At the level of diagnosis and risk stratification, exosomes are capable of simultaneously providing multi-level information^[35]. At the nucleic acid level (e.g., mRNA/miRNA), signatures often correlate more closely with pathway activity and functional states; at the protein level, immune checkpoints and profiles related to invasion, migration, and angiogenesis can indicate disease activity and immune context; meanwhile, lipid composition and metabolism-related characteristics may reflect tumor metabolic reprogramming and structural differences in membranes^[36]. Given the multi-omic nature of exosomes, single molecules are often insufficient to capture disease heterogeneity; therefore, a more rational approach involves constructing composite signatures and performing stratified validation across different stages or risk subgroups^[37,38].

Given the high heterogeneity and significant risk of metastasis associated with melanoma, the clinical demand for continuous longitudinal monitoring is particularly acute^[39,40]. The potential utility of exosomes in prognostic assessment is primarily manifested in their ability to: reflect fluctuations in tumor burden and systemic inflammatory/immune responses; indicate directional molecular changes prior to radiographically evident progression; and indirectly signal an increased risk of recurrence or metastasis through alterations in signaling pathways associated with matrix remodeling, immunosuppression, and angiogenesis^[41-43].

Compared with synthetic nanomaterials, exosomes originate from the cellular membrane system, and their lipid bilayer structure and membrane protein composition generally confer superior biocompatibility and stability in vivo^[44]. The receptors, adhesion molecules, and glycosylation patterns on the exosomal surface may influence target cell recognition, uptake, and intracellular trafficking pathways, thereby exhibiting delivery behaviors that more closely mimic physiological processes in certain contexts^[45]. Furthermore, the ability of exosomes to encapsulate diverse cargos—including nucleic acids, proteins, and small molecules—provides a versatile platform for the development of multi-component intervention strategies^[46-48]. It must be emphasized that exosomes are not entirely inert "empty shells"; their endogenous components may actively participate in signaling, which offers potential advantages while necessitating early evaluation of their safety profiles and application boundaries^[49,50].

In melanoma, insufficient response and acquired resistance remain the primary bottlenecks; consequently, the strategic positioning of exosomal delivery leans toward synergistic action^[51]. When combined with immune checkpoint inhibitors, the delivery of immunomodulatory molecules to improve immune infiltration and effector T-cell function theoretically offers the potential to increase response rates or delay resistance. Within the context of BRAF/MEK targeted therapies, compensatory interventions can be designed to address resistance-related pathways and post-treatment microenvironmental rebound. Furthermore, when integrated with local modalities such as radiotherapy, ablation, or photothermal/photodynamic therapy, exosomes may facilitate the transition from "local control" to "systemic immune response" by enhancing immunogenic cell death and antigen presentation^[52-54]. The commonality among these synergistic pathways is that exosomal delivery is not intended to replace existing therapies; instead, it seeks to amplify the systemic benefits of established regimens through more refined interventions at the local or cellular level, with efficacy validated by quantifiable incremental gains^[55].

Whether exosome-mediated delivery is truly viable in vivo depends on the controllability of its biodistribution and clearance^[56]. In practice, uptake by the reticuloendothelial system often limits the effective dose reaching the tumor; furthermore, variations in membrane composition and surface molecular profiles resulting from different sources and preparation protocols may lead to biased tissue distribution and undermine inter-study reproducibility. Simultaneously, the non-inert nature of exosomes implies that they may exert unintended regulatory effects in non-target tissues, necessitating vigilance regarding disturbances to immune homeostasis, particularly in the context of immunotherapy^[57-59]. Moreover, the route of administration (e.g., intravenous, local injection, or subcutaneous/lymphatic pathways) has a decisive impact on exposure and safety, directly shaping its application scenarios and development strategies^[56,57]. Consequently, effective exosomal delivery is not an inherent property; it requires a systematic trade-off between source selection, administration strategy, and downstream optimization to achieve relatively predictable in vivo performance^[58].

