A new horizon of precision medicine: combination of the microbiome and extracellular vesicles

Over several decades, the disease pattern of intractable disease has changed from acute infection to chronic disease accompanied by immune and metabolic dysfunction. In addition, scientific evidence has shown that humans are holobionts; of the DNA in humans, 1% is derived from the human genome, and 99% is derived from microbial genomes (the microbiome). Extracellular vesicles (EVs) are lipid bilayer-delimited nanoparticles and key messengers in cell-to-cell communication. Many publications indicate that microbial EVs are both positively and negatively involved in the pathogenesis of various intractable diseases, including inflammatory diseases, metabolic disorders, and cancers. Microbial EVs in feces, blood, and urine show significant differences in their profiles between patients with a particular disease and healthy subjects, demonstrating the potential of microbial EVs as biomarkers for disease diagnosis, especially for assessing disease risk. Furthermore, microbial EV therapy offers a variety of advantages over live biotherapeutics and human cell EV (or exosome) therapy for the treatment of intractable diseases. In summary, microbial EVs are a new tool in medicine, and microbial EV technology might provide us with innovative diagnostic and therapeutic solutions in precision medicine.


INTRODUCTION
The approaches to precision medicine defined by the Precision Medicine Initiative 1 and National Research Council 2 allow doctors and researchers to predict more accurately which treatments and preventive strategies will work in patients with a particular disease. These definitions indicate that precision medicine can guide the most effective health care decisions for a given patient and thus provide the best quality of therapy while reducing unnecessary medical interventions 3 . Precision medicine enables the individualization of patient care from diagnosis to therapy selection and monitoring using a patient's own data in engineering technology to identify individualized treatment strategies using population-wide data 4 .
With the genomic revolution of the 21st century came the ability to rapidly sequence entire genomes in a matter of days at increasingly more affordable costs. In addition, artificial intelligence (AI) platforms have been used to optimize combination therapy in preclinical and clinical trials 4 . Thus, the prospect of deciphering an individual's unique genome as a roadmap for precision medicine has become a tantalizing goal for scientists and physicians around the world. However, several challenges have been encountered when applying human genomic strategies in precision medicine and in translation from the bench to the clinic. To address these issues, technological improvements, including long-read sequencing methods, algorithms for indel and structural variants, graph reference approaches, and standardization of nomenclature, have been suggested 4 . The variety of next-generation methods through which precision medicine solutions have been crafted include genetic sequencing, proteomic analysis, and AI strategies. The genomic analysis ultimately remains the core foundation of most precision medicine applications.
The human genome and human microbiome projects have revealed that microbial protein-encoding genes are 360 times more abundant than human genes 5 . In 1991, Lynn and Fester 6 defined the term "holobiont", which is derived from Ancient Greek hólos ("whole") and the word biont for a unit of life. In 1994, Jefferson 7 defined the term "hologenome" when he introduced the hologenome theory of evolution at a plenary lecture at Cold Spring Harbor Laboratory. The hologenome theory of evolution reboots elements of Lamarckian evolution and recasts the individual human as a community or a holobiont that is the host plus all of its symbiotic microbiota species 8,9 . Variation in the hologenome may encode phenotypic plasticity of the holobiont and can be subject to evolutionary changes caused by selection and drift if portions of the hologenome are transmitted between generations with reasonable fidelity. One of the important outcomes of recasting the individual as a holobiont subject to evolutionary forces is that genetic variation in the hologenome can be brought about by changes in the host genome and microbiome, including new acquisitions of microbes, horizontal gene transfers, and changes in microbial abundance within hosts. Thus, humans might be holobionts, with 99% of their DNA attributed to microbial genomes (the microbiome) and 1% of their DNA attributed to the human genome, according to the hologenome theory of evolution.
Extracellular vesicles (EVs) are lipid bilayer-delimited particles that are naturally released from almost all types of cells and cannot replicate. The vast majority of EVs are smaller than 200 nm. EVs, according to the biogenesis route, are divided into exosomes, ectosomes, and apoptotic bodies 10 . EVs carry cargoes of proteins, nucleic acids, lipids, metabolites, and even organelles from parent cells. Most cells that have been studied to date are thought to release EVs, including archaeal, bacterial, fungal, plant, and human cells 11 . Microbial EVs (MEVs) appear to be key messengers in host cell−microbiota communication. Recent experimental evidence shows that MEVs are key messengers in the communication between host cells and the microbiota. In this review, we will address MEVs as a new factor in precision medicine, focusing on 1) MEV biogenesis and functions, 2) the importance of MEVs in disease pathogenesis, and 3) the application of MEVs in the development of diagnostic and therapeutic solutions for precision medicine.

