 12 h light/dark cycle at the animal laboratory of Ren Ji Hospital. DU145‐luc cells treated with EVs were suspended in PBS and injected into the left ventricle of the nude mice. At 0, 3, and 6 days after intracardiac injection, EVs were injected into the tail veins of the mice. The animals were subjected to bioluminescence imaging at Day 9 after the injection of cancer cells, after which the tumors were harvested, measured, and prepared for downstream analysis. Animals showing signs of being moribund or losing over 20% of their baseline weight were terminated before the experiment's predetermined end. Mice were euthanized with carbon dioxide.

4.23. Bioluminescence Imaging

In vivo bioluminescence imaging was conducted using a Xenogen IVIS Spectrum imager, which is equipped with a highly sensitive, cooled CCD camera housed in a light‐tight enclosure. Mice were administered luciferin (200 mg/kg, 40901, Yeasen) via intraperitoneal injection 10 min before imaging. The animals were then anesthetized with 1% isoflurane and positioned on a warmed platform within the imaging chamber, followed by a 1 min imaging session. Regions of interest (ROI) were defined using standardized rectangular areas for quantification. The resulting signal was quantified in photons per second (ph/sec) using Living Image software v.4.2 (Xenogen).

4.24. Statistical Analysis

All statistical analyses were performed using the GraphPad Prism software (version 8.0). Data are presented as the mean ± SD from at least three independent experiments. The normality of data distribution was assessed using the Shapiro‐Wilk test. For comparisons between two groups, if the data passed the normality test, an unpaired two‐tailed Student's t‐test was used. If the variances between groups were significantly different (as determined by an F‐test), Welch's correction was applied. For non‐normally distributed data, the non‐parametric Mann‐Whitney U test was employed. For comparisons among multiple groups, one‐way analysis of variance (ANOVA) was used for data meeting the assumptions of normality and homogeneity of variance (assessed by Brown‐Forsythe test). If the overall ANOVA was significant, post‐hoc comparisons were conducted using Tukey's test for comparisons of all groups, or Dunnett's test when comparing multiple groups to a single control group. If data were non‐normally distributed or variances were heterogeneous, the Kruskal‐Wallis test was used, followed by Dunn's post‐hoc test. Statistically significant outcomes were indicated as follows: ^*, p < 0.05; ^**, p < 0.01; ^***, p < 0.001; ****, p < 0.0001; and ns for non‐significant data.

Funding

This study was supported by the National Natural Science Foundation of China Projects (82103485, 82472142, 82227801, 82373358), the Major Natural Science Research Projects of Shanghai Municipal Education Commission (2023ZKZD23), the Shanghai Top Priority Research Center Project (2023ZZ02014), the Innovative research team of high‐level local universities in Shanghai, the Shanghai Rising‐Star Program (23QA1405900), the Shanghai Municipal Health Commission Outstanding Young Talents Program (2022YQ014), Shanghai Professional Technical Service Platform (23DZ2291000), Shanghai Jiao Tong University School of Medicine Institute of Molecular Medicine Clinical⁺ Excellence Program (2025ZYA‐001), the Physician‐Scientist Development Award from Shanghai Immune Therapy Institute, and the Shanghai Jiao Tong University 2030 Initiative.

Conflicts of Interest

The authors declare no conflict of interest.

Supporting information

Supporting File 1: advs73377‐sup‐0001‐SuppMat.docx.

Supporting File 2: advs73377‐sup‐0002‐Supplementary Table.xls.

Supporting File 3: advs73377‐sup‐0003‐Raw data of figures.xls.

Supporting File 4: advs73377‐sup‐0004‐clean metadata for analysis#Uff08RNA SEQ of EV‐treated PCa cells#Uff09.xlsx.

Data Availability Statement

The sequencing data have been deposited in the Genome Sequence Archive (GSA) of the National Genomics Data Center (accession number: PRJCA047512, https://ngdc.cncb.ac.cn/search/specific?db = bioproject&q = PRJCA047512). The raw single cell sequencing data are deposited in GSA‐human (accession number: HRA014199). The count matrix data have been deposited in OMIX, China National Center for Bioinformation/Beijing Institute of Genomics, Chinese Academy of Sciences (accession number: OMIX012217). Other data generated in this study are available in Additional files or from the corresponding author upon reasonable request.