EVs were isolated [Figure 1] from L. plantarum^[16], kindly provided by Dr. Daniel (Institut Pasteur de Lille, France). Cultures were grown in 5 L of De Man–Rogosa–Sharpe (MRS; Merck, Germany) broth supplemented with Tween 80 (MRS-T) in closed bottles without shaking at 37 °C (Innova 4230 Incubator, Eppendorf New Brunswick, Germany). The culture was inoculated with an overnight starter culture at a 1:200 ratio and incubated until an optical density at 600 nm (OD600) of 1.5 was reached. OD600 = 1.5 corresponds to the exponential growth phase of L. plantarum [Supplementary Figure 1] and was selected to obtain robust vesicle yield while minimizing contamination from membrane fragments released during stationary-phase cell lysis. Bacterial cells were then removed via centrifugation (6,000 × g, 20 min, 4 °C; Heraeus, Thermo Fisher Scientific, USA), followed by filtration through a 0.22-µm sterile vacuum bottle-top filter (Steritop, Merck). The resulting cell-free supernatant was concentrated approximately 13-fold using an Amicon Ultra Stirring Cell (Merck) equipped with 300 kDa ultrafiltration discs.

Crude LpEVs were obtained via ultracentrifugation (UC) of the concentrated supernatant at 150,000 × g for 3 h at 4 °C (Beckman Coulter, USA; 45Ti rotor) and resuspended in 500 µL of 25 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid buffer containing 0.15 M NaCl (HEPES-N). The EVs were further purified via SEC using qEVoriginal/35 nm Gen 2 columns coupled to an Automatic Fraction Collector (IZON Science, France), following the manufacturer’s instructions. Eight fractions of 400 µL each were collected [Supplementary Figure 2A and B]. Fractions enriched in homogeneous particles - typically corresponding to the purified collection volume (PCV) 1.1-1.3 - were pooled and stored at 4 °C for immediate use or at -20 °C for long-term storage.

Mock EVs were prepared following the same procedure, using uninoculated 5 L of MRS-T medium. After incubation at 37 °C for 8 h (the time required for bacterial cultures to reach an OD600 of 1.5), the medium was processed in parallel with the inoculated samples, including centrifugation, filtration, concentration, UC, SEC, and fraction pooling (PCV 1.1-1.3).

The long-term storage stability of LpEVs was evaluated following a modified protocol of Görgens et al.^[19]. After UC (150,000 × g, 3 h; SW41Ti rotor), LpEVs were resuspended in one of five buffers: (i) 25 mM HEPES with 0.9% NaCl (HEPES-N); (ii) 25 mM HEPES with 0.9% NaCl and 0.2% BSA (HEPES-NA); (iii) phosphate-buffered saline (PBS); (iv) PBS with 0.2% BSA (PBS-A); or (v) PBS with 25 mM HEPES and 0.2% BSA (PBS-HA). Aliquots were prepared in triplicate and stored at 4, -20, or -80 °C. Samples were analyzed at baseline (time 0) and after 1, 6, and 12 months. Each aliquot was thawed only once for size and concentration measurements using the Zetasizer; after measurement it was discarded. All measurements were performed from a single stock preparation to ensure consistency across analyses.

The physical stability of L. plantarum and LpEVs was evaluated under varying salt, pH, and detergent conditions relevant to the human physiological environment. Experiments were performed in triplicate (biological replicates).

Salt stability: LpEVs (1.2 × 10¹¹ particles/mL in 50 µL of HEPES-N) were used as the starting material. Concentrated NaCl was added to achieve final concentrations of 0.25, 0.35, 0.5, and 1.0 M. Samples were mixed and analyzed for particle size and concentration within 30 min. In parallel, overnight bacterial cultures were diluted 1:200 in MRS-T supplemented with the same NaCl concentrations.

pH stability: LpEVs (0.5 × 10¹⁰ particles) were pelleted via UC (150,000 × g, 3 h; SW41 rotor), resuspended in 50 µL of HEPES-N adjusted to pH values of 5.7-8.2 (in 0.5-unit increments), and analyzed within 30 min as described above. In parallel, overnight bacterial cultures were diluted 1:200 in MRS-T adjusted to the same pH using citrate–phosphate buffer.

Detergent stability: LpEVs (1 × 10¹¹ particles/mL in 90 µL of HEPES-N) were incubated with SDS at final concentrations of 0, 0.1, 0.6, 2.2, and 8.9 mM and analyzed within 30 min. In parallel, overnight bacterial cultures were diluted 1:200 in MRS-T containing the same SDS concentrations.

