LMP1 Protein from the Epstein-Barr Virus Is a Structural CD40 Decoy in B Lymphocytes for Binding to TRAF3 *

Epstein-Barr virus is a human herpesvirus that causes infectious mononucleosis and lymphoproliferative malignancies. LMP1 (latent membrane protein-1), which is encoded by this virus and which is essential for transformation of B lymphocytes, acts as a constitutively active mimic of the tumor necrosis factor receptor (TNFR) CD40. LMP1 is an integral membrane protein containing six transmembrane segments and a cytoplasmic domain at the C terminus that binds to intracellular TNFR-associated factors (TRAFs). TRAFs are intracellular co-inducers of downstream signaling from CD40 and other TNFRs, and TRAF3 is required for activation of B lymphocytes by LMP1. Cytoplasmic C-terminal activation region 1 of LMP1 bears a motif (PQQAT) that conforms to the TRAF recognition motif PVQET in CD40. In this study, we report the crystal structure of this portion of LMP1 C-terminal activation region-1 ( 204 PQQATDD 210 ) bound in complex with TRAF3. The PQQAT motif is bound in the same binding crevice on TRAF3 where CD40 is bound, providing a molecular mechanism for LMP1 to act as a CD40 decoy for TRAF3.

The success of viral infection depends on effective evasion of the cell death machinery of the host. This is a formidable task for the pathogen because the response to infection is complex in mammals. The immune response to viruses may involve apoptosis, or, in some cases, the host defense may incorporate the expression of survival and pro-inflammatory genes to avoid the serious side effects associated with the apoptotic response. One evasion tactic used by the oncogenic herpesviruses is a virally encoded ortholog of the anti-apoptotic regulator Bcl-2. For example, Epstein-Barr virus (EBV) 4 encodes two Bcl-2 orthologs (BHRF1 and BALF1) (1,2), and Kaposi sarcoma-associated ␥-herpesvirus murine hepatitis virus type 8 expresses Kaposi sarcoma bcl-2 (3,4). These viral proteins block release of cytochrome c from mitochondria, an early and essential step in the apoptotic cascade. Viral mimicry is also evident when mammalian viruses express orthologs of the IAPs (inhibitors of apoptosis proteins) that regulate caspase enzymes to initiate and effect protein degradation and ultimately cell death. African swine fever virus encodes the viral IAP ortholog (5), and serpin CrmA derived from poxviruses (6 -8) inhibits several caspases.
The tumor necrosis factor family of cytokines and their receptors play a key role in regulating both the innate and adaptive immune responses to viral pathogens. Not surprisingly, therefore, the regulatory elements in cell death or tumor necrosis factor pathways are frequently targeted by viruses in death escape strategies (reviewed in Refs. 9 and 10). Viral proteins have evolved that modulate the tumor necrosis factor receptors (TNFRs) or, in some cases, even mimic the receptors or associated signaling molecules.
EBV is a human herpesvirus that causes infectious mononucleosis and lymphoproliferative malignancies such as AIDS-related lymphoma, Burkitt lymphoma, Hodgkin disease, and nasopharyngeal carcinoma (11)(12)(13)(14)(15). LMP1 (latent membrane protein-1) is encoded by this virus and is essential for transformation of B lymphocytes (16). This protein acts as a constitutively active mimic of the TNFR CD40 (17)(18)(19). CD40 is expressed on B lymphocytes and, after ligation, activates B cells by interaction with intracellular TNFR-associated factors (TRAFs) (20). This association leads to signaling through NF-B and JNK pathways to activate expression of anti-apoptotic genes.
