Interactive Sites in the MyD88 Toll/Interleukin (IL) 1 Receptor Domain Responsible for Coupling to the IL1 (cid:1) Signaling Pathway*

Myeloid differentiation factor MyD88 is the essential adaptor protein that integrates and transduces intracellular signals generated by multiple Toll-like receptors including receptor complex for interleukin (IL) 1 (cid:1) , a key inflammatory cytokine. IL1 (cid:1) receptor complex interacts with MyD88 via the Toll/IL1 receptor (TIR) domain. Here we report structure-function studies that help define the MyD88 TIR domain binding sites involved in IL1 (cid:1) -in-duced protein-protein interactions. The MyD88 TIR domain, employed as a dominant negative inhibitor of IL1 (cid:1) signaling to screen MyD88 TIR mutants, lost its suppress-ing activity upon truncation of its Box 3. Accordingly, mutations of Box 3 residues 285–286 reversed the dominant negative effect of the MyD88 TIR domain on IL1 (cid:1) induced and NF (cid:2) B-dependent reporter gene activity and IL6 production. Moreover, mutations of residues 171 in helix (cid:3) A, 195–197 in Box 2, and 275 in (cid:1) E-strand had similar functional effects. Strikingly, only mutations of residues 195–197 eliminated the TIR-TIR interaction of MyD88 and IL1 receptor accessory protein (IL1RAcP), whereas substitution of neighboring canonical Pro 200 by His was without effect. Mutations in Box 2 and 3 pre-vented homotypic MyD88 oligomerization via TIR domain. Based on this structure-function

Consistent with these studies, IL1␤ is one of the most potent inflammatory cytokines responsible for fever, leukocytosis, thrombocytosis, and production of IL6 and other cytokines (1)(2)(3). The signals generated by IL1␤ binding to its cognate receptor complex, formed by two type I transmembrane proteins, IL1 receptor I (IL1RI) and IL1 receptor accessory protein (IL1RAcP), are transduced by their cytoplasmic segments denoted as the Toll/IL1 receptor (TIR) domain. TIR domain is shared with Drosophila Toll, mammalian TLRs, and cytoplasmic adaptors exemplified by MyD88 (6).
MyD88 adaptor integrates signals flowing from IL1 receptor/ IL1RAcP and from an array of other TLRs (6,7). This initial IL1 receptor-MyD88 adaptor interaction evoked by IL1␤ is a critical step in its signaling to the nucleus and, therefore, represents a potential target for new anti-inflammatory agents. MyD88 has a bipartite structure composed of an aminoterminal Death domain and a carboxyl-terminal TIR domain with a short intervening linker segment (6). Upon IL1␤ stimulation, IL1RI⅐IL1RAcP complex recruits MyD88 via its TIR domain (8). In addition, IL1RI-associated kinases are recruited to an IL1RI⅐IL1RAcP complex including IRAK (9, 10), IRAK-2 (11), IRAK-4 (12) and IRAK-M (13). Our current understanding of IL1 receptor complex-MyD88 adaptor interaction is limited.
Here we report studies that help to establish the molecular determinants of MyD88 TIR domain interactions in IL1␤ signaling pathway.
In terms of its structural features, the MyD88 TIR domain contains three highly conserved motifs denoted Box 1, 2, and 3 (Figs. 1 and 2). Box 2 forms a loop denoted the BB loop that contains an invariant proline residue at position 200, which, in other receptors and adaptors (namely, TLR2, TLR4, IL1RAcP, and MAL/TIRAP), is essential for their signaling function (6,7). For example, mutating this residue to histidine in TLR4 renders C3H/HeJ mice hyporesponsive to lipopolysaccharide (LPS) (14). This canonical example indicates that conserved structural motifs in TIR domain of TLRs and their adaptors play a highly significant role in proinflammatory ligand-initiated intracellular interactions between TLRs and their adaptors. Depending on the recognition of distinct ligands by TLRs, the preferential usage of its adaptors may require different interacting sites in TIR domain of the same adaptor or an alternative adaptor. The latter applies to TLR3, which requires its adaptor, TRIF, rather than MyD88 for signaling by viral double-stranded RNA. Conversely, TRIF mediates signaling induced by interaction of LPS with TLR4 in the absence of MyD88 (7). We hypothesized that signaling evoked by IL1␤ through its cognate receptor complex may depend on different interactive sites on TIR domain of MyD88 than the recently reported sites involved in MyD88 interaction with TIR domains of TLR2 and TLR4 (15).
