Hematologic malignancies (HMs) represent a heterogeneous group of hematopoietic neoplasms commonly characterized by the abnormal production of blood cells. It is possible that the shelter provided to malignant cells by the cancer-modulated niche contributes to refractory cases and relapse (36). Exploring the exact alterations of the “hijacked” niche and how these promote disease progression could shed light on potential new cancer therapies. Indeed, dissecting the intercellular communication networks among malignant cells, normal cells, and their surrounding microenvironment could provide valuable information on the optimal way to target tumor-permissive niches. Here, we discuss the roles of EVs in HMs, focusing on the cargoes transferred, the genes that are regulated, and how the biological behaviors of recipient cells are altered, as well as the mechanisms by which EVs contribute to disease development.

Tumor-derived EVs

Tumor-derived EVs affect tumor cells and subpopulations. Tumor-derived EVs are involved in the maintenance of cancer stem cells, metastasis, and resistance to chemotherapeutic drugs, and there is accumulating evidence of the direct and indirect roles of tumor-derived EVs in HMs. Tumor-derived EVs have been shown to directly modify the behavior of malignant cells in several types of HMs. For example, diffuse large B cell lymphoma tumor cell lines and patient samples were composed of flow cytometry–defined side population (SP) cells and non-SP cells. SP cells were characterized as weakly positive Hoechst 33342–stained cells that were postulated to be leukemia stem cells (LSCs) in HMs. The transfer of Wnt3a-containing EVs was involved in the cell state transition of non-SP into SP cells. Specifically, SP cells provided EV-Wnt3a to neighboring non-SP cells, resulting in activation of the canonical Wnt signaling pathway in recipient cells (37). EVs derived from a human chronic myeloid leukemia (CML) cell line (LAMA84) enhanced tumor growth both in vitro and in vivo by providing antiapoptotic molecules and TGF-β (38). Therefore, targeting of the EV autocrine effect is implicated as a potential therapy strategy. We also confirmed that blocking EV maturation and secretion by acute myeloid leukemia (AML) cells through Vps33b knockout/knockdown suppressed AML cell growth and prolonged disease progression in both a mouse model and patient samples (17). Similarly, lentivirus-mediated knockdown of Rab27a also decreased the EV levels and prolonged AML mouse survival (39). In addition, miR-34c-5p downregulated RAB27B, thus inhibiting the EV-mediated transfer and consequently increasing the senescence and eradication of LSCs through p53/p21/cyclin-dependent or p53-independent pathways (40). The EV-mediated autocrine effect was also observed in patient plasma. Comparison of the protein cargoes from indolent and progressive chronic lymphocytic leukemia (CLL) cells revealed that S100A9 protein levels in plasma EVs increased significantly with disease progression, thus contributing to disease progression via activation of the NF-KB pathway (41).

Collectively, these observations demonstrate the contribution of tumor-derived EVs to the organization of tumor cell populations and disease progression. The past decade has witnessed extraordinary advances in our understanding of how the interaction between cancer cells and their microenvironment contributes to disease progression and overall survival. As an important cell-cell communication mechanism, tumor-derived EVs were often found in these communication scenarios (42).

Effects of tumor-derived EVs on BM niche components. The BM niche appears to be remodeled in various HMs. The remodeled niche exhibits common features, such as increased hypoxia, angiogenesis, inflammation, and metabolic reprogramming (36). The concept of tumor-derived EVs as potent mediators of intracellular interactions is now widely accepted and has led to an increase in studies focused on deciphering these communication networks. Indeed, emerging evidence has confirmed that tumor-derived EVs actively contribute to formation of the tumor-permissive niche (Figure 2).

Roles and functional cargoes of tumor-derived EVs. Tumor-derived EVs, released by malignant cells, act as mediators of signals between malignant cells and hematopoietic cells (HSCs, macrophages, etc.) as well as non-hematopoietic cells (stromal cells, endothelial cells, osteoblasts, etc.). The crosstalk among various cell types via tumor-derived EVs results in remodeling the behaviors of recipient cells.

Endothelial cells. Angiogenesis is a common feature of tumors. Tumor-derived EVs contribute to endothelial cell (EC) remodeling during angiogenesis in a variety of HMs. Multiple myeloma (MM) cells were found to regulate multiple pathways, resulting in increased BM EC line (STR10) viability, enhanced angiogenesis, and immunosuppression in a murine model, which further facilitated MM progression (43). MM-EV–contained Piwi-interacting RNA-823 (piRNA-823) was essential for the EC modulation required to support the growth of MM cells (44). Human AML cell–derived VEGF-containing EVs were responsible for glycolysis-mediated vascular remodeling and chemoresistance acquisition in AML (45). Furthermore, EVs derived from a CML cell line (LAMA84) induced a rapid reduction of CXCL12 and VCAM1 expression on ECs (46). Additionally, in acute promyelocytic leukemia (PML), EVs contained high levels of PML retinoic acid receptor-α transcripts. EV treatment resulted in the acquisition of procoagulant and tissue factor antigen in ECs (47). Since tumor blood vessels are key targets for therapeutic management, deciphering the mechanisms of EC remodeling in HMs is an important focus of research.

