Induction of innate immune responses by Escherichia coli and purified lipopolysaccharide correlate with organ- and cell-specific expression of Toll-like receptors within the human urinary tract


*For correspondence. E-mail Agneta.Richter.; Tel. (+46) 8728 7425; Fax (+46) 8342 651.


Mucosal epithelial linings function as physical barriers against microbes. In addition, they participate in the first line of host defence by production of a variety of proinflammatory mediators when exposed to microbes and microbial agents. Here, we use a human urinary tract infection model to demonstrate that organ- and cell-specific innate responses induced by lipopolysaccharides (LPS) present on Gram-negative bacteria correlates with the expression of Toll-like receptor 4 (TLR4). The presence of TLR4 on human bladder epithelial cells enables them to rapidly respond to bacterial infections in vitro and in vivo. In contrast, TLR4 is not expressed on human proximal tubule cells isolated from the renal cortex, which may explain the cortical localization of bacteria in pyelonephritis. TLR4-negative renal epithelial cells, A498, are non-responsive to purified LPS, however, they respond to viable bacteria via a mannose-sensitive attachment-mediated pathway. To identify LPS components recognised by bladder epithelial cells, a bacterial lipid A mutant and LPS of different chemotypes were tested. Full interleukin 8 induction required hexa-acylated lipid A and was decreased by between 50% and 70% in the presence of O-antigen. Taken together, we propose that multiple independent pathways, which are organ- and cell-specifically expressed, mediate bacterial recognition and determine the outcome of innate responses to infection.


The urinary tract is exposed to microbes in different ways: the urethra and bladder are most frequently exposed, while the renal pelvis, medulla and cortex are exposed to a lesser degree. The intestinal microflora serve as a reservoir for uropathogenic Escherichia coli and, occasionally, these bacteria enter the urinary tract in which they may cause cystitis and/or pyelonephritis (Sokurenko et al., 1999). The inflammatory role of E. coli lipopolysaccharide (LPS) has been extensively characterized in sepsis models, but little is known about this molecule as a virulence factor in urinary tract infection. LPS is an amphipathic molecule anchored to the outer membrane by lipid A and the acylation state of lipid A has been previously shown to be a major determinant for immunogenicity. The final step of lipid A biosynthesis is performed by WaaN, an enzyme that adds myristic acid to the penta-acylated lipid A, thus completing the lipid A structure (Karow and Georgopoulos, 1992). Bacteria harbouring a mutation in waaN are unable to induce E-selectin and cytokine release in endothelial and myeloid cells (Somerville et al., 1996). Furthermore, the same mutation in Salmonella typhimurium allows growth of bacteria to very high numbers within the mouse, before the mouse clears the infection (Khan et al., 1998; Low et al., 1999). This demonstrates that acylation of lipid A is critical for septic shock and death in mice.

In contrast to the highly variable O polysaccharide, lipid A and the inner core structures of LPS are well conserved among Gram-negative species and are therefore recognised as pathogen-associated molecular patterns. Recognition of these patterns by specific receptors induces the innate immune response within a host (Janeway, 1989; Medzhitov and Janeway, 1997). Recently, LPS was shown to signal via Toll-like receptor 4 (TLR4), peptidoglycan and lipoproteins via TLR2, while the ligands for the seven additional human TLRs are unknown (Poltorak et al., 1998; Aliprantis et al., 1999; Takeuchi et al., 1999; Underhill et al., 1999). All human TLRs contain an intracellular domain highly homologous to the interleukin 1 (IL-1) receptor (IL-1R) and, accordingly, TLRs and IL-1R utilize the same signalling pathway (Rock et al., 1998). Stimulation of TLR4 induces nuclear translocation of NF-κB and subsequent transcription of proinflammatory cytokines and co-stimulatory molecules (Medzhitov et al., 1997). This suggests a central role for TLRs in the host's innate recognition, and provides a link between the innate and the adaptive immune systems.

