Mutations Leading to X-linked Hypohidrotic Ectodermal Dysplasia Affect Three Major Functional Domains in the Tumor Necrosis Factor Family Member Ectodysplasin-A
2001; Elsevier BV; Volume: 276; Issue: 22 Linguagem: Inglês
10.1074/jbc.m101280200
ISSN1083-351X
AutoresPascal Schneider, Summer L. Street, Olivier Gaide, Sylvie Hertig, Aubry Tardivel, Jürg Tschopp, Laura Runkel, Konstantinos Alevizopoulos, Betsy Ferguson, Jonathan Zonana,
Tópico(s)Oral and Maxillofacial Pathology
ResumoMutations in the epithelial morphogen ectodysplasin-A (EDA), a member of the tumor necrosis factor (TNF) family, are responsible for the human disorder X-linked hypohidrotic ectodermal dysplasia (XLHED) characterized by impaired development of hair, eccrine sweat glands, and teeth. EDA-A1 and EDA-A2 are two splice variants of EDA, which bind distinct EDA-A1 and X-linked EDA-A2 receptors. We identified a series of novel EDA mutations in families with XLHED, allowing the identification of the following three functionally important regions in EDA: a C-terminal TNF homology domain, a collagen domain, and a furin protease recognition sequence. Mutations in the TNF homology domain impair binding of both splice variants to their receptors. Mutations in the collagen domain can inhibit multimerization of the TNF homology region, whereas those in the consensus furin recognition sequence prevent proteolytic cleavage of EDA. Finally, a mutation affecting an intron splice donor site is predicted to eliminate specifically the EDA-A1 but not the EDA-A2 splice variant. Thus a proteolytically processed, oligomeric form of EDA-A1 is required in vivo for proper morphogenesis. Mutations in the epithelial morphogen ectodysplasin-A (EDA), a member of the tumor necrosis factor (TNF) family, are responsible for the human disorder X-linked hypohidrotic ectodermal dysplasia (XLHED) characterized by impaired development of hair, eccrine sweat glands, and teeth. EDA-A1 and EDA-A2 are two splice variants of EDA, which bind distinct EDA-A1 and X-linked EDA-A2 receptors. We identified a series of novel EDA mutations in families with XLHED, allowing the identification of the following three functionally important regions in EDA: a C-terminal TNF homology domain, a collagen domain, and a furin protease recognition sequence. Mutations in the TNF homology domain impair binding of both splice variants to their receptors. Mutations in the collagen domain can inhibit multimerization of the TNF homology region, whereas those in the consensus furin recognition sequence prevent proteolytic cleavage of EDA. Finally, a mutation affecting an intron splice donor site is predicted to eliminate specifically the EDA-A1 but not the EDA-A2 splice variant. Thus a proteolytically processed, oligomeric form of EDA-A1 is required in vivo for proper morphogenesis. ectodysplasin-A tumor necrosis factor X-linked hypohidrotic ectodermal dysplasia ectodysplasin-A1 receptor X-linked ectodysplasin-A2 receptor Fas ligand amino acids single-stranded conformation polymorphism Chinese hamster ovary polymerase chain reaction phosphate-buffered saline wild type enzyme-linked immunosorbent assay The ED1 gene encodes a protein, ectodysplasin-A (EDA),1 recently recognized to be a member of the tumor necrosis factor (TNF) superfamily of ligands. Mutations within the ED1 gene cause an X-linked recessive disorder, hypohidrotic or anhidrotic ectodermal dysplasia (ED1, XLHED) (Mendelian inheritance in man 305100), involving abnormal morphogenesis of teeth, hair, and eccrine sweat glands. Various splice forms of the ED1 transcript have been detected, but two isoforms differing only by two amino acids, EDA-A1 (391 aa) and EDA-A2 (389 aa), contain a TNF homology domain (1Bayes M. Hartung A.J. Ezer S. Pispa J. Thesleff I. Srivastava A.K. Kere J. Hum. Mol. Genet. 1998; 7: 1661-1669Crossref PubMed Scopus (181) Google Scholar, 2Monreal A.W. Zonana J. Ferguson B. Am. J. Hum. Genet. 1998; 63: 380-389Abstract Full Text Full Text PDF PubMed Scopus (185) Google Scholar, 3Yan M. Wang L.C. Hymowitz S.G. Schilbach S. Lee J. Goddard A. de Vos A.M. Gao W.Q. Dixit V.M. Science. 2000; 290: 523-527Crossref PubMed Scopus (236) Google Scholar). EDA is a type II transmembrane protein with a small N-terminal intracellular domain and a larger C-terminal extracellular domain containing a (Gly-X-Y)19 collagen-like repeat with a single interruption and a C-terminal TNF homology domain (Fig. 1 A). The TNF homology domain is similar to other members of the TNF family, consisting of 10 predicted anti-parallel β-sheets linked by variable loops (Fig. 1 A). TNF family ligands homotrimerize to form a pear-shaped quaternary structure able to bind a receptor molecule at each monomer-monomer interface (4Hymowitz S.G. O'Connell M.P. Ultsch M.H. Hurst A. Totpal K. Ashkenazi A. de Vos A.M. Kelley R.F. Biochemistry. 2000; 39: 633-640Crossref PubMed Scopus (228) Google Scholar, 5Banner D.W. D'Arcy A. Janes W. Gentz R. Schoenfeld H.J. Broger C. Loetscher H. Lesslauer W. Cell. 1993; 73: 431-445Abstract Full Text PDF PubMed Scopus (974) Google Scholar). The closest EDA homologues in the TNF family are BAFF/BLyS, APRIL, and TWEAK, although none of them contains collagen-like repeats (6Moore P.A. Belvedere O. Orr A. Pieri K. LaFleur D.W. Feng P. Soppet D. Charters M. Gentz R. Parmelee D. Li Y. Galperina O. Giri J. Roschke V. Nardelli B. Carrell J. Sosnovtseva S. Greenfield W. Ruben S.M. Olsen H.S. Fikes J. Hilbert D.M. Science. 