Acquired Resistance of Escherichia coli to Complement Lysis by Binding of Glycophosphoinositol-Anchored Protectin (CD59)

Abstract

Protectin (CD59) is a glycophosphoinsitol (GPI)-anchored defender of human cells against lysis by the membrane attack complex of complement. In this study, we examined whether protectin released from human cell membranes can incorporate into the surface of gram-negative bacteria. Analysis by using radiolabeled protectin, immunofluorescence, flow cytometry, and whole-cell enzyme-linked immunosorbent assay demonstrated that protectin bound to nonencapsulated Escherichia coli EH237 (Re) and EH234 (Ra) in a calcium-dependent manner. The incorporation required the GPI-phospholipid moiety since no binding of a phospholipid-free soluble form of protectin was observed. Mg²⁺ did not enhance the binding, and a polysialic acid capsule prevented it (strain IH3080 [O18:K1:H8]). Bound protectin inhibited the C5b-9 neoantigen expression on complement-treated bacteria. Protection against complement lysis was observed in both a colony counting assay and a bioluminescence assay, where viable EH234 bacteria expressing the luciferase gene emitted green light in the presence of the luciferine substrate. In general, two- to four-times-higher serum concentrations were needed to obtain 50% lysis of protectin-coated versus noncoated bacteria. The results indicate that protectin can incorporate in a functionally active form into the cell membranes of the two nonencapsulated deep rough E. coli strains studied.

The human complement system plays a major role in resistance against microbial infections, as it is involved in both specific and nonspecific immunity (3). The defense action of complement is mediated via a general enhancement of the inflammatory response opsonizing molecules and by cytolytic membrane lesions on foreign targets. After initiation and amplification, complement activation leads to the formation of C5b-9 complexes that can insert into the membranes of serum-sensitive bacteria and cause cell death through the collapse of membrane potential (2). Polymerization of the terminal complement component C9 has been shown to be necessary for optimal killing of some gram-negative bacteria (12, 32).

As the complement system is a powerful defense system of the host, any pathogenic microbe coming into contact with human blood or plasma must have developed mechanisms to evade complement attack. Most gram-negative bacteria are sensitive to the lytic action of complement in fresh human serum, whereas some are resistant and therefore more virulent. Mechanisms of complement resistance include the steric barrier of the bacterial capsule and lipopolysaccharide (LPS) side chains of gram-negative bacteria that hinder the access of the membrane attack complex (MAC) to the bacterial membrane (9). Bacterial proteases can inactivate complement proteins or inhibit their accumulation on bacterial surfaces (28, 30). Some bacteria have membrane proteins that interfere with the assembly of the terminal C5b-9 complex and prevent lethal outer cell membrane damage (10, 11).

The human complement system is in most cases well controlled by the host, and inappropriate activation and host cell destruction are prevented. The control is mediated by soluble inhibitors and by specific cell membrane glycoproteins. The membrane proteins are either inhibitors of the C3 and C5 convertase enzymes (decay-accelerating factor [DAF], membrane cofactor protein [MCP], and complement receptor type 1 [CR1]) or regulators of MAC (protectin and C8bp) (23). Protectin (CD59) is linked to cell membranes via a glycosylphosphatidylinositol (GPI) anchor and inhibits cell lysis by preventing the C5b-8 complex-catalyzed insertion and polymerization of C9 (4, 21, 27). The effective cytolysis-inhibiting form of protectin has a glycophospholipid tail, but the phospholipid part is lost from the soluble form (18, 19). It has recently been shown that the lipid-tailed form of protectin is capable of transferring from one cell surface to another in vitro (37) and in vivo (15).

To avoid complement attack, microbes have developed various ways to utilize complement regulatory proteins. Sialic acid can enhance the binding the complement regulator factor H to C3b on the microbial surface and prevent amplification of the alternative pathway (5, 22, 24). Group A streptococcal protein M can also bind factor H and block the alternative pathway via a similar mechanism (7). Some blood parasites have been found to acquire complement regulatory molecules from the host, and some even produce analogs of the human regulatory proteins (13). As an example, the C3 convertase inhibitor DAF (CD55) has been shown to be transferred from human erythrocytes to Schistosoma mansoni worms (8). Whether a similar interaction occurs between GPI-anchored proteins and bacteria is not known.

