Two-Amino Acid Molecular Switch in an Epithelial Morphogen That Regulates Binding to Two Distinct Receptors

See allHide authors and affiliations

Science  20 Oct 2000:
Vol. 290, Issue 5491, pp. 523-527
DOI: 10.1126/science.290.5491.523


Ectodysplasin, a member of the tumor necrosis factor family, is encoded by the anhidrotic ectodermal dysplasia (EDA) gene. Mutations in EDA give rise to a clinical syndrome characterized by loss of hair, sweat glands, and teeth. EDA-A1 and EDA-A2 are two isoforms of ectodysplasin that differ only by an insertion of two amino acids. This insertion functions to determine receptor binding specificity, such that EDA-A1 binds only the receptor EDAR, whereas EDA-A2 binds only the related, but distinct, X-linked ectodysplasin-A2 receptor (XEDAR). In situ binding and organ culture studies indicate that EDA-A1 and EDA-A2 are differentially expressed and play a role in epidermal morphogenesis.

Members of the tumor necrosis factor receptor (TNFR) superfamily are involved in a number of physiological and pathological responses by activating a wide variety of intracellular signaling pathways. In a database search based on sequence similarity (1), XEDAR was initially identified as a member of the TNFR superfamily.

The deduced amino acid sequence of XEDAR contains three cysteine-rich repeats and a single transmembrane region (Fig. 1A). XEDAR lacks an NH2-terminal signal peptide. The presence of an in-frame upstream stop codon in both human and mouse cDNA clones indicated that the sequence shown in Fig. 1A represents the full-length open reading frame (ORF) (1). To confirm that XEDAR was indeed a membrane protein, we transfected MCF7 cells with either an NH2- or COOH-terminal Flag-tagged version of XEDAR (Fig. 1B). In permeabilized cells, the expression of both tagged proteins was readily detected by anti-Flag immunostaining. However, in the absence of permeabilization, cell surface staining was only observed in cells transfected with NH2-terminal–tagged XEDAR, consistent with it being a type III transmembrane protein with an extracelluar NH2-terminus and a cytoplasmic COOH-terminus (2). The XEDAR gene was localized to the X chromosome (3).

Figure 1

XEDAR is a type III transmembrane protein of the TNFR superfamily. (A) Amino acid sequence of human XEDAR, Troy, and EDAR (21). The cysteine-rich repeat regions are underlined, and the predicted membrane-spanning region is doubly underlined. Conserved residues are boxed. (B) MCF-7 cells transiently transfected with XEDAR with either NH2-terminal Flag tag (Flag-XEDAR) or COOH-terminal Flag tag (XEDAR-Flag) were immunostained with M2 antibody to Flag (Sigma), either with or without permeabilization with 0.5% Triton X-100.

XEDAR failed to bind many known ligands of the tumor necrosis factor (TNF) superfamily, including APRIL, Blys/TALL-1, 4-1BBL, CD27L, CD30L, CD40L, FasL, GITRL, OX-40L, RANKL, TNF-α, or TRAIL/APO2L (3). Hypohidrotic (anhidrotic) ectodermal dysplasia (EDA) is a disorder characterized by the abnormal development of hair, teeth, and eccrine sweat glands (4). Mutations in the X-linked EDA gene, a distantly related member of the TNF-ligand superfamily, are responsible for most of the clinical cases studied to date (5, 6). A minority of patients with the EDA phenotype display an autosomal inheritance pattern that is due to mutations in a distinct gene, termed downless (dl) in mice and DL in humans (7, 8). The gene for EDA encodes various isoforms of a type II transmembrane protein, termed ectodysplasin. The longest form, EDA-A1, encodes a 391-residue protein with a domain similar to TNF at the COOH-terminus (7–9). Except for a two-residue deletion (Glu308 and Val309 in this domain), EDA-A2 is identical to EDA-A1. Genomic analysis of the EDA gene reveals that EDA-A2 is generated through the use of an alternative internal splice donor site (10). The Tabby (Ta) gene represents the ortholog of human EDA (11, 12). Virtually superimposable genomic organization and transcript profile have been described for Ta (9), implying evolutionary conservation.

