The Atherogenic Bacterium Porphyromonas gingivalis Evades Circulating Phagocytes by Adhering to Erythrocytes

Abstract

A relationship between periodontitis and coronary heart disease has been investigated intensively. A pathogenic role for the oral bacterium Porphyromonas gingivalis has been suggested for both diseases. We examined whether complement activation by P. gingivalis strain ATCC 33277 allows the bacterium to adhere to human red blood cells (RBCs) and thereby evade attack by circulating phagocytes. On incubation with normal human serum, the P. gingivalis strain efficiently fixed complement component 3 (C3). Incubation of bacteria with washed whole blood cells suspended in autologous serum resulted in a dose- and time-dependent adherence to RBCs. The adherence required functionally intact complement receptor 1 (CR1; also called CD35) on the RBCs and significantly inhibited the uptake of P. gingivalis by neutrophils and B cells within 1 min of incubation (by 64% and 51%, respectively) and that by monocytes after between 15 min and 30 min of incubation (by 66% and 53%, respectively). The attachment of C3b/iC3b to bacterium-bearing RBCs decreased progressively after 15 min, indicating that conversion of C3 fragments into C3dg occurred, decreasing the affinity for CR1 on RBCs. We propose that P. gingivalis exploits RBCs as a transport vehicle, rendering it inaccessible to attack by phagocytes, and by doing so plays a role in the development of systemic diseases.

Mucosal and skin barriers and the innate immune system, including the complement system, neutrophils, monocytes, and their derivatives, form the first line of defense against Gram-negative bacteria. Failure of these defense systems to clear bacteria from the site of infection may lead to systemic spread of the bacteria and consequent disease. Bacteremia occurs after scaling and after dental hygiene procedures in patients with periodontal disease but also in the presence of an apparently healthy periodontium (22, 41). Certain bacteria, most notably Porphyromonas gingivalis, have been implicated in the pathogenesis of chronic periodontitis as well as aggressive periodontitis (42), cardiovascular diseases (21, 28), diabetes (5, 24), rheumatoid arthritis (10, 27), and pulmonary diseases (17, 36).

There is strong evidence to support the hypothesis that atherosclerosis is an inflammatory disease (39), and a relationship between periodontal infections and cardiovascular disease has been suggested (3, 14). P. gingivalis has been investigated intensively as a possible pathogen for both conditions (4, 15, 21). For example, binding of the bacterium's fimbriae to endothelial cells mediates its invasion of aortic and cardiac endothelium (11). The endothelial cells may become apoptotic (45) or may, together with vascular smooth muscle cells, secrete proatherogenic mediators, including interleukin-1 (IL-1), tumor necrosis factor alpha (TNF-α), IL-6, monocyte chemotactic protein 1 (MCP-1), vascular endothelial growth factor, and P-selectin (12, 44, 46). These mediators activate circulating neutrophils and monocytes and induce release of procoagulant factors from the endothelium. Mice injected with wild-type P. gingivalis develop atheromatous lesions in the aorta, and markedly fewer lesions are observed if metronidazole is administered before injection or if a fimbria-deficient strain of P. gingivalis is used (1). DNA from P. gingivalis has been found in specimens of atherosclerotic plaques removed from carotid and femoral arteries during surgery (23, 43), suggesting that the innate immune system is occasionally incapable of limiting P. gingivalis infections to the periodontal tissues. It is not clear how the bacterium evades the immune system, but some strains of P. gingivalis produce gingipains, a group of cysteine proteases that partly degrade bound C3 and IgG, making the bacterium relatively resistant to phagocytosis in the absence of specific antibodies (7, 13).