The aim of this study was to examine whether protectin with a phospholipid moiety can bind to the outer cell membrane of gram-negative bacteria in a functionally active form. We demonstrate that protectin binds to two nonencapsulated deep rough mutant strains of Escherichia coli in a Ca²⁺-dependent manner and inhibits the formation of cytolytic complement lesions on the bacteria.

MATERIALS AND METHODS

Bacteria.

E. coli EH237 (LPS chemotype Re) (36), EH234 (LPS chemotype Ra) (1), and encapsulated strain IH3080 (O18:K1:H8) (33) were kindly provided by M. Vaara at our department. The luciferase gene expression vector pCSS962 (lucGR) (16, 39) and a helper plasmid pGB3 (17) were kindly provided and cloned into the E. coli EH234 and JM103 (LPS chemotype Ra) by M. Karp at the University of Turku, Turku, Finland. Bacteria were grown at 37°C on Luria broth. Following overnight growth, the bacteria were washed three times (1,750 × g, 8 min), resuspended in buffer, and concentrated to approximately 2 × 10¹⁰ bacteria/ml for binding analyses. For fluorescence-activated cell sorting analysis, the bacteria were resuspended to approximately 10⁹/ml.

Isolation of CD59.

A lipid-tailed form of protectin (CD59_E) was purified from human erythrocytes, and a soluble form (CD59_U) was purified from human urine, as described earlier (21). Both proteins were radiolabeled with Na[¹²⁵I] as described previously (18). The ¹²⁵I-CD59_E (initial activity, 2 × 10⁷ cpm/μg) stock preparation contained 0.02% Nonidet P-40 detergent. ¹²⁵I-CD59_U (10⁷ cpm/μg) did not contain detergent, but for comparative binding assays it was reconstituted with Nonidet P-40 at concentrations equivalent to those used with ¹²⁵I-CD59_E.

Buffers and reagents.

Veronal-buffered saline (VBS; 3.2 mM diethyl barbituric acid, 1.8 mM diethyl barbituric acid sodium salt, 0.15 M NaCl [pH 7.3]), VBS containing MgCl₂ (0.5 mM) and CaCl₂ (0.15 mM) (VBS²⁺), or phosphate-buffered saline (PBS) was used as the buffer. In some instances, 0.05% bovine serum albumin (BSA) or 0.05% Tween 20 (T) was added to VBS (VBS-BSA), VBS²⁺ (VBS²⁺-BSA), or PBS (PBS-T) to reduce nonspecific reactions. In the whole-cell enzyme-linked immunosorbent assay (ELISA), 0.5% BSA in VBS was used as a blocking reagent.

BRIC229 (immunoglobulin G2b mouse monoclonal antibody (MAb) against protectin (International Blood Group Reference Laboratory, Elstree, United Kingdom) and mouse MAb (IgG2a) against a C5b-9 neoepitope (Quidel, San Diego, Calif.) were used as primary antibodies, and mouse MAbs (IgG) AF1 and AF-16.1 (gifts from M. Kaartinen at our department) were used as control antibodies. Fluorescein isothiocyanate (FITC)- and peroxidase-conjugated rabbit antibodies against mouse Igs (Jackson ImmunoResearch Laboratories, West Grove, Pa.) were used as secondary antibodies. Normal human serum (NHS) was obtained from healthy laboratory personnel. Heat-inactivated NHS (NHSi) was prepared by treatment at 56°C for 30 min.

¹²⁵I-CD59 binding tests.

Three different strains of E. coli (EH237, EH234, and IH3080) were incubated with different concentrations of ¹²⁵I-CD59_E and ¹²⁵I-CD59_U (0.5 ng to 0.5 μg/10⁹ bacteria in a final volume of 100 μl of VBS-BSA) at 37°C for 30 min with gentle shaking. After washes, the cell-bound and free ¹²⁵I-CD59 were separated by centrifuging (5,000 × g, 1 min) the mixtures through 20% sucrose (250 μl) in narrow (0.4-ml) test tubes. The bottom parts of the tubes containing the cells were cut out, and radioactivities in both pellets and supernatants were counted. The experiments were performed in duplicate and repeated three times.

