Exploring the role of polymers to overcome ongoing challenges in the field of extracellular vesicles

Abstract

Extracellular vesicles (EVs) are naturally occurring nanoparticles released from all eucaryotic and procaryotic cells. While their role was formerly largely underestimated, EVs are now clearly established as key mediators of intercellular communication. Therefore, these vesicles constitute an attractive topic of study for both basic and applied research with great potential, for example, as a new class of biomarkers, as cell‐free therapeutics or as drug delivery systems. However, the complexity and biological origin of EVs sometimes complicate their identification and therapeutic use. Thus, this rapidly expanding research field requires new methods and tools for the production, enrichment, detection, and therapeutic application of EVs. In this review, we have sought to explain how polymer materials actively contributed to overcome some of the limitations associated to EVs. Indeed, thanks to their infinite diversity of composition and properties, polymers can act through a variety of strategies and at different stages of EVs development. Overall, we would like to emphasize the importance of multidisciplinary research involving polymers to address persistent limitations in the field of EVs.

1. INTRODUCTION

Cells communicate and exchange information by several mechanisms, such as direct cell‐to‐cell contact, secretion of soluble factors, but also, as now well established, by the secretion of extracellular vesicles (EVs). EVs structure consists in a shell of phospholipid bilayer, including surface and transmembrane proteins, surrounding a hydrophilic core (lumen) containing lipids, carbohydrates, proteins, as well as a set of nucleic acids. They appear as carrier systems operating from inside a donor cell towards either the surface or the inside of target cells. EVs are therefore involved in a wide range of physiologic and pathologic events (van Niel et al., 2018). In healthy cells, EVs are commonly classified into two subcategories according to their biogenesis: (i) ectosomes (100 nm–1 µm), which bud directly from the plasma membrane of healthy cells; (ii) exosomes, the smallest EVs (50–150 nm), which are released by the fusion of multivesicular endosomal bodies (MVBs) with the plasma membrane. Regarding the complexity and heterogeneity of these vesicles, also associated with the difficulty of current isolation methods to fully sort and select one or another EVs populations, we chose, in this review, to use the generic term EVs, following ISEV guidelines (Théry et al., 2018). This term was selected even if the name exosome or ectosome or microvesicle was specified in some described works.

Over the last 20 years, discoveries and interest in the EVs field have exploded, transforming EVs from vulgar cellular ‘garbage‐bags’ to key players in intercellular communication (van Niel et al., 2018). In parallel to the growing interest of academic researchers, a rising number of companies have emerged to exploit EVs as therapeutic agents (Evox Therapeutics, Codiak BioSciences, Carmine Therapeutics, Capricor Therapeutics, EverZom, Ciloa, RoosterBio, Omnispirant…). Recently, several partnerships have been concluded between these start‐ups and key players in the pharmaceutical sector (i.e., Lilly, Takeda or Bayer) (Zipkin, 2020). EVs are particularly promising as they can exert some of the therapeutic effects of their producing cells, without the constraints associated with the use of living cells. In this context, one of the most widely used EVs source is mesenchymal stromal cells (MSCs), since vesicles produced by these cells were reported to carry some of the beneficial properties of their cells of origin, such as antibacterial, anti‐inflammatory and anti‐fibrotic effects (Cosenza et al., 2017; Lai et al., 2010; McCarthy et al., 2020; Romanelli et al., 2019). In parallel, their role of natural biomolecules nanocarrier make them interesting candidates as drug delivery systems (Elsharkasy et al., 2020; Herrmann et al., 2021). Indeed, EVs possess a natural ability to diffuse into dense extracellular matrices (Lenzini et al., 2020), to interact with target cells both in vitro and in vivo (Saux et al., 2020; Smyth et al., 2015) and to deliver, at picomolar amount, several biomolecules able to induce an effect in the target cell (Ratajczak et al., 2006; Valadi et al., 2007). Recent reports revealed that in vivo miRNA delivery efficacy could be 10 to 300‐fold more important with EVs compared to lipid nanoparticles (Reshke et al., 2020) and more to 100‐fold in vitro (Murphy et al., 2021). However, to achieve such efficiencies, RNA molecules must be loaded inside EVs, and not coated onto EVs (using lipid siRNA (Didiot et al., 2016) or aggregated as often observed with standard electroporation processes). The history of the nucleic acids loading, such as siRNA, into EVs is a particularly interesting example of the research evolution in this field. Indeed, if RNA loading into EVs was initially described as efficient using electroporation (Alvarez‐Erviti et al., 2011), the excitement died down with the discovery of aggregated RNA particles induced by the electroporation process itself (Kooijmans et al., 2013). However, due to the high potential of these vesicles and the development of increasingly effective characterization methods, key players in the community have been able to come up with appropriate loading methods. This led, over the following decade, to the high transfection efficiencies mentioned above. Alongside the complexities associated with exogenous molecules loading, the beneficial effects of unmodified EVs are also sometimes debated. Indeed, depending on the method of isolation and characterization, the results are sometimes contradictory as heterogeneity of vesicle population could hinders EVs clear identification and subsequent behaviour. Therefore, reproducibility is still a major challenge in this field, with a need for standardized and reproducible methods. In addition to the main limitations already mentioned, issues of production and isolation scalability, development of standard storage conditions, as well as instability after in vivo administration are other barriers still hindering the development of EVs as relevant therapeutics. Therefore, in spite of EVs potential, as new systems coming from living sources, several technical barriers must be overcome. In this context, thanks to their versatility, we identified polymers as one of the most valuable contributors that can help take up these challenges for EVs, even though they are often not identified as such till now.

Polymers are macromolecules composed of a monomer (or monomeric unit) repetition which defines the degree of polymerization. They can differ from their origin, that is, natural or synthetic, but also from their chemical nature, molecular weight, dispersity of polymer chains, nature and position of reactive sites on the polymer skeleton, and functionality. To help readers unfamiliar with polymers, some educational boxes have been created to describe specific knowledges associated with the polymer field and needed to fully understand this review. The different polymer chemical processes were explained in BOX 1 with their limits, experimental imperatives and toxicity of the reactants. This box helps to further understand toxicology issues, problems of purification, eventual degradation, and incompatibility with EVs in health applications. Among polymers associated with EVs, 40% were from natural sources versus 60% from synthetic ones. This difference matters since it impacts the polymer properties or applications as detailed in BOX 2. Lastly, in BOX 3 we analysed the vast functionalization of polymers reported through different chemistries and various macromolecular architectures.

Synthesis of polymers

All monomers cannot polymerize relying on the same chemical process depending the cyclic or acyclic nature, the presence or not of electron donor or withdrawing groups.

Step‐growth polymerization (SGP) consists in the reaction between two co‐monomers owing antagonist reactivities (monomer X and monomer Y in Figure BOX 1a) such as alcohol and acid or amine and acid groups to produce polyamides, polyesters, polyurethanes (i.e., PET, PBAE, PGSA). SGP offers some advantages such as the absence of potentially toxic residual reactants (initiator, terminating agent and catalyst) or the possibility to growth polymer chains only by heating, with or without organic solvent. The main limitation is the requirement of a very high conversion rate and the necessity to perfectly respect of stoichiometry of respective monomers to reach high molecular weights.

Starting from cyclic monomers (lactames, lactones, oxazolines…) in some cases tense cycles (ethylene oxide, aziridine), ring‐opening polymerization (ROP) can happen (Fig. BOX 1b). Depending on the nature of the initiator, poly(caprolactone) (PCL) and poly(ethylene oxide) (PEO) also named poly(ethylene glycol) (PEG) were synthesized by anionic ROP (AROP), poly(glycolic acid) (PGA) and polyoxazolines by cationic ROP (CROP), and poly(lactic acid) (PLA) by organo‐mediated ROP. The methodology is often limited to relatively low molecular weights (50 kDa) because of the occurrence of transfer reactions but it gives access to many different natures of polymers.

Many acrylates are used in the biomedical field because various functional substituents (R) of the unsaturation are marketed in large quantities (Fig. BOX 1c). Nevertheless, the conventional radical polymerization (CRP) also named Free Radical Polymerization (FRP) requires potential toxic radical initiator and the process does not allow sharp control of the molecular weights (Das et al., 2022). To overcome this limit, controlled and/or living processes have been successfully developed with ATRP (Atom Transfer Radical Polymerization) (Lathwal et al., 2021), NMP (Nitroxide Mediated Polymerization) or RAFT (Reversible Addition‐Fragmentation Transfer) polymerization. Unfortunately, all these processes require a supplementary metal and ligand for ATRP, transfer agent for RAFT or nitroxide reactant for NMP. As alternative to thermal FRP (Fig. BOX 1c), the photopolymerization is largely spread using the same monomers (i.e., (meth)acrylates), and a photoinitiator with a low toxicity (i.e., Irgacure 2959). Interestingly, the photopolymerization is often used instead of the appropriate terms of photo‐‐crosslinking.

This brief introduction to the synthesis of polymers points out that each chemical process of polymerization has pros and cons. Some of them require a catalyst which can be detrimental for toxicology issues (e.g., polyesters with metallic catalyst based on tin (DBTDL for PLA)). The same toxicity can be noted with ATRP using copper catalyst. In other cases, a limit can be the toxicity of monomeric precursor as ethylene oxide for PEG or aziridine for PEI. So, the well‐defined character and the high purity by simple purification process of the synthetic polymers are two main advantages for biomedical applications.

Figure BOX 1 Different processes of polymer synthesis: (a) the step‐growth polymerization using two co‐monomers bearing complementary chemical functions named X and Y, (b) the ring‐opening polymerization (ROP) starting from cyclic monomer and using anionic initiator (AROP), cationic initiator (CROP) and organocatalyst (organoROP), c) the free radical polymerization (FRP) and the controlled radical polymeriation (RAFT…) of (meth)acrylates, (meth)acrylamides.

Natural polymers versus synthetic polymers

The natural polymers gather polysaccharides (glycan, chitosan, cellulose, hyaluronic acid (HA), pullulan and alginate) and proteins and other natural polypeptides (silk fibroin (SF), gelatin) while synthetic polymers range from polyesters ((poly(ethylene terephthalate) (PET), poly(caprolactone), poly(lactic acid) (PLA), poly(lactic‐co‐glycolic acid) (PLGA), poly(glycerol sebacate) (PGSA), poly(b‐amino ester) (PBAE)) to N‐rich polymers including synthetic polypeptides (poly(L‐lysine) (PLL) also named poly(ε‐L‐lysine) (EPL)), polyamides (poly(vinylpyrrolidone) (PVP), poly(N‐vinyl caprolactam) (PNVCL), poly(amido amine)), polyamines (poly(ethyleneimine) (PEI)). The gold standard of the synthetic polymers is poly(ethylene oxide) (PEO) also named poly(ethylene glycol) (PEG) which represents 25% from the 40% of synthetic polymers listed in this review. Also worthy of note is the exclusive synthetic character of the polyesters whereas N‐rich polymers were found to belong both classes of polymers.

Other anecdotic or specific synthetic polymers were reported with aptamer, silicone (poly(dimethylsiloxane) (PDMS) (Guo et al., 2021), poly(acrylic acid) (PAA), thermosensitive polymers such as poly(N‐isopropylamide) (PNIPAM), or electroactive polymers including poly[(9,9‐dihexyl‐2,7‐(2‐cyanodivinylene)‐fluorenylenyl‐2,7‐diyl)] (CN‐PDHFV) by Zhu et al. for detection (Zhu et al., 2021). All the polymers are electrical insulators except conjugated polymers characterized by the alternating of single and double bonds permitting the delocalisation of pi electrons along the macromolecular structure and consequently the electrical conduction. Conjugated polymers belong to the electroactive polymers which exhibit a change in size or shape under electric field. The most common applications of this type of material are in actuators and sensors.

The table below illustrates some polymers and their main applications but it does not represent an exhaustive list nor intend to sum up all polymers from this review. Indeed, several significant trends emerge from the global inventory of the petro‐ and bio‐sourced polymers used with EVs, with mostly water‐soluble and biodegradable polymers. As the term biodegradable commonly implies two different senses depending on the media and the applications, a clarification is be provided. The first sense refers to the polymer ability to be decomposed in the environment. It applies more to plastics used in the current industrial applications and can be nuanced by oxo‐degradable, compostable and fragmentable properties. The second sense corresponds to the ability of a polymer to be decomposed by biological activity and action of cells. It is more employed in biomedical field which correspond to the meaning in this review. The term degradable should be preferred when the polymer degradation results from the action of water (hydrolysis) or when the mechanism of chain scission is unknown or not proven as cell mediated. Back to the table (Fig BOX 2), the petro‐based polymers are more attractive for manufacturing or production due to their well‐defined character and hydrophobic behaviour. By contrast, the biomacromolecules (polysaccharides and proteins) are mainly used for hydrogels benefiting to their biocompatibility and polyfunctionality. The values of glass transition temperature (Tg) and melting temperature (Tm) reveal the variety of nature of polymers coming from amorphous to semi‐crystalline, from thermoplastic (PET, PLA…) to elastomer (silicone, COC) or thermoset. Even if they do not directly used, these latter data give some information on (i) the ability of a polymer to the self‐organization and a further crystallinity (case of the polymers with Tm), favouring some specific interactions, (ii) the rigidity of the backbone (high Tg) influencing the penetration into lipid membrane or (iii) the deformability of soft materials as evidenced by a low value of Tg.

Figure BOX 2 Nature and properties of the main polymers associated to EVs. Summary of various polymers for EVs research with their origin (petro‐based or natural), their properties (hydrosolubility, glass transition or melting temperatures, degradability) and main applications.

Macromolecular design of polymers

Non‐modified polymers

Some pristine polymers are employed for their capacity to interact with other polymers or particles thus in this case no modification is required (non‐modified). A typical application is the modification of the surface of EVs using layer‐by‐layer (LbL) deposition of polymers bearing opposite charged polymers such as PLL and HA or PLL and PAA (Jhan et al., 2021). This concept is also used for the surface modification of inorganic nanoparticles quantum dot (Pramanik et al., 2019) or Fe₃O₄ magnetic beads (Liao et al., 2022) or MPC (Mori et al., 2019). In fact, the pristine polymers can be employed without any modification but many EVs associations often require more sophisticated macromolecules such as amphiphilic properties and with specific anchoring groups to be inserted in EVs membrane.

Polymer functionalization

Linear architecture

Lipid‐polymer. Polymers can be easily decorated by lipids like DSPE (DSPE‐PEG) (Choi et al., 2019), DMPE (DMPE‐PEG) (Kooijmans et al., 2016), cholesterol (Chol‐Pullulan (Sawada et al., 2020), Chol‐PEG‐DNA (Lathwal et al., 2021)). These associations consist in the PEGylation of lipids which correspond to the coupling of PEG and a natural hydrophobic chain to generate an amphiphilic polymer.

The terminal or pendant reactive functionalization of polymers are also performed for potential future coupling steps. For instance, amine and alcohol groups of gelatin are methacrylated to elaborate bioink for 3D printing as described by Hu et al (Hu et al., 2021), and Chen et al. (Chen et al., 2019). The most common chemical modifications are the terminal modification of the end groups of PEG chain (PEG‐CO₂H, PEG‐SH, PEG‐NH₂, PEG‐CH = CH₂). These functionalisations allow different kind of couplings including thiol‐ene coupling between unsaturated chain and thiol group (DMPE‐PEG‐ CH = CH₂ and NB‐SH) (Fig. BOX 3.1a) as well as thiol‐epoxy (Fig. BOX 3.1e) (Zhu et al., 2021). But, the main issue of this reaction is the requirement of an excess of thiol to offset the consumption of thiol in unwanted S‐S bridges. However, the separation of the excess of thiol reactants is often complicated. We also note the use of voluntary excess of one of the two reactants, named off‐stoichiometry thiol‐ene (OSTE), especially employed to fabricate microfluidic devices and lab‐on‐chips (Carlborg et al., 2011). The click chemistry between azide and alkyne groups is an alternative coupling (HA bearing pendant alkyne chains and 4,4′‐diazido‐2,2′‐stilbenedisulfonic acid) (Zhu et al., 2021) (Fig. BOX 3.1b). The drawback of this strategy relies in the use of copper catalyst difficult to remove although recent improvements help to overcome it by means of beads of copper easily recovered. Thio‐Mickael coupling has also been reported with the reaction between an amphiphilic copolymer based on PEG terminated by a thiol unit and a methacrylate or maleimide reactant (Kooijmans et al., 2016). Finally, the isocyanate‐amine coupling was described with DSPE‐PEG‐NH₂ and mannose‐NCO (Fig. BOX3.1f) (Choi et al., 2019).

Branched polymers. Some functionalisations can act as nodes of macromolecular networks and generate crosslinked nanoparticles or chemical hydrogels. Various strategies of crosslinking have been envisaged with the copolymerization of methacrylate and dimethacrylate monomers, the use of four‐arm‐PEG‐SH and a bridging Ag salt (Lin et al., 2021), the reaction between the amine groups of PEI with F127‐polycitrate (Wang et al., 2021), base Schiff crosslinking resulting from the reaction of methyl cellulose modified by aldehyde pendant units and amine groups of chitosan (Wang et al., 2020; Wang et al., 2019). We noted the variety of chemical approaches.

