Exosome biogenesis begins with an endocytic event at the plasma membrane. The most well-described mechanism for exosome formation is driven by the endosomal sorting complex required for transport (ESCRT), along with a multimolecular machinery that operates at the endosomal membrane during the various stages of exosome biogenesis. Although ESCRT complexes (TSG101, ALIX and VPS proteins) regulate membrane budding and cargo selection, current evidence indicates they do not directly recognize circRNA sequences (31). The selective loading of circRNAs into exosomes is predominantly mediated by RNA-binding proteins (RBPs), such as hnRNPA2B1, YBX1, FUS and QKI, which bind specific sequence motifs or secondary structures, including the exosome-enrichment motif 5'-GMWGVWGRAG-3' reported in circRNAs (32). These RBPs recruit circRNAs to sites of intraluminal vesicle formation, while ESCRT machinery facilitates vesicle scission (33). Emerging evidence suggests that AML blasts exhibit an increased exosomal output and an enhanced enrichment of oncogenic circRNAs compared to normal hematopoietic cells, likely due to dysregulated RBP expression, altered MVB dynamics and leukemic stress pathways (34). However, AML-specific circRNA-sorting mechanisms remain incompletely defined and warrant further investigation. Exosomes are synthesized from multivesicular bodies (MVBs). The formation of MVBs can occur through ESCRT-independent pathways and is influenced by lipid components such as plasma-membrane-derived lipids and ceramides (35). Additionally, MVB formation is regulated by syndecan heparan sulfate proteoglycans and the cytoplasmic adaptor protein syntenin. Rab GTPases and soluble N-ethylmaleimide-sensitive factor attachment protein receptors facilitate either the fusion of MVBs with lysosomes for degradation or their docking at the cell periphery for exosome secretion (36). Moreover, research indicates that the elevated expression of p53, pyruvate kinase M2 and tumor suppressor-activated pathway 6 enhances exosome secretion in tumor cells (37). Exosome release has also been shown to be influenced by changes in the microenvironment, including altered pH and the accumulation of intracellular Ca²⁺ ions (37). Exosomes are taken up by target cells through several mechanisms, such as endocytosis and direct membrane fusion (35). It has been demonstrated that exosome entry can be inhibited under certain conditions; for example, the internalization of exosomes by ovarian cancer cells via multiple endocytic pathways is significantly reduced at 4˚C (38). The mechanisms of exosome biogenesis are illustrated in Fig. 1.

The biogenesis of circRNAs occurs through the back-splicing of pre-mRNA, in which a downstream splice donor site (5'splice site) attaches to an upstream splice acceptor site (3'splice site), and through transcription mediated by RNA polymerase II. Various circRNAs can be synthesized by the alternative back-splicing of the same sequence (39). However, the exact mechanisms that determine circRNA formation have not yet been fully delineated. CircRNAs are classified into three main categories based on their structure and mode of circularization: Circular intronic RNAs (ciRNAs), exonic circRNAs (ecircRNAs), and exon-intron circRNAs (EIciRNAs). EcircRNAs consist of one or more exons and are sometimes generated through alternative splicing that are primarily located in the cytoplasm (40). Several models have been proposed to explain the biosynthesis of ecircRNAs, including RBP-mediated circularization, lariat-driven circularization and intron-pairing-driven circularization. In the lariat-driven model, partial RNA folding occurs during pre-mRNA transcription, leading to exon skipping as the RNA folds. These structural changes can produce lariat intermediates containing initially non-adjacent exons along with their introns. CircRNAs are generated after intron sequences are removed through splicing within the lariat structure (41). In the intron-pairing-driven model, reverse complementary sequences within introns on both sides of the pre-mRNA facilitate complementary base pairing, mediating circularization. RBPs acting as trans-acting activators or inhibitors also play key roles in regulating circRNA production. DHX9, an abundant nuclear RNA helicase and ALU-element-binding protein, functions as an antagonist of circRNA biosynthesis. The lLoss of DHX9 increases circRNA production by allowing the formation of RNA pairs flanking circularized exons (40). Additionally, several splicing factors can bind specific RNA motifs and regulate circRNA biogenesis. Quaking (QKI), a mesenchymal splicing factor, is crtiical for generating circRNAs during epithelial-mesenchymal transition. The RBP FUS regulates back-splicing reactions and thus controls circRNA formation in mouse embryonic-stem-cell-derived motor neurons. While numerous RBPs, such as QKI, DHX9, hnRNPA2B1 and FUS regulate circRNA circularization, AML-specific alterations in these RBPs remain poorly defined (42). Available transcriptomic datasets indicate that DHX9 is not downregulated in AML, but is often upregulated; however, to date, at least to the best of our knowledge, no study has demonstrated a direct link between DHX9 expression and increased cellular or exosomal circRNA production in leukemic cells (43). Similarly, although QKI is a major enhancer of back-splicing, recurrent QKI mutations have not been reported in AML, and changes in QKI expression across AML subtypes have not been shown to modify circRNA biogenesis. Moreover, although specific AML driver lesions [e.g., FLT3-ITD, NPM1 mutations, t(8;21)] have been associated with altered circRNA expression patterns, there is currently no evidence that these mutations directly disrupt intronic complementary sequences or intron-pairing mechanisms that determine circRNA formation (25,44). Thus, the influence of AML-associated RBPs and genomic lesions on circRNA circularization and exosomal loading remains an important but unresolved area for future investigation (34). EIciRNAs are primarily located in the nucleus and consist of exons and introns that are retained during splicing. These EIciRNAs play a crucial role in regulating the transcription of their parental genes (45). During the formation of ecircRNAs, the introns flanking the exons are typically spliced out; however, when these introns are retained, EIciRNAs are formed. Some lariat structures are produced during splicing from introns where most of them are rapidly degraded through debranching (40). A schematic illustration of the biogenesis of exosomal circRNAs is presented in Fig. 2.

