Structural and Functional Evidence for the Role of the TLR2 DD Loop in TLR1/TLR2 Heterodimerization and Signaling*

The Toll/Interleukin-1 receptor (TIR) domain of the Toll-like receptors (TLRs) plays an important role in innate host defense signaling. The TIR-TIR platform formed by the dimerization of two TLRs promotes homotypic protein-protein interactions with additional cytoplasmic adapter molecules to form an active signaling complex resulting in the expression of pro- and anti-inflammatory cytokine genes. To generate a better understanding of the functional domains of TLR2 we performed a random mutagenesis analysis of the human TLR2 TIR domain and screened for TLR2/1 signaling-deficient mutants. Based upon the random mutagenesis results, we performed an alanine scanning mutagenesis of the TLR2 DD loop and part of the αD region. This resulted in the identification of four residues crucial for TLR2/1 signaling: Arg-748, Phe-749, Leu-752, and Arg-753. Computer-assisted energy minimization and docking studies indicated three regions of interaction in the TLR2/1 TIR-docked heterodimer. In Region I, residues Arg-748 and Phe-749 in TLR2 DD loop were involved in close contacts with Gly-676 in the TLR1 BB loop. Because this model suggested that steric hindrance would significantly alter the binding interactions between DD loop of TLR2 and BB loop of TLR1, Gly-676 in TLR1 was rationally mutated to Ala and Leu. As expected, in vitro functional studies involving TLR1 G676A and TLR1 G676L resulted in reduced PAM3CSK4 mediated NF-κB activation lending support to the computerized predictions. Additionally, mutation of an amino acid residue (TLR2 Asp-730) in Region II also resulted in decreased activity in agreement with our model, providing new insights into the structure-function relationship of TLR2/1 TIR domains.

The first line of defense against any invading microbe is provided by the Innate Immune system. Cells of the innate immune system sense and respond to microbial products via the Tolllike receptor family. Toll-like receptors (TLR) 2 are an evolu-tionarily conserved family of cell surface molecules that participate in innate immune recognition of pathogen-associated molecular patterns (PAMPs) (1). To date 11 mammalian homologues of these receptors have been found (TLR1-11) (2) and individual members of the TLR family recognize a diverse array of microbial components. For example, the first human toll like receptor to be identified, TLR4, senses lipopolysaccharide (LPS) while TLR2 on the other hand senses diacylated or triacylated lipopeptides after heterodimerizing with either TLR6 or TLR1, respectively. The primary function of the TLRs is to alert the immune system to the presence of pathogenic microorganisms. Upon recognition of specific microbial components these receptors turn on a complex series of signaling events leading to the production of numerous immunologically important cytokines, chemokines, and effector molecules. Additionally, microbial products also induce the production of pro-inflammatory cytokines, such as IL-1, TNF-␣, and IL-12 and the expression of co-stimulatory molecules on professional antigen presenting cells that are necessary for the activation of T and B cells. Thus, in addition to directly controlling the microbial infection, the innate immune response is also instructive to the adaptive immune response (2).
The conserved cytoplasmic TIR (Toll/IL-1 receptor) domains of the IL-1 and Toll-like receptors are the critical focal point for the generation of ligand-induced cytoplasmic signaling cascades. It is generally believed that the TIR domains serve to promote homotypic protein-protein interactions between receptor chains and with additional cytoplasmic adapter molecules to form an active signaling complex. For signaling all the TLRs utilize one or more of the four known TIR-containing adaptor molecules: MyD88, TIRAP/MAL, TRIF, and TRAM (reviewed in Refs. 3 and 4). Despite the critical role of the TIR domain in coordinating the initial cytoplasmic signaling events relatively little is known regarding how homotypic TIR-TIR interactions are formed. Several studies have pointed to the important role of a conserved proline within the conserved "BB loop" as being critical for the generation of downstream signals. Indeed, mutation of this proline to a histidine in TLR4 is responsible for the loss of LPS-responsiveness in the C3H/HeJ mouse (5) and acts as a dominant negative mutant. Mutation of the conserved res-idue within TLR2 (Pro-681) was demonstrated to result in an inability to recruit MyD88 (6) but surprisingly has no effect on the ability of TLR4 to bind MyD88 (7). A peptidomimetic based on the BB loop of the MyD88 (8) clearly showed the role of BB loop in signaling of a TIR domain protein (IL-1) as the mimic could block downstream signaling. Recently, Loiarro et al. (9) reported that an eta-peptide derived from the BB-loop region of the MyD88 resulted in the inhibition of homodimerization of MyD88 and thereby signaling.