In summary, as natural membranous nanocarriers, exosomes provide a versatile platform for melanoma delivery and combination therapy, owing to their biocompatibility, functionalized surfaces, and capacity to carry diverse types of cargo^[55]. However, their clinical feasibility is collectively determined by in vivo drug distribution and clearance, immunological safety, and the consistency of large-scale manufacturing and quality control^[54]. Consequently, a more rational translational positioning is to view exosomal delivery as a complementary component of combination therapy.

The PD-1/PD-L1 axis is central to immune evasion in melanoma^[69]. In melanoma, PD-L1 is not only expressed on the surface of tumor cells and multiple immune cell populations, but is also closely associated with responses to immune checkpoint inhibitors and clinical prognosis^[70-73]. Recent studies have demonstrated that melanoma cells can deliver biologically active PD-L1 into the circulation via sEVs, thereby establishing a systemic “circulating immune checkpoint” and challenging the traditional notion that PD-L1 activity is confined to the local TME^[74-77]. In metastatic melanoma models, PD-L1 displayed on the surface of tumor-derived exosomes can directly bind to PD-1 on CD8⁺ T cells, suppress TCR signaling, reduce cytokine secretion, and impair cytotoxic functions, thereby promoting tumor progression^[74]. Blockade of sEVs-PD-L1 has been shown to restore T cell activity in vivo, induce systemic antitumor immune responses, and establish immunological memory, highlighting its independent and critical role in global immune regulation^[30].

At the bone marrow stage, melanoma-derived EVs can directly act on normal immature myeloid cells (IMCs)^[78]. Studies based on murine models have shown that danger-associated molecular patterns (DAMPs) carried by tumor EVs activate TLR4 signaling in IMCs, driving their differentiation into immunosuppressive myeloid-derived suppressor cells (MDSCs) and markedly upregulating PD-L1 expression, thereby suppressing T cell activation and proliferation^[78,79].

In addition, melanoma-derived EVs profoundly influence the differentiation and functional properties of antigen-presenting cells (APCs). Early studies have demonstrated that melanoma EVs inhibit the normal differentiation of monocytes into dendritic cells (DCs) and instead induce the formation of a CD14⁺ immature myeloid cell population^[80,81]. This process is partially dependent on the enrichment of TGF-β signaling within EVs, ultimately leading to impaired DC maturation, reduced antigen-presenting capacity, and compromised T cell proliferation and effector functions^[81-83]. Tumor-derived exosomes isolated from the plasma of melanoma patients similarly drive APCs toward a suppressive phenotype, with their immunomodulatory effects closely associated with the enrichment of immunosuppressive factors such as TGF-β within exosomes^[82,84-86].

In summary, melanoma-derived sEVs amplify and transform myeloid cell functional reprogramming that is initially confined to the local tumor site into a systemic, body-wide immunosuppressive effect through multilevel, cascade-like signal transmission and cellular reprogramming, thereby establishing a structurally stable and self-sustaining immunosuppressive barrier within the tumor microenvironment^[87-89], as summarized in Table 1.

To meet the increased demand for oxygen and nutrients associated with rapid proliferation, metastatic cancer cells can induce angiogenesis, remodel pre-existing vasculature, and recruit bone marrow-derived cells (BMDCs). Peinado et al.^[100,101] demonstrated that melanoma-derived exosomes confer a pro-angiogenic phenotype to BMDCs through horizontal transfer of the tyrosine kinase receptor MET, thereby promoting the formation of highly vascularized PMNs in organs prone to metastasis such as the lung^[100,101]. Subsequent validation studies further confirmed that sEVs derived from B16F10 melanoma mildly promote the formation of lung and bone metastatic lesions; although no significant reduction in metastatic burden was observed following treatment with MET-downregulated sEVs, a decreasing trend was still evident, indicating an important regulatory role of MET in this process^[102].

In addition to the vascular system, melanoma relies heavily on lymphatic routes to achieve early metastatic dissemination. Clinical studies have demonstrated that lymphangiogenesis is significantly associated with sentinel lymph node metastasis and reduced disease-free survival (DFS) in melanoma patients^[103]. Hood et al. were among the first to propose that melanoma-derived sEVs can be transported through lymphatic vessels to lymph nodes, where they remodel lymph node architecture prior to the arrival of tumor cells and guide melanoma cell migration toward “hotspot” regions enriched in tumor-derived exosomes^[104].

