Classical OMVs form by outward budding of the outer membrane through several non-mutually exclusive mechanisms. These include localized weakening of Braun’s lipoprotein (Lpp)-peptidoglycan (PG) links^[30], accumulation of PG fragments or misfolded proteins in the periplasm^[39], and altered phospholipid transport through the VacJ/Yrb ATP-binding cassette (ABC) system^[40]. Lipid A remodeling and reduced outer membrane protein A (OmpA)-PG attachment can also favor membrane curvature and release^[31]. In Escherichia coli, YdcZ interacts with YdiH and BssR to regulate vesicle formation and selective protein loading^[41]. In Vibrio fischeri, rotation of sheathed flagella promotes OMV release, providing evidence for a mechanical contribution to biogenesis^[42]. Different pathways can coexist within a bacterial population. This heterogeneity complicates attribution of specific cargo and functions to individual vesicle subpopulations.

Beyond orderly blebbing, phage-derived endolysins can trigger explosive lysis, as demonstrated in Pseudomonas aeruginosa. Released membrane fragments can reassemble into vesicles^[32]. OIMVs contain both outer and inner membranes and can enclose cytoplasmic material. OIMVs have been observed in Neisseria gonorrhoeae, P. aeruginosa, and Acinetobacter baumannii; double-membrane protrusion was directly visualized in N. gonorrhoeae^[21]. Double-membrane topology therefore does not by itself establish an explosive-lysis origin. The relative contribution of these pathways to the intestinal vesicle pool remains unknown. Density-gradient and size-exclusion purification alone do not reliably assign biogenesis subtypes. The Minimal Information for Studies of Extracellular Vesicles 2023 (MISEV2023) recommendations include bacterial EVs, but subtype assignment requires organism- and preparation-specific structural evidence^[43].

The thick PG layer of Gram-positive bacteria was once thought to prevent vesicle release^[33]. However, CMVs have been isolated from Staphylococcus aureus, Bacillus subtilis, and Lacticaseibacillus casei^[34-36]. Autolysin-mediated PG hydrolysis facilitates passage through the cell wall, as demonstrated for Sle1 in S. aureus^[34]. Phenol-soluble modulins (PSMs) promote membrane remodeling and vesicle release; PSM-deficient mutants show reduced production in the studied conditions^[34,37]. Prophage-encoded holin-endolysin systems can also promote vesiculation through cell-wall damage^[44]. In B. subtilis, endolysin-generated openings permit cytoplasmic membrane protrusion and vesicle release, followed by membrane failure and cell death^[35]. The relative contribution of each mechanism depends on the strain and growth conditions.

Vesicle production is not a fixed constitutive process, but rather, it responds dynamically to environmental cues, a property that is directly relevant to the gut ecosystem. Antibiotic exposure, nutrient limitations, oxidative and envelope stress, temperature shifts, and host immune pressure can alter OMV production rates, size distributions, and cargo loading preferences^[39,45-49]. For example, cysteine or other essential amino acid deprivation stimulates vesicle production in Neisseria meningitidis and Francisella spp., respectively^[45,48], whereas hydrogen peroxide and elevated oxygen tension promote OMV release in P. aeruginosa and N. meningitidis^[46,50]. Antimicrobial peptides, including polymyxin-like peptides, can likewise induce vesiculation as an adaptive envelope-stress response^[51,52].

Temperature also affects vesicle production and cargo loading; however, the direction and magnitude of these effects vary among bacterial species. In S. aureus, temperature variation alters EV abundance, size, and protein and RNA cargo, emphasizing the context-dependent nature of thermal regulation^[49]. Subinhibitory exposure to the β-lactam antibiotic ceftazidime increases membrane-vesicle formation in A. baumannii^[53], while gentamicin exposure increases OMV abundance and alters vesicular protein composition in susceptible extraintestinal pathogenic E. coli^[54].

These observations may have implications for patients receiving long-term antibiotic therapy, although their relevance to BEVs production in the human gut remains to be established. Within the gastrointestinal tract, the local chemical environment varies dramatically along the cephalocaudal axis: the pH, bile salt concentration, oxygen tension, and antimicrobial peptide gradients each impose selective pressures on the resident microbiota. Host-like changes in pH, divalent-cation availability, and antimicrobial exposure have been shown to affect OMV production, size, and lipid composition in Salmonella^[47]. However, the ways in which these gradients shape the vesiculation behavior of individual gut commensals and pathobionts in situ remain largely unexplored. Single-species in vitro assays capture only a fraction of the regulatory complexity; future work employing microfluidic gut-on-chip models and in vivo vesicle-tracking reporters will be essential for mapping the spatiotemporal dynamics of BEVs production across intestinal niches.