Several studies have demonstrated that circRNAs function in AML by either promoting or inhibiting tumor progression, and modulating the biological behavior of malignant cells, including proliferation, invasion, migration, and metastasis (53,54). Exosomal circRNAs perform all of the aforementioned roles in cancer due to their ability to mediate intercellular communication. When exosomal circRNAs are taken up by target cells, they influence the physiological and pathological states of those cells (55). Numerous studies have examined the mechanisms through which exosomal circRNAs contribute to cancer progression and development. Exosomal circRNAs are circular RNAs enclosed within exosomes where small extracellular vesicles released by cells into bodily fluids such as saliva, blood and urine. These exosomal circRNAs participate in cell-cell communications, other biological processes and disease progression (56). Circ-ITCH has been shown to promote the progression of hepatocellular carcinoma by sponging miR-214. By sequestering miR-20b-5p, circ-ITCH reduces its inhibitory effect on the target gene PTEN, leading to activation of the PTEN/PI3K/AKT signaling pathway, which enhances cell proliferation and inhibits apoptosis (57). Exosomal circRNAs further promote cell proliferation and survival by modulating the expression of genes involved in cell-cycle progression and apoptosis. CircRNA_0014130, for instance, has been identified as an oncogenic driver in AML by sponging miR-139-5p/miR-141-3p and activating the β-catenin signaling pathway, thereby contributing to the development of AML (25). Similarly, circRNA_100290 has been shown to be transferred from AML cells to bone marrow stromal cells, where it upregulates CXCL16 expression. CXCL16, in turn, supports the progression and invasion of AML cells within the bone marrow microenvironment, promoting disease development (58). Exosomal circ_0004136 was among the first circRNAs implicated in AML: The knockdown of circ_0004136 was shown to reduce AML cell viability, migration, invasion and promoted apoptosis, whereas its overexpression was shown to enhance proliferation and invasive behavior in vitro (59). Mechanistic studies link circ_0004136 to the miR-570-3p/TSPAN3 axis. As circ_0004136 was detected in AML serum exosomes, it is considered a promising non-invasive biomarker and a potential therapeutic target (60). While earlier studies focused on intracellular circ_0009910(61), more recent research indicates that it is enriched in AML-derived exosomes: Exosomal circ_0009910 promotes the proliferation, reduces the apoptosis and regulates the cell cycle of AML cell lines (61). The functional mechanism involves the sponging of miR-5195-3p, leading to the upregulation of GRB10, which drives leukemic cell survival and proliferation. These data indicate that exosomal circ_0009910 may contribute to AML pathogenesis via intercellular signaling through EVs (51). Exosomal circRNAs affect the antitumor immune microenvironment and promote immune evasion and thus regulate immune response in AML (62). CircRNA_0000190 upregulates CD96 by sponging miR-142-3p, which suppresses proliferation and the cytotoxic activity of natural killer cells (61). Certain circRNAs activate pro-inflammatory signaling pathways or promote cytokine and chemokine secretion by immune cells or stomal cells. These inflammatory mediators provide a tumor-promoting environment that enhances AML cell survival and proliferation.

