The origin of Mtb EVs is thought to be closely linked to the cytoplasmic membrane. Their enrichment in lipid components unique to the cytoplasmic membrane - including LAM, lipomannan (LM), phenolic glycolipids (PGL), and phosphatidylinositol mannosides (PIMs) - suggests that their formation initiates at the plasma membrane^[11]. Electron microscopic observations further reveal that Mtb EVs exhibit substantial heterogeneity in diameter (approximately 60-300 nm) and electron density^[11]. The structure of some Mtb EVs appears to be in the process of being released from the cell surface, implying potential interactions with the cell wall during their biogenesis and the possibility of carrying envelope components of varying layers, thereby contributing to the diversity in composition and morphology^[21]. However, controversies persist regarding the origin of Mtb EVs, necessitating additional research to confirm their source. Meanwhile, EVs release is a universal trait of the Mycobacterium genus, not restricted to specific strains or pathogenic variants^[11]. Nevertheless, EVs derived from pathogenic mycobacteria display striking specificity in lipoprotein composition. For instance, lipoproteins account for 20% of total proteins in bacillus Calmette-Guérin (BCG) EVs and 10% in Mtb H37Rv EVs - proportions far higher than those in whole-cell lysates (1.4% and 2.6%)^[11]. In contrast, the absence of lipoproteins in Mycobacterium smegmatis EVs suggests that the active sorting and enrichment of these proteins distinguishes the vesicles of pathogenic mycobacteria from their non-pathogenic counterparts^[11]. The biogenesis and regulation of Mtb EVs and HC EVs are summarized in Figure 1.

Proteomic analyses have uncovered another characteristic of the composition of Mtb EVs. A pioneering study by Prados-Rosales et al. was the first to systematically identify the protein composition of Mtb H37Rv EVs, identifying 48 proteins and highlighting the significant enrichment of lipoproteins as a core feature^[11]. With advances in mass spectrometry technology, Lee et al. substantially expanded the number of identified Mtb EVs proteins to 287, encompassing diverse functional categories such as metabolism, signal transduction, and lipid metabolism^[22]. Notably, these Mtb EVs proteins show substantial overlap with naturally occurring antigen complexes widely used in TB research (e.g., Ag60)^[23], providing a solid molecular foundation for understanding the immunological functions of Mtb EVs. Beyond proteins, Mtb EVs also contain a variety of structural and immunomodulatory lipids. Lipids in Mtb EVs are predominantly composed of polar lipids, including phosphatidylinositol, acylated phosphatidylinositol mannosides, cardiolipin (CL), and phosphatidylethanolamine (PE), and carry the key immunomodulatory molecule LAM^[11]. These lipid components not only constitute the structure of Mtb EVs but also participate in host immune recognition and regulation. Additionally, bacterial genomic DNA has been identified within Mtb EVs^[11]. Although Mtb DNA in the host cytoplasm can initiate innate immune responses, the specific immunological roles and underlying mechanisms of EVs-delivered DNA remain to be further explored. No studies have systematically identified whether Mtb EVs carry RNA molecules; future research clarifying their RNA composition is expected to reveal new layers of bacterial gene regulation and host interactions mediated by Mtb EVs. Simultaneously, it remains unknown whether the cargo profile of Mtb EVs released by Mtb during different infection stages undergoes dynamic changes. Systematic analysis of the cargo of Mtb EVs throughout the entire infection process in future studies may uncover the refined strategies by which Mtb adapts to host immune pressure from a novel “temporal regulation” perspective.

The biogenesis of Mtb EVs is governed by an intricate, multi-factorial regulatory network, whose complexity has been unraveled from diverse research perspectives. A core component of this network is VirR: its deletion enables Mtb to induce heightened immune responses in vivo, resulting in reduced bacterial burden^[21]. Subsequent studies revealed that loss of VirR causes profound abnormalities in cell wall structure, increasing its permeability and ultimately driving the massive release of Mtb EVs^[24]. Independent VirR-independent pathways have also been reported - for instance, in a mutant strain deficient in the pstA1 gene, over 60 genes sustain RegX3 activation to significantly promote Mtb EVs release^[25]. Future research could further identify key effectors driving Mtb EVs secretion from these RegX3-regulated genes. Collectively, these findings indicate that Mtb EVs biogenesis is controlled by specific signaling pathways and responds to distinct intracellular states. Notably, Mtb may restrict excessive Mtb EVs production via factors such as VirR and pstA1, thereby avoiding overexposure of antigens to the host immune system - a potential strategy for immune evasion within the host. In-depth investigations into Mtb EVs regulatory factors in the future will not only uncover the fundamental principles of Mtb EVs biogenesis but also hold promise for disrupting the pathogen’s precise control over Mtb EVs release through interventions targeting key nodes such as VirR or RegX3. Such interventions could enhance host recognition and clearance of the pathogen, offering novel therapeutic targets for anti-TB treatments. Meanwhile, Mtb EVs release represents an active strategy for the pathogen to cope with stress. During infection, the host initiates a “nutritional immunity” strategy to limit Mtb’s iron acquisition, creating an iron-restricted environment critical for Mtb survival^[26]. Under this stress, Mtb has evolved multiple countermechanisms, with the role of Mtb EVs in this context increasingly being elucidated. A study by Prados-Rosales et al. observed that Mtb releases significantly higher levels of EVs under iron-depleted conditions compared to iron-replete environments^[27]. Iron limitation robustly promotes Mtb EVs secretion, with regulatory mechanisms involving factors such as Rv3371 and the dynamin-like proteins IniA/IniC. Studies have shown that under iron restriction, Rv3371 expression is upregulated, and it promotes Mtb EVs release by altering cell wall lipid composition^[28]. Similarly, under iron deprivation or isoniazid stress, the iniBAC operon is upregulated, and its encoded dynamin-like IniA/IniC complex directly mediates membrane fission and Mtb EVs release in a GTP-dependent manner^[29]. It is important to recognize that the regulatory pathways described above - particularly the roles of Rv3371 and the IniA/IniC complex in orchestrating stress-induced vesiculogenesis - have been delineated almost exclusively through in vitro experimentation. These culture-based studies have yielded valuable mechanistic insights into the molecular machinery governing Mtb EVs biogenesis. However, a significant knowledge gap persists: whether these same regulatory cascades are operative within the complex and heterogeneous microenvironment of infected host tissues - where Mtb encounters dynamic gradients of iron, oxygen, and other limiting nutrients - has yet to be demonstrated. Direct in vivo evidence confirming that these spatiotemporal dynamics operate inside granulomas, necrotic lesions, or other tissue niches remains conspicuously absent, underscoring the need for future investigations employing physiologically relevant infection models.

Release of Mtb EVs is not a random cellular leakage or evolutionary relic but an active biological program tightly regulated by sophisticated genetic and biochemical networks, serving the pathogen’s survival needs across diverse physiological and stress conditions. The focus of future research should not only be on identifying specific effector molecules and their functions within candidate gene clusters downstream of RegX3 but also on systematically deciphering the topological structure of the global regulatory network connecting key nodes - including VirR, RegX3, iron restriction responses, and cell wall stress sensing - by integrating multi-omics data and quantitative biological approaches from a systems biology and dynamic regulation perspective. Building on this, a quantitative investigation into the dynamic reprogramming of this regulatory network across different infection models and stages will be of great importance for clarifying how environmental signals are sensed, integrated, and ultimately translated into precise regulation of Mtb EVs biogenesis pathways. This will reveal the molecular logic underlying the disruption and re-establishment of the balance between “restricted release” and “promoted release” states. Targeted interventions to modulate this dynamic balance - whether by enhancing inhibitory nodes to impair the pathogen’s environmental adaptability or by relieving inhibition to overexpose pathogen antigens - may disrupt the immune homeostasis established by Mtb within the host, thereby providing new therapeutic breakthroughs to dismantle its persistent infection mechanisms.

As explicitly stated in a key position paper by the International Society for Extracellular Vesicles (ISEV) in 2014, significant terminology confusion existed during the early stages of the field’s development. Terms such as “microparticle”, “ectosome”, and “exosome” were frequently used interchangeably in the literature without uniform standards^[30]. Notably, the term “exosomes” - owing to its rapid popularity - was often employed indiscriminately as an umbrella term for all EVs subtypes, obscuring the potential biogenetic and functional differences between distinct HC EVs subpopulations^[30]. Therefore, HC EV subtypes have been uniformly classified and named - including exosomes, microvesicles, and apoptotic bodies - based on criteria such as their biogenetic pathways, size, density, and protein composition^[5].