1999; 285: 260-263Crossref PubMed Scopus (1006) Google Scholar, 7Schneider P. MacKay F. Steiner V. Hofmann K. Bodmer J.L. Holler N. Ambrose C. Lawton P. Bixler S. Acha-Orbea H. Valmori D. Romero P. Werner-Favre C. Zubler R.H. Browning J.L. Tschopp J. J. Exp. Med. 1999; 189: 1747-1756Crossref PubMed Scopus (1130) Google Scholar, 8Hahne M. Kataoka T. Schroter M. Hofmann K. Irmler M. Bodmer J.L. Schneider P. Bornand T. Holler N. French L.E. Sordat B. Rimoldi D. Tschopp J. J. Exp. Med. 1998; 188: 1185-1190Crossref PubMed Scopus (459) Google Scholar, 9Chicheportiche Y. Bourdon P.R. Xu H. Hsu Y.M. Scott H. Hession C. Garcia I. Browning J.L. J. Biol. Chem. 1997; 272: 32401-32410Abstract Full Text Full Text PDF PubMed Scopus (561) Google Scholar). All four ligands contain consensus sequences for proteolytic cleavage by furin within their extracellular domain. In the case of EDA, two overlapping consensus sites are located between the transmembrane and the collagen-like domains (Fig. 1 A). EDA-A1, but not EDA-A2, has been shown to specifically bind to EDAR, a member of the TNF receptor superfamily that, like most members of the TNF receptor family, activates the NF-κB and c-Jun N-terminal kinase pathways (3Yan M. Wang L.C. Hymowitz S.G. Schilbach S. Lee J. Goddard A. de Vos A.M. Gao W.Q. Dixit V.M. Science. 2000; 290: 523-527Crossref PubMed Scopus (236) Google Scholar, 10Tucker A.S. Headon D.J. Schneider P. Ferguson B.M. Overbeek P. Tschopp J. Sharpe P.T. Development. 2000; 127: 4691-4700PubMed Google Scholar). Mutations in DL (EDAR), the human homologue of the murine downless locus, produce an identical phenotype to loss of function of EDA (11Headon D.J. Overbeek P.A. Nat. Genet. 1999; 22: 370-374Crossref PubMed Scopus (308) Google Scholar, 12Monreal A.W. Ferguson B.M. Headon D.J. Street S.L. Overbeek P.A. Zonana J. Nat. Genet. 1999; 22: 366-369Crossref PubMed Scopus (323) Google Scholar). XEDAR, another member of the TNF receptor superfamily that also activates the NF-κB pathway, binds EDA-A2 but not EDA-A1. Although EDA-A1 and EDA-A2 are closely related splice variants, the respective proteins appear to have different patterns of expression in mouse skin and hair follicles (3Yan M. Wang L.C. Hymowitz S.G. Schilbach S. Lee J. Goddard A. de Vos A.M. Gao W.Q. Dixit V.M. Science. 2000; 290: 523-527Crossref PubMed Scopus (236) Google Scholar). Intracellular signals elicited by EDA in vivo rely at least in part on the activation of NF-κB, because a rare form of HED associated with immunodeficiency (HED-ID) correlates with mutations in NEMO/IKK-γ, an essential component of the NF-κB pathway (13Zonana J. Elder M.E. Schneider L.C. Orlow S.J. Moss C. Golabi M. Shapira S.K. Farndon P.A. Wara D.W. Emmal S.A. Ferguson B.M. Am. J. Hum. Genet. 2000; 67: 1555-1562Abstract Full Text Full Text PDF PubMed Scopus (419) Google Scholar).In order to get insight into the structure-function relationship of EDA, we identified 44 mutations (17 of which have not been reported previously) in unrelated families with XLHED and studied their effect on the properties of EDA in vitro. The mutations clustered in three functionally important domains as follows: a TNF homology domain necessary for receptor binding, a bundle-forming collagen domain, and a cleavage site for a furin protease. This indicates that the receptor binding ability of EDA and also its oligomerization and proteolytic processing to a soluble form are critical events for its action in vivo.DISCUSSIONIn this study, mutations in the EDA gene were detected in 63% of the families with XLHED, which is lower than the 95% rate we found previously (2Monreal A.W. Zonana J. Ferguson B. Am. J. Hum. Genet. 1998; 63: 380-389Abstract Full Text Full Text PDF PubMed Scopus (185) Google Scholar) by the direct sequencing of affected males. The lower detection rate is the consequence of two factors. The use of a single set of conditions lowers the sensitivity of SSCP analysis, as only 11 of the 15 known mutations run as controls could be detected under these conditions (73%). In addition, this study included 28% of families with "affected" females only, which was not the case in our previous study. Indeed, the detection rate was lower in families with female probands (45%), and this may well be due to genetic heterogeneity for autosomal forms of HED.A number of point mutations are located within the TNF homology domain of EDA, but only one of them (Y343C) affects a residue which, based on structural homology with known ligand-receptor structures (4Hymowitz S.G. O'Connell M.P. Ultsch M.H. Hurst A. Totpal K. Ashkenazi A. de Vos A.M. Kelley R.F. Biochemistry. 2000; 39: 633-640Crossref PubMed Scopus (228) Google Scholar, 5Banner D.W. D'Arcy A. Janes W. Gentz R. Schoenfeld H.J. Broger C. Loetscher H. Lesslauer W. Cell. 1993; 73: 431-445Abstract Full Text PDF PubMed Scopus (974) Google Scholar), is predicted to interact with the receptor. Indeed, this mutation abolished receptor binding without affecting the trimeric structure of EDA, although we cannot exclude indirect conformational effects. All other mutations are predicted to have indirect effect on receptor binding, e.g. by altering the folding of EDA. It is noteworthy that four independent mutations (G291W, G291R, A356D, R357P) occurred in two short loops at the bottom of EDA (loops BCand FG, see Fig. 1 A). The affected amino acids are probably crucial for proper folding of the monomer, as mutation A356D resulted in insoluble EDA-A1 and EDA-A2. Another group of mutations (H252L, S374R) seems to affect the stability of the trimer, because the resulting proteins contain a proportion of monomers. The propensity of unglycosylated subunits to form larger aggregates support the idea that glycosylation of some TNF family members promotes their solubility (18Schneider P. Bodmer J.L. Holler N. Mattmann C. Scuderi P. Terskikh A. Peitsch M.C. Tschopp J. J. Biol. Chem. 1997; 272: 18827-18833Abstract Full Text Full Text PDF PubMed Scopus (158) Google Scholar). Interestingly, one of the mutations (S374R) destroys a potential N-glycosylation site without affecting the glycosylation of EDA, suggesting that this particular