There are six characterized TRAF proteins numbered sequentially 1-6. In TRAF2-6, two N-terminal zinc-binding domains that bear a ring finger and five zinc finger motifs are followed by a conserved TRAF domain at the C terminus. The TRAF domain mediates binding to the cytoplasmic portion of TNFRs or signaling activators. TRAFs are coinducers of downstream signaling with a range of binding affinities for various TNFRs (21). For example, CD40 binds to TRAF2, TRAF3, and TRAF6 to control B cell proliferation, growth, and differentiation (Refs. [22][23][24][25]reviewed in Ref. 26). Binding recognition is mediated by specific contacts of a recognition sequence by residues in a hydrophobic crevice on the TRAF domain. The specific contacts of three TNFRs have been defined in crystal structures of CD40, the lymphotoxin ␤-receptor (LT␤R), and the BAFF receptor (BAFF-R) in complex with TRAF3 (27)(28)(29) or with peptides of the motifs in complex with TRAF2 (30). From these structural studies, a recognition motif has been revealed in a shared consensus sequence, PXQXT or (P/S/T/A)X(Q/E)E. The binding motifs from CD40, LT␤R, and BAFF-R as well as the downstream regulator TANK (31) are each accommodated in the same binding crevice on TRAF3, and this binding interface is structurally and functionally adaptive (32).
LMP1 closely mimics signaling events and effector functions of CD40 in B lymphocytes, including activation of NF-B and JNK, modulation of adhesion and co-stimulatory molecules, and secretion of antibodies and cytokines (reviewed in Refs. 22 and 33). LMP1 is an integral membrane protein (386 residues) that contains a short cytoplasmic N-terminal region (residues 1-24), six transmembrane segments (residues 25-186), and a cytoplasmic C-terminal tail (CCT; residues 187-386) (34). The N-terminal region is essential for insertion into the membrane. The transmembrane segments oligomerize within the membrane and mediate constitutive activation. The CCT contains two subdomains that have been implicated in LMP1 signaling, C-terminal activation region (CTAR) 1 (residues 194 -232) and CTAR2 (residues 351-386) (35), and these two regions play distinct yet overlapping roles in EBV-associated lymphoproliferation (36). The CCT of LMP1 binds to TRAF1-3 and TRAF5 by recognition of a sequence in CTAR1 ( 204 PQQAT 208 ) that conforms to the consensus sequence PXQXT for TRAF recognition. In contrast, CTAR2 has been proposed as a site for binding the TNFR-associated death domain protein (TRADD) and receptor-interacting protein (RIP) 37,38). In addition, CTAR2 influences the recruitment of TRAFs to membrane rafts by CTAR1 (36). Interestingly, we have shown previously that TRAF3 mediates signaling through direct interactions with CTAR1 and may also play a role in the interactions between CTAR1 and CTAR2 (36). Moreover, our data indicate that TRAF3 is required and essential for certain CTAR1-mediated effects in B cells, the principal target of EBV.
LMP1-mediated signaling is critical for EBV-associated pathogenesis, and we have shown recently that TRAF3 is required for LMP1 activation of B cells (39). Although this activation mimics that effected by CD40 and although both LMP1 and CD40 bind to TRAF3, experiments in B cell lines deficient in TRAF3 have shown that LMP1 and CD40 differentially use TRAF3 and that TRAF3 is required for LMP1 signaling. LMP1 signaling remains intact in TRAF2-deficient B cells (36,39). To understand the molecular basis for binding of LMP1 to TRAF3, we crystallized a peptide representing part of CTAR1 of LMP1 and bearing the TRAF recognition motif in complex with TRAF3. The structure reveals the intermolecular contacts, and we report the direct comparison of these contacts with those seen in a crystal structure of the CD40-TRAF3 complex (27).

EXPERIMENTAL PROCEDURES
Crystallographic Analysis-For co-crystallization of LMP1 with TRAF3, crystals of TRAF3 were grown as described previously (40) after tryptic digestion of the protein to shorten the long N-terminal helix. Crystals formed in space group P321 in hanging drops from solutions of 0.1 M MES (pH 6.5) containing 15% polyethylene glycol 4000. The crystals grew to a size of 500 ϫ 500 ϫ 25 m at room temperature and diffracted to 2.7-Å resolution.