MICROBIAL EV BIOGENESIS
The biogenesis of EVs is a very tightly regulated process governed by multiple signaling molecules and begins with receptor activation that is unique for each cell type 12 . Eukaryotic EV biogenesis is well characterized, whereas MEV biogenesis has only recently been elucidated 13 . According to the method of biogenesis, eukaryotic cell-derived EVs are classified as exosomes, ectosomes (or shedding vesicles), and apoptotic bodies 14,15 , while prokaryotic cell-derived EVs are classified as ectosomes and apoptotic bodies 16 (Fig. 1). Both gram-negative and gram-positive bacteria produce ectosomes, known as outer membrane (OM) vesicles (OMVs) and membrane vesicles (MVs), respectively 16 . The key milestones of bacterial EVs are as follows (Fig. 2): (1) gramnegative bacteria-derived OMVs were first discovered in 1966 17 , (2) the fact that bacterial EVs contain nucleic acids was reported in 1989 18 , (3) gram-positive bacteria-derived EVs were first included in a publication in 2009 19 , and (4) since 2010, much evidence has shown that both gram-negative and gram-positive bacteria release EVs that are positively or negatively involved in disease pathogenesis.
When OMVs are released from gram-negative bacteria during their proliferation, the OM bulges outwardly, and the connection between the OM and peptidoglycans is broken. The next steps are an accumulation of peptidoglycan fragments and misfolded proteins in the periplasmic space and enrichment of speciesspecific membrane curvature-inducing molecules 16,20 . However, this mechanism cannot sufficiently explain OMV biogenesis. Thus, new mechanisms for OMV formation have been reported. Several studies have shown that overexpressed and misfolded periplasmic proteins can cause budding, and specific proteins have been found to be enriched in or excluded from OMVs, suggesting a more sophisticated biogenesis pathway. The accumulation of periplasmic proteins, as well as the accumulation of curvatureinducing OM proteins, induces additional budding events 21,22 . Another study showed that the VacJ/Yrb ATP-binding cassette transport system was involved in OMV formation. In that report, a decrease in or deletion of the VacJ or Yrb genes was found to result in phospholipid accumulation in the outer leaflet of the OM. This asymmetric expansion of the outer leaflet then caused an outward bulging of the OM. Then, further enrichment of the positive and negative curvature-inducing phospholipids in both leaflets supported the budding of the OM. Ultimately, the released OMV was enriched in phospholipids incorporated into the outer leaflet of the vesicle membrane 20,23 .
Gram-positive bacteria lack an OM and have a much thicker peptidoglycan cell wall. Thus, until 2009, it was assumed that gram-positive bacteria could not release EVs. However, recent experimental evidence has shown that gram-positive bacteria release MVs 19,24,25 . The biogenesis mechanisms of MVs during the proliferation of gram-positive bacteria have not been as clearly elucidated compared to those of gram-negative bacterial ectosome biogenesis. One hypothesis is that MVs may be forced through the cell wall by turgor pressure after their release from the plasma membrane 26 . Another hypothesis is that cell walls loosened by enzymes may enable the release of MVs. For example, Staphylococcus aureus (S. aureus)-derived MVs carry peptidoglycan-degrading enzymes, which can modify thick gram-positive peptidoglycan cell walls 19 . Thus, membrane fluidity and cell wall integrity, affected by peptidoglycan cross-linking, might be critical for the release of MVs 25 . The third hypothesis is that EVs may pass through channels such as protein channels or structural cables from the intracellular to the extracellular environment 26 .
Regarding the biogenesis of apoptotic bodies, much evidence has revealed that bacterial cell apoptosis is induced in response to environmental stress, biofilm formation, and genetic transformation for the benefit of the entire colony rather than the individual cell 27 . Once the programmed cell death signaling pathways are activated, DNA damage and fragmentation are induced, and subsequently, cell membrane blebs, are induced, resulting in cell fragmentation and the release of apoptotic bodies 15 .

Microbial EV components and physiological functions
Nanometer-sized MEV particles contain a variety of functional components, including lipids, proteins, nucleic acids, and metabolites, within a spherical phospholipid bilayer and play a key role in cell-to-cell communication 10,16,21,26 .
Membrane components: lipids, carbohydrates, and proteins. Lipids in MEVs can play roles as drug delivery vehicles. The bilayer lipids of OMVs mainly contain lipopolysaccharide (LPS) and OM lipids 28 . Liposomal vesicles can be incorporated into cultured mammalian cells with no cytotoxic effects. The variety of glycolipids and glycoproteins in liposomal vesicles suggests the potential to modify the cellular composition and introduce biologically active molecules into cells 16,29 . In addition, the asymmetric distribution of EV lipids suggests that EVs could be synthesized in lipid domains, which could potentiate a lipoprotein-dependent effect on cargo loading 25 . For example, lipid domains, such as functional membrane microdomains, were found to exist in the membranes of Bacillus subtilis 30,31 . Therefore, EVs have been considered advantageous as pharmacological vehicles.
