CD46 (Membrane Cofactor Protein) Acts as a Human Epithelial Cell Receptor for Internalization of Opsonized Uropathogenic Escherichia coli¹

Bacterial infection of the urinary tract is a common clinical problem, estimated to affect 40–50% of women at least once in their lifetime, 20% of whom will develop recurrent infections (1). The most important causative organism is Escherichia coli (E. coli), which is responsible for 70–80% of all urinary tract infections (UTIs).³ E. coli has several well-recognized characteristics that increase its virulence, such as fimbrial adhesins (type 1, P, S, and Dr fimbriae) and toxins (α-hemolysin, cytotoxic necrotizing factor 1). In addition, uropathogenic strains are usually resistant to serum bactericidal activity (2). Perhaps less well understood is how host factors influence disease outcome. Understanding the interactions between E. coli and host-defense mechanisms may provide new insight into the success of this organism and provide a rationale for immune-based therapies for disease prevention.

The epithelial cell lining of the urinary tract is the first point of contact for potential pathogens. The ability of uroepithelial cells to internalize bacteria has been recognized for some time (3). Epithelial cells from both the kidney (3, 4, 5) and the bladder (6, 7, 8) have the capacity to internalize bacteria into membrane-bound vacuoles. The importance of bacterial fimbriae, particularly type 1 fimbriae, in this process has been demonstrated (9). Internalization is increasingly recognized as an important feature of bacterial infection. It can enhance bacterial survival by providing protection from host immune defenses and allow pathogens greater access to deeper tissues. A recent study also has shown that intracellular E. coli can form a reservoir within the bladder mucosa that may serve as a source for recurrent acute infections, a well-recognized clinical problem (10).

However, the uroepithelial cell is not simply a passive target for infection but may actively participate in the innate immune response. Epithelial cells have the capacity to produce chemotactic cytokines (11), enhancing neutrophil recruitment, and also can produce components of the complement system, including C3 (12, 13).

Complement is an important element in the innate immune response against many bacterial pathogens. Several observations suggest that complement plays an important role in the pathogenesis of ascending UTI. First, production of complement proteins within the kidney is increased in response to cytokines associated with acute infection (14, 15). Second, most uropathogens are resistant to killing by complement, suggesting evolutionary pressure on bacteria to develop resistance. Third, previous studies have shown that decomplementation by Cobra Venom Factor (CVF) decreases the degree of tissue damage during renal infection (16, 17, 18). Furthermore, we reported recently that mice deficient in C3, the pivotal component of the complement cascade, are resistant to ascending infection (19). This may, in part, be explained by the observation that, in vitro, bacteria coincubated with serum were internalized by mouse proximal tubular epithelial cells at a rate 10-fold higher than that of nonopsonized bacteria.

Human cells express both complement receptors (CR) and regulatory proteins, both of which have the capacity to bind to C3 and mediate signal transduction. C3 is deposited on the surface of pathogens, acting as a ligand for receptors on phagocytic leukocytes. Less is known about the ability of C3-opsonized pathogens to interact with receptors on epithelial cells, and the functional consequence of such C3-receptor–ligand interaction. Colonic and cervical epithelial cells can internalize opsonized HIV (20) and Neisseria gonorrhoea (21), respectively. This occurs through binding to CR3. However, these receptors have limited distribution and have not been identified on renal tract epithelial cells.

In contrast, complement regulatory proteins, such as CD46 and CD55, are ubiquitous. CD46, membrane cofactor protein, is the cellular receptor for several pathogens, including Neisseria meningitidis (22), measles virus (23), human HSV 6(24), group A Streptococci (25), and group B adenovirus (26). For many of these pathogens, binding to CD46 leads to internalization and mediates increased pathogenicity. These pathogen–CD46 interactions rely on specific pathogen-expressed proteins rather than host-derived C3. However, a recent report has suggested that opsonized Klebsiella may interact with pulmonary epithelial cells via CD46(27). CD55 is bound directly by Dr fimbriated uropathogenic E. coli, a pathogen particularly associated with ascending UTI during pregnancy (28).

