Exosomes as Natural Nanocarriers for RNA-Based Therapy and Prophylaxis

Exosomes are natural nanocontainers actively secreted by the body’s cells and transmitting molecular signals of various types to recipient cells. Cellular mechanisms of exosomes’ biogenesis involve specific sorting of RNA for incorporation into them. As a result, the molecular composition of exosomes is closely related to the donor cell’s functional state, and this makes exosomes an important diagnostic and prognostic marker in a number of diseases (primarily oncological). The ability of exosomes to transport biologically active molecules and to protect the cargo from degradation makes them nearly ideal candidates as delivery carriers of RNA in therapeutic or prophylactic regimes. Potential of exosomal surface functionalization enables improved targeting to specific organs, tissues and cells. However, the development of an effective technology for RNA’s loading into exosomes cannot be considered resolved. This review is focused on experimental data on the use of exosomes as vehicles for the delivery of therapeutic and prophylactic RNAs. We briefly consider the biogenesis and functions of exosomes, focusing on those biological properties that make them formidable candidates in the race to develop effective delivery carriers. Furthermore, we describe various techniques of cargo loading into exosomes. Prospects of exosomes application as therapeutic delivery system for siRNAs, miRNAs, and long RNAs are considered.


Introduction
RNA-based therapeutics and vaccines are rapidly progressing areas of medical science and making their first steps towards clinical implementation. Agents with potential medical applications include both short and long RNAs. Therapeutic short RNAs include siRNAs, miRNA mimics, and antisense miRNA antagonists [1,2]. Potentially therapeutic long RNAs include mRNAs, long non-coding RNAs, and circular RNAs [3][4][5].
Messenger RNA vaccines have been developed recently as a new tool for immunotherapy for oncological diseases [6,7]. They have also been developed as means of infection prevention, with a number of advantages over classical protein vaccines [8]. The current COVID-19 pandemic has sped up progress in the field. Two SARS-CoV-2 mRNA vaccines have received accelerated approval by US and EU regulators for clinical use, and several others are undergoing clinical trials [9].
Despite obvious advances in this area over the past few years, many challenges remain for RNA-based therapy and prophylaxis. For siRNAs and miRNAs, eventual off-target effects with silencing of unexpected genes are a concern, despite powerful bioinformatic Rab7 [63], Rab11 [64], Rab35 [65], and Ral [66,67]. Membrane fusion is mediated by SNARE proteins, such as VAMP7 [68] and YKT7 [69].
Knowledge of the mechanisms of recipient cell exosomal uptake is of great importance for the development of exosome-based delivery vehicles. Clearly, natural exosomes are capable of preferentially binding specific target cell types and inducing specific phenotypic or biochemical changes. For instance, exosomes from primary neurons are internalized only by other neurons, whereas those from neuroblastoma cells are internalized predominantly by glial cells [70]. Similarly, exosomes derived from various cell types (mesenchymal stromal cells (MSC), monocytes), when administered to a co-culture of both cell types, are captured only by homotypic cells [71]. This specificity is based on the protein repertoire exposed on the surfaces of both the exosome and recipient cell. These molecules include, for instance, integrins [72]; heparan sulfate proteoglycans [73]; T cell immunoglobulin and mucin domain containing protein 4 (Tim4) [74]; and sialoadhesin CD169 [75].
Exosomal internalization by recipient cells apparently occurs via several endocytic pathways, such as a clathrin-dependent pathway; a caveolin-dependent pathway; macropinocytosis; and a lipid raft-mediated pathway. Direct membrane fusion is also possible (reviewed in: [30]). While direct fusion provides immediate cytosolic delivery of exosomal cargo, endocytosed exosomes enter the recipient cell's endosomal compartment. It is assumed that their escape from the endosome lumen by membrane fusion is promoted by low endosomal pH [76].
The essential steps of biogenesis, release and uptake of exosomes are summarized in Figure 1. When forming exosomes, ESCRT-0 binds the monoubiquitinated proteins, clusters them in specific patches on the endosomal membrane and passes to ESCRT-1. ESCRT-1 interacts with ubiquitinated proteins and activates ESCRT-2, which leads to formation of secondary invaginations of the late endosome and movement of ubiquitin-labeled proteins into ILV. ESCRT-2 recruits ESCRT-3, which exerts budding off of ILV, with MVB formation. MVB is subjected to either lysosomal degradation or plasma membrane fusion, with release of ILV as exosomes. A number of Rab GTPases and SNARE proteins are involved into exosome secretion. Exosomes contain a specific set of proteins and RNAs of several types. Exosomes can interact with recipient cell in several ways, such as receptor-ligand interaction, direct membrane fusion, macropinocytosis, or endocytosis. Within recipient cell, exosomes are either degraded or release their cargo to the cytosol, thereby modifying the recipient cell functions.
