Extracellular vesicles (EV) are a very diverse group of cell-derived, lipid bilayer enclosed vesicles released by almost all kind of living cells, and this process has been highly conserved throughout evolution [1]. There are several subtypes of EV that have distinctive structural, biochemical properties and composition depending of their intracellular origin that, in turn, affect their function [2]. EV are highly heterogeneous, which in great part is responsible for hindering the characterization and description of their properties and functions [1], and are often classified in terms of their generation mechanism. The first type includes the EV released by dying apoptotic cells, which are called apoptotic bodies. Apoptotic bodies have a wide range of sizes and exhibit different compositions and features from EV derived from living cells, and are not discussed in this rentryview. The second type of EV are directly formed by outward budding of the plasma membrane and are called microvesicles, ectosomes, ectovesicles or, generically, shedding vesicles [3] (Figure 1). The third type, exosomes, are secreted via the fusion of multivesicular bodies (MVB) with the plasma membrane ^[4][5][4,5] and have diameters ranging 30–150 nm (depending of the estimation technique [6]), smaller than shedding vesicles (100 nm to >1 μm) (Figure 1). MVB are subcellular organelles containing intralumenal vesicles (ILV) ^[7][8][7,8] that are components of the endolysosomal system, which also comprises early endosomes, late endosomes and lysosomes ^[9][10][9,10]. MVB are formed by inward budding from the external membrane of late endosomes and successive pinching off of budding vesicles into the lumenal space of late endosomes (Figure 1). ILV present in MVB are called exosomes when they are released into the extracellular medium [11].

The term exosomes (initially used to define shedding vesicles with diameters ranging from 40 to 1000 nm [45]) was later adapted to define nanovesicles of endosomal source that are liberated by fusion of MVB with the plasma membrane, as a means to discard specific out of date constituents during red cell maturation [4], and since then this proposal has been widely accepted by the scientific community, although not fully standardized yet. However, growing evidence supports that EV in general, and exosomes in particular, have much wider biological functions than removal of certain unwanted proteins, and all these EV are involved in intercellular communication, both locally and systemically, since they may transfer their cargo (proteins, lipids, miRNAs) between cells, and also may trigger new cues in recipient or target cells ^{[1][10][12][46]}[1,10,12,46]. Thus, EV have been shown to affect the physiology of neighbouring target cells in diverse ways, from inducing cell signaling upon cell surface receptor triggering, to generating new properties in the target cells after acquisition of novel receptors, enzymes or genetic material contained into the EV ^[2][47][2,47].