Lipids, carbohydrates, and proteins in the MEV membrane are crucial for specific targeting. In particular, the physical state of the lipids in EVs is important for determining the pathways of EVs 16 . The EV lipid bilayer is adapted to different target environment conditions, such as pH, for optimal function 32 . The carbohydrates in the cell membrane include receptors for glycoproteins, suggesting that the tissue distribution and cellular uptake of phospholipid EVs could be controlled by carbohydrate determinants on the EV surface 33,34 . Modification of the EV surfaces with specific synthetic glycolipids significantly affected the tissue distribution and stability of the EVs in mice 35,36 . Furthermore, the recognition systems of cell membrane carbohydrates include receptors for galactose-terminated glycoproteins in hepatocytes, 6-phosphomannosyl residue-containing glycoproteins in fibroblasts, and mannose-terminated glycoproteins in macrophages 16,[37][38][39] . These membrane glycoprotein receptors are essential for specific targeting. Generally, membrane proteins are essential components for EV functions, such as immunity, targeting, and pathogenicity, and EV proteins can play significant roles in regulatory processes, cellular responses, host−microbe interactions, and targeting 32 . OM proteins have been shown to participate in flagellum assembly, pore formation, transport of specific substrates, OM stabilization, antibiotic resistance, proteolysis, and host bacteria interaction 28 .
Luminal components: proteins, nucleic acids, and metabolites. Proteomics studies have shown that there are a number of proteins in the membrane or luminal space of MEVs, which demonstrate complex protein organization 19,25,40,41 . Most EV membrane proteins originate from the cytoplasmic membrane, while EV luminal proteins are cytoplasmic proteins packaged during vesiculogenesis 25 . For example, S. aureus EVs (SaEVs) were found to have more than 200 proteins associated with EVs, and 160 of these proteins were cytoplasmic proteins 42 . Furthermore, αhemolysin is localized in the lumen of SaEVs 43 . α-Hemolysin in EVs induces necrotic cell death via toxin entry into the cytoplasm of keratinocytes, whereas soluble α-hemolysin induces keratinocyte death. Thus, α-hemolysin in the EV lumen enhances keratinocyte death and the evasion of host immune defenses, illustrating the clinical significance of EVs in luminal protein delivery 43 .
Previous studies have shown that MEVs carry DNA, rRNA, tRNA, mRNA, and sRNA 16,25 . The majority of RNAs in EVs are sRNAs and are noncoding 16,44 . In addition, a large proportion of sRNAs in EVs are from uncharacterized intergenic regions and may also modulate gene expression in target cells 44,45 . However, there is much less evidence regarding the properties and functions of the nucleic acids in MEVs 44,46 . Further studies using next-generation sequencing (NGS) and transcriptome sequencing are needed to ascertain the content and function of MEV nucleic acids 25 . Previous studies suggest that the nucleic acids in MEVs could interfere with the host transcriptome and inhibit protein synthesis 25 .
The nucleic acids in EVs mediate horizontal gene transfer and immunomodulatory effects 16 . To affect gene expression in target cells directly, RNAs in EVs enter the cytoplasm. Once in the cytoplasm, microbial RNAs might interact with or be modified by target cell factors. Complementarity between sRNAs in EVs and RNAs in target cells might induce the attenuation or enhancement of gene expression via transcription, translation, or mRNA processing. The mechanisms involved likely vary greatly depending on the specific RNAs in EVs and the presence of specific target cell factors. mRNAs in EVs might be translated to produce microbial proteins in target cells 44 . Microbial nucleic acids are sensed by either endosomal or cytoplasmic receptors in host cells 47 . Several MEVs trigger pattern recognition receptor signaling to promote protective immunity, a property that has been harnessed for vaccine development 44 .
Recently, it has been reported that EVs are metabolically active 16 . Although emerging evidence suggests that EVs can act as metabolic regulators, the involvement of EVs in metabolic activity and the existence and function of metabolites in MEVs have not been completely characterized 16 . Thus, it is necessary to further investigate the metabolites in MEVs to demonstrate their potential metabolic functions.

Microbial EVs in pathogenesis
Previous studies elucidated the role of MEVs in disease development and the pathogenesis of diseases, including skin, lung, and gut diseases, metabolic diseases, and cancers (Table 1).
Skin diseases. Atopic dermatitis (AD) is a chronic inflammatory disease of the skin and is defined as eczematous lesions with pruritus and xerosis. AD skin lesions show distinct features, such as abnormal skin barrier function with epidermal hyperplasia and S. aureus colonization. Abnormal skin barrier function induced by the death of keratinocytes is one of the major causative factors of AD 43 . Skin inflammation has been associated with SaEVs 43,48 . SaEVs increase IL-6, IL-17, serum IgG1, IgE, thymic stromal lymphopoietin, macrophage inflammatory protein-1α, and eotaxin levels and recruit mast cells and eosinophils in the skin 48 . A previous study on the metagenomics of MEVs showed that Staphylococcus accounted for the highest proportion of bacterial genera in AD patients, and the proportion of Staphylococcus significantly decreased after treatment 49 .
If the skin barrier is disrupted, pathogens can penetrate the human body. Therefore, skin barrier disruption is a major cause of AD exacerbation 43 . α-Hemolysin can induce keratinocyte cell death, consequently enhancing skin penetration of highmolecular-weight allergens. SaEV-associated α-hemolysin was found to induce keratinocyte necrosis and induce epidermal thickening and eosinophilic inflammation in the dermis. SaEVs were internalized into the cytoplasm of keratinocytes, and EVs efficiently delivered α-hemolysin to the cytoplasm. These findings suggest that EV-associated toxins are key molecules in disease pathogenesis 43 . Therefore, SaEVs are useful in the development of in vivo models of AD 49 . Propionibacterium acnes (P. acnes) EVs induce dysregulated epidermal differentiation in human keratinocytes via TLR2-mediated signaling pathways. P. acnes EVs may play pivotal roles in the early stages of the development of acne vulgaris 50 .
