Direct binding of Fas-associated death domain (FADD) to the tumor necrosis factor-related apoptosis-inducing ligand receptor DR5 is regulated by the death effector domain of FADD.

Members of the tumor necrosis factor superfamily of receptors induce apoptosis by recruiting adaptor molecules through death domain interactions. The central adaptor molecule for these receptors is the death domain-containing protein Fas-associated death domain (FADD). FADD binds a death domain on a receptor or additional adaptor and recruits caspases to the activated receptor. Tumor necrosis factor-related apoptosis-inducing ligand (TRAIL) signals apoptosis through two receptors, DR4 and DR5. Although there is much interest in TRAIL, the mechanism by which FADD is recruited to the TRAIL receptors is not clear. Using a reverse two-hybrid system we previously identified mutations in the death effector domain of FADD that prevented binding to Fas/CD95. Here we show that these mutations also prevent binding to DR5. FADD-deficient Jurkat cells stably expressing these FADD mutations did not transduce TRAIL or Fas/CD95 signaling. Second site compensating mutations that restore binding to and signaling through Fas/CD95 and DR5 were also in the death effector domain. We conclude that in contrast to current models where the death domain of FADD functions independently of the death effector domain, the death effector domain of FADD comes into direct contact with both TRAIL and Fas/CD95 receptors.

Members of the tumor necrosis factor (TNF) 1 superfamily of receptors induce a variety of cellular responses including apoptosis, cellular differentiation, and proliferation. A subfamily of these receptors contains a death domain (DD) that is essential for transducing the apoptotic signal. Fas/CD95 is the best characterized member of this family. Binding of Fas ligand (FasL) to a preformed Fas/CD95 trimer (1) results in dimerization of two trimers (2) and higher levels of oligomerization (3). These activated receptors signal the apoptotic response by recruiting FADD to the cytoplasmic DD of the receptor to form the death-inducing signaling complex (DISC). FADD consists of two distinct domains, a DD that binds to the DD of Fas/CD95 and a death effector domain (DED) that binds to DEDs on caspase-8 and caspase-10 (4). The DD and DED have similar structural folds consisting of six anti-parallel ␣-helices, and both form globular protein structures whose only known function is to interact with other proteins (5). Thus, it is thought that binding of ligand to Fas/CD95 results in the recruitment of FADD through DD interactions followed by caspases through DED interactions. The induced proximity of two or more initiator caspases in a complex with FADD and the receptor results in their dimerization and activation (6,7). Proteolytic processing leads to a fully processed, active form of the caspase that can dissociate from the receptor complex. Once activated, these caspases can then cleave and activate effector caspases such as caspase-3 and other substrates to induce the characteristic phenotypes associated with apoptosis.
Activation of TNF receptor 1 requires an additional adaptor protein, TRADD. Binding of TNF␣ to TNF receptor 1 results in the recruitment of TRADD (8) followed by FADD (9,10), again through DD interactions. Similar to the Fas/CD95 receptor, recruitment of FADD to the complex leads to the recruitment and activation of caspase-8. Of note, recent studies suggest that the TRADD/FADD/caspase-8 complex dissociates from the receptor during TNF-induced apoptosis, implying an additional layer of regulation (11,12). Nonetheless, FADD is essential for TNF-induced caspase activation because FADD-deficient cells fail to undergo apoptosis when treated with TNF␣ and a dominant negative form of FADD that consists of only the DD can block TNF-induced death (13,14).
There is much interest in TNF-related apoptosis inducing ligand (TRAIL) because of its reported ability to induce apoptosis in tumor cells without affecting normal cells (15,16), suggesting that it may be useful for treating cancer. There are two "signaling" receptors DR4 and DR5, which, similar to Fas/ CD95, contain death domains. In addition, there are two "decoy" receptors, DcR1 and DcR2, that have truncated or completely absent DDs (17), suggesting that the DD is essential for apoptotic signaling through TRAIL. FADD is necessary for signaling through DR4 and DR5 because dominant negative FADD blocks TRAIL signaling (14), and FADD-deficient Jurkat cells do not undergo TRAIL-induced death (18). FADD, caspases-8 and caspase-10 are recruited to the TRAIL DISC, but TRADD was not detectable in this complex (19,20). These reports led to a model similar to that of Fas/CD95 in which FADD binds directly to both TRAIL receptors rather than through an adapter molecule (21).