site is not recognized by the N-glycosyltransferase. A single mutant (S374R) retained some binding activity to EDAR. Although preferential binding of the unglycosylated (and most probably aggregated) isoform was observed, the interaction appeared specific, suggesting that residual activity may be associated with this mutation in vivo. In general, no apparent phenotype/genotype correlation was observed, but the family with missense mutation S374R had two affected males and an affected maternal grandfather with isolated hypodontia. Whether residual activity of this mutant may account for the milder phenotype, and whether there is truly a tissue-specific difference in the function of this mutant protein remains to be determined. Finally, one mutation (T378M) affected secretion and aggregation of EDA-A1 and EDA-A2 in a strikingly different manner. The structural reason for this differential behavior is unclear. In summary, all missense mutations in the TNF homology domain result in abolished or much impaired binding of EDA-A1 to EDAR and EDA-A2 to XEDAR.A number of mutations occurring outside the TNF homology domain did not affect binding to the receptor in an in vitro assay, indicating that the interaction of EDA with its receptor(s) is necessary but not sufficient for its function in vivo. In particular, the integrity of the collagen domain appears to be functionally essential, and we provided evidence that it may serve to multimerize EDA trimers. This is in strong support of the hypothesis that EDA belongs to the C1q as well as to the TNF family of proteins (19Mikkola M.L. Pispa J. Pekkanen M. Paulin L. Nieminen P. Kere J. Thesleff I. Mech. Dev. 1999; 88: 133-146Crossref PubMed Scopus (106) Google Scholar). C1q family members are characterized by the presence of a C-terminal globular trimeric domain, with striking structural homology to TNF (20Shapiro L. Scherer P.E. Curr. Biol. 1998; 8: 335-338Abstract Full Text Full Text PDF PubMed Google Scholar), which is prolonged by a collagen triple helix further assembling into an N-terminal bundle structure, giving rise to a highly multimeric superstructure. In the TNF family, it has been shown that the activity of soluble trimers can be dramatically increased by antibody-mediated multimerization, thereby mimicking the membrane-bound form of the ligand (16Schneider P. Holler N. Bodmer J.L. Hahne M. Frei K. Fontana A. Tschopp J. J. Exp. Med. 1998; 187: 1205-1213Crossref PubMed Scopus (696) Google Scholar, 21Grell M. Douni E. Wajant H. Lohden M. Clauss M. Maxeiner B. Georgopoulos S. Lesslauer W. Kollias G. Pfizenmaier K. Cell. 1995; 83: 793-802Abstract Full Text PDF PubMed Scopus (1144) Google Scholar). A highly multimeric structure of EDA would provide a powerful means for the soluble protein to signal through high valency receptor clustering. In line with this hypothesis, we found that a naturally occurring point mutation (G207R) in the collagen domain completely abolished the bundle effect. It is, however, likely that the collagen domain also serves additional functions; a number of families with XLHED displayed in frame deletion of 2 or 4 GlyXY repeats in the predicted bundle domain of the collagen triple helix, i.e. before the interruption in the GlyXY repeats. Deletions of 2 GlyXY repeats can also be found C-terminal to the interruption. Surprisingly, the activity of recombinant proteins containing these types of deletions was indistinguishable from wild type, at least in the model systems utilized. The deletions may specifically affect multimerization of the collagen domain under in vivo conditions. Alternatively, the collagen domain may have additional functions, e.g.interaction with other proteins. It is well known for C1q that the collagen domain interacts with the serine proteases C1r and C1s to form C1, the first component of the serum complement system, and with a number of other proteins, including membrane-bound receptors (22Eggleton P. Reid K.B. Tenner A.J. Trends Cell Biol. 1998; 8: 428-431Abstract Full Text Full Text PDF PubMed Scopus (80) Google Scholar). Further investigations are required to understand the molecular mechanism underlying loss of function of EDA in these particular deletion mutants.15 families with XLHED displayed 5 distinct mutations in the furin consensus recognition sequences of EDA, demonstrating an important functional role for this 7-aa sequence. The release of soluble EDA upon proteolytic processing is an expected event, as mRNAs for EDA and EDAR are not expressed in adjacent cells but rather in spatially distinct tissues, at least in the developing tooth (10Tucker A.S. Headon D.J. Schneider P. Ferguson B.M. Overbeek P. Tschopp J. Sharpe P.T. Development. 2000; 127: 4691-4700PubMed Google Scholar). EDA contains two overlapping furin recognition sequences (RVRR and RNKR spanning aa 153–156 and 156–159, respectively). Because Arg156 is part of both sequences, it is hardly surprising that mutation R156C completely abolished EDA processing. Mutation R153C also affected the cleavage of EDA, but in a less dramatic manner. Although this mutation destroys only one of the two furin sites, this must be sufficiently disturbing to prevent efficient release of EDA from the cells naturally expressing it. These cells may express low amounts of furin or furin isoforms whose specificity may extend further than the canonical tetrapeptide recognition sequence. However, as mutations affecting the first furin domain invariably yield Cys residues, and as these Cys residues appear to form novel disulfide bridges, it is possible that this novel structural constraint prevents proper recognition of the remaining intact furin site. It also clearly appears from this study that, beside the furin cleavage sites, there are no alternative sites for solubilization of