It has been firmly established by extensive research [88,89] that a significant fraction of these RNA molecules: (i) is specifically loaded into donor cell exosomes; (ii) is passed to recipient cells intact; and (iii) when passed, modulates recipient cell function. Taken together, these characteristics make exosomes a nearly ideal biomimetic carrier for RNAbased therapeutics. The ability of exosomes to transfer and specifically deliver various RNA types to target cells [90] is of great importance in several areas. These include understanding of the fundamental mechanisms of multiple diseases and developing approaches for their treatment and prevention, including gene therapy.
Crucially, exosomes overcome the immunogenicity issue, which is an insurmountable hurdle for many other delivery vehicles [27,91]. Exosomes exhibit longer sustained circulation in mice, compared to liposomes, due to the surface presence of CD47, a molecule that prevents fast capture by monocytes or tissue macrophages [92]. The small size of exosomes (30-120 nm) also diminishes their clearance rate by macrophages. In addition, exosomes are small enough to pass through fenestrae in liver capillaries [93], which permits extravasation. Obviously, effective extravasation is a key step for exosome-delivered cargo to exert its effects.
Exosomal surfaces can be engineered to enhance their targeting to certain cell types. For example, exosomal membrane protein Lamp2b has been fused to the rabies virus RVG peptide which is known to bind specifically to neurons. In result, RVG-exosomes acquired the ability to cross the blood-brain barrier in mice and deliver siRNA therapeutics to brain tissue [94]. Similarly, exosomes have been engineered to carry Epstein-Barr virus glycoprotein 350. These exosomes specifically bind CD19+ B cells, but not other peripheral blood mononuclear cells [95].
In another work, HEK293 cells were modified to stably overexpress the GE11 peptide fused with the transmembrane domain of the platelet-derived growth factor receptor. GE11 binds specifically to the epidermal growth factor receptor (EGFR), which is abundantly expressed in tumors of epithelial origin. GE11-containing exosomes from these donor cells were able to deliver therapeutically relevant let-7a miRNA specifically to EGFR-expressing xenograft breast cancer tissue in RAG2 −/− mice [96].
Although the aforementioned recombinant protein-based technology for exosome targeting is effective, it is also very expensive, time consuming and labor intensive. Therefore, simpler, click chemistry approaches to exosomal surface modification have been developed. In this way, functional groups are attached to the exosomal surface via covalent bonds. Typically, exosome modification by click chemistry is based on a reaction between alkyne and azide groups (copper-catalyzed azide alkyne cycloaddition), which forms a triazole linkage [97]. By click chemistry, exosomes were functionalized with RGE peptide targeting neuropilin-1. As a result, these exosomes, loaded with curcumin, displayed high tropism for glioblastoma cells and featured positive therapeutic effects [98].
Another approach to adding required molecules to exosomal surfaces, called the 'post-insertion' technique, entails exosome mixing with lipid micelles conjugated to the molecules of interest. For example, anti-EGFR nanobodies have been fused to PEGylated phospholipids. Subsequent mixing of these nanobody-PEG-lipid micelles with exosomes at elevated temperatures resulted in effective transfer of nanobody-PEG-lipids to the exosomal membrane [99]. In general, this "exosome mimetic" strategy is capable of providing a wide range of exosomal surface modifications [100].
Apart from exosomal surface remodeling, cargo RNA can be engineered in various ways to achieve efficient packaging into exosomes. For example, incorporation of a "zipcode" (homing) sequence into the 3 UTR of the cargo RNA enables recognition by specific RBPs associated with MVB-targeting [101].