As a typical intracellular pathogen, Mtb profoundly alters the physiological state of host cells such as macrophages following infection^[31]. Infected cells release a repertoire of HC EVs, which act as complex messengers in host-pathogen interactions, with cargo exhibiting dual host and pathogen characteristics. These vesicles not only carry host proteins, nucleic acids, and lipids that reflect cellular activation or functional status but also frequently encapsulate Mtb antigens and lipids, thereby playing multifaceted roles in mediating immune regulation, promoting bacterial survival, and facilitating long-distance dissemination^[32]. Throughout this review, we use the term “HC EVs” as a collective designation encompassing all host-derived vesicle subtypes - exosomes, microvesicles, and apoptotic bodies. However, it is important to note that the literature we cite varies considerably in whether individual studies distinguish among these subtypes. When the original study provides clear evidence for a particular subtype (e.g., through specific isolation methods, marker analysis, or electron microscopy), we adopt the corresponding term in our functional descriptions. When the original work examines total EVs or mixed populations without subtype resolution, we retain “HC EVs” and indicate the composition or methodological limitations accordingly, so that readers can clearly differentiate between subtype-specific and population-level findings.

Among the diverse subtypes of HC EVs, exosomes have emerged as the most extensively studied in TB research, attributed to their unique biogenetic pathway, specific cargo sorting mechanisms, and stability in bodily fluids. Exosomes are nanoscale EVs with a diameter of 30-150 nm, originating from multivesicular bodies (MVB) within the endocytic pathway^[33]. Their membrane structure is enriched in tetraspanins (CD9, CD63, CD81), sphingolipids, cholesterol, and unsaturated lipids, and they carry biogenesis-related proteins such as TSG101 and ALIX, as well as immune-associated components including major histocompatibility complex (MHC) molecules^[34,35]. During Mtb infection, exosomes released by host cells have been found to be significantly enriched with pathogen-specific cargo, such as bacterial-derived lipoproteins, virulence factors, and host stress-response proteins [Table 1]^{[11,22,27,32,36-44]}.

Exosome biogenesis is a highly programmed process involving early endosome formation, MVB maturation, and final secretion. First, the plasma membrane of the cell undergoes initial invagination to form early endosomes (EE)^[33]. As EE mature into late endosomes (LE) and further develop into MVB, their membranes bud inward to generate intraluminal vesicles (ILV), with the majority of exosome cargo captured and encapsulated into ILV during this “invagination” process^[45]. For protein cargo - including soluble Mtb-derived proteins - sorting and loading into exosomes are primarily mediated by the coordinated action of ESCRT-dependent and ESCRT-independent pathways^[46]. Smith et al. have shown that Mtb-derived proteins loaded into exosomes are mainly soluble proteins actively secreted or released by Mtb into host cells^[41]. Regardless of how Mtb soluble proteins enter the cell, they can be directed to a ubiquitination-dependent pathway; this ubiquitination tag is subsequently recognized by the ESCRT system, guiding their entry into ILV^[41]. Meanwhile, Mtb infection triggers rapid phosphorylation of AKT, activating both ESCRT-dependent and ESCRT-independent pathways. However, whether this represents active hijacking by Mtb or a host cell response to Mtb infection remains to be further elucidated^[47]. RNA cargo enters ILV primarily through active sorting mechanisms mediated by RNA-binding proteins recognizing specific sequence motifs^[45]. For Mtb RNA in exosomes released by infected host cells, Cheng and Schorey reported that the Mtb SecA2 protein secretion system is crucial for delivering Mtb RNA into host cell-derived exosomes^[48]. Mtb infection can alter the composition of both host and bacterial RNA in host cell exosomes^[49]. Nevertheless, the underlying logic for the upregulation and downregulation of different RNAs in exosomes released by Mtb-infected host cells remains unclear. The fate of MVB determines exosome release: some fuse with lysosomes for degradation, while others are transported to the plasma membrane under the regulation of Rab GTPases and release exosomes through SNARE protein-mediated membrane fusion^[50]. Exosome release is an actively regulated process dependent on Rab GTPase-directed transport, SNARE protein-mediated membrane fusion, and ESCRT complex assistance, with precise control by multiple intracellular and extracellular signals^[45]. During Mtb infection, the level of VAMP5 in exosomes from TB patients is significantly increased and decreases following treatment, indicating an association between VAMP5 and disease activity^[51]. VAMP5 belongs to the SNARE protein family and acts as a regulator of exosome release^[52]. Arya hypothesized that after Mtb infects macrophages, VAMP5 expression on MVB membranes is regulated; it forms specific SNARE complexes by binding to STX1 or STX4 and SNAP47 on the plasma membrane, driving MVB-plasma membrane fusion and thereby promoting exosome secretion^[51,53].

Unlike exosomes, the sorting of their luminal cargo relies more on lipid anchoring of proteins (e.g., myristoylation or palmitoylation) rather than the canonical ubiquitination pathway^[54]. Anchored proteins can spontaneously accumulate in specific microdomains of the plasma membrane and further selectively encapsulate certain cytoplasmic proteins, RNAs, and other cargo into the ectosome lumen by forming higher-order polymers or protein networks^[54]. They carry bioactive molecules such as proteins and nucleic acids, participating in various physiological and pathological processes including immune regulation, cell differentiation, and infection responses^[55]. Due to their cargo sorting mechanism that favors membrane-associated proteins, microvesicles may serve as efficient carriers for specific membrane proteins (e.g., co-stimulatory molecules, adhesion molecules), a conjecture supported by studies by Ramachandra et al.^[40]. As terminal products of apoptosis, apoptotic bodies function as core “antigen transporters” in TB immunity. Phosphatidylserine exposed on their surface acts as an “eat-me” signal, which can be recognized and phagocytosed by macrophages, enabling gentle antigen delivery without triggering significant inflammation^[44,56]. A study has shown that apoptotic bodies produced by macrophages pre-treated with interferon-gamma (IFN-γ) after apoptosis can effectively enhance the antigen-presenting process^[57]. However, how Mtb interferes with their formation and cargo sorting, as well as their role in the reactivation of latent infections, remains to be systematically elucidated.

The key characteristics of each HC EVs subtype - including cellular origins, representative cargo molecules, and major immunological functions - are summarized in Table 2 for quick reference^{[15,17,32,40,43,44,48,49,57-68]}. Compared with the rapid development of exosome research, the exploration of microvesicles and apoptotic bodies in translational applications for TB is significantly lagging. How to accurately decipher their specific cargo composition and immunological fate - i.e., whether they cooperate with the host to clear pathogens or are exploited by pathogens to promote immune evasion? Answering these questions is a prerequisite for achieving precise regulation of their functions and comprehensively understanding HC EVs-mediated host-pathogen interactions.

The reliability of EVs research largely depends on efficient isolation and accurate characterization methods. Current isolation techniques face multiple challenges: EVs in biological samples often coexist with impurities such as protein aggregates and lipoproteins, making co-precipitation difficult to avoid^[5]; the high heterogeneity of EVs themselves in terms of particle size and surface markers further increases purification complexity^[69]. Additionally, the impact of different isolation strategies on EVs integrity, recovery rate, and functional properties reduces the comparability of research results. However, there is currently no universal isolation method applicable to all samples, and appropriate strategies must be selected based on EVs sources and research objectives^[34].

For the isolation of Mtb EVs, Rao et al. proposed that their purification primarily involves three steps: induction (culture in minimal medium), collection (centrifugation, filtration, and concentration), and purification (obtaining high-purity Mtb EVs via ultracentrifugation combined with OptiPrep density gradient centrifugation)^[70]. Studies have shown that different methods exert significant effects on the yield, composition, and function of Mtb EVs. For instance, Wang et al. demonstrated that commercial kit-based methods yield higher quantities of Mtb EVs, while the EXODUS system produces Mtb EVs with more diverse components^[71]; Ryan et al. evaluated four approaches - ultracentrifugation, density gradient ultracentrifugation, qEV size exclusion chromatography, and Capto™Core size exclusion chromatography - among which the combination of qEV and OptiPrep density gradient centrifugation exhibited superior performance in terms of purity^[72]. They also noted that although Mtb EVs isolated by different methods are similar in morphology and size, significant differences exist in their protein composition, yield, and reproducibility^[72]. Notably, multiple techniques are often combined in practical research to improve the quality of Mtb EVs isolation. For the isolation of specific subpopulations within HC EVs, such as exosomes, more refined strategies are required. Differential centrifugation combined with ultracentrifugation is currently the most commonly used method. Due to differences in size and sedimentation coefficients among exosomes, microvesicles, and apoptotic bodies, exosomes are typically enriched at a centrifugal force of 100,000 × g, while microvesicles and apoptotic bodies can be collected at lower centrifugal forces^[40,44]. Furthermore, regarding the accurate separation of Mtb EVs and exosomes, Athman successfully isolated the two in distinct fractions using sucrose density gradient centrifugation^[58].

However, achieving higher standards of EVs isolation through advanced technologies remains a current technical bottleneck that urgently requires further exploration. The precise isolation of various EV subpopulations is not only a sophisticated technical operation but also a critical prerequisite for in-depth functional research and the advancement of clinical applications. In accordance with the Minimal Information for Studies of Extracellular Vesicles 2023 (MISEV2023) guidelines issued by ISEV, we have compiled Table 3, summarizing the principles, advantages, disadvantages, and applicable scenarios of commonly used isolation techniques for Mtb EVs and HC EVs in TB research.