Recently, we made the surprising discovery that specific mutations in the FADD DED could prevent binding to Fas/ CD95 (22). This suggests that the FADD DED contributes to the interaction with the Fas/CD95 receptor and that the DED and DD of FADD do not function independently as suggested in previous models. Here we show that mutations in the DED of FADD also prevent binding of FADD to DR5 and are unable to rescue TRAIL signaling in FADD-deficient cells. Compensating second site mutations that restore binding of such a FADD mutation to DR5 were identified in the DED of FADD and restore TRAIL-induced activation of caspases, further indicating a role for the DED in coordinating the FADD-DR5 interaction. Using a reverse two-hybrid approach we were able to identify only one mutation in the FADD DD that showed differential binding between Fas/CD95 and DR5. These data suggest that the same residues in both the DED and DD of FADD regulate binding of FADD to Fas/CD95 and DR5. Plasmids-pGB14 was made by cloning the ADH1 promoter, Gal4 DNA-binding domain, multiple cloning site, ADH1 terminator cassette from pGBKT7 into the KpnI and SacI sites of pRS314. DR5 (amino acids 209 -412), Fas (amino acids 177-335), and full-length catalytically inactive caspase-8 were made by PCR on the corresponding cDNA (23) and by cloning the product into pGB14. Full-length TRADD was cloned into pBTM-116. pACT3 and pcDNA-Puro plasmids have been described previously (22). Full-length FADD and FADD-DD (amino acids 79 -208) were cloned into pEGFP-C2 or pACT3. The cytoplasmic domains of DR5 (amino acids 209 -412) or Fas (amino acids 177-335) were cloned into pEGFP-C2. Amino acids 1-208 of FADD or 272-469 of DR5 were cloned C-terminally to FLAG. A more complete description of plasmids, maps, and sequences are available upon request.
Immunoprecipitation-HeLa cells were transfected with 2 g of pEGFP construct plus 2 g of FLAG or FLAG-FADD using FuGENE 6 (Roche Applied Science) in a 10-cm plate. After 18 h the cells were lysed in Triton X-100 lysis buffer (50 mM Tris, pH 8.0, 150 mM NaCl, 5 mM EDTA, 1% Triton X-100, 1ϫ protease inhibitors), and the soluble fraction was incubated with 30 l of M2-agarose (Sigma) for 4 h a 4°C. The beads were washed four times in Tris-buffered saline, and precipitated GFP fusions were detected by immunoblot.
DISC Immunoprecipitations-20 ϫ 10 6 BJAB cells were treated with 1.0 g/ml nonspecific mouse IgG (Sigma) or monoclonal antibody 631 (R & D Systems, Minneapolis, MN) cross-linked with an equal amount of anti-mouse Fc (Sigma) for 30 min. The precipitations were carried out using protein A/G-agarose beads essentially as described (19), and the bound proteins were detected by immunoblot.
Caspase Activation-Jurkat cells were seeded at a density of 1.0 ϫ 10 6 cells/ml in growth medium ϩ 1 g/ml cycloheximide. The cells were incubated for 6 h with FasL (2 ng/ml) or TRAIL (50 ng/ml cross-linked with an equal amount of anti-His 6 ), washed once with phosphatebuffered saline, and then harvested in Triton X-100 lysis buffer then analyzed by immunoblot.
Modeling of FADD Mutations-Free energies were obtained by using the thermodynamic cycle of (wild type protein Ϫ mutant protein) Ϫ (wild type residues Ϫ mutant residues). To determine the free energy for each mutant state and the wild types, the solvated energy was computed from snapshots taken from a 500 ps Molecular Dynamics simulation (one for each mutant/wild type) and averaged over each trajectory. Residue energies were obtained by taking the protein simulations, deleting all of the protein except the mutated residue(s) and computing the solvated energy for each snapshot. The generalized born molecular volume with surface area (26,27) approach was used to obtain the solvation energy and added to the CHARMM22 (28) force field energy to obtain the solvated energy for each snapshot.

RESULTS
FADD Binds Directly to DR5-We performed DR5 DISC immunoprecipitations to see whether FADD precipitated in the endogenous DR5 DISC. BJAB cells were stimulated for 30 min with either nonspecific IgG or an agonistic monoclonal antibody against DR5 (␣DR5), and the DR5 DISC was precipitated. Consistent with previous reports (19,20), we were able to detect both FADD and caspase-8 in cells stimulated with ␣DR5 but not in cells stimulated with IgG (Fig. 1A).
Although it has been suggested that FADD binds directly to DR5 (21), other reports indicate that an adaptor protein such as TRADD (29) or DAP3 (23) might be involved in recruiting FADD to the DR5 DISC. We therefore used a yeast two-hybrid assay to test for interaction between the cytoplasmic domain of DR5 and full-length FADD. Fas/CD95 was used as a positive control because it is known to interact directly with FADD (30). Both Fas/CD95 and DR5 interacted with FADD in yeast, suggesting that DR5 is recruited directly to the activated TRAIL receptor complex (Fig. 1B).
