Innate immunity recognizes invading micro-organisms and triggers a host defence response. However, the molecular mechanism for innate immune recognition was unclear. Recently, a family of Toll-like receptors (TLRs) was identified, and crucial roles for these receptors in the recognition of microbial components have been elucidated. The TLR family consists of 10 members and will be expanding. Each TLR distinguishes between specific patterns of microbial components to provoke innate immune responses. The activation of innate immunity then leads to the development of antigen-specific adaptive immunity. Thus, TLRs control both innate and adaptive immune responses.
The host defence response to pathogens depends on the immune system. The immune system in mammals consists of innate and adaptive immunity. Adaptive immunity is a highly sophisticated system that is mediated by antigen-specific T and B cells, and is observed only in vertebrates. In contrast, innate immunity is conserved from invertebrates to vertebrates. Even invertebrates harbouring only innate immunity have an effective host defence system. The mechanism for recognition of non-self (antigen) by adaptive immunity has been investigated intensively and the major mechanisms have been clarified. However, the mechanism for innate immune recognition of non-self has long remained unclear.
Studies of the host defence system in fruit flies provided the first clue as to the mechanism of innate immune recognition. In Drosophila, a family of Toll receptors plays an important role in combating the invasion of pathogens (Hoffman et al. 1999). Subsequently, homologues of Drosophila Toll were identified in mammals, and are now termed Toll-like receptors (TLRs) (Medzhitov & Janeway 1997). TLRs compose a large family with at least 10 members. Evidence is now accumulating that TLRs play an important role in the recognition of components of pathogens and subsequent activation of innate immunity, which then leads to development of adaptive immune responses (Medzhitov & Janeway 2000; Aderem & Ulevitch 2000). In this review, we will focus on recent advances in the study of innate immune recognition involving Toll-like receptors.
Host defence response in Drosophila
Drosophila, although they have no adaptive immunity, show an effective host defence response against the invasion of micro-organisms through synthesis of anti-microbial peptides (Hoffman et al. 1999). The anti-fungal peptides have been shown to be induced through activation of the signalling pathways via Toll, which was originally identified as a receptor responsible for dorso-ventral patterning in the developing embryo (Lemaitre et al. 1996). Subsequently, a family protein of Toll, 18-wheeler, has been shown to regulate the production of anti-bacterial peptides (Williams et al. 1997). Adult fruit flies which are mutated in Toll and 18-wheeler are susceptible to infection by fungi and bacteria, respectively (Lemaitre et al. 1996; Williams et al. 1997). Thus, in Drosophila, at least two Toll family proteins discriminate between pathogens, and induce effective host defence responses (Hoffman et al. 1999). The Toll family has now been expanded to include eight members in Drosophila (Tauszig et al. 2000). The function of these newly identified members remains unclear.
Innate immune recognition by Toll-like receptors in mammals
The cytoplasmic portion of Toll-like receptors shows a high similarity to that of the IL-1 receptor family, and is now called the Toll/IL-1 receptor (TIR) domain. In spite of this similarity, the extracellular portions of both receptors are structurally unrelated. The IL-1 receptor is characterized by the presence of an Ig-like domain, whereas Toll-like receptors bear leucine-rich repeats in the extracellular domain. Important roles for TLRs in the recognition of components of micro-organisms are rapidly being clarified (Fig. 1).
TLR4 is essential for the recognition of lipopolysaccharide
Lipopolysaccharide (LPS) is a major component of the outer membrane of Gram-negative bacteria. Two mouse strains, C3H/HeJ and C57BL10/ScCr, are known to be LPS-hyporesponsive. Two independent groups analysed the gene responsible for hypo-responsiveness to LPS, and found mutations in the Tlr4 gene in these strains (Poltorak et al. 1998; Qureshi et al. 1999). In the C3H/HeJ mouse strain, a point mutation in the cytoplasmic region of TLR4 resulted in an amino acid change from proline to histidine. This single amino acid conversion has been shown to result in defective TLR4-mediated signalling (Hoshino et al. 1999). Another LPS hypo-responsive strain, C57BL10/ScCr, was shown to be null mutated in the Tlr4 gene (Poltorak et al. 1998; Qureshi et al. 1999). The generation by gene targeting of mice lacking TLR4 also demonstrated the LPS hyporesponsiveness of TLR4-deficient mice, confirming that TLR4 is a long-sought receptor that is essential for the recognition of LPS (Hoshino et al. 1999).