More than 50 years ago, it was known that pneumococci and Treponema pallidum sensitized with antibodies bind to human red blood cells (RBCs) in a process requiring a heat-labile factor which was presumed to be complement (29). This phenomenon was referred to as immune adherence. Later, it was demonstrated that RBCs from humans and other primates (unlike most animal species, including rodents) bear complement receptor 1 (CR1) grouped in clusters consisting of 2 to 15 receptors (6, 35), allowing the RBCs to bind immune complexes (ICs) and bacteria opsonized with complement. In the presence of normal serum, ICs released from RBCs cannot rebind to RBCs (30). They also have a decreased affinity for rebinding to neutrophils or mononuclear cells, depending on how long they were associated with RBCs (26). Immune adherence to RBCs is generally considered beneficial because it allows the transport of complement-opsonized microorganisms or ICs to the liver and spleen, where they are eliminated by the fixed macrophage system (29, 40).

We have previously shown that the binding of complement-opsonized ICs to RBCs restricts the binding of the ICs to circulating neutrophils and monocytes (31). We hypothesize now that complement-activating oral pathogenic bacteria may use RBCs as a transport vehicle, allowing the systemic spread of the bacteria and preventing circulating neutrophils and monocytes from attacking the bacteria. Such a mechanism may contribute to a pathogenic role of P. gingivalis in the development of systemic diseases.

MATERIALS AND METHODS

Cells and serum.

For isolation of blood cells, peripheral blood was collected by venipuncture into lithium-heparin tubes (BD Bioscience, Brondby, Denmark) from 15 healthy donors attending the blood bank at Copenhagen University Hospital, Rigshospitalet, Denmark, and the School of Dentistry, University of Copenhagen, Copenhagen, Denmark (all Caucasian; 7 men and 8 women; median age, 42 years; age range, 18 to 72 years). The whole blood cells were washed twice in phosphate-buffered saline (PBS) (Invitrogen, Auckland, New Zealand). For some experiments, leukocytes were isolated by lysis for 10 min in Easy Lyse ammonium chloride buffer (Dako, Glostrup, Denmark) diluted 1:20 in distilled water, followed by centrifugation (300 × g at 24°C).

Sera were isolated from blood samples from the same donors in dry Vacutainer tubes (BD Bioscience) after deposition in the dark for 1 h and centrifugation at 700 × g for 30 min. All combinations of washed whole blood cells and serum were autologous.

All donors gave informed consent, and the study was approved by the local ethics committee.

Preparation of bacteria.

P. gingivalis strain ATCC 33277 was obtained from the Oral Microbiology Section, School of Dentistry, Faculty of Health Sciences, University of Copenhagen, Copenhagen, Denmark. It was cultured on Trypticase soy (TS) blood agar plates (Oxoid Ltd., Basingstoke, Hampshire, England) containing 5 mg/liter hemin and 50 μg/liter vitamin K. The TS blood agar plates were incubated anaerobically in the presence of 10% CO₂, 10% H₂, and 80% N₂ for 5 days in an MAC-3 incubator (DW Scientific Ltd., West Yorkshire, United Kingdom). The bacteria were harvested and diluted in brain heart infusion (BHI) medium (Oxoid Ltd., Basingstoke, Hampshire, United Kingdom) and then reincubated as described above for 24 h. The bacteria were labeled with 2 μM 5,6-carboxyfluorescein-diacetate-succinimidyl-ester (CFSE) (Molecular Probes, Poortgebouw, Holland) for 15 min at 4°C in the dark and centrifuged twice for 10 min (4,400 × g at 4°C), followed by resuspension in PBS immediately before use. Unless otherwise stated, 1 × 10⁷ bacteria were added to the cell preparations.

Assessment of binding of bacteria to RBCs or leukocytes.

A total of 1 × 10⁷ CFSE-labeled bacteria were incubated for various periods with washed blood cells or isolated leukocytes suspended in 70% (vol/vol) autologous serum (see the detailed description of each experiment in the figure legends). The reaction was stopped by transfer of 10-μl samples to 1 ml PBS on ice, and the binding of bacteria to RBCs, neutrophils, monocytes, and B cells was assessed by flow cytometry (see below).

Antibodies.

Attachment of split products of C3 to P. gingivalis was measured using fluorescein isothiocyanate (FITC)-conjugated polyclonal rabbit anti-human C3d (recognizing native C3, C3b, iC3b, and C3d; Dako, Copenhagen, Denmark) at a final dilution of 1:50.

