Escherichia coli-Derived Outer Membrane Vesicles Relay Inflammatory Responses to Macrophage-Derived Exosomes

ABSTRACT Extracellular vesicles are considered to be an inflammatory factor in several acute and chronic inflammatory diseases. The present study shows that exosomes from macrophages (Mφ) infected with live Escherichia coli induced secretion of proinflammatory factors by uninfected Mφ. Inflammatory responses induced by exosomes derived from Mφ infected with heat-inactivated E. coli or lipopolysaccharide were significantly weaker than those elicited by outer membrane vesicles (OMVs) released from live E. coli. Proteome analysis of exosomes from Mφ infected with live or heat-inactivated E. coli revealed that E. coli proteins OmpA, GroL1, DegP, CirA, and FepA are candidate triggers of exosome-mediated inflammatory responses. OMVs from a cirA-deleted strain suppressed exosome-mediated inflammatory responses by uninfected Mφ. The C terminus of the CirA protein (residues 158 to 633), which was relayed from E. coli-derived OMV to Mφ-derived exosomes, promoted exosome-mediated inflammatory responses by uninfected Mφ. These results suggest an alternative mechanism by which extracellular vesicles from E. coli OMV-elicited Mφ transmit proinflammatory responses to uninfected Mφ.

the outer membrane of Gram-negative bacteria, is one of the primary drivers of sepsis in the absence of live bacteria and has been investigated as one of the triggers of lethal shock (3). LPS triggers production of proinflammatory factors such as tumor necrosis factor alpha (TNF-a), interleukin 1 beta (IL-1b), and IL-6 by immune cells (4). These cytokines, which mediate the initial response of the innate immune system to injury or infection, are produced primarily by macrophages (Mw ). Mw are the first line of defense against bacteria and release cellular signaling molecules and various proinflammatory cytokines and inflammatory mediators, including nitric oxide (NO) and cyclooxygenase-2 (COX-2) (5). Moreover, extracellular vesicles (EVs) released by Mw play a clear role as inflammatory mediators. Exosomes, EVs derived from the endolysosomal pathway in mammalian cells, have a unique lipid and protein makeup (5,6). Moreover, several cytokines, including TNF-a (7) and IL-18 (8), were reported to be contained within exosomes. Mw -derived exosomes induce production of inflammatory factors by endothelial cells by decreasing the level of microRNA contained within exosomes (9). Therefore, exosomes are novel intercellular proinflammatory factors (10). However, the role of exosomes transmitted from infected donor Mw to uninfected recipient Mw is unknown. Bacteria also release EVs. EVs of Gram-negative bacteria are specifically called outer membrane vesicles (OMVs) (11). OMVs have a role in pathophysiological functions in both bacterium-bacterium and bacterium-host interactions. OMVs harbor LPS together with proteins, lipids, and virulence factors, suggesting that OMVs mediate the cytosolic localization of LPS on host cells (12). OMVs produced by Helicobacter pylori, Neisseria gonorrhoea, and Pseudomonas aeruginosa enter host epithelial cells and subsequently detect the NOD1 receptor, resulting in immune responses (13). All of the OMVs derived from Treponema denticola, Tannerella forthia, and Porphyromonas gingivalis were reported to increase TNF-a, IL-8, and IL-1b cytokine secretion (14). However, the mechanism by which bacterial OMVs elicit inflammatory responses in host cells has yet to be fully understood. In the present study, we considered the possibility that bacterial OMVs influence the ability of host cells to initiate or expand inflammation during bacterial infections, and we focused on Mw -derived exosomes that trigger inflammatory responses in uninfected Mw during infection with nonpathogenic E. coli or treated with E. coli-derived OMVs. Furthermore, we identified a key E. coli protein in exosomes that triggers the inflammatory responses.

