Structural requirements for signal-induced target binding of FADD determined by functional reconstitution of FADD deficiency.

FADD is a key adaptor modulating several signaling pathways such as apoptosis induced by Fas (CD95) and tumor necrosis factor receptor 1, and cell proliferation induced by mitogens. Whereas mutations in Fas disrupt its binding to FADD and cause autoimmune lymphoproliferative (lpr) syndromes, a FADD deficiency blocks embryonic development in mice. To delineate the multifunction of FADD in vivo, we performed functional reconstitution analysis by introducing wild type and mutant FADD into FADD-/- cells or FADD-/- mice lacking the endogenous FADD. An lpr-like FADD mutant, V121N, was reported previously as being defective in Fas binding in vitro. However, we found that in mice V121N can bind to Fas and is functional in signaling apoptosis. Unexpectedly, this lpr-like mutant FADD failed to support mouse development, indicating that the death domain of FADD has an additional function required for embryogenesis, which is independent of that required for receptor-induced apoptosis. Further mutagenesis was targeted at charged residues in the FADD death domain, presumably mediating electrostatic interactions with Fas. We showed that the target binding and apoptosis signaling functions of FADD were not affected when mutations were introduced to a majority of the charged residues. In one exception, replacing arginine 117 with an uncharged residue disrupted target binding and apoptosis signaling, but restoring the positive charge at position 117 failed to reconstitute the FADD function. Therefore, in vivo target binding of FADD involves an additional mechanism distinct from electrostatic interaction.

During development, programmed cell death or apoptosis is required for proper organogenesis and generation of complex multicellular tissues (1,2). In adult organisms, homeostasis in the lymphoid system and other organs and tissues is controlled in part by cell death induction (3). In general, two major cell death pathways exist in mammals (4). One is controlled by intrinsic signals, which involves the Bcl-2 family proteins operating primarily in the mitochondria (3,5,6), and a second so-called "extrinsic" pathway initiated by type I transmembrane proteins such as Fas (Apo-I or CD95), tumor necrosis factor (TNF) 1 receptor I (TNF-R1), and TNF-related apoptosisinducing ligand receptors (TRAIL-Rs) (7). During early mutational studies, a sequence of about 80 amino acid residues within the intracellular sequence of Fas or TNF-R1 was shown to be critical for the apoptotic function of these two receptors and was thus designated the "death domain" (DD) (8,9). These DDs have a weak homology to ankyrin repeats and may represent a novel protein-protein interaction motif. The Fas-associated death domain-containing protein FADD was identified when searching for molecules that bind to the intracellular domain of Fas (10 -12). FADD contains a DD at the carboxyl terminus, which presumably interacts directly with the DD of Fas. In the TNF-R1 signaling pathway, another DD-containing adaptor protein TRADD is involved, which has been shown to associate with FADD (13). A second protein-protein interaction module called the death effector domain (DED) is present at the amino terminus of FADD, which binds to the DED in pro-caspase-8 (FLICE or MACH) (14,15). Activation of the apical caspase-8, upon triggering of the Fas signaling, leads to processing of downstream caspases such as caspase-3, -6, and -7 (16 -18) or the BID protein initiating the mitochondriamediated cell death pathway (19,20).
Fas-induced apoptosis is essential for maintaining lymphoid homeostasis and suppressing autoimmunity (21). Inactivation of the Fas gene blocks lymphocyte apoptosis, leading to development of autoimmune lymphoproliferative (lpr) diseases (22). Some of the mutations resulting in defective signaling were mapped to the DD of Fas, and these lpr mutant Fas proteins are incapable of binding to FADD (11,23), indicating that Fas-FADD interaction is essential for cell death signaling. To understand the physiological function of FADD, we and others previously deleted the FADD gene in mouse germ cells (24,25). These FADD-deficient mice die in utero possibly because of heart defects, a phenotype not present in Fas-deficient mice (26). Embryonic stem cell-derived FADD-deficient primary T cells or FADDdeficient Jurkat T lymphoma cells are resistant to Fas-induced cell death (24,27). Other studies showed that expression of a dominant negative FADD mutant can block TNF-induced cell death and that a FADD deficiency inhibits TRAIL-induced cell death responses (13, 28 -32). Therefore, FADD appears to be a common signaling adaptor shared by multiple death receptors.
The early embryonic lethality phenotype of FADD-deficient mice is intriguing, given that mice lacking Fas, TNF-R1, or TRAIL-R develop normally (26,(33)(34)(35). No embryonic defects were reported in double mutant mice containing null alleles of both Fas and TNF-R1, besides the accelerated autoimmune lymphoproliferative disease (36). Whether collective defects in multiple receptor-induced signaling result in a developmental blockage remains to be investigated further. It is possible that FADD has additional functions, independent of apoptosis mechanisms. In a RAG-1 Ϫ/Ϫ blastocyst complementation analysis, FADD deficiency inhibited lymphocyte development (24). Furthermore, FADD-deficient T cells are defective in both apoptosis induced by Fas as well as proliferation responses induced by mitogens. Therefore, FADD appears to have a dual function in lymphocytes and likely in embryonic cells as well. The diverse function of FADD may be mediated by its selective interaction with various protein partners, including novel ones yet to be identified.
