Extracellular vesicles: a rising star for therapeutics and drug delivery

Extracellular vesicles (EVs) are nano-sized, natural, cell-derived vesicles that contain the same nucleic acids, proteins, and lipids as their source cells. Thus, they can serve as natural carriers for therapeutic agents and drugs, and have many advantages over conventional nanocarriers, including their low immunogenicity, good biocompatibility, natural blood – brain barrier penetration, and capacity for gene delivery. This review first introduces the classification of EVs and then discusses several currently popular methods for isolating and purifying EVs, EVs-mediated drug delivery, and the functionalization of EVs as carriers. Thereby, it provides new avenues for the development of EVs-based therapeutic strategies in different fields of medicine. Finally, it highlights some challenges and future perspectives with regard to the clinical application of EVs. Graphical Abstract

EVs is a term used to describe a population of heterogeneous vesicles ranging from 40 to 1000 nm in size, encapsulated by lipid bilayers, and released by various cell types [ 19 ]. In fact, there are several subtypes of EVs, and this classification is based on their size, biogenesis, and the expression or absence of specific proteins [ 20 ]. EVs can be broadly classified into three groups: exosomes, microvesicles, and apoptotic vesicles. The process of their formation is illustrated in Fig. 1 . With progress in research on EVs, other types of EVs, such as ectosomes, microparticles, and oncosomes have also been identified [ 21 ]. Since the contents of different EVs are highly heterogeneous depending on the recipient or source cells and the biogenesis of their subtypes is difficult to elucidate [ 22 ], we focused on the three broader categories mentioned above (exosomes, microvesicles, and apoptotic vesicles). The differences between the three main types of EVs are summarized in Table 1 . Table 1 Main types of extracellular vesicles and the differences among them Type and size (nm) Origin Biogenesis Density (g/m) Appearance Major pathway Biomarkers Contents Refs. Exosome 50–150 Endosomal membrane The fusion of multivesicular bodies and plasma membranes 1.13–1.18 Cup-shaped ESCRT-dependent Tetraspanins (CD9, CD81, CD63, CD82) Alix TSG101 HSP70 ESCRT Nucleic acids (mRNA, miRNA, Pre-miRNA, snRNA, mtDNA, dsDNA, etc.) Lipids Proteins (cytoskeletal, heat shock, nuclear enzyme, etc.) Amino acids Metabolites MHC [ 28 , 40 , 232 , 235 ] Microvesicles 100–1000 Plasma membrane Shedding from the plasma membrane 1.04–1.07 Cup-shaped Ca 2+ -dependent CD40 Integrin Selectin Nucleic acids (miRNAs, etc.) Lipids Membrane protein enzymes Growth factor- receptors, Cytokines Chemokines [ 33 , 34 , 39 , 40 , 232 , 236 ] Apoptotic bodies 1000–5000 Plasma membrane, Endoplasmic reticulum Direct outward budding of the cell membrane in dying cells 1.16–1.28 Heterogeneous Apoptosis-related pathway Histone proteins, Annexin V Thrombospondin C3b Nucleic acids (mRNA, miRNA, fragments of DNA), Lipids Cell organelles [ 33 , 34 , 40 , 232 , 236 ] Notably, diverse culture conditions, methods of isolation, and purification protocols may result in the formation of different subpopulations. Further, the overlap between vesicle size and the lack of specific biogenetic markers for the identification of EVs subtypes have led to conflicting definitions in the literature [ 23 ]. Therefore, the International Society for Extracellular Vesicles (ISEV) recommends the use of “extracellular vesicles” as the universal fate method for defining the cellular release of vesicles [ 22 ]. Fig. 1 Formation and release of exosomes, microvesicles, and apoptotic vesicles Exosomes Exosomes are relatively small extracellular vesicles, ranging in size from approximately 50–150 nm in diameter [ 24 ]. Their biogenesis occurs via the endocytic endosomal pathway, in which the cytoplasmic membrane buds inward, leading to the capture of membrane molecules and the formation of early endosomes within cells [ 19 , 25 ]. During subsequent maturation, early endosomes fuse to form late endosomes, resulting in the invagination of the endosomal membrane into the lumen to form intracellular vesicles (ILVs), which in turn form multivesicular bodies (MVBs) [ 26 , 27 ]. MVBs may fuse with lysosomes, which results in their degradation, or fuse with the plasma membrane and subsequently release exosomes into the extracellular compartment as cytosolic vesicles [ 22 , 28 , 29 ]. Studies have demonstrated that MVBs formation is mediated by two different pathways. The first pathway is associated with the ESCRT endosomal sorting complex (ESCRT-0, -I, -II, and -III and Vps4 complexes) [ 30 ]. ESCRT-0 degrades ubiquitinated cargo, while ESCRT-I and ESCRT-II are responsible for the formation of endosomal membrane buds. ESCRT-III surrounds the neck of the formed vesicle, and Vps4 plays a role in the rupture of the membrane and finally the formation of the MVBs luminal vesicle [ 31 , 32 ]. The second pathway does not rely on the ESCRT mechanism but is instead dependent on a lipid component of the endosomal membrane, which contains many sphingolipids. However, these sphingolipids represent substrates for neutral sphingomyelinase 2 (nSMase2). On the endosomal membrane, nSMase2 transforms sphingolipids into ceramides, subsequently inducing microdomains to merge into larger structures, leading to domain budding and ILVs formation [ 33 – 35 ]. This process involves the transport and signaling of many proteins, especially RAS-related proteins. In this process, Rab27A and Rab27B are key regulatory proteins, and Rab27A has also been found to be associated with the fusion of the MVBs with the plasma membrane [ 36 ].

Apoptotic vesicles are protrusions formed during programmed cell death via the bubbling of the apoptotic cell membrane, followed by disintegration. They have a particle size of 1000–5000 nm [ 44 – 46 ]. The membranes of apoptotic cells express markers that promote phagocytosis by macrophages or surrounding cells before rupture, thereby clearing apoptotic vesicles and regulating the immune system [ 47 ]. Phosphatidylserine (PS), for example, translocates from the inner lobe to the outer leaflet in the early stages of apoptosis and is believed to act as an “eat me” signal. In addition, PS may also serve as a tumor marker [ 48 , 49 ]. However, most studies have focused on exosomes and microvesicles, and apoptotic vesicles have rarely been examined in the context of nanomedicine, likely owing to their size heterogeneity. Meanwhile, in tumor cells, EVs can serve as signaling tools, regulating many cellular processes, including the promotion of cell growth, invasion [ 50 ], the stimulation of angiogenesis [ 51 ], and drug resistance [ 52 ]. In normal cells, they are involved in coagulation [ 53 ], regeneration [ 54 ], and immune regulation [ 55 ], among other processes. Since EVs have a complex composition with multiple physiological functions, their own intrinsic functions should be considered during their application as therapeutic vehicles [ 56 ]. EVs released into the tumor microenvironment (TME) or bodily fluids are taken up by receptor cells. The known pathways of EVs entry into cells are direct fusion with receptor cell membranes, receptor-mediated endocytosis, lipid raft interaction, reticulin interaction, and macrophage phagocytosis [ 57 ]. However, it has been reported that exosomes are mainly internalized by non-dependent lipid raft-mediated endocytosis and not via direct membrane fusion. Similar to other nanocarriers, internalized EVs can undergo endosomal escape and subsequently release the cargo into the cytosol. However, the acidic environment during endosomal escape and in the lysosomal pathway may lead to cargo degradation [ 58 ]. Overall, the mechanism of EVs uptake is complex and needs to be evaluated in greater depth.