García-Silva et al. further systematically elucidated the molecular mechanisms underlying sEV-mediated lymphangiogenesis, demonstrating that metastatic melanoma sEVs enriched in NGFR (p75NTR) are internalized by lymphatic endothelial cells, activating MAPK/ERK and NF-κB pathways to promote lymphangiogenesis^[105]. Genetic ablation of NGFR or pharmacological inhibition of its signaling effectively suppresses lymph node metastasis and improves survival in murine models^[105].

Within the tumor microenvironment, sEVs have been demonstrated to induce the transdifferentiation of normal fibroblasts into cancer-associated fibroblasts (CAFs) and to extensively remodel the ECM^[106,107]. Studies show that sEVs derived from B16F0 murine melanoma promote the upregulation of CAF markers, including α-smooth muscle actin (α-SMA) and fibroblast activation protein (FAP), in NIH/3T3 fibroblasts while significantly enhancing their migratory capacity^[94,106]. In co-culture systems, fibroblasts conditioned by these sEVs further facilitate melanoma cell proliferation and migration^[96,106].

Melanoma sEVs are enriched in multiple oncogenic microRNAs, including miR-21 and miR-155-5p. miR-21 suppresses the expression of tissue inhibitor of metalloproteinases-3 (TIMP3) while upregulating multiple matrix metalloproteinases (MMPs), thereby markedly enhancing ECM degradation^[108,109]. In addition, membrane-type 1 matrix metalloproteinase (MT1-MMP), a key enzyme involved in matrix breakdown, has been identified in melanoma EVs released via a melanosome-associated secretory pathway, suggesting a melanoma-specific vesicle release mechanism tightly linked to ECM destruction^{[107,109,110]}.

Metabolic reprogramming represents another hallmark of melanoma progression. sEVs induce a "reverse Warburg effect" in stromal cells: exosomal miR-155 and miR-210 reprogram fibroblasts toward aerobic glycolysis, producing metabolites like lactate to fuel tumor cell oxidative phosphorylation^[111]. In this context, melanoma-derived sEVs are increasingly recognized as key mediators that induce and coordinate this “reverse” metabolic coupling between stromal and tumor compartments.

In reciprocal tumor-stroma communication, adipocyte-derived sEVs likewise exert profound effects on melanoma metabolism. Lazar et al. reported that adipocyte-derived sEVs enhance the migratory and invasive capacities of melanoma cells and are enriched in proteins associated with fatty acid oxidation (FAO)^[112]. In the presence of adipocyte sEVs, intracellular FAO levels in melanoma cells are markedly increased, offering a metabolic explanation for the association between obesity and poor melanoma prognosis^[112]. Subsequent studies further demonstrated that FAO-related proteins and their fatty acid substrates can be directly transferred to melanoma cells via sEVs, establishing a form of intercellular metabolic support^[112,113]. Collectively, these findings indicate that melanoma not only remodels the tumor microenvironment through its own secreted sEVs but can also acquire and exploit sEVs derived from distant metabolic organs, such as adipose tissue, to obtain exogenous metabolic resources that fuel tumor progression.

As a class of bio-nanomaterials, exosomes are highly sensitive to sample processing conditions and detection platforms^[23]. Regarding EV isolation techniques, while ultracentrifugation remains the standard method, approaches more commonly adopted in clinical settings include SEC, polymer-based precipitation, ultrafiltration, and immunoaffinity capture. Each method has its advantages and limitations, and the choice should be based on research objectives and clinical practicality^[122].

In addition to peripheral blood-based sampling, ocular biofluids also provide clinically relevant proximal sources of tumor-derived vesicles in uveal melanoma, as illustrated in Figure 4B.