To test whether FADD could bind directly to DR5 in mammalian cells, we performed immunoprecipitation experiments with full-length FADD and the cytoplasmic domains of DR5 and Fas/CD95. Empty FLAG vector or FLAG-tagged FADD was transfected into HeLa cells with GFP-tagged Fas/CD95 or DR5 cytoplasmic domains. FLAG complexes were immunoprecipitated, and interaction was detected by immunoblotting for GFP. As shown in Fig. 1C, both Fas/CD95 and DR5 co-precipitated with FLAG-FADD but not with empty vector. We also observed a modest decrease in the amount of GFP-DR5 remaining in the lysate after immunoprecipitation. We conclude that similar to the Fas/CD95 model, FADD binds directly to DR5.
The Death Effector Domain of FADD Modulates Binding to DR5-Using a reverse two-hybrid strategy, we previously identified mutations in the DED of FADD that prevent binding to Fas/CD95 (22). These mutations do not disrupt overall protein FIG. 1. FADD binds directly to DR5. A, BJAB cells were stimulated for 30 min with nonspecific IgG or an agonistic DR5 antibody (␣DR5), and the endogenous DR5 DISC was precipitated. Both FADD and processed caspase-8 co-precipitate in cells stimulated with ␣DR5 but not in cells stimulated with IgG. B, the cytoplasmic domains of Fas/CD95 or DR5 fused to the Gal4 DNA-binding domain were tested for interaction with FADD in a directed two-hybrid assay. A color change indicates that FADD is able to interact with both Fas/CD95 and DR5 in yeast. C, immunoprecipitation (IP) experiments were performed to test for interaction between FADD and DR5 or Fas/CD95 in vivo. Both GFP-DR5 and GFP-Fas precipitated with FLAG-FADD, indicating a direct interaction. Whole cell lysates were blotted with anti-FLAG and anti-GFP to show equal transfection. structure because the mutated proteins still bind both TRADD and caspase-8 and are able to transduce caspase signaling in response to TNF␣ in mammalian cells (22). Using a directed yeast two-hybrid assay, we tested these FADD mutations to see whether they affected binding to DR5. A mutation in FADD at arginine 71, which is located in the loop between helices 5 and 6 of the DED (31), to either tryptophan or alanine prevented binding to DR5 (Fig. 2A). This suggests that similar to Fas/ CD95, the DED of FADD participates in binding to DR5.
To test whether FADD DED mutations could transduce TRAIL signaling, we stably expressed GFP, FADD, FADD (R71W), or FADD (R71A) in FADD-deficient Jurkat cells (24). Jurkat cells express very little DR4 so almost all TRAIL signaling is through DR5 (19). FADD-deficient Jurkat-GFP cells do not express any FADD protein that can be detected by immunoblotting, and the expression level of wild type FADD and the FADD mutations was similar between each cell line (Fig. 2B). The earliest signaling event after ligand binding is the recruitment of FADD and processing of caspase-8; FADD mutations that cannot bind to DR5 should therefore be unable to rescue TRAIL-induced caspase-8 processing. These cells were treated with TRAIL, and caspase-8 processing was measured by immunoblot. The cells expressing wild type FADD showed cleavage of caspase-8 in response to TRAIL, whereas cells expressing GFP or the FADD DED mutations did not (Fig.  2C). We also measured caspase-3 processing because it is cleaved and activated by caspase-8. Caspase-3 was cleaved only in cells expressing wild type FADD (Fig. 2C). Therefore, FADD proteins containing the DED point mutations that prevent binding to DR5 cannot rescue the phenotype associated with FADD deficiency. These data indicate that the binding phenotype observed in yeast correlates with signaling ability in mammalian cells and the DED of FADD modulates binding to DR5.
Mutations That Restore Binding to DR5 Are Located in the DED of FADD-We next sought to identify secondary compensating mutations that would restore binding of FADD DED mutations to DR5. Using a forward two-hybrid approach, we performed a second round of random mutagenesis on FADD (R71A) and screened for second site mutations that restore the binding activity of FADD (R71A) (22). The second site mutations were located in helix 5 of the FADD DED: glutamate 61 to lysine, leucine 62 to phenylalanine, and glutamate 65 to lysine. These mutations restored binding of FADD (R71A) to both DR5 and Fas/CD95 (Fig. 3A).
