Extracellular vesicles as emerging targets in cancer: recent development from bench to bedside

Abstract

Extracellular vesicles (EVs) have emerged as important players of cancer initiation and progression through cell-cell communication. They have been recognized as critical mediators of extracellular communications, which promote transformation, growth invasion, and drug-resistance of cancer cells. Interestingly, the secretion and uptake of EVs are regulated in a more controlled manner than previously anticipated. EVs are classified into three groups, (i) exosomes, (ii) microvesicles (MVs), and (iii) apoptotic bodies (ABs), based on their sizes and origins, and novel technologies to isolate and distinguish these EVs are evolving. The biologically functional molecules harbored in these EVs, including nucleic acids, lipids, and proteins, have been shown to induce key signaling pathways in both tumor and tumor microenvironment (TME) cells for exacerbating tumor development. While tumor cell-derived EVs are capable of reprogramming stromal cells to generate a proper tumor cell niche, stromal-derived EVs profoundly affect the growth, resistance, and stem cell properties of tumor cells. This review summarizes and discusses these reciprocal communications through EVs in different types of cancers. Further understanding of the patcment of tumor specific therapeutics. This review will also discuss the translational aspects of EVs and therapeutic opportunities of utilizing EVs in different cancer types.

2. Characteristics and biogenesis of EVs

2.1. EVs are defined as exosomes, microvesicles, and apoptotic bodies

EVs are a family of membrane-wrapped vesicles released from cells to the extracellular space. They are known to transport messages from donor cells to recipient cells, mediating intercellular communications (1). The discovery of EVs can be dated back to 1967 when Peter Wolf identified “Platelet dust” by electron microscopy as a subcellular fraction derived from platelets (2). The first annotation of EVs in cancer patients was reported in 1978 by Friend et al. (3). They described the presence of “rare, pleomorphic membrane-lined particles” in the extracellular space of a cell line established from a patient with Hodgkin’s disease. Two years later, Poste et al. found that fusion of plasma membrane vesicles from a highly metastatic melanoma cell line could increase the lung metastasis rate of an otherwise poorly metastatic melanoma cell line (4), suggesting a regulatory role of EVs in recipient cells. However, little attention was paid to these tiny particles until 1983, when two studies showing the recycling of transferrin receptors (TfRs) in reticulocytes through EVs were published consecutively (5, 6). Using labeled antibodies, Johnstone et al. found that TfRs were selectively packaged into vesicles released by cells during the maturation of reticulocytes. Just one month later, Harding et al. reported close connection of endocytosis with the recycling of TfRs in reticulocytes. Their work suggested that transferrin receptors could be either recycled to the membrane through multivesicular endosomes (MVE) or released to the extracellular space in the form of vesicles derived from MVE. More importantly, the balance shifted from recycle to release during the maturation of reticulocytes, indicating a regulatory mechanism for vesicle secretion and cargo sorting. This work provided the first evidence that EVs are an essential part of the biological mechanisms that maintain cellular activity. In 1987, Johnstone redefined “exosome” to refer to these vesicles with a diameter of ~50 nm (7, 8). Thereafter, several studies reported the functional behaviors of EVs in diverse types of cells. In 1996, Raposo et al. found that B lymphocytes secrete major histocompatibility complex (MHC) class II bound exosomes, which induced antigen-specific T cell responses, suggesting their roles in immune system mobilization (9). This pioneering work opened a new era of EV research in immunology. In 2006, it was reported that exosomes and MVs from embryonic stem cells enclosed enriched amount of mRNAs of several pluripotent transcription factors. These mRNAs are engulfed and translated into proteins by hematopoietic progenitor cells to enhance survival and expansion (10). This finding suggests that cells can exchange genetic information through EVs and exert their biological functions. Furthermore, Valadi et al. found that exosomes also contain microRNAs (miRNAs) that can be transferred into recipient cells (11). They also observed packaging of selective nucleic acids into exosomes, which indicated a potential sorting process during exosome loading. Recently, EVs were also found to contain both mitochondrial DNAs and chromosomal DNAs (12–14), which could indicate a novel mechanism of genetic exchange from donor cells to recipient cells. In conclusion, the presence of nucleic acid in EVs provides an opportunity for identifying new diagnostic tools for diseases with genetic mutations, such as cancer (15).

Given the relatively short history of EVs, their nomenclature has yet to be unified. Many different names have been used to refer to these particles, including microparticles, plasma membrane vesicles, membrane blebs, and ectosomes (16). Currently, a widely accepted standard classification separates the EVs into three groups according to their cellular origins: exosomes, MVs, and ABs. Exosomes are generated through the endocytic pathway. Invaginations of late endosomes lead to the formation of MVBs containing small vesicles. Fusion of MVBs with the plasma membrane releases exosomes to the extracellular space. Exosomes are generally saucer-shaped, between 40 and 100 nm in diameter (17). Exosome membrane is a lipid bilayer comprising enriched sphingomyelin and decreased phosphatidylcholine, with a random distribution of phosphatidylethanolamines as compared to the plasma membrane (18). In terms of cargoes carried by exosomes, proteins such as tetraspanins (e.g. CD63, CD9, and CD81), endosome-associated proteins (e.g. Alix and Tsg101) and antigenic peptide binding proteins (e.g., HSC70 and HSP90) have been found to be enriched in exosomes (19–22), and they are widely used as markers of exosomes (17, 23). However, their expressions are not restricted to exosomes since they can be detected in other vesicles (24, 25), and in some cases, they are more encriched in MVs than exosomes (25, 26). Nucleic acid is another important content carried by exosomes. Functional mRNAs and miRNAs have been reported as important intercellular messengers associated with exosomes (11, 27). Other than RNA, both mtDNA and genomic DNA have been found in exosomes, although there is not sufficient evidence to support the idea that they are actively functional instead of simply passengers (12–15). In addition to exosomes, cells also release EVs by directly budding and fission of the plasma membrane. This class of vesicles, termed MVs, have a diameter ranging from 100–1000 nm (23, 28). MVs have a similar composition of lipids as the plasma membrane but lack the asymmetric distribution seen in the plasma membrane (29). MVs also carry proteins, mRNAs, and miRNAs. Currently, there is no clear evidence indicating a specific marker or cargo associated with MVs compared to exosomes (30). A potential exclusive phenotype of MVs may be a lack of DNA since there is no clear evidence indicating the presence of DNA in MVs except studies with inconsistencies in the nomenclature used to indicate exosomes and MVs. The third category of EVs is ABs. Unlike other EVs, ABs are not released during the normal cellular process. They are only formed by the fragmentation of cells during apoptosis, a process of programmed cell death. As a result, ABs consists of cytoplasm, intact organelles, as well as nuclear fragments (31), which indicates that ABs carry cellular constituents inherited from apoptotic cells. The major significance of ABs is the obliteration of dead cells without ensuing inflammatory reactions since ABs are subsequently phagocytized by other cells. However, it has also been reported that ABs are indeed involved in intercellular communications (32–34).

The field of EV is rapidly developing, and new types of EVs are emerging such as large oncosome (35) and spheresome (36). Large oncosome is a group of EVs with atypically large size (1–10 μm in diameter). Despite the difference in size, large oncosome shares similarities with MV in biogenesis. Similar to MV, large oncosome is also a type of membrane-shedded vesicle that is originated from plasma membrane outward budding. The production of large oncosome is associated with amoeboid migration of cancer cells. The non-apoptotic membrane blebs, which are extruded and retracted to drive the amoeboid movement of cancer cells, were found to budded from palsma membrane to form large oncosomes. Interestingly, biogenesis of MV has also been found to be promoted by RhoA-induced amoeboid phenotype of cancer cells (37). Regradless of these similarities, large oncosome has its unique features. It is specifically generated by cancer cells (38), and its formation can be stimulated by activation of epidermal growth factor receptor and Akt1 (39). Further analysis revealed a link between chromosomal loss of DRF3 and biogenesis of large oncosome. It was found that actin nucleating protein Drf3 could inhibit the release of large oncosome from cancer cells, and chromosomal DRF3 locus is frequently deleted in metastatic prostate cancer (PCa). Another new type of EV is termed spheresome. Spheresome was first identidied in the tissue of gastrointestinal stromal tumors by scanning transmission electron microscopy (36). These vesicles have lipid bilayer membrane with 40–125 nm in diameter. Distinct from exosomes and MVs, spheresomes are not released through fusion of MVBs or plasma membrane budding, and they are released as a cluster wrapped by a large sphere with lipid bilayer membrane. This sphere is around 0.5 to 1.5 μm in diameter and named multivesicular spheres. Spheresomes are further released by multivesicular spheres in the extracellular space. These findings of new EVs released by cancer cells are particularly exciting as they may reveal novel mechanisms of cancer development and help to identify new biomarkers with high specificity.

2.2. Biogenesis, uptake, and cargo sorting of EVs are tightly regulated

The biogenesis of EVs is tightly controlled by cells through various mechanisms. Exosomes can be generated by the Endosomal Sorting Complex Required for Transport (ESCRT)-dependent or -independent pathways (40). Cancer cells modulate these pathways to facilitate growth and metastasis. For example, metastatic cancer cells were found to deliver angiogenic miR-210 through neutral sphingomyelinase 2 (nSMase2)-dependent exosome biogenesis (41). The MV biogenesis was also found to be well controlled in cancer cells. Tumor cells are capable of switching between MV-secreting phenotype and invadopodia phenotype to facilitate invasion when a different microenvironment was encountered (37). On firm matrices, Rac1 is activated and cancer cells display mesenchymal mode of migration with invadopodia formation, whereas on compliant and deformable matrices, Rac1 is inactivated and RhoA is activated. Cancer cells adopt an amoeboid phenotype with the release of MVs, which is dependent on Rho-mediated regulation of myosin light chain phosphatase. Hypoxia was found to induce the release of both exosomes and MVs (42, 43). HIF induced under hypoxia promotes RAB22A expression, which positively regulates MV formation (42). The uptake of EVs in cancer cells is also found to be tightly regulated. It was shown that the uptake of exosomes in hepatic stellate cells was dependent on integrin αvβ3, integrin α5β1, or heparan sulfate proteoglycan (HSPG) (44). Mantle cell lymphoma-derived exosomes were found to be selectively incorporated by malignant and non-malignant B-lymphocytes, but not by T lymphocytes or NK cells. This selectivity is mediated by the cholesterol/lipid rafts-dependent endocytosis pathway (45). An important role of lipid rafts is concentrating signaling proteins into this microdomain, one of which is annexin, which was shown to mediate exosome endocytosis (46, 47). Since the regulatory mechanisms of EV biogenesis and uptake have been discussed in detail previously (17, 48–50), this review focuses on cargo sorting in EVs, particularly in exosomes and MVs. ABs are not discussed due to their nature as cell fragments.

Exosomes have been found to selectively incorporate their cargo. For example, as a major component in exosomes, miRNA has been found to be selectively enriched in exosomes. Compared to that in parental cells, the proportion of miRNA is higher in exosomes (51). Different miRNAs can be packaged into exosomes at different efficiencies (52). Astonishingly, cancer cells can take advantage of these sorting mechanisms. Exosomes derived from metastatic cells were shown to have higher enrichment of miRNAs compared to exosomes from non-metastatic cells (53). This may be contributed by a compensatory mechanism in which excess miRNAs were sorted into exosomes depending on the abundance of target genes (54). However, this selective process needs the involvement of machinery proteins capable of sending particular cargoes into exosomes. Currently, such sorting machineries can be divided into four categories: ESCRT-related sorting, lipid-related sorting, membrane protein-related sorting, and others (Figure 1).

Figure 1. Pathways of cargo sorting in exosomes.