PD-L1, which is abundantly present in exosomes (exo-PD-L1), has been demonstrated to play a significant role in immunotherapy. PD-L1 is the target of immune checkpoint inhibitors, such as PD-1 inhibitors. Exo-PD-L1 can be regarded as an extension of the PD-1/PD-L1 axis, forming a systemic immunosuppressive layer in bodily fluids. This layer weakens the function of CD8⁺ T cells and promotes tumor immune evasion^[74].

The functional role of exo-PD-L1 as a decoy implies that PD-L1 molecules on exosomes can competitively bind to the PD-1 receptor, thereby inhibiting the immune system's attack on the tumor. This mechanism may explain why some melanoma patients with high plasma PD-L1 levels still exhibit poor responses to immunotherapy. Although the efficacy of immune checkpoint inhibitors theoretically relies on PD-1/PD-L1 binding, exo-PD-L1 may interfere with this effective immune response through its decoy function.

With ongoing clinical research, the dynamic changes in exo-PD-L1 levels are increasingly recognized as a real-time monitoring indicator that more accurately reflects the systemic immune state. Multiple cohort studies have found that plasma exo-PD-L1 levels are closely associated with tumor burden, degree of immunosuppression, and clinical outcomes. High baseline levels are often observed in patients with greater tumor burden and poorer prognosis, while a significant decline during early treatment (approximately 2-6 weeks) is commonly seen in patients who achieve objective responses or disease control^[123].

Beyond exo-PD-L1, several other immunosuppressive biomarkers detectable in exosomes provide valuable insights into the immune microenvironment. Taking EV-CD73 as an example, CD73 is a key molecule in the adenosinergic immunosuppressive pathway^[124]. By catalyzing the conversion of AMP to adenosine, CD73 inhibits effector T cell function while promoting the activity of regulatory T cells (Tregs) and MDSCs^[125]. This mechanism plays a significant role in shaping the immunosuppressive tumor microenvironment^[126]. Studies have shown that the expression of CD73⁺ EVs in the plasma of melanoma patients is significantly higher than in healthy controls, and their adenosine-generating capacity can suppress lymphocyte proliferation and IFN-γ secretion^[127].

More importantly, clinical research has found that high or increasing levels of CD73⁺ EVs before or during early treatment in patients receiving PD-1 inhibitors are associated with treatment resistance and poorer prognosis^[127]. Therefore, CD73⁺ EVs may serve as an early biomarker predictive of resistance and represent a potential therapeutic target for combination strategies, such as co-administration with A2A/A2B receptor antagonists, to enhance the efficacy of PD-1 inhibitors^[128].

TGF-β-associated exosomes are another important marker in the immune microenvironment. TGF-β promotes fibrotic stromal remodeling, Treg recruitment, and exhaustion of effector T cells, mechanisms closely linked to immune escape and an immunosuppressive phenotype^[129]. Multiple studies indicate that high levels of EV-TGF-β before or early in treatment suggest limited benefit from immunotherapy and poorer outcomes. Evidence supports including EV-TGF-β in risk stratification models for immunotherapy^[87,130].

uPAR⁺ EVs are strongly associated with tumor invasiveness and metastatic potential^[131]. Research has found that in patients with metastatic melanoma, early increases in uPAR⁺ EV levels during treatment are significantly correlated with drug resistance and shorter overall survival (OS). These exosomes, taken up by endothelial cells, can promote angiogenesis and increase vascular permeability, thereby facilitating conditions for tumor metastasis^[132].

Therefore, combined assessment of biomarkers such as EV-CD73, EV-TGF-β, and EV-uPAR allows for a more precise characterization of the immunosuppressive microenvironment and identification of resistance mechanisms during immunotherapy^[133]. Particularly in the course of treatment, when persistently high or rising exo-PD-L1 levels coincide with upregulation of EV-CD73 and/or EV-TGF-β, and ctDNA levels do not decrease or show signs of rebound, this pattern may indicate a high risk of immune escape and treatment resistance [Figure 4C]^[87,127,134].