In the present study, we have examined the relationship between human renal tract epithelial cells and opsonized uropathogenic E. coli. We report that the mechanism for bacterial invasion of human bladder and renal tubular epithelial cells is complement dependent. Moreover, we find that CD46 expressed on epithelial cells plays a critical role in the internalization process. Finally, we demonstrate that infected urine contains sufficient quantities of C3 to allow this CD46-mediated internalization.

The uropathgenic E. coli strain J96 was provided by Dr. R. Welch, (University of Wisconsin, Madison, WI). It is a serum-resistant, hemolysin-secreting E. coli strain that expresses both type 1 and P fimbriae (29). A Dr fimbriated pyelonephritic strain IH11128 (O75:K5:H⁻) was provided by Dr. B. Nowicki (University of Texas, Galveston, TX). Infected urine samples were obtained from the Department of Microbiology, St. Guy’s and St. Thomas’ National Health Service Foundation Trust (London, U.K.). Bacteria were grown in 10 ml of static Luria-Bertani broth at 37°C for 16 h before use in experiments. For each experiment, bacterial concentration was standardized by photospectrometry at 600 nm, and colony numbers were confirmed by using serial dilutions and plating to agar plates. The human PTEC line was a gift from Dr. L. C. Racusen (Johns Hopkins University School of Medicine, Baltimore, MD) (30). The cells were grown in DMEM-F12 supplemented with 5% FCS, 5 μg/ml insulin, 5 μg/ml transferin, 5 ng/ml sodium selenium, 100 U/ml penicillin, and 100 μg/ml streptomycin. The bladder epithelial cell line J82 was obtained from the American Type Culture Collection. Cells were grown in Eagle’s minimum essential medium with nonessential amino acids and sodium pyruvate supplemented with 10% FCS, 2 mM glutamine, 100 U/ml penicillin, and 100 μg/ml streptomycin.

Normal human serum (NHS) was obtained from five healthy volunteers. After collection, serum was pooled and stored at −70°C for up to 3 mo. Complement activation experiments were performed in GVBS⁺⁺ buffer (Veronal-buffered saline (VBS) with 0.1% gelatin, 0.15 mM CaCl₂, and 1.0 mM MgCl₂), in which all complement pathways are active. EDTA-GVBS⁻buffer (VBS with 0.01M EDTA) inhibits all complement pathways. Complement activity in serum was inactivated by incubation at 56°C for 30 min (heat-inactivated serum, HINHS). Alternatively, C3 in serum was depleted by treatment with CVF by incubating 1 ml of NHS with 20 μg of CVF at 37°C for 60 min (31). Complement inactivation was confirmed by loss of hemolytic activity using standard methodology (data not shown).

Bacteria were opsonized as described previously (32). Briefly, 2 × 10⁸ CFU J96 were washed and then incubated in GVBS⁺⁺ containing 5% NHS, HINHS, or CVF-treated NHS at 37°C for 20 min. Bacteria were washed and then stained with rabbit anti-human C3 (1/100 dilution; Serotec) or mouse monoclonal anti-human complement fragments iC3b (cross-reactive with C3b) and anti-human C4d (10 μg/ml; Quidel) for 1 h at 25 °C. Bacteria were washed again and then stained with FITC-labeled secondary Abs (goat anti-rabbit IgG (1/100 dilution; Serotec) or goat anti-mouse IgG (1/100 dilution; Sigma-Aldrich). After washing, bacteria were fixed with 1% paraformaldehyde and analyzed by flow cytometry. The intensity of staining for C3 was analyzed using CellQuest software (BD Biosciences), using the median cell fluorescence as a measure of staining intensity (median fluorescence intensity).