Given the obvious importance of the TIR domain in the control and coordination of innate immune responses, it is surprising that, with the exception of the BB loop, there is a relative lack of information regarding functional domains within the cytoplasmic portion of Toll receptors that are essential for signaling activity. Potential clues as to the importance of several conserved amino acids within TIR domain were provided by Slack et al. (10) who performed alanine scanning mutagenesis of a number of residues within the three conserved TIR domain boxes in the type I IL-1 receptor. Of the twelve individual mutations made, four resulted in decreased cell surface expression and of those four; only two could be demonstrated to have decreased abilities to signal for NF-B or SAPK activation. Ronni et al. (11) published a detailed alanine-scanning mutagenesis study of the TLR4 TIR domain which identified two structural surfaces that were required for TLR4-dependent signaling in macrophages. However, no information was provided regarding a molecular basis for the loss of function. Finally, Tao et al. (12) have provided crystallographic data demonstrating in the crystal packing the TIR domains of C713S mutant of TLR2 formed asymmetric dimers and that the critical BB loop can adopt different conformations within the structure (PDB accession: 1O77). Interestingly, of the five chains in the crystallographic asymmetric unit, two molecules showed interaction between the ␣D helix and DD loops of one with the ␣B helix and BB loop of another. Besides packing forces, the interacting surface was also held together by an interchain S-S linkage, but the authors argued for a minimal role of the latter. In all the crystal and homology-modeled structures of the TIR domains, the DD loop is located on the opposite side of the BB loops (6,7,12). Because TLR2 is not functional as a homodimer and considering the conserved homology in the BB region among TLRs, this observation and supporting functional studies led them to suggest that TLR1 or TLR6 can form similar heterodimeric structures with TLR2. It is important to note here that Xu et al. (6) reported that gel-filtration and dynamic light-scattering experiments carried out to understand the oligomerization state of the isolated TIR domains in solution indicated a low affinity for self-association of the TIR domains. Additionally, based on molecular modeling studies, Dunne et al. (7) predicted that the DD loops of adaptor molecules (Mal and MyD88) could be involved in molecular recognition and subsequent signaling. In summary, a clear understanding of the role of the DD loop of the TLR2 molecule in innate immunity is still elusive. In this study we report a structural and functional analysis of the TLR2 TIR domain and based on computational and experimental methods propose a model for how the TLR2 DD loop may interact with the TLR1 BB loop.

DNA, Plasmids, and Reagents
pcDNA5 FRT/TO-TLR2 was generated by moving the TLR2 coding region as a HindIII and BamHI fragment from pcDNA3.1-TLR2 (13) to the same sites in pcDNA5FRT/TO. pDHA-TLR2 was generated by PCR amplifying the TLR2 gene from pcDNA3.1 TLR2 using primers with ApaI (5Ј-AAGGGCC-CTCTCCAAGGAAGAATCCTCC-3Ј) and PstI (5Ј-AACTGCA-GCTAGGACTTTATCGCAGCTC-3Ј) restriction enzyme sites at N and C terminus, respectively and ligating to the same sites in pDisplay HA mouse TLR6 (14) (a gift from D. Underhill, Institute for Systems Biology, Seattle, WA). Wild-type TLR1 plasmid (pFLAG-CMV-TLR1) was a gift from P. Tobias (The Scripps Research Institute, Dept of Immunology, La Jolla, CA). pFLAG-CMV2 human MyD88 was made by inserting the human MyD88 coding region into the HindIII and SmaI sites of pFLAG-CMV2. The human MyD88 coding region was PCR amplified using forward primer with HindIII (5Ј-TTATAA-GCTTGCTGCAGGAGGTCCCGGCGC-3Ј) and a bluntended reverse primer (5Ј-AATTTCTAGATCAGGGCAGG-GACAAGGCGTTG-3Ј). All the point mutations were made using QuikChange Site-directed Mutagenesis kit from Stratagene. Plasmids were verified by restriction mapping and sequencing (University of Virginia Biomolecular Research Facility, Charlottesville). The MIP-3␣ luciferase reporter plasmid was a gift from A. C. Keates (Harvard Medical School, Boston, MA), IL-8 luciferase was a gift from N. Mukaida (Kanazawa University, Japan), NF-B luciferase and pEGFPN1 were obtained from Clontech (Mountain View, CA). PAM 3 CSK 4 was obtained from EMC Microcollections (Tübingen, Germany). All other regents were obtained from Sigma.

Random Mutagenesis
Random mutations were generated using GeneMorph PCR Mutagenesis kit (BD Biosciences). Using PCR primers Forward: (5Ј-TGATCCTGCTCACGGGGGTC-3Ј) and Reverse: BamHI (5Ј-AAGGATCCCTAGGACTTTATCGCAGCTCTCAG-3Ј) a 554-bp region of TLR2 containing the TIR domain was amplified under mutagenic conditions resulting in 1 to 3 random mutations per 500 bp. The resulting PCR product was digested with PpuMI and BamHI then cloned at the same sites in pcDNA5-TLR2. Individual clones were then screened for correct restriction pattern and tested for their ability to signal in response to PAM 3 CSK 4 in HEK 293 transient transfection system. Clones resulting in reduced activity were sequenced using a TLR2 internal primer (5Ј-GCAAATTACCTGTGTGACTC-3Ј) to identify the mutations.

