Dominant Negative Mutants of TRAF3 Reveal an Important Role for the Coiled Coil Domains in Cell Death Signaling by the Lymphotoxin-β Receptor*

Ligation of the lymphotoxin-β receptor (LTβR) recruits tumor necrosis factor receptor-associated factor-3 (TRAF3) and initiates cell death in HT29 adenocarcinoma cells. The minimal receptor binding domain (TRAF-C) defined by two hybrid analyses is not sufficient for direct recruitment to the ligated receptor. A series of TRAF3 deletion mutants reveal that a subregion of the coiled coil motif is required for efficient recruitment to the LTβR. Furthermore, the ability of TRAF3 to self-associate maps to an adjacent subregion. A TRAF3 deletion mutant that lacks the N-terminal zinc RING and zinc finger motifs, but retains the coiled coil and TRAF-C motifs, competitively displaces endogenous TRAF3 from the LTβR. A second TRAF3 mutant that lacks the receptor binding domain, yet contains the TRAF3 self-association domain, prevents TRAF3 homodimers from being recruited to the LTβR. Both of these mutants have a dominant negative effect on cell death and demonstrate that the recruitment of TRAF3 oligomers is necessary to initiate signal transduction that activates the cell death pathway.

The lymphotoxin-␤ receptor (LT␤R 1 ), a member of the tumor necrosis factor receptor (TNFR) superfamily, binds the cell surface form of LT (1). Cell surface LT is comprised of two subunits, LT␣ and ␤, arranged as a heterotrimer of LT␣ 1 ␤ 2 stochiometry (2). LT␤R is expressed on most cell types, including cells of epithelial and myeloid lineages but is conspicuously absent on T and B lymphocytes (3). LT␤R is implicated as a critical mediator of lymphoid organogenesis and of the formation of germinal centers during immune responses (4,5). Signaling through LT␤R activated with LT␣ 1 ␤ 2 or agonist anti-LT␤R antibodies induces cell death of certain adenocarcinoma tumor cells (6), stimulates chemokine secretion (7), and up-regulates the expression of integrins (8). LT␤R ligation can also regulate gene expression by activation of nuclear factor B (NFB) (9), a transcription factor important for inflammation and protecting cells from apoptotic death.
Signaling through LT␤R, and other TNFR family members that lack a death domain (10), involves interactions with TNF receptor-associated factors (TRAFs). There are six known TRAFs identified to date (named TRAF1 through 6) (11)(12)(13)(14)(15). TRAF molecules display a distinct arrangement of structural motifs with multiple zinc binding motifs at the N terminus, which includes a RING finger and several zinc fingers (16,17). TRAF1 is a noted exception in that it is lacking an N-terminal RING finger (11). The N-terminal region of TRAFs 2, 5, and 6 is the effector domain that controls activation of NFB (14,15,18). The C-terminal half of a TRAF molecule is known as the "TRAF" domain that displays a high degree of sequence homology that defines the TRAF family of signal transducers. The TRAF domain can be subdivided into N-terminal (TRAF-N) and C-terminal (TRAF-C) domains. The TRAF-N domain contains a region of ␣-helix-forming potential, which in TRAF3 and TRAF5 is an extended coiled coil domain (14). The TRAF domain is thought to be responsible for binding to the cytoplasmic tail of the TNFRs as defined by protein interaction assays, such as yeast-2 hybrid or GST fusion proteins (12,18,19). TRAF3 interacts with several members of the TNFR superfamily including LT␤R, CD40, CD30, and TNFR80 (CD120b) (9,12,20), as well as LMP1, a multitransmembrane protein encoded by the dominant oncogene of Epstein-Barr virus (21), and I-TRAF (TANK) (22), a cytosolic protein whose function is presently unclear. TRAF3 self-associates forming homo-oligomers (23), a feature observed for other TRAFs. TRAF3 association with the LT␤R is ligand-dependent and occurs within a minute following addition of LT␣ 1 ␤ 2 or agonist antibodies, whereas its association with LMP1 is constitutive (9,12).