Figure BOX 3.1 Chemistry used for chemical modification of polymers used with EVs or for coupling to EVs. The red bonds result from the coupling reaction (a–d) using unsaturated monomers in presence of thiol or RN₃ reactants, and (e–g) epoxy isocyanate and aldehyde units reacted with thiol and amine reactants.

Various macromolecular architectures

Amphiphilic block copolymers. The hydrophobic section of amphiphilic polymer is not always a lipid chain, it could also be a hydrophobic block polymer (Fig. BOX 3.2). For all, most of the time, their synthesis occurred by polymerization using a macroinitiator. The copolymers result from the polymerization of two or three co‐monomers reacting together to produce polymeric chains bearing both the two or three corresponding repetitive units. The statistical, alternative or block configuration of the copolymers are predicted by their reactivity ratio. The widespread composition is the block copolymers allowing segregation of phases and self‐organization in water for instance. Diblock copolymers (PLGA, poly(ε‐caprolactone‐co‐lactide)) (Liang et al., 2022), triblock copolymers such as the copolymers based on PEG and poly(propylene glycol) (PPO) sold under the trade name of Pluronic (the most used are F‐127 and 123) or PLGA‐b‐PEG‐b‐PLGA (Fig. BOX 3.1) are currently employed.

Non linear (ramified/branched) architectures. The examples of chemical modification of the previous part mainly concern linear polymers (Fig. BOX 3‐2A) but many other macromolecular architectures have been described to overcome EVs issues. The building of the architecture depends on the functionality (equivalent to number of reactive functions) of the precursors leading either to branched or linear structure. Considering this aspect, a common character between natural polymers is the polyfunctionality in amine groups (chitosan, gelatin), carboxylic acid groups (HA), alcohol groups (polysaccharides including pullulan) thus explaining their branched structure. However, it can induce some inconveniences such as a low solubility, a high viscosity. By contrast, most of the synthetic polymers have only few reactive sites along the backbone or in terminal position thus creating more linear structure. The most widespread of ramified architectures are the graft polymers which can be elaborated according to different strategies. The first one consists in the connection of a polymer chain to a main polymer chain bearing multi reactive units (grafting onto, Figure BOX 3.2.a). Another strategy lies in the use of a monomer that can be polymerized from reactive sites along a polymer chain changing the architecture from linear polymer to graft polymer by growing of the pendant chains (grafting from, Figure BOX 3.2.a). Both strategies have their advantages and drawbacks. Problem of reactivity between antagonist polymers or ill‐defined ramified polymer chains can take place. An elegant synthetic route has been investigated by S. Lathwal et al. using a macroinitiator based on DNA for ATRP. The functionalization of EVs was also undertaken by the ATRP process (Lathwal et al., 2021). Two strategies to engineer EVs polymer hybrids either tether preformed DNA block copolymers onto EVs membrane (‘grafting‐to’) using single‐stranded DNA or by grafting polymers directly from the exosomal surface (‘grafting‐from’) using double‐stranded DNA initiator. Finally, sophisticated architectures like star polymers composed of various number of arms (Lin et al., 2021) can be reported. Denser polymeric materials can also be prepared by building of a network (Figure BOX 3.2.b) based on either interactions between complementary functions located along the linear or ramified polymer chains or covalent bonds.

Figure BOX 3.2 (a) Macromolecular architectures of copolymers ranging from linear to star and graft compositions. The two strategies, grafting from and grafting onto, to elaborate graft copolymers were illustrated from reactive polymer in presence of monomer or polymer, respectively. (b) The two types of crosslinking were reported using covalent bonds for chemical crosslinking and interactions for physical crosslinking.

Therefore, this review, addressed to the EVs community, aims at identifying how polymers have already and could, in the future, contribute to circumvent some of the barriers existing in the EVs field. This state‐of‐the‐art, looking at the connections between EVs and polymers of different nature and structure, assessed all the different stages of EVs development from production, enrichment and storage to therapeutic use including surface modifications and scaffolds association (Figure 1).

FIGURE 1.

Examples of connections existing between EVs and polymers at different steps of EVs development. A broad diversity of polymers is used at each step of EVs life cycle, as cell scaffold (2D or 3D), for cell transfection, for EVs collection, purification, separation, to facilitate EVs detection, and in a set of strategies for EVs therapeutic applications (surface modification, scaffolds association). PEG, poly(ethylene glycol); PS, polystyrene; PEI, poly(ethylene imine); PLL, poly(L‐lysine); PET, poly(ethylene terephthalate); PP, poly(propylene); PES, poly(ethersulphone); PVDF, poly(vinylidene difluoride); COC, Cyclic olefin copolymer; PEO, poly(ethylene oxide); PPO, poly(propylene oxide).

2. INPUT OF POLYMERS IN EVS PRODUCTION, ENRICHMENT AND DETECTION

2.1. EVs production

2.1.1. Polymer based 3D scaffolds for EVs production

To isolate EVs from cell culture, static two‐dimensional (2D) cell culture conditions are frequently used. In this context, polystyrene (PS) has been a fundamental substrate for the culture of adherent animal and human cells for several decades due to its optical clarity, relative ease of manufacture and low production cost. PS has largely replaced glass for cell work, for example, for imaging (Lerman et al., 2018). Nevertheless, the 2D production is not easy to scale‐up because of reproducibility issues, high surface required to expend high cell number, and long processing times leading to high production costs. Thus, bioreactors, such as stirred suspension flasks or tanks, can be used to maintain cells in suspension and allow medium homogenization. With adherent cells, microcarriers are required to allow cell survival and to increase available surface area (Jorgenson et al., 2018; Roberts et al., 2019; Yuan et al., 2014). This culture option offers the opportunity of scalability and working with high cell culture volumes. These systems, initially employed to amplify cells for cell therapy, were also used to support cells that will produce EVs. Currently, it is mainly commercial microcarriers of different nature (dextran, gelatin, coated polystyrene) which are often used, that is, Cultispher^® Macroporous gelatin‐coated microcarrier beads (Cytiva); Cytodex^® dextran microcarriers (Cytiva), Synthemax^® II coated PS microcarriers (Corning), SoloHill^® Star plus positively coated PS microcarriers (Pall Life Sciences). In general, studies have reported an increase in the amount of EVs produced per cell in agitated bioreactor compared to static 2D conditions, irrespective of the nature of the microcarriers (Coffin et al., 2021; de Almeida Fuzeta et al., 2020; Gazeau et al., 2017; Haraszti et al., 2018; Phelps et al., 2022). Interestingly, when reported, the biological activity of EVs produced in 3D bioreactors tends to be higher than when the EVs are produced under 2D static culture conditions (Haraszti et al., 2018, Patel et al., 2019).

Besides commercial microcarriers, other polymer‐based materials were used as scaffolds to allow EVs production. Molly Stevens's group designed a scaffold produced by stereolithography of E‐Shell 300 (acrylate‐based material) to obtain 3D‐print scaffolds with 0.5 mm small pillars spaced 1 mm apart in a multi‐layered shelf geometry (Patel et al., 2019). This scaffold was coated with a gelatin solution to allow human dermal microvascular endothelial cells (HDMECs) seeding. Cells were cultured on this 3D scaffold under static or dynamic (4 mL/min) conditions. Interestingly, a 10,000‐fold increase in EV production yield (NTA) was observed after 3 days under dynamic conditions compared to static ones and 2D flasks. This tendency of increased EV production by applying mechanical forces (direct flow stimulation and mechanical stretching) was also observed with other cell source [dental pulp stem cells (DPSCs), mesenchymal stem cells (MSCs), or skeletal muscle cells (SkMCs)] seeded onto PDMS elastic scaffolds (Guo et al., 2021). Indeed, when PDMS was covered with fibronectin, DPSCs produced 11‐fold higher EVs yields when a mechanical stretching was applied (vs. non‐stimulated cells on the same scaffold). When a 1 mL/min flow was applied a 37‐fold higher production was found.

It is well‐known that polymers can act as pertinent cell scaffolds that can modify supported cell behaviour, including their secretive properties. Regarding previous examples, it seems that the variety of polymers (in terms of nature, mechanical properties, surface chemistry, shape) can be modulated in order to stimulate EVs production from the supported cells. Looking at polymer based microcarriers and scaffolds for EVs production, synthetic polymers are often coated with natural polymers or polymer fragments (gelatin), to increase cell adhesion and survival. Indeed, the drawback associated with synthetic polymer surface is that the cell adhesion is generally low. Therefore, it requires a subsequent coating step with a natural or a positively charged polymer. As an example, hMSC were reported to adhere to PLGA microspheres to a lesser extent than to poly‐D‐Lysine or fibronectin coated PLGA microspheres (Delcroix et al., 2011; Garbayo et al., 2009; Kandalam et al., 2020; Karam et al., 2015; Morille et al., 2013; Morille et al., 2016; Rmaidi et al., 2021). In this context, the use of natural collagen microcarriers, as developed in our team (Azria et al., 2018; Salvador et al., 2022), could be a more biomimetic option to explore. Of course, natural polymers also expose limitations regarding mass manufacturing and interbatch reproducibility. So that to date, it still represents a major limitation of collagen use for mass production of microcarriers. Interestingly, even if the influence of the 3D microcarrier composition on the nature of produced EVs is probably an important factor, to our knowledge it has not been extensively explored till now. Additionally, regardless of their composition, such microcarriers could be an interesting option to modify EVs secreting cells and therefore EVs content, through association and controlled release of a set of biomolecules. In this context, several groups, including ours, demonstrated that scaffold properties or encapsulation of biomolecules into scaffolds could influence behaviour of cells supported by these scaffolds (Caldwell et al., 2020; Morille et al., 2013; Morille et al., 2016; Morille et al., 2015; Qazi et al., 2017; Raisin et al., 2016; Salvador et al., 2022). Du et al. cultured MSCs with a chitosan‐based hydrogel encapsulating and slowly releasing nitric oxide (NO) (Du et al., 2017). Indeed, NO molecules are known to elevate the expression of proangiogenic cytokines in MSCs and to promote an angiogenic response. However, NO instability prevents its use as proangiogenic treatment. To circumvent this issue, the encapsulation method was investigated. Interestingly, the EVs produced in these culture conditions showed improved proangiogenic properties. Thus, this strategy based on the design of supporting biomaterials to control EVs content is a particularly interesting one, which has not yet been widely explored.

2.1.2. Treatment of producing cells with polymer‐based nanoparticles to control EVs composition

Nanoparticles for priming

During the production of EVs from cultured cells, different stimuli (inflammatory cytokines treatment, hypoxia, applying mechanical stress or even FBS depletion) (Najar et al., 2018; Sheng et al., 2008) can be used to increase the production yield or modify their content. In a few cases, polymers were used as molecules used to ‘prime’ producing cells. Indeed, it is well known that nanoparticle (NP) interactions with the cell surface could result in cell stimulation, such as the triggering of cytokine production (Cui et al., 2005; Elsabahy & Wooley, 2013). In the same way, it could increase EVs production yield. Emam et al. (2018) showed that treatment of tumour cells (B16Bl6 and C26 cell lines) with synthetic lipid nanoparticles (liposomes) of different lipid compositions (more than 10 formulations) associated or not with PEG were able to influence EVs production. Interestingly, whatever the lipid composition, PEG polymers were shown to suppress secretion of EVs from cancer cells, probably by decreasing cell stress due to NP endocytosis. However, it is important to note that EVs quantification was done by dosage of total protein assay after isolation with either a precipitation kit or a differential ultrafiltration protocol (without any purification step). Therefore, it is possible that the stress induced by the treatment with these nanoparticles has an impact on total protein secretion, without necessarily being selective for EVs. Nevertheless, the authors highlighted that treating tumours with PEGylated liposomes can limit EVs production from tumour cells, as these tumour vesicles were reported to induce tumour progression.

Internalization of positively charged iron oxide NPs encapsulated in poly(lactic‐co‐glycolic acid)‐b‐poly(ethylene imine) (PLGA‐PEI) by MSCs were also shown to increase EVs secretion (Park et al., 2020). Indeed, the increase of EVs release (evaluated regarding expression of CD81, CD9, CD63 by Western Blot) was reported especially when positively charged NPs (PLGA‐PEI +35 mV) were used to treat cells. Nevertheless, it is important to notice that an important cell viability decrease (∼40%) was observed after positive NP treatment. The authors reported that the production of EVs was probably due to the amplification of autophagy by the NPs introduced into the endosome. This resulted in an increase in autophagosomes and the accumulation of MVB in the cells. Thus, despite the fundamental interest of this study, this method seems difficult to apply to produce EVs for therapeutic application.

Therefore, priming EVs with polymer nanoparticles to increase production level is a strategy that has been rarely used to date, as the benefits in terms of production are difficult to measure. The increase in EVs production is probably a sign of cell stress. Thus, the EVs content is maybe not as beneficial as the one found with cell non‐treated with polymer NPs or this parameter should, at least, be checked. In addition, the NPs used for the treatment must be eliminated for the EVs production phase. Otherwise, depending on their size, they could be identified and counted as EVs, leading to a bias in the interpretation of the results.

Treatment of producing cells with polymers to modify EVs content and surface

To allow exogeneous biomolecules loading into EVs, two main strategies can be carried out (i) transfection of producing cells to obtain modified EVs, which is called ‘pre‐production loading’ or (ii) modification of EVs after isolation or ‘post‐production loading’ relying on physicochemical methods. The first strategy, relying on manipulation of producing cells, requires an efficient nucleic acid vector. While viral or lipid‐based reagents are frequently used for cell transfection, poly(ethyleneimine) (PEI) is generally the gold standard polymer. Thus, it was recently used for the transfer of plasmid DNA encoding a set of fusion proteins (GFP, nanoluciferase) (Heath et al., 2019), for gene editing proteins Streptococcus pyogenes (Sp) CRISPR (Clustered Regularly Interspaced Short Palindromic Repeats) associated protein 9 (Cas9) nuclease (SpCas9) spCas9/sgRNA (Lainšček et al., 2018) or Cre CRISPR (Somiya & Kuroda, 2022; Zhang et al., 2020) recombinase for gene edition. Transfection could also be performed to modify EVs surface proteins to control their biodistribution. This strategy generally lies in creating fusion proteins to associate targeting moieties to EVs membrane proteins such as LampB2 (Alvarez‐Erviti et al., 2011; Bellavia et al., 2017; Mentkowski & Lang, 2019), CD63 (Stickney et al., 2016) or lactadherin. Thus, to generate pancreas targeting EVs, with PEI as a transfection reagent, C1C2 domain was fused to p88, a peptide with known affinity for the human and mouse membrane small ion transport regulator (FXYD2)γa (Komuro et al., 2021). The same transfection reagent was recently employed to create EVs exposing monoclonal antibodies targeting the Epithelial Growth Factor Receptor (EGFR) at their surface (Komuro et al., 2022). Finally, still aiming at modifying the EVs surface, a high amount of DSPE‐PEG‐Biotin was added to treat producing cells (Wang et al., 2017). After isolation with microfluidic chips, some EVs were evidenced to be positive for biotin, with no information on the actual yield of this modification. However, this strategy seems difficult to control and requires large quantities of polymers.

2.2. EVs enrichment

One of the main challenge in the EVs field is to obtain a pure, reproducible population of EVs from cell supernatants or biofluids. In this context, numerous kits have been launched in the market. Depending on the EVs source (biofluids vs. cell culture supernatant, cell type, biofluid origin), production conditions (presence or absence of serum, depleted or not with EVs) and final use of isolated vesicles (diagnosis or therapeutic use), different protocols can be performed. EVs can be isolated based on their size, density or affinity to specific molecules, such as surface antigens, enriched or not in some EVs subpopulations.

2.2.1. Enrichment in whole EVs suspensions

Besides the historical gold standard differential ultracentrifugation (dUC) (Théry et al., 2006), isolation methods based on ultrafiltration microfiltration devices or tangential flow filtration and/or size exclusion chromatography (SEC) were increasingly implemented as all resulted in a good equilibrium between EVs yield and purity of resulting suspensions (Staubach et al., 2021). SEC stationary phases applied to isolate EVs all consisted in crosslinked polymers: dextran (Sephadex^®), agarose (Sepharose^®), polyacrylamide (Bio‐gel^® P), or allyldextran (Sephacryl^®) (Huang et al., 2020; Liangsupree et al., 2021; Sidhom et al., 2020). Filtration, ultrafiltration, microfiltration or tangential flow filtration rely on filter membranes, generally composed of cellulose but more recently on poly(ethersulfone) nanomembranes defined by molecular mass or size exclusion range pore size (Molecular weight cut off (MWCO) in Da or nm). The MWCO is selected to concentrate EVs and discard contaminants (such as proteins). Separation efficiency is greatly dependent on the quality of filter membranes and the uniformity of the membrane pore size distribution. The advanced version of filtration is the tangential filtration. It is particularly interesting to avoid clogging as it is one of the main limitations of this method. Thus, hollow fibre filter modules used for tangential flow filtration were generally constituted of polymer membranes (polysulfone, poly(ether sulfone)). Nevertheless, filtration is poorly applicable to complex suspension as high protein content can induce membrane clogging. This approach is more adapted to EVs isolated from cell culture supernatant than to isolation from complex biofluids leading to low recovery rate and/or low separation efficiency.

