Structurally Distinct Recognition Motifs in Lymphotoxin-β Receptor and CD40 for Tumor Necrosis Factor Receptor-associated Factor (TRAF)-mediated Signaling*

Lymphotoxin-β receptor (LTβR) and CD40 are members of the tumor necrosis factor family of signaling receptors that regulate cell survival or death through activation of NF-κB. These receptors transmit signals through downstream adaptor proteins called tumor necrosis factor receptor-associated factors (TRAFs). In this study, the crystal structure of a region of the cytoplasmic domain of LTβR bound to TRAF3 has revealed an unexpected new recognition motif, 388IPEEGD393, for TRAF3 binding. Although this motif is distinct in sequence and structure from the PVQET motif in CD40 and PIQCT in the regulator TRAF-associated NF-κB activator (TANK), recognition is mediated in the same binding crevice on the surface of TRAF3. The results reveal structurally adaptive “hot spots” in the TRAF3-binding crevice that promote molecular interactions driving specific signaling after contact with LTβR, CD40, or the downstream regulator TANK.

Members of the tumor necrosis factor (TNF) 1 superfamily of cytokines and receptors initiate signal transduction pathways that regulate multiple cellular processes including cell survival or death (1), which are important processes shaping the development and homeostasis of many organs. The TNF receptors utilize at least two distinct mechanisms to activate pathways mediating cell death or survival. The intracellular signaling domains of some receptors in this family, such as Fas, contain a helical death domain that couples to the apoptotic machinery through interactions with adaptors, such as FADD (Fas-asso-ciated death domain), that also contain death domains. Other related TNF family receptors utilize adaptors in the TNFRassociated factor (TRAF) family of RING/zinc finger proteins (2) that mediate signal transduction through protein kinases, such as NF-B-inducing kinase, to activate transcription factors, such as NF-B.
The lymphotoxin-␤ receptor (LT␤R), a member of the TNFR superfamily, represents one of the TNF receptors that use TRAF adaptors. LT␤R mediates an essential signaling system for the development, organization, and differentiation of lymphoid tissue (3). This receptor is expressed on the surface of cells in the parenchyma and stroma of most lymphoid organs but is conspicuously absent on T-and B-lymphocytes. The LT␤R is activated by two distinct but related ligands: LT␣␤ (4) and LIGHT, which are transiently expressed on the surface of activated lymphocytes (5). Receptor ligation leads to activation of two distinct forms of NF-B and expression of genes involved in regulating immune processes that contribute to cell survival, cell migration, and other events, including expression of genes encoding chemokines, integrins and cytokines, B cell-activating factor, and interferon-␤, (6,7). Paradoxically, LT␤R induces apoptosis of some epithelial tumors (8 -10). The molecular mechanisms by which LT␤R signals control distinct cell fates are still inadequately understood.
The trimeric TNF family ligands initiate signal transduction by clustering receptors at the cell surface, followed by recruitment of different TRAF adaptors to the cytoplasmic domain of the receptor, thus promoting assembly of signaling complexes. LT␤R directly binds to several TRAFs, including TRAF2, TRAF3, and TRAF5 (9,11,12). TRAF2 and TRAF5 binding propagates signals leading to activation of NF-B (12). In contrast, binding of TRAF3 is implicated in the induction of apoptosis of tumor cells (9,10) and acts as a negative regulator of NF-B activation. Thus, determination of cell fate may occur, in part, at the level of recruitment of the adaptor proteins.
Six TRAF proteins have been identified in humans and mice, numbered sequentially TRAF1-6. TRAF2-6 contain two Nterminal zinc-binding domains, including a ring finger and five zinc finger motifs followed by a conserved TRAF domain. It is this conserved TRAF C-terminal domain that mediates binding to the cytoplasmic domains of TNFRs or signaling activators. TRAFs serve as co-inducers of downstream signaling, through a complex set of related and yet unique interactions with a range of binding affinities for the various TNFRs (13). For example, the well characterized CD40 receptor binds to TRAF2, TRAF3, and TRAF6 to control B cell proliferation, growth, and differentiation (14 -17). These TRAFs each recognize the motif PVQET in CD40. Similarly, the downstream regulator TANK (TRAF-associated NF-B activator or I-TRAF), which functions as a modulator of NF-B activation (18,19), contains a PIQCT recognition motif homologous to the TRAF-binding sequence in CD40 (18).