Despite all advantages of exosomes listed above, their clinical implementation for the drug delivery (including RNA delivery) is still challenging. First, safety concerns should be addressed more rigorously, including potential immunogenicity and tumorigenicity of exosomes. Naturally, exosomes are biologically active entities aimed at the intercellular signaling, particularly within the immune system. Although shown to be less immunogenic than many artificial carriers, exosomes inherently exert immunomodulatory effects, depending on the donor cell type. Therefore, the appropriate choice of the producing cells is of key significance for the exosome application. Most often, autologous dendritic cells (DC), macrophages, mesenchymal stromal cells (MSC), fibroblasts, as well as HEK293-T cells are utilized as exosome donors [102,103]. Of note, MSC-derived exosomes are capable to promote the tumor vascularization [104], while mature DC-derived exosomes have immunostimulatory properties [105].
The second obstacle currently complicating the clinical implementation of exosomes is the drawback of purification technologies, which leads to instability in the composition of exosome preparations. In terms of the physico-chemical properties (size, buoyant density, molecular composition), exosomes partially overlap with other nanocomponents of the cellular secretome, which fundamentally hinders their reproducible separation in a completely pure form. There is a significant discrepancy in the assessment of the composition of exosomal proteins and RNAs, depending on the purification methods used by the researchers [22]. Obviously, these differences in molecular composition further affect the functional impact of the exosome preparation on recipient cells during exosomal delivery of therapeutic RNAs.
Finally, the problem of scaling up the production of exosomes for clinical purposes has not yet been solved. The level of exosome secretion by producer cells of any type is low, and scaling from research to clinical practice incurs huge material and labor costs. In this sense, ultracentrifugation and polymer precipitation are the simplest and most productive methods for isolating exosomes, but they rather do not provide a sufficient purity of the preparation. More complex approaches, such as gradient fractionation, size exclusion chromatography, and especially antibody-based affinity separation, make it possible to obtain more purified products, but their cost is much higher, and the yield of exosomes is lower, which makes technological scale-up difficult. The search for an acceptable compromise is ongoing, and, clearly, the further research in the field is required.
To summarize, exosomes are biocompatible, non-toxic RNA delivery vehicles with low immunogenicity. They are able to provide: extravasation of cargo RNA; internalization by target cells; and endosomal escape. There are broad opportunities for the functionalization of exosomal surfaces to increase delivery targeting, and various methods for modification of RNA cargo also exist. These features make exosomes a highly promising carrier for the delivery of both therapeutic RNAs [106] and prophylactic RNA vaccines [107,108], although extensive further research is necessary on the donor cell selection, exosome safety, optimization and scaling-up of purification protocols, before exosome-based drug delivery is implemented into clinics.

Techniques for RNA Loading into Exosomes
Methods for RNA loading into exosomes include both: straightforward approaches for RNA incorporation into previously isolated, purified exosomes; and more sophisticated cellular engineering techniques. In the latter, expression of the target RNA by exosomeproducing cells, and the intracellular transport of this RNA into the forming exosomes, occur first. Isolation and purification of target RNA-containing exosomes then proceeds.
In principle, the simplest method of loading into the lumen of an exosome is a mere co-incubation [exosomes with cargo molecules], during which molecules diffuse through the lipid bilayer of the exosome. This approach permits sufficiently effective penetration into exosomes for some therapeutically important hydrophobic substances, including doxorubicin [109], curcumin [110], and paclitaxel [111]. In contrast, however, RNA of any type is a negatively charged molecule, which precludes its spontaneous diffusion through the exosomal membrane's lipid bilayer. In order to improve RNA's interaction with the exosomal membrane, chemical modifications can be performed. Promising results have been obtained by conjugation of siRNA with cholesterol [112,113]. This makes RNA molecules more lipophilic and ensures their efficient incorporation into exosomes.