In parallel to these size and density isolation protocols, a set of polymeric reagents (based on PEG) has been commercialized (e.g., ExoQuick™ or Total Exosome Isolation™) to isolate EVs via polymer‐based enrichment. This type of kit probably represented the first “historical” source of interactions between polymers and EVs. ExoQuick™ precipitation solution is typically prepared as a 50% by weight solution of PEG, for example, PEG‐8000, although lower concentrations ranging from 30% to 50% can also be employed (Antes et al., 2013). The liquid sample and the precipitation solution are most typically combined in a volume ratio of about 5:1. Interestingly, as reported in ExoQuick™’s patent, the PEG polymer does not appear to require any particular purity, and the choice of purity is dependent on the user requirements. Regardless of the purity, the average molecular weight is defined, PEGs with molecular weights (MW) between 6000 and 10,000 Da are mostly used for the precipitation of EVs. Indeed, PEG MW influences the precipitation: the higher the MW of PEG, the lower the concentration required for precipitation, as evidenced with virus precipitation (Polson et al., 1964; Yamamoto et al., 1970). PEGs of higher MW (PEG 20,000 and PEG 40,000) cannot be selected because they give highly viscous solutions, which render centrifugation and other manipulations difficult. Variation in PEG MW can also optimize the precipitation of EVs according to their origin. As an example, among EVs isolation reagents proposed by System Biosciences, a precipitation solution comprising PEG 8000 is recommended for the isolation of microvesicles from urine or conditioned cell culture media (ExoQuick™), whereas for the isolation of EVs from blood serum or other body fluids, a precipitation solution comprising PEG 10,000 is preferred (ExoQuick™). Nevertheless, it should be notified that the theoretical rationale is not clearly described and is probably not based solely on PEG MW. PEG has the advantage over polyelectrolyte polymers that it usually interacts more strongly with biological materials than other non‐ionic polymers. This method, based on PEG precipitation, has proved its value for virus isolation for many decades now. Indeed, the use of PEG to precipitate rather than to partition viruses between two immiscible aqueous polymer phases was reported by Hebert (1963). They successfully concentrated both rod‐shaped and spherical plant viruses with a solution of PEG and sodium chloride. pH and NaCl concentrations were evidenced as key parameters influencing precipitation (Leberman, 1966; Polson et al., 1964). If the pH is close to the isoelectric point (pI) of the virus, lower concentration of PEG will be required to trigger precipitation. At a fixed pH value, the higher the NaCl concentration is, the lower is the PEG concentration required for precipitation (Leberman, 1966). Despite the number of data available for more than 75 years now, the mechanism of precipitation of particles by PEG remains unclear. Several teams proved that it cannot be explained by the formation of complexes (Yamamoto et al., 1970; Zeppezauer & Brishammar, 1965). Experiment with PEG radiolabelling showed that the PEG content of the precipitated particles does not exceed 2% of the PEG awarded, similarly to the 2.1% of liquid volume trapped by the precipitate. The hypothesis based on the precipitation of particles by PEG results from the particle dehydration is not supported either. Indeed, if this hypothesis was right, it might be expected that lower MW PEG, on account of their higher osmotic pressure, would dehydrate to a greater extent than higher ones. However, the opposite occurs, since low PEG concentrations of high MW are required to precipitate particles. One of the main accepted hypothesis is that PEG sterically excludes particles from part of the solution (Iverius & Laurent, 1967). Based on this assumption, the amount of excluded volume depends solely on polymer concentration and is independent of particle concentration or any other factor. The results obtained by Yamamoto et al. (1970) confirmed this hypothesis. Indeed, between a particle concentration of 10⁵–10¹³ particles/mL, they observed that bacteriophage precipitation efficiency did not change significantly using the same PEG concentration. The correlation between particle size and PEG concentration also supports the exclusion hypothesis. However, if exclusion is probably the prevailing phenomenon, other factors may also be involved. It is well established that the volume occupied by a ‘protein’ particle depends on its polyelectrolyte character, its current conformation as well as its solvation. All these elements are influenced by ionic strength, pH and temperature. Thus, the extent of the excluded volume is a function of polymer concentration (according to the exclusion theory) but at a fixed polymer concentration it might be relatively small or large for a given particle, depending on the conditions affecting the volume of the particle. All these theories, coming from the virus field, are totally applicable to EV's one. This also evidenced how, depending on EVs sub type and therefore protein content, some particles can be precipitated whereas others cannot. To conclude with the PEG precipitation method, this strategy has the advantage of not requiring expensive or specialized equipment, nor any particular technical knowledge. Moreover, this method could be seen as less detrimental to EVs integrity compared to dUC (Andreu et al., 2016). Nevertheless, as this method was initially employed to isolate protein and/or viral particles; it therefore induces co‐isolation of contaminating molecules or particles (albumin, immunoglobulins, viral particles, immune complexes) (Atha & Ingham, 1981). Reproducibility issues could also occur and the obtained EVs suspensions generally present some remaining PEG molecules associated to EVs as well as EVs aggregation (Konoshenko et al., 2018; Li et al., 2017; Salih et al., 2014; Taylor & Shah, 2015).

In parallel to ‘PEG only’ based precipitation, another method exists for separating particles by means of high‐molecular‐weight polymer solutions. This alternative strategy consists in using a liquid‐liquid two‐phase system with different phase volumes by adding two different high molecular weight water‐soluble polymers to the particle suspension. Separation then takes place thanks to an unequal distribution of particles between the phases, depending mainly on the hydrophilicity and lipophilicity of the particles (or molecules) to be isolated. If the composition of the smaller‐volume phase is preferred by the particles, separation and concentration will occur (Albertsson, 1970). Such method refers to a system known as aqueous two‐phase system (ATPS). It can be obtained by mixing PEG and salt, or two incompatible polymers such as PEG and dextran. Phase separation appears above a critical concentration resulting in two aqueous phases, each enriched with one of the components. In regards with the soft conditions used (low interfacial, tension, high water content) and its simple extraction procedure, this system has been widely employed for extraction, separation, purification, and enrichment of biomolecules (Iqbal et al., 2016), including viruses and more recently EVs (Kırbaş et al., 2019; Shin et al., 2015; Slyusarenko et al., 2021). Following the ATPS precipitation method, EVs partition was mostly affected by polymer hydrophobicity, rather than by the polymer MW. Thus, even if PEG is generally described as a hydrophilic polymer, it is less hydrophilic than dextran. Moreover, it is well‐known that the hydrophobicity of PEG and the hydrophilicity of dextran increase in proportion to their molar mass. Therefore, as EVs were reported to expose a rather hydrophilic surface (Desmet et al., 2017; Iqbal et al., 2016), in a PEG‐dextran mix, they present a higher partition preference towards the dextran‐rich phase. Interestingly, ATPS isolation method exposed higher recovery yield of EVs compared to other methods (77% vs. <16% for dUC and 40% for ExoQuick™) (Shin et al., 2015). Unfortunately, contaminant proteins were reported to expose the same repartition behaviour. Worthy of note, the repetition of the isolation steps could avoid this co‐repartition and increase the purity of the isolated EVs (Kim et al., 2015). Kırbaş et al. (2019) further optimized the ATPS based dextran workflow by adding washing steps, and fully characterizing obtained EVs following ISEV guidelines (Théry et al., 2018). They evidenced that a two‐step washing process with distilled water was able to remove all contaminating proteins. In addition, it was also possible to purify EVs from different non‐protein contaminants such as fatty acids or phenol red, by contrast to traditional EVs purification methods. This ATPS isolation‐based method is therefore particularly promising, especially to create new kits for isolating EVs from complex fluids. However, the potential adverse effects of the dextran present in the EVs‐rich phase should be investigated in the future, as this polymer can interfere with common analytical techniques such as western‐blot or RNA isolation, and therefore hinder characterization of obtained EVs.

2.2.2. Enrichment of EVs subpopulations

When EVs sub‐classes isolation is required, immunoaffinity techniques should be performed. Currently, immunoaffinity is generally associated with (magnetic) beads (i.e., MACS Technology based EVs isolation kits, Miltenyi Biotec) or biosensor surface (see following Section 2.3). These methods can be efficient and selective when minimal EVs amounts are required, but they generally result in low EVs recovery rates and are not well suited for scalability due to their high cost. Indeed, conventional immune based EVs capture relies on beads or chromatography phase functionalized with ‘conventional’ monoclonal antibodies exposing elevated production costs and low homogeneity after functionalization (Bradbury & Plückthun, 2015; Nakai et al., 2016). Moreover, these limitations are amplified by the fact that a mix of monoclonal antibodies is generally required. Alternatives are currently under investigation such as the use of single domains antibodies (VHH). It gathers many advantages as it is simple to engineer, to evaluate and to optimize with in silico modelling due to their small sequence. To associate VHH to affordable chromatographic matrix, a copolymer based on glycidyl methacrylate monomer was chosen relying on the chemical versatility of its epoxy groups allowing VHH coupling on its primary amine groups (via glutaraldehyde linkers on dry aminated copolymer) (Filipović et al., 2022). A mix up to five different VHH constructs were added to the polymer to generate a polyclonal matrix. EVs suspensions from varying complexities (HEK 293 cell culture supernatant's or plasma) were evaluated. A combination of moderate binding affinity and low VHH density led to efficient EVs capture, but above all enabled their release under gentle conditions. Interestingly, no lipoprotein markers were identified in isolated EVs. This is an example of the interest of polymers to offer a cost‐effective and simple solution for immunoaffinity methods development.

Aside from protein expression, it is now well established that carbohydrates, conjugated to protein or lipids present at EVs surface are others important key players in EVs functions. Also worthy of note that EVs glycans can be a novel source of EVs biomarkers (Martins et al., 2021; Royo et al., 2019; Williams et al., 2018). Thus, a polymer‐based EVs precipitation approach was modified to select specific sub population of EVs depending on their surface glycans content thanks to hydrazide functionalized thermosensitive polymers. This method allowed to recover N‐glycosylated EVs from complex samples of human plasma containing abundant unglycosylated protein and/or N‐glycoprotein not associated to EVs (Bai et al., 2018) (Figure 2a). The strategy relied on hydrazide chemistry to oxidize hydroxyl groups on oligosaccharides structure of N‐glycosylated proteins in aldehydes, using sodium periodate. These aldehydes will then be covalently coupled with hydrazine groups. This mode of action is a known basis to functionalize solid phase (gold or silica or magnetic particles) in hydrophilic interaction liquid chromatography. To avoid steric hindrance and high mass transfer resistance found in such solid phase extraction of glycoproteins, the authors investigated a pH responsive soluble enrichment method for N‐glycosylated proteins. To do so, they copolymerized thermosensitive monomeric units, N‐isopropylacrylamide, to methyl acrylate easily modified in a second time by hydrazine. The soluble polymer formed a liquid phase homogeneous system to facilitate interaction with N‐glycosylated proteins at room temperature. After coupling to the oxidized glycoproteins, the copolymer was precipitated by decreasing temperature above its LCST (34°C). Thus, the resulting precipitate could then be recovered by a low‐speed centrifugation (10,000 g), just as with PEG precipitation. After supernatant removal and washing steps to discard nonspecific protein attachment, the conjugates were re‐dissolved in conditions for EVs release from polymer. This strategy aimed to determine glycosylation profiles of EVs isolated from healthy or glioblastoma patient plasmas, allowing identification of 25% more N‐glycosylated proteins‐EVs compared to commercial hydrazine beads. Twenty‐six N‐glycosylated proteins over‐expressed in the glioma patient group were identified highlighting the interest of this approach for diagnosis.

FIGURE 2.

Schematic representation of different polymer's inputs in the field of EVs detection/enrichment (non‐exhaustive list). Polymers are used for a wide range of applications in the field of EVs detection as they can induce precipitation of specific subpopulation of glycosylated EVs (a) or be used to create super resolution probe (b), as well as to increase the specificity of immunocapture devices sensitivity (c–f). LSCT, lower critical solution temperature; poly(NIPAM)‐co‐hydrazide; CN‐PPV, poly[2‐methoxy‐5‐(2‐ethylhexyloxy)‐1,4‐(1‐cyanovinylene‐1,4‐phenylene)]; Zn‐TPP, zinc‐tetraphenylporphyrin; PCBM, phenyl‐C61‐butyric acid methyl ester; CN‐PDHFV, poly[(9,9‐dihexyl‐2,7‐(2‐cyanodivinylene)‐fluorenylenyl‐2,7‐diyl); Pd‐TPBP, meso‐tetraphenyltetrabenzoporphine palladium; ZrOCl₂, Zirconyl chloride; SA‐g‐PE, Sodium Alginate grafted Glycidyl Propargyl ether; P(DES‐BFPA), poly(bis‐ferrocenoyl propargylamide); EGFR, Epidermal Growth Factor Receptor; mAb, monoclonal antibody; RAFT, Reversible Addition−Fragmentation chain‐Transfer Polymerization; ROP, Ring‐opening polymerization; EDOT, 3,4‐ethylenedioxythiophene; EDOTSAc, 2‐(2,5,8,11‐tetraoxadodecyl)‐2,3‐dihydrothieno[3,4‐b]; EDOTEG, 2,3‐dihydrothieno[3,4‐b] (Cosenza et al., 2017; van Niel et al., 2018) dioxin‐2‐yl) methyl) ethanethioate; EpCAM, Epithelial Cell Adhesion Molecule; HER2, Human Epidermal Growth Factor Receptor‐2; PEG, poly(ethylene glycol); PAMAM NP, Poly(amidoamine) dendrimer nanoparticle.

2.3. EVs storage

Pharmaceutical development of EVs also requires the identification of optimal storage conditions, since the systematic use of fresh EVs is a strong limitation to their clinical use. Until now, the MISEV2018 guidelines recommended the storage of EVs in PBS at −80°C in siliconized vessels to prevent adherence of EVs to surface at −80°C (Witwer et al., 2013). Regarding this EVs stability, several published works were found in contradiction to one another and sometimes speculative (Kalra et al., 2013; Lee et al., 2016; Tang et al., 2012). This was probably due to differences in vesicle type to evaluate stability, and read out procedures. For example, H22 tumour‐derived EVs were reported to be unaffected by sunlight exposure, acidic or alkaline conditions, shaking and temperature (room temperature and 37°C) (Tang et al., 2012). However, these results can be questioned as EVs were counted by flow cytometry through fixation on 3 µm latex beads, using an indirect gating strategy. Bosch et al. evidenced that the addition of trehalose was a valuable asset to stabilize EVs derived from beta pancreatic cells during freezing (Bosch et al., 2016). Our team also confirmed its value to protect MSC derived EVs and we showed that trehalose was also interesting as stabilizer for freeze‐drying (Saux et al., 2020). Since then, the addition of trehalose for cryopreservation was also proved to preserve the morphology and function of MSC derived exosomes and macrovesicles (Budgude et al., 2021). The impact of lyophilization on EVs was further investigated by Charoenviriyakul et al. who demonstrated that EVs lyophilized with trehalose and stored at room temperature did not modify the composition, physico‐chemical properties and function of protein and DNA loaded (Charoenviriyakul et al., 2018). Overall, trehalose is the most common polymer used as cryoprotectant but mannitol also belonging to the sugar family was tested (Bari et al., 2019; Bosch et al., 2016). Several reviews detailed the impact of storage temperature on EVs quality and functions and the different options and polymers available such as PEG or polyvinylpyrrolidone (PVP) to preserve the EVs (Bahr et al., 2020; Jeyaram & Jay, 2018; Kusuma et al., 2018; Qin et al., 2020; Yuan et al., 2021). More recently, El Andaloussi group highlighted the importance of proteins such as human albumin combined with trehalose to stabilize EVs during storage short and long terms (Görgens et al., 2022). In parallel to proteins, that can be seen as natural statistic polymers, the use of surfactant molecules (PEG based nonionic surfactant) was a good option to stabilize therapeutic proteins but until now it was rarely used to stabilize EVs. However, it could be interesting to consider it as it can avoid EVs adsorption onto collection tube walls (until 22% in recovery improvement after freezing) (van de Wakker et al., 2022).