Binding of the DD of FADD to the DDs of Fas and TRADD could readily be detected in yeast cells, indicating direct contacts between these proteins during signal transduction in mammalian cells (10 -13). Defining the protein interaction motif in FADD and characterization of FADD mutants lacking one function but retaining others could facilitate delineation of the multiple functions of FADD in vivo by generating hypomorphic mutant mice. In NMR spectroscopic studies, the DD of FADD appears to fold into a structure consisting of six antiparallel ␣-helices (37,38). Previous mutational analysis suggested ␣-helices 2 and 3 as an interface interacting with Fas and TRADD by an electrostatic mechanism (37). The lpr mutation causes a local conformational change in the DD of Fas, resulting in defective FADD binding ability (39), and an lpr-like mutant of FADD reportedly fails to bind to Fas or TRADD (10,37). In this study, we performed in vivo functional reconstitution analysis by introducing wild type and mutant forms of FADD into FADD Ϫ/Ϫ mice or cells lacking the endogenous FADD. We demonstrated that the lpr-like mutant FADD retained normal capability in apoptosis signaling but lost another function essential for embryogenesis. We then performed additional mutagenesis and identified a site in the DD of FADD essential for signal-specific binding to Fas and TRADD and for mediating apoptosis. Therefore, the DD of FADD has a dualistic function that could be dissociated in vivo.

EXPERIMENTAL PROCEDURES
Genomic Reconstitution of FADD Ϫ/Ϫ Mice-The 12-kb EcoRV genomic DNA containing the promoter region, two coding exons, and the intron of the mouse FADD gene was subcloned from a 23-kb genomic DNA inserted in a cosmid clone (12). This wild type FADD minigene (FADD WT ) was previously shown to reconstitute normal embryonic development of FADD Ϫ/Ϫ mice (40). Site-directed mutagenesis was performed using mutant oligonucleotide primers in a PCR. The codon encoding valine 121 was changed to the one encoding an asparagine residue. The resulting FADD V121N mutant and FADD WT minigenes were injected into wild type mouse embryos to produce FADD ϩ/ϩ FADD V121N or FADD ϩ/ϩ FADD WT mice. These mice were then crossed with heterozygous FADD knock-out (FADD ϩ/Ϫ ) mice described previously (24). The resulting FADD ϩ/Ϫ FADD V121N or FADD ϩ/Ϫ FADD WT mice were backcrossed with FADD ϩ/Ϫ mice. The offspring were genotyped by Southern blots using genomic DNA isolated from tails. In timed pregnancy analysis, female mice at various gestation stages were sacrificed and embryos isolated. Genotyping was performed by Southern blots, using genomic DNA extracted from these embryos. To prepare mouse embryonic fibroblasts (MEFs), heads and guts were surgically removed. The rest of the embryonic tissue was incubated in a 0.25 mM trypsin solution for 15 min at 37°C and passed thorough a 23-gauge needle several times. The resulting cells/tissues were cultured in DMEM plus 10% fetal bovine serum. All animal studies were approved by the Institutional Review Board at Thomas Jefferson University.
Mutagenesis and Retrovirus-mediated Gene Transfer-Full-length murine FADD cDNA was described previously (12) and was used as a template in PCR-mediated amplification with mutant oligonucleotide primers. Some of the mutagenesis was performed with a QuikChange II kit (Stratagene). The wild type and mutant FADD cDNA was cloned into a MSCV-based CMV-IRES-GFP vector (Provided by Dr. Bill Sha, University of California, Berkeley). Recombinant viruses were packaged by cotransfecting human embryonic kidney 293T cells with pCL-Eco (Imgenix) by calcium phosphate-mediated transfection (Promega Profection Mammalian Transfection System). Virus supernatants were collected 48 h post-transfection and were used to infect E8.5 FADD Ϫ/Ϫ MEF cells (32) in the presence of 10 g/ml Polybrene. The infection was performed by the addition of recombinant virus-containing medium collected from packaging cultures. Cells were washed three times with PBS and cultured in fresh DMEM containing 10% fetal bovine serum. Infection efficiency was examined by fluorescence microscopy and flow cytometry 48 h after infection. Stable clones expressing GFP were established using limiting dilution and plating in 96-well plates. GFP ϩ cells were isolated using a high speed cell sorter (Moflo, Cytomation).
