Trends in Immunology
Volume 24, Issue 10, October 2003, Pages 528-533
Journal home page for Trends in Immunology

Leucine-rich repeats and pathogen recognition in Toll-like receptors

https://doi.org/10.1016/S1471-4906(03)00242-4Get rights and content

Abstract

Toll-like receptors (TLRs) are the major cell-surface initiators of inflammatory responses to pathogens. They bind a wide variety of pathogenic substances through their ectodomains (ECDs). Here, we ask: what is the structural basis for this interaction? Toll-like receptor ECDs comprise 19–25 tandem copies of a motif known as the leucine-rich repeat (LRR). No X-ray structure of a TLR-ECD is currently available but there are several high-resolution LRR-containing proteins that can be used to model TLRs. We suggest that the basic framework of TLRs is a horseshoe-shaped solenoid that contains an extensive β-sheet on its concave surface, and numerous ligand-binding insertions. Together, these insertions and the β-sheet could provide a binding surface that is 10-fold greater in area than binding surfaces in antibodies and T-cell receptors.

Section snippets

LRR consensus sequence

The leucine-rich motif was first described in 1985 as a 24-residue repeated sequence with characteristically spaced hydrophobic residues [16]. Since then, several closely related LRR subtypes have been identified in a variety of proteins (reviewed in 10, 17). Using these known motifs, we have identified all LRRs in the ten human TLR-ECDs (Table 1). The prevailing LRR consensus sequence for the TLRs is the 24-residue motif shown in Figure 1a. Although the majority of LRRs conform to the TLR

LRR in 3D

The 3D structures of several LRR-containing proteins are now known. In all, each individual LRR forms a loop (Fig. 1d, e), and the juxtaposition of several LRR loops produces a coil or solenoid-like structure (Fig. 1b,c). The conserved hydrophobic residues of the LRR consensus motif (Fig. 1a) point inward and form the core of the solenoid (Fig. 1d,e). Residues 1–10 in the LRR motif (shaded green in Fig. 1a) are present in all LRR subtypes and form a β-strand within each LRR. The remaining

PAMP binding sites formed by insertions in LRR loops

As mentioned earlier, the repeats in TLR-ECDs frequently contain insertions following positions 10 and 15 in the TLR consensus sequence. The insertion at position 10 occurs in a loop that connects the β-face with the convex surface (Fig. 1e). Insertions at this position would therefore be expected to lie in proximity to the β-sheet. Insertions after position 15 might also contact the β-face but are more likely to be located near the convex surface. At that position, they could interact with a

TLR7, TLR8 and TLR9

TLR9 recognizes and is essential for the inflammatory response to bacterial DNA and oligodeoxynucleotides that contain unmethylated CpG sequences (reviewed in [2]). TLR7 and TLR8 are highly homologous to TLR9 but their natural ligands have yet to be discovered. The alignment of selected LRRs from human TLR7, TLR8 and TLR9 to the TLR consensus sequence is shown in Figure 2. Leucine-rich repeat 1 is a classical TLR consensus repeat, whereas LRR3 provides an example of a shortened repeat with a

TLR5

Toll-like receptor 5 is essential for the inflammatory response to flagellins from Gram-negative bacteria [2], and, recently, Mizel et al. demonstrated that flagellin binds directly to the TLR5-ECD [25]. Toll-like receptor 5 contains 20 contiguous, easily located LRRs, among which five (LRR-7, -9, -14, -15 and -17) contain insertions after position 15 (Table 1). Truncation studies [25] showed that the flagellin-binding site in TLR5 is located between residues 386 and 407, placing it in LRR14.

TLR4

The ECD of TLR4 forms a complex with an accessory protein, MD-2, and initiates the innate response to LPS from Gram-negative bacteria 2, 26. In severe infections, the TLR4 response to LPS results in septic shock and death [26]. From our analysis, the TLR4 ECD contains 21 LRRs (Table 1) of which the first eight and last nine are easily identified. The intervening sequence of ∼100 residues cannot easily be aligned with the LRR consensus motif. One possible arrangement, shown in Figure 2, contains

Capping the solenoids

All TLR-ECDs contain non-LRR sequences at their N- and C-termini. After accounting for signal peptides, the remaining N-terminal sequences that flank the LRRs in TLR2–5 and TLR7–9 bear resemblance to the N-terminal cap of CD42b [11]. The N-terminal residues of CD42b form a β-hairpin, with a disulfide bond connecting the β-strands. This type of structure stabilizes the LRR solenoid by capping the hydrophobic core of the first LRR.

In contrast to the N-terminal caps, which are variable among TLRs,

Concluding remarks

Thirty years ago, it was shown that the structural basis for antibody diversity derived from a rigid scaffold that could accommodate the enormous variety of sequences found in the hypervariable regions of antibodies. The scaffold was created by a structure that is now known as the immunoglobulin domain. Had the immunoglobulin fold been observed in other proteins at that time, it would have been possible from sequence data to construct a reasonably correct model for an antibody before its

Acknowledgements

We thank Stephen Shaw (NCI, EIB) and Thang Chiu (NIDDK, LMB) for critical review of the manuscript.

References (35)

Cited by (638)

View all citing articles on Scopus

These authors contributed equally.

View full text