Outer membrane vesicles secreted by pathogenic and nonpathogenic Bacteroides fragilis represent different metabolic activities

Numerous studies are devoted to the intestinal microbiota and intercellular communication maintaining homeostasis. In this regard, vesicles secreted by bacteria represent one of the most popular topics for research. For example, the outer membrane vesicles (OMVs) of Bacteroides fragilis play an important nutritional role with respect to other microorganisms and promote anti-inflammatory effects on immune cells. However, toxigenic B. fragilis (ETBF) contributes to bowel disease, even causing colon cancer. If nontoxigenic B. fragilis (NTBF) vesicles exert a beneficial effect on the intestine, it is likely that ETBF vesicles can be utilized for potential pathogenic implementation. To confirm this possibility, we performed comparative proteomic HPLC-MS/MS analysis of vesicles isolated from ETBF and NTBF. Furthermore, we performed, for the first time, HPLC-MS/MS and GS-MS comparative metabolomic analysis for the vesicles isolated from both strains with subsequent reconstruction of the vesicle metabolic pathways. We utilized fluxomic experiments to validate the reconstructed biochemical reaction activities and finally observed considerable difference in the vesicle proteome and metabolome profiles. Compared with NTBF OMVs, metabolic activity of ETBF OMVs provides their similarity to micro reactors that are likely to be used for long-term persistence and implementing pathogenic potential in the host.


Results
Comprehensive proteome analysis of NTBF and ETBF vesicles. For both B. fragilis strains (ETBF and NTBF), full genome sequencing was performed. Detailed information about the alignment of the two sequenced genomes (BOB25 = ETBF and JIM10 = NTBF) can be found in Supplementary Information, Table S1-1, S1-3 and Figure S1. The procedure for vesicle isolation, purification, counting of the vesicle number and size distribution can also be found in the Supplementary Information and Figure S2.
Isolated and purified vesicles, as well as bacterial cells, of both strains independently were analyzed by HPLC-MS/MS. Analysis of the proteome of bacterial cells was performed as the combined proteomes of the membrane and cytoplasm cell fractions. SDS-PAGE analysis was provided for all samples ( Figure S3). HPLC-MS/ MS analysis of ETBF and NTBF OMVs was performed for three independent biological repeats with three technical repeats for each sample. HPLC-MS/MS analysis of ETBF and NTBF cells fractions was performed for two independent biological repeats with three technical repeats for each sample As a result, 49 HPLC-MS/MS runs were carried out during comprehensive proteome analysis of NTBF and ETBF vesicles and cells, and 2,430,656 MS/MS spectra were obtained. In total, 17,644 unique peptides were identified against the database constructed by B. fragilis BOB25 proteins (ProtDB). Additionally, 1,465 proteins were identified by 2 or more peptides across all experiments (cells and vesicles). Score thresholds and FDR are presented in Table S2. The complete list of the identified peptides and proteins are provided in Tables S3-1-S3-7 and S4, respectively. These tables contain information about the peptide sequence, PTM, PSM score, charge, m/z value, protein sequence coverage and number of identified unique peptides per protein.
As a result, 823 proteins were identified in both ETBF and NTBF OMV preparations (Table S5). For further analysis, we used only 431 proteins, which were identified in at least two of the three independent preparations of B. fragilis OMVs. For ETBF vesicles, we identified 392 proteins, and for NTBF vesicles, we identified 291 proteins. To predict the subcellular localizations of the proteins isolated from ETBF and NTBF OMV, PSORTb 3.0 was used (Fig. 1A) 36 .
Proteins identified in ETBF and NTBF OMVs were mostly similar (Table S5). Among the detected outer membrane proteins, we identified the group of proteins named TonB proteins. TonB-dependent receptors and TonB-linked outer membrane proteins are the part of the transporter complexes including these outer membrane proteins that are likely to import large degradation products of proteins (e.g., RagA) or carbohydrates (e.g., SusC) as nutrients 6,37 .
