Structural basis for the multiple interactions of the MyD88 TIR domain in TLR4 signaling

Abstract

Myeloid differentiating factor 88 (MyD88) and MyD88 adaptor-like (Mal) are adaptor molecules critically involved in the Toll-like receptor (TLR) 4 signaling pathway. While Mal has been proposed to serve as a membrane-sorting adaptor, MyD88 mediates signal transduction from activated TLR4 to downstream components. The Toll/Interleukin-1 receptor (TIR) domain of MyD88 is responsible for sorting and signaling via direct or indirect TIR−TIR interactions between Mal and TLR4. However, the molecular mechanisms involved in multiple interactions of the TIR domain remain unclear. The present study describes the solution structure of the MyD88 TIR domain. Reporter gene assays revealed that 3 discrete surface sites in the TIR domain of MyD88 are important for TLR4 signaling. Two of these sites were shown to mediate direct binding to the TIR domain of Mal. Interestingly, Mal-TIR, but not MyD88-TIR, directly binds to the cytosolic TIR domain of TLR4. These observations suggested that the heteromeric assembly of TIR domains of the receptor and adaptors constitutes the initial step of TLR4 intracellular signal transduction.

Myeloid differentiating factor 88 (MyD88) is a cytosolic adaptor protein that plays essential roles in both innate and acquired immune responses by mediating signal transduction pathways that are initiated by Toll-like receptors (TLRs) and IL-1 and IL-18 receptors (IL-1R and IL-18R). MyD88 consists of an N-terminal death domain (DD) (approximately 90 aa residues), a C-terminal Toll/Interleukin-1 receptor (TIR) domain (approximately 150 aa residues), and a short connecting linker (1). In innate immune responses, the TIR domain of MyD88 has pivotal functions in the formation of signal initiation complexes involving the cytosolic domain of TLRs. The best characterized pathway is the TLR4 pathway, in which the cytosolic TIR domain of LPS-stimulated TLR4 interacts with the TIR domain of MyD88 (MyD88-TIR), in cooperation with another TIR-containing adaptor protein, MyD88 adaptor-like (Mal). Subsequently, signal is transmitted to the IL-1 receptor-associated kinase (IRAK) through an interaction between the death domains of MyD88 and IRAK. This eventually activates the transcription factors NF-κB and activator protein 1 (AP-1) via a phosphorylation cascade (2).

MyD88 has been reported to be involved in signaling pathways initiated by all TLRs thus far reported, with the exception of TLR3 (3). Of the MyD88-dependent pathways involving TLR2, 4, 5, 7, and 9, only the TLR2 and TLR4 pathways require Mal for efficient signal transduction (4). TLR4 also possesses the MyD88-independent signaling pathway, which comprises other TIR-containing cytosolic adaptors, TIR domain-containing adaptor inducing IFN-β (TRIF), TRIF-related adaptor molecule (TRAM), and sterile α and huntingtin-elongation-A subunit-TOR (HEAT) Armadillo motifs (SARM) (5). Therefore, in general, specific complexes involving more than one TIR-containing adaptor are likely to be required for initiation of each TLR signal transduction pathway.

Recently, Kagan and Medzhitov (6) revealed that MyD88 and Mal have distinctly different roles in TLR4 signaling: MyD88 serves as an essential “signaling adaptor,” which transmits signals from ligand-activated TLRs to downstream factors to initiate kinase-dependent signaling cascades, while Mal functions as a “sorting adaptor,” which recruits MyD88 to the plasma membrane via its PIP2 binding domain to promote interaction between MyD88 and activated TLR4 beneath the membrane. Indeed, Mal was shown to be dispensable for TLR4 signaling when MyD88 is fused to a PIP2 targeting domain. In the TLR4-TRIF pathway, TRAM has been proposed to serve as a sorting adaptor, which delivers TRIF to a specific membrane portion via its myristoylation site (7). These findings suggest that specific combinations of “sorting” and “signaling” TIR-containing adaptors might be involved in TLR signaling pathways.

