Naturally Produced Outer Membrane Vesicles from Pseudomonas aeruginosa Elicit a Potent Innate Immune Response via Combined Sensing of Both Lipopolysaccharide and Protein Components

Bacterial strains.The P. aeruginosa strains used were the laboratory strains PAO1 and PAO1ΔfliC and the clinical isolate S470. PAO1ΔfliC was provided by Dan Wozniak (Ohio State University). Strain S470 was provided by J. R. Wright (Duke University) and is a minimally passaged, nonmucoid cystic fibrosis clinical isolate.

Vesicle preparation.Vesicles were purified from cell free supernatants by the method described by Bauman and Kuehn (3). Briefly, P. aeruginosa was grown in LB broth to early stationary phase. Cells were pelleted and supernatants were concentrated via a 100-kDa tangential filtration concentration unit (Pall-Gellman) to ∼500 ml. The retentate was centrifuged and filtered through a 0.45-μm-pore-size Durapore polyvinylidene difluoride filter (Millipore) to remove the remaining bacteria. Vesicles were precipitated from the cell-free supernatant with ammonium sulfate, pelleted, dialyzed overnight with HEPES (50 mM, pH 7), concentrated (50-kDa MWCO Centriplus; Millipore), and adjusted to 45% Opti-Prep/HEPES-NaCl. Opti-Prep gradients were layered over the 2-ml crude vesicle samples at concentrations of 40, 35, 30, 25, and 20%. Gradients were ultracentrifuged (100,000 × g, 16 h), and 1-ml fractions were removed from the top. Fractions were analyzed for vesicle protein content by SDS-15% PAGE and Ruby Red staining. Pure vesicles were recovered from pooled peak fractions of the Opti-Prep gradients, reconstituted into phosphate-buffered saline (PBS; 130 mM NaCl, 8 mM Na₂HPO₄, 2 mM KCl, 1 mM KH₂PO₄ [pH 7.4]). The protein concentration of the vesicle preparations was determined by Bradford assay.

To calculate the quantity of LPS in each sample, serial dilutions of LPS standards and purified vesicles were separated by SDS-15% PAGE and stained for LPS with a PRO-Q Emerald LPS staining kit (Molecular Probes). The density of the LPS B bands was determined by using ImageJ software, and the LPS content in the vesicles was calculated by using the standard curve generated from the pure LPS samples.

LPS purification.LPS from P. aeruginosa was purified by a modified version of the method of Zaidi et al. (55). Briefly, cells were lysed in 10% sarcosine, debris was pelleted, and LPS was precipitated overnight with 4 volumes of 95% ethanol. LPS was pelleted by ultracentrifugation, resuspended in PBS, and digested with DNase, RNase, and pronase. The digested material was ultracentrifuged to pellet the LPS, and the LPS was lyophilized.

Fluorescent labeling.Purified vesicles were fluorescently labeled by incubation with Alexa Fluor 555-succinimidyl ester (AF555 [Invitrogen], in 0.1 M Na₂CO₃ [pH 9]) according to the manufacturer's instructions for 1 h at 25°C with mixing. Free AF555 was removed from labeled vesicles by three washes in PBS (150,000 × g, 30 min).

Cell culture.RAW264.7 or MH-S mouse macrophage cells were grown in Kaighn's F-12K media with 10% fetal bovine sera and penicillin-streptomycin-amphotericin B (Fungizone). Confluent cells were harvested and seeded into wells of a 96-well plates at a concentration of 5 × 10⁴ cells/well. Vesicles, purified LPS, or sham PBS treatment (10-μl doses of each) were added to each well, followed by incubation until time of assay.

Polymyxin B treatment.Purified vesicles or LPS were preincubated with 0.1 mg of polymyxin B/ml for 30 min at room temperature prior to addition to macrophages. Polymyxin B (50 μg/ml) was also added to the macrophage cell culture media for experiments using the polymyxin B-pretreated vesicles or LPS.