2. Extracellular Vesicles from T Lymphocytes

IS formation by T lymphocytes subsequent to TCR binding to antigen bound to MHC on the APC surface is a very dynamic, plastic and critical outcome involved in antigen-specific, cellular and humoral immune responses ^[48][49][93,94]. IS establishment integrates signals and combines molecular interactions leading to a proper and antigen-specific immune response ^[50][95]. There are two main groups of IS built by T lymphocytes leading to quite different, although essential, immune effector functions ^[48][50][51][93,95,96]. The interaction of helper T lymphocytes (Th), generally CD4⁺ cells, with MHC-II-bearing APC causes T lymphocyte activation (cytokine secretion, proliferation, etc.). In contrast, naïve CTL, usually CD8⁺ cells, recognize antigen-associated MHC-I on APC and become activated or “primed” to proliferate in the first phase, and kill target cells bearing the antigen in the second, effector phase. In the effector phase, primed CTL similarly form IS with target cells (virus-infected cells or tumor cells) leading to specific killing. Thus, the functional outcomes produced by the formation of an effective, mature IS include activation (naïve CTL and Th lymphocytes), killing (primed CTL), and functional anergy or apoptosis induction ^[52][97] once the effector phase is finished. IS formation induces the convergence of T lymphocyte secretion granules towards the MTOC and, almost simultaneously, MTOC polarization and secretion granules move towards the central supramolecular activation cluster (cSMAC) at the IS ^[49][53][94,98]. T lymphocyte secretion granules include cytokine-containing secretion granules in Th lymphocytes, cytolytic granules/secretory lysosomes in CTL, and MVB in Th lymphocytes and CTL. This dedicated mechanism appears to specifically endow the immune system with a superbly tuned tactic to enhance the efficiency of decisive secretory effector roles of T lymphocytes, while diminishing nonspecific, cytokine-controlled stimulation, target cell killing and activation induced cell death (AICD) of bystander cells ^[54][99]. CTL cytolytic granules are secretory lysosomes that have an MVB structure, and their degranulation causes ILV secretion as nanosize “extracellular vesicles” at the synaptic cleft made at the CTL-target cell interface, formerly described by Peters et al. ^[55][100]. Although CTL-secreted vesicles were not referred in this study as canonic exosomes, their creation and mode of exocytosis warrants this classification [10] (see above). Cytolytic granules contain perforin and granzymes that are located in a lumenal, electron dense core characterized by EM and also accumulate in the ILV ^[56][77]. In addition, it was shown that ILV and their derived exosomes contained, apart from the proapoptotic proteins perforin and granzymes, the exosome marker CD63 and molecules relevant for CTL-target cell interaction, such as TCR and CD8 ^[56][55][77,100], demonstrating that most of the cytotoxic factors exocytosed into the cleft between CTL and the target cell are membrane-enveloped or exosome-associated. However, release of perforin and granzymes in soluble form coming from the lumenal core cannot be excluded ^[56][57][77,101]. Perforin, which is inactive in the acidic environment of secretory lysosomes, is activated by neutral pH and Ca²⁺ at the synapse and polymerizes and forms a transmembrane pore that allows the entry of granzymes into the target cell; granzymes trigger caspase-dependent and independent cell death [44]. It was hypothesized that the presence of the TCR complex, CD8, and possibly other relevant molecules on these nanovesicles displaying their extracellular portions facing outwards, may ensure unidirectional delivery of lethal factors to target cells, since it was proposed that ILV released into the synaptic cleft bind specifically to the relevant antigen-MHC-I complex on the target cell membrane, and not to the CTL itself or to bystander cells ^[58][79]. This model would explain not only why a CTL does not kill itself, but also why bystander cells, which are in close proximity but do not bear the proper antigen, are not killed ^[58][79]. Subsequently, it was demonstrated that newly synthesized Fas ligand (FasL) is also stored in the limiting membrane of CTL secretory lysosomes and that polarized degranulation controls FasL delivery to the T cell surface, which is consistent with the role of a FasL-dependent pathway in CTL-mediated cytotoxicity ^[59][102]. In addition, it was shown that FasL can also be sorted from the MVB limiting membrane to ILV and hence to exosomes upon T lymphocyte activation, since T lymphocyte activation induced 100–200 nm “microvesicles” secretion including proapoptotic FasL and Apo2L ^[60][103] (Figure 2 and Figure 3). These microvesicles were subsequently characterized as canonic exosomes, since they arose from FasL⁺Apo2L⁺ ILV after MVB fusion with the plasma membrane ^[61][56]. Exosomal FasL and Apo2L, with the same topology as cell surface FasL and Apo2L, can bind to their respective death receptors on the surface of target cells, or effector T lymphocytes themselves, inducing caspase-dependent apoptosis ^[62][63][104,105] (Figure 3). Proapoptotic exosomes are thus involved in AICD of effector T lymphocytes, which constitutes an important suicide or fratricide mechanism participating in the downregulation of T cell-dependent immune responses ^[61][60][64][56,103,106]. Another major contribution was to demonstrate that inducible, polarized exosome secretion occurred at the IS formed by living Th lymphocytes and APC [18], as occurred in the IS formed by CTL ^[56][55][77,100] (Figure 3). The hypothesis derived from all these publications that TCR activation of the effector T cell may induce the release of CD63⁺ exosomes bearing TCR was formally demonstrated by using TCR agonists to activate Jurkat Th, CTL and CD4⁺ lymphocytes ^[19][69][19,54]. Taken together, these reports constitute a major milestone in the exosome field since they demonstrate that TCR stimulation triggers inducible exosome secretion by T lymphocytes (both in CTL and Th cells) ^[2][12][2,12]. However, it is remarkable that, depending on the stimulation regime (absence or presence of co-stimulation signals), CD4⁺ T cell activation promotes the differential release of distinct EV subpopulations ^[70][110]. Adding more complexity to the EV field, CD63-enriched shedding vesicles or ectosomes directly budding from the Th lymphocyte plasma membrane and accumulating at the IS formed with a B lymphocytes acting as APC have been described [42] (Figure 3). These synapse-induced shedding vesicles were enriched in TCR and, upon endocytosis by APC, were capable of signaling via pMHC-II stimulation ^[42][71][42,111] (Figure 3), suggesting these synaptic ectosomes may facilitate the activation of B cells and other APC presenting the cognate pMHC-II. In this report, centrally accumulated TCR were located on the surface of extracellular microvesicles that bud at the IS centre. Members of the ESCRT-I family such as TSG101 sort TCR for inclusion in shedding vesicles, whereas VPS4 mediates microvesicle scission from the T-cell plasma membrane. TSG101 interference reduced EV production, whereas VPS4 function disruption rendered budding vesicles unable to undergo fission from the plasma membrane [42] (Figure 3). However, neither the existence of T cell-derived shedding vesicles in the CTL IS, nor a proapoptotic role for these EV, has been demonstrated yet. In addition, although nSMase has been shown to be involved in plasma membrane budding in several cell types ^[13][72][13,91], its participation in EV release from T lymphocytes has not been reported yet, in contrast with its role in ILV biogenesis ^[73][59].