Density gradient centrifugation (DGC) is a modified ultracentrifugation method based on the size, shape, mass, and density of EVs. In a density gradient solution, when the centrifugal force of each particle is balanced with the buoyant force, different components accumulate at different positions in the top-to-bottom gradient due to their different densities [ 17 , 68 ]. Accordingly, exosomes can be separated from other components in the sample. Generally, sucrose or iododiol is used to generate the density gradient [ 69 ]. Arab et al. used differential centrifugation and additional Optiprep ™ density gradient ultracentrifugation to extract EVs released from Daphnia primordial microglia (Fig. 2 b). They determined the protein content of the isolated EVs using mass spectrometry and found that the use of DGC eliminated contaminants and limited the effects of co-separating protein aggregates and other membrane particles present during the separation [ 70 ]. Iwai et al. used DGC to isolate EVs from human saliva and compared them to those isolated from a cell culture supernatant. They found that the volume and density of saliva EVs were 47.8 ± 12.3 nm and 1.11 g/mL, respectively, while those of cell EVs were 74 ± 23.5 nm and 1.06 g/mL, respectively. Thus, the volume of saliva EVs was lower and their density was higher [ 71 ]. DGC requires separation across different gradients, and time is required to reach equilibrium at each point in the gradient. Hence, this process is relatively time-consuming and time-sensitive, but the operation is relatively simple.

The principle of ultrafiltration (UF) is based on molecular size [ 77 ]. This process is similar to conventional filtration methods in that larger particles are retained by the filter while smaller particles pass through one or more membranes with different pore sizes or MWCOs (molecular weight cut-offs) [ 68 , 78 ]. Tangential flow filtering (TFF) is also based on this principle and is an advanced form of UF. Paterna et al. used TFF to isolate EVs from microalgae, using cutoff values of 650, 200, and 20 nm (Fig. 2 d), To improve sample purity and yield, they evaluated different technical parameters and conditions to improve EVs separation. The optimized TFF-based bioprocess was found to be suitable for large-scale production [ 79 ]. Busatto et al. compared UC and TFF with regard to EVs isolation from MDA-MB-231 breast cancer cell cultures. They found that TFF was superior in terms of yield and the removal of individual macromolecules and aggregates [ 80 ]. He et al. developed a method to optimize UF by introducing a 0.22µm filter and a dialysis membrane with an MWCO of 10,000 kDa to remove extracellular microbubbles larger than 200 nm [ 81 ]. Parimon et al. reported the isolation of EVs from bronchoalveolar lavage fluid using centrifugal UF with 100 kDa MWCO nanomembrane filtration unit and found that this process provided a smaller and more homogeneous distribution of EVs than UC and DGC [ 82 ]. Cardoso et al. reported that UF and SEC increased the ability to recover small EVs (sEVs) per ml of media by approximately 400 times compared to UC [ 83 ]. However, when UF is used, the membrane is prone to clogging as well as shear forces, which can destroy EVs during filtration. Nevertheless, the operation is simpler.

The polymer precipitation method involves mixing the relevant biofluid with a polymer solution, followed by centrifugation at a low speed to promote exosome precipitation [ 60 , 89 ] and obtain exosomes (Fig. 2 f) [ 60 ]. Ludwig et al. added 50% polyethylene glycol (PEG) of different molecular weights to achieve final concentrations of 6, 8, 10, 12, and 15% PEG and 75 mM NaCl. After optimizing the method, they finally chose to add PEG 6000 and NaCl to achieve a concentration of 10% PEG and 75 mM NaCl. After incubation for 8 h, a significant amount of bovine serum proteins could be removed. However, studies show that EVs samples prepared with PEG may still retain a certain percentage of non-EV-related molecules [ 90 ]. Juan A et al. reported that the precipitation of exosomes from the cell culture medium using PEG prior to the SEC step can improve the resolution of conventional SEC methods because PEG precipitates soluble proteins. Moreover, the combination of polymer-based precipitation with SEC (Pre-SEC) methods can help in separating individual cell types secreting EVs isoforms. Today, commercial kits based on polymer co-precipitation are available for EVs isolation [ 91 ]. Jenni et al. used the miRCURY™ Exosome Isolation Kit to obtain abundant EVs-specific miRNAs from plasma. They found that the precipitation-based method was not sufficient to purify the EVs-containing miRNA cargo from plasma. Although a portion of vesicle-free miRNAs could be removed, vesicle-free miRNAs remained predominant in plasma EVs precipitates isolated by this method [ 92 ]. Romero et al. compared the performance of UC, the PEG method, and two commercial kits (Exoquick ® and PureExo ® ) in the isolation of gDNA-EVs from healthy donor blood. They found that the PEG method could increase gDNA yields and reduce cost and time [ 93 ]. In addition to polymer-based precipitation, charge-based precipitation can be used for EVs separation. Deregibus et al. used positively charged fish sperm proteins to induce EVs precipitation in the serum, saliva, and cell supernatants. In their study, EVs resuspension was facilitated when fish sperm proteins were precipitated using 35,000 Da PEG, and the recovery of precipitated EVs was more efficient than that of EVs obtained via ultracentrifugation using charge. The precipitation method avoids the need for expensive equipment and is suitable for the isolation of EVs from small biological samples [ 94 ]. Tan et al. proposed the use of ammonium sulfate for salting to isolate EVs from skim milk, achieving purity and yield comparable to those of UC, while EVs isolated using the ExoQuick kit were of lower purity. And they also verified the relevant function of the EVs as therapeutic carriers [ 95 ]. In the precipitation method, contamination of the polymer may reduce purity. Although the kits are expensive, they are easy to handle and readily available.