Quantitative Reverse Transcriptase (RT)-PCR
Total RNA was purified using the TRIzol reagent (Invitrogen). RT of 0.5 g of total cellular RNA was performed in a final volume of 20 l containing 1ϫ final first-strand buffer, 1 mM each dNTPs, 20 units of placental RNase inhibitor, 5 M random hexamers, and 9 units of Moloney murine leukemia virus RT (Invitrogen). After incubation at 37°C for 45 min, the samples were heated for 5 min at 92°C to end the reaction and stored at Ϫ20°C until PCR use. cDNA (2 l) was subjected to real-time, quantitative PCR using the MJ Research Opticon system with SYBR Green I (Molecular Probes, Eugene, OR) as a fluorescent reporter. Duplicate PCR reactions were performed for each sample, and the average threshold cycle number was determined using the Opticon software. Levels of MIP-3␣ and IL-8 expression normalized to HGPRT levels were determined using the formula 2 (Rt-Et) , where Rt is the threshold cycle for the reference gene (HGPRT), and Et is the threshold cycle for the experimental gene (⌬⌬CT method). Data are thus expressed as arbitrary units. Sequences of primers used: MIP-3␣ (F: 5Ј-CTGG-CCAATGAAGGCTGTGA-3Ј, R: 5Ј-ACCTCCAACCCCAGCA-AGGT-3Ј), IL-8 (F: 5Ј-GGCAGCCTTCCTGATTTCTG-3Ј, R: 5Ј-GGGGTGGAAAGGTTTGGAAGT-3Ј) and hypoxanthine guanine phosphoribosyl transferase (HGPRT) (F: 5Ј-TTGGAAA-GGGTGTTTATTCCTCA-3Ј, R: 5Ј-TCCAGCAGGTACGC-AAAGAA-3Ј).

Cells, Cell Culture, Transfection, and FACS
HEK 293 cell line was obtained from American Type Culture Collection and cultured in Dulbecco's modified Eagle's medium (Mediatech, Herndon, VA) ϩ10% fetal bovine serum (Hyclone, Logan, UT). The HEK 293/FlpIn cell line which was engineered for use with the FlpIn recombinase system was purchased from Invitrogen and maintained in RPMI ϩ10% fetal bovine serum plus Zeocin (Invitrogen) as recommended. HEK 293 cell line was transfected using Lipofectamine 2000 (Invitrogen) per manufacturer's recommendations.
Stable Cell Lines of pDHA-TLR2 clones were made by transfecting HEK 293 cell lines with respective DNA using Lipofectamine 2000 and selecting with G418. For making stable cell lines of pcDNA5/FRT/TO-TLR2 clones, the HEK 293/FlpIn cells were transfected in the same way as described above with the respective pcDNA5-TLR2 DNA along with recombinase expressing plasmid pOG44 (Invitrogen) in 1:9 ratio. Hygromycin-B (Invitrogen)-resistant clones were picked and screened for surface expression with FACS.
To confirm the surface expression of WT or mutant TLR2 on stable cell lines, cells were washed with cold PBS, blocked with cold 0.1% bovine serum albumin in PBS and stained with PEconjugated monoclonal antibody against TLR2 (TLR2.1 from eBiosciences, San Diego, CA) for 20 min at room temperature. After a wash with cold 0.1% bovine serum albumin in PBS cells were analyzed by FACS along with positive (WT-TLR2) and negative control (HEK 293 cells) using FACS Calibur (Becton Dickinson) flowcytometer and FLOWJO software. Procedure for checking the surface expression of the clones using transient transfection was similar except that pEGFPN1 plasmid (Clontech) was cotransfected with the TLR2 mutant or WT TLR2 plasmids (3:1 ratio) into HEK 293 cells using Lipofectamine 2000. Cells were harvested after 36 h, washed with PBS, blocked, and stained as described above. FACS was done by gating on GFP-positive cells. Mean fluorescence intensity (MFI) for the PE channel was calculated for the GFP-positive cells.