All experiments were repeated three times, and each sample was prepared in triplicate (technical replicates). Bacterial growth under all conditions was monitored by measuring OD600 at 0, 3, 5, 6, 7, and 8 h in 96-well plates using a plate reader. The OD600 value at time 0 was used as the baseline and subtracted from subsequent measurements.

Experimental details have been submitted to the EV-TRACK knowledge base (EV-TRACK ID: EV250022)^[26].

Statistical analyses were performed using GraphPad Prism version 9.5.1, unless otherwise specified. Comparisons between groups were conducted using one-sample t-tests, ordinary one-way ANOVA, Dunnett’s multiple-comparisons test with a single pooled variance, or two-way ANOVA, as appropriate. The statistical tests used are indicated in the corresponding figure legends. A P-value of P ≤ 0.05 was considered statistically significant.

L. plantarum was cultured in MRS-T, and LpEVs were harvested from 5 L cultures at an OD600 of 1.5, corresponding to the mid-logarithmic growth phase. Following filtration to remove cells and debris, crude EVs were isolated via UC. Further purification was performed using SEC [Figure 1], resulting in eight fractions of 400 µL each (Supplementary Figure 2A and B shows a representative chromatogram). Typically, fractions 1-3 (PCV 1.1-1.3) were pooled, resulting in a final volume of approximately 1.2 mL of purified LpEVs per batch.

Dynamic light scattering (DLS) analysis using a Zetasizer revealed a mean particle size of 73.9 nm (SD = 6.9 nm) and an average particle concentration of 1.64 × 10¹² particles/mL (SD = 0.91 × 10¹² particles/mL) across five independent batches (five biological replicates) [Supplementary Table 1]. The average total yield per 5 L culture was 1.97 × 10¹² LpEVs (SD = 1.09 × 10¹²). Representative particle size and concentration data from one batch are shown in Figure 2A and B. Purification via SEC markedly reduced the protein and PGN content in the purified EV preparations compared with crude EVs obtained via UC alone [Supplementary Figure 2C and D]. To contextualize LpEV production relative to bacterial biomass, the colony-forming unit (CFU) count of the parent bacteria was determined in the same 5 L culture used for EV isolation. At an OD600 of 1.5, the culture contained 6.7 × 10¹² CFUs and yielded 1.2 × 10¹² LpEVs, corresponding to an approximate bacterium-to-EV ratio of 6:1.

Because the determination of EV size largely depends on the analytical method used^[8], cryo-EM was employed as a complementary technique to assess the size distribution of LpEVs [Figure 2C]. Compared with DLS measurements obtained using the Zetasizer, cryo-EM analysis showed that LpEVs predominantly ranged within 20-50 nm in diameter.

Transmission electron microscopy (TEM) [Figure 2D and Supplementary Figure 3] and cryo-EM [Figure 2E] were used to visualize the morphology of LpEVs and assess the presence of aggregates or bacterial debris. TEM images showed intact, spherical vesicles that frequently appeared in clusters [Figure 2D], as further confirmed by wide-field TEM views [Supplementary Figure 3]. Cryo-EM provided definitive confirmation of vesicle identity and revealed a well-defined bilayer membrane even in the smallest EVs [Figure 2E], underscoring their structural integrity.

Because MRS-T contains components derived from animal products, it was important to assess whether EVs or EV-like particles originating from the medium itself could confound the experimental results. To address this, MOCK-EVs were isolated from 5 L of uninoculated MRS-T and processed using the same protocol as for bacterial cultures. DLS analysis showed that MOCK-EVs were larger than LpEVs [Supplementary Figure 4A], with a concentration of 2.12 × 10¹⁰ particles/mL - approximately 100-fold lower than that observed in L. plantarum cultures [Supplementary Figure 4B]. TEM revealed only a few particles in MOCK-EV samples, exhibiting morphological features distinct from those of LpEVs [Supplementary Figure 4C]. No protein bands were detected when the maximum possible sample volume (20 µL) was loaded onto the SDS–PAGE gel [Supplementary Figure 4D]. Moreover, the protein content of the MOCK-EV samples, determined via both BCA and Bradford assays, was below the detection limit (data not shown). These findings confirm that EV-like particles present in MRS-T are negligible in both abundance and protein content and are therefore unlikely to interfere with the interpretation of LpEV-specific experimental results.