To test whether EDA-A1 or EDA-A2 could be the ligand for XEDAR, we generated XEDAR-hFc, containing the extracellular domain of XEDAR fused to human immunoglobulin G1 Fc (hFc) (13,14). Surprisingly, XEDAR-hFc bound cells that were transfected with full-length EDA-A2 but not cells transfected with EDA-A1, despite there only being a two–amino acid difference between these two molecules (Fig. 2A). The dl gene responsible for the autosomal form of EDA is predicted to encode a TNFR member, termed EDAR (7, 8). Genetic evidence suggests that EDAR is likely the receptor for EDA. We generated EDAR-hFc, consisting of the extracellular domain of EDAR fused to hFc (13). EDAR-hFc bound cells transfected with EDA-A1, the ligand possessing the extra two-residue insert (Glu308 and Val309), but not cells transfected with EDA-A2 (Fig. 2A). This was the opposite of what was observed with XEDAR-hFc, which bound EDA-A2 but not EDA-A1. The reciprocal binding pattern was also demonstrable with soluble forms of ligands and cell surface–expressed receptors (Fig. 2B). Tagged soluble ligands were generated (15,16) with human placental alkaline phosphatase (AP) or Flag-tag fused to the COOH-terminal regions of EDA-A1 and EDA-A2. AP-EDA-A1 and AP-EDA-A2 specifically stained cells transfected with EDAR or XEDAR, respectively (Fig. 2B). The binding of AP fusion proteins was specifically blocked by the corresponding Flag-tagged versions. Specific interactions of EDAR/EDA-A1 and XEDAR/EDA-A2 were confirmed by coimmunoprecipitation experiments (Fig. 2C). We also measured the binding of soluble ligand to immobilized receptor-Fc fusion protein (Fig. 2D). EDA-A2 exhibited an affinity for XEDAR similar to that of EDA-A1 for EDAR, both binding with an apparent dissociation constant of ∼600 pM. Additionally, there was essentially no cross binding, even at high concentrations of ligand. Thus, the insertion of two amino acids in the ligand unequivocally dictated receptor binding specificity.

Figure 2

Specific interactions of EDAR with EDA-A1 and XEDAR with EDA-A2. (A) COS-7 cells transfected with EDA-A1 or EDA-A2 were incubated with EDAR-hFc or XEDAR-hFc and stained with biotinylated goat antibody to human Fc and Cy3-streptavidin. (B) COS-7 cells transfected with either EDAR or XEDAR were stained with AP-EDA-A1 or AP-EDA-A2 (16). Where indicated, cells were preincubated with purified Flag-tagged ligand. (C) EDAR-hFc, XEDAR-hFc, or TNFR1-hFc was incubated with Flag-EDA-A1, Flag-EDA-A2, or conditioned medium contain- ing Flag-AP-EDA-A1 or Flag-AP-EDA-A2. The immunoprecipitated receptor-Fc fusion proteins were subjected to anti-Flag Western blot. (D) Dosage-dependent binding of Flag-EDA-A1 and Flag-EDA-A2 to XEDAR-hFc (top) and EDAR-hFc (bottom) (14). (E) Molecular surfaces of models of EDA-A1 and EDA-A2 colored according to their calculated surface potential from –10 kT (red) to 0 kT (white) to 10 kT (blue). EV, Glu308Val309. The expected receptor contact regions are outlined in green. This figure was made with the program GRASP (22).

To better understand the nature of this molecular switch, we constructed models of EDA-A1 and EDA-A2 by sequence alignment and solved x-ray structures of family members, including lymphotoxin (17) and Apo2L/TRAIL (18, 19) as a template. The models (Fig. 2E) showed that the two-residue insert in EDA-A1 was predicted to be on the surface of the protein in an area expected to interact with the receptors (3). The extracellular region of XEDAR displays the highest sequence similarity to Troy (20), a recently identified orphan TNFR family member (Fig. 1A). Troy, however, did not bind either EDA-A1 or EDA-A2 (3).