RESULTS
Exosomes derived from live E. coli treated-Mu promote inflammatory responses by naive Mu. First, we examined whether exosomes derived from Mw infected with E. coli induce expression of inflammatory mediators. Mouse RAW264.7 Mw were infected for 9 or 18 h with live or heat-inactivated E. coli (DH5a strain) at a multiplicity of infection (MOI) of 5. Both live and heat-inactivated E. coli led to a strong increase in expression of inducible NO synthase (iNOS) and COX-2 proteins; however, expression of these proteins did not increase in noninfection (Fig. 1a). Expression of mRNA-encoding iNos and Cox-2 also increased in Mw infected with live and heat-inactivated E. coli, but not in uninfected Mw (Fig. 1b). Next, we isolated exosomes from cell culture media by ultracentrifugation. The amount of protein in exosomes from cells infected with live or heat-inactivated E. coli (live-exo or HI-exo) for 18 h was 2.26-fold or 2.40-fold higher than that in exosomes from uninfected cells (none-exo) (Fig. 1c). Expression of integrin b1, heat shock protein 90 (HSP90), and glyceraldehyde-3-phosphate dehydrogenase (GAPDH), all of which are exosome marker proteins, was similar in all three exosomes (i.e., live-exo, HI-exo, and none-exo) (Fig. 1d). In addition, live-exo showed high expression of CD63 and CD9. Next, the particle size of each of these exosomes was analyzed using NanoSight. The diameter of live-exo (135.8 6 55.6 nm) was a little larger than that of HI-exo (113.6 6 41.5 nm) and none-exo (112.4 6 58.0 nm) (Fig. 1e). When RAW264.7 Mw were used as recipient cells, the live-exo from Mw infected with live E. coli for 9 or 18 h showed a marked increase in expression of iNOS and COX-2 (Fig. 1f). In addition, live-exo increased expression of mRNA encoding iNos (9 h, 168-fold; 18 h, FIG 1 Increase in the proinflammatory mediators by exosomes from macrophages that were infected with live E. coli. After mouse RAW264.7 macrophages (Mw ) were infected for 9 or 18 h with live (L) or heat-inactivated (HI) E. coli K-12 strain (EC) at a multiplicity of infection (MOI) of 5, the exosomes (none-exo, live-exo, and HI-exo) were prepared from their culture media of cells. (a) The protein expression of inducible nitric oxide synthase (iNOS) and cyclooxygenase-2 (COX-2) in whole-cell lysates was measured by Western blotting. b-Actin was used as loading control. (b) The mRNA levels of iNos and Cox-2 in bacteria-infected cells were analyzed by quantitative real-time PCR and normalized to 18S rRNA. Values are expressed as the mean 6 SD (n = 4) of at least three independent biological replicates. (c to e) Exosomes (live-exo or HI-exo) from donor Mw , which were infected with live or HI E. coli for 9 or 18 h, were isolated by ultracentrifugation. Exosomes (none-exo) from uninfected Mw were collected as control. (c) The amounts of exosomes were measured by protein concentration. (d) The expression of exosome markers, including integrin b1, heat shock protein 90 (HSP90), CD63, glyceraldehyde-3-phosphate dehydrogenase (GAPDH), and CD9 in exosomes from donor Mw , which were infected with live or HI E. coli for 18 h, were detected by Western blotting. (e) Particle size of exosomes derived from donor Mw , which were infected with live or HI E. coli for 18 h, was analyzed by NanoSight. Values are expressed as the mean 6 SD. (f, g) RAW 264.7 cells (recipient cells; 5 Â 10 5 cells) were incubated with exosomes from donor cells (1 Â 10 6 cells). (f) After incubation for 24 h, the expression of iNOS and COX-2 in whole-cell lysates was measured by Western blotting. b-Actin was used as loading control. (g) After incubation with each exosome for 24 h, the mRNA levels of iNos and Cox-2 were analyzed by quantitative real-time PCR and normalized to 18S rRNA. Values are expressed as the mean 6 SD (n = 3 to 4) of at least three independent biological replicates. One-way ANOVA and Tukey's test were used for statistical analysis; *, P , 0.05; **, P , 0.01; ***P , 0.005; ****, P , 0.001; ns, not significant.

53-fold) and
Cox-2 (9 h, 471-fold; 18 h, 65-fold) significantly compared with exosomes from untreated Mw (Fig. 1g). In contrast, none-exo and HI-exo failed to increase expression of iNOS and COX-2 protein and mRNA in recipient Mw . These data show that expression of inflammatory mediators by uninfected Mw was increased by exposure to exosomes from donor Mw infected with live E. coli, but not those infected by heat-inactivated E. coli.
E. coli-derived OMVs promote exosome-mediated inflammatory responses. Next, we focused on OMVs released by live E. coli rather than the bacteria themselves; this is because we did not prepare OMVs from heat-inactivated E. coli. To examine whether OMVs from live E. coli trigger inflammatory responses by recipient Mw via donor Mw -derived exosomes, we prepared OMV pellets from E. coli culture media by ultracentrifugation at 100,000 Â g (Fig. 2a). The amount of OMVs (as protein contents) in live E. coli cultured for 3 to 24 h peaked at 12 h (Fig. 2b). Expression of OmpA, a marker of OMVs, was also highest in OMVs from E. coli cultured for 12 h (Fig. 2c). Transmission electron microscopy (TEM) showed that OMVs were about 100 nm in diameter (Fig. 2d). The amount of OMVs and OmpA decreased in a time-dependent manner after 12 h; therefore, we prepared OMVs from E. coli cultured for 12 h. After incubation of donor Mw with E. coli culture medium (med) for 9 h, the supernatant (sup), or the pellet (OMVs) (the latter two were obtained from E. coli culture media by ultracentrifugation at 100,000 Â g), the Mw medium was exchanged with fresh medium. Exosomes were then collected