Mesenchymal stromal cells and descendant cells. Experimental evidence has demonstrated that tumor-derived EVs can broadly modulate MSC proliferation, differentiation potential (mainly referring to osteogenesis and adipogenesis), hematopoietic supportive function, and metabolic profiling. In turn, these alterations modulated disease progression. AML-EV treatment increased Dickkopf-1 (DKK-1) expression and decreased osteogenesis of MSCs in an AML mouse model, providing direct evidence of the function of AML-EVs in vivo. More pertinently, leukemogenesis was largely accelerated after mice were pretreated with AML-EVs (39). AML-EVs also modulated the sensitivity of malignant cells to chemotherapy by altering MSC function. Mechanistically, AML-EVs elicited an unfolded protein response (UPR), which increased osteogenic priming of MSCs, potentially through the transfer of BMP2 (48). The UPR activated PERK/eIF2/ATF4 signaling during osteoblast differentiation followed by upregulation of the expression of genes that are essential for osteogenesis (49). Suppression of osteolineage cell function by tumor-derived EVs was also observed in MM and systemic mastocytosis (50–52).

In knockin syngeneic AML/acute lymphoblastic leukemia (ALL) mouse models, tumor-derived EVs increased the expression of adipose triglyceride lipase (ATGL) and hormone-sensitive lipase (HSL) enzymes in adipocytes, resulting in increased lipolysis, which supported leukemia cell expansion (53). Similarly, miR-92a-3p–containing EVs derived from a CML cell line (K562) attenuated adipogenesis of an adipose-derived MSC cell line (ADSCs) by inhibiting CCAAT/enhancer binding protein-α (C/EBPα). This inhibition of adipogenesis by tumor-derived EVs is postulated as a major mediator of cancer-associated cachexia (54).

EVs derived from primary AML patient cells and human leukemia/lymphoma cell lines (HEL, HL-60, MOLM-14, and U937) were internalized by a murine stromal cell line (OP9), resulting in increased proliferation and an altered growth factor secretion pattern in recipient cells. AML-EV–contained IGF-1R mRNA contributed to these changes (55). Furthermore, the same group demonstrated that AML-EVs downregulated critical retention factors (SCF, CXCL12) in stromal cells (56). Expression of JAG1 and SCF was also decreased after exposure to AML/myelodysplastic syndrome–derived EVs. This effect was partially abrogated by treatment with GW4869, which, as an inhibitor of the neutral sphingomyelinase SMPD2, blocks EV generation (57). Similar effects of tumor-derived EVs were found in other HMs (CML/CLL/MM/adult T cell leukemia/lymphoma [ATL]) (58–64). For example, CLL-EVs upregulated IL-8 expression in MSCs (60). Furthermore, miR-7977–containing EVs reduced the ability of human MSCs to support normal hematopoiesis via PCBP1 (57). Coincidentally, tumor-derived EV–contained miR-7977 modulated the Hippo/YAP signaling pathway in recipient MSCs, indicating its involvement in the increase in leukemia-supporting stroma growth (65).

Tumor-derived EVs can also regulate the metabolic state of BM stromal cells, which, in turn, become more supportive of malignant cells. Following internalization of ALL-EVs, a human stromal cell line (HS-5) showed a reduced oxygen consumption rate and increased extracellular acidification rate. These reprogrammed MSCs secreted an excess of lactate into the extracellular fluid, which is speculated to be the preferred energy source of tumor cells (66).

Taken together, these findings jointly illustrate that the modulations of MSCs caused by tumor-derived EVs not only constrain their capacity to support hematopoiesis, but also force them to become a shelter for leukemia cells.

Osteoclasts. Reduced bone volume is a shared characteristic of multiple HMs. In addition to the reduction in osteolineage-forming cells, the recruitment and abnormal activation of osteoclasts also contribute to bone loss (67). Culturing with MM-derived EVs improved the migration and differentiation of primary human osteoclasts and increased expression of osteoclast markers (68). Treatment of mice with EVs derived from a murine MM cell line (5TGM1) promoted osteoclast formation and blocked osteoblast differentiation, which were attributed to the EV-contained DKK-1 protein. Intriguingly, GW4869-induced blockade of EV secretion not only increased cortical bone volume, but also sensitized myeloma cells to bortezomib (52). EV-contained EGFR ligand was subsequently shown to contribute to this phenomenon (69).

