All living cells are capable of secreting EVs. The production of EVs was initially described as a mechanism for the clearance of unnecessary compounds from the cells ^[1][7]. In 1967, researchers identified small lipid vesicles derived from whole serum as well as from platelets after ultracentrifugation. This material was initially referred to as “platelet dust” and later identified as composed of microparticles. In 1971, Aaroson described the membranous structures as EVs ^[2][8]. EVs of different sizes were visualized using electron microscopy and it took a long time before the hypothesis that considered EVs as experimental “artifacts” was finally excluded. In 1981, Trams observed that shed “microvesicles” collected from the culture medium of glioblastoma cell lines had a membrane composition similar to specific plasma membrane domains and that these vesicles induced specific effects on receiving cells ^[3][9]. Despite the initial evidence, for many years, EVs have been largely overlooked. Only one decade ago, the production of EVs started to be recognized as a new possible mechanism of intercellular communication ^[4][10]. EVs have heterogeneous structures delimited by a lipid bilayer and because they lack a functional nucleus, cannot self-replicate. EVs are divided into two categories based on dimension: medium/large EVs (m/lEVs) and small EVs (sEVs). These two populations differ in size, in the mechanism of their formation and in cargo. Concerning the cargo, all EVs can transport molecules with biological activity such as proteins, lipids, and nucleic acids, including DNA and different kinds of RNAs, such as mRNA and microRNA (miRNA). miRNAs are small non-coding molecules able to bind to complementary sequences in the 3′-untranslated regions (3′UTRs) of target mRNAs, defining the post-transcriptional regulation of different genes ^[5][11].

Much has been learned on the content of EVs but the identification of cargo specific for small or medium/large vesicles is an ongoing area of intense research.

Because of their capacity to exchange molecules between cells, EVs act as signaling structures both under physiological and pathological conditions. Specifically, in pathological conditions, EVs might act in favor of disease, such as in multiple sclerosis ^[6][12], Alzheimer’s ^[7][8][13,14], prion ^[9][15], and Huntington’s disease ^[10][16]. These aspects, together with the identification of m/lEVs in different body fluids, increased the interest in research to elucidate the vesicle functions in different contexts ^[11][17]. In addition, EVs transport biologically active factors across the blood–brain barrier (BBB) and the choroid plexus ^[12][18], making them potential tools for the early diagnosis of neurological disease ^[13][19] and potential vehicles for non-invasive therapies ^[14][20].

According to the physical characterization and to the guidelines of the International Society of Extracellular Vesicles (ISEV) ^[15][21], m/lEVs comprise all the vesicles with a diameter larger than 200 nm. They are produced by the direct invagination of the plasma membrane and then released into the extracellular space ^[16][22].

Each cell type changes the EVs composition according to its physiological state, regardless of its size, with peculiar lipids, proteins, and nucleic acids content ^[17][23]. The process of plasma membrane blebbing is typical of m/lEVs, even if the underlying mechanisms remain incompletely understood ^[17][23].

Recently, it was reported that the process involved in membrane curvatures during EV formation requires the interaction of the arrestin domain-containing protein-1 (ARRDC1) with the endosomal tumor susceptibility protein 101 (TSG101) and that this interaction leads to the transfer of TSG101 from the endosomal compartment to the plasma membrane. The presence of TSG101 in the plasma membrane allows the release of m/lEVs containing TSG101 and ARRDC1, together with other molecules. Other mechanisms are involved in the process of plasma membrane curvature. For example, the enrichment of proteins at the cellular periphery and the pressure generated by their interaction could contribute to shaping changes and curvature. This process indicates that the enrichment of protein at the site of m/lEVs budding might be a stimulus sufficient to start vesicle formation ^[18][24]. Another important factor is the alteration in the lipid composition that modifies membrane rigidity and curvature ^[19][25], key events for m/lEV formation. Phospholipids are made up of hydrocarbon tails and large head groups that give them a conical shape, with an irregular distribution. The membrane curvature is determined by the distribution of phospholipids in the plasma membrane but also by the presence of aminophospholipid translocases, such as flippases and floppases, similarly to what is described for the Golgi vesicles ^[20][26].