A, ESCRT-related sorting. Proteins belonging to ESCRT family regulate cargo sorting from cells to exosomes. Binding of syndencan to ALIX-ESCRT complex results in incorporation of syndecan, VEGF and HGF. SIMPLE affects CD63 and ALIX expression in exosome. VPS proteins controls loading of selective cargoes such as THPO and ANGPTLs. B, Lipid-related sorting. The nSMase catalyzes ceramide formation from sphingolipids. Ceramide is associated with Ago2 and miRNA sorting. C, Membrane protein-regulated sorting. Tetraspanins and integral membrane proteins regulate exosomal cargo sorting. Membrane anchors also control the selectivity of exosomal cargo. D, Other sorting mechanisms. Other proteins (such as hnRNPs) and modification (uridylation of miRNA and N-linked glycosylation of protein) also regulate cargo sorting.

a. ESCRT-related sorting

ESCRT-related machineries are important components of endocytic recycling pathways and were found to be involved in exosome biogenesis (55, 56). These components were also reported to involve the cargo sorting into exosomes (Figure 1A). For example, syndecan was found to be sorted into exosomes by binding to syntenin, which formed a complex with ALIX and ESCRT-III (57–59). Heparanase, an enzyme that trims heparan sulfate chains on syndecan, promotes internalization of syndecans by regulating engagements and increasing binding to the complex. This complex is also capable of sorting several other cargoes, including CD63 correlated with exosome secretion and higher loading of proteins such as syndecan-1, VEGF, and HGF (61). Vacuolar Protein Sorting (VPS) proteins are also important components of ESCRT machineries (62). VPS33B was found to regulate the production of total exosomes by interacting with the GDI2/RAB11A/RAB27A pathway. Furthermore, it was found to be associated with loading of selective cargoes, such as thrombopoietin (THPO), and angiopoietin-like proteins (ANGPTLs) into secretory exosomes, while exosomal IGFBP2 was not affected by VPS33B knockdown (63). The small integral membrane protein of the lysosome/late endosome (SIMPLE), a functional partner of ESCRT, was found to reside within the intraluminal vesicles of MVBs and exosomes. Overexpression of SIMPLE increased the exosome yield, as well as the expression of exosomal marker proteins such as CD63 and Alix (64).

b. Lipid-related sorting

Another sorting mechanism is based on lipid raft-based microdomains (Figure 1B). These microdomains are enriched with sphingolipids, cholesterol, cytoplasmic proteins covalently modified by saturated fatty acids, and cell surface proteins attached via a glycosyl phosphatidylinositol anchor (65). These components of lipid raft microdomains were also found in the membranes of exosomes (66). More importantly, the high concentration of sphingolipids resulted in ceramide formation via nSMase, which was shown to be involved in cargo sorting during exosome biogenesis (67). In addition, production of exosomes and secretion of miRNAs was found to be regulated by nSMase2 (68). Kosaka et al. found that knockdown of nSMase2 or treatment with nSMase2 inhibitor decreased both total exosome production and exosomal miRNA expression (41, 68). In neuronal cell GT1–7, the prion protein was found to be selectively sorted into exosomes by nSMase2, while the sorting of the disease associated misfolded prion protein was independent of nSMase2 (69). Furthermore, in colorectal cancer (CRC) cells, the KRAS mutation is involved in differential miRNA sorting from parental cells into exosomes (70). miR-100 level was increased in exosomes derived from KRAS-mutant CRC cells while miR-10b was selectively enriched in wild-type exosomes. A subsequent study revealed that activated KRAS induced Ago2 phosphorylation, resulting in suppressed Ago2 interaction with endosomes and secretion into exosomes (71). With the deficiency in Ago2-based miRNA sorting, an alternative sorting mechanism must be present in KRAS-mutant cells. It was found that nSMase2 inhibitor could decrease miR-100 secretion through exosomes in a mutant KRAS-dependent manner, suggesting the functional link between mutant KRAS and nSMase2 in miRNA sorting. More importantly, there is evidence showing selectivity of nSMase2-mediated miRNA sorting. Inhibition of nSMase2 in KRAS-wild type cells did not cause retention of miR-10b in cells. These results imply distinct mechanisms present in cells for cargo sorting.

c. Membrane protein-regulated sorting

The family of tetraspanin is considered a marker of exosomes, and tetraspanin-enriched microdomains have a role in protein cargo selection in exosome biogenesis. By mass spectrometry, intracellular interactome of tetraspanin-enriched microdomains was shown to have a considerable overlap with protein cargoes in exosomes from human lymphoblasts. When CD81 was knocked out, CD81-associated proteins such as Rac were not present in exosomes (72). Sorting of metalloprotease CD10 into exosomes was also found to be dependent on tetraspanin protein CD9 (73). Moreover, Epstein Barr virus (EBV)-associated oncogene-latent membrane protein 1 (LMP1) was shown to be sorted into exosomes from EBV-infected lymphoblastoid cells in a CD63-dependent manner. The direct interaction between CD63 and LMP1 resulted in endosomal-exosomal trafficking of this protein complex and activation of the NF-κB signaling pathway (74). In addition to tetraspanins, other plasma membrane proteins and membrane trafficking modifications are also known to regulate cargo sorting (Figure 1C). Attachment of plasma membrane anchors could incorporate highly oligomeric cytoplasmic TyA into exosomes and MVs (75). Interestingly, the integral membrane protein CD43 was also shown to direct the sorting of proteins such as Dicer into exosomes (53, 75).

d. Other sorting mechanisms

Several other proteins, including the RNA binding protein heterogeneous nuclear ribonucleoprotein (hnRNP), have been found to be involved in cargo sorting into exosomes (Figure 1D). The sumoylated hnRNPs were revealed to regulate the sorting of miRNAs into exosomes. Sumoylated hnRNPA2B1 recognizes CCGA or CCCU motifs in miRNAs and thus mediates sorting of miRNAs into exosomes derived from T lymphocytes (76). Another hnRNP, hnRNP-Q, was also found to regulate miRNA loading into exosomes in hepatocyte. GGCU, a common sequence motif in miRNA, is recognized by hnRNP-Q and thus sorted into exosomes (77). The sorting of long non-coding RNAs (lncRNAs) is also regulated by hnRNP. In renal cancer, lncARSR was found to be sorted by hnRNPA2B1 into exosomes. Incorporation of exosomes by recipient cells increased the Sunitinib resistance, as lncARSR competitively bound miR-34 and miR-449 to promote AXL and c-MET expression (78). Major vault protein (MVP) was also found to regulate the sorting of miR-193a into exosomes from colon cancer cells (79). Furthermore, Y-box protein 1 was shown to sort selective miRNA populations, including miR-223, into exosomes from HEK293T cells (80). Interestingly, the TfRs were known to be released from exosomes during the maturation of erythrocytes. It was found that the Hsc70 sorts TfRs into exosomes by interacting with the cytosolic domains of the TfRs (81). Nedd4 family-interacting protein 1 (Ndfip1) was also found to increase protein loading into exosomes. Nedd4, Nedd4-2, and Itch, which were absent from exosomes, were detected in exosomes after expression of Ndfip1 in parental cells (82). The modification of cargoes also has a regulatory effect on sorting. Packing of miRNA into exosomes was reported to be related to the 3′ uridylation of miRNAs. By small RNA profiling of B cells and their exosomes, it was found that B cells preferentially sorted 3′ end uridylated miRNAs into exosomes while keeping 3′ end adenylated isoforms in cells (83). Similarly, N-linked glycosylation was also found to regulate the sorting of glycoprotein into exosomes (84). While these sorting-related modifications have been observed, the exact sorting machineries have not been identified, which warrants further investigation of the systemic interactome in exosomal cargo sorting.

e. Cargo sorting in MVs

The vesicle-associated membrane protein 3, a member of the v-SNARE family, was found to regulate the sorting of proteins, including type 1 matrix metalloprotease into MVs through interaction with tetraspanin CD9 (85). Sorting of mRNAs into EVs is related to a zipcode-like 25 nucleotide sequence in the 3′-untranslated region (3′UTR) of mRNAs. This zipcode has a “CTGCC” core on a stem-loop region together with a miRNA-binding site. Incorporation of this zipcode into 3′UTR of one mRNA led to the enrichment of mRNA in EVs (59). The miRNA processor, RNA-induced silencing complex (RISC), was reported to sort miRNA in EVs. Knockout of AGO2, a major functional protein of RISC complex, resulted in differential change of sorting efficiency among endogenous miRNAs (52). Interestingly, Alix, which was found to be involved in cargo sorting in exosome biogenesis, was also shown to bind Ago in EV, resulting in selective miRNA sorting (86). However, these studies on RNA sorting did not distinguish between exosomes and MVs. Therefore, further study is needed to address whether these sorting mechanisms are shared in exosomes and MVs or there is a vesicle type specific mechanism.

2.3. Isolation of EVs

Due to the overlapping sizes among exosomes, MVs, and ABs, and the lack of specific markers to distinguish each type, obtaining EVs with high purity and quality remains a challenge. Currently, there are multiple established approaches for purifying EVs from biofluids or cell culture media (CCM) (Table 1). Depending on the fluid type and the purpose of subsequent analysis, the appropriate selection of isolation method is a key step for accurate assessment of EVs.

Table 1.

Methods to Isolate EVs

Exosome isolation method	Mechanism	Advantage	Disadvantage	Most suitable Application	Examples
Differential centrifugation	Distinct molecules in biofluid are separated by differential centrifugal force.	As the most common approach used to isolate EVs, it is cost effective and highly flexible in the scale of production. It is able to separate ABs, MVs, and exosomes from biological fluid.	It requires specialized instruments such as ultracentrifuges and rotors. The isolation efficiency is subject to change based on different instruments. The multi-step process is also relatively time consuming. Biological fluid with high viscosity needs dilution before ultracentrifugation.	Protein, mRNA, miRNA	NA
Density gradient centrifugation	Combining ultracentrifuge and density gradient could exclude molecules with different densities from EVs.	Density gradient further purifies EVs from co-precipitation such as protein aggregates, apoptotic bodies, or nucleosomal fragments.	Other than shared disadvantages with ultracentrifugation, density gradient centrifugation needs extra washing steps. It doesn’t separate certain lipoproteins and virus vesicles since they share the similar densities.	Functional study; Protein; mRNA; miRNA	OptiPrep
Ultrafiltration	Using membranes with different pore sizes to seperate EVs.	This method is able to enrich EVs from small biofluid volume. The procedure is relatively easy and quick.	Causing retention of exosomes because of membrane -protein binding or saturation of membrane pores. Using force filtration potentially leads to artificial vesicles or vesicle rupture.	NA	PureExo; DiagExo
Polymer-based precipitation	Using volume-excluding polymers to wrap and aggregate Evs, which could be precipitated at low speed centrifugation.	Specialized instruments are not needed. The process of isolation is easy to handle. It is high-throughput and efficient for EV isolation.	It is compromised in purity with contaminants such as lipoproteins. The presence of polymers in precipitation might affect following analyses.	miRNA	miRCURY; ExoQuick and Total Exosome Isolation Reagent
Size exclusion chromatography	SEC is based on a column of beads with different pore sizes. EVs are separated and collected as molecules with different hydrodynamic radii pass through the column at different rates.	Usually only single step is needed to isolate EV by SEC from biofluid. The integrity and biological activity of the molecules isolated are minimally altered during the process of SEC.	SEC doesn’t concentrate EVs. Optimization and re-equilibration are needed before running, which limit the efficiency of SEC.	Functional study; Protein; mRNA; miRNA	Izon qEV SEC
Immunoaffinity capture	EVs are isolated based on antibodies recognizing antigens expressed on the surface of vesicles.	It can be used to isolate total EVs or a specific group of EVs expressing unique antigen.	Immunoaffinity capture is not suited for large sample volume. Heterogeneity in antigen expression and dynamic in epitope masking affect the isolation efficiency. The specific antibody needs to be available.	Subgroup analysis	MACSPlex Exosome Kit; Exo-Flow
Microfluidic device	Using distinct or combined technologies such as immune affinity, SEC and filtration to study the EVs in biofluid at microscale.	Small sample volume needed, high effeciency, lower cost in sample analysis and high sensitivity.	The cost of device is high. It still needs development and optimization to be a clinical tool.	Clinical diagnosis	iMER; nPLEX; NanoDLD

a. Differential centrifugation

The most common method for the isolation of EVs is differential centrifugation. By applying differential centrifugal force to the liquid sample, particles are separated based on their different densities or sizes. Differential centrifugation has been widely utilized for isolating exosomes, MVs, and ABs from suspensions. Usually, liquid suspensions are first centrifuged at 300 ×g for 10 min to remove cells. Then ABs are isolated as pellets after centrifugation at 2000 ×g for 10–20 min (87, 88). The supernatant is collected and then used for MV isolation by centrifugation at 16,500 ×g for 20 min (87, 89). To isolate the exosomes, the supernatant is further centrifuged at 120,000 ×g for longer than 70 min (87, 89). The pellets from the last step are considered exosomes with some contamination of aggregated proteins and lipoproteins. One additional washing step with PBS is usually applied to further purify the exosomes. In many cases, a 0.22-μm filter is used before proceeding to the exosome isolation step to exclude particles larger than 200 nm in diameter (89). As the most common method for isolating EVs, differential centrifugation has several advantages compared to other methods. It is a very flexible approach that can be used in most circumstances. By changing the rotors and tubes used in the ultracentrifuge, samples with different volumes can be processed. It is in general not limited to the types of the liquid samples. Liquids with low viscosity can be directly centrifuged while those with high viscosity can be diluted before being processed. This protocol of EV isolation has been widely accepted and optimized by many researches. However, the disadvantages of differential centrifugation are also obvious. It requires access to an ultracentrifuge and rotors. The process is relatively time-consuming and labor-intensive. An extra dilution step is needed for samples with high viscosity. Additionally, ultracentrifugation may cause the damage and aggregation of EVs (88, 90), which may affect the subsequent analysis, such as particle size and number estimate, in EV treatment in vitro and in vivo. More importantly, isolated exosomes have been reported to be contaminated with aggregated proteins (91).