again from Mw culture medium after 9 h. We found that expression of iNOS and COX-2 proteins by donor Mw treated with med, sup, or the OMV pellet was higher than that by untreated Mw (Fig. 2e). In recipient Mw , expression of iNOS and COX-2 increased markedly after exposure to exosomes from Mw treated with the med (relative intensity of iNOS, 1.0, and COX-2, 1.0) and OMVs (iNOS, 2.27; COX-2, 1.59), but not by exosomes from Mw treated with sup (iNOS, 0; COX-2, 0.04) (Fig. 2f). These results suggest that OMVs derived from live E. coli increase the exosome-mediated inflammatory responses by uninfected Mw . Next, we confirmed the relationship between OMV concentration and exosome-mediated inflammatory responses. In donor Mw , the levels of iNOS and COX-2 proteins in Mw (8 Â 10 6 cells) treated with a low concentration of OMVs (0.5 mg/7 mL) were the same as in Mw (8 Â 10 6 cells) treated with a large concentration of OMVs (25 mg/7 mL) (see Fig. S1a in the supplemental material). In contrast, exosomes from Mw treated with a high concentration of OMVs (5, 10, or 25 mg/7 mL) showed increased levels of iNOS and COX-2 proteins, while exosomes from Mw treated with a low concentration of OMVs (0.5 or 1.0 mg/ 7 mL) did not (Fig. S1b). These results indicate that there is a threshold concentration of OMVs that triggers exosome-mediated inflammatory responses. Next, we examined whether LPS promoted exosome-mediated inflammatory responses; this is because OMVs derived from the outer membrane of E. coli contain a large amount of LPS (12). LPS increased expression of iNOS and COX-2 proteins by donor Mw , whereas exosomes from Mw treated with LPS did not increase protein expression by naive Mw (Fig. S2a and b). LPS increased expression of iNOS and COX-2 in a dose-dependent manner; the data suggest that LPS at a dose of .1 ng/ mL in culture medium is required to increase expression of inflammatory mediators (Fig. S2c). In contrast, when we measured the concentration of LPS in exosomes from Mw treated with live E. coli, it was 0.0025 6 0.0001 ng/mL of culture medium (Fig. S2d). We examined the levels of expression of inflammatory mediators by using DLPS-OMVs derived from E. coli (ClearColi), which deleted the two secondary acyl chains of the normally hexa-acylated LPS, in recipient cells. DLPS-OMVs increased the expression of iNOS and COX-2 proteins by donor Mw (Fig. S2f). Exosomes from Mw treated with DLPS-OMVs also increased the mRNA levels of inflammatory mediators by naive Mw . These results suggest that exosome-mediated inflammatory responses are triggered by molecules other than LPS.
Furthermore, we examined whether exosomes from mouse bone marrow-derived macrophages (BMDMs) treated with E. coli-derived OMVs enhanced inflammatory responses. E. coli-derived OMVs increased the expression of iNOS and COX-2 proteins by donor Mw of BMDM (Fig. S3a). Mouse BMDM-derived exosomes increased the mRNA levels of inflammatory mediators by naive BMDM (Fig. S3c).
Proteins contained within exosomes promote inflammatory responses. To examine the components within exosomes that increase expression of inflammatory mediators, exosomes from Mw infected with live E. coli were treated with proteinase K, DNase, or RNase in the presence or absence of Triton X-100. Live-exo treated with proteinase K in the presence or absence of Triton X-100 failed to increase expression of iNOS protein and mRNA significantly ( Fig. 3a and b). However, exosome-mediated increases in expression of COX-2 protein and mRNA were inhibited by proteinase K in the presence of Triton X-100. In contrast, DNase and RNase did not suppress the exosome-mediated increase in expression of iNos and Cox-2 mRNA (Fig. 3a). These results show that the proteins contained within exosomes are an important factor that triggers exosome-mediated inflammatory responses. Therefore, we used MS analysis to examine proteins contained in three exosomes (none-exo, live-exo, and HI-exo) from Mw infected with or without live or heat-inactivated E. coli. E. coli proteins and/or mouse Mw proteins were detected in all three exosomes. Data shown earlier demonstrated that increased exosome-mediated inflammatory responses in recipient Mw were dependent on the concentration of OMVs in donor Mw (Fig. S1); therefore, we focused on proteins expressed by E. coli. Sixty-six E. coli proteins were detected from live-exo. Among these, only three proteins were also found in HI-exo ( Fig. 3c; Table 1). OmpA, GroL1, DegP, CirA, and FepA proteins were identified as candidate triggers of exosome-mediated inflammatory responses because large numbers of their peptide fragments were detected by mass spectrometry (MS). Aldehyde-alcohol dehydrogenase (ADHE) was the most abundant peptide fragment detected in all E. coli proteins from live-exo; however, ADHE was removed as a candidate because mouse ADHE was contained in all three exosomes from Mw . Single gene-deleted E. coli suppresses exosome-mediated inflammatory responses. To determine which factors are important for triggering exosome-mediated inflammatory responses, we next examined whether OMVs from single gene-deleted E. coli strains (i.e., DompA, DfepA, DcirA, and DdegP) obtained from the Keio library triggered exosome-mediated inflammatory responses. Since groE1 is a lethal gene, we did not examine it in this study. Moreover, strain DompC was used as a negative control because OmpC is an abundant protein in OMVs. Density gradient ultracentrifugation was used to confirm expression of the OmpA, FepA, CirA, DegP, and OmpC proteins in OMVs. All proteins were most abundant in fraction 9 (Fig. S4a). The density on