Polymer coating on vessels such as with low binding coating can also help avoiding adherence of EVs just as it did for proteins. Indeed, the literature found on this matter for proteins, can be adapted to EVs as the same interactions can apply due to EVs composition. Proteins can dynamically be adsorbed on material surfaces leading to their denaturation. Therefore, polymers are used to coat material surfaces by providing a hydration layer to decrease protein adsorption and preserve their integrity. To evaluate the potential of polymers to create this hydration layer, Whitesides research group elaborated ‘rules’ to consider: (i) presence of a polar functional group, (ii) presence of hydrogen bond acceptor group, (iii) absence of hydrogen bond donor group, and (iv) the absence of net charge (Chapman et al., 2000; Ostuni et al., 2001). Among the potential hydrophilic polymers selected for that matter, we can find poly(ethylene glycol), polyglycerols, poly(propylene sulfoxide), polyoxazolines, polyvinylpyrrolidones or zwitterionic polymers (Rabinow et al., 1995; Wei et al., 2014). Nevertheless, the recent studies comparing EVs recovery (particle and protein content) after short (7 days) and long (2 years) term storage in regular polypropylene microtubes, versus DNA LoBind^® (Eppendorf) or low protein binding (Thermo scientific) ones, did not evidenced any significant difference in particle or protein recovery (Görgens et al., 2022; van de Wakker et al., 2022). The only significant particle lost was found when EVs were stored in glass tube, non‐coated with Sigmacote^® (chlorinated organopolysiloxane forming a hydrophobic film on glass tubes). This last treatment led to a recovery similar to the one obtained with standard polypropylene tubes.

2.4. Detection and selection of specific EVs subtypes

2.4.1. Polymers in the design of microfluidic chips

Immunoaffinity on‐chip EVs isolation could improve recovery yields, decrease sample volume requirements as well as processing times compared to standard protocols such as dUC. Until now the chips employed, such as ExoChip™ were manufactured with the ‘gold standard’ microfluidic polymer, PDMS (Kanwar et al., 2014; Zhang et al., 2016). It was functionalized with a set of biomolecules (anti CD63 antibodies, phosphatidyl serine (PS)‐specific protein) offering the selection of specific EV subsets. PDMS has also been functionalized with graphene oxide polydopamine nano‐interface to increase EVs captured on surface to reach a limit of detection (LOD) of 1.10⁶ EV/mL (Zhang et al., 2016). Even if these devices represent interesting options especially in terms of detection limits, PDMS chips as diagnostic tools face large‐scale production issues. Alternatively, PDMS‐free microfluidic devices were investigated based on cyclic olefin copolymers (COC) using off‐stoichiometry thiol‐ene chemistry (OSTE). Thermoplastics can also be used as they are more prone to mass‐production and easy to modify regarding surface chemistry (Bajo‐Santos et al., 2023; Jackson et al., 2014; O'Neil et al., 2016; ONeil et al., 2016). Thus, microchips made of COC were manufactured by micro‐injection moulding to create herringbone structure. Such herringbone‐chips were shown to capture 60% more EVs (tumour‐specific EVs‐RNAs) compared to flat surface, leading to a LOD of 11.10⁶ EV/mL. Nevertheless, a limitation of such microfluidic chips possessing herringbone mixers was the operation time. Indeed, the volumetric flow rate was usually around 0.5 µL/min, so it would require 3 h to analyse 100 µL of fluid, without considering EVs analysis time. To decrease process time, others design of COC microfluidic chips containing a dense array of micropillars with a 10 µm diameter and 10 µm spacing were created (Wijerathne et al., 2020). These devices, called EVs Micro Affinity Purification (MAP), were decorated with CD8 antibodies, to trap EVs released from CD8 (+) T‐cells, expressing this antigen. They were evidenced as key players in central nervous system homeostasis, stroke pathology, and subsequent stroke recovery. EVs‐MAP were therefore useful for diagnosing acute ischemic stroke patients. Thus, operating at 10 µL/min, these devices were shown to be able to process 100 µL of plasma in 10 min. Thanks to the association with gene profiling via droplet digital PCR, differences in mRNA profile of strokes patients were identified with a total processing time of 220 min which is a time compatible with clinical management of stroke side effects. Still in the context of microfluidic, but outside the scope of the chip manufacturing, the addition of a small PEG amount (0.1%) in the circulating media was shown to control the viscoelastic forces exerted in the media. This resulted in an 80% of EVs recovery with a 90% separation between large and small EVs (but no information on protein contamination) (Liu et al., 2017).

2.4.2. Polymers to improve immunocapture sensitivity, selectivity and modularity

To increase detection specificity and to avoid nonspecific adsorption on chips, the polymer printing technique allowed creating EVs binding cavities using PC3 derived EVs as a template (Mori et al., 2019). Gold chips were coated with a mixed self‐assembled monolayer bearing Nickel‐nitrilotriacetic acid and an ATRP initiator bearing 2‐bromoisobutyryl reactive group. EVs were then immobilized on this gold substrate already functionalized with CD9 antibodies (Figure 2c). Methacryloyl disulphide groups were then introduced onto EVs using a cell‐surface modifier bearing an oleyl group (to allow EVs membrane anchorage), a thiol group (to be transformed into a free thiol after the removal of the template EVs) and thiol‐reactive methacrylate (to allow ATRP with a biocompatible monomer 2‐methacryloyloxyethyl phosphorylcholine (MPC)) on a “grafting from” strategy (BOX 3.2A). This polymer matrix around antibodies‐EVs complexes, estimated at 20 nm thickness, avoid non‐specific interactions and retain free thiol groups that will be thereafter used to introduce reporter fluorescent molecules (Alexa Fluor 467). EVs template removal from CD9 IgG was then performed thus creating EVs‐binding cavities wherein free thiol groups were functionalized with Alexa 467. The linkage of EVs present on the suspension to be analysed on mAb will therefore induce a fluorescence quenching. Using such system, the LOD was estimated at 6 pg protein/mL, revealing an extremely high detection potential. The authors evidenced this with already purified EVs (isolated from PC3 cell line supernatants), but also with EVs in tear drops without previous isolation methods. Body fluids from cancer patients are currently being tested to confirm the reliability of this liquid biopsy technique. To our knowledge, no update was published until now. The same polymer imprinting strategy using EVs template was also engaged to trap EVs on magnetic Fe₃O₄ microparticles (Liao et al., 2022). To do so, carboxyl‐modified Fe₃O₄ magnetic microparticles were co‐incubated with EVs from human serum isolated by PEG precipitation, serving as an EVs template. Monomers of methacrylic acid, NIPAAm and acrylamide were mixed and polymerized using redox co‐initiators. EVs templates were then eluted before incubation of plasma samples onto microparticles exposing cavities complementary to EVs in terms of shape and chemical composition. Afterwards, a fluorescently labelled CD63 aptamer was added. Thanks to this method, the authors reported a LOD of 1.10⁶ EV/mL. Nevertheless, bound EVs were not fully characterized.

EVs secreted from lung cancer cells (A4549) (over‐expressing EGFR) were immobilized onto gold chips functionalized with anti‐EGFR mAb to create an electrochemical sensor able to translate the chemical signal into an electrical signal (Figure 2d) (Zhu et al., 2021). For enhanced sensitivity, signal amplification strategy based on polymerization amplification was carried out. Indeed, the polymerization mechanism, in which an initiation step drives the inclusion of many monomers in a continuously growing polymer chain, is in itself an efficient signal amplification pattern. When initiation is coupled to a molecular recognition event, it provides a means for the development of highly sensitive bioassays leading to a strong amplification of the electrochemical signal (Hu et al., 2020). In this study, sodium alginate (SA) grafted glycidyl propargyl ether (SA‐g‐GPE) was associated to EVs phospholipids via a ‘phosphate‐Zr⁴⁺‐carboxyl chemistry’ based on hydroxyl groups of SA. This initiated the ROP (see BOX 1) of GPE while alkyne groups on lateral chains of SA‐g‐GPE provided reaction sites for subsequent click reaction to graft electroactive polymers bis‐ferrocenoyl propargylamide. The authors reported that their method allows LOD of 1.49 × 10² particle/mL, outperforming classic EVs sensors (from 10⁵ to 10⁹ NP/mL). Even if ROP was performed at 60°C, which can be deleterious for EVs integrity, this study was particularly original and one of the two examples of polymerization from EVs.

Conductive terpolymers P(EDOT‐co‐EDOTSAc‐co‐EDOTEG) where EDOT means 3,4‐ethylenedioxythiophene, EDOTEG for 2,3‐dihydrothieno[3,4‐b] [1,4] dioxin‐2‐yl) methyl) ethanethioate, (EDOTSAc) for 2‐(2,5,8,11‐tetraoxadodecyl)‐2,3‐ dihydrothieno[3,4‐b] [1,4] dioxine were coated on carbon cloth substrate to form electrochemical devices able to capture and release EVs using a simple electrochemical redox process (Figure 2e) (Ashraf et al., 2022). The EDOTEG part enhanced the hydrophilicity and avoided nonspecific interactions with the substrate, while EDOT served as a spacer between the other two monomers. The EDOT terpolymer third arm with an acetylthiomethyl moiety (EDOTSAc) was able to electrochemically convert into a ‘free’ thiol sensitive thus leading to reversible oxidation/reduction cycles at +1.0 V and −0.8 V (vs. Ag/AgCl), respectively. This redox process allowed the capture of EVs and their release on demand. Carbon clothes of the substrate were first functionalized by [P(EDOT‐co‐EDOTSAc‐co‐EDOTEG)] electrodeposition. After reductive thiol activation and attachment of a thiolated linker, mercaptohexanoic acid, a HER2 specific antibodies were then conjugated via a reaction with 1‐ethyl‐3‐((3‐dimethylaminopropyl)‐carbodiimide) (EDC) and n‐hydroxysuccinimide (NHS). EVs from SKBR3 breast cancer cells line (HER2+) and MCF7 cell lines (HER2‐) were then co‐incubated (1 h, RT) on these substrates before being released by a ‘simple’ application of a potential of −0.8 V for 120 sec. SKBR3‐EVs were efficiently captured and released (70%) by contrast to MCF7‐derived EVs, indicating the specificity of the device. It would now be interesting to see if these promising devices could capture EVs in a complex suspension or in crude biological samples versus the isolated EVs suspensions gone through dUC and SEC.

Immunoaffinity based assay to capture tumour‐derived EVs, was designed by coating the capture surface with poly(amidoamine) (PAMAM) dendrimer nanoparticles (∼9 nm) inducing attachment of several antibodies (Poellmann et al., 2020). These flexible nanoparticle structures being deformable, allowed accommodation of binding domains and therefore increased in affinity. An epoxide functionalized glass side was first coated with partially carboxylated generation 7 PAMAM dendrimers (Figure 2f). Thereafter, a PEG mixture was coated to the dendrimers and/or to remaining free epoxide groups. This PEG mixture was composed of heterobifunctional PEG tethers for conjugation with another layer of dendrimers and 2 kDa methoxy‐PEG (mPEG) to block nonspecific adsorption. Finally, a second layer of carboxylated PAMAM dendrimers was added on PEG chains. These last dendrimers were functionalized with fluorescently labelled antibodies (rhodamine‐conjugated anti‐epithelial adhesion molecule (aEpCAM)). Enhanced binding was shown on a fully functionalized system (vs. bare glass, PEG only or single layer dendrimer) resulting in a 4‐fold increase in EVs capture for head and neck cancer patient plasma. Interestingly, this increase in flexibility could also avoid EVs standard deformation issues that can arise with classic immunocapture (LeClaire et al., 2021).

To improve immunocapture and decrease nonspecific interaction, PS beads were coated with a zwitterionic polymer (PMPC) constituted of methacrylates associated to a phosphoryl polar group linked to hydrophobic polymer units such as n‐butyl methacrylate (Yoshida et al., 2018). PMPC have been designed as membrane mimicking polymers and are frequently used as coating agents to avoid non‐specific binding of proteins onto material surfaces. To discriminate EVs from pathological tissues from the ones released by healthy cells in biologic samples, aptamers targeting EpCAM (marker of circulating tumour cells) were conjugated to PMPC. EVs were isolated to form EpCAM positive (HT‐29, HCT15) and negative (HEK293 and HT‐1080) cell supernatants and co‐incubated with PS surface functionalized or not with PMPC +/− EpCAM aptamer. Curiously, EVs were observed on uncoated PS plates, irrespective of the presence of EpCAM on them. PMPC coating significantly reduced this non‐specificity while PMPC‐EpCAM functionalized surfaces retained ‘substantial’ amounts of EVs EpCAM after two washings.

2.4.3. Polymers for EVs labelling probes stabilization and imaging

Another field of polymer application lies in optimization of EVs labelling probes for imaging or super‐resolution imaging. In this way, quantum dots, based on Cesium‐lead‐halide perovskite were associated to triblock copolymers PS‐b‐poly(ethylene‐stat‐butylene)‐b‐PS (PS‐PEB‐PS) and PEG‐PPG‐PEG to circumvent water solubility issues (Pramanik et al., 2019). This stabilization allowed water soluble perovskite quantum dots nanocomposites (∼20–25 nm) to be functionalized with anti‐CD63 to bind EVs and stained them for imaging.

Identification of different EVs subtypes is a major challenge in the field. To achieve high‐throughput super resolution mapping of protein on EVs surface, the optical properties of the super resolution probe must be optimized according to the structure. Indeed, due to the small size of EVs and their heterogeneity in tetraspanin expression, it is difficult to resolve EVs surface structure efficiently using conventional super‐resolution probes, which possess sufficient imaging resolution or low switch‐on frequency (Dempsey et al., 2011). To achieve high‐throughput super‐resolution imaging of EVs surface tetraspanins (CD9, CD63, CD81), Jiang et al. (2021) investigated a new class of photoswitching polymer dots (Pdot) based on the principle of N‐P‐N transistors. These Pdots (∼12 nm) were formed by nanoprecipitation of semi‐conductive polymers (poly[(9,9‐dihexyl‐2,7‐(2‐cyanodivinylene)‐fluorenylenyl‐2,7‐diyl)](CN‐PDHFV) or poly[2‐methoxy‐5‐(2‐ethylhexyloxy)‐1,4‐(1‐cyanovinylene‐1,4‐phenylene)] (CN‐PPV)) and metalloporphyrins. When the semiconducting polymers were excited by a 488 nm laser, it transferred electrons to phenyl‐C61‐butyric acid methyl ester (PCBM) and metalloporphyrin. Accumulated holes in the semiconducting polymer phase caused an almost complete quenching of the Pdot emission. By contrast, when metalloporphyrins were excited by a 405 nm laser, it transferred electrons to the semiconducting polymers, transiently reducing hole population in the polymer phase and increasing the duty of the Pdots. To adapt the signal in regards to protein level expression, different combinations of semi‐conducting polymers and porphyrins were chosen because they offered optimal optical properties and exposed different chemical structure and level of energy. Two types of Pdots were designed (Figure 2b): (i) Pdot 1 with CN‐PDHFV polymer doped with 10% meso‐tetraphenyltetrabenzoporphine palladium (Pd‐TPBP) and 20% PCBM, and (ii) Pdot 2 with CN‐PPV doped with 10% zinc‐tetraphenylporphyrin (Zn‐TPP) and 20% PCBM. Such Pdots, exhibiting spontaneous blinking and photoactivation response to 405 nm excitation, were conjugated to anti‐tetraspanin antibodies and used to label tetraspanins depending on their expression levels (i.e., CN‐PPV/10 % Zn‐TPP/20 % to label low‐expression CD81 and CN‐PDHFV/10% Pd‐TPBP/20% PCBM Pdots to label CD63). 3D super resolution images of seminal EVs estimated the average copy number of CD63, CD81, and CD9 to be 12.6, 1.6, and 16.6 per EVs, respectively. The exceptional brightness of the Pdots allowed the hollow structure of exosomes and the spatial distribution of the tetraspanins to be resolved with high precision (<5 nm uncertainty).

As illustrated by the various examples described above, the applications of polymers in the field of EVs detection are multiple. It is obvious that this field of application in detection is the one where the diversity of the polymer used is the greatest. Indeed, many different polymers are synthesized here to meet a specific need of the field. One of the major applications concerns their ability to increase specificity, based on antifouling properties to avoid non‐specific adsorption, to enhance antibody mobility. An interesting and innovative trend focuses on associating microfluidic with conductive polymers to amplify electrochemical signals and create sensitive electrochemical sensors. However, the major current challenge here is probably scalability, which should be tackled before using these technologies in an affordable way.

3. EVS SURFACE FUNCTIONALIZATION WITH POLYMERS FOR THERAPEUTIC APPLICATIONS

Despite their potential compared to synthetic nanoparticles such as liposomes (Le Saux et al., 2021; van der Koog et al., 2022), EVs also face pharmacokinetic issues to be handled before being used as rational drug delivery systems (DDS). As an example, the convincing results obtained by Kamerkar et al. to inhibit tumour progression with EVs carrying a siRNA targeting the oncogene KRAS require daily injections (Kamerkar et al., 2017; Mendt et al., 2018). This clearly illustrated one of the issues encountered with use of EVs in non‐autologous model, which is the poor plasma stability of non‐modified particles (Morishita et al., 2015; Smyth et al., 2015; Wiklander et al., 2015). Polymers can help overcoming this major issue, but the choice of the polymer strongly depends on its role towards EVs. To improve EVs stability after administration and/or to control EVs biodistribution, different strategies relying on EVs surface modifications have therefore been investigated (Figure 3). This will be described in the following sections.

