Aggregatibacter actinomycetemcomitans Outer Membrane Vesicles Are Internalized in Human Host Cells and Trigger NOD1- and NOD2-Dependent NF-κB Activation

Abstract

Aggregatibacter actinomycetemcomitans is an oral and systemic pathogen associated with aggressive forms of periodontitis and with endocarditis. We recently demonstrated that outer membrane vesicles (OMVs) disseminated by A. actinomycetemcomitans could deliver multiple proteins, including biologically active cytolethal distending toxin (CDT), into the cytosol of HeLa cells and human gingival fibroblasts (HGF). In the present work, we have used immunoelectron and confocal microscopy analysis and fluorescently labeled vesicles to further investigate mechanisms for A. actinomycetemcomitans OMV-mediated delivery of bacterial antigens to these host cells. Our results supported that OMVs were internalized into the perinuclear region of HeLa cells and HGF. Colocalization analysis revealed that internalized OMVs colocalized with the endoplasmic reticulum and carried antigens, detected using an antibody specific to whole A. actinomycetemcomitans serotype a cells. Consistent with OMV internalization mediating intracellular antigen exposure, the vesicles acted as strong inducers of cytoplasmic peptidoglycan sensor NOD1- and NOD2-dependent NF-κB activation in human embryonic kidney cells. Moreover, NOD1 was the main sensor of OMV-delivered peptidoglycan in myeloid THP1 cells, contributing to the overall inflammatory responses induced by the vesicles. This work reveals a role of A. actinomycetemcomitans OMVs as a trigger of innate immunity via carriage of NOD1- and NOD2-active pathogen-associated molecular patterns (PAMPs).

INTRODUCTION

Both Gram-negative and Gram-positive bacteria can extend their pathogenicity by releasing membrane vesicles (MVs), which represent a very basic and relevant mode of protein export from bacteria that has been referred to as “type zero” secretion (1). Via MVs, bacteria can expose host cells to relatively high concentrations of toxins and additional virulence factors without the requirement of a close contact between the bacterial and target mammalian cells (2–9). MVs have also several defensive functions, including a role in antimicrobial peptide resistance (6, 10). Reports that membrane vesicles are produced in vivo during infection, that patient sera show reactivity to MV antigens, and that circulating MVs cause sepsis in animal models together support the idea that such vesicles may have a pivotal role in effecting a toxic response in the host, beyond that provided by the infecting microorganism itself (11–13). In chronic localized infections, such as periodontitis, a common dental disease resulting in irreversible alveolar bone and attachment loss around teeth and eventual tooth loss (14), membrane vesicles may represent an important source of inflammatory stimulants both locally and systemically upon entry into the circulation (15).

Aggregatibacter actinomycetemcomitans is an oral and systemic human pathogen that is associated with aggressive forms of periodontitis and with endocarditis (16–18). The mechanisms by which A. actinomycetemcomitans causes alveolar bone resorption and systemic disease are not entirely understood. Outer membrane vesicles (OMVs) released by this organism carry multiple proteins, and density gradient centrifugation supports the notion that there may be subpopulations of vesicles exhibiting slight variations in protein composition (7, 19). Further to earlier reports revealing the OMV association of leukotoxin (LtxA) and a concomitant leukotoxic activity of A. actinomycetemcomitans OMVs (19, 20), we have demonstrated that the OMVs could simultaneously deliver multiple proteins, including OmpA and biologically active cytolethal distending toxin (CDT), into HeLa cells and human gingival fibroblasts (HGF) (7). Hence, OMVs released by this species likely play a role in periodontal disease by delivering biologically active toxins and additional virulence factors into susceptible cells of the periodontium. It has been suggested that in the presence of A. actinomycetemcomitans, CDT toxicity may play a part in the early pathogenesis of periodontitis (21, 22). This would be in accordance with OMVs promoting damage in the sulcular/junctional epithelium. Besides proteins, the lipid components of A. actinomycetemcomitans OMVs have been identified, revealing lipopolysaccharide (LPS) as one of the predominant lipid constituents (19).

It was not known if A. actinomycetemcomitans OMVs can be internalized into the interior of nonphagocytic human cells via endocytic pathways, analogously to vesicles derived from other species, such as Brucella abortus, Escherichia coli, Helicobacter pylori, Pseudomonas aeruginosa, Porphyromonas gingivalis, and Vibrio cholerae (23–28). In addition to protein delivery, internalization of OMVs represents an important mechanism to expose the host cells to pathogen-associated molecular patterns (PAMPs), e.g., peptidoglycan (PGN) fragments and LPS, which are associated with intact OMVs and/or with fragments from ruptured vesicles within the cell and are recognized by intracellular host pattern recognition receptors (PRRs). For example, vesicles from V. cholerae, H. pylori, P. aeruginosa, and Neisseria gonorrhoeae were demonstrated to deliver PGN to host cells to activate nucleotide-binding oligomerization domain-containing protein 1 (NOD1) and/or NOD2, two cytosolically expressed members of the NOD-like receptor (NLR) family (23, 29). Moreover, there is evidence supporting the notion that internalization of OMVs is a prerequisite for induction of a proinflammatory response (23).