Lung diseases. The main causative factor of chronic inflammatory diseases in the lung has been suggested to be exposure to chemical substances such as cigarette smoke and air pollution 51,52 because of the inherent pharmacological or toxic effects of such chemical substances. However, the etiology of a large proportion of chronic inflammatory diseases is not related to these factors. Biological factors, including allergens, viruses, and bacterial factors, can cause severe inflammatory reactions 52 .
Airway exposure to MEVs triggers two main pathophysiological mechanisms, Th17, and Th1 responses, based on whether the parent bacterial cell is extracellular or intracellular, respectively. Specifically, extracellular bacteria-derived EVs generally cause neutrophilic inflammation through IL-17 release by polarized Th17 cells, whereas intracellular MEVs induce mononuclear inflammation through IFN-γ produced by polarized Th1 cells 53 . The pathogenic MEV-induced neutrophilic inflammatory response leads to airway hyperreactivity and fibrosis that contribute to asthma development. The combination of increased elastase production and fibrosis induced by neutrophilic inflammation can cause emphysema, which is a key pathology of chronic obstructive pulmonary disease (COPD). Furthermore, intracellular MEVs lead to Th1 polarization and subsequent IFN-γ-induced mononuclear inflammation, leading to increased elastase production in the alveoli and causing pulmonary tuberculosis-associated emphysema 53 .
In several in vivo tests, MEVs caused severe airway inflammation, which resulted in severe asthma and emphysema 52 . Previous studies have shown that repeated exposure to Escherichia coli (E. coli) EVs (EcEVs) in the airways led to IL-17-dependent neutrophilic inflammation and emphysema in conjunction with elastase upregulation 54 . Repeated airway exposure to SaEVs induced both Th1 and Th17 immune responses with IFN-γ and IL-17, respectively, and neutrophilic pulmonary inflammation was increased primarily through TLR2 engagement 55 . EcEVs and SaEVs induced MUC5AC expression in human airway epithelial cells. MUC5AC, a major mucin, is overexpressed in chronic inflammatory airway diseases and causes airway obstruction, increases susceptibility to infection, and decreases pulmonary function 56 . In addition, Pseudomonas aeruginosa EVs increase pulmonary inflammation through activation of the TLR2 and TLR4 pathways through dose-dependent increases in chemokines (CXCL1 and CCL2) and cytokines (IL-1β, TNF-α, IL-6, and IFN-γ), as well as increases in neutrophils and macrophages 57 . In addition, EVs from Bacteroides ovatus, Bacteroides stercoris, and Prevotella melaninogenica induce IL-17 expression and consequently promote Th17 cell development associated with lung fibrosis 58 .
Indoor dust containing MEVs induces both Th1 and Th17 responses associated with neutrophilic pulmonary inflammation 59,60 and over 50% of the patients with childhood asthma were found to be sensitized to MEVs in indoor dust, suggesting that MEVs in indoor dust could be an important causative factor in the pathogenesis of childhood asthma 59 . Moreover, the antiindoor dust EV IgG antibody was found to be significantly associated with the risk of asthma, COPD, and lung cancer (LC) 61 . These results suggest the potential of MEVs as a major risk factor in the development of inflammatory diseases and cancer in the lung.
Gastrointestinal disease. Inflammatory bowel disease (IBD) is characterized by chronic inflammation of the gastrointestinal tract. Although the exact etiology of IBD is under investigation, disruption of the normal intestinal flora and normal intestinal function have been linked to the development of IBD. Alteration of the commensal gut microbial ecosystem, which interacts with gut epithelial cells and plays significant roles in the maintenance of the mucosal barrier, induces dysfunctional immunomodulation and metabolic activity in the gut 62 . Such disruptions lead to reductions in mucus barrier functionality, including decreased tight junctions in the intestinal epithelial lining. Since the integrity of the mucosal barrier is significantly reduced, inflammatory agents such as microbial components in the intestinal lumen can induce mucosal inflammation associated with IBD 63,64 .
The NOD-like receptor (NLR) family, which is a family of cytoplasmic pattern recognition receptors, initiates inflammation by activating signaling pathways, including NF-κB signaling, followed by induction of caspase-1-mediated cleavage of IL-1β/ IL-18 and production of gasdermin, which causes pyroptotic cell death via the formation of pores in the plasma membrane. NLRs or NLR-associated proteins have been found to be genetically associated with susceptibility to IBD 65 . Vibrio cholerae EVs and EcEVs activate NOD1 signaling 66,67 , resulting in NF-κB signaling and a pro-inflammatory immune response. NOD1 activation by MEVs can lead to IBDs, such as Crohn's disease 68 .
Helicobacter pylori (H. pylori) EVs (HpEVs) may play a significant role in the development of various gastric diseases through IL-8 production and NF-κB activation. HpEVs may play a crucial role in the pathogenesis of gastric disease associated with H. pylori 69 .