FIGURE 3.

Example of EVs surface modification using polymers. Illustration of different strategies that can be or have been investigated to modify EVs surface: post‐insertion (a), fusion with synthetic particles (b) or layer by layer polymer association (c).

3.1. Lipopolymer anchorage

The most encountered surface modification to escape the innate immune system and increase plasma stability is the addition of PEG, known as PEGylation. This coating was used to improve nanoparticle steric stabilization and prolong blood half‐life by avoiding protein opsonization and uptake from innate immune cells. This increased blood concentration is generally correlated to the well‐known enhanced permeability and retention (EPR) effect, which promotes nanovector accumulation (due to imperfect tumoral neovessel epithelium) and retention (due to absence of lymphatic drainage) in tumour tissue (Maeda et al., 2000). However, mainly described in the 90′s, it is important to note that the EPR effect is now quite controversial. Indeed, pathophysiological heterogeneity of tumours and their vasculature, size differences between human solid tumours and those of murine tumour models, and waste production by tumours (thus clogging neovasculature gaps) are others explanations for the poor interindividual and inter‐tumour reproducibility observed in human (Fang et al., 2011; Harrington et al., 2001; Nagy et al., 2010). Altogether with a poor ability of synthetic nanoparticles to deliver their payload toward wanted cells (Wilhelm et al., 2016), this unreproducible EPR effect could explain the final rather disappointing results observed with PEGylated liposomes over these last years. However, the natural properties of EVs to diffuse in biological matrices and to be internalized by cells, would probably allow them to overcome this issue encountered with synthetic vectors.

Among the possible options for polymer association, EVs are generally combined with amphiphilic copolymer composed of a lipophilic part to allow anchoring in EVs membrane and a hydrophilic block which shapes a stabilizing corona (Figure 3a). This strategy was applied to EVs for the first time by Kooijmans et al. (2016). First, lipopolymers with two different anchors: 1,2‐Dimyristoyl‐sn‐glycero‐3‐phosphoethanolamine anchor (DMPE) for DMPE‐PEG₂₀₀₀‐maleimide and with 1,2‐Distearoyl‐sn‐glycero‐3‐phosphoethanolamine anchor (DSPE) for DSPE‐PEG₂₀₀₀ were mixed to form PEG micelles. These micelles were conjugated with nanobodies targeting epidermal growth factor receptors (EGFR) to create nanobody‐PEG₂₀₀₀ micelles. The resulting nanobody‐PEG₂₀₀₀ micelles were incubated at 40°C with EVs from mouse neuroblastoma cells. The nanobodies incorporation to EVs was confirmed by western blotting and then further characterized by TEM and NTA indicating the preservation of EVs integrity and morphology. After intravenous injection into tail vein of nude immunodeficient mice, PEGylated EVs showed an increased circulation time to 60 min compared to 10 min without PEGylation, but with no modification on the distribution in organs (liver, spleen, kidneys, lungs, brain, tumour). In another study, EVs tumour accumulation of naked versus PEGylated EVs (DSPE‐PEG₂₀₀₀) from C26 colon cancer cells (Exo‐C26) and B16BL6 melanoma cells (Exo‐B16BL6) was investigated (Emam et al., 2019). In vitro, while naked Exo‐C26 or Exo‐B16BL6 were more uptaken by their producting cells, this uptake was significantly (p < 0.001) decreased with PEGylation of the autologous EVs, proportionally to PEG concentration (ratio of EVs to PEG micelles from 1:10 to 1:100). Both EVs sources, with or without PEGylation were then intravenously (i.v.) injected in balb/c mice bearing C26 tumours. Interestingly, none of the naked EVs were able to reach the tumour, highlighting the importance of polymer shielding. Both PEGylated EVs (from C26 and B16BL6) accumulated in the C26 tumour from 4 to 96 h with higher accumulation for the autologous EVs. Looking at tumour cell uptake, the autologous PEGylated EVs showed significantly (p < 0.001) higher uptake in C26 cells, tumour associated macrophages and T‐cells. Despite evidencing the interest of EVs polymer coating, these results also illustrate the cell‐type tropism effect and its matter for tumour cell targeting. The same group recently focused on production of anti‐PEG IgM after single and multiple injections of PEGylated EVs with different EVs amounts (Emam et al., 2021). They proved that the i.v. administration of PEGylated EVs induced anti‐PEG IgM but not anti‐PEG IgG production by contrast to what was observed for PEGylated synthetic nanoformulations (liposomes, micelles, proteins). Nevertheless, the injection of a second dose of PEGylated EVs after 5 days led to the well‐known accelerated blood clearance, and to a more important relocalization of EVs in the liver versus tumours, has already observed for PEGylated liposomes (Dams et al., 2000). Therefore, this study highlights the need for in‐depth investigations of immunological responses to PEG for therapeutic use but also evidences differences in EVs versus liposomes behaviour even with a similar PEG coating. Shi et al. (2019) worked on EVs from 4T1 breast cancer cells conjugated with longer PEG chains (mPEG₅₀₀₀‐NHS). After i.v. injection on immunocompetent balb/c mice, PEGylated EVs exposed a prolonged blood circulation (7.9% injected dose (ID)/g at 1 h, 5.6% ID/g at 4 h and 1.8% ID/g at 24 h) whereas bare EVs remained around 1% ID/g whatever the time. Modified EVs also showed less liver uptake and higher tumour accumulation mostly due to the EPR effect. Also advocating for enhanced plasma stability due to EPR but with different type of vesicles, M1 macrophage derived EVs mimetic nanovesicles (MNV) obtained by cell extrusion method were tested for in vivo improved tumour targeting (Baek et al., 2022). The authors proved that PEGylated MNVs with DSPE‐PEG₂₀₀₀ enhanced their tumour uptake by 7‐fold after 24 h compared to unmodified MNVs. However, no proof of enhanced blood circulation over time and no quantification of injected MNVs were provided: only fluorescence tracking in vivo and ex vivo imaging and quantitative analysis of tumour/liver ratio was given. In another study focusing on tumour delivery, EVs from melanoma cells modified by DSPE‐2000 and loaded with doxorubicin (DOX) were intravenously administered in a mice model of B16.F10 murine melanoma (Patras et al., 2022). Interestingly, loaded DOX was efficiently delivered to tumour cells using PEGylated EVs compared to liposomal reference (PEGylated liposomes) inducing a higher antitumor activity mostly. This can also be explained by the cell‐type tropism effect as already observed by Emam et al. (2019). Lastly, an original rapid and on demand functionalization of EVs method was designed with PEG₂₀₀₀ based cholesterol‐modified DNA tethers and complementary DNA block copolymers (Lathwal et al., 2021). This creative strategy was further developed in BOX 3. This tethering method enhanced the EVs stability under various storage conditions and allowed precise control over EVs surface interactions, cellular uptake and preserved bioactivity. Worthy of note, 12 h after i.v. injection, PEGylated EVs showed increased circulation (24% of the injected dose) compared to unmodified ones cleared after 3 h. In addition, the plasma stability of cholesterol‐modified DNA tethered EVs was increased without altering the distribution profile in organs.

In parallel to their use for plasma stabilization of EVs described above, EVs surface modifications were also performed to control EVs biodistribution using targeting ligands (Table 1). Overall, looking at the literature of EVs post‐insertion with lipoPEG, in most cases ligands are added for specific cell targeting as summarized in Table 1. In the already described study by Kooijmans et al. (2016), DMPE‐PEG₂₀₀₀ and DSPE‐PEG₂₀₀₀ micelles were conjugated with nanobodies targeting the epidermal growth factor receptor (EGFR) via a thiol‐maleimide coupling (see BOX 3) to enhance specific binding to EGFR‐overexpressing tumour cells. In vitro, this surface modification allowed recovering cell interactions (Neuro2A and A451 cell lines) that were lost due to only PEG surface modification, and induced a significant uptake increase on EGFR expressing cells (A451). Unfortunately, in vivo, no specific tumour accumulation was observed even after injection of 2–5 × 10¹⁰ EVs/mouse in the tail vein. The biodistribution profile was unchanged compared to non modified EVs with a quick accumulation of EVs in elimination organs liver and spleen. The authors evocated a technical limitation due to the lack of sensitivity of the detection methods used in vivo that could have hindered the detection of EVs in tumours (fluorescent detection of DiR labelled EVs). In a different strategy, DSPE‐PEG₂₀₀₀‐mannose post‐inserted in EVs were used for a selective uptake in dendritic cells via the mannose receptor, CD206 (transmembrane protein primarily expressed in antigen presenting cells) and compared to DSPE‐PEG₂₀₀₀‐Biotin (Choi et al., 2019). In vitro, compared to control fluorescent probe only, DSPE‐PEG₂₀₀₀‐mannose modified EVs showed a 2.1‐fold increased dendritic cell uptake while in vivo an 1.8 fold enhanced lymph node accumulation was observed after subcutaneous administration. However, no significant differences were noticed compared to naked EVs on these two aspects. Similarly, DSPE‐PEG₂₀₀₀‐Folate was associated to EVs to target folate receptor targeting overexpressed on triple‐negative breast cancer cells (Yu et al., 2019). Thus, DSPE‐PEG₂₀₀₀‐folate post‐inserted EVs loaded with erastin as active molecule were able to inhibit proliferation and migration of cancer cells in vitro thanks to efficient delivery of erastin with folate PEGylated EVs. These promising results should be confirmed with an in vivo study. Embryonic stem cell derived EVs have also been post‐inserted with DSPE‐PEG functionalized with cyclo(Arg‐Gly‐Asp‐D‐Tyr‐Lys) peptide c(RGDyK) targeting 𝛼_v𝛽₃ integrin receptor overexpressed in tumour cells (Zhu et al., 2019). Also, when i.v. administered in mice, RGD functionalized EVs, loaded with anti‐cancer drug Paclitaxel (PTX), present a nearly two‐fold increased targeting ability toward an orthotopic model of glioblastoma cells (U87MG) compared to non‐modified EVs (in vivo fluorescence tracking of DiR labelled EVs). The therapeutic potential was then evaluated in the same model. After daily intravenous injection of the different EVs formulation, a Kaplan Meyer analysis clearly shown an increase in median survival time when mice were treated with PTX loaded EVs functionalized with RGD peptide (from 20 days survival with PBS injection to nearly 30 days with cRGD‐EVs‐PTX). Interestingly, PTX loaded EVs, without surface modification, even if in less extend, also led to an increase in mice survival (to 27 days). These results tend to confirm the natural ability of unmodified EVs to cross the blood‐brain barrier as well as a natural tropism to tumour cells. Nevertheless, even if these results can be seen as promising, the daily injection is clearly a main limitation in a therapeutic point of view. Still with the aim of cancer cell targeting, DSPE‐PEG₂₀₀₀‐hyaluronan was post‐inserted into milk derived EVs to target CD44 specific ligand overexpressed in a set of human cancer cell lines (breast cancer MDA‐MB‐231, MCF‐7 and lung cancer cells A549) (Li et al., 2020). A qualitative increased cell internalization was observed in vitro (compared to non‐functionalized EVs) only in MDA‐MB‐231 and A549 cells based on fluorescent microscope analysis. A lower expression of CD44 was proposed to explain the less important DOX accumulation in other cell lines. Thus, the anticancer activity of DOX was significantly more efficient once loaded in EVs, only on A459 at 10 ng/µL with a slight decrease (≈20%) in cell survival compared to the free drug. Lastly, DSPE‐PEG₂₀₀₀ was grafted with aminoethylanisamide (AA) ligand targeting sigma receptors overexpressed in lung cancer cells (Kim et al., 2018). Macrophages derived EVs post‐inserted with DSPE‐PEG₂₀₀₀‐AA showed a 2‐fold increased uptake in vitro compared to non modified EVs. in vivo, after i.v. administration, a co‐localisation of targeted EVs with pulmonary metastases was observed, in a more extent manner compared to non modified EVs. Finally, surface modified EVs loaded with PTX seemed to induce a spectacular decrease in the metastase area (≈ 90% decrease compared to PTX only). Nevertheless, it should be notified that EVs modification (surface modification and PTX loading), led to significant changes in EVs physico‐chemical properties, as a size increase from 110.8 ± 4.1 nm for naïve EVs to 280.8 ± 3.1 nm for modified ones was observed. Unfortunately, these values here questioned the nature of tested nanoparticles.

TABLE 1.

Example of targeting ligands used to functionalize lipoPEG.

Ligand	Grafting method to anchor	Targeted antigens	Targeted cells	References
Nanobodies (R2 and EGa1)	Conjugation by thiol‐ maleimide coupling	Epidermal growth factor receptor	Cancer cells (A431)	(Kooijmans et al., 2016)
Mannose	Chemical conjugation via NHS activated PEG	Mannose receptors	Dendritic cells	(Choi et al., 2019)
Folate	Commercial	Folate receptors alpha	Triple‐negative breast cancer	(Yu et al., 2019)
cRGDyK	Commercial	𝛼v𝛽3 integrin receptors	Glioblastoma (U87MG)	(Zhu et al., 2019)
Hyaluronan	Coupling hyaluronan with DSPE‐PEG‐amine with EDC/NHS	CD44 receptors	Cancer cells (MDA‐MB‐231, MCF7, A459)	(Li et al., 2020)
Aminoethylanisamide	DSPE‐PEG‐AA	Sigma receptors	Lung cancer cells (3LL‐M27)	(Kim et al., 2018)

To extend a more versatile platform of EVs functionalization, Antes et al. (2018) post‐inserted DMPE‐PEG₅₀₀₀‐streptavidin (STVDN) onto cardiosphere‐derived cells EVs. This STVDN terminal group allowed any biotinylated molecule, for example, fluorescent molecules (FITC, Quantum dots 655) but also targeting peptide (cardiac or ischemia homing peptide) to be coupled to the EVs. In vitro, the targeting effect seemed to be dependent on the couple targeting peptide—targeted cells. Indeed, STVDN associated with a muscle targeting peptide showed a significant 2‐fold increased uptake into myoblast cells (H2K mdx myoblast) while when an ischemic peptide was associated, only a low increase in uptake was observed in cells neonatal rat ventricular myocytes. However, when ischemic peptide modified vesicles (1.10⁸ EVs/animal) were i.v. administered in a rat model of ischemia, even if most of EVs type were localized in major elimination organs after 48 h, the EVs distribution in heart was slightly significantly higher with ischemic peptide EVs.

Finally, to conclude on the interest of EVs surface modification for targeting, if some targeting strategies have proven their efficiency in vitro, convincing results are still missing in vivo, mainly due to a major EVs localization towards liver and spleen. Therefore, before considering EVs targeting, the main challenge probably consists now in finding a solution to avoid mass elimination by the mononuclear phagocyte system.

3.2. Hybrid EVs and polymer containing nanoparticles

Another strategy for EVs surface modification is based on their fusion with synthetic vesicles or nanoparticles from different types (lipids or polymers). Fusion strategies, for various applications (drug delivery mainly), have already been covered in recently published reviews (Richter et al., 2021; Rodríguez & Vader, 2022). Therefore, this section will focus on studies dealing with fusion associating polymers for surface modifications.

The fusion of EVs with liposomes can lead to the creation of hybrid lipid vesicles constituted of both synthetic and natural lipids (Figure 3b). Different methods were used to obtain such hybrid particles, such as extrusion, freeze‐thaw (F/T) cycles or PEG‐mediated fusion. Extrusion relied on physical treatment by forcing the passage of vesicles through polycarbonate membranes of different pore size. When the operational temperature is slightly above the phase transition temperature of phospholipids, and if vesicles are larger than pores size, vesicles can be rapidly reshaped into smaller‐size vesicles after being ruptured by the shear of membrane pores. Thus, EVs from ovarian cancer cells were fused with PEGylated liposomes DLin‐MC3‐DMA:1,2‐dipalmitoyl‐sn‐glycero‐3‐phosphocholine (DPPC):cholesterol:18:1 Biotinyl PE:DMG‐PEG; 0.3:0.3:0.355:0.015:0.03; M:M (Evers et al., 2022). EVs and siRNA were introduced during the hydration step of lipids film and the obtained suspension (therefore constituted of EVs/siRNA/liposomes) was extruded through 1.0, 0.1 and 0.05 µm polycarbonate membranes. EVs were added at different ratios of EVs protein to total lipid: 1:100 and 1:50 (protein/total lipid; w/w). The resulting hybrids presented intrinsic properties of EVs and improved gene‐silencing efficacy due to siRNA loading. Hybrids were physicochemical characterized, evidencing a significant decrease in surface charge and siRNA loading capacity while adding EVs in the suspension, two signs suggesting the efficient creation of hybrid NPs. This was confirmed by the increased capture of hybrid NPs by beads functionalized with CD9, CD63, CD81 correlated to the proportion of EVs in the initial mix, as well as by the colocalization of a fluorescent synthetic lipid, initially present in liposomes, with hybrids particles captured on these CD9 beads.