Many TNFR family members activate the transcription factor nuclear factor κB (NF-κB). Upon transfection into 293 EBNA (293E) cells, XEDAR induced NF-κB activation (3). When the amount of input DNA was low, the small amounts of expressed XEDAR induced only minimal activation of NF-κB. However, NF-κB could be activated by exposure to the appropriate ligand (Fig. 3A). Even in the absence of XEDAR transfection, there was a weak activation of the NF-κB upon addition of purified Flag–EDA-A2 protein or transfection of EDA-A2, suggesting that 293E cells express low levels of endogenous receptor for EDA-A2 (Fig. 3, A and B). Treatment of untransfected 293E cells with purified Flag–EDA-A2 but not with Flag-EDA-A1 resulted in increased phosphorylation of inhibitor of nuclear factor κB–α (IκB-α), and this effect of Flag–EDA-A2 was specifically blocked by XEDAR-hFc but not by EDAR-hFc (Fig. 3C). Taqman analysis revealed a high level of XEDAR transcript in 293E cells. Additionally, a monoclonal antibody specific for XEDAR blocked EDA-A2–induced IκB-α phosphorylation (3). Thus, endogenous XEDAR is responsible for mediating NF-κB activation in 293E cells following exposure to EDA-A2.

Figure 3

Signal transduction mediated by EDAR and XEDAR. (A and B) Interaction of EDA-A2 with XEDAR results in activation of NF-κB. In (A), purified Flag-EDA-A1 or Flag-EDA-A2 was used; in (B), the cells were cotransfected with full-length EDA-A1 or EDA-A2. (C) Activation of NF-κB by treatment of EDA-A2 in untransfected 293E cells. Total cell lysates were subjected to Western blot using phosphospecific IκB-α antibody (New England BioLabs, Beverly, Massachusetts). (D) Analysis of XEDAR binding TRAF6 and activating NF-κB. XEDAR and its mutants were transiently transfected into 293E cells to measure their activities of NF-κB activation. Glutathione S-transferase fusion protein containing the cytoplasmic region of XEDAR or its variants was used to measure the interaction with Flag-TRAF6. E256K, Glu256 → Lys256; ED, extracellular domain; TM, transmembrane; CD, cytoplasmic domain; WT, wild type. (E) Ligand-dependent interaction between XEDAR and TRAF6. Low levels of differentially tagged XEDAR and TRAF6 were cotransfected into 293E cells. As indicated, cells were treated with purified Flag-EDA-A2 for 10 min before cells were lysed. IP, immunoprecipitation; WB, Western blotting. (F) Binding of EDA-A1 to EDAR results in activation of NF-κB in MCF7 cells. (G) Mutation in the cytoplasmic region of EDAR impairs EDAR-mediated NF-κB activation in 293E cells (21). (H) XEDAR but not EDAR mediates activation of Elk1 in 239E cells (PathDetect, Stratagene). Expression of receptor proteins was shown by anti-Flag Western blotting.

Members of the TNF receptor–associated factor (TRAF) family mediate signaling by TNFRs. The cytoplasmic region of XEDAR bound TRAF1, TRAF3, and TRAF6 (3). The TRAF6 binding site was mapped to amino acids 252 to 260, and replacing Glu256 with Arg abolished the binding of TRAF6 to XEDAR. Because the binding of TRAF6 (but not other TRAFs) to XEDAR correlated with activation of NF-κB (3), TRAF6 is likely a key adapter molecule in transducing XEDAR-mediated NF-κB signaling (Fig. 3D). In cells transfected with low amounts of TRAF6 and XEDAR, the interaction between the molecules was marginal. However, treatment with soluble Flag–EDA-A2 increased the recruitment of TRAF6 to XEDAR (Fig. 3E).