Two monoclonal antibodies (MAbs) were raised in-house: anti-human C3 clone f1-23 reacted with C3 (recognizing C3b and iC3b bound to bacteria as well as C3 and C3c), and anti-human C3c MAb f1-7 recognized a neo-epitope exposed only on iC3b and C3c. To test the reactivity of these MAbs, 20 mg of each was covalently coupled to CNBr-activated Sepharose, essentially as described previously (38), and used as a purification matrix. Five milliliters of cobra venom factor in vitro-activated normal human serum (activated as described previously [37]) was applied to the matrix, which was subsequently washed with PBS. Bound proteins were eluted with 0.5% citric acid and analyzed by reducing SDS-PAGE and Coomassie blue staining or by immunoblotting with polyclonal antibodies against C3 (A0062 and A0063; Dako, Glostrup, Denmark).

The anti-CR1 MAb HB8592 (IgG1 isotype; American Type Culture Collection, Manassas, VA) was used for blockade of CR1 at a concentration of 1 μg/ml. Anti-glycophorin A (Immunotech, Marseilles, France) was used as an IgG1 isotype control at 1 μg/ml.

Allophycocyanin (APC)-conjugated anti-CD15 MAb (0.3 μg/ml) was used for labeling of neutrophils, phycoerythrin (PE)-anti-CD19 MAb was used for labeling of B cells (0.1 μg/ml), peridinin chlorophyll protein (PerCP)-anti-CD45 MAb was used for labeling of total leukocytes, and—in separate experiments—PerCP-anti-CD14 MAb (0.1 μg/ml) was used for labeling of monocytes (all from BD Bioscience). Incubations took place for 30 min at 37°C in the dark.

Assessment of complement activation by P. gingivalis.

A total of 1 × 10⁷ bacteria were added to samples of 70 μl of serum in low-absorbing polypropylene tubes (Life Technologies, Paisley, United Kingdom) and supplemented with PBS to a final volume of 100 μl for 30 min at 37°C. To stop further complement activation, the samples were placed on ice, and 4 μl 0.5 M EDTA (Invitrogen, Auckland, New Zealand) was added. After two washes in PBS at 4°C, the volume was adjusted to 100 μl. Sulfo-NHS-LC-biotin (Perbio Science, Bonn, Germany) dissolved in PBS was added to a final concentration of 0.3 mg/ml. One hundred microliters of the mixture was added to the bacterial suspension and incubated for 30 min at room temperature. After a wash at 4°C, the volume was adjusted to 100 μl, and 2 μl streptavidin-PE (Pierce Biotechnology, Rockford, IL) and 1 μl FITC-conjugated polyclonal rabbit anti-human C3d (see below) were added. Following incubation in the dark for 30 min at room temperature and two washes at 4°C, the stained bacteria were suspended in 500 μl sterile filtered distilled water, and the binding of C3 fragments to the bacteria was analyzed by flow cytometry.

For assessment of C3b and iC3b fragments on bacteria bound to RBCs, the bacteria were incubated with whole blood cells for various periods in the presence of 70% (vol/vol) autologous serum. The reaction was stopped by transfer of 10-μl samples to 1 ml PBS on ice. The cells were incubated for 1 h with the MAbs f1-23 and f1-7 to a final MAb concentration of 14 μg/ml, followed by incubation with PE-conjugated rabbit anti-mouse IgG antibodies (Dako, Glostrup, Denmark) (final concentration, 1 μg/ml).

Flow cytometry.

Flow cytometry was performed using a FACSCalibur flow cytometer (BD Bioscience). In experiments assessing C3 attachment to P. gingivalis, the bacteria were identified by morphological characteristics (forward and side light scatter) and by staining with sulfo-NHS-LC-biotin and PE-streptavidin, whereas specific C3 fragment deposition was measured by staining with FITC-conjugated MAbs.