Freeze/thaw cycles treatment on EVs was used to trigger a membrane rearrangement occurring during thermal phase transitions (Heimburg, 2007; Stewart et al., 2018). Indeed, when temperature goes down to subzero temperature, ice crystal formation results in volume expansion due to changes in the arrangement of hydrogen bonds. F/T technique was used to fuse macrophages derived EVs with liposomes of different compositions (DOPC, DOPS, DOTAP) +/− DSPE‐PEG (Sato et al., 2016). The cellular uptake of thus obtained hybrids was found similar to unmodified EVs one (except for DOTAP one) while hybrids constituted of EVs/DOPS/DSPE‐PEG were two‐fold more uptaken. As PEGylation on liposomes is usually creating a steric barrier avoiding interactions with target cells, this result was not expected. The authors hypothesized that PEG could (i) reduce electrostatic repulsions that could exist between negatively charged particles and cell membrane or (ii) that it could induce membrane fusion. Relying on the same strategy, CD47‐expressed EVs from fibroblasts cells were fused with thermosensitive liposomes composed of DPPC/DSPE‐PEG/MSPC loaded with docetaxel using freeze‐thaw cycles technique (Lv et al., 2020). The hybrids bearing CD47 retained the thermosensitive property with release of loaded docetaxel at 42°C perfectly compatible with hyperthermic intraperitoneal chemotherapy targeted. Moreover, the hybrids showed in vivo accumulation in tumours increased by 3.3 and 2.1‐fold compared to liposomes and EVs, respectively, thanks to enhanced circulation time in blood (half‐life: 3.5, 6.2 and 8.1 h for liposomes, EVs and hybrids, respectively).

Finally, membrane fusion was triggered using PEG, which has the ability to create tight contacts between membranes, especially by dehydrating lipid bilayers, a method already used to trigger cell‐cell fusion (Lentz, 2007). This PEG‐mediated fusion strategy was investigated by Piffoux et al. (2018) working with EVs from human umbilical vein endothelial cells (HUVEC) and liposomes composed of phosphatidylcholine and phosphatidylethanolamine. The choice of these phospholipids naturally present in cells and EVs membrane was done to limit change in EVs lipids composition after fusion. The fusion was conducted thanks to the addition of 10% (w/v) of PEG. Different PEG concentrations (0–30 mol% w/v) and molecular weights were tested (3000, 6000 and 8000 g/mol), the PEG 8000 at 30% mol being selected as two‐fold more efficient for fusion while maintaining the membrane stability and integrity. Interestingly, when PEGylated liposomes were used (by adding DSPE‐PEG₂₀₀₀ to the initial lipid mix), a reduced fusion was found (from 23% to 12%), as expected regarding the well‐known shielding role of PEG. Therefore, despite interesting for drug loading, this strategy may lead to low yield for EVs PEGylation.

The idea of fusing EVs with synthetic system was also investigated with polymer nanoformulations. Thus, branched low molecular weight PEI/siRNA complexes were mixed with EVs from different cell lines (prostate, ovarian, colon carcinoma and osteosarcoma) through a 15 min incubation step followed by a 3 min US bath treatment (Zhupanyn et al., 2020). Complex formation was assumed based on physico‐chemical features of the obtained systems compared to native EVs (i.e., size and surface charge increase from 110 nm and −16 mV for Evs to 180 nm and + 25 mV for the complexes). The labelling of EVs and siRNA led to a co‐localization of signal in complexes by flow cytometry. The authors showed increased knockdown efficacy and storage stability of PEI/siRNA complexes when associated to EVs.

3.3. Polymer adsorption (layer‐by‐layer)

In parallel to polymer lipid anchorage, adsorption of polymer at EVs surface mainly relying on electrostatic interactions was also evaluated as illustrated Figure 3c. Tamura et al., worked with cationized pullulan to modify MSC derived EVs surface (Tamura et al., 2017). Indeed, cationized pullulan has the ability to target hepatocyte asialoglycoprotein receptors overexpressed in the liver which enhanced therapeutic effect MSC‐EVs. In vivo the pullulan modified EVs accumulated in the liver and showed enhanced anti‐inflammatory effects with significantly lower necrotic area in liver and plasma alanine aminotransferase compared to unmodified ones. Similarly, the surface of EVs from human embryonic kidney cells was modified with carbonate apatite and glycan cationic polymers: poly‐L‐Lysine (PLL)‐lactose and PLL‐N‐acetylglucosamine for improved delivery to targeted cells (Matsuki et al., 2021). Overall, a selective delivery to parenchymal hepatocytes was achieved in vitro and in vivo with (PLL)‐lactose due to the galactose group that induced recognition of the asialoglycoprotein receptors on the cells. Lastly, EVs from mouse macrophage cells were coated with cationic nanogel particles using amino group‐modified cholesterol‐bearing pullulan (Sawada et al., 2020). The success of the formulation relied on the electrostatic interactions but also the hydrophobic interactions via cholesteryl group and lipid domain for efficient cellular delivery.

In addition to the different strategies described below, a combination of hybrid formation and layer by layer modification was used by Jhan et al. to allow co‐delivery of siRNA and anticancer drug DOX in multilayered EVs hybrids nanoparticles (Jhan et al., 2021). The EVs were co‐extruded through 200 nm with zwitterionic POPC in order to create the hybrid NPs. To protect siRNA in hybrid core, siRNA was loaded by electroporation at this step. Afterwards, the LbL occurred with first the polycation PLL to coat the weakly anionic hybrid surface then the polyanion PAA enabling DOX loading and lastly the cationic poly (ß‐amino ester) to facilitate cell interactions and siRNA delivery. Each step of the multilayered engineered EVs was characterized (zeta potential, size and concentration by NTA analysis) to ensure the elaboration of the hybrid. The resulting hybrid EVs led to efficient co‐delivery of siRNA and DOX into lung adenocarcinoma cells (A459).

As exposed in this section focusing on surface modification, the main polymers used for stabilization of synthetic vectors were almost all PEG based and commercially available. Thus, to date, no specific polymer has been designed and evaluated specifically for EVs. All studies intending to modify EVs surface for targeting were, to our knowledge, based only on a first step of lipid PEG post‐insertion, with no example identified using other strategies such as hybridization with synthetic particle of layer‐by‐layer modification. In addition, it is interesting to notice that fusion strategies mainly rely on interaction with lipid‐based systems and polymers‐EVs fusions were scarcely evaluated. Moreover, only some charged polymers other than PEG such as PLL (positive charge) and HA, PAA (negative charge) were also used in layer‐by‐layer process for the modification of EVs surfaces.

This almost exclusive use of the PEG (with 11 out of 13 papers inventoried) for surface modification of allogenic EVs is therefore questionable. Indeed, nowadays a growing concern has risen on PEG overuse (Knop et al., 2010). It has been clearly established that the immune system can directly generate PEG specific antibodies triggering accelerated elimination of PEGylated formulation, including LNP‐mRNA commercial vaccine formulations, after multiple injections, which decreases their efficacy (Carreño et al., 2022; Guerrini et al., 2022; Hamad et al., 2008; Ju et al., 2023; Yang et al., 2016). This anti‐PEG IgM effect was illustrated for EVs with the recent study of Emam et al. (2021) and reinforced by the pre‐existing anti‐PEG antibodies in patients and healthy humans due to daily use of PEG products (household cleaners, processed food and cosmetics). Overall, these effects greatly endanger the therapeutic efficacy in addition to the well‐known PEG dilemma where PEGylated nanoparticles showed low intracellular internalization and low endosomal escape resulting in a poor therapeutic delivery in cells (Fang et al., 2017). Therefore, the diversification of hydrophilic polymer sources for PEG alternatives is mandatory to avoid such effects. In recent reviews, categories and characteristics of six PEG alternatives including as polyesters, poly(oxazolines) or polyacrylamides were discussed (Hoang Thi et al., 2020; Knop et al., 2010; Yao et al., 2023). EVs stabilized with one of these polymer alternatives should be deeply investigated for intravenous delivery of stable EVs with efficient therapeutic delivery. However, it is important to again point out that the surface modifications of EVs must be carefully considered in order not to alter the natural properties of these vesicles.

4. EVS ASSOCIATION TO POLYMER‐BASED SCAFFOLDS FOR THERAPEUTIC APPLICATIONS

In this section, we will investigate how EVs have been associated with various polymeric scaffolds (hydrogels, fibres and microspheres) to allow their retention in their site of action, as well as their controlled release (Figure 4). We will focus here on EVs loading techniques, interactions with polymer‐based scaffolds and release profile. Different fabrication techniques such as 3D printing or electrospinning will also be discussed.

FIGURE 4.

Association of EVs to different scaffolds for therapeutic applications. EVs can be associated to hydrogels (a) from different compositions before (left), or after (right) the formation of hydrogel depending on the condition required for gelation. For microsphere association (b), EVs can be encapsulated, loaded into microspheres (MS) (left) or coated onto MS surface (right). Regarding nanofibre association (c), different types of association have been explored: loading or coating, as evaluated with microsphere, but also a ‘mussel like’ association using adhesive protein.

4.1. EVs association with synthetic polymer‐based hydrogel scaffolds

Hydrogel are hydrophilic three‐dimensional (3D) cross‐linked polymer networks, able to absorb and retain large quantities of water. Various types of interactions can structure them by combination in a same system: that is, covalent cross‐links, electrostatic interactions, hydrogen bonds, hydrophobic interactions, physical entanglements of individual polymer chains. Hydrogels can be synthesized in several ways: monomer polymerization followed by cross‐linking or use of polymers followed by subsequent cross‐linking. Depending on the synthesis strategy and the nature of the polymer (i.e., structure, cross‐linking density), it is possible to obtain customized properties, such as biodegradability, mechanical resistance (addition of inorganic compounds) and response to chemical or biological (endogenous or exogenous) stimuli. Hydrogels are therefore widely used in healthcare, particularly in regenerative medicine (Hu et al., 2019; Slaughter et al., 2009). In association with EVs, hydrogels have been employed for their 3D structure and their resulting rheological and mechanical properties similar to native extracellular matrices (Rosales & Anseth, 2016). Depending on the nature and therefore degradation rate of hydrogel matrices, different EVs release profiles can be found. Reviews have already reported the existing interaction between hydrogel and EVs, but generally focusing on the type of application (cartilage, bone, kidney, cardiac repair or wound healing) (Ju et al., 2023; Safari et al., 2022). By contrast, in this review, we decided to study how EVs were associated to scaffold, as this could strongly impact their biological activity, and discuss the type of interactions responsible for their retention and release from these scaffolds. In the following section, studies will be discussed regarding EVs association strategies: impregnation into already formed hydrogel, anchoring or gelation entrapment.

4.1.1. EVs association after gelation process

Impregnation in preformed hydrogels

The first and most accessible method is the impregnation of EVs which consists in first preparing the scaffold and then adding the EVs by immersion or soaking. This strategy was used to associate EVs to natural polymers such as chitosan and silk acting as a sponge‐like scaffold (Shi et al., 2017; Xu et al., 2018). After freeze‐drying the hydrogel sponge, EVs derived from gingival MSCs or plasma platelet‐rich (suspended in PBS) were added to rehydrate these scaffolds resulting in their insertion into the microstructure. The hydrogel sponges found application in both cases to efficiently promote wound healing. However, no information was provided regarding the yield of EVs associated into the scaffold or their release profile. Hyaluronic acid (HA), another natural polymer, was used to embed EVs from MSC to be injected in the pericardial cavity to treat coronary heart disease with a minimally invasive injection (Cheng et al., 2022). Interestingly, EVs retention at the injection site was significantly higher with HA matrix compared to injection of a ‘simple’ EVs suspension. A sustained release was observed over 36 h thus preventing peritoneum leaking post injection. Also working on injectable HA hydrogel for cardiac repair Pezzana et al., incorporated umbilical cord mMSC derived EVs in an HA gel using two syringes connected with a luer‐type adaptor for optimal dispersion of EVs within the polymer matrix (Pezzana et al., 2022). After conducting a deep study on HA rheological properties, EVs release and properties, the authors demonstrated that EVs were homogeneously dispersed in the scaffold, from which they can be sustained released over 10 days with no alteration of the EVs bioactivity. Once administered in vivo in a rat model, the EVs hydrogel preserved cardiac function, improved angiogenesis and decreased apoptosis and fibrosis compared to HA scaffold or EVs suspension alone.

Some studies were found to use multiple scale‐based scaffolds working with synthetic polymers. For example, a multiscale macro‐/micro‐/meso‐porous scaffold using Pluronic F127 as mesoporous template, poly(acrylic acid) microspheres as microporous template in a macroporous polyurethane sponge was designed to trigger bone regeneration (Niu et al., 2017; Tang et al., 2016). EVs from BM‐MSC were added dropwise, and then lyophilized within the scaffold. After rehydration, BM‐EVs were uniformly distributed on the inner surface of the micropores, thus serving as shelters. The release profile evidenced a burst release in the first week followed by a slow sustained release over 28 days. The therapeutic benefit, evaluated through the study of bone regeneration, was evidenced for EVs embedded scaffold compared to EVs suspension (Liu et al., 2021). Another synthetic polymer, broadly used for health application, PLA, was 3D printed with different pore size and porosity. A coating of positively charge PEI was then added to associate negatively charged EVs from human gingival MSC onto this scaffold and improve adhesiveness through electrostatic interactions. In vivo, the 3D‐PLA scaffold ensured mechanical stability during its degradation. A better internalization in cells was reached thanks to the PEI coating of EVs then activating local bone induction in order to improve bone tissue regeneration (Diomede et al., 2018). Finally, still for bone regeneration purposes, Tao et al. prepared a LbL self‐assembly of hyaluronic acid and poly‐L‐lysine coated on a β‐tricalcium phosphate scaffold crosslinked by EDC/NHS (Tao et al., 2022). Again, the presence of a cationic polymer (here PLL) induced electrostatic interaction to maintain EVs into the scaffold, which led to a subsequent sustained release over 35 days. These EVs were loaded, relying on the EXPLOR technology, with the protein ZEB1 (Zinc Finger E‐Box Binding Homeobox 1) to promote angiogenesis‐dependent bone formation. This system, despite being complex, was able to enhance angiogenesis and osteogenesis impaired by diabetes mellitus (DM), but also to inhibit the abnormal formation caused by DM in vitro. In vivo, only the full system was able to repair a critical‐size cranial defect in DM rats, while scaffold loaded with non‐loaded or non‐surface modified EVs did not (Yim et al., 2016).

Lastly, among the biological scaffolds for tissue engineering, decellularized extracellular matrix (dECM) are obtained from human or animal organs and tissues by removal of cellular components via decellularization processes. On their own, dECM can be defined as a 3D scaffold containing extracellular macromolecules (i.e., collagen, elastin) and matricellular proteins (Zhang et al., 2022). As dECM scaffold mimics natural ECM and allows sustained bioactives molecules, their use in association with synthetic polymers is interesting to increase the biomimetic properties of these synthetic scaffolds, while synthetic polymers can help ECM shaping control. The first dECM/synthetic scaffold used for EVs association relied on porous pneumatic microextrusion composite combining synthetic PLGA, magnesium hydroxide (MH) and decellularized porcine kidney extracellular matrix (kECM (Ko et al., 2021)). EVs from primed MSC (TNF‐α, IFN‐γ) were added as a solution to rehydrate the scaffold after freeze drying to boost anti‐inflammatory/anti fibrotic properties of the scaffold and increase kidney regeneration. MH was used as an antacid to treat the acidic environment caused by the degradation of PLGA scaffold while kECM mimicked the kidney tissue microenvironment. In vivo, the obtained composite scaffold efficiently improved the regeneration and maintenance of a functional kidney tissue. dECM association to synthetic polymer is also interesting to enable dECM 3D printing for fine control of the final scaffold structure, as dECM itself did not have the proper properties to be 3D printed. Thus, a decellularized small intestinal submucosa (SIS) was combined with mesoporous bioactive glass (MBG), using Pluronic P123 as structure directing agent and sodium alginate as thickener and porogen (Hu et al., 2021). EVs from BM‐MSC were associated with the scaffold by simple impregnation in the scaffold media after 3D printing. After hydrogel formation, sustained release over 14 days was observed. Overall, this scaffold gathered suitable porosity, biocompatibility and haemostasis ability. Tested in rat's model of diabetic wound, the EVs loaded in scaffold allowed the proliferation, migration of HUVEC and enhanced angiogenesis enabling diabetic wound healing.