The crystal structures of the TRAF domain of TRAF3 in complex with the recognition motif PVQET from CD40 (20) and also in complex with the binding motif PIQCT from TANK (21) revealed that TANK and CD40 bind to the same binding pocket on the surface of the TRAF3 domain, supporting the hypothesis that TANK and CD40 compete for the TRAF site. Similarly for TRAF2, peptides from a number of TNFRs that contain a PXQXT or (P/S/T/A)X(Q/E)E motif bind in the homologous crevice (22).
The cytoplasmic domain of the LT␤R is 194 residues in length and contains a large proline-rich region (ϳ60 residues) near the C terminus that is responsible for initiating NF-B activation and apoptosis (23). A series of deletion mutants localized the binding site for TRAF2, TRAF3, and TRAF5, but the minimal binding motif 389 PEEGDPG 395 showed limited homology to the conserved recognition motif in other receptors that engage TRAFs, including CD40, HVEM, CD27, EDAR, or B cell-activating factor receptor. Thus, it was unclear how this sequence would establish molecular contacts to propagate signal transduction (13).
To develop a detailed understanding of the molecular mechanisms that drive TRAF-mediated signaling from the lymphotoxin-␤ receptor, we present here structural studies of TRAF3 in complex with a fragment of the cytoplasmic domain of LT␤R. The crystal structure clearly identifies the recognition motif that binds to the conserved TRAF domain. Comparison with the molecular interactions seen in the TRAF3/CD40 complex (20) reveals that the primary intermolecular contacts are made in the same surface binding crevice on TRAF3 that accommodates CD40 (20) and TANK (21). Mutational analyses of residues at the contact interface identify unique interactions that provide specific recognition of TRAF3 for LT␤R that are discrete from those that drive specific recognition for CD40 or TANK.

EXPERIMENTAL PROCEDURES
Crystallographic Analysis-TRAF3 used for this study was the shortened molecule, produced by tryptic cleavage that removed 36 residues from the N terminus. This truncated molecule was shown previously to produce crystals with marked improvement in diffraction (24). The crystals were grown as described (24) in hanging drops by vapor diffusion from reservoir solutions containing 15% polyethylene glycol 4000 in 0.1 mM MES, pH 6.5. The crystals formed at room temperature in space group P321 to a size of 500 ϫ 500 ϫ 25 m in 2-3 days.
A synthetic peptide corresponding to residues 385 PYPIPEEGDPGP-PGLSTPHQEDGK 408 from the cytoplasmic domain of LT␤R was soaked into TRAF3 crystals. The crystals were cryoprotected with 25% glycerol and flash frozen. Diffraction data were collected at the Stanford Synchrotron Radiation Laboratory beamline BL9-1 at Ϫ175°C using a MAR345 image plate detector. The data were processed using DENZO and SCALEPACK (25). A summary of the data collection statistics is presented in Table I.
The structure of the complex was first refined using the native truncated TRAF3 atomic coordinates. Refinement was carried out with simulated annealing in CNS (26). An iterative process of model building in the program O (27) and refinement in CNS was used to construct the model for the complex. After refinement F o Ϫ F c difference maps and OMITMAPS (28) were used to fit the peptide. Clear electron density was visible for backbone atoms in the early maps to position the peptide at the binding crevice on TRAF3. The final model included all of the residues in the peptide. Refinement statistics are presented in Table I. The R factor and R free were 26.6% and 31.5%, respectively. Molecular images for the figures and electrostatic surfaces were prepared with MOLMOL (29) and SPOCK (30).
Peptide Synthesis-The LT␤R peptide was synthesized with Fmoc (N-(9-fluorenyl)methoxycarbonyl) chemistry on Rink's amide methoxybenzhydrylamine resin using an Advanced Chemtech 350 multiple peptide synthesizer. The peptide was acetylated at the N terminus and amidated at the C terminus. The peptide was cleaved from the resin and deprotected with 95% trifluoroacetic acid, 2.5% H 2 O, and 2.5% triisopropylamine. After cleavage, the crude peptide was precipitated with ice-cold diethyl ether and purified by high pressure liquid chromatography on a reverse phase C18 Cosmosil column developed with a water, acetonitrile, 0.1% trifluoroacetic acid gradient. The final product was analyzed and confirmed by mass spectrometry.