NOD1 and NOD2 have emerged as pivotal intracellular sensors for bacterial infection in mammalian innate immunity, inducing immune responses and the elimination of bacteria through activation of nuclear factor NF-κB and resulting neutrophil recruitment (30–32). NOD1 is reactive to peptidoglycan peptides containing diaminopimelic acid (DAP) (33, 34), whereas NOD2 responds to muramyl dipeptide (MDP) (35, 36). It has been suggested that MDP is critical for the NOD2-stimulatory activity of peptidoglycan of Gram-negative bacteria (37). NOD1 and NOD2 are functionally expressed in various cell types of oral tissues, e.g., epithelial cells, gingival fibroblasts, and periodontal ligament cells (38–40). In vivo evidence that stimulation of NOD2 reduced P. gingivalis-induced periodontal inflammation and alveolar bone loss supports a crucial role of NLRs in preventing periodontitis development (41). Hence, low NOD-stimulatory activity might be helpful for bacterial survival in the periodontal pocket (42). On the other hand, NOD-mediated innate immune responses targeting oral bacteria can also induce alveolar bone resorption (43). This was demonstrated in a recent study in which NOD1 was found to be essential for bone loss in a mouse ligature-induced periodontitis model (44). Interestingly, this observation was linked to the identification of an A. actinomycetemcomitans-like (>60% coding sequence identity) mouse commensal (termed NI1060) which accumulated at the site of ligature placement and acted as a significant NOD1 activator (44). Based on these observations, it was proposed that NI10160, and possibly the human pathogen A. actinomycetemcomitans, can trigger bone loss via stimulation of NOD1 (44).

Here we show that A. actinomycetemcomitans OMVs can be internalized into human cells to deliver antigens that partly activate NOD1 and NOD2.

MATERIALS AND METHODS

Bacterial strains and growth conditions.

In this study, we used the A. actinomycetemcomitans serotype a smooth-colony strain D7SS (45), a ΔltxA ΔcdtABC mutant derivative of D7SS (46), and the highly leukotoxic strain JP2 (serotype b) (47). The strains were routinely cultivated in air supplemented with 5% CO₂ at 37°C as previously described (48) on blood agar plates (5% defibrinated horse blood, 5 mg hemin/liter, 10 mg vitamin K/liter, Columbia agar base).

Cell lines, culturing, and treatment conditions.

Cells were cultivated at 37°C in a 5% CO₂ atmosphere in air. HeLa cells (ATCC CCL-2) and human gingival fibroblasts (HGF) (49) were cultured in advanced minimum essential medium (MEM) (Invitrogen) supplemented with 2 mM glutamine, 10% fetal calf serum, and 50 μg/ml gentamicin. Human embryonic kidney (HEK293T) cells were cultured in Dulbecco modified Eagle medium (DMEM) supplemented with 10% heat-inactivated fetal calf serum and 100 units/ml each penicillin and streptomycin. Prior to experimental treatment, cells were washed with phosphate-buffered saline (PBS) and treated with trypsin (250 μg/ml in PBS) for 5 min at 37°C. Trypsin-treated cells were collected by centrifugation and then resuspended in MEM containing gentamicin (final concentration, 50 μg/ml). Human myeloid monocytic THP1-Blue cells (InvivoGen) were maintained in very-low-endotoxin (VLE) RPMI 1640 medium (Biochrom) supplemented with 10% fetal calf serum, 2 mM glutamine, 100 units/ml each penicillin and streptomycin, and 100 μg/ml Zeocin (InvivoGen).

Isolation of OMVs.

Outer membrane vesicles (OMVs) were isolated from A. actinomycetemcomitans cells harvested from an average of 10 blood agar plates, using ultracentrifugation as described earlier (7). OMV pellets were washed twice with PBS (85,000 × g, 2 h, 4°C) using a 70 Ti rotor (Beckman Instruments Inc.), and then used as the OMV preparation. The yield of OMVs was estimated by quantifying vesicle preparations for protein content using a Picodrop (Picodrop Ltd.) as in our earlier work (8). To assess the reproducibility of OMV preparations, samples were validated by atomic force microscopy (AFM), and SDS-PAGE. OMVs were also checked for absence of bacterial contamination by cultivating small aliquots on blood agar plates in air supplemented with 5% CO₂, at 37°C for 3 days.

OMV internalization assays.