EVs tethering in preformed hydrogels

To control in EVs association and subsequent release from scaffold, another approach relies on anchoring EVs in a preformed hydrogel scaffold using a tethering strategy (Figure 5a). Thus, the EVs surface modification described in Section 3, can be used to control their association with hydrogels. One compelling example of this strategy was recently illustrated by Yerneni et al., with non‐covalent cholesterol‐modified DNA tethers on EVs derived from non‐activated macrophages cells (Yerneni et al., 2022). They employed a ‘grafting‐from’ strategy using modified EVs as ATRP initiator as previously observed in section surface modification above (see BOX 3 for more detail). Methacrylated PEG as monomer and dimethacrylated PEG as crosslinker were in situ photo‐copolymerized starting from the EVs to form a macroscopic hydrogel. The presence of a photocleavable group (p‐nitrophenyl) between the DNA and the cholesterol moiety allows triggering release of EVs upon light stimuli. Overall, the EVs release was extended to a month thanks to the tethering compared to physically entrapped EVs in gelatin methacrylated (GelMa). Interestingly, the modification of crosslinking density of the hydrogel can also influence the release profile, in addition to the photocleavable group controlling the release upon light stimuli. In another study, integrins present at MSC‐EVs surface were used to bind with RGD peptide hydrogels of methacrylated alginate by photo‐polymerization as illustrated Figure 5a (Huang et al., 2021). The designed hydrogel was able to encapsulate, tether and retain the EVs over 7 days while maintaining the structural integrity and osteoinductive functionality of the EVs. The last example in the literature was found with a thiolated CP05 peptide capable of binding on EVs with the second extracellular loop of CD6 (a tetraspanin enriched on the EVs surface) (Zou et al., 2021). The resulting thiolated EVs (from human UC‐MSC) was covalently anchored to the hydrogel via an epoxy/thiol ‘click’ reaction in situ. The hydrogel was composed of thiolated hyaluronic acid (HA‐SH) with hyperbranched epoxy macromer grafted by an aniline tetramer (EHBPE‐AT) as crosslinker for the hydrogel construction. The hydrogel was obtained after mixing EHBPE‐AT and EVs with HA‐SH and CP05 and showed efficient encapsulation, controllable gelation kinetics, shear‐thinning injectability, and excellent cytocompatibility. Finally, EVs loaded hydrogel possessed conductivity for myocardium and dynamic stability adapting to heartbeats leading to improved cardiac function and promoting blood vessel regeneration in a rat model of myocardial ischemia.

FIGURE 5.

Example of strategies used to cross‐link hydrogels associated with EVs. Hydrogels can be reticulated and EVs anchored to the polymer network using a tethering strategy such as with integrin/RGD interactions as described in (a). EVs can be trapped thanks to thermo‐responsive polymers inducing gelation with temperature increase such as with Pluronic F127 (b). Another possible stimulus is photo‐responsivity with methacrylate grafted onto gelatin able to reticulate under UV and associated EVs in solution (c). Finally, self‐healing hydrogel can also entrapped EVs as illustrated with Schiff‐base linkage between aldehyde function grafted on methyl cellulose and amino group of chitosan (d).

4.1.2. EVs association before or during gelation process

This EVs entrapment strategy before gelation, is, up to now, the main method used to associate EVs with scaffolds. The concept is based on the gelation process, that is, the establishment of the hydrogel network, to entrap EVs within the matrix. The use of smart polymers able to create weak interactions or transform under external stimulations such as with temperature and light allowed creating stimuli responsive hydrogels. Moreover, with these smart polymers, the gelation is also potentially reversible.

EVs association to thermo‐responsive polymer‐based hydrogels

Polymer solutions can be prepared at a temperature below the gelation point, with a viscosity appropriate for EVs homogeneous association, before the gelation process can occur in situ after in vivo injection. The gelation upon temperature can be done with thermo‐responsive polymers or thermogelation. Thermo‐responsive polymers (Figure 5b) are capable of a modification of their shape above a characteristic temperature named low critical solution temperature (LCST). These intrinsic properties of the polymer result from the enhancement of polymer‐polymer interactions instead of polymer‐water interactions producing a partial insolubility of the polymer and the contraction of the hydrogel. By contrast, the thermogelation is due to a dehydration of the polymer chains in water with temperature increased depending on polymer concentration and presence of salts. Thus, the gelation from solution to hydrogel can be triggered using temperature as a stimulus either using thermo‐responsive polymers (with specific LCST) or thermogelation (with polymer chain dehydration) that will form hydrogel at body temperature.

Thermogelation is the most widely used strategy in the healthcare field, relying mainly on the use of poloxamers (Pluronic) capable of forming hydrogels at body temperature. Poloxamers are non‐ionic surfactants with amphiphilic block copolymers made of the hydrophilic poly (ethylene oxide) (PEO) and the hydrophobic poly(propylene oxide) (PPO) structured in triblock (PEO‐PPO‐PEO). The poloxamer forms micelles in water and once the critical micellar temperature is reached the thermogelation occurs due to the dehydration of the PPO block. Thus, the gelation properties can be tuned depending on the hydrophilic‐lipophilic balance and PPO chain length and also the physical blends of poloxamers and excipients (Shriky et al., 2020). Poloxamers with hydrophilic HLBs, such as Pluronic F127 (equivalent to Poloxamer‐407), are the most widely used in the healthcare sector. Indeed, prepared at 20% (w/v) their gelation can be obtained in a range of temperature of 25–30°C. This property enables an easy manipulation of the polymer solution at temperature lower than 20°C, and a quick gelation process at body temperature. So, logically, these poloxamers were also the most widely used to retain and control EVs release. Immunosuppresent PD‐L1 positive EVs, obtained either by genetic engineering or IFN‐γ priming of producing cells, were associated to F127 to allow wound healing in a mouse skin excisional wound model (Figure 5b) (Su et al., 2020). Interestingly, when PD‐L1‐EVs were combined with F127 thermoreactive gel, this treatment significantly accelerated wound contraction and re‐epithelialization during the inflammation phase. Poloxamer can also be associated to fine‐tune the gelation temperature. Thus, Pluronic F127 was combined to Pluronic F68 (called Poloxamer‐188) to obtain a gelation temperature around 33°C, and an optimized viscosity for polymer solution easy handling and injectability before administration (Liu et al., 2016). Using this Pluronic association, platelet‐rich plasma‐derived EVs were loaded to the polymer solution and then entrapped in the matrix after gelation at 37°C for a continuous release over 28 days (Zhang et al., 2022). After injection into subtalar joint in vivo in mice, EVs locally retained in the hydrogel inhibited apoptosis and hypertrophy of chondrocytes and enhanced their proliferation to delay the development of subtalar osteoarthritis. To increase binding between molecules, Pluronic F127 (gelation at 20°C) was associated with alginate and dextran (gelation at 19°C) (Chen et al., 2022). The pluronic F127/alginate/dextran solution was loaded with EVs from AD‐MSC resulting in the formation of a thermogel at body temperature. The positively charged lysine‐dextran component of the hydrogel also facilitated the association of EVs for a sustained release in situ. Tested in rat models of peripheral nerve damage, the hydrogel loaded with EVs was applied around the nerve at the coaptation site before wound closure. Three months later, the peripheral‐nerve was regenerated and the contraction force of leg muscles was also recovered. Similarly, to associate EVs through electrostatic interactions, a cationic polymer, the polycitrate‐PEI, was combined with F127 thanks to dynamic hydrogen bonding (Wang et al., 2021). Combining these polymers, the sol to gel transition was induced above 25°C. This association led to a long‐term retention and sustained release over 50 days and improvement of motor functional recovery after spinal cord injury in rats.

Thermogelation was also obtained using natural polymers such as chitosan, collagen and cellulose in combination with salts, thus relying on hydrogen bonding, electrostatic and hydrophobic interactions for hydrogel formation. To do so, hydrogel was formed at body temperature using chitosan and sodium glycerophosphate (Shen et al., 2020; Zhang et al., 2018). The addition of glycerophosphate salt to chitosan solution impacts the interactions for sol‐gel transition with (i) reduction of electrostatic repulsions thus increasing chitosan interchain hydrogen bonding, (ii) chitosan‐phosphate electrostatic attractions (ammonium vs. phosphate groups) and (iii) chitosan/chitosan hydrophobic interactions enhanced by structuring action of glycerol interaction in water (Chenite et al., 2000). This gel allowed EVs retention at the site of injection (periodontal) and their release over 7 days. Relying on the same system, Q. Tang et al. worked with chitosan associated to 1% of gelatin and glycerol 2‐phosphate disodium salt hydrate for sustained release of iPSC‐MSC EVs. This promising system, promoted the repair of damaged rat corneal epithelium and stromal layers (Tang et al., 2022). Another natural polymer, collagen, was used to create an EVs containing patch (Liu et al., 2018). Patches were obtained by punching, using a dermal biopsy punch, into commercial Gelfoam (Pfizer) to obtain 7 mm cylinders. These patches were thereafter loaded with an acidic collagen solution (2 mg/mL) subsequently neutralized and 37°C heated to trigger gelationand an EVs suspension in PBS. The resulting hydrogel patch was able to slowly and sustainably deliver EVs secreted by cardiomyocytes derived from iPSCs (iCM‐EV). This EVs retention system, when implanted onto infarcted rat hearts, enabled significant reduction of infarct size 4 weeks post‐infarction. Finally, the last example was with a biocompatible methylcellulose‐based hydrogel (porous structure) composed of disodium hydrogen phosphate for a gelation temperature of 32°C (Xing et al., 2022). The temperature played a role in the kinetics of EVs derived from endothelial cells, to trigger a sustained release over 5 days at 29°C and 8 days at 37°C due to a burst of release at lower temperature. This profile was adequate for the therapeutic application of critical limb ischemia (CLI) where the body temperature is lower than in healthy humans. Genetically engineered‐EVs, overexpressing vascular endothelial growth factor (VEGF) and transcription factor EB (autophagy regulator) associated to this injectable thermoresponsive hydrogel were more stable in vivo and can significantly improve limb function after CLI.

For thermo‐responsive polymers, only two were found to be used with EVs relying on their LCST to induce gelation. In both cases, LCST values were below physiological temperature so when injected into the body they transitioned from liquid to gel form an in situ hydrogel without additional chemical reaction, cross‐linker or external stimuli. PEG‐poly(ε‐caprolactone‐co‐lactide) with a LCST at 32°C was prepared to create hydrogel containing stem cell‐derived EVs suitable for administration at room temperature by intratunical injection for erectile dysfunction treatment (Liang et al., 2022). After the EVs encapsulation and sol‐gel transition at 37°C, a sustained release of 2 weeks was observed. In a rat model of erectile dysfunction, EVs were retained at the injection site and inhibits cavernous smooth muscle cell apoptosis, favoured the recovery of vascular endothelial cells of the cavernous sinus and thus helped to accelerate the recovery of erectile function. The second polymer poly (N‐vinyl caprolactam) with a LCST at 31°C was used by Das et al., for encapsulation and sustainable delivery of EVs from fibrosarcoma to induce clot lysis tested in vitro as promising for effective treatment for thromboembolic diseases (Das et al., 2022).

EVs association with UV‐responsive polymer‐based hydrogels

As described in BOX 3, the most well‐known function used to allow UV crosslinking is the acrylate function forming polymer such as methacrylated gelatin (GelMA) based hydrogel. Thus, HUVEC‐derived EVs in solution with GelMA were entrapped during photo‐crosslinking under UV stimuli (Figure 5c). After this, EVs were sustainably released over 7 days in vitro. The hydrogel, loaded with EVs, was cutaneously applied in a wounded rat model. After 14 days an accelerated re‐epithelialization, a promotion of collagen maturity and an improved angiogenesis suitable for wound healing were observed (Zhao et al., 2020). In a different therapeutic context, aldehyde‐functionalized chondroitin sulphate was introduced into GelMa to entrap BM‐MSC EVs and promote cartilage repair after growth plate injury. This hydrogel could significantly promote the synthesis of ECM due to the doping of chondroitin sulphate. In vitro and in vivo, this scaffold was reported to exert anti‐inflammatory effects through the slow release of EVs by regulating the polarization of macrophages toward a M2 profile (Guan et al., 2022). GelMa hydrogel and microneedle array patch were associated to create an MSC‐EVs hydrogel microneedles patch to induce in situ spinal cord injury (Han et al., 2022). After local implantation of this patch to the site of spinal cord injury, the sustained release of EVs over 28 days led to improved recovery of nerves function in a rat model. To enhance mechanical and biological properties of hydrogels, laponite nanoclay was added into the photo‐crosslinkable GelMa (Hu et al., 2020). Laponite is a synthetic smectite nanomaterial capable of generating colloidal like suspensions in aqueous environment. This is worth noticing that laponite loaded in polymeric hydrogel has received an FDA approval (Pullin, n.d.) setting their potential for clinical use. More especially, it has been investigated for bone tissue engineering applications due to the osteoinductive properties of its degradation products (Man et al., 2022). A solution of GelMa, laponite nanoclay and EVs from UC‐MSC was prepared and injected in rats before triggering gel cross‐linking by UV light. The sustainable release of EVs, correlated to the degradation of the hydrogel over 31 days with nanoclay (compared to 11 days without nanoclay), led to improved cartilage regeneration. This enhancement of GelMa mechanical properties using nanoclay has been recently exploited to improve printability with better gel shape fidelity in 3D printing (Man et al., 2022). The addition of nanoclay also played a role in the release kinetics of osteoblasts‐derived EVs over 14 days and increased mineralisation of the extracellular matrix. Similarly, HA was methacrylated and mixed with EVs MSCs before photo‐crosslinking to form an hydrogel that was intrapericardial (iPC) injected in pigs. After iPC injection in a myocarde infarctus pig model, the EVs loaded HA hydrogel were shown to spread in the pericardial cavity while MSC EVs were uptaken by epicardiac cells. This EVs hydrogel improved cardiac morphology and suppressed transition to heart failure in a pig model of MI (Zhu et al., 2021). The same approach was performed with dECM to allow its 3D printing. EVs MSC‐EVs, ECM and GelMa were mixed before the stereolithography process leading to a homogeneous 3D printed scaffold entrapping EVs for cartilage repair (Chen et al., 2019). The EVs release was sustained over 14 days in vitro and 7 days in vivo in a rabbit model of osteochondral defect. In vivo, the full 3D printed system (ECM/GelMa/EVs) seem to successfully increase chondrocyte migration, to induce synovial macrophage polarization toward an M2 phenotype to be translated into an enhancement of cartilage regeneration. Relying on a different functionalization to induce UV cross linking, HA and gelatin were functionalized with photoinduced crosslinking (Liu et al., 2017). BM‐MSC‐derived EVs were entrapped and retained in this hydrogel for 14 days in vitro. In a rabbit articular cartilage defect model, the EVs loaded in hydrogel were injected in the articulation and promoted migration and proliferation of these cells compared to hydrogel without EVs. Only the HA hydrogel with EVs was able to promote cartilage repair and regeneration after 12 weeks in vivo. Finally, a photocurable adhesive polymer containing poly (glycerol sebacate) acrylate was used to deliver EVs from human induced pluripotent cardiac progenitor cells (Hamada et al., 2020). The originality of this work lies in the hydrophobic and viscous nature of the polymer enables on‐demand adhesion to biological tissue using light in this case for heart adhesion.

Physically crosslinked polymer‐based hydrogels

The first type of reaction for the establishment of physically crosslinked hydrogel is with reversible Schiff base linkage (see BOX 3 for further details on this reaction). Hydrogels of polysaccharides were prepared by mixing aldehyde modified methylcellulose and chitosan grafted PEG thanks to Schiff base linkage between the aldehyde function and amino group of chitosan (Wang et al., 2020) as illustrated Figure 5d. Placental MSC derived EVs were added to methylcellulose solution and entrapped in the network during the mixing with chitosan solution within a gelation time of 300 s. This biocompatible injectable self‐healing hydrogel was tested in vivo in diabetic rat model and enabled wound healing with full recovery of skin structure and function within 15 days. However, no specific evaluation of EVs release was conducted only the biological effect was considered. A multifunctional hydrogel composed of Pluronic F127 (F127), oxidative hyaluronic acid (OHA), and poly‐ε‐L‐lysine (EPL) was formed thanks to Schiff base linkage between OHA and EPL, and to the thermoresponsive property of F127 (see section thermoresponsive above ) (Wang et al., 2019). The positive charge of EPL promoted tissue integration, antibacterial activity but also triggered interaction with negatively charged EVs. In parallel, OHA allowed water retention ability, while F127 provided heat‐sensitive gelation, and Schiff's base bonds (between OHA and EPL) created self‐healing ability. Furthermore, these base bonds can be broken in a weak acidic environment leading to a slight increase of EVs release over time, to reach ∼90% at pH 5.5 at day 21 (vs. ∼80% at pH 7.5). In vitro, this system induced proliferation, migration and angiogenesis of HUVECs while the in vivo study evidenced faster granulation tissue formation, re‐epithelialization and collagen remodelling leading to faster diabetic wound healing (compared to hydrogel w/o EVs). Similarly, HA was modified with aldehyde or adipodihydrazide functions to trigger hydrogel association through a Schiff base reaction (Li et al., 2020). Laminin‐derived adhesive peptides were combined with the hydrogel, through a laminin‐integrin interactions, thanks to the integrins exposed at EVs surface. Thus, this association led to a sustained release of EVs reaching 90% over 15 days. This EVs loaded hydrogel was evaluated in vivo in a severe rat long‐span spinal cord transection model. Twenty‐eight days after tissue implantation with this EVs loaded adhesive hydrogel, a reduced lesion cavity and an increased myelin sheaths were observed. As a results, the successful functional recovery evidenced the potential of this hydrogel in central nervous system diseases treatment. Lastly, Schiff base reaction was also performed by S. Yang et al., to cross‐link hyaluronic acid and alginate to create a hydrogel for bone repair (Yang et al., 2020). This biomaterial also included nanosized hydroxyapatite (HAP), the main inorganic compound of natural bone and teeth, embedded in the hydrogel. This addition of HAP brought mechanical strength and 3D porous microstructure. Afterwards, EVs derived from the human umbilical cord MSC were incorporated with the polymer solution and entrapped during gelation. Their release was sustained over 14 days. In vivo, in a calvarial defect model, EVs loaded hydrogel exposed enhanced bone repair compared to hydrogel only.