Mice lacking TRAF3 by gene deletion exhibit a profound failure in homeostasis of the hemopoietic system and other organ systems, and exhibit a specific defect in T cell responses to antigens (24). It is unclear how TRAF3 signals these functions; however, a role for TRAF3 is beginning to emerge as a regulator of cell viability. Deletion of the N-terminal half of TRAF3 creates a dominant negative mutant that effects LT␤Rmediated death but not activation of NFB (9) indicating a bifurcation in the LT␤R signaling pathway. This mutant did not alter death induced by Fas, clearly distinguishing those pathways initiated by the death domain receptors and the TRAF binding receptors, all of which can induce apoptosis in different contexts (9). Additional evidence linking TRAF3 to cell growth regulation was shown with a dominant negative mutant that suppressed tumor growth mediated by CD40 or LMP1 (21). Overexpression of TRAF3 does not directly activate NFB, unlike TRAF2, 5, or 6 (14, 23, 25), but TRAF3 was shown to suppress NFB activity induced by CD40, TNFR80, and LMP1 (23,26). The fact that activated NFB inhibits apoptosis (27,28) further implicates TRAF3 as a regulator of cell death.
Our previous study of a TRAF3 mutant (⌬1-367) that blocked cell death signaled by LT␤R indicated that the mutant protein did not displace endogenous TRAF3 from the LT␤R but was recruited to the LT␤R at an equimolar ratio with endogenous TRAF3 (9). The simplest explanation of this data was that the TRAF3 mutant-blocked recruitment of an effector molecule, possibly TRAF3 itself, implying that self-association of TRAF3 may be crucial for propagation of signals for cell death. To address this issue we have made several deletion mutants of TRAF3 to assess their function in cell death mediated by LT␤R by measuring their ability to directly bind LT␤R, to self-associate and to form signaling complexes with the LT␤R. The results indicate that the formation of TRAF3 oligomers is required for signal transduction through LT␤R and that the coiled coil region of TRAF3 is critical for recruitment to the LT␤R.

EXPERIMENTAL PROCEDURES
Cells and Reagents-Recombinant human TNF (29) and soluble LT␣ 1 ␤ 2 (30) produced with a truncated version of LT␤ lacking the cytosolic and transmembrane domains were provided by Jeffrey Browning (Biogen). Anti-Fas mouse IgM monoclonal CH11 was obtained from MBL, Nagoya, Japan. M2 and M5 anti-Flag IgG1 monoclonal antibodies were obtained from Eastman Kodak Co. HT29.14s cells were obtained from Jeffrey Browning (6). 293 (human embryonic kidney) cells were obtained from ATCC. Myc-LT␤R-293 cells were generated by infection with retrovirus expressing the Myc-epitope-tagged LT␤R as described below. Overexpression of Myc-LT␤R was confirmed by flow cytometric analysis of Myc-LT␤R-293 cells with an anti-Myc monoclonal antibody 9E10 (data not shown). All cells were maintained in Dulbecco's modified Eagle's medium (Life Technologies, Inc.) with 10% fetal bovine serum.
Plasmid Construction-The Flag epitope-tagged TRAF3 (Flag-TRAF3) mutants were created by PCR amplification from TRAF3 coding sequences using the pSG5-LAP1 vector as template (12) and oligonucleotide pairs as shown below. The resulting PCR products were digested with EcoRI/BamHI and ligated into the EcoRI/BamHI site of pFlag-CMV-2 (Kodak). The following primer pairs were used for T3 ( The coding regions of the Flag-TRAF3 mutants were inserted into the Moloney murine leukemia virus-based vector pBABEpuro, which contains a puromycin resistance gene (31). TRAF3 mutants in pFlag-CMV-2 were used as a template for PCR. The PCR products were generated using the following primer pair; 5Ј-GTGGAGCGAGATCTAC-CATGGACTAC-3Ј and 5Ј-GATACCGTCGACTGGAGTGGCAACTT-3Ј. PCR products were digested with BglII/SalI and ligated into the BamHI/SalI site of pBABEpuro.