fraction 9 (1.16 g/mL) was lower than that reported for OMVs (1.2 g/mL) (15,16) (Fig. S4b). Next, OMVs from each single gene-deleted E. coli strain were prepared by centrifugation. OMVs derived from the DompA, DfepA, DcirA, DdegP, and DompC strains contained five proteins (Fig. S4c). The increased expression of iNOS and COX-2 proteins by donor Mw exposed to OMVs was the same for the wild-type (WT) strain and the five single gene-deleted strains (Fig. 4a). In recipient Mw , expression of iNOS and COX-2 proteins increased in the presence of exosomes from Mw treated with OMVs from the DompA, DompC, DfepA, and DdegP strains ((DompA-exo, DompC-exo, DfepA-exo, and DdegP-exo, respectively) (Fig. 4b)). In contrast, expression of COX-2 protein by exosomes from Mw -treated OMVs from the DcirA strain (DcirA-exo) decreased 0.51-fold in comparison with treatment by OMVs from the WT strain (WT-exo). Expression of iNOS by DcirA-exo decreased 0.52-fold compared with that by OMVs from the WT strain, albeit not significantly. Similar to protein levels, the increase in expression of iNos and The number of OMV proteins is shown in the bar graph (n = 1). The experiment was repeated with at least three independent biological replicates. (c) OMVs from 1.2 mL of culture medium were prepared by ultracentrifugation. The protein expression of OmpA, an OMV marker, was detected by Western blotting. (d) OMVs were visualized with a transmission electron microscope. (e) After RAW264.7 cells were incubated with or without Ec-med, Ec-sup, or OMVs for 9 h, cells were incubated with fresh medium for 1 h. After we changed fresh medium again, cells were incubated for 9 h. The protein expression of inducible nitric oxide synthase (iNOS) and cyclooxygenase-2 (COX-2) in whole-cell lysates was measured by Western blotting. b-Actin was used as loading control. (f) After RAW264.7 cells (recipient cells; 5 Â 10 5 cells) were incubated with each exosome from donor cells (5 Â 10 5 cells) for 24 h, the protein expression of iNOS and COX-2 in whole-cell lysates was measured by Western blotting. b-Actin was used as loading control. The protein levels of iNOS, COX-2, and b-actin were quantitatively analyzed by using CS Analyzer 3.0 as image analysis software. Relative intensity was expressed as the mean 6 SD (n = 3) of at least three independent biological replicates. One-way ANOVA and Tukey's test were used for statistical analysis; *, P , 0.05; ***, P , 0.005; ****, P , 0.001. Cox-2 mRNA decreased 0.56-fold and 0.32-fold, respectively, in recipient Mw exposed to DcirA-exo compared to WT-exo (Fig. 4c). The increase in iNOS and COX-2 protein and mRNA in recipient Mw treated with OMVs from the DompC strain was significantly higher than after treatment with OMVs from the WT strain (iNOS protein, 11.2-fold; iNos mRNA, 206-fold; COX-2 protein, 3.3-fold; and cox-2 mRNA, 370-fold). In addition, expression of TNF-a and IL-1b was decreased by DcirA-exo to a greater extent than by WT-exo (Fig. S5). These results suggest that the cirA-deleted strain suppressed exosome-mediated inflammatory responses.
Complementation of the cirA gene promotes exosome-mediated inflammatory responses. Next, we examined whether WT-exo contained the CirA protein. WT-exo, but not none-exo, contained the CirA protein together with exosome marker proteins HSP90, CD63, and GAPDH (Fig. 5a). To confirm that CirA is a key factor in exosomemediated inflammatory responses, a pTV-cirA vector expressing CirA protein, or a pTV-118N empty vector was transformed into the E. coli DcirA strain. In bacterial cells and OMVs, the protein levels of CirA in the cirA gene-complementary strain (DcirA/cirA) were the same as those in the WT strain (Fig. 5b). The increase in iNOS and COX-2 proteins in donor Mw treated by OMVs was the same after exposure to the WT, DcirA (a, b) Exosomes from live E. coli-infected macrophages (Mw ) were treated with proteinase K (150 mg/ mL), DNase (100 mg/mL), or RNase (100 mg/mL) in the presence or absence of 0.01% Triton X-100. Mouse RAW264.7 Mw was incubated with each exosome (0.1 mg/mL). (a) After incubation of cells for 24 h, the mRNA levels of inducible nitric oxide synthase (iNos) and cyclooxygenase-2 (Cox-2) were analyzed by quantitative real-time PCR and normalized to 18S rRNA. Values are expressed as the mean 6 SD (n = 4) of at least three independent biological replicates. One-way ANOVA and Tukey's test were used for statistical analysis; **, P , 0.01; ***, P , 0.005; ****, P , 0.001. (b) After incubation of cells for 24 h, the protein expressions of iNOS and COX-2 were measured by Western blotting. b-Actin was used as loading control. (c) Exosomes (none-exo, live-exo, or HI-exo) from Mw uninfected or infected with live or heat-inactive (HI) E. coli were analyzed by mass spectrometry. Venn diagrams show the number of identified proteins.    