To form the complex, synthetic peptides of various lengths representing the minimal region in CTAR1 of LMP1 implicated for TRAF3 recognition were soaked into TRAF3 crystals. Each peptide contained the PQQAT TRAF recognition motif of LMP1, but differed in the number of flanking residues. The peptides were 202 PHPQQATDDSGHESDSN-SNEGRHH 225 , 202 PHPQQATDDSGHESDSNSN 220 , and 203 HPQQAT-DD 210 . Each peptide was tested in separate experiments. Peptides were dissolved in water and soaked into TRAF3 crystals, and then the crystals were cryoprotected with 25% polyethylene glycol and 20% glycerol and flash-frozen for data collection. Diffraction data were collected at the Stanford Synchrotron Radiation Laboratory beamline 11-3 at Ϫ175°C using a Q4 image plate detector. The data were processed using DENZO and SCALEPACK (41). The data collection statistics are summarized in TABLE ONE.
The structure of the complex was refined using the atomic coordinates of native truncated TRAF3 (40) implementing simulated annealing in CNS (59). An iterative process of refinement in CNS and model building in the program O (42) was used to construct the model of the complex. After refinement, difference maps (F o Ϫ F c and 2F o Ϫ F c ) and OMITMAPS (43) were used to fit the peptide. Clear electron density was visible for only the shortest peptide, in the TRAF3 binding crevice for backbone atoms. After several rounds of refinement, annealing, and model adjustment, all atoms were clearly placed in density for residues 204 -210. These residues were included in the final model. Refinement statistics for the complex are presented in TABLE ONE. For the final structure, the R factor and R free values were 20.6 and 25.6%, respectively. Graphic images and electrostatic surfaces presented in the figures were prepared with MOLMOL (44) and SPOCK (45).
Peptide Synthesis-Peptides for the complex were designed to correspond to the sequence in LMP1 that contains the binding site for TRAF3. The peptides acetyl-HPQQATDD-amide, acetyl-PHPQQAT-DDSGHESDSNSN-amide, and acetyl-PHPQQATDDSGHESDSNSN-EGRHH-amide were synthesized using Fmoc (N-(9-fluorenyl)methoxycarbonyl) chemistry with diisopropylcarbodiimide/hydroxybenzotriazole coupling on Rink's amide (methoxybenzhydrylamine) resin with an Advanced ChemTech 350 multiple peptide synthesizer. The peptides were cleaved from the resin and deprotected by treatment with trifluoroacetic acid/water/triisopropylsilane (95:2.5:2.5) for 2 h at room temperature. The cleaved peptides were precipitated and washed with cold diethyl ether. After drying, the peptides were dissolved in aqueous acetonitrile and purified on a preparative C 18 column (Cosmosil 5C18-AR, 20 ϫ 250 mm; Phenomenex, Torrance, CA) with detection at 210 nm using a Gilson HPLC apparatus. The peptides were separated from impurities using a linear gradient of 0 -40% solvent B over 40 min (solvent A ϭ 0.1% trifluoroacetic acid in water and solvent Bϭ 0.1% trifluoroacetic acid in 90% acetonitrile) at a flow rate of 8 ml/min. Pure peptides, as judged by their elution as single peaks by HPLC on analy- tical C 18 columns (Vydac 218TP54 and 238TP54, each 5 m, 4.6 ϫ 250 mm), were confirmed by matrix-assisted laser desorption ionization time-of-flight mass spectrometric analysis with an Applied Biosystems Voyager System 6264. DNA Constructs-DNA oligonucleotide primers were obtained from Integrated DNA Technologies (Coralville, IA). The DNA construct used to stably express human CD40 (hCD40)-LMP1 has been described previously (46). The hCD40-LMP1 mutants P204A, Q206A, T208A, D209A, and D210A were generated by PCR splicing by overlap extension (SOEing) (47) using hCD40-LMP1 as the template, and hCD40 -5Ј Stable Transfection of the Mouse B Cell Line-The mouse B cell line M12.4.1 has been described previously (48,49). Cells were cultured in RPMI 1640 medium supplemented with 10% fetal calf serum, 10 M ␤-mercaptoethanol, and antibiotics as described (18). M12.4.1 cells were stably transfected by electroporation as described (48). Cell lines transfected with hCD40-LMP1 or its mutants were selected and maintained with 400 g/ml Geneticin (Invitrogen). Surface expression of hCD40-LMP1 and its mutants was determined by immunofluorescence flow cytometry, and expression-matched clones were selected for experiments as described previously (46).