The effect of Ca²⁺ and Mg²⁺ on binding of protectin to E. coli.

Stock solutions of CaCl₂ and MgCl₂ were prepared in distilled H₂O. The cations were diluted into VBS to obtain eight final concentrations ranging between 0 and 30 mM. Three E. coli strains (EH234, EH237, and IH3080) were incubated with ¹²⁵I-CD59_E or ¹²⁵I-CD59_U (0.4 μg/10⁹ bacteria in a final volume of 100 μl) in VBS-BSA containing various concentrations of Ca²⁺ or Mg²⁺ for 30 min at 37°C with gentle shaking. After incubation and washes, the cell-bound and free ¹²⁵I-CD59 were separated as described above. To control for possible cation-induced precipitation of protectin, the binding experiments were repeated in the absence of bacteria. All experiments were done in duplicate and repeated three times.

Indirect immunofluorescence microscopy.

CD59_E (0.4 μg) was incubated with 10⁹ E. coli EH237 cells in VBS-BSA with or without Ca²⁺ (2.5 mM) as described above. Bacteria were washed three times with VBS-BSA, spread on microscope slides, and allowed to air dry. The samples were fixed in cold acetone (−20°C) for 8 min. After being washed three times, the slides were incubated for 30 min at room temperature with the primary antibody (BRIC229) against protectin (6 μg/ml). After three washes, the samples were treated with the corresponding FITC-conjugated secondary antibody. Control stainings were performed by omitting CD59_E or the primary antibody or by incubating the bacteria with the irrelevant primary antibody AF1. The indirect immunofluorescence slides were mounted with Mowiol (6) and examined on an Olympus BX50 standard microscope equipped with a filter specific for FITC fluorescence. The slides were photographed on Kodak Tri-X 400 Pro film.

Flow cytometry.

CD59_E was allowed to bind to the EH237 bacteria as described above. After three washes with VBS-BSA, the bacteria were incubated for 30 min at 37°C with the BRIC229 antibody (3.3 μg/ml) against CD59. After washing, the secondary FITC-conjugated antibody was added and the bacteria were incubated at 37°C for 30 min, washed three times, and examined immediately by flow cytometry. Control stainings were performed by omitting CD59 or the primary antibody. The experiment was done in triplicate and repeated twice. All samples were examined with a FACScan 440 (Becton Dickinson, San Jose, Calif.) flow cytometer with an argon laser tuned to 488 nm at a power output of 15 mW. The data were analyzed with the Lysys II software supplied by Becton Dickinson.

Bactericidal assay.

E. coli EH237 was incubated with or without CD59_E (0.01 μg/2 × 10⁵ bacteria in 100 μl). Bacteria were washed with VBS²⁺-BSA and incubated with 0, 1.7, 5, or 17% NHS or NHSi in a final volume of 500 μl. After washing, serial 10-fold dilutions of bacterial suspensions were made in saline, and 900 μl of each dilution was plated on Luria agar plates. After a 15-h incubation at 37°C, CFU were counted. The survival of bacteria in NHS was calculated relative to that in NHSi. The experiment was done in duplicate and repeated twice.

Bioluminescence assay for bacteriolysis.

The effect of CD59 on serum sensitivity of the bacteria was also studied in E. coli EH234 and JM103. A structural gene for luciferase from the Jamaican click beetle Pyrophorus plagiophthalamus was cloned and expressed in both strains. Incubations with CD59_E (0.4 μg/10⁹ bacteria in 1 ml) and NHS or NHSi were performed as described above. The survival of bacteria in NHS was calculated relative to that in NHSi. In controls, CD59_E was omitted. Strain JM103 and the bioluminescence method used have been described earlier (34, 35). Briefly, live bacteria expressing the luciferase enzyme illuminate in the presence of the luciferine substrate. Bacterial death leads to the loss of enzyme production and activity, resulting in a decrease in illumination. The substrate, 0.25 mM d-luciferine in sodium citrate (pH 5.0), was added 90 min before measurement of luminescence. A Luminoskan EL1 luminometer and Biolise software (Labsystems, Helsinki, Finland) were used for data collection and analysis. The samples were analyzed in quadruplicate, and the experiment was repeated four times.

