The desired properties of drug carriers include efficient cellular entry, near-natural physicochemical properties, and the ability to evade immune responses (58, 59). NVs are derived from natural and synthetic vesicular carriers. The types of natural lipid NVs include exosomes, virosomes, bacterial ghosts, and erythrocyte ghosts (60). Conversely, synthetic NVs were created to mimic the physicochemical properties of liposomes (61). Liposomes contain a lipid bilayer surrounding an aqueous core to allow the encapsulation and protection of hydrophilic molecules such as miRNA or DNA (62).

Liposomes have been widely used in drug delivery because their structure can effectively entrap various drugs and then transport cargo to target sites (63). This approach has demonstrated strong therapeutic efficacy in some cancer types (64, 65). However, liposomes have poor selectivity for cancer cells, resulting in severe systemic side effects (66). Conjugating liposomes with specific molecules, such as ligands, antibodies, or small molecules, improves selectivity and cellular targeting (67–69). By mimicking EV properties, synthetic NVs created from biomimetic phospholipid bilayers result in several improvements such as increased solubility, prolonged action, reduced toxicity, and lower adverse effects (66, 70–72). Nonetheless, the issues limiting the utility of synthetic NVs are immunorecognition as foreign substances and immune clearance by phagocytic cells (73).

In this regard, RBCEVs have proven extremely safe, and they can be used as robust carriers clinically because of their biocompatibility (74). Regarding biosafety, biocompatibility, efficiency, accessibility, and cost-effectiveness, RBCEVs are superior to conventional RNA delivery systems such as tripartite formulations with RNA, cationic polymers, and anionic liposome-encapsulated neutral lipopolyplexes (15, 75). Although conventional RNA delivery systems such as lipid nanoparticles are more stable than RBCEVs, they cause toxic side effects, and they are rapidly cleared from the circulation (76). RBCEVs have been used as carriers for RNA-based therapeutics to facilitate the effective delivery of both short RNA molecules and long mRNA molecules to their target sites for cancer therapy (47). RBCEVs loaded with RNA molecules display long-term stability and retain their functional capacity for long periods (77). Moreover, RBCEVs have great potential in drug delivery platforms because they can penetrate anatomical barriers and display sufficient binding (78, 79). This outstanding drug delivery platform carries special properties that make it suitable for drug delivery approaches (15). Further development of cancer-targeting peptide- or antibody-coated RBCEVs may result in improved target specificity and reduced adverse side effects in normal tissues. In addition to therapeutic agent delivery, RBCEVs can be applied to deliver ultra-small superparamagnetic iron oxide particles into human bone marrow mesenchymal stem cells for cellular magnetic resonance imaging to increase the performance of stem cell therapies (74). In addition, ^99mTc has been delivered to white blood cells via RBCEVs to observe organ inflammation in a mouse model using a gamma camera (80). This novel strategy using RBCEVs as delivery vehicles overcomes the limitations of traditional imaging including low intracellular labeling efficiency and biosafety concerns (74).

Production upscaling is perhaps easier for synthetic NVs. However, it should be noted that RBCEVs can be easily prepared at a relatively low cost from RBC units available in blood banks. Chemical induction (using modalities such as calcium ionophores) to enhance RBCEV release is an interesting scaling-up strategy for large-scale preparation and clinical applications (22, 81). Moreover, RBCEVs retain their stability and efficiency of delivery without any harmful effects even after multiple freeze–thaw cycles (82). As previously mentioned, CD47 expressed on the surface of RBCEVs prevents phagocytosis through an interaction with SIRPα (54), thus supporting the stability of RBCEVs after intravenous administration. In addition, EVs can efficiently penetrate the blood–brain barrier (83). Signaling molecules on the RBC membrane, which is a component of RBCEVs, can inhibit immune cell engulfment via an interaction between CD47 and SIRPα and defend against complement system attack via C8 binding protein, homologous restriction protein, DAF, membrane cofactor protein, complement receptor 1, and CD59 (84–86). This property of RBC membranes was applied to coat nanoparticles to increase their half-lives in blood circulation for drug delivery (87). Notably, it is feasible to prepare autologous RBCEVs for therapeutic agent loading (88).