The specificities of TIR–TIR interactions between adaptors, and between adaptors and TLRs, define the formation of various complexes that initiate TLR signaling pathways. However, little is known about the mechanism of heteromeric interactions between TIR domains. The crystal structures of cytosolic TIR domains of the membranous receptors, TLR1, TLR2, TLR10, and IL-1RAPL have been reported (8–10), and the homomeric TIR interfaces observed in the crystals have been described. However, the functional relevance of these homomeric interactions remains obscure because the formation of a homomeric dimer in these TIR domains has not been observed in solution (9, 10). Based on crystal structures and mutational data, several structural models have been proposed for heteromeric TIR–TIR interactions, which commonly suggest the importance of the so-called BB loop in these interactions (11, 12).

The present study describes the solution structure of MyD88-TIR using NMR spectroscopy. The isolated domain was shown to exist as a monomer in solution state on the basis of size-exclusion chromatography, although full-length MyD88 forms a dimer, which appears to be mediated via homomeric interactions within its death domain. By combining in vitro mutational binding experiments with an NF-κB reporter system in mammalian cells, 2 surface sites were identified as binding interfaces for the TIR domain of the sorting adaptor Mal, one of which includes a critical residue for MyD88 function, whose mutation causes the pyogenic bacterial infections (13). Furthermore, the in vitro binding experiments demonstrated that MyD88-TIR does not directly bind to the cytosolic TIR domain of TLR4, while Mal-TIR does. The distal location of the Mal binding sites on the MyD88-TIR surface suggests that the TIR domain of MyD88 simultaneously interacts with 2 Mal-TIR molecules, which may provide a highly efficient scaffold for signal transduction.

Results

Structure Determination of the TIR Domain of MyD88.

Based on sequence comparison between TIR domains, a region was selected that comprised residues 148–296 of human MyD88 and was used for structure determination by solution NMR spectroscopy (see Fig. S1). In the buffer used for structural studies, MyD88-TIR resided in a monomeric state, as assumed from size-exclusion chromatography. However, the death domain including internal domain (DD+ID) of MyD88 existed in a dimeric state (see Fig. S2). Therefore, the reported MyD88 dimerization was likely mediated by DD+ID but not by the TIR domain (14). The TIR domain structure of human MyD88 (residues 157–296), which presented the lowest overall energies in the 20 final structures generated by calculations, is shown in Fig. 1A. Statistics for the final 20 conformers are summarized in Table S1, which shows that the rmsd for the coordinates of backbone heavy atoms (N, Cα, and C') of residues 157–185, 188–194, and 203–295 is 0.45 Å. The N-terminal 9 residues (148–156) displayed random coil propensity, which was characterized by the lack of medium-to-long range NOESY cross-peaks for the region. Hence, these residues were omitted from the figure and statistics for clarity.

Fig. 1.

Solution structure of the TIR domain of human MyD88. (A) A representative ribbon drawing of the NMR structure of MyD88, generated with MOLMOL 2K.2. The notation of the secondary structures (βA-βE, αA-αC, and αE) is based on TLR TIR domains. (B and C) A superimposed representation of MyD88-TIR (sky blue) and the crystal structure of the TIR domain of TLR2 (orange). Although the β-sheet cores are similar (rmsd = 0.90 Å), there are differences in some regions.

Previous studies have indicated that 3 short, sequence motifs, called box 1–3 motifs, which are (F/Y)DA, RDXXPG, and FW, respectively, are conserved between TIR domains (see Fig. S1) (15). Of these, the box 2 motif, which resides in the so-called BB loop region, has been suggested to be important for TIR–TIR interactions and specificities (8, 16). In the calculated conformers of MyD88-TIR, a section of the BB loop, namely residues 194–208, was not well converged. With the exception of Gly-201, Val-204, and Ser-206, the backbone amide resonances of these residues were not identified in 2D ¹H-¹⁵N hetero-nuclear single quantum coherence (HSQC) spectra, although resonances of some side chain protons were observed and assigned in HCCH-TOCSY and 3D ¹³C-edited NOESY spectra. This region appeared to not form a single, definite structure, as judged from observations of a relatively few number of long-range NOEs. A {¹H}-¹⁵N heteronuclear NOE experiment showed that some main-chain amide groups in the BB loop region, namely Gly-201, Val-204, and Ser-206, displayed NOE values of 0.67, 0.55, and 0.67, respectively, which were less than the average value for residues 157–297 (0.76 ± 0.09). These results implied that the BB loop was mobile in solution. The presumed conformational flexibility in the BB loop region might exert broadening effects due to chemical exchange, which results in absence of the backbone amide resonances of residues 195–200, 202, and 203. In addition, the turn region comprising residues 185–188, which follows helix αA, was poorly defined as the main chain amide resonances of Thr-185, Asp-186, Tyr-187, and Arg-188 could not be identified in the HSQC spectrum.