Transfection of EDAR into 293E cells also resulted in a dose-dependent activation of NF-κB (3). Unlike 293E cells, untransfected MCF7 cells do not respond to either EDA-A1 or EDA-A2 by activating the NF-κB pathway (Fig. 3F). In MCF7 cells, transfection of EDAR alone resulted in a dose-dependent activation of NF-κB. Synergistic activation of NF-κB was found upon cotransfection with EDA-A1. Furthermore, EDAR bearing the point mutation (Arg420 → Gln420) identified in a patient with an autosomal form of EDA (8) or the point mutation (Glu379 → Lys379) identified in dl mutant mice (7) was incapable of NF-κB activation (Fig. 3G), suggesting that engagement of the NF-κB pathway is critical for EDAR function. The cytoplasmic region of EDAR failed to bind any known TRAF molecules (3). Transient transfection of 293E cells with XEDAR, but not EDAR, activated the mitogen-activated protein kinase pathway as assessed by a luciferase-based Elk1 transcriptional assay (Fig. 3H).

We examined the expression of EDAR and XEDAR in developing mouse skin by in situ hybridization (Fig. 4A). At embryonic day 14 (E14), EDAR transcripts were clearly present in the basal cells of the developing epidermis, with elevated focal expression in placodes. XEDAR expression was barely discernable at this stage. However, by E16 and E17, both receptors were expressed in large amounts in the maturing follicles. By postnatal day 1 (P1), the pattern of expression was confined to the hair follicles.

Figure 4

Expression of EDAR, XEDAR, EDA-A1, and EDA-A2 in mouse skin. (A) Expressions of EDAR and XEDAR in mouse skin were examined by in situ hybridization (23). The digoxigenin-labeled riboprobes for mouse EDAR and XEDAR were synthesized corresponding to nucleotides 12 to 493 of the EDAR ORF and nucleotides 37 to 767 of the XEDAR ORF, respectively (1 and 3). The same sections were also immunostained with anti-keratin 5 (green) (2 and 4). (B and C) In situ staining of mouse skin with EDAR-hFc and XEDAR-hFc fusion proteins (red) (24). In (C), the mouse skin section was stained twice with anti-keratin 5 (green). A transverse view of P1 hair follicles shows that XEDAR-hFc and EDAR-hFc bind distinct regions. (D) Stimulation of epidermis downgrowths by EDA-A1 and EDA-A (25). Skin was dissected from CD-1 mouse embryos between E13.5 and E14.5 and cultured for 3 days in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum. Purified recombinant Flag-EDA-A1, Flag-EDA-A2, or Flag-Blys (control) was used at a final concentration of 0.2 μg/ml. Skin sections were then immunostained with anti-keratin 5 to assess the number of epidermal downgrowths. The average number of downgrowths per unit length of epidermis was calculated on the basis of a total of 18 organ cultures in three different experiments.

We used the receptor Fc-fusion proteins to determine endogenous expression of EDA-A1 and EDA-A2 proteins in mouse skin. EDAR-hFc bound epidermis and hair follicles, but not dermal cells. Binding of XEDAR-hFc appeared later than that of EDAR-hFc and was mainly restricted to hair follicles (Fig. 4B). Detailed examination showed that EDA-A2 was more concentrated in the central core of the developing hair follicle, whereas EDA-A1 expression of EDA-A1 was circumferential (Fig. 4C). Regardless, the distinctive temporal and spatial expression of EDA-A1 and EDA-A2 suggest that they may have distinct roles in development of the hair follicle. In a skin organ culture system, both EDA-A1 and EDA-A2 increased the number of epidermal invaginations that are a precursor stage to mature hair follicles. EDA-A1 had a slightly higher activity than EDA-A2 (Fig. 4D).

Members of the TNFR family are important in immunity and inflammation. Our study indicates the involvement of this family in morphogenesis.

  • * These authors contributed equally to this work.

  • To whom correspondence should be addressed. E-mail: dixit{at}


View Abstract

Stay Connected to Science

Editor's Blog

Navigate This Article