Metabolic diseases. Commensal bacteria in the gastrointestinal tract play an important role in nutrient absorption and the fermentation of dietary fibers. In addition, the gut microbiota provides a critical function in the production of short-chain fatty acids (SCFAs) for energy sources and epigenetic regulation via histone deacetylase inhibition 70 . Gut microbiota dysbiosis resulting in irregular SCFA production can lead to dysregulated host metabolic functionality. Moreover, MEVs can participate in the development of metabolic diseases. Pseudomonas panacis EVs promoted by a high-sucrose/fat diet can induce type 2 diabetes (T2D) via the induction of insulin resistance in insulin-responsive organs. Pseudomonas panacis EVs blocked the insulin signaling pathway in both skeletal muscle and adipose tissue, thereby promoting glucose intolerance in skeletal muscle, while these EVs induced typical diabetic phenotype characteristics, such as glucose intolerance after glucose administration or systemic insulin injection 71 .
In one study, Porphyromonas gingivalis EVs equipped with proteases gingipains were translocated to the liver. Subsequently, hepatic glycogen synthesis in response to insulin was decreased, and high blood glucose levels were maintained. Porphyromonas gingivalis EVs also attenuated insulin-induced Akt/GSK-3β signaling in a gingipain-dependent fashion in hepatic HepG2 cells. The delivery of gingipains mediated by Porphyromonas gingivalis EVs elicited changes in glucose metabolism in the liver and contributed to the progression of T2D 72 . These results suggest a close relationship between MEVs and metabolic disease.
Cancers. MEVs are known to be associated with tumors, and cancer patients show different MEVs than healthy people. Previous studies have elucidated that Th17 cells play an important role in pathogen clearance during host defense reactions but induce tissue inflammation in the pathogenesis of autoimmune disease 73 . Recent studies reported that Th17 cells played an important role in LC pathogenesis 74,75 , and the immune effector T cells in smallcell LC patients included a large proportion of Th17 cells 76 . Airway exposure to SaEVs and EcEVs triggers the Th17 response. Generally, OMVs induce IL-17 production through polarized Th17 cells and thus cause neutrophilic inflammation. The neutrophilic inflammatory response induces epithelial cell dysplasia and the induction of matrix metalloproteinase expression, possibly leading to LC. COPD also increases the risk of LC 53 .
The abundances of H. pylori and HpEVs in gastric cancer patients were significantly higher than those in healthy people. HpEVs contain CagA and VacA proteins that can induce the production of TNF-α, IL-6, IL-1β, and IL-8. In addition, HpEVs induce the expression of IFN-γ, IL-17, and EV-specific IgGs in mice. In summary, HpEVs might be involved in the pathogenesis of gastric cancer, mainly via the induction of the production of inflammatory mediators by gastric epithelial cells after selective uptake by the cells 77 .
Central nervous system (CNS) diseases. MEVs can penetrate the blood-brain barrier (BBB) through three possible mechanisms: (1) receptor-mediated transcytosis, (2) paracellular passage in disease states, or (3) via EVs in infected immune cells. Once MEVs are inside brain tissues after penetrating the BBB, the components and cargo in EVs act as ligands of innate immune receptors, such as TLRs, and the NALP3 inflammasome, and activate the inflammatory immune response 78 .
A previous study showed that Aggregatibacter actinomycetemcomitans EVs can cross the mouse BBB and are present in the brain 24 h after intracardiac injection, inducing TNF-α production via extracellular RNAs through the TLR-8 and NF-κB signaling pathways 79 . In the mouse brain, increased TNF-α may cause neuroinflammatory diseases, such as Alzheimer's disease 79 . Similarly, Aggregatibacter actinomycetemcomitans EVs can successfully deliver extracellular RNAs to brain monocytes/microglial cells to cause neuroinflammation associated with the upregulation of IL-6 through NF-κB activation 80 . Oral administration of Paenalcaligenes hominis EVs increased bacterial 16 S rDNA levels in the hippocampus, causing cognitive impairment in mice, whereas oral administration of LPS caused more severe cognitive impairment. Paenalcaligenes hominis EV treatment increased NF-κB+/Iba1+, LPS+/Iba1+, and IL-1R+ cell counts and reduced BDNF expression in the hippocampus while increasing IL-1β expression in the blood 81 . These findings provide evidence that some MEVs are important in the pathogenesis of neuroinflammatory diseases.
Recently, we performed microbiome analysis of EVs in blood and brain tissue samples from brain tumor (BT) patients and healthy controls (HCs) 82 . Blood EV microbiome analysis at the phylum level showed that the abundance of Firmicutes was significantly higher in BT patients than in HCs, whereas those of Proteobacteria and Actinobacteria were significantly lower. In addition, at the genus level, the proportions of EVs from Ruminococcaceae UCG-014, Lachnospiraceae NK4A136, Lactobacillus, Ruminococcaceae UCG-013, Ruminiclostridium 6, and Peptoclostridium were significantly higher but those from Escherichia-Shigella, Blautia, Bifidobacterium, Streptococcus, and Sphingomonas were significantly lower in BT patients than in HCs. Moreover, brain tissue microbiome analysis at the genus level showed that the abundance of EVs from Bacteroides and Erysipelatoclostridium was significantly higher but those from Chloroplast(c), Prevotella 9, and Saccharibacteria(p) were significantly lower in BT patients than in HCs 82 . These findings suggest that MEVs might be closely linked to the pathogenesis of CNS diseases, although further research is needed.