In experiments assessing the binding of P. gingivalis to RBCs, the bacteria were identified as CFSE-positive events, RBCs were identified as CD45-negative events within a morphological (forward scatter/side scatter) RBC gate, neutrophils were identified as CD15⁺ cells, B cells were identified as CD19⁺ cells, and monocytes were identified either as CD14⁺ cells or, when PerCP labeling of leukocytes was used, as CD45⁺ events within a morphological monocyte gate. Cell Quest Pro software (BD Biosciences) was used for analysis, and the binding of bacteria was quantified on the basis of the mean fluorescence intensity (MFI).

Assessment of bacterial viability.

For assessment of viability, bacteria were prepared and incubated with whole blood cells suspended in autologous serum as described above. Following incubation for 1, 5, 10, 15, and 30 min, the blood cells were washed three times, cultured on TS blood agar plates (Oxoid Ltd., Basingstoke, Hampshire, England) containing 5 mg/liter hemin and 50 μg/liter vitamin K, and incubated under anaerobic conditions as described above. As a negative control, blood cells not exposed to bacteria, but otherwise treated identically, were used. The viability of bacteria was assessed after 7 days.

Statistics.

GraphPad Prism for Windows v. 4.00 (GraphPad, San Diego, CA) was used for statistical analysis. The Wilcoxon matched pair test was used for comparison of paired data, and two-way analysis of variance was used for comparisons of differences between curves for the binding of P. gingivalis in the presence and absence of anti-CR1 antibody. P values of <0.05 were considered significant.

RESULTS

P. gingivalis activates complement and binds to RBCs in a complement-dependent manner.

To assess whether P. gingivalis ATCC 33277 was capable of activating complement, the bacterium was incubated with untreated sera obtained from 10 healthy humans or with the same sera subjected to complement inactivation by heating to 56°C for 1 h. As a positive control, we incubated the same sera with a different oral bacterium, Fusobacterium nucleatum, which is known to activate complement efficiently (34). As shown in Fig. 1, significant incorporation of C3 fragments occurred on P. gingivalis, although to a lesser extent than that on F. nucleatum.

FIG. 1.

Activation of complement by P. gingivalis. The binding of fragments of C3 to P. gingivalis ATCC 33277 (P.g.; circles) following incubation for 30 min with sera from 10 healthy individuals was assessed by flow cytometry, using FITC-conjugated polyclonal rabbit anti-human C3d (recognizing native C3, C3b, iC3b, and C3d) as the detecting antibody. F. nucleatum (F.n.; squares) was used as a positive control. The data shown are net values after subtraction of background fluorescence observed in the presence of the same sera preheated to 56°C for 1 h to achieve complement inactivation. **, P < 0.005; ***, P < 0.001 for the null hypothesis that no binding of C3 occurs.

To test whether C3 fragments deposited on P. gingivalis targeted the bacterium to RBCs bearing clusters of CR1 (6, 35), CFSE-labeled bacteria were added to whole blood cells suspended in autologous serum to allow complement opsonization in situ. Binding of P. gingivalis to RBCs was clearly visible when 1 × 10⁷ bacteria were added to blood cells from 50 μl blood (data not shown). Since this volume contains approximately 25 × 10⁷ RBCs, bacteria were not available for more than 4% of the RBCs (see marker M1 in Fig. 2A). The excess of RBCs relative to bacteria presumably mimics the physiological conditions under which bacteria enter the bloodstream in vivo. Kinetic analysis of P. gingivalis adherence to RBCs showed rapid binding: about half-maximum binding was reached within 1 min. This initial pattern was followed by progressive uptake at a lower rate over the next 30 min (Fig. 2B). Addition of the anti-CR1 antibody HB8592 to the reaction mixture completely abrogated the binding to RBCs, clearly demonstrating the dependency of CR1 in the process (Fig. 3).

FIG. 2.

