Differential Requirements for Tumor Necrosis Factor Receptor-associated Factor Family Proteins in CD40-mediated Induction of NF-κB and Jun N-terminal Kinase Activation*

CD40 is a member of the tumor necrosis factor receptor family that mediates a number of important signaling events in B-lymphocytes and some other types of cells through interaction of its cytoplasmic (ct) domain with tumor necrosis factor receptor-associated factor (TRAF) proteins. Alanine substitution and truncation mutants of the human CD40ct domain were generated, revealing residues critical for binding TRAF2, TRAF3, or both of these proteins. In contrast to TRAF2 and TRAF3, direct binding of TRAF1, TRAF4, TRAF5, or TRAF6 to CD40 was not detected. However, TRAF5 could be recruited to wild-type CD40 in a TRAF3-dependent manner but not to a CD40 mutant (Q263A) that selectively fails to bind TRAF3. CD40 mutants with impaired binding to TRAF2, TRAF3, or both of these proteins completely retained the ability to activate NF-κB and Jun N-terminal kinase (JNK), implying that CD40 can stimulate TRAF2- and TRAF3-independent pathways for NF-κB and JNK activation. A carboxyl-truncation mutant of CD40 lacking the last 32 amino acids required for TRAF2 and TRAF3 binding, CD40(Δ32), mediated NF-κB induction through a mechanism that was suppressible by co-expression of TRAF6(ΔN), a dominant-negative version of TRAF6, but not by TRAF2(ΔN), implying that while TRAF6 does not directly bind CD40, it can participate in CD40 signaling. In contrast, TRAF6(ΔN) did not impair JNK activation by CD40(Δ32). Taken together, these findings reveal redundancy in the involvement of TRAF family proteins in CD40-mediated NF-κB induction and suggest that the membrane-proximal region of CD40 may stimulate the JNK pathway through a TRAF-independent mechanism.


From the Burnham Institute, La Jolla, California 92037
CD40 is a member of the tumor necrosis factor receptor family that mediates a number of important signaling events in B-lymphocytes and some other types of cells through interaction of its cytoplasmic (ct) domain with tumor necrosis factor receptor-associated factor (TRAF) proteins. Alanine substitution and truncation mutants of the human CD40ct domain were generated, revealing residues critical for binding TRAF2, TRAF3, or both of these proteins. In contrast to TRAF2 and TRAF3, direct binding of TRAF1, TRAF4, TRAF5, or TRAF6 to CD40 was not detected. However, TRAF5 could be recruited to wild-type CD40 in a TRAF3-dependent manner but not to a CD40 mutant (Q263A) that selectively fails to bind TRAF3. CD40 mutants with impaired binding to TRAF2, TRAF3, or both of these proteins completely retained the ability to activate NF-B and Jun N-terminal kinase (JNK), implying that CD40 can stimulate TRAF2-and TRAF3-independent pathways for NF-B and JNK activation. A carboxyl-truncation mutant of CD40 lacking the last 32 amino acids required for TRAF2 and TRAF3 binding, CD40(⌬32), mediated NF-B induction through a mechanism that was suppressible by co-expression of TRAF6(⌬N), a dominant-negative version of TRAF6, but not by TRAF2(⌬N), implying that while TRAF6 does not directly bind CD40, it can participate in CD40 signaling. In contrast, TRAF6(⌬N) did not impair JNK activation by CD40(⌬32). Taken together, these findings reveal redundancy in the involvement of TRAF family proteins in CD40-mediated NF-B induction and suggest that the membrane-proximal region of CD40 may stimulate the JNK pathway through a TRAF-independent mechanism.
CD40 is a member of the tumor necrosis factor receptor (TNFR) 1 superfamily, which triggers pleiotrophic biological responses in B-cells, including up-regulation of adhesion proteins (intercellular adhesion molecule; CD56), expression of immune co-stimulatory molecules (B7; CD80), clonal expansion (proliferation), immunological memory (cell survival), affinity-maturation of antibodies, and Ig class switching (1,2). CD40 may also play a role in some aspects of inflammatory responses involving nonlymphoid cells, such as intercellular adhesion molecule (ICAM) expression on activated endothelial cells (3,4) and prostaglandin secretion by fibroblasts (5), Moreover, CD40 may provide a conduit through which the immune system can suppress growth and induce apoptosis of CD40-expressing epithelial cancers (6).
The ct domain of huCD40 is only 62 amino acids (aa) long and lacks any motifs that might suggest a biochemical mechanism for transducing signals into cells. Signaling by this receptor appears to be initiated by the binding of TRAF family proteins to CD40ct. TRAFs represent a group of structurally similar adapter proteins that link the cytosolic domains of some TNFR family cytokine receptors to downstream signaling pathways. All members of this family contain a conserved C-terminal ϳ150-aa region known as the TRAF domain, which mediates their interactions with the cytosolic domains of specific TNF family cytokine receptors (reviewed in Ref. 7). Among the six known TRAF family proteins, all except TRAF4 have been reported to associate with CD40ct (8 -13). TRAF family members can self-oligomerize, and some form heterocomplexes with each other (13,14). TRAF2, TRAF5, and TRAF6 induce NF-B and Jun N-terminal kinase (JNK) activation when overexpressed in cells, whereas TRAF1, TRAF3, and TRAF4 do not (15)(16)(17)(18)(19). It is unclear whether those TRAF family members that fail to activate NF-B and JNK have other unidentified signaling functions versus operating as competitors that preclude recruitment of other TRAFs to CD40ct and other TNF family receptors after ligand binding.