Myc-LT␤R-pBABE was constructed by replacing the N-terminal 27 amino acids of human LT␤R with a type I leader sequence and a c-Myc decapeptide epitope followed by subcloning into pBABEpuro using the following strategy. Human LT␤R cDNA sequence was generated by PCR using a 5Ј primer containing an AlwNI site (5Ј-CCGGTCCA-GAAGCTGTCGCAGCCCCAGGCGGTGCCT-3Ј), a 3Ј primer containing a SalI site (5Ј-GGAACGCGTCGACCCGTCAGTCATGGGTGATA-AATTGGT-3Ј), and LT␤R cDNA as a template (1). The 1255-base pair LT␤R PCR product was digested with SalI and subjected to partial digestion with AlwNI to produce a 1235-base pair AlwNI/SalI fragment. The c-Myc cassette complete with leader sequence was obtained by linearization of sLT␤-Myc in pCDM8 on the 3Ј side of c-Myc with AlwNI followed by ligation of the c-Myc coding region, in frame, to the 1235base pair AlwNI/SalI LT␤R fragment (1). The ligation mix was subjected to PCR using a 5Ј primer containing a BglII site 5Ј-GGAACGCGTCGACCCGTCAGTCATGGGTGA-3Ј and the 3Ј primer 5Ј-CCGTCCCACCATCGGGCGCGGATCAGATCT-3Ј to amplify the desired full-length 1400-base pair Myc-tagged LT␤R PCR product. The PCR product was purified, cut with BglII and SalI, and ligated into the BamHI and SalI sites of pBABEpuro to create Myc-LT␤R-pBABE. PCR was performed using pfu DNA polymerase (Stratagene, La Jolla, CA). PCR and ligation products were purified using Wizard PCR Preps DNA purification system (Promega, Madison, WI).
Retrovirus Production-Moloney retroviral virions were produced as described (32). Briefly, the Phoenix (NX) amphotrophic packaging cell line was plated at 2.5 ϫ 10 6 cells/10-cm 2 culture dish for 18 -24 h before transfection (33). Cells were transfected with 7 g of the desired pBA-BEpuro construct using a standard calcium phosphate protocol as described (34) except chloroquine (25 M final) was added to the cells 5 min prior to addition of the CaPO 4 DNA precipitate. For production of control virus supernatants used for mock infections, Phoenix cells were transfected with the pBABEpuro vector. After 24 h the cells were gently washed, and fresh medium was added. Virus-containing supernatant was harvested at 24 h after transfection and stored at 4°C. For virus infection HT29.14s or 293 cells plated at 1 ϫ 10 6 /10-cm 2 dish (approximately 30% confluency) were infected with 3 ml of viral supernatant containing 5 g/ml polybrene for 3 h. The virus supernatant was diluted with 5 ml of fresh medium containing 5 g/ml polybrene, and the cells were then cultured overnight. 48 h after infection, cells were selected with puromycin (1 g/ml). Typically 50 -90% of the infected cells survived puromycin selection, and all assays described were performed within 1-2 weeks following selection.
In Vitro Binding and Immunoprecipitation Assays-TRAF proteins were generated in vitro using a TnT T7 coupled wheat germ extract kit (Promega) and [ 35 S]methionine (1175 Ci/mmol, in vitro translation grade, ICN, Costa Mesa, CA). A T7 promoter was introduced 5Ј of the TRAF coding region by PCR using the TRAF mutants in pFlag-CMV-2 as template and the primers 5Ј-TAATACGACTCACTATAGGGAGGTC-TATATAAGCAG-3Ј and 5Ј-CTAGAGATGGGGTCACAGGGATG-3Ј. The PCR products (ϳ1 g each) were used directly as templates for the in vitro transcription/translation reaction. Construction of the fusion proteins between LT␤R cytoplasmic domain and glutathione S-transferase (LT␤R-GST) have been described (9). Binding assays using LT␤R-GST or GST (2 g of protein/reaction) bound to glutathione-Sepharose beads were incubated with 18 fmol of each 35 S-labeled protein in 0.5 ml of binding buffer (1% Nonidet P-40, 10% glycerol in phosphate-buffered saline) as described (12). Dried gels were visualized using PhosphorImaging (Molecular Dynamics Storm). 35 [S]methionine-TRAF3 was quantitated by trichloroacetic acid precipitation and scintillation counting of the in vitro translation product. A reference gel relating density volume of the band image to radioactivity was generated by Molecular Dynamics ImageQuant software. The moles of TRAF3 or mutants in the receptor-bound fraction were then calculated from the image density. Nonspecific binding of 35 S-TRAF3 or mutants to GST alone was Ͻ0.04% of the input.