BEVs preparations contain neuroactive metabolites and associated metabolic enzymes, suggesting potential vesicle-associated metabolic activity. A metabolomic and metaproteomic study of stool-derived and ex vivo microbiota-derived vesicles reported neuroactive metabolites and enzymes associated with glutamine-glutamate-γ-aminobutyric acid (GABA) metabolism^[57]. GABA was also detected in vesicles from a cultured Bacteroides finegoldii isolate. These findings establish cargo association but do not demonstrate neurotransmitter synthesis within circulating vesicles. Metabolite annotations, particularly xenobiotic-like signals such as gabapentin, require confirmation before being interpreted as products of bacterial biosynthesis. Functional enzyme activity has been demonstrated in Bacteroides-derived OMVs enriched in acidic glycosidases and proteases^[59]. Separately, Bacteroides species synthesize sphingolipids that contribute to intestinal homeostasis and host immune regulation^[60], while microbial tryptophan metabolites can influence CNS-relevant aryl hydrocarbon receptor (AhR) signaling^[17]. Vesicular packaging and sustained production of these metabolites at distant sites require direct validation.

Microbe-associated molecular patterns (MAMPs) are an important class of immunologically active BEVs cargo. Gram-negative OMVs contain LPS and can deliver PG fragments to host cells. These components engage distinct receptors, including Toll-like receptor 4 (TLR4) and nucleotide-binding oligomerization domain-containing protein 1 (NOD1); the cited epithelial-cell study directly demonstrated NOD1 activation^[61]. Gram-positive vesicles can contain lipoteichoic acid and lipoproteins that engage Toll-like receptor 2 (TLR2)^[38,62]. Vesicle-associated presentation can alter cellular delivery and receptor activation, but its potency depends on the preparation, dose metric, and target cell^[62]. The number of intact BEVs reaching brain parenchyma under physiological conditions remains unknown.

LPS preparations differ in their biological activity. Lipid A acylation and phosphate position influence TLR4 responses, so LPS from different bacteria should not be treated as functionally interchangeable^[63]. Bulk endotoxin measurements alone do not resolve lipid A structure or predict all host responses. Structural analysis and functional receptor assays can therefore complement measurements of circulating LPS-positive BEVs^[64]. Capsular polysaccharide A (PSA) from Bacteroides fragilis provides another example of context-dependent immune signaling. PSA-containing OMVs engage dendritic-cell TLR2 and promote regulatory T-cell responses and interleukin-10 (IL-10) production^[55]. This finding complements earlier evidence that PSA contributes to host immune maturation^[65]. It does not establish that PSA activity is exclusive to vesicular presentation. Ligand identity, molecular structure, and delivery context should all be considered when evaluating BEV-mediated neuroimmune signaling.

BEVs carry small noncoding RNAs (sRNAs), including regulatory RNAs and transfer RNA-derived fragments. Selected RNAs can retain biological activity after uptake by host cells, although cargo selection and function vary among species and preparations^[66]. Enterotoxigenic and nontoxigenic B. fragilis OMVs differ in their molecular composition and metabolic activity^[67]. P. aeruginosa OMVs deliver an sRNA to human airway epithelial cells and reduce interleukin-8 (IL-8) secretion in the studied system^[58]. OMV-associated extracellular RNAs from the oral pathogen Aggregatibacter actinomycetemcomitans promote tumor necrosis factor alpha (TNF-α) production in human macrophages, and brain-associated RNA signal has been reported after administration in mice^[68]. These findings should not be interpreted as proof of intact gut-derived vesicle entry into brain parenchyma. The abundance, host targets, and intracellular mechanism of transferred bacterial RNAs require further investigation, including whether particular effects depend on host Argonaute proteins^[58,66].