Stromal and AML cells communicate through exosomal circRNAs in the bone marrow microenvironment (63). This intercellular crosstalk influences disease progression, drug resistance and immune evasion crucial for the development of AML. Exosomes containing specific circRNAs that are secreted by AML cells can be transferred to neighboring stromal cells. For example, circRNA_100290 has been shown to be transferred from AML cells to bone marrow stromal cells. Once internalized, circRNA_100290 promotes the expression of CXCL16, which is involved in cell migration and invasion. Exosomal circRNAs adapt to the stromal-cell environment that influence their interactions with AML cells and modify the surrounding microenvironment (64). By regulating the expression or activation of specific genes or signaling pathways within stromal cells, circRNA contribute to shaping conditions that support leukemic growth (Fig. 3, top panel). In a 2023 study analyzing public AML circRNA microarray datasets (GSE94591 and GSE163386), circZBTB46 was among the most consistently upregulated circRNAs in patients with AML compared to healthy controls. Although not yet functionally characterized in vitro or in vivo, its overexpression suggests potential relevance for AML biology, warranting further studies (65).

In recent years, the diagnosis of cancer has been a major focus of scientific research worldwide. Various accessible methods for the diagnosis of cancers have continued to emerge (61,66), using samples collected from tears, urine (67), semen, ascites, synovial fluid, cerebrospinal fluid, amniotic fluid (68) and feces. These sample types are easier and less invasive to obtain, and once biomarkers are detected, biopsies can be performed to analyze tumor tissues. This highlights the advantages of using exosomes as potential biomarkers. Moreover, circRNAs, as part of exosomal cargo exhibit characteristics such as structural diversity, conservation, stability and specificity in expression and localization (69).

Exosomal circRNAs have emerged as promising biomarkers as they offer several advantages, such as abundance, stability and disease specificity (69). They are considered as strong diagnostic candidates due to their differential expression in diseased AML and healthy individuals. For example, the expression of circRNA_0014130 has been shown to be high in patients with AML compared with healthy controls (25). CircRNA_0014130 has been reported to exhibit better diagnostic accuracy, with more than 90% sensitivity and specificity (25). The identification of exosomal circRNAs may help in early detection of AML resulting in timely intervention and improved patient outcomes. CircRNA_0014130 has been detected in the plasma exosomes of AML patients at early stages of disease, suggesting it to be an early diagnostic biomarker (25). These results are promising; however, multi-center validation and larger cohorts are required to confirm reproducibility and clinical applicability (70). Exosomal circ_0006896 has been shown to be upregulated in AML, and promotes proliferation, chemoresistance and immune-evasive behavior in preclinical models (71). Recent reviews and meta-analyses have provided the increasing evidence that exosomal cargo, such as circRNAs, modulates immune checkpoints and that engineered exosomes are being developed to augment immune checkpoint inhibitor and adoptive-cell therapies (72).

Studies have indicated that exosomal circRNAs may be used to monitor disease development and treatment response in patients with AML. Variations in the expression levels of specific exosomal circRNAs throughout the course of therapy can provide key insight into therapeutic efficacy and disease status (73). Monitoring circRNA levels enables the assessment of treatment response and the detection of disease recurrence (51). The isolation and detection of exosomal circRNAs from peripheral blood samples offer a minimally invasive approach for AML diagnosis and monitoring (74). Certain exosomal circRNAs exhibit prognostic value as they are associated with survival outcomes in AML patients (75). For example, circRNA_0009910 correlates with poor prognosis as higher expression levels are linked to reduced overall survival and ehanced risk of relapse (76). Exosomal circRNAs may also help in determining the treatment response and clinical outcomes in patients undergoing treatment. Expression levels of certain exosomal circRNAs is associated with disease advancement and treatment resistance in AML patients providing information related to disease severity (25). Exosomal circRNAs can also be used to monitor minimal residual disease in patients with AML following therapy, which provides information related to disease recurrence and treatment response. Patients with hepatocellular carcinoma exhibiting complete remission following chemotherapy exhibit decreased levels of circ-ITCH, suggesting its role as a predictive biomarker of the treatment response (77).