As expected, we identified 19 periplasmic proteins as common for the two types of vesicles, and most of them were hydrolases. 5 identified periplasmic proteins were unique for ETBF OMVs and 2 were observed only in NTBF OMVs including 2′, 3′-cyclic nucleotide 2′-phosphodiesterase and alkaline phosphatase. Among the cytoplasmic membrane proteins of ETBF OMV, preprotein translocase subunit SecD and succinate dehydrogenase flavoprotein subunit are involved in the citrate cycle 38 . Cell division protein FtsH observed in ETBF OMV is an ATP-and Zn 2+ -dependent metalloprotease, which is anchored to the cytoplasmic membrane via two transmembrane segments in such a way that the very short amino-and the long carboxy termini are exposed to the cytoplasm 39 . For NTBF vesicles, we found only one cytoplasmic membrane protein-thiol-disulfide interchange protein, which is a protein with disulfide isomerase and oxidoreductase activity 40 . Interestingly, we identified flagellar motor protein MotA in both types of vesicles. Considering that B. fragilis has no flagellum, this protein may act as an H + channel that allows the flux of protons outside the bacterial cell to maintain the required cellular pH and membrane charge 41 .
The essential difference among the proteins identified in the OMVs of both strains was observed for cytoplasm-located proteins ( Fig. 1A and Table S5-1). Compared with NTBF, we identified components of the RNA translation system, including elongation factor G, Ts, Tu, translation initiation factor IF-2 and multiple ribosomal proteins in ETBF OMV. Since the ribosomal proteins detection in OMVs preparation could be the result of contamination by cell debris, additional step of sucrose purification was added before proteome analysis (described in Supplementary Information). Resulted proteomic data obtained by using two protocols (with or without sucrose purification) were compared. As expected, we didn't observed most of ribosomal proteins and  Table S5-6). Only 50 S ribosomal protein L7/L12 was detected in NTBF OMV. Moreover, we identified the proteins involved in DNA replication and transcription. According to our results, NTBF OMVs contained only two components of the RNA translation system-elongation factor G and Tu-and did not comprise any proteins involved in DNA replication or transcription. Additionally, the identified transcription and translation components included several molecular chaperones involved in the process of RNA degradation that were common for the two types of vesicles (Table S5-2). Several stress-induced proteins, such as pirin, non-heme ferritin, thioredoxins and phosphate starvation protein PhoH, were also observed among the cytoplasm-located proteins of ETBF OMVs.
Clostripain is one of the unique proteins identified in both types of vesicles. According to a recent publication, Clostripain or Fragipain was described as the protein contributing to fragilysin toxin maturation 42 . NTBF also contains clostripain, but it is not active because of an insertion mutation of the clostripain gene 43 . According to our sequencing data of the NTBF (JIM10) clostripain gene, it contains an amino acid mutation (Q/L) at the C-end of the protein, possibly modulating its activity. The results of quantitative proteome analysis are shown in Table 1.
Virulence factors associated with OMVs. According to the genome data, B. fragilis contains a pathogenicity island (BfPAI) located within the conjugative transposon CTn86 of 63,282 bp 44 . After accurate analysis of the ETBF OMV proteome, we have found several proteins belonging to the pathogenicity island that could be delivered to the host cells via OMVs: TonB-dependent receptor (VU15_RS14380), hypothetical protein (VU15_ RS14430), which is potentially outer membrane protein OmpA according to homology analysis provided by protein blast, and choloylglycine hydrolase (VU15_RS14400) ( Fig. 2 and Table S1-3). Interestingly, choloylglycine hydrolase has been previously found to show special activity against the taurine conjugates of dihydroxy bile acids 45 . Enterotoxin-fragylisin (VU15_RS14340) was also found to be transported by the vesicles in our previous publication 46 . According results of proteogenomic analysis we determined more potential virulence factors which could be essential for ETBF pathogenity. But these data should be clarified since they can be the result of sequencing errors and subsequent annotations, respectively (Table S1-2 and Figure S4).