Proteinase K treatment.Purified vesicles were treated with 0.1 mg of proteinase K/ml overnight at 37°C with agitation. Postincubation, soluble proteins and excess proteinase K were removed from vesicles by three washes in PBS (150,000 × g, 30 min).

Inhibition of TLR9.RAW264.7 cells were pretreated for 1 h with 3 μg of nuclease-resistant phosphorothioate oligodeoxynucleotide (iCpG DNA; ODN 2088 5′-TCCTGGCGGGGAAGT-3′)/ml to inhibit TLR9 signaling (50).

Real-time PCR.Total RNA was harvested from cells by using a Qiagen RNeasy kit and DNase treated to remove contaminating DNA. Harvested RNA was quantitated, and 1 μg of RNA was reverse transcribed to cDNA using oligo(dT) primers and Superscript II reverse transcriptase (Invitrogen). This cDNA was purified from reaction components by using a Qiagen DNA clean-up kit and reconstituted in a final volume of 50 μl. Then, 1 μl of purified cDNA was used as the template for each real-time PCR. Real-time PCR was performed on a Bio-Rad iCycler using the iQ SYBR green Supermix. Primers for target genes are presented in Table 1. Gene expression was normalized to actin production in each sample, and the fold induction was determined by using the ΔΔC_T method (32).

ELISA.Cytokine protein production (macrophage inflammatory protein 2 [MIP-2], IL-1β, IL-6, and TNF-α) was quantified by enzyme-linked immunosorbent assay (ELISA; R&D Systems/BD) according to the manufacturer's instructions.

Confocal microscopy.All fluorescence microscopy reagents were purchased from Molecular Probes/Invitrogen unless otherwise stated. MH-S mouse alveolar macrophages were seeded overnight onto coverslips in 24-well tissue culture plates (10⁵ cells per well), followed by incubation with fluorescently labeled vesicles for 15 min at 37°C. All subsequent steps were carried out on ice using ice-cold Dulbecco phosphate-buffered saline (PBS) for washes. After incubation with vesicles, cells were washed twice to remove unbound vesicles or LPS. Cell exteriors were labeled with the FluoReporter cell surface biotinylation kit, washed twice, and incubated with AF633-conjugated streptavidin. Cells were then washed, fixed in neutral buffered formalin, mounted with ProLong AntiFade reagent, and visualized on a Leica SP5 confocal microscope.

Statistics.All experiments were performed with n ≥ 4. Statistical comparisons between groups were performed using single-factor analysis of variance, followed by Tukey's post hoc test. The statistical significance was set at P ≤ 0.05.

Macrophages are more sensitive to outer membrane vesicles than to pure LPS.Within the lung, macrophages are critical to sensing and initiation of the inflammatory host response to P. aeruginosa (22). In order to analyze how sentinel macrophages trigger inflammation in response to outer membrane vesicles secreted by P. aeruginosa, we developed an in vitro assay to measure proinflammatory cytokine gene expression by macrophages. Outer membrane vesicles were collected from the overnight culture supernatant of clinical P. aeruginosa strain S470 and purified by density gradient. Proinflammatory genes (MIP-2, KC, TNF-α, IL-1β, IL-6, and iNOS) were selected to be assayed based on their previously described upregulation during active P. aeruginosa infection (5, 11, 21, 22, 41, 42, 45, 51, 52). RAW 264.7 macrophages were coincubated with vesicles for 3 h, and gene expression was assayed by real-time PCR and normalized to actin production in each sample. The fold induction was determined by using the ΔΔC_T method (32). To determine the minimal dose of vesicles required to trigger a response, we used a 10-fold dilution series of vesicle doses which were standardized to LPS content. A 3-h exposure elicited optimal upregulation of gene expression (data not shown), and therefore this condition was used in subsequent experiments.