Construction of LT␤R and TRAF3 Mutants-Site-specific mutations were introduced into the cDNA sequences of human LT␤R and TRAF3 in the pCDNA3-mycLT␤R (10) and pSG5-FLAG-TRAF3 vectors (31), respectively, using the QuikChange site-directed mutagenesis kit (Stratagene). The following mutants in the cytoplasmic domain of LT␤R were made: P387A, P389A, D393A, E390A/E391A, and E390A/E391A/ D393A. For TRAF3, the following single mutants were made: R393A, Y395A, L432E, F448E, S454A, S455A, S456A, and F457E. The residue numbers for TRAF3 correspond to those used in the report of the crystal structure of TRAF3 (20). The cytoplasmic domain of LT␤R or the LT␤R mutants was PCR-cloned into the pGEX4T-1 vector (Amersham Biosciences) to create GST-LT␤R fusion proteins using the following primers: 5Ј-ggaattcaagagccacccttctctctgc-3Ј and 5Ј-cctcgaggtcagtcatgggtgataaattgg-3Ј containing EcoRI and XhoI sites, respectively. Each construct was verified by DNA sequencing.
Protein Expression-GST and GST fusion proteins were expressed in bacteria (BL-21) from pGEX plasmids and purified using glutathione-Sepharose 4B beads (Amersham Biosciences). Briefly, bacterial cultures in the log phase of growth were induced with 0.1 mM isopropyl-␤-D-thiogalactopyranoside for 4 h at 25°C. For some mutants, induction time was reduced to 1 h to prevent the formation of protein aggregates. The bacterial pellets were resuspended in lysis buffer (1% Triton X-100, 1.5% Sarkosyl, 0.5 mM phenylmethylsulfonyl fluoride, and 10 mM dithiothreitol in PBS) and sonicated on ice to lyse the cells. The lysates were cleared by centrifugation and incubated with glutathione-Sepharose 4B beads for 2 h at 4°C, and the beads were washed once with PBS, twice with PBS and 0.1% Triton X-100, and twice with co-immunoprecipitation buffer (10 mM Hepes, pH 7.6, 142 mM KCl, 2.5 mM MgCl 2 , 1 mM EGTA, 1 mM dithiothreitol, 0.2% Nonidet P-40, and protease inhibitors) and resuspended in co-immunoprecipitation buffer. GST fusion proteins were analyzed by SDS-PAGE/Coomassie Blue staining and normalized to ϳ1 g/5 l by adding "empty" beads. GST-CD40 and GST-TANK (TRAF-interacting motif) were expressed from the pGEX-CD40 (32) and pGEX-TANK (21), respectively. Binding of LT␤R to TRAF3 was tested by isothermal titration calorimetry, but the binding was weak and not measurable by the calorimeter. Therefore in vitro binding assays were utilized.
In Vitro Binding Assays-For the protein binding assays, the lysates were prepared from 293T cells transfected by the calcium phosphate method with pSG5-FLAG-TRAF3 or the FLAG-tagged TRAF3 mutants, pSG5-FLAG-TRAF2 (33), or pCEP-hemagglutinin-TRAF5 (34) in coimmunoprecipitation buffer (0.5 mg of total protein/ml) and precleared with glutathione-Sepharose beads. Immobilized GST fusion proteins (ϳ5 g) were added and incubated for 1 h at 4°C. The beads were washed twice with detergent buffer (10 mM Hepes, pH 7.6, 142 mM KCl, 2.5 mM MgCl 2 , 1 mM EGTA, 1 mM dithiothreitol, 0.2% Nonidet P-40, and protease inhibitors), washed twice with PBS, and resuspended in 25 l of Laemmli buffer. The proteins were analyzed by SDS-PAGE and immunoblotting with anti-FLAG antibody (M2; Sigma) for TRAF2 and

RESULTS AND DISCUSSION
TRAF3/LT␤R Complex-As reported earlier, the TRAF domains of TRAF3 (20), TRAF2 (35,36), and TRAF6 (37) share a conserved folding pattern that incorporates an independently folded C-terminal domain (residues 348 -504). This domain folds into an eight-stranded ␤-sandwich that is formed from two layers of ␤-sheet that each contain four antiparallel strands. TRAF3 is an oligomeric molecule that functions as a trimer in solution. The trimer is stabilized by coiled-coil interactions (10) between the long helices that are N-terminal of the conserved TRAF C-terminal domain (Fig. 1). In the crystal, one monomer is the asymmetric unit, and the three subunits are related by 3-fold crystallographic symmetry.