Binding of P. gingivalis ATCC 33277 to human RBCs. (A) Flow cytometry histogram showing the adherence of CFSE-labeled P. gingivalis to CD45-negative RBCs after 5 min of incubation of 1 × 10⁷ bacteria with whole blood cells suspended in autologous serum. M1 represents the bacterium-binding fraction of the RBCs. (B) A total of 1 × 10⁷ CFSE-labeled P. gingivalis bacteria were incubated for various periods with whole blood cells suspended in autologous serum. The MFI for the binding of bacteria to RBCs are shown as medians and interquartile ranges for eight experiments.

FIG. 3.

Adherence of P. gingivalis to RBCs is CR1 dependent. A total of 1 × 10⁷ CFSE-labeled P. gingivalis ATCC 33277 bacteria were incubated with 50 μl PerCP-anti-CD45-labeled whole blood cells suspended in autologous serum in the presence of the anti-CR1 MAb HB8592 (IgG1 isotype) (open circles), anti-glycophorin A (anti-GPA) as an IgG1 isotype control (closed squares), or no antibody (−MAbs; closed circles). The adherence of P. gingivalis to CD45-negative RBCs under these conditions is shown as the median and range for four experiments. The P value signifies the probability for identical curves, as determined by two-way analysis of variance.

RBCs act as a buffer inhibiting uptake of P. gingivalis by circulating leukocytes.

In contrast, with the rapid adherence of P. gingivalis to RBCs, the binding of the bacterium to neutrophils, monocytes, and B cells in preparations of whole blood cells occurred at a lower, almost constant rate over the entire observation period (closed symbols in Fig. 4A to C).

FIG. 4.

Red blood cell-mediated restriction of uptake of P. gingivalis by peripheral leukocytes. A total of 1 × 10⁷ CFSE-labeled P. gingivalis ATCC 33277 bacteria were added to cells derived from 100 μl normal whole blood (+RBCs) or from 100 μl whole blood deprived of red blood cells (−RBCs) by lysis in ammonium chloride buffer. Autologous serum was present during the entire incubation, at a concentration of 70% (vol/vol). The binding of the bacteria to CD15⁺ neutrophils (A), CD14⁺ monocytes (B), and CD19⁺ B cells (C) at various time points is shown, expressed as MFI values, and the corresponding ratios for the binding occurring in the presence and absence of red blood cells are also shown (D to F). Circles and error bars represent medians and interquartile ranges for eight experiments. (*), P = 0.09; *, P < 0.05; **, P < 0.01; ***, P < 0.005.

When RBCs in the blood samples were lysed, however, all three leukocyte populations took up bacteria at a considerably higher rate (open symbols in Fig. 4A to C). Indeed, at any time point within the first 30 min, markedly larger numbers of bacteria were associated with neutrophils, monocytes, and B cells in the absence of RBCs.

As shown in Fig. 4D, the presence of RBCs significantly inhibited the binding of bacteria to neutrophils, by a median of 64% at 1 min (P < 0.008) and 60% at 5 min (P < 0.02), which decreased gradually to 40% (P < 0.004) at 15 min. At 30 min, inhibition was much lower, at 23%, and was not significant (P < 0.10).

The uptake of P. gingivalis by monocytes was also inhibited by the presence of RBCs (Fig. 4E). Although the median inhibition was 80% after 1 min and 75% after 5 min of incubation, these results were not significant (P < 0.10) due to wide variability in the data. After 15 and 30 min, however, significant RBC-mediated inhibition was observed, amounting to 66% (P < 0.03) and 53% (P < 0.008) inhibition, respectively.

A distinct pattern of inhibition was observed for B cells (Fig. 4F). After an initial reduction in B-cell binding of P. gingivalis to 51% after 1 min of incubation (P < 0.03), RBC-mediated inhibition remained borderline significant until 15 min (P = 0.05) but turned into an RBC-mediated enhancement (by 65 to 76%) of binding after 60 min in 4 of 8 tested donors.

Conversion of P. gingivalis-associated C3 fragments into C3dg during binding to RBCs.