Taken together, RBCEVs display advantages over conventional drug carriers in terms of high biocompatibility with limited immunogenicity, simple scaling-up, and high stability (15). Nonetheless, there are some disadvantages of RBCEVs for drug delivery. RBCEVs require a robust isolation method to separate them from blood cells and contaminating proteins. The heterogeneity of EV populations, including differences in EV size in the isolates, is nearly unavoidable depending on the isolation methods. A systematic comparison of EV isolation methods on the quality and quantity of plasma EVs indicated that ultracentrifugation (the gold standard) was the most appropriate method. Ultracentrifugation provided better EV purity compared to ExoQuick (System Biosciences), Total Exosome Isolation (TEI, Invitrogen), size exclusion chromatography (qEV), ultrafiltration, and exoEasy (Qiagen, membrane-based affinity binding) (89). However, the highest recovery rate was yielded by qEV (~60%), while ultracentrifugation and ultrafiltration yielded ~40% recovery rate (89). Polymer-based precipitation had impurity particles while exoEasy kit caused fusion and aggregation of EVs during the isolation process (89). Microfluidic and antibody selection platforms based on antigen-specific capture were successfully applied to isolate tumor-specific EVs (90). Unfortunately, microfluidic platform could separate EVs in a relatively small amount, i.e., 100 EVs per 1 μl and may cause EV aggregation during the isolation process (90). An optimized protocol for RBCEV preparation including an additional quality-control step is required to minimize batch effects and ensure the reproducibility of RBCEV applications. Notably, there is no study to clarify the normal range of EV concentration in the human body and it is unclear how various EV distributions in healthy or disease states might affect the efficacy of RBCEV-based therapy. For example, neuronal-enriched EV levels had not changed between healthy individuals and patients with Alzheimer's disease (91), so further study of RBCEV therapy in this disease context have no confounding from other EV distribution. Human lactoferrin could promote EV releasing from human adipose-derived stem cells (92), so the diseases with evidence of plasma lactoferrin changes might affect the interpretation of RBCEV therapeutic efficacy. The relationship between (exogenous) therapeutic RBCEVs and (endogenous) EV distribution should be clarified in future studies.

Currently, which cell types represent the best sources of EVs for drug delivery remains unclear. Because EVs carry the membrane ligands and receptors of their parental cells, different cell types may produce EVs with differing delivery proficiency and targeting selectivity (81). RBCs, endothelial cells, monocytes, granulocytes, and platelets have been reported as cell sources for EV-based drug delivery (93). Conversely, fibroblast- and dendritic cell-derived EVs are not stably obtained from all subjects (94, 95), whereas cancer cell lines may release EVs that promote tumor development (96, 97). Various circulating cell type-derived EVs, especially those derived from nucleated cells, might contain genetic material, leading to horizontal gene transfer to recipient cells (98). Whole plasma is a major source of EVs that is easily obtained and readily available. However, whole plasma-derived EVs are heterogeneous, and they may contain several (unknown) substances (42).

Blood exosomes were engineered by co-embedding of drug and cholesterol-modified miR-21 inhibitor with high payloads into the lipid bilayer of exosomes (99). Moreover, superparamagnetic molecules and targeting proteins/peptides were loaded into the exosome membrane using ligand-receptor coupling and electrostatic interactions to enhance delivery to tumor cells and then inhibit tumor growth (99, 100). For example, in Parkinson's disease treatment, the engineering blood exosomes were applied to loaded dopamine by a saturated solution incubation method into blood transferrin receptor positive exosomes which were purified by multiple superparamagnetic nanoparticles labeled with transferrin (101, 102). Additionally, in type 2 diabetes mellitus treatment, a potential therapeutic peptide BAY55-9837 was loaded into exosome and coupled with superparamagnetic iron oxide nanoparticles with pancreas islet targeting activity to increase insulin secretion (103).