The kinetics of C3 activation and degradation was assessed by C3 Western blotting. First, J96 were incubated with 5% NHS in GVBS⁺⁺ buffer at 37°C for varying time points. At the end of the incubation period, the bacteria were washed in EDTA-GVBS⁻ to stop further complement activation. Bacterial-bound complement proteins were eluted with 4 mM sodium carbonate and 46 mM sodium bicarbonate (pH 9.2) for 2 h at 37°C. Bacteria were removed by centrifugation. Eluted proteins were separated by 10% SDS-PAGE under reducing conditions and transferred to a Hybound-c Extra membrane (Amersham Biosciences). The membrane was sequentially incubated with blocking buffer at 4°C overnight, rabbit anti-human C3c (1/1000; DakoCytomation), and peroxidase-conjugated goat anti-rabbit IgG (1/5000; DakoCytomation). The membrane was then developed using the ECL system (Amersham Biosciences).

Urine samples were collected from patients with acute UTI (provided by the Department of Microbiology, St. Guy’s and St. Thomas’ National Health Service Trust) and healthy volunteer controls (six samples). The healthy volunteer and patient groups were of similar age and were sex matched. C3 concentration was measured by ELISA as described previously (33). In brief, 96-well plates (Nunc) were coated overnight at 4°C with sheep anti-human C3c (The Binding Site) diluted 1/200 in PBS. After blocking with PBS containing 1% BSA, the plate was incubated with appropriately diluted urine samples, followed by rabbit anti-human C3c (DakoCytomation) diluted 1/3000, then peroxidase-conjugated goat anti-rabbit IgG (DakoCytomation) diluted 1/5000. Each Ab incubation was performed in 100 μl of PBS containing 1% BSA, 0.1% Tween at 37°C for 1 h, and followed by washing in PBS containing 0.1% Tween. The enzyme activity was read after incubation with O-phenylenediamine by measuring absorbance at 490 nm. Purified C3 (1.0 mg/ml; Quidel) was used to generate a standard curve. The limit of sensitivity of this assay was 0.1 ng/ml. Urinary proteins also were separated by 10% SDS-PAGE under reducing conditions and Western blotted as above.

PTECs were grown on glass coverslips and infected with E. coli J96 for 3 h. After washing, cells were prefixed in 2.5% glutaraldehyde in PBS, washed, and postfixed with 1% osmium tetroxide. For scanning EM, cells were dried using a critical point dryer to achieve minimal cell distortion. Cells were then coated with gold and examined using a scanning electron microscope (model S520; Hitachi) at 15 kV of accelerating voltage. For transmission EM, fixed cells were dehydrated in increasing concentrations of ethanol at 60°C, postfixed, stained with lead citrate, and viewed on a transmission electron microscope (model H7000; Hitachi).

Epithelial cells were seeded into 24-well plates and grown to confluence. Overnight cultures of J96 were adjusted to an OD of 0.01 at 600 nm (1 × 10⁷ CFU/ml). Culture medium was replaced with 900 μl of prewarmed DMEM-F12 in the presence of 5% NHS or HINHS, and epithelial cells were infected with 100 μl of the bacterial suspension. Bacterial contact with host cells was increased by centrifugation of plates at 600 × g for 5 min. After 3 h of incubation at 37°C, the total number of bacteria present and the number of bacteria bound to and internalized by cells were measured in wells run in parallel. Total bacterial number was measured by adding 2× lysis buffer (2% Triton X-100) to each well and plating out lysate. Bacteria bound to PTECs were measured by lysing cells after vigorous washing to remove unattached bacteria. This would include internalized bacteria, but because binding exceeded internalization by ∼100-fold, no correction was made. The percentage of bacteria bound was calculated as the number of bacteria bound to the epithelial cell monolayer divided by the total number of bacteria present. To assess the number of internalized bacteria, after washing, the cells were incubated for 1 h in medium containing 100 μg/ml gentamicin to kill extracellular bacteria, then lysed. The percentage of bacteria internalized was calculated as the number of bacteria surviving incubation with gentamicin divided by the total number of bacteria present.