The surface lipid composition of L. plantarum and LpEVs was analyzed using the semi-quantitative ToF-SIMS method. The analysis revealed a decrease in prenols and an enrichment of fatty acids on the surface of LpEVs compared with the parent bacteria [Figure 2F]. Conversely, levels of glycerophospholipids, sphingolipids, and sterols showed only minor differences between the bacteria and their EVs. The presence of LTA, a key structural component of Gram-positive bacteria, was confirmed in both the parent cells and LpEVs via western blotting [Figure 2G]. ZP measurements performed in 1 mM HEPES buffer revealed distinct differences in surface charge between the bacteria and their EVs [Figure 2H]. The bacterial surface exhibited a strong negative charge of -37.38 mV (SD = 0.25 mV), whereas LpEVs displayed a considerably less negative charge of -10.8 mV (SD = 1.01 mV). These results indicate substantial differences in the surface properties of LpEVs and their parental cells, which may critically influence their stability, interactions, and biological functionality.

LpEVs were routinely stored in HEPES-N buffer at -20 °C. To evaluate their long-term stability under different conditions, we analyzed LpEV samples stored in HEPES-N at three temperatures: 4, -20, and -80 °C [Figure 2I and J]. At 4 °C, signs of instability became evident after 12 months, including significant fluctuations in particle concentration (P ≤ 0.05). Conversely, samples stored at -20 and -80 °C remained stable for up to one year. We further tested four alternative buffer formulations for their ability to preserve LpEVs: (i) PBS, (ii) PBS-A, (iii) PBS-HA, and (iv) HEPES-NA [Supplementary Figure 5]. At 4 °C, none of the buffers maintained LpEV stability beyond six months; complete degradation occurred, as no reliable measurements could be obtained (low data quality reported by the Zetasizer) [Supplementary Figure 5A]. Conversely, storage at -20 °C [Supplementary Figure 5B] and -80 °C [Supplementary Figure 5C] preserved vesicle integrity across all buffer conditions. Notably, buffers containing albumin led to an apparent increase in particle size, suggesting aggregation or interaction effects at 4 °C. Based on these results, we recommend storing LpEVs at -20 °C in 25 mM HEPES-N buffer as the optimal condition for maintaining vesicle integrity and concentration during long-term storage.

Given the apparent physical robustness of LpEVs, we evaluated their stability under various environmental conditions that mimic those in the human body. Equivalent quantities of LpEVs were exposed to buffers of different compositions, and their particle size and concentration were measured using the Zetasizer. LpEVs remained stable in size and concentration across a pH range of 5.7-7.7. However, at pH values above 7.7, a significant decrease (P < 0.01) in LpEV concentration was observed, indicating sensitivity to alkaline conditions [Figure 3A]. LpEVs also demonstrated high tolerance to salt stress with respect to their size and concentration when incubated in NaCl concentrations up to 1 M [Figure 3B]. When exposed to the negatively charged detergent SDS, LpEVs maintained stable size and numbers at SDS concentrations up to 0.6 mM. At higher SDS concentrations, both particle size and concentration changed significantly (P < 0.01); however, LpEVs remained detectable even under these conditions, suggesting that complete vesicle degradation did not occur [Figure 3C].

Given the demonstrated stability of LpEVs across a broad range of pH values, salt concentrations, and negatively charged detergents, we next examined how these environmental conditions affect the growth of the parent bacteria. L. plantarum was initially cultured in MRS-T, and overnight cultures were subsequently inoculated into fresh media modified to reflect varying pH, NaCl, and SDS concentrations [Figure 4]. Under standard conditions (pH 5.8), the bacteria exhibited robust growth. However, deviations from this pH significantly impaired bacterial proliferation, with significant reductions observed from 5 h post-inoculation (P < 0.0001, two-way ANOVA) [Figure 4A]. Similarly, increasing the NaCl concentration above 0.15 M led to dose-dependent growth inhibition, with significant differences detected at all tested concentrations from 6 h onward (P < 0.001, two-way ANOVA) [Figure 4B]. Exposure to SDS also markedly suppressed bacterial growth; even at 0.6 mM, SDS significantly reduced growth at 5, 6, 7, and 8 h post-inoculation (P < 0.0001, two-way ANOVA), with higher concentrations exerting progressively stronger inhibitory effects [Figure 4C].

As L. plantarum encounters bile salts during gastrointestinal transit, we investigated bile as a physiologically relevant stressor influencing bacterial growth, EV production, and EV-associated cargo. As shown above, environmental factors such as pH, salinity, and detergents affect bacterial growth dynamics. Extending this to bile, we examined the effects of increasing concentrations of bovine bile in MRS-T.