Immunoprecipitation and Western Blot
HEK 293T cells (10 6 ) in a 10-cm tissue culture plate were transfected with the indicated plasmid DNA (10 g each) using Lipofectamine 2000. 48 h after transfection cells were washed once with ice-cold 1ϫ PBS (Mediatech, Inc., Herndon VA) and lysed in ice-cold lysis buffer (10 mM Tris, pH 7.5, 150 mM NaCl, 1 mM EDTA, pH 8.0, 1% v/v Nonidet P-40, and 10% v/v glycerol), this buffer was supplemented before use with 1ϫ (final) Halt protease inhibitor (PI) mixture (Pierce). To remove cell debris lysates were centrifuged at 12,000 ϫ g for 15 min. at 4°C. Clear lysates were put in a rotary mixer at 4°C with 4 g of a rabbit polyclonal antibody against the HA epitope tag (Y-11; Santa Cruz Biotechnology, Inc., Santa Cruz, CA) for 2 h followed by the addition of 20 l of protein A-Sepharose bead slurry (Pierce). After washing in lysis buffer, the beads were suspended in 2ϫ SDS gel loading dye, heated at 95°C for 10 min, followed by centrifugation at 12,000 ϫ g for 15 min. Samples and dual color Precision plus Protein standards (Bio-Rad) were run on 10% PAGE and transferred to nitrocellulose membrane. HA-tagged TLR2 was detected using anti-HA (262K) mouse monoclonal antibody (Cell Signaling Technologies, Inc., Danvers, MA) and FLAG-MyD88 with anti-FLAG M2 mouse monoclonal antibody (Sigma). Secondary antibody was horseradish peroxidase-conjugated anti-mouse IgG (Cell Signaling Technologies, Inc.). Blots were developed using ECL plus Western blotting detection system, scanned and analyzed by Storm 840 and using software ImageQuant 5.2, all from Amersham Biosciences (GE Health Care).

Computational Methodology
Energy Minimization-All calculations were performed using the precompiled executables of Tinker molecular modeling suite of programs v. 4.0 for Windows platform (15,16). The PDB coordinate files 1FYW and 1FYV were used as the starting structures for the TIR domain of the TLR2 and TLR1 molecules, respectively (6). Prior to energy minimization of these structures, the covalent seleniums were replaced by sulfur atoms. The corrected structures were then minimized using the Newton program and all-atom AMBER 99 force-field to a rootmean square (RMS) gradient convergence of 0.05 kcal/mol/Å. The calculations were carried out in implicit dielectric conditions of 4.0 to mimic the receptor-like conditions where the role of water and salts can be presumed minimal (17,18). The resultant coordinates were confirmed to lack any racemization of chiral centers or cis-trans isomerization during optimization. A comparative plot of the backbone dihedrals confirmed no large alterations occurred in the structures during optimization for both the molecules. The minimized coordinates were then used for graphical analysis and docking studies.
Protein-Protein Docking and Co-minimization-High resolution rigid body protein-protein docking between the TIR domains of TLR1 and TLR2 was done employing Global Range Molecular Matching (GRAMM) program v. 1.03 (19,20). Employing a grid-step of 1.7, repulsion factor of 30.0, atomic radius for potential range, gray mode projection and 10 degrees for angle of rotations, 500 lowest energy matches were written out. From this output, only those docked structures were sorted which satisfied two conditions: 1) the ␤-sheet backbone of TIR domains of TLR2 and TLR1 were in same plane, and 2) the N termini were in parallel orientation i.e. toward the membrane. The most stable low energy structure satisfying the imposed requirements (E Ϫ E 0 Ͻ 2 kcal/mol) was considered for co-minimization experiments. For co-minimization, implicit water conditions were used by considering a dielectric value of 80. A cubic box size of one edge equal to 250 Å was used and the van der Waals calculations were cut-off at 15 Å. Employing Newton program and all-atom AMBER99 force field, the coordinates of the TLR2/1 complex were minimized to a RMS gradient convergence of 0.05 kcal/mol/Å. As before, the final coordinates were confirmed to lack any racemization of chiral centers or cis-trans isomerization during optimization and were used for analyzing close contacts between the molecules.

RESULTS AND DISCUSSION
Random Mutagenesis and Functional Screening of TLR2 TIR Domain-The only detailed structure/function analysis of any TLR TIR domain was done for TLR4 by Ronni et al. (11) and very little is known about the amino acid residues and subdomains of the human TLR2-TIR, which are essential for function. Thus, we decided to undertake a detailed structure/function analysis of the TLR2 TIR domain. Unlike the alanine substitution approach used by Ronni et al. (11) we decided to perform random mutagenesis followed by a functional screen to identify the crucial residues involved in signaling. Random mutagenesis of TIR-TLR2 (amino acids 607-784) was done using a PCR-based commercially available kit, and the resulting product was cloned back into pcDNA5-TLR2. PCR conditions were optimized to result in 2-3 mutations per 500 bp (for details see "Materials and Methods"). The resulting random mutagenic (RM) clones were then screened for their abilities to activate an NF-B Luc reporter construct in response to PAM 3 CSK 4 (TLR2 agonist) in the HEK 293 transient transfection assays. Clones which demonstrated reduced activity (compared with WT) were sequenced to identify the mutation(s). This screen resulted in the identification of numerous RM clones with reduced activity, 17 of which are shown in Table 1. The inactive mutants can be basically grouped in three sets: First, clones that had mutations resulting in the stop codons within the coding region which include RM 27, 44, 51, 53, 54, 60, and 61. Signaling in these mutants is blocked implying the need of complete integrity of the TIR domain in TLR2 for the signaling to occur. The second set of clones had mostly single or multiple mutations in and around the well characterized BB loop. Under this category the mutants were RM5, RM18, RM37, RM38, RM57, RM62, and RM73. These mutants also show aberrant signaling supporting the previously pointed out crucial role of the integrity of BB loop for receptor functionality (6). The third group of clones RM6, RM30, and RM67, had single or multiple mutations outside