For metabolic labeling of HT29.14s cells and direct immunoprecipitation of TRAFs with the anti-Flag monoclonal M2, HT29.14s were plated at 1.5 ϫ 10 6 cells/10 cm 2 , 18 h prior to labeling with [ 35 S]methionine (Express Label, NEN life Science Products, 45.3 TBq/mmol) as described (9). Detergent lysates were prepared with 1% Nonidet P-40 nonionic detergent. Extracts were precleared twice with 25 l of protein G-Sepharose beads and 5 g of mouse IgG1. Immunoprecipitation of Flag-tagged proteins was then carried out by the addition of 5 g of M2 and 25 l of protein G-Sepharose beads followed by mixing for 2 h. The beads were washed, and the proteins were eluted and subjected to SDS-PAGE as described (9). Dried gels were visualized by PhosphorImaging.
To detect TRAF3 binding to LT␤R in HT29.14s cells, Western blot analysis was performed as described previously (9). Briefly, HT29.14s cell monolayers containing 10 7 cells were treated with 1 nM recombinant soluble LT␣ 1 ␤ 2 (30) for 15 min at 37°C. Cell monolayers were extracted with a 1% Triton X-100 containing buffer and subjected to immunoprecipitation with goat anti-LT␤R IgG. After SDS-PAGE, the gel was blotted onto polyvinylidene fluoride membrane, and TRAF3 was detected with a rabbit anti-TRAF3 antiserum (9). The TRAF3 antiserum was used at a 1:1000 dilution and showed no cross-reactivity in Western blots with TRAF1, 2 or 5.
For demonstration of TRAF3 self-association 293 cells (2 ϫ 10 6 ) were transfected with 7 g of pSG5-TRAF3 along with 7 g of the indicated TRAF3 mutant in pFlag-CMV-2. DNA content was normalized to 14 g total using the appropriate parent vector. Cells were harvested 48 h after transfection as described above and immunoprecipitated with N-terminal specific anti-TRAF3 antibody (rabbit anti-CRAF1, [H2O] Santa Cruz Biotechnology Inc. Santa Cruz, CA). Flag-TRAF3 proteins were detected by Western analysis using Immobilon-P transfer membrane (Millipore, MA). After transfer the membrane was blocked in 2% gelatin in TBS (50 mM Tris, pH 7.4, 150 mM NaCl) for 2 h at 37°C followed by blocking with 7.5% skim milk in TBS for 2 h at room temperature. Flag-TRAF3 proteins were detected using an anti-Flag monoclonal antibody M5 (Santa Cruz Biotechnology, Inc.) at 2 g/ml in TBS-Tween (0.05% Tween 20) for 30 min at room temperature. After two brief washes in TBS-Tween, membranes were treated with sheep anti-mouse Ig (horseradish peroxidase conjugated, Amersham) and illuminated using an ECL detection kit (Amersham).
To detect binding of the Flag-TRAF3 mutants to LT␤R in 293 cells, Myc-LT␤R-293 cells were transfected with 7 g of the indicated TRAF3 plasmid. 48 h after transfection the cells were treated with LT␣ 1 ␤ 2 for 15 min prior to harvesting. Cell lysates were subjected to immunoprecipitation with a goat anti-LT␤R antibody (9) and subjected to Western blot analysis as described above. For preparation of whole cells extract, 5 ϫ 10 5 cells were lysed by boiling in lysis buffer (1% SDS, 50 mM Tris, pH 6.8, 20 mM EDTA, 50 mM NaF, 1 mM orthovanadate, 1 mM phenylmethylsulfonyl fluoride, 2.5% 2-mercaptoethanol) for 15 min.

TRAF3 Domains Required for Binding LT␤R-TRAF3
is comprised of an N-terminal zinc RING and zinc finger motif and a C-terminal TRAF domain that is divided into an N-and C-terminal region (Fig. 1). Separating these domains is a region of high ␣-helix potential, the coiled coil or leucine zipper domain, which precedes the TRAF-N domain. The coiled coil domain of TRAF3 is predicted to be tripartate in structure with three distinct coil regions (Fig. 1) (35, 36) where coils 1 and 2 are separated by a single amino acid. Deletion mutants of TRAF3 that roughly demarcate the domains of TRAF3 were studied for their ability to bind LT␤R (Fig. 1). LT␤R-GST binds full-length TRAF3 with high efficiency with Ͼ20% of the input TRAF3 represented in the bound fraction (Fig. 1B). LT␤R bound to the RING finger deletion mutant T3(⌬1-113) and the RING/zinc finger mutant T3(⌬1-258) with efficiency comparable to wild type TRAF3. By contrast deletion of coils 1 and 2 of the coiled coil domain with T3(⌬1-339) resulted in a Ͼ90% loss of TRAF3 binding to LT␤R-GST illustrating the importance of coils 1 and 2 to TRAF3-receptor interactions. Deletion of coil 3 resulted in an additional loss of binding demonstrating that the TRAF-C domain is only capable of weakly interacting with LT␤R. T3(⌬382-568) and T3(aa114 -381) contain coils 1 and 2 and failed to bind LT␤R-GST suggesting that sequences, which are C-terminal to all three coils, are necessary to support high efficiency binding of TRAF3 to LT␤R.