EDA. In particular, the sequence RRER (aa 69–72) and the basic motives KNKK and KGKK (aa 171–174 and 175–178) were not cleaved in our expression system. The latter two motifs are encoded in the small exon 4 and are also found in the sequences of Tweak and APRIL. However, mutations in these sequences have not been described so far in association with XLHED.A single splice site mutation was detected in the current study (IVS8 G+5 → A), which affects the splice donor site of exon 8 utilized to generate the 391-aa EDA-A1 isoform of the ligand. EDA-A2 utilizes an alternate splice site 6 base pairs 5′ to this site. This mutation probably interferes with splicing of EDA-A1 but not EDA-A2, as computer analysis by HSPL (prediction of splice sites in human DNA sequence) demonstrates a complete loss of the A1 but not of the A2 donor site. This together with the fact that genetic defects in EDA and in EDAR both lead to identical phenotypes indicate a crucial role for EDA-A1/EDAR interactions during morphogenesis. The role of the parallel EDA-A2/XEDAR interaction is less well established. If at all involved in hair, sweat gland, and teeth formation, it is not able to rescue a genetic deficiency in EDAR. In addition, there is no evidence to date for mutations in XEDAR being associated with the HED phenotype. XEDAR may play a distinct role in skin development, which does not translate into an HED phenotype upon dysfunction. Alternatively, inactivation of XEDAR might be lethal, but this would only be possible if it binds another ligand beside EDA-A2 or fulfills a vital, ligand-independent function. TROY/TAJ is a close sequence homologue of XEDAR, which is also expressed in the developing skin (23Kojima T. Morikawa Y. Copeland N.G. Gilbert D.J. Jenkins N.A. Senba E. Kitamura T. J. Biol. Chem. 2000; 275: 20742-20747Abstract Full Text Full Text PDF PubMed Scopus (102) Google Scholar, 24Eby M.T. Jasmin A. Kumar A. Sharma K. Chaudhary P.M. J. Biol. Chem. 2000; 275: 15336-15342Abstract Full Text Full Text PDF PubMed Scopus (102) Google Scholar). The precise functional roles of XEDAR and TROY and their interplay with the EDAR pathway remain to be determined. The ED1 gene encodes a protein, ectodysplasin-A (EDA),1 recently recognized to be a member of the tumor necrosis factor (TNF) superfamily of ligands. Mutations within the ED1 gene cause an X-linked recessive disorder, hypohidrotic or anhidrotic ectodermal dysplasia (ED1, XLHED) (Mendelian inheritance in man 305100), involving abnormal morphogenesis of teeth, hair, and eccrine sweat glands. Various splice forms of the ED1 transcript have been detected, but two isoforms differing only by two amino acids, EDA-A1 (391 aa) and EDA-A2 (389 aa), contain a TNF homology domain (1Bayes M. Hartung A.J. Ezer S. Pispa J. Thesleff I. Srivastava A.K. Kere J. Hum. Mol. Genet. 1998; 7: 1661-1669Crossref PubMed Scopus (181) Google Scholar, 2Monreal A.W. Zonana J. Ferguson B. Am. J. Hum. Genet. 1998; 63: 380-389Abstract Full Text Full Text PDF PubMed Scopus (185) Google Scholar, 3Yan M. Wang L.C. Hymowitz S.G. Schilbach S. Lee J. Goddard A. de Vos A.M. Gao W.Q. Dixit V.M. Science. 2000; 290: 523-527Crossref PubMed Scopus (236) Google Scholar). EDA is a type II transmembrane protein with a small N-terminal intracellular domain and a larger C-terminal extracellular domain containing a (Gly-X-Y)19 collagen-like repeat with a single interruption and a C-terminal TNF homology domain (Fig. 1 A). The TNF homology domain is similar to other members of the TNF family, consisting of 10 predicted anti-parallel β-sheets linked by variable loops (Fig. 1 A). TNF family ligands homotrimerize to form a pear-shaped quaternary structure able to bind a receptor molecule at each monomer-monomer interface (4Hymowitz S.G. O'Connell M.P. Ultsch M.H. Hurst A. Totpal K. Ashkenazi A. de Vos A.M. Kelley R.F. Biochemistry. 2000; 39: 633-640Crossref PubMed Scopus (228) Google Scholar, 5Banner D.W. D'Arcy A. Janes W. Gentz R. Schoenfeld H.J. Broger C. Loetscher H. Lesslauer W. Cell. 1993; 73: 431-445Abstract Full Text PDF PubMed Scopus (974) Google Scholar). The closest EDA homologues in the TNF family are BAFF/BLyS, APRIL, and TWEAK, although none of them contains collagen-like repeats (6Moore P.A. Belvedere O. Orr A. Pieri K. LaFleur D.W. Feng P. Soppet D. Charters M. Gentz R. Parmelee D. Li Y. Galperina O. Giri J. Roschke V. Nardelli B. Carrell J. Sosnovtseva S. Greenfield W. Ruben S.M. Olsen H.S. Fikes J. Hilbert D.M. Science. 1999; 285: 260-263Crossref PubMed Scopus (1006) Google Scholar, 7Schneider P. MacKay F. Steiner V. Hofmann K. Bodmer J.L. Holler N. Ambrose C. Lawton P. Bixler S. Acha-Orbea H. Valmori D. Romero P. Werner-Favre C. Zubler R.H. Browning J.L. Tschopp J. J. Exp. Med. 1999; 189: 1747-1756Crossref PubMed Scopus (1130) Google Scholar, 8Hahne M. Kataoka T. Schroter M. Hofmann K. Irmler M. Bodmer J.L. Schneider P. Bornand T. Holler N. French L.E. Sordat B. Rimoldi D. Tschopp J. J. Exp. Med. 1998; 188: 1185-1190Crossref PubMed Scopus (459) Google Scholar, 9Chicheportiche Y. Bourdon P.R. Xu H. Hsu Y.M. Scott H. Hession C. Garcia I. Browning J.L. J. Biol. Chem. 1997; 272: 32401-32410Abstract Full Text Full Text PDF PubMed Scopus (561) Google Scholar). All four ligands contain consensus sequences for proteolytic cleavage by furin within their extracellular domain. In the case of EDA, two overlapping consensus sites are located between the transmembrane and the collagen-like domains (Fig. 1 A). EDA-A1, but not EDA-A2, has been shown to specifically bind to EDAR, a member of the TNF receptor superfamily that, like most members of the TNF receptor family, activates the NF-κB and c-Jun N-terminal kinase pathways (3Yan M. Wang L.C. Hymowitz S.G. Schilbach S. Lee J. Goddard A. de Vos A.M. Gao W.Q. Dixit V.M. Science. 