It has been demonstrated that the expression levels of certain circRNAs before treatment may determine the response to targeted therapy thereby acting as a predictive biomarker for targeted therapies. For example, it has been shown that higher levels of Circ_0059706 indicate a poor response to tyrosine kinase inhibitor (TKI) therapy and a shorter progression-free survival in AML (78,79). Exosomal circRNAs exhibit diagnostic and prognostic potential across various types of cancer. For example, the study conducted by Zheng et al confirmed that exosomal circRNAs derived from the serum of patients with gastric cancer served as diagnostic markers for gastric cancer (80). Exosomal circ_0072083 expression was previously shown to be upregulated in patients with glioma compared to healthy controls (81). Similarly, the dysregulated expression of exosomal hsa_circ_0064428 in hepatocellular carcinoma is associated with clinical characteristics, such as WBC count and bone marrow blast percentage (82). Elevated levels of circPAN3 have been shown to be associated with poor survival rates of patients with AML, exhibiting its prognostic significance (83). Likewise, Yan and Yan (84) reported that the overexpression of exosomal circRNA 100199 indicates severe clinical outcomes, including resistance to chemotherapy and a shorter survival duration. It has also been shown that the inhibition of exosomal circ_0072083 suppresses AML cell proliferation and induces apoptosis, highlighting its potential as a therapeutic target (29). Thus, targeting exosomal circRNAs holds considerable therapeutic promise in the management of AML (Fig. 3, bottom right panel).

CircRNAs play critical roles in cancer development, including breast cancer, gastric cancer and AML. Exosomes are used as delivery vehicles for circRNAs, which exhibits a promising therapeutic strategy for the treatment of AML. Exosomes are generated through the endosomal pathway and contain biomolecules, including lipids, proteins and nucleic acids (DNA and RNA). Specific RNA motifs help in the loading of circRNAs into exosomes (73). Using motif prediction software, circRNAs with motif 5'-GMWGVWGRAG-3' were predicted to be enriched in exosomes, observed by circRNA-seq data from human blood exosomes (85). CircRNAs contribute to the pathogenesis of AML by regulating key signaling pathways and gene expression. Delivering specific circRNAs via exosomes is expected to regulate these pathways and reduce leukemic cell proliferation and survival. Exosome isolation can be achieved through several techniques, such as ultracentrifugation, size-exclusion chromatography and precipitation. Exosomes are characterized on the basis of morphology, size and protein markers, including CD9, CD63 and CD81 along with their molecular contents (86). Preclinical studies with AML animal models were used to evaluate the efficacy of exosomal circRNA delivery. These studies involve systemic administration of exosomes loaded with specific circRNAs, followed by monitoring leukemic burden, survival outcomes, and potential adverse effects. Although the therapeutic application of exosomal circRNA delivery in AML is promising, several challenges remain, including the scalability of exosome production, optimization of loading efficiency, safety issues associated with systemic administration and the need to harness specific properties of exosomes to deliver circRNAs effectively Addressing these limitations may lead to the development of AML therapies capable of overcoming current treatment barriers and improving patient outcomes (87).