EVs isolation based on chemical affinity has been extensively studied. Affinity-based methods can be divided into two broad types: targeted EVs capture and non-targeted EVs capture. For the targeted capture of EVs, various biomolecules on the surface of EVs are targeted using a combination of high-affinity antibodies, aptamers, and peptides. Meanwhile, non-targeted capture uses lipid probes, phosphatidylserine (PS), and TiO 2 to extract EVs based on the high affinity of lipids on the EVs surface [ 17 ]. The use of conventional spherical immunomagnetic beads can result in a lower concentration and recovery efficiency due to the smooth surface and rigid interfacial modifications of the beads. Sun et al. proposed a new technique using lipid labeling and magnetic beads by first inserting a lipid motif DSPE-PEG1000-TCO, which labels EVs in plasma, and then using bioorthogonal click chemistry to immobilize TCO-labeled EVs onto tetrazine (TZ)-grafted microspheres (clicklets). The EVs on the clicklets were then separated by centrifugation, and finally, the mRNAs in the EVs were analyzed using reverse-transcription digital PCR (RT-dPCR) (Fig. 2 h). Unlike immunoaffinity-based EVs labeling, which is limited by the number of specific antigens (CD63, CD81, CD9) on the EVs surface, lipid-based EVs markers are independent of EVs surface antigens. The simultaneous combination of these markers with RT-dPCR allowed for the quantification of oncogenic changes in Ewing sarcoma and pancreatic cancer, demonstrating its potential clinical value in monitoring treatment responses and disease progression [ 100 ]. Cheng et al. developed immunomagnetic hedgehog particles (IMHPs) to capture and release exosomes. These particles were used to capture exosomes from MCF-7 cells, and the capture rate was as high as 91.7%. In contrast to exosomes obtained via UC, the exosomes obtained through this method maintained their structural integrity and showed good biological activity [ 101 ]. Yang et al. exploited the phosphatidylserine-rich surface of exosomes and immobilized peptide ligands on SiO 2 microspheres to induce specific interactions and isolate exosomes, and isolated exosomes from serum using this method [ 102 ]. Brambilla et al. proposed the use of a DNA-directed immobilization (DDI) strategy for the isolation of EVs via the immobilization of anti-CD63 antibodies bound to vesicles on magnetic particles. That is, the surface and antibody were functionalized using complementary oligonucleotides for the release of EVs via the DNAse I-catalyzed enzymatic cleavage of the double-stranded DNA linker. Conventional methods are based on antigen–antibody destruction and alterations to their physical properties by heating or ultrasound, and treatment with organic solvents, bases, and other chemical methods can damage EVs. However, the proposed reversible DNA link could allow the release of EVs following enzymatic cleavage, overcoming this problem while enabling enhanced affinity for anti-CD63 antibodies [ 103 ]. Thuy et al. prepared a chimeric nanocomposite consisting of lactoferrin-coupled dendrimer-modified magnetic nanoparticles based on a combination of electrostatic interaction with the EVs surface, physical adsorption, and biological recognition to isolate EVs without the need for centrifugation and antibody affinity. The EVs could be separated from cell cultures and clinical specimens within 30 min, but could not be distinguished from other types of EVs [ 104 ].

As drug delivery carriers, EVs can be loaded with therapeutic drugs, including nucleic acids and chemotherapeutic drugs [ 108 ]. They are a promising delivery agent for carrying exogenous drugs due to their low immunogenicity and good biocompatibility. The methods of drug loading can be roughly classified as drug loading before isolation (pre-loading) and drug loading after isolation (post-loading). The characterization, loading rates, and functions of EVs loaded using different methods, such as incubation, ultrasound, and transfection, are summarized in Table 3 . Table 3 Loading methods for extracellular vesicles EVs loading Source Loading methods Size (nm) Zeta potential(mV) Cargos Loading efficiency Functions Refs. Pre-loading ADMSCs Incubation sEV-CUR: 74.05 ± 2.52 N/A Curcumin 82.26 ± 5.25% Excellent anti-oxidative and anti-apoptotic capacity; Favorable bioavailability; Controlled release [ 110 ] HEK293T Incubation EVs (ICG/PTX): 149.9 ± 5.2 -30.2 ICG/PTX ICG: 60.7%; PTX: 51.9% Simultaneous therapy and high accumulation at the tumor site; High encapsulation efficiency and cellular uptake; Photo-stability and storage stability [ 111 ] HL-60; dHL60; MCF-7; THP-1 Infection HL-60: 170.5 ± 49.4; dHL-60: 246.8 ± 19.5 N/A Penicillin/ PTX /MCP-1/MiR-16/Cas-9-GFP/Cas9 N/A High encapsulation efficiency and production efficiency; Low immunogenicity [ 112 ] MSCs Infection and incubation 60–150 N/A CTX/TRA-IL 15.43 ± 0.44% Synergistic effects and few side effects [ 113 ] ADSCs Infection 30–150 N/A NT-3 siRNA N/A Stable and functional delivery [ 114 ] HUVECs USMB N/A N/A CTG/BSA-FITC N/A High encapsulation efficiency and improved EVs production [ 115 ] Post-loading RAW 264.7 Incubation 100–200 N/A HA/CV/D-OX N/A Polarization to M1 macrophages; High cellular uptake; Excellent antitumor effect [ 117 ] THP-1 Incubation A15-Exo: 94.1 ± 104.4 A15-Exo: − 9.68 ± 0.29 DOX NA Targeting ability; high yield; efficient release [ 118 ] A549 Incubation DOX/LND-16k: 93.2 ± 24.2; DOX/LND-120k: 70 ± 11.1 DOX/LND-16k: -15.2; DOX/LND-120k: -15.9 DOX/LN-D DOX/LND-16k: 4.16 ± 1.9%; DOX/LND-120k: 2.77 ± 0.35% Excellent anticancer effect (DNA damage, ATP inhibition, and ROS generation) [ 119 ] Macrophage Sonication 115.0 ± 8.3 N/A TPP1 N/A High loading capacity; Sustained release; Bio-inspired; Non-viral and favorable stability [ 243 ] DOX: 162.2 ± 1.6; PTX: 129.4 ± 2.3 N/A DOX/PT-X N/A Superior intracellular accumulation and drug accumulation in cancer cells; Low immunogenicity and favorable stability [ 120 ] hMSCs Electroporation ~ 210 ~ -10 GPX4 siRNA 16.6% Magnetic targeting; BBB penetration ability; Synergistic ferroptosis therapy; Good biocompatibility and safety [ 123 ] 4T1 Extrusion MSNs: 125 ± 15; E-MSNs: 150 ± 11 MSNs: 20.5 ± 1.2; ID@MS-Ns: -5.8 ± 1.5; ID@E-MSNs: -28.9 ± 3 ICG/DOX N/A High cellular uptake and long-term retention; Synergistic chemo-photothermal therapy; High purity; Favorable biocompatibility [ 125 ] MSCs Extrusion 135.9–194.9 -7.23 Curcumin 75.53% Relieves neuroinflammation [ 126 ] 4T1 Extrusion 173 N/A DOX 7.4% Prominent biocompatibility; Synergistic photothermal properties; Controlled release drug; Good stability [ 127 ] U87; U251 Saponin U87-sEVs: 76.71 ± 21.7; U251-sEVs: 79.11 ± 28.77 ~ -12.5 DOX N/A Excellent anti-oxidative and anti-apoptosis ability; Favorable bioavailability; Controlled release; Outstanding stability [ 129 ] U251-GMs; SF7761- GMs Microfluidics U251GMs: 150 SF7761 GMs: 100 N/A DOX U251GMs: 31.98%; SF7761 GMs: 19.7% Homing effect; Simple; Efficient setup; Adjustable condition [ 132 ] Human plasma Acoustofluidics N/A N/A DOX ~ 30% High loading efficiency; One-step process; Rapid encapsulation [ 133 ] Pre-loading In a nutshell, the “Pre-loading” method involves cargo loading before EVs isolation and typically has useful therapeutic effects [ 109 ]. When research on EVs first started, the drug-loading efficiency and encapsulation methods of EVs attracted extensive attention. During cell growth, cells continue to communicate with each other via EVs. During this phase, drugs can be taken up and secreted by the cell using these vesicles. As a result, the drug is loaded into the vesicles. The most common pre-loading methods are incubation, infection, and ultrasound combined with microbubbles (USMB). Among the various strategies for drug loading via incubation, direct incubation is the simplest method. Xu et al. fabricated a new drug delivery system called sEV-CUR. In this system, curcumin (CUR) was incubated with adipose-derived mesenchymal stem cells (ADMSCs), and sEV-CUR particles were harvested using UC (Fig. 3 a). The average diameter and zeta potential of the EVs remained largely consistent following incubation, and the loading efficiency was 82.26% ± 5.25%. Notably, sEV-CUR showed excellent therapeutic function (anti-oxidative stress and anti-apoptosis ability), favorable bioavailability, controlled release, and improved stability [ 110 ]. EVs are well-known to be both hydrophobic and hydrop