b. Density gradient centrifugation

To further improve the purity of the exosomes isolated by differential centrifugation, an extra step using sucrose or iodixanol cushion/gradient can be applied (91). Density gradient can greatly improve the purity of isolated exosomes (92). However, it is not able to separate high-density lipoproteins (HDLs) from the exosome sediments (93). Due to the fact that HDLs also carry proteins and miRNAs (94, 95), it is important to exclude the noise from HDLs while studying exosomes.

c. Ultrafiltration

Another common approach to isolate EVs is ultrafiltration, which separates EVs based on their size. Usually, it combines membrane concentrator and size exclusion filters to separate particles with different diameters. The most commonly used filters have pore sizes of 0.8 μm, 0.22 μm, and 0.1 μm. Lobb et al. reported the successful isolation of exosomes from CCM and plasma by using a 100,000 kDa membrane after filtration through 0.22 μm filters (90). Use of low protein binding hydrophilized polyvinylidene difluoride membrane can further increase the yield while decreasing the co-purified proteins (96). While ultrafiltration is less time-consuming compared to ultracentrifugation, it also has several limitations that should be considered. Ultrafiltration potentially decreases the yield of EVs as they can saturate and block the pores of the membrane. The choice of membranes also greatly impacts the quality of the isolated EVs (90). During the filtration steps, high pressure may lead to the leakage of unwanted particles or even rupture of EVs (97).

d. Polymer-based precipitation

Volume-excluding polymers, such as polyethylene glycol (PEG), have been commonly used for virus precipitation. Because of the similarities in biophysical properties between exosomes and viruses, these polymers can also precipitate exosomes (98, 99). Several commercial products, such as ExoQuick (System Biosciences) and Total Exosome Isolation (Thermo Fisher Scientific) also use polymer-based precipitation to isolate exosomes. These products simplify the exosome isolation process with decreases in centrifugation steps and speed. The yields obtained are also higher compared to those by other approaches. However, polymer-based precipitation tends to precipitate larger particles and lipoproteins. Aggregated polymers on the exosomes interfere with subsequent analyses. Furthermore, isolation of EVs from plasma by this approach requires additional thromboplastin treatment.

e. Size-exclusion chromatography (SEC)

Recently, SEC has been reported to be used for EV isolation with high purity and integrity (100–102). SEC separates molecules in solution by their sizes. Using adsorbent materials packed in a column, small molecules are trapped and their passage through the column is impaired. They are eluted in later fractions while large molecules are present in earlier fractions. Several different types of resins have been used in SEC, such as Sepharose CL-2B (100), Sepharose CL-4B (103), and Sephacryl S-400 (101). Compared to other methods, SEC has a significant advantage in its applicability for different types of fluids, particularly for serum and plasma. It also leads to less lipid and protein contamination in isolated EVs (102, 104). More importantly, it minimally alters the characteristics of the isolates, retaining biologically active and morphologically intact EVs (102, 105). However, SEC also has several disadvantages. As EVs are isolated by serial fractions, pre-optimization is needed to determine the fraction of interest. SEC may also dilute the EVs due to the elution process. Additional concentration, such as centrifugal filtration with 100 kDa membrane, may be needed for subsequent analyses (102).

f. Immunoaffinity

EVs can also be isolated according to their surface markers by immunoaffinity. For example, antibodies targeting CD9, CD63, and CD81 are conjugated onto magnetite beads, which selectively capture EVs expressing these membrane proteins. Commercial products such as MACSPlex Exosome Kit (Miltenyi Biotec) and Exo-Flow (System Biosciences) use this technique for exosome isolation. This approach is especially useful for purifying a subpopulation of EVs derived from a specific type of cells, such as cancer cells (106). However, the heterogeneity in antigen expression and dynamic in epitope masking may affect the isolation efficiency. It is also costly when dealing with large volume of samples.

g. Selection of appropriate isolation method

While various approaches of EVs isolation are available, choosing an appropriate method can greatly impact the consistency of results and the subsequent analyses. Studies have been carried out to compare the approaches in different parameters. For overall yield, volume-excluding polymers, such as miRCURY (Exiqon), ExoQuick (System Biosciences), and Total Exosome Isolation Reagent (Thermo Fisher Scientific), were found to generate higher yield than ultracentrifugation, ultrafiltration (PureExo) and density gradient centrifugation in serum, cell culture media, and milk (107–110). However, despite the improvement in yield, polymer-based precipitation results in compromised purity (107, 108). This is to be considered carefully, especially when the downstream analysis is related to the protein content of the isolated exosomes. It was reported that volume-excluding polymers co-isolated various types of protein factors. In serum, high levels of albumin contamination were found in precipitations (90), while for urine, abundant Tamm–Horsfall glycoprotein (THP), a common protein excreted in ordinary urine, was co-isolated by volume-excluding polymers (111). In studies of the protein content within exosomes, ultracentrifugation and SEC (Izon qEV columns) showed superior results as indicated by particles per μg of protein and expression of exosomal protein markers (90, 102, 107, 111). Density gradient centrifugation (OptiPrep) showed high purity of exosomes and high quality of exosomal proteins, but the yield was low, possibly due to the loss of particles during the centrifugation (90, 107). If a specific subgroup of exosomes is being studied, immunoaffinity capture is also an approach with high-quality protein yield (112). When studying the bioactivity of exosomes or using exosomes as a drug carrier, the isolation method also needs to be carefully optimized. Attention should be paid to the use of volume-excluding polymers since they co-isolate polymers. It was also reported that large aggregates were formed using volume-excluding polymers (113), and toxicity in cell treatment was observed (102), suggesting that they are not suitable for bioactivity assays in vitro and in vivo. Ultracentrifugation and ultracentrifugation combined with a cushion also generated some degree of aggregation, while density gradient centrifugation generated exosomes with the highest dispersion, indicating that this approach is superior for maintaining the bioactivity of the exosomes. SEC was also found to generate exosomes with preserved biophysical and functional properties. These exosomes showed less toxicity in cell treatment and better bio-distribution in mice (102, 114), suggesting that SEV isolated exosomes are suitable for in vitro and in vivo functional assays. mRNA is another important cargo in exosomes that has drawn attention from researchers and clinicians because of its role in exosome-regulated cancer progression. The impact of isolation methods on mRNA carried by exosomes has been studied. Ultracentrifugation could generate high quality of mRNA in exosomes with a reasonable yield (111). Ultracentrifugation combined with a cushion or density gradient centrifugation gave high quality of mRNA, but the yields were compromised compared to other approaches (107, 111). Density gradient centrifugation (OptiPrep) could lead to a distinct mRNA profile of exosome from CCM compared to ultracentrifugation and volume-excluding polymers, possibly due to contamination of the RNA binding proteins (107). Compared to mRNA, small RNA such as miRNA was found to be more enriched in exosomes. Numerous studies have been carried out to study the role of exosomal miRNA in cancer development. However, the method of exosome isolation needs to be optimized and standardized due to its impact on the miRNA profile in exosomes. Ultracentrifugation was found to be a suitable approach for studying most exosomal miRNA due to the high quality and relatively good yield (111, 115). Exosomes isolated by volume-excluding polymers were shown to have low purity, but surprisingly the miRNA content was not affected. The miRNA profiles were consistent between volume-excluding polymers and ultracentrifugation, while volume-excluding polymers outcompeted other methods in yield (111, 115). This suggests the possible use of volume-excluding polymers for miRNA profiling when the bioactivity of exosomes is not required for the subsequent study.

Every approach of EV isolation described above (Table 1) has its advantages and disadvantages. Selecting an approach appropriate for the samples and optimizing the method are important for accurate assessment for downstream analyses. Combination of different approaches can be used to further improve the quality of EVs. For example, when additional filtration is combined with volume-excluding polymers, the purity of isolated exosomes is further improved. A modified protocol using ExoQuick (with DL-dithiothreitol to reduce THP, increased reagent used, and increased centrifuge speed) was also reported to increase the quality of miRNA, mRNA, and protein in isolated exosomes (111). As a burgeoning subject in cancer research, the study of EVs also adds to the improvement of isolation methods. For example, System Biosciences Inc. provides updated kits such as ExoQuick- PLUS and ExoQuick-LP for removal of contamination carried by volume-excluding polymers alone, although the efficiency of these products and the quality of isolates need to be further tested. Other more advanced devices such as microfluidic platforms (see section 4) are being developed and have great potential for the easy isolation and assessment of EVs, especially for clinical use. With the optimizing and standardizing of isolation methods, new biological functions of these small particles may be further disclosed.

3. Roles of EVs in cancer progression

3.1. EVs act as carriers of macromolecules for intercellular communication between cancer cells

In recent years, the number of publications on the role of EVs in cancer development and progression has exponentially increased. Cancer cells have been found to release higher quantities of EVs compared to normal cells, and more EVs can be detected in the circulation of cancer patients (22, 116, 117). Oncogenic components have been found in the EVs released by cancer cells and they trigger various signaling pathways in recipient cells once incorporated. One major approach for EVs to affect cancer progression is to innately transfer pathogenic components between cancer cells. Conferring these components, which include proteins, mRNA, miRNA, DNA, and lipids, phenocopies behaviors from aggressive cancer cells to more indolent cancer cells (Figure 2 and Table 2).

Figure 2. Intercellular communication through EVs.

EVs-mediated intercellular communication occurs between cancer cells. Aggressive cancer cells secrete exosomes and MVs containing oncogenic factors to transform indolent cancer cells into aggressive phenotypes. EVs also mediate the intercellular communication between cancer cells and TME cells. EVs derived from cancer cells modulate TME cells thus generating a favorable microenvironment for the growth and metastasis of cancer cells. TME cells could also secrete EVs to affect cancer cells, inducing aggressive phenotypes in indolent cancer cells.

Table 2.

EVs act as carriers of macromolecules for intercellular communication between cancer cells

Type of cancer	Type of EVs	Key molecules	Role	Reference
Bladder Cancer	Exosome	EDIL-3	Exosomes from high grade bladder cancer cells promote migration of bladder cancer cells by activation of EGFR.	(121)
Prostate Cancer	Exosome	MDR-1/P-gp	Exosomes from docetaxel-resistant prostate cancer cells confer docetaxel-resistance to non-resistant cells, inducing the motility, invasion, proliferation and anchorage-independent growth.	(124)
		metalloproteinases, TGF-β2, TNF1α, IL6, TSG101, Akt, ILK1, and β - catenin	Exosomes from prostate cancer cells under hypoxia enhance the invasiveness and stemness of naive PCa cells by targeting of epithelial adherens junction pathway.	(118)
Breast cancer	Exosome	miR-10b	Metastatic breast cancer cells secrete miR-10b containing exosomes to non-metastatic cells to promote the invasion by targeting HOXD10 and KLF4.	(130)
		miR-221/222	Exosomal miR-221/222 derived from tamoxifen resistant MCF7 targets P27 and ERa in recipient cells to enhanced tamoxifen resistance.	(131)
		miR-100, miR-222 and miR-30a	Exosomes from adriamycin and docetaxel resistant cells confer chemoresistance by transferring miRNAs to sensitive cells.	(132)
		Lnc UCA11	Exosomal lnc UCA1 confers tamoxifen resistance from resistant cells to sensitive cells.	(137)
		EDIL-3	EDIL3 containing exosomes enhance cell invasion and accelerates lung metastasis in vivo through integrin-FAK signaling cascade.	(120)
		Hsp90α	Hsp90α from invasive cancer cell derived exosomes activates plasmin to increase cancer cell mobility.	(119)
		P-gp	Docetaxel-resistance could be transferred between MCF7 breast cancer cells through Pgp contained exosomes.	(125)
		Unknown	Exosomes from aggressive cancer cells transfer phenotypic traits to less aggressive cancer cells.	(278)
	MV	EMMPRIN	EMMPRIN from tumor cells derived MVs activates p38/MAPK pathway to promote cancer invasion.	(279)
		RAB22A	MVs from breast cancer cell under hypoxia promote focal adhesion formation, invasion, and metastasis in naive breast cancer cells.	(42)
Renal cell carcinoma	Exosome	LncARSR	Exosome-mediated transmission of lncARSR, which acts as a sponge for miR-34 and miR-449, can confer sunitinib resistance to sensitive cells.	(78)
Gastric cancer	Exosome	Unkown	Gastric cancer exosomes promote cancer cell proliferation by activation of PI3K/Akt and MAPK/ERK pathways.	(280)
Hepatocellular cancer	Exosome	miR-584, miR-517c, miR-378, etc.	HCC cell-derived exosomes suppress TAK1 expression in recipient cells to enhance transformed cell growth.	(133)
Glioma	Exosome and MV	mRNA, miRNA and angiogenic proteins	Glioblastoma derived EVs confer transcripts and angiogenic proteins to promote the tumor growth.	(142)
	MV	EGFRvIII	Cancer cells share EGFRvIII through MVs, leading to the transfer of oncogenic activity	(126)
Oral squamous cell carcinoma	Exosome	miR-21	Exosome derived hypoxic OSCC cells increase the invasion and migration of normoxic cells by transferring miR-21.	(43)