Alternatively, pore formation in the lipid bilayer for the entry of RNA into the exosome is possible. Several physical methods use such a mechanism. The most commonly used option for membrane pore formation is exosome electroporation, in which a mixture of exosomes and cargo RNA is pulsed with an electric field. Electroporation parameters require optimization depending on donor cell type. Typically, applied voltages range from 150 to 700 V [114,115], with exosome concentrations ranging from 0.07 to 0.5 µg/µL [5,94,116]. There are a number of reports on the successful loading of siRNAs [94,[117][118][119][120][121] and miR-NAs [122][123][124][125] into exosomes by electroporation, followed by in vitro or in vivo small RNA delivery into cells, with silencing of target genes. Cumulative experience, however, has also revealed some disadvantages of exosome electroporation, such as damage to vesicle integrity, aggregation, and low RNA loading efficiency [126]. Some authors did not achieved satisfactory loading of small RNAs by exosome electroporation [127]. Most fluorimetrybased estimates of small RNA loading efficiency into electroporated exosomes indicate about 20-25% of small RNA is captured by exosomes [116]. Furthermore, according to Kooijmans et al. [126], electroporation leads to massive aggregation of siRNA-containing exosomes (obviously not suitable for subsequent siRNA delivery). Trehalose medium minimizes the aggregation [128]. When using it, the value of RNA capture decreases to 0.05% [126]. This low loading efficiency significantly increases material consumption and the final cost of preparing of RNA-containing exosomes. Data on the applicability of electroporation for loading longer mRNAs into exosomes are also inconsistent. In one publication, Cas9 mRNA (4521 nucleotides) was successfully loaded into exosomes using electroporation with an insertion efficiency of 18% [129]. Other authors, however, have reported the impossibility of HChrR6 mRNA inclusion into exosomes by electroporation and the need to use cell-based technologies for loading for this purpose [130].
In addition to electroporation, other simple methods have been used for the formation of pores in the exosomal membrane to load small RNAs, such as sonication [131] and freeze-thaw cycles [132].
Further options are provided by: repeated extrusion of mixtures (exosomes, cargo molecules) through a syringe-based lipid extruder with 200 nm or 400 nm pore size [133,134]; or by exosome treatment with the pore-forming agent saponin [135,136]. Currently, these methods are used for loading of rather small cyclic compounds than for RNA.
Another approach that has been successfully applied for loading RNA into exosomes is transfection of exosomes using commercial liposomal reagents (HiPerFect, Lipofectamine, ExoFect), with formation of exosome/liposome hybrids. Exosome/liposome fusion occurs via co-incubation and can be further enhanced by repeated freezing-thawing of the mixture [137]. The exosome/liposome hybrid approach has been applied to exosome loading with miRNAs [138][139][140] and siRNAs, [141,142]. Importantly, according to published data, transfection of exosomes with lipid carriers is feasible not only for the inclusion of short RNAs, but also for much longer molecules, such as: mCherry mRNA (approx. 800 nucleotides) [143]; Antares2 mRNA (2000 nucleotides) [144]; as well as massive CRISPR-Cas9 plasmid expression vectors [145].
However, the exosome/liposome hybrid approach has also drawn criticism. After co-incubation, separation of exosomes from unbound lipid vesicles is either not carried out at all [139,140], or is performed by filtration. Such filtration is based on assumed differences in the linear sizes of exosomes and liposomes [141]. However, the two types of vesicles not only coexist in solution, but actively interact leading to hybrid formation. Therefore, taking also into account the generally high lability of lipid systems, complete separation is fundamentally impossible. Transfection reagents themselves are capable of binding RNA and delivering it to cells. Therefore, interpreting the results of RNA (or DNA) delivery into cells by exosome/liposome hybrids requires caution, especially as it relates to long mRNA molecules and CRISPR-Cas9 expression plasmids. The exosome is not a hollow vesicle, but is filled with a range of endogenous proteins and RNAs; this complicates incorporation of exogenous plasmids or mRNA into it.
The aforementioned physical and chemical methods deal with isolated exosomes. Cell-engineering methods involve delivery of the target RNA into donor cells, followed by incorporation of this RNA into exosomes during their biogenesis. Several cell-engineering approaches of this kind have been proposed.

1.
In the simplest case, cells are transfected directly with target RNAs: miRNA, [146,147]; siRNA [148]; antisense miRNA antagonist [149]; or in vitro transcribed mRNA [150]. Then, the loading of transfected RNAs into exosomes occurs. After harvesting and purification of exosomes secreted by donor cells, exosome-based RNA delivery into target cells has been demonstrated, including therapeutically relevant functional effects of RNAs on the recipient cells.