Reliable EVs research relies on comprehensive and accurate characterization of isolated products, with strategies tailored to the biological origin of EVs and specific scientific questions. Systematic and precise characterization of Mtb EVs and HC EVs serves as a critical foundation for subsequent mechanistic exploration and translational research. The MISEV2023 guidelines propose a multi-layered characterization framework applicable to EVs from diverse species and cellular sources, but its practical implementation requires adjustments based on the biological characteristics of specific EVs subtypes^[73].

Characterization of Mtb EVs should focus on their unique cell wall components, such as the identification of LAM and mycolic acids via specific antibodies (e.g., Western Blot) or lipidomic analysis, alongside confirmation of their lipid bilayer structure using electron microscopy^[73]. Additionally, density gradient centrifugation is recommended to reduce interference from co-isolated protein aggregates, with detailed documentation of bacterial culture conditions emphasized - given their significant impact on EVs composition and function^[73]. For the characterization of HC EVs, multi-layered, complementary techniques are required to verify their identity, purity, and cellular origin. First, particle concentration and size distribution should be assessed using techniques such as nanoparticle tracking analysis (NTA), while transmission electron microscopy or cryo-electron microscopy can visualize their typical morphology. In terms of biochemical characteristics, the marker framework proposed by Minimal Information for Studies of Extracellular Vesicles 2018 (MISEV2018) should be adopted: detecting at least one transmembrane or glycosylphosphatidylinositol (GPI)-anchored protein (e.g., CD9, CD63, CD81) and one cytoplasmic protein (e.g., TSG101, Alix) as positive EVs markers. Simultaneously, verification of negative or significantly low expression of one intracellular organelle marker (e.g., Calnexin, Grp94, or cytochrome c) is necessary to evaluate sample purity and exclude contamination by cellular debris^[73]. Ideally, the proteomic profile of EVs should be compared with that of parental cell lysates, with “blank medium” processed through the same isolation workflow set as a negative control. Only through the integration of the aforementioned physical and biochemical evidence can the isolated particles be confirmed as high-purity HC EVs of host cell origin, providing a reliable basis for subsequent functional studies and clinical applications.

Research on these Mtb EVs and HC EVs - whether the standardization of isolation and characterization techniques or the in-depth dissection of their biogenesis and functions - serves as a critical cornerstone for fully deciphering the pathogenesis of TB and developing novel diagnostic and therapeutic strategies.

As important immunomodulatory factors, Mtb EVs exert a broad range of effects - they not only influence infected cells but also regulate the functions of distant uninfected immune cells. As will be further discussed in Section “Discussion: allies or foes? Re-evaluating the immunological functions of Mtb EVs and HC EVs”, the net immunological outcome of Mtb EVs exposure is highly context-dependent, varying with ligand dose, cellular target, and infection stage; the immunosuppressive effects described below should therefore be interpreted as one facet of a dynamic and tunable regulatory system, rather than as an intrinsic and invariant property of Mtb EVs. Studies have demonstrated that Mtb EVs can be internalized by macrophages and dendritic cells (DCs), thereby exerting dual regulatory roles: on one hand, they promote DC maturation and antigen presentation through a Toll-like receptor 2 (TLR2)-dependent pathway, characterized by the upregulated expression of MHC-I, MHC-II, and the costimulatory molecule CD86, which in turn induces T-cell responses. This process is associated with pathogen-associated molecular patterns (PAMPs) that act as potent TLR2 agonists^[15]; on the other hand, Mtb EVs can significantly inhibit IFN-γ-induced MHC-II expression in macrophages^[15]. At first glance, these two observations appear contradictory, as both are mediated via TLR2 signaling. However, as we will systematically dissect in Section “Discussion: allies or foes? Re-evaluating the immunological functions of Mtb EVs and HC EVs”, this apparent paradox can be resolved by considering that the functional outcome of TLR2 engagement depends on ligand dose, duration of exposure, and the cellular context - parameters that vary substantially between the DC maturation assay (short-term, low-dose stimulation) and the macrophage MHC-II suppression model (sustained, high-dose exposure). This effect is likely closely related to lipoproteins such as LprG and LpqH inhibiting CIITA transcription via the TLR2/MYD88 signaling pathway^[15,36,58]. As TLR2 agonists, these lipoproteins initiate a negative feedback regulatory loop through sustained TLR2/MYD88 signaling, with the core consequence of suppressing the transcription of CIITA - the “master regulator” of MHC-II molecule expression^[75]. Early studies revealed that LAM, a key virulence factor, can directly act on CD4⁺ T cells and assist Mtb in evading host clearance through multiple potential mechanisms, such as interfering with the T-cell receptor (TCR) signaling pathway, inhibiting CD4⁺ T-cell activation, and even inducing T-cell anergy^[76]. However, how LAM is delivered to T cells remained an unresolved question until Athman et al. proposed, based on their research findings, that Mtb EVs serve as critical carriers responsible for transporting inhibitory components such as LAM^[58]. Lipoglycans including LAM carried by Mtb EVs can be directly delivered to CD4⁺ T cells, inhibiting interleukin-2 (IL-2) secretion and cell proliferation by interfering with the phosphorylation of key molecules (e.g., Lck, ZAP-70, and LAT) in the TCR signaling pathway^[58]. Notably, the inhibitory effect of Mtb EVs is significantly stronger than that of purified LAM at the same dose^[58]. Given that Mtb EVs also contain other lipoglycans (e.g., LM and PIMs) that have been proven to possess independent inhibitory activities, their potent suppressive effect is likely derived from the synergistic effect of these lipoglycans^[58]. Upon delivery of LAM to T cells, it triggers the upregulation of GRAIL protein^[77]. GRAIL is an E3 ubiquitin ligase that negatively regulates IL-2 transcription, promoting T cells to enter a state of anergy^[77]. Additionally, Mtb EVs may act as carriers to deliver LAM to different compartments to facilitate immune escape, such as phagosomes, to inhibit phagosome maturation^[78].

By efficiently delivering their cargo of various lipoproteins and lipoglycans to host immune cells, Mtb EVs play a crucial role in inhibiting macrophage antigen presentation and inducing T-cell exhaustion, providing a new perspective for understanding how Mtb evades host immune clearance. However, it must be acknowledged that although existing studies have initially revealed the potential pathways through which Mtb EVs regulate host immune responses, their underlying molecular mechanisms remain far from fully understood, and many links urgently require in-depth elaboration. As a complex molecular ensemble, the immunosuppressive effects of Mtb EVs are likely not the result of independent actions of a single component but rather stem from the synergy and additive effects among different components. Their total amount, proportion, and spatiotemporal distribution may collectively determine the final outcome of the immune response. As suggested by Athman et al., multiple lipoglycans carried by Mtb EVs may collectively contribute to the inhibition of T-cell activation, potentially producing a stronger suppressive effect than individual components^[58]. The specific mode of this “synergistic effect” remains to be clarified. In-depth elucidation of the mechanisms underlying Mtb EVs-mediated immunosuppression is indispensable for comprehensively revealing the pathophysiological processes of TB and explaining the core links of its chronic infection and immune escape.

One of the key strategies for Mtb to achieve immune escape is to interfere with multiple links of the host immune system by precisely regulating the composition and function of HC EVs. As key messengers for intercellular communication, HC EVs can accurately deliver a variety of bioactive molecules such as proteins, lipids, and nucleic acids to target cells, thereby reshaping the immune microenvironment at multiple molecular levels. In the complex immune context of TB infection, HC EVs, especially exosomes and microvesicles, collectively form an elaborate immune regulatory network that assists the pathogen in achieving immune escape.