To compare the recruitment efficiency of the TRAF3 mutants in intact cells, Myc-LT␤R-293 cells were transfected with each of the Flag-TRAF3 deletion mutants. Full-length Flag-TRAF3 co-immunoprecipitated with the LT␤R as expected (Fig. 2A). TRAF3 mutants T3(⌬1-113) and T3(⌬1-258), which contain the entire coiled coil domain, were also recruited to the receptor-signaling complex in agreement with the in vitro binding studies. Surprisingly, T3(⌬382-568) and T3(⌬1-339), two mutants which do not bind LT␤R-GST efficiently in vitro, are recruited strongly to LT␤R in 293 cells. Both T3(⌬382-568) and T3(⌬1-339) contain coil 3, suggesting that coil 3 can support recruitment of these molecules to the receptor when other cellular proteins are present. Deletion of coil 3 in T3(⌬1-388) prevents recruitment to LT␤R further illustrating the importance of coil 3 to TRAF3-receptor interactions. Interestingly, T3(aa114 -381) contains all three coils and is expressed at high levels but is not recruited to the receptor. Deletion mutants T3(⌬382-568) and T3(⌬1-339) contain coil 3 and are capable of binding LT␤R only in the presence of other cellular proteins. We reasoned that these mutant proteins may bind LT␤R by oligomerizing with endogenous full-length TRAF3. To determine if coil 3 is capable of supporting TRAF3 self-oligomerization, full-length Flag-TRAF3 was overexpressed in 293 cells along with several different TRAF3 deletion mutants. The TRAF3 mutants and full-length Flag-TRAF3 were expressed at comparable levels as determined by Western blot analysis of whole cell extracts with an anti-Flag antibody (data not shown). Full-length TRAF3 was immunoprecipitated from the cell lysates with an anti-TRAF3 antibody that recognizes an N-terminal epitope in TRAF3 that is not present in the TRAF3 deletion mutants. Flag-TRAF3 proteins were then detected by Western blot analysis with an anti-Flag antibody (Fig. 3). TRAF3 mutants T3(aa114 -381), T3(⌬1-113), T3(⌬1-258), and T3(⌬1-339) all contain coil 3 and co-immunoprecipitate with full-length TRAF3. Mutants lacking coil 3, such as T3(aa114 -273) and T3(⌬1-388), lose the ability to oligomerize with full-length TRAF3. Thus, self-association of TRAF3 requires amino acids 339 -388. These results support the hypothesis that TRAF3 deletion mutants that contain coil 3 bind to the LT␤R-signaling complex via their association with endogenous full-length TRAF3.