the BB loop. Of all these random mutant clones we chose RM18, RM 30, and RM67 for a more detailed analysis. These clones were stably transfected into HEK 293 Flp-In cells and checked for surface expression by FACS using the TL2.1 antibody. All the stable clones expressed approximately equivalent levels of TLR2 on the surface (Fig.  1A). These clones were then examined by real-time RT-PCR analysis for their abilities to induce IL-8 (Fig. 1B) and MIP-3␣ (Fig. 1C) mRNA in response to PAM 3 CSK 4 . These data indicated that the expression of both genes was almost completely inhibited in RM18 and RM30 and significantly repressed (ϳ80%) in the case of RM67. RM18 contains a single mutation in the BB loop (I685F), which, based upon previous reports of the critical role of the BB loop in mediating TLR2-MyD88 interactions, would be expected it to be inactive (6). The reasons for the loss of activity observed for the other two clones are less clear. RM30 is a double mutant (F701V and K743E) and RM67 which has a single mutation L762Q, which also decreases receptor activity significantly. Interestingly, according to the crystal structure 1FYW the amino acids in both these mutants i.e. Lys-743 and Leu-762 lie on the opposite face of the TIR domain from BB loop whereas F701V is buried deep inside the molecule, is not surface exposed, and is in fact a valine in TLR4. Leu-762 is a conserved residue among various TLRs and is pres- ent at the base of the ␣D region whereas Lys-743 is present in the DD loop of TLR2 (6). In the TLR2 molecule, the less conserved DD loop is also very mobile, as evident from the high R values for that region in the crystallographic data. Though Dunne et al. (7,21) have suggested that the adaptor molecules like Mal and MyD88 might bind/dock via their DD loop, no such argument has been reported for TLR2 molecule. As mentioned earlier, based on the contact surface among two chains in the C713S mutant TLR2 crystal, Tao et al. (12) implied that DD loop of TLR2 can interact with the BB loop of the another TLR to form TLR2⅐TLRx heterocomplex. However, no work identifying the key residues involved in the DD loop mediated TLR2 activity has been reported to date. Given the potential important role of the DD loop in mediating TLR2-dependent responses we decided to perform an alanine scanning mutagenesis of the TLR2 TIR domain in this region. Alanine Scanning Mutagenesis of the DD Loop of TLR2-The DD loop is present as a flexible loop on connecting the fourth ␣-helix and fourth ␤-sheet of the TIR domain. Using a sitedirected mutagenesis approach, all the amino acids in the DD loop region (Glu-738 to Phe-749) were individually changed to alanine, except residue 744, that being alanine itself. When tested for ability to signal, two of the mutants (R748A and F749A) were found to have reduced ability to induce NF-B Luc construct in response to PAM 3 CSK 4 in the transient transfection assay. The activities of R748A and F749A were ϳ60 and 40% less, respectively, as compared with the wild-type TLR2 ( Fig. 2A). Surprisingly, the K743A substitution had no significant negative effect on NF-B activation in contrast to the effects observed above with the double F701V/K743E (RM30) mutant. The reason for this effect is unclear but may suggest that although not solvent accessible the F701 may play an essential role in stabilizing TIR domain secondary structure and/or the additional mutation of the surface Lys-743 to Glu alters the DD loop region by reversing the electrostatic potential. Because Phe-749 was at the end of the DD loop we generated mutations to screen five residues further downstream in ␣D region, i.e. Cys-750 to Lys-754. Notably, this series also included Arg-753, whose naturally occurring mutation (polymorphism) to glutamine has been shown to result in decreased recognition of some bacterial peptides and also predispose to Staphylococcal and Mycobacterial infections (22). When screened for activity in response to PAM 3 CSK 4 , two of the mutants L752A and R753A were found to show reduced activity ( Fig.  2A). All these clones were checked for surface expression by FACS and found to be similarly expressed on the surface (Fig.  2B), thereby ruling out this reason for their inability to signal. We further tested some of the mutants (R748A, F749A, L752A, and R753A) for their abilities to induce MIP-3␣ and IL-8 promoter activities in transient transfection assays. Similar to the results of the screening with NF-B Luc all mutants were found to have reduced ability to activate these promoters in response to PAM 3 CSK 4 (Fig. 2C). We also generated cell lines for these four mutants by stably transfecting them in HEK 293 cells. Individual clones were then screened for surface expression of mutant TLR2 protein using FACS and clones which gave surface expression equivalent to the WT-TLR2 were used for further study (Fig. 3A). These mutant TLR2 cell lines were then stimulated with PAM 3 CSK 4 for 8 h and the levels of IL-8 and MIP-3␣ mRNA were analyzed by RT-PCR (Fig. 3, B and C,  respectively). MIP-3␣ and IL-8 expression for all the four mutants (R748A, F749A, L752A, and R753A) were found to be reduced, which correlated with the transient transfection assays. As a result of our alanine scanning mutagenesis of the DD loop and downstream region we identified four crucial residues which appeared to be required for optimal signaling in response to PAM 3 CSK 4 . Our results for the TLR2 DD loop are in general agreement with those reported by Ronni et al. (11) in their alanine scanning mutagenesis of the human TLR4 protein. Results from that study also indicate that the double mutants EK775-776AA and QQ781-782AA show reduced NF-B Luc activity (Ͻ10% of WT); these residues correspond to DD loop residues in the TLR4 TIR domain. Although the DD loop is so mobile that one cannot resolve it well in the crystallographic analysis, and its sequence is not well-conserved among the TLR family, it nevertheless appears to mediate the formation of an active signaling complex for at least two members of the family: TLR2 and TLR4. Herein, our work identified four residues in DD loop and ␣D helix of TLR2 which played critical roles in signaling via the TLR2/1 heterotypic complex.