Fibroblasts. The cancer-associated fibroblast is also an important niche component that is correlated with the survival of patients (70). Primary human myeloma cells modulated miR-27b-3p and miR-214-3p expression in fibroblasts through the release of EVs, which triggered proliferation and apoptosis resistance in myeloma fibroblasts via the FBXW7 and PTEN/AKT/GSK3 pathways, respectively (71). Shuttling of hTERT mRNA (the transcript of the telomerase enzyme) from Jurkat cells (human acute T lymphocyte leukemia cell line) via EVs transformed telomerase-negative fibroblasts into telomerase-positive cells, inducing increased proliferation, extension of lifespan, and the postponement of senescence (72). AML-EVs entered bystander fibroblast cells, resulting in increased proliferation and VEGF expression (55).

Macrophages. M2-like macrophage induction and recruitment contribute to the formation of the immunosuppressive niche in tumors (73). Exposure to K562-derived EVs reduced NO and ROS levels in macrophages, and EV-treated macrophages were polarized to the M2-like phenotype, accompanied by elevated secretion of TNF-α and IL-10 (74). Furthermore, recent work confirmed that human primary MM cell–derived EVs also modulated the polarization toward M2-like macrophages. More importantly, abundant EV-contained miR-16 targeted the NF-κB canonical pathway, thus contributing to the M2-like macrophage polarization, and indicating that miR-16 overexpression represents a target for therapeutics with enhanced sensitivity (75).

In brief, these observations have shed light on the cellular components of the niche that are modulated by tumor-derived EVs; however, other noncellular niche components that are also modulated await further exploration. For instance, while tumor-derived EVs have been reported to be actively involved in matrix degradation in solid tumors (76), their participation in extracellular matrix remodeling in HMs remains poorly understood. Tackling this barrier would help to clarify the mechanisms underlying tumor infiltration and metastasis.

Tumor-derived EVs and normal HSPCs. In many HMs, tumor cell infiltration is often accompanied by lethal cytopenia as a result of the impaired function of HSPCs. The profound suppression of HSPCs is caused not only indirectly by a less supportive niche (56, 57), but also directly through the action of tumor-derived EVs. The clonogenicity of HSPCs was attenuated by direct trafficking of AML-EVs containing microRNAs such as miR-150 and miR-155, which were sufficient to suppress murine HSPC clonogenicity, potentially by targeting the translation of the transcription factor MYB (77). We demonstrated that residual HSCs in leukemic mice were more quiescent than their counterparts in nonleukemic hosts (78). Later studies revealed that EVs impact the fate of HSCs via EV-dependent mechanisms. EV-contained miR-1246, which directly targeted the mTOR pathway and protein synthesis in HSCs, was shown to contribute to reversible quiescence and persistent DNA damage in murine HSCs (32). Similarly, AML-EVs carrying miR-4532 repressed normal hematopoiesis in human CD34+ HSPCs through activation of the LDOC1-dependent STAT3 signaling pathway (79). More importantly, hematopoietic progenitor cell differentiation was also compromised by EVs isolated from AML patient plasma through inhibition of dipeptidyl peptidase 4 (DPP4) in vitro (80). Therefore, exploring tumor-derived EV cargoes is likely to yield strategies that benefit hematopoietic regeneration and thus ameliorate cytopenia in HMs.

Tumor-derived EVs affect the immune niche and immunotherapy. In addition to their capacity for niche modulation and HSPC repression, tumor-derived EVs have also been reported to contribute to immune suppression in various tumors (81). Here, we discuss whether and how tumor-derived EVs interfere with antitumor immunity in HMs. EVs released by B cell lymphoma cells carried CD20 that functions as a decoy target in rituximab treatment, thereby allowing cancer cells to escape treatment (82). On the other hand, EVs isolated from Burkitt’s lymphoma cell line (Jurkat and Raji cell) culture supernatants downregulated NKG2D receptor–mediated cytotoxicity and impaired NK cell function in vitro, thus indicating that EVs induced immune evasion in HMs (83). Moreover, EVs isolated from AML patient sera contained TGF-β1, membrane-associated major histocompatibility complex class I chain–related genes A/B (MICA/MICB), and myeloid blast markers, suggesting that they were probably secreted by leukemia blasts and potentially contributed to immune suppression. Confirmation that expression of the activating receptor NKG2D and NK cell activity decreased after treatment with AML serum–derived EVs further validated this hypothesis. More importantly, these impacts on NK cells were reversed by TGF-β1 neutralizing antibody treatment (84). The level of TGF-β1 in EVs might reflect a response to chemotherapy (85). In a phase I clinical trial, EVs isolated from AML patient sera blocked the antileukemia cytotoxicity and other functions of a human NK lymphoma cell line (NK-92), inducing the failure of adoptive cell transfer therapy (86). These observations indicated that removing EV-contained TGF-β would benefit immune restoration in patients. Interestingly, lentiviral shRNA-mediated silencing of TGF-β1 in both murine lymphocytic leukemia cell line (L1210) and secreted EVs reversed the immune repression effect in vitro and in vivo (87).