To monitor internalization of OMVs in HeLa cells and HGF, we used the lipophilic membrane dye PKH26 analogously to its use in earlier studies (10). Vesicles were labeled using the PKH26 red fluorescent cell linker kit for general cell membrane labeling (Sigma-Aldrich) according to the manufacturer's instructions. After labeling, OMVs were washed with PBS (85,000 × g, 2 h, 4°C) using a 70 Ti rotor (Beckman Instruments Inc.) to remove unbound dye. Aliquots of 1 ml (5 × 10⁴ cells per ml) of HeLa cells and HGF were then treated for 6 to 72 h with PKH26-labeled OMVs such that the final OMV protein concentration was 13 μg/ml. Control treatment corresponded to PBS which had been preincubated with PKH26 in the same manner as the OMVs. Where applicable, vesicles were incubated with cells in the presence of the inhibiting agent filipin III (Sigma-Aldrich) at a final concentration of 10 μg/ml as in earlier studies (7, 8, 50). The inhibitor monensin (Sigma-Aldrich) was used at a final concentration of 2 μM or 10 μM. To estimate the effect of these inhibitors on cell viability, a neutral red uptake assay was carried out as described previously (51). Similar to the procedure in other studies (27, 52), to minimize detrimental effects on the cells, we assessed the effect of the inhibitors mainly by pretreating the target cells for 30 min. The cells were then washed twice, and the medium was replaced prior to the incubation with OMVs. In addition, where applicable, the inhibitor was also present during the incubation with vesicles. After incubations, cell samples were fixed with 2% paraformaldehyde in PBS (pH 7.3) for 10 min. Nuclei were stained with 4′,6′-diamidino-2-phenylindole (DAPI) (Sigma-Aldrich) at a final dilution of 1:5,000. The endoplasmic reticulum (ER) was labeled using ER-Tracker Green (Bodipy FL Glibenclamide; Molecular Probes) at a final concentration 2 μM (25 min, 37°C). For antibody detection, fixed cells were incubated (60 min, 37°C) with a rabbit polyclonal antiserum specific for whole cells of A. actinomycetemcomitans serotype a (53) at a final dilution of 1:100. After three washes with PBS, the cells were incubated with fluorescein isothiocyanate (FITC)-conjugated goat anti-rabbit IgG (Invitrogen) (final dilution, 1:400) for 30 min at 37°C in the dark. Following three washes with PBS, the coverslips were subjected to confocal microscopy as described below.

Confocal microscopy.

For analysis by confocal microscopy, coverslips were mounted with Mowiol (Scharlau Chemie S.A., Sentmenat, Spain) containing antifade (p-phenylene diamine). Confocal microscopy was carried out using a Nikon D-Eclipse C1 confocal laser with a Nikon Eclipse 90i microscope. Images were captured with a Nikon color camera (24 bit), using aplan Apo Nikon 60× and 100× objectives. Fluorescence was recorded at 405 nm (blue; DAPI), 488 nm (green; FITC, ER-Tracker Green, and Alexa Fluor 488-phalloidin), and 543 nm (red; PKH26). Z-stack images were taken at 0.1-μm steps covering 2.8 to 5.8 μm using EZ-C1 3.80 imaging software. The images were adjusted and assembled in Adobe Photoshop 13.0. Levels of red and green fluorescence were determined using ImageJ on projected confocal stacks, analyzing single confocal slices of identical sizes from within the perinuclear region of 20 cells. Quantification of fluorophore colocalization (i.e., overlapping pixels) in confocal stacks was done using NIS-Elements AR 3.2 software (Nikon). The Pearson colocalization coefficients (54) were calculated from quantitative data obtained from three confocal stacks.

Electron microscopy.

For analysis by electron microscopy, 5 × 10⁴ HeLa cells in 1 ml were incubated for 24 h with OMVs (final OMV protein concentration, 30 μg/ml). Subsequently, cells were fixed in 2% paraformaldehyde in 0.1 M phosphate buffer (pH 7.4) for 2 h. The fixative was removed, and the cells were washed 3 times with PBS and infused with 2.3 M sucrose for 2 h at 4°C. The blocks were mounted on a specimen holder and frozen in liquid nitrogen. Ultrathin cryosections were prepared at −110°C on a Leica EM UC7/EM FC7 (Leica) with a diamond knife. Immunogold labeling was performed as described earlier (55) using a rabbit polyclonal antibody specific for whole A. actinomycetemcomitans serotype a cells (53) or Staphylococcus aureus sortase A (Abcam) (final dilutions, 1:100). As conjugate, 10-nm gold-coupled goat anti-rabbit secondary antibodies were used (BBI Solutions GAR10; 1:20). The sections were examined in a Jeol 1230 transmission electron microscope (TEM). Digital images were captured using a Gatan MSC 600CW instrument.

Assays for NF-κB activation.

NOD1- and NOD2-dependent NF-κB activation by the OMVs was analyzed as described previously (23, 56), using an LPS-unresponsive, cell-based NF-κB reporter assay. Briefly, HEK293T cells were transiently transfected with expression plasmids for NOD1, NOD2, and β-galactosidase together with an NF-κB luciferase reporter construct. Assessment of NF-κB activation in human myeloid cells was conducted using THP1-Blue cells (InvivoGen) as described earlier (23). This reporter cell line expresses the secreted embryonic alkaline phosphatase (SEAP) under the control of an NF-κB promoter. In THP1-Blue cells, gene silencing of NOD1 and NOD2 was carried out using small interfering RNA (siRNA) in cells differentiated with phorbol 12-myristate 13-acetate (PMA) as described previously (23). In all assays, cells were incubated for 16 h with OMVs or with NOD1/2 elicitors (TriDAP and MDP, respectively) (InvivoGen) as controls.