Enzymes and functional metabolic pathways. As a result of comparative proteome analysis for both types of vesicles (NTBF OMVs and ETBF OMVs), we identified 5 classes of enzymes: hydrolases, isomerases, lyases, oxidoreductases, and transferases ( Fig. 3A and Table S5-3).
To determine the probable protein sorting contributing to the well-defined set of proteins for both types of OMVs, we compared the relative representations of OMVs and bacterial cell proteins for both B. fragilis strains (Tables S3-1-S3-7, S4 and S5-5). As expected, there was no correlation of the protein relative representation between the vesicles and cell proteomes (Fig. 4). We identified an increase of the different proteins amounts of both types of OMVs, which were represented as the minor part of the cell proteins, suggesting that there is a special sorting mechanism that contributes to the individual differences between the two types of vesicles. For ETBF OMVs, we identified 96 proteins that were not detected in ETBF cells, including chitobiase, serine protease, several TonB-dependent receptors and others. We observed the same phenomena comparing NTBF OMVs and NTBF cells (Table S5-5).
Comprehensive metabolome analysis of NTBF and ETBF vesicles. Using metabolite standards, we performed HPLC-MS/MS analysis of 95 metabolites that are highly represented in bacterial cells (Table S6) and supplemented it by GS-MS analysis. In particular, we detected the components of amino acid metabolism, nucleotide and nucleoside metabolism, glycolysis, the TCA cycle and several cofactors ( Fig. 5A and Table S6-1). The  (Table S6-1). Moreover, according to our GC-MS data, both types of OMVs contain the oxidized and reduced forms of free fatty acids (Table S6-4).
Comparative quantitative metabolomic analyses (Tables 2 and S6-2) showed the decrease of the metabolites amount in ETBF vesicles compared with those in NTBF, indicating the existence of active metabolic pathways ( Fig. 5B-D). Importantly, we identified 6 essential amino acids that were increased in NTBF vesicles against ETBF OMVs. Creatinine and creatine, glycerate-2P, components of the TCA cycle (fumarate and malate), cofactor FAD and GMP were increased in NTBF OMVs (Table 2). We detected D-glycerate-2-P accumulation in NTBF OMVs, indicating the presence of glycolysis gaps and partial activity of this pathway. The detected decrease in the TCA cycle intermediates and specific cofactor FAD could be the result of cycle activity in ETBF vesicles. We identified increasing GMP amounts in NTBF vesicles, which could be the result of the RNA degradation process followed by purine nucleotide catabolism in NTBF vesicles, as confirmed by specific enzymatic activity in vesicles.
Metabolites may also be trapped in the vesicles as well as the proteins. To verify it, we compared the result of the metabolome analyses of ETBF and NTBF cells and OMV (Table S6-2). As expected, we observed a weak cell-OMV metabolic correlation (Fig. 6). The following metabolites were not detected in the cells and were highly represented in both types of OMVs: L-histidine, mannitol, nicotinic acid, sucrose, xanthosine, D-fructose-1,6-PP, and trehalose. We also identified unique metabolites exclusively represented in NTBF OMVs: cytidine and gluconate. Compared to the OMV metabolite composition, B. fragilis cells were mostly similar, but several metabolites were detected exclusively in ETBF or NTBF cells (Fig. 6).    ETBF and NTBF OMV metabolism reconstruction. To build comprehensive metabolic networks for OMV metabolism, we used the proteomic and metabolomic data combined with proteogenomic annotation data provided previously and developed in our laboratory using the metabolic visualization software "Pathways editor" (unpublished). For metabolites that could participate in non-annotated B. fragilis reactions (for example, arginine/ornithine transformation), we hypothesized the proteins that could carry out these reactions. We have mapped the enzyme on the map when the product of the reaction was detected. Finally, we obtained a full metabolic map of ETBF and NTBF OMVs, including different interrelated metabolic pathways, as confirmed by HPLC-MS/MS and GC-MS data. All of the detected intermediates and cofactors of the reconstructed reactions, as well as the interrelation of each pathway, were indicated in the final metabolic maps (Figs 7 and 8).