To determine whether the presence of serum acted through bacterial opsonization or altered PTEC phenotype, bacteria or cells were pre-exposed to 5% NHS or HINHS for 1 h at 37°C. The rate of internalization was then measured as described above but in the absence of additional serum.

The expression of CD46 and CD55 was detected by FACS and immunocytochemistry using rabbit anti-CD46 and rabbit anti-CD55 (obtained from Dr. J. Atkinson and Dr. D. Lublin, Washington University School of Medicine, St. Louis, MO, respectively). To silence CD46 expression PTECs (seeded in a 24-well plate) were transfected with short interfering RNA (siRNA) (sense sequence: 5′-GGAGCCACCAACAUUUGAATT-3′; anti-sense sequence: 5′-UUCAAAUGUUGGUGGCUCCTC-3′) or silencer negative control siRNA (Ambion Europe) using an siRNA transfection kit (siPORT Amine; Ambion). After 48 h, mRNA degradation and reduction in proteins synthesis were confirmed by RT-PCR (5′-GCTGCTCCAGAGTGTAAAGTGG-3′ and 5′-AACAATCACAGCAATGACCC-3′) and Western blotting (using rabbit anti-CD46). Image J software (National Institutes of Health, Bethesda, MD) was used to quantify band intensity on PCR gels and Western blots. Transfected cells were further used for bacterial binding and internalization assays.

To reduce surface expression of CD55, PTECs were treated enzymatically to remove GPI-linked membrane proteins. PTECs (seeded in 24-well plate) were incubated with 500 μl of PBS containing varying concentrations of phosphatidylinositol-specific phospholipase C (PI-PLC; BIOMOL) at 37°C for 1 h. PI-PLC-treated cells were further used for bacterial internalization assay.

Soluble CD46 (generated from MCP-C2 isoform, a gift from Dr. J. Rojo, Centro de Investigaciones Biológicas, Consejo Superior de Investigaciones Cientificas, Madrid, Spain; Ref. 27) at a final concentration of 100 μg/ml was added to PTECs with bacteria and serum, and the internalization assay continued as described above. BSA was used in control experiments.

Data were analyzed by Student’s t test for comparison of two variables, ANOVA with Bonferroni posttest for multiple comparisons, Mann-Whitney U test or Fischer’s exact test. Values of p < 0.05 were regarded as significant.

Opsonization of bacteria by complement, in particular, the pivotal component C3, is a fundamental element of the innate immune response against infection. When the uropathogenic E. coli strain J96 was exposed to NHS, there was rapid opsonization with C3 protein in a concentration-dependent manner, with maximum opsonization seen at serum concentrations >10% (Fig. 1,a). To find out the kinetics of C3 activation and degradation, we have performed C3 Western blotting using elute from bacteria incubated with 5% NHS for varying time points. Maximal C3b deposition on E. coli was achieved at 5 min and persisted for up to 3 h. iC3b was evident after 1 min, the amount increasing with time (Fig. 1 b). C3b and iC3b are the major forms of C3 deposited on E. coli surface. These C3 opsonins can potentially interact with a broad range of host membrane complement binding proteins including C3 regulatory proteins and the CR1–CR4. J96 is not sensitive to complement-dependent lysis, with no reduction in the number of viable bacteria in serum concentrations up to 50%.