strain, and DcirA/pTV and DcirA/cirA strains (Fig. 5c). However, in recipient Mw , the increase in iNOS and COX-2 mediated by DcirA/cirA-exo was the same as the mediated by WT-exo, while that mediated by DcirA-exo and DcirA/pTV-exo was lower than that mediated by WT-exo (Fig. 5d). The increase in the mRNA levels of iNos, Cox-2, TNF-a,   IL-1b, and IL-6 mediated by DcirA/cirA-exo and WT-exo was higher than that mediated by DcirA-exo ( Fig. 5e; Fig. S6c). Therefore, CirA plays an important role in exosomemediated inflammatory responses. The C-terminal protein of CirA promotes exosome-mediated inflammatory responses. To identify the CirA region responsible for the increase in protein and mRNA expression of proinflammatory mediators in recipient Mw , we divided the CirA protein into two fragments, an N-terminal fragment comprising amino acids 1 to 157 (N-CirA) and a Cterminal fragment comprising amino acids 158 to 633 (C-CirA) (Fig. 6a). Expression of N-CirA or C-CirA in protein in DcirA strains (DcirA/N-cirA or DcirA/C-cirA) transformed with the pTV-N-cirA or pTV-C-cirA strains was observed in bacterial cells and OMVs (Fig. 6b). The increase in expression of iNOS and COX-2 in donor Mw treated by OMVs was the same in the WT, DcirA (iNOS, 1.00-fold; COX-2, 0.87-fold), DcirA/N-cirA (iNOS, 0.88-fold; COX-2, 1.18fold), and DcirA/C-cirA (iNOS, 0.74-fold; COX-2, 1.00-fold) strains (Fig. 6c). In recipient Mw , expression of iNos and Cox-2 by DcirA/C-cirA-exo (iNos, 1.07-fold; Cox-2, 2.58-fold) was as same as that by WT-exo, while that by DcirA-exo (iNos, 0.37-fold; Cox-2, 0.23-fold) and  were incubated with OMVs (10 mg/7 mL) for 9 h. After the culture medium was replaced with fresh medium containing antibiotics, cells were incubated for 1 h. After replacement with fresh media containing antibiotics again, cells were incubated for 9 h. The protein expression of inducible nitric oxide synthase (iNOS) and cyclooxygenase-2 (COX-2) in whole-cell lysates was measured by Western blotting. b-Actin was used as loading control. CS Analyzer 3.0 was used for image analysis software. Relative intensity was expressed as the mean 6 SD (n = 3) of at least three independent biological replicates. (d) Exosomes were isolated from cell culture medium by ultracentrifugation. RAW264.7 cells (recipient cells; 5 Â 10 5 cells) were incubated with each exosome from donor cells (5 Â 10 5 cells). After incubation for 12 h, the mRNA levels of iNos and Cox-2 were analyzed by quantitative real-time PCR and normalized to 18S rRNA. Values are expressed as the mean 6 SD (n = 6 to 9) of at least three independent biological replicates. One-way ANOVA and Tukey's test were used for statistical analysis; *, P , 0.05; **, P , 0.01; ***, P , 0.005; ****, P , 0.001. DcirA/N-cirA-exo (iNos, 0.31-fold; Cox-2, 0.07-fold) was lower (Fig. 6d). These results show that C-terminal fragment of CirA triggers inflammatory responses via E. coli-derived OMVs and Mw -derived exosomes.

DISCUSSION
EVs are membrane-bound vesicles from cells. EVs produced during an infection can be derived from pathogens or the host. Here, we show that E. coli-derived OMVs, and exosomes derived from infected Mw , rely on inflammatory responses to naive Mw . Expression of iNOS and COX-2 by naive Mw increased after exposure to exosomes from live E. coliinfected Mw , but not Mw infected by dead (heat-inactivated) E. coli (Fig. 1). As for live E. coli, expression of inflammatory mediators by naive Mw was increased by exosomes derived from Mw , which were activated by E. coli-derived OMVs in the absence of bacterial cells (Fig. 2). These data suggest that E. coli-derived OMVs and Mw -derived exosomes trigger inflammatory responses in Mw far from the local area of bacterial infection. OMVs, nanostructures ranging from 20 to 300 nm in diameter, are released by almost all Gram-negative bacteria (17,18). The cargoes include an array of materials derived from the parental bacterium, including outer membrane, periplasmic, and cytosolic proteins, peptidoglycans, LPS, DNA, and RNA (17). Some pathogenic bacteria secrete virulence factors within OMVs, which interact with pathogenically relevant human cells. The OMVs released by enterohemorrhagic E. coli O157 include Siga toxin 2a, cytolethal distending toxin V, and hemolysin (19). OMVs from Enterotoxigenic E. coli also include physiological, active, heat-labile enterotoxin (16). Helicobacter pylori-derived OMVs contain the oncogenic cytotoxin-associated CagA protein, which is a major virulence factor. It is thought that OMVs facilitate direct delivery of virulence factors to target host cells rather than to the external environment, where diffusion has less functional impact (20). So what effect do nonpathogenic bacteria-derived OMVs have on host cells? A previous proteomics analysis of E. coli DH5a-derived OMVs showed that inclusion of particular proteins within OMVs does not appear to be strictly dependent on their abundance, suggesting that specific protein-sorting mechanisms operate during production of OMVs (15). We identified major components of the outer membrane protein, OmpA, and low-abundance outer membrane protein, FepA, in E. coli-derived OMVs. However, it was unclear whether proteins contained in OMVs from nonpathogenic bacteria also have cytotoxic effects on host cells. In contrast, OMV-contained LPS activates cytosolic caspase-11 (12). LPS is an abundant component of OMVs from pro-and nonpathogenic Gram-negative bacteria. Our results show that LPS-treated donor Mw showed increased expression of inflammatory mediators, while exosomes from LPS-treated Mw failed to increase expression of these mediators in recipient Mw (see Fig. S2b in the supplemental material). Therefore, LPS from OMVs does not play a role in exosome-related inflammatory responses. More than 1 ng/mL LPS increased expression of iNOS in naive Mw (Fig. S2c). However, the LPS concentration in Mw -derived exosomes isolated from Mw culture medium was 0.0026 6 0.0001 ng/mL (Fig. S2d). Thus, LPS in both OMVs and exosomes did not promote the inflammatory responses by exosomes from Mw that were infected with E. coli. In relation to inflammatory responses relayed by the two EVs, we found that proteins in exosomes were the key factors, while DNA and RNA did not have a role (Fig. 3a and b). The increase in iNOS protein expression mediated by Mw -derived exosomes was decreased by treatment with