EV-DNA refers to tumor-derived DNA fragments encapsulated within nanoscale vesicles. Its primary advantages include high stability due to protective encapsulation and a relatively high tumor-origin fraction. Compared to cell-free DNA (cfDNA), EV-DNA can more effectively preserve tumor-specific genetic information^[135,136]. This protection makes EV-DNA a more stable analyte for liquid biopsy applications, particularly advantageous for detecting low-frequency mutations^[88].

Furthermore, EV-DNA can be used for confirming mutational profiles before and after targeted therapy and for tracking clonal evolution. Especially in patients where repeat tissue biopsies are challenging, EV-DNA serves as a complementary source to ctDNA, improving the success rate of mutation detection. During long-term targeted therapy, it can help identify newly emerging resistant clones. For example, in some patients undergoing targeted therapy, rapid changes in the mutational profile detected in EV-DNA may indicate the emergence of new resistance mechanisms, information which ctDNA alone might fail to capture in a timely manner^[89].

In contrast to EV-DNA, which focuses on the detection of structural mutations, EV-RNA primarily conveys functional information^[137]. EV-RNA, including mRNA, miRNA, and lncRNA, reflects transcriptional activity within tumor cells and offers unique advantages in tracking non-mutational resistance mechanisms. Unlike ctDNA, which lacks functional readouts, EV-RNA can reveal tumor-associated changes such as transcription factor activation and gene splice variants, which are part of the adaptive evolution of cancer cells^[138].

Through the analysis of EV-RNA, clinicians can identify key pathways involved in immune escape and detect gene splice variants at an early stage, enabling timely adjustment of treatment strategies to prevent the further development of resistance. Unlike ctDNA, EV-RNA captures more complex functional adaptations, illustrating how tumor cells may overcome therapeutic pressure through transcriptional regulation^[138].

ctDNA and circulating tumor-derived extracellular vesicles (ctEV, encompassing EV-DNA and EV-RNA) together form a multi-dimensional liquid biopsy framework with significant potential for tracking resistance mechanisms, monitoring minimal residual disease (MRD), and evaluating immunotherapy efficacy^[139]. ctDNA primarily provides structural information such as tumor mutational burden, clonal expansion, and copy number variations. In contrast, ctEV, via its nanovesicular carriers, integrates DNA, RNA, and membrane protein-based functional biomarkers, thereby supplementing information on functional states such as immune evasion and metastatic potential [Figure 4D]^[108,140].

The major differences and complementary features of ctDNA and exosome-related biomarkers are summarized in Table 2.

EV-miRNAs are recognized as biomarkers encapsulated within nanoscale vesicles, holding significant promise for early diagnosis and staging of melanoma^[143]. Utilizing high-throughput screening and machine learning, multiple studies have identified EV-miRNA panels capable of distinguishing melanoma from benign pigmented lesions, as well as early-stage from late-stage disease^[144]. For instance, one study reported that a plasma-based EV-miRNA panel comprising miR-200c-3p, miR-144-3p, and miR-221-3p demonstrated favorable AUC values for melanoma diagnosis and staging. Notably, it showed high sensitivity and specificity (AUC > 0.85) for early-stage detection^[144].

EV-proteins represent a nanoscale extension of traditional serum biomarkers, offering significant diagnostic and prognostic value. For example, classical melanoma markers such as S100B and melanoma inhibitory activity (MIA) have been detected in circulating EVs from patients with advanced melanoma, suggesting their potential utility as EV-based biomarkers^[108,145]. Studies indicate that EV-S100B achieves an AUC value of approximately 0.85 in melanoma diagnosis, highlighting the advantage of this nanoscale protein biomarker^[146].

Furthermore, research on communication proteins like EV-Connexin-43 demonstrates that their expression levels are significantly correlated with DFS and OS in melanoma^[142]. ROC analysis shows that EV-Connexin-43 yields an AUC of about 0.77-0.78, supporting its potential as a supplementary marker for prognostic stratification and treatment intensity decisions^[142]. Through small EV proteomics, researchers continue to identify candidate signature panels associated with metastasis risk and immune evasion—including adhesion molecules, cytoskeletal proteins, and metabolic enzymes—providing features for multi-marker panels and machine learning models^[124].