Epithelial linings are frequently exposed to microbes and epithelial cells are known to play an active role in mucosal immunity upon bacterial attachment (Hedlund et al., 1996; Mulvey et al., 1998). This might occur via adhesion-mediated signalling by receptor interaction or via attachment-facilitated presentation of conserved microbial patterns to TLRs on epithelial cells. The recent discovery of TLR4 as the receptor for LPS prompted us to study the contribution of attachment-mediated versus TLR4-mediated microbial recognition in epithelium originating from different organs. The urinary tract infection (UTI) model is suitable for this purpose as we can investigate the response from microbe-exposed bladder epithelium, and compare that with the response from renal epithelium.


IL-8 response of uroepithelial cells to E. coli infection

Following infection by E. coli strain W3110, the cellular response of epithelial cell lines derived from human bladder (T24) and kidney (A498) were investigated. The culture supernatants were analysed for cellular release of the proinflammatory cytokine IL-8 using enzyme-linked immunosorbent assay (ELISA) (Table 1). The presence of bacteria stimulated release of IL-8 from both cell lines, although at different levels. E. coli strain W3110 express type 1 fimbriae, but not P-pili or the exotoxin α-haemolysin (data not shown). Type 1-mediated binding to mannosylated proteins present on cells within the urinary tract can be blocked by the addition of mannose to the medium of tissue culture cells. Incubation of cells in 3% methyl alpha-d-mannopyranoside showed that the IL-8 response from A498 cells was indeed attachment-mediated, as mannose inhibited IL-8 release by 37.3%. In contrast, the response from T24 cells was only decreased by 5.7%, suggesting that other dominating mechanisms for bacterial stimulation of these cells exist.

Table 1.  IL-8 release from epithelial cells following E. coli stimulation.
Cell lineStimuliIL-8 (ng ml−1) ± SE% inhibition by mannose
T24None2.8 ± 0.2 
E. coli59.4 ± 1.8 
E. coli+ mannose56.0 ± 1.75.7%
A498None0.10 ± 0.01 
E. coli4.79 ± 0.5 
E. coli+ mannose3.0 ± 0.437.3%

LPS induces proinflammatory cytokines in bladder but not in renal epithelial cells

LPS is a major contributor to the induction of inflammation and sepsis in mammals. To elucidate whether LPS act as inducers of the non-attachment-mediated phenotype observed in T24 cells, experiments were performed in which different commercially obtained LPS preparations were used to stimulate the cells. Rough LPSLCD25 (same chemotype as LPS on strain W3110) and deep rough LPSD31m4 were both potent inducers of IL-6 and IL-8 (Fig. 1). Interestingly, IL-6 release was decreased by 70% and IL-8 by 60% when smooth LPSO55:B5 was used to stimulate the cells compared with LPSLCD25. Renal cells did not respond to any of the LPS preparations tested, although up to 5 μg of LPS ml−1 were used in an attempt to stimulate the cells. This is in sharp contrast to bladder epithelial cells, in which 5 ng ml−1 LPSLCD25 induced a full response of IL-8 (20 ± 2.6 ng ml−1).

Figure 1.

Bladder epithelial cells respond to LPS, while renal epithelial cells are non-responsive. ELISA measurements of IL-6 (A) and IL-8 (B) in confluent T24 and A498 cells as a response to E. coli LPS of different chemotypes; O55:B5 express O-antigen as well as outer and inner core sugars (black bars), LCD25 express outer and inner core sugars (striped bars) and the deep rough D31m4 express inner core sugars (bars with grids). Unstimulated cells are presented as white bars. Cytokine concentrations were determined in culture supernatants 6 h after stimulation. Numbers are means ± SE of three independent experiments.

Bladder and renal epithelial cells respond differently to under-acylated lipid A

To elucidate whether bladder and renal epithelial cells recognise the state of lipid A acylation, we isolated LPS from E. coli strain W3110 (hexa-acylated lipid A) and the isogenic mutant MLK1067 (penta-acylated lipid A). Figure 2A shows that T24 cells stimulated with 5 μg of LPSW3110 release fivefold more IL-8 than cells stimulated with underacylated LPSMLK1067. Similar results were obtained when 5 ng ml−1 of the two LPS preparations were used as stimuli (data not shown). As expected, renal cells were unresponsive to purified LPS.

