The Ectodermal Dysplasia Receptor Activates the Nuclear Factor- k B, JNK, and Cell Death Pathways and Binds to Ectodysplasin A*

The ectodermal dysplasia receptor (EDAR) is a recently isolated member of the tumor necrosis factor receptor family that has been shown to play a key role in the process of ectodermal differentiation. We present evidence that EDAR is capable of activating the nuclear factor- k B, JNK, and caspase-independent cell death pathways and that these activities are impaired in mutants lacking its death domain or those associated with anhidrotic ectodermal dysplasia and the downless phenotype. Although EDAR possesses a death domain, it did not interact with the death domain-containing adaptor proteins TRADD and FADD. EDAR successfully inter-acted with various TRAF family members; however, a dominant-negative mutant of TRAF2 was incapable of blocking EDAR-induced nuclear factor- k B or JNK activation. Collectively, the above results suggest that EDAR utilizes a novel signal transduction pathway. Fi-nally, ectodysplasin A can physically interact with the extracellular domain of EDAR and thus represents its biological ligand.

was amplified by reverse transcription-polymerase chain reaction using total RNA derived from human prostate gland as a template. The upstream primer used for amplification was 5Ј-CGCGGGATCCCGCCCT-ATTGAATTTCTTC-3Ј, and the downstream primer used was 5Ј-CGC-GGTCGACCTAGGATGCAGGGGCTTCAC-3Ј. The amplified product was digested with BamHI and SalI enzymes and subsequently cloned into a modified pFastBAC1 vector (Life Technologies, Inc.), which contained a Myc epitope tag downstream of a baculovirus gp67 signal peptide. The sequences of all constructs were confirmed by automated fluorescent sequencing. Constructs encoding NIK and its mutants (5), the IKK2 mutant (6); IB␣-S32A/S36A (7); and murine TRAF1, murine TRAF2 and its mutant, TRAF3, I-TRAF, FADD, DR5, CD40, and NF-B/luciferase reporter constructs have been described previously (8 -10) and were obtained from the indicated sources. The dominant-negative IKK1 mutant was kindly provided by Dr. Richard Gaynor.
Electrophoretic Mobility Shift Assay-293T cells (3 ϫ 10 5 ) were transfected with 2 g of various constructs in each well of a sixwell plate. After 36 h, nuclear extracts were prepared, and electrophoretic mobility shift assay was performed essentially as described previously (9).
Luciferase Reporter Assays-The NF-B reporter assay was performed essentially as described previously (9). Briefly, 293T cells were transfected in duplicate in a 24-well plate with the various test plasmids along with an NF-B/luciferase reporter construct (75 ng/well) and a Rous sarcoma virus promoter-driven ␤-galactosidase reporter construct (pRcRSV/LacZ; 75 ng). Twenty-four to thirty hours later, cells were lysed, and extracts were used for the measurement of luciferase and ␤-galactosidase activities, respectively. Luciferase activity was normalized relative to ␤-galactosidase activity to control for the difference in the transfection efficiency.
For the c-Jun transcriptional activation assay, 293 EBNA cells (1.2 ϫ 10 5 ) were transfected in duplicate with various expression constructs (500 ng) along with a fusion transactivator plasmid containing the yeast Gal4 DNA-binding domain fused to transcription factor c-Jun (pFA-c-Jun) (50 ng), a reporter plasmid encoding the luciferase gene downstream of the Gal4 upstream activating sequence (pFR-luc) (500 ng), as well as a Rous sarcoma virus/LacZ (␤-galactosidase) reporter construct (75 ng). Cells were lysed 24 h later, and luciferase assay was performed as described previously (10).