Previous studies have shown that a 17-aa segment of CD40ct ( 250 PVQETLHGCQPVTQEDG 266 ) is sufficient for activating NF-B and binding TRAF2 and TRAF3 (8). A T254A alanine substitution mutation within CD40ct completely abolishes TRAF2, TRAF3, and TRAF5 binding, suggesting that these three TRAF family members bind an overlapping site (12,13,20). In contrast, a TRAF6-binding motif has been localized to an upstream, membrane-proximal portion of CD40ct (12,13,21). Given that TRAF2, TRAF5, and TRAF6 are all capable of activating NF-B and JNK, redundancy in the signal-initiating mechanisms used by CD40 may exist. If TRAF family proteins are to serve as potential therapeutic targets for modulating CD40 action, it is important to define the individual contributions of each TRAF family member that binds this receptor. We therefore created CD40ct mutants that selectively lost the abil-ity to bind specific TRAFs, with the goal of determining the resulting effects on downstream CD40-mediated signal transduction events.

MATERIALS AND METHODS
Plasmids-The plasmid pGEX4T1-CD40 encoding huCD40ct fused to GST (10) was employed for creation of CD40 mutants using either one-step or two-step polymerase chain reaction methods. These pGEX-CD40 mutants were digested with MscI and XhoI, and then DNA inserts corresponding to the mutagenized CD40ct regions were subcloned in place of the corresponding wild-type (WT) sequences in pcDNA3-huCD40, a plasmid encoding full-length huCD40.
TRAF cDNAs in pcDNA3 were in vitro transcribed/translated with L-[ 35 S]methionine (Amersham Pharmacia Biotech) using reticulocyte lysates (Promega). Immobilized GST fusion proteins (ϳ1 g/5 l) were incubated with 5 l of in vitro translation (IVT) products in a total volume of 200 l in Co-IP buffer for 1 h at 4°C. Beads were washed four times in 1 ml of Co-IP buffer, and proteins were eluted in 20 l of Laemmli solution for SDS-PAGE/autoradiography analysis. Results were quantified with a Bio-Rad phosphorimaging device and expressed relative to WT CD40.
Alternatively, lysates were prepared from transfected 293 cells in Co-IP buffer (0.5 mg of total protein/ml) and precleared with glutathione-Sepharose, and immobilized GST-fusion proteins (ϳ5 g) were added to 500 l for incubation at 4°C for 4 h. After washing, the beads were resuspended in 20 l of Laemmli buffer, and the proteins were analyzed by SDS-PAGE/immunoblotting. Experiments involving recombinant TRAF domains were performed similarly, using Co-IP buffer and GST fusion proteins at 1 g/l.
Cell Culture and Transfections-293T and HeLa cells were obtained from ATCC (Rockville, MD) and cultured in Dulbecco's modified Eagle's medium-high glucose medium (Life Technologies, Inc.) supplemented with 10% fetal bovine serum (Hyclone), 1 mM L-glutamine, and antibodies. Cells in 10-cm dishes at 80% confluence were transfected (10 g of total DNA) using a calcium phosphate method. The plasmid pcDNA3-␤Gal (1 g) was included in all transfections to adjust for differences in transfection efficiency, as assessed by ␤-galactosidase assay (31). Cells were lysed in 1 ml of Co-IP buffer.
Reporter Gene Assays-For NF-B reporter gene assays, 293T or HeLa cells were calcium phosphate-transfected with 3 g (total) of plasmid DNA at 60 -80% confluency in 12-well plates. After 1.5-2 days, cells were lysed in 0.1 ml of Promega lysis buffer. Lysates were measured for luciferase activity using a luminometer (EG & G Berthold) and Promega luciferase kits, normalizing relative to ␤-galactosidase.
Immunocomplex Kinase Activity Assay for JNK-293T or HeLa cells in 10-cm dishes were subjected to lipofection with various wild-type or mutant pcDNA3-CD40 plasmids (5 g) and 1.2 g of pcDNA3-FLAG-JNK1. Whole cell lysates were prepared using 1 ml of JNK kinase buffer assay (32), and 300 g of protein in 1 ml was employed for immunoprecipitation using 20 l of anti-FLAG antibody-Sepharose (Sigma), followed by in vitro kinase assay (32). Phosphorylated GST-c-Jun bands in gels were analyzed with a Bio-Rad phosphorimaging device. Comparable expression of FLAG-JNK1 and CD40 constructs was confirmed by immunoblotting using anti-CD40ab (Santa Cruz Biotechnology, Inc., Santa Cruz, CA) or anti-FLAG-M 2 (Eastman Kodak Co.) antibodies.