TRAF Recruitment to hCD40-LMP1 and Its Mutants in Detergentinsoluble Microdomains (Rafts) and Co-immunoprecipitation-Binding
of LMP1 to TRAF3 was tested in cell-based assays, but was too weak to permit measurements by isothermal titration calorimetry, as we have seen with other TRAF3 complexes (LT␤R or BAFF-R) (28,29). Instead, assays were developed that mimic the in vivo binding situation and that provide the capability to measure significant differences in binding. Stably transfected M12.4.1 subclones (2 ϫ 10 7 cells) were stimulated in a total volume of 1 ml with 10 g of anti-hCD40 Ab (clone G28-5, to trigger signaling through hCD40-LMP1) or isotype control monoclonal Abs for 10 min at 37°C to induce recruitment of TRAFs to membrane rafts and to allow formation of LMP1 signaling complexes. Detergentsoluble and -insoluble raft lysates were prepared as described previously (36). The lysates were immunoprecipitated with protein G-Sepharose beads (Amersham Biosciences AB, Uppsala, Sweden) pre-armed with anti-hCD40 Ab (clone G28-5) as described previously (36). Aliquots of the immunoprecipitates were separated by SDS-PAGE and electroblotted onto nitrocellulose membranes (ProTran, Schleicher & Schüll). Immunoblot analysis was performed as described previously (36). Generally, blocking and secondary Ab incubations were done for 1 h at room temperature, whereas primary Ab incubation was done overnight at 4°C. A chemiluminescent substrate (Pierce) was used to detect horseradish peroxidase-labeled Abs on Western blots. The same protein blot was stripped and re-immunoblotted sequentially with different antibodies. Bands of immunoblots were quantitated using a low light imaging system (LAS-1000, FUJIFILM Medical Systems USA, Inc., Stamford, CT).

RESULTS AND DISCUSSION
TRAF adaptor proteins are trimeric assemblies that are stabilized by coiled-coil interactions of elongated N-terminal ␣-helices. At the end of these helices, a conserved C-terminal TRAF domain exists with a folding pattern that is structurally maintained in TRAF3 (27), TRAF2 (51,52), and TRAF6 (53). This independently folded domain is an eightstranded ␤-sandwich formed by two layers of ␤-sheet that each contain four antiparallel strands and that enclose a hydrophobic core (Fig. 1). In the TRAF3 crystals, one monomer is the asymmetric subunit, and the three structurally identical subunits are related by crystallographic 3-fold symmetry.
Residues 348 -504 in TRAF3 form the TRAF domain. In the mushroom-shaped molecule, intermolecular contacts are made between the C-terminal TRAF domain at one end of the trimer and typical coiledcoil interactions at the other end. Because of the shape of the molecule, there are large solvent channels in the crystal lattice along the length of the extended helices. To form the complex, synthetic peptides corresponding to the TRAF-binding region of LMP1 were soaked into existing TRAF3 crystals. These peptides varied in length from 24 to 8 residues, but each contained the PQQAT motif for TRAF3 binding: 202 PHPQQATDDSGHESDSNSNEGRHH 225 , 202 PHPQQATDDSGH-ESDSNSN 220 , 203 HPQQATDD 210 . The structure of each complex was solved, but density for the peptide was strong and clearly defined only for the short 8-residue fragment. In the case of the longer peptides, some density was apparent in the TRAF3 binding crevice, but there was no evidence of ordered peptide, and residues could not be placed with confidence (data not shown). Furthermore, there was no extra nonprotein density anywhere around the TRAF3 surface. This was surprising because we have shown in several previous studies that peptides as long as 24 residues in length can be accommodated in TRAF3 crystals in a restricted solvent "cave" located at the binding crevice on the TRAF domain (27,28,31). The structure presented here is the complex with the short peptide representing residues 203-210.
