Report

A Toll-like Receptor That Prevents Infection by Uropathogenic Bacteria

See allHide authors and affiliations

Science  05 Mar 2004:
Vol. 303, Issue 5663, pp. 1522-1526
DOI: 10.1126/science.1094351

Abstract

Toll-like receptors (TLRs) recognize molecular patterns displayed by microorganisms, and their subsequent activation leads to the transcription of appropriate host-defense genes. Here we report the cloning and characterization of a member of the mammalian TLR family, TLR11, that displays a distinct pattern of expression in macrophages and liver, kidney, and bladder epithelial cells. Cells expressing TLR11 fail to respond to known TLR ligands but instead respond specifically to uropathogenic bacteria. Mice lacking TLR11 are highly susceptible to infection of the kidneys by uropathogenic bacteria, indicating a potentially important role for TLR11 in preventing infection of internal organs of the urogenital system.

TLRs represent a family of transmembrane proteins characterized by multiple copies of leucine-rich repeats in the extracellular domain and a cytoplasmic Toll/IL-1 (interleukin-1) receptor homology domain (TIR) (1, 2). Currently 10 TLRs have been reported in mammalian species, and these appear to recognize distinct pathogen-associated molecular patterns (PAMPs). Of these known TLRs, TLR2, TLR3, TLR4, TLR5, and TLR9 have been extensively characterized (1, 2). TLR1, TLR6, TLR7, and TLR8 have not yet been shown to independently impart signals after recognition of specific microbial products. Heterodimerization between certain TLRs also helps to increase the diversity of PAMPs that can be recognized (3, 4).

The completion of the human and mouse genome sequence has provided the opportunity to determine whether the mammalian TLR family extends beyond the 10 known members. To identify as yet uncharacterized TLRs, we used the sequence of the TIR domain of TLR4 to search National Center for Biotechnology Information (NCBI) databases. The sequence encoding TLR11 was first detected as an expressed-sequence tag (EST) from a mouse liver EST database. Sequential searches using this EST ultimately led to the recovery of the mouse genomic sequence encoding TLR11. The GENSCAN program (5) was used to predict the putative open reading frame (ORF), and hypothetical translation predicted a 907 amino acid protein with the hallmarks of known Toll receptors, including a leucine-rich domain, a transmembrane domain, and a TIR domain (1, 2) (Fig. 1A and fig. S1A). Boot-strapping phylogenetic analysis indicated close similarity to TLR5, which suggests that these two TLRs might represent an evolutionarily conserved subgroup within the TLR family (fig. S1B).

Fig. 1.

Expression pattern of TLR11. (A) Alignment of the amino acid sequence of cytoplasmic domains of known Toll-like receptor family members with TLR1 to TLR11. Alignments were performed using the Clustal algorithm and Boxshade. Three regions (boxes 1, 2, and 3) are conserved across all TIR domains and are thought to be important for signaling. (B) Multiple tissue Northern analysis to determine expression pattern of TLR11 mRNA. TLR11 is predominantly expressed in kidney and liver, with considerably lower levels of expression in spleen and heart. A β-actin probe was used as a control for RNA loading. (C) Localization of TLR11 mRNA in mouse tissues by in situ hybridization. TLR11 localization is shown by incubation with the antisense probe (AS) of TLR11 in liver, kidney, and bladder. Control incubations with the TLR11 sense probe (S) were negative.

TLR11 sequences are present in the genomes of many mammalian species, including humans. We have also detected a 2.7-kb mRNA in Northern blots of human kidney cell lines using a mouse TLR11 cDNA probe (6). However, the human genome sequence in the NCBI database indicates the presence of stop codons in the putative TLR11 ORF (fig. S2). Stop codons are also seen in the TLR11 genomic sequence of human 293 and Jurkat cells, which suggests that TLR11 might not be expressed in humans (fig. S2). Recently, a common polymorphism was observed in the human TLR5 sequence that results in the introduction of a stop codon and prevents the expression of TLR5 in certain individuals (7). It is therefore possible that the stop codons in TLR11 might also represent a form of genetic polymorphism.