Statistical analysis of data.

Statistical analysis of the data was performed using GraphPad Prism 4.03 (GraphPad). Data are expressed as means ± standard deviations (SD). Means were compared using the two-tailed Student t test. P values of less than 0.05 were regarded as statistically significant. For analysis of the HEK data shown in Fig. 6A, two-way analysis of variance (ANOVA) with Bonferroni correction was used.

FIG 6.

A. actinomycetemcomitans OMVs trigger NOD1- and NOD2-dependent NF-κB activation. (A) HEK293T cells transiently transfected with expression plasmids encoding NOD1 or NOD2 and an NF-κB luciferase reporter system were incubated for 16 h with the indicated amounts of OMVs isolated from the A. actinomycetemcomitans strains D7SS, D7SS ΔltxA ΔcdtABC, and JP2. NF-κB activity is shown as mean + standard error of the mean (SEM) from two different OMV preparations (each measured in triplicates). **, P < 0.01; ***, P < 0.0001 (two-way ANOVA with Bonferroni correction, compared to control treatment [Ctrl]). (B) Titration of the activation of the D7SS OMVs toward NOD1 (left panel) and NOD2 (right panel) in HEK293T cells. The same NF-κB luciferase reporter system as employed for panel A was used. Nonsignificant (ns) differences for OMVs relative to the synthetic elicitor TriDAP and MDP (two-sided Student t test, n = 3) are indicated. Data represent mean + SD. (C) NOD1, NOD2, or both were silenced in THP1-Blue cells by transfecting specific siRNA duplexes for 72 h. Cells were subsequently stimulated with A. actinomycetemcomitans OMVs as indicated or with TriDAP (25.6 μM) or MDP (10 μM) as controls. NF-κB activity was determined by measuring SEAP activity in the supernatant after 16 h. Error bars indicate SD (n = 3). Ctrl, not stimulated. Results from one representative experiment out of two using two different OMV preparations are shown. *, P < 0.05 (two-sided Student t test).

RESULTS

A. actinomycetemcomitans OMVs are internalized into the perinuclear region of HeLa cells and HGF.

To investigate if A. actinomycetemcomitans OMVs are internalized in human cells, HeLa cells and human gingival fibroblasts (HGF) were incubated for 6 h to 72 h with strain D7SS vesicles labeled with the lipophilic membrane dye PKH26 (red fluorescence). We then performed confocal microscopy analyses to detect the internalized OMVs. This revealed an apparent perinuclear localization of labeled vesicles within both HeLa cells and HGF, which appeared to accumulate during incubation as judged by quantitative assessment of the OMV-associated red fluorescence (Fig. 1A and 2A). In contrast, we detected only weak background levels of red fluorescence in cell samples subjected to control treatment (buffer) in the presence of PKH26 (Fig. 1B and 2B). These observations are consistent with internalization of OMVs in both types of treated human cells.

FIG 1.

Internalization of A. actinomycetemcomitans OMVs in HeLa cells. Cells were incubated with PKH26-labeled strain D7SS OMVs (red fluorescence) (A) or with PBS (buffer) (B) for 6 h, 24 h, and 72 h. After treatment, cells were fixed and incubated with an antibody specific for whole cells of A. actinomycetemcomitans serotype a. Bound antibodies were detected using an FITC conjugate (green). Nuclei were stained with DAPI (blue). Merged images show the labeling with all three fluorophores. Confocal Z-stack projections are included in all images. The cross lines indicate the positions of the xz and yz planes. Bars, 10 μm. Colocalization of OMVs (red) and A. actinomycetemcomitans antigens (green) is indicated by small white arrows. Bar graphs show quantitative analysis of PKH26, FITC, and PKH26/FITC colocalization in the HeLa cell samples incubated with OMVs for the indicated times. PKH26 and FITC values represent arbitrary units of pixel intensity for red and green fluorescence, respectively, determined using ImageJ. PKH26/FITC colocalization is expressed as Pearson's coefficient (PC) values for red and green pixels. Shown are the means ± SD of data collected from 20 cells.

FIG 2.

Internalization of A. actinomycetemcomitans OMVs in human gingival fibroblasts. HGF were incubated with PKH26-labeled strain D7SS OMVs (red fluorescence) (A) or with PBS (buffer) (B) for 6 h, 24 h, and 72 h. After incubation, cells were fixed and incubated with an antibody specific for whole cells of A. actinomycetemcomitans serotype a. Bound antibodies were detected using an FITC conjugate (green). Nuclei were stained with DAPI (blue). Merged images show the labeling with all three fluorophores. Confocal Z-stack projections are included in all images. The cross lines indicate the positions of the xz and yz planes. Bars, 10 μm. Colocalization of OMVs (red) and A. actinomycetemcomitans antigens (green) is indicated by small white arrows. The bar graphs show quantitative analysis of PKH26, FITC, and PKH26/FITC colocalization in the HGF cell samples incubated with OMVs for the indicated times. PKH26 and FITC values represent arbitrary units of pixel intensity for red and green fluorescence, respectively, determined using ImageJ. PKH26/FITC colocalization is expressed as Pearson's coefficient (PC) values for red and green pixels. Shown are the means ± SD of data collected from 20 cells.