Physically crosslinked hydrogel was also designed based on the association of sodium alginate and calcium chloride. Indeed, a sodium alginate (with high content of guluronic acid) solution was mixed with EVs derived from AD‐MSC followed by the addition of calcium acting as physical crosslinker (Shafei et al., 2020). A full release was subsequently observed after 170 h, and directly correlated with hydrogel degradation rate. This hydrogel was then tested in vivo in a rat model of full‐thickness skin excision and was able to improve wound healing to a significant extent compared to control or alginate hydrogel only. Still relying on calcium alginate, Xiao et al., designed oxidized alginate crosslinked to di‐aldehyde gelatin (Xiao et al., 2021). AD‐MSC derived EVs were encapsulated by addition of calcium chloride with bonds (between Ca²⁺ vs. COO⁻ and CH = N) to facilitate bladder regeneration in rats model. Lastly, hydroxyapathite/chitosan composite hydrogel was investigated for the encapsulation of EVs derived from miR‐126‐3p overexpressing synovium MSC (Li et al., 2016). EVs were added in an acetic acid solution of chitosan (CS) with hydroxyapatite (HAP) and the entrapment was triggered by addition of NaOH solution (pH∼11–12) inducing gelation of chitosan. Regarding these harsh conditions, the integrity and stability of EVs in the final system is questionable. However, this CS‐HAP hydrogel loaded with EVs led to successful angiogenesis and re‐epithelization in vivo in rat model compared to the hydrogel without EVs.

Finally, the following section will describe hydrogels gelation triggered by electrostatic, hydrophobic interactions and dynamic bonding. An injectable hyaluronic acid hydrogel relying on host‐guest interactions of adamantane‐modified HA and β‐cyclodextrin‐modified HA was built (Chen et al., 2018). EVs derived from endothelial progenitor cells were encapsulated in the hydrogel for intramyocardial delivery and slowly released over 21 days. This hydrogel enhanced peri‐infarct angiogenesis and myocardial haemodynamics in a rat model of myocardial infarction. Another original study was dedicated to fibrin gel with embedded EVs from BM‐MSC thanks to immediate gelation by mixing of thrombin and fibrinogen (Yu et al., 2020). The local delivery of EVs in this fibrin gel was a successful strategy for tendon repair. Furthermore, hydrogel can also be obtained through dynamic coordination such as with Ag⁺‐S (Lin et al., 2021). Using a four‐arm thiol‐PEG with silver nitrate solution (AgNO₃) it creates dynamic coordination bonding together with simultaneous interactions between Ag inducing gel formation. Even if the bonds were broken during injection, it can be self‐repaired thanks to the dynamic coordination. Thus, this hydrogel gathers self‐healing, degradable, injectable and antibacterial properties interesting for clinical application. Mi et al., also worked on dynamic bonding, with hydrazone self‐crosslinking applied to hyaluronic acid hydrogel (Mi et al., 2022). Two derivatives were synthesized, one aldehyde and quaternary ammonium modified HA and one hydrazide modified HA that mildly but quickly reacted in situ after mixing. The resulting hydrogel possessed adhesive properties and allowed a sustained release of engineered (with miR‐26a‐5p transferred into EVs using the CD9‐HuR fusion protein) endothelial cell‐derived EVs for appropriate bone repair.

Overall, the strategy of associating EVs with hydrogels allowed to retain EVs at a define site and, to some extent, achieve a controlled release as a function of the density of the hydrogel network. The use of smart polymers, responsive to different environmental stimuli (i.e., pH, redox potential, enzyme, temperature, light, magnetic field or ultrasound) could offer an ‘on‐demand’ EVs release. These cleavable functional groups causing destabilization of hydrogel can be in the repeating unit of the polymer, in terminal position or at the junction between two polymeric blocks. In the future, dual or multi‐responsive polymers triggering specific changes in conformation, interactions or self‐assembly upon different stimuli could be an interesting option. The release of EVs could result from a shrinking of the hydrogel as a consequence of an increase of bonds or interactions provoking expulsion of water and consequently EVs. By contrast, the release of EVs can be triggered by the expansion of hydrogel under stimulus by bond breaks coming from a decrosslinking. The broad functionality of polymers can be fully exploited to cover more stimuli as it has been well studied over the years. As an example, cinnamate or coumarin dyes can be exploited as reversible UV‐sensitive groups capable of dimerization and de‐dimerization. These multi‐stimuli hydrogels entrapping EVs started to emerge, as recently described by Jiang et al., using glucose/reactive oxygen species dual‐responsive hydrogel via crosslinking of phenylboronic acid‐modified HA and PVA (Jiang et al., 2022).

4.2. EVs association with polymer nanofibres

The term nanofibres refers to a fibre with a diameter below 100 nm and a collection of nanofibres is often called a mat or a mesh. These nanofibres are usually obtained by electrospinning where a strong electric field (kV range) is applied to disperse liquids into fine jets (Williams et al., 2018). PLGA is one of the most used polymers when looking at electrospinning. Thus, EVs derived from human MSC (AD‐MSC or BM‐MSC) were mixed with a PLGA /metal‐organic framework (MOF) composite scaffold. MOFs are assembled using metal ion or cluster nodes and functional organic ligands through coordination bonds. Recently, the functionalization of scaffolds with MOFs has been explored as an interesting strategy for bone tissue regeneration relying on the release of metal ions (e.g., Mg²⁺, Cu²⁺) in cells, governing multiple cellular functions, including cell signalling, growth, metabolism and proliferation. Either magnesium‐gallate‐based MOF (Kang et al., 2022) or copper‐based MOF (Xu et al., 2023) were associated with the scaffolds. EVs were fixed on the synthesized composite scaffold by physical embedding and electrostatic interactions with the PLGA/MOF. Compared to PLGA scaffold, in both cases the EVs release from PLGA/MOF was more gradual and slower over 7 days, indicating the efficacy of MOF for immobilization and retaining EVs. In a rat models of calvarial bone defect, the porous polymer meshes associated to both AD‐MSC or BM‐MSC derived EVs were able to accelerate the bone regeneration compared to the EVs free scaffold. Using an original mussel‐inspired immobilization strategy X. Xing et al. created a bioactive electrospun silk fibroin (SF)/poly (ε‐caprolactone) (PCL) scaffold coated with EVs from MSCs (Xing et al., 2021). This SF/PCL scaffold was prepared by electrospinning and coated with polydopamine before being immersed in an EVs solution mimicking mussels (here EVs) that attached to their surface materials with a secreted adhesive protein (here polydopamine). A sustained release of EVs from the scaffold was observed over 25 days. in vitro, this system led to promotion of osteogenesis and angiogenesis triggering bone regeneration in a rat model of calvarial bone defect.

Among electrospinning strategy, only one study demonstrated the possibility of associating EVs into the polymer matrix before electrospinning (Trindade et al., 2021). The polymer matrix was made of polyvinylpyrrolidone, a fast dissolving material, as proof‐of‐concept. An NTA analysis was done after the fibre dissolution in PBS, but unfortunately, the presence of polymer forming nanometric auto assemblies hinders the interpretation of the results. Nevertheless, a tendency of increase in particle number was observed. A more complete characterization in terms of vesicle identity and function would support the results of this study.

4.3. EVs association with microspheres

Microencapsulation processes, often used to protect and control chemicals or biomolecules, were also adapted to EVs. As for biomolecules, one of the challenges of this strategy is to preserve the physical and functional integrity of EVs during the microencapsulation process. For example, triblock copolymer microspheres (MS) of PLGA‐PEG‐PLGA were produced by a water‐in‐oil‐in‐water double emulsion‐solvent evaporation methods relying on dual flow‐focusing junction microfluidic device, a low shear stress process (Swanson et al., 2020). In this study, EVs derived from dental pulp stem cells were added in the intern aqueous phase, the PLGA‐b‐PEG‐b‐PLGA in ethyl acetate in oil phase and milliQ water constituted the continuous aqueous phase. This method allowed formation of homogeneous 10 µm MS suspensions which were then physically immobilized onto nanofibrous poly (L‐lactic acid) (PLLA) scaffold. The controlled delivery of EVs loaded MS (EVs‐MS) from the nanofibres was sustained over 2 weeks. EVs release from MS showed an initial burst release (∼20% release) followed by a sustained release profile, delivering a consistent dose of EVs for up to 10 weeks. Interestingly, by acting on the hydrophilicity of the copolymer composing MS, EVs release profile can be modulated: the burst release was more important when using the more hydrophilic compositions. Once EVs‐MS were immobilized on PLLA scaffolds, EVs release exposed an attenuated burst phase (∼7%) compared to EVs‐MS alone in suspension (∼20%). The physical integrity (NTA, TEM) was evidenced after release. However, once implanted in a mouse model of calvarial bone defect without exogenous cell seeding, only the scaffold functionalized with EVs‐MS significantly increased bone volume evaluated by Micro‐computed tomography (μCT). Also working on MS of PLGA, Gao et al., adsorbed EVs on the surface of the porous MS thanks to polydopamine coating (Gao et al., 2022). In this case, the EVs derived from stem cells of human teeth, were sustainably released over 21 days and induced bone regeneration once tested in rat models. Finally, EVs were also associated with pharmacology active microcarriers (PAM) for myocardial repair. These MS were constituted by a matrix of PLGA‐Pluronic188‐PLGA triblock copolymer and covered by fibronectin/poly‐D‐lysine. This interesting biomimetic coating allowed EVs fixation through interaction receptors (FN, α5β1) expressed at EVs surface (Riaud et al., 2021). PAMs were designed to encapsulate hepatocyte growth factor (HGF) and coated with EVs before being were injected locally in a rat model of ischemia/reperfusion. The intracardial injection of the EVs‐PAMs‐HGF system was found to be successful for both cardioprotection and cardiac repair by decreasing fibrosis.

In conclusion, this section covered three types of polymer scaffolds in association with EVs: hydrogels, nanofibres and microspheres (Figure 4). Regarding the above presented data, hydrogels scaffolds are the most commonly used to ensure sustained release of EVs at their site of actions compared to the two others. This can be explained with the broad knowledge and diversity of hydrogel constructs (Ho et al., 2022) upon polymer nature and functionalization for stimuli responsiveness. The processes to obtain nanofibres and MS are more complex to carry out and could require harsh physical and chemical conditions (temperature, shear stress, solvent) compared to hydrogels but could also lead to more control over the EVs retention and release.

Overall, a wide range of polymers enabled the design of scaffolds for EVs administration. The family of polyesters had shown various applications: the electrospinning of PCL, versatile applications of PLGA in scaffolds including MS, association with metal organic‐framework (MOF), porous pneumatic microextrusion composite, etc. The polysaccharides such as chitosan, cellulose, HA, pullulan and alginate were mainly employed for hydrogel building with their numerous amine, acid or alcohol groups acting as nodes of the network. Their ability to form hydrogen bonds allows interactions at the origin of the increased viscosity of the polymeric material in which EVs could be sequestered. The polypeptides such as silk fibroin and gelatin also elaborated hydrogels. Some original investigations were noticed for gelatin as bioink in regards with the stereolithography technique and silk fibroin for electrospinning. Finally, PEG is often associated with PPO giving amphiphilic copolymers such as poloxamers for hydrogels for thermogelation.

Lastly, regarding EVs for therapeutic applications, the most common route of administration is, for now, i.v. or local injection. For future transfer to human and improved patient compliance other routes of administration should be considered. Thus, nasal or pulmonary routes have started to be investigated (Dinh et al., 2020; Popowski et al., 2022; Shi et al., 2021; Worthington & Hagood, 2020). Moreover, the oral route, already explored with non‐modified EVs (Song et al., 2022) could probably be investigated using a polymer scaffold protecting EVs from pH or enzyme attacks and inducing release (burst, delayed or prolonged) depending on the requirement. Whatever the administration pathway, either through surface modification or scaffolds association, there is no doubt that the use of polymers to facilitate EVs administration, as it could be used to stabilize biomolecules or synthetic nanoparticles, is still in its infancy.

5. CONCLUSION AND PERSPECTIVES

This review clearly highlights that EVs and polymers worlds are strongly interconnected, from the pre‐production modification stage, through production, characterization and detection to their therapeutic use (Figure 1 and BOX 2). Regarding the pre‐production stage, polymers are mainly represented by transfection agents with PEI being the most encountered one. In the EVs production and isolation steps, we mainly identified polymers already described for cell culture (2D and 3D), or polymers in isolation devices such as filtration membranes or SEC stationary phase. In this context, polymers are particularly important to reach EVs production amount and purity compatible with a therapeutic production. For these applications, petro‐based polymers were the most used ones, due to their compatibility with a commercial scale production (BOX 2). It is probably in the field of EVs detection that the broadest variety of polymers was found, responding generally to specific requirements: isolation of an EVs subset (e.g. glycosylated EVs), increasing antibodies specificity and presentation for EVs immunocapture, or stabilizing fluorescent probes for super resolution imaging. However, despite the broad polymer diversity found in the EVs detection field, the therapeutic application is the most explored one in terms of published studies. Even though this number of studies is quite important, PEG remained the main polymer used, especially for surface modification. Nevertheless, as already disclosed in Section 3.3, this almost exclusive reliance on PEG is a growing concern and called for polymers alternatives. Interestingly, for hydrogel production, biomacromolecules such as polysaccharides and proteins are mainly selected, thanks to their biocompatibility and polyfunctionality.

Aside from polymers nature, the chemical process for synthesis, the modifications, and the functionalization of polymers must be considered. Indeed these aspects were unfortunately poorly described in the majority of studies reviewed here. Generally speaking, synthetic polymers are more attractive because they are more structurally controllable with reproducible synthesis in comparison to natural polymers which can differ upon the sourcing. However, for both types of polymers, that is, natural or synthetic, the residual toxicity due to the synthesis conditions (unreactive monomer, catalyst, ligand, initiator…) by using organic solvents or the presence of metallic residues coming from catalysts must be taken into account. To do so, the substitution of toxic solvents by green ones such as cyrene, supercritical CO₂, or the use of organocatalysts instead of metallic ones constitute great options. Other improvements rely on the use of microwaves or ultrasounds to reach a polymerization avoiding the presence of a catalytic system. These synthesis and purification processes, avoiding extended purification steps should probably be favoured in the future, as they are more suitable for industrial translation, as well as for health application.

In this review, we intended to focus on the nature of the interactions between EVs and scaffolds. In view of the interactions diversity that have been reported (Section 4), it is difficult to identify and prefer one type of interaction or another. By contrast, it is interesting to see that, depending on the therapeutic requirement, the association of EVs with scaffolds can be adapted and lead to customizable release profiles. When it comes to the timing of association of EVs with scaffold (i.e., before or after gelation for hydrogel, encapsulated or coated for microspheres and nanofibres), few studies reported a significant influence of a strategy versus another on EVs physical integrity or biological functions. Although it is intuitive to think that physical stresses are deleterious for EVs, and that their association should avoid fabrication processes requiring harsh conditions, this has not especially been reported in the reviewed studies. Lastly, in general, the characterization of EVs release was not always extensive, particularly in regards to their biological functions. Therefore, it would be of interest to call for a more exhaustive characterization of associated and released EVs, following the EVs community's guidelines.

Overall, this review showed that the link between the EVs and the polymer worlds is very diverse, depending on the development stage (production, characterization, administration). This association has helped resolving certain EVs related problems for several decades now, and undoubtedly could benefit from it even more, with plenty more combinations of polymer and EVs to be discovered.

AUTHOR CONTRIBUTIONS

Laurianne Simon: Conceptualization; Visualization; Writing—original draft; Writing—review & editing. Vincent Lapinte: Conceptualization; Supervision; Visualization; Writing—original draft; Writing—review & editing. Marie Morille: Conceptualization; Funding acquisition; Project administration; Supervision; Visualization; Writing—original draft; Writing—review & editing.

CONFLICT OF INTEREST STATEMENT

The authors declare that they have no conflicts of interest.

Simon, L. , Lapinte, V. , & Morille, M. (2023). Exploring the role of polymers to overcome ongoing challenges in the field of extracellular vesicles. Journal of Extracellular Vesicles, 12, e12386. 10.1002/jev2.12386