Analysis of Binding Sites for TRAF Family Proteins on
CD40 -Alanine substitution mutations and C-terminal truncations of huCD40ct were produced as GST fusion proteins and used for protein binding assays employing IVT 35 S-TRAF1, -2, -3, -4, -5, and -6 ( Fig. 1 and data not shown). Similar experiments were also performed using cell lysates from 293T cells that had been transiently transfected with plasmids producing epitope-tagged TRAF1, -2, -3, -4, -5, or -6, followed by immunoblot detection of bound TRAFs ( Fig. 2 and data not shown). Consistent with previous reports (8 -10), binding of TRAF2 and TRAF3 to GST-CD40 was easily detectable (Figs. 1 and 2). In contrast, TRAF5 failed to bind GST-CD40 altogether or was seen at levels only slightly above background, as determined by comparisons with GST nonfusion, GST-p75-nerve growth factor receptor (NGFR) ct domain, or other GST-control proteins ( Fig. 2C and data not shown), whereas TRAF6 binding was completely absent (Fig. 2C). Binding of TRAF5 and TRAF6 to the ct domains of other TNFR family members, e.g. GST-LT␤R and GST-p75-NGFR, respectively, was readily detectable under similar conditions (not shown), implying that TRAF5 and TRAF6 produced by IVT or by expression in 293T cells were competent to bind at least some TNF family receptors. TRAF1 and TRAF4 also did not bind GST-CD40ct (not shown), although they did interact with control GST fusion proteins such as GST-cIAP1 (33) and GST-LT␤R (23).
As described previously (8,13) point mutations in a conserved PXQX(T/S) core motif found in several TRAF-binding TNFR members ( 250 PVQET 254 ) significantly impaired binding of both TRAF2 and TRAF3. The T254A mutation abolished binding of TRAF2 and TRAF3 completely, whereas alanine substitution mutants of other residues in the 250 PVQET 254 motif reduced but did not entirely abrogate binding of these TRAFs (Fig. 1). Interestingly, Pro 250 and Val 251 appeared to be relatively more important for TRAF2 than TRAF3 binding, implying differences in the specificity of these proteins for residues in the PVQET core motif of CD40. Alanine substitutions in a region ( 229 PKQEPQE 235 ) located upstream of, but having similarity to, the PVQET core motif had no effect on TRAF2 or TRAF3 binding.
Differences in the requirements for TRAF2 and TRAF3 binding to GST-CD40 were revealed by mutations affecting residues distal to the PVQET motif. For example, alanine replacements within the 262 TQEDGK 267 region demonstrated that Gln 263 is crucial for TRAF3 binding but not TRAF2 (Fig. 1). Although TRAF2 binding was relatively insensitive to alanine substitutions within 262 TQEDGK 267 of CD40ct, a G266A mutation did reduce binding by approximately half. In contrast, TRAF3 binding was slightly enhanced by the G266A mutation ( Fig. 1), again indicating differences in the structural features of CD40ct binding by TRAF2 and TRAF3.
Similarly, C-terminal truncation mutations of CD40ct provided further evidence of different requirements for TRAF2 and TRAF3 binding. Deleting 11, 13, or 15 aa from the C terminus of CD40ct reduced in vitro TRAF2 binding to ϳ40% of control levels, whereas TRAF3 binding was unaffected by removal of the C-terminal 11 aa and only modestly reduced by deletion of the last 13 aa of CD40ct. Deletion of another 2 amino acids (⌬15 aa), thus removing Gln 263 and Glu 264 , completely abrogated TRAF3 binding (Fig. 1). These data are consistent with the alanine substitution results, which identified Gln 263 as an important residue for TRAF3 binding, suggesting that Gln 263 lies near or at the carboxyl boundary of the minimal TRAF3 binding site within CD40ct. We conclude therefore that TRAF2 and TRAF3 bind overlapping but nonidentical sites in the ct domain of CD40.
Aptamer Peptide Library Screening Reveals TRAF2 Binding to Peptides with Homology to the C-terminal 11 aa of CD40ct-A combinatorial library of constrained 20-residue peptides displayed by the active-site loop of E. coli-derived thioredoxin was screened for TRAF2 binders using a yeast two-hybrid method (28). Peptides that bound specifically to TRAF2 were identified by secondary screens against irrelevant bait proteins, eliminating false positives. Most of the identified peptides shared sequence similarity with each other and with sequences within the C-terminal 11 aa of huCD40 (Fig. 3). This 11-aa region is perfectly conserved in the human and mouse CD40 proteins, containing an ISVQE motif, which is resembled by several of the TRAF2-binding peptides. Some of the peptides were more similar to motifs in the C terminus of CD30 or LMP1 than to CD40 (Fig.  3), all of which have been previously implicated in TRAF2 binding (35)(36)(37). None of the peptides identified by aptamer library screening bound TRAF3 (not shown), arguing that the recognition of peptide ligands by TRAF2 and TRAF3 differs. The peptide aptamer library screening results, therefore, indirectly support the idea that optimal binding of TRAF2 to CD40 is facilitated by residues in the C-terminal 11 aa of CD40ct.