Internalization of A. actinomycetemcomitans OMVs represents a mechanism for intracellular antigen exposure.

To investigate if internalization of OMVs was accompanied by delivery of antigens to the host cells, we used an antibody specific to whole A. actinomycetemcomitans serotype a cells in the confocal microscopy analyses. This antibody recognizes multiple epitopes, including GroEL and peptidoglycan-associated lipoprotein (PAL) (57, 58). According to our results (Fig. 1A and 2A), OMV epitope-associated green fluorescence having an apparent perinuclear distribution was detected in both HeLa cells and HGF incubated with PKH26-labeled vesicles for 6 to 72 h. Notably, in the perinuclear regions of both cell types there was strong colocalization between the PKH26-labeled OMVs and internalized A. actinomycetemcomitans antigens (i.e., overlapping red and green pixels), supporting that the vesicles carried cargo into the host cells. Cell type-specific features may have affected the kinetics of this colocalization as judged by the nonidentical patterns of red and green fluorescence accumulation from 6 h to 72 h in HeLa cells compared to HGF (Fig. 1A and 2A). In contrast to the case for OMV-treated cells, only background levels of green fluorescence were detected in cells subject to control treatment (buffer) (Fig. 1B and Fig. 2B). To corroborate these observations, we conducted electron microscopy analysis using immunogold labeling and an antibody directed against whole A. actinomycetemcomitans serotype a cells. As shown in Fig. 3A, gold particles were observed in association with OMVs, confirming that epitopes on the vesicles were accurately detected by the antibody used. In contrast, no gold particles were observed when the OMVs were assessed using a control antibody, i.e., the antibody specific for S. aureus sortase A (Fig. 3B). Analysis of ultrathin sections of HeLa cells incubated with OMVs for 24 h revealed the presence of internalized OMV antigens (Fig. 3D to G), although in these cell samples the outlines of OMVs appeared indistinct, suggesting that internalized vesicles may be subject to degradation. In contrast to the case for OMV-treated HeLa cells, no or very few gold particles were detected in samples from cells subjected to control treatment (buffer) for 24 h (Fig. 3C). Based on these results, we concluded that upon internalization of A. actinomycetemcomitans OMVs there was delivery of vesicle-associated antigens to the interior of the host cells.

FIG 3.

Electron microscopy and immunogold labeling of internalized A. actinomycetemcomitans OMVs. In immunoelectron micrographs showing OMVs alone, vesicles were detected using polyclonal antibodies specific for whole cells of A. actinomycetemcomitans serotype a (A) or S. aureus sortase A (B). The serotype a whole-cell-specific antibody was used for assessment of ultrathin cryosections of HeLa cell samples: cells incubated with PBS (C) and cells incubated with OMVs for 24 h (D to G). The regions indicated by boxes in panels D and F are shown enlarged in panels E and G, respectively. The arrowheads in panel G indicate a Golgi structure and gold deposition in the vicinity. Bars, 0.2 μm.

To further test this idea, we investigated whether inhibition of OMV internalization would result in decreased antigen delivery. For this, we used monensin, a monovalent ionophore that prevents acidification of intracellular compartments such as endosomes and blocks receptor-mediated endocytosis by the rise in intraendosomal pH that abrogates receptor-ligand dissociation (59–61). HeLa cells pretreated (30 min) or not pretreated with monensin were incubated for 24 h with PKH26-labeled A. actinomycetemcomitans OMVs. As shown by confocal microscopy and quantitative analysis of fluorescence (Fig. 4A), monensin clearly suppressed internalization of OMVs (reduced to approximately 25% ± 4% [SD]), suggesting that they were subject to endocytic uptake. Moreover, there was a concomitant decrease in the level of internalized A. actinomycetemcomitans antigens (reduced to approximately 49% ± 17% [SD] compared to when monensin was absent), consistent with OMV internalization contributing to the antigen delivery. This likely explains the lack of cell enlargement, which is caused by OMV-associated CDT (7), in the presence of monensin (Fig. 4A). This notion was supported by prolonged incubation (up to 72 h) of the HeLa cells with OMVs, showing a concentration-dependent inhibitory effect of monensin (see Fig. S1 in the supplemental material). As judged by neutral red staining (Fig. 4C), monensin had no apparent detrimental effects on the cell viability under these conditions. Hence, these results collectively support the idea that A. actinomycetemcomitans OMVs carrying bacterial antigens are internalized into human cells and that this interaction serves as a mechanism for intracellular antigen delivery.

FIG 4.