Outer membrane vesicles contain significant quantities of LPS; however, the membrane context of vesicle LPS might activate a different cytokine response than extracted, pure LPS. To compare both the amount and the type of response, cytokine levels were measured for macrophages exposed to equivalent concentrations of LPS either purified or as part of a pure vesicle preparation. As seen in Fig. 1A, both vesicles and pure LPS stimulated significant transcription of MIP-2, TNF-α, IL-1β, and IL-6. Other genes assayed, such as KC and iNOS, did not exhibit significant upregulation in response to either vesicles or LPS (data not shown). Proinflammatory gene expression was clearly dose dependent. At the highest dose (1 μg of LPS content), both treatments triggered similar expression levels by all cytokines assayed, indicating that this may be a saturation level response (data not shown). However, sequential dilution of doses demonstrated an increasing divergence in the response with vesicles triggering a more profound cytokine response than pure LPS as the dose decreased. Since secreted vesicles comprise a small fraction (ca. 1 to 2%) of the total outer membrane protein of a growing culture (3), these lower doses represent a physiologically relevant point of comparison.

• Open in new tab
• Download powerpoint

FIG. 1.

The inflammatory response of macrophages to outer membrane vesicles is more potent than to pure LPS. RAW 264.7 mouse macrophages were incubated with vesicles or purified P. aeruginosa LPS for 3 h. S470 vesicles and LPS doses were created from 10-fold serial dilutions of stocks containing 1 μg of LPS. (A) Gene expression was assayed by real-time PCR. (B) Protein secretion was assayed from culture supernatants by ELISA. The bars represent the standard error of the mean (SEM) (n = 4). *, P < 0.05; **, P < 0.01.

The four cytokines that demonstrated significant transcriptional upregulation were further analyzed by ELISA for protein secretion into culture supernatant over the course of a 3-h exposure to either vesicles or purified LPS. Figure 1B shows that similar levels of MIP-2 were secreted over this time period in response to both LPS and vesicles, indicating that recognition of LPS within vesicles may be the dominant stimulus needed to generate this cytokine. Neither treatment stimulated significant production of IL-1β, even through significant amounts of IL-1β RNA transcripts were present, as assayed by real-time PCR. In contrast, both TNF-α and IL-6 were secreted at significantly higher levels in response to outer membrane vesicles than LPS. Thus, P. aeruginosa vesicles are potent stimulators of macrophages that lead to the upregulated secretion of MIP-2, TNF-α, and IL-6.

CpG DNA does not significantly contribute to the inflammatory response to vesicles.Outer membrane vesicles from Pseudomonas have been observed to contain bacterial DNA, which is a ligand for endosomal TLR9 (43). To investigate the contribution of CpG DNA in the response to vesicles, TLR9 signaling was inhibited by pretreatment of RAW 264.7 cells with an inhibitory oligonucleotide (iCpG), which is a competitive inhibitor of CpG binding to TLR9. As seen in Fig. 2, inhibition of TLR9 signaling with iCpG had no effect on cytokine responses to S470 vesicles. No significant change in the cytokine transcriptional profiles was seen at any of the vesicle doses tested, indicating that CpG DNA is not a significant trigger of inflammatory responses to vesicles.

• Open in new tab
• Download powerpoint

FIG. 2.

CpG DNA is not a major contributor to the inflammatory response to outer membrane vesicles. The contribution of CpG DNA was determined by preincubating RAW 264.7 cells with CpG inhibitory oligodeoxynucleotides prior to the addition of vesicles. Gene expression was assayed by real-time PCR normalized to actin production in each sample. The bars represent the SEM (n = 4). Statistical analysis revealed no significant differences between the pairs of treatments for each gene product.