Effect of bound protectin on C5b-9 deposition.

A modification of a whole-cell ELISA method for C3 deposition onto mycobacteria (29) was used for analysis of protectin binding and C5b-9 complex formation on E. coli. Strains EH237 and IH3080 were incubated with or without CD59_E (2 μg/10⁹ bacteria in a final volume of 2 ml of VBS²⁺) in the presence of 2.5 mM Ca²⁺. The bacteria were washed with VBS²⁺-BSA, divided into aliquots, and incubated with 0, 1.7, 5, and 17% NHS or NHSi in a final volume of 500 μl. The bacteria were then washed again and resuspended to 10⁹/ml. Then 50-μl aliquots were dispensed into the wells of microtiter plates and allowed to dry overnight at 37°C. Each well was washed three times with PBS-T, incubated with 75 μl of 0.5% BSA in PBS for 1 h at 37°C to block nonspecific protein binding sites, and washed again. Six parallel wells of each kind (EH237 or IH3080 with or without CD59_E; four concentrations of NHS or NHSi) were incubated for 1 h at 37°C with or without 50 μl of the primary antibody against CD59 (1.3 μg/ml) or the C5b-9 neoepitope (1.9 μg/ml). The wells were washed, and the secondary peroxidase-conjugated antibody (1.6 μg/ml) was added. The plates were incubated for 1 h at 37°C and washed thoroughly with PBS-T. Phenylenediamine dihydrochloride substrate (70 μg/100 μl in ureoperoxidase; Dakopatts, Glostrup, Denmark) was added, and the reaction was stopped after 5 min with 50 μl of 20% sulfuric acid. The absorbances were read by using a 492-nm filter on an ELISA reader (Labsystems Multiskan MCC/340). Absorbances in wells with NHSi-incubated bacteria were subtracted as background in the C5b-9 ELISA, and absorbances in wells with CD59-uncoated bacteria were subtracted as background in the CD59 ELISA. The means and standard deviations (SD) of the absorbances for six identical parallel wells of each type were calculated, and the correlation between bound CD59 and C5b-9 neoepitope expression was analyzed by linear regression.

RESULTS

The ability of protectin to bind to the surface of gram-negative bacteria was first examined by using ¹²⁵I-labeled protectin in a direct binding assay. In the absence of added divalent cations, phospholipid-tailed protectin bound to the nonencapsulated E. coli strain EH237 (Re chemotype) but not to the nonencapsulated strain EH234 (Ra chemotype) or to the encapsulated strain IH3080. With an input of 0.5 μg of protectin per 10⁹ bacteria, strain EH237 bound 5.4% ± 1.1% (mean ± SD) of the protein available, whereas binding to neither EH234 (1.8% ± 0.2%) nor IH3080 (1.0% ± 0.01%) exceeded significantly the background (1.0% ± 0.5%). Protectin binding was dependent on the glycolipid moiety since no binding of soluble protectin (CD59_U) to any of the E. coli strains was observed (binding of <1%).

Since the binding of phospholipids to cell membranes is known to be dependent on divalent cations (14), we next investigated the effects of Ca²⁺ and Mg²⁺ on CD59 binding to E. coli. The addition of increasing concentrations of Ca²⁺ resulted in a significant increase in the binding of phospholipid-tailed protectin to both nonencapsulated strains (EH234 and EH237) (Fig. 1A). A detectable increase in binding occurred at 0.5 mM Ca²⁺. At a physiological (plasma) concentration of Ca²⁺ (2.5 mM), an average of 50% of the lipid-tailed protectin bound to the nonencapsulated strains. The binding reached a maximum at a Ca²⁺ concentration of 30 mM, at which point approximately 70% of protectin had bound to the bacteria. Ca²⁺ did not induce binding of protectin to the encapsulated strain IH3080. Concentrations of Mg²⁺ as high as 30 mM did not affect the binding of lipid-tailed protectin to the E. coli strains studied (Fig. 1B). No binding of soluble protectin to the bacteria was observed in the presence of either Ca²⁺ (Fig. 1C) or Mg²⁺ (not shown).