Structure of the LMP1-TRAF3 Complex-The structure of the LMP1-TRAF3 complex at 2.8-Å resolution is presented in Fig. 1 (A and B). One LMP1 peptide was bound to each of the three subunits in TRAF3, and the structure of the peptide was identical at each of the three sites in the trimer, related by strict crystallographic 3-fold symmetry. The binding site for LMP1 is located in the same crevice on TRAF3 that accommodates other TNFRs, including CD40, LT␤R, and BAFF-R (27)(28)(29). There was clear density for the polypeptide backbone atoms of LMP1 peptide 204 PQQATDD 210 , and residues in the PQQAT motif could be positioned unambiguously in the electron density (Fig. 1C). Density for the N-terminal histidine was weak and fragmented, so this residue was omitted from the model. Density was clear and continuous for the rest of the peptide, except for the side chain of Asp 209 , where the density was broken and poorly defined, probably because this residue does not make contact with TRAF3 and is flexible in the complex. In contrast, the density for the side chain of the adjacent aspartic acid, Asp 210 , was strong and clearly defined. The intermolecular contacts between TRAF3 and LMP1 observed in the crystal structure were further examined by site-directed mutagenesis. LMP1 self-aggregates through its six transmembrane domains and thus is constitutively active when it is expressed on cells (15,33,54). It has been shown previously that only the CCT of LMP1 is required for post-aggregation delivery of signals (18,22,33,54). To better determine the recruitment and binding of TRAF molecules by LMP1 signaling in B cells, we previously generated a chimeric molecule (hCD40-LMP1) composed of the extracellular and transmembrane domains of hCD40 FIGURE 1. Structure of the LMP1-TRAF3 complex. The complex is shown schematically, with the TRAF3 trimer represented in a ribbon diagram and each subunit colored separately. LMP1 is shown as a gray ball-and-stick model. One molecule of LMP1 binds to each TRAF3 subunit in a crevice at the edge of the TRAF3 ␤-sandwich domain. The TRAF3 trimer is stabilized by coiled-coil interactions between long helices that are at the N terminus of each TRAF3 monomer. In A, the location of the cell membrane would be at the top of the image. In B, the view from the top of the trimer is shown. The 3-fold symmetry is apparent, illustrating that the TRAF3 subunits and the LMP1 molecule are identical and related by strict crystallographic symmetry. In C, the model of LMP1 is displayed in a 2F o Ϫ F c density map contoured at 2.8-Å resolution. Clear strong electron density was visible to define the polypeptide backbone and the orientation of the side chains for the residues labeled: PQQATDD. In D, the contacts for LMP1-TRAF3 recognition are shown. This is a close-up view of the intermolecular contacts in the LMP1-TRAF3 complex, with TRAF3 shown as an orange ribbon and contact residues shown as gray ball-and-stick models. LMP1 is shown as a green ball-and-stick model. Critical contact residues are labeled, and the labels are underlined for residues from TRAF3. Intermolecular and intramolecular hydrogen bonds are drawn as red and green dotted lines, respectively. and the CCT of LMP1 (46). This chimeric molecule signals indistinguishably from LMP1 but with controllable initiation, and like wild-type LMP1, its aggregation localizes the hybrid receptor to plasma membrane rafts (36,46,55). To evaluate the individual contributions of residues of the PQQATDD motif of LMP1 in TRAF3 binding, we mutated each contact residue seen in the crystal structure to alanine within the context of the hCD40-LMP1 chimeric molecule. These chimeric molecules were stably transfected into the M12.4.1 mouse B cell line, and expression-matched clones were selected by immunofluorescence flow cytometry and used in the co-immunoprecipitation study.