In the two cases above, the RNA (delivered directly to cells, or expressed from a plasmid or viral vector) does not carry any exosome-guiding sequence. Apparently, its loading into the exosomes of the donor cell from the cytoplasm occurs through non-specific engulfment as part of secondary invaginations of the endosomal membrane during the MVB formation. With this in mind, the abundance (number of copies) of RNA cargo available for loading is critical. Expression plasmids and viral vectors typically provide a high level of transcription, while the cytosolic concentration of directly delivered RNAs (both small RNAs and mRNAs) may be insufficient for its efficient packaging into exosomes of a donor cell. In addition, exogenous RNA delivered to the cytoplasm arrives at the general pool of cellular RNAs and inevitably enters cellular functional RNA pathways, such as gene silencing (for small RNAs), protein translation (for mRNAs), and, finally, degradation pathways. This can affect the state of donor cells, including the production/composition of their exosomes. As known, degradation of exogenous RNA can be slowed down by introducing chemically modified nucleotides into it [158].

3.
To overcome the aforementioned limitations, more specifically, RNA sequence and a protein that binds and directs it to exosomes can be engineered. For example, a C/D box can be inserted into the 3' UTR of the target mRNA, which is recognized by archaeal ribosomal protein L7Ae. In turn, the L7Ae sequence is fused with the C-terminus of the CD63 exosomal marker protein. Co-transfection of these two constructs provides highly specific targeting of the mRNA of interest to exosomes [159].
A similar principle was implemented in the Targeted and Modular EV Loading (TAMEL) platform proposed by Hung and Leonard [101]. In it, exosomal proteins (Lamp2b, or CD63) were fused with MS2 bacteriophage coat protein capable of binding a certain type of RNA secondary structure (MS2 stem loop). Overexpression of these fusion proteins in cells, and simultaneous lentiviral transduction of model dTomato mRNA (1.8 kb, containing several MS2 stem loops), resulted in efficient, targeted delivery of this mRNA to exosomes.
Various techniques for RNA loading into exosomes are summarized in Figure 2. Techniques for RNA loading into exosomes. RNA loading into exosomes can be achieved by either exosome donor cell-based or isolated exosome-based approaches. Donor cell-based methods involve target sequence delivery to the exosome donor cell via direct RNA delivery or via plasmid transfection or viral transduction, with the following intracellular packaging of target RNA to exosomes and further exosome harvesting. Loading into exosomes can be enhanced by RNA sequence engineering, including zipcode sequence or specific secondary structure joining. RNA-binding proteins (RBPs) can be engineered too, to recognize these RNA motifs. Isolated exosome-based methods are the several physical methods for exosomal membrane poration (electroporation, sonication, freeze-thaw cycles), as well as chemical methods for RNA penetration through the exosomal membrane. Chemical methods include strengthening of RNA affinity to the exosomal membrane via RNA-cholesterol conjugation, saponin-based pore formation and liposome-mediated RNA packaging into exosomes (exosome-liposome hybrid method).

Exosomal Delivery of Therapeutically Relevant RNAs
The first results on successful exosome-mediated siRNA delivery to target cells and subsequent specific gene silencing in them were obtained a decade ago [94,114]. Alvarez-Erviti et al. engineered exosomes to carry Lamp2b-RVG peptide, electroporated BACE1 siRNA into them and achieved a significant decrease in BACE1 expression in mouse brain tissue by systemic injection of these exosomes [94]. Walgreen et al. performed siRNA electroporation into plasma exosomes from the peripheral blood of healthy donors and demonstrated selective gene silencing of MAPK-1 in the target human monocytes and lymphocytes [114].
In recent years, advantages of exosomes as carriers have been repeatedly used experimentally in works on the delivery [both in vitro and in vivo] of therapeutically significant RNAs (most often small RNAs, such as siRNAs, miRNAs, miRNA antagonists) aimed at treating a number of diseases, such as oncology, neurological and neurodegenerative diseases, and infections.
Below, we consider the most substantial experimental results of in vivo delivery of therapeutically important RNAs by exosomes, in order to characterize the current state of the art.

siRNAs
Successful in vivo experiments on the use of exosome-delivered therapeutic siRNAs have in most cases been aimed at various neurological diseases, including Huntington's disease, Alzheimer's disease and spinal cord injury, as well as tumors of various origins. In the case of neurological diseases, RVG-Lamp2b modified exosomes were predominantly used to overcome the blood-brain barrier (BBB) and blood-spinal cord barrier (BSCB) and to deliver siRNA to the brain and spinal cord. In oncological diseases, targeting of siRNAcontaining exosomes to tumor tissue can be achieved by modifying them with the RGD-LAMP2b fusion protein. RGD peptide has a high affinity to integrin αvβ3 and specifically delivers exosomes to tumors in vivo [160]. For cancer treatment, siRNA targets include genes critically important for growth of tumors, such as S100A4, c-Met, PAK4, KRAS, etc. Of note, currently a phase 1 clinical trial of metastatic pancreatic adenocarcinoma treatment by MSC-derived exosomes with siRNA to mutated KRAS is in progress [NCT03608631].
