Structural Reorganization of the Toll-Like Receptor 8 Dimer Induced by Agonistic Ligands

See allHide authors and affiliations

Science  22 Mar 2013:
Vol. 339, Issue 6126, pp. 1426-1429
DOI: 10.1126/science.1229159

Dissecting TLR8 Interactions

Toll-like receptors (TLRs) activate the innate immune system in response to invading pathogens. TLR7 and TLR8 recognize single-stranded RNA from viruses and also contribute to the pathogenesis of autoimmune diseases. Tanji et al. (p. 1426) now report the crystal structure of the unliganded TLR8 ectodomain and the TLR8 ectodomain bound to three different small-molecule agonists. Ligand binding to preformed TLR8 dimers induced conformational changes that brought the C-terminal domains closer together, presumably initiating downstream signaling.


Toll-like receptor 7 (TLR7) and TLR8 recognize single-stranded RNA and initiate innate immune responses. Several synthetic agonists of TLR7-TLR8 display novel therapeutic potential; however, the molecular basis for ligand recognition and activation of signaling by TLR7 or TLR8 is largely unknown. In this study, the crystal structures of unliganded and ligand-induced activated human TLR8 dimers were elucidated. Ligand recognition was mediated by a dimerization interface formed by two protomers. Upon ligand stimulation, the TLR8 dimer was reorganized such that the two C termini were brought into proximity. The loop between leucine-rich repeat 14 (LRR14) and LRR15 was cleaved; however, the N- and C-terminal halves remained associated and contributed to ligand recognition and dimerization. Thus, ligand binding induces reorganization of the TLR8 dimer, which enables downstream signaling processes.

The Toll-like receptors (TLRs) are a family of pattern-recognition receptors that recognize microbial components and initiate subsequent immune responses (1). TLRs recognize multiple pathogen-associated molecular patterns and are expressed predominantly on cells of the immune system, such as macrophages, dendritic cells, neutrophils, and monocytes (2). Both TLR7 and TLR8 are involved in the recognition of single-stranded RNA (ssRNA) from various viruses (37), as well as small interfering RNAs (811). TLR7 and TLR8 also mediate the recognition of self RNA that is released from dead or dying cells and contribute to the pathogenesis of autoimmune diseases, such as systemic lupus erythematosus (12).

Phylogenetic analysis has revealed that TLR7, TLR8, and TLR9 form a subfamily of proteins that each contain an extracellular domain of >800 residues and share functional and structural features (13, 14) (fig. S1). Before ligand recognition, both TLR8 and TLR9 exist as preformed dimers; ligand-induced conformational changes in the preformed dimer have been demonstrated for TLR9 (15). This preformed dimer contrasts with the ligand-induced dimerization mechanism of other TLRs (1620).

Both TLR7 and TLR8 are believed to recognize uridine- and guanosine-rich ssRNA (3, 4). Moreover, several small-molecule compounds have been identified as TLR7 and TLR8 activators (21, 22). Some imidazoquinoline derivatives, such as resiquimod (R848), are recognized by both human TLR7 and TLR8 (21); whereas guanosine nucleotides analogs such as loxoribine, and other imidazoquinoline derivatives such as imiquimod (R837), selectively activate TLR7 but not TLR8 (2224). With molecular masses of 200 to 300 daltons, these compounds are the smallest of the TLR ligands and are therefore attractive therapeutic targets for various infectious diseases. In fact, imiquimod is approved for treatment of external genital warts, actinic keratosis, and basal cell carcinoma. The lack of structural knowledge of TLR7 and 8 hinders understanding of the molecular mechanisms underlying activation of signaling and specific ligand recognition. In particular, an understanding of why and how TLR7 and TLR8 are activated by small ligands remains largely unknown.

We sought to determine the crystal structures of an unliganded preformed dimer and ligand-induced activated forms of human TLR8. The TLR8-R848 complex structure was initially determined using two heavy-atom derivatives (Pt and Cs) (fig. S2); the other structures were determined by the molecular replacement method, using the refined TLR8-R848 complex coordinates (table S1 and S2). TLR8 contains 26 leucine-rich repeats (LRRs), which is the largest number of LRRs among TLRs whose structures have been reported (Fig. 1A). Because TLR7, TLR8, and TLR9 are predicted to have a long insertion region (~40 amino acids) between LRR14 and LRR15, we chose to observe this insertion region (residues 442 to 481), hereafter referred to as the Z-loop.