To analyze the function of this TLR, the cDNA encoding TLR11 was cloned from a mouse liver library (8). In vitro transcription/translation of the cloned TLR11 resulted in a 97-kD protein, which is similar to the size of the other known TLRs (GenBank accession number AY531552) (1, 2). Using the TIR domain of TLR11 cDNA as a probe, we examined the pattern of TLR11 expression in mouse tissues by Northern blot analysis. TLR11 was strongly expressed only in liver and kidney (8) (Fig. 1B), a pattern that is distinct from the other known TLRs (1, 2), which suggests that TLR 11 might play a specific role in these organs. TLR11 expression was low in the spleen, a tissue that expresses abundant levels of most of the other TLRs. Consistent with this, we failed to detect substantial amounts of TLR11 in most blood cell types, with the exception of macrophages (6). In situ staining of isolated sections of the liver, kidney, and bladder with biotinylated TLR11 probes, as well as in situ hybridization, revealed that TLR11 was strongly expressed in the epithelial cells of the kidney and bladder (Fig. 1C). TLR11 was also detected in liver epithelial cells, but the level of expression was lower.

To determine whether TLR11 is capable of initiating signal transduction pathways that lead to the activation of nuclear factor κB (NF-κB) and activating protein 1 (AP-1), we overexpressed a CD4-TLR11 fusion construct in 293 cells stably transfected with an NF-κB or AP-1 reporter construct (293-luciferase cells) (9). Previous studies have shown that CD4 fusion constructs incorporating the TIR domain of other TLRs can induce the spontaneous activation of NF-κB in transfected cells (10). Expression of CD4-TLR11 led to NF-κB activation at levels similar to that seen with CD4-TLR4 (Fig. 2, A and B), which suggests that TLR11 is able to activate the classical TLR signal transduction pathway to NF-κB and AP-1. In addition, the NF-κB target gene, tumor necrosis factor α (TNF-α), was expressed at a high level (Fig. 2C), similar to that seen upon transfection of CD4-TLR4 (6).

Fig. 2.

TLR11-induced activation of transcription measured by reporter-gene expression and endogenous cytokine production. (A and B) Constitutively active TLR11 activates NF-κB and AP-1. The 293 cells were transiently transfected with expression vectors for CD4/TLR11 or CD4/TLR4 fusion constructs. In the constructs, the cytosolic domain of the TLRs (the TIR domain) was fused to the extracellular portion of CD4. The amount of DNA transfected was equalized with empty expression vector, which was also used in the control, along with either an NF-κB (A) or an AP-1 (B) luciferase reporter construct. NF-κB–induced and AP-1–induced luciferase activity were measured using a luminometer. (C) Transfection of RAW 264.7 macrophages with a CD4/TLR11 expression vector. The production of TNF-α was detected by flow cytometry. The blue region indicates TNF-α expression in untransfected cells, whereas the green line represents TNF-α produced in cells transfected with CD4/TLR11. (D) A P804H mutant form of TLR11 does not activate NF-κB upon overexpression. CD4-TLR4 and CD4-TLR11 fusion proteins with the critical proline residue in the TIR domain (P712 in TLR4 and P804 in TLR11) mutated to histidines were overexpressed in 293-luciferase cells. Only the wild-type fusion proteins, but not the mutants, led to the activation of NF-κB. (E) Dominant-negative MyD88 (DN-MyD88), dominant-negative IRAK (DN-IRAK), and dominant-negative TRAF6 (DN-TRAF6) constructs inhibit CD4/TLR11-mediated NF-κB activation. The 293-luciferase cells, stably transfected cells with the NF-κB luciferase reporter construct, were cotransfected with CD4/TLR11 and wild-type or dominant-negative versions of MyD88, IRAK, and TRAF6.