TRAF2 and TRAF3 Do Not Efficiently Compete for Binding to CD40 -To determine whether TRAF2 and TRAF3 compete for binding to CD40ct, competitive binding assays were performed using the TRAF domains of TRAF2 and TRAF3, which were produced as recombinant proteins in bacteria and purified. The TRAF domain of TRAF2 or TRAF3 was preincubated at a 10-fold molar excess relative to GST-CD40ct to saturate its binding sites on CD40, and then various amounts of the opposing TRAF domain were added in 1-90-fold molar ratios compared with GST-CD40, and binding was monitored 1 h later by SDS-PAGE/immunoblotting. In pilot experiments, both TRAF domains reached saturation binding when added at ϳ5-10-fold molar excess over GST-CD40, with ϳ15-20% of the input TRAF binding. Despite presaturation of binding sites for TRAF2 on GST-CD40ct, TRAF3 bound in a concentration-dependent manner, and vice versa (Fig. 4). Furthermore, the titration of the other TRAF did not lead to a reduction in retention of the preincubated TRAF. Even when TRAF domains were added at ϳ9-fold molar excess of the opposing TRAF and at ϳ90-fold excess of GST-CD40, no apparent reduction in the amounts of TRAF prebound to GST-CD40 was seen (Fig. 4). Similar results were obtained in experiments where TRAF2 and TRAF3 were added simultaneously to GST-CD40 in various molar ratios (not shown). We conclude, therefore, that TRAF2 and TRAF3 do not efficiently compete for binding sites on CD40ct, despite the apparent overlap in some regions required for their recognition of CD40ct.
TRAF5 Is Recruited to CD40ct by Interactions with TRAF3-TRAF3 and TRAF5 can bind to each other but not to other known TRAF family members in vitro and in yeast-two-hybrid FIG. 1. Summary of in vitro protein binding assay results obtained using GST-CD40 mutants. The binding of IVT 35 S-TRAF2 or 35 S-TRAF3 to WT GST-CD40 or various GST-CD40 mutants was assayed. Data represent mean Ϯ S.D. (n ϭ 3) based on phosphorimager analysis from SDS-polyacrylamide gels and are expressed as a percentage relative to WT GST-CD40. Results obtained using a GST control protein are also indicated. The sequence of huCD40ct is presented, and alanine substitution mutations are indicated in boldface type. C-terminal truncation mutants lacking the last 11, 13, or 15 aa of huCD40 are shown at the far right. The hatched boxes are aligned with the approximate regions of the CD40ct sequence that appear to be important for TRAF2 (top) or TRAF3 (bottom) binding.
assays (13). 2 We therefore asked whether TRAF5 could be indirectly recruited to CD40ct through interactions with TRAF3. To test this idea, GST-CD40ct was incubated with 35 S-labeled ("hot") TRAF5 in the presence or absence of IVT unlabeled ("cold") TRAF3 (Fig. 5A). As a control, GST-CD40 (Q263A) that fails to bind TRAF3 was also employed. Only a small amount (ϳ0.5%) of the input 35 S-TRAF5 bound to GST-CD40 in the absence of TRAF3 (Fig. 5A). In contrast, when increasing amounts of TRAF3 were added, progressively larger amounts of 35 S-TRAF5 were associated with GST-CD40. In contrast, TRAF3 did not promote association of TRAF5 to the GST-CD40(Q263A) mutant. Moreover, the addition of TRAF2 or TRAF6 to these assays had no effect on TRAF5 association with GST-CD40 (not shown). We conclude therefore that while TRAF5 displays little affinity for CD40ct, it can be recruited to this receptor in a TRAF3-dependent manner.
TRAF3 Modulates NF-B Induction by TRAF5-Given that TRAF3 binds TRAF5, we wondered whether TRAF3 could modulate NF-B induction by this TRAF family member. Transfection of 293T cells with TRAF2, TRAF5, or TRAF6 induced increases in NF-B activity, as determined by assays using FIG. 2. Examples of in vitro binding assay results obtained using GST-CD40 mutants. In vitro binding assay results obtained with selected mutants of CD40ct are presented. A, GST-CD40 and CD40 mutant proteins immobilized on glutathione-Sepharose were incubated with IVT 35 S-TRAF2 and -TRAF3. Bound TRAFs were detected by SDS-PAGE/autoradiography. Binding ratios represent quantification of the data by phosphorimager, expressed relative to binding to WT CD40 (set at 1.0). p75-NGFR was included as a negative control (far right). One-tenth (A) or one-twentieth (B) of the input 35 S-TRAF protein was run directly in gels (far left) for comparison. B, similar binding assays were performed using GST-CD40 fusion proteins and lysates from 293T cells that had been transiently transfected with plasmids producing HA-tagged TRAF2 or TRAF3. Bound TRAFs were analyzed by SDS-PAGE/immunoblotting using antibodies specific for TRAF2 or TRAF3 followed by 125 I-Protein-A and autoradiography. Note that endogenous TRAF2 and TRAF3 as well as HA-tagged TRAF2 and TRAF3b were detected by the antibodies. HA-tagged protein results were used for calculation of binding ratios. Loading of equivalent amounts of each GST fusion protein was confirmed by stripping and reprobing blots with anti-GST monoclonal antibody (not shown). C, binding of immobilized GST-CD40 to 35 S-labeled in vitro translated TRAF5 and TRAF6 (left) or to epitope-tagged TRAF5 and TRAF6 derived from lysates of transfected 293T cells (right) was tested as above. pUC13-4xNF-B-Luc (Fig. 5, B-D). In contrast, transfection of 293T cells with TRAF1 or TRAF3 did not result in NF-B induction (not shown), confirming the specificity of these reporter gene assay results. Co-transfecting 293T cells with fixed amounts of TRAF5-encoding plasmid DNA and progressively increasing relative amounts of a TRAF3-producing plasmid resulted in an initial slight decrease in NF-B activity at lower TRAF3 plasmid concentrations followed by a pronounced enhancement (approximate doubling) of NF-B activity at higher ratios of TRAF3 to TRAF5 plasmids. In contrast, TRAF3 had only modest effects on TRAF2 and essentially no effect on TRAF6-mediated NF-B induction. Taken together, these findings provide indirect evidence of functional interaction of TRAF3 with TRAF5 in cells.