2000; 290: 523-527Crossref PubMed Scopus (236) Google Scholar, 10Tucker A.S. Headon D.J. Schneider P. Ferguson B.M. Overbeek P. Tschopp J. Sharpe P.T. Development. 2000; 127: 4691-4700PubMed Google Scholar). Mutations in DL (EDAR), the human homologue of the murine downless locus, produce an identical phenotype to loss of function of EDA (11Headon D.J. Overbeek P.A. Nat. Genet. 1999; 22: 370-374Crossref PubMed Scopus (308) Google Scholar, 12Monreal A.W. Ferguson B.M. Headon D.J. Street S.L. Overbeek P.A. Zonana J. Nat. Genet. 1999; 22: 366-369Crossref PubMed Scopus (323) Google Scholar). XEDAR, another member of the TNF receptor superfamily that also activates the NF-κB pathway, binds EDA-A2 but not EDA-A1. Although EDA-A1 and EDA-A2 are closely related splice variants, the respective proteins appear to have different patterns of expression in mouse skin and hair follicles (3Yan M. Wang L.C. Hymowitz S.G. Schilbach S. Lee J. Goddard A. de Vos A.M. Gao W.Q. Dixit V.M. Science. 2000; 290: 523-527Crossref PubMed Scopus (236) Google Scholar). Intracellular signals elicited by EDA in vivo rely at least in part on the activation of NF-κB, because a rare form of HED associated with immunodeficiency (HED-ID) correlates with mutations in NEMO/IKK-γ, an essential component of the NF-κB pathway (13Zonana J. Elder M.E. Schneider L.C. Orlow S.J. Moss C. Golabi M. Shapira S.K. Farndon P.A. Wara D.W. Emmal S.A. Ferguson B.M. Am. J. Hum. Genet. 2000; 67: 1555-1562Abstract Full Text Full Text PDF PubMed Scopus (419) Google Scholar). In order to get insight into the structure-function relationship of EDA, we identified 44 mutations (17 of which have not been reported previously) in unrelated families with XLHED and studied their effect on the properties of EDA in vitro. The mutations clustered in three functionally important domains as follows: a TNF homology domain necessary for receptor binding, a bundle-forming collagen domain, and a cleavage site for a furin protease. This indicates that the receptor binding ability of EDA and also its oligomerization and proteolytic processing to a soluble form are critical events for its action in vivo. DISCUSSIONIn this study, mutations in the EDA gene were detected in 63% of the families with XLHED, which is lower than the 95% rate we found previously (2Monreal A.W. Zonana J. Ferguson B. Am. J. Hum. Genet. 1998; 63: 380-389Abstract Full Text Full Text PDF PubMed Scopus (185) Google Scholar) by the direct sequencing of affected males. The lower detection rate is the consequence of two factors. The use of a single set of conditions lowers the sensitivity of SSCP analysis, as only 11 of the 15 known mutations run as controls could be detected under these conditions (73%). In addition, this study included 28% of families with "affected" females only, which was not the case in our previous study. Indeed, the detection rate was lower in families with female probands (45%), and this may well be due to genetic heterogeneity for autosomal forms of HED.A number of point mutations are located within the TNF homology domain of EDA, but only one of them (Y343C) affects a residue which, based on structural homology with known ligand-receptor structures (4Hymowitz S.G. O'Connell M.P. Ultsch M.H. Hurst A. Totpal K. Ashkenazi A. de Vos A.M. Kelley R.F. Biochemistry. 2000; 39: 633-640Crossref PubMed Scopus (228) Google Scholar, 5Banner D.W. D'Arcy A. Janes W. Gentz R. Schoenfeld H.J. Broger C. Loetscher H. Lesslauer W. Cell. 1993; 73: 431-445Abstract Full Text PDF PubMed Scopus (974) Google Scholar), is predicted to interact with the receptor. Indeed, this mutation abolished receptor binding without affecting the trimeric structure of EDA, although we cannot exclude indirect conformational effects. All other mutations are predicted to have indirect effect on receptor binding, e.g. by altering the folding of EDA. It is noteworthy that four independent mutations (G291W, G291R, A356D, R357P) occurred in two short loops at the bottom of EDA (loops BCand FG, see Fig. 1 A). The affected amino acids are probably crucial for proper folding of the monomer, as mutation A356D resulted in insoluble EDA-A1 and EDA-A2. Another group of mutations (H252L, S374R) seems to affect the stability of the trimer, because the resulting proteins contain a proportion of monomers. The propensity of unglycosylated subunits to form larger aggregates support the idea that glycosylation of some TNF family members promotes their solubility (18Schneider P. Bodmer J.L. Holler N. Mattmann C. Scuderi P. Terskikh A. Peitsch M.C. Tschopp J. J. Biol. Chem. 1997; 272: 18827-18833Abstract Full Text Full Text PDF PubMed Scopus (158) Google Scholar). Interestingly, one of the mutations (S374R) destroys a potential N-glycosylation site without affecting the glycosylation of EDA, suggesting that this particular site is not recognized by the N-glycosyltransferase. A single mutant (S374R) retained some binding activity to EDAR. Although preferential binding of the unglycosylated (and most probably aggregated) isoform was observed, the interaction appeared specific, suggesting that residual activity may be associated with this mutation in vivo. In general, no apparent phenotype/genotype correlation was observed, but the family with missense mutation S374R had two affected males and an affected maternal grandfather with isolated hypodontia. Whether residual activity of this mutant may account for the milder phenotype, and whether there is truly a tissue-specific difference in the function of this mutant protein remains to be determined. Finally, one mutation (T378M) affected secretion and aggregation of EDA-A1 and EDA-A2 in a strikingly different manner. The structural reason for this differential behavior is unclear. In summary, all missense mutations in the TNF homology domain