Thus, metabolic reconstruction for ETBF OMVs contained major metabolic pathways, such as glycolysis, the pentose phosphate pathway, the TCA cycle, amino acid metabolism, purine and pyrimidine metabolism, including active DNA and RNA synthesis and the lipid peroxidation defense mechanism. The most filled pathways in the reconstructed map of ETBF OMVs were glycolysis and the TCA cycle. We detected several metabolites formed in the TCA cycle: malate, succinate, fumarate and cofactor FAD, as well as pyruvate carboxylase and phosphoenolpyruvate carboxykinase enzymes, combining glycolysis with the TCA cycle. Moreover, we identified the most important enzymes involved in the rate-limiting stage of the TCA cycle (citrate synthesis): citrate synthase, malate dehydrogenase and aconitate hydratase. Such metabolic representation of the TCA cycle and glycolysis enzymes and metabolites in OMVs indicates active energy production in vesicles. We detected nucleotide and nucleoside metabolism enzymes in OMVs, including polynucleotide phosphorylase, 5′-nucleotidase and NDP-reductase; however, in the OMV metabolome, we identified only UDP as a single accumulation point. Along with these findings, we detected AMP, CMP and GMP and their metabolism products: adenosine, guanosine, xanthosine, cytidine, adenine and guanine. Simultaneously, pyrimidine metabolism (purine and cytidine) is active as marked on the map.
Based on the obtained proteo-metabolomic data, we reconstructed the lipid peroxidation defense pathway for ETBF OMVs, combining glutathione peroxidation and thioredoxin redox systems. In addition to the detected free long-chain fatty acids, we observed the hydroxyl derivative of fatty acids in the OMV metabolome, which could indicate the activity of membrane lipid peroxidation defense processes in OMVs.  Table S6-2). The table located in the upper left corner contains metabolites that were not detected in cells but only in OMVs. The table in the right bottom contains the decryption of the KEGG nomenclature. The full data represented in Table S6.
Scientific RepoRts | 7: 5008 | DOI:10.1038/s41598-017-05264-6 Despite the detection of all 20 amino acids in ETBF OMVs, we have not identified most enzymes involved in amino acid metabolism, excluding several enzymes, catalyzing amino acid transformation. Among them, we identified the enzyme catalyzing histidine decarboxylation to histamine. Histamine is a well-investigated intermediate of histidine metabolism acting as a neurotransmitter in human organisms and is involved in local immune responses as well as the regulation of physiological function in the human intestine 47,48 . Another interesting finding was γ-aminobutyric acid (GABA) and its biosynthesis intermediates α-ketoglutarate and glutamate detection. GABA is the main inhibitory neurotransmitter in the mammalian central nervous system. Detected GABA biosynthesis components in ETBF OMVs indicate the OMV transmission role in the microbiota-host interaction, via modulating the host's neurophysiological and immunological response as well as previously described histamine production 49 . NTBF OMV metabolome and proteome do not include the components of main pathways. Despite the lack of meaningful enzymes in NTBF OMVs, we detected the same composition of amino acids, including essential, monophospho-nucleotides and the products of their degradation, confirming the described B. fragilis nutrition function. Moreover, during metabolic reconstruction, we detected the pathway with "enzymatic gaps"-the absence of one or more enzymes in the reconstructed metabolic pathway in the presence of most of its enzymes. To fill such gaps in reconstruction and identify novel protein-coding genes, we performed proteogenomic analysis. Thus, we successfully detected and marked on the map fructose-1,6-bisphosphatase (fbp), catalyzing phosphate cleavage from fructose-1,6-bisphosphate in glycolysis. Using the described approach, we identified the following enzymes in B. fragilis OMVs: histidine kinase, nucleoside (xanthosine) permease and flavodoxin. Previously, we detected xanthosine in the NTBF and ETBF OMV metabolome, but there were no xanthosine precursors and specific catabolic enzymes for xanthosine biosynthesis. Xanthosine permease (XapB), identified by proteogenomic analysis, successfully solved this problem because it catalyzed xanthosine import into the OMVs from the media. Histidine kinase detected and mapped on the reconstructed metabolic map plays a major role in signal transduction in prokaryotes and is needed for the cellular adaptation to environmental conditions and stresses 50 . We have also identified the equally important stress adaptation enzyme-flavodoxin-known as an important contributor to metronidazole B. fragilis sensitivity 51 .