Effect of inhibitors on internalization of A. actinomycetemcomitans OMVs in HeLa cells. (A) HeLa cells pretreated for 30 min with monensin (+Mo) (10 μM), or filipin III (+Fi) (10 μg/ml) were incubated with PKH26-labeled OMVs for 24 h. Control cells were not subjected to pretreatment with inhibitor. (B) HeLa cells incubated with PKH26-labeled OMVs for 24 h in the presence or absence of filipin III (+Fi) (10 μg/ml). After treatment, cells were fixed and incubated with an antibody specific for whole cells of A. actinomycetemcomitans serotype a. Bound antibodies were detected using an FITC conjugate (green). Nuclei were stained with DAPI (blue). Merged images show the labeling with both fluorophores. Confocal Z-stack projections are included in all images. The cross lines indicate the positions of the xz and yz planes. Bars, 10 μm. Colocalization of OMVs (red) and A. actinomycetemcomitans antigens (green) is indicated by small white arrows. The bar graphs in panels A and B show quantitative analysis of PKH26, FITC, and PKH26/FITC colocalization in treated HeLa cell samples. PKH26 and FITC values represent arbitrary units of pixel intensity for red and green fluorescence, respectively, determined using ImageJ. PKH26/FITC colocalization is expressed as Pearson's coefficient (PC) values for red and green pixels. Shown are the means ± SD of data collected from 20 cells. *, P < 0.0001 for treatment with OMVs in the presence versus absence of the indicated inhibitor. (C) Effect of filipin III and monensin on HeLa cell viability. Neutral red uptake was quantified in cells incubated for 24 h in the presence (+Fi) (10 μg/ml) and absence of filipin III and in cells incubated for 30 min in the presence (+Mo) (10 μM) and absence of monensin. The results shown are means of numbers of viable cells (percentage of control, i.e., inhibitor absent) ± SD from three experiments.

Internalization of A. actinomycetemcomitans OMVs is partially dependent on plasma membrane cholesterol.

As studies with E. coli and H. pylori vesicles support that plasma membrane cholesterol is critical for toxin and antigen delivery via vesicle internalization and for OMV-induced innate immune responses (26, 29), we investigated whether there was cholesterol-dependent uptake of A. actinomycetemcomitans OMVs. For this, HeLa cells were pretreated (30 min) with filipin III, a cholesterol-sequestering agent (62), prior to incubation for 24 h with PKH26-labeled OMVs. According to confocal microscopy and quantitative analysis of fluorescence (Fig. 4A), sequestration of cholesterol using filipin III caused a partial inhibition of the internalization of both OMVs and A. actinomycetemcomitans antigens, which was reduced to approximately 74% ± 14% (SD), whereas their colocalization appeared to be unaffected. Essentially the same results were obtained when filipin III was present in the medium throughout the 24 h of incubation with OMVs (Fig. 4B). As judged by neutral red staining (Fig. 4C), filipin III had no apparent detrimental effects on the cell viability under these conditions. Hence, based on these results, we concluded that internalization of OMVs at least partially depended on the presence of cholesterol in the plasma membrane.

Internalized A. actinomycetemcomitans OMVs colocalize with the ER.

As internalized OMVs and vesicle antigens were colocalized in the perinuclear region (Fig. 1A and 2A) and as some internalized OMV antigens appeared in the vicinity of Golgi structures (Fig. 3F and G), we conducted costaining of the endoplasmic reticulum (ER) to test whether vesicles colocalized with the ER. According to confocal microscopy analysis (Fig. 5A), there was colocalization between PKH26-labeled OMVs (red fluorescence) and ER-Tracker Green (green fluorescence) in HeLa cells, which increased upon prolonged incubation with vesicles (up to 72 h) as judged by quantitative image analysis. Consistent with absent OMV-associated red fluorescence, such colocalization was not observed in cells subject to control treatment (buffer) (Fig. 5B). These observations suggest that the internalized OMVs at least to some extent were trafficked to the ER.

FIG 5.

Internalized A. actinomycetemcomitans OMVs colocalize with the endoplasmic reticulum. HeLa cells were incubated with PKH26-labeled strain D7SS OMVs (red fluorescence) (A) or with PBS (buffer) (B) for 6 h, 24 h, and 72 h. After treatment, fixed cells were stained with DAPI (blue; nuclei), and ER Tracker Green (green; endoplasmic reticulum). Merged images show the labeling with all three fluorophores. Confocal Z-projections are included in all images. The cross lines indicate the positions of the xz and yz planes. Bars, 10 μm. Colocalization of OMVs (red) and ER-Tracker Green is indicated by small white arrows. The bar graphs show quantitative analysis of PKH26, and PKH26/ER-Tracker Green colocalization in HeLa cell samples incubated with OMVs for the indicated times. Fluorescence values represent arbitrary units of red and green pixel intensity, respectively, determined using ImageJ. PKH26/ER-Tracker Green colocalization is expressed as Pearson's coefficient (PC) values for red and green pixels. Shown are the means ± SD of data collected from 20 cells.