Our laboratory has recently developed an approach to better detect the recruitment and binding of endogenous TRAFs to rafts upon LMP1 signaling in B cells (36,39). In this method, non-ionic detergent-soluble proteins are first extracted with 1% Brij, which does not disrupt rafts, and then Brij-58-insoluble proteins assembled in rafts are resolubilized with 1% Nonidet P-40 supplemented with 60 mM octyl glucopyranoside and 0.1% SDS (both octyl glucopyranoside and SDS ensure the solubilization of rafts). Brij 58-soluble and -insoluble (raft) lysates are subsequently analyzed by co-immunoprecipitation and immunoblotting. Using this approach, we determined the association of TRAF3 with the above hCD40-LMP1 mutants in comparison with wild-type hCD40-LMP1 in B cells. After stimulation with anti-hCD40 Ab to trigger signaling through hCD40-LMP1, most cellular TRAF3 was recruited by wild-type hCD40-LMP1 to detergent-insoluble membrane rafts (Fig. 2) (36,39). Interestingly, we consistently observed that a significant amount of TRAF3 was also co-immunoprecipitated with wild-type hCD40-LMP1 in the detergent-soluble fraction stimulated with an isotype control Ab (Fig. 2) (36, 39). One possibility is that TRAF3 may be able to bind to unaggregated (monomeric) hCD40-LMP1 because of its particularly high avidity for the cytoplasmic tail of LMP1. During the immunoprecipitation procedure, all samples were incubated with protein G beads pre-armed with anti-hCD40 Ab (clone G28-5), which would aggregate or cross-link hCD40-LMP1 in the lysates to the beads. Hence, another possibility is that the aggregated hCD40-LMP1 on the protein G-clone G28-5 beads may mimic the clustering of hCD40-LMP1 triggered by anti-hCD40 Ab stimulation in live cells and thus may recruit TRAF3 in the lysates during the immunoprecipitation procedure. Therefore, the amount of TRAF3 that co-immunoprecipitated with each mutant in both lanes of the Brij-58-soluble fraction stimulated with the isotype control Ab (iso, S) and the Brij-58-insoluble raft fraction stimulated with anti-hCD40 Ab (␣-h, I) shown in Fig. 2 reflects the ability of this mutant to bind to TRAF3. Our results demonstrate that, in M12.4.1 mouse B cells, recruitment and binding of TRAF3 were dramatically diminished by substitution of Ala for Pro 204 , Gln 206 , and Thr 208 and moderately decreased by substitution of Ala for Asp 210 , but were not affected by the D209A mutation (Fig. 2). Corroborating our co-immunoprecipitation data, a previous study using in vitro pull-down experiments with glutathione S-transferase fusion proteins also showed that mutations of Pro 204 , Gln 206 , Thr 208 , and Asp 210 have important effects in dampening TRAF3 association (50). These findings indicate that Pro 204 , Gln 206 , Thr 208 , and Asp 210 (but not Asp 209 ) of LMP1 are critical for binding TRAF3. Gln 206 , Thr 208 , and Asp 210 participate in hydrogen bond interactions with TRAF3, whereas Pro 204 participates in van der Waals interaction in a hydrophobic pocket in the TRAF3 binding crevice. Substitution of alanine for the 3 polar residues in the LMP1 motif would prevent formation of key hydrogen bonds. Substitution of Ala for Pro 204 apparently affects the strength of the hydrophobic interactions at that site in the motif or perhaps may affect the folding pattern of the LMP1 motif in a more general manner that diminishes binding.