During the preparation of this manuscript, Rossi et al. released the solution structure of the TIR domain of human MyD88 in the Protein Data Bank (PDB ID: 2JS7). The overall folding was identical to our findings, despite minor differences. However, a detailed comparison of these 2 structures would not be appropriate, because of substantial differences in solution conditions, such as organic additives and pH. The structure of Rossi et al. was determined in a buffer containing 5% acetonitrile at pH 5.0, while the present structure determination was performed in a buffer at pH 6, with no organic solvent.

Structural Description of the TIR Domain of MyD88, and Comparison with Other TIR Domains.

The MyD88 TIR domain structure (residues 157–296) comprised a central 5-stranded parallel β-sheet (βA–βE) surrounded by 4 α-helices (αA–αC and αE) (Fig. 1A). As predicted, the global fold was similar to what was observed in previously determined crystal structures of TIR domains of receptors TLR1, TLR2, TLR10, and IL-1RAPL (PDB codes: 1FYV, 1FYW, 2J67, and 1T3G, respectively). Of the known structures, the MyD88 TIR domain exhibited highest sequence similarity to TLR2. Fig. 1 B and C show superimposed representations of the TIR domain structures from MyD88 and TLR2. While the β-sheet cores displayed high structural similarity, as indicated by an rmsd value for backbone N, Cα, and C' of 0.90 Å, several regions displayed notable conformational differences. The largest structural discrepancy was observed in the region from the BB loop (Ser-194-Ala-208 of the MyD88 TIR domain) to αB (Fig. 1C). The BB loop was exposed to solvent in MyD88 and TLR2, but the direction in which the loops orient was markedly different. This was mainly due to a structural difference in the C-terminal region of the BB loop (residues 205–208), which precedes αB of MyD88. These residues adopted an extended conformation in MyD88, whereas corresponding residues are involved in an α-helix (αB) in TLR2. Therefore, αB of MyD88 was much shorter than TLR2. Another major conformational difference was the lack of an α-helix in the region between strands βD and βE (residues 257–273) of the MyD88 TIR domain (see Fig. S1). This region adopted an extended and a short helical coil conformation in MyD88, but the corresponding residues formed an α-helix (αD) in TLR2 (positions 266–270 in MyD88 numbering).

Cell-Based Functional Assays of the MyD88 TIR Domain in TLR4 Signaling Pathways.

To explore residues that are important for function of the MyD88 TIR domain, we performed mutational analysis of the domain, using dominant negative effects of ectopically expressed isolated TIR domain (17). For the assay, a luciferase reporter system for NF-κB activation was constructed in HEK293 cells, where MyD88-TIR or mutants harboring single amino acid substitutions of surface residues, was ectopically expressed. Expression of MyD88-TIR, which lacked the N-terminal death domain and thus was supposedly unable to transmit signals, suppressed LPS-induced luciferase expression. This effect was presumably due to inhibition of signaling pathways by competitive binding of the isolated TIR domain to signaling components that interact with endogenous MyD88. When a TIR mutant harboring a substitution of a functionally important residue is expressed, such suppression is alleviated, which leads to higher LPS-induced luciferase activity than observed with a TIR domain harboring the wild-type sequence. It should be noted that the dominant negative effect of ectopically expressed TIR domain has been used for functional analysis of some key residues of MyD88-TIR in IL-1 signaling (11).