First, in terms of protein composition, Mtb can alter the expression of various antibacterial proteins in host-derived exosomes. For instance, the expression levels of proteins such as CAMP, FCN3, and scavenger receptors are downregulated, which is consistent with previous reports, suggesting that this may be a result of Mtb maintaining intracellular survival, though the specific regulatory mechanisms remain unclear^[42]. Second, Mtb can also utilize exosomes to actively eliminate vital immune cells. Infected cells package the bacterial protein Rv1818c into exosomes, which, upon delivery to T cells, induce mitochondria-dependent apoptosis, thereby actively clearing surrounding T cells^[79]. Whether Mtb hijacks the exosome system to deliver more immune escape factors can also serve as a direction for future research. Beyond proteins, Mtb’s regulation of nucleic acid components in exosomes is particularly crucial. Transcriptomic data from serum exosomes of active tuberculosis (ATB) patients show significant inhibition of multiple immune-related signaling pathways, indicating that Mtb may achieve immune escape by regulating the composition of exosomal RNA^[80]. A recent study confirms this view: miR-146a-5p expression is significantly increased in exosomes derived from BCG-infected macrophages, and uptake of these miR-146a-5p-enriched exosomes by macrophages can suppress inflammatory responses^[59]. MiR-146a-5p in exosomes targets the 3’ UTR regions of IRAK1 and TRAF6, inhibits their protein expression, thereby blocking the activation of the nuclear factor kappa B (NF-κB) signaling pathway, and ultimately leading to decreased expression of key pro-inflammatory cytokines such as interleukin-1 beta (IL-1β), interleukin-6 (IL-6), and tumor necrosis factor-alpha (TNF-α), impairing the bactericidal capacity of macrophages^[59]. Exosomes are also used to reshape the immune microenvironment. Singh et al. reported that exosomes released by Mtb-infected macrophages can be internalized by uninfected macrophages, activating recipient cells via the TLR2/MyD88 signaling pathway and selectively inhibiting the expression of their critical transcription factor CIITA^[81]. Similarly, mannose-capped lipoarabinomannan (ManLAM) of Mtb can induce mast cells to produce chemokine-containing exosomes that recruit macrophages and promote the high expression of M2 macrophage markers^[82]. Neutrophil-derived microvesicles can also interfere with the immune function of macrophages. When Duarte et al. treated Mtb-infected macrophages with these microvesicles, the antibacterial activity of macrophages was significantly inhibited^[83]. The authors pointed out that this inhibitory effect may be related to the anti-inflammatory properties of neutrophil microvesicles reported previously - namely, they can induce macrophages to produce anti-inflammatory cytokines such as transforming growth factor-β (TGF-β) while suppressing the release of pro-inflammatory mediators such as TNF-α and IL-8^[83]. The authors speculate that neutrophil-derived microvesicles may exert different effects in different environments: for example, they can stimulate inflammation in the circulation but exhibit anti-inflammatory properties in the context of Mtb infection^[83]. Their specific cargo composition and functional pathways still await systematic elucidation.

By hijacking host exosomes and microvesicles, Mtb has constructed a sophisticated, multi-layered “toolkit” for immune regulation. These HC EVs (including exosomes and microvesicles, with the evidence base for each subtype specified above) not only act as carriers of molecular cargo but also as long-range executors of pathogen instructions, acting synergistically in space and time: exosomes systematically reprogram the host’s innate and adaptive immune responses through the delivery of bioactive molecules such as proteins and nucleic acids, achieving long-term immunosuppression; while neutrophil-derived microvesicles act rapidly at the infection front, directly impairing the antibacterial function of macrophages. Deciphering this complex “communication network” not only holds important theoretical significance for a deep understanding of the pathogenesis of TB but also may provide key targets for the development of novel EVs-based diagnostic biomarkers and immune intervention strategies.

As key components for Mtb to regulate the microenvironment and implement immune intervention in the host, the biogenesis and functional diversity of Mtb EVs under iron-limiting conditions have become important entry points for understanding the mechanism of persistent infection by this pathogen. Early studies using lipidomic analysis revealed that Mtb EVs induced by low iron are enriched in mature mycobactin, whereas those under high iron conditions are dominated by precursor forms^[27]. Further experiments demonstrated that such Mtb EVs can act as functional siderophores to complement the growth defect of siderophore synthesis-deficient strains in low-iron environments, and this function depends on the mycobactin content in Mtb EVs and the iron-binding capacity of mycobactin^[27,84]. This finding not only reveals the direct role of Mtb EVs in iron acquisition but also suggests that they may serve as a group cooperation strategy for Mtb in iron-limited microenvironments, supporting the persistent survival of Mtb in host microenvironments with severe iron restriction^[27]. Recent proteomic studies have further expanded the understanding of the functional diversity of Mtb EVs. Quantitative mass spectrometry-based analyses showed that in addition to siderophores, Mtb EVs selectively enrich a wide range of effector molecules, including immune regulatory proteins, anti-apoptotic factors, metabolic enzymes, and various known antigens^[36]. These molecules collectively form a multifunctional effector platform, reflecting Mtb’s multi-layered adaptation strategies such as immune evasion, immunosuppression, and immune stimulation under iron-limited host environments^[36]. However, current research still has several important limitations. First, existing conclusions are mainly based on in vitro culture models, and the functional integration and spatiotemporal dynamics of Mtb EVs in real infection scenarios have not been directly verified in host cells or animal infection systems. Second, although omics data have revealed the complex composition of Mtb EVs, the interpretation of their functional mechanisms is mainly based on bioinformatics predictions and literature associations, lacking functional experimental verification for key putative factors; nor has it been confirmed in cell or animal infection models whether Mtb EVs components can truly mediate immune escape, regulate host metabolism, or affect infection outcomes. Chiplunkar’s study on Mycobacterium avium also provides evidence for the adaptation mechanism of Mtb EVs under harsh environments in the host^[85]. They used an in vitro “phagosome model” simulating the metal ion composition of phagosomes for Mycobacterium avium. In this model, Mycobacterium avium not only significantly enhanced the release of Mycobacterium avium EVs but these EVs also exhibited unique virulence characteristics in protein composition, including Eis protein, PE/PPE family proteins, etc.^[85]. Notably, double-stranded DNA (dsDNA) was also detected in the study, and it was proposed that dsDNA may be involved in various biological processes such as biofilm formation, genetic exchange, and host immune regulation^[85]. We have observed the specific enrichment of these virulence-related proteins in Mtb EVs under iron-limiting conditions. However, a more in-depth scientific question worthy of exploration is: through what biological mechanisms do these Mtb EVs, with strictly selected cargo, systematically enhance Mtb’s adaptability to extreme host environments?

Notably, this stress-induced nature and cargo plasticity of Mtb EVs are not unique to iron-limiting environments. Recent studies have shown that when Mtb is exposed to sub-inhibitory concentrations of traditional anti-TB drugs (such as rifampicin, isoniazid, ethambutol) or antimicrobial peptides, it also significantly triggers the release of Mtb EVs^[37]. Similar to iron limitation-induced EVs enriched in mycobactin, these antibiotic-induced EVs also exhibit stressor-specific lipid and protein compositions. This finding forms an insightful parallel with the response pattern of Mtb EVs under iron limitation. Together, they reveal a core principle: Mtb can assemble and release Mtb EVs according to the specific type of environmental stress encountered. Whether coping with iron starvation (enriching siderophores), antibiotic attack (possibly enriching metabolism-related proteins or lipid barriers), or simulating phagosome environments (enriching virulence proteins such as Eis), Mtb EVs act as a medium for bacteria to export specific stress products extracellularly, perform community cooperation, or interfere with the host. Therefore, combining the study by Davids with research under iron limitation, we can propose a hypothesis: the biogenesis and cargo sorting mechanisms of Mtb EVs may be regulated by a set of core environmental sensing and signaling pathways, which can activate different cargo loading programs in response to different stress signals. When encountering a specific factor that threatens survival, this program is biased towards loading components that counteract the adverse effects of this factor to maintain the basic survival of Mtb. The key to future research lies in clarifying the specialized molecular switches behind this universal stress response platform, deciphering how Mtb precisely perceives the presence of different specific factors and translates them into targeted cargo sorting signals, and verifying how these “customized” EVs synergistically act in the complex in vivo infection environment to ultimately promote the persistent survival and immune escape of Mtb.