3. Chimeric Antigen Receptor (CAR) T Cells and CAR T Cell-Derived EV

3.1 Cancer Therapeutic Approaches

Cancer Therapeutic Approaches

The role of EV from immune cells, including T lymphocytes, in anti-tumor immunity has been recently and exhaustively reviewed ^[74][75][130,131]; therefore, reswearcher focus on potential therapeutic uses of CAR T cell-derived EV. CAR T lymphocyte-based immunotherapy has proven to be a promising treatment of patients suffering several refractory cancer diseases ^[76][132]. CAR consists of an extracellular domain that confers antigen-recognition specificity, a transmembrane domain, and an intracellular signaling domain (CD3ζ) that provides activation signals to T lymphocytes ^[76][132]. However, several challenges preclude the use of adoptively-transferred CAR T cells, including their low efficacy against solid tumors, immunosuppression by tumor microenvironment, poor T cell persistence, T cell dysfunction or exhaustion, cytokine release syndrome (CRS) ^[76][77][132,133] and immune effector cell-associated neurotoxicity syndrome (ICANS). CRS is associated with supraphysiologic cytokine production and massive in vivo T cell expansion ^[78][134], whereas the cause of ICANS remains poorly understood, although appears to be related with direct central nervous system toxicity by the CR T cells, diffusion of inflammatory cytokines through the blood-brain barrier and the disfunction of this barrier caused by CAR T cells and/or cytokines ^[79][135]. It is out of the scope of this rentryview to summarize the preclinical and clinical uses of CAR T cells for cancer therapy (there are more than 250 clinical trials testing CAR T cells; please refer to recent and superb reviews on this topic ^[76][77][132,133]), thus researchwer will focus on trials using CAR T cell-derived EV.