Although TLR4 has been established as an essential receptor for the LPS-induced activation of immune cells, several additional molecules have also been shown to be involved in the recognition of LPS. CD14 is a glycosylphosphatidylinositol (GPI)-anchored molecule which is preferentially expressed in monocytes/macrophages and neutrophils, and which is postulated to bind to LPS trapped by LPS binding protein (LBP). CD14-deficient mice showed a reduced response to LPS, demonstrating that CD14 is critically involved in the recognition of LPS (Haziot et al. 1996; Moore et al. 2000). Furthermore, a physical association between CD14 and TLR4 has recently been demonstrated (Jiang et al. 2000). The involvement of CD11b/CD18 (also called Mac-1) in LPS responsiveness has also been reported. In CD11b/CD18-deficient mice, the LPS-induced production of cyclo-oxygenase 2 (COX-2) and IL-12 was partially diminished (Perera et al. 2001). Miyake and colleagues identified MD-2 as a molecule that physically associates with the extracellular portion of TLR4 and enhances LPS responsiveness. They used Ba/F3 cell lines that had no ability to respond to LPS, and showed that the enforced expression of both TLR4 and MD-2, but not TLR4 alone, conferred the LPS responsiveness (Shimazu et al. 1999; Akashi et al. 2000). Miyake and colleagues (Miyake et al. 1995) further identified RP105, which bears leucine-rich repeats that are structurally related to TLRs in the extracellular portion. RP105 is expressed in B cells, but not in macrophages. Accordingly, B cells from RP105-deficient mice demonstrated a severely reduced response to LPS (Ogata et al. 2000). They also showed a functional association of TLR4 and RP105, indicating that RP105 is closely involved in the recognition of LPS, together with TLR4 in B cells. Thus, several additional components are implicated in the recognition of LPS, indicating that the functional LPS receptor consists of a large complex of several molecules. Additionally, the expression of RP105 is observed in B cells but not in macrophages, whereas the expression of the TLR4-MD-2 complex is hardly detected in B cells despite its strong expression on mouse peritoneal macrophages (Akashi et al. 2000; Nomura et al. 2000). From the distinct expression patterns of RP105 and MD-2, we suspect that the type of LPS receptor complex varies between B cells and macrophages.
TLR2 is essential for the recognition of lipoproteins and peptidoglycan
TLR2 was first implicated in the recognition of LPS by studies on the over-expression of TLR2 in human embryonic kidney 293 cells (Yang et al. 1998; Kirschning et al. 1998). However, subsequent studies in mice lacking TLR2 or chinese hamster ovary (CHO) fibroblasts genetically lacking functional TLR2 demonstrated that TLR2 is not involved in the recognition of LPS (Heine et al. 1999; Takeuchi et al. 1999b). Over-expression of TLR2 in 293 cells seems to confer a response to the TLR2 ligand contaminating the preparation of LPS used. Indeed, re-purification of LPS demonstrated that TLR4, but not TLR2, is a receptor for LPS (Hirschfeld et al. 2000; Tapping et al. 2000). However, it is still intriguing whether TLR2 is involved in all kinds of LPS or not. Two recent reports have indicated that the response to LPS from Leptospira interrogans or Porphyromonas gingivalis is mediated by TLR2 (Hirschfeld et al. 2001; Werts et al. 2001). Even in these papers, the authors could not rule out the possibility that very small amounts of the TLR2 ligand had contaminated the LPS preparation. More experiments will be required to settle this matter.
TLR6 participates in the discrimination of lipoproteins
The roles of TLR1 and TLR6, both of which are highly homologous (Takeuchi et al. 1999a), have recently been revealed. Functional cooperation of TLR1 and TLR2 in the response to components of Neisseria meningitidis was shown in a study on the ectopic expression of TLR1 and TLR2 in HeLa cells (Wyllie et al. 2000). Aderem and colleagues analysed the role of TLR6 by expressing the dominant negative form in the RAW264 macrophage cell line, and demonstrated that TLR6 cooperates with TLR2 to detect the specific pattern of peptidoglycan or a modulin secreted from Staphylococcus aureus (Ozinsky et al. 2000; Hajjar et al. 2001). They also demonstrated that TLR6 associates with TLR2. The generation of TLR6-deficient mice has revealed that TLR6 functionally associates with TLR2 and discriminates between microbial lipopeptides (Takeuchi et al. 2001). In this study, TLR6-deficient macrophages did not show any inflammatory response to mycoplasma-derived dipalmitoyl lipopeptides. However, these cells responded normally to tripalmitoyl lipopeptides derived from other bacteria. In contrast, TLR2-deficient macrophages showed no response to either type of lipopeptides. This also indicates the existence of other TLRs that functionally associate with TLR2 to recognize tripalmitoyl lipopeptides. Thus, some TLRs seem to associate with other TLRs to discriminate between a subtle difference in the microbial components.