Interestingly, several properties of RBCEVs allow them to overcome the limitations of other cell source-derived EVs. First, RBCs are easily obtained and stored for prolonged periods after blood transfusion. Second, RBCEVs have long been present as a hidden component of transfused RBCs, which highlights their safety and biocompatibility. Third, RBCEVs have a low risk of horizontal gene transfer during delivery because RBCs lack nuclear and mitochondrial DNA (44). Finally, RBCEV release can be triggered by several processes, such as membrane complement activation and calcium influx. This EV release process can be applied to produce a large number of RBCEVs for experimental and clinical applications (81). Nonetheless, RBCEVs should be considered a blood product, and as such, blood group compatibility must be considered. In this regard, autologous RBCs can be an ideal source of EVs to avoid blood group incompatibility or immunorecognition.

For allogeneic treatments in patients with cancer, RBCEVs are safer than plasma EVs because cancer and immune cells generally release an extremely large number of cancer-promoting EVs into the circulation (96, 97).

A comparison of drug delivery characteristics between RBCEVs and non-RBC-derived EVs is presented in Table 3.

EVs, as natural carrier systems, efficiently deliver complex molecules including proteins, nucleic acids, lipids, and sugars similarly as their parental cells (109, 110). Moreover, because they have similar membrane properties as their parental cells, EVs can easily internalize into parental cells as well as target cells via clathrin-independent endocytosis and macropinocytosis (109). The different cell types may use different EV uptake pathways such as membrane fusion, phagocytosis, micropinocytosis, and endocytosis (111). EVs may bind to surface receptors of targeted cells, trigger intracellular signaling cascades, and then mediate EV uptake depending on EV composition and origin (112). For example, monocyte-derived dendritic cells could uptake the milk-derived EVs via dendritic cell-specific intercellular adhesion molecule-3-grabbing non-integrin (DC-SIGN) and mucin1 (MUC1) protein interaction and phagocytosis, but not the EVs derived from other sources or lacking MUC1 (113). Furthermore, EV uptake capability depends on the recipient cell types but not the donors (114). For instance, EV uptake of human colon carcinoma cells was mediated by clathrin-dependent endocytosis, but that in human lung carcinoma cells was mediated through neither clathrin- nor caveolin-dependent endocytosis (114). However, RBCEVs can internalize into cancer cells through their primary membrane components (i.e., phospholipids) (14, 115). To reduce the loss of drugs/molecules during transport to target cells, RBCEVs are designed to internalize drugs/molecules and reach the target cells without inducing immune system attack and drug/molecule loss, leading to increased treatment efficacy (116).

Drugs/small molecules are easily loaded into EVs. For example, the anti-inflammatory agent curcumin was loaded into exosomes via incubation at 22°C for 5 min (105). Curcumin-loaded exosomes were more stable than free curcumin in vivo following intranasal administration (105, 116). In addition, a heat shock technique for bacterial cell transfection (incubation on ice for 30 min followed by 42°C for 60 s) and five rounds of electroporation at 500 V using a 10-ms pulse were used to load miR-15a mimic/inhibitor into exosomes (117, 118). The EV loading protocol should be optimized to account for differences in properties among different EV sources (119). The size of EVs may also influence the size and number of loaded drugs/molecules (120). MVs can carry larger amounts of linear and plasmid DNA than EVs following electroporation (120). Additionally, smaller linear dsDNA (<750 bp) was loaded to EVs (85 ± 41 nm) at higher amounts than larger dsDNA (>1,000 bp) (120). Furthermore, miRNA loading into EVs has been optimized via incubation at 22°C for 2 h at pH 2.5 (121). Notably, unlike miRNA loading methods of RBCEVs that were comprehensively evaluated and optimized (15, 121), the protocols for small molecule drug-loading into RBCEVs require further studies in a systematic manner.