Coimmunoprecipitation Assays-For studying in vivo interaction, 2 ϫ 10 6 293T cells were plated in a 100-mm plate and cotransfected 18 -24 h later with 5 g/plate of each epitope-tagged construct by calcium phosphate coprecipitation. A hemagglutinin-tagged green fluorescent protein-encoding plasmid was also included in some experiments. Twenty-four hours post-transfection, cells were lysed in 1 ml of lysis buffer containing 0.1% Triton X-100, 20 mM sodium phosphate (pH 7.4), 150 mM NaCl, and one EDTA-free mini-protease inhibitor tablet (Roche Molecular Biochemicals)/10 ml. Cell lysates (500 l) were incubated for 1 h at 4°C with 10 l of FLAG or control mouse Ig beads precoated with a supersaturated casein solution. Beads were washed twice with lysis buffer; twice with wash buffer containing 0.1% Triton X-100, 20 mM sodium phosphate (pH 7.4), and 500 mM NaCl; and again with lysis buffer. Bound proteins were eluted by boiling, separated by SDS-polyacrylamide gel electrophoresis, transferred to a nitrocellulose membrane, and analyzed by Western blotting. Essentially a similar procedure was used for experiments involving coimmunoprecipitation of TRADD and TRAFs, except cells were lysed in a modified radioimmune precipitation assay buffer containing 150 mM NaCl, 50 mM Tris (pH 7.4), 1% Nonidet P-40, 0.1% sodium deoxycholate, and 1 mM EDTA, and the beads were washed extensively in the above buffer containing 1 M NaCl.
Receptor-Ligand Interaction Assays-293T cells were transfected with expression plasmids encoding EDAR and TAJ immunoadhesins, and 12 h post-transfection, the medium was changed to 293SFM (Life Technologies, Inc.). Supernatants containing the secreted immunoadhesin were collected 48 h later and stored in aliquots at Ϫ70°C until use. Myc-EDA I protein was produced by infection of Sf9 insect cells with the corresponding baculovirus construct following the manufacturer's instructions (Life Technologies, Inc.). Supernatant containing the secreted protein was collected 48 h post-infection, filtered, and stored at Ϫ70°C until use.
For coimmunoprecipitation assay, equal volumes (500 l) of the immunoadhesin and Myc-EDA supernatants were mixed in a buffer containing 50 mM Tris (pH 7), 150 mM NaCl, and 0.1% Triton X-100 and incubated at 4°C overnight with gentle shaking. Supernatants were subsequently divided into two halves and immunoprecipitated with goat anti-mouse IgG1 beads (Sigma) or control antibody beads precoated with supersaturated casein solution. After extensive washing, the bound proteins were eluted by boiling, separated by SDS-polyacrylamide gel electrophoresis, transferred to a nitrocellulose membrane, and analyzed by Western blotting.
For enzyme-linked immunosorbent assay, 5 l of a control supernatant or the supernatant containing Myc-EDA I were immobilized overnight at 4°C in the wells of a microwell plate in Na 2 CO 3 /NaHCO 3 buffer (pH 9.6). After washing with Tris-buffered saline containing 0.05% Tween (TBST), 5 l of supernatants containing FLAG-tagged immunoadhesins (or a control supernatant) diluted in Tris-buffered saline were added to the wells and incubated for 4 h at room temperature. After washing, goat anti-mouse peroxidase (1:2000 in TBST) was added for 1 h. Color was developed using o-phenylenediamine dihydrochloride (Sigma), and the absorbance was measured at 490 nm.

RESULTS
EDAR Activates the NF-B Pathway-We began by testing the ability of EDAR to activate an NF-B-driven luciferase reporter construct upon transient transfection in 293T cells. As shown in Fig. 1A, transfection of EDAR in these cells led to significant activation of the NF-B pathway that was comparable in magnitude to that induced by TNFR1. In comparison, TAJ/TROY, another TNFR family member that is highly expressed in skin during embryonic development (11,12), failed to activate the NF-KB pathway. NF-B activation by EDAR was further confirmed using an electrophoretic mobility shift assay (Fig. 1B).
Mechanism of EDAR-induced NF-B Activation-TRAF2 has been known to mediate NF-B activation by various TNFR family members, and TANK/I-TRAF has been known to regulate this process (13). Therefore, we investigated the roles of TRAF2 and TANK/I-TRAF in the activation of the NF-B pathway by EDAR. An N-terminal deletion mutant of TRAF2 (14) that could effectively block NF-B activation by CD40 was ineffective in blocking NF-B induction by EDAR ( Fig. 2A).
Similarly, EDAR-induced NF-B activation was not affected by coexpression of TANK/I-TRAF ( Fig. 2A). Collectively, these results suggest either that TRAF2 and TANK/I-TRAF are not involved in NF-B activation via EDAR or that they play a functionally redundant role in this process. We have also tested the ability of dominant-negative mutants of the receptor-interacting protein and TRADD to block EDAR-induced NF-B, but have failed to observe any significant inhibitory effect (data not shown).