Differently, sEVs are vesicles with a diameter smaller than 200 nm ^[15][21]. At the beginning of the 1980s, a pathway that includes endocytosis was described for sEV formation. It includes the internalization of extracellular ligands and cellular components that will then be transferred again to the plasma membrane and/or degraded ^[21][27]. Upon degradation, early endosomes will be the first to be produced, followed by late endosomes ^[22][28], which accumulate intraluminal vesicles (ILVs), or multivesicular endosomes (MVBs), inside. Proteins and lipids contained in the MVBs contribute to the curvature of the early endosomal membrane. Mostly, MVBs fuse with lysosomes and their content is degraded by hydrolases. In other cases, they can fuse with clusters of the plasma membrane enriched in tetraspanin CD63, lysosomal-associated membrane proteins (LAMP1 and LAMP2) and other molecules typical of late endosomes, releasing the content into the extracellular space ^[23][24][29,30].

To exert biological effects, EVs can either be internalized or activate receptor–ligand signaling on the surface of target cells. In the last few years, the mechanisms of EVs–cell interaction have also been investigated, taking advantage of drugs or antibodies to block specific signaling pathways. These studies revealed that EVs can be internalized by target cells through different mechanisms, such as clathrin-mediated endocytosis, phagocytosis, micropinocytosis and plasma or endosomal membrane fusion ^{[25][26][27][28][29]}[31,32,33,34,35], in addition to protein–protein interactions mediated by tetraspanins (as CD63, CD9 and CD81), integrins and immunoglobulins, proteoglycan, and lectins ^[30][36].

2. Microglial EVs in Brain Tumors

In GBM, microglia and infiltrating macrophages represent about 30% of total cells in the tumor mass, contributing to the early anti-tumor immune response and later playing a role in supporting cancer growth ^[31][32][57,58]. In fact, brain tumor cells attract microglia by secreting factors such as cytokines, growth factors, chemokines and colony-stimulating factors ^[32][58], which induce phenotypical and genetical switch of microglia toward a pro-tumorigenic ally. Microglia and infiltrating macrophages, modified by cancer cells, are defined as tumor-associated myeloid cells (TAMCs) ^[33][59]. In addition to secreted molecules, microglia-tumor communication is also mediated by gap junctions, tunneling nanotubes and EVs ^[34][60]. EVs make possible communication along distant sites and, importantly, also permit bidirectional cell-to-cell communication ^[34][60]. As stated before, EVs can deliver not only soluble proteins but also a wide variety of coding and non-coding RNAs that can alter the gene expression of the target cells ^[34][60].