In special lineages of melanoma, such as uveal melanoma, EVs isolated from tumor-proximal biofluids (e.g., aqueous humor, vitreous) often carry a high abundance of tumor-associated information^[147]. While the proteomic and miRNA signatures of these EVs overlap with those in peripheral blood, their enrichment is significantly higher. This confers a considerable advantage for "proximal nanoscale readouts" of localized lesions^[148]. Studies show that EV concentrations in tumor-proximal fluids are typically greater than in peripheral blood, substantially improving the tumor signal-to-background noise ratio. This makes them particularly suitable for diagnosis, post-operative assessment of residual disease, and recurrence monitoring^[149].

In summary, nanoscale readouts such as EV-miRNA and EV-proteins primarily reflect the phenotypic characteristics and microenvironmental state of tumors, whereas ctDNA conveys information on mutational burden and clonal expansion. By integrating these with imaging and traditional serum biomarkers, a multimodal framework encompassing "structure + function + behavior" can be constructed, enabling more precise whole-course patient management [Figure 4E]^[150]. In practical clinical pathways, EV-based liquid biopsy is likely to be initially adopted as a complementary tool alongside ctDNA and conventional markers within high-level medical centers. Its incremental value at specific clinical decision points can then be progressively validated through real-world cohort studies^[151].

The combined analysis of EV-miRNA, EV-proteins, and ctDNA offers significant advantages in addressing clinical challenges. For instance, EV-miRNA can aid in distinguishing melanoma from benign lesions, EV-proteins enhance the diagnosis of advanced tumors, and ctDNA provides insights into mutational burden and clonal dynamics. Integrating these layers of information allows for more accurate early diagnosis, treatment stratification, and recurrence monitoring. Particularly in scenarios where imaging and traditional serum biomarkers yield ambiguous results, EVs liquid biopsy delivers additional clinical value, holding broad application prospects in early melanoma detection and the dynamic monitoring of immunotherapy response^[152].

Passive loading strategies rely on the natural interactions between drug molecules and exosomes or endogenous cellular pathways, offering operational simplicity but generally lower loading efficiency, and mainly include co-incubation and co-culture approaches. Representative examples are shown in Figure 6A-C (tumor-derived EV-based doxorubicin delivery and antitumor evaluation), Figure 6D and E (engineered blood exosome-based gene/chemo co-delivery characterization), and Figure 6F-M (producer-cell engineering/co-culture-associated loading strategies using TRAIL-expressing and triptolide-loaded engineered exosomes)^[162-164].

Co-incubation is one of the simplest passive loading methods, in which exosomes are incubated with therapeutic agents under defined conditions to enable drug encapsulation via diffusion, lipid membrane fusion, or transmembrane transport mechanisms^[165]. In melanoma therapy, doxorubicin-loaded exosomes generated by co-incubation exhibit favorable biocompatibility and low immunogenicity, thereby reducing systemic toxicity while enhancing drug accumulation at tumor sites^[166]. Nevertheless, the drug-loading capacity of this method is limited and remains inefficient for macromolecular or hydrophilic therapeutics. For instance, when doxorubicin was loaded into HBc virus-like particle nanoparticles, incubation at 70 °C for 90 min achieved high loading efficiency, highlighting temperature as a critical determinant of encapsulation efficiency^[167].

This strategy involves genetically modifying donor cells (e.g., MSCs, HEK-293T cells) to stably express fusion proteins composed of targeting elements and exosomal membrane proteins such as Lamp2b or CD63, thereby enabling the continuous production of targeted exosomes at the source^[168-170]. Its advantages include uniform targeting functionality and avoidance of potential carrier damage during subsequent purification.

To target integrin αvβ3, which is overexpressed on melanoma cell surfaces, researchers fused the high-affinity cyclic RGD (cRGD) peptide with the extracellular domain of lysosome-associated membrane protein 2B (Lamp2b), successfully generating targeted engineered exosomes [Figure 7A-F]^[171]. In B16-F10 melanoma-bearing mouse models, iRGD-modified exosomes loaded with doxorubicin exhibited over threefold higher fluorescence intensity at tumor sites compared to unmodified exosomes, accompanied by significant tumor growth inhibition, prolonged survival, and markedly reduced systemic toxicity^{[169,172,173]}.