Considering the early hypothesis that the presence of both TCR and proapoptotic molecules (FasL, Apo2L, perforin, granzymes A and B) on T cell-derived exosomes would confer on them both antigenic specificity and guiding cytotoxicity ^[56][58][55][77,79,100], making them potent vectors to deliver proapoptotic cues to target cells bearing the cognate antigen, several strategies have been developed to test the use of CAR T cell-derived exosomes for cancer therapy. The fact that TCR activation boosts the secretion of CTL-produced exosomes, and the presence of the TCR/CD3ζ complex in these exosomes (see above) ^[69][54] would reinforce this approach. TCR/CD3ζ complexes endocytosed after recognition of the pMHC-II complexes are targeted to MVB, then to ILV and hence exosomes ^[69][54]. A crucial prerequisite for using CAR T cell-derived exosomes to specifically induce tumor cell death is the presence on exosomes of the CAR molecule, since its antibody-derived, antigen-binding variable fragment endows them with tumor cell specificity, as occurs with CAR T cells ^[76][80][132,136]. Since CAR is an artificial molecule containing the CD3ζ intracellular signaling and localization domain, it was unknown whether CAR was present in exosomes and shedding vesicles, and whether CAR conferred a specific cytotoxic effect, until recent reports characterized CAR expression and function in exosomes ^[81][55] and shedding vesicles ^[82][137]. This prerequisite has been endorsed in several preclinical studies by using exosomes and/or EV produced by CAR T cells ^[81][83][84][55,80,138] (Table 1).

Table 1.

 CAR T cell and cell-derived EV: preclinical studies.

Target Molecule	EV-Producing Cell	EV Types	EV Phenotype	Anti-Tumor Mechanism	Target Cell
EGFR, HER2 [81]	EGFR, HER2 [55]	Human CAR T cells (?)	1	Exosomes	CAR	+	, CD3	+	, CD63	+	, perforin	+	, granzyme B	+	, CD45	−	, CD28	−	Perforin/ granzyme B	2	EGFR	+	, HER2	+	human breast cancer cells
HER2 [84]	HER2 [138]	Human CAR T cells CD4	+	(46%) CD8	+	(49%)	EV (small EV, probably exosomes plus larger EV)	3	CAR	+	, CD3	+	, CD63	+,	granzyme B	+	Granzyme B	2	HER2	+	human breast cancer cells, ovarian cancer cells
Mesothelin [83]	Mesothelin [80]	Human CAR T cells CD4	+	(58%) CD8	+	(31%)	−Probably exosomes	4	CAR	+	, CD3	+	, CD63	+	, perforin	+	, granzyme B	+	Perforin/ granzyme B	2	Triple negative human breast cancer cells
++	CD19 [85]	CD19 [139]	Human CAR HEK293 cells	Probably exosomes	4	CAR	+	, CD63	+	, CD81	+	Indirect induction of proapoptotic genes in target cells	CD19	+	human B cell leukemia
Efficiency against solid tumors	+/−	++	CD19 [82]	CD19 [137]	Human CAR HEK293 cells	Probably shedding vesicles	4	CAR	+	, annexin V binding (PS exposure)	MYC Gene disruption mediated by CRISPR/Cas9	CD19	+	human B cell leukemia cell lines
Immunosuppression by tumoral PD-L1	+	−	Mesothelin CD19 [86]	Mesothelin CD19 [140]	Human and mouse CAR T Cells (?)	1	EV	4	Unknown	5	Contain RN7SL1	Recruitment of endogenous anti-tumor immunity byRN7SL1	Mouse melanoma expressing human CD19

¹ No data available regarding CD4⁺/CD8⁺ subpopulations. ² Proposed mechanism, but not formally demonstrated. ³ Unfractionated EV; more criteria to differentiate exosomes from shedding vesicles are needed. ⁴ More criteria to differentiate exosomes from shedding vesicles are needed. ⁵ No phenotypic analyses of EV were performed.