Bacterial amyloids and proteolytic virulence factors have potential relevance to neurodegeneration, but evidence for their vesicular delivery differs. Curli produced by E. coli and Salmonella species form β-sheet-rich fibrils and can promote aggregation of host amyloidogenic proteins. Gut bacterial curli accelerated α-synuclein aggregation and worsened motor impairment in a mouse model^[69]. That study does not establish a BEV-dependent mechanism or delivery of curli into the CNS by intact vesicles. Nonvesicular amyloids and peripheral host responses remain possible contributors. Resolving the vesicular contribution will require strains with altered vesicle biogenesis or cargo loading, with controls for bacterial growth, lysis, curli production, and nonvesicular components.

Proteolytic virulence factors provide more direct examples of biologically active vesicle-associated cargo. Porphyromonas gingivalis produces the lysine-specific gingipain Kgp and arginine-specific gingipains RgpA/RgpB in soluble and OMV-associated forms^[70]. Gingipains detected in postmortem AD brain tissue correlate with tau and ubiquitin pathology. In P. gingivalis-challenged BALB/c mice, gingipain inhibition reduced cerebral bacterial burden and neurodegenerative changes^[71]. The human findings are associative, whereas the mouse experiment supports a mechanistic contribution in that model. Neither establishes a causal role for gingipains in human AD or their delivery into the brain by intact vesicles^[72]. Outside the CNS, Fusobacterium nucleatum OMVs containing Fusobacterium adhesin A (FadA) aggravated rheumatoid arthritis in an experimental study^[73]. S. aureus vesicles also induced atopic dermatitis-like inflammation^[74]. These peripheral examples justify investigating cargo-specific effects but do not establish corresponding CNS mechanisms.

Current cargo inventories derive mainly from bulk preparations and a limited range of culturable species. Important questions remain about the in vivo proteome, metabolome, and RNA content of gut BEVs, and whether particular cargo molecules coexist within individual vesicles. Molecular heterogeneity, including lipid A acylation, RNA modifications, and amyloid conformations, may further influence biological activity. Single-vesicle methods can help resolve selected cargo combinations, but each platform has limits in sensitivity, specificity, and throughput^[75]. Combining vesicle analysis with single-cell and spatial profiling of recipient host cells could connect cargo exposure to cellular responses. Such measurements must be paired with perturbation and transport experiments to test causality.

BEVs may cross the intestinal epithelium through paracellular passage or cellular uptake and transport. Barrier disruption can facilitate passage, although transport has also been studied in models without overt epithelial injury. Patients with intestinal barrier dysfunction have elevated circulating levels of LPS-positive BEVs^[64]. B. fragilis OMVs are taken up by dendritic cells and induce immunoregulatory responses^[55]; this finding does not establish microfold-cell transcytosis as their route across the epithelium. Bacteroides thetaiotaomicron OMVs undergo rapid epithelial uptake and can traverse primary intestinal epithelial monolayers^[78]. A separate reporter study demonstrated transfer of microbial biomolecules to host cells in vivo^[76]. After oral administration of labeled B. thetaiotaomicron OMVs, a biodistribution study detected the strongest extraintestinal signal in the liver^[78].

Once in the bloodstream, BEVs may be taken up by the mononuclear phagocyte system, although BEV-specific pharmacokinetic data remain limited. Intraperitoneally administered E. coli OMVs accumulated in the liver, lungs, spleen, and kidneys^[85]. This experiment describes exogenous administration rather than endogenous gut-to-brain delivery. General principles established in mammalian EV studies provide a conceptual framework but should not be presented as BEV-specific pharmacokinetic measurements^[86]. Plasma half-life, organ-specific uptake, and the brain-associated fraction of endogenous gut-derived BEVs remain poorly quantified. Tissue detection of vesicular components or labels does not alone establish intact-vesicle entry into brain parenchyma. Translocated OMVs can nevertheless affect distant organs; gut pathobiont-derived OMVs promote liver inflammation and fibrosis in primary sclerosing cholangitis-associated inflammatory bowel disease^[87].

The BBB contains tightly connected brain microvascular endothelial cells that restrict entry into brain parenchyma. Several candidate routes or interfaces may contribute to CNS exposure, but their anatomical location and supporting evidence differ.