Western Blotting and Immunoprecipitation Assays-MEF cells were detached from culture plates by trypsinization and lysed in a buffer containing 50 mM Tris-HCl, pH 8.0, 150 mM NaCl, 1 mM dithiothreitol, 1 mM EDTA, 10 mM ␤-glycerophosphate, 1 mM sodium vanadate, 0.1 mM NaF, 1% Nonidet P-40, 0.2 mM phenylmethylsulfonyl fluoride, and a protease inhibitor mixture (Roche Applied Science). Cell debris was removed by centrifugation at 13,000 ϫ g for 5 min at 4°C, the supernatant (20 g of total proteins) boiled in a standard SDS-containing sample buffer, and proteins resolved by 12% SDS-PAGE. Yeast cells were pelleted from 1 ml of culture (A 600 0.4ϳ0.8), and resuspended in 50 l of a lysis buffer containing 5% SDS, 40 mM Tris-HCl, 0.1 mM EDTA, 0.2 mM phenylmethylsulfonyl fluoride, and a protease inhibitor mixture (Roche Applied Science). After mixing with 40 l of glass beads, the yeast cell were incubated for 10 min at 70°C, and the tubes were vortexed vigorously for 1 min. The cell extracts were then collected after centrifugation at 13,000 ϫ g for 5 min, and proteins resolved by a 12% SDS-PAGE. Proteins were blotted onto Protran membranes (Schleicher & Schuell) that were blocked with nonfat milk (5%) in PBST (2 mM NaH 2 PO 4 , 8 mM Na 2 HPO 4 , pH 7.4, 65 mM NaCl, and 0.1% Tween 20), incubated with primary polyclonal rabbit anti-mouse FADD antibodies previously prepared (12) and secondary anti-rabbit antibodies conjugated to horseradish peroxidase in 5% milk and PBST. To probe for caspase-8 in the DISC, monoclonal rat anti-mouse caspase-8 (clone 1G12, Alexis) and secondary anti-rat antibodies conjugated to horseradish peroxidase were used. The SuperSignal West Pico or Dura chemiluminescent reagent (Pierce) was used for detecting immobilized FADD proteins.
Stable FADD Ϫ/Ϫ MEFs infected with the wild type or mutant FADDcontaining viruses were detached from culture plates by trypsinization and were resuspended in 5 ml of DMEM containing 10% fetal calf serum. Agonistic monoclonal anti-Fas antibodies (Jo2, Pharmingen) were added (10 g), followed by a 15-min incubation at 37°C. These treated cells were washed with ice-cold PBS and lysed for 1 h at 4°C in a lysis buffer containing 50 mM Tris-HCl, pH 8.0, 150 mM NaCl, 1 mM dithiothreitol, 1 mM EDTA, 10 mM ␤-glycerophosphate, 1 mM sodium vanadate, 0.1 mM NaF, 1% Nonidet P-40, 0.2 mM phenylmethylsulfonyl fluoride, and a protease inhibitor mixture (Roche Applied Science). After centrifugal removal of cell debris, additional anti-Fas antibodies were added (1 g) to the supernatant, and protein complexes were precipitated using protein A-agarose beads (Pierce) by rotational mixing at 4°C for 3 h. The beads with protein complex were washed four times with ice-cold lysis buffer. Proteins were eluted by boiling in a standard sample buffer for 3 min and fractionated by 12% SDS-PAGE. Western blotting using polyclonal anti-FADD antibodies was performed to detect the Fas-associated wild type or mutant FADD proteins.
Yeast Two-hybrid Assays-The full-length wild type or mutant forms of FADD were cloned into pACTII (Clontech), and the intracellular domain of the wild type or lpr mutant Fas or the full-length TRADD was cloned into pAS1-CYH2 (41). The recombinant pACTII and pAS1-CYH2 plasmids were cotransformed into Y190 yeast cells. Transformants were selected by culturing in minimal medium containing necessary amino acid supplements but without leucine or tryptophan. Target binding of FADD or its mutants was determined by measuring the ␤-galactosidase activity using X-gal in filter lifting assays or by colorimetric assays using ONPG as a substrate, according to protocols described previously (41).
Cytotoxicity Assays-MEF cells were grown in DMEM supplemented with 10% fetal calf serum, 100 g/ml streptomycin sulfate, and 100 IU/ml penicillin. Cells were seeded in triplicate in 96-well plates at a density of 1.5 ϫ 10 4 cells/well in 100 l of DMEM with 5% fetal calf serum. After a 10-h incubation at 37°C (about 70% confluence), cells were treated with 1 g/ml anti-Fas antibodies in the presence of 1 g/ml cycloheximide (Sigma) in DMEM containing 5% fetal calf serum. 16 h poststimulation, cell death was determined by propidium iodide uptake assays using a flow cytometer (Coulter EPICS XL) as described previously (12). For TNF-induced apoptosis, cells were treated with 100 ng/ml mouse recombinant TNF-␣ (Genzyme Corp.) and 10 g/ml cycloheximide for 16 h, and cell death was also determined by propidium iodide uptake assays.