Fig. 1

Structures of human TLR8. (A) Schematic representation of the domain organization of extracellular regions of human TLR8 (hTLR8). LRRs are indicated by numbered boxes. The protruding LRR loops and the long insertion region between LRR14 and LRR15 (Z-loop) that are characteristic of the TLR7-9 family are indicated by curved lines above the LRRs. The regions missing from the structural model are indicated by red dashed lines. The N-terminal and C-terminal halves of TLR8 are shown in light green and light orange, respectively. (B) Monomer structure of the hTLR8-CL097 complex showing the lateral face (left) and the convex face from the N-terminal side (right). The bound CL097, N-glycan residues, and disulfide bonds are shown in stick representations. The O, N, and S atoms are colored red, blue, and orange, respectively. The C atoms of CL097 and N-glycans are shown in yellow and gray, respectively. The N and C termini of each fragment are shown as spheres. (Left) Front and (right) side views of the preformed inactivated state (C) and ligand-induced activated state (D). TLR8 and its dimerization partner TLR8* are green and cyan, respectively. The CL097 molecules in the activated dimerization interface are illustrated by space-filling representations. The distances between the two Cα atoms of each C terminus are indicated.

Although purified TLR8 was found cleaved at the Z-loop, the N terminus (residues 31 to 432) and C terminus (residues 458 to 817) of the TLR8 monomer were directly associated with each other (Fig. 1B). The consensus β strands of LRR14 and LRR15 interacted to form a β-sheet structure (fig. S3A), which was positioned within the concave face of TLR8 (fig. S3B). At the convex face of the structure, LRR13 interacted with LRR15 directly (fig. S3A); as a result, the loop region of LRR14 located after the β strands was excluded from the main body of the LRR structure (fig. S3A). Although the loop region of LRR14 and the first half of the Z-loop (residues 433 to 457) were missing from the electron density maps, most likely because of the disorder of this portion, the latter half of the Z-loop (after Asp458) was well ordered in the concave face of the LRR structure and was stabilized by extensive contact with LRR3 to LRR18 (fig. S3B). At the end of the Z-loop, we observed a single turn of the α helix, which was stabilized by a disulfide bridge between Cys479 (within the Z-loop) and Cys509 (within LRR16). These two Cys residues are highly conserved between the TLR7, 8, and 9 proteins (fig. S1).

Gel filtration analysis revealed that unliganded TLR8 was dimeric in solution (fig. S4), which suggested that the inactivated protein creates a preformed dimer. Consistent with the gel filtration results, the crystal structure of unliganded TLR8 exhibited a dimeric form (Fig. 1C). Throughout this Report, we indicate the second TLR8 and its residues in the dimeric TLR8 with asterisks. The two TLR8 monomers were separated at the C-terminal regions by ~53 Å. The length of this separation likely precludes proper association of the cytoplasmic domain; therefore, this dimeric structure represents the inactivated form of TLR8.

We also determined the crystal structures of TLR8 in complex with three different agonistic chemical ligands, namely, CL097, CL075, and R848 crystal forms 1 to 3 (table S2). Despite TLR8’s being in complex with different ligands and displaying differential molecular packing in the crystal structures, we observed essentially the same dimeric structures of TLR8 for each of the three ligands tested (fig. S5), which excludes the possibility of artificial dimer formation in the crystals. The crystallographic asymmetric unit of the TLR8-CL097 complex contained two copies of the TLR8 monomer, which were assigned as components of the TLR8 dimer (Fig. 1D). Superposition of the two TLR8 molecules in the asymmetric unit yielded a small root mean square deviation (RMSD) value of 0.4 Å, which indicated that the two molecules are essentially identical. The two TLR8 monomers in the dimer, related by noncrystallographic (NCS) two-fold symmetry, were positioned in the lateral face of the LRR structure, in a manner similar to the positioning of other ligand-induced TLR structures (1620). Upon ligand binding, the two C termini were brought into close proximity (~30 Å) (Fig. 1D and fig. S5), which would enable the subsequent dimerization of the intracellular Toll–interleukin-1 receptor (TIR) domain and downstream signaling, which, in turn, suggests that the ligand-bound dimer is a signaling complex.