a. Transfer of protein

The active proteins carried by EVs can be directly transferred into the new host cell where they perform functional bioactivities. In PCa, higher activity of metalloproteinases and increased oncogenic signaling molecules were found in hypoxic cancer cells by LC/MS/MS (118). Exosomes from these PCa cells under hypoxia enhance invasiveness and stemness of naïve PCa cells by transferring molecules targeting the epithelial adherens junction pathway. In breast cancer, Hsp90α carried by invasive cells can be conferred to other breast cancer cells via exosomes. In the recipient cells, hsp90α interacts with plasminogen activator, possibly through the bridge of annexin II, and activates plasminogen, which increases cancer cell mobility (119). Other than the delivery of cytoplasmic molecules, the membrane-bound proteins on EVs can also serve as ligands for the receptors expressed on the recipient cells. In the breast cancer, membrane-bound EDIL-3 on exosomes from donor cancer cells was shown to bind and activate the integrin-FAK signaling cascade to promote cell invasion and accelerate lung metastasis (120). Furthermore, exosomes from high-grade bladder cancer cells were found to contain increased levels of protein EDIL-3. This protein could also activate the EGFR signaling in the recipient cells, promoting the migration of cancer cells (121). However, in this case, it is unknown whether EDIL-3 on the exosome membrane directly activates EGFR, or if it needs to be incorporated and recycled as a free ligand. Other than promoting invasion and migration, transfer of protein contents between cancer cells through EVs also confers drug resistance. P-glycoprotein (P-gp), a transporter protein associated with the efflux of cancer therapy drugs (122), is a typical example that was found to be involved in EV-mediated phenocopy of drug resistance in cancers. The docetaxel-resistant PCa cells expressed increased levels of MDR-1/P-gp (123). MDR-1/P-gp could be carried and transported by exosomes to non-resistant cells where they induce the drug resistance, motility, invasion, proliferation, and anchorage-independent growth (124). In breast cancer, it was found that docetaxel-resistance could be transferred between MCF7 breast cancer cells through P-gp-containing exosomes (125). Besides exosomes, MVs also deliver pathogenic signaling molecules between cancer cells. EGFRvIII was found to be shared between glioma cells through MVs (126). Acquisition of EGFRVIII in recipient cells resulted in the transfer of oncogenic activity such as activation of MAPK and Akt, increase of VEGF and Bcl-x(L) levels, decrease of p27 levels, and induction of spindle-like morphology and anchorage-independent growth. Breast cancer cells under hypoxia were found to have increased expression of RAB22A, which promoted the biogenesis of MVs. These MVs could be incorporated by naïve cancer cells and promote focal adhesion formation, invasion, and metastasis of the recipient cells (42).

b. Transfer of RNA

Intercellular exchange of RNAs is another approach of trait transfer between cancer cells. It was found that small RNAs ranging from 18–28 nt were particularly enriched in exosomes (127). The most abundant small RNAs found in exosomes were miRNA and rRNA (127, 128). Transfer of miRNAs by exosomes can result in translational repression by 3′UTR binding in recipient cancer cells (129). In breast cancer, enriched miR-10b was secreted in exosomes from metastatic cells. Non-metastatic cells could uptake these exosomes and increase the invasive ability through miR-10b-induced HOXD10 suppression (130). Drug resistance can also be modulated by exosomal miRNAs. MiR-221/222 derived from tamoxifen-resistant MCF7 breast cancer cells can be transferred to sensitive cells through exosomes. In the recipient cells, the miRNAs enhanced tamoxifen resistance by targeting the expression of P27 and ERα (131). MiR-222, together with miR-100 and miR-30, could also confer chemoresistance to adriamycin and docetaxel between breast cancer cells by targeting tumor suppressor genes such as PTEN (132). The transfer of miRNAs between cancer cells has also been discovered in other types of cancers, including hepatocellular cancer (133), oral squamous cell carcinoma (43), esophageal cancer (134), and ovarian cancer (135). In general, these transferred miRNAs act as oncogenic factors that target tumor suppressor genes and induce aggressive phenotypes in recipient cells.

Other than small RNAs, lncRNA is another category of non-coding RNAs (ncRNAs) found to be carried by EVs. LncRNA in exosomes derived from cancer cells often harbors enriched miRNA seed regions (136). In renal cell carcinoma, exosome-mediated transmission of lncARSR, which acts as a sponge for miR-34 and miR-449, can confer sunitinib resistance to sensitive cells (78). In breast cancer, tamoxifen-resistant cancer cells had increased expression of the lncRNA UCA1, which was sorted into exosomes. Uptake of exosomal UCA1 induced tamoxifen resistance in sensitive cancer cells (137). Even though the detailed mechanism was not revealed in this study, UCA1 was previously found to induce aggressive phenotype and drug resistance by suppression of p27 (138) or acting as sponge for miR-143 (139) and miR-204-5p (140).

Transfer of ncRNAs may modulate the gene expression in an indirect way, while delivery of mRNA could directly result in the increased translation of this transcript. It was reported that mRNA transferred by exosomes is active and can be translated into functional proteins in the recipient cells (11). Jurkat T Cells were found to contain a specific splice variant (GlnRS ^ΔiABD) of GlnRS mRNA in exosomes and this transcript variant can be translated into a protein product (141). A study from Breakefield’s lab also indicates that exosomes and MVs from glioblastoma contain RNAs and could promote the growth of human U87 glioma cells (142). In glioma, mRNA, miRNA, and proteins were found to be selectively packaged into exosomes and MVs, which showed a promoting effect on tumor growth (142).

c. Transfer of macromolecules by ABs

Currently, there is no clear evidence indicating that ABs mediate the conferring of pathogenic components between cancer cells. However, they are also known to carry functional molecules such as mRNAs, miRNAs, DNA, and proteins that can trigger signaling pathways in recipient cells (143–146). Further studies are needed to decipher the role of ABs in intercellular communication between cancer cells.

3.2. EVs facilitate intercellular communication between cancer cells and microenvironmental cells to promote tumor progression

TME has a significant role in the development and metastasis of cancer cells (27, 147, 148). In general, TME is composed of extracellular matrix, fibroblasts, adipocyte, blood vessels, lymphatic vascular networks, and other associated tissue cells (147, 149). The interaction between cancer and TME cells is an essential process for the survival of cancer cells. The crosstalk between cancer cells and TME can be established through direct cell-cell contact or secreted factors (150). EVs, by carrying ample biologically active molecules, were also found to mediate the distant communication between cancer and TME cells (Figure 2 and Table 3) (151). Transfer of cellular components from cancer cells to TME cells helps the establishment of a niche for cancer growth. On the other hand, TME-derived EVs are also utilized by cancer cells to reinforce the growth, survival, and mobility of cancer cells.

Table 3.

EVs facilitate intercellular communication between cancer cells and microenvironmental cells to promote tumor progression

Type of cancer	Type of EVs	Interacting Cells	Role	Reference
Breast cancer	Exosome	Astrocyte	Astrocyte-derived exosomes transfer miR-19a to inhibit PTEN in cancer cell to support cancer growth in brain.	(182)
		Fibroblast	Breast cancer cells trigger NOTCH-MYC signaling in CAF to increase RNAs of RN7SL1. Exosomal RN7SL1 acts as damage-associated molecular patterns to modulate immune cells and cancer cells.	(153)
		Fibroblast	Fibroblasts secrete CD81 containing exosomes to promote autocrine Wnt-planar cell polarity signaling that induces protrusive activity and motility of breast cancer cells.	(154)
		Stroma	Noncoding transcripts and transposable elements in stromal exosomes activate STAT1 through pattern recognition receptor RIG-I and expand therapy-resistant tumor-initiating cells.	(281)
		Mesenchymal Stem Cells	Mesenchymal stem cell-derived exosomes deliver miR-16 to inhibit VEGF expression in cancer cells, leading to decreased angiogenesis.	(165)
		Bone marrow mesenchymal stem cell	Mesenchymal stem cell-derived exosomes deliver miR-222/223 to promote quiescence and drug resistance. Targeting miR-222/223 sensitized breast cancer cells to carboplatin-based therapy.	(170)
			Exosomal miR-23b induces cancer cell dormancy by targeting MARCKS	(171)
		Adipose tissue derived mesenchymal stem cells	Cancer cell-derived exosomes induce adipose tissue derived MSCs into tumor-associated myofibroblasts, which release tumor-promoting factors. This process is mediated by phosphorylation of SMAD2.	(173)
		Pre-adipocyte	Preadipocyte-derived exosomes promote tumorigenesis and cancer stem cell properties. Shikonin inhibits cancer growth by targeting SOX9 signaling through miR-140.	(282)
		Macrophages	IL-4 activated macrophages transfer miR-223 to promote the invasion of breast cancer cells via targeting the Mef2c-b-catenin pathway.	(189)
	MV	Normal epithelial cells/Fibroblast	MVs derived from MDAMB231 cells induce transformation of normal epithelial and fibroblast cells by transferring tissue transglutaminase and fibronectin.	(155)
		Platelet	MVs from platelets transfer integrin CD41 to the surface of breast cancer cells and enhance their adhesion to endothelial cells; MVs also promote invasion by up-regulation of CXCR4, p42/44, AKT, and MMPs.	(192)
Prostate cancer	Exosome	Adipose-derived mesenchymal stem cells	Prostate cancer cell-derived exosomes induce tumorigenic reprogramming of adipose stem cells by trafficking of oncogenic factors and downregulating tumor suppressors.	(175)
		Bone-marrow mesenchymal stem cell	Exosome TGFβ trigger mesenchymal stem cell differentiation into myofibroblasts, which support tumor cell growth.	(172)
		Fibroblasts	Exosomal TGFβ induces fibroblast differentiation to myofibroblasts, which are pro-angiogenic and tumor-promoting.	(156, 157)
		Cancer-associated fibroblast	CAF-derived exosomes promote tumor growth under nutrient stressed conditions by delivering intact metabolites.	(283)
		Lymph node stroma cells and lung fibroblasts	Exosomal miR-494 and miR-542-3p target cadherin-17 to increase matrixmetalloproteinase expression in niche cells.	(284)
	Large oncosome	Fibroblasts	Large oncosomes derived from prostate cancer cells contain active AKT1. Internalization of large oncosome by fibroblasts activates MYC and reprograms fibroblast to support angiogenesis and tumor growth.	(159)
pancreatic cancer	Exosome	Adipocyte	Exosomal adrenomedullin from cancer cell induce lipolysis through p38 and ERK1/2, which causes weight loss of cancer patients.	(191)
		Endothelial Cell	Uptake of Tspan8-CD49d containing exosome induces several angiogenesis-related genes to enhance endothelial cell proliferation.	(176)
		Kupffer cells	Exosome from pancreatic ductal adenocarcinomas induces the release of transforming growth factor ß by KCs and fibronectin production by hepatic stellate cells, which are critical to recruit macrophage for the formation of pre-metastatic niche.	(183)
		Cancer-associated fibroblast	CAFs exposed to gemcitabine significantly increase the release of exosomes, which contain chemoresistance-inducing factor Snail and miR-146a.	(158)
	Exosome and MV	muscle cell	tumor-derived EVs induce apoptosis of skeletal muscle cells through delivery of miR-21	(184)
ovarian cancer	Exosome	adipose tissue derived mesenchymal stem cells	Cancer cell-derived exosomes induce myofibroblasts like phenotype in adipose tissue derived MSCs, through SMAD2 or AKT phosphorylation.	(174)
		Lymph Node Cells	Exosomes go to lymph node and induce premetastatic niche by inducing expression of factors responsible for cell recruitment, matrix remodeling, and angiogenesis.	(285)
Melanoma	Exosome	Bone marrow cells	Melanoma cell derived exosomes educate bone marrow progenitors by transferring MET. Targeting Rab27A could decrease cancer growth and metastasis.	(190)
	MV	Fibroblasts	MVs from melanoma cells induce VCAM-1 in tumor-associated fibroblast, which is involved in melanoma cell attachment to fibroblast.	(161)
Renal cell carcinoma	Exosome	Endothelial cells	Exosomes from CD105-positive cancer stem cells contain angiogenic mRNAs and miRNAs to promote lung endothelial cells, establishing the pre-metastatic niche.	(177)
Lung cancer	MV	Endothelial cells; Stroma fibroblasts	Lung cancer cells secrete more MVs under non-apoptotic doses of hypoxia and irradiation. These MVs activate and chemoattract fibroblasts and endothelial cells, increasing metastasis of lung cancer cells	(160)
		Platelet	Platelet derived MVs transfer integrin CD41 to promote growth, invasion and angiogenesis of lung cancer cells. These MVs also chemoattract cancer cells.	(188)
Colorectal cancer	Exosome	Endothelial cells	CRC cell-derived exosomes transport cell cycle-related mRNAs to endothelial cells to promote angiogenesis.	(286)
Bladder Cancer	Exosome	Endothelial cells	Exosomes from high grade bladder cancer contain EDIL-3, which promotes angiogenesis, and migration of endothelial cells and bladder cancer cells.	(121)
Glioma	Exosome and MV	Endothelial cells	GBM Cancer cells under hypoxia release TF/VIIa containing exosomes to activate hypoxic endothelial cells. Glioblastoma exosomes and MVs can deliver functional RNA to HBMVECs to promote angiogenesis.	(142, 178)
	MV	normal fibroblasts	MVs derived from U87 glioma cells induce transformation of normal fibroblasts by transferring tissue transglutaminase and fibronectin.	(155)
Transformed fibroblast	AB	Fibroblast	Apoptotic bodies from transformed rat embryonic fibroblasts deliver oncoaenes to mouse embryonic fibroblasts.	(32)