TLR9 is essential for the recognition of bacterial CpG DNA
In addition to the cell wall components of pathogens, bacterial DNA is known to have activity to stimulate immune cells. The immunostimulatory activity of bacterial DNA is attributed to the presence of unmethylated CpG motifs. CpG motifs in vertebrate genomic DNA are observed at reduced frequency and highly methylated, which leads to no immunostimulatory activity. Synthetic oligodeoxynucleotides (ODNs) containing unmethylated CpG motifs also activate immune cells. The administration of CpG DNA induces potent Th1-like immune responses that are protective against several infectious and immune disorders in animal models. Therefore, CpG DNA is now anticipated to be a clinically useful adjuvant for a variety of vaccines against infectious diseases, cancer and allergy (Wagner 1999).
The critical involvement of TLR9 in the recognition of CpG DNA has been demonstrated through the generation of TLR9-deficient mice (Hemmi et al. 2000). The cellular response to CpG DNA, including B cell proliferation, macrophage production of inflammatory cytokines, and dendritic cell maturation, were all abolished in TLR9-deficient mice. The CpG DNA-induced activation of intracellular signalling molecules, including NF-κB, JNK and IRAK, was also compromised in TLR9-deficient mice. Accordingly, TLR9-deficient mice were resistant to CpG DNA-induced lethal shock syndrome.
CpG DNA has been shown to be recognized in the endosome after nonspecific uptake into the cells (Wagner 1999, 2001). It remains unclear whether TLR9 is expressed on the cell surface or in the endosome. Flow cytometric analysis using monoclonal antibodies against TLR1, TLR2 or TLR4 clearly demonstrated that these TLRs are expressed on the cell surface (Shimazu et al. 1999; Yang et al. 1999; Akashi et al. 2000; Visintin et al. 2001). These findings appear to suggest that TLR9 is expressed on the cell surface like other TLRs. However, CpG DNA-induced activation of signalling cascades such as JNK and IRAK was observed with delayed kinetics when compared with LPS-induced activation in normal macrophages (Hemmi et al. 2000). This indicates that CpG DNA is not recognized by the same mechanism as in LPS, and seemingly activates cells after its uptake, indicating that TLR9 may exist in the endosome. Bacterial DNA should be exposed to the outside through digestion of bacteria in the phago-endosome before being recognized by TLR9. Thus, the existence of TLR9 in the endosome seems very reasonable. TLR2, which is usually expressed on the cell surface, has been shown to be recruited to phagosomes after stimulation with zymosan (Underhill et al. 1999a; Ozinsky et al. 2000). Therefore, TLR2 may recognize ligands that are naturally secreted or detached from bacteria on the cell surface and ligands that are exposed after digestion in the phago-endosomes.
TLR5 is responsible for recognition of flagellin
The mouse Tlr5 gene has been shown to map to a distal region of chromosome 1, a region where a Salmonella-susceptibility locus presents (Sebastiani et al. 2000). In Salmonella-susceptible MOLF/Ei mice, the expression of TLR5 mRNA was reduced, indicating that TLR5 is implicated in the recognition of components of Salmonella (Sebastiani et al. 2000). Aderem and colleagues found that culture supernatants of Gram-positive and Gram-negative bacteria had the ability to stimulate CHO cells expressing human TLR5. They purified the culture supernatants of Listeria monocytogenes, and identified flagellin as a TLR5-stimulating fraction (Hayashi et al. 2001). Flagellin is a monomeric constituent of bacterial flagella, a polymeric rod-like appendage extending from the outer membrane of bacteria, and has immunostimulatory properties (Eaves-Pyles et al. 2001). Aderem and colleagues also demonstrated that flagellated bacteria, but not non-flagellated bacteria, activated TLR5, indicating that flagellin is a specific ligand for TLR5 (Hayashi et al. 2001).
Additional ligands for TLRs
TLRs appear to recognize viral infection in addition to bacterial invasion. The relation between virus and TLRs was first demonstrated in a report in which the mechanism of immune evasion by poxviruses was examined. Two types of vaccinia viruses have been shown to share an amino acid similarity with the TIR domain, which causes the interference of TLR-mediated intracellular signalling (Bowie et al. 2000). From this finding, we can speculate that the TLR-mediated pathway is involved in the host defence response against infection by vaccinia virus, and the virus in turn develops an evasion system by suppressing TLR signalling. Subsequently, the involvement of TLR4 in the recognition of virus was demonstrated. In TLR4-mutated C3H/HeJ and C57BL10/ScCr mice, the inflammatory response against proteins of respiratory syntcytial virus (RSV) was severely reduced. Accordingly, TLR4-mutated mice did not clear RSV effectively (Kurt-Jones et al. 2000).