Flagellin stimulates MIP-2 and IL-6 secretion in response to outer membrane vesicles.Since flagellin is present in the protein profiles of P. aeruginosa outer membrane vesicles and is a potent innate immune stimulator via both TLR5 and the NOD-like receptor IPAF (3), we investigated the contribution of flagellin to the cytokine response to vesicles using vesicles from strain PAO1 and its isogenic flagellin mutant (ΔfliC). We quantified the flagellin bands from SDS-PAGE profiles of both PAO1 and S470 derived vesicles using densitometric analysis and demonstrated that these preparations contained equivalent levels of flagellin (data not shown). We could not use RAW 264.7 cells for these experiments because these cells do not express TLR5 (53). Therefore, we used the MH-S alveolar macrophage cell line, which is known to express TLR5. We noted first that the MH-S cell line exhibited much lower sensitivity to vesicles than did RAW 264.7 cells: the lowest vesicle dose (0.001 μg of LPS content) elicited only modest transcriptional increases for any of the genes assayed, these increases did not result in measurable cytokine levels by ELISA and, unlike RAW 264.7 cells, the MH-S cells did not secrete TNF-α in response to any of the vesicle doses (Fig. 3 and data not shown). However, ELISA analysis indicated that vesicles did induce both MIP-2 and IL-6 secretion and that these were significantly lower in response to ΔfliC vesicles across multiple doses. Thus, the flagellin component of vesicles contributes significantly to the stimulation of MIP-2 and IL-6 secretion by macrophages.

• Open in new tab
• Download powerpoint

FIG. 3.

Flagellin contributes to the MIP-2 and IL-6 responses to outer membrane vesicles. The contribution of flagellin was determined using MH-S mouse macrophage cells incubated with vesicles from the PAO1 or PAO1ΔfliC strains. (A) Fold changes in gene expression were assessed by real-time PCR normalized to actin production. (B) Protein secretion was assayed from culture supernatants by ELISA. The bars represent the SEM (n = 4). *, P < 0.05; **, P < 0.01.

Strain-specific proteins impact the inflammatory responses to vesicles.We previously demonstrated that the protein composition of vesicles varied with bacterial strain and can affect their association with host cells (2, 3). Therefore, we investigated whether the macrophage responses would also differ for vesicles from different strains of P. aeruginosa. A proteomic comparison of vesicles from the clinical isolate S470 and lab strain PAO1 demonstrated some significant differences in their protein composition (3). The ratios of LPS to protein content for these two vesicle preparations were highly similar (data not shown), so any observed differences in the responses would likely be the result of protein composition rather than LPS concentration.

Cytokine upregulation and secretion were assayed in response to vesicles from S470 and PAO1. As shown in Fig. 4A significant differences in gene expression by real-time PCR analysis were detected for the lowest dose (0.001 μg of LPS content). At this dose, vesicles from strain S470 upregulated significantly higher levels of all three cytokines assayed. Greater differences were detected when cytokine levels were measured by ELISA (Fig. 4B). MIP-2 secretion was significantly higher in response to PAO1 vesicles, whereas the secretion of both TNF-α and IL-6 was greater in response to S470 vesicles. TNF-α secretion above the baseline of untreated cells was not detected at the lowest dose for either vesicle treatment, nor were significant IL-6 levels detected in response to the lowest dose of PAO1 vesicles. These data indicate that strain-specific vesicle components affect both the nature and the sensitivity of the macrophage proinflammatory cytokine response.

• Open in new tab
• Download powerpoint

FIG. 4.

The cytokine responses to outer membrane vesicles are bacterial strain dependent. RAW 264.7 mouse macrophage cells were incubated with vesicles from clinical isolate S470 or lab strain PAO1 for 3 h. Gene expression was assayed by real-time PCR (A), and protein secretion was assayed from culture supernatants by ELISA (B). The bars represent the SEM (n = 4). *, P < 0.05; **, P < 0.01.

LPS and vesicle proteins act synergistically to generate a distinct inflammatory response.Two different biochemical treatments were used to investigate the contribution of particular vesicle components in more detail. First, vesicles were pretreated with polymyxin B to neutralize the contribution of LPS. Polymyxin B binds to the lipid A portion of LPS, inhibiting sensing by the TLR4 complex (27). Second, vesicles were incubated overnight with proteinase K to degrade vesicle proteins. Figure 5A shows the effect of proteinase K treatment on the vesicle proteins. The protein profile of the vesicles was significantly altered, with many high-molecular-weight bands disappearing after protease treatment. All proteins were not fully degraded, as they were likely to be protected from protease treatment by the vesicular membrane.