To test this hypothesis, we undertook our studies focused on the potential role of Box 1, 2, and 3 of MyD88 TIR domain in * This work was supported in part by United States Public Health Service National Institutes of Health Grants HL69542, HL62356, and HL68744. The use of core facilities in this study was supported by National Institutes of Health Grants 2P30CA68485 to the Vanderbilt Ingram Cancer Center and 5P30DK058404-03 to the Vanderbilt Digestive Disease Research Center. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
‡ IL1␤ signaling. Within these three boxes, we focused on residues representing AA loop, BB loop, and EE loop that were reported to participate in interactions of MyD88 with TLR2 and TLR4 (15). The functional consequences of this mutational analysis were monitored by NFB-dependent reporter gene activity and by IL1␤-induced expression of the endogenous gene that encodes inflammatory cytokine IL6. Our structurefunction studies of MyD88 TIR domain led to the development of a three-dimensional docking model of MyD88 and IL1RAcP interaction mediated by their respective TIR domains.
Plasmids and Reagents-The NFB-luciferase reporter construct (NFB-luc) containing five B elements was provided by Dean Ballard (Vanderbilt University, Nashville, TN). The Renilla thymidine kinase luciferase reporter construct (RL-TK luc) was purchased from Promega. The AU1-tagged MyD88-expressing plasmid was a gift from Marta Muzio (Mario Negri Institute, Milan, Italy) (16). All MyD88-TIR constructs were cloned into pcDNA3.1. An AU1 tag or a Myc tag was introduced at the amino terminus of MyD88 or MyD88-TIR by PCR. IL1RAcP with a Myc tag at the amino terminus was cloned by reverse transcription-PCR into pcDNA3.1. All constructs were verified by sequencing.
Mutagenesis-The mutated MyD88 and MyD88 TIR domain sequences were generated using an in vitro site-directed PCR mutagenesis method and subcloned into plasmid pcDNA3.1 (Invitrogen) as described previously (17). Briefly, PCR was utilized with a supercoiled double-stranded DNA template and two synthetic complementary oligonucleotides containing the desired mutation and followed by removal of methylated parental DNA template with DpnI. The nicked DNA containing the desired mutations was transformed into the DH5␣ strain of competent Escherichia coli. All the mutants were confirmed by DNA sequencing and subsequently tested in transiently transfected HEK 293T cells.
Transient Transfection of HEK 293T Cells and NFB Reporter Gene Activity-The cDNAs for all MyD88 and MyD88 TIR domain mutants were inserted into the pcDNA3.1 vector that drives transcription from a cytomegalovirus promoter enhancer and contains an AU1 epitope tag for immunodetection of transiently expressed mutants in HEK 293T cells. Transfection of HEK 293T cells was performed with the indicated cDNAs by a conventional calcium phosphate method. One day before transfection, cells were seeded at a density of 2.5 ϫ 10 5 ml Ϫ1 /100-mm plate. After 18 h, the culture medium was replaced with fresh medium, and after 24 h, the cells were treated as indicated, harvested, and submitted to subsequent analysis. When the cells were stimulated with IL1␤, the culture medium was replaced by fresh medium containing IL1␤ (CellSciences, Canton, MA) at 10 ng/ml or as indicated otherwise and either further cultured for 6 h or left untreated for the indicated period of time. For NFB reporter gene activity assay, HEK 293T cells co-transfected with NFB-luc and RL-TK luc plasmids were harvested and submitted to subsequent dual-luciferase assay according to manufacturer's protocol (Promega).