Barrier disruption is one candidate route. Oral administration of P. gingivalis OMVs reduced brain tight junction gene expression and occludin protein abundance and was associated with NOD-like receptor family pyrin domain-containing 3 (NLRP3) inflammasome activation, tau phosphorylation, and memory impairment in mice^[81]. These observations do not establish direct gingipain-mediated cleavage of BBB claudin-5 or occludin. A comparison of P. gingivalis and its OMVs found different inflammatory responses and tissue-associated signals after exposure^[82]. Another study linked OMV exposure to BBB dysfunction and glycogen synthase kinase 3β-associated tau phosphorylation^[88]. Together, these findings support barrier dysfunction as a possible contributor to CNS exposure, but do not resolve whether intact vesicles, released cargo, or secondary inflammatory mediators cross the barrier.

Receptor-mediated transcytosis is a candidate route across an intact BBB, but BEV-specific receptor mechanisms remain incompletely defined. Studies of mammalian-cell-derived EVs found that transport depended on vesicle origin and could change after systemic LPS exposure^[89]. These findings are not direct evidence for BEVs transcytosis. OmpA contributes to invasion of brain microvascular endothelial cells by intact E. coli^[90], but this does not establish an equivalent receptor-dependent pathway for OMVs. The relevant receptors and in vivo kinetics of gut-resident BEVs transport remain unresolved.

The choroid plexus is a candidate route across the blood-cerebrospinal fluid (CSF) barrier rather than the BBB proper. After repeated intraperitoneal administration of P. gingivalis OMVs, gingipain immunoreactivity was detected around cerebral ventricles, in the choroid plexus, and in ependymal cells^[91]. This distribution does not identify the transported entity or the route taken. Entry from choroid plexus capillaries into CSF would still require passage across the choroid plexus epithelial barrier. Cathepsin B-associated stress-activated protein kinase/c-Jun N-terminal kinase (SAPK/JNK) signaling has been implicated in downstream pathology after OMV exposure^[92], but does not establish a transport mechanism.

A further hypothesis is transport within migrating leukocytes, sometimes termed a Trojan horse mechanism. Postmortem detection of bacterial 16S ribosomal RNA (rRNA) gene sequences does not identify vesicles or establish this route^[93]. Such signals may reflect bacterial components, intact bacteria, or contamination. BEV-associated neuroinflammatory changes in the brain and trigeminal ganglion likewise do not identify the carrier or route of entry^[94]. Direct evidence that leukocytes transport native gut-derived BEVs across the BBB remains insufficient.

Mechanistic BBB studies have emphasized oral pathobionts, but relevant evidence also exists for gut commensals. B. thetaiotaomicron BEVs have been studied in multicellular intestinal and BBB models, with vesicle-associated signals detected in recipient neuronal cells^[80]. Gut microbiota-derived preparations have also been tested in epithelial and brain endothelial transport models^[57]. More recently, colonization with B. thetaiotaomicron engineered to produce NanoLuc-labeled BEVs yielded reporter activity in distal organs, including the brain^[79]. This advances evidence for endogenous microbial cargo delivery. It does not establish the precise in vivo route or continuously visualize intact vesicles crossing the BBB. Receptor dependence, vesicle integrity during transit, and quantitative delivery to brain cell types remain unresolved.

Transport experiments also require attention to labeling, dose, and purity. Lipophilic dyes such as PKH26 and DiR can yield signals from free dye or dye-derived particles, while labeled cargo can separate from vesicles. Dye-only controls, removal of free label, independent membrane and cargo reporters, and cellular or ultrastructural localization strengthen interpretation^[43]. Experimentally concentrated preparations may not reproduce endogenous exposure, which is poorly quantified. Mixed vesicle subtypes and co-isolated material further limit attribution to intact OMVs. These limitations preclude equating brain-associated signal with verified vesicle entry into parenchyma.

The vagus nerve may transmit BEV-induced signals from the gut without requiring BBB crossing. Evidence for this mechanism rests largely on a study of Paenalcaligenes hominis EVs in mice, in which bilateral subdiaphragmatic vagotomy attenuated cognitive impairment^[77]. Vagotomy also changes gut motility, immune regulation, and neuroendocrine signaling. The result therefore supports vagal involvement but does not demonstrate retrograde axonal transport of intact BEVs. Helicobacter pylori OMVs have been linked to AD-related pathology through complement component 3-C3a receptor (C3-C3aR) signaling, with hematogenous and vagal mechanisms proposed^[95]. Gastric infection studies provide additional, indirect support for gut-to-brain inflammatory signaling^[96]. The relative contributions of intact vesicles, released cargo, circulating mediators, and neural activation remain unresolved.