proteinase in the presence or absence of Triton X-100. These results show that the key factors that increase expression of iNOS are on the outside and/or inside exosomes. In contrast, the factors inducing COX-2 were presence inside the exosomes because increased expression of COX-2 by Mw -derived exosomes was decreased by protease with Triton X-100, but not without Triton X-100. We identified the CirA protein as an important factor that relays inflammatory signals. Indeed, the CirA protein was visualized in exosomes from Mw treated with E. coli-derived OMVs (Fig. 5a). Moreover, exosomes from Mw treated with OMVs derived from the DcirA strain caused a slight increase in iNOS expression while strongly suppressing expression of COX-2 ( Fig. 4b and Fig. 5e). These results show that proteins other than CirA on the membrane of exosomes must be involved in increased expression of iNOS. CirA, which is a receptor of colicin Ia (21,22), is expressed in E. coli-derived OMVs (15). CirA is a TonB-dependent transporter; spanning of the periplasmic space is facilitated by the inner membrane protein TonB (22). The transmembrane b-barrel, consisting of residues 164 to 663, contains 22 beta strands connected by 11 long extracellular loops and 10 short periplasmic loops. The N-terminal plug domain interacts with the interior of the b-barrel. The TonB box, comprising residues 26 to 163, resides at the N-terminal end of the plug domain. The TonB box is required for killing of cells by colicin Ia and plays an important role in transport of colicin Ia by CirA. Therefore, we confirmed whether the N-terminal plug domain was required for relay of inflammatory responses from E. coli-derived OMVs to Mw -derived exosomes. The inflammatory responses were dependent on the Cterminal transmembrane b-barrel, not on the N-terminal plug domain (Fig. 6). Exosomes from Mw treated with OMVs increased expression of TNF-a, IL-1b, and IL-6 in a CirA-dependent manner (Fig. S6b and Fig. S7). Thus, CirA in Mw -derived exosomes acts as a proinflammatory factor during infection with E. coli. It is not known whether CirA receptor is present in Mw . In contrast, OmpA, an outer membrane protein of E. coli, interacts with a 95-kDa glycoprotein on brain microvascular endothelial cells (Egcp), which allows E. coli to breach the blood-brain barrier (23). In the present study, we found that OmpA played no role in exosome-mediated inflammatory responses during infection with E. coli. The OmpA receptor, Egcp, might be not expressed by Mw . Another outer membrane protein, OmpC, did not play a role in exosome-mediated inflammatory responses. When E. coli lacking OmpC were cultured under different glucose concentrations, expression of CirA on the outer membrane changed (24). Here, we showed that deletion of OmpC did not alter expression of CirA by OMVs (Fig. S3c). Moreover, OmpC was not directly related to exosome-mediated inflammatory responses, but might be a target for inhibiting them. The protein content of OMVs from the DompC strain was one-tenth that of the WT strain, suggesting that decreased expression of OmpC might suppress release of OMVs (Fig. S4b). Because inflammatory responses were dependent on the dose of OMVs, reducing the amount of OMVs may be one effective way to stop relay of exosome-mediated inflammatory responses from Mw to Mw (Fig. S1). Here, we treated donor Mw with the same amount of protein (10 mg/8 Â 10 6 cells) and found that OMVs from the DompC strain strongly increased expression of iNOS and COX-2. We considered that high concentrations of E. coli-secreted molecules, including LPS, were contained within OMVs from the DompC strain. TonB-dependent outer membrane transporters include FecA, the ferric citrate transporter; BtuB, the vitamin B 12 receptor; FhuA, the ferrochrome receptor; and CirA-FepA, the ferric-enterochelin receptor (25,26). In our study, we confirmed that the protein components of the TonB-dependent transporter are present in E. coli-derived OMVs. CirA and FepA share the highest sequence similarity (30.8%) within the TonB-dependent group. Moreover, the association between CirA and FepA could be extended to encompass a close evolutionary relationship (25). Both the CirA and FepA transporters play a role in colicine binding receptor and the uptake of ferric iron (Fe 31 ) complexed with siderophores. CirA was identified as a receptor of microcin that is of a lower molecular weight than colicins (27). Thus, the function of CirA and FepA are not exactly the same. On host cells, CirA relays inflammatory responses from donor Mw to recipient Mw , while FepA did not. Some studies show that the release of extracellular microvesicles, increased by the lack of iron in a culture system, triggers a stress response in bacterial species such as Haemophilus influenzae, Mycobacterium tuberculosis, and uropathogenic E. coli (28)(29)(30). We found that the amount of OMVs released by E. coli also increased as the iron concentration in the culture medium decreased (Fig. S8). We cultured E. coli under iron-limiting conditions (0.2 mM Fe 31 ) to prepare OMVs. Iron depletion increased expression of CirA in E. coli, resulting in the taking up of CirA into OMVs. E. coli K-12 strain-derived OMVs, which were analyzed by Reimer et al., did not contain the TonB-dependent transporters, including CirA (31); this must be because E. coli was cultured in iron-rich LB broth. Indeed, we found that the expression of CirA in bacterial cells decreased in an iron dose-dependent manner when more than 10 mM iron was added to the E. coli culture medium (Fig. S8). This phenomenon was also observed in Salmonella.