result in abolished or much impaired binding of EDA-A1 to EDAR and EDA-A2 to XEDAR.A number of mutations occurring outside the TNF homology domain did not affect binding to the receptor in an in vitro assay, indicating that the interaction of EDA with its receptor(s) is necessary but not sufficient for its function in vivo. In particular, the integrity of the collagen domain appears to be functionally essential, and we provided evidence that it may serve to multimerize EDA trimers. This is in strong support of the hypothesis that EDA belongs to the C1q as well as to the TNF family of proteins (19Mikkola M.L. Pispa J. Pekkanen M. Paulin L. Nieminen P. Kere J. Thesleff I. Mech. Dev. 1999; 88: 133-146Crossref PubMed Scopus (106) Google Scholar). C1q family members are characterized by the presence of a C-terminal globular trimeric domain, with striking structural homology to TNF (20Shapiro L. Scherer P.E. Curr. Biol. 1998; 8: 335-338Abstract Full Text Full Text PDF PubMed Google Scholar), which is prolonged by a collagen triple helix further assembling into an N-terminal bundle structure, giving rise to a highly multimeric superstructure. In the TNF family, it has been shown that the activity of soluble trimers can be dramatically increased by antibody-mediated multimerization, thereby mimicking the membrane-bound form of the ligand (16Schneider P. Holler N. Bodmer J.L. Hahne M. Frei K. Fontana A. Tschopp J. J. Exp. Med. 1998; 187: 1205-1213Crossref PubMed Scopus (696) Google Scholar, 21Grell M. Douni E. Wajant H. Lohden M. Clauss M. Maxeiner B. Georgopoulos S. Lesslauer W. Kollias G. Pfizenmaier K. Cell. 1995; 83: 793-802Abstract Full Text PDF PubMed Scopus (1144) Google Scholar). A highly multimeric structure of EDA would provide a powerful means for the soluble protein to signal through high valency receptor clustering. In line with this hypothesis, we found that a naturally occurring point mutation (G207R) in the collagen domain completely abolished the bundle effect. It is, however, likely that the collagen domain also serves additional functions; a number of families with XLHED displayed in frame deletion of 2 or 4 GlyXY repeats in the predicted bundle domain of the collagen triple helix, i.e. before the interruption in the GlyXY repeats. Deletions of 2 GlyXY repeats can also be found C-terminal to the interruption. Surprisingly, the activity of recombinant proteins containing these types of deletions was indistinguishable from wild type, at least in the model systems utilized. The deletions may specifically affect multimerization of the collagen domain under in vivo conditions. Alternatively, the collagen domain may have additional functions, e.g.interaction with other proteins. It is well known for C1q that the collagen domain interacts with the serine proteases C1r and C1s to form C1, the first component of the serum complement system, and with a number of other proteins, including membrane-bound receptors (22Eggleton P. Reid K.B. Tenner A.J. Trends Cell Biol. 1998; 8: 428-431Abstract Full Text Full Text PDF PubMed Scopus (80) Google Scholar). Further investigations are required to understand the molecular mechanism underlying loss of function of EDA in these particular deletion mutants.15 families with XLHED displayed 5 distinct mutations in the furin consensus recognition sequences of EDA, demonstrating an important functional role for this 7-aa sequence. The release of soluble EDA upon proteolytic processing is an expected event, as mRNAs for EDA and EDAR are not expressed in adjacent cells but rather in spatially distinct tissues, at least in the developing tooth (10Tucker A.S. Headon D.J. Schneider P. Ferguson B.M. Overbeek P. Tschopp J. Sharpe P.T. Development. 2000; 127: 4691-4700PubMed Google Scholar). EDA contains two overlapping furin recognition sequences (RVRR and RNKR spanning aa 153–156 and 156–159, respectively). Because Arg156 is part of both sequences, it is hardly surprising that mutation R156C completely abolished EDA processing. Mutation R153C also affected the cleavage of EDA, but in a less dramatic manner. Although this mutation destroys only one of the two furin sites, this must be sufficiently disturbing to prevent efficient release of EDA from the cells naturally expressing it. These cells may express low amounts of furin or furin isoforms whose specificity may extend further than the canonical tetrapeptide recognition sequence. However, as mutations affecting the first furin domain invariably yield Cys residues, and as these Cys residues appear to form novel disulfide bridges, it is possible that this novel structural constraint prevents proper recognition of the remaining intact furin site. It also clearly appears from this study that, beside the furin cleavage sites, there are no alternative sites for solubilization of EDA. In particular, the sequence RRER (aa 69–72) and the basic motives KNKK and KGKK (aa 171–174 and 175–178) were not cleaved in our expression system. The latter two motifs are encoded in the small exon 4 and are also found in the sequences of Tweak and APRIL. However, mutations in these sequences have not been described so far in association with XLHED.A single splice site mutation was detected in the current study (IVS8 G+5 → A), which affects the splice donor site of exon 8 utilized to generate the 391-aa EDA-A1 isoform of the ligand. EDA-A2 utilizes an alternate splice site 6 base pairs 5′ to this site. This mutation probably interferes with splicing of EDA-A1 but not EDA-A2, as computer analysis by HSPL (prediction of splice sites in human DNA sequence) demonstrates a complete loss of the A1 but not of the A2 donor site. This together with the fact that genetic defects in EDA and in EDAR both lead to identical phenotypes indicate a crucial role for EDA-A1/EDAR