Computer-assisted Docking of TIR Domains of TLR2 and TLR1-To gain structural insights into how the DD loop region may play a role in TLR2 signaling, we performed molecular modeling studies of the interaction of TLR2 with its heterologous partner TLR1. The crystal structures of both the TLR1 and TLR2 TIR domains have been previously resolved (6). The current understanding of the TLR2 signaling complex is that it forms a heterodimer with TLR1 to sense the synthetic lipopeptide PAM 3 CSK 4 resulting in the recruitment of two adapter proteins MyD88 and MAL/TIRAP to the cytoplasmic TIR domain. Previously, MyD88 has been shown to interact in vitro with the TLR2 BB loop (6) and computerized docking studies suggest that both the monomers of MyD88 and MAL/TIRAP interact with the TLR2 BB loop (7). Furthermore, as stated earlier, crystallographic evidence from TLR2 homomultimers suggests that the DD loop and BB loop may form points of contact between two molecules (12). Based on these previous reports, we hypothesized that the DD loop side of the TLR2 TIR domain might be interacting with TIR domain of TLR1 and that an inability to form such interaction might be the reason for the decreased activity observed with the TLR2 DD loop alanine mutants. To test our hypothesis we performed computer-assisted docking studies of the two TIR domain interactions. The previously resolved crystal structures of TLR2 and TLR1 PDB files 1FYW (TIR-TLR2) and 1FYV (TIR-TLR1) were used as the starting molecules. As described under "Materials and Methods," both molecules were individually energy minimized to a RMS gradient convergence of 0.05 kcal/mol/Å. This was followed by a high resolution rigid body protein-protein docking between the TIR domains of TLR1 and TLR2 using the GRAMM program. Using a grid-step of 1.7, repulsion factor of 30.0, atomic radius for potential range and gray mode projection to calculate interacting surfaces provided good selection criteria. Using only 10 degrees for angle of rotations allowed us to map the surface in the lowest possible grid. 500 lowest energy combinations were generated from the rigid body docking. Graphical analysis of docking output indicated a high (87 structures out of 100 lowest energy combinations) propensity for the TLR1 molecule to interact with the DD loop face of the TLR2 molecule. To sort out functionally meaningful orientations representing the complex between TIR domains of TLR2 and TLR1, we postulated that the N-terminal ends of both the components should be oriented toward the cell membrane. The 9th lowest energy docked complex was compatible with our assumption. Here, it is of importance to mention that identical results were obtained on docking minimized coordinates of TLR2 on the template of minimized coordinates of TLR1 and vice versa. This low energy complex was minimized again by employing a cubic BOX condition (a ϭ 250 Å). During minimization, implicit dielectric conditions representing aqueous surroundings were employed. The final minimized heteromeric complex was used to analyze the interacting surface and is shown in Fig. 4A.
Detailed analysis of this structure indicated three possible interacting faces between the two molecules. In Region I, the TLR2 DD loop residues Arg-748 and Phe-749 were closely associated with Gly-676 of TLR1 BB loop (Fig. 4C). This contact region was in agreement with the possibility suggested by Tao et al. (12). Analysis suggested that this region in the energy minimized complex was predominantly stabilized by three nonconventional C-H⅐⅐⅐O bonds, namely: two bifurcated TLR1 Gly-676 C ␣ -Hs⅐⅐⅐OϭC TLR2 Arg-748 and a TLR1 Gly-676 CϭO⅐⅐⅐H-C ␣ TLR2 Phe-749 with distances of 2.9 and 2.9, and 3.0 Å, respectively. Because in solution these molecules will undergo substantial motion, we abstained from overanalyzing our static H-bonding features. Interestingly in our model of the heterotypic complex, the critical Pro-675 in the BB loop of TLR1 was positioned close to the DD loop of TLR2. We speculate that as shown for TLR2, a single mutation of this proline to histidine will drastically reverse the electrostatic potential of the contact point and negatively effect the TLR2/1 mediated immune response. Another observation worth noting is the unique backbone conformational features across the Gly-676 residue. In the minimized complex, and torsion angles for this Gly were found to be 65°and Ϫ4°, respectively. All other natural L-␣-amino acids will generally find this backbone geometry unfavored. Whether this is the very reason behind the evolutionary conservation of Gly at this position in the BB loops of TLRs and adapter proteins will remain an interesting query for structural biologists.