Western Blotting and Indirect Immunofluorescence-To check the protein expression after transient transfection, HEK 293T cells were seeded (2.5 ϫ 10 5 ml Ϫ1 ) onto 100-mm dishes 24 h prior to transfection with combinations of plasmids (20 g, total) or as indicated, using calcium phosphate method. Thirty-six hours after transfection, cells were washed by the addition of 5 ml of ice-cold phosphate-buffered saline. Cells were lysed on ice for 10 min in lysis buffer containing 150 mM NaCl, 2 mM EDTA, 10% glycerol, 0.1% Nonidet P-40, 0.2 mM phenylmethylsulfonyl fluoride, 0.2 mM Na 3 VO 4 , and 1 g/ml Ϫ1 leupeptin. Cell lysate proteins (50 g) were separated by SDS-PAGE and then analyzed by Western blotting. Monoclonal antibodies against the epitope tags c-Myc and AU1 were obtained from Covance Company (Princeton, NJ). For indirect immunofluorescence, 10 5 transfected HEK 293T cells were cytocentrifuged onto a glass slide and fixed with 3.5% paraformaldehyde. After washing with phosphate-buffered saline, cells were permeabilized with 0.25% Triton X-100 for 10 min and then probed with anti-AU1 antibody followed by Rhodamine Red-X-labeled goat anti-mouse IgG antibody (Jackson ImmunoResearch Lab, West Grove, PA) as described previously (18). Slides with stained cells were mounted in Poly/Mount (Polysciences, Warrington, PA) and analyzed in an Olympus fluorescence microscope using a ϫ100 oil immersion lens.
Measurement of IL6 Expression Using Cytometric Bead Array Assay-One million MRC-5 cells were transfected with Cell Line Nucleofector Kit R (Amaxa, Gaithersburg, MD) program U23, following the manufacturer's protocol. After 8 h, cells were washed with Hanks' balanced salt solution, and fresh media were added. Cells were stimulated with IL1␤ for 6 h, and culture medium was collected and analyzed for production of cytokine IL6. Analysis of IL1␤-induced expression of cytokine IL6 in human fibroblast MRC-5 cells was performed using the Human Inflammation Kit (BD Biosciences) according to the manufacturer's protocol.
Immunoprecipitation-HEK 293T cells were plated at the density of 5 ϫ 10 6 cells/100-mm plate. Twenty-four hours later, cells were transfected with either (a) 10 g of Myc-IL1RAcP and 10 g of AU1-MyD88 or MyD88 mutants or (b) 10 g of Myc-MyD88 and 10 g of AU1-TIR or TIR mutants using calcium phosphate method. Fresh medium was added 18 h later, and cells were incubated for 24 h. In co-immunoprecipitation of IL1RAcP and MyD88, the cells were treated with 100 ng/ml IL1␤ for 5 min. The cells were washed with phosphate-buffered saline and resuspended in 400 l of hypotonic gentle lysis buffer (10 mM Tris-HCl, pH 7.5, 10 mM NaCl, 10 mM EDTA, 0.5% Triton X-100, 1 mM phenylmethylsulfonyl fluoride, 1 M aprotinin, and 1 M leupeptin). Lysates were initially precleared with normal mouse serum in combination with protein A-Sepharose 4 fast flow beads (Amersham Biosciences), and then 200 g of lysate was incubated with 10 g of anti-Myc at 4°C overnight. Protein A-Sepharose 4 fast flow slurry (10 l) was added and incubated for 3 h at 4°C. Beads were washed eight times with 0.5 ml of wash buffer (10 mM Tris-HCl, pH 7.5, 150 mM NaCl, 10 mM EDTA, 0.5% Triton X-100, 1 mM phenylmethylsulfonyl fluoride, 1 M aprotinin, and 1 M leupeptin). Beads were resuspended in 20 l of 2ϫ Laemmli buffer. The samples were fractionated on 15% SDS-PAGE, transferred to a polyvinylidene difluoride membrane, and subjected to immunoblotting analysis with anti-AU1 antibody. As a control, 50 g of total cell lysate was applied to Western blot with both anti-c-Myc and anti-AU1 antibodies.
Modeling Studies-The three-dimensional models of MyD88 TIR and IL1RAcP TIR domains were established by a comparative (homology) computation modeling method using Swiss-Model: An Automated Comparative Protein Modeling Server (swissmodel.expasy.org//SWISS-MODEL.html) (19,20). At least six Protein Data Bank records are available within known three-dimensional structure of TIR domain from highly homologous members of human TLR family. TIR domain of TLR2 with 31% alignment identity and Ͼ50% similarity (21) was used for comparative modeling of MyD88 TIR domain (Protein Data Bank code 1O77). TIR domain of TLR1 with 33% identity and 66% similarity (22) was used to generate the three-dimensional model of IL1RAcP (Protein Data Bank code 1FYV). All modeling processes were done in four steps: template selection (BLASTP2), target-template alignment (SIM), model building (ProModII), and energy minimization (GRO-MOS96). Templates, selected by sequence identity with the target sequence, were then prepared to create the core of the model by averaging of the backbone atom positions. Atoms with significantly deviating position were excluded from modeling process. The Constraint Space Programming (CSP) protocol was used for generation of insertion coordinates. The best loop was selected using a score scheme including force field energy, steric hindrance, and specific or nonspecific interaction, i.e. hydrogen bond formation or dipole-dipole interaction. Side chains were reconstructed by weighting positions of corresponding residues in the template structure and their iso-steric replacement. The final model was optimized by steepest descent energy minimization with GRO-MOS96. The visualizations of the three-dimensional model were performed with the DeepView (Swiss-PdbViewer) freeware program, available for download from the Swiss-Model webpage.