Bile concentrations used in this study (0%-2%) were selected to span the physiological range encountered in the human small intestine, where luminal bile concentrations typically range from approximately 0.2% to 2% (w/v) under fed conditions^[27,28]. A concentration of 0.25% was ultimately selected for detailed EV analysis as it represents a sublethal yet clearly stressful condition within this physiological range, causing approximately 50% inhibition of L. plantarum growth [Figure 5A] while still permitting sufficient bacterial growth and EV production for downstream characterization. EVs produced under bile stress (LpEVs^BILE) were isolated and characterized using the same protocol as for EVs produced under standard conditions (LpEVs), with harvesting conducted at OD600 = 1.5. Although the total particle yield of LpEVs^BILE was comparable to that of LpEVs, DLS analysis revealed a significant increase (P < 0.01) in mean vesicle size, from 82.79 (SD = 11.63 nm) to 118.1 nm (SD = 8.35 nm) [Figure 5B]. Additionally, ZP measurements showed a more negative surface charge for LpEVs^BILE (-17.09 mV, SD = 1.33 mV) than for LpEVs (-10.99 mV, SD = 2 mV), suggesting alterations in membrane composition or structure [Figure 5C]. Analysis of vesicular RNA content using a Bioanalyzer revealed an enrichment of RNA in LpEVs^BILE, increasing from 2.21 ng per 1 × 10¹¹ EVs in LpEVs to 155.56 ng per 1 × 10¹¹ EVs in LpEVs^BILE. Following RNase treatment, both vesicle types retained sRNA fragments < 100 nucleotides, whereas untreated samples (particularly LpEVs) contained a broader range of RNA sizes, including longer fragments [Figure 5D]. No DNA was detected in either EV preparation using our assay (data not shown). However, this does not rule out the presence of low-abundance vesicle-associated DNA below our method’s detection limit.

To investigate the protein cargo of LpEVs and LpEVs^BILE, we performed SDS–PAGE, quantitative protein assays using three independent methods, and subsequent proteomic analysis. SDS–PAGE revealed a distinct protein band at approximately 120 kDa in LpEVs^BILE, whereas no visible bands were detected in LpEVs under identical loading conditions or in the MOCK-EVs^BILE sample [Figure 5E]. Given the typically low protein yield from bacterial EVs - particularly for LpEVs - this likely reflects protein concentrations below the detection limit of SDS–PAGE rather than a true absence of protein. Consistently, quantitative protein assays showed a significantly higher total protein content in LpEVs^BILE than in LpEVs (P ≤ 0.05 for Pierce 660 nm, P < 0.01 for Bradford, P < 0.001 for BCA) [Figure 5F], with no significant variation among the three methods used.

FTIR spectroscopy was employed to investigate bile-induced metabolic changes in L. plantarum [Figure 6A] and its EVs [Figure 6B], building on previous studies demonstrating the utility of this technique for analyzing biochemical profiles of cells and vesicles^[25]. Spectral data were recorded in the 4,000-500 cm^-1 range and preprocessed before analysis. The effect of bile on LpEV spectra was evident across all three major spectral regions - proteins (1,720-1,500 cm^-1), fatty acids (3,020-2,800 cm^-1), and polysaccharides (1,200-900 cm^-1) - but was most pronounced in the protein region [Figure 6B]. A distinct new peak appeared at 1,550 cm^-1 in LpEVs^BILE that was absent in LpEVs (indicated by an arrow in Figure 6B). This wavelength is characteristic of chemical groups containing amide bonds, carboxylates, or nitro compounds. In contrast, spectral differences in the bacterial samples were primarily observed in the regions associated with fatty acids and polysaccharides [Figure 6A]. Baseline- and vector-normalized spectra from two additional biological replicates of bacteria and EVs are provided in the Supplementary Figure 6A and B. Additional control experiments (MOCKEV^BILE and MRST ± 0.244% bile) showed negligible FTIR spectral interference from bile acids or medium components [Supplementary Figure 7A and B], confirming that the observed LpEV^BILE signatures arise from vesicular material rather than bile- or medium-derived artifacts.

To further investigate the impact of bile on the spectral profiles of L. plantarum and LpEVs, HCA was performed. This analysis revealed clear clustering of samples according to the presence or absence of bile in the culture medium, whether based on the entire spectral range [Supplementary Figure 6C] or specific regions: fatty acids [Figure 6C], proteins [Figure 6D], and polysaccharides [Figure 6E]. Although the overall impact of bile on the spectral heterogeneity of L. plantarum/L. plantarum^BILE and LpEVs/LpEVs^BILE was comparable when analyzed across the full spectrum [Supplementary Figure 6C], region-specific differences became apparent [Figure 6C-E]. Spectral heterogeneity between LpEVs and LpEVs^BILE was more pronounced in the fatty acid and protein regions, whereas bacterial cells exhibited greater variability in the polysaccharide region.