As described above, attachment of C3 fragments to the surfaces of P. gingivalis cells resulted in a rapid, CR1-dependent binding of bacteria to RBCs. To assess the kinetics of incorporation and degradation of P. gingivalis-associated C3 fragments, we used MAbs f1-23 (recognizing both C3b and iC3b) and f1-7 (recognizing a neo-epitope exposed on iC3b and C3c only) as detection antibodies. These specificities were confirmed by immunoaffinity purification of cobra venom factor-activated normal serum (Fig. 5).

FIG. 5.

Reactivities of MAbs used for measurement of C3 fragment deposition on RBCs. MAbs f1-7 and f1-23, raised for detection of C3 fragments, were coupled to CNBr-activated Sepharose columns and used for purification of C3 from cobra venom factor-activated normal serum. The eluted proteins were analyzed by SDS-PAGE (under reducing conditions) and Coomassie staining (left) or by immunoblotting with polyclonal anti-C3 antibodies as detection antibodies (right). Staining of the eluates from both columns by either method revealed bands with apparent molecular masses of approximately 75 kDa, 40 kDa, and 25 kDa, corresponding to the β-chain and the α2 and α1 portions of C3. Blotting of the f1-7 eluate also revealed an iC3b-specific α1-C3d portion (∼63 kDa), and the absence of a band corresponding to the intact α-chain (110 kDa) shows that f1-7 does not react with C3 or C3b. Conversely, the eluted fractions from the MAb f1-23 immunoaffinity purification produced a band of 110 kDa, corresponding to an intact α-chain, which suggests that this antibody also recognizes C3b.

As shown in Fig. 6, the association of the C3b/iC3b-specific MAb f1-23 with the RBCs bearing P. gingivalis closely reflected the association of the bacterium with the RBCs. After a maximum at 15 min, binding gradually declined, however, indicating that a gradual conversion of P. gingivalis-associated C3 fragments into C3dg occurs after 15 min at 37°C. The iC3b-specific MAb, on the other hand, bound to the RBCs 3 to 5 min later, presumably reflecting the time required for conversion of C3b into iC3b, and only declined after 30 min of association with the P. gingivalis-bearing RBCs. The major proportion of RBCs that did not bind P. gingivalis stained negative for the MAbs (not shown).

FIG. 6.

Degradation kinetics of C3 fragments attached to P. gingivalis on RBCs. A total of 1 × 10⁷ CFSE-labeled P. gingivalis bacteria were added to PerCP-anti-CD45-labeled washed whole blood cells obtained from 20 μl blood and suspended in autologous serum (final serum concentration, 90%). MAbs f1-7 and f1-23 were used as primary antibodies to detect C3 fragments associated with the RBCs. The binding of these MAbs to the fraction of RBCs positive for binding the CFSE-labeled bacteria (corresponding to M1 area in Fig. 2A) is shown as the median for quadruplicate experiments.

Viability of P. gingivalis released from RBCs.

To examine whether P. gingivalis bacteria released from RBCs were viable, the bacteria were incubated with whole blood cells suspended in autologous serum for 1, 5, 10, or 30 min. After being washed thoroughly, the cells were cultured on TS agar plates under anaerobic conditions for 7 days (see the supplemental material). Growth of P. gingivalis was observed in four of four experiments at all time points, indicating that bacteria carried by RBCs were viable.

As a negative control, whole blood cells not exposed to bacteria, but otherwise treated identically, were used. No growth was observed under these conditions (see the supplemental material).

DISCUSSION

Associations between periodontal infections and systemic diseases have been subject to intense investigation for the last few decades (3, 14). P. gingivalis is a major oral pathogen associated with periodontitis (42) as well as with a number of systemic diseases (21, 28). It is believed that the bacterium can spread via the bloodstream and possibly invade aortic and cardiac endothelium (11, 15).

In accordance with other groups (7), we report here that P. gingivalis ATCC 33277 efficiently fixes C3. Moreover, we show for the first time that complement-opsonized P. gingivalis adheres to RBCs via CR1. This challenges the general view that P. gingivalis and other atherogenic bacteria normally spread in plasma.