Microbial EVs as biomarkers
Liquid biopsy, also known as fluid biopsy or fluid phase biopsy, is the sampling and analysis of nonsolid biological tissue. MEVs are a newly emerging biomarker that can be attested for by liquid biopsy. Moreover, AI-based analysis of clinical and MEV data enables the development of a more accurate prediction of disease status (Fig. 3).
Commensal microbes secrete MEVs for cell-to-cell communication. MEVs penetrate the mucosal barrier, circulate throughout the body, and then are excreted via feces, urine, and exhaled air in their intact forms, unlike live microbes, which are restricted to the mucosal lumen or skin surface 71 . Certain MEVs act as etiological agents of various diseases, while some MEVs have a protective role in disease pathogenesis. Thus, MEVs circulating in our body provide us with good information for health and disease status.
An organism's evolutionary history is documented in its genome. rDNA encoding rRNA changes slowly and is useful for investigating relationships between taxa that diverged hundreds of millions of years ago. In contrast, mitochondrial DNA evolves rapidly and is more useful for exploring more recent evolutionary events. MEVs harbor 16S rDNA, which provides taxonomy information 83 . However, there are some issues in comparing published microbiome data. First, the regions targeted during the investigation of the microbiota are different for the 16S rDNA and NGS platforms. The diversity and abundance of the bacterial mock community are different between the V1-V3 region and V3-V4 region of the 16S rRNA or MiSeq platform (Illumina, USA) and the Roche 454 platform (Roche, Switzerland) 84 . Second, microbiome information differs among the databases used for taxonomic assignments. There are several public databases, such as SILVA, RDP, Greengenes, EzTaxon, and NCBI, and these databases have different taxonomic classifications 85 . Third, these public databases are updated continuously, and unassigned taxa have the potential to be assigned in the future.
Disease diagnosis and prediction goals, such as risk assessment, early diagnosis, treatment response prediction, and disease monitoring, are important for the reduction of mortality and the increase of quality of life 86,87 . MEVs are novel biomarkers for in vitro diagnostics. Several 16S rDNA amplicon studies using serum, urine, feces, and skin-fluid samples have shown significant differences in MEV composition between healthy and patient groups ( Table 2). The clinical performance of the diagnostic models as evaluated by area under the curve (AUC) analysis using external validation set was 0.97, 1.00, 0.95, 0.93, 0.87, 0.82, 1.00, and 1.00 for asthma 88 94 , respectively. Furthermore, diagnostic models were developed to predict the prognosis of hepatic diseases 91 . MEVbased models, in combination with microbiome and metabolome data, provided the best performance; the performance of colorectal cancer models using both metagenomic and metabolomic data was better than that of models using microbiome or metabolome data alone 90 . Moreover, EV microbiome data provided better performance when combined with tumor markers; the diagnostic performance in ovarian cancer was better using both MEV data and the tumor marker CA-125 than either biomarker alone 95 .
The presence of antibodies against MEVs can also provide good information for the diagnosis of disease. The presence of IgG antibodies against EVs in indoor dust is closely associated with asthma, COPD, and LC 61 . The levels of IgGs against pathogenic MEVs, such as EVs from S. aureus, Acinetobacter baumannii, Enterobacter cloacae, and Pseudomonas aeruginosa abundant in indoor dust, showed good AUC values: 0.78 for asthma, 0.79 for COPD, and 0.81 for LC. Thus, MEV-based diagnostic technologies, including MEV composition assessment and immunoassays to assess MEVs, can provide information on exposure to etiological agents 96 .

Therapeutic potential of microbial EVs
That commensal bacteria play an essential role in human health is beyond doubt, and it is now widely accepted that humans are holobionts or supra-organisms whose collective metabolic potential exceeds the sum of the individual eukaryotic and prokaryotic components. Probiotics or live biotherapeutics (LBPs) are live microorganisms that, when administered in adequate amounts, confer a health benefit on the host and are widely used in many countries in clinical practice. Paraprobiotics are currently referred to as "modified probiotics", "inactivated probiotics", "nonviable probiotics", or "ghost probiotics" [97][98][99] . Paraprobiotics are a newly emerging modality, an immobilized version of probiotics that has gained traction in recent years due to concerns about the possibility of low tolerance of probiotics or LBPs, especially in pediatric populations and severely ill or immunocompromised patients 100 . The mechanism of action (MoA) of LBPs and parabiotics is not well understood, although the possible MoAs include immune system regulation and interference with pathogen attachment to host cells. Limited research has found that immobilized parabiotics release key bacterial components, including lipoteichoic acids, peptidoglycans, and exopolysaccharides, that exhibit key immunomodulatory effects and antagonizing effects against pathogens 100 .