interactions during morphogenesis. The role of the parallel EDA-A2/XEDAR interaction is less well established. If at all involved in hair, sweat gland, and teeth formation, it is not able to rescue a genetic deficiency in EDAR. In addition, there is no evidence to date for mutations in XEDAR being associated with the HED phenotype. XEDAR may play a distinct role in skin development, which does not translate into an HED phenotype upon dysfunction. Alternatively, inactivation of XEDAR might be lethal, but this would only be possible if it binds another ligand beside EDA-A2 or fulfills a vital, ligand-independent function. TROY/TAJ is a close sequence homologue of XEDAR, which is also expressed in the developing skin (23Kojima T. Morikawa Y. Copeland N.G. Gilbert D.J. Jenkins N.A. Senba E. Kitamura T. J. Biol. Chem. 2000; 275: 20742-20747Abstract Full Text Full Text PDF PubMed Scopus (102) Google Scholar, 24Eby M.T. Jasmin A. Kumar A. Sharma K. Chaudhary P.M. J. Biol. Chem. 2000; 275: 15336-15342Abstract Full Text Full Text PDF PubMed Scopus (102) Google Scholar). The precise functional roles of XEDAR and TROY and their interplay with the EDAR pathway remain to be determined. In this study, mutations in the EDA gene were detected in 63% of the families with XLHED, which is lower than the 95% rate we found previously (2Monreal A.W. Zonana J. Ferguson B. Am. J. Hum. Genet. 1998; 63: 380-389Abstract Full Text Full Text PDF PubMed Scopus (185) Google Scholar) by the direct sequencing of affected males. The lower detection rate is the consequence of two factors. The use of a single set of conditions lowers the sensitivity of SSCP analysis, as only 11 of the 15 known mutations run as controls could be detected under these conditions (73%). In addition, this study included 28% of families with "affected" females only, which was not the case in our previous study. Indeed, the detection rate was lower in families with female probands (45%), and this may well be due to genetic heterogeneity for autosomal forms of HED. A number of point mutations are located within the TNF homology domain of EDA, but only one of them (Y343C) affects a residue which, based on structural homology with known ligand-receptor structures (4Hymowitz S.G. O'Connell M.P. Ultsch M.H. Hurst A. Totpal K. Ashkenazi A. de Vos A.M. Kelley R.F. Biochemistry. 2000; 39: 633-640Crossref PubMed Scopus (228) Google Scholar, 5Banner D.W. D'Arcy A. Janes W. Gentz R. Schoenfeld H.J. Broger C. Loetscher H. Lesslauer W. Cell. 1993; 73: 431-445Abstract Full Text PDF PubMed Scopus (974) Google Scholar), is predicted to interact with the receptor. Indeed, this mutation abolished receptor binding without affecting the trimeric structure of EDA, although we cannot exclude indirect conformational effects. All other mutations are predicted to have indirect effect on receptor binding, e.g. by altering the folding of EDA. It is noteworthy that four independent mutations (G291W, G291R, A356D, R357P) occurred in two short loops at the bottom of EDA (loops BCand FG, see Fig. 1 A). The affected amino acids are probably crucial for proper folding of the monomer, as mutation A356D resulted in insoluble EDA-A1 and EDA-A2. Another group of mutations (H252L, S374R) seems to affect the stability of the trimer, because the resulting proteins contain a proportion of monomers. The propensity of unglycosylated subunits to form larger aggregates support the idea that glycosylation of some TNF family members promotes their solubility (18Schneider P. Bodmer J.L. Holler N. Mattmann C. Scuderi P. Terskikh A. Peitsch M.C. Tschopp J. J. Biol. Chem. 1997; 272: 18827-18833Abstract Full Text Full Text PDF PubMed Scopus (158) Google Scholar). Interestingly, one of the mutations (S374R) destroys a potential N-glycosylation site without affecting the glycosylation of EDA, suggesting that this particular site is not recognized by the N-glycosyltransferase. A single mutant (S374R) retained some binding activity to EDAR. Although preferential binding of the unglycosylated (and most probably aggregated) isoform was observed, the interaction appeared specific, suggesting that residual activity may be associated with this mutation in vivo. In general, no apparent phenotype/genotype correlation was observed, but the family with missense mutation S374R had two affected males and an affected maternal grandfather with isolated hypodontia. Whether residual activity of this mutant may account for the milder phenotype, and whether there is truly a tissue-specific difference in the function of this mutant protein remains to be determined. Finally, one mutation (T378M) affected secretion and aggregation of EDA-A1 and EDA-A2 in a strikingly different manner. The structural reason for this differential behavior is unclear. In summary, all missense mutations in the TNF homology domain result in abolished or much impaired binding of EDA-A1 to EDAR and EDA-A2 to XEDAR. A number of mutations occurring outside the TNF homology domain did not affect binding to the receptor in an in vitro assay, indicating that the interaction of EDA with its receptor(s) is necessary but not sufficient for its function in vivo. In particular, the integrity of the collagen domain appears to be functionally essential, and we provided evidence that it may serve to multimerize EDA trimers. This is in strong support of the hypothesis that EDA belongs to the C1q as well as to the TNF family of proteins (19Mikkola M.L. Pispa J. Pekkanen M. Paulin L. Nieminen P. Kere J. Thesleff I. Mech. Dev. 1999; 88: 133-146Crossref PubMed Scopus (106) Google Scholar). C1q family members are characterized by the presence of a C-terminal globular trimeric domain, with striking structural homology to TNF (20Shapiro L. Scherer P.E. Curr. Biol. 1998; 8: 335-338Abstract Full Text Full