To investigate further qualitative differences in protein composition, we performed a comparative proteomic analysis of L. plantarum cultures grown with and without 0.25% bile, along with the corresponding EVs produced under these conditions. The proteomic workflow demonstrated excellent reproducibility across biological and technical replicates [Supplementary Figure 8A]. Consistent with the quantitative protein data [Figure 5F], LpEVs^BILE contained significantly higher total protein content and a greater number of identifiable proteins than LpEVs (1,533 proteins SD = 30.04 vs. 855.7 proteins SD = 53.91, P < 0.0001) [Supplementary Figure 8B]. For mass spectrometry, 250 ng of trypsin-digested peptide was analyzed for L. plantarum, L. plantarum^BILE, and LpEVs^BILE, whereas the maximum possible volume of LpEVs was used due to their limited protein content. PCA showed a well-structured separation of all sample groups [Figure 7A]. However, difference between Lp and Lp^BILE were subtle. LpEVs were separated from the other groups reflecting major compositional shift. Strong separation in PC1 and PC2 of LpEVs and LpEVs^BILE indicates that bile alters the EV protein composition. In bacterial cells, a total of 1,871 proteins were identified in both L. plantarum and L. plantarum^BILE, with only 56 showing altered abundance [Figure 7B]. Among these, BSH, a key enzyme mediating bile acid deconjugation^[29], showed no significant change (P ≤ 0.05, fold-change cutoff of ≤ -2 or ≥ 2). Seventy-one proteins were unique to L. plantarum, whereas 22 were detected exclusively in L. plantarum^BILE. In contrast, the EV proteome exhibited substantially greater variation. Across LpEVs and LpEVs^BILE, 1,627 proteins were identified, with 101 showing significant differential abundance (P ≤ 0.05 and a fold-change cutoff of ≤ -2 or ≥ 2) [Figure 7C]. Notably, BSH was significantly upregulated in LpEVs^BILE compared with LpEVs. Furthermore, 82 proteins were unique to LpEVs, whereas 784 were detected exclusively in LpEVs^BILE, indicating that bile stress induces a pronounced shift in EV cargo composition.

To investigate the identity of the prominent protein band observed exclusively in LpEVs^BILE by SDS-PAGE [Figure 5E], we re-examined the proteomic dataset for proteins within the corresponding molecular weight range of 105 to 125 kDa. Several proteins were detected exclusively in LpEVs^BILE and were absent from control LpEVs, suggesting that their incorporation into EVs is induced by bacterial exposure to bile rather than reflecting constitutive EV cargo. These include DNA polymerase III subunit alpha, carbamoyl phosphate synthase pyrimidine-specific large chain, a hypothetical membrane protein, cell surface protein (CscB family), Type I restriction enzyme endonuclease subunit (R protein), Sigma54 activator (mannose PTS operon regulator), 3′-5′ exonuclease DinG, and DNA translocase FtsK. These proteins are involved in core cellular processes including DNA maintenance, gene regulation, and metabolism/transport, suggesting that bile stress may trigger selective packaging of functionally significant cargo into LpEVs. While these candidates represent plausible identities for the band observed in bile-exposed samples, definitive assignment would require targeted mass spectrometric analysis of the excised gel band, which was beyond the scope of the current study.

Protein–protein interaction network and functional enrichment analyses performed using STRING revealed significant enrichment of proteins associated with molecular functions such as catalytic, ligase, hydrolase, and oxidoreductase activities [Figure 7D]. Enriched terms with adjusted P ≤ 0.05 (FDR) were considered statistically significant. Biological processes significantly enriched in LpEVs and LpEVs^BILE included the metabolism of pyridine-containing compounds, tRNA and RNA processing, noncoding RNA processing, and vitamin metabolic pathways [Supplementary Table 2]. Compared with the bacterial proteome background, LpEVs^BILE were predominantly enriched in cytosolic and cytoplasmic proteins [Supplementary Table 2]. Keyword-based annotation further revealed enrichment in flavoproteins, tRNA processing, ligases, oxidoreductases, transcription-associated proteins, and cytoplasmic proteins [Supplementary Table 2]. Of the 30 flavoproteins identified in L. plantarum, 24 were also detected in LpEVs^BILE.