Results from luciferase assays of mutant-expressing cells upon LPS stimulation are shown in Fig. 2A. LPS addition to HEK293 cells expressing MD2 and TLR4 resulted in an approximately 3-fold increase in reporter activity, which was consistent with a previous report (18). Alanine substitution of 5 residues, Arg-196, Asp-197, Arg-217, Lys-282, or Arg-288 resulted in significantly reduced inhibitory effects in LPS-induced luciferase activity. Interestingly, these 5 residues are closely associated with the box 1, 2, and 3 motifs, which are highly conserved across TIR domains (see Fig. S1). Arg-196 and Asp-197 are located within the box 2 motif. The side chains of Lys-282 and Arg-288 form a continuous protein surface with box 3 forming residues, and Arg-217 is located distant in the sequence, but proximal in space to the box 1 motif (Fig. 2B). Hence, we designated the sites that these residues form as Site II, Site III, and Site I, respectively.

Fig. 2.

Functional sites for signaling and binding sites for Mal. (A) The NF-κB reporter gene assay of MD2 cotransfected with LPS-induced (1.0 μg/ml) 293-hTLR4A-HA cells. In these graphs, each column indicates relative luciferase activity of stimulated cells over nonstimulated cells. Color code: black bars indicate a significant increased NF-κB activity, compared with the wild-type group. The statistical significance of differences in luciferase activities between wild type and mutants was analyzed using the Dunnett's multiple comparison test. Statistical significance was assumed to be P < 0.05. (B) Results of the functional assays of LPS/TLR4 signaling presented on the 3D structure of the TIR domain of MyD88. The results of the functional assays are mapped onto the molecular surface of the MyD88 TIR domain. The amino acid residues judged to be significant by the luciferase assay are shown in red, while nonsignificant residues are shown in light brown. The conserved motifs of boxes 1–3 (FDA of box1, VLPG of box2, FW of box3) are shown in blue. (C) Assay of binding of Mal TIR domain and MyD88 TIR domain wild type or mutants. Alanine substitutions in Site II (R196A) or Site III (R288A) in MyD88 resulted in reduced interaction with Mal. The double alanine substituted mutant of Site II and Site III caused complete abolition of the interaction with Mal. (D) Cysteine substitution in R196 also caused reduced interaction with Mal. (E) Plots of relative integrated area of NMR signals, which were derived from NMR titration data, in a function of added volume of Mal-TIR into ¹⁵N MyD88-TIR sample. The best-fit lines to the data, assuming a simple 1:1 complex model, are also shown. Apparent dissociation constants calculated are also indicated with standard errors (see Fig. S3). The black box, red circle, blue triangle, and green triangle indicate the dissociation curves of wild-type, R196A, R288A, and R196A-R288A, respectively.

Binding Sites of MyD88-TIR for Mal.

Because Site I, Site II, and Site III of the MyD88 TIR domain are important for TLR4-mediated cellular responses following LPS stimulation, the involvement of these sites was examined in direct binding to the TIR domain of Mal (Mal-TIR), the sorting adaptor in MyD88-dependent TLR4 pathways (19). The effect of alanine substitution of Arg-196, Arg-217, or Arg-288 (which forms Site II, Site I, and Site III of the MyD88 TIR domain, respectively) on interactions with Mal-TIR was analyzed by GST pull-down assay. Mal-TIR was pulled down by wild-type MyD88-TIR. However, substitution of either Arg-196 or Arg-288 resulted in moderate, but significant, decreases in MyD88-TIR affinity for Mal-TIR. The MyD88-TIR affinity of the Arg-196 and Arg-288 double-substituted mutant for Mal-TIR was completely abolished (Fig. 2C). In contrast, the Arg-217 mutation had no significant effect. The results, therefore, suggested that Site II and Site III, but not Site I, contributed to the interface with Mal-TIR. In addition, Arg-196 substitution by cysteine, which was detected in MyD88 deficiency patients (13), also caused reduced interaction with Mal (Fig. 2D). The effect of alanine substitution of those arginine residues on interactions with Mal-TIR was also examined by observing 2D ¹H-¹⁵N correlation NMR spectra of ¹⁵N-labeled MyD88-TIR and its derivative in the absence or presence of various concentrations of nonlabeled Mal-TIR. The ¹⁵N-labeled MyD88-TIR signals uniformly decreased upon titration of Mal-TIR (see Fig. S3). Signal attenuation was presumably due to increased apparent molecular weight upon complexation or chemical exchange (Fig. 2E). A MyD88-TIR mutation of either Arg-196 or Arg-288 caused moderate effects. However, double mutations of these residues resulted in large effects on signal attenuation, suggesting that signal attenuation was due to interactions between TIR domains of MyD88 and Mal. The apparent dissociation constants were estimated from the signal attenuation of wild-type and mutant MyD88 as previously described (Fig. 2E) (20). These results indicated that contributions from Site II and Site III to the interaction were comparable to each other. It should be noted that the effect of tested alanine substitution on the structure of the TIR domain was minor and only limited to the region close to the mutational sites, as judged from the 2D spectra of those mutants. Thus, these substitutions do not affect the opposite functional surface. Therefore, allosteric effect was neglected in interpreting the data.