The Three-dimensional Docking Model of MD88 TIR and IL1RAcP TIR Interaction-Many aspects, such as molecular surface, geometry, surface topology, charge distribution, electrostatic field, and residue localization, were considered before an arrangement was constructed. First, a possible position of two interacting TIR domains was prepared manually in stereoviewed mode with the PSSHOW program (SYBYL-Tripos package). When the proper position was chosen, computation was conducted by merging receptor (IL1RAcP-TIR) into adaptor (MyD88-TIR), coordinates of the backbone atoms were then frozen, and the heavy atom aggregate was created and optimized. In the next step, hydrogen atoms were added, and the final model was optimized by energy minimization followed by molecular dynamics. Both steps were performed with the SANDER program (AMBER software package) using integral Newtonian equation of motion with 2000 cycles each. Once the computation process was completed, the coordinates were transformed into a Protein Data Bank file. The three-dimensional docking models of dot-surface and contact surface were obtained with PSS-HOW software. The ribbon structure of docking model was prepared with DeepView (Swiss-PdbViewer).
Statistical Analysis-Statistical differences between mean values were analyzed using the two-sided Student's t test.

RESULTS AND DISCUSSION
The IL1␤ signaling pathway depends on an orchestrated interplay of intracellular protein-protein interactions (1-3). MyD88 plays a pivotal role in these interactions by directing the flow of signals from IL1␤-occupied cognate receptor complex to downstream signal transducers (2,3). Within MyD88, the TIR domain provides an interacting surface for heterotypic interaction with the TIR domain of IL1RAcP (3). Therefore, we embarked on structure-function analysis of MyD88 TIR domain that is essential for the transduction of IL1␤ signaling to downstream effector(s). Our stepwise strategy consisted of analysis of the secondary and tertiary structure of MyD88 TIR domain. These data were obtained on the basis of the available crystal structure of TLR2 TIR domain (21). Drawing from these modeling studies, two series of mutagenesis experiments were carried out. In the first set of experiments, construct containing only TIR domain of MyD88, as its dominant negative inhibitor, was mutated, and the expressed mutants were screened for their inhibitory effect on IL1␤-induced signaling to the nucleus. In the second set of experiments, selected TIR residues were mutated in full-length MyD88 to assess the impact of specific replacements on heterotypic interaction of MyD88 with IL1RAcP and homotypic MyD88 oligomerization mediated by its TIR domain. The results of this structure-function analysis led us to the development of the three-dimensional docking model of MyD88 interaction with IL1RAcP mediated by their respective TIR domains.
Structural Characterization of MyD88 TIR Domain-MyD88 TIR domain is modeled on the basis of the crystal structure solved for TLR2 (21). It has an ␣-␤-fold similar to that of the bacterial chemotaxis protein CheY and contains three highly conserved motifs termed Boxes 1, 2, and 3 (3). On the basis of primary structure alignment obtained with the program T-COFFEE, presented in Fig. 1, the crystal structure of human TLR2 TIR domain (Protein Data Bank code 1O77) was selected as the best template for MyD88 TIR domain with 31% identity and Ͼ50% similarity. Results obtained from the server were visualized using the DeepView program (Swiss-PdbViewer). As shown in Fig. 2, the secondary structure of TIR domain consists of five ␤-strands (␤A, ␤B, ␤C, ␤D, and ␤E) forming the core of molecule that is surrounded by five ␣-helices (␣A, ␣B, ␣C, ␣D, and ␣E). The five loops (AA, BB, CC, DD, and FF) form specific links between ␤-strand and corresponding ␣-helix.