Thus, CirA appears to be selectively incorporated into OMVs from some CirA-expressing bacteria under iron-limited conditions. Furthermore, a low iron concentration would not only increase in the expression of CirA but also enhance the inflammatory responses relayed by OMVs and exosomes during E. coli infection of the host. Within hours of human infection, the concentration of iron in extracellular fluid and plasma decreases dramatically (32,33). This is thought to be a critical host defense strategy against bacterial pathogens (34). In contrast, our results suggest that iron-limiting conditions upregulate CirA expression, resulting in relay of inflammatory responses by OMVs and exosomes.
In the present study, we identified Mw -derived exosomes by measuring particle size and confirming marker proteins. However, it must contain other extracellular vesicles of different origin than the exosome. Budden et al. (35) characterized the extracellular vesicle released by human macrophage THP-1 cells upon activation of the NLRP3 inflammasome. Therefore, we examined whether E. coli-derived OMVs induced NLRP3 activation, resulting in activation of caspase-1 and pyroptosis via processing gasdermin D. Caspase-1 was activated in mouse BMDM after stimulation by E. coli-derived OMVs (Fig. S9f). In contrast, cleaved caspase-1 was not detected in RAW264.7 cells because ASC, a factor in the inflammasome complex, was reported to be completely absent in RAW264.7 cells (Fig. S9a). Furthermore, E. coli-derived OMVs are known to be sensed by caspase-11 and to activate the cell death-inducing protein gasdermin D independently of caspase-1 (12). In fact, the expression of caspase-11 was stimulated in both RAW264.7 cells and mouse BMDM; however, gasdermin D was activated only in mouse BMDM (Fig. S9b, c, f, and g). Intracellular ATP levels were decreased by E. coli-derived OMVs, while cell proliferation measured by 3-(4,5-dimethyl-2-thiazolyl)-2,5-diphenyl-2H-tetrazolium bromide (MTT) assay resulted in an increase by E. coli-derived OMVs (Fig. S9d, e, i, and j). In addition, the extracellular lactate dehydrogenase, as cell death, was not detected in culture medium by Mw treated with or without E. coli-derived OMVs (data not shown). These data suggest that the dose of E. coliderived OMVs used in this study caused little pyroptosis. In fact, more than 50 mg of OMVs was reported to cause pyroptosis (12).
We demonstrate that the CirA protein in the OMVs from live E. coli is incorporated into Mw -derived exosomes and increases protein expression of inflammatory mediators. However, proteins other than CirA may play roles in exosome-mediated inflammatory responses. In particular, increased expression of iNOS was partially attenuated by the DcirA E. coli strain. Other factors will be clarified in a future study.
Plasmids and bacterial strains. Cloning of the E. coli cirA gene (GenPept accession no. NP_416660) was amplified using Ex Taq with the primers pMD-cirANdeI-s and pMDcirANotI-a and E. coli (WT) genome DNA as the template. The pMD-cirA plasmid was ligated with the amplified fragments and pMD T vector by using Ligation Mighty Mix (TaKaRa). The pET-cirA plasmid was ligated with the fragments cirANdeI-NotI of pMD-cirA digested by NdeI and NotI restriction enzymes and the NdeI-NotI-digested pET-22b(1) plasmid by using ligation mix (TaKaRa). The cirA-6ÂHis fragments were amplified with the primers pTV-cirANcoI-s and pTV-cirANcoI-a and pET-cirA as the template. The pTV-cirA plasmid was ligated with the amplified fragment of cirA-6ÂHis and NcoI-digested pTV plasmid by the In-Fusion HD. The fragment of N/C-cirA was amplified with plasmids pET-NcirANcoI-s and pET-NcirANotI-a, or pET-CcirANcoI-s and pET-CcirANotI-a, and pET-cirA plasmid as the template. The pET-N/CcirA plasmids were ligated with amplified fragments N/CcirA, and NcoI-digested pET plasmid by In-Fusion HD. The N/CcirA-6ÂHis fragments were amplified with the primers pTV-N/CcirANcoI-s and pTV-N/CcirANcoI-a and pET-N/CcirA as the template. The expression vectors (pET-ompA, pET-fepA, pET-cirA, and pET-degP) were transformed into the competent E. coli strain BL21(DE3) strain. Bacteria were incubated in LB broth with 100 mg/mL ampicillin at 37°C with shaking until the optical density at 600 nm (OD 600 ) reached about 0.6. After incubation of bacteria at 22°C overnight with 0.1 mM IPTG, the bacterial cells were pelleted at 2,000 Â g at 4°C for 30 min, and the supernatant was removed. The cells were lysed by lysis buffer (20 mM Tris-HCl, pH 8.0, 500 mM NaCl, 10% [vol/vol] glycerol, and 8 M urea) and were