For crystallization, we used tryptic digestion to shorten the N-terminal helices by 36 residues, and this procedure markedly improved the diffraction properties of the crystals (21,24). In the crystal lattice, the predominant intermolecular contacts involve the TRAF C-terminal ␤-sandwich domain, and the crystals are susceptible to twinning. In addition, large solvent channels exist along the length of the extended N-terminal helices. Tryptic digestion truncates these helices, thereby reducing the solvent content of the crystals and eliminating the twinning. To form the complex, a synthetic peptide encompassing the TRAF-binding region of the cytoplasmic domain of LT␤R was dissolved and diffused into the TRAF3 crystals. The peptides bound to a surface crevice on the ␤-sandwich domain with a stoichiometry of 1:1, as shown in Fig. 1. Because of the crystallographic symmetry, the structure of the peptide is identical in each of the three sites in the complex. The LT␤R sequence represented in the synthetic peptide was 385 PYP-IPEEGDPGPPGLSTPHQEDGK 408 . After the final refinement, all of the residues were ordered and clearly defined in the electron density (Fig. 2). The "alternating pattern" of large and small side chains in the sequence assisted the unambiguous positioning of the LT␤R sequence at the TRAF3-binding crev-ice. In particular, strong density in the initial maps for the side chain of Asp 393 dictated the assignment of this density to the aliphatic portion of the aspartic acid, whereas assignment of the flanking Gly 392 or Pro 394 was clearly not appropriate. Similarly, a large region of electron density directed the positioning of the side chain for Tyr 386 rather than the two adjacent prolines. After the final cycles of refinement, the electron density was clear for the polypeptide backbone as well as 11 residues with large side chains.
The long LT␤R peptide was located in a restricted solvent "cave" that exists at the binding crevice on TRAF3. The overall dimensions of this cave are 15 ϫ 19 ϫ 22 Å, and the space in this portion of the crystal lattice is sufficient to accommodate the large peptide so that the conformation observed in the complex reflects the actual conformation and is not an artifact of crystal packing (20,21).
Structure of LT␤R-The results of the refinement are presented in Table I. When LT␤R binds to TRAF3, this region of the cytoplasmic domain assumes a reverse turn configuration, with the change in direction occurring through residues 394 PG-PPG 398 . The two flanking strands are in an extended conformation, and there are no intramolecular contacts made between these strands. The N-terminal sequence of the fragment binds at the same binding crevice where other TNFRs bind to both TRAF3 (20) and TRAF2 (22,36). All of the intermolecular interactions between LT␤R and TRAF3 are made with residues 388 IPEEGD 393 . (One set of intermolecular contacts involving the C-terminal strand of this segment of LT␤R is observed with a symmetry-related TRAF3 molecule in the crystal lattice.) Comparison of TRAF3 Binding Motifs in LT␤R and CD40 -LT␤R and CD40 both bind to TRAF3 and propagate cellular signals. These receptors bear no sequence or structural resemblance in their cytoplasmic domains. However, their recogni- tion motifs bind to the same binding crevice on the surface of TRAF3 despite the fact that the motifs are presented in completely different folding patterns. In Fig. 3, the structure of the TRAF3-binding fragment of LT␤R is compared with the corresponding segment of CD40. Although both structures can loosely be described as reverse turns, the structural resemblance is obscure. It was known from previous work that the cytoplasmic domain of CD40 contains little secondary structure yet assumes an ordered structure when bound to TRAF3 (20). Given the large number of prolines in the cytoplasmic region of LT␤R (33 of 193 residues), it is possible that this domain similarly lacks defined secondary structure. Yet, when bound to TRAF3, this segment does fold in a defined manner as seen in the complex in the crystal. In CD40, an intramolecular contact involving Thr 254 established the configuration of the reverse turn and thus the orientation of a TRAF3-specific intermolecular contact (20). In LT␤R, no intramolecular interactions within the peptide were observed that stabilize the folding of the LT␤R fragment, and no contacts were seen with TRAF3 other than those at the recognition motif.