We introduced FADD molecules with these double mutations into FADD-deficient Jurkat cells (Fig. 3B) and measured caspase processing after treatment with TRAIL. Cells expressing FADD (R71A) did not show caspase-8 or caspase-3 processing when treated with TRAIL, whereas cells expressing FADD (R71A) along with a second site compensating mutation in the DED showed strong TRAIL-induced caspase cleavage (Fig. 3C). Thus, DR5-induced processing of caspase-8 and caspase-3 is prevented by mutations in the DED of FADD and second site mutations that are also in the DED restored processing. Fur-  2. The DED of FADD regulates binding to DR5. A, a directed yeast two-hybrid assay was used to test for interaction of DR5, Fas, TRADD, and caspase-8 with empty vector (pACT3), wild type FADD, or FADD DED mutations. Changes in arginine 71 to either alanine or tryptophan prevented interaction with both DR5 and Fas/CD95 while retaining interaction with TRADD and caspase-8. B, FADD-deficient Jurkat cells were stably transfected with GFP vector, wild type FADD, or FADD mutants. The level of FADD protein was determined by immunoblot. C, stable Jurkat cells were left untreated or stimulated with TRAIL, and caspase-8 and caspase-3 processing was measured by immunoblot. Only cells expressing wild type FADD showed any processing, indicating that arginine 71 is required for FADD binding to DR5. thermore caspase processing occurs only in response to treatment with the ligand, indicating that it is in response to receptor activation. The same second site mutations also restored Fas/CD95-induced caspase cleavage (Fig. 3D). These data suggest that FADD uses the same surface of the DED, specifically helix 5, to bind both DR5 and Fas/CD95.
Full-length FADD Binds Better to DR5 than the Death Domain Alone-Dominant negative FADD, which consists of the DD alone, can inhibit signaling through TRAIL, indicating that the DD is sufficient for binding to DR5 when overexpressed (14). Because the DED is important for the interaction between DR5 and FADD, we reasoned that full-length FADD might bind better than the DD alone. Constructs expressing either the FADD-DD or full-length FADD were tested for interaction with DR5 in yeast. To measure this interaction we used ␤-galactosidase assays, which allow us to quantitate each interaction. We observed about a 20% increase in binding of DR5 to full-length FADD compared with the DD alone (Fig. 4A). We performed immunoprecipitation experiments with FLAGtagged DR5 and GFP-tagged FADD or FADD-DD to measure this interaction in mammalian cells. Full-length FADD coprecipitated with DR5 to a much greater extent than the death domain alone (Fig. 4B). Thus, both the DD and the DED of FADD contribute to the interaction with DR5.
The Death Domain of FADD Can Discriminate between DR5 and Fas/CD95-Because all of our data indicate that binding of FADD to DR5 and Fas/CD95 is very similar, we attempted to identify residues in FADD that are required for binding to one receptor but not the other. We performed a reverse two-hybrid screen to identify mutations in FADD that prevent binding to Fas/CD95 but retain binding to DR5. We screened more than 10 million randomly mutated FADD molecules but were able to identify only a single mutation that discriminated between Fas/CD95 and DR5 binding. A change in valine 108 to glutamate (V108E) in the DD of FADD prevents binding to Fas/ CD95 but does not alter binding to DR5, TRADD, or caspase-8 (Fig. 5A). We also screened for mutations in FADD that prevent binding to DR5 but retain interaction with Fas/CD95 but were unable to find such a mutation. This implies that other than valine 108, the same residues that are required for FADD binding to DR5 are also required for Fas/CD95 binding.
To determine whether FADD (V108E) rescues signaling in response to activation of DR5 or Fas/CD95 in mammalian cells, we introduced this FADD mutation into FADD-deficient Jurkat cells. The expression level of FADD (V108E) along with cells expressing wild type FADD or GFP is shown in Fig. 5B. These cells were treated with TRAIL or FasL, and caspase processing was measured by immunoblot. GFP cells did not show caspase processing when treated with TRAIL or FasL, whereas cells expressing FADD showed both caspase-8 and caspase-3 processing (Fig. 5C). Cells expressing FADD (V108E) underwent caspase processing in response to TRAIL but not when treated with FasL (Fig. 5C). Thus, FADD (V108E) is able to bind DR5 and transduce TRAIL signaling but is unable to bind Fas/CD95 or transduce signaling through FasL. DISCUSSION We and others have reported that the DED of death domain proteins can regulate binding of the DD (22,32). Previously we identified mutations in the DED of FADD that prevent binding to Fas/CD95. These FADD mutations did not cause gross conformational changes because binding to TRADD and caspase-8 was still intact. Here we show that mutations in the DED of FADD also prevent binding to DR5. Cells expressing FADD with a mutation in the DED were unable to transduce TRAIL signaling, whereas second site compensating mutations within the DED were able to restore FADD binding to DR5 and rescue TRAIL signaling. Computer modeling indicates that these mutations do not disrupt the overall protein structure because the effect on free energy for most mutations was small (Fig. 6A). FADD R71A L62F is the only mutation with a significant change in free energy, but this mutation actually leads to a more stable structure. Interestingly the second site mutations in FADD that rescued binding to DR5 also rescued binding to Fas/CD95, suggesting that FADD uses the same surface of the DED for binding to both receptors.