CAR T cell-derived exosomes exhibit excellent capability for use as direct aggressors in immunotherapy, since ex vivo-produced human exosomes transporting human EGFR and HER2-specific CAR have powerful in vivo activity against EGFR⁺ and HER2⁺ human tumor cells in xenograft models ^[81][55]. These CAR T cell-derived exosomes specifically induce apoptosis of tumor cells expressing the antigens recognized by CAR on the cell surface but do not kill tumor cells that do not express these antigens ^[81][55]. Comparable results have been obtained using HEK-derived exosomes, showing that only when exosome entry into cells is mediated via binding to the CD19 antigen on the surface of CD19⁺ B-cells does pro-apoptotic signaling occur resulting in selective cytotoxicity ^[83][84][80,138]. Certainly, the evoked apoptotic mechanism does not depend on FasL, Apo2L, perforin and granzymes, as for CAR T-cell derived exosomes, since HEK293 cells do not express these molecules. More recently, an interesting paper demonstrated an in vivo role of the RNA component of signal recognition particle 7SL1 (RN7SL1, a non-coding RNA that activates interferon-IFN-stimulated genes) contained in EV from CAR T cells in tumor rejection ^[86][140]. In this approach, the CAR construct contains downstream of the CAR sequence a U6 promoter that drives the transcription of human RN7SL1 and, optionally, a 5′LTR promoter driving the transcription of a peptide antigen. The authors demonstrate that RN7LS1-containing CAR T cells deliver RN7SL1 in EV in vivo, that orchestrates endogenous immune activation to improve responses against the tumor. EV released by CAR T cells upon CAR engagement transfer RN7SL1 to endogenous immune cells (myeloid cells, DC and T cells), but not to the tumor cells, and EV release and RN7LS1 transfer is inhibited by a nSMase inhibitor. RN7LS1 inside target cells activates a RIG-I-dependent, IFN-mediated inflammatory response and induces transcription of IFN-stimulated genes. RN7LS1 delivery in EV to immune cells improves immunostimulatory properties of myeloid and DC cells that, in turn, effectively activate the function of endogenous CD8 T cells against the tumor. All these immune cells, acting together, may trigger solid tumor rejection even in case of CAR-recognized antigen loss by the tumor. Although in this research neither EV nature or composition, nor the presence and contribution of CAR and proapoptotic proteins in the EV to tumor rejection, have been established, this strategy opens new venues based on improved endogenous immunity against tumor cells conferred by EV, since CAR T cells can now co-deploy antigenic peptides with RN7SL1 released by EV to enhance their efficacy against tumor cells, even when tumors lack adequate neoantigens ^[86][140]. Table 1 summarizes the most relevant features of the preclinical trials involving CAR T cell and HEK293-derived EV. Most of these trials have been performed ex vivo or in vivo using xenograft models in mice.

The advantages of CAR T cell-derived exosomes as cell-free immunotherapy are their independence of CAR T cell life span and division, their stability, some obvious logistic issues, the low risk of collateral toxicity (i.e., CRS incidence) when contrasted to CAR T cells, and the fact that exosomes lacking PD-1 (in contrast to PD-1-expressing T cells) are refractory to PD-L1 immunosupression by the tumor ^[81][55] (Table 2). In this context, it is remarkable that CAR T cell-induced CRS is one of the most harmful complications that follow infusion of the CAR T cells and occurs in approximately two thirds of CAR T cell recipients, generally within 10 days after cell infusion. This life-threatening complication is generally ascribed to uncontrolled release of cytokines from CAR T cells ^[80][136]. Moreover, exosomes could be distributed through the blood circulation and other body fluids as supported by the abundance of exosomes found in most body fluids. In addition, exosomes have the ability to cross certain biological barriers, such as the blood–brain barrier and blood-tumor barrier, as documented by the presence of tumor cell-originated exosomes in body fluids ^[80][136] (Table 2).

Table 2.

 CAR T cell and CAR T cell-derived EV. A comparison.

Event	CAR T Cells	CAR T Cell-Derived EV
Cytokine releasing syndrome	++	−
Neurotoxicity	++	−
Cross the blood barrier
Immunological memory
+	1	(?)	2

¹ Depends on the use of central memory or effector memory CAR T cells. ² Not formally established.

The fact that such a cell-free “exosome therapy” involves the on-bench stimulation and expansion of the effector CAR T cells ^[81][55] may allow the genetic modification and/or pharmacologic treatment of the cultured cells to increase T cell activation, and/or exosome biogenesis, to enhance EV production. In addition, exosome collection and purification ex vivo involve the elimination of effector CAR T cells, and also bystander or contaminating cells, before exosome infusion ^[81][55]. Thus, this strategy would circumvent the undesirable possibility of transducing, for instance, CAR to residual tumoral cells during T cell manufacturing that may lead to provocation of resistance to CAR T cell therapy by unintentional transduction of a single leukemic B cell, as reported in ^[87][141]. These findings illustrate the need for purging residual contaminating tumor cells from engineered CAR T cells or using alternatives such as cell-free, exosome-based therapies.