FIG. 1.

Effects of Ca²⁺ and Mg²⁺ on binding of protectin to E. coli EH234, EH237, and IH3080. Three strains of E. coli were incubated (30 min at 37°C) with ¹²⁵I-CD59_E (0.4 μg/10⁹ bacteria) in VBS in the presence of the indicated concentrations of Ca²⁺ (A) or Mg²⁺ (B). Note a dose-dependent enhancing effect of Ca²⁺ on the binding of CD59_E to the nonencapsulated strains EH234 (Ra) and EH237 (Re) but not to the encapsulated strain IH3080 (O18:K1:H8). Mg²⁺ has no effect on the binding of lipid-tailed protectin to the bacteria examined. (C) Comparison of binding of ¹²⁵I-CD59_E and soluble ¹²⁵I-CD59_U to E. coli EH237 in the presence of 2.5 mM Ca²⁺. Unlike CD59_E, soluble CD59_U fails to bind to the rough, nonencapsulated strain EH237. bd, background.

In the next set of experiments, we examined the binding of lipid-tailed protectin to individual bacteria within a population of bacterial cells by using indirect immunofluorescence microscopy and flow cytometry. Microscopic visualization showed that bacteria incubated with CD59 in the presence of Ca²⁺ stained positive for protectin (Fig. 2A), whereas in the absence of Ca²⁺ the staining was clearly less intense (Fig. 2B). When CD59_E (Fig. 2C) or the primary antibody (Fig. 2D) was omitted or an irrelevant primary MAb was used (Fig. 2E), no staining was seen. Flow cytometric analysis of CD59 binding to strain EH237 showed that CD59 bound to the bacteria (Fig. 3). Bacteria incubated with CD59_E showed a 2.1-times-greater mean fluorescence intensity. When the primary antibody was omitted, no difference in fluorescence between the CD59-coated and the uncoated bacteria was seen.

FIG. 2.

Demonstration of binding of the glycolipid-tailed CD59_E to E. coli EH237 by indirect immunofluorescence microscopy. In this assay, 0.4 μg of CD59_E was incubated with 10⁹ bacteria in the presence (A) or absence (B) of 2.5 mM of Ca²⁺ for 30 min at 37°C. Washed bacteria were incubated with a mouse MAb (BRIC229) against CD59 and further with FITC-conjugated secondary antibody. Control stainings were performed by omitting CD59 (C) or the primary antibody (D) or by incubating the bacteria with an irrelevant primary antibody (E). Washed bacteria were mixed with the mounting medium, spread on microscope slides, and covered with a coverslip. A membranous staining of EH237 for CD59 can be seen in panel A. The staining for CD59 is weak in the absence of Ca²⁺ (B) and negative in the controls (C to E). Bar = 5 μm.

FIG. 3.

Analysis of CD59 binding to E. coli EH237 by flow cytometry. A 0.4-μg aliquot of CD59_E was incubated with 10⁹ bacteria (EH237) in the presence of 2.5 mM of Ca²⁺ at 37°C for 30 min. The washed bacteria were incubated with an antibody against CD59 (mouse MAb BRIC229) and exposed to FITC-conjugated secondary antibody (rabbit anti-mouse IgG). A total of 10,000 cells were counted, and histograms showing counts per channel (y axis) relative to fluorescence intensity (x axis) are shown. Bacteria incubated with CD59 show a mean intensity of 8.0, whereas the native bacteria show a mean intensity of 3.8 for green fluorescence.