During TB infection, the secretion of HC EVs is regulated by Mtb, and accumulating evidence indicates that exosomes contribute to establishing a favorable intracellular survival environment for Mtb through the delivery of functional molecules. In the early stage of infection, Mtb inhibits host cell apoptosis through multiple mechanisms to maintain a stable intracellular environment required for its replication^[86]. As a critical molecular delivery system, exosomes play a central role in this process. Among the various molecules carried by exosomes, miRNAs are pivotal for regulating the host microenvironment^[6,87,88]. Mtb infection not only downregulates miR-20b-5p in the cytoplasm of macrophages but, more importantly, triggers the specific sorting and packaging of miR-20b-5p into exosomes by the cells. Zhang et al. hypothesized that Mtb “clears” the pro-apoptotic factor miR-20b-5p from cells by hijacking exosomes, which indirectly leads to the upregulation of its target protein Mcl-1, thereby inhibiting host cell apoptosis^[63]. Although this study was the first to report the downregulation of miR-20b-5p in macrophages after Mtb infection and detected the presence of this miRNA in secreted exosomes, it did not clarify the quantitative relationship and causal logic between the two. Therefore, the biological significance of miR-20b-5p in exosomes remains unclear: does it passively reflect the overall decrease in intracellular levels, or does it represent an active hijacking of the exosome system by Mtb to eliminate unfavorable intracellular miRNAs for its own survival? Studies have shown that the total amount of miRNAs in exosomes released by infected cells is significantly reduced, while these miRNAs are not downregulated in the cells themselves^[49]. This evidence supports the notion that miRNAs are primarily targeted by Mtb hijacking, but this key mechanism awaits further in-depth investigation. Current research mostly remains at the level of correlative description, and exploring the underlying causal relationships will be an important direction for future studies. Additionally, whether these exosomes carrying specific miRNA cargo, released by infected cells as potential intercellular messengers, can be internalized by surrounding host cells (such as other macrophages, lymphocytes, or lung epithelial cells) to alter the function of recipient cells, regulate the immune microenvironment, and thereby influence the course of infection - all of these require further exploration and verification. This debate over “active hijacking” vs. “passive reflection” is not unique to miR-20b-5p but is further highlighted in Guio et al.’s study on miR-125a-5p. By comparing with healthy controls, Guio’s research identified a specifically dysregulated set of miRNAs in exosomes from patients with ATB and latent tuberculosis (LTB)^[89]. Among them, miR-125a-5p was the only miRNA downregulated in blood exosomes from patients with LTB, ATB, and lung adenocarcinoma (PA)^[89]. Most differentially expressed miRNAs are highly expressed in normal lung tissue, suggesting their potential origin from lung tissue^[89]. Combining existing literature, they hypothesized that this downregulation may inhibit autophagy by relieving its suppression on STAT3 and anti-apoptotic proteins^[89,90]. However, miR-125a-5p expression is significantly induced and upregulated in macrophages infected with Mycobacterium avium^[90]. If the presence of miR-20b-5p in exosomes is the result of active sorting, it strongly supports the hijacking hypothesis - i.e., Mtb clears this intracellular threat by upregulating its export; if it is merely passive loading, it is just a reflection of the overall decrease in intracellular levels. To clarify this issue, future studies need to accurately compare the relative proportions of miRNAs in infected cells and their secreted exosomes. Moreover, both “hijacking” and “reflection” imply that exosome-mediated intracellular and extracellular distribution of miRNAs is a highly dynamically regulated process during infection, with similar complexity observed in other molecules such as miR-125a-5p. In addition to apoptosis, autophagy - an important mechanism for the host to clear intracellular Mtb - has also become an important target of exosome-mediated immune escape. Yuan et al. found that Mtb infection induces upregulated expression of miR-18a in macrophages and their exosomes^[62]. This miRNA targets the ataxia-telangiectasia mutated (ATM) gene, inhibits AMPK phosphorylation, thereby impairing the expression of the key autophagic protein LC3-II and autophagosome formation, ultimately suppressing host protective autophagy^[62]. However, there is no evidence yet whether exosomes containing miR-18a can be internalized by surrounding infected cells to create an environment conducive to Mtb survival. In contrast, the latest research has thoroughly investigated and revealed a pathway driven by exosomal miR-766-3p. Mechanistic studies by Cui et al. demonstrated that exosomes isolated from the plasma of ATB patients can be internalized by A549 lung epithelial cells and deliver their cargo miR-766-3p to recipient cells^[60]. miR-766-3p can directly bind to and target the 3’-UTR of the mRNA encoding natural resistance-associated macrophage protein 1 (NRAMP1), leading to a significant downregulation of NRAMP1 protein expression^[60]. This finding not only reaffirms the cross-cellular transport function of exosome-mediated miRNAs but also provides a mechanistic basis for understanding how Mtb regulates immune responses in distant cells through the host circulatory system. Metabolic reprogramming is another important strategy for Mtb to achieve persistent infection. BCG-infected human monocyte-derived macrophages release a set of specific exosomal miRNAs (e.g., miR-1224, -1293, -484), which mainly target host cell pathways such as fatty acid biosynthesis, central carbon metabolism, and amino acid metabolism. By remodeling the host metabolic network, these miRNAs enhance bacterial survival^[91]. Whether the differential expression of exosomal miRNAs during Mtb infection is a passive reflection of host cell metabolic disorders or an active strategy adopted by the pathogen to manipulate the immune microenvironment remains to be determined.

Although current studies have initially revealed the important functions of exosomes in TB infection, several challenges remain: first, the “active hijacking” vs. “passive reflection” mechanisms of exosomal miRNAs require rigorous verification; second, more research is needed to confirm whether exosomes carrying regulatory molecules can be internalized by surrounding cells and exert subsequent effects; finally, the causal relationship between the dynamic changes in exosome content and their functional outputs needs systematic clarification.

Mtb EVs play complex and pivotal roles in host immune regulation. Through their enriched PAMPs such as lipoproteins and lipoglycans, they act as potent TLR2 agonists, directly participating in the activation and modulation of innate immunity^{[11,15,58,98]}. They promote inflammatory responses by activating the TLR2 signaling pathway in macrophages^[11]. Yamaguchi et al. further clarified that Mtb EVs, via the binding of TLR2 ligands to receptors in mouse models, activate downstream NF-κB signaling to promote the secretion of inflammatory cytokines such as TNF-α, IL-6, and IL-1β^[98]. Beyond pro-inflammatory cytokines, Mtb EVs can also induce the expression of monocyte chemoattractant protein 1 (MCP1/CCL2), effectively driving the migration and chemotaxis of monocytes^[98,99]. Collectively, these studies demonstrate that Mtb EVs activate multiple signaling pathways through pattern recognition receptors such as TLR2, not only initiating pro-inflammatory responses but also coordinating the migration and recruitment of immune cells, thereby playing important roles in the immunopathogenesis of TB infection.

In earlier periods, HC EVs (especially exosomes) were considered the main mediators of macrophage inflammatory effects, and numerous studies have provided evidence to support this view. As early as 2007, a study by Bhatnagar et al. revealed the role of exosomes in initiating and amplifying inflammatory responses^[17]. Exosomes released by infected macrophages contain PAMPs such as LAM and 19-kDa lipoprotein. When internalized by neighboring macrophages, they can activate NF-κB and mitogen-activated protein kinase (MAPK) via the TLR/MyD88 signaling pathway, drive pro-inflammatory cytokine secretion, and recruit neutrophils and macrophages in vivo^[17]. Singh et al. observed that chemokines produced after exosomes carrying Mtb components (e.g., ManLAM) activate inflammatory responses and form a chemical gradient, which can effectively guide the directional migration of macrophages, neutrophils, and T cells in vitro^[100]. Subsequent studies further confirmed the universality of exosomes as inflammatory signals. Exosomes derived from cells infected with different mycobacteria (including Mycobacterium avium and Mycobacterium smegmatis) can activate resting macrophages via the TLR pathway through their cargo of bacterial glycopeptidolipids (GPLs) or host Hsp70, promoting pro-inflammatory cytokine release and co-stimulatory molecule expression^[18,101]. Similarly, exosomes derived from Mtb-infected mesenchymal stem cells can also synergistically activate macrophage NF-κB and p38 MAPK via the TLR2/4-MyD88 pathway, upregulating pro-inflammatory mediator expression, with this effect being time-dependent and peaking at 72 h post-infection^[102]. In addition to protein components, nucleic acid components in exosomes also possess immunomodulatory functions. Exosomes derived from Mtb-infected macrophages are enriched in immune-related RNAs, and their purified RNAs can specifically induce recipient macrophages to secrete pro-inflammatory mediators and chemokines, excluding interference from proteins and lipids^[49]. Notably, a study by Athman et al. clarified the division of labor between EVs of different origins in pro-inflammatory responses^[58]. They found that Mtb EVs released by infected macrophages, rather than host-derived exosomes, are the main carriers of TLR2 agonists such as Mtb lipoproteins and lipoglycans, thereby playing a dominant role in activating uninfected macrophages to produce pro-inflammatory cytokines such as TNF-α^[58]. This indicates that in TB infection, the TLR2-dependent pro-inflammatory response driven by specific PAMPs synthesized by the bacteria themselves is mainly mediated by Mtb EVs rather than exosomes^[58]. More importantly, they pointed out that previous studies failed to distinguish between the two due to technical limitations, potentially misattributing activity to exosomes^[58]. This may explain why Mtb EVs and exosomes exhibited similar mechanisms in earlier studies. This study warns that when analyzing the functions of Mtb EVs and HC EVs in infection systems, their cellular origins must be strictly distinguished; otherwise, deviations in mechanism interpretation may occur. In the inflammatory regulatory network of TB infection, microvesicles, as an important subclass of HC EVs, also actively participate in the progression of inflammation. Studies have shown that Mtb infection can significantly promote macrophages to release microvesicles, which can also stimulate recipient macrophages to secrete pro-inflammatory cytokines such as TNF, IL-8, and MCP-1, and synergistically disrupt the integrity of the respiratory epithelial cell monolayer, thereby facilitating the transepithelial migration of inflammatory cells^[65]. Further proteomic studies by Hare et al. found that type I interferon-induced proteins (e.g., ISG15, IFIT1-3) are significantly enriched in microvesicles from infected cells. When microvesicles are internalized by immature macrophages, they can actively induce recipient cells to upregulate the expression of type I interferon-responsive genes^[64].