a. Cancer-associated fibroblasts (CAFs)

Fibroblasts, the most abundant cells in connective tissues, form the tissue framework by secreting extracellular matrix components (152). Recently, Nabet et al. reported that the activated NOTCH-MYC signaling by breast tumor cells in fibroblasts induced CAFs, which have increased amount of unshielded RNAs of RN7SL1 (153). These RNAs are secreted by CAFs through exosomes, acting as damage-associated molecular patterns that bind to pattern recognition receptors in recipient cells. This signaling cascade then drives an inflammatory response in immune cells and promotes growth, metastasis, and therapy resistance in cancer cells. CAFs also secrete CD81-containing exosomes to promote autocrine Wnt-planar cell polarity signaling that induces protrusive activity and motility of breast cancer cells (154). Breast cancer cells were also found to be capable of transforming the fibroblasts through MVs. These MVs contained tissue transglutaminase and fibronectin, which could activate the AKT/ERK signaling pathway in fibroblasts, thus leading to anchorage-independent growth and enhanced survival capability (155). In PCa, exosomes derived from cancer cells promoted differentiation of fibroblasts to myofibroblasts (156, 157). This effect was mediated by membrane-bound TGFβ on the exosomes through SMAD-dependent signaling. Interestingly, soluble TGFβ was not able to trigger differentiation of fibroblasts, suggesting differential roles of membrane TGFβ from exosomes and soluble TGFβ. On the other hand, CAFs could also support PCa growth under nutrient-stressed conditions by directly transferring intact metabolites such as amino acids (158). Exosome derived from CAFs could also confer drug resistance. It was found that CAFs are resistant to gemcitabine in pancreatic cancer, and gemcitabine treatment increased exosome release from CAFs. These exosomes could endow drug resistance to pancreatic cancer cells by transferring Snail and miR-146a (158). Other than exosomes and MVs, the interaction between cancer cells and fibroblasts has also been reported to be mediated by ABs (32). ABs from transformed rat embryonic fibroblasts delivered oncogenes to mouse embryonic fibroblasts, resulting in aneuploidy and oncogenic change. Besides exosomes, MVs and ABs, large oncosomes also carry oncogenic messages mediating the communication between cancer cells and fibroblasts (159). In PCa, large oncosomes derived from cancer cells were shown to contain active AKT1. Uptake of large oncosome by fibroblasts leads to activation of MYC, reprograming the fibroblast to support angiogenesis and tumor growth. The interactions between cancer cells and fibroblast cells have also been reported in other cancers, such as melanoma and glioma (155, 160, 161) (Table 3).

b. Mesenchymal stem cells (MSCs)

MSCs are another major component of stromal cells. They are known to differentiate into mesenchymal tissues such as bone, cartilage, and fat (162). In some aspects, MSCs share many similarities with fibroblasts, such as cell surface markers and pluripotency (163). Similar to fibroblasts, MSCs greatly impact cancer progression by interacting with cancer cells (164), and EVs play important roles in mediating this interaction. For example, in breast cancer, exosomes derived from MSCs contain miR-16, which targets the expression of VEGF in breast cancer cells upon uptake (165). Low VEGF production from breast cancer cells resulted in decreased angiogenesis. MSCs were originally identified in bone marrow (166), where they were known to be quiescent but metabolically active (167). These MSCs, together with hematopoietic stem cells, form a pre-metastatic niche for cancer cells (167–169). One function of this niche is the transfer of the quiescent trait to cancer cells by EVs. As in breast cancer, MSC-derived exosomes delivered miR-222/223 to promote cancer cell quiescence and drug resistance (170). Targeting miR-222/223 sensitized breast cancer cells to carboplatin-based therapy. MiR-23b, another exosomal miRNA secreted by bone marrow MSCs, was also found to induce dormancy in metastatic breast cancer cells by targeting MARCKS (171). The communication between bone marrow-derived MSCs and cancer cells is bidirectional. Not only do cancer cells take in exosomes, they also secrete exosomes to educate MSCs in bone. In PCa, exosomes from cancer cells were shown to trigger MSC differentiation into pro-angiogenic and pro-invasive myofibroblasts by the carried TGFβ. However, soluble TGFβ could not induce the same change in MSCs (172). Aside from bone marrow-derived MSCs, adipose-derived MSCs were also found to interact with cancer cells through EVs. Exosomes secreted by cancer cells facilitate tumor growth by inducing MSC differentiation into tumor-associated myofibroblasts through the activation of the SMAD dependent pathway (173, 174), which is similar to that found in bone marrow-derived MSCs (172). Furthermore, PCa cell-associated exosomes can promote cancer progression by directly inducing neoplastic reprogramming of adipose-derived MSCs and the transformed adipose-derived MSCs have PCa-like mesenchymal-to-epithelial transition (MET) morphology and cellular markers, such as CK8, CK5/18, and the PCa-specific marker, AMACR (175). This change is possibly caused by the trafficking of oncogenic miRNAs, transcripts, and proteins into MSCs.

c. Endothelial cells

Endothelial cells are another type of TME cells that have been widely reported to interact with cancer cells. In pancreatic cancer, Tspan8 contributed to the selective sorting of angiogenic cargoes, such as CD106 and CD49d into exosomes. Uptake of these exosomes by endothelial cells resulted in increased angiogenesis (176). In renal cell carcinoma, exosomes from CD105-positive cancer stem cells had the ability to induce activation of endothelial cells and to promote angiogenesis and lung metastasis in vivo (177). In glioma, exosomes and MVs promoted angiogenesis by directly transferring RNA and proteins into endothelial cells (142). Moreover, EVs derived from glioma cells were reported to carry angiogenic factors such as EGFRvIII and TF/VIIa, which can be incorporated by endothelial cells (142, 178).

d. Tissue-specific microenvironmental cells

Other than microenvironmental cells found to communicate with various types of cancer cells, cancer cells also interact with tissue-specific cell types to support growth and metastasis of cancer cells. When breast cancer cells reach the brain, they interact with brain microenvironment cells, such as astrocytes. Both direct contact and soluble factors were reported as the patterns of interaction (179–181), and exosomes were also found to be involved in this communication. Astrocytes were shown to secrete miR-19a-containing exosomes to breast cancer cells in the brain (182). This miRNA directly binds the 3′UTR of PTEN mRNA and decreases the expression of this tumor suppressor. The loss of PTEN resulted in an increase of CCL2 secretion from cancer cells, which recruited myeloid cells to enhance cancer cell growth. In pancreatic cancer, exosomes from cancer cells were selectively incorporated by Kupffer cells in the liver. Macrophage migration inhibitory factor carried by exosomes induced release of TGFβ from Kupffer cells, which promoted fibronectin deposition and subsequently recruited bone marrow-derived macrophages and neutrophils in the liver, forming the pre-metastatic niche (183). In another study, exosomes and MVs from pancreatic cancer were also found to be engulfed by muscle cells (184). The miRNA in the EV, miR-21, activated the Toll-like 7 receptor to promote apoptosis of myoblasts, which led to cancer cachexia.

Reprograming TME cells by tumor cells through cell-cell communication has been frequently observed in multiple malignancies. Interestingly, this communication seems to be established by cancer cells though modulation of membrane compositions on EVs. Breast and pancreatic cancer cell lines that metastasize to different organs produced exosomes harboring organotropic integrins so that these exosomes could induce pre-metastatic niches at the preferred sites (185). Lipid composition also contributes to selective uptake of exosomes in recipient cells (186). Exosomes from mantle cell lymphoma patients were selectively untaken by B-lymphocytes mediated by a cholesterol-dependent pathway, while no apparent internalization in T lymphocytes or NK cells was observed (45). Besides delivery of bioactive molecules to trigger signaling pathways in recipient cells, EVs also bridge cancer cells and microenvironment cells as chemoattractants. Exosomes from chronic myelogenous leukemia cells were found to be chemoattractant for endothelial cells (187).Under non-apoptotic doses of hypoxia and irradiation, MVs derived from lung cancer cells were shown to attract endothelial cells (160). Furthermore, MVs released by platelets were found to attracted lung cancer cells (188). EVs also mediated the interaction between cancer cells and diverse environmental cells such as platelets, adipocytes, bone marrow cells, and immune cells (189–192). A summary of these interactions in different types of cancers mediated by exosomes, MVs, and ABs is presented in Table 3.

4. Translational applications of EVs in cancer

4.1. EV as a biomarker for cancer diagnosis

a. Level of EVs as a diagnostic marker

Emerging evidence suggests various translational utility of EVs. Cancer cells secrete more EVs than normal cells (193, 194), and consequently, patients with cancer have higher levels of EVs compared to healthy individuals (195–200). With the increased level of EVs in circulation and the packaging of cancer-related molecules, these vesicles may serve as promising biomarkers for cancer diagnoses. The number of EVs in the blood has been proven to be an indicator for multiple types of cancers. While cancer cells secrete more EVs compared to normal cells, higher grade cancer cells secrete more EVs than lower grade cancer cells (143–145). In colorectal cancer, the level of exosomes in cancer patients is statistically higher than that in healthy controls and it is positively correlated with the level of CEA. Colorectal cancer patients with poorly differentiated tumors and shorter overall survival have increased numbers of exosomes in the plasma (199). The level of MVs is also found to be an indicator for PCa (198). Prostasomes, MVs secreted by normal and malignant prostate acinar cells, were found to be elevated in the blood of PCa patients. Their numbers were 2.5- to 7-fold higher compared with that in the benign group, and were higher in patients with high-grade (Gleason scores more than 7) than in those with low grade (Gleason scores less than 6) PCa. In PCa, the number of ABs from cancer tissue also correlated with the disease progression (201), suggesting the poteintal of AB as biomaker. Other than these typical EVs, the newly identified EV, large oncosome, has also been shown to be a specific marker for cancer diagnosis (38). In mouse model of prostate cancer, large oncosomes were detected in both tumor tissues and circulation. More importantly, the abundance of large oncosome was found to be correlated with tumor progression in both human and mouse. These findings suggest that the level of EVs can serve as circulating biomarkers for cancer diagnosis and prognosis.