Both immunoprecipitation and two-hybrid experiments indi- FIG. 4. DR5 binds to full-length FADD better than the death domain alone. A, the interaction of DR5 with full-length FADD or the DD alone was tested in yeast using quantitative ␤-galactosidase assays. DR5 showed no interaction with empty vector (pACT3) but was able to interact with both of the FADD constructs. However, the interaction with full-length FADD was about 20% higher than the DD alone. B, HeLa cells were transfected with FLAG-DR5 along with GFP, GFP-FADD-DD, or GFP-FADD; FLAG complexes were precipitated, and interaction was measured by immunoblotting for GFP. Full-length FADD interacted with DR5 to a greater extent than the DD alone. IP, immunoprecipitation.
FIG. 5. Valine 108 is essential for binding of FADD to Fas/CD95. A, a reverse two-hybrid screen identified valine 108 as essential for binding to Fas/ CD95. Interaction with other death domain-containing proteins was determined by a directed yeast two-hybrid assay. B, the level of exogenous FADD or FADD (V108E) in FADD-deficient Jurkat cells is shown by immunoblot. C, Jurkat cells expressing GFP, FADD, or FADD (V108E) were stimulated with TRAIL or FasL, and caspase processing was measured by immunoblot. FADD (V108E) is able to transduce TRAIL signaling but not FasL signaling. cate a direct interaction between FADD and DR5. Although it is possible that cellular proteins were mediating this interaction in the immunoprecipitation, our data using FADD DED mutations provide additional evidence for a direct interaction. Mutations of arginine 71 prevented signaling through TRAIL but still allowed for binding to TRADD, suggesting that TRADD is not involved in the FADD-DR5 interaction. In addition, second site mutations that restored interaction of FADD R71A to DR5 in yeast also rescued TRAIL signaling. It would be very unlikely to see the same response in yeast and mammalian cells if there were not a direct interaction between FADD and DR5.
Our data suggest a model in which the DED of FADD comes into direct contact with the receptor. Mutations that disrupt the FADD-receptor interaction were in the loop region-flanking helix 5, and the second site compensating mutations were all in helix 5 of the DED (Fig. 6B). Had the effects of the DED on binding to DR5 been allosteric, we would have expected to find compensating mutations in the DD. In addition, we show that DR5 binds more efficiently to full-length FADD than it does to the FADD DD alone. We therefore suggest that in the context of the full-length FADD, helix 5 of the DED comes into direct contact with the receptor.
Although the requirements for FADD binding to DR5 and Fas/CD95 are very similar with regards to the DED, we identified valine 108 in helix 2 of the DD as necessary for binding to Fas/CD95 but dispensable for binding to DR5, TRADD and caspase-8 (Fig. 6C). Because we were unable to identify any other mutations in FADD that could discriminate between DR5 and Fas/CD95, we reason that FADD uses the same surface to bind both receptors. Berglund et al. (33) identified a patch of charged residues on the surface of FADD that was necessary for binding to Fas/CD95. Valine 108 is near this patch, suggesting that the Fas/CD95-binding surface of FADD is larger than the binding surface for DR5. Taken together, our data indicate that FADD uses the same surface of the DED for binding to DR5 and Fas/CD95, whereas regions within the DD can confer specificity for each receptor.
FIG. 6. The location of FADD mutations. A, the effect of each mutation on the overall structure of FADD was determined by energy minimizations. The change in free energy for each mutation along with the standard error is shown for each mutation. The only mutation to show a large effect was R71A L62F, which led to a more stable structure. B, FADD DED mutations were modeled onto the solved structure of the FADD DED (31). Arginine 71, which is required for the FADD-DR5 interaction, flanks helix 5, and the compensating mutations were in helix 5, suggesting a direct role for helix 5 in the FADD-DR5 interaction. C, residues previously shown to be important for the FADD-Fas/ CD95 interaction (red and blue) along with valine 108 (green) were modeled onto the solved structure of the FADD DD (33).