Recent data on exosomal delivery of therapeutic siRNA in vivo are summarized in Table 1.

miRNAs
Dysregulation of miRNA profiles is a key component of a wide range of diseases including those of a cardiovascular, infectious, oncological, or neurological nature. Compensation for these pathological disturbances is a promising therapeutic strategy. Unlike siRNAs, miRNAs have multiple targets in the genome to perform "fine tuning" of gene expression. Exosomes are natural carriers which transfer a multitude of donor cell miRNAs. These endogenous exosomal miRNAs themselves can be of high therapeutic significance. For example, MSC-derived exosomes contain more than 150 pre-miRNAs and miRNAs [170]. Such exosomes have shown therapeutic potential in the treatment of multiple diseases including liver fibrosis, acute kidney injury, myocardial infarction, Alzheimer's disease, type 1 diabetes, and asthma. Several clinical trials (phases I, II, III) of MSC exosome-based therapies are now ongoing [171]. For severe COVID-19, MSCc exosome treatment resulted in reliable improvement in patient oxygenation, neutrophil count reduction, and declines in acute phase reactants (C-reactive protein, ferritin, Ddimer) [172,173], and the key role of several miRNAs (miR-126, miR −30b-3p, miR −145, miR −27a-3p) has been established [174].
In addition, exosomes from various cells (most often MSCs) with exogenously loaded miRNAs have been repeatedly tested experimentally to treat cancer and several neurological disorders, with proven amelioration in disease course. Table 2 represents data on the experimental treatment of various types of cancer and neurological disorders by exosome-delivered miRNAs.

mRNAs
Although the huge potential of mRNA-based therapy is widely acknowledged, current experience in delivery of therapeutic mRNAs by exosomes is rather limited. The large size of mRNA significantly hampers its incorporation into exosomes, in comparison with small RNAs. There is one publication that deals with model mRNA encoding fluorescent protein, loaded into exosomes by exosomal-liposomal hybrid method [143].
More therapeutically relevant, sophisticated cellular engineering techniques of specific mRNA design and targeting are typically used. Maugeri and co-authors delivered mRNA encoding human erythropoietin (hEPO-mRNA) to epithelial cells HTB-177 using LNP. Then, hEPO-mRNA was revealed in EVs secreted by these donor HTB-177 cells. After intravenous administration of hEPO-mRNA-containing EVs to mice, hEPO protein was detected in plasma and number of organs in 2 h after injection [180].
In Kojima's group work, catalase mRNA fused with C/Dbox for effective exosomal targeting was engineered [159]. In this study, designer exosomes carrying RVG-Lamp2b effectively delivered catalase mRNA across the BBB, and finally decreased 6-OHDA-mediated neuroinflammation in mouse model of Parkinson's disease.
In another study, exosomes aimed at HER2+ve breast cancer treatment were developed. For exosome targeting to the cancer cells, HEK293 cells were transfected with plasmid encoding EVHB chimeric protein, which contains high-affinity anti-HER2 scFv antibody. This protein is capable of being exhibited on the EV surface. These cells were also transfected with a plasmid-encoding modified E. coli enzyme HChrR6. This enzyme is capable of converting the prodrug (CNOB) into the cytotoxic drug (MCHB). In transfected cells, HChrR6 mRNA was loaded into HER2-targeting exosomes. When administered in vivo, these exosomes caused effective HChrR6 expression in tumor cells, with subsequent near-complete growth arrest of BT474 xenografts after CNOB prodrug treatment [130].
In a recent publication by Tsai et al., exosome-packaged mRNA encoding the S and N proteins of SARS-CoV2 was used to experimentally vaccinate mice. The results obtained by the authors indicate the post-vaccinal appearance of specific anti-S and anti-N antibodies in the blood plasma, as well as T cell responses to N and S proteins [144]. These data confirm the possibility of using exosomes as a means of delivering COVID-19 mRNA vaccines.