LT␤R/TRAF3 Interactions-The residues in the first strand of the LT␤R reverse turn represent the critical contacts for binding, and thus the recognition motif for TRAF3 is structurally defined as IPEEGD. These contacts are made in a shallow binding crevice at the edge of the ␤-sandwich. The intermolecular contacts in this binding crevice are quite similar to those seen when CD40 or TANK binds in complex with TRAF3 (20,21), even though the sequences of the recognition motifs are quite different from the LT␤R motif. In CD40, binding to TRAF3 is through contacts from the PVQET motif, and for TANK, recognition is made through a PIQCT motif.
The detailed molecular interactions of this motif with TRAF3 are shown in Fig. 4. In LT␤R, Ile 388 and Pro 389 are accommodated in a hydrophobic pocket lined by TRAF residues Phe 448 and Pro 471 , and the aliphatic portion of the Asp 451 side chain.
The side chain carboxyl group of Glu 390 makes a hydrogen bond with the main chain carbonyl of Ala 467 . This hydrogen bond may be critical to stabilize the orientation of LT␤R when docked to TRAF3. The carboxyl group of Glu 390 is also in position to make a hydrogen bond with the hydroxyl of Ser 456 . A salt bridge is formed between the side chains of Glu 391 and Arg 393 on TRAF3. Finally, a critical hydrogen bond is formed between the side chain of Asp 393 and TRAF3 Tyr 395 .
LT␤R/TRAF3 Recognition-The contacts seen in the crystal structure of the complex were tested by site-directed mutagenesis and protein binding assays. The residues were selected based on inspection of the model of the complex, and clusters of substitutions were made to define the critical residues in the 388 IPEEGD 393 recognition motif. The carboxylate residues in the sequence were changed to alanine and tested for binding to TRAF3. When alanine was substituted for the tandem glutamic acids (E390A,E391A), binding to TRAF3 was diminished (Fig.  5a). However, when Asp 393 was mutated to alanine along with the substitution of alanine for the tandem glutamic acid residues (E390A,E391A,D393A), binding was completely abolished. Thus, the three carboxylate side chains play important roles in TRAF3 recognition, and in this context, the contact of Asp 393 with Tyr 395 on TRAF3 appears to enhance tight binding. Interestingly, as shown in Fig. 5, mutant LT␤R with substitution of alanine for this aspartic acid (Asp 393 ) alone retained binding activity. In murine LT␤R, the corresponding sequence is IPEEGA, yet all the critical contact residues in TRAF3 that mediate binding of LT␤R are absolutely conserved in murine and human proteins. Accordingly, we would anticipate that mutation of Glu 390 -Glu 391 in murine LT␤R would be sufficient to ablate binding to TRAF3.
Before the structural studies, it was difficult to predict the actual TRAF3 recognition motif in this region of the cytoplasmic domain of LT␤R ( 385 PYPIPEEGDPG 395 ) from inspection of the sequence alone because homology could be inferred in at least two modes with the consensus sequence PXQXT, which is present in TNFRs that bind to TRAFs. For example, the sequences PIPEE or PEEGD each bear similarities to the recognition motifs seen in TNFRs or TANK (21,22). Thus, the recognition motif observed in the complex was unexpected and

FIG. 3. Comparison of the TRAF-binding regions of LT␤R and CD40.
For this image, the TRAF3 recognition motifs in each molecule were superimposed to maximize the structural homology in these molecular features. In this view, the TRAF3-binding crevice would be to the right. The molecules are represented by ball-and-stick models, with LT␤R on the left and CD40 on the right. Although both of the molecules assume reverse turn configurations, the comparison reveals that the recognition motifs are embedded in completely different structural contexts in the two receptors. has now been confirmed by the mutagenesis experiments. Substitution of alanine for Pro 387 or Pro 389 did not affect binding to TRAF3 (Fig. 5a), which might have been expected if one of the sequences PIPEE or PEEGD were, in fact, the recognition motif. The motif has been defined as IPEEG, but an additional critical contact at Asp 393 suggests that the hexapeptide sequence IPEEGD is required for LT␤R/TRAF3 recognition.