Postbiotics, also known as metabiotics, biogenics, or simply metabolites, are soluble factors secreted by live microbes or released after microbial lysis, and they provide physiological benefits to the host 101 . Postbiotics are a newly emerging modality in therapy. Cell-to-cell communication is tightly regulated, and its disruption contributes to disease development. Intercellular communication can occur either via cell-to-cell contact or via soluble factors 102 . Soluble factors include proteins and small molecules that bind specific receptors on target cells, which then trigger a signaling cascade that will affect the cell phenotype. However, more recently, another process also involved in this type of communication has been highlighted. In addition to producing specific factors, microbial cells can send a package of information that is enclosed by a cell membrane, also called MEVs. Moreover, recent scientific evidence has shown that certain MEVs have protective effects against disease development or progression 16,[103][104][105] . Thus, MEVs belong to the category of postbiotics and are a new modality for next-generation therapy to complement current therapeutics, such as small molecules, proteins, monoclonal antibodies, and cell therapeutics (Table 3).
Therapy for immune and inflammatory diseases. Akkermansia muciniphila (A. muciniphila) is considered a beneficial anaerobic bacterium living in the large bowel. A previous study 104 showed that A. muciniphila EVs (AmEVs) were decreased in dextraninduced colitis mice vs. healthy mice. In vitro pretreatment with AmEVs improved the production of the proinflammatory cytokine IL-6 by EcEV-stimulated intestinal epithelial cells. In addition, AmEVs protected against dextran-induced IBD phenotypes, including bodyweight loss, decreased colon length, and inflammatory cell infiltration of the colon wall. Oral therapy with AmEVs protected against LPS-induced gut barrier disruption, and this beneficial effect was mediated by the AMPK signaling pathway 106 .
Lactobacillus EVs have the possibility to be developed as a medicine for chronic inflammatory diseases such as IBD and AD 105,107 . Lactobacillus plantarum (L. plantarum) EVs (LpEVs) have anti-inflammatory effects and prevent SaEV-induced cell death 107 . Oral administration of LpEVs has a therapeutic effect against SaEVinduced atopic dermatitis 107 . LpEVs promoted the differentiation of human monocytic THP1 cells toward an anti-inflammatory M2 Fig. 3 Application of microbial EVs for precision medicine. Precision medicine goals, including risk assessment, screening, prediction, and monitoring, are based on analysis using an artificial intelligence (AI) algorithm through human microbial EV big data and clinical data. Highrisk individuals and diseased individuals can be identified by risk assessment, and preventive therapy can reduce disease risk. Diseased individuals identified by screening and diagnosis are suggested therapeutics based on the results of response prediction and monitoring. These processes can realize microbial EV-based health care solutions.    Choi et al. 105 Lactobacillus kefir (L. kefir), Lactobacillus kefiranofaciens (L. kefiranofaciens), Lactobacillus kefirgranum (L. kefirgranum)

Inflammatory bowel disease
Treatment of TNF-α-stimulated Caco-2 cells with each kefirderived Lactobacillus EV type (L. kefir, L. kefiranofaciens, and L. kefirgranum) reduced both mRNA expression and the IL-8 level in association with inhibition of TNF-α signaling by reducing the phosphorylation of p65, a subunit of NF-κB. Subsequent administration of kefir-derived Lactobacillus EVs into IBD mice alleviated bodyweight loss and rectal bleeding and enhanced stool consistency. Histological examination showed that kefir-derived Lactobacillus EVs substantially reduced the infiltration of transmural leukocytes and loss of goblet cells within the colon, and the serum level of myeloperoxidase was significantly lower in the EV-treated group than control group.

Seo et al. 109
Lactobacillus sakei (L. sakei) Immune system disease IgA production in Peyer's patch cells was enhanced by L.  108 . Furthermore, LpEV treatment significantly induced IL-1β, GM-CSF, and IL-10 expression in human skin organ cultures 108 . In the case of IBD, the administration of Lactobacillus paracasei EVs reduced weight loss, maintained colon length, decreased the disease activity index, and protected against dextran-induced IBD 105 . Lactobacillus paracasei EVs inhibited signaling via NF-kB, which is a major transcription factor inducing expression of the inflammatory cytokines IL-6 and TNF-α, and the expression of inducible nitric oxide synthase and cyclooxygenase-2, which is mediated by the unfolded protein response of the endoplasmic reticulum (ER) (UPR ER ) against ER stress 105 . Moreover, EVs derived from Lactobacillus kefir, Lactobacillus kefiranofaciens, and Lactobacillus kefirgranum inhibited the production of IL-8 from TNF-α-activated Caco-2 cells, reduced rectal bleeding, and improved stool consistency in IBD mice 109 . In addition, it was reported that Lactobacillus sakei EVs play crucial roles in the immune system through enhanced IgA production in Peyer's patch cells 110 .
Bifidobacterium longum EVs were reported to alleviate food allergy symptoms mainly by inducing apoptosis of mast cells in the small intestine 111 . In terms of the components in Bifidobacterium longum EVs participating in mast cell apoptosis, extracellular solute-binding protein in the EVs specifically reduced the number of mast cells and alleviated diarrhea in mice with food allergies.