b. EV-derived ncRNA as a diagnostic marker

Tumor cells also produce distinct EVs compared to normal cells, and cancer cell derived-EVs inherit content from cancer cells. Because of the enwrapped cancer-specific molecules, these EVs may be used as an indicator for differential diagnosis of cancers. miRNA is the most studied biomarker in EVs due to the high enrichment of small RNA in these particles. It was reported that EVs contribute to the majority of miRNAs found in serum and saliva (202, 203). In PCa, 12 exosomal miRNAs were found to be differentially expressed between PCa patients and healthy donors. While the level of only one miRNA, miR-181a-2*, was significantly decreased in PCa patients, levels of other miRNAs were increased in the blood (204). Among the increased miRNAs, miR-107 and miR-574-3p were also found to be at higher concentration in urine exosomes of PCa patients compared to that in healthy individuals, indicating that these exosomes may serve as non-invasive diagnostic makers for PCa. The authors also found that levels of miR-141 and miR-375 in exosomes were increased in PCa patients with recurrences compared to those in non-recurrent patients. In colorectal cancer, seven miRNAs, including let-7a, miR-1229, miR-1246, miR-150, miR-21, miR-223, and miR-23a, were found to be upregulated in tumor-bearing patients compared to those in healthy individuals (205). The levels of these seven miRNAs were decreased after surgical resection of tumors in the patients. Compared to known tumor markers such as CEA and CA19-9, these seven miRNAs were more sensitive for identifying tumor presence, suggesting their potential use for early detection of colorectal cancer. In addition to the presence of specific miRNAs, the total miRNA amount in lung cancer was also reported to be increased in cancer cell-derived exosomes compared to that in control, indicating the diagnostic utility of miRNA levels in exosomes (206). LncRNA is another category of ncRNA wrapped in EVs. The level of certain lncRNAs was also found to be changed in the blood of cancer patients. Level of long intergenic non-protein coding RNA 152 (LINC00152) was found to be elevated in plasma of gastric cancer patients compared to those in healthy control groups (207). It had a sensitivity of 48.1% and specificity of 85.2% in the diagnosis of gastric cancer.

c. EV-derived mRNA as a diagnostic marker

Besides ncRNAs, coding RNA transcripts are also present in EVs. In PCa, tumor-specific mRNAs were detected in exosomes from the urine of patients. The mRNAs of PSA and TMPRSS2:ERG were found in the exosomes from the urine of newly diagnosed but untreated PCa patients. However, they were not present in exosomes from the urine of patients receiving androgen deprivation therapy or patients with an impaired/nonfunctional prostate (208). Surprisingly, no prostasomes (MVs) were present in urine from PCa patients, while no exosomes were found in the urine of healthy donors, possibly due to the increased biogenesis of exosomes in the maligned tissue. In another study, mRNAs of ERG (including TMPRSS2: ERG), PCA3, and SPDEF in urine exosomes plus standard of care (SOC), which includes prostate-specific antigen (PSA) level, age, race, and family history, were found to more accurately discriminate PCa with Gleason scores higher than 7 from lower grade PCa or benign lesions (209). Combination of exosomal RNA analysis with SOC increased the value of receiver operating characteristic curve (AUC) to 0.77, compared to 0.66 for SOC alone and 0.61 for PSA alone. This strategy has been commercialized by Exosome Diagnostics and is available for clinical diagnosis as a non-invasive screening system of PCa, named ExoDx™ Prostate(IntelliScore).

d. EV-derived DNA as a diagnostic marker

DNA mutation is a hallmark of cancer, which makes exosome-derived DNA an excellent tool for identifying cancer. Exosomes were found to harbor double-stranded DNA (dsDNA) (15). Melanoma cell-derived exosomes were found to have the BRAF(V600E) mutation in dsDNA, consistent with the parental cell lines. In non-small cell lung cancer (NSCLC), the dsDNA in exosomes derived from several cell lines were found to have EGFR mutations. In NOD/SCID mice, when cancer cells were implanted, the same DNA mutations in the cancer cells could be identified in circulating exosomes. Similarly, exosomes from pancreatic cancer cell lines and patient serum carry genomic DNA spanning all chromosomes with mutated KRAS and P53 (210). The inherited DNA mutation in exosomes suggests they may serve as precise biomarkers for early identification of cancer. The genetic information provided by dsDNA in exosomes may also provide a guide for targeted therapy in cancer.

e. EV-derived protein as a diagnostic marker

Proteins carried by EVs can also serve as biomarkers in cancer. In ovarian cancer, claudin-4 was expressed in both cancer cells and cancer cell-shed exosomes (211). The exosomal claudin-4 had a sensitivity of 51% in ovarian cancer detection compared to 71% of CA125. This suggests exosomal claudin-4 could potentially complement CA125 as a diagnostic for ovarian cancer. In pancreatic cancer, glypican-1 (GPC1) was found to be specifically enriched in exosomes (212). GPC1 in serum exosomes could distinguish cancer patients from healthy donors. Furthermore, the level of GPC1+ exosomes also correlated with tumor burden and survival of pancreatic cancer patients. Tumors carrying specific KRAS mutations could also be detected in GPC-1+ exosomes, indicating its role as a prediction and prognostic biomarker for pancreatic cancer. Combination of different components within exosomes could result in high sensitivity in cancer detection. In pancreatic cancer, cancer-initiating cell markers (CD44v6, Tspan8, EpCAM, MET, and CD104) had a sensitivity of 0.96 and a specificity of 0.86, while pancreatic cancer-associated exosomal miRNAs had a sensitivity of 0.81 and specificity of 0.94. A combination of both testing approaches achieved a sensitivity of 1 (213).

f. Future perspectives of EV-based diagnosis

A Large number of EVs can be released from a single cancer cell, suggesting that EV is a signal amplifier for cancer detection. More importantly, they inherit the disease markers from parental cancer cells and can be easily accessed from blood, urine, or saliva. These properties of EVs make them the perfect targets for tracing cancer cells. EV-based diagnosis enables the possibility of cancer identification at a very early stage. Continuous efforts are underway to identify EVs as diagnostic markers. A summary of biomarkers associated with cancer cell-derived EVs is shown in Table 4. Various research institutions and companies, such as Exosome Diagnostics, are actively working on the development of EV-based diagnosis. One successful innovation is the ExoDx Prostate(IntelliScore) system for PCa screening. As aforementioned, compared to the conventional SOC method, IntelliScore achieves higher sensitivity and specificity in detecting high-grade PCa without the need of invasive tissue collection. A study involving 774 patients to test the IntelliScore system using urine exosomes was published before the commercial launch of the system (209). The company also provides the lung cancer diagnostic system ExoDx Lung(ALK) as a tool to detect EML4-ALK fusion transcript in serum exosomes, while other systems including ExoDx Lung(T790M) and ExoDx Lung(EGFR) will be available in late 2017. These serum exosome-based diagnostic tools can provide guidance for targeted therapy in lung cancer patients. The efficiency of these strategies is currently being validated in clinical trials (NCT03236675).

Table 4.

Diagnostic Value of EVs in Cancer

Type of cancer	Marker	Type of EVs	Conclusions	Source	Reference
Colorectal Cancer	Level of circulating exosomes	Exosome	Level of exosomes in cancer patients is statistically higher than in healthy controls. It correlates with CEA and poorly differentiated tumors and shorter overall survival.	Plasma; CCM	(199)
	let-7a, miR-1229, miR-1246, miR-150, miR-21, miR-223, and miR-23a	Exosome	Serum exosomal levels of seven miRNAs are significantly higher in CRC patients, cell lines compared to healthy controls.	Serum; HCT116, HT-29, RKO, SW48, and SW480	(205)
Prostate Cancer	miRNAs	Exosome and MV	12 miRNAs are differentially expressed in prostate cancer patients compared with controls; Among them, miR-107 and miR-574-3p can serve as urine marker. 11 miRNAs, including miR-141 and miR-375, are higher in prostate cancer patients with metastases compared with those without metastases.	Plasma; Serum; Urine	(204)
	miR-1290 and miR-375	Exosome	Plasma exosomal miR-1290 and miR-375 are associated with poor overall survival of CRPC patients.	Plasma	(287)
	Frequency of apoptotic bodies	AB	The level of apoptotic bodies in tumor tissue is positively correlated with Gleason grade.	Tumor tissue	(201)
	mRNA of ERG, PCA3, and SPDEF	Exosome	ExoDx Prostate IntelliScore gene signature in urine exosome plus SOC could increase the sensitive of detecting PCa with Gleason scores higher than 7 in a non-invasive way.	Urine	(209)
	mRNA of PCA-3 and TMPRSS2:ERG	Exosome	The mRNAs of PCA-3 and TMPRSS2:ERG in urine exosome can be used as the biomarkers for diagnosis and prognosis.	Urine	(208)
	PTEN	Exosome	Prostate cancer cells down-regulate PTEN by incorporating PTEN into exosomes, which can be used as a blood biomarker for diagnosis.	plasma and DU145, PC-3	(288)
	PDCD6IP, FASN, XPO1 and ENO1	Exosome	By using high performance LC-FTMS, exosomal proteins PDCD6IP, FASN, XPO1 and ENO1 are discovered as new candidate biomarkers for prostate cancer.	PC346C and VCaP	(289)
	Survivin	Exosome	Prostate cancer patients express higher level of Survivin in blood exosome compared to BPH and healthy controls. Exosomal Survivin levels are also higher in patients with relapse on chemotherapy.	Plasma; serum	(290)
	miR-34a	Exosome	Decreased exosomal miR-34a indicates higher resistance to Docetaxel in prostate cancer.	22Rv1, DU145, and PC3	(291)
			The prostasomes in the serum can be detected by a sensitive and specific
	Level of prostasomes	MV	assay-proximity ligation assay (PLA). By PLA, higher level of prostasomes can be detected in prostate cancer patients compared to healthy donors. The amount of prostasome in serum is also correlated to prostatectomy Gleason scores.	Plasma	(198)
	Level of large oncosomes	Large oncosome	Large oncosome could be detected in human tumor tissues, but not in benign compartments. In mouse model of prostate cancer, large oncosome could be identified in tumor tissues and circulation. The abundance of large oncosome correlated with tumor progression.	Plasma; Tumor tissue	(38)
Ovarian cancer	Claudin-4	Exosome	Claudin-4-containing exosomes can be detected in the serum of ovarian cancer patients. Claudin-4 and CAI 25 are strongly correlated.	Plasma; Cell lines	(211)
Esophageal Squamous Cell Carcinoma	miR-21	Exosome	Exosomal miR-21 is significantly higher in patients with ESCC and it is positively correlated with disease progression.	Serum	(203)
Pancreatic cancer	CD44v6, Tspan8, EpCAM, MET and CD104; miR-1246, miR-4644, miR-3976 and miR-4306	Exosome	Pancreatic cancer patients have higher CD44v6, Tspan8, EpCAM, MET, CD104, miR-1246, miR-4644, miR-3976 and miR-4306 in serum exosome. Combined evaluation of miRNAs and protein in exosome for diagnosis achieves high sensitivity and specificity.	Serum; Cell lines	(213)
	Transcripts of Apbb1ip, Daf2, Foxp1, Incenp, BCO31781, and Gng2;	Exosome	Transcriptomic biomarkers of Apbb1ip, Daf2, Foxp1, Incenp, BCO31781, and Gng2 are up-regulated in both serum and saliva-derived exosomes of pancreatic tumor-bearing mice.	Salivary and serum from mouse	(292)
	Genomic DNA	Exosome	Serum exosomes of pancreatic cancer patients carry genomic DNA spanning all chromosomes with mutated KRAS and P53.	Serum; T3M-4 and Panc-1	(210)
	Glypican-1	Exosome	Glypican-1 positive serum exosomes distinguish pancreatic cancer patients from healthy controls. The level of glypican-1 positive serum exosomes correlates with tumor burden and the survival of patients.	Serum	(212)
Multiple cancers	dsDNA	Exosome	DsDNA is present in exosome, representing the whole genomic DNA from cancer cell. Exosomal dsDNA can be used to identify mutations and cancers.	Plasma; Cell lines	(15)
Lung cancer	Level of circulating exosomes and exosomal miRNAs	Exosome	Lung cancer patients have increased level of circulating exosome. MiR-17-3p, miR-21, miR-106a, miR-146, miR-155, miR-191, miR-192, miR-203, miR-205, miR-210, miR-212, hsa-miR-214 are upregulated in NSCLC.	Plasma	(206)
	miRNAs	Exosome	MiR-378a, miR-379, miR-139-5p, and miR-200b-5p are up-regulated in patients with lung adenocarcinomas or lung granulomas. Expression of exosomal miR-151a-5p, miR-30a-3p, miR-200b-5p, miR-629, miR-100, and miR-154-3p can be used to distinguish lung adenocarcinomas from lung granulomas.	plasma	(293)
Melanoma	CD63, Caveolin-1	Exosome	Plasma exosomes expressing CD63 or caveolin-1 are higher in melanoma patients compared to healthy donors. Caveolin-1+ plasma exosomes have higher sensitivity compared to CD63+ plasma exosomes.	Plasma; Me501 and MeBS	(22)
Gastric cancer	Let-7	Exosome	Exosomal let-7 miRNA family is up-regulated in metastatic gastric cancer cell line.	Cell lines	(294)
	LINC00152	Exosome	Exosomal LINC00152 is higher in plasma of gastric cancer patients compared to healthy controls.	Plasma	(207)