The NIK and IKK serine/threonine kinases have been shown to be involved in the activation of the NF-B pathway by the members of the TNFR and interleukin-1 receptor families (15). To determine the role of these proteins in EDAR-induced NF-B activation, we took advantage of the dominant-negative inhibitors of these kinases. As shown in Fig. 2B served effect. EDAR-induced NF-B activation was also effectively blocked by dominant-negative mutants of IKK1 (IKK1-K44M) and IKK2 (IKK2-K44M), respectively (Fig. 2C).
As IKKs function by mediating inducible phosphorylation and degradation of IB proteins, we tested the ability of a dominant-negative mutant of IB␣ (IB␣-S32A/S36A) to block NF-B induction by EDAR. This mutant contains serine-toalanine substitutions at amino acids 32 and 36, respectively, and is resistant to phosphorylation-induced degradation of IB␣ (16). As shown in Fig. 2D, NF-B induction by EDAR was effectively blocked by IB␣-S32A/S36A. Taken together, the above results suggest that EDAR activates NF-B by NIK-and IKK-induced phosphorylation and degradation of the IB␣ protein.
Mutagenesis Analysis of EDAR-induced NF-B Activation-We used C-terminal deletion mutagenesis to map the domain of EDAR responsible for NF-B activation. Deletion mutants EDAR⌬C38 and EDAR⌬C94, which are missing the C-terminal 38 and 94 amino acids, respectively, were only minimally effective in NF-B activation (Fig. 3, A and B). The EDAR⌬C38 mutant possesses a partial death domain, whereas the EDAR⌬C94 mutant is missing it entirely (Fig. 3A). These results suggest that the death domain plays a crucial role in NF-B activation by EDAR. A complete lack of NF-B activation was also observed upon expression of mutants EDAR⌬C164 and EDAR⌬C223, which are missing additional C-terminal sequences of the cytoplasmic domain (Fig. 3

, A and B).
We next tested whether two EDAR mutations seen in association with anhidrotic ectodermal dysplasia could affect the ability of EDAR to activate the NF-B pathway. The E379K mutation is an autosomal recessive mutation in the death domain of murine EDAR and is responsible for the spontaneous downless Jackson phenotype, whereas the R420Q mutation has been detected in the death domain of human EDAR in a family with autosomal dominant anhidrotic ectodermal dysplasia (2,3). We used site-directed mutagenesis to generate the corresponding mutants of the human EDAR gene. As shown in Fig.  3 (B and C), whereas the E379K mutant retained significant residual ability to activate the NF-B pathway, the R420Q mutant demonstrated a more severe loss of this activity. These results suggest that the recessive phenotype of the E379K mutant may be due to the need for two mutant alleles to significantly influence NF-B signaling. We would like to further point out that, in addition to the R420Q mutation, a nonsense mutation in the cytoplasmic domain of EDAR has also been detected in a family with autosomal dominant anhidrotic ectodermal dysplasia (2). This mutation (R358ter) results in the production of a truncated protein that is missing the C-terminal 90 amino acid residues and that closely resembles the deletion mutant EDAR⌬C94, which failed to activate the NF-B pathway (Fig. 3, A and B). Taken together, the above results suggest that the impaired ability to activate NF-B may be a key determinant in the pathogenesis of anhidrotic ectodermal dysplasia.
EDAR Activates the JNK Pathway-In addition to NF-B activation, different members of the TNFR family are also known to activate the JNK pathway. Therefore, we tested the ability of EDAR to activate this pathway using a luciferasebased c-Jun transcriptional activation assay. In this assay, luciferase expression is driven by JNK-mediated phosphorylation of the activation domain of transcription factor c-Jun that is fused to the Gal4 DNA-binding domain. As shown in Fig. 4A, expression of EDAR in the 293 EBNA cells led to modest activation of the JNK pathway. In contrast to the situation with NF-B activation, the JNK-inducing ability of EDAR was relatively weak as compared with that of TAJ/TROY. Activation of the JNK pathway by EDAR was further confirmed using a pull-down kinase assay based on in vitro phosphorylation of GST-c-Jun (Fig. 4B). We used deletion and point mutagenesis to map the region of the EDAR cytoplasmic domain responsible for JNK activation. These studies revealed that, although the EDAR⌬C38 and EDAR⌬C94 deletion mutant have some residual JNK activation ability, almost a complete lack of this ability is present in deletion mutants EDAR⌬C164 and EDAR⌬C223 (Fig. 4C). Thus, the putative death domain of EDAR is essential for both NF-B and JNK activation.