To determine whether TLR11 uses the same signal transduction pathway for NF-κB activation as do other TLRs, wild-type and dominant-negative versions of intermediates in the TLR-signaling pathway, MyD88, interleukin-1 receptor–associated kinase (IRAK), and TNF receptor–associated factor 6 (TRAF6) (11), were transfected into 293-luciferase cells along with CD4-TLR11. All of the dominant negatives, but not the wild-type proteins, strongly inhibited NF-κB activation (Fig. 2D), which suggests that overexpression of TLR11 activates the same general signaling pathway as the other TLRs (11). It has been previously reported that a specific proline residue in the TIR domain is crucial for signaling by most TLRs (12), and changing the residue abolishes the ability of TLRs to activate NF-κB. TLR11 also has a proline residue at this position, and we found that mutating the proline to histidine (P804H) abolished the ability of CD4-TLR11 to activate NF-κB when expressed in 293-luciferase cells (Fig. 2E).

We then tested the ability of known TLR ligands, including lipopolysaccharide (LPS), peptidoglycan (PGN), and polyinosinic:polycytidylic acid (polyI:C) (2), to activate NF-κB in 293-luciferase cells transfected with TLR11 (Fig. 3A). None of the tested ligands were able to activate signaling through TLR11. Although we have not yet identified the natural ligand for TLR11, we hypothesized that its expression in kidney and bladder might mean that it is involved in responses to bacteria that cause infections of the urinary tract. Most uropathogenic bacteria are strains of Escherichia coli (uropathogenic E. coli, or UPEC), although certain strains of Klebsiella, Staphylococcus, Enterococcus, Proteus, and Pseudomonas have also been found in pathogenic isolates (13, 14). We cultured some of these bacterial strains, along with certain nonpathogenic strains of E. coli as controls. Heat-killed bacteria from these cultures were used to stimulate 293-luciferase cells stably expressing TLR11 (8) (Fig. 3, B and C). The uropathogenic strains, E. coli 8NU (15), E. coli NU14 (16), E. coli HLK120, and E. coli AD110 (17), showed striking activation of NF-κB in these cells compared with the nonpathogenic strains, E. coli BL21, E. coli DH5α, Enterococci, and Listeria (Fig. 3, B and C). Because we already know that the TLR11-expressing 293 cells do not respond to most of the known TLR ligands, our results strongly suggest that uropathogenic strains of E. coli contain a potential ligand for TLR11.

Fig. 3.

Uropathogenic bacterial lysates stimulate TLR11-expressing cells. (A) TLR11 is not stimulated by most known TLR ligands. The 293-luciferase cells were transiently transfected with TLR2, TLR4, TLR5, TLR11, or empty expression vectors. Luciferase activity in cells was measured following treatment with 2 μg ml–1PGN, 100 ng ml–1 LPS, 200 ng ml–1 Flagellin, 25 μg ml poly(I:C) (dsRNA), 100 ng ml–1 CpG DNA, or untreated (control) cells. (B and C) The 293-luciferase cells stably transfected with TLR2 or TLR11 were treated with 70 μl ml–1 of heat-killed bacteria from the indicated saturated bacterial cultures or with LB alone (control). Data are representative of three independent experiments.

To test the hypothesis that TLR11 has a more specialized role in immunity to UPECs in the kidney, and to more generally understand the biological role of TLR11, we generated mutant mice lacking TLR11 using homologous recombination (8) (fig. S4). Heterozygous animals appeared normal and, upon breeding, homozygous knockout animals were obtained in the expected Mendelian ratio. Thioglycollateelicited peritoneal macrophages were isolated from wild-type and TLR11–/– mice and tested for response to the well-characterized TLR ligands, LPS, PGN, and polyI:C (2). All these ligands stimulated the wild-type and the knockout macrophages equally well (Fig. 4A), confirming our previous findings that TLR11 does not respond to known TLR ligands. Macrophages were then stimulated with heat-killed E. coli 8NU (a uropathogenic strain) and DH5α (a nonpathogenic strain). Macrophages from the TLR11–/– mice showed a dramatically reduced response to 8NU compared with wild-type cells, whereas the response to DH5α was almost unaffected (Fig. 4A). This experiment strongly suggests that TLR11 recognizes a component in 8NU that is absent in DH5α, and is consistent with our findings in TLR11-expressing cell lines.