Direct functionalization of exosomes refers to a class of engineering strategies in which targeting molecules or functional moieties are anchored onto the exosomal membrane after isolation and purification through chemical or physical approaches. This method does not require genetic modification of donor cells, offering high modularity, operational flexibility, and broad applicability, making it widely adopted for constructing precise exosome delivery systems.

Chemical conjugation is one of the most established direct modification methods. It primarily exploits reactive functional groups exposed on exosomal membrane proteins or lipids, such as amino, carboxyl, or thiol groups, to covalently link targeting ligands or functional molecules^[174]. Bioorthogonal click chemistry (BCC), such as azide-alkyne cycloaddition, is considered an ideal tool for exosome surface modification due to its mild reaction conditions, high selectivity, and excellent biocompatibility^[175]. For example, using copper-free BCC, researchers metabolically engineered donor cells to produce surface-modified EVs, which demonstrated effective tissue targeting and accumulation in rheumatoid arthritis and various tumor models, including breast, colon, and prostate cancers^[176]. In melanoma, exosomes can potentially be functionalized via DBCO-NHS chemistry on their surface and subsequently conjugated with azide-modified PD-L1 antibodies, generating exosomes that target PD-L1-overexpressing melanoma cells. This strategy blocks the PD-1/PD-L1 immune checkpoint while enabling specific drug accumulation within tumor cells, enhancing the synergistic efficacy of immune checkpoint inhibitors and targeted therapeutics^[177-180]. Importantly, this approach preserves the native targeting properties and bioactivity of exosomes while avoiding structural or functional damage to the exosomal membrane, making it suitable for in vivo imaging and precise therapy^[122].

Additionally, Bolaños et al. conjugated gold nanorods to B16F-derived exosomes and chemically modified them with methotrexate (MTX). Leveraging MTX's high affinity for folate receptors overexpressed in melanoma cells, this engineered system achieved a gold accumulation in tumor tissue of 5.5% of the injected dose, compared with ~1% using conventional nanodelivery systems^[181]. Consistent with the updated Figure 7G-J, Duan et al. developed a pYEEIE peptide-functionalized Rhodiola rosea-derived exosome-like nanovesicle system for targeted doxorubicin delivery, in which the pYEEIE ligand and RELN carrier were used to enhance melanoma-targeted accumulation and antitumor efficacy [Figure 7G-J]. Similarly, Chen et al. developed anti-CD20 antibody-conjugated, doxorubicin-loaded exosomes, which demonstrated therapeutic potential in targeting CD20+ A375 and WM266 melanoma models^[182].

Membrane fusion is another important direct functionalization technique. By co-processing exosomes with synthetic liposomes or engineered cell membrane vesicles under extrusion conditions, physical fusion and functional integration of distinct membrane components can be achieved^[183-185]. For example, dendritic cell-derived exosomes fused with B16-F10 melanoma cell membranes via micro-extrusion produced hybrid vesicles with biomimetic properties. This hybrid system retained MHC molecules from dendritic cell exosomes and transmembrane proteins from melanoma cell membranes, enabling efficient homing to subcutaneous melanoma and lung metastases, and acting as an antigen presentation platform to enhance CD8⁺ T cell activation and proliferation, effectively inhibiting tumor growth and metastasis^[186]. Furthermore, Gao et al. designed T cell-derived photoreactive hybrid γδ-T cell exosomes for targeted photodynamic therapy. Upon light irradiation, these exosomes induced melanoma cell apoptosis and generated reactive oxygen species (ROS) to achieve synergistic antitumor effects^[187].

To further enhance exosome penetration and targeting specificity in tumor tissues, researchers have developed multistage targeting systems and stimulus-responsive smart targeting systems, simulating programmed in vivo delivery and enabling precise recognition of the tumor microenvironment.