CircRNAs can be loaded into exosomes either by transfecting donor cells (e.g., mesenchymal stem cells) with circRNA-expressing plasmids or through direct loading methods such as electroporation. Transfection-based approaches allow the efficient loading of circRNAs into exosomes during their natural biogenesis within donor cells. Direct loading methods offer simplicity and flexibility, but may exhibit lower loading efficiency and can potentially alter exosomal properties. Overall, different loading techniques vary in efficiency and may have distinct downstream effects. Electroporation and donor-cell transfection are the most commonly used approaches: Optimized electroporation protocols report apparent loading efficiencies on the order of 20-35% (compared with passive incubation); however, electroporation can produce RNA precipitation artifacts and may require rigorous controls (88). Donor-cell engineering (expressing circRNA constructs or RNA-binding domain fusions) often yields higher functional loading and preserves exosome integrity (89). Once loaded with circRNAs, exosomes can be administered systemically or locally to patients with AML (90). After any loading method, EV quality control (particle count/size, TEM morphology and tetraspanin markers, such as CD9/CD63) and immune-safety testing (PBMC cytokine release, in vivo cytokine panels) are essential as cargo and surface modifications can alter immunogenicity (91). Exosomes possess inherent tropism toward tumor cells, including AML cells, due to the presence of specific surface proteins and ligands. Additionally, the modification of exosomal surface proteins or incorporation of targeting peptides can further enhance their specificity for AML cells. Once internalized, exosomes release their cargo including circRNAs into the cytoplasm of AML cells. These circRNAs can then exert regulatory effects; for example, circRNAs functioning as miRNA sponges can sequester oncogenic miRNAs involved in the progression of AML, while other circRNAs may modulate the expression of key AML-associated genes or signaling pathways (92). For targeted delivery to AML blasts, surface functionalization approaches directed at AML antigens (e.g., CD33) have been reported and are promising; however, preclinical data testing exosomal circRNA delivery together with standard AML drugs (cytarabine, FLT3 inhibitors) are currently limited, and combination experiments are an important future direction (93).

Exosomes facilitate communications between AML cells and various other cells within the microenvironment, including endothelial cells, immune cells and bone marrow stromal cells (26,92). For instance, AML-derived exosomes enriched with miR-155 have been shown to promote angiogenesis by targeting endothelial cells (94). These exosomes can enhance angiogenesis, a crucial process for tumor growth and metastasis by transporting pro-angiogenic factors, such as VEGF and FGF-2, which stimulate endothelial cell proliferation and blood vessel formation. Exosomes also play a key role in immune evasion by suppressing anti-tumor immune responses. AML-derived exosomes can carry immunosuppressive molecules, including TGF-β and PD-L1, which inhibit cytotoxic T-cell activity and thereby promote immune escape (95).

These exosomes can enhance angiogenesis, a process crucial for tumor growth and metastasis by transporting pro-angiogenic factors, such as VEGF and FGF-2, which stimulate endothelial cell proliferation and blood vessel formation. Exosomes aid in the suppression of antitumor immune responses and thus play a crucial role in immune evasion. AML-derived exosomes can carry immunosuppressive molecules, such as TGF-β and PD-L1 inhibiting cytotoxic T-cell activity and thereby promoting immune escape (96). Among these, exosomal circRNA_0056616 was shown to promote AML cell migration and proliferation by regulating the miR-625-5p/HOXA13 axis (97). The prevalence and clinical significance of these individual axes in metastatic AML remain unquantified. The majority of supporting data are derived from in vitro experiments, small discovery datasets, or from non-AML tumor studies; no cohort-level studies that report the proportion of patients with metastatic AML with the dysregulation of circRNA_0056616 axis have been identified (98). Metastatic AML typically signifies an advanced disease stage and a higher leukemic burden, both of which contribute to treatment resistance and reduced survival rates (99). The extramedullary manifestations of AML can also pose diagnostic challenges as their symptoms and clinical presentations may resemble those of other conditions or primary malignancies affecting the same organs (100). The accurate diagnosis of AML metastasis requires comprehensive evaluation, including imaging studies (e.g., CT scans, MRI), bone marrow biopsy, and, in some cases, tissue biopsy of the involved extramedullary site. Metastatic AML often necessitates tailored treatment approaches based on the extent and location of disease involvement. Apart from standard chemotherapy, patients with central nervous system metastasis may require intrathecal chemotherapy or cranial irradiation to control AML. Similarly, extramedullary AML lesions may require treatments, such as radiation therapy or surgical removal in combination with systemic therapy (101). AML metastasis increase the risk of post-treatment relapse, including metastasis to the central nervous system. During remission in the bone marrow, residual leukemic cells present in extramedullary sites may serve as a reservoir for relapse. Therefore, to reduce the risk of relapse strategies are to be designed to remove of prevent extramedullary disease (102). Patients with metastatic AML require close monitoring and surveillance to detect disease recurrence or progression, especially in extramedullary locations. Regular imaging and clinical assessments are critical for early relapse detection and timely intervention. Furthermore, the presence of metastatic disease may influence both the frequency and intensity of post-remission monitoring strategies. Effective symptom management and supportive care targeted toward alleviating pain, addressing neurological deficits and managing organ dysfunction remain critical components of comprehensive AML care in the metastatic setting (100).