completely  E. coli culture and preparation of Ec-med, Ec-sup, and OMVs. The E. coli DH5a strain, E. coli K-12 BW25113 strain (WT), and E. coli deletion mutants were cultured at 37°C in LB broth. Bacteria were prepared at an OD 600 of 0.0008 by Dulbecco's modified Eagle medium (DMEM) containing 4.5 g/L glucose and 0.2 mM Fe 31 (Fujifilm, Osaka, Japan) and then were cultured at 37°C for 12 h under 5% CO 2 conditions. ClearColi (E. coli BL21(DE3) mutant lacking the oligosaccharide chain of LPS; Lucigen, Buenos Aires, Argentina) was gifted from Sohkichi Matsumoto (Department of Bacteriology, Niigata University Graduate School of Medical and Dental Sciences). Similarly, ClearColi bacteria were cultured at 37°C for 27 h under 5% CO 2 conditions. After centrifugation at 2,000 Â g for 30 min, the bacterial culture medium was collected as supernatant. After we centrifuged the bacterial culture medium at 10,000 Â g at 4°C for 30 min, the supernatant was filtrated by a 0.2-mm filter (Kurabo). After we ultracentrifuged the filtrate (Ec-med) using a 50.2 Ti rotor (Beckman) at 100,000 Â g at 4°C for 3 h, the supernatant (Ec-sup) was removed, and phosphate-buffered saline (PBS) was added in pellets. After we ultracentrifuged the filtrate at 100,000 Â g at 4°C for 2 h, pellets (OMVs) were suspended by PBS.
Density gradient of OMVs. After we ultracentrifuged the filtrate at 100,000 Â g at 4°C for 3 h, pellets suspended by PBS were used as OMVs. Moreover, the OMVs were purified using an OptiPrep (Axis-Shield Diagnostics, XA, Scotland) density gradient. A discontinuous iodixanol gradient was prepared by diluting a stock solution of OptiPrep (60%, wt/vol) with 10 mM Tris, pH 7.5, containing 0.25 M sucrose to generate 40%, 20%, 10%, and 5% (wt/vol) iodixanol solutions. The discontinuous iodixanol gradient was generated by sequentially layering 3.0 mL of 40% (wt/vol) iodixanol solutions, 3.0 mL of 20% (wt/vol) iodixanol solutions, 3.0 mL of 10% (wt/vol) iodixanol solutions, and 2.5 mL of 5% (wt/vol) iodixanol solutions in a centrifuge 344060 tube (Beckman). A 0.3-mL volume of OMVs in PBS was overlaid on discontinuous iodixanol gradient and ultracentrifuged using an SW40 Ti rotor (Beckman) at 160,000 Â g at 4°C for 16 h. Twelve fractions of 1 mL each were collected from the top of the iodixanol solution and used for analysis such as Western blotting.
TEM analysis. OMV samples were applied to the carbon film TEM grids, which were created by carbon evaporation using VE-2020 (Vacuum Device, Tokyo, Japan). After staining with 1% uranyl acetate, samples were observed in a Talos F200C G2 (Thermo).
Donor cell culture. RAW264.7 cells (8 Â 10 6 cells) were cultured in 7 mL of DMEM containing 10% fetal bovine serum (Biowest, Tokyo, Japan) and 100 units/mL penicillin and 100 mg/mL streptomycin (Gibco/ Thermo, Tokyo, Japan). After preincubation for 24 h, the medium was changed to fresh DMEM without antibiotics. Bone marrow was flushed from the femur and tibia of C57BL/6 mice (male, 5 weeks old), and cells were plated in dishes with DMEM containing 10% fetal bovine serum (FBS), 4.5 g/L glucose, and 20% conditioned medium from the supernatants of L929 (LC14) fibroblasts secreting macrophage colony-stimulating factor. Bone marrow cells were differentiated into macrophages in 7 to 10 days and were then used in experiments (36). All mice were maintained under specific-pathogen-free conditions in the animal facilities of Kyoto Prefectural University. After live or heat-inactivated E. coli DH5a and E. coli BW25113 were added to cells at an MOI of 5 for 9 h, culture media were replaced with fresh DMEM medium containing antibiotics, and the cells were incubated for 1 h. After the medium was replaced again with fresh DMEM medium containing antibiotics, cells were incubated for 9 h. After Ec-med, Ec-sup, or OMVs were added to cells for 9 h, culture media were replaced with fresh DMEM medium containing antibiotics, and the cells were incubated for 1 h. After the medium was replaced again with fresh DMEM medium containing antibiotics, cells were incubated for 9 h. The culture medium was collected for preparation of exosomes, and cells were used for analyzing immunoblotting or real-time PCR.
ATP measurement. RAW264.7 cells and mouse bone marrow-derived macrophages (BMDMs) (3 Â 10 3 cells/well) were spread in a 96-well plate and cultured for 24 h at 37°C. After E. coli-derived OMVs (0.143 mg/0.1 mL/well) were added to cells for 9 h, culture media were replaced with fresh DMEM medium containing antibiotics, and the cells were incubated for 1 h. After the medium was replaced