Figure 2.

Bladder and renal epithelial cells respond differently to penta-acylated lipid A. Concentration of IL-8 in supernatants from human bladder (T24) and renal (A498) epithelial cells stimulated with (A) 5 μg ml−1 LPS isolated from E. coli strains W3110 (black bars) or MLK1067 (striped bars) and (B) cells infected by E. coli (3 × 106 cfu ml−1) strains W3110 (black bars) or MLK1067 (striped bars) 6 h post infection. Unstimulated cells are presented as white bars. Numbers are means ± SE of three independent experiments.

It is possible that the presentation of purified LPS to target cells differs from the way native LPS is presented. Wild-type bacteria (W3110) and the isogenic lipid A mutant (MLK1067) were used as stimuli to investigate whether the acylation state of lipid A was discriminated when LPS was present within intact bacterial membranes. We found that T24 cells released fivefold more IL-8 upon infection with wild-type bacteria than the penta-acylated mutant (Fig. 2B). This result is in parallel with the result obtained using purified LPS, and indicates that LPS is similarly recognised in soluble and membrane-bound form. Furthermore, the decreased response of bladder epithelial cells to MLK1067 is consistent with the established role of lipid A as the most immunogenic moiety of LPS (Somerville et al., 1996). No dramatic effect was observed in kidney-derived epithelial cells exposed to either wild-type or mutant bacteria.

Responsiveness to bacteria is correlated with the expression of Toll-like receptor 4

We hypothesised that the cell-specific responses to E. coli and E. coli LPS could be explained by different expression patterns of Toll-like receptors. Total RNA from cell lines T24 and A498 was isolated, subjected to reverse transcription-polymerase chain reaction (RT-PCR) and subsequently analysed in our custom-designed multiplex PCR. This multiplex PCR included primers specific for TLR1–5, the macrophage-specific marker CD14 and the internal control GAPDH (Fig. 3). Bladder epithelial cells express TLR2, TLR3 and TLR4, while renal epithelial cells express TLR3 and TLR5. Hence, the difference in cytokine response between LPS-stimulated T24 and A498 cells could be explained by the absence of TLR4 in renal epithelial cells. Several previous studies show that LPS signals via TLR4 in a variety of cells (Hoshino et al., 1999; Faure et al., 2000; Ogata et al., 2000). Therefore, we used the macrophage-like cell line THP-1 to investigate whether LPS responsiveness correlated with TLR4 expression. We found that, in addition to the macrophage marker CD14, these cells express the same TLRs as bladder epithelial cells. Taken together, these results suggest that cell-specific expression of TLR4 is the major determinant for LPS recognition and the subsequent induction of inflammatory responses in a subset of human cells.

Figure 3.

Expression of Toll-like receptors in different cell lines. Total RNA isolated from bladder epithelial (T24), renal epithelial (A498) and macrophage-like (THP-1) cells were subjected to RT-PCR. The mRNA expression pattern for TLR1-5, CD14 and GAPDH was visualised using multiplex PCR.

TLR expression in vivo

Epithelial cells of the proximal tubules were recently shown to be the renal cell type that bacteria initially interact with during pyelonephritis (Uhlén et al., 2000). A multiplex PCR analysis of isolated human proximal tubule cells (Soderhall et al., 1997) revealed that these cells do not express TLR4 or CD14, while TLR2, 3 and 5 were expressed (Fig. 4). This pattern is similar to the TLR4-negative A498 cells, but differs from the pattern of whole kidney mRNA. However, the kidney is a multicellular organ that includes TLR4-positive endothelial cells as well as other cell types with unknown TLR expression patterns (Zhang et al., 1999). Human bladder mucosa revealed that bladder epithelium expresses a similar pattern of receptors as T24 cells (TLR 2, 3 and 4), as well as TLR5 and CD14.