Binding of uropathogenic bacteria to renal tract epithelium is critical for bacterial persistence in the urinary tract and therefore for infection. The E. coli strain J96 binds to tubular epithelial cells. However, binding was increased by >10-fold if a source of complement (5% NHS) was present throughout the assay (Fig. 2, a and b). This effect required complement activity, as it was not seen when the assay was performed in the presence of HINHS (1.75 ± 0.11 vs 0.13 ± 0.02 bacteria bound per cell, NHS vs HINHS). Preopsonization of the bacteria with complement before adding to cells also increased bacterial binding (Fig. 2,c), whereas pre-exposure of epithelial cells had no effect (Fig. 2 d), confirming that an interaction between complement and the bacteria was essential for this effect. Complement-mediated damage to the epithelial cell during the assay was excluded by the absence of significant LDH release from cells when incubated with NHS (data not shown). This suggests that bacterium-bound complement acts a ligand binding to an epithelial cell receptor, thereby increasing bacterial binding.

Human tubular epithelial cells have the capacity to internalize bacteria. J96 is internalized by the human tubular epithelial cell line used in this study. As with binding, if internalization is assessed in the presence of NHS as a source of complement the rate of internalization is increased (Fig. 3, a and b). This occurred in a serum concentration-dependent manner, being evident at serum concentrations >1% (Fig. 3,c). The effect of complement on internalization was not unique to tubular epithelial cells. The same effect was also seen with bladder epithelial cells (Fig. 3 d) and glomerular epithelial cells (data not shown).

As with bacterial binding to epithelial cells, the effect of complement was due to opsonization of bacteria rather than an effect on the epithelial cells. Pre-exposure of bacteria to complement proteins led to an increase in bacterial internalization (Fig. 3,e). However, pre-exposure of epithelial cells had no effect on the rate of bacterial internalization (Fig. 3 f).

To confirm that this was a specific affect of complement activation, complement in serum was inactivated by CVF treatment. CVF depletes C3 and therefore prevented C3 deposition on bacteria (Fig. 4,a) but did not reduce the deposition of another potential opsonin C4 (Fig. 4,b). Pretreatment of NHS with CVF abolished its ability to augment bacterial internalization by epithelial cells to the level seen with HINHS (Fig. 4 c). C4b did not substitute for C3b in enhancing bacterial internalization, possibly due to lesser amounts of C4b deposited on the bacterial surface in the absence of the amplification provided by the alternative pathway.

For this effect to be relevant to clinical infection. bacterial opsonization needs to occur within the urinary tract. C3 was detected in the urine of patients with acute UTI by Western blotting. Intact C3 (α- and β-chains, 115 and 75 kDa, respectively) as well as the activation product C3 α2 was present in infected urine samples (Fig. 5,a, Lane 4). The C3 concentration (corrected for urinary creatinine concentration) in the urine of 6 normal controls and 20 patients with active UTI are shown in Fig. 5 b. There was a significant increase of ∼15-fold in the mean C3/Cr ratio during infection (mean 2.6 and 38.6 mg/μmol Cr in control and infected urine, respectively; p < 0.005). The maximum C3 concentration achieved was 14.4 μg/ml, which equates to ∼1.4% of the C3 concentration in serum. This is above the level required for demonstrable C3 opsonization of bacteria and suggests that sufficient C3 is present in urine to lead to functionally significant opsonization of bacteria within the urinary space during infection.

We next determined whether urinary complement proteins can opsonize bacteria. Cells and bacteria were removed from infected urine by centrifugation. Laboratory grown bacteria (free from prior contact with serum proteins) were then added to the cleared urine, incubated at 37°C for 2 h, and then extensively washed. Surface-bound C3 was eluted and Western blotted. C3 activation products could be detected from the bacterial surface that must be derived from the urine (Fig. 5 a, Lane 5). These observations support the ability for C3 opsonization of bacteria to occur during clinical infection.

It is clear from these data that complement proteins on bacterial surfaces can interact with ligands on epithelial surfaces. The membrane expression of proteins capable of binding C3 was, therefore, determined. Both CD46 (Fig. 6, a and b) and CD55 (Fig. 6, c and d) were present on the epithelial cell surface. However, CR1–CR4 were not detected on cells by FACS, and no evidence of gene expression could be detected by RT-PCR (data not shown).