Therapy for metabolic diseases. A. muciniphila LBP and parabiotics (or lysates) have been shown to be beneficial in reducing fat mass and glycemia in T2D patients. The benefits involved the interaction of the A. muciniphila OM protein Amuc_1100 with host cell TLR2 112 . Interestingly, it was shown that AmEVs were decreased in T2D patients vs. HCs 113 . Moreover, oral therapy with AmEVs had a therapeutic effect against metabolic changes, such as the obesity and glucose intolerance induced by a high-fat diet. These metabolic benefits are likely linked to the beneficial effect of AmEVs on gut barrier integrity, which is mainly mediated by AMPK signlaing 113 . Taken together, the beneficial effects of A. muciniphila on metabolic phenotypes might be mediated by AmEVs.
Immunomodulatory therapy for malignant diseases. Cancer immunotherapies such as immune checkpoint inhibitors and chimeric antigen receptor therapies are now being recognized as a promising approach to overcome various malignant diseases 114 . MEVs are nanosized particles that can easily penetrate the gut mucosal barrier and readily interact with immune cells in the gut, giving them great therapeutic potential in the immunology −oncology area. A previous study reported that systemic administration of genetically modified EcEVs significantly and effectively induced antitumor immune responses and fully eradicated established tumors without notable adverse effects 115 . Furthermore, that study showed that systematically administered EcEVs targeted and accumulated in the tumor tissue and subsequently induced the production of the antitumor cytokines CXCL10 and IFN-γ. Moreover, the study found that IFN-γ-deficient mice did not induce such an EV-mediated immune response, suggesting that this antitumor effect is dependent on IFN-γ.
Therapy for CNS diseases. Recent evidence indicates that the gut microbiota plays a key role in regulating mental disorders, but the mechanisms by which the gut microbiota regulates brain function are controversial. L. plantarum is a gram-positive bacterium present in dairy, fermented vegetables, and the gut 116 . Recent studies have reported the beneficial effects of L. plantarum on stress-induced behavioral dysfunction 117 ; for example, it improves cognitive deficits 118 , attenuates anxiety behavior 119 , and reduces oxidative stress markers 120 . In addition, we investigated whether LpEVs have a therapeutic effect against stress-induced depression 117 . The study showed that LpEVs protected against and reversed the depressive behaviors induced by psychological stress. The beneficial effect of LpEVs was linked to the overexpression of neurotrophic factor BDNF and the Nt4/5 and Sirt1 genes, which are decreased by the stress hormone glucocorticoid 117 . Collectively, these findings suggest that secreted L. plantarum EVs show highly beneficial effects and LpEVs can be used to develop novel therapies that penetrate the BBB and modulate CNS function.

CONCLUDING REMARKS
We reviewed MEVs as a new tool for precision medicine, especially focusing on the role of MEVs in disease pathogenesis and as biomarkers and MEV therapeutic potential. The pathogenesis of currently intractable diseases is related to cellular aging and increased reactive oxygen species (ROS). Intracellular ROS cause loss of proteostasis, mitochondrial dysfunction, genome instability, and telomerase exhaustion, thereby leading to aging-related Table 4. Comparison of microbial EV therapy vs. live biotherapeutics (LBP) and human cell EV (exosome) therapies.

Microbial EVs
Live biotherapeutics (LBP) Human cell EVs (Exosomes) diseases, such as immune, metabolic, and neurodegenerative diseases and cancers. According to changes in disease patterns, medical needs have shifted from caring for sick patients to promoting health, from invasive to noninvasive diagnosis methods, and from toxic or expensive drugs to safe drugs with reasonable costs. To meet medical needs, advances in biological data and AI technologies that enable disease prediction and tailored therapy are needed in the near future. Despite intense microbiome research, there is still much to learn about the intrinsic mechanisms of microbes, especially those of effector microbial products exchanged between the microbiota and host. An exact understanding of microbiota-to-host communication is essential for developing translation strategies employing microbial products for human health. In this context, the study of MEVs is receiving increasing interest. Evidence from nonclinical studies provides us with an opportunity to develop novel first-inclass therapeutic modalities: microbiome-related nanotherapeutics.
MEVs are commensal nonliving materials that act on the human immune system to induce protective responses. Studies have shown that MEVs are absorbed into and distributed in various organs after oral administration, unlike exosomes (human cellderived EVs). Furthermore, MEVs have a unique ability to target distant organs and effectively penetrate cellular organelles compared with LBPs (live microbes), which are generally restricted to the mucosal or skin surface. Moreover, MEVs can be conveniently administered orally or topically, as opposed to biologics requiring intravenous infusion or subcutaneous injection. LBPs, exosomes, and MEVs are considered to be promising next-generation therapies to complement or replace current biologics, such as protein, monoclonal antibody, cell, and gene therapies. Among the candidates for next-generation therapies, MEVs offer a variety of advantages over LBPs and exosomes in terms of pharmacology (pharmacokinetics and pharmacodynamics), safety, and manufacturing, as summarized in Table 4.
The study of MEVs is rapidly evolving. In the near future, it is expected that research efforts in this area will contribute to the understanding of the role of MEVs in disease pathogenesis, which will provide an essential, solid basis for precision medicine solutions, such as in vitro diagnostics based on circulating MEV profiling and therapeutics based on EVs from beneficial bacteria as postbiotics.