In terms of its tertiary structure, as depicted in Fig. 2, MyD88 TIR domain has a globular shape. Among three highly conserved motifs, Box 1, located at the amino terminus of TIR domain, forms a part of ␤A-strand. Box 2 makes the second part of the BB loop, whereas Box 3 creates the first part of the ␣E-helix, which is located at the carboxyl terminus of MyD88 TIR domain. As shown in Fig. 2B, distribution of charged residues indicates that the molecular surface of MyD88 TIR domain is mostly positively charged (blue), with a few distinct negatively charged knobs (red). They surround a larger swatch of negatively charged surface (red), formed by three loops (AA, BB, and part of DD) and ␣C-helix. Moreover, the BB loop projects from the globular TIR domain, forming a quasi-plane on its surface.
NFB Reporter Gene Activity Assay Indicates that Box 3 of MyD88 TIR Domain Is Involved in IL1␤-induced Signaling-NFB reporter gene activity assay was used to test MyD88 TIR domain as a dominant negative inhibitor of IL1␤-induced signaling to the nucleus. As documented in Fig. 3, HEK 293T cells transfected with plasmid containing NFB-dependent luciferase gene responded to sub-nanogram doses of IL1␤. This activation reached the maximum at 1 ng/ml IL1␤, attesting to the high sensitivity of transfected HEK 293T cells to this inflammatory cytokine. Consistent with prior studies (23,24), the MyD88 TIR domain, used as a dominant negative inhibitor of IL1␤ signaling, almost completely suppressed its activating effect on NFB reporter gene over a wide range of IL1␤ concentrations (Fig. 3). We engineered a deletion mutant of MyD88 TIR domain to establish the utility of the NFB reporter gene activity assay for screening MyD88 TIR domain mutants for their inhibitory effect on IL1␤-induced signaling. The deleted segment encompassed Box 3 ( 282 KSWF-WTRLAK 291 ) located at the carboxyl terminus of MyD88 TIR domain. This deletion caused the loss of the dominant negative inhibitory function of MyD88 TIR (Fig. 3B), suggesting that Box 3 was essential for IL1␤-induced signaling. The carboxylterminal deletion mutant was expressed in transfected cells at a level comparable with that of the intact MyD88 TIR domain (see Fig. 3B, inset).
Mutagenesis of MyD88 TIR Domain: Loss of Its Dominant Negative Inhibitory Activity toward IL1␤-induced Signaling-The involvement of Box 3, as compared with Boxes 1 and 2, in signaling induced by IL1␤ and mediated by the MyD88 TIR domain was analyzed in the first series of mutagenesis experiments. Mutations of TIR domain (residues 152-296) included two bulky hydrophobic residues (Phe 285 and Trp 286 ) in a highly conserved short motif in Box 3 (SWFWTRL) and proline at position 200. The canonical P712H mutation in TLR4 renders

MyD88 Structure and IL1␤ Signaling
C3H/HeJ mice hyporesponsive to LPS (14). This highly conserved proline residue is located at position 200 in the MyD88 TIR domain (15). It was mutated to histidine in MyD88-TIR domain (residues 152-296 (P200H)). Furthermore, alignment of members of the TLR family reveals two short motifs in Box 1 (PERFDAF) and Box 2 (DRDVLPG) that are conserved along with Box 3 motif in MyD88 TIR domain. Alanine substitutions were primarily based on selection of charged residues either in the conserved region or predicted to be on the surface. In addition, two mutations, Val 204 and Gln 229 , were in the region predicted to interact with MAL (15). The expression of mutants varied as compared with that of wild-type MyD88 TIR domain (data not shown). Five mutants listed in Table I were expressed at a level comparable with that of wild-type MyD88 TIR domain (see Fig. 4B, inset). The expression of other mutants was  reduced or undetectable, presumably due to misfolding and/or degradation. The mutants listed in Table I were screened for potential inhibitory effect on NFB reporter gene activation following stimulation with IL1␤.