The dimerization interfaces of the unliganded and the ligand-induced dimers bury ~1290 Å2 and ~2150 Å2 of the accessible surface area of the TLR8 structures, respectively (Fig. 2 and figs. S6A and S7). The unliganded form of TLR8 has a dimerization interface spanning from LRR8 to LRR18. In fact, the dimerization interface is formed by several hydrogen bonds (LRR8, LRR14-15, and LRR18) (table S3) and hydrophobic cores consisting of Tyr348, Phe261, Tyr567*, and Phe568*, and Phe405, Tyr353, Val378, Phe494*, Phe495*, and Pro432*, which are spread around LRR8, LRR11– 15, and LRR18 (fig. S6A). Ligand binding induced the local conformational changes of the loop structures of LRR5, LRR8, and LRR17–20, whereas the overall structures superpose well with an RMSD value of 1.1 Å (fig. S6B). The dimerization interface of ligand-induced dimer is divided into two regions, the protein-protein interface and the ligand-mediated interface, which account for ~75% and ~25% of the buried surface area, respectively (Fig. 2 and fig. S7). The protein-protein interface involves both hydrogen bonds and hydrophobic interactions and is widely spread around LRR5, LRR8, and LRR14–20 (Fig. 2). Hydrogen-bonded pairs of Asn491 O–Arg541*, Ala514 O–Arg541*, and Glu427–His566*, and van der Waals contact of Phe494–Phe494* and Ser516–Ser516* were observed near the two-fold NCS axis (Fig. 2, top panel, and table S3). At the peripheral region of the dimerization interface (Fig. 2, bottom panel), extensive van der Waals contacts and hydrogen bonds were observed. The ligand-mediated interface is composed of residues of LRR11–14 and LRR16–18 as described below.

Fig. 2

Dimerization interface of the ligand-induced TLR8 dimer. (Left, middle) A surface representation of the protein-protein interface (orange) and ligand-mediated interface (red). (Right, middle) Top view of the ligand-induced TLR8 dimer along the two-fold NCS axis. Magnified views of the central dimerization interface around the two-fold NCS axis (top) and the peripheral dimerization interface (bottom). Hydrogen bonds in the magnified views are indicated by dashed lines.

Two clear electron densities corresponding to the ligand were observed in the crystals of TLR8 complexed with chemical ligands (fig. S8A). The two chemical ligands were located in the dimer interface of TLR8 and the two positions were related by the NCS two-fold axis, where the first is close to LRR11–14 and LRR16*–18*, and the second is close to LRR11*–14* and LRR16–18 (Figs. 2 and 3). These ligands were recognized by interactions with multiple amino acids from both TLR8 and TLR8* and act as molecular glues bridging the two TLR8 molecules. All of the ligands tested occupied the same position when the proteins were superposed (fig. S8B). The most noticeable interaction between TLR8 and these ligands was that of the benzene rings of imidazoquinoline (R848) or thiazoloquinolone (CL097, CL075) stacked on the side chain of Phe405 (Fig. 3, A to C). The amidine group of the quinoline moiety formed hydrogen bonds with the side chain of Asp543 with excellent geometry. The N atoms of the imidazole or thiazole moieties formed hydrogen bonds with Thr574 N atoms. The 2-propyl (CL075) or 2-ethoxymethyl (CL097 and R848) substituent of these ligands protruded into the small hydrophobic pocket formed by Phe346, Tyr348, Gly376, Val378, Ile403, Phe405, Gly572*, and Val573*. These hydrophobic interactions may be important for the agonistic activity of chemical ligands targeting TLR8, because R837 (known as imiquimod) (fig. S9A), an imidazoquinoline compound similar to R848, specifically activates TLR7 but not TLR8 and lacks the 2-substituent in the imidazole ring. CL097 forms a water-mediated contact with the carbonyl oxygen of Ser352.