Finally, we tested the ability of the two EDAR mutants known to be associated with anhidrotic ectodermal dysplasia to activate the JNK pathway. As in the situation with NF-B activation, the E379K mutant was almost half as effective as the wild-type protein in JNK activation, whereas a more severe impairment of JNK activation was seen with the R420Q mutant.
TRAF2 has been shown to play an essential role in JNK activation via various members of the TNFR family (17,18). However, as shown in Fig. 4D, a dominant-negative mutant of TRAF2 that could effectively block CD40-induced JNK activation failed to block JNK activation via EDAR. These results suggest either that TRAF2 is not involved in JNK activation via EDAR or that it plays a functionally redundant role in this process. Finally, EDAR-induced JNK activation was effectively blocked by the JNK-binding domain of JIP1 (Fig. 4E), a recently described inhibitor of the JNK pathway (19).
EDAR Induces Caspase-independent Cell Death-As discussed above, EDAR is known to possess a region in its cytoplasmic domain with partial sequence homology to the "death domain" present in the apoptosis-inducing members of the TNFR family. Previous studies have demonstrated that transient transfection of EDA, the putative ligand for EDAR, in MCF7 cells leads to cellular rounding and detachment, which are not inhibited by the caspase inhibitor Z-VAD-fmk (20,21). Consistent with these results, transient transfection of EDAR in 293T, 293 EBNA, or MCF7 cells led to cellular rounding and detachment, two features suggestive of cell death (Fig. 5, A-C). However, unlike TNFR1-transfected cells, cellular rounding and detachment were relatively delayed features of EDARtransfected cells (24 versus 30 h). Finally, unlike TNFR1-transfected cells, those cells transfected with EDAR failed to demonstrate membrane budding, a feature associated with caspasedependent cell death (Fig. 5A).
We were next interested in testing whether EDAR-expressing cells actually undergo cell death. To test this hypothesis, we used nuclear staining with YOPRO-1, a cell-impermeable DNA-intercalating dye. As shown in Fig. 5D, although the majority of vector-transfected cells failed to show nuclear staining with this dye, a large number of EDAR-transfected cells stained positively, suggesting a lack of membrane integrity indicative of cell death. However, unlike the TNFR1-trans- fected cells, those cells dying in response to EDAR failed to show nuclear fragmentation, another key feature of caspasedependent cell death (Fig. 5D).
The lack of caspase activation during EDAR-induced cell death was characterized further by using several known inhibitors of this pathway. EDAR-induced cell death was not blocked by Z-VAD-fmk and t-butoxycarbonyl-Asp-fmk, two synthetic cell-permeable caspase inhibitors, and by CrmA and p35, two virally encoded caspase inhibitors. In contrast, all these caspase inhibitors effectively blocked cell death induced by TNFR1 (Fig. 6, A and B). Similarly, EDAR-induced cell death was not blocked by MRIT/cFLIP, MC159L, and dominant-negative mutants of caspase-8 (caspase-8-C360S) and FADD, suggesting that EDAR uses a FADD-and caspase-8-independent pathway for inducing cell death (Fig. 6C). Finally, EDAR-induced cell death was not blocked by a dominant-negative IB␣ mutant or JBD-JIP1, suggesting the lack of involvement of the NF-B and JNK pathways in this process (data not shown).
Caspase-3 is one of the executioner caspases of the caspase cascade and is activated during apoptosis induced by death receptors belonging to the TNFR1 family (22). We used a chromogenic assay, based on caspase-3-mediated cleavage of the chromogenic substrate Z-DEVD-p-nitroanilide, to test the activation of caspase-3 during EDAR-induced cell death. As shown in Fig. 6D, cell lysates from TNFR1-or DR4-transfected cells demonstrated caspase-3 activation, whereas EDAR-transfected cells failed to do so. Collectively, the above results suggest that cellular rounding and eventual cell death induced by EDAR overexpression are not mediated by a caspase-dependent mechanism.