In the immunopathological process of TB, Mtb EVs and HC EVs exert synergistic effects in the initiation, amplification, and regulation of pro-inflammatory responses by mediating intercellular communication. Mtb EVs, enriched in PAMPs such as lipoproteins, mainly activate NF-κB via Toll-like receptor 2-dependent signaling pathways, inducing macrophages to produce key pro-inflammatory cytokines such as TNF-α and IL-6, as well as chemokines such as monocyte chemoattractant protein-1, driving the recruitment of myeloid cells to infection sites. HC EVs, including exosomes and microvesicles, synergistically promote the formation of an inflammatory microenvironment by delivering bacterial components or host-encoded immunomodulatory molecules. Among them, exosomes can carry host molecules such as Hsp70, enhancing the production of pro-inflammatory mediators via the TLR/MyD88 pathway; they can also transport specific microRNAs to limit excessive inflammation by negatively regulating the NFAT5/NF-κB signaling axis. Microvesicles, on the other hand, promote the expression of interferon-stimulated genes by upregulating type I interferon-induced proteins, further enhancing chemokine secretion. Together, Mtb EVs and HC EVs form an elaborate regulatory network that maintains a dynamic balance between establishing effective antibacterial immunity and preventing tissue damage, with their specific molecular mechanisms requiring further clarification.

In the immune process of TB, the precise regulation of various immune cells is crucial for activating effective immune responses. As the primary host cells of Mtb, macrophages play a central role in initiating and controlling anti-TB immunity, and their activation status directly determines the outcome of infection^[103]. The host also responds to infection through macrophage death pathways, and macrophage apoptosis has been proven to be an important host defense mechanism^[103]. Both Mtb EVs and HC EVs occupy a core position in the regulation of macrophages.

First, we elaborate on the relevant mechanisms of exosomes. On one hand, exosomes lay the foundation for establishing early immune defense lines by promoting the functional maturation and directional recruitment of macrophages. Studies have shown that macrophage-derived exosomes can induce the differentiation of monocytes into macrophages by activating the ERK1/2-MK-2-NF-κB signaling pathway, and upregulate adhesion-related genes to promote cell maturation^[47]. Exosomes from immunostimulated macrophages are enriched in endoplasmic reticulum aminopeptidase 1 (ERAP1) and various cytokines, which can significantly enhance the phagocytic activity and NO synthesis capacity of recipient macrophages^[104]. On the other hand, exosomes finely regulate cellular autophagy and apoptosis programs by delivering nucleic acid substances. Host and bacterial RNAs carried by host cell-derived exosomes released from Mtb-infected macrophages can be delivered to uninfected macrophages, where these exosomal RNAs can be translated into proteins such as LCN2 to exert antibacterial functions^[49]. These exosomes can also directly enhance their ability to clear intracellular bacteria by precisely regulating cellular autophagy and apoptosis programs. Mtb RNA can be delivered into exosomes released by infected macrophages through Mtb’s SecA2 secretion system; after being internalized by recipient macrophages, it can activate RIG-I/MAVS pathway-mediated LC3-associated phagocytosis and enhance phagosome-lysosome fusion^[48]. Meanwhile, specific miRNAs precisely target and regulate key autophagic nodes. A study by Xu et al. confirmed that miR-107 delivered by exosomes enhances mitochondrial reactive oxygen species (ROS) production and promotes autophagy by inhibiting Wnt16 to activate the Wnt/β-catenin pathway^[105]. In terms of clearing infected cells, natural killer (NK) cell-derived exosomes induce apoptosis by delivering cytotoxic proteins (perforin, granulysin, etc.) to activate the caspase cascade^[106]. Although this study confirmed that NK-exo carries typical NK cell cytotoxic proteins such as perforin (PFN), granulysin (GNLY), and granzyme A (Gzm-A), the direct roles and relative contributions of these specific molecules in mediating the aforementioned phenotypes such as pro-apoptosis, anti-inflammation, and ROS elevation require further investigation^[106]. The miR-1249-3p carried by NK-exo synchronously activates mitochondrial and death receptor-mediated apoptosis pathways by targeting the SKOR-1 gene, ultimately clearing infected cells through the activation of Caspase-3 and PARP^[68]. After being delivered to infected macrophages, miR-1249-3p contained in these cell-derived exosomes directly binds to and inhibits the expression of the SKOR-1 gene, thereby relieving its inhibitory effect on cell apoptosis^[68].

The role of microvesicles in the activation and regulation of macrophages has also been gradually revealed. Microvesicles released by Mtb-infected neutrophils significantly activate macrophages by increasing the production of TNF-α, IL-6, and IL-10, as well as upregulating the expression of the co-stimulatory molecule CD86^[107]. Alvarez-Jiménez found that macrophages treated with microvesicles from infected neutrophils not only enhance macrophage activation but, more importantly, can significantly upregulate autophagy levels^[107]. By inducing the expression of the autophagy marker LC3-II and its colocalization with intracellular bacteria, macrophage clearance of Mtb is enhanced^[107].

In the immune regulatory network of TB infection, exosomes and microvesicles collectively form a multi-layered defense system by regulating macrophage differentiation, autophagy, and apoptosis. This provides a vital mechanistic basis for the host to resist pathogens and points out directions for the development of novel immune intervention strategies.

Nucleic acidsWith the implementation of multiple independent studies, a large number of nucleic acids derived from hosts or pathogens have been identified as potential diagnostic biomarkers for TB. Host miRNAs are the most extensively studied biomarkers to date. Multiple independent studies have confirmed that specific miRNAs in the blood exosomes of ATB patients show a significant upregulation trend. For instance, the combined detection of miR-484, miR-425, and miR-96 exhibits superior diagnostic efficacy [area under the curve (AUC) = 0.78] compared to single indicators, demonstrating the advantage of multi-marker combined detection^[129]. As research advances, more high-performance single miRNA biomarkers have been successively discovered, such as miR-185-5p (AUC = 0.750) and miR-766-3p (AUC = 0.896), with the latter approaching clinical application standards^[60,130]. Diagnostic strategies are evolving from single indicators to multi-dimensional integration. Hu et al. found that six exosomal miRNAs (miR-20a, miR-20b, miR-26a, miR-106a, miR-191, and miR-486) are significantly upregulated in the plasma of TB patients; when combined with electronic health records, this panel exhibits excellent performance in distinguishing pulmonary TB from tuberculous meningitis^[131]. Nevertheless, it is important to emphasize that while multi-miRNA panels demonstrate robust statistical associations with disease status, these correlations do not automatically equate to functional causality. The mechanistic interpretations of these miRNA signatures - particularly whether they reflect active bacterial manipulation of host pathways or merely represent general host cellular stress responses secondary to infection - remain largely unresolved. Extensive experimental validation, including loss- and gain-of-function studies in relevant cell types and longitudinal animal models, is still required to determine the functional significance of these EVs-encapsulated molecules. The transition from promising correlative biomarkers to established functional indicators therefore represents a critical gap that future investigations must address. Meanwhile, research focus has expanded from miRNAs to other non-coding RNAs (ncRNAs). Through bioinformatics analysis and experimental validation, Fang et al. identified four significantly downregulated long non-coding RNAs (lncRNAs) (NONHSAT101518.2, NONHSAT067134.2, NONHSAT148822.1, and NONHSAT079957.2) in the serum exosomes of ATB patients, which display high diagnostic value^[132]. In addition to host nucleic acids, direct detection of Mtb-derived RNA fragments offers theoretically high specificity. Researchers identified Mtb-specific small RNAs (sRNAs) (ASdes and MTB-miR5) in serum EVs (isolated by size-exclusion chromatography, yielding a mixed population enriched in exosomes and small microvesicles); these biomarkers are exclusively detected in pulmonary tuberculosis patients, including those initially smear- and culture-negative but later converted positive, indicating promising potential for early diagnosis^[133]. Similarly, Mtb ribosomal RNA fragments (16S-L1 and 16S-L2) identified from plasma exosomes are significantly upregulated in TB patients, with a combined diagnostic AUC of 0.7197^[134].

Proteins and lipidsMtb-specific proteins and lipids serve as direct evidence of its presence. Targeting these unique molecules enables precise diagnosis of the pathogen. For example, the cell wall lipid LAM has become a key basis for urine testing in HIV-infected individuals^[135]; meanwhile, accurate quantification of specific antigen peptides such as the Ag85 complex and ESAT-6/CFP-10 in patients’ blood or urine using high-sensitivity mass spectrometry has also demonstrated excellent diagnostic value^[136]. Mtb infection triggers intense immune and inflammatory responses in the human body, which are reflected in changes in protein and lipid metabolism.