Fig. 4.

TLR11–/– mice fail to recognize and respond to uropathogenic E. coli. (A) Thioglycollate-elicited peritoneal macrophages were isolated from wild-type and knockout mice. The macrophages were tested for response to the TLR ligands, LPS, PGN, and poly I:C, and the E. coli strains 8NU and DH5α, by measuring secretion of TNF-α. (B) Infection of kidneys of TLR11–/– mice by UPEC. Significantly greater numbers of bacterial clones (more than 10,000 times as many) were observed in the kidneys of knockout TLR11 mice compared with wild-type mice, even though equal numbers of bacteria were used to infect the animals. (C) Histology of infected bladder and kidney from the TLR11 wild-type and knockout mice. The kidneys from wild-type mice showed significantly greater inflammatory response (estimated by numbers of infiltrating leukocytes) compared with knockout mice, even though the number of bacteria was far lower. However, the inflammation of the bladders was similar in wild-type and knockout mice.

We next examined whether the absence of TLR11 would render mice uniquely susceptible to infection by UPECs. Wild-type and knockout mice were infected intraurethrally with UPEC (E. coli 8NU) and analyzed 96 hours after infection. The urinary bladder and kidneys were removed, and half of each organ was used for determining bacterial titers by plating on LB-agar plates after homogenization, whereas the remainder of the tissue was used to generate slices for histology. Although in each case, the bladder was almost equally infected by UPEC in wild-type and knockout animals, the kidney was massively infected only in the knockout animals (Fig. 4B). Quantitation revealed that there were at least 10,000 times as many bacteria in the kidneys of the knockout animals. Histological examination showed a greater inflammatory response, as measured by infiltration of leukocytes, in wild-type kidneys compared with kidneys from TLR11–/– mice (Fig. 4C).

In summary, in this manuscript we report the identification of a Toll-like receptor that appears to recognize UPECs and to protect the kidneys from ascending infection by UPECs. The urogenital system represents a unique situation where internal organs are potentially continually exposed to pathogenic microorganisms. Previous studies have shown that TLR4, which recognizes LPS, and hence Gram-negative bacteria in general, is expressed in the urinary tract and the bladder epithelium but not in the kidney epithelium (18, 19). Consistent with its expression in these tissues, it has been shown that mice with mutated TLR4, the C3/HeJ strain (20), are unable to mount a robust inflammatory response when infected intra-urethrally with UPECs (19). TLR11 is abundantly expressed in the bladder, where it probably shares the burden of responding to UPECs with TLR4, but in the kidney, TLR11 alone appears to be responsible for mounting innate immune responses. The massive infection of the kidney observed in the TLR11 knockout mice supports the hypothesis that TLR11 provides a barrier that prevents uropathogenic bacteria from infecting the kidneys.

An intriguing aspect of our study is its implication for response to UPECs in humans. Based on the sequence of the human genome in the NCBI database, and the genomic sequence of some human cell lines, it appears that humans might not express full-length TLR11 protein. It is possible that the stop codons in the ORF of human TLR11 might represent a form of genetic polymorphism, similar to the situation observed for TLR5 in which a stop codon within the ORF of human TLR5 in many individuals makes them incapable of responding adequately to flagellated bacterium (7). A systematic analysis of TLR11 sequences will be necessary to determine whether TLR11 is absent from the human population or only from a subpopulation. However, it is tempting to speculate that one of the reasons humans are particularly susceptible to urinary tract infections is because the absence of TLR11 has removed a defense pathway with the unique ability to specifically recognize UPECs.

Supporting Online Material

www.sciencemag.org/cgi/content/full/303/5663/1522/DC1

Materials and Methods

Figs. S1 to S4

References

References and Notes

View Abstract

Navigate This Article