• Open in new tab
• Download powerpoint

FIG. 5.

LPS and vesicle proteins act synergistically to generate a distinct inflammatory response. The ability of the proteins present on vesicles to trigger a host response was determined by treating vesicles with proteinase K or polymyxin B. (A) SDS-PAGE of S470 outer membrane vesicles protein profiles before and after proteinase K digestion. Macrophages were exposed to treated vesicles for 3 h and responses were assayed by real-time PCR (B) or ELISA (C). The bars represent the SEM (n = 4). *, P < 0.05; **, P < 0.01.

Macrophages were incubated with either of these treated vesicle preparations, and cytokine expression was assayed by real-time PCR and ELISA. As seen in Fig. 5B and C, polymyxin B treatment of vesicles dramatically depressed both cytokine transcriptional and protein responses of all genes assayed. The significant inhibition of cytokine transcription by polymyxin B treatment was surprising given the higher potency of whole vesicles compared to LPS (Fig. 1B). Polymyxin B-treated LPS, used as a control for these experiments, generated minimal transcriptional responses by the macrophages (data not shown). These results indicate that LPS acts as an adjuvant in the reaction of macrophages toward vesicles, greatly enhancing the response to the non-LPS components of vesicles.

In comparison to polymyxin B, proteinase K treatment of vesicles resulted in a more selective dampening of cytokine responses. Proteinase K treatment depressed IL-6 production most significantly for all tested doses of vesicles, as measured by both reverse transcription-PCR and ELISA (Fig. 5B and C), indicating the likelihood that the production of this cytokine is triggered in response to outer membrane protein recognition. However, for several cytokines, induction levels increased with protease treatment of the vesicles. Secreted MIP-2 levels were higher than that seen in response to untreated vesicles (Fig. 5C). TNF-α secretion was most dramatically impacted by proteinase K treatment at the lowest vesicle dose (0.001 μg of LPS content), whereas significantly higher TNF-α secretion was seen compared to the response to either intact vesicles or pure LPS. Together, these data suggest that vesicles interact with macrophages in a manner that triggers a distinct, strain-specific response that reflects recognition of a combination of vesicle constituents.

Sensing of both LPS and vesicle proteins is required for internalization by macrophages.Based on their activation of distinct cytokine responses, we predicted that LPS, vesicles, and treated vesicles might interact differently with macrophages. We developed a confocal microscopy assay to compare how our preparations of vesicles associated with macrophages. MH-S alveolar macrophages were treated with untreated, proteinase K-treated, or polymyxin B-treated, Alexa Fluor 555-labeled S470 vesicles for 15 min. Vesicles were fluorescently labeled after proteinase K treatment, whereas polymyxin B treatment was performed on fluorescently labeled vesicles. After vesicle coincubation, the cells were washed to remove unbound material, surface biotinylated, and incubated with Alexa Fluor 633-streptavidin.

The different LPS-containing reagents yielded substantially distinct localization profiles (Fig. 6), giving insight into the molecular basis for the distinct pattern of activation of the macrophages. Untreated S470 vesicles appeared as punctate particles that were not labeled by streptavidin and thus were likely to have been internalized by the macrophages. Polymyxin B treatment resulted in few vesicles interacting with macrophages. The few vesicles that were observed were not labeled by streptavidin and thus were internalized by the macrophages. Proteinase K-treated vesicles adhered to the surfaces of macrophages and were labeled by streptavidin and thus not internalized. These data indicate that vesicle LPS is required for initial surface binding of vesicles to the macrophage surface, whereas recognition of vesicle protein triggers internalization.

• Open in new tab
• Download powerpoint

FIG. 6.

Macrophage recognition of outer membrane vesicle proteins is required for internalization by macrophages. MH-S alveolar macrophages were exposed to native, proteinase K-treated, or polymyxin B-treated Alexa Fluor 555-labeled vesicles for 15 min. At the indicated time points, the unbound label was washed off, and the cells were surface labeled with biotin and Alexa Fluor 633-streptavidin (blue) and visualized by confocal microscopy.