CD55 is a GPI-linked protein and can be removed from the cell surface by treatment with PI-PLC (Fig. 7,a). However, reduction in the level of surface CD55 expression by ∼50% did not affect the rate of bacterial internalization (Fig. 7,b). One possible explanation for this result is that, even at the highest concentration of PIPLC, only ∼50% of the surface CD55 is removed. However, PI-PLC treatment was sufficient to reduce the internalization of Dr fimbriated strain 1H11128(34). This E. coli strain is dependent on Dr fimbriae binding to CD55 to mediate internalization (Fig. 7 c). This would suggest that PI-PLC treatment does induce a functionally significant reduction in cell surface CD55.

A recent study has suggested that CD46 may be responsible for the internalization of opsonized bacteria by lung epithelium. CD46 is expressed widely throughout the urinary tract. CD46 expression on tubular epithelial cells was reduced by transfection of cells with CD46 siRNA. A reduction was observed at siRNA concentrations >25 nM, but silencing was most pronounced at 100 nM, reducing mRNA levels by 45% and protein synthesis by 73% (Fig. 8, a and b). Because CD46 is required to control complement activation, it is possible that suppression of CD46 would lead to an increase in cell damage when exposed to serum proteins. LDH release after exposure to NHS did not differ from cells with suppressed CD46 expression, compared with that released by cells transfected with a random sequence siRNA.

A significant reduction in the rate of internalization was seen when CD46 levels were suppressed (Fig. 8,c). There was an ∼8-fold reduction in internalization, accounting for the majority of the complement-mediated increase. The siRNA dose-dependent reduction in CD46 expression was mirrored by a reduction in bacterial internalization (Fig. 8 d).

Purified soluble CD46 provided another means of blocking C3–CD46 interactions. It binds C3b and, therefore, has the capacity to competitively inhibit the interaction between C3 opsonic fragments deposited on the bacterial surface and epithelial cell surface-bound MCP. The addition of 100 μg/ml CD46 led to a reduction in the internalization of E. coli by 70.4%, whereas the presence of BSA (used as a control) did not affect the rate of internalization (Fig. 9,a). C3 FACS analysis showed that incubation of bacteria with serum in the presence of soluble CD46 did not affect C3 deposition, compared with bacteria incubated with serum alone (Fig. 9 b).

The results presented here offer several key insights into the mechanism of invasion of the human urinary tract by pathogenic E. coli. A previous study has shown that the murine urinary tract is susceptible to infection by complement resistant E. coli that use C3 to enhance infection of the upper urinary tract (19). Our current study establishes the relevance of this complement-mediated internalization to human UTI. First, we show that human renal tract epithelial cells demonstrate complement-mediated internalization. Second, we demonstrate an increase in the concentration of C3 in infected urine permissive for this effect. Third, our data suggest that the human complement regulatory protein, CD46 (MCP), is the likely receptor on human urinary tract epithelium that mediates the internalization of opsonized E. coli.

Internalization of uropathogenic bacteria by epithelial cells has been recognized for some time (3). However, the potential for internalization to influence the outcome of infection is only now becoming clear. Intracellular bacteria, persistent within the urinary tract, are protected from the host innate immune response. Bacteria can proliferate intracellularly, developing into pod-like structures that can act as a source of recurrent infection (10). Internalization may also provide a route by which bacteria can access deeper structures and establish tissue-invasive infection. Bacterial fimbriae appear to be essential for internalization, not only mediating binding but also promoting the intracellular cytoskeletal reorganization required to internalize the bacteria (35). Type 1 fimbriae, in particular, have been implicated in this process (9, 35).

As with other studies, we found bacterial internalization occurred in the absence of complement activation, suggesting that complement is nonessential for this process. However, the rate of uptake is increased 10-fold when the bacteria are preopsonized with complement or when complement is added at a later stage during the incubation with target epithelial cells. Binding of bacteria to epithelial cells is also increased by C3 opsonization. It is possible that the increased internalization of opsonized bacteria simply reflects greater bacteria–cell contact. However, we go on to show that internalization involves a specific ligand–receptor interaction.