As discussed above, TLR2 amino acid residues Arg-748 and Phe-749 were independently demonstrated in the functional studies to be important for optimal TLR2-mediated responses. This functional correlate provides additional confidence in the validity of the computerized docking studies. Additional modeling experiments in which the Arg-748 and Phe-749 were singly substituted for alanines confirmed the disruption of this "native-like" interaction. Additionally we observed that Leu-752 and Arg-753 are involved in stabilizing the local structure of TLR2-DD loop, since substitution to alanines in the computerized models resulted in significant deformations of local structure thereby likely inhibiting the activity. As per both crys-tal structure of TLR2 and our energy minimized structures, both the residues Leu-752 and Arg-753 are intricate members of the ␣D helical region. Because Ala also has a high propensity for forming ␣-helix, as expected, we did not observe any loss in the helical architecture of the region in the mutants. While, the role of Leu side-chain could not be determined from the modeling data, the positively charged guanidine group of the Arg side-chain played a major role in stabilizing the 747 QRF 749 region of the TLR2 DD loop. Hence in the R753A mutant studied, the truncation of the charged group of Arg led to conformational changes across the QRF region which eventually led to an altered contact topology between the two proteins during co-minimization. Notably, polymorphism of residue Arg-753 (R753Q) has been linked with the susceptibility to Staphylococcal, Borellia sp., and Mycobacterium sp. and has also been associated with acute rheumatoid fever in children (23)(24)(25)(26)(27). Thus our modeling studies also suggest a potential explanation for the observed decreased activity of the R753Q polymorphism.
Two other regions of interfacial contact between the two proteins were named as region II and III. In region II, most noticeably the TLR2 D730 comes in close proximity to His-646 and Asn-700 of TLR1 via a network of interwoven backbone and side-chain hydrogen bonds. In fact our modeling data showed a strong hydrogen bond (2.3 Å) between the side-chain O ␦ N700 of TLR1 and backbone NH of His-646, which appears to play a role in maintaining the tertiary structure of the TLR1 in the complex. The same O ␦ of N700 was also involved in the intermolecular hydrogen bonding with the backbone NH of the Asp-730 in the TLR2 molecule (2.7 Å). The third component in this network was the strong intermolecular electrostatic interaction between the oppositely charged side-chains of His-646 of TLR1 and Asp-730 of TLR2 (2.3 Å). Close inspection also suggested that there was another region of possible contact between the two molecules. ThisregiontermedasregionIIImainlysuggestedinteractionbetween the Tyr-737 of TLR1 and the C-tail helix of the TLR2 molecule. Con- To critically evaluate the regions II and III from our modeling results, we generated additional mutations in TLR2, which we reasoned to be involved in TLR1/TLR2 heterodimer formation. Mutations made in TLR2 were: D730A (possible interacting residue of TLR2 in region II) and N777A (possible interacting residue of TLR2 in Region III). In addition several multiple mutants were generated to assess possible cooperative interactions between the three regions: double mutants D730A/F749A, F749A/N777A, and triple substitution D730A/F749A/N777A. All of these mutants were tested for their surface expression by FACS in transient transfection system in HEK 293 cells and were found to be approximately equivalently expressed on the cell surface (Fig. 5B). These mutants were then tested for their ability to induce NF-B, MIP-3␣, and IL-8 luciferase promoter constructs in HEK 293 transient transfection system in response to PAM 3 CSK 4 . Results shown in Fig.  5A indicated that substitution of alanine for Asp-730 reduced activity of all the three reporters by about 50 to 60% indicating that Asp-730 may in fact participate in TLR1/TLR2 heterodimer formation as suggested by the computerized docking studies. On the other hand, we could not confirm a role for Asn-777 in stabilizing the heterodimer as the alanine substitu-tion had no effect on reporter activities. Both the double mutants D730A/F749A and F749A/N777A resulted in reduced activity. In the case of F749A/N777A, activity was equivalent to that of single mutation, F749A (ϳ40% reduction). The double D730A/F749A substitution was slightly more effective at inhibiting reporter gene responses than either of the single mutants alone, suggesting the potential for cooperativity between Regions I and II in forming an efficient signaling complex.