An lpr-like Mutation in the DD of FADD Had No Effect on Apoptosis but Disrupted a Function Required for Mouse Em-
bryonic Development-To understand the multifunction of FADD, we initially proposed to separate one function from the others in vivo by generating hypomorphic FADD mutant mice. For this purpose, point mutations could be introduced into FADD to generate mutants that lose some function but retain others. Because mice lacking each of the FADD-dependent death receptors develop normally (26,(33)(34)(35), it is possible that mutations disabling the cell death function of FADD would still allow a partial or complete embryogenesis. The cell death function of both Fas and FADD were investigated previously by site-directed mutagenesis analyses. In one of the lpr mutant mice, a point mutation in the Fas gene results in the replacement of an isoleucine residue at position 225 by an asparagine residue within the intracellular DD of the Fas receptor protein (23) (Fig. 1). This lpr (I225N in the mouse and V238N in the human) mutation results in defective intracellular cell death signal transduction induced by Fas, causing lymphoproliferation diseases. It was demonstrated previously by yeast twohybrid assays that the wild type, but not the lpr mutant Fas protein, binds directly to FADD (11). In a sequence similarity comparison, the valine 121 residue within the DD of FADD aligns with the isoleucine 225 residue in the DD of the mouse Fas and valine 238 in the human Fas (10, 11) (Fig. 1). It was therefore suggested that an lpr-like mutation in FADD, replacing the valine 121 with asparagine (V121N), would disrupt the function of FADD. In one of the previous reports, a full-length V121N mutant of FADD was generated and shown to lack binding to Fas in a mammalian cell transient transfection assay (10). In another report, an in vitro binding assay using a BIAcore was performed, which could not detect binding of V121N mutant of FADD to the Fas DD (37). Therefore, we initially determined the in vivo effect of the V121N mutation of FADD by performing a genomic reconstitution analysis in mice.
In a previous study, transgenic expression of a 12-kb wild type FADD minigene restored normal development of the homozygous knock-out FADD Ϫ/Ϫ mice, which otherwise die around embryonic day 9 (E9) (24,40). This FADD minigene was isolated as a 12-kb fragment from the murine FADD gene locus and contained the promoter region and two coding exons ( Fig. 2A). We introduced the V121N mutation into the fulllength 12-kb FADD minigene and expressed the resulting mutant minigene FADD V121N in FADD Ϫ/Ϫ mice by transgenesis. In agreement with the previous report (40), the wild type minigene FADD WT could fully reconstitute normal development of FADD Ϫ/Ϫ mice because crosses between FADD ϩ/Ϫ and FADD ϩ/Ϫ FADD WT mice produced normal viable FADD Ϫ/Ϫ FADD WT offspring at expected Mendelian frequencies, as determined by Southern blot genotyping using adult mice tail DNA (Fig. 2B). In contrast, we detected no viable FADD Ϫ/Ϫ FADD V121N pups in analyses of 120 offspring produced from crosses between FADD ϩ/Ϫ mice and FADD ϩ/Ϫ FADD V121N mice. Although detectable at or before E15, the FADD Ϫ/Ϫ FADD V121N embryos were absent at later developmental stages, indicating that FADD V121N is partially functional and could only support embryogenesis for an additional 4 -5 days.
In further analyses, we prepared MEFs from E14.5 embryos of the FADD Ϫ/Ϫ FADD V121N genotype and induced apoptosis in these cells. In control experiments, stimulation with anti-Fas antibodies or TNF resulted in little cell death responses in E8.5 FADD Ϫ/Ϫ MEFs (Fig. 2C). In contrast, MEFs prepared from FADD Ϫ/Ϫ FADD V121N embryos were killed effectively, similar to the control E14.5 FADD ϩ/Ϫ MEFs (Fig. 2C), indicating that the V121N mutant protein functions as well as the wild type counterpart. This result was unexpected because defective Fas binding properties in this mutant FADD has been shown previously in transient expression or in vitro binding assays (10,37). Thus, we performed further ex vivo analysis by stable expression of the V121N mutant FADD in E8.5 FADD Ϫ/Ϫ MEFs lacking the endogenous FADD, using a retrovirus gene transfer system. Stable clones were established expressing ei- ther the wild type FADD control or V121N mutant to similar levels. To detect signal-induced protein-protein interaction, we performed coimmunoprecipitation (co-IP) assays. After stimulation with anti-Fas antibodies, the wild type FADD was detected in association with the membrane-bound Fas (Fig. 3A).
In similar experiments, the V121N mutant FADD protein was also found to be capable of binding to Fas as well as the wild type counterpart. Signal-specific recruitment of caspase-8 to the DISC was revealed by co-IP and Western blotting using anti-caspase-8 antibodies (Fig. 3A). To analyze apoptosis responses, cells were treated with anti-Fas antibodies or TNF. Although FADD Ϫ/Ϫ MEFs infected with the control viral vector are highly resistant to death induced by either anti-Fas antibodies or TNF, those reconstituted with either the wild type or V121N mutant were killed equally well (Fig. 3B). These results suggest that the V121N mutation has no obvious effect on the target binding or apoptosis signaling function of FADD.