Fluxomic experiments. To test carbohydrate transporters and metabolic enzymes activity in both types of
OMVs, we used fluxomic experiments with isotope-labeled glucose ( 13 C 6 -D-glucose) as the substrate, supplemented with the dilution buffer. We incubated both types of OMVs and isotope-labeled glucose ( 13 C 6 -D-Glucose) containing solution at 37 °C for 30 min, 1 and 3 hours. As a result, labeled glucose metabolites such as 13 C 4 -D-erythrose-4P and 13 C 3 -Glycerol-3P were detected only in ETBF OMV. D-Erythrose-4-phosphate is a unique pentose phosphate pathway intermediate, participating in the formation of D-fructose-6-phosphate. Glycerol-P-13 C is usually used by bacterial cells for lipid synthesis and could be produced by the dehydrogenation of D-glyceraldehyde-3-phosphate. We detected a stable decrease of D-glyceraldehyde-3-phosphate at the third hour of incubation in ETBF OMVs, indicating the end of the metabolic process. During metabolic map reconstruction, we included the reaction of D-fructose-6-P phosphorylation by pfkA that was not identified but was obviously active according to HPLC-MS/ MS fluxomic data. We have not observed any metabolic activity in NTBF OMVs (Fig. 9).

Discussion
Today, there are several hypothetical models of OMV formation and mechanisms of protein sorting 52 . However, thus far, there is no exact data about the mechanism of the protein packing in OMVs. Moreover previously it has been shown, that gram negative bacteria can form two types of vesicles, consist of single and double membranes. Double membranes vesicles contain cytoplasm and cytoplasmic membrane proteins [53][54][55] . We also observed this phenomenon in Bacteroides fragilis ( Figure S5) and after all during ETBF and NTBF OMVs proteome analysis we found that the main difference was the presence of large number of cytoplasmic proteins in ETBF OMVs. We also observed ribosomal proteins but during sucrose gradient treatment we have lost most of them, confirming that its existance can be the result of the contamination by cell debris. But we have detected ribosomal proteins only in ETBF OMVs but not in NTBF OMVs before using protocol with sucrose gradient. Given the fact that ribosomal proteins are the major part of the cellular proteins we suggest that its detected in ETBF OMVs can be the result not only contamination by cell debris but the existence of different types of OMVs (including single and double membrane OMVs) which possible can be lost during sucrose gradient purification. Among the identified cellular proteins of ETBF OMV, there are a large number of enzymes, which contribute to hydrolase activity. Species from the genus Bacteroides carry multiple systems to bind to the epithelium and degrade surface located polysaccharides 30 . B. fragilis is also involved in charity towards other bacterial types deprived of the polysaccharide utilization opportunity. In addition to proteomic results obtained in our experiments, hydrolase and peptidase activity were confirmed by metabolomic data where we have detected high represented essential amino acids and non-phosphorylated carbohydrates. Another class of identified enzymes was oxidoreductases. Among the identified oxidoreductases, the components of oxidative stress defense mechanism (glutathione peroxidase and lipid hydroperoxide peroxidase) were the most important. Glutathione peroxidase requires adapting to the oxidative stress that accompanies an inflammatory response 56 . Several stress-induced proteins, including pirin, non-heme ferritin, thioredoxins and others, were also observed. Pirin, which was found exclusively in ETBF OMVs is stress induced in Cyanobacteria and may