As enriched carriers of disease-related information, exosomes provide a highly valuable biomarker library for TB diagnosis, with value manifested at two levels: First, at the host response level, proteomic analysis can screen for host proteins closely associated with infection status. For instance, Zhang performed proteomic profiling of serum exosomes from ATB patients and identified 123 differentially expressed proteins, among which changes in LBP (increased), CD36 (decreased), and MHC-I (decreased) are diagnostically relevant^[117]. Wang et al. revealed differential characteristics of infections caused by different mycobacteria: by comparing the plasma exosomal protein profiles of patients infected with M. abscessus, M. intracellulare, and Mtb, they identified protein panels such as C1R/C4B/C9 and SERPINA5/SERPING1 that can effectively distinguish non-tuberculous mycobacterial infections from Mtb infections^[137]. Second, more directly, unique components of Mtb itself have been detected in exosomes. Jiang et al. measured three Mtb antigens (MPT64, Ag85A, and CFP-10) in the plasma exosomes of ATB patients, LTB patients, recovered individuals, and healthy controls. They found that these three antigens are specifically enriched in the exosomes of ATB and LTB patients, confirming their diagnostic potential^[138]. VAMP5 in HC EVs is significantly upregulated in TB patients and downregulated after treatment, indicating its potential as a TB biomarker^[51]. Additionally, the exosome-enriched Hsp16.3 protein is stably present in the plasma of ATB patients but not detected in latent infection or healthy populations, suggesting its value as a highly specific blood biomarker^[139]. Notably, this exosome-based detection technology significantly enhances the performance of traditional biomarkers. Dahiya et al.’s study demonstrated that detecting LAM in urine exosomes using high-sensitivity immuno-polymerase chain reaction (I-PCR) exhibits superior diagnostic efficacy compared to traditional ELISA, opening new avenues for non-invasive diagnosis^[140].

In the future, organically combining these biomarkers from different levels to construct a multi-dimensional diagnostic model will be the critical direction for achieving ultra-early and high-precision diagnosis of TB.

Distinguishing ATB from LTBBased on WHO guidelines, the accurate distinction between ATB and LTB is a core decision point in TB prevention and control, directly determining drastically different clinical management strategies^[1]. This differentiation is crucial for preventing missed diagnoses of active cases that delay treatment, avoiding drug resistance risks caused by single-drug prophylaxis, and optimizing the allocation of public health resources. Emerging EVs research provides a novel perspective to address this issue; by decoding the pathogen- and host-specific molecular information carried by EVs, it is expected to establish a more precise diagnostic pathway in the future.

At the RNA level, Lyu et al. compared the miRNA profiles in serum exosomes from healthy individuals, LTB patients, and ATB patients via small RNA sequencing. They identified 250 differentially expressed miRNAs (including 130 specifically expressed ones), and validated 4 LTB-specific miRNAs (hsa-let-7e-5p, hsa-let-7d-5p, hsa-miR-450a-5p, and hsa-miR-140-5p) as well as 5 TB-specific miRNAs (hsa-miR-1246, hsa-miR-2110, hsa-miR-370-3p, hsa-miR-28-3p, and hsa-miR-193b-5p)^[141]. These differentially expressed miRNAs may participate in disease progression by regulating host signaling pathways, providing a molecular basis for distinguishing infection status. Detection of Mtb protein components carried by HC EVs offers more direct evidence. Mehaffy et al. detected multiple Mtb peptides in serum HC EVs (isolated by ultracentrifugation, representing a mixed vesicle population) from individuals with LTB infection using multiple reaction monitoring mass spectrometry (MRM-MS). The results showed that at least one Mtb peptide was detected in 95% of individuals, with the peptide SVF from glutamine synthetase (GlnA1) identified in 82% of samples^[142]. A study by the Zheng team in a non-human primate model found that the co-localization pattern of LAM and LprG on the surface of HC EVs has discriminatory value: LTB only shows elevated LprG signals, while ATB exhibits significant enhancement of both LAM and LprG signals^[143]. This indicates that their co-localization can serve as a key biomarker panel for distinguishing active disease from latent infection. Finally, at the metabolic response level, host lipid reprogramming also provides important information. Lyu’s team’s lipidomic study identified 7 lipid biomarkers with diagnostic value, among which the free fatty acid FFA (20:1) performed particularly well in distinguishing ATB from LTB, with an AUC of 0.952, 100% specificity, and 88.9% sensitivity^[144].

Collectively, these studies demonstrate that HC EVs integrate multi-dimensional information such as pathogen proteins, host miRNAs, and metabolic lipids to construct a molecular diagnostic system capable of accurately distinguishing TB infection status, providing a new technical pathway for optimizing TB clinical management.

Localization, identification, and classification of TBThe accurate diagnosis of different types of TB has long been a clinical challenge, and exosomal miRNAs have shown great potential as promising biomarkers. Multiple studies have revealed their specific expression profiles in distinguishing TB types: Kushwaha et al.’s study on the Indian population found distinct miRNA expression profiles in serum exosomes of patients with HIV-tuberculosis coinfection (HIV-TB), extrapulmonary tuberculosis, and pulmonary tuberculosis alone (PTB). Among them, hsa-let-7i-5p, hsa-miR-24-3p, and hsa-miR-92a-3p were significantly differentially expressed in HIV-TB coinfected patients; hsa-miR-20a-5p, hsa-let-7e-5p, and hsa-miR-26a-5p were specifically downregulated in extrapulmonary tuberculosis (EPTB) patients; while hsa-miR-21-5p, hsa-miR-19a-3p, hsa-miR-19b-3p, and hsa-miR-146a-5p were significantly expressed in PTB patients^[145]. Sun et al. identified 62 differentially expressed miRNAs in macrophage exosomes from spinal tuberculosis (STB) patients via high-throughput sequencing, including 28 upregulated and 34 downregulated ones. Further validation by quantitative reverse transcription PCR (qRT-PCR) showed that miR-125b-5p was significantly upregulated in the STB group, with an area under the receiver operating characteristic curve (AUC) of 0.881.^[146]. These studies collectively confirm that HC EVs miRNAs, as highly promising biomarkers, hold tremendous exploration value and application potential in achieving precise classification diagnosis of TB.

Distinguishing TB from other pulmonary diseases (e.g., lung cancer)In clinical practice, the differential diagnosis between TB and lung cancer is of great significance. Due to their potential coexistence or sequential occurrence, coupled with overlapping imaging manifestations, the differential diagnosis and subsequent treatment plans face significant challenges^[147]. In this context, identifying novel diagnostic biomarkers capable of effectively distinguishing these two diseases is crucial for improving the accuracy of clinical decision-making.

Molecular analysis based on HC EVs provides a new solution. Studies have shown that the miRNA profiles of HC EVs exhibit significant specificity under different disease states. Guio et al. found via serum exosome miRNA sequencing that miR-125a-5p is perturbed in both TB and lung cancer, while miR-20a-5p and miR-584 are specifically upregulated in LTB and ATB, respectively^[89]. Research has also shifted to pleural effusion, a key sample, to achieve precise differentiation between TB and other lesions. The miRNA characteristics of HC EVs in pleural effusion also show discriminatory value: the Zhang team identified 99 upregulated and 91 downregulated miRNAs, among which hsa-miR-200b-3p and hsa-miR-182-5p are highly expressed in malignant pleural effusions, suggesting these exosomal miRNAs may serve as potential biomarkers for distinguishing malignant pleural effusions (MPE) from tuberculous pleural effusions (TPE)^[148]; the Wang team found 9 miRNAs (including miR-205-5p, miR-483-5p, miR-375, and members of the miR-200 family) enriched in pleural effusions from lung adenocarcinoma patients, and another 3 miRNAs (miR-148a-3p, miR-451a, miR-150-5p) that can distinguish tuberculous from benign pleural effusions, filling the diagnostic gap of only distinguishing cancerous from non-cancerous pleural effusions^[149]. In addition to nucleic acid biomarkers, proteins and metabolites in HC EVs also provide important discriminatory information. Mtb-specific antigens MPT51 and MPT64 enriched in HC EVs from pleural effusions show significantly better detection performance than direct use of pleural effusion samples (PF-ELISA), and can be used for the diagnosis of pleural tuberculosis^[150]. At the metabolic level, Luo et al. analyzed the metabolic characteristics of different EVs subsets in pleural effusions and identified a four-molecule biomarker panel consisting of phenylalanine, leucine, phosphatidylcholine, and sphingomyelin, which exhibits extremely high diagnostic efficacy^[151].

HC EVs show good adaptability in multi-disease differentiation scenarios, and their cargo profiles can effectively reflect disease-specific host responses and pathogen information. However, as noted in the mechanistic discussions in Sections “Roles of HC EVs in creating a favorable intracellular survival environment” and “Potential involvement of Mtb EVs and HC EVs in Mtb drug resistance”, many of the miRNA and protein signatures associated with TB diagnosis currently rest on correlative evidence. While these profiles exhibit promising statistical power in cross-sectional studies, extensive functional and experimental verification is still required to determine whether the observed changes represent active bacterial manipulation of host pathways or are merely secondary manifestations of general host cellular stress. This distinction is not merely academic: it has direct implications for the biological interpretability of diagnostic biomarkers and their potential utility as therapeutic targets. We therefore urge the field to complement continued biomarker discovery with rigorous mechanistic studies that causally link EVs cargo alterations to specific pathogen-host interactions, thereby transforming statistical associations into functionally validated indicators.