To test if T3(⌬1-258) should also act in a dominant negative fashion to block LT␤R-mediated cell death, HT29.14s cells expressing T3(⌬1-258) or T3(⌬1-339) were tested for their sensitivity to cell death by LT␣ 1 ␤ 2 . Both mutants were significantly less sensitive in repeated assays to LT␣ 1 ␤ 2 (7-10-fold reduction in the IC 50 ) than mock-infected cells (Fig. 5A). The sensitivity to TNF was unaltered in these mutants compared with mock-infected lines (Fig. 5B). These data indicate specific inhibition of LT␤R-mediated cytotoxicity by both of the TRAF3 dominant negative mutants by distinct mechanisms of inhibition. The high efficiency of gene transfer by retrovirus, resulting in the survival of most of the cells during puromycin selection, eliminates possible artifacts due to selection of rare clones and indicates the phenotype is due to the expressed gene. DISCUSSION TRAF3 requires the coiled coil domain for efficient interactions with LT␤R and the zinc RING and zinc finger domains for signaling cell death. Previous studies have reported that various zinc RING and zinc finger deletion mutants of TRAF3 bind CD40 (37), CD30 (20), and LMP1 (12). With a series of TRAF3 deletion mutants, we identified a region of TRAF3 located between amino acids 258 -339 (coils 1 and 2) that is critical for efficient binding to the LT␤R. Molecular modeling predicts two leucine zipper-like coiled coils (coils 1 and 2) located within this region (amino acids 267-339) (35,36). These coils are separated by a single cysteine residue and may act together to define a functional LT␤R binding domain. However, this coiled domain  (⌬1-258), or mock-infected cells (Vector) were treated with (ϩ) or without (Ϫ) soluble LT␣ 1 ␤ 2 for 15 min. LT␤R was immunoprecipitated from cell lysates with goat anti-LT␤R Ig and subjected to SDS-PAGE. TRAF3 was then detected by Western blot analysis with an anti-TRAF3 antibody specific for the RING finger domain of TRAF3 (9). and the C-terminal 186 amino acids of the TRAF-C domain are co-dependent on each other for binding GST-LT␤R. Neither the coiled domain nor the TRAF-C domain can independently support high efficiency binding to GST-LT␤R. Previous studies using the yeast-2 hybrid assay reported that the TRAF domain of TRAF3, residues 415-567, was sufficient for binding CD40 (37). By contrast we found that an even larger region of TRAF3, exemplified by mutant T3(⌬1-339), could not support efficient binding to GST-LT␤R. This apparent discrepancy could be due to differences in the binding stringency between the yeast-2 hybrid assay and in vitro binding assays. Alternatively, the cytoplasmic domains of CD40 and LT␤R may interact with different regions of TRAF3. By comparison, the TRAF-C domain of TRAF2 in the yeast-2 hybrid complementation assay associates with the cytoplasmic tail of CD30 but failed to coimmunoprecipitate with TNFR80, suggesting that CD30 and TNFR80 may also recognize and bind distinct regions of TRAF2 (18,38).
Our data suggest that the RING finger domain of TRAF3 may be required for correct folding of coils 1 and 2 and subsequent interaction of TRAF3 with LT␤R. The RING finger deletion mutant T3(aa114 -381) associates with full-length TRAF3 but does not co-precipitate with LT␤R by its association with endogenous TRAF3. However, T3(⌬382-568), which differs from T3(aa114 -381) only by the addition of the RING finger, efficiently co-immunoprecipitates with LT␤R. Many factors such as upstream and downstream sequences have been reported to affect the proper folding of coiled coil domains. Examples include influenza hemagglutinin that requires dis-sociation of its globular head group to form a coiled coil structure (39), and the bZip transcription factors that need DNA to form coiled coils (40). Removal of the RING finger in T3(aa114 -381) may result in an altered coiled structure that prevents recruitment to the receptor.
Self-association of TRAF3 into oligomers requires a specific coiled region within its TRAF-N domain. Deletion of 49 amino acids from the N-terminal side of T3(⌬1-339) resulted in a complete loss of TRAF3 self-association and identified a region between amino acids 339 -388 that is critical for TRAF3 selfassociation. This domain contains a leucine zipper-like coiled coil (coil 3) between amino acids 345-367 that may be involved in TRAF3 oligomerization (35). T3(⌬1-339) contains the TRAF3 self-association domain and binds full-length TRAF3, but lacks coils 1 and 2 that are necessary for receptor binding. This protein failed to bind LT␤R in vitro yet co-immunoprecipitated with LT␤R in cell lysates. We reasoned that this TRAF domain mutant was recruited to the LT␤R in cells by forming oligomers with endogenous TRAF3. Using a computer algorithm (Multi-coil) that predicts whether coiled coil domains prefer forming dimers or trimers, coil 3 of TRAF3 is predicted to form trimers (41), a feature also found in TRAF4 and 5 suggesting that these proteins also forms oligomers.