Thus, exosome-mediated delivery of mRNAs is of proven therapeutic relevance. At the same time, however, complexities present with loading and targeting technologies based on cell engineering complicate clinical implementation. Table 3 summarizes data on the exosome-based mRNA delivery.

Conclusions and Future Perspectives
Exosomes are natural nanocontainers actively secreted by the body's cells, thereby transmitting molecular signals of various types to recipient cells. Cellular mechanisms of exosome biogenesis involve specific sorting of RNA and proteins for incorporation. As a result, the molecular composition of exosomes is closely related to the functional state of the donor cell; this makes exosomes important diagnostic and prognostic markers in a number of diseases (primarily oncological). The ability of exosomes to protect their cargo from degradation in the extracellular environment, and to transfer it to recipient cells, makes them effective candidates for the delivery of RNA in therapeutic or prophylactic regimes. The possibility of exosomal surface functionalization enables improved delivery targeting.
There is no doubt about the significant potential of exosomes as a carriers of gene therapy agents. In recent years, there has been an explosive growth in research interest in exosome-based drug delivery, and several clinical trials of exosome-encapsulated siR-NAs are in progress currently. However, there are certain obstacles to the more active implementation of exosome technology in clinical practice.
Despite their proven low immunogenicity, exosomes are always biologically active entities aimed at modulating recipient cell function. For example, tumor-derived exosomes prepare a niche for metastasis [181]. In this context, safety concerns around exosome-based therapies need to be addressed more thoroughly.
The question of choosing the type of donor cells requires special attention. Exosomes are not biologically inert particles, but perform regulatory and communicative (including immunomodulatory) functions, and their activity depends on the type and functional state of parental cells. Currently, human MSC, immature DC, and HEK293-T cell lines are most often used as exosome donors. In the future, a "universal exosome donor" could be developed, which is optimal for packaging and delivery of various RNA cargoes. In this sense, of great interest is the work of Kojima and colleagues [159], who created an "exosome production booster" based on HEK293-T, suitable for further engineering of exosome-targeted RNAs of interest.
Clearly, further development of specific ligands that determine the dispatching of exosomes to tissues and target cells is necessary. In this regard, there are examples of striking success, such as RVG peptide-based targeting of exosomes to neuronal tissue [94] or GE11peptide-based targeting of exosomes to EGFR-positive tumors [96]. However, a huge therapeutic potential of exosome-encapsulated RNAs definitely requires a significant expansion of this list. Apparently, monoclonal antibody-based aiming of exosomes could be a promising, albeit expensive, solution.
The development of a technology for loading of RNAs into exosomes cannot be considered resolved. Plain pore-forming approaches dealing with isolated exosomes are attractive due to their methodological simplicity; however, their usage has often resulted in uncontrolled aggregation and damage to the integrity of both the exosomes themselves and their cargo, especially long mRNA molecules. In general, cell engineering methods seem to be more promising, although currently expensive and difficult to scale.
Exosome purification technology and its scalability is also challenging at the moment. It is well known that the vesicular secretome of the cell is heterogeneous. EVs of different types differ significantly in biogenesis, molecular composition, and functions. It is probable that some of them are "debris bags" for the cell, as was postulated in early studies, while others perform specific communicative tasks. At the same time, vesicles of different types overlap in size and buoyant density, and reproducibly obtaining completely purified homogeneous preparations of exosomes is currently extremely complicated. The most effective and specific purification method is the immunoaffinity technique; however, the high cost of antibodies makes it rather difficult to scale it for clinical application. Undoubtedly, further efforts of researchers are needed to optimize exosome purification technologies and their industrial scaling.
Finally, more fundamental knowledge is needed regarding the biological roles of exosomes in the body including: kinetics; overcoming of (or limitation by) biological barriers; and molecular mechanisms of endosomal escape. The roles of multiple molecules delivered by exosomes have not been sufficiently studied. For example, the functions of a number of exosomal lncRNAs and circRNAs remain rather elusive, yet such understanding is obviously necessary to realize the therapeutic potential that exosomes represent.
Author Contributions: Conceptualization, A.G. and A.V.; writing-original draft preparation, L.P., A.G. and A.B.; writing-review and editing, A.V. All authors have read and agreed to the published version of the manuscript.
Funding: This research was funded by the Russian Science Foundation, grant number 20-15-00228.