The earliest event in TLR signaling is the recruitment of MyD88 into the receptor complex. The decreased signaling activity observed with our mutant TLR2 molecules would suggest that we have compromised the formation of an appropriate heterodimeric signaling complex of TLR2 and TLR1. In the light of these observations we postulated that the TIR domain mutants might be inhibiting the recruitment of the MyD88 into the receptor complex. To test our hypothesis we performed co-immunoprecipitation experiments. HEK 293 cells were transiently co-transfected with either HA-tagged WT-TLR2 or HA-TLR2 double mutant (D730A and F749A) along with FLAG-tagged MyD88. Prior to immunoprecipitation of HA-Tagged TLR2, cells were stimulated with PAM 3 CSK 4 for 15 min to induce the redistribution of MyD88 to the receptor complex. Results as shown in Fig. 5 C indicate that MyD88 was recruited to WT-TLR2 but not to the D730A/F749A double mutant. Two control samples (vector and FLAG-MyD88 alone) show that the interaction was specific to WT-TLR2 and there was negligible binding of FLAG-MyD88 to the anti HA antibody and the beads. These results indicate that the TLR2 TIR domain mutants described here likely block signaling by preventing the formation of an appropriate platform for the efficient recruitment of adapter proteins into the complex.
Evaluating the Role of Gly-676 in TLR1 in TLR2/1-mediated Signaling-As detailed above, in our minimized docking model, the two backbones C ␣ Hs Gly-676 of TLR1 were involved in a bifurcated non-conventional hydrogen bonding with Arg-748 of TLR2. At the same time, the carbonyl group of Gly-676 of TLR1 was involved in a similar interaction with the C ␣ H of Phe-749 of TLR2. If this observation is true, then addition of an alkyl group on the Gly residue will result in decrement in the efficiency of native-like complexation and hence the mediated signaling. Thus two mutations G676A and G676L were generated to test this hypothesis. We expected the Leu mutant to exhibit higher degree of suppression in activity because of bulkier side-chain than the Ala one. TLR1-G676A and TLR1-G676L were tested in transient transfection assays in HEK 293 cells for their ability to induce NF-B luciferase reporter construct in response to PAM 3 CSK 4 . Because HEK293 cells express low levels of endogenous TLR1 we tested the activity of the TLR1 mutants to act as dominant negative mutants. The results of these experiments shown in Fig. 6 indicated that both the mutants (G676A and G676L) negatively affected the ability of the cells to respond to PAM 3 CSK 4 as both the mutants inhibited NF-B activation by ϳ60%. These results though did not exhibit the increased suppression from Ala to Leu substitution, but they definitely strengthened our in silico data and support the hypothesis that the interaction between the TLR1 and TLR2 TIR domains is facilitated by these residues in the BB loops and the DD loops of the two molecules, respectively.
Importantly, the docking of the two TIR domains as described here leaves the BB loop of TLR2 available for binding to MyD88 and/or TIRAP/MAL as suggested in the modeling studies of Dunne et al. (7). The results of our studies also suggest, as was previously suggested by Dunne et al. (7) that the role of the BB loop in TLR signaling may be different between the different receptors. Clearly the P714H mutation in the TLR4 BB loop abolishes activity however; the mechanism behind that effect remains elusive as this mutant is still capable of binding MyD88 and TIRAP/MAL. In contrast, the corresponding mutation in TLR2 (P681H) results in the inability of the molecule to bind MyD88 (6). Based upon the studies reported herein and our modeling data discussed earlier, it is likely that the same mutation in TLR1 (P675H) would disrupt the ability of TLR1 and TLR2 to efficiently heterodimerize. Previously, Sandor et al. (28) demonstrated using chimeric TLR1 and TLR2 molecules that the TIR domains of the two molecules are non-redundant. In that study, NF-B activation in response to araLAM or PAM 3 CSK 4 was observed only when both the TLR1 and TLR2 TIR domains were present. A recent study by Brown et al. (29) used yeast two hybrid experiments to demonstrate that MyD88 binds to the TIR domain of TLR2 and not TLR1. Their finding is consistent with our results indicating that the TLR1 BB loop is essential for heterodimerization with TLR2. In the same study, Brown et al. could not demonstrate a functional interaction between the TIR domains of TLR1 or TLR6 and TLR2 providing additional evidence that the affinities of the individual TLR TIR domains for each other are quite weak. Likewise, we have been unable to demonstrate a functional interaction between TLR2 and TLR1 TIR domains in a mammalian two hybrid system (data not shown). These results are consistent with studies from Meng et al. (30), which demonstrated that mutant TLR2 molecules lacking the entire cytoplasmic domain could still be co-immunoprecipitated with WT TLR2. Together these data support a model wherein the primary region for dimerization between TLRs is located in the transmembrane region.
Finally, Brown et al. (29) demonstrated a unique interaction between TLR1-TIR and HSP60 and HSP75. It will be interesting to determine the region of TLR1 to which these proteins bind in light of our studies indicating that the TLR1 DD loops remains accessible in the heterodimer. Our results support the conclusion that the TLR1 and TLR2 TIR domains are unique in their structures and provide for the first time a structural basis for the functional interaction between these two TLRs.
Acknowledgment-We thank Dr. J. W. Ponder for freely distributing the Tinker Suite of programs for academic purposes.