Our data suggest that bacterium-bound C3 interacts with CD46 on the epithelial cell surface, presumably transducing an intracellular signal and resulting in increased internalization. CD46 is recognized as the cellular receptor for many pathogens, including measles virus (36), human herpesvirus 6(24), some adenovirus strains, Group A Strepococci (25), and Neisseria species (37). Binding of these pathogens to CD46 transduces a signal through the cytoplasmic tail of CD46. There are two potential cytoplasmic tails of CD46 (Cyt 1 and Cyt 2) that are generated by differential splicing and appear to have different roles in signal transduction. The cytoplasmic tail of CD46 can be phosphorylated (38) with subsequent activation of both kinases (39) and phosphatases (40). However, the exact signaling pathway(s) activated through CD46 is incompletely understood and may differ in different cell types.

The CD46-mediated internalization described in this report differs from those previously described in that in this case host derived C3 is used by the organism to engage CD46. Exploitation of C3 by cell-invasive bacteria has been described with several organisms. For example, C3-opsonized Mycobacterium tuberculosis (41), Neisseria gonorrhoea (21), and the HIV (20) exhibit facilitated entry into macrophages and epithelial cells. These pathogens use surface-bound C3 to engage the classical opsonic receptors, in particular CR3. However, we were unable to identify these receptors on urinary tract epithelial cells by either FACS or PCR analysis. Instead, our data, using gene silencing and competitive blocking, suggest that bacterium-bound C3b-cell surface CD46 was responsible for this effect. Similarly, it has been suggested that CD46 may mediate the uptake of C3 opsonized Klebsiella pneumoniae into pulmonary epithelial cells. In contrast, CD55 does not appear to mediate internalization of opsonized E. coli. This is despite its known capacity to mediate the internalization of Dr fimbriated strains, which was again demonstrated in this report.

For this process to occur in vivo, uropathogenic E. coli would need to be opsonized in the urinary space. The urine of healthy individuals contains very little C3. However, as shown in this study, the level of urinary C3 increases significantly during infection, possibly as a result of LPS-mediated stimulation of local epithelial synthesis of C3 or due to transepithelial leak of serum complement proteins. The local production of C3 may originally have served as an important defense against complement-sensitive organisms. However, the majority of pyelonephritic isolates are resistant to complement lysis, suggesting selective pressure and, therefore, a fundamental role for complement. It appears that complement-resistant strains have manipulated this system of defense, resulting in a mechanism by which pathogens can potentially evade other extracellular immune defense mechanisms.

The results described in the present study show that intracellular infection of human epithelial cells by urinary tract E. coli may be mediated by the complement system. Although our results are compatible with either a protective mechanism clearing bacteria from the urinary tract or a mechanism for enhancing the pathogenicity of E. coli, the published data in mice favor a pathogenic effect of complement use. Mice deficient in the complement component C3 have striking resistance to ascending infection of urinary tract. In fact, <10% of C3-null mice with bladder inoculation of E. coli J96 develop ascending infection, compared with 60% of wild-type mice, consistent with a benefit for the organisms in the presence of urinary opsonization (19).

In conclusion, our results depict a novel mechanism of intraepithelial invasion by a common urinary pathogen that is dependent on the detection of bound C3 by the complement regulatory protein, CD46. Based on previous research in mice, it is likely that this complement-mediated uptake enhances the pathogenicity of the organism. C3 opsonization acts in conjunction with fimbrial adhesins to maximize cell invasion. Novel approaches to the treatment of E. coli infection depend on understanding the mechanism of infection, including how host factors, such as the complement system, are exploited by uropathogenic bacteria.

The authors have no financial conflict of interest.