The Function of Site I in TLR4 Signaling.

Site I could serve as an interaction site with cytoplasmic TIR domain of TLR4 (TLR4-TIR), as previously suggested (6). However, an interaction between TIR domains of MyD88 and TLR4 was not detected (Fig. 3A). The GST pull-down experiment in the present study further demonstrated that the DD of IRAK4 exhibited no detectable binding activity to MyD88-TIR, but rather bound to MyD88 that lacked the TIR domain (Fig. 3B). This indicated that Site I was not involved in interactions with the downstream effector IRAK4, although results from a previous study showed that a small region of TIR, which included box 1 motif, ID, and DD of MyD88, interacts with IRAK4 (21).

Fig. 3.

Direct interactions of MyD88, Mal, TLR4, and IRAK4. (A) Binding assay of the TLR4 TIR domain with the MyD88 TIR domain or Mal TIR domain. GST-TLR4-TIR binds Mal-TIR but not MyD88-TIR. (B) Binding assay of the IRAK4 death domain with the MyD88 TIR domain or death domain. IRAK4-DD binds the MyD88-DD+ID but not MyD88-TIR.

Discussion

The TIR domain is typically composed of 135–160 aa residues, with sequence conservation ranging from 20 to 30%. While the hydrophobic core residues are conserved, the surface exposed residues vary greatly between TIR domains. Consequently, the distribution of surface electrostatic potential differs significantly between TIR domains (22), possibly underlying the differences in binding specificity. For TIR domains of membranous receptors, 4 structures from TLR1, TLR2, TLR10, and IL-1RAPL have been reported (8–10). When comparing these, the BB loop of MyD88 displayed the largest structural difference (Fig. 1C). The BB loop region has been proposed to be important for interactions between TIR domains. Thus, structural deviation of the BB loop, and differences in surface electrostatic potential, might reflect specificities of TIR–TIR interactions. It should be noted that the isolated MyD88-TIR domain existed as a monomer in solution state, while some TIR domains have been reported to form a dimer in crystal structure (9, 10). On the other hand, full-length MyD88 is known to form a dimer, which seems to be mediated via homomeric interactions between the death domains (see Fig. S2).

The present study identified 3 functional surface sites (Sites I–III) of MyD88-TIR that were important for the LPS-activated TLR4-signaling pathway. Two of these sites, Sites II and III, served as binding sites for Mal-TIR (Fig. 2 C and E). Results from the GST pull-down and NMR titration experiments suggested that these 2 sites equally contributed to interactions between MyD88-TIR and Mal-TIR. The Site II-forming residues Arg-196 and Asp-197 were located in the BB loop, and were highly conserved across TIR domains (see Fig. S1). Another Mal binding site, Site III, which was formed by 2 basic residues, Lys-282 and Arg-288, flanks the box 3 comprised of FW motif, creating a positively charged surface patch. Because basic amino acids were conserved at positions 282 and 288 (see Fig. S1), this positively charged patch appeared to be common in TIR domains. The present data revealed that Arg-217 in Site I played a crucial role in the TLR4-mediated cellular response to LPS stimulation but was not involved in direct binding to Mal-TIR (Fig. 2C). In the GST pull-down experiments, direct interaction of MyD88-TIR with either the TIR domain of TLR4 or the death domain of IRAK4 was not observed (Fig. 3). This was consistent with previous observations that MyD88 does not directly bind to the cytosolic domain of TLR4 (23). Therefore, Site I is unlikely involved in MyD88 interaction with any of the known possible binding partners, such as Mal, TLR4, and IRAK4, but might serve as a contact surface with a yet unidentified MyD88 binding protein or specific membrane portion. The functional role of Site I in TLR4 signaling remains to be clarified in further studies.