Multistage targeting systems activate different functional modules sequentially, mimicking hierarchical drug delivery mechanisms in vivo. For example, in the first targeting stage, the broad affinity of cRGD peptides for integrin αvβ3 allows exosomes to initially anchor on the endothelium of tumor neovasculature. In the second targeting stage, a matrix metalloproteinase-2 (MMP-2) sensitive peptide linker connects a cell-penetrating peptide to the exosome surface. Upon entering the tumor microenvironment, highly expressed MMP-2 specifically cleaves this linker, exposing and activating the cell-penetrating peptide, thereby facilitating exosome transvascular and dense ECM penetration and subsequent tumor parenchyma entry. Studies show that the tumor penetration depth of such multistage systems is over 2.5-fold greater than that of single-targeting systems^[163,188]. Similarly, Kang et al. constructed an active targeting system (ASL) by combining a membrane-anchoring group (BODIPY)-spacer (PEG)-targeting ligand (cRGD peptide)^[172]. In a B16-F10 melanoma mouse model, the system enabled precise tumor localization via specific cRGD-integrin interactions, integrating in vivo imaging, active targeting, and drug delivery functions [Figure 7K-O]^[172].

By precisely regulating delivery routes and the tumor immune microenvironment, multistage and stimulus-responsive smart exosome systems significantly enhance antitumor targeting and therapeutic efficacy, demonstrating broad clinical potential in immunotherapy and precision treatment of highly invasive tumors such as melanoma^[189]. Representative examples of active loading and hybrid-engineering strategies include: electroporation-based exogenous loading of miR-195-5p into EVs for melanoma-targeted therapy sensitization [Figure 8A-F]^[190], ultrasound-assisted delivery in a murine melanoma model [Figure 8G-J]^[191], and fusion-based exosome-liposome hybrid nanoplatform co-delivering TP and miR497, included as a cross-tumor technical example of hybrid construction and evaluation [Figure 8K-N]^[183].

To synergistically enhance cytotoxic tumor-cell killing and immune activation, engineered exosome-based systems may be designed either to improve targeted chemotherapy or to combine therapeutic delivery with immune microenvironment modulation. For example, pYEEIE-RELNs-DOX achieved targeted doxorubicin delivery and antitumor efficacy in melanoma-bearing mice^[192], whereas exosome-mediated co-delivery of transforming growth factor-β receptor 1 kinase inhibitors and toll-like receptor 7/8 agonists provides a cross-tumor example of combining vesicle-based delivery with immunomodulatory therapy^[193]. These examples support the rationale for exosome-assisted combination therapy, but melanoma-specific chemoimmunotherapy designs still require direct validation in appropriate preclinical models.

Exosomes are highly efficient carriers for the delivery of nucleic acid-based therapeutics, but the current cited evidence should be framed more cautiously. In melanoma, EVs loaded with miR-195-5p have been shown to sensitize melanoma cells to MAPK-targeted kinase inhibitors and to reduce tumor growth in vivo^[190]. Beyond melanoma-specific EV data, engineered tumor-derived exosomes have also been explored in other tumor contexts to modulate chemoresistance and metastatic behavior^[194], while non-EV RNAi nanoplatforms such as peptidomimetic lipid nanoparticles have demonstrated that STAT3 silencing can downregulate PD-L1 and inhibit melanoma growth^[195]. Together, these findings support the conceptual feasibility of nucleic-acid-based nanodelivery, but they do not yet establish a mature exosome platform co-delivering BRAF V600E-specific siRNA and MEK inhibitors in melanoma.

By integrating imaging probes and therapeutic agents on the surface or within exosomes, real-time monitoring of treatment processes and quantitative evaluation of therapeutic efficacy can be achieved. In image-guided surgery, melanoma-targeting exosomes labeled with near-infrared (NIR) fluorescent dyes (such as Cy5.5 and DiR) can be directed to tumor tissues, enabling intraoperative, high-contrast visualization of micrometastatic lesions and metastatic lymph nodes. This approach assists surgeons in accurately delineating surgical margins and improves the completeness of tumor resection^[196,197].