EVs are biocompatible and have an ideal natural structure and hydrophilic core. Hence, they are increasingly being used as drug carriers or therapeutic agents and are expected to serve as valuable nanocarriers for clinical use. Although their surface expresses intact transmembrane proteins (CD81 and CD91) and integrins (CD51 and CD61) with homing and targeting functions, their targeting ability remains weak. EVs show heavy accumulation in the liver and spleen and are subsequently cleared by macrophages. Therefore, their modification can facilitate the delivery of cargo to target cells [ 134 – 137 ]. The current surface functionalization strategies can be divided into two categories: endogenous and exogenous. The methods used for modifying EVs, related strategies, sources of EVs, and functions after modification are summarized in Table 4 . Table 4 Methods and strategies for the modifications of EV surfaces Surface modification Strategy EVs source Functions Refs. Genetic engineering Infection with PGMLV-PA6 virus expressing both the CXCR4 protein and GFP MSCs Allowed more MSC-derived exosomes to nest around the target region [ 139 ] Transfection via the pCDH-GFP vector HuCMSCs Decreased ATP concentration; increased adenosine levels; and reduced spinal cord inflammation [ 140 ] Transfection via the recombinant adenoviral vector GFP-CTF1 encoding CTF1 BMSCs Increased proangiogenic activity and the rates of successful pregnancy outcomes [ 141 ] Transfection via a pCDNA-MEG3 vector OS cells Inhibition of osteosarcoma growth [ 143 ] Introduction of pcDNA3.1(-)-RGD-Lamp2b into cells by electroporation HEK293T Enhanced tumor site targeting [ 145 ] Transfection via a miR-31-5p lentiviral vector HEK293 Promoted the healing of diabetic wounds [ 144 ] Co-transfection of a reporter plasmid and miR-181b mimics using Lipofectamine Human umbilical cord mesenchymal stem cells (HuCMSCs) Enhanced M2 polarization; inhibited inflammation; and promoted osseointegration [ 244 ] Transfection via an XStamp-PDGFA lentiviral vector Neural stem cells (NSCs) Improved potential for CNS injury targeting [ 245 ] Transfection via an LV-iRGD-Lamp2b lentiviral vector Human cord blood MSCs (cbMSCs) Enhanced targeting to tumor sites [ 246 ] Transfection via a Lenti-XStamp-PDGFA lentiviral vector Neural stem cells (NSCs) Enhanced targeting efficiency for central nervous system lesions [ 247 ] Transfection via the pRBP-Lamp2b-HA-hygro vector using Lipofectamine 2000 HEK293 cells Anti-inflammatory effects [ 248 ] Transfection via the iRGDC1-EGFP-Lamp2b virus HEK293T Enhanced tumor targeting [ 249 ] Bioorthogonal chemistry The azide groups were bioorthogonally labeled with DBCO-Cy5 via bioorthogonal click chemistry A549 Exosome tracking and imaging [ 148 ] DBCO reacted with azide or azide-containing methionine analogs via bioorthogonal click chemistry B16F10 Regulation of exosome composition and binding of exosomes for intracellular delivery [ 147 ] DBCO-Exo was linked to a c(RDGyK) peptide with an azide moiety via copper-free click chemistry MSCs Improved targeting of lesion sites [ 149 ] Copper-free click chemistry was used with AlexaFlour®488 (AF488)-azide PANC-1 B16F10 HEK293 Achieved quantification of intracellular tracking and intracellular uptake [ 150 ] Physical modification Membrane extrusion method HCC Enhanced targeting ability and improved siRNA transfection efficiency [ 153 ] Lipid membrane fusion Sf9 insect cells Enhanced targeting capabilities [ 154 ] Membrane extrusion method SKOV3-CDDP Enhanced targeting capabilities [ 155 ] Membrane extrusion method L-929 Depleted cells and homing effects [ 156 ] Membrane fusion technology using the freeze–thaw method CT26 Allowed immune evasion, enhanced targeting ability, and acted as a drug carrier [ 157 ] PEG-mediated membrane fusion HUVECs Widely used in studies on the mechanism of membrane fusion [ 158 ] Membrane extrusion method J774A.1 Enhanced targeting ability and acted as a drug carrier [ 159 ] Incubation-mediated membrane fusion HEK293FT Enabled efficient wrapping of clustered regularly interspaced short palindromic repeats (CRISPR)/CRISPR-associated protein 9 (Cas9) in exosomes [ 250 ] Incubation HEK293T Improved targeting capabilities [ 251 ] Electrostatic interaction MSCs Targeted hepatocyte asialoglycoprotein receptors [ 160 ] Extrusion method Bone marrow MSCs Improved targeting ability and promoted angiogenesis [ 161 ] Chemical modification Bioconjugate chemistry Human leukemia monocytic cell line (THP-1) Promoted blood–brain barrier penetration and improved targeting [ 164 ] Copper-free click chemistry M2-BV2 Provided rapid and effective recruitment and differentiation transformation of neural stem cells [ 165 ] Cycloaddition reaction of sulfonyl azide U-87 MG Improved targeting [ 166 ] Hydrophobic insertion method ADMSCs Improved targeting [ 167 ] Lipid insertion method BMSCs Improved targeting [ 168 ] PDA self-polymerization and thiol-Michael addition reactions L929 Facilitated