act as a quercetinase in Escherichia coli, the functions of pirin orthologs in prokaryotes remain mostly uncharacterized 57 . Non-heme ferritin, identified in ETBF OMV has a special oxidoreductase activity, binding ferric iron providing cellular iron ion homeostasis 58 . Thioredoxins are proteins that act as antioxidants by facilitating the reduction of other proteins by cysteine thiol-disulfide exchange 59 . PhoH is activated in response to phosphate limitation 60 . Transcriptome analysis revealed that the deletion of phoB affected the expression of 585 genes (more than 4-fold change) in B. fragilis, which included genes for stress response (chaperons and heat shock proteins), virulence (capsular polysaccharide biosynthesis) and phosphate metabolism 61 . Vesicles production increases under stress conditions 62 . Previously it has been shown that B. fragilis treated with bile salts overproduces fimbria and OMVs. Moreover bile salts enhance bacterial co-aggregation, bacterial-intestinal epithelial cell adhesion, biofilm formation and antimicrobial resistance of B. fragilis 63 . Possible OMVs is important mechanism involved in biofilm formation as it was previously described for Xylella fastidiosa 64 . The increase in OMV production containing multiple oxidoreductases may be also the stress adaptation mechanism at the expense of increasing the protein yield and promoting the distal neutralization of adverse environmental factors. We believe that secreted OMVs may surround bacteria and protect them from stress similar to a "shield" helping to adapt to the host immune reaction during infection.
OMV is also a mechanism for various virulence factors, especially toxin delivery. Some examples of OMV-associated toxins include: Shiga toxin from Shigella dysenteriae, LT from enterotoxigenic E. coli, leukitoxin from Actinobacillus actinomycetemcomitans, Apx toxin from Actinobacillus pleuropneumoniae and the best known toxin of Helicabacter pylori-VacA, and one of the last finding -toxins and antitoxins of Xylella fastidiosa OMVs 64-68 . Among the identified ETBF OMV proteins, we detected several virulence factors produced from the bacterial pathogenicity island including TonB-dependent receptor, hypothetical protein that, according to homology analysis (protein blast), belong to the membrane protein OmpA and choloylglycine hydrolase. In our recent study, we detected the major virulence factor -Fragilysin associated with OMVs 46 . Fragilysin is a Bacteroides fragilis toxin (bft2) that promotes colonic injury and inflammation through E-cadherin cleavage 69,70 . Until recently, there was no information of possible proteases that can provide fragilysin maturation. Clostripain-cysteine protease was found as a main factor that contributed to toxin maturation 42,43 . An unexpected result was to the identification of clostripain in OMVs, suggesting possible toxin maturation carried out in vesicles. Patatin was another potential virulence factor that was observed in OMVs. Patatin in ETBF OMVs does not belong to the pathogenicity island but is a potential virulence factor. First described as a plant storage glycoprotein with lipid acyl hydrolase activity, patatin was found in the Toxoplasma gondii genome and was associated with host-microbe interactions in bacteria 71 . The latter study has shown that ubiquitin, which can signal for protein degradation via the proteasome in eukaryotic cells, activates patatin-like phospholipases from Pseudomonas aeruginosa 72 . Considering that patatin presumably has lipolytic activity, intracellular patatin activated by host ubiquitin may dramatically interact with lipids contained in eukaryotic cell organelles.