Taking into account that it is not known whether CAR signaling or the involvement of any of the exosomal proapoptotic molecules (perforin, granzymes, FasL, etc.) mediate the therapeutic effects described, these findings require further confirmation, and to formally establish the contribution of these molecules, although they convincingly support the use of exosomes as biomimetic nanovesicles in antitumor therapy ^[81][55]. Recent reviews have dealt with the proapoptotic mechanisms evoked by the proapoptotic molecules present in EV on target cells; please refer to these for further details ^[74][88][130,142].

It is remarkable that most of the strategies directed to unveil the role of the molecular components involved in exosome biogenesis/degradation and/or release in T cells (ESCRT, tetraspanins, Rabs, MAL, ISGylation, SNAREs, etc.) are mainly based on either the expression of dominant negative mutants or RNA interference, which leads, in a vast majority of the approaches, to EV secretion inhibition ^[6][89][6,64]. Thus, although these approaches have been useful in establishing the necessity of diverse components for exosome secretion, very few of these approaches have led to an increase of exosome secretion ^{[17][18][19][89][90]}[17,18,19,64,143]. In principle, the positive modulation of exosome secretion ex vivo could be a useful strategy to enhance the effectiveness of exosomes for subsequent in vivo applications. Among all the potential approaches designed to enhance on bench exosome secretion, perhaps the regulation of the lipid pathways and their metabolites (DAG, ceramide) involved in exosome biogenesis appears the most feasible approach, due the existence of quite specific pharmacologic agents suitable for exosome induction ex vivo. Thus, increasing DAG levels using DGK inhibitors (i.e., R59949) [19], or ceramide levels using sphingomyelin synthase inhibitors (i.e., D609) ^[91][60], which have been shown to increase exosome secretion, may constitute useful tools for this strategy.

Although the timing and modes of activation and maturation are different, both CTL and Natural Killer cells (NK) utilize an overlapping arsenal consisting of cytotoxic effector proteins including FasL ^[59][102], perforins, granzymes and granulysin contained in their secretory lysosomes ^[92][93][78,144] (recently reviewed in ^[88][142]). The molecular mechanisms controlling the maturation and traffic of the secretory lysosomes and the polarized secretion of exosomes towards the synapse are, in great part, common to both cell types ^[44][59][94][44,102,145]. Thus, it is conceivable that approaches directed to increase exosome secretion in CTL ex vivo can also be useful to boost exosome secretion by CAR NK cells. However, although TCR activation enhances the constitutively low secretion of exosomes by T lymphocytes, resting NK cells secrete pro-apoptotic exosomes with no differences in the amounts of exosomes or marker expression relative to activated NK cells via NK activating and inhibition receptors ^[92][95][78,146]. NK cell expansion ex vivo increases exosome secretion ^[95][146], as occurs in T lymphocytes, which constitutes an useful strategy to produce on a large scale exosomes that may lead to new preclinical and clinical applications ^[95][146]. Different to T lymphocytes, NK cell recognition of their targets is not controlled by antigen specificity but rather through the integration of signals evoked by activating and inhibitory receptors, activated by a myriad of ligands in the target cells, which requires a deeper knowledge of the complex NK biology and signaling before any experimental design. This fact probably has led to CAR NK therapy being less developed than CAR T cell therapy in general ^[96][147], and therefore there are less pre-clinical trials using CAR NK-derived EV in particular ^[95][146]. However, the use of CAR NK cells has some advantages (and disadvantages) when compared with CAR T cells ^[96][147]. Although the use of CAR NK cells in clinical trials has been less extended in comparison to CAR T cells, in principle it is conceivable the use of NK-derived EV in future anti-tumor therapies.