TRAF2 and -3 Are Expendable for CD40-mediated NF-B Induction-We employed CD40ct mutants to correlate the binding of specific TRAFs with NF-B induction. For these experiments, WT or mutant CD40 proteins were expressed in 293T cells by transient transfection (Fig. 6A), which results in ligand-independent aggregation and signaling of CD40 (15). Indirect immunofluorescence fluorescence-activated cell sorting analysis confirmed that all CD40 mutants were expressed on the cell surface at levels comparable with WT CD40 (not shown).
With the exception of a mutant of CD40 that lacks its entire ct domain (CD40⌬ct), all mutants of CD40 tested retained the ability to induce NF-B at levels at least equivalent to or closely approaching those obtained with WT CD40 (Fig. 6A). This included mutants of CD40 with selective impairments in TRAF2 (e.g. ⌬11 aa) or TRAF3 (e.g. Q263A) binding, as well as mutants that failed to bind both TRAF2 and TRAF3 (e.g. ⌬15 aa, ⌬32 aa, T254A). Moreover, the CD40 ⌬15 aa mutant, which fails to bind both TRAF2 and TRAF3, consistently resulted in higher levels of NF-B induction compared with WT CD40 (Fig.  6A), although it was not expressed at higher levels than WT CD40 (not shown). These observations thus indicate that neither TRAF2 nor TRAF3 are essential for CD40-mediated induction of NF-B. They also indicate that the membrane-proximal region of CD40ct retained in the ⌬32 aa C-terminal truncation mutant is sufficient for NF-B induction.
TRAF2 and -3 Are Not Essential for CD40-mediated JNK Activation-We next explored the effects on JNK activation of CD40 mutations that alter TRAF binding. HeLa cells were transiently co-transfected with pCMV-M 2 FLAG-JNK1 and expression plasmids, producing WT CD40 or various CD40 mutants. After stimulation with CD40L for 25 min, immunoprecipitations were performed with anti-FLAG antibody, and JNK in vitro kinase activity was measured using GST-C-Jun-(1-79) as an exogenous substrate (Fig. 6B). Similar to the results obtained with NF-B, all mutants of CD40 induced JNK activation at levels comparable with WT CD40. Immunoblot analysis confirmed production of the FLAG-JNK1 protein and of the CD40 mutants at similar levels in all samples (not shown).
TRAF6 Mediates NF-B Induction but Not JNK Activation by the Membrane-proximal Region of CD40 -Although we were unable to find evidence of direct binding of TRAF6 to CD40ct, the membrane-proximal region of this receptor has been implicated in TRAF6 recruitment to CD40 receptor complexes by other groups (12,13,38). Moreover, Glu 235 within this membrane-proximal region of CD40 has been reported to be critical for association of TRAF6 with CD40 receptor complexes (12,13,21). Therefore, we examined the effects of a E235A mutation within the context of either the full-length CD40 protein or the CD40(⌬32) truncation mutant, which contains only the membrane-proximal region that we found was sufficient for induction of NF-B and for activation of JNK (Fig. 6).
As shown in Fig. 7, the CD40 (E235A) mutant was substantially impaired in its ability to induce NF-B compared with WT CD40 when expressed in 293T cells. In contrast to E235A, other alanine substitutions within the membrane-proximal region (e.g. Q231A and Q234A), which reportedly do not substantially impair TRAF6 association with CD40 receptor complexes (21), did not reduce NF-B induction or JNK activation by FIG. 5. Functional interaction of TRAF3 with TRAF5. A, IVT unlabeled ("cold") TRAF3 and 35 S-labeled ("hot") TRAF5 were co-incubated in various relative amounts as indicated (v/v) with GST-CD40ct or GST-CD40(Q263A) proteins immobilized on glutathione-Sepharose. After washing, CD40ct-bound proteins were eluted and analyzed by SDS-PAGE/autoradiography. A volume of the TRAF5 IVT mix equivalent to 0.5% of the total input into binding assays was loaded directly into gels as a control (far left). Note that a small percentage of the input TRAF5 may bind to CD40(Q263) independently of TRAF3, as indicated by the faint band in the rightmost lane of the lower panel. However, background binding to control GST-proteins also is observed when large amounts of IVT TRAF5 are input (not shown). B-D, 293T cells in six-well dishes were transiently transfected with fixed amounts (1 g) of plasmids encoding TRAF5 (top), TRAF2 (middle), or TRAF6 (bottom) and variable amounts (0.25-3.0 g) of a TRAF3-producing plasmid (gradient) as indicated, together with 0.25 g of the NF-B reporter plasmid pUC13-4xNF-B-luc and 0.25 g of pCMV-␤Gal. NF-B activity was measured from cell lysates 2 days later by luminometer-based luciferase assays (relative luciferase units; RLU), with normalization for ␤-galactosidase activity (mean Ϯ S.D.; n ϭ 3). Immunoblot analysis confirmed production of the transfected TRAFs and excluded TRAF3mediated alterations in the levels of TRAF2, TRAF5, and TRAF6 as an explanation for the results (not shown). CD40 in 293T cells (Fig. 6), thus providing a specificity control. Remarkably, when the E235A mutation was placed into the CD40(⌬32) truncated protein, NF-B induction was completely abolished (Fig. 7A). Similar data were obtained with respect to JNK activation (Fig. 7B). Anti-CD40 immunoblot analysis indicated that the impaired function of the E235A mutants was not attributable to lower levels of expression compared with WT CD40 and CD40(⌬32) (not shown). Thus, Glu 235 defines a region of CD40 that makes important contributions to both NF-B induction and JNK activation within the context of the full-length receptor. Moreover, this region is essential for NF-B induction and JNK activation when the C-terminal TRAF2 and TRAF3 interaction sites are missing from CD40.