A. actinomycetemcomitans OMVs induce NOD1- and NOD2-dependent NF-κB activation.

To test if intracellular antigen delivery to host cells by A. actinomycetemcomitans OMVs also includes exposure of peptidoglycan, we investigated if they may trigger the intracellular PRRs NOD1 and NOD2, which are well-known cytosolic receptors for bacterial peptidoglycan fragments. To this end, we employed a cell-based NF-κB gene reporter assay in human HEK293T cells. OMVs isolated from the strain D7SS induced NF-κB activation in a dose-dependent manner in cells expressing NOD1 or NOD2 (Fig. 6A), whereas a highly significant activation of NOD2 was observed. Mock-treated cells were found to be virtually unresponsive toward the OMVs (Fig. 6A). This observation showed that internalized A. actinomycetemcomitans OMVs deliver NOD1- and NOD2-active peptidoglycan and revealed a role of the vesicle internalization in presenting these PAMPs to cytosolic NLRs. Similar results were obtained when assessing OMVs from a ΔltxA ΔcdtABC mutant derivative of D7SS and from the highly leukotoxic serotype b strain JP2 (Fig. 6A). Notably, a low capacity to activate NOD1 was observed for all OMV samples tested, which reached significant levels only when using higher doses of JP2 vesicles. This excluded that this type of mechanism depended on leukotoxin or CDT. In order to estimate the concentration of NOD1- and NOD2-active peptidoglycan in the D7SS OMVs, we titrated these against the synthetic NOD1 and NOD2 agonists TriDAP and MDP, respectively. This showed that 5 μg/ml of D7SS OMVs induced NOD1-mediated NF-κB in the same range as 25 nM TriDAP and that the NOD2-mediated response induced by 1 μg/ml of OMVs was not significantly different from those induced by 2.5 and 5 nM MDP (Fig. 6B).

To substantiate our findings, we used the human myeloid cell line THP1-Blue, which responds to multiple PAMPs. PMA-differentiated macrophage-like THP1-Blue cells were incubated with different concentrations of A. actinomycetemcomitans OMVs. As indicated in Fig. 6C, the vesicles induced a robust dose-dependent NF-κB activation in these cells, which was far higher than that observed with pure NOD1 or NOD2 ligand alone. Depletion of NOD1 or NOD2 using siRNA blunted the response to TriDAP or MDP, respectively, to background levels, showing functional knockdown of NOD1 and NOD2 in our experimental setting (Fig. 6C). Notably, in line with a recent publication demonstrating a particular role of NOD1 in sensing of A. actinomycetemcomitans and in subsequent alveolar bone loss (44), we observed that mainly NOD1 contributed to the overall inflammatory responses induced by OMVs in THP1-Blue cells (Fig. 6C). The high NF-κB activation observed with the vesicles and the only partial dependence on NOD1 was largely due to LPS as shown by inactivation of LPS using the LPS-sequestering agent polymyxin B (data not shown). Taken together, these results strongly suggest that A. actinomycetemcomitans OMVs contain PGN and that NOD2 and in particular NOD1 contribute to inflammatory responses induced by these vesicles in human cells.

DISCUSSION

In the present study, we have demonstrated that A. actinomycetemcomitans OMVs carrying bacterial antigens are internalized into human cells and that this interaction serves as a mechanism for intracellular antigen exposure, including NOD1- and NOD2-active PAMPs.

Our confocal microscopy data revealed a perinuclear localization of the internalized vesicles in HeLa cells and human gingival fibroblasts, which to the best of our knowledge is the first demonstration of A. actinomycetemcomitans OMV uptake into nonphagocytic host cells. Internalization of OMVs in human cells, which may also allow further intracellular trafficking of the vesicles, is in accordance with several recent studies on OMVs from various bacterial species, including the human pathogens B. abortus, E. coli, H. pylori, P. aeruginosa, P. gingivalis, and V. cholerae (23–28) and also nonvirulent E. coli K-12 strains (63). Colocalization of A. actinomycetemcomitans OMVs with the ER is consistent with findings with vesicles from enterotoxigenic E. coli and P. aeruginosa (24, 26). However, similar to observations with the P. aeruginosa OMVs (50), intracellular delivery of A. actinomycetemcomitans vesicle cargo was unaffected by inhibition of retrograde transport (7), suggesting that passage through this pathway is not crucial for the antigen delivery. Based on the apparent rupture of internalized OMVs observed in electron microscopy, which has also been noticed regarding P. gingivalis vesicles (25), it cannot be excluded that vesicles might be subject to degradation inside the host cells. This does not rule out the possibility that internalized A. actinomycetemcomitans antigens are associated with fragments from ruptured vesicles, such as lipids, which may facilitate their immunostimulatory activity (29).