The ability of bacterium-bound protectin to prevent complement-mediated bacteriolysis was studied by counting viable bacteria (strain EH237) after exposure to increasing concentrations of NHS and by measuring changes in the luminescence of luciferase-transfected bacteria (strain EH234). When measured by viable counts, 50% of the EH237 bacteria with bound protectin survived exposure to 1.0% NHS, compared with only 12% survival of the native bacteria (Fig. 4). The bacteria with bound protectin needed at least three-times-higher serum concentrations than native bacteria for 50% killing. In the bioluminescence assay, more than 90% of the EH234 bacteria with bound protectin survived exposure to 1.0% NHS, compared with 50% survival of the native bacteria (Fig. 5). The bacteria with bound protectin needed at least twice the amount of serum as native bacteria for 50% lysis. Consistent results were obtained from four independent experiments. The bioluminescent strain JM103 was used as an internal control in the bioluminescence method, and the shift in complement resistance after protectin coating could also be seen with this strain (not shown).

FIG. 4.

Protection of E. coli EH237 against complement lysis by CD59_E. Human erythrocyte protectin was incorporated into E. coli EH237 as for Fig. 2, whereafter the bacteria were incubated (30 min, 37°C) with indicated concentrations of NHS. The ability of bacterium-bound CD59_E to prevent complement-mediated bacteriolysis was studied by counting the CFU of serial 10-fold dilutions of the bacterial suspensions. The survival of serum-treated bacteria is expressed as percentage of surviving bacteria exposed to NHSi. After incorporation of CD59_E, the survival (at 1.7% of NHS) rose from 4 to 45%. Mean (±SD) values of duplicates are shown in this representative example of two similar experiments.

FIG. 5.

Analysis of the ability of bacterium-bound CD59_E to prevent cell death by using the bioluminescent recombinant E. coli strain EH234. A gene (lucGR) encoding the luciferase enzyme was cloned and expressed in E. coli EH234 (Ra). The recombinant bacteria were incubated with CD59_E or VBS as for Fig. 2 and treated with increasing concentrations of NHS. d-Luciferin, the substrate for luciferase, was added, and the luminescence of the bacteria was measured after 60 min with a luminometer. In the presence of the luciferase enzyme and luciferine substrate, the bacteria emit green light but cell death leads to the loss of enzyme activity and light emission. Mean (±SD) values of quadruplicates are shown in this representative example of four similar experiments.

The ability of membrane-bound protectin to prevent deposition of the complement MAC was studied with a whole-cell ELISA. After treatment with serum at concentrations up to 3%, the CD59-coated bacteria expressed significantly less (P < 0.05) C5b-9 neoepitope than the uncoated bacteria (Fig. 6A). As shown in Fig. 6B, C5b-9 deposition correlated negatively with the amount of CD59 binding to the EH237 strain (r = 0.973, P < 0.05).

FIG. 6.

(A) Inhibition of MAC assembly (as measured by C5b-9 neoepitope expression) by incorporation of CD59_E into E. coli EH237 studied by whole-cell ELISA. The glycolipid-anchored protectin was incorporated into E. coli EH237, and the bacteria were incubated with the indicated dilutions of NHS. Washed bacteria were incubated with a mouse MAb against a C5b-9 neoepitope (or CD59, in panel B) and further with a peroxidase-conjugated secondary antibody (rabbit anti-mouse). Results from control incubations performed by omitting the primary antibody have been subtracted as background (optical density [OD] of <0.4). Results are shown as mean ± SD (n = 6). The significances of differences in MAb binding were examined by the two-tailed paired Student’s t test. ∗∗∗, P < 0.001; ∗∗, P < 0.01. (B) Correlation between CD59_E binding and C5b-9 assembly on E. coli EH237 analyzed by simple linear regression analysis. The binding of CD59_E and C5b-9 epitope expression on the bacteria were found to be inversely proportional.

DISCUSSION

Results of this study show that bacterial outer cell membranes can act as acceptors for the GPI-anchored protein protectin (CD59). In the presence of Ca²⁺ ions, protectin can bind to nonencapsulated E. coli and protect the bacteria against complement lysis. In earlier studies, it was demonstrated that phospholipid-tailed protectin can transfer from one human cell surface to another both in vitro (37) and in vivo (15). The binding of protectin to bacteria is somewhat analogous to the binding of bacterial lipid A-containing LPS to human cells (26). Bacteria and host cells thus appear to be capable of transferring their lipid-anchored molecules from one cell type to another. This may result in both beneficial and adverse biological consequences for both cell types.