Figure 4.

In vivo expression of Toll-like receptors in the urinary tract. Total RNA from bladder mucosa (bladder), whole kidney (kidney) and isolated human proximal tubule cells (hPTC) were used in a multiplex PCR to investigate the expression pattern of TLRs in vivo.


Some organs are immunologically privileged sites, i.e. they do not elicit immune responses. For instance, the brain and the anterior chamber of the eye are sites in which tissue can be grafted without inducing graft rejection. Instead of eliciting a destructive immune response, these organs induce tolerance or a response that is non-destructive to the tissue (Janeway et al., 1999). Although the kidney is not an immunologically privileged site, we hypothesized that its major role in maintaining constant ion concentrations during homeostasis is of such importance that only tightly regulated local inflammatory responses are allowed. This is in sharp contrast to the immunocompetent urinary bladder, which is equipped with mechanical clearing mechanisms as well as other systems to rapidly sense and respond to the presence of bacteria.

In this study we examined whether there exists an organ-specific response to bacteria within the urinary tract and, if so, the possible mechanism(s) for this. Our data show that the bladder epithelial cells respond avidly to purified LPS, while renal epithelial cells are non-responsive. In accordance with previous publications, we found that native, hexa-acylated lipid A is required for maximal cytokine induction in the bladder cells. More surprisingly, we found that LPS with an intact O-antigen is less potent as an inducer of IL-6 and IL-8 than rough LPS. The molecular weight of smooth LPS is six times greater than that of rough LPS (25 kDa versus 4 kDa; G. Widmalm, personal communication). Rough LPS (LCD25) could be diluted 1000-fold and still induce comparable amounts of IL-8. Hence, the increased stimulatory effect of rough LPS is not as a result of higher molar concentrations. We suggest that one function of the O-antigen is to protect core sugars and lipid A from being recognised by the innate immune system, providing an explanation as to why O-antigen is expressed on most clinical UTI isolates of E. coli.

Poltorak et al. (1998) identified TLR4 as the receptor mediating LPS-induced signalling via NF-κB. Here, we used multiplex PCR to show that cells derived from the epithelial lining of human bladder express TLR4. This guarantees efficient induction of the proinflammatory response, as shown by IL-6 and IL-8 release of LPS-stimulated T24 cells. The inability of A498 cells to induce proinflammatory cytokines upon LPS stimulation is probably because of the lack of TLR4 expression. Consistent with the in vitro data, TLR4 could not be detected in isolated human proximal tubule cells.

We compared whether cells recognise viable bacteria by a similar mechanism as when they are exposed to soluble LPS. When E. coli was used to infect the tissue culture cells, we found that both bladder and renal cells responded by cytokine release, although by different degrees. To investigate a possible effect of bacterial attachment, internalization and intracellular multiplication on the cellular response, experiments were performed in the presence of mannose. Binding of Type 1 fimbriae to epithelial cells was recently demonstrated as being crucial for attachment, bacterial invasion, as well as intracellular multiplication (Martinez et al., 2000). The IL-8 response from A498 cells was inhibited by blocking the binding of type 1 fimbriae, whereas the response from T24 cells was barely affected. Thus, the low response observed in A498 cells when exposed to bacteria is attachment-mediated rather than TLR4-mediated. In contrast, adhesion onlyplays a minor role in mediating cytokine induction by T24 cells. This result strongly suggests that different recognition and signalling systems exist in these cells.