Our present data revealing strong colocalization between internalized OMVs and antigens detected using an antibody specific for whole A. actinomycetemcomitans cells are consistent with the ability of the vesicles to transport bacterial antigens such as PAMPs into host cells. Our experiments using the ionophore monensin, which is an inhibitor of receptor-mediated endocytosis (60), supported that the OMV internalization contributed in delivering A. actinomycetemcomitans antigens to the perinuclear region and that the vesicles may be subject to endocytic uptake. Results from several studies support the idea that OMVs are internalized via multiple endocytic pathways, including macropinocytosis and clathrin- and caveola-dependent endocytosis (64). A possible explanation for such observations may be that endocytic pathways differ based on the size of the particle (65, 66). Thus, as A. actinomycetemcomitans, similar to many additional bacterial species, disseminates vesicles of heterogenous sizes (approximately 10 to 300 nm in diameter) (3, 7, 19, 48), differentially sized subpopulations of bacterial OMVs may be internalized via discrete pathways of endocytosis. A. actinomycetemcomitans OMV uptake was only partly reduced by filipin III, an inhibitor of caveola-dependent endocytosis (62, 67, 68), preventing OMV colocalization with caveolin protein (69). It thus seems unlikely that the vesicles enter host cells via this pathway. A number of earlier studies have shown that OMVs are internalized via cholesterol-rich lipid rafts in the plasma membrane (26, 29). Although filipin III appeared to prevent fusion of A. actinomycetemcomitans OMVs with the plasma membranes of target cells (7), the weak inhibitory effect of this agent on vesicle internalization and antigen delivery observed in the present work is consistent with findings with P. aeruginosa vesicles (24). This may be explained by the fact that although it disrupts cholesterol-rich lipid rafts, filipin III does not remove the cholesterol entirely from the plasma membrane (62).

Our results obtained by using NF-κB reporter assays in HEK293T and THP1-Blue cells supported that OMV internalization may represent an important mechanism for intracellular antigen exposure by A. actinomycetemcomitans. Upon incubation of the HEK293T cells with A. actinomycetemcomitans OMVs, we observed a strong activation of the cytoplasmic PRRs NOD1 and NOD2, supporting the presence of NOD1- and NOD2-active peptidoglycan in the vesicles. This is consistent with earlier in vitro studies using the HEK293T cell reporter assay, revealing that soluble A. actinomycetemcomitans PGN and heat-killed cells of this species act as potent inducers of both NOD1 and NOD2 (42). Conversely, another recent report using the same cell-based reporter assay suggested that the tested A. actinomycetemcomitans strain mainly triggered NOD1 (44). The reason(s) for this discrepancy is not known; however, A. actinomycetemcomitans strain-dependent differences cannot be excluded, as the latter study, in contrast to the others, used strain JP2, which produces high levels of leukotoxin (47). Notably, in accordance with that study, we observed that OMVs obtained from JP2 exhibited higher activity toward NOD1 than the other vesicle samples (Fig. 6A). It is possible that membrane channels formed by leukotoxin (70) may act as a port of entry for PGN fragments into the cytosol, as demonstrated earlier with Streptococcus pneumoniae pneumolysin and S. aureus alpha-toxin, which alleviated uptake of Haemophilus influenzae PGN to trigger NOD1 (71). We observed that NOD1 significantly contributed to the OMV-induced inflammatory response in THP1-Blue cells. Notably, the A. actinomycetemcomitans OMVs induced a very strong inflammatory response in these cells, which was mainly due to LPS. In contrast to HEK293T cells, which do not express many pathogen recognition receptors, including most Toll-like receptors, myeloid THP1 cells do express most PRRs and thus are suited to gain insight into the contribution of NOD1/2 to the overall responses induced by A. actinomycetemcomitans OMVs. These data suggest that NOD1-dependent sensing of OMV-derived PAMPs may have a physiological relevance in immunocompetent, i.e., myeloid cells. Thus, on the basis of the recent report by Jiao et al. (44) indicating that A. actinomycetemcomitans triggers bone resorption mainly via NOD1, intracellular delivery of PGN via vesicles could act as a direct inducer of periodontal bone loss. It is conceivable that such a functional role of vesicles may be evident mainly in oral species possessing peptidoglycan with high NOD1-stimulatory activity, such as A. actinomycetemcomitans and the closely related mouse commensal NI1060. In contrast to these Pasteurellaceae species, Bacteroidetes organisms, including P. gingivalis, possess peptidoglycan exhibiting low levels of NOD1-stimulatory activity (42, 43).

To the best of our knowledge, our present report provides the first evidence that OMVs released from an organism of the Pasteurellaceae family are internalized into nonphagocytic human host cells to induce NOD1- and NOD2-dependent signaling. It is possible that this type of mechanism may be relevant in additional Pasteurellaceae spp., considering that OMVs capable of modulating the host's immune responses are liberated by several of the species associated with inflammatory conditions in different mammalian hosts, e.g., Actinobacillus pleuropneumoniae, H. influenzae, Mannheimia haemolytica, and Pasteurella multocida (72–75). Further studies assessing their full immunogenic potential in animal models may generate a more complete picture of the roles of A. actinomycetemcomitans OMVs in periodontal inflammation and attachment loss.