Binding of the glycolipid-tailed form of the complement inhibitor protectin to E. coli was dependent on the phospholipid tail since no binding of the soluble urinary form of protectin was observed. It is likely that the association is due to a hydrophobic interaction between the glycolipid tail and the lipid A-rich bacterial outer membrane. Since human cell membranes analogously bind only phospholipid-tailed protectin, it is probable that the phospholipid moiety becomes incorporated into the bacterial outer membrane in a manner similar to that for eukaryotic cell membranes. The persistence of bound protectin on the bacteria during washing procedures suggests stable incorporation rather than simple adsorption.

When present, a capsule forms a steric barrier on the surface of the bacterium, and the probability of a lipid-rich micelle to pass the barrier and enter the outer membrane is low. In our experiments, the capsule apparently prevented the binding of GPI-anchored protectin, but the nonencapsulated strains of E. coli were found to bind protectin. LPS core oligosaccharides also seemed to hinder the binding of protectin to the bacteria in the absence of exogenously added Ca²⁺ ions, suggesting partial steric hindrance by the oligosaccharide. This finding further indicates that the bacterial outer membrane rather than the core oligosaccharide is the principal protectin binding structure. The role of protectin as a receptor for the bacteria examined was ruled out by the fact that no binding of the soluble form of protectin was observed.

In a previous study, it was shown that incubation of Salmonella typhimurium with artificial bilayer phospholipid vesicles resulted in a significant transfer of lipids from vesicles to the bacteria in the presence of Ca²⁺ (14). Ca²⁺ is probably important in maintenance of the structural organization of the outer membrane which, unlike eukaryotic cell membranes, contains mostly lipid A and relatively little phospholipid. We studied the effect of two divalent cations, Ca²⁺ and Mg²⁺, on the binding of protectin to the bacteria. Interestingly, Ca²⁺ enhanced the binding up to 10-fold whereas Mg²⁺ had no effect. The binding of protectin to bacteria could already be seen at a physiological extracellular (plasma) concentration of Ca²⁺, but more binding occurred at higher Ca²⁺ concentrations. The amount of protectin available to E. coli in the presence of Ca²⁺ was estimated to result in a density similar to that on cultured human endothelial cells (20, 31). Very low concentrations of Ca²⁺ could, in fact, have inhibited protectin binding. At higher concentrations, Ca²⁺ ions may affect the bacterial surface charge and destabilize the outer membrane to allow vesicle fusion into it. On the other hand, the ability of Ca²⁺ to bind to the phosphate head groups of phospholipids could affect the micelles themselves. Ca²⁺ may decrease the size of protectin micelles and increase the probability of their passing through the LPS barrier.

The ability of bacterium-bound protectin to protect the microbes against complement-mediated lysis was examined by two independent methods. The bactericidal assay is a traditional and well-described method based on the growth capacity of the bacteria. The bacterial bioluminescence method is based on measuring changes in the metabolism of the bacteria. It enables analysis of multiple parallel specimens and their equal processing. Similar results were obtained by the two methods. The binding of protectin resulted in a significant increase in the resistance of the bacteria against complement-mediated lysis. In addition to inhibiting bacteriolysis, the bacterium-bound protectin was found to inhibit the expression of the C5b-9 neoepitope on the bacteria. This finding suggests that protectin inhibited complement similarly as on eukaryotic cells.

The granular expression pattern of protectin often seen in areas of tissue damage suggests that an inflammatory reaction may induce instability of GPI-anchored molecules on cell membranes (25, 38). Shedding of protectin together with its GPI-lipid moiety could allow its incorporation into neighboring cells, including bacteria. Transfer of protectin to bacteria could therefore occur in vivo and lead to an acquired complement resistance of bacteria. This phenomenon might have in vivo relevance, e.g., in chronic infections caused by Neisseria, Chlamydia, or Mycoplasma species. The simultaneous incorporation of two or more complement regulators like GPI-anchored DAF and protectin could result in a further increase in the complement resistance of the bacteria. The transfer of complement-inhibiting surface molecules from host cells to infecting microbes thus provides another mechanism whereby microbes can evade the host immune system.