To further explore the potential involvement of TRAF6 in mediating signals from this membrane-proximal region of CD40, co-transfection experiments were performed in which a trans-dominant inhibitory mutant of TRAF6 was co-expressed in 293T cells with the CD40(⌬32) protein, which lacks the TRAF2 and TRAF3 binding sites. This mutant of TRAF6 is missing the N-terminal region containing the RING domain and the zinc fingers, previously reported to be necessary for NF-B induction and JNK activation by TRAF proteins (12, FIG. 6. Analysis of NF-B and JNK inactivation by CD40 mutants. A, 293T cells in 12-well dishes were transiently transfected with 3 g of pcDNA3 control plasmid ("vector") or pcDNA3 plasmids encoding CD40 or various CD40 mutants as indicated, together with 0.25 g of pUC13-4xNF-B-Luc and 0.25 g of pCMV-␤Gal. Relative NF-B activity was assessed using lysates prepared 2 days later by luciferase assays with normalization for ␤-galactosidase (relative fluorescence units; RFU) (mean Ϯ S.D.; n ϭ 3). Expression of CD40 mutants at comparable levels was confirmed by indirect immunofluorescence using anti-huCD40 monoclonal antibody and fluorescence-activated cell sorting analysis (not shown). B, HeLa cells were transiently transfected with either pcDNA3 control or pcDNA-CD40 plasmids and pCMV-M 2 FLAG-JNK1. After ϳ1.5 days, cells were stimulated with 2 g/ml soluble CD40L for 25 min. As an additional control, JNK1-transfected HeLa cells were treated with 1.2 mJ/m 2 UV irradiation and lysed 1 h later. Lysates (300 g of protein) were immunoprecipitated using anti-FLAG antibody. Immune complex in vitro kinase assays were performed using [␥-32 P]ATP and GST-c-Jun-(1-79), and 32 P-labeled GST-Jun was analyzed by SDS-PAGE/autoradiography (top). Withholding CD40L resulted in no increase in JNK activation in all cases where control vector, WT, or mutant CD40 was transfected (not shown). The -fold induction of JNK activity relative to control-transfected cells is indicated, in relative intensity units (RIU). Immunoblot analysis using anti-FLAG antibody M 2 and anti-CD40 antibody confirmed production of FLAG-JNK and CD40 proteins at similar levels in all samples (not shown).

FIG. 7. TRAF6 involvement in NF-B induction but not JNK activation by membrane-proximal region of CD40. Cell transfections
were performed and the effects on either NF-B induction (A) or JNK activity (B) were assessed as in Fig. 6. Plasmids (3 g each) used for transfections encoded the following proteins: wild-type CD40 (WTCD40), CD40 missing the last 32 amino acids (⌬32 aa), CD40 in which Glu235 was replaced with Ala (E235A), CD40 with both E235A substitution and ⌬32 carboxyl-truncation mutations (⌬32 aa ϩ E235A), HA-tagged TRAF2 lacking residues 1-247 (T2⌬N), or Myc-tagged TRAF6 lacking residues 1-274 (T6⌬N). Dominant-negative TRAF2(⌬N) or TRAF6(⌬N) plasmids were co-transfected with CD40(⌬32) using plasmid ratios of 0.08:1, 0.25:1, and 0.75:1 (w/w) for NF-B assays (A) and at ratios of 0.25:1 and 0.75:1 for JNK assays (B). In B, autoradiography analysis of in vitro phosphorylated GST-JUN is shown, as well as immunoblot analysis of the cell lysates (25 g of total protein) using anti-HA and anti-Myc antibodies to detect the epitope-tagged TRAF2(⌬N) and TRAF6(⌬N) proteins. Data are representative of two (JNK) or three (NF-B) experiments. 15). As shown in Fig. 7A, transfection of various amounts of plasmid DNA encoding the same amounts of TRAF6(⌬N) protein produced a concentration-dependent reduction in NF-B induction by CD40(⌬32). In contrast, TRAF2(⌬N), a dominantnegative inhibitor of TRAF2 (15), had relatively little effect on NF-B induction by CD40(⌬32). Despite its inhibitory effect on NF-B induction by CD40(⌬32), overexpression of the TRAF6(⌬N) protein did not interfere with JNK activation in these experiments (Fig. 7B). Immunoblot analysis confirmed expression of both the TRAF2(⌬N) and TRAF6(⌬N) proteins at levels severalfold higher than the endogenous TRAF proteins and also verified consistent expression of CD40(⌬32) (Fig. 7B  and data not shown). These findings therefore suggest that while TRAF6 may be important for mediating NF-B induction by the membrane-proximal region of CD40, alternative pathways for activation of JNK appear to exist. DISCUSSION Our analysis of TRAF family protein interactions with GST-CD40 fusion proteins indicates that TRAF2 and TRAF3 are the major CD40 binding