Akkermansia muciniphila-derived extracellular vesicles (AmEVs) are among the best-studied commensal BEVs with reported neuroprotective effects. In a postoperative cognitive dysfunction model induced by intestinal ischemia-reperfusion, AmEVs suppressed microglial activation through TLR2/4-associated signaling. Pharmacological microglial depletion abolished their reported protection of BBB integrity^[97]. Earlier work showed improved intestinal barrier function and increased tight junction protein expression after AmEV treatment^[98]. In a colitis-associated cognitive impairment model, AmEVs improved intestinal and BBB integrity, reduced hippocampal inflammation, and improved behavioral performance. They reduced excessive colonic serotonin while restoring hippocampal serotonin levels, indicating tissue- and disease-dependent regulation^[24]. Garlic exosome-like nanoparticle exposure also altered the activity of A. muciniphila-derived OMVs, which improved high-fat-diet-associated metabolic dysfunction in mice^[99]. This metabolic study should not be interpreted as direct evidence of cognitive benefit. In a smoking-exposure model, treatment with AmEVs or exogenous indole-3-lactic acid improved cognitive outcomes in association with AhR-dependent microglial regulation^[100]. This does not by itself establish delivery of indole-3-lactic acid within intact AmEVs.

Neuroprotective effects are not restricted to Akkermansia. Lactiplantibacillus plantarum (formerly Lactobacillus plantarum)-derived EV treatment reduced ischemic neuronal injury and increased host microRNA-101a-3p (miR-101a-3p), implicating a c-Fos/transforming growth factor beta (TGF-β) pathway^[101]. These findings support modulation of host microRNA signaling; they do not demonstrate delivery of bacterial vesicle-encapsulated microRNAs.

Oral gavage of P. hominis EVs induced cognitive impairment and colitis in mice. Subdiaphragmatic vagotomy attenuated the cognitive deficits without significantly reducing colitis^[77]. Hippocampal fluorescein isothiocyanate (FITC) signals were more frequently detected after labeled EVs than labeled LPS, and vagotomy reduced EV-associated signals. However, the investigators did not detect intact EVs in the hippocampus by transmission electron microscopy. Label-associated signals therefore support exposure but do not establish intact vesicle transport. FITC can label molecules that later become detached or released; it should not be conflated with the micelle-forming behavior of lipophilic dyes. A structural study identified penta- and tetra-acylated lipid A in P. hominis LPS and found weaker TLR-dependent activity than for E. coli LPS^[102]. This finding motivates tests of non-LPS cargo but does not establish its contribution to the neurological phenotype.

P. gingivalis OMV administration has been associated with hippocampal and cortical signals, reduced tight junction protein expression, glial activation, increased interleukin-1 beta (IL-1β), tau phosphorylation, and NLRP3 inflammasome activation in mice^[81]. A broader review summarizes the systemic effects of P. gingivalis OMVs^[103]. These findings support model-specific neuropathological effects but do not establish a vesicular cause of human AD. OMV-mediated gut-brain signaling has also been proposed as a contributor to PD-associated inflammation; direct causal evidence remains limited^[104].

Moreover, two complementary studies implicate gut-derived BEVs in AD-related pathology. In germ-free App^NL-G-F mice, fecal BEVs from specific pathogen-free (SPF)-housed donors increased amyloid-β (Aβ) plaque burden and influenced microglial Aβ uptake and synaptic parameters^[83]. In amyloid precursor protein/presenilin 1 (APP/PS1) mice, prolonged intragastric administration of AD patient-derived fecal OMVs impaired cognition, enhanced neuroinflammation and Aβ production, and disrupted intestinal barrier integrity and short-chain fatty acid homeostasis^[84]. These findings suggest that gut-derived BEVs preparations can influence AD-related pathological processes independently of administered live bacteria.

Several gaps limit causal interpretation. Many experiments use purified, concentrated preparations, and the corresponding endogenous exposure is often unknown. Strain, cargo, isolation method, and host state can alter the outcome. Preparations may contain several vesicle subtypes, free cargo, and co-isolated bacterial material^[43,105]. Evidence that BEVs administration changes CNS phenotypes does not establish intact vesicle entry into the brain. Peripheral immune responses, altered barrier function, and neural signaling may contribute. Defined producer strains, appropriate nonvesicular controls, quantitative cargo measurements, and direct tracking of intact vesicles are needed to resolve these alternatives.