LT␤R Binding to TRAF2 and TRAF5-Binding of LT␤R to TRAF2 and TRAF5 was expected to be closely similar to the interactions seen in the complex of this receptor with TRAF3. The critical contact residues at the TRAF3/LT␤R binding interface are identical in TRAF2, TRAF3, and TRAF5. In fact, when the LT␤R mutants were tested for binding to TRAF2 and TRAF5, the results showed interesting distinctions (Fig. 5a). For example, only TRAF5 lost binding when Pro 387 was changed to alanine.
Furthermore, TRAF5 binding was significantly reduced by mutation of the two adjacent glutamates, Glu 390 -Glu 391 . In contrast, although TRAF3 binding was somewhat reduced by this tandem mutation, binding of TRAF2 was unaffected. Even when binding to TRAF3 was completely abolished by the triple LT␤R mutant E390A/E391A/D393A, this mutant retained some binding activity for TRAF2. We previously noted some differences in the mode by which TRAF3 and TRAF2 bind to CD40 (20). In that study, a critical contact was made between Tyr 395 in TRAF3 and CD40, but in TRAF2, this tyrosine participates in another set of intermolecular interactions with CD40 (22,36). Asp 393 in LT␤R forms a hydrogen bond with Tyr 395 in TRAF3. Perhaps the differences in sensitivity to substitution of Asp 393 may involve this tyrosine. For TRAF5, there is no crystal structure available, so the interpretation of distinctions in recognition must await future structural studies. Clearly, the binding interfaces for LT␤R on TRAF2, TRAF3, and TRAF5 are located in the surface crevice of the corresponding TRAF domain, but the mutagenesis results suggest that the molecular contacts are not identical.
Contact Residues in the TRAF3 Binding Crevice for TNFRs-The binding crevice on TRAF3 where LT␤R and other TNFRs bind contains three principal subregions, as illustrated in Fig.  6. To further test the contacts seen in the crystal structure, TRAF3 residues that line the binding crevice were mutated, and the resultant mutant proteins were tested for binding to LT␤R. In the first subregion, Arg 393 and Tyr 395 make critical contacts with Glu 391 and Asp 393 in the recognition motif and were shown by mutagenesis to be essential for LT␤R recognition. Importantly, when Arg 393 or Tyr 395 in TRAF3 were changed to alanine (Fig. 5b), binding was reduced (R393A) or completely lost (Y395A). This result directly parallels the response when their interacting partners in LT␤R were mutated. Arg 393 forms a salt bridge with carboxylate OE2 of Glu 391 in LT␤R, representing an important intermolecular contact. In the case of Tyr 395 , the phenolic hydroxyl of this TRAF3 residue participates in a hydrogen bond with the side chain carboxylate OD2 of Asp 393 in LT␤R. When one of the interacting side chains is mutated in either of the molecules, binding of the partner molecule is reduced or totally lost. Thus, these contacts are critical for LT␤R/TRAF3 recognition.
The second subregion of the TRAF3 crevice involves a cluster of three serines that was first described for TRAF2/CD40 interactions ("serine tongs") (36). These serines have the potential in both TRAF2 and TRAF3 to form hydrogen bonds with the third residue in the recognition motif of CD40. In the TRAF3/LT␤R complex, the side chain of Glu 390 is within hydrogen-bonding distance of Ser 456 in the cluster. Binding of LT␤R was not affected when alanine was substituted for these serines (Fig. 5b). The results suggest that contact with the hydrophilic serine cluster may not be critical for binding.
The third subregion is lined by five hydrophobic residues. The N-terminal residue in the recognition motif is accommodated in this pocket. In LT␤R, that residue is Ile 388 . When Phe 448 or Phe 457 in the hydrophobic pocket were mutated to glutamic acid (F448E,F457E), binding of LT␤R was blocked completely (Fig. 5b). In the wild type TRAF3, the side chains of these two phenylalanines are within van der Waals' distance of the side chain of Ile 388 in LT␤R. The introduction of a charged residue into this hydrophobic pocket apparently changes the character of the environment so that the receptor domain no longer binds. It is interesting to note that this hydrophobic interaction is critical for recognition, because substitution for just one of the phenylalanines disrupts binding. Another hydrophobic residue in the pocket, Leu 432 , is located more than 8 Å from Ile 388 . Substitution of glutamic acid for this leucine does not affect binding of LT␤R. Thus, two intermolecular contacts have been defined that promote recognition of LT␤R by TRAF3, one being a network of salt-bridged and hydrogen-bonded contacts and the second involving hydrophobic interactions that permit a deep penetration of the LT␤R motif deep into the TRAF3 crevice.