Despite the great promise of EV-based diagnosis, efficient capture of cancer cell-derived EVs apart from normal cell-derived EVs is still challenging. Conventional methods of EV isolation are time-consuming and labor-intensive, preventing the broad application of EV-based diagnosis in clinical practice. Thus, development of new approaches aimed at efficient isolation of EVs is emergent. Currently, microfluidic platforms are being actively developed and optimized for EV isolation and analysis. These devices are usually in a chip format and have built-in components for capturing EVs by filtration (214) or immunoaffinity (215, 216). In addition to isolating EVs, these devices have integrated components for further analysis of EVs. One of the integrated functions is the quantification of EVs (214, 216) since the number of EVs is an important biomarker of cancer. In addition, determining the content of EVs is also very valuable for diagnosis and prognosis, and microfluidic chips have been developed to profile the molecular markers carried by EVs by Hakho Lee and Ralph Weissleder. They reported the iMER platform, which captures EVs, and performs on-chip RNA isolation and profiling (217). They validated this device on detection of MGMT and APNG, the drug-resistance markers in glioblastoma multiforme (GBM). More importantly, the mRNA levels in EVs detected by iMER correlated with those in cancer cells, indicating the potential of iMER in predicting drug response in GBM patients. They also developed nPLEX to profile the protein expression in exosomes from cancer patients (218). This nPLEX chip has a lattice of 44 × 32 nanoholes per sensing unit, with a nanohole diameter of 200 nm. Facilitated by antibodies, theses nanoholes specifically capture exosomes and binding can be monitored by measuring either wavelength shifts in the light spectrum or intensity changes at a fixed wavelength. By modifying the capture antibodies, this chip is able to quantitatively detect exosome proteins. Meanwhile, IBM also developed a new lab-on-a-chip/nano-DLD technology aimed at rapid cancer diagnostics with automated exosome isolation. Other than the small microfluidic chips that could be used for diagnosis of a specific type of cancer, more integrated systems aimed at comprehensive screening of cancers by EVs are being developed. As the leader in the liquid biopsy market, Exosome Diagnostics developed an integrated exosome analysis system, Shahky™, for exosomal protein capture and quantitative analysis. They also recently announced the launch of MedOncAlyzer 170, which examines both exosomal RNA and circulating tumor DNA. By combining both exosomal RNA and ctDNA, this system could identify cancers in both early and late stages. The mutations in COSMIC, a catalogue of 8,2000 somatic mutations in oncogenic variants of 170 cancer-related genes, could be detected in a plasma sample less than 0.5 mL. With high sensitivity and specificity, these integrated systems could be powerful tools for disease diagnosis, and particularly complementing current diagnostics for the early stages of cancer. The landscape of cancer diagnosis will undergo rapid change with the involvement of EV-based techniques.

4.2. Therapeutic applications of EVs in Cancer

a. Target EVs in cancer

Therapeutic applications of EVs have also been actively explored. Consistent with the nature of increased secretion from malignant cells, more tumor-promoting roles of EVs have been revealed compared to tumor-suppressive roles. Thus, direct targeting of EVs has been tested for the treatment of cancer. Several drugs have been found with the ability to inhibit biogenesis of exosomes in multiple studies (Table 5). Of them, GW4869 is widely used to eliminate exosomes. Sphingomyelinase mediates the formation of ceramide from sphingomyelin and ceramide is essential for the budding of intraluminal vesicles (ILVs) into MVBs (67). GW4869, as an inhibitor of nSMase2, was shown to decrease the biogenesis of exosomes in several cell lines (130, 219, 220). Similar to GW4869, other nSMase2 inhibitors, including spiroepoxide, glutathione, and manumycin-A (67, 221), were also found to inhibit the exosome biogenesis in various reports. However, inhibition of nSMase2 did not lead to exosome suppression in all cell lines (222), suggesting an alternative regulation of exosome biogenesis in different cell types. Another important regulator of exosome biogenesis is calcium (223). It was found in a hematopoietic cell line, K562, that monensin could enlarge MVBs and increase the production of exosomes. This effect was achieved by the increase of Ca²⁺ as the Na+/H+ exchanger, monensin, reversed activity of the Na⁺/Ca²⁺ exchanger (224). Furthermore, transferrin, which was reported to increase the intracellular Ca²⁺ level, was also able to induce exosome release in K562 cells. Interestingly, dimethyl amiloride, an inhibitor of the H+/Na+ and Na+/Ca²⁺ exchangers, was found to inhibit exosome release in K562 cells. Other than nSMase2 and calcium, exosome biogenesis was also found to be related with several other signaling pathways. Brefeldin A-inhibited guanine nucleotide-exchange protein (BIG)2 was found to be associated with the release of type I tumor necrosis factor receptor carried by exosomes from human vascular endothelial cells (225). Use of brefeldin A has been reported to successfully decrease exosome production from cells (119, 221). The regulatory effect of BIG2 on exosome release is dependent on two distinct mechanisms. One is through the activation of class I ADP-ribosylation factors (ARFs) 1 and 3 (168), while another is through the role of BIG2 as a kinase-anchoring protein for RIIβ that localizes protein kinase A signaling in exosome trafficking (226). In addition, ARF 6 was found to be associated with exosome biogenesis, particularly for syntenin-positive exosomes (58). ARF6 induces budding of ILVs in MVBs through its effector phospholipase D2 (PLD2) and treatment of PLD2 inhibitor CAY10594 on MCF7-suppressed exosome production (58). However, ARF6 was also found to regulate shedding of MVs from cancer cells (227). The specificity of CAY10594 needs to be further addressed to clarify its role in exosomes and MVs. While the above reagents target the biogenesis of exosomes, the uptake of EVs can also be suppressed to block the oncogenic roles of EVs on recipient cells. Disruption of actin cytoskeletal filaments by Cytochalasin D or Lantrunculin A has been shown to suppress exosome uptake (228). Treatment with heparin, a competitive inhibitor of cell surface receptors dependent on HSPG coreceptors, decreased exosome uptake in bladder cancer cells (229). However, these reagents also inhibit endocytosis in target cells, and therefore, the possible negative side effects of these reagents should be considered. Other than exosomes, drugs targeting MV release have also been developed. While the biogenesis of MV occurs by out-budding of the plasma membrane, constituents of membranes have been explored for their roles in this process. Actin is a cytoskeleton protein positively involved in membrane dynamics, cell mobility, and contraction, and it is essential for MV biogenesis. Inhibition of actin-related proteins suppressed the release of MVs from cells. One of the MV inhibitors with this mechanism of action is calpeptin (230, 231), an inhibitor of calpain. Calpain-mediated destabilization of the cortical actin cytoskeleton is essential for MV biogenesis. Another important regulator of actin filaments is RhoA. Inhibition of RhoA pathway by targeting the factor downstream of RhoA, Rho-associated, coiled-coil containing protein kinase, resulted in a decrease of MV release (232, 233). Due to the shared membrane trafficking machineries or regulatory pathways, some reagents could lead to inhibition of both exosomes and MVs. For example, exosome release was found to be regulated in a calcium-dependent manner (223); however, calcium also regulates phospholipid redistribution, a process related to MV production (234). Calcium chelator EGTA was found to inhibit both exosome and MV production from cells (223, 235). Currently, these EV-inhibiting reagents are limited to research use only. In addition to inhibition of EV release, these reagents also induce off-target changes, as many of them are major regulators of cell functions. The translational implication of EV-targeted reagents requires further understanding of the regulatory pathways specific to each EV.

Table 5.

Drugs targeting EV

Drug	Target	Direction	Mechanism of Action	Reference
dimethyl amiloride	Exosome	Inhibition	inhibitor of the H+/Na+ and Na+/Ca2+ exchangers	(223)
Transferrin	Exosome	stimulation	transferrin receptor (increase cytosolic Ca2+)	(223)
Monensin	Exosome	stimulation	a Na+ ionophore (increase cytosolic Ca2+)	(223)
GW4869/spiroepoxide/glutathione/manumycin-A brefeldin A	Exosome	Inhibition	target nSMase2 for ceramide formation	(67, 221, 222)
	Exosome	Inhibition	Inhibitor of BIG2	(226)
CAY10594	Exosome	Inhibition	Inhibitor of PLD2	(58)
chlorpromazine	Exosome	Inhibition (uptake)	block clathrin-mediated endocytosis	(295)
Cytochalasin D/Lantrunculin A	Exosome	Inhibition (uptake)	disrupt actin cytoskeletal filaments	(228)
Heparin	Exosome	Inhibition(uptake)	Inhibitor of cell surface receptors dependent on HSPG co-receptors	(229)
apilimod	Exosome	stimulation	PIKfyve	(296)
CI-1033/PF-00299804	Exosome	stimulation	ErbB/EK	(297)
Cl-amidine	Exosome and MV	Inhibition	Inhibit protein deimination of cytoskeletal actin	(298)
Bisindolylmaleimide-I	Exosome and MV	Inhibition	protein kinase C inhibitor preventing the externalization of phosphatidylserine	(298)
Imipramine	Exosome and MV	Inhibition	Inhibitor of acid sphingomyelinase aSMase	(298, 299)
EGTA	Exosome and MV	inhibition	calcium chelation	(223, 235)
MßCD	Exosome and MV	Inhibition	membrane cholesterol depletion	(300, 301)
D-pantethine	MV	Inhibition	blocks the translocation of the aminophospholipid phosphatidylserine from the inner to the outer membrane leaflet	(302, 303)
Y-27632	MV	Inhibition	Target RhoA mediated action mobilization and MV formation	(232, 233)
calpeptin	MV	inhibition	inhibitor of calpain mediated destabilization of cortical actin cytoskeleton	(230, 231)

b. EV-based cancer vaccine

Cancer cell-secreted EVs could also be used for vaccine immunotherapy. It was found that both tumor cell-derived EVs and immune cell-derived EVs have a role in immune activation. More importantly, antigen-presenting EVs from B lymphocytes and dendritic cell containing MHC class I and class II complexes could also stimulate both CD4+ and CD8+ T cells (9, 236, 237). Other than presenting antigens directly, EVs have also been found to transfer antigenic peptides to antigen-presenting cells (APCs) (238), or exchange functional peptide-MHC complexes between APCs (236), thus mediating the indirect antigen presenting process. For example, exosomes released from tumor cells could be recognized and engulfed by dendritic cells for antigen presentation, which is capable of inducing potent CD8+ T cell-dependent antitumor effects (239). These exosomes also contain enriched immunostimulatory factors such as tetraspan proteins CD63 and CD82, which is a possible reason for the superior anti-tumor effect compared to that of whole dendritic cells (240). Exosome released from dendritic cells could act as signal extender by transmitting antigen to other dendritic cells (241). These findings suggest that exosomes could be used as a cell-free vaccine for immunotherapy in cancer patients (Figure 3). In a phase I clinical study of cancer vaccine using dendritic cell-derived exosomes (dexosome), tumor-specific antigens-MAGE peptides were loaded into dexosomes from NSCLC patients (242). These dexosomes were given back to patients and were well-tolerated. More importantly, these vaccines activated immune response and induced long term stable disease in some patients. A similar study was carried out in 15 late stage melanoma patients. MAGE 3 peptides were loaded onto autologous DC derived-exosomes and then used as cancer vaccine. The dexosome-based cancer vaccine was well-tolerated, and in 4 patients some extent of therapy response was observed; in particular, partial response was observed in 1 patient (243). However, in this trial, no significant T cell response was observed. These two studies suggest the safety and feasibility of using exosomes as cell-free cancer vaccines. Furthermore, an intriguing finding in both trials was the increase of NK activity following immunization, suggesting dexosomes contained stimulatory factors for innate immune cells (242, 243). To further verify the safety and immunotherapeutic effect of dexosomes, a phase II clinical trial was performed in 22 late stage NSCLC patients after first line chemotherapy (244). The dexosomes in this study was termed second generation since it was optimized to boost both NK cell and T cell response. To achieve this purpose, the exosomes were isolated from IFN-γ matured DC before antigen loading. These second generation dexosomes were well tolerated and 7 patients experienced stabilized disease during 4 months, which didn’t meet the primary endpoint that 50% patients achieve progression-free survival by 4 months post-chemotherapy. However, dexosome-enhanced NK cell response was observed in patients with prolonged survival and it correlated with the expression of MHC II and BAG6, which are known to be activators of NK cells, on the dexosomes (245–247). However, there was no clear evidence of activation of tumor-specific T cell response in patients, which might be caused by several reasons. One may be the immunosuppressive phenotype associated with these patients since they were previously vaccinated with metastatic NSCLC. Other than dexosomes, exosomes from other sources have also been shown to have immunomodulatory functions. Tumor cell-derived exosome (TEX) was also found to be immunogenic with antigen presenting functions (239). It was reported to trigger tumor specific DC and T cell responses in vitro and in vivo in multiple types of cancers (248, 249). However, use of TEX in a clinical setting has its limitation due to the source of cancer cells. In some cases of cancers, ascites could be an alternative source of TEX. In a phase I clinical trial in 40 advanced CRC patients, exosomes from ascites were shown to induce tumor-specific antitumor cytotoxic T lymphocyte response when combined with GM-CSF treatment (250). Similar to other reports, treatment with exosomes in CRC patients was well-tolerated. However, ascites is limited by the types of cancer and clinical presentation. Moreover, TEX was also found to be immunosuppressive in some cases (251, 252). With these complications, especially oncogenic compositions in TEX, its use as a cancer vaccine needs further investigation to verify its feasibility and safety. Exosome-based immunotherapy requires the loading of antigen, which can be achieved in both indirect and direct ways (242). In the indirect approach, peptides are added into DC culture, so they can be loaded onto MHC proteins expressed on the cell surface. Later antigen-presenting exosomes could be purified from the CM of these peptide-treated DC (253). In the direct loading approach, antigen-presenting exosomes are generated by acid elution. DC-derived exosomes are mixed with peptide in acetate buffer of pH 5.1 and assembly of MHC and peptide is achieved by neutralizing the buffer to pH 7.0 (253).