In addition to the proteome research of OMVs, for the first time, we performed metabolomic assay of ETBF and NTBF OMVs; according to combined proteomic and metabolomic data, we reconstructed metabolic pathways for both types of vesicles. During comparative metabolome analysis of bacterial cells and OMVs, we detected the inverse correlation between metabolites represented in cells and OMVs. We hypothesized that biochemical processes in vesicles are still active after secretion. After the detailed reconstruction of metabolic maps, we determined the huge difference between the metabolic activities of the two types of vesicles. Metabolic pathways (glycolysis, the pentose phosphate pathway, the TCA cycle and nucleotide and nucleoside metabolism) reconstructed in ETBF OMVs were fully represented in ETBF OMVs in contrast to the NTBF vesicles. We carried out the fluxomic experiments using isotope-labeled substrates to confirm the activity of glycolysis. During those experiments, we have confirmed enzyme activity only in ETBF OMV but not in NTBF OMVs. Our finding and reconstructed maps indicate that ETBF OMVs could exist as an active, self-sufficient biological system. In this way, ATP and NADH-the main energy source in biochemical reactions-could be generated and fully used during ETBF vesicle persistence. Simultaneously, the generated ATP is not detected in the vesicle metabolome as the end product. We suggest that ATP and NADH could be utilized in vesicles during carbohydrate import by special ATP-dependent transporters or phosphorylation reactions such as phosphoribosyl pyrophosphate or fructose-1,6-bisphosphate synthesis or the TCA cycle. Several proteins that were detected in OMVs may utilize ATP during biochemical activity: mobility proteins, polynucleotide phosphorylase (PNPase) and histidine kinase 73,74 . PNPase was found in Salmonella enterica as a factor contributing to bacterial invasion and intracellular replication by affecting the mRNA levels of a subset of virulence genes 75 . It should be noted that PNPase activity is modulated by the ATP in E. coli. We hypothesized that the ATP generated in ETBF OMVs regulates the activity of virulence factors, transport to other microorganisms, and cell-cell communication signals, as well as determines the distal pathogenicity of B. fragilis. Moreover in our previous study we have shown that ETBF toxin which is associated with OMVs causes epithelial cells contact disruption 46 . Thus the metabolic activity of the ETBF OMVs can facilitate its long-term contact with epithelial cells, increasing the damage level. However despite the possible role of ATP in mentioned processes it should be confirmed by future experiments.
In summary, we have characterized the vesicles of two genetically similar strains of B. fragilis and found a fundamental difference between them. NTBF OMVs have only enzymes needed for polysaccharide utilization, contributing to the nutrition function among bacterial types that lack this capability. By contrast, ETBF OMVs have active enzymes that possibly provide their stability and capability for long persistence, as well as contribute to virulence factors delivery. So we hypotized, that ETBF OMVs, during long persistence, can preserve and support enzyme activity by functioning as micro reactors and transmitting the information among other microbiota cells in the host. But this suggestion should be supported by further investigations in future.

Material and Methods
Bacterial strains and growth conditions. Enteropathogenic Bacteroides fragilis (ETBF) (B. fragilis strain BOB25) 44 and non-toxigenic B. fragilis 323-J-86 (clinical isolates kindly provided from the Federal Research and Clinical Centre of Physical-Chemical Medicine Federal Medical Biological Agency, Moscow, Russia) were grown on blood agar plates containing either 5% defibrinated horse blood or brain heart infusion broth supplemented with hemin (5 g/ml) under anaerobic conditions. OMV purification. OMVs were isolated and purified from 24-h cultures of Bacteroides fragilis (BOB25 and JIM10) by multiple filtration and ultracentrifugation followed by visualization using transmission electron microscopy (TEM). Detailed information about the methods of OMV isolation, purification and visualization by TEM can be found in the Supplementary Information.
Cell fractionation. Cell fractionation was performed as described by Lindmark et al. 76 . Precipitated proteins from cell fractions were collected by centrifugation at 12,000 × g, washed with acetone, dried and dissolved in Laemmli sample buffer. The protein concentrations were quantified using a 2D-quant kit (GE Healthcare Life Sciences, USA).