Advances have shown that immunotherapy is a promising approach for HMs, with several studies reporting that tumor-derived EVs can be successfully combined with adoptive T cell therapy. Tumor-derived EVs were internalized and presented by dendritic cells, inducing a potent CD8+ T cell–dependent antitumor effect on syngeneic and allogeneic murine tumors (88). The cytotoxicity of cytotoxic T lymphocytes was increased by exposure to leukemia-derived EVs that contained high levels of HSP70 and ICAM1, thereby enhancing leukemia antigen presentation (89, 90). However, in a later study, EVs were shown to induce immune escape by upregulating PD-L1 expression. Transcriptome and proteome analyses of human primary CLL-EVs revealed an abundance of noncoding Y RNA hY4, the transfer of which contributed to an increased release of CCL2, CCL4, and IL-6, as well as upregulating PD-L1 expression on monocytes (91). In particular, PD-L1–containing EVs from melanoma cells were sufficient to inhibit CD8+ T cells in vitro and in vivo, thus facilitating the progression of melanoma (92). Similarly, PD-L1–positive EVs from patient plasma induced T cell exhaustion after chimeric antigen receptor T cell therapy in CLL (93).

Immune therapy based on tumor-derived EVs is still in the proof-of-concept stage. Although tumor-derived EV molecules inherited from the parental cells could function as tumor-specific antigens, further profiling and investigation of their efficiency are required. The profound negative effects on immune cells and interference in immune therapy emphasize the importance of caution in the application of tumor-derived EV–based immune therapy.

Niche cell–derived EVs

Given that cell-cell communication is a “two-way street,” researchers have focused on deciphering the roles of EVs derived from certain niche cell types. In a CML mouse model, miR-126 was transferred from ECs to CML-LSCs via EVs. Furthermore, conditionally knocking out miR-126 from ECs delayed leukemia progression and improved survival (94). However, the extent to which EV–miR-126 contributes to the overall transfer of miR-126 is unknown. In addition, in vitro cultures indicated that MSC-EVs protected AML cells against the cytotoxic effects of tyrosine kinase inhibitors (95, 96). EVs derived from the MSCs of primary MM patients (MM BM MSCs) and healthy volunteers (BM MSCs) caused opposing effects on tumor growth when transferred to MM cells, as MM BM MSC-EVs were found to promote MM tumor growth while BM MSC-EVs inhibited growth. It can be speculated that these opposing effects can potentially be explained by differences in the contents of microRNAs (miRNAs) and oncogenic proteins in the EVs (97). Another study showed that BM MSC-EVs increased proliferation and drug resistance in human MM cells (98). The uncertainty and contradictory conclusions among studies of EVs are an inevitable result of not only the heterogeneity of niche cells, but also differences in experimental factors such as models or culture conditions, EV isolation methods, doses, and intervals of administration. Genetically manipulated animal models are useful for clarifying these contradictions and systemically analyzing the function of specific cell type–derived EVs in vivo (99). We recently conducted a systematic exploration of the effects of specific BM niche cell–derived EVs using a conditional Vps33b-knockout mouse model and showed that EC-EVs accelerated AML progression. Mechanistically, we found that EC-EVs contained a high level of ANGPTL2, which bound to the PIRB receptor on AML cells and enhanced leukemia development. Furthermore, blocking the secretion of ANGPTL2-containing EVs from ECs delayed the progression of AML (Figure 3) (100). Thus, our research indicates the value of conditional knockout mouse models for exploring cell type–specific EV function in other systems and conditions, leading to a deeper understanding of the physiological and pathological roles of EVs.

EVs derived from ECs accelerate the progression of AML. Various cellular components in the BM niche secrete EVs. Niche cell–specific conditional Vps33b-knockout mouse models confirmed that EC-derived EVs accelerated AML progression (17). EC-EVs contained a high level of ANGPTL2, which bound to the PIRB receptor on AML cells and further enhanced leukemia development via the p-SHP2/p-CREB pathway (100). MVB, multivesicular body; ILV, intraluminal vesicle; TSPAN, tetraspanin; ESCRT, endosomal sorting complex required for transport; SEV, small extracellular vesicle.