The 2 Mal binding sites of MyD88-TIR, Sites II and III, are distantly located from each other and are on opposite molecular surfaces. Thus, it is impossible to assume that one Mal-TIR can make simultaneous contact with both MyD88-TIR sites. In addition, contributions from Site II and Site III to the interaction were shown to be comparable to each other. Thus, assuming 2 Mal-TIR molecules would bind to one MyD88-TIR molecule, we constructed a complex model between MyD88-TIR and Mal-TIR using a molecular docking method similar to previous studies (12, 24). The model indicated that Sites II and III residues are well situated at the interface centers of each Mal-TIR. In addition, all noncritical residues, including Arg-217 (Site I), avoided the interfaces, which supported validity of this model (Fig. 4). Previous reports have shown that P125H mutation or S180L polymorphism of Mal causes decreased interactions between Mal and TLR4 or TLR2, respectively, but has no effect on interactions between Mal and MyD88 (25, 26). Moreover, the TRAF-6 binding site on Mal was shown to include Glu-190 (27). These 3 Mal residues were not included in the Mal-MyD88 complex model interfaces (Fig. 4). Therefore, the model suggested that MyD88-TIR binding might not interfere with interactions between Mal-TIR and TLR4, TLR2, or TRAF-6. Recently, Tyr-86 phosphorylation of Mal was shown to negatively regulate interactions with MyD88 (28). In addition, Tyr-86 mutation, not Tyr-106 or Tyr-159, significantly altered affinity of Mal to MyD88. These observations were consistent with the present complex model, in which Tyr-86, not others, was at the molecular interface (Fig. 4). Moreover, the model predicted that Tyr-86 phosphorylation might perturb MyD88 interface steric complementarity of Mal and result in an electrostatic repulsion to the acidic surface of MyD88-TIR that Mal binds to.

Fig. 4.

Structural model of signaling complex formed by MyD88 and Mal. The Mal binding sites, Sites II and III residues, are shown in red, and noncritical residues for signaling and Site I residue are shown in light blue. The positions of the previously reported functional residues (P125, S180, and E190) are shown as orange spheres, and the phosphorylation sites in Mal for signaling, Tyr-86, Tyr 109, and Tyr-159, are shown as yellow spheres.

Recently, one of the Site II residues, Arg-196, has just been found to be mutated to cysteine in the new primary immuno-deficiency (MyD88 deficiency) patients (13). This mutation did not cause destabilization of the MyD88 protein, but showed a significant decrease of the direct binding ability between MyD88-TIR and Mal-TIR (Fig. 2D). Patients with the MyD88 deficiency were highly susceptible to Gram-positive bacteria, while they showed normal resistance to other kinds of pathogens, such as Gram-negative bacteria and viruses. The phenotypes suggest that TLR2 signaling is more critical than other self-defense systems in early life (3) because the other innate immune signaling pathways have a kind of redundancy, acting as alternative signaling pathways; i.e., TLR2 and TLR4 signaling needs MyD88 and Mal to signal, but TLR4 has other MyD88-independent pathways with TRIF (5). It suggests that the Gram-positive bacterial recognition system is much more dependent on TLR2/MyD88/Mal signaling. The loss of interaction between MyD88 and Mal caused by the mutation would be a critical molecular mechanism for MyD88 deficiency patients.