3. EV-Mediated Communication between GBM and the Brain Tumor Environment

GBM is the most aggressive brain tumor and represents an important research and medical challenge ^[35][1]. EVs released by GBM cells have specific cargos that favor tumor propagation. They contain oncogenes, such as the epidermal growth factor receptor (EGFRvIII), that induce the expression of other oncogenes (i.e., p27 and Bcl-xL) and the activation of pro-tumoral pathways (i.e., AKT and ERK1/2 phosphorylation) ^[36][61]. Another mechanism of EV-mediated progression of glioma is the transfer of the RNA-binding motif 11 (RBM11) ^[37][62], a pro-tumoral protein that promotes invasion and proliferation ^[38][63] by apoptotic glioma cells. Additionally, EVs derived from glioma stem cells (GSCs) contribute to maintaining the GBM cellular heterogeneity (a peculiarity of high malignancy) ^[39][64]. In addition to the horizontal propagation of the oncogenic activity, glioma-released EVs also convey materials to non-tumoral cells ^[40][65], such as astrocytes, neurons, and endothelial and immune cells. Astrocytes stimulated by GBM-derived EVs acquire a tumor-supportive phenotype ^[41][6]. In addition, GBM-derived EVs enhance the proliferation and the transformation of astrocytes by RNAs able to reprogram metabolic activity ^[42][66]. Conversely, exosome-mediated transport of miR-19a from astrocytes to tumor cells critically results in PTEN downregulation and tumor growth ^[43][44][67,68]. GBM-derived EVs can regulate neuronal excitability ^[45][69] and miRs transfer (miR-148a and miR-9-5p) promotes angiogenesis in endothelial cells ^{[46][47][48][49]}[70,71,72,73]. In bidirectional communication, brain endothelial cells release EVs containing tetraspanin CD9 to GBM ^[50][74], enhancing tumor progression through the inhibition of ubiquitination of IL6 receptor gp30 and promoting the activation of the signal transducer and activator of transcription 3 (STAT3) ^{[51][52][53][54]}[75,76,77,78]. Endothelial-derived EVs overexpressing the tumor suppressor esophageal cancer-related gene-4 (ECRG4) inhibit glioma cell proliferation ^[55][79]. GBM-derived EVs suppress the activation of T lymphocytes, partially through the programmed death ligand-1 (PD-L1) ^[56][80], a key player of the tumor immune escape in many cancers including GBM ^[57][58][81,82]. Macrophages stimulated with GBM-derived EVs acquire a pro-tumoral phenotype (i.e., overexpression of Arg-1 and IL-10 and downregulation of iNOS and TNF-α), in part through miR-10b-5p ^[59][83]. Furthermore, GBM-released EVs also block the clonal proliferation of T cells by transferring CD73 ^[60][84], which inhibits aerobic glycolysis ^[61][85]. GBM-derived EVs are involved in the initiation of the tumor-supportive TAMC phenotype ^[62][86]. TAMCs efficiently engulf EVs, as visualized in vivo, in glioma-bearing mice ^[63][87]. GBM-derived EVs increase the phagocytic activity of microglia towards the extracellular matrix, creating free space in the brain parenchyma ^[64][88]. One of the mechanisms involved is the transfer of the membrane type 1-matrix metalloproteinase (MT1-MMP) ^[64][88], which degrades the proteins of the extracellular matrix. GBM-derived EVs affect TAMC proliferation both in vivo and in vitro ^[65][89]. GBM-derived EVs contain miR-21 ^[66][90] which downregulates the target tumor-suppressive gene Btg2 (B cell translocation gene 2) ^[67][91] in microglial cells. Moreover, they transfer the Wilms tumor-1 (WT1) protein to microglial cells, upregulating the expression of thrombospondin-1 (Thbs-1), a key promoter of angiogenesis ^[68][92]. GSCs represent a cell subpopulation relevant for the tumor resistance to radiations and release EVs able to polarize microglial cells towards a pro-tumoral phenotype, specifically by transferring miR-504 that inhibits the putative onco-suppressor gene grb10 (growth factor receptor-bound protein 10) ^[66][90]. miR-504 also increases the pro-tumoral markers CD209 and TGF-β and decreases the expression of the anti-tumoral marker genes CD86 and TNF-α ^[66][90]. GBM-released EVs reprogram microglia towards a pro-tumoral phenotype also through the long noncoding RNA (lncRNA) associated with temozolomide (TMZ) (lnc-TALC). lnc-TALC competes with miR-20b-3p by inducing the expression of Stat3, which, in turn, activates the expression of the DNA repair enzyme O⁶-methylguanine-DNA methyltransferase (MGMT) ^[69][93]. lnc-TALC also increases Arg-1, CD163, TGFβ, IL4, and IL10 ^[70][94], enhancing the pro-tumoral phenotype of microglia. On the other side, several pieces of evidence demonstrate that microglial-released EVs can affect the TME and the effects induced by microglial EVs vary with the activation state of these cells ^[71][95]. Specifically, m/lEVs released by LPS/INFγ-activated microglia contain transcripts for several inflammation-related genes that reduce the pro-tumoral efficacy of TAMCs. In contrast, m/lEVs released by IL4-stimulated microglia enhance the pro-tumoral phenotype of TAMCs ^[71][95]. The biogenesis process and the cargo differ in sEVs and m/lEVs. In fact, in contrast with m/lEVs, sEVs released by both LPS/INFγ and IL4-stimulated microglial cells exert anti-tumoral effects in a mouse model of glioma, reducing tumor mass and prolonging mice survival ^[72][96]. Similarly, sEVs released by unstimulated microglial cells reduce the invasion of glioma cells in 3D-spheroid cultures ^[73][97]. The anti-glioma effect of microglial-derived sEVs can be mediated by miR124. This specific miR enhances the expression of the glutamate transporter GLT1 on astrocytes, with increased clearance of glutamate in the synaptic cleft. Glioma cells are in fact able to release neurotoxic amounts of glutamate that promote neuronal death and permit the invasion of tumor cells ^[72][96]. EV-mediated delivery of miRNAs could represent a promising approach to contrast the growth of brain tumors. In a 3D-microfluidic GBM microenvironment, miR-124-loaded EVs reduce tumor growth and inhibit microglia polarization towards a pro-tumoral phenotype ^[74][98]. Furthermore, microglia EVs can be engineered to deliver drugs to glioma cells, taking advantage of their ability to target the brain tumor mass ^[75][99]. Specifically, microglia cells were engineered to release EVs enriched in paclitaxel (PTX), a pro-apoptotic compound that blocks the cell cycle at the G2/M phase ^[76][100]. Loading PTX in EVs overcomes its low BBB permeability ^[77][101] and could be used for other drugs with a reduced capability to enter brain parenchyma.