SDS PAGE and in-gel trypsin digestion of protein samples. Forty micrograms of each OMV sample
(NTBF and ETBF) and 40-60 µg of cytoplasmic and membrane fractions were mixed with Laemmli sample buffer and were separated by SDS-PAGE. The resultant gels were cut into small (1 × 1 mm) pieces and were transferred into sample tubes. The samples were then subjected to in-gel trypsin digestion. Finally, the obtained peptides were dried in a vacuum and were dissolved in 3% ACN with 0.1% FA solution prior to HPLC-MS/MS analysis. Detailed information about the methods of SDS PAGE and in-gel trypsin digestion of protein samples can be found in the Supplementary Information.

HPLC-MS/MS for proteome analysis. HPLC-MS/MS proteome analysis of OMVs was performed using
a TripleTOF 5600+ mass spectrometer equipped with a NanoSpray III ion source (ABSciex, Canada) coupled to a NanoLC Ultra 2D+ nano-HPLC system (Eksigent, Dublin, CA). Comparative proteome analysis of vesicles and cells was performed using an Ultimate-3000 HPLC system (Thermo Scientific)

Label-free protein quantification. For comparative analysis of the protein amount, Exponentially
Modified Protein Abundance Indexes (emPAI) were calculated 77 . Data were normalized using the scaling method. Proteins were considered to be significantly different according to unpaired two-tailed Student's t-test (p-value < 0.05) with Benjamini and Hochberg adjustment for p-values.

LC-MS analysis and metabolite identification.
Metabolite extraction was carried out using the cold methanol method with several modifications (described in the Supplementary Information) [78][79][80] . Metabolite analysis was performed using an LC-MS-8030 triple quadrupole liquid chromatography system (Shimadzu, Kyoto, Japan). The instrument control and data processing were performed using the workstation "LabSolutions LCMS" Version 5.75 (Shimadzu Corporation, Kyoto, Japan). The metabolites were analyzed in the multiple reaction monitoring (MRM) modes, with two transitions per compound for identification and quantification purposes. The identification of qualitative differences in metabolites isolated from the OMVs of ETBF and NTBF was performed using paired multiple-adjusted t-test. To account for multiple testing and to control the false discovery rate (FDR), the Benjamini-Hochberg procedure was used 81 . A cut-off value for FDR (q < 0.2) was applied according to previous metabolomic studies [82][83][84] . Detailed information about LC-MS parameters and metabolite identification can be found in the Supplementary Information.

GC-MS method for metabolite analysis and identification.
For the GS-MS method, the procedure for OMV metabolite extraction was the same as that described for LC-MS analysis in the Supplementary Information with several modifications. The resultant freeze-dried samples were dissolved in 20 µl of pyridine and were converted to trimethylsilyl derivatives. Gas chromatography-mass spectrometry and GC-MS data analysis of silylated samples were analyzed using an Agilent 7890 gas chromatograph interfaced with an mass selective detector 5975 C (MSD). The data were processed with UniCrom (New Analytical Systems (Belarussia)) and quantified with the AMDIS (http://chemdata.nist.gov/mass-spc/amdis/downloads/) and NIST8, Wiley9 and GOLM (http:// gmd.mpimp-golm.mpg.de) software's. Fluxome experiments. Solutions containing freshly isolated ETBF and NTBF OMVs were separated into three equal parts. One microgram of D-Glucose 13 C was added to previously prepared OMV-containing solutions and was incubated at 37 °C for 30 min, 1 and 3 h. One microgram each of glucose 1, 3 (C13) and glucose 6 (C13) was also added to OMV-free 150 mM NaCl solution as the negative control. After incubation, all of the OMV preparations were subjected to ultracentrifugation at 100,000 g for 2 h (Optima L-90K ultracentrifuge; Beckman Coulter). The supernatant was discarded, and then the pellet was dissolved in 150 mM NaCl and prepared for metabolome extraction as described earlier in the "Extraction of metabolites" section.