Exogenous modifications refer to the direct modifications of the EVs membrane and can be performed using physical or chemical methods. Physical modification Physical methods involve the use of physical forces and processes such as ultrasound, incubation, extrusion, and freeze-thaw methods to temporarily disrupt the lipid structure of the vesicles. With these methods, the vesicles self-assemble into their original structure when the forces disappear [ 152 ]. Further, the vesicles can also be modified via electrostatic interactions or weak forces between EVs and functional molecules (e.g., fusion of lipid membranes) [ 134 ]. Zhou et al. isolated tumor-derived EVs (TDEVs) from hepatocellular carcinoma (HCC) cells and subsequently hybridized them with lipid nanovesicles to obtain innovative nanovesicles (LEVs), which were generated by the fusion of TDEV membranes with phospholipids. LEVs showed better targeting abilities due to the “homing” effect and a 1.7-fold higher siRNA transfection rate than liposomes [ 153 ]. Raga et al. exploited the presence of Gp64 in the viral particles of Sf9 insect cells and their membrane fusion under acidic conditions. Sf9 insect cell EVs, which also express the functional membrane protein PD-1 on their surface and can be actively targeted by Cx43, were isolated. The fusion of Gp64-expressing PD-1 EVs with small molecule liposomes contributed to the generation of heterozygous EVs with further biomedical applications [ 154 ]. Li et al. designed a novel pH-sensitive bionanoplasmic nanosystem comprising SKOV3-CDDP-derived exosomes hybridized with cRGD peptide-modified liposomes via the membrane fusion technique. Exosomes were vortexed with liposomes after vortex sonication in a vacuum vortex and finally extruded using a 200 nm polycarbonate membrane filter. A dual-targeting effect including homologous targeting and cRGD could be achieved [ 155 ]. Sun et al. proposed the preparation of EL-CLD hybrids by the extrusion of fibroblast-derived exosomes (E) with liposomes (L) loaded with clodronate (CLD) (Fig. 4 e), and the EL-CLD obtained had a good drug release potential [ 156 ]. Cheng et al. used the freeze-thaw method to mix temperature-sensitive liposomes with genetically engineered exosomes. The mixture was frozen at −80 °C for 15 min and rewarmed at 37 °C for 15 min. After three cycles, lipid membrane-fused nanovesicles (hGLV) were obtained [ 157 ]. Max et al. developed a method for fusing EVs with functionalized liposomes in the presence of PEG and retaining the natural properties of EVs. The drug delivery efficiency of the heterogeneous EVs was 3–4-fold greater than that observed with free drug and liposome delivery [ 158 ]. Sagar et al. crossed mouse macrophage-derived J774A.1 sEVs with liposomes to obtain engineered hybrid exosomes (HEs) that retain the advantages of endogenous sEVs, which target tumor sites, and liposomes, which exhibit significant flexibility for surface modification as potential drug delivery carriers. The disadvantages of both types of particles could be overcome, and their advantages could be combined, to generate an effective hybrid drug delivery tool [ 159 ]. Tamura et al. used cationic pullulan polysaccharides to modify exosomes via electrostatic interactions [ 160 ]. Hu et al. hybridized platelet membranes with stem cell-derived exosomes to enhance binding and accumulation in damaged tissues [ 161 ]. Li et al. prepared platelet-like membranes via the fusion of platelet membranes with bone marrow MSC-derived EVs using extrusion for improving the ability to target the lesion site (Fig. 4 f) [ 162 ]. Modification of the EVs surface by physical means avoids the introduction of foreign impurities and improves the purity of EVs while causing less damage to them. However, some complex auxiliary equipment is often required during the preparation process. Moreover, the extrusion method may lead to variations in the dimensions of EVs.

The application of EVs as drugs or delivery vehicles for the treatment of cancer, neurodegenerative diseases, and regeneration is prompted by their unique biocompatibility, low immunogenicity, and stability [ 46 ]. EVs have fewer limitations with respect to the safety and feasibility of cell transplantation than cell-based therapies in the context of regenerative medicine [ 172 ]. The application of EVs as therapeutic agents or carriers in different areas of medicine is summarized in Table 5 . Table 5 Strategies and doses of EVs for different diseases Treatment strategies Diseases Drugs Dose of EVs in vivo Administration Refs. Immunotherapy Breast cancer N/A 200 µg per mouse Intravenous injection [ 176 ] Immunotherapy Gastric cancer N/A 5 mg/kg (100 µL) Intravenous or intratumoral injection [ 177 ] Immunotherapy Liver cancer siRNA N/A Intratumoral injection [ 178 ] Immunotherapy Pancreatic cancer Galectin-9; siRNA ~ 10 8 exosomes Intravenous injection [ 179 ] Chemotherapy Glioma TMZ; DHT N/A Intravenous injection [ 181 ] Chemotherapy Triple-negative breast cancer DOX N/A Intravenous injection [ 258 ] Chemotherapy Glioma DOX 0.15 mg/mL (400 µL) Intravenous injection [ 182 ] Phototherapy Liver cancer; Breast cancer N/A 50 mg/mL (100 µL) Intravenous injection [ 184 ] Phototherapy Gastric cancer Proton pump inhibitor (PPI) N/A Intravenous injection [ 186 ] Gene Therapy Triple-negative breast cancer Anti-miRNA-21; anti-miRNA-10b 3 ~ 4 × 10 11 NPs (150 µL ) Intravenous injection [ 188 ] Combination therapy Glioblastoma AQ4N 200 µg (100 µL) Intravenous injection [ 191 ] Other Lung cancer Dinaciclib 200 µg Intravenous injection [ 173 ] Other Nasopharyngeal carcinoma N/A NPC tumor-bearing mice: 30 µg/mouse C666-1/NPC43 tumor-bearing mice: 20 µg/mouse Intravenous or intraperitoneal injection [ 194 ] Other Glioblastoma Verrucarin A (Ver-A) N/A Intravenous injection [ 195 ] Gene therapy Parkinson’s Disease (PD) Glial-cell-line-derived neurotrophic factor (GDNF) 3 × 10 9 particles/10 µL/mouse Intranasal injection [ 200 ] Gene therapy PD Anti-TNF-α antisense oligonucleotide (ASO) 2.7 × 10 9 particles Intravenous injection [ 202 ] Gene therapy PD shRNA Minicircles 150 µg (100 µL) Intravenous injection [ 204 ] Other PD N/A N/A Intravenous injection [ 201 ] Gene therapy Bone regeneration Bone morphogenetic protein-2 ( BMP2 ) gene 50 µg (1.8 µg/µL) In situ injection into the bone defect [ 212 ] Other Periodontal disease Minocycline 230 µg sEV Injection into the site of defect [ 209 ] Other Hair loss N/A 200 µg Milk-exo in 100 µL saline Intradermal injection [ 210 ] Other Wound healing N/A 100 µg (100 µL) Subcutaneous injection [ 211 ] Other Bone regeneration N/A 50 µg/per cranial defect Implantation into the defect [ 213 ] Other Shoulder stiffness (SS) N/A 50 µL EV (20 µg/mL or 50 µg/ml, equal to final particle from 1.2 × 10 8 to 3 × 10 8 ) Intra-articular injection [ 216 ] Other Inflammation Piceatannol N/A Intravenous injection [ 217 ] Other Rheumatoid arthritis (RA) Dexamethasone sodium phosphate N/A Intravenous injection [ 218 ] Cancer therapy Cancer is associated with high morbidity and mortality, and traditional treatments such as surgery and radiotherapy have certain limitations (e.g., easy recurrence and severe side effects). However, some of these problems can be addressed using nano-drug delivery systems. EVs are widely used in cancer treatment because of their ability to penetrate tissues, tumor tropism, and low immunogenicity [ 173 ]. Further, as with cell membrane coating technologies, the CD47 expressed on the membrane can aid in immune evasion, preventing the engulfment of nanoparticles [ 174 ]. Immunotherapy Immunotherapy has emerged as a powerful clinical strategy for the treatment of cancer. However, the key challenge in the widespread implementation of immunotherapies against cancer remains the controlled regulation of the immune system, as these therapies have serious side effects, including autoimmunity and non-specific inflammation. Thus advanced biomaterials and drug delivery systems [ 175 ], such as the use of macrophage exosomes could effectively harness immunotherapy and increase its efficacy while reducing toxic side effects. Macrophages are usually found in two phenotypes: classically activated M1 cells and alternatively activated M2 cells. The M1 phenotype of macrophages exerts antitumor effects, while the M2 phenotype exerts pro-tumor effects. Nie et al. coupled anti-CD47 with anti-Sirpα antibodies (aCD47 and aSirpα) on M1 macrophage-derived exosomes via pH-sensitive linkers. aCD47 recognizes tumor cells by identifying the CD47 on their surface and thereby actively targets the tumor. As a result, the “don’t eat me” signal disappears, leading to the enhanced phagocytosis of macrophages. Meanwhile, M1 exosomes can also transform pro-tumor M2 macrophages into antitumor M1 macrophages, while the M1-Exo synergistic antibodies exert antitumor effects [ 176 ]. Neutrophils are innate immune cells that d