Fig. 3

Ligand recognition sites of TLR8. (A), (B), and (C) Residues involved in the interaction of TLR8 with CL097, CL075, and R848, respectively. The C atoms of the ligand molecules are yellow. Water molecules mediating the ligand recognition are indicated by red filled circles, and hydrogen bonds by dashed lines. The chemical structures of these ligands are shown at top right in each panel. (D), (E), and (F) NF-κB activity of human TLR8 mutants, stimulated by CL097, CL075, and R848, respectively. The reactivity of wild type and various mutants of TLR8 were analyzed by an NF-κB–dependent green fluorescent protein reporter assay using Ba/F3 cells. Data represent the fold NF-κB induction, calculated as mean fluorescence intensity (MFI) of stimulated cells divided by MFI of nonstimulated cells. Dotted lines indicates the fold induction as 1. Data shown are representative of three independent experiments.

To determine the functional importance of the specific residues recognizing the chemical ligands, we mutated these residues to Ala and examined the ability of the mutant proteins to activate nuclear factor–κB (NF-κB) (Fig. 3, D to F). The Phe405, Asp543, Tyr348, Val520, and Thr574 residues made contact with the ligand, and mutation of these residues to Ala either completely abolished or significantly weakened their ability to activate NF-κB. In contrast, mutation of other residues, including Arg429, Asp545, and Tyr353, resulted in either a retention or a slight reduction in NF-κB activity, which suggested that these residues are not essential for binding to the chemical ligands. NF-κB activity was notably up-regulated after mutation of Val378 to Ala, particularly in response to R848; therefore, Val378 may play a role in fine recognition or discrimination of nucleoside modifications.

Ligand binding opens up the TLR8 structure by 15 Å in the top lateral face of the LRRs, which enables the residues to recognize the ligand properly. Concomitantly, the dimerization interface is reorganized (Fig. 4). In the unliganded form, the interactions between LRR8 and LRR18*, as well as LRR11–13 and LRR14*–15*, are the major contributors to formation of the dimerization interface. Upon ligand binding, the dimerization interface is rearranged; LRR8 changes its binding partner to LRR18*–20*, LRR18* to LRR11, and LRR11–13 to LRR17*–18*. These rearranged interfaces include both the protein-protein and ligand-mediated interfaces. Concomitantly, Phe405 (within LRR13), which makes a hydrophobic contact with Phe494* (within LRR15*) in the unliganded form, in turn, interacts with the ligand in the liganded form (fig. S10). In addition, the ligand-induced conformational change enables formation of new interactions between LRR5 and LRR20*. As a result, the two C termini are brought into closer proximity, from 53 to 30 Å (CL097). The agonistic chemical ligand activates a preformed TLR8 dimer by inducing a conformational change in TLR8 ectodomains that, in turn, lead to an activated configuration of the cytoplasmic TIR signaling domains.

Fig. 4

Conformational change in TLR8 induced by binding to agonistic ligands. The unliganded, inactivated form of TLR8 (A) transforms into the activated form (B) upon ligand binding. The overall conformational change is illustrated schematically (C). The conformational changes are represented by ring rotations and hinge motions, both of which are indicated by gray arrows.

In conclusion, the TLR8 unliganded and liganded structures described in this study represent a milestone for understanding the signaling mechanism of the TLR7 to 9 family and will contribute to the development of therapeutic antiviral or modulating agents targeting the TLR7 to 9 subfamily.

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 to S13

Tables S1 to S3

References (2542)

References and Notes

  1. Acknowledgments: The NF-κB luciferase reporter plasmid, pELAM1-luc, containing the human E-selectin promoter region was kindly provided by T. Muta. We thank the beamline staffs at Photon Factory and SPring-8 for their assistance with data collection. This work was supported by funding from a Japanese Ministry of Education, Culture, Sports, Science and Technology Grant-in-Aid (U.O., K.M., and T.S.), Takeda Science Foundation, and Mochida Memorial Foundation for Medical and Pharmaceutical Research (U.O.). The data presented in this manuscript are tabulated in the main paper and the supplementary materials. The coordinate and structure factor data of human TLR8 (unliganded form), TLR8-CL097, TLR8-CL075, TLR8-R848 (form1), TLR8-R848 (form2), and TLR8-R848 (form3) have been deposited to the Protein Data Bank (PDB) under the PDB IDs 3W3G, 3W3J, 3W3K, 3W3L, 3W3M, and 3W3N, respectively. We all declare that none of the authors have a financial interest related to this work.
View Abstract

Navigate This Article