Comparison of LMP1 Versus CD40 Contacts with TRAF3-LMP1 mimics signaling events and effector functions of CD40 in B lymphocytes (18,19,46,54), and TRAF3 appears to be a major adaptor protein required to transmit LMP1 signals while acting as a negative regulator for CD40 signals (56,57). We demonstrated recently that TRAF3 is actually required for activation of B cells by LMP1 and that CD40 and LMP1 use TRAF3 in different ways (39). The cytoplasmic C-terminal regions of CD40 and LMP1 each bear a TRAF recognition motif (PXQXT) that binds in the same binding crevice on the surface of TRAF3. The sequences are closely similar, PVQET in CD40 and PQQAT in LMP1, and the structural features of the pentapeptide motifs are also similar (Fig. 3). LMP1 and CD40 bind in the same crevice on TRAF3, and the intermolecular interactions involving proline and glutamine in the motif are the same. Although the recognition motifs are accommodated in a similar mode, there is a molecular adaptation of the TRAF3 surface illustrated by changes in the electrostatic surface of hCD40-LMP1 and hCD40-LMP1 mutants P204A, Q206A, T208A, D209A, and D210A were stimulated with 10 g/ml anti-hCD40 Ab (␣-h) to trigger signaling through these chimeric receptors or with an isotype control Ab (iso) for 10 min. Detergent-soluble (S) and -insoluble (I) raft lysates were prepared. The lysates were incubated with anti-hCD40 Ab (clone G28-5) to immunoprecipitate the chimeric receptors. The immunoprecipitates were analyzed by immunoblotting for TRAF3 and LMP1. B, histogram analysis of TRAF3 binding to hCD40-LMP1 mutants. TRAF3 and hCD40-LMP1 bands on immunoblots were quantitated using a low light imaging system. The amount of TRAF3 in each lane was normalized to the intensity of the corresponding hCD40-LMP1 band. The graph depicts the results of two independent experiments (mean Ϯ S.D.). The binding crevice is shown as a molecular surface colored according to electrostatic characteristics, with blue corresponding to negative regions, red representing positive regions, and white corresponding to neutral regions. The view of the two complexes is in the same orientation for direct comparison. The surface residues of TRAF3 undergo conformational adjustments as a mechanism for molecular adaptation (32), and the effect of the adaptations can be seen in the differences in electrostatic patterns when the two receptors are bound. Overall, the conformation of the recognition motif in LMP1 mimics that in CD40.
TRAF3 calculated when bound to LMP1 versus CD40 (Fig. 3). This is consistent with our previous observations that the TRAF3 binding crevice contains structurally adaptive "hot spots" that undergo conformational adjustments in the binding crevice to provide a unique shape and electrostatic characteristic for each binding partner (32).
The overall structural similarity of the recognition motifs in LMP1 and CD40 facilitates docking in the binding crevice, but there are also distinct differences in the binding patterns that must be considered in light of the functional differences that are known to result from binding events with these two proteins. It should be noted that extensive comparisons are not possible because the segment of LMP1 that has been structurally defined is short (8 residues) compared with the portion of the cytoplasmic domain of CD40 (20 residues) that was determined in complex with TRAF3 (27). With the longer segment from CD40 as well as similar fragments from LT␤R (28) and BAFF-R (29), secondary structural features of the peptides were determined to define binding of the receptors in the context of either a hairpin or extended "boomerang" conformation. CD40 binds to TRAF3 in a hairpin or reverse turn configuration. The hairpin is stabilized by an intramolecular contact made by the threonine in the consensus motif (27). In LMP1, the equivalent threonine (Thr 208 ) does make an intramolecular contact with the side chain of the neighboring residue Asp 210 . It is not possible with the present data to predict whether LMP1 also assumes a hairpin/reverse turn configuration upon binding TRAF3, as we observed for CD40 (27). Both CD40 and LT␤R bind TRAF3 in reverse turn configurations, and each of these receptors bears 1 or 2 prolines at the turn. LMP1 does not have proline(s) in the equivalent sequence and, in this respect, is more like the sequence pattern seen in BAFF-R and the downstream regulator TANK, which bind to TRAF3 as extended chains. This is of particular interest given the recent report that, like LMP1, BAFF-R utilizes TRAF3 as a positive signal regulator and shows unique binding features (58).