Phototherapy, which includes photothermal therapy (PTT) and photodynamic therapy (PDT), is a promising non-invasive strategy for cancer treatment [ 183 ]. In PTT, heat treatment is used to damage the microvasculature at the tumor site. Platelets target the damaged blood vessels as well as the tumors owing to their high surface expression of P-selectin, which recognizes CD44 receptors on the surface of cancer cells. Hence, Ma et al. combined exosomes derived from platelets with photothermal-sensitive liposomes and added glucose oxidase (GOx) and ferric ammonium (FAC) to create a laser-controlled nanoplatform called FG@PEL. In this system, GOx and FAC could enhance the Fenton response at the tumor site and damage tumor cells, while the photothermal effect resulted in both vascular damage to achieve cascade targeting effects and accelerated -OH production by increasing GOx activity through warming. In an in vivo experiment, mice were inoculated with mouse hepatocellular carcinoma cells (H22) on one side of their body, following which five treatments were administered (Fig. 7 a). The most effective tumor growth inhibition was observed in the FG@PEL group (Fig. 7 b–c). Subsequently, GPX-4, an indicator of ferroptotic death in tumor tissue, was detected. The reduced expression of GPX-4 in the FG@PEL + Laser group indicated that this combination induced ferroptotic death following light irradiation (Fig. 7 d). It also significantly enhanced the release of immunogenic cell death (ICD)-related molecules, thus initiating an immune response to inhibit tumor development (Fig. 7 e–f). Subsequently, a lung metastasis model was established. Fewer lung nodules were observed in the FG@PEL + Laser group (Fig. 7 g). Finally, a bilateral tumor model was established, and treatment at the primary tumor site led to reduced tumor volume and a prolonged lifespan in the FG@PEL + Laser with anti-PD-1 treatment group [ 184 ]. Liu et al. introduced black phosphorus quantum dots (BPQDs) into exosomes (EXO) via electroporation to construct BPQDs@EXO nanospheres. The small BPQDs had a high photothermal conversion efficiency and good biocompatibility, and the EXO membrane could protect BPQDs from external oxygen and water, preventing their degradation in physiological fluids. Meanwhile, the homing effect of the EXO membrane allowed BPQDs@EXO to effectively accumulate within tumors. In in vivo experiments, BPQDs@EXO inhibited the growth of distant tumors in mouse models of bladder cancer more obviously after PTT, and no recurrence was detected [ 185 ]. Zhu et al. combined PDT with glutamine metabolic therapy and found that this multimodal therapy was successful in strongly inhibiting tumor growth [ 186 ]. Du et al. used CD47-functionalized exosomes loaded with a ferroptosis inducer (Erastin, Er) and a photosensitizer (Rose Bengal, RB) and performed PDT using 532 nm laser irradiation. They found that the Er/Rb@ExoCD47 (laser) group exhibited the best tumor suppression, highest levels of total ROS and lipid peroxidation, and low toxicity. Thus, they could serve as a useful strategy for the treatment of malignancies [ 187 ]. Phototherapy is a promising tool for cancer treatment, but some photothermal agents/materials and photosensitizers are prone to degradation and instability (e.g., BP). Hence, coating with EVs could help overcome these defects and achieve better therapeutic effects. Fig. 7 FG@PEL-based PTT for inhibiting tumor progression and lung metastasis in a 4T1 primary tumor mouse model in vivo. a Schematic diagram of FG@PEL-based PTT for inhibiting tumor progression in an H22 primary model. b–c Tumor volumes H22 tumor-bearing mice after different treatments. d Immunofluorescence assay examining the expression of GPX-4 in tumor tissues isolated from the aforementioned mice. Scale bar: 100 μm. e–f Levels of ATP, HMGB1, calreticulin, and eif-2a. g Lungs excised from 4T1 orthotopic tumor mouse models were used to evaluate the inhibitory effect of different treatments on pulmonary metastasis. H&E staining was performed on whole lungs isolated from these mice. Scale bar: 2000 μm. H&E staining of partial lung tissue. Scale bar: 100 μm. Reprinted with permission from Ref [ 184 ]

Monotherapy is often ineffective in killing tumor cells. Hence, researchers have adopted combination therapy for antitumor treatment. Glioblastoma (GBM) is a highly aggressive brain tumor. Through an analysis of clinical specimens, an increase in M2-like macrophages was observed in GBM patients. Therefore, Wang et al. proposed the use of M1-like macrophage-derived EVs (M1EVs) to modulate the tumor microenvironment (TME), and after aggregation at the tumor site, simultaneously polarize M2 into M1 and carry the chemotherapeutic drug bansoxadone (AQ4N) for post-release at the tumor site. The surface of EVs was modified with bis(2,4,5-trichloro-6-aminophenyl)oxalates (CPPO) and chlorin e6 (Ce6) to achieve synergistic effects of immunomodulation, chemically inspired photodynamic therapy, and hypoxia-triggered chemotherapy. A low M2/M1 ratio was positively correlated with low tumor proliferation and improved survival outcomes (Fig. 9 a–c). The targeting effect of the EVs was confirmed using two-photon imaging, which revealed that M1EVs signals were abundantly present at the tumor site (Fig. 9 d). The in vitro transwell co-culture system demonstrated that M1EVs loaded with different drugs could cross the BBB, and the integrity of the cell layer was confirmed using the tight junction marker protein ZO-1. The fluorescence intensity of DiD in the lower chamber was measured, and the penetration rate was found to be time-dependent, reaching approximately 30% at 8 h. The entry of M1EVs into multicellular tumor spheroids (MCTSs) improved with time after 24 h of incubation (Fig. 9 e). Thus, macrophage-derived EVs showed great potential in the treatment of brain diseases [ 191 ]. Huang et al. designed lung-homing tumor-derived exosomes (Tex) hybridized with paclitaxel liposomes (Liposome-PTX) and incorporated PEG-gold nanorods (GNR) to achieve synergistic antitumor effects. Tex preferentially targeted the lung tissue, and GNR could treat tumors via thermal ablation after infrared (IR) irradiation, triggering both antitumor immune responses and PTX chemotherapy. Thus, this system showed potential as a clinical treatment option for advanced breast cancer recurrence and metastasis [ 192 ]. Pan et al. extracted exosomes from urine, achieving a high purity rate. They embedded synthetic nanoparticles within purified exosomes to inhibit tumor growth by blocking EGFR/AKT/NF-κB signaling via synergistic low-dose chemo-dynamic kinetics [ 193 ]. Due to the heterogeneity of tumors and the complex TME, researchers have shifted their focus from monotherapy to a combination therapy approach. The plasticity and load-bearing capacity of EVs enables the incorporation of multiple combination therapy modalities. Further, their inherent biocompatibility and targeting properties enable the delivery of multiple anti-cancer drugs or materials that can exert synergistic effects. Fig. 9 a Immunostaining of clinical histological sections obtained from normal tumor-adjacent tissues and low- (LGG: diffuse astrocytoma, n = 22) and high-grade gliomas (HGG: anaplastic astrocytoma, n = 20; glioblastoma multiforme, n = 22); M1 macrophages (iNOS), M2 macrophages (CD163), and cell proliferation (Ki67) are visualized. Quantitative analysis of the corresponding M2/M1 ratios is shown on the right side. The proliferation marker Ki67 was positively correlated with the M2/M1 ratio. Scale bars: 50 μm. b M2/M1 ratio analysis of 167 HGG and 522 LGG samples acquired from The Cancer Genome Atlas (TCGA) database. Each dot represents a single individual. c Survival curves of glioma patients from TCGA database. The OncoLnc tool was used to explore the correlations of survival with M2/M1 ratios. d In vivo time-lapse two-photon images of the diffusion of M1EVs, M0EVs, EMVs, and PEG NPs across microvascular endothelial cells of the brain at 48 h after i.v. injection (left). Tetramethyl-rhodamine isothiocyanate-Dextran was used to label blood vessels (red). M1EVs, M0EVs, and EMVs were labeled with DiO (green); PEG NPs were labeled with FITC (green); corresponding formulation distributions in tumor tissue are also shown (right). Scale bars: 50 μm. Immunofluorescence images of histological sections showing M2 (CD163, green) and M1 macrophages (iNOS, red) (left), and quantitative analysis of M2/M1 ratios (right) at 48 h after i.v. injection. Scale bars: 50 μm, (n = 3). e CLSM images and surface plots showing DiD-labeled M1EVs penetrating MCTS (top) and the corresponding fluorescence signal intensities across the spheroids (bottom). Reprinted with permission from Ref [ 191 ]