We used deletion and point mutagenesis to map the region in the EDAR cytoplasmic domain responsible for induction of cell death. As shown in Fig. 6E, the EDAR⌬C38 deletion mutant, which retains a partial death domain, demonstrated significant residual ability to induce cell death, whereas the EDAR⌬C94 mutant, which completely lacks the death domain, possessed only a minor cell death-inducing ability. However, a complete lack of this activity was present in the deletion mutants lacking the death domain, such as EDAR⌬C164 and EDAR⌬C223. Finally, the point mutants E379K and R420Q demonstrated significant residual cell death-inducing ability.
EDAR Interacts with TRAFs and NIK, but Fails to Interact with FADD or TRADD-The death domain-containing adaptor proteins TRADD and FADD have been shown to play an essential role in signaling via various death domain-containing receptors of the TNFR family (4). Therefore, we tested the ability of these proteins to interact with EDAR using a coimmunoprecipitation assay. EDAR failed to coimmunoprecipitate FADD and TRADD when overexpressed with them in 293T cells (Fig. 7, A and B). Control experiments, performed in parallel, confirmed successful coimmunoprecipitation of FADD with Fas and TRADD with TNFR1, thereby demonstrating the validity of the assay.
In addition to the death domain-containing adaptor proteins, different members of TRAF family have been shown to interact with various members of the TNFR family (13). Therefore, we tested the ability of these proteins to interact with EDAR using coimmunoprecipitation assay. EDAR successfully coimmunoprecipitated murine TRAF1, murine TRAF2, and TRAF3 when these proteins were coexpressed in 293T cells (Fig. 7, C-F). Deletion mutagenesis revealed that the C-terminal 94 amino acids, encoding the death domain, were not essential for interaction of EDAR with murine TRAF1. Consistent with this hypothesis, the death domain point mutants E79K and R420Q were as effective as the wild-type protein in coimmunoprecipitating murine TRAF1. Finally, EDAR successfully coimmunoprecipitated with NIK, a protein known to be involved in NF-B activation by various members of the TNFR family (Fig.  7G). However, we have so far failed to detect an interaction between a GST fusion protein containing the EDAR cytoplasmic domain and in vitro transcribed and translated TRAF2 or NIK (data not shown). Similarly, no interaction has been detected between the cytoplasmic domain of EDAR and TRAF2 or NIK using a yeast two-hybrid assay (data not shown). Collec- tively, the above results suggest that the interaction between EDAR and TRAFs or NIK might be facilitated by the presence of intermediate bridging proteins present in the 293T cells.
EDA Is the Ligand for EDAR-EDA is believed to be the ligand for EDAR based on the similarity in the clinical features of genetic disorders resulting from the mutations in these genes (2,23,24). However, EDA has never been shown to bind to EDAR. To test the ability of EDA to bind to EDAR, we generated a baculovirus construct containing the extracellular receptor-binding domain of EDA fused to an N-terminal Myc epitope tag. As shown in Fig. 8A, the Myc-EDA I construct contained the second and third collagenous repeats of the EDA-A1 isoform, in addition to its TNF homology domain. Myc-tagged soluble EDA proteins was collected from the supernatant of baculovirus-infected insect cells and tested for its ability to bind to EDAR-Fc or mTAJ-Fc in a coimmunoprecipitation assay. As shown in Fig. 8B, EDAR-Fc successfully coimmunoprecipitated the Myc-EDA I protein, whereas mTAJ-Fc failed to do so. The ability of Myc-EDA I to bind to EDAR was further confirmed using enzyme-linked immunosorbent assay (Fig. 8C). DISCUSSION During the process of skin differentiation, the mitotically active cells of the basal epithelium cease proliferating and then migrate outwards and undergo terminal differentiation (25). The NF-B proteins are initially present in the cytoplasm of basal cells, but later migrate to the nuclei of suprabasal cells, suggesting a role for NF-B activation in the switch from proliferation to growth arrest and differentiation (25). This hypothesis is supported by the result of a recent study involving a functional blockade of NF-B via the expression of a dominant-negative mutant of IB␣ in transgenic murine and human epidermis. This mutant resulted in the production of hyperplastic epithelium due to increased thickness of the suprabasal squamous layer (25). More recently, targeted disruption of the IKK1 gene has resulted in a similar phenotype (26 -28). As the IKK1-deficient keratinocytes exhibited near normal IKK activation in response to a number of pro-inflammatory stimuli, such as TNF-␣, interleukin-1, and lipopolysaccharide, these results have led to the suggestion that IKK1 is critical for IB-dependent activation of NF-B in response to an as yet unidentified developmental signal that triggers keratinocyte differentiation (26 -28). In the present study, we demonstrate that EDAR-induced NF-B activation is effectively blocked by a dominant-negative IKK1 mutant, suggesting that EDAR may be the missing developmental signal required for keratinocyte differentiation. The key role of EDAR-induced NF-B activation in the process of ectodermal differentiation is also supported by our results demonstrating the impaired ability of EDAR mutants associated with anhidrotic ectodermal dysplasia to activate the NF-B pathway. However, these mutants also demonstrated impaired ability to activate the JNK and cell death pathways. Thus, it is conceivable that defects in EDAR-mediated JNK and cell death pathways may also contribute to the clinical and pathological phenotype of anhidrotic ectodermal dysplasia.