It is of special interest that the TIR domain of Mal, and not MyD88, directly interacted with the TIR domain of TLR4 (Fig. 3A) as described in the previous predicted docking model (29). This observation raises the possibility that Mal-TIR might simultaneously bind to the TIR domains of TLR4 and MyD88, and thereby mediate association as previously predicted (5). Because Mal has been shown to be dispensable for TLR4 signaling when MyD88 is artificially fused to a PIP2 targeting domain (6) there is the possibility that weak interactions between TIR domains of MyD88 and TLR4 mediates signal transduction. Alternatively, an unidentified alternate Mal-independent pathway could contribute to signaling as previously discussed (30). Previous mutation analysis suggests that Mal Pro-125 contributes a binding interface with TLR4 (26). This residue was located distal to the putative MyD88 interface in the present MyD88:Mal complex model and was therefore consistent with the hypothesis that Mal-TIR mediates TIR–TIR interactions between TLR4 and MyD88.

Conclusion

Structure determination combined with functional assays of human MyD88-TIR revealed that 3 sites, which are related to conserved boxes 1–3 of the domain, were important for the LPS/TLR4 pathway. Two of these sites were located at opposite surfaces of the molecule and were shown to mediate direct interaction with Mal-TIR. Thus, the 2 independent binding sites served by MyD88-TIR might contribute to formation of higher order TIR–TIR complexes, which may result in amplification of TLR signal activation. Identification of the key residue in MyD88, which is a direct interacting residue for Mal, is of the clinical significance because one of these residues was shown to be critical for the primary immunodeficiency syndrome. Distribution of the 3 functional sites dispersed on the molecular surface of MyD88-TIR suggested that MyD88 provided multiple interaction surfaces to protein factors that form the signal initiation complex at the cytosolic TLR4 domain. Knowledge of the sites revealed in this study will facilitate further identification of factors and mechanisms used in TLR signaling pathways.

Materials and Methods

Sample Preparation.

The portion of the human MyD88 gene encoding the TIR domain (amino acid residues 148–296) was cloned into the vector pGEX-5X-3 (GE Healthcare). This vector was transformed into Escherichia coli BL-21 (DE3) (Novagen). The TIR domain of MyD88, which was expressed as a GST (GST) fusion protein, was first purified by glutathione Sepharose 4B FF (GE Healthcare) affinity chromatography, and the GST-tag was removed by digestion with Factor Xa (GE Healthcare). Subsequently, the TIR domain was purified by gel filtration (Sephacryl S-100 HR 26/60 column; GE Healthcare) and cation-exchange chromatography (Mono-S column; GE Healthcare). Using the purification protocol, ¹⁵N-labeled, and ¹³C, ¹⁵N-doubly-labeled monomeric TIR domain of MyD88 wild-type proteins were prepared. The protein sample buffer was replaced by 20 mM potassium phosphate buffer (pH 6.0) containing 0.1 mM EDTA and 10 mM DTT. The final protein sample concentration for typical NMR experiments was approximately 0.3 mM.

NMR Spectroscopy.

All NMR spectra were recorded at 25 °C on a Bruker DRX500 or DRX800 spectrometer equipped with a cryogenic probe. For assignment of backbone and side chain ¹H, ¹³C, and ¹⁵N resonances, a series of triple-resonance experiments were conducted (31). Distance restraints for structure calculations were obtained from 3D ¹⁵N-edited NOESY and 3D ¹³C-edited NOESY experiments, with a mixing time of 150 msec. NMR spectra were processed with NMRPipe software (32) and analyzed using Sparky (33). The pulse sequence used to obtain 2D {¹H}-¹⁵N steady-state NOE spectra has been previously described (34). The {¹H}-¹⁵N NOE values were determined from ratios of peak intensities with or without a 3 sec ¹H-saturation applied before each scan: NOE = I_sat/I_unsat.

Structure Calculation.

Automated NOESY cross-peak assignment and iterative structure calculation were performed using CYANA version 2.1 (35). The obtained assignment of NOESY cross-peaks was manually validated, and the final structure calculation was performed using CNS version 1.1 (36). Surface electrostatic potentials were calculated using MOLMOL 2K.2 (37).

Cell Culture.