Neurodegenerative diseases are chronic disorders of the central nervous system and usually result from the intracellular accumulation of misfolded/aggregated mutant proteins. These abnormal protein aggregates impair mitochondrial function and induce oxidative stress, resulting in neuronal cell death. In turn, neuronal damage causes chronic inflammation and neurodegeneration [ 196 , 197 ]. However, the BBB limits the accumulation and entry of drug molecules into the central nervous system. This precludes an effective concentration of drug molecules from reaching the brain tissue, compromising efficacy. Therefore, in the treatment of such diseases, drug delivery is challenging. There is an urgent need to develop drug delivery systems that can traverse the BBB [ 198 ]. Current efforts to improve drug delivery across the BBB are focused on enhancing drug entry into the brain and limiting drug loss [ 199 ]. Zhao et al. first used a strategy involving the intranasal injection of EVs loaded with the neurotrophic factor GDNF into transgenic Parkin Q311(X)A mice. They then evaluated motor function over a year after treatment and observed improved mobility, reduced neuroinflammation, and increased neuronal survival, without any systemic toxicity [ 200 ]. Wang et al. proposed the concept of “independent module/cascade function”, in which nanoparticles with movement/chemotaxis abilities obtained from L-arginine acted as artificial modules that could bind to natural exosome modules (Fig. 11 a). The guanidine group in arginine can react with iNOS and ROS in the PD microenvironment to generate nitric oxide (NO), providing motility to the engineered exosomes and increasing the degradation of α-synuclein (α-syn) aggregates, thereby promoting neuronal cell growth and enabling a functional disease treatment cascade. The effects of the strategy were examined in an MPTP-induced mouse model of PD (Fig. 11 b) using the open-field test (parameters examined: walking trajectory, total walking distance, and average speed) and the pole test (parameters examined: time taken to climb atop the pole) (Fig. 11 c–d). GAP-43, an indicator of neuronal growth, was found to be significantly upregulated in the substantia nigra (SN) region of the mouse brain. Further, α-syn aggregates in the SN were also found to be significantly downregulated (Fig. 11 e). Based on these results, it appeared that the artificial module drives the natural module to cross the BBB and target damaged neuronal cells and mitochondria for the effective treatment of PD [ 201 ]. In contrast to exosomes produced by living cells, apoptotic vesicles produced by apoptotic cells are more useful. If manipulated, the apoptotic process can be controlled, and the vesicles can be loaded with drugs (such as nucleic acids) more efficiently than other EVs. Wang et al. first screened a subset of brain metastatic cells and eventually developed drug-loaded small apoptotic vesicles (Sabs) using melanoma cells. These vesicles carried anti-TNF-α antisense oligonucleotides. It was feasible to use Sabs as a novel EVs vector for drug delivery in vivo, especially for highly efficient siRNA or microRNA delivery, which is valuable for some in vivo biological applications [ 202 ]. Xue et al. proposed that MSC-derived exosomes can help in the treatment of PD by promoting ICAM1-associated angiogenesis. The presence of α-syn aggregates is a pathological feature of PD. Hence, α-syn downregulation is a potential strategy for PD treatment. While siRNA can achieve these effects, it has a short efficacy [ 203 ]. Izco et al. designed an shRNA microloop (shRNA-MCs) that can prolong the efficacy of siRNA and used RVG-exosomes as carriers for siRNA delivery to the brain, reducing the aggregation of α-syn and loss of dopaminergic neurons [ 204 ]. Hence, this system showed great potential in the treatment of neurodegenerative diseases. Kojima et al. developed exosomal transfer into cells (EXOtic) device that enables the efficient production of exosomes in engineered mammalian cells. The implantable exosome-producing cells could deliver therapeutic mRNA in vivo and reduce neuroinflammation and neurotoxicity in an in vitro model of PD [ 205 ]. Wang et al. used Cur to obtain Exo-Cur after ultracentrifugation following incubation with RAW264.7 macrophages. The addition of lymphocyte function-associated antigen 1 (LFA-1) on the surface of the exosomes enabled them to cross the BBB. The solubility and bioavailability of Cur were also enhanced. After targeted brain delivery, Cur could alleviate Alzheimer’s disease (AD) symptoms by activating AKT/GSK-3β and thereby inhibiting Tau protein phosphorylation [ 206 ]. Qi et al. also developed plasma exosomes carrying quercetin for the treatment of AD, improving both drug bioavailability and brain targeting [ 207 ]. This nanoformulation reduced cognitive dysfunction in mouse models of AD. EVs are widely used in the treatment of neurodegenerative diseases because of their inher