It is possible that the co-IP assay could not detect subtle differences in the signal-specific Fas binding affinity between the wild type and V121N mutant FADD, and possible minor defects in the mutant protein may not result in reduced cell death responses. Thus, we performed additional analysis using a yeast-two hybrid system. In filter lift assays, Fas-FADD interaction activated expression of ␤-galactosidase, which induces chromatic conversion of X-gal, resulting in blue coloring of yeast colonies (Fig. 3C). We used the lpr mutant Fas protein as a control because it fails to interact with FADD (Fig. 3C), a result in agreement with a previous report (11). The full-length V121N mutant cDNA was transformed into yeast cells containing the wild type Fas. Colonies of yeast cells expressing Fas and the V121N mutant FADD turned blue at a rate similar to those expressing Fas and the wild type FADD (Fig. 3C). As expected, colonies of yeast cells expressing the lpr mutant of Fas and the V121N mutant of FADD did not turn blue. Western blots with yeast extract confirmed protein expression of the wild type and lpr mutant Fas as well as the wild type and the V121N mutant FADD (Fig. 3C). We then performed a relatively quantitative analysis to determine the Fas-FADD binding affinity in colorimetric assays using ONPG as a ␤-galactosidase substrate. As shown in Fig. 3D, yeast cells expressing Fas and FADD or Fas and V121N contain an equivalent level of the ␤-galactosidase activity. In this colorimetric analysis, binding of the lpr mutant Fas to either the wild type or the V121N mutant FADD was not detectable (Fig. 3D). These results demonstrate that the lpr mutation in Fas results in a nonfunctional protein, whereas a lpr-like mutation in FADD has no effect on the function required for Fas signaling.

Further Mutagenesis and Functional Reconstitution Analysis in FADD Ϫ/Ϫ MEFs Revealed Arginine 117 as Being Critical for Fas Binding and Apoptosis
Signaling-The DD-containing fragment of FADD interacts with the Fas intracellular domain in yeast, indicating direct contact between the two polypeptides. The DD of FADD may assume a three-dimensional structure consisting of six antiparallel ␣-helices (37,38), similar to that of the Fas DD (42). Because the lpr-like mutant of FADD, V121N, is functional in Fas binding and apoptosis signaling, we searched for other mutants that may lack the Fas binding ability. According to previous in vitro analysis, Fas-FADD association is mediated by electrostatic interaction between opposing charges on residues localized on the interfaces of the DDs (37). In particular, charged residues such as Arg-110, Lys-113, Arg-117, Glu-118, and Glu-123 within the DD of FADD were suggested to contribute significantly in mediating FADD-Fas interaction. We therefore used full-length FADD cDNA to generate a series of mutants by site-directed mutagenesis in which the charged residues were replaced by alanine. To test the function in vivo, we introduced these mutants into FADD Ϫ/Ϫ MEFs, using the retroviral vector CMV-IRES-GFP. In this vector, expression of genes of interest is driven by the cytomegalovirus (CMV) promoter and the green fluorescent protein (GFP) gene, translated from the internal ribosomal entry site (IRES), can be used as a gene transfer marker.
Replication-defective recombinant MSCVs were prepared in packaging cells by transient transfection and were used to transduce FADD Ϫ/Ϫ MEFs. Two days after transduction, MEFs were found to express various levels of GFP as detected by flow cytometric analysis (Fig. 4A) and were then fractionated into GFP-negative (GFP Ϫ ) and GFP-positive (GFP ϩ ) populations by cell sorting. Expression levels of the wild type and mutant FADD protein in the mixed culture were detected by Western blots (top, Fig. 4B). The GFP ϩ population was then isolated by high speed sorting; single GFP ϩ cells were plated  1 and 2). An lpr-like mutation was introduced into this FADD minigene, replacing valine 121 with asparagine (V121N) encoded by the second exon. B, genotyping of the offspring from crosses between FADD ϩ/Ϫ and FADD ϩ/Ϫ FADD WT or FADD ϩ/Ϫ FADD V121N mice was performed by Southern blotting using mouse tail DNA. WT, wild type FADD alleles. K/O, FADD knock-out alleles. C, MEFs were prepared from FADD Ϫ/Ϫ FADD V121N embryos at E14.5 and stimulated with anti-Fas antibodies or TNF. Apoptosis were analyzed by propidium iodide uptake assays using a flow cytometer. FADD Ϫ/Ϫ (E8.5) and FADD ϩ/Ϫ (E14.5) MEFs were used as controls. and expanded to establish multiple stable clones for each mutant FADD. After Western blot analysis (bottom, Fig. 4B), mutant stable clones containing FADD antigen levels similar to the wild type FADD-expressing MEF clones were used in further analysis. To detect signal-specific interaction between FADD and Fas, we performed co-IP assays. Cells were stimulated with an agonistic anti-Fas antibody, and cell lysates were prepared. After the cell surface Fas was precipitated using anti-Fas antibodies and fractionated by SDS-PAGE, the presence of FADD in the DISC was detected by Western blotting using anti-FADD antibodies. In FADD Ϫ/Ϫ MEFs transduced with control viral vectors, no FADD protein was detected (first two lanes, Fig. 5A). In untreated cells transduced with the wild type FADD cDNA, there was a basal level interaction between FADD and Fas, and this interaction is enhanced dramatically after stimulation with anti-Fas antibodies (Fig. 5A). Among the FADD mutants tested, R110A, K113A, E118A, and E123A bind efficiently to Fas upon cell death induction with anti-Fas stimulation (Fig. 5A). One obvious exception is the R117A mutant, which failed to bind to Fas upon Fas activation. In a control experiment, we performed Western blots with anti-FADD antibodies using whole cell lysates. As shown at the bottom in Fig.  5A, the wild type FADD and the five mutant proteins were expressed at a similar level. Therefore, a lack of signal-induced binding of the R117A mutant to Fas is not the result of reduced protein expression.