Of particular significance is the result with the P200H mutant (T3), analogous to the Pro/His mutation in TLR4, which is responsible for the LPS hyporesponsiveness of C3H/HeJ mice (14). Similar loss of signaling in other TIR-containing molecules such as MAL/TIRAP (25) and IL1RAcP (26,27) has been reported, indicating that the invariant proline in the BB loop of Box 2 in these molecules is one of the interactive sites for other TIR domain-containing proteins. In striking contrast, a similar mutation (P200H) in MyD88 TIR domain, tested within a range of input concentrations, did not change the dominant negative effect of MyD88 TIR on IL1␤-induced NFB reporter gene activation in 293T cells (Fig. 4A). These cells showed a similar level of expression of wild-type and mutant proteins (Fig. 4B, inset). This result is consistent with recent modeling studies of the interaction of MyD88 with TLR2 and TLR4 (15), which suggested that the highly conserved proline residue may not participate in the protein-protein interactions of MyD88 with TLRs.
In contrast to the canonical P200H mutation, the following mutants displayed Ͼ2.5-fold loss of the dominant negative effect on IL1␤-stimulated NFB reporter gene activation as compared with the wild-type MyD88 TIR domain (Fig. 4B): D171A in helix ␣A (T1), triple mutant D195A/R196A/D197A (T2) in Box 2, D275A in ␤E-strand (T4), and double mutant F285A/W286A (T5) in Box 3. The result with the F285A/W286A mutant is consistent with the loss of inhibition displayed by Box 3-deleted MyD88 TIR domain (Fig. 3B). All these mutants and the wild-type MyD88 TIR domain were expressed at a comparable level in HEK 293T cells (Fig. 4B, inset).
These mutants were chosen for further validation of the functional significance of mutated MyD88 TIR domain residues. We selected IL1␤-induced expression of an endogenous gene that encodes inflammatory cytokine, IL6, in human fibroblast MRC-5 cells for testing MyD88 TIR domain mutants. The expression of the IL6 gene is regulated by NFB, and its mobilization by IL1␤ is dependent on MyD88 (1-3). Upon stimulation of MRC-5 cells with IL1␤, an inflammatory cytokine, IL6, was expressed. This expression of endogenous IL6 gene was suppressed 2.5-fold by the dominant negative TIR domain of MyD88 (Fig. 4C). The observed degree of inhibition of endogenous IL6 gene expression is smaller than suppression of NFB reporter gene activity by the MyD88 TIR domain in HEK 293T cells, most likely due to the lower transfection efficiency of MRC-5 cells, which minimizes the impact of potential inhibitors on IL6 expression. Nevertheless, inhibition of IL1␤-induced IL6 production was dependent on residues Asp 171 , Asp 275 , Asp 195 /Arg 196 /Asp 197 and Phe 285 /Trp 286 because alanine substitutions reduced the dominant negative effect of wild-type MyD88 TIR domain. These functional studies of IL␤induced IL6 production are consistent with NFB-dependent reporter gene activation (Fig. 4B). Thus, our first series of mutagenesis experiments identified interactive sites within the MyD88 TIR domain responsible for coupling IL1␤ signaling to NFB translocation to the nucleus and induction of endogenous IL6 gene.
Mutagenesis of Full-length MyD88 Reveals an Interactive Site for Direct Contact with IL1RAcP-It is still unknown which of the interactive sites identified in the MyD88 TIR domain are responsible for direct contact of MyD88 with the IL1 receptor complex subunit, IL1RAcP, which is indispensable for IL1␤ signaling (26). Alternatively, these interactive sites within MyD88 could participate in IL1␤-induced oligomerization of MyD88 through homotypic interactions mediated by its TIR domain.
To sort out these possibilities, a second series of mutagenesis experiments was conducted. Five selected mutations were engineered in full-length MyD88 to allow a comparative analysis of its direct interaction with IL1RAcP (Table II). HEK 293T cells were co-transfected with Myc-IL1RAcP and AU1-MyD88 or its mutant constructs. We used immunoprecipitation followed by Western blotting to assess the effect of mutated residues on the receptor TIR-adaptor TIR interaction. As demonstrated in Fig. 5 A and B) and cytokine IL6 production (C) assays. A, concentration-dependent inhibition of NFB reporter gene activity by wild-type and T3 mutant (P200H). B, inhibition of NFB reporter gene activity by mutants T1, T2, T4, and T5. HEK 293T cells were transfected with MyD88-TIR or its mutant constructs (1-15 g in A and 12.5 g in B) together with NFB-luc and RL-TK. After 24 h, IL1␤ was added at concentration of 10 ng/ml for 6 h. Cells were harvested and split for both reporter assay and Western blotting. NFB reporter gene activity was analyzed and normalized on the basis of RL-TK activity. The figure shown was a representative of three experiments. Insets in A and B, expression levels of TIR and TIR mutants as determined by Western blotting. C, inhibition of IL6 production by T1, T2, T4, and T5 mutants. MRC-5 cells were transiently transfected with TIR mutants and stimulated with 10 ng/ml IL1␤ for 6 h. IL6 production was analyzed with cytometric bead array as described under "Experimental Procedures." Data represent combined results from three independent experiments done in triplicate. Error bars indicate the S.E. **, p Ͻ 0.01; *, p Ͻ 0.05, by two-sided Student's t test.