Structural Basis for Specific Binding of LT␤R, CD40, or TANK in the TRAF3 Crevice-To probe the molecular basis for specific recognition, we tested the TRAF3 mutants for binding to CD40 and TANK, for which we had crystal structures in hand (20,21). As shown in Fig. 5b, three TRAF3 mutations caused reduction or loss of binding to each of these three proteins: Y395A, F448E, and F457E. The Y395A TRAF3 mutant failed to bind to LT␤R, and its ability to interact with CD40 and TANK was dramatically reduced. Tyrosine 395 participates in critical hydrogen bonds in each of the complexes. Although mutation of the clustered serines in TRAF3 did not affect binding of LT␤R, two of these serines were shown to be FIG. 5. Mutations in LT␤R at contact residues alter binding to TRAF3. a, lysates from 293T cells overexpressing FLAG-tagged TRAF3, FLAG-tagged TRAF2, or hemagglutinin-tagged (HA) TRAF5 were incubated with GST-LT␤R mutant proteins immobilized on glutathione-Sepharose beads. The bound proteins were analyzed by SDS-PAGE and immunoblot analysis using antibodies specific to the tag portion of the TRAF proteins. Binding was compared with wild type GST-LT␤R as a positive control or GST alone or GST-TNFR1 as negative controls. Single mutants P387A, P389A, and D393A still bound to TRAF3. b, mutation of contact residues in TRAF3 affects binding to LT␤R, CD40, and the TRAF-interacting motif domain of TANK. Lysates from 293T cells overexpressing mutants of FLAG-tagged TRAF3 were incubated with wild type GST-LT␤R, GST-CD40, or GST-TANK (TRAF-interacting motif) immobilized on glutathione-Sepharose beads, and the bound proteins were analyzed as in a.
important for CD40 or TANK binding. (Of course, it is possible that substitution of a single serine within this cluster simply promotes an alternate hydrogen-bonded or water-mediated arrangement with another serine in the cluster.) All three proteins interact with Tyr 395 through hydrogen bonds. This appears to be a critical contact for LT␤R but may not be absolutely essential for binding CD40 or TANK. In contrast, the integrity of the hydrophobic pocket is critical for binding of LT␤R, CD40, and TANK, as evidenced by the complete loss of binding when Phe 448 and Phe 457 are substituted by glutamic acid. The change in character of the binding pocket by this mutation and its drastic effects on binding demonstrate that hydrophobic interactions are key in TRAF-mediated signaling complexes.
Polar or hydrophilic interactions may modulate specific interactions with individual TNFRs or regulators. Our results indicate that mutations of key sites in these two subregions variably affect binding to LT␤R, CD40, or TANK. When Arg 393 is changed to alanine, binding is most dramatically affected in CD40. Although hydrogen bonds to this arginine are possible in LT␤R or TANK, no hydrogen bond was made in the CD40/ TRAF3 complex. The dramatic effect seen in the present study is therefore associated with an accompanying structural change involving TRAF3 surface residues. Future studies will define more clearly the role of Arg 393 in CD40 recognition by TRAF3. Interestingly, the role of the serine cluster in recognition of specific binding partners is highlighted by the mutational analyses. Substitution of alanine for Ser 454 did affect binding to TANK, whereas mutation of Ser 456 to alanine caused nearly complete loss of binding of either CD40 or TANK. In the CD40/TRAF3 complex, the potential for hydrogen bonding exists for Ser 454 with Glu 252 in CD40. In the TANK/TRAF3 complex, a hydrogen bond is formed between Ser 454 and NE2 in Gln 182 in TANK. The results of both the arginine and serine cluster mutations suggest that other configurational changes may occur in the binding crevice when a hydrophilic residue is changed to the small hydrophobic alanine. Indeed, the results suggest that the TRAF3-binding crevice is structurally adaptive.