To test whether these FADD mutants could transduce Fasmediated death signaling, we treated the wild type and mutant MEF clones with anti-Fas antibodies. FADD Ϫ/Ϫ MEFs transduced with the control CMV-IRES-GFP vector remained resistant to apoptosis, whereas those transduced with the wild type FADD-expressing viruses become sensitive to Fas-induced cell death (Fig. 5B). Similarly, MEFs expressing R110A, K113A, E118A, and E123A were killed as well as the control MEFs expressing the wild type FADD. In contrast, the R117A amino acid replacement dramatically disrupted the signaling capability of FADD in the Fas-induced cell death response (Fig. 5B). This result could be explained by a lack of binding of the R117A mutant protein to Fas as indicated in the co-IP experiment (Fig. 5A). In a further analysis, we replaced Arg-117 with a positively charged lysine residue. After transduction of the full-length R117K mutant into FADD Ϫ/Ϫ MEFs, we examined the Fas binding capability by co-IP assays and performed apoptosis induction analysis. Interestingly, this R117K mutant protein was also defective in signal-specific binding to Fas (Fig.  5C). In a functional assay, the R117K mutant FADD retained less than 26% of its wild type counterpart in the cell death signaling capability (Fig. 5D). These results indicated that Arg-117 is critical for Fas binding and apoptosis signaling, and restoring the positive charge characteristics present in arginine by using a lysine residue could not fully reconstitute these functions of FADD in vivo.
Functional Reconstitution Analysis of TNF-induced Death Responses-Interaction between FADD and TRADD could not readily be detected by co-IP assays in mammalian cells unless the proteins are overexpressed (13). This is probably because of the relative transient and/or weak interaction between the two proteins. Indeed, when a quantitative measurement was performed using a yeast two-hybrid system, the binding affinity between FADD and TRADD is more than 10-fold weaker than that between FADD and Fas (Fig. 6A). Similar to co-IP assays using recombinant retrovirus-transduced MEF cells, interac- 3. Ex vivo analysis of the V121N mutant of FADD. A, the full-length wild type (FADDwt) or V121N mutant of FADD was stably expressed in E8.5 FADD Ϫ/Ϫ MEFs by retrovirus-mediated gene transfer. Co-IP analysis showed that the V121N mutant FADD protein can bind to Fas after treatment with anti-Fas antibodies. Untreated (UT) cells were used as controls. Recruitment of caspase-8 to the DISC was revealed by co-IP assays using anti-caspase-8 antibodies. B, E8.5 FADD Ϫ/Ϫ MEF reconstituted with the V121N mutant can be killed by anti-Fas antibodies or TNF. FADD Ϫ/Ϫ MEFs infected with the viral vector or those expressing the wild type FADD were used as controls. C, yeast two-hybrid analysis can detect interaction between Fas and the V121N mutant of FADD, as indicated by a colony filter-lifting assay (top). Blue colonies (converted to a gray scale) indicated interactions between two proteins. The wild type FADD and lpr mutant Fas were used as controls. Expression of FADD and Fas proteins in yeast cells was confirmed by Western blots (bottom). D, the relative Fas binding affinity of the V121N mutant was determined by colorimetric assays in yeast. The wild type FADD and lpr mutant Fas were used as controls.