RBCEVs may have some cellular effects because they participate in several biological processes including oxidative stress, inflammation, NO homeostasis, thrombosis, and foam cell formation (122). In oxidative stress, RBCEVs can upregulate NADPH oxidase expression via the excessive production of ROS by activated neutrophils through respiratory burst (123, 124). This may change cytoskeletal and cell membrane asymmetry, leading to Oxi-ERY formation and hemolysis. This can ultimately cause cholesterol release, lipid peroxidation production, and protein and iron aggregation, thereby inducing vascular cell damage (24, 125). Furthermore, during inflammation, components on the membrane of RBCEVs, including cholesterol (induces inflammation reaction), iron and myeloperoxidase (catalyst and source of ROS production, respectively), hemoglobin (activates pro-inflammatory transcription factor), and phospholipase A₂ (hydrolyzes phospholipid, resulting in inflammatory mediator production), may cause vascular inflammation, leading to coronary heart disease (21, 125–127). In NO homeostasis, RBCEVs can induce NO synthase, resulting in excessive NO production, enhanced ROS production, increased erythrocyte adhesion, and increased endothelial cell damage and dysfunction (128, 129). During thrombosis, RBCEVs have pro-coagulant activity, providing a site (i.e., phosphatidylserine) for prothrombinase assembly to accelerate the coagulant cascade from prothrombin to thrombin-mediated clot formation (130). When aged or damaged RBCs enter suicidal death (eryptosis), cell shrinkage and cell membrane blebbing and scrambling lead to phosphatidylserine (“eat me” marker) exposure on the outer cell surface and then induce macrophage engulfment, thereby stimulating foam cell formation (24, 131, 132). Additionally, cholesterol on the RBC membrane can trigger foam cell formation (133). However, these aforementioned causes of vascular damage may occur when blood vessels contain high numbers of RBCEVs (134).

Drug-loaded RBCEVs are designed to significantly reduce side effects on normal cells (135). For example, RBCEVs containing miR-125b anti-sense oligonucleotides effectively antagonized oncomiRs and suppressed tumorigenesis without any observable side effects in breast cancer (15). In addition, RBCEVs induce pro-coagulant activity in vitro, but the effect on thrombotic complications after blood transfusion is unknown (136).

In prior research, fluorescently labeled EVs could be up taken and accumulated by every cell type (137). However, EVs contain parts of the plasma membrane as their parental cells, including surface ligands and receptors. This fact highlights that EVs have specific interactions with target cells through several mechanisms such as direct fusion with the plasma membrane, endocytosis, binding to cell surface receptors and docking at the cell surface (111, 138, 139). For these mechanisms, interactions between the surface proteins of EVs and those of recipient cells, such as that between syncytin and its receptor major facilitator superfamily domain 2a, are required (140, 141).

RBCEVs bound to the target amino acid sequence can deliver drugs to specific cancer cells (135). Moreover, Plasmodium falciparum-infected RBC-derived RBCEVs loaded with drugs produced better therapeutic efficacy against malaria in vitro than normal RBCEVs loaded with drugs and free drugs (142). The specific interaction between EVs and cells depends on the origin of the EVs and the target cells under active processes (115).

The factors influencing RBCEV release include (i) RBC storage conditions (i.e., several weeks at 4°C in additive solutions), (ii) donor variability, and (iii) the leukoreduction method (89). First, during RBC period, ATP concentrations inside RBCs decrease, resulting in membrane skeleton destabilization and intracellular calcium increase and leading to vesiculation (143). Moreover, the loss of endogenous anti-oxidants during RBC storage causes a number of proteins to undergo oxidative degradation such as spectrin, beta-actin, glyceraldehyde-3-phosphate dehydrogenase, and band 4.1, leading to vesiculation processes (144, 145). Furthermore, oxidative modification has been observed in the hemoglobin-beta chain, which affects the function of hemoglobin (146). Similarly, storage at 4°C can inhibit the ATP-dependent activity of Na⁺/K⁺ cationic pumps, resulting in increased Na⁺ and Ca²⁺ concentrations inside cells and subsequently increased RBC vesiculation (147). Conversely, the size of RBCEVs changes during storage from 100 nm after 5 days up to 200 nm after 42 days (148). Second, donor-specific factors depend on the hematological profile, which affects the basal number of RBCEVs and level of hemolysis during packed RBC preparation. Third, the leukoreduction method affects the size and number of RBCEVs, as RBCEVs obtained via whole-blood filtration had a smaller diameter (<200 nm) and higher total count than those prepared using the buffy coat method (149).

The RBC half-life is 58 ± 1.5 days (150), whereas the clearance half-time of RBCEVs in peripheral circulation after injection using ¹²⁵I-tagged RBCEVs is 44 min (83). In addition, RBCEVs remain stable and intact even after multiple freeze–thaw cycles, and they have long-term stability at −80°C without effects on the moiety, uptake, and genetic material loading capacity (135).