three-dimensional docking model of MyD88 and IL1RAcP and verify the strategic position of these three residues as a main interactive site on the surface of the MyD88 TIR domain for its binding to the TIR domain of IL1RAcP.
The Development of the Three-dimensional Docking Model of MyD88-IL1RAcP Interaction-The three-dimensional docking model (Fig. 6) was developed by optimized superposition of two mutually interacting TIR domains of MyD88 and IL1RAcP. The negatively charged side of MyD88 TIR (see Fig. 2B) was selected as a possible interface of the molecule. This side contains Asp 195 /Arg 196 /Asp 197 residues that are essential, on the basis of mutagenesis studies (Fig. 5), for MyD88-TIR heterotypic interaction with IL1RAcP-TIR. Then, a suitable positively charged site of IL1RAcP was selected to conduct the modeling computation process. This modeling was based on geometry optimization by energy minimization followed by molecular dynamic computation using the program SANDER (AMBER software package). Ribbon structure was developed with the DeepView program, whereas the molecular surface of associated proteins was determined by PSSHOW (SYBIL-Tripos software package). Separation surface indicates that there is no crossing of molecular surfaces, and the distance between them is within the range of 0.4 -4.7 Å, whereas their topology is diverse and contains several deep pockets.
Development of this three-dimensional model allowed us to verify the contribution of the triplet of functionally important residues Asp 195 /Arg 196 /Asp 197 to the binding reaction with IL1RAcP TIR domain. An analysis of the tertiary structure of two TIR domains that participate in the docking model indicates that three mutated residues (Asp 195 /Arg 196 /Asp 197 ), which are responsible for a loss of MyD88 binding to IL1RAcP, are involved in the interaction with residues 527-534 of IL1RAcP previously identified to play a key role in the IL1␤ signaling pathway (26,27). Thus, our study identified a complementary site on the MyD88 TIR domain that contributes to its interaction with the IL1RAcP TIR domain. We therefore postulate that the negatively charged "knob," partially composed of BB loop on the surface of the MyD88 TIR domain (Fig.  2B), fits into the positively charged lysine patch formed by residues 527, 530, and 532 of the IL1RAcP TIR domain previously identified by Radons et al. (26,27) as essential for IL1␤ signaling.
Interactive Sites Involved in Oligomerization of MyD88 through Homotypic Interaction of Its TIR Domain-Following IL1␤-induced interaction of IL1RAcP with MyD88, this adaptor oligomerizes and interacts with downstream signal transducers (2,3,7). Homotypic oligomerization of MyD88 due to its forced expression resulted in robust activation of NFB reporter gene activity observed in the absence of IL1␤ stimulation. This receptor-independent effect of ectopically expressed MyD88 oligomers was abolished by co-transfected MyD88-TIR domain (data not shown). Therefore, we examined the direct interaction of full-length MyD88 co-expressed with MyD88 TIR domain or its mutants. We co-transfected HEK 293T cells with full-length MyD88 that contained c-Myc epitope tag along with the wild-type or mutated TIR domain that contained AU1 epitope tag. As demonstrated in Fig. 7, P200H mutant bound to full-length MyD88 to a similar extent as wild-type MyD88 TIR domain. However, mutations in Box 2 (D195A/R196A/D197A) and Box 3 (␤E-strand D275A and F285A/W286A) caused a loss of binding to MyD88, suggesting that these mutated residues constitute the interactive sites for homotypic oligomerization of the MyD88 TIR domain. Thus, the interacting site in Box 2 composed of residues Asp 195 /Arg 196 /Asp 197 potentially has an additional function that may encroach on the ability of MyD88 to interact with IL1RAcP. However, in a cascade of signaling steps induced by IL1␤, heterotypic interaction of IL1RAcP TIR domain with MyD88 TIR domain precedes oligomerization of MyD88. The latter depends on homotypic binding mediated by its TIR domain. Therefore, the interactive site in Box 3 is not likely to be involved in binding of MyD88 to IL1RAcP. Rather, this site participates in homotypic MyD88 oligomerization and possibly other transactions involving downstream signal transducers. This interpretation is consistent with the loss of inhibition of IL1␤-induced signaling by the MyD88 TIR domain upon truncation of its carboxyl-terminal segment that contains Box 3.