Structurally Adaptive "Hot Spots" in the TRAF3 Binding Crevice for TNFRs-LT␤R-mediated signaling through the NF-B pathway is propagated through binding to TRAFs. Here we have shown that LT␤R, CD40, and the downstream regulator TANK each bind to the same crevice on TRAF3. This observation stimulates an important question regarding the specificity of binding recognition and the trigger of the TRAF3 signaling response. These three molecules bind at the same surface pocket, and similarly, peptides bearing the motif FIG. 6. Hot spots in the TRAF3 protein interaction binding crevice. The solvent-accessible surface of the TRAF ␤-sandwich domain is shown. The view is from the side of one subunit in the same view as Fig. 1. In this orientation the cell membrane is located at the top of the image, and the N-terminal helix that participates in the coiled-coil interactions upon trimer formation is at the bottom of the image. (The helix is clipped in the image.) The binding crevice is at the edge of the ␤-sandwich, and the view here is directly into the crevice. The area that contains the contact residues for TNFRs or TANK regulator proteins is enclosed in a box. In A, the surface is colored according to hydrophobicity. The yellow color intensity is proportional to increasing hydrophobic character. This binding crevice is mainly hydrophobic in nature. In B, the area that is boxed in A is now colored to highlight three separate hot spots for protein-protein interactions; each is colored separately. Critical residues in each hot spot are labeled and correspond to TRAF3 residues that were mutated in this study (see Fig. 5). The first hot spot is polar in nature with two charged residues and is colored red. The second hot spot is hydrophilic (colored green) with a serine cluster. The last hot spot, which is hydrophobic in character, is shown in yellow. C, TRAF3 protein-protein interface is structurally adaptive. The TRAF3 binding crevice is compared as it exists when bound in complex with LT␤R (left panel), CD40 (middle panel) (20), or TANK (right panel) (21). (The view into the binding crevice is as in B.) In each panel, the binding surface is colored according to electrostatic potential and is presented in the same orientation for comparison. Areas with negative, positive, or neutral character are depicted in red, blue, or white, respectively. This comparison illustrates how the flexibility and adaptability of contact side chains in the TRAF3 binding crevice alter the overall shape and character of the interface. The side chains that contact the partner protein undergo conformational adjustments when binding to TNFRs or TANK regulatory protein.
PXQXT from a number of TNFRs bind to the homologous crevice on TRAF2 (22). Is it possible that TRAF molecules are structurally adaptive and flexible at the protein/protein interface? We have defined hot spots in the TRAF3 protein interface (Fig. 6, A and B) corresponding to residues that provide the same principal contacts for each of several different binding partners (38). In a comparison of the configurations of conserved residues at these hot spots, we noted that adaptability is mediated by adjustments of side chains, particularly in the orientation of the polar residues that contact the interacting molecule directly; i.e. Asp 399 , Arg 393 , and Tyr 395 . In the comparison, substantial movements of phenylalanines also occur in the hydrophobic subregion of the binding crevice. The conformational adjustments do not involve significant movements of the polypeptide backbone, as reflected by the fact that the root mean square deviation for TRAF3 ␣-carbon atoms in the three complexes was 0.88 Å (38). Yet the character of the binding surface changes demonstrably in response to different binding partners. The electrostatic surface of TRAF3 as it exists when bound in complex with LT␤R, CD40, or TANK is compared in Fig. 6C. Such adaptations at the protein interaction interface are known in proteins that bind functionally diverse partners (39 -41). The "plasticity" or "flexibility" of protein residues is apparent in the molecular interactions (42,43) and may influence binding affinity. In the case of LT␤R, signaling through two different NF-B pathways (7) may involve similar adaptations at the interface, with distinct responses from the adaptable TRAF molecule. At least 18 different receptors in the TNFR family can engage TRAFs, suggesting that selection of an adaptable binding crevice in TRAF3 may be a parallel evolutionary event to compensate for gene duplicative mechanisms driving expansion of the TNFR family. The flexibility of the TRAF-binding site may also represent an advantageous evolutionary adaptation, serving as a rather rigorous defense against viral pathogens that target TNFR signaling pathways (44). The answer awaits more structural studies on complexes of linked signaling molecules in the pathway.