Signalling pathways via Toll-like receptors
MyD88-dependent pathway is essential for inflammatory response
The signalling pathway via the TLR family is highly homologous to that of the IL-1 receptor (IL-1R) family. Both TLR and IL-1R interact with an adaptor protein MyD88 in their TIR domains. When stimulated, MyD88 recruits IL-1 receptor associated kinase (IRAK) to the receptor. IRAK is activated by phosphorylation and then associates with TRAF6, leading to the activation of two distinct signalling pathways, JNK and NF-κB (Muzio et al. 1997, 1998; Wesche et al. 1997; Burnsm et al. 1998; Medzhitov et al. 1998). MyD88-deficient mice did not show any response to the IL-1 family cytokines, demonstrating that MyD88 is an essential adaptor in the IL-1-mediated signalling pathway (Adachi et al. 1998). A critical role for MyD88 in the TLR-mediated signalling pathway has also been shown in MyD88-deficient mice. MyD88-deficient mice showed no inflammatory responses to LPS, including the macrophage production of inflammatory mediators, B cell proliferation, and endotoxin shock (Kawai et al. 1999). The cellular responses to peptidoglycan and lipoproteins were abolished in MyD88-deficient mice (Takeuchi et al. 2000a,c). Furthermore, MyD88-deficient cells did not show any response to CpG DNA (Hacker et al. 2000; Schnare et al. 2000). Finally, MyD88-deficient mice were resistant to flagellin–induced shock syndrome (Hayashi et al. 2001). These findings demonstrated that MyD88 is essential for the inflammatory responses mediated by all the TLR family. Indeed, MyD88-deficient mice are highly susceptible to infection by Staphylococcus aureus (Takeuchi et al. 2000b). Interestingly, MyD88-deficient mice housed in specific-pathogen free conditions are healthy. However, when housed in a conventional facility where several pathogens existed, many of the young MyD88-deficient mice suffered from infectious disorders such as abscess and died before weaning (personal observations). This indicates that the MyD88-dependent activation of innate immunity is very important for prevention of infectious disorders, even when immunogloblin (e.g. adaptive immunity) is supplied maternally.
Roles of the MyD88-independent pathway
Although an essential role for the MyD88-dependent pathway in the inflammatory response was clearly demonstrated in MyD88-deficient mice, an unusual observation was made. In MyD88-deficient macrophages, lipoproteins- and CpG DNA-induced activation of JNK and NF-κB was not observed, in accord with a lack of inflammatory response to both microbial components (Takeuchi et al. 2000a; Hacker et al. 2000). However, LPS-induced activation of JNK and NF-κB was observed in MyD88-deficient macrophages at a level which was equivalent to that in wild-type cells, although delayed (Kawai et al. 1999). Of course, the LPS-induced activation of JNK and NF-κB was not observed in TLR4-deficient macrophages (Takeuchi et al. 1999b). These observations indicate the existence of a MyD88-independent pathway(s) in TLR4-mediated signalling, and raise several questions; Why is the MyD88-independent pathway required? What roles does it have? Why is it specific to the LPS signalling? What molecules does it rely on? Recent studies have provided some answers. The LPS-induced maturation of dendritic cells (DCs), as determined by expression of co-stimulatory molecules and enhancement of T cell allo-stimulatory activity, was normally observed in MyD88-deficient mice, but not in TLR4-deficient mice. This indicates that the LPS-induced maturation of DCs relies on the MyD88-independent pathway (Kaisho & Akira 2001; Kaisho et al. 2001). Analysis of LPS-induced genes in MyD88-deficient macrophages led to the identification of IFN-regulated genes, such as those encoding IP-10 and GARG16. LPS stimulation of MyD88-deficient macrophages has also been shown to lead to the activation of IRF-3 (unpublished data). Thus, the physiological function of the MyD88-independent pathway is now becoming clear.