The TRAF-C domain mutant T3(⌬1-388) provides insight into the structural requirements necessary for the TRAF-C domain to self-associate. This mutant failed to form oligomers with TRAF3. TRAF2 behaves in an analogous fashion in that the TRAF2-C domain also fails to independently oligomerize with TRAF2 in a cellular coexpression system (18), although a similar mutant complements itself in the yeast-2 hybrid assay (19). However, TRAF2 oligomerization is restored when its TRAF-C domain is fused to the RING and zinc finger domains (18). This evidence suggests that N-terminal sequences may be required for functional conformation of the TRAF-C domain. We suggest that fusion of the TRAF-C domain of TRAF3 to the DNA binding domain of GAL4 may have provided the necessary structural integrity that allowed for TRAF-C self-association in yeast-2 hybrid analysis (19). In addition, the GAL4 DNA binding domain forms dimers in yeast-2 hybrid analysis providing added stability to protein-protein interactions that utilize multimeric associations (42). For example, the dimerization of the TRAF-C domain initiated by the GAL4 DNA binding domain may provide an interface for a third TRAF-C domain to bind. In the absence of coil 3 or GAL4 to anchor the TRAF-C domains together, TRAF-C interactions may be relatively weak and undetectable in most other binding assays.
The inhibition of LT␤R-mediated cell death by both T3(⌬1-258) and T3(⌬1-339) suggest that recruitment of TRAF3 oligomers to the cytoplasmic tail of LT␤R is required for signaling cell death. T3(⌬1-258) binds LT␤R as a self-associated oligomer and competitively displaces endogenous TRAF3 from the receptor. However, the inhibition of cell death by T3(⌬1-339) is by a distinct mechanism. The T3(⌬1-339) protein is recruited to the LT␤R as an oligomer with the wild type TRAF3 molecule but does not competitively displace endogenous TRAF3 from LT␤R like T3(⌬1-258). This implicates that the formation of TRAF3 homo-oligomers or the recruitment of TRAF3 homo-oligomers to the signaling complex is necessary to propagate the death signal originating from LT␤R. However, TRAF3 is recruited to the LT␤R by a monoclonal antibody that does not by itself trigger death indicating that TRAF3 recruitment alone is not sufficient to induce death (9). This suggests that additional processes are required to propagate signals for the death pathway. These results have lead to the hypothesis that the zinc RING and zinc fingers of TRAF3 are effector domains that initiate or regulate the cell death signal from the LT␤R.
Ligation of the LT␤R initiates activation of NFB and cell death by distinct signaling pathways (9). LT␤R-mediated activation of NFB may involve recruitment of TRAF2 or TRAF5 (14), whereas TRAF3 appears to play a role in cell death. One effect of NFB activation is to act as a negative regulator of apoptosis (27,28), a finding that suggests a role for TRAF3 in regulation of NFB. In support of this contention, overexpression of TRAF3 inhibited LMP1-mediated NFB activation apparently by competitive displacement of TRAF2 from LMP1 (26). Previously we found that overexpression of a dominant negative TRAF3 mutant that blocked cell death (T3⌬1-367) had no detectable affect on NFB activation in these cells (9). Furthermore, overexpression of T3(⌬1-258), the TRAF3 mutant that competitively displaces wild-type TRAF3, did not make the cells more responsive to LT␣ 1 ␤ 2 -induced death as would be expected for an inhibitor of NFB activation (28). Thus, NFB activation via the LT␤R is probably not regulated by TRAF3 through competitive displacement of TRAF2 or TRAF5. This conclusion raises the question of what factor or process controls recruitment of different TRAFs to the LT␤R. One explanation is that TRAF2 and TRAF5 bind LT␤R at sites distinct from those required for TRAF3. Alternatively, selective recruitment may be achieved by regulation of TRAF3 expression in different cell types. TRAF3 is abundantly expressed in most cells (43), 2 whereas TRAF5 is poorly expressed in tumor cells such as HT29, but is relatively abundant in nontransformed cells such as fibroblasts and follicular dendritic cells (44). In addition, multiple forms of TRAF3 exist in different cell types suggesting that additional mechanisms may be at work to achieve selective recruitment and activation. 2 It is becoming clear that signaling through TNFR cytoplasmic domains involves complex aggregations of various signaling molecules, each with distinct functional domains. The studies presented here provide a better understanding of the functional domains of TRAF3, a putative mediator of cell death for nondeath domain receptors. Identification and characterization of proteins that participate in LT␤R-TRAF3 signal transduction may yield new insights into understanding pathways that lead to apoptosis.