Taken together, our results identify key residues in the MyD88 TIR domain that are responsible for its heterotypic interaction with IL1RAcP. In addition, we identified interactive sites for homotypic oligomerization of MyD88. These protein-protein interactions evoked by IL1␤ are essential for its signaling to the nucleus mediated by NFB and other proinflammatory stressresponsive transcription factors. Mutations identified on the interacting surface of the MyD88 TIR domain are functionally important because they interfere with the induction of endogenous gene that encodes IL6. This inflammatory cytokine, along with IL1␤, is responsible for cardinal signs of systemic inflammation: fever, leukocytosis, thrombocytosis, acute phase protein response, and tissue injury (1,5). The development of the docking three-dimensional model of MyD88-IL1RAcP binding, in which a cluster of highly charged residues in Box 2 plays a key role, reaffirms their strategic role in contacting complementary site on IL1RAcP TIR domain. This site is composed of several positively  5. Interaction of MyD88 mutants with IL1RAcP. HEK 293T cells were transiently transfected with Myc-IL1RAcP (10 g) and fulllength MyD88 or its mutants (10 g). After 24 h, the cells were stimulated with 100 ng/ml IL1␤ for 5 min, harvested, and immunoprecipitated as described under "Experimental Procedures." Protein samples bound to anti-Myc antibody protein A beads were subjected to SDS-PAGE, and the immunoprecipitated MyD88 or its mutants were monitored by immunoblotting with anti-AU1 antibody (top panel). Expression of MyD88 and its mutants (middle panel) and IL1RAcP (bottom panel) was analyzed by immunoblotting. charged residues identified previously in EE loop residues 527-534 (26,27). Thus, it is not surprising that the invariant Pro 200 is inconsequential for MyD88-ILRAcP interaction. However, canonical mutation of the invariant proline to histidine in TLR4, which abolishes its signaling by LPS (14), indicates that different ligands and their cognate TLRs utilize MyD88 in a structurally distinct way. In addition to TLR4, a proline to histidine mutation attenuated signaling mediated by TLR2, MAL, and IL1RAcP (14,25,26). We interpret the latter result as indicative of invariant proline playing a significant role in reshaping the EE loop in IL1RAcP or in interactions with the TIR domain of IL1RI or with signaling molecules other than MyD88.
In summary, our data indicate that following stimulation with IL1␤, the MyD88 TIR domain binds to the IL1RAcP TIR domain via a highly charged interactive site composed of residues 195-197 within the BB loop of Box 2. We postulate, on the The HEK 293T cells were transiently transfected with c-Myctagged full-length MyD88 (5 g) and AU1-tagged MyD88-TIR (5 g) or AU1tagged MyD88-TIR mutants (5 g). After 24 h, the cells were harvested and immunoprecipitated as described under "Experimental Procedures." Protein samples bound to anti-AU1 antibody protein A beads were subjected to SDS-PAGE and immunoprecipitated.
MyD88-TIR or MyD88-TIR mutants were detected with anti-c-Myc antibody (top panel). Expression of MyD88-TIR and MyD88-TIR mutants (middle panel) or MyD88 (bottom panel) was analyzed by immunoblotting. basis of the three-dimensional docking model developed herein, that this interactive site is complementary to the previously identified site composed of residues 527-534 within the EE loop of the IL1RAcP TIR domain (26,27). Despite its proximity to the interactive site, invariant Pro 200 of the MyD88 TIR domain does not play a role in IL1␤-induced signaling. However, residues located in Box 3 are essential for subsequent homotypic interaction of MyD88 TIR domain. In a broader context, the results suggest that the IL1␤ signaling pathway differs from other ligand-initiated TLR intracellular signaling by usage of distinct structural motifs within the MyD88 TIR domain. Further mapping of the MyD88 surface will expand our understanding of its role in integrating signals derived from a variety of Toll-like receptors.