CpG DNA-induced activation of immune cells appears to depend completely on the MyD88-dependent pathways, because the CpG DNA-induced response was completely abolished in MyD88-deficient mice. However, a recent report indicated the involvement of the catalytic subunit of DNA-dependent protein kinase (DNA-PKcs) in CpG DNA-induced immune cell activation (Chu et al. 2000). DNA-PKcs is a member of the phosphatidyl-inositol-3 kinase (PI3K) family, and originally implicated in the process of repair of DNA double-stranded breaks caused by stress-induced damage from ionized radiation and by programmed DNA rearrangement (called VDJ recombination) during the development of T and B cells. Raz and colleagues demonstrated that macrophages from DNA-PKcs-deficient mice showed a severely reduced production of inflammatory cytokines in response to CpG DNA. They further showed that CpG DNA-induced activation of DNA-PKcs leads to the activation of NF-κB. At present, a connection between the TLR9-MyD88-dependent pathway and DNA-PKcs is not easy to imagine. There would have to be some molecule responsible for cross-talk between TLR9 and DNA-PKcs in the CpG DNA-mediated signalling pathway.
The occurrence of MyD88-independent activation of NF-κB in TLR2 signalling has recently been demonstrated in the human monocytic cell line THP-1 or 293 cells expressing TLR2 (Arbibe et al. 2000). Stimulation of TLR2 with heat-killed Staphylococcus aureus caused the recruitment of active RacI and PI3K to the cytoplasmic portion of TLR2. This induced the activation of Akt followed by activation of the p65 subunit of NF-κB independently of IκBα degradation. These authors mentioned that TLR1, TLR2 and TLR6 possess a PI3K binding motif in their cytoplasmic portions. This result suggests the existence of a unique signalling pathway in the TLR2 signalling.
Role of TLRs in conducting activation of adaptive immunity
In mammals, the activation of innate immunity is followed by the activation of adaptive immunity, which is characterized by the clonal expansion of antigen-specific T and B cells. Instructions for the activation of adaptive immunity are largely provided by DCs. Immature DCs residing in the periphery have a high capacity for endocytosis, which is suitable for antigen uptake. The immature DCs are activated by and maturate in response to microbial components. Mature DCs in turn lose the capacity for endocytosis, migrate into lymph node and interact with naive T cells to initiate the adaptive immune response (Banchereau & Steinman 1998). Expression of TLR1, 2, 4 and 5 is observed in immature DCs, but decreased as DCs matured (Visintin et al. 2001). TLR3 is expressed only in mature DCs (Muzio et al. 2000). Additionally, several bacterial components such as LPS, CpG DNA, lipoprotein and the cell wall skeleton of Mycobacteria have been shown to induce the maturation of DCs via TLR (Hemmi et al. 2000; Kaisho et al. 2001; Tsuji et al. 2000; Hertz et al. 2001). From these findings, we can speculate that TLRs are critically involved in the maturation of DCs and initiate the activation of adaptive immunity (Reis e Sousa 2001) (Fig. 2). However, several questions need to be answered before a full understanding of the roles of TLRs in adaptive immunity can be obtained. First, all of the microbial components that activate TLRs induce the production of IL-12 by DCs, leading to the development of Th1 cells. However, adaptive immunity exhibits Th2 responses as well as Th1 responses. It remains unclear how innate immunity regulates the balance between Th1 and Th2 cell development. Second, it is worth noting that, although TLR1, 2, 4 and 5 are expressed in immature DCs, only TLR3 is preferentially expressed in mature DCs (Muzio et al. 2000). This implies that TLR3 has a unique function. Elucidation of the role of TLR3 in DC function will reveal a new function of TLRs. Finally, phagocytosis is important in combating microbial invasion (Aderem & Underhill 1999). It remains unclear whether phagocytosis precedes the TLR-mediated activation of innate immune cells. In the case of TLR9, CpG DNA-induced activation would occur after phagocytosis, because CpG DNA is recognized in the endosome. However, several TLRs, including TLR1, TLR2 and TLR4 at least, are expressed on the cell surface, suggesting that naturally secreted products from pathogens are recognized before phagocytosis. Elucidation of the relationship between TLRs and phagocytosis will shed light on the mechanism for triggering the activation of innate immunity.
The discovery of Toll-like receptors provided us with an important clue as to the mechanism of innate immunity, which has long remained undiscovered. As we described, the mandatory role of TLRs in the recognition of microbial components is rapidly being revealed. This will hopefully be followed by elucidation of the mechanism of activation of innate immunity and the connection between innate and adaptive immunity.
We thank N. Tsuji and E. Horita for excellent secretarial assistance. This work was supported by grants from the Special Coordination Funds of the Ministry of Education, Culture, Sports, Science and Technology, the Naito Foundation, and the Novartis Foundation for the Promotion of Science.