proteins among the six known TRAF members. Although we cannot exclude the possibility that other TRAF family proteins can bind to CD40ct with affinities sufficient for detection in yeast two-hybrid assays (12) or in more sensitive in vitro protein binding assays (13), it is clear that TRAF2 and TRAF3 bind to CD40 far more efficiently than other TRAFs (38), thus suggesting that they are likely to be the predominant members of the family that are capable of direct interactions with this receptor in vivo. Other TRAFs such as TRAF5 and TRAF6 may nevertheless participate in CD40mediated signal transduction, but their interactions with CD40 receptor complexes are presumably mediated by interactions with other proteins. In this regard, TRAF5 can be indirectly recruited to CD40 via interactions with TRAF3, which is analogous to the previously reported recruitment of TRAF1 to the ct domain of TNFR2 by hetero-oligomerization with TRAF2 (14). Moreover, evidence for indirect interactions of TRAF6 with cytokine receptors comes from the interleukin-1 receptor, where TRAF6 recruitment is mediated by MyD88 (19,39). Thus, we presume that TRAF6 is recruited to CD40-containing receptor complexes through an unidentified adaptor protein analogous to MyD88, thus explaining the discrepancies among various groups who have examined TRAF6 interactions with CD40 (21,38). We cannot exclude the possibility, however, that post-translational modifications of TRAF5 and TRAF6, which were not provided by our protein interaction systems, can permit these TRAFs to bind CD40ct under some conditions. Mutational analysis of residues required for TRAF2 and TRAF3 binding within CD40ct suggests that overlapping but nonidentical sites are recognized by these proteins. Both TRAF2 and TRAF3 share a requirement for the 250 PVQET 254 motif in huCD40 but display striking differences in their dependence on other residues distal to this region. Motifs resembling 250 PVQET 254 in CD40 are found in the ct domains of several TNF family receptors that are known to bind TRAF2 and TRAF3, as well as in the LMP1 transforming protein of Epstein-Barr virus (35,40,41). In contrast, the downstream regions are less well conserved among these receptors. Nevertheless, an apparent requirement for distal motifs similar to those seen in the carboxyl-terminal CD40 region has been reported for TRAF2, but not TRAF3, binding to CD30 and LMP-1 (35)(36)(37). Moreover, the complete conservation of this distal region in human and murine CD40 is consistent with the notion that this segment is important for CD40 function. Finally, after this work was completed, the x-ray crystal structure was reported of the TRAF domain of TRAF2 in a complex with a 9-mer peptide corresponding to the analogous distal region of TNF-R2, confirming that this segment makes critical contacts with the receptor-binding pocket of the TRAF domain (42). This same peptide, however, does not bind TRAF3 (43), indicating differences in the specificities of TRAF2 and TRAF3 for binding sites on TNF family receptors.
Interestingly, the last 11 aa of huCD40, which were found to be important for binding TRAF2 but not TRAF3, fall outside a 17-aa segment (residues 250 -266) previously reported to be sufficient for NF-B induction and for binding TRAF2 and TRAF3 (8). These and other data (44) suggest that the requirement of this distal region of CD40 for TRAF2 binding may be relative, perhaps improving the affinity of TRAF2 interactions with CD40 but not absolutely required for at least low affinity interactions. In this regard, TRAF2 has been shown to be capable of interacting at least weakly with 5Ј-mer peptides comprising the PVQET motif (13). Nevertheless, our results from screening of a library of peptide aptamers using TRAF2 and the recently determined structure of TRAF2 (42) argue that this TRAF member can recognize peptide sequences resembling the distal portion of CD40. The apparent interaction of TRAF2 with two regions of CD40, only one of which is also recognized by TRAF3, may underlie the observation that TRAF2 and TRAF3 did not effectively compete with each other for binding to CD40 in vitro. This observation has important implications for understanding the relative roles of these two TRAF family members in CD40 signaling. Indeed, although not examined here, it is even conceivable that TRAF2 and TRAF3 could bind cooperatively to CD40 receptor complexes (45).