EVs have a wide range of therapeutic applications. In addition to the above-mentioned diseases, EVs can also be used to achieve anti-inflammatory effects and fibrosis inhibition. For example, Luo et al. found that BMSC-EVs significantly inhibit the fibrotic process and improve inflammatory phenomena and shoulder mobility by transducing let-7a-5p and inhibiting TGFBR1 [ 216 ]. Gao et al. proposed a method based on nitrogen cavitation to isolate EVs from cells, with a 16-fold higher yield and fewer subcellular organelles and genetic materials compared to naturally secreted EVs, and preserved cell membrane proteins. They found that the loading of anti-inflammatory drugs into EVs significantly attenuates lipopolysaccharide (LPS)-induced acute lung inflammation/injury and sepsis. The nitrogen cavitation method can be used to isolate EVs from any cell and produce EVs with a high yield, reproducibility, and scalability. These EVs can act as novel targeted delivery vehicles, and the increased yield could support clinical applications [ 217 ]. Yan et al. isolated exosomes from RAW 264.7 cells and loaded dexamethasone (Dex) into the exosomes using electroporation. The exosomes were surface modified with FA-PEG-Chol, which could actively target joint inflammation and reduce inflammation, improving the therapeutic effect of Dex on rheumatoid arthritis [ 218 ]. Ma et al. demonstrated that HuCMSC-EVs can act as bioactive agents to alleviate fibrosis in ligamentum flavum cells by delivering miR-146a-5p and miR-2213p, thereby inhibiting hypertrophy [ 219 ]. Han et al. obtained EVs from adipose-derived stem cells and used them in a thioacetamide-induced liver fibrosis model. They observed a significant reduction in collagen deposition and the restoration of liver function after treatment [ 220 ]. This model has also been applied to hypertension-related diseases. For example, Wang et al. reported the involvement of EVs secreted by endothelial cells damaged by hypertension via mechanical elevation in arterial wall remodeling, guiding renewed research into the treatment of diseases related to vascular wall remodeling [ 221 ]. Furthermore, EVs have also been used in the treatment of eye diseases. For example, Moisseiev et al. used saline and human MSC (hMSC)-derived exosomes to treat mice with oxygen-derived retinopathy (OIR) and found that the exosomes inhibited retinal thinning and alleviated retinal ischemia [ 222 ]. These findings indicate that good progress has been made in the application of EVs for achieving anti-inflammatory effects, fibrosis inhibition, and ocular management. By taking advantage of the natural properties of EVs, stem cell-derived EVs with anti-inflammatory and anti-fibrotic activities have been generated. These EVs could exert synergistic effects in combination with loaded anti-inflammatory drugs [ 223 ]. The value of EVs in the treatment of ocular diseases is also becoming more apparent. In conclusion, EVs can serve as excellent nano-delivery vehicles or therapeutic drugs in their own right, and can be effective against a variety of diseases through the use of different therapeutic approaches. As delivery vehicles, EVs have relevant antigenic surface molecules that allow lesion targeting and BBB permeability, making them a novel therapeutic tool for the treatment of neurodegenerative diseases. In order to achieve better efficacy, different therapeutic tools can be combined, or EVs can be modified to obtain other functions. Thus, EVs show great therapeutic promise.

While EVs hold great promise as therapeutic and drug delivery agents for disease treatment, and some EVs are already in clinical trials for disease treatment applications (clinical trials for EVs-based treatments are summarized in Table 6 ), but scalable production in compliance with regulations remains challenging [ 230 ]. Few EVs formulations are currently being translated clinically, and many problems need to be overcome before their universal clinical application. Table 6 EVs-based therapeutics in clinical trials Phase and number Source/Sampling Condition or disease Dose or period of administration Results or primary outcome measures Refs. Not Applicable n = 20 Human amniotic mesenchymal stem cells Hair Loss; Alopecia Exosome (100e10 particle) injections with an interval of 14 days during two months Change in mean total hair density (hair/cm 2 ) NCT05658094 Phase 1 Phase2 n = 80 Human Placenta Mesenchymal Stem Cells Fistula Perianal In 3 weekly episodes Safety of injected exosomes NCT05402748 Phase 2 Phase3 n = 60 MSCs SARS-CoV2 Infection Intravenous injection twice, in day 1 and day 7 of 14 days Time to clinical improvement (days) NCT05216562 Phase 1 Phase2 n = 80 Human Placenta Mesenchymal Stem Cells Perianal Fistula in Patients With Crohn’s Disease 5 mL of exosome solution Safety of injected exosomes NCT05499156 Phase 1 n = 24 Allogenic Adipose Mesenchymal Stem Cells Coronavirus 5 times aerosol inhalation of MSCs-derived exosomes (2.0 × 10E8 nano vesicles/3 ml at Day 1, Day 2, Day 3, Day 4, Day 5) Adverse reaction (AE) and severe adverse reaction (SAE) NCT04276987 Phase 1 n = 30 Platelet rich plasma Chronic Low Back Pain; Degenerativ-e Disc Disease 2 mL of exosomes Visual analog scale (VAS) NCT04849429 Phase 2 n = 30 MSCs Knee; Injury; Meniscus (Lateral) (Medial) 1 million cells/kg Exosome Evaluation of Knee Functions NCT05261360 Not Applicable n = 30 N/A Exosome Post-stroke Dementia; Acupunctur-e N/A concentration of Exosome NCT05326724 Phase 1 Phase 2 n = 30 MSCs COVID-19 Twice a day for 10 days inhalation of 3 mL special solution contained 0.5 ~ 2 × 10 10 of nanoparticles (exosomes) Number of Participants With Non-serious and Serious Adverse Events During Trial NCT04491240 Phase 2 n = 41 Tumor Antigen-loaded Dendritic Cell Non-Small Cell Lung Cancer Intradermal injections once a week during 4 consecutive weeks Progression free survival NCT01159288 Early Phase 1 n = 9 Dendritic cells, macrophages, Tumor cells Recurrent or Metastatic Bladder Cancer N/A Clinical response rate NCT05559177 Phase 1 n = 13 Tumor cells Malignant Glioma of Brain 10 to 20 million IGF-1R/AS ODN treated tumor cells, encapsulated in diffusion chambers (maximum of 10), and re-implanted in the patient’s abdomen within 24 h after the surgery for a 24-hour period To establish the safety profile of a combination product with an optimized Good Manufacturing Practices AS ODN in the treatment of patients with recurrent malignant glioma with concomitant assessment of any therapeutic impact NCT01550523 Phase 1 n = 38 Wharton’s Jelly Mesenchymal Stem Cells Chronic Ulcer Conditioned Medium gel for 2 weeks Knowing the success rate of chronic ulcer healing in patients undergoing wound care with conditioned medium NCT04134676 Phase 2 n = 102 Bone marrow COVID-19, Acute respiratory distress syndrome (ARDS) 10 mL, which is 800 billion EVs. 15 mL, which is 1.2 trillion EVs Evaluation of 60 day mortality rate NCT04493242 Phase 1 n = 28 MSCs Metastatic Pancreatic Adenocarci-noma ; Pancreatic Ductal Adenocarci-noma; Stage IV Pancreatic Cancer over 15 ~ 20 min on days 1, 4, and 10. Treatment repeats every 14 days for up to 3 courses in the absence of disease progression or unacceptable toxicity. Participants who respond may continue 3 additional courses Maximum Tolerated Dose Determined by Dose Limiting Toxicity NCT03608631 Phase 1 Phase 2 n = 20 MSCs Segmental; Fracture - Bone Loss N/A Adverse effects associated with the therapy NCT05520125 Phase 1 Phase 2 n = 81 BMSCs ARDS; Human 10 mL, 15 mL The incidence of serious adverse events NCT05127122 Not Applicable n = 25 Autologous blood Otitis Media Chronic; Temporal Bone N/A Change in Inflammation Surface Area NCT04281901 n = 300 Explore the source of extracellular vesicles Traumatic Brain Injury N/A The type and content of circulating extracellular vesicles NCT05279599 Phase 2 Phase 3 n = 100 Autologous blood Otitis Media Chronic N/A Change of tympanic membrane perforation size NCT04761562 Phase 1 n = 10 MSCs Burns 1 × 10 4 MSCs for each cm 2 Primary Objective NCT05078385 N/A BMSCs Covid19; ARDS; Hypoxia Cytokine Storm Intravenous Infusion over 60 min N/A NCT04657458 Phase 1 Phase 2 n = 60 BMSCs Covid19; Postviral Syndrome; Dyspnea ExoFlo 15 mL (10.5 × 10 8 EVs) Increased distance on Six Minute Walk Test (6MWT) NCT05116761 n = 50 Peripheral blood samples ARDS Human N/A 28 day mortality NCT05061212 Not Applicable n = 10 Autologous serum Ulcer Venous 3 weeks once a week Changes in the ulcer area from