tion affinity between R117A and Fas in the yeast was reduced by more than 90% compared with that between the wild type FADD and Fas (Fig. 6A). The R117A mutant almost completely lost the TRADD binding capability. In contrast, the R110A, K113A, E118A, and E123A mutant bound to Fas or TRADD in yeast as well as the wild type FADD (data not shown). To determine the signaling capability of FADD mutants, we stimulated MEF cells with TNF to induce cell death. MSCV-IRES-GFP vector-transduced control cells are resistant to this treatment, whereas the wild type FADD-expressing MEFs were killed effectively by TNF (Fig. 6B). The K113A and E123A mutations had little effect, and the R110A and E118A only slightly weakened the function of FADD in mediating cell death signaling induced by TNF (Fig. 6B). However, the R117A mutation almost completely abolished TNF-induced cell death responses, apparently because of a lack of the TRADD binding ability in this mutant protein as shown in the yeast two-hybrid assays (Fig. 6A). Therefore, the Arg-117 residue is important for the function of FADD required for signaling pathways induced by Fas and TNF-R1. DISCUSSION Several lines of evidence revealed in previous studies have established an essential role of FADD in death receptor-initiated apoptosis. In particular, FADD-deficient cells are resistant to cell death induced by Fas, TNF-R1, and TRAIL-R (24, 25, 29 -32). Although deletion of each individual death receptor had no obvious effect on mouse development, FADD deficiency blocked embryogenesis during early gestation. These results suggest that FADD may play an additional role in death receptor-independent pathways essential for embryonic development. In this study, we investigated the multiple functions of FADD by reconstitution of FADD deficiency using wild type and mutant forms of FADD. To this end, the data suggest that FADD signals regulating embryonic development and receptorinduced apoptosis are modulated by the same DD and could be uncoupled using partial loss-of-function mutants of FADD.
The DD represents a novel protein-protein interaction motif present in a number of proteins with diverse biological functions (43,44). Alanine scanning mutagenesis of the DDs demonstrated that many of the amino acids conserved between Fas and TNF-R1 are critical for the cytotoxic signal (8,9). The solution structure of the Fas DD, as determined previously by NMR spectroscopy, consists of six antiparallel, amphipathic ␣-helices (42). The DD of FADD has a comparable three-dimensional structure of a six-antiparallel helical bundle (37,38). Although the primary amino acid sequence and the threedimensional structure of the DD of FADD appear to be similar to those of Fas, the modulation mechanism mediated by the two DDs may be distinct during signaling. This is indicated by the fact that the lpr mutation in the Fas DD (I225N in the mouse and V238N in the human) resulted in a defective protein with pathological consequences (23), whereas a similar mutation in FADD (V121N) had no effect on apoptotic signal transduction (Fig. 2). Structural studies of the human lpr mutant (V238N) of the Fas DD indicate that ␣-helix 3 is unfolded, thus abolishing its interaction with FADD (39). However, the V121N mutant protein of FADD binds to Fas as well as the wild type protein ex vivo or in vivo (Fig. 3), indicating that this mutation unlikely altered the structure of the Fas binding motif in FADD. This result contrasts that of the previous studies using a transient overexpression system or in vitro peptide binding assays showing a lack of Fas binding in the V121N mutant protein (10,37). Possible indirect effects caused by varying transfection efficiencies and protein expression levels in the transient system were avoided in this study by stable expression of the wild type and mutant FADD in cells lacking the endogenous FADD (Figs. 4 and 5). It is possible that the properties of the DD fragment of FADD in vitro may not necessarily be the same as that of the intact FADD protein in vivo. The three-dimensional structure of the native, intact FADD protein has yet to be analyzed. Although functional in target binding and apoptosis signaling, the V121N mutant FADD appears to have a partial loss of a function essential for mouse embryogenesis. It is possible that the V121 residue is part of a motif interacting with distinct proteins independently of the cell death function. In this regard, the V121N mutant would be of use in further investigation of the novel mechanism regulated by FADD. Future studies will be focused on proteomic analysis of proteins that bind to the wild type FADD, but not to the V121N mutant protein.