The advantage of exosome-based therapy is their ability to mediate intercellular communication and deliver therapeutic agents to specific targets. Due to the small size of exosomes, their content is protected and thus they act as efficient therapeutic markers (103). In several studies, exosomes have shown improved therapeutic benefits in cancer treatment compared with the direct use of chemotherapeutic agents (104). Exosomes also offer potential solutions to limitations associated with current drug delivery approaches, particularly those involving programmable RNAs, which often suffer from low uptake efficiency and substantial cytotoxicity, rendering them unsuitable for clinical application (105). To address these challenges, researchers have demonstrated and validated a novel strategy for generating large quantities of RBC-derived EVs intended for RNA drug delivery (106). Furthermore, gold nanoparticle-based targeting systems have been used to engineer specific exosome types capable of precise and selective elimination of cancer cells (107-109). While exosomes containing endogenous circRNAs may contribute to cancer progression, exosomes loaded with therapeutic circRNAs or custom-engineered siRNAs against pathogenic circRNAs hold considerable promise for inhibiting tumor growth. Delivering circRNAs that suppress cancer hallmarks and promote apoptosis may further enhance the development of precise, effective therapeutic strategies (110).

Exosomal circRNAs affect treatment response by regulating processes such as drug resistance, apoptosis and immune evasion and thus have emerged as important regulators of therapeutic responses in AML. CircRNA_000911 has been implicated in chemoresistance in by sponging miR-449a and upregulating HDAC1, which reduces apoptosis and promotes cell survival (111). CircRNAs may also regulate drug efflux pump encoding genes that leads to alteration in the intracellular concentration of chemotherapeutic agents and thus contributes to resistance. Exosomal circRNAs can also regulate apoptosis-related pathways affecting sensitivity of AML cells to treatment-induced cell death. Moreover, exosomal circRNAs play critical roles in the response to targeted therapies in AML, including genetically targeted drugs and TKIs. By regulating the expression of pro- or anti-apoptotic genes, circRNAs can shift the balance between cell survival and apoptosis (68). For instance, circRNA_0004015 promotes resistance to TKI therapy by upregulating anti-apoptotic proteins and suppressing chemotherapy-induced apoptosis (112). The overexpression of such circRNAs can enhance immune evasion and tumor progression by inhibiting T-cell activation and cytokine secretion (76).

Exosomal circRNAs can regulate molecular pathways that affect drug metabolism, cell survival and apoptosis thereby influencing the efficacy of targeted therapies (29). They also modulate the expression of genes encoding drug targets or signaling molecules involved in precision therapy (75). Additionally, exosomal circRNAs can alter drug metabolism and pharmacokinetics by affecting intracellular uptake and bioavailability. For example, circPAN3 has been shown to be associated with development of AML drug resistance doxorubicin through regulating autophagy (83). Exosomal circRNAs may serve as predictive biomarkers of treatment resistance and disease relapse. Reduced intracellular drug accumulation and decreased sensitivity to chemotherapeutic agents (25), combined with treatment-induced changes in circRNA expression, can indicate the emergence of resistant clones and the need for alternative therapeutic strategies. Integrating exosomal circRNA profiles with clinical parameters, such as cytogenetics, molecular mutations and patient demographics significantly enhances their predictive value. Multivariate analyses incorporating circRNA levels and conventional prognostic factors have yielded improved risk stratification and more accurate assessments of the treatment response (113). For instance, the increased expression of circ_100053 enhances CML cell resistance to cytarabine by promoting proliferation and inhibiting apoptosis (48).

Exosomal circRNAs may also regulate gene expression programs involved in AML progression and therapy response (75). Certain circRNAs, such as circ-foxo3 and circ-0004277, exhibit an altered expression in de novo AML and are associated with disease characteristics (51). Their stability renders exosomal circRNAs attractive candidates for reliable diagnostic biomarkers that support early detection and disease monitoring. Identifying exosomal circRNAs involved in the pathogenesis of AML may also reveal therapeutic opportunities aimed at targeting their expression or function. Furthermore, exosomal circRNAs can exert immunomodulatory effects, influencing interactions between AML cells and the immune system. Understanding these associations may have key implications for AML immunotherapy strategies (116) (Fig. 3, bottom left panel).