Similar to the epithelial linings of the airways and the gastrointestinal tract, the urinary tract acts as a bacterial port of entry into the human body. In order to control bacterial multiplication, an immediate innate response is required at the mucosal linings. Based on the results presented here, we hypothesize that, although the urinary bladder can combat bacteria in various ways, the major cytokine response is as a result of induction of the TLR4-signalling pathway upon exposure to LPS. The capacity of bladder epithelial cells to respond to soluble LPS enables them to sense an infection at an early stage. This mechanism is not present within renal cells owing to their lack of TLR4 expression. However, other pathways exist for bacteria-induced cell stimulation; once bacteria reach the site of proximal tubules, α-haemolysin induced [Ca2+]i oscillations and bacterial attachment, both resulting in increased release of IL-6 and IL-8 (Uhlen et al., 2000). This arrangement enables them to maintain the delicate function of ion homeostasis, which is too important to be disturbed by occasional exposure to LPS. However, once bacteria reach the renal tubules, the condition is serious enough for cellular induction of proinflammatory cytokines. Thus, the importance of the organ in maintaining homeostasis determines which pathways it utilizes for bacterial recognition.

Experimental procedures

Bacterial strains and growth conditions

E. coli strains W3110 and MLK1067 (W3110 waaN) were used in this study (Karow and Georgopoulos, 1992). Both strains express type-1 fimbriae but not P-pili or α-haemolysin (data not shown). Static overnight cultures of strains W3110 and MLK1067 were grown in Luria–Bertani (LB) medium at 37°C. Bacteria were washed thoroughly in phosphate-buffered saline (PBS) before addition to the tissue culture medium at a density of 3 × 106 colony-forming units (cfu) ml−1.

LPS preparations

Strains W3110 and MLK1067 grown in 2.5 l of LB medium were harvested in late log phase. Bacteria were resuspended in 0.9% NaCl and incubated for 12 h at 4°C. Following one wash, bacteria were resuspended in dH2O (20 g l−1). This suspension was heated to 68°C, mixed with equal amounts of warm phenol (68°C) and incubated for 12 h at 4°C. Following 72 h dialysis against water (Spectra/Por, Mw cut-off 12–14 kDa, Spectrum), the sample was freeze-dried. LPS was resuspended in dH2O (25–35 mg ml−1) and briefly centrifuged. The supernatant was ultracentrifuged (80 000 g) for 20 h. The obtained pellet was resuspended in dH2O and purity was controlled on a silverstained Laemeli gel. Purified LPS (O55:B5, LCD25 and D31m4 LPS) were from List Biochemicals. All LPS preparations were used at 5 μg ml−1 and sonicated for 5 min before addition to cells.

Cell stimulation

The human epithelial cell lines T24 (ATCC HTB-4, human bladder carcinoma) and A498 cells (ATCC HTB-44, human kidney carcinoma) were grown in 24-well cell-culture plates in RPMI-1640 medium supplemented with 10% fetal calf serum, 25 mM HEPES and 2 mM l-glutamine (Life Technologies) at 37°C in 5% CO2. At confluency, cells were washed before the addition of medium with or without stimuli. Supernatants were collected 6 h post infection and analysed using ELISA (Diaclone) for proinflammatory cytokines IL-6 and IL-8.

RNA isolation, RT-PCR and multiplex PCR

Human proximal tubule cells were isolated as previously described (Soderhall et al., 1997) and stored at −80°C. The mucosa of a human bladder was dissected and snap-frozen before RNA was isolated. The human monocytic cell line THP-1 (ATCC TIB-202) was differentiated to macrophage-like cells during 48 h in the presence of 160 nM of the phorbol ester PMA (Sigma). Then, cells were washed and cultured for 60 h in the absence of PMA. T24 and A498 cells were cultured as above. Total RNA was isolated by TRIZOL (Life Technologies). RT-PCR was performed according to the manufacturer's instructions (Roche). The custom-designed multiplex PCR reactions contained primers specific for Toll-like receptors 1–5 and CD14, with GAPDH as the internal standard. The reactions were analysed on 2% agarose gels. mRNA from human kidney was obtained from Invitrogen.


This work was supported by grants from the programmes ‘Glycoconjugates in Biological Systems’ sponsored by the Swedish Foundation for Strategic Research (F.B.), the Swedish Cancer Foundation (S.N.), Bristol-Myer Squibbs (S.N.), the Medical Research Council (S.N. and A.R.D.) and the Wenner-Gren Foundation (A.R.D.).