We have observed that EDAR-induced NF-B activation could be blocked by a dominant-negative mutant of IKK2. However, no obvious ectodermal defect has been reported in IKK2-deficient animals (29,30). These results might be explained by the functionally redundant role of IKK2 in EDARinduced NF-B activation. Premature embryonic lethality of the IKK2-deficient animals might have also prevented the manifestation of the ectodermal defects in these animals (29,30).
Our results further suggest that, like the other death domain receptors of the TNFR family, the death domain of EDAR plays a key role in the activation of the NF-B, JNK, and cell death pathways. However, the putative death domain of EDAR possesses only a weak sequence homology to the classical death domains present in the known apoptosis-inducing death receptors and does not interact with either TRADD or FADD. Therefore, we tend to favor the hypothesis that the death domain of EDAR may be involved in NF-B, JNK, and cell death pathways by acting as a more general protein recruitment domain, as has been suggested recently (3,31). Although EDAR could interact with various TRAF family members and NIK in the coimmunoprecipitation assay in 293T cells, we have so far failed to detect an interaction between these proteins in mammalian cell-free systems, suggesting that the interaction between EDAR and TRAFs or NIK might be facilitated by the presence of intermediate bridging proteins. Future studies aimed at isolation of the adaptor proteins that directly bind to the cytoplasmic domain of EDAR will greatly enhance our understanding of EDAR signaling in the process of ectodermal differentiation.
Previous studies had demonstrated that transfection of an EDA expression construct in MCF7 cells led to cell rounding and detachment, resembling the morphology of cells undergoing cell death (20,21). However, these morphological changes could not be blocked by the cell-permeable caspase inhibitor Z-VAD-fmk, suggesting the lack of a role for caspase activation in this process. In the present study, we have similarly demonstrated that transfection of EDAR in 293T, 293 EBNA, and MCF7 cells also leads to cell rounding and detachment, followed by cell death. Cells dying in response to EDAR overexpression do not show any morphological or biochemical features of caspase activation, suggesting that EDAR induces cell death by using a caspase-independent mechanism. Such a caspaseindependent form of cell death has been described previously for several death domain-and non-death domain-containing members of the TNFR family, and it remains to be seen whether EDAR shares a common mechanism of cell death induction with them (11,(32)(33)(34)(35)(36)(37)(38)(39). Several potential mediators of caspase-independent cell death have been recently described, such as Bax, nitric oxide, and apoptosis-inducing factor (40 -44). It will be interesting to test the involvement of these proteins in EDAR-induced cell death. It is also conceivable that cell death induced by EDAR is a consequence of cellular detachment from the plate.
Finally, in this report, we demonstrate for the first time that extracellular domains of EDAR and EDA can physically interact with each other. EDA is unique among the ligands of the TNF family in possessing three collagenous repeat domains, in addition to a TNF homology domain (20,21,24,45,46). Like other ligands of the TNF family, EDA is expressed in a trimeric form, and it is conceivable that the collagenous repeats of EDA help in this process (45). Several alternatively spliced isoforms of EDA that lack one or more of its subdomains have been described recently (20,24,46). Future studies should address the role of each of these subdomains of EDA in its interaction with EDAR.