Human embryonic kidney (HEK) 293-hTLR4A-HA cells were purchased from Invivogen. These cells were cultured in Dulbecco's modified Eagle's medium (high glucose-containing DMEM, Invitrogen) supplemented with 10% heat-inactivated FBS (Sigma), penicillin (100 U/ml), and streptomycin (100 pg/ml). All cells were incubated at 37 °C in a humidified atmosphere of 5% CO₂.

Vector Preparations.

A cDNA encoding the TIR domain (amino acid residues 148–296) that was tagged at the N terminus with a myc-epitope was cloned into the plasmid vector pcDNA3.1+ (Invitrogen). Mutants of the MyD88 TIR domain were generated using the GeneEditor in vitro Site-Directed Mutagenesis System (Promega). Mutants of each of 25 charged polar amino acid residues (Asp, Glu, Arg, Lys, and His) substituted by alanine were generated. Mutants with poor expression were not included to avoid possible misinterpretation of the loss of dominant negative inhibitory effect. The MD2 construct was also cloned into pcDNA3.1+. A pGL3-Basic Vector (Promega) containing 4 kB binding sites, which was used in the NF-κB luciferase reporter assay, and a Renilla luciferase reporter vector used as an internal control in the assay were gifts from Drs. Sewon Ki and Tetsuro Kokubo (Yokohama City University, Yokohama, Japan).

NF-κB Reporter Gene Activity.

293-hTLR4A-HA cells were transfected with pcDNA3.1+ control vector or pcDNA3.1+ myc-MyD88 TIR domain (wild type or mutant), pcDNA3.1+ MD2, NF-κB luciferase reporter vector, and Renilla luciferase reporter vector, using Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions. After transfection, the cells were stimulated with LPS O127 (1.0 μg/ml, Sigma) and incubated for 6 hours. Luciferase reporter gene activity was analyzed using the Dual-Luciferase Reporter Assay System (Promega). The inhibitory effect of each TIR mutant expression was assessed in at least 3 independent experiments. The statistical significance of differences in luciferase activities between wild type and mutants in the NF-κB reporter assays was analyzed using Dunnett's multiple comparison test. Statistical significance was assumed to be P < 0.05.

GST Pull-Down Assay.

The TIR domain of MyD88 wild type and mutants (R196A, R196C, R217A, R288A, and R196A-R288A) was purified as GST-fusion proteins. These expression vectors were generated by subcloning the pcDNA3.1+myc-tagged MyD88 TIR domain into pGEX 5X-1 (GE Healthcare). The DD+ID of MyD88 (amino acid residues 18–141) and TLR4-TIR were also purified as GST-fusion proteins. The GST-fusion proteins were purified by glutathione Sepharose 4B FF (GE Healthcare) affinity chromatography. The TIR domain of human Mal, as well as the DD of IRAK4, was purified using a modified previously reported method (22, 38). These purified proteins were incubated with Glutathione Sepharose 4B (GE Healthcare) for 3 hours. After 4 wash steps with wash buffer (20 mM potassium phosphate buffer (pH 6.0), 100 mM KCl, 0.1 mM EDTA, 10 mM DTT, and 0.5% Triton X-100), the resin was analyzed by SDS polyacrylamide gel electrophoresis and Coomassie Brilliant Blue staining.

NMR Titration.

One 15-μl aliquot of 100 μM nonlabeled Mal-TIR was added to 150 μl of 20 μM ¹⁵N-labeled MyD88-TIR or its alanine substituted mutants up to 1.5 molar equivalent of ¹⁵N MyD88-TIR. At each titration point, 1D ¹H-{¹⁵N} and 2D ¹H-¹⁵N SOFAST-HMQC spectra were measured. Quantification of NMR signal attenuation in the titration experiments, and evaluation of apparent dissociation constant (K_d^app) for the interaction, are described in SI Materials and Methods.

Docking Studies Between MyD88 and Mal.

Structure modeling of the TIR domain of Mal was performed using the MyD88-TIR structure as a template on molecular operating environment (MOE) software (39, 40). The docking simulation was performed on AutoDock without any specific restraints between the molecules as previously reported (24, 41, 42). (See detailed method of the docking study in SI Materials and Methods.)