In the previous study, DD fragments of FADD containing the R110A, K113A, R117A, E118A, and E123A mutations were expressed and purified from bacteria. In analysis using a BIAcore instrument, these mutant peptides were shown to lack binding to the DD fragment of Fas also expressed and purified from bacteria (37). It was suggested that the charged residues within ␣-helices 2 and 3 are important for establishing electrostatic interactions with oppositely charged residues present in the DD of Fas or TRADD (37). Because the V121N mutation has no effect on FADD-Fas or FADD-TRADD interactions, we searched for mutants lacking these functions by targeting the five charged residues in the DD of FADD. Earlier studies showed that transient transfection of mammalian cells with wild type FADD cDNA can induce massive apoptotic and necrotic cell death independent of Fas or caspases (10,45). To avoid undesired nonspecific effects caused by overexpression, we stably expressed wild type and mutant FADD by using a retroviral vector in fibroblasts cells lacking the endogenous FADD. In contrast to the previous report, our results showed that four of the five charged residues (Arg-110, Lys-113, Glu-118, and Glu-123) in the DD of FADD are dispensable in establishing the apoptotic signal-specific contact with activated Fas or TRADD (Fig. 5). Although the Arg-117 residue appears to play a critical role in modulating the FADD-Fas interaction, restoring the positive charge using a lysine residue at position 117 in the death domain did not fully reconstitute the function of FADD (Fig. 5). The charge distribution in ␣-helices 2 and 3 of the DD of FADD is similar to that of the DD of Fas (37,38,42), and thus electrostatic repulsion, rather than attraction, may occur between the two proteins unless the interaction is antiparallel. However, the relative in vivo orientation and positioning of the DDs in Fas and FADD is unclear. Our data suggest that electrostatic interactions may not be a major mechanism modulating in vivo activation signal-induced Fas-FADD or TRADD-FADD associations. Analysis of the threedimensional structure of a complex between the DDs of the Drosophila Tube and Pelle protein has indicated a nonelectrostatic interaction mechanism (46). It is possible that basal interactions exist between complementary structures present in the DDs of FADD and Fas, and triggering through trimeric Fas ligand engagement promotes stabilization of the membrane-proximal protein complex to sustain the apoptotic signaling response. The R117A mutant retained only a minimal interaction with either Fas or TRADD and was defective in apoptosis signaling (Fig. 5). This debilitating mutation apparently did not lead to destabilization of the FADD protein (Figs. 4 and 5), and according to previous studies, the overall folding of R117A and other FADD mutants was not altered (47). Therefore, the Arg-117 residue in the DD of FADD likely represents a point of contact with Fas and TRADD. Further evidence may lay in the co-crystal structure of the FADD DD and Fas DD complex, which has yet to be determined. Future studies will be carried out to analyze the effect of the R117A mutation in vivo. FIG. 5. Arginine 117 is critical for binding of FADD to Fas. A, wild type (FADDwt) and five mutants replacing charged residues individually in the DD of FADD were stably expressed in FADD Ϫ/Ϫ MEFs, and signal-specific Fas binding ability was determined by co-IP assays after stimulation with anti-Fas antibodies (top). FADD Ϫ/Ϫ MEFs containing the viral vector or FADDwt-expressing viruses were used as untreated (UT) controls. Total FADD protein levels in stable clones used were determined by Western blotting using whole cell extracts (bottom). B, FADD Ϫ/Ϫ MEFs containing the viral vector and viruses expressing FADDwt or mutants were stimulated with anti-Fas antibodies for 16 h, and apoptosis was determined by flow cytometry. C, the positive charge at position 117 in the DD of FADD was restored by using a lysine residue to replace arginine, and the Fas binding ability of the resulting R117K mutant FADD was determined in FADD Ϫ/Ϫ MEFs by co-IP assays after stimulation with anti-Fas antibodies (top). Whole cell extracts were analyzed by Western blotting to detect total FADD proteins expressed in stable clones (bottom). D, apoptosis in FADD Ϫ/Ϫ MEFs expressing the wild type and Arg-117 mutant FADD were induced by stimulation with anti-Fas antibodies and analyzed using a flow cytometer.
FIG. 6. TRADD binding and TNF-induced cell death signaling analysis. A, the relative TRADD binding ability of the wild type and R117A mutant FADD was analyzed by determining expression levels of ␤-galactosidase (␤-gal) in a yeast two-hybrid system (right). Fas binding was determined similarly and used as controls (left). B, to determine the function in apoptosis signaling induced by TNF, FADD Ϫ/Ϫ MEFs expressing the wild type and mutant FADD were treated with recombinant TNF-␣ for 16 h, and cell death was analyzed by flow cytometry.
It will be interesting to determine whether the R117A mutant FADD, if functional during embryogenesis, would cause accelerated autoimmune lymphoproliferation diseases in mice as seen in mice lacking both Fas and TNF-R1.
The pleiotropic effect caused by a FADD deficiency in mice reflects the diverse functions of FADD in multiple pathways. The apoptosis-independent function of FADD was first indicated by studying T cell antigen receptor-induced proliferation responses (24). Whether FADD is required for proliferation signal transduction in embryonic cells remains to be determined. The diverse function of FADD may be because this adaptor interacts with multiple proteins participating in the modulation of various signaling processes. To understand the molecular mechanisms involved in the complex function of FADD, mutants with partial loss of function would be useful especially for in vivo analysis. Using the Val-121 and Arg-117 mutants, this study revealed a dualistic function for the DD of FADD involved in both receptor-induced apoptosis as well as a novel pathway important for mouse embryonic development. Further analysis of the V121N mutant mouse embryos would help an understanding of the nature of the nonapoptotic pathway mediated by FADD. It will be interesting to test the in vivo effect of the Arg-117 mutants in mice lacking the endogenous FADD and whether this mutation leads to hypomorphic phenotypes in mice.