Figure 3. Therapeutic Applications of EVs.

Exosomes purified from tumor cells and dendritic cells are used as tumor vaccine. By loading tumor specific antigens to exosomes, they can activate and mobilize immune cells to kill tumor cells. Exosomes derived from tumor cells, dendritic cells, MSCs and HEK293 cells can be used as drug carrier for cancer theray. Specifc ligands or receptors are expressed in exosomes, if not, can be introduced into exosomes to achieve targeted delivery. Loading drugs into exosomes results in better biodistribution, stablility and less side effects. Inhibition of EV biogenesis by drugs is also reported to suppress the tumor growth.

c. EV-based drug delivery system

Another therapeutic application of EVs includes use as a drug delivery system (Figure 3). As drug carriers, exosomes outcompete ABs and MVs. There are several advantages associated with exosome-based drug carriers. First, their size distribution is smaller and more homogenous compared to that of MVs and ABs. The advantage in size enables better biodistribution and bioavailability of exosomes in the human body. Escape of lung clearance and permeability across the blood-brain barrier were reported for exosome-based drug carriers (200, 254). In addition, exosomes could achieve targeted delivery compared to synthetic nanoparticles. Exosomes could recognize and thus deliver cargoes to specific types of cells by the ligands or receptors expressed on the membrane. One of these bridges is the tetraspanin family, which is widely expressed on exosome membrane and acts as an exosome marker (255). Tetraspanin 8, an organizer of microdomains in the membrane, was found to be expressed in a rat pancreatic cancer cell line and associated with integrin α4 (256). Expression of tetraspanin 8-integrin α4 complex resulted in selective uptake of exosomes by endothelial and pancreas cells, with CD54 as the ligand. Other integrins were also found to be associated with organotropic metastasis of multiple cancers (185). Integrin β₄ was associated with lung metastasis while integrin β5 was associated with cancer liver metastasis. Targeted delivery of exosomes from different cell lines to different organs mediated by integrin receptors resulted in pre-metastatic niches responsible for organotropic metastases. In another study, MUC1 was expressed on the exosome from breast milk and responsible for the targeted delivery to monocyte-derived dendritic cells via the CD209 receptor in the recipient cells (257). Other than the natural targeting proteins carried by exosomes, artificial targeting components can be easily engineered onto exosomes. While for synthetic nanoparticles, targeting components must be loaded directly, in exosomes they can be assembled using either direct or indirect approaches. As mentioned in dexosome, peptides can be loaded onto either exosomes or parental cells (242, 253). Compared to synthetic nanoparticles, the approach of modification of exosomes through parental cells is much easier and cost efficient. Ectopic expression of targeting components can be introduced into parental cells that generate a significant number of engineered exosomes. For example, a vector coding targeting peptides fused with a transmembrane protein, such as Lamp2b, was introduced into parental cells to enhance the targeted delivery (258). Secondary, the source of exosome is abundant. Drug carrier exosomes have been reported to be isolated from various types of cells, including HEK-293 cells, immature DC, MSCs, and cancer cells (259–263). Furthermore, the exosome-based drug carrier is less immunogenic compared to synthetic nanoparticles. Different therapeutic reagents including protein, siRNA, miRNA, and chemicals can be incorporated into exosomes via electroporation, chemical-based transfection, modification of parental cells, or direct incubation (264), which increases the bioactivity and achieve targeted delivery in patients. It was reported that exosomal delivery of curcumin increased the stability and the blood concentration of the drug, inducing specific delivery to inflammatory cells (261). As a follow up of this work, James Graham Brown Cancer Center initiated two phase I clinical trials to test the therapeutic effect of plant exosomes on head and neck cancer as well as colon cancer (NCT01668849, NCT01294072).

5. Discussion

EVs are emerging targets for cancer diagnosis and therapy. The physiological and pathological roles of EVs are under active study. However, successful translation from bench to bedside still needs a deeper understanding of these particles that were previously considered cell debris. However, current studies of EVs still lack the standardization of isolation and analysis. A primary reason is a lack of the specific markers of exosomes and MVs, which hampers accurate demarcation of their identities. Exosome makers, such as tetraspanins, could also be found in MVs (19), and past studies may lack the accurate authentication of the particles studied, which is a potential cause of controversial results from different publications. With various approaches now available, the isolation methods have been shown to have a great impact on subsequent analysis (111). Thus, standardization of the isolation method is critical for an accurate assessment of EVs. In this review, we discussed the advantages and disadvantages of currently available isolation techniques and their impact on subsequent analyses, which we hope will be helpful to standardize the isolation protocols in this research field, although the identity of isolated particles still needs to be verified due to the variations in morphologies of EVs. While exosomes usually have diameters ranging from 40–100 nm, ones with diameters larger than 100 nm have also been identified (250). Different subpopulations of exosomes were discovered with diverse sizes, cargoes, and morphologies (19, 24, 265–267). These findings suggest that classification of EVs requires combinatory analyses of sizes, makers, and shapes.

When quantifying the RNA content in exosomes and MVs, especially small RNAs, the normalization of data may be a critical point. Currently, there are three strategies to normalize qRT-PCR readings of small RNA in EVs. They can be normalized to average Cq values of all arrays. While this approach has the highest accuracy, it needs a large number of arrays and is more suitable for global gene profiling. Another strategy is to normalize the expression to endogenous reference genes. However, there is no well accepted “housekeeping” transcript for small RNA in EVs. Selection of endogenous reference genes needs optimization and verification on a case-by-case basis. The third approach is using an exogenous spike-in synthetic oligonucleotide. This strategy is very controversial. It could provide an estimate of efficiency in RNA extraction and reverse transcription. However, for using them for normalization, a fixed amount of spike-in needs to be added into exactly the same amount of RNA from different samples, which is not practical. Furthermore, this approach cannot compensate the variations in RNA quality resulted from sample handling before the spike-in is added (268). Thus, normalization of small RNAs from EVs requires great caution, and selection of reference genes needs to be carefully determined and verified. Another concern is the EV concentration used in the functional study. The exosome concentration in plasma has been reported to be 10⁹–10¹¹ exosome/ml (269, 270). When translated into protein, serum and plasma have the concentrations at 50–200 μg/ml while freeze-thaw can increase the concentration to 50–500 μg/ml (269, 271). Thus, exosome treatment should ensure that the concentration used is within the physiological range.

While the translational application of EV as a diagnostic tool has been rapidly developed and established, the therapeutic application of EVs still lags behind. Even though exosome targeted therapy has been shown to have an anti-cancer effect in specific cases in pre-clinical models (182, 272), efficacy and safety in clinical use need further study. Currently, there is no available drug that can be used to safely and specifically inhibit EV biogenesis in patients. More investigation is needed to picture the layout of the machineries involved in EV biogenesis, cargo sorting, and EV uptake. Understanding how cancer cells coordinate these machineries to generate EVs with tumor-promoting effect could provide the opportunities to find specific regulators involved in EV-associated cancer development, which helps the development of drugs targeting EVs in cancer without significant side effects. A potential direction is looking at the different subpopulations among the same type of EVs. Cancer cells were found to secrete different subpopulations of exosomes harboring different cargoes and functions (273, 274). Deciphering the differences between these subpopulations is necessary for fully understanding the roles of cancer cell derived exosomes. It is also crucial to distinguish cancer-promoting exosomes and cancer supportive exosomes since both have been identified in cancer cells. This would help in identifying specific targets for exosome-targeted therapy. It also helps in identifying the appropriate population of tumor cell-derived exosomes that can be used as a tumor vaccine. At the same time, understanding the differences between EVs from same cells would also optimize the efficacy of a dexsome-based vaccine. It was reported that MHC II was more enriched on larger EVs (100–200 nm) compared to that on the smaller EVs (30–50 nm) (56), which might be a cause of the failure of exosome-based cancer vaccine in inducing T cell-specific responses (243, 244). Selection of right EV population could enhance the immunostimulatory effect of these vaccines, enabling their better clinical use. Currently, more studies are conducted on roles of exosomes in cancer, and less emphasis is placed on MVs and ABs. However, MVs and ABs are also known to impact cancer progression, and have regulatory pathways distinct from those of exosomes (275, 276). Therefore, understanding differential roles of exosomes, MVs, and ABs in cancer is equally important, which is likely to reveal novel diagnostic and therapeutic opportunities.

EVs as drug carriers are also being actively explored. Exosomes from stem cells have been identified with therapeutic effect in tissue repair, immune modulation, and anti-inflammation (277). They were also found to be candidates for carriers of anti-cancer drugs. RoosterBio and Exopharm recently announced the Exomere project that aims at developing clinical-grade stem cell-derived EVs for therapeutics. Another company, MDimune, is developing BioDrone™, an anti-cancer drug delivery system based on EVs. With more active studies on EVs and rapid development in technology, this field of study may change the landscape of the fight against cancer in the near future.

Abbreviations

EV(s)

extracellular vesicle(s)

MV(s)

microvesicle(s)

AB(s)

apoptotic bodie(s)

TME

tumor microenvironment

TfRs

transferrin receptors

MVE

multivesicular endosomes

MHC

major histocompatibility complex

miRNAs

microRNAs

PCa

prostate cancer

ESCRT

Endosomal Sorting Complex Required for Transport

nSMase2

neutral sphingomyelinase 2

HSPG

heparan sulfate proteoglycan

VPS

Vacuolar Protein Sorting

THPO

thrombopoietin

ANGPTLs

angiopoietin-like proteins

SIMPLE

small integral membrane protein of the lysosome/late endosome

CRC

colorectal cancer

EBV

Epstein Barr virus

LMP1

latent membrane protein 1

hnRNP

heterogeneous nuclear ribonucleoprotein

lncRNAs

long non-coding RNAs

MVP

Major vault protein

Ndfip1

Nedd4 family-interacting protein 1

3′UTR

3′-untranslated region

RISC

RNA-induced silencing complex

CCM

cell culture media

HDLs

high-density lipoproteins

PEG

polyethylene glycol

SEC

size-exclusion chromatography

THP

Tamm–Horsfall glycoprotein

ncRNAs

non-coding RNAs

CAFs

Cancer-associated fibroblasts

MSCs

mesenchymal stem cells

MET

mesenchymal-to-epithelial transition

LINC00152

long intergenic non-protein coding RNA 152

SOC

standard of care

PSA

prostate-specific antigen

dsDNA

double-stranded DNA

NSCLC

non-small cell lung cancer

GPC1

glypican-1

GBM

glioblastoma multiforme

ILVs

intraluminal vesicles

BIG

Brefeldin A-inhibited guanine nucleotide-exchange protein

ARF(s)

ADP-ribosylation factor(s)

PLD2

phospholipase D2

PLA

proximity ligation assay