Prognosis assessment and therapeutic efficacy monitoring of TB are critical links in disease control, which are essential for identifying high-risk individuals with progression potential, accurately evaluating the effectiveness of treatment regimens, and predicting recurrence. However, existing biomarkers still rely on complex technologies and have limitations in application among children, HIV-coinfected patients, and extrapulmonary TB patients, highlighting an urgent need for developing novel non-invasive, high-precision, and easily scalable monitoring tools^[152].

In terms of therapeutic monitoring for drug-resistant TB, Carranza et al. identified miR-let-7e-5p with temporal expression characteristics in serum exosomes. The expression level of miR-let-7e-5p undergoes regular changes during treatment progression, especially at and after the time point of sputum culture conversion to negative, demonstrating its potential as a dynamic indicator for therapeutic efficacy monitoring^[153]. Meanwhile, studies detected Mtb-specific RNA sequences such as 30 kDa and 5KST in exosomes of 33% of multidrug-resistant tuberculosis (MDR-TB) patients even after the intensive phase of treatment^[153]. The authors proposed combining host-derived miR-let-7e-5p with pathogen-derived RNA as a “dual biomarker”: abnormal expression of host miRNA coupled with persistent positivity of pathogen RNA strongly indicates the risk of treatment failure or future recurrence^[153].Another study compared the plasma exosomal miRNA expression profiles among MDR-TB patients, drug-susceptible tuberculosis (DS-TB) patients, and healthy controls, identifying 7 miRNAs with significantly differential expression in MDR-TB patients. Further focus was placed on three key molecules: hsa-miR-122-5p_R-1, hsa-miR-23b-3p_R+1, and hsa-miR-15a-5p_R-1^[154]. Exosomal miR-122-5p, KNG1, and C3 also exhibit diagnostic value in drug-resistant TB^[96]. Beyond nucleic acids and proteins, the exosomal lipidome has emerged as a sensitive indicator of treatment response. Exosomal lipid compositions (e.g., sphingomyelin, triglycerides) undergo significant changes after effective treatment and tend to return to normal, indicating their potential as biomarkers for therapeutic monitoring^[42].In LTB, the expression levels of complement component 4 binding protein alpha chain (C4BPA) and S100 calcium-binding protein A9 (S100A9) in plasma exosomes of patients with effective LTB treatment are significantly downregulated. ROC analysis showed that the combined AUC of these two molecules is 0.78, suggesting their potential as biomarkers for evaluating the effectiveness of LTB preventive treatment^[155]. Overall, by integrating multi-dimensional signals from both the host and pathogen, exosomes provide an unprecedented systematic monitoring perspective for TB therapeutic efficacy assessment and prognostic management.

In the development of HC EVs-based diagnostic tools, attention must be paid to the impact of population and regional specificity on diagnostic outcomes. Studies by Kushwaha et al. on the Indian population and Guio et al. on the Peruvian population observed certain regional characteristics^[89,145]; miRNA expression profiles, as well as protein and lipid compositions, may be influenced by population genetic backgrounds and regional environmental factors. Whether biomarkers that perform excellently in one cohort will experience diagnostic efficacy drift when generalized to other regions remains unproven. Future research must emphasize external validation in large-scale, multi-center cohorts across different endemic areas to avoid diagnostic bias and ensure the universality and stability of the tools.

The influence of the host’s immune status on diagnosis also deserves attention. Jin’s team evaluated the diagnostic efficacy of six exosomal miRNAs (miR-20a, miR-20b, miR-26a, miR-106a, miR-191, and miR-486) in individuals with HIV-TB coinfection. They found that these miRNAs were significantly upregulated in the plasma of HIV-TB coinfected patients, with more pronounced expression changes in extrapulmonary TB^[118]. Subsequent subgroup analysis revealed that the diagnostic performance of these miRNAs was superior in patients with a CD4⁺ T-cell count ≥ 200 cells/μL, suggesting that the host’s immune status may affect exosomal miRNA expression levels^[118]. This finding serves as an important reminder: for populations with comorbidities (e.g., HIV-infected individuals, diabetic patients, or the elderly), HC EVs-based diagnostic models may require specific calibration or the development of independent biomarker panels. This not only relates to diagnostic accuracy but also directly impacts the interpretation of therapeutic efficacy monitoring results. Notably, antigens derived from BCG vaccines or other mycobacteria may cross-react with HC EVs proteins, affecting the specificity of serological tests^[23,156]. Whether such cross-reactivity will challenge the specificity of HC EVs-based serological assays or antigen capture experiments, leading to false positives in latent infection or misdiagnosis of ATB, requires further investigation. Additionally, most existing studies focus on cross-sectional comparisons, resulting in limited understanding of the dynamic evolutionary patterns of HC EVs throughout the entire course of TB infection. Comprehensive analysis of the dynamic changes in HC EVs from various sample sources may address many current challenges, such as whether HC EVs exhibit characteristic changes prior to TB recurrence, whether the dynamic changes of HC EVs biomarkers after treatment initiation can predict the final outcome, and whether EVs characteristics after cure can predict recurrence risk.

Beyond these population- and disease-specific considerations, the clinical translation of HC EVs-based TB applications faces profound systemic hurdles rooted in methodological heterogeneity and technical standardization gaps. A fundamental issue lies in the ambiguous subtype attribution that pervades many diagnostic studies: although serum or plasma “exosomal” miRNA panels have shown promising statistical correlations with disease status, the isolation methods commonly employed - typically differential ultracentrifugation or precipitation-based kits - do not achieve complete separation of exosomes from microvesicles and apoptotic bodies. Consequently, markers frequently reported as “exosomal” may actually reflect contributions from mixed vesicle populations, a caveat that does not negate their diagnostic value but critically complicates mechanistic interpretation and cross-study comparability. On a broader level, the lack of standardized isolation and characterization protocols across laboratories represents an equally formidable bottleneck. Different research groups employ vastly divergent methods - including ultracentrifugation, density gradient centrifugation, size-exclusion chromatography, precipitation-based kits, and microfluidic devices - each yielding EVs with substantially different purity, recovery rates, and functional activity. This methodological heterogeneity severely compromises cross-study reproducibility and has been documented to produce divergent conclusions regarding EVs cargo composition and immunomodulatory activity. Although the MISEV2023 guidelines provide a comprehensive minimal information framework encompassing recommended positive and negative markers, characterization requirements, and reporting standards, adherence within the TB field remains inconsistent. Compounding these issues, preanalytical variables - such as blood collection tube types, time from sample collection to processing, freeze-thaw cycles, and storage conditions - can profoundly affect EVs yield, integrity, and molecular cargo profiles, yet most TB studies fail to report these parameters in sufficient detail, potentially explaining some of the discrepant biomarker results between independent cohorts. Furthermore, batch-to-batch variability in EVs preparations poses a major hurdle, particularly for therapeutic applications where consistent dosing and cargo loading are critical, and scaling production from laboratory settings to Good Manufacturing Practice grade introduces additional challenges including cell source consistency, culture condition optimization, and purification reproducibility. Collectively, these challenges are interconnected: the inability to isolate pure EVs subtypes undermines functional attribution, the absence of standardized protocols precludes meaningful inter-laboratory comparisons, and unstandardized preanalytical handling introduces unpredictable variability that obscures true biological signals. To move the field forward, future diagnostic studies should adopt more stringent subtype-specific isolation protocols - such as immunoaffinity capture or optimized density gradient centrifugation - and rigorously adhere to MISEV2023 recommendations for marker verification, ensuring that biomarker candidates can be precisely attributed to their vesicle subpopulations of origin. Concurrently, the establishment of well-characterized EVs reference materials derived from standardized Mtb-infected macrophage cultures or pooled patient plasma, coupled with external quality assessment programs analogous to those in clinical chemistry, would enable benchmarking of isolation efficiency and functional assays across centers, while the adoption of FAIR (Findable, Accessible, Interoperable, Reusable) data principles would facilitate meta-analyses and accelerate the identification of robust biomarkers.

In summary, the value of HC EVs in TB diagnosis and treatment is systematic - not only do they carry pathogen information for direct detection, but they also reflect the host’s immune and metabolic status, and further integrate host and pathogen information in therapeutic efficacy monitoring. However, their future clinical translation must be built on overcoming population heterogeneity, clarifying immune-related influences, avoiding cross-reactivity, and addressing the systemic methodological and standardization challenges outlined above. Only through global collaboration and the establishment of standardized large-scale biobanks and databases can this highly promising technology be transformed into a reliable tool that truly serves the precise prevention and control of TB. Simultaneously, AI-driven HC EVs analysis is emerging as a cutting-edge direction in liquid biopsy for oncology^[157]. In the future, leveraging machine learning and deep learning algorithms of artificial intelligence holds the potential to enhance both the analytical efficiency and diagnostic accuracy of exosome-based detection for TB, thereby promising advancements in disease diagnostics overall.