LMP1 Makes Unique Contacts with TRAF3-LMP1 makes two additional key contacts at or near the recognition motif that are not made by CD40. Thr 208 in the consensus pentapeptide forms a hydrogen bond with Asp 399 in TRAF3. This hydrogen bond is also made by the conserved threonine in the motif in TANK (31). In CD40, the equivalent of this threonine is engaged in the stabilizing intramolecular interaction within the hairpin and does not make any direct contacts with TRAF3 (27). Another interesting contact is made between LMP1 Asp 210 and 2 residues (Tyr 395 and Arg 393 ) in TRAF3. The hydrogen bonds formed in this network apparently provide stability to the complex because substitution of Ala for Asp 210 diminished binding to TRAF3 (Fig. 2). In CD40, the equivalent residue (2 residues downstream of the recognition motif) is histidine, which does not contact TRAF3. Interestingly, Asp 209 does not make an intermolecular contact with TRAF3. This observation was confirmed by mutagenesis in which substitution of Ala for Asp 209 did not affect binding.
Our results contrast with those from studies with TRAF2 (30) in which Asp 209 in LMP1 was reported to be the contacting residue, whereas Asp 210 , which is in close proximity, did not form hydrogen bonds with residues 393 and 395. Although binding of LMP1 to TRAF2 is considerably weak, our LMP1 mutant D210A also showed reduced binding to TRAF2, whereas D209A retained binding (data not shown), consistent with our results for TRAF3 and implicating Asp 210 as the key contact for both TRAF2 and TRAF3.
Stabilizing Interactions with TRAF3-It is possible within the TRAF crevice for hydrogen bonds to form with an aspartic acid adjacent to the PXQXT motif, as we have shown in crystal structures of LT␤R and TANK complexes with TRAF3 (28,31). In LT␤R, within the context of the sequence IPEEGD, the aspartic acid participates as a sixth residue for recognition by forming a hydrogen bond with Tyr 395 in TRAF3. In the complex of TANK and TRAF3, we noted that, within the context of the sequence PIQCTD, the aspartic acid makes hydrogen bonds with Tyr 395 and Arg 393 in TRAF3, similar to the pattern we have reported here for Asp 210 in LMP1. The additional hydrogen bonds provided by these aspartic acids may serve to strengthen binding over contacts made only by residues in the consensus motif, as seen with CD40. In the case of TANK, this downstream regulator competes with CD40 for binding to the same binding crevice on TRAF3. Stronger binding affinity was proposed to have important implications for release of TRAFs from CD40 or for modulation of TANK-mediated inhibition of NF-B activation by CD40.
The question of whether LMP1 competes with CD40 for the TRAF3 binding crevice awaits future experiments in cell-based assays, which are beyond the scope of this study. But our work to date has already given us some insight that stronger binding of LMP1 to TRAF3 compared with CD40 is critical for the ability of LMP1 to transform B lymphocytes. We have demonstrated previously that LMP1 exhibits stronger binding to TRAF3 relative to CD40. In B lymphocytes, LMP1 recruits the majority (ϳ80%) of cellular TRAF3, whereas CD40 engagement recruits only ϳ30% of cellular TRAF3 (39,46,50). Furthermore, the amount of TRAF3 that co-immunoprecipitates with CD40 is dramatically reduced by the presence of LMP1, suggesting that LMP1 may sequester cellular TRAF3, making it unavailable for CD40 (39). The additional hydrogen bonds provided by Asp 210 in the LMP1-TRAF3 complex strengthen binding and stabilize the complex and thus may represent the molecular basis for understanding these binding properties.
In summary, we have reported the structural details of the molecular interactions that are made when LMP1 binds to TRAF3 and have shown that more hydrogen-bonded contacts are formed than exist in the CD40-TRAF3 complex. In particular, Asp 210 forms key intermolecular hydrogen bonds that do not exist when CD40 binds to TRAFs. The stability of the LMP1-TRAF3 complex may play a role in the dysregulated signaling and sustained B cell activation caused by LMP1. Presentation of LMP1 as a structural mimic of CD40 may be an effective viral strategy to introduce a molecular decoy for the CD40-binding site on TRAF3, influencing downstream signaling in B lymphocytes.