The data presented here suggest that neither TRAF2 nor TRAF3 is absolutely required for NF-B induction or JNK activation in cells, in agreement with previous reports (12,21,46,47). This observation does not imply that TRAF2 is unimportant for NF-B induction or JNK activation but merely indicates redundancy of mechanisms. The TRAF2/TRAF3-independent pathway for NF-B induction and JNK activation maps to the membrane-proximal first 30 aa of CD40ct. In some physiological contexts, however, either the TRAF2-binding or membrane-proximal regions of CD40ct may be quantitatively more important for NF-B or JNK activation for a variety of reasons, including tissue-specific differences in TRAF family expression (23,38,48). Moreover, in some types of cells, CD40 might require both of these pathways either for (a) generating sufficient amounts of NF-B or JNK activity to elicit certain biological responses or (b) triggering activation of different members of the Rel or mitogen-activated protein kinase kinase kinase families that might be required for cell-specific responses that are not discerned by generic reporter gene assays (18,49). Indeed, cells from TRAF2 knockout mice exhibit defective JNK activation but not NF-B induction in response to TNF (50). Similarly, B-cells from transgenic mice overexpressing a TRAF2(⌬RING) protein exhibit defects in CD40L-mediated activation of JNK but not NF-B (51). It seems likely, therefore, that the functional requirements for individual TRAFs may vary depending on both the cell type and the particular TNF family receptor.
Unlike TRAF2, TRAF5, and TRAF6, the TRAF3 protein fails to activate NF-B and JNK when overexpressed in cells, thus leaving uncertain its role in signal transduction. TRAF3 was initially reported to interfere with NF-B activation induced by overexpression of CD40, TNFR2, LMP1, or TRAF2 (41). However, our CD40 mutant with selective impairment of TRAF3 binding (e.g. Q263A) did not exhibit enhanced NF-B activation compared with WT CD40. TRAF3 has also been implicated in cell growth control, although the mechanism is uncertain. Overexpression of TRAF3, for example, has been reported to inhibit growth of epithelial cancer cells, whereas TRAF3⌬N mutants can partly negate growth suppression induced by CD40, LMP1, or LT␤R in carcinoma cells (52,53). Expression of TRAF3(⌬N) mutants in a B-cell lymphoma line has also been reported to interfere with selected CD40-mediated signal transduction events but not growth inhibition (54). It remains uncertain, however, whether these effects of overexpression of TRAF3(⌬N) mutants are a reflection of a requirement for TRAF3 for these CD40-mediated events, versus the result of interference with other TRAF family members.
In this report, we describe two novel functions for TRAF3. First, TRAF3 can modulate NF-B induction caused by overexpression of TRAF5, with higher TRAF3:TRAF5 ratios resulting in a marked elevation in NF-B activity. Thus, contrary to prior reports, which suggested an inhibitory role for TRAF3 in NF-B induction, under at least some circumstances TRAF3 can enhance NF-B activation. Second, TRAF3 can recruit TRAF5 to the CD40 receptor complex, apparently by forming TRAF3-TRAF5 hetero-oligomers (13).
Although our data suggest that TRAF6 does not directly bind CD40, the membrane-proximal region of CD40, which is retained in the CD40(⌬32) mutant, reportedly mediates TRAF6 association with CD40 receptor complexes (21). By analogy to TRAF6 involvement in interleukin-1 receptor signaling, where an adaptor protein, MyD88, is required for TRAF6 recruitment (39,55,56), we presume that an unidentified linking protein is necessary for bridging TRAF6 to CD40. The presence or activity of a hypothetical TRAF6-binding adaptor protein could be tissue-or stimulus-specific, accounting for differences in the ability of various groups to demonstrate TRAF6 association with CD40 receptor complexes. Previous reports involving overexpression of dominant-negative mutants of TRAF6 have suggested that TRAF6 plays a role in CD40-mediated induction of NF-B and ERK1 (12,57). Two observations made here also support a role for TRAF6 in CD40 induction of NF-B. First, a CD40 mutant (E235A), which reportedly fails to recruit TRAF6 (21), displayed reduced ability to active both NF-B and JNK within the context of the full-length CD40 protein and was almost completely inactive when evaluated within the context of the CD40(⌬32) truncation mutant, which fails to bind other TRAFs. Second, co-expression of TRAF6(⌬N) with CD40(⌬32) markedly reduced CD40-mediated NF-B induction. Thus, we speculate that the membrane-proximal region (first 30 amino acids) of CD40 contains a binding site for a factor that recruits TRAF6 to receptor complexes, thereby promoting NF-B induction. In contrast to NF-B, however, TRAF6(⌬N) did not suppress JNK activation by CD40(⌬32). This observation suggests that a TRAF-independent pathway for JNK activation may be stimulated by the membrane-proximal region of CD40. It remains to be determined whether this pathway for JNK activation involves the same hypothetical adaptor protein that recruits TRAF6 to CD40 receptor complexes, but the finding that the E235A mutation suppressed both NF-B and JNK activation is consistent with this idea. Further analysis of the components of CD40 receptor complexes and the mechanisms that link CD40 to downstream effectors such as the protein kinases NIK, RIP2, ASK1, and GCK (18,34,54,58) are required to clarify the role of TRAF6 and other TRAF family proteins in CD40 signal transduction.
TRAF2, TRAF5, and TRAF6 are all capable of inducing NF-B induction and JNK activation. The data provided here and in other reports indicating that the cytosolic domain of CD40 contains at least three distinct sites that permit the direct or indirect recruitment of TRAF2, TRAF5, and TRAF6 to receptor complexes implies redundancy in the mechanisms by which CD40 signals. Multiple variables, including tissue-specific differences in the repertoire of TRAF family proteins ex-pressed in cells and differential preferences among individual TRAF family members for binding specific downstream kinases, however, may permit cell context-dependent responses to CD40 that account for its pleiotrophic actions and explain its differential effects on various cell lineages.