Binding of P. gingivalis to RBCs reduced the uptake of the bacterium by neutrophils and monocytes to approximately 1/3 and 1/5 within the first minute of incubation, and the inhibitory effect lasted for 15 to 30 min. This is analogous to previous findings that binding of complement-opsonized ICs to RBCs restricts the uptake of ICs by phagocytes (31), preventing neutrophils from being activated (33).

It remains to be established whether adherence of P. gingivalis to RBCs is beneficial for the host or for the bacterium. Adherence of bacteria to RBCs has long been considered instrumental in the clearance of bacteria by fixed liver macrophages (Kupffer cells) (29). However, clearance by Kupffer cells is slow, as demonstrated by Davies et al.: approximately 25% of complement-activating ICs capable of binding to RBCs remain in circulation 1 h after injection into healthy humans (9). Moreover, P. gingivalis is thought to be capable of adhering to endothelial cells via its fimbriae (8, 11, 20). Taking these observations together with our demonstration here that C3 associated with P. gingivalis-bearing RBCs is gradually degraded to C3dg, which decreases its affinity for CR1, we hypothesize that a substantial proportion of P. gingivalis bacteria carried by RBCs in the circulation might be transferred to endothelial cells rather than to liver macrophages.

The conversion of C3b into iC3b, which is known to cause a 100-fold decrease in affinity for CR1 (19), was already apparent within 5 min, as shown by the discrepancy between the curves for MAbs f1-23 (recognizing iC3b and C3b) and f1-7 (recognizing iC3b) in Fig. 6. The impaired binding of the two MAbs to P. gingivalis-bearing RBCs after 20 to 30 min reflects the conversion of iC3b to C3dg, which results in a complete loss of affinity for CR1 (19). By analogy, ICs bound to RBCs are gradually lost in the presence of normal serum (18, 25, 26, 30), and once released from RBCs, ICs cannot rebind to RBCs (26) and show a decreasing tendency to rebind to neutrophils or mononuclear cells (26). It is noteworthy that CR1 itself acts as a cofactor in the factor I-catalyzed degradation of C3b into iC3b and, finally, into C3dg (25).

Interestingly, ICs released from RBCs show a tendency to rebind to B cells (31), and these cells, unlike other leukocytes, express CR2, which is the receptor for the degradation product of C3d (32). In the present study, we observed that the binding of P. gingivalis to B cells in four of eight cases followed a pattern compatible with rebinding to B cells. Hence, the binding of bacteria to B cells was higher at later time points in the presence of RBCs.

It is noteworthy that P. gingivalis released from RBCs seems to be perfectly viable and thus presumably able to colonize endothelial walls and various tissues. Adherence of P. gingivalis to endothelial cells has been shown to depend on major and minor fimbriae on the bacterial surface (8, 11). Accordingly, atheromatous lesions can be induced in mouse aortas by in vivo injection with wild-type P. gingivalis, whereas significantly fewer lesions are observed when fimbria-deficient strains of the bacterium are injected (1). P. gingivalis invasion of endothelial cells leads to increased autocrine production of IL-6, a cytokine with a well-established role in the regulation of atherosclerotic disease (16). Toll-like receptor 2 (TLR2) and CD14 receptors on the endothelium have been suggested to interact with bacterial fimbriae (8), but a recently published study suggests that lipid contamination of isolated fimbriae accounts for much of the TLR2 activation observed in previous reports (2).

In conclusion, we demonstrate here that P. gingivalis ATCC 33277 fixes C3 and readily adheres to RBCs via CR1 and that rapid degradation of C3 fragments into iC3b, and presumably C3dg, occurs on the RBC surface. We show that adherence of P. gingivalis to RBCs restricts uptake of the bacterium by neutrophils and monocytes, and we propose that P. gingivalis exploits RBCs as a transport vehicle and privileged site that is inaccessible to attack by circulating phagocytes. This mechanism may underlie a pathogenic effect of P. gingivalis in a number of systemic diseases. Further investigation is needed to test this hypothesis.