Text PDF PubMed Google Scholar), which is prolonged by a collagen triple helix further assembling into an N-terminal bundle structure, giving rise to a highly multimeric superstructure. In the TNF family, it has been shown that the activity of soluble trimers can be dramatically increased by antibody-mediated multimerization, thereby mimicking the membrane-bound form of the ligand (16Schneider P. Holler N. Bodmer J.L. Hahne M. Frei K. Fontana A. Tschopp J. J. Exp. Med. 1998; 187: 1205-1213Crossref PubMed Scopus (696) Google Scholar, 21Grell M. Douni E. Wajant H. Lohden M. Clauss M. Maxeiner B. Georgopoulos S. Lesslauer W. Kollias G. Pfizenmaier K. Cell. 1995; 83: 793-802Abstract Full Text PDF PubMed Scopus (1144) Google Scholar). A highly multimeric structure of EDA would provide a powerful means for the soluble protein to signal through high valency receptor clustering. In line with this hypothesis, we found that a naturally occurring point mutation (G207R) in the collagen domain completely abolished the bundle effect. It is, however, likely that the collagen domain also serves additional functions; a number of families with XLHED displayed in frame deletion of 2 or 4 GlyXY repeats in the predicted bundle domain of the collagen triple helix, i.e. before the interruption in the GlyXY repeats. Deletions of 2 GlyXY repeats can also be found C-terminal to the interruption. Surprisingly, the activity of recombinant proteins containing these types of deletions was indistinguishable from wild type, at least in the model systems utilized. The deletions may specifically affect multimerization of the collagen domain under in vivo conditions. Alternatively, the collagen domain may have additional functions, e.g.interaction with other proteins. It is well known for C1q that the collagen domain interacts with the serine proteases C1r and C1s to form C1, the first component of the serum complement system, and with a number of other proteins, including membrane-bound receptors (22Eggleton P. Reid K.B. Tenner A.J. Trends Cell Biol. 1998; 8: 428-431Abstract Full Text Full Text PDF PubMed Scopus (80) Google Scholar). Further investigations are required to understand the molecular mechanism underlying loss of function of EDA in these particular deletion mutants. 15 families with XLHED displayed 5 distinct mutations in the furin consensus recognition sequences of EDA, demonstrating an important functional role for this 7-aa sequence. The release of soluble EDA upon proteolytic processing is an expected event, as mRNAs for EDA and EDAR are not expressed in adjacent cells but rather in spatially distinct tissues, at least in the developing tooth (10Tucker A.S. Headon D.J. Schneider P. Ferguson B.M. Overbeek P. Tschopp J. Sharpe P.T. Development. 2000; 127: 4691-4700PubMed Google Scholar). EDA contains two overlapping furin recognition sequences (RVRR and RNKR spanning aa 153–156 and 156–159, respectively). Because Arg156 is part of both sequences, it is hardly surprising that mutation R156C completely abolished EDA processing. Mutation R153C also affected the cleavage of EDA, but in a less dramatic manner. Although this mutation destroys only one of the two furin sites, this must be sufficiently disturbing to prevent efficient release of EDA from the cells naturally expressing it. These cells may express low amounts of furin or furin isoforms whose specificity may extend further than the canonical tetrapeptide recognition sequence. However, as mutations affecting the first furin domain invariably yield Cys residues, and as these Cys residues appear to form novel disulfide bridges, it is possible that this novel structural constraint prevents proper recognition of the remaining intact furin site. It also clearly appears from this study that, beside the furin cleavage sites, there are no alternative sites for solubilization of EDA. In particular, the sequence RRER (aa 69–72) and the basic motives KNKK and KGKK (aa 171–174 and 175–178) were not cleaved in our expression system. The latter two motifs are encoded in the small exon 4 and are also found in the sequences of Tweak and APRIL. However, mutations in these sequences have not been described so far in association with XLHED. A single splice site mutation was detected in the current study (IVS8 G+5 → A), which affects the splice donor site of exon 8 utilized to generate the 391-aa EDA-A1 isoform of the ligand. EDA-A2 utilizes an alternate splice site 6 base pairs 5′ to this site. This mutation probably interferes with splicing of EDA-A1 but not EDA-A2, as computer analysis by HSPL (prediction of splice sites in human DNA sequence) demonstrates a complete loss of the A1 but not of the A2 donor site. This together with the fact that genetic defects in EDA and in EDAR both lead to identical phenotypes indicate a crucial role for EDA-A1/EDAR interactions during morphogenesis. The role of the parallel EDA-A2/XEDAR interaction is less well established. If at all involved in hair, sweat gland, and teeth formation, it is not able to rescue a genetic deficiency in EDAR. In addition, there is no evidence to date for mutations in XEDAR being associated with the HED phenotype. XEDAR may play a distinct role in skin development, which does not translate into an HED phenotype upon dysfunction. Alternatively, inactivation of XEDAR might be lethal, but this would only be possible if it binds another ligand beside EDA-A2 or fulfills a vital, ligand-independent function. TROY/TAJ is a close sequence homologue of XEDAR, which is also expressed in the developing skin (23Kojima T. Morikawa Y. Copeland N.G. Gilbert D.J. Jenkins N.A. Senba E. Kitamura T. J. Biol. Chem. 2000; 275: 20742-20747Abstract Full Text Full Text PDF PubMed Scopus (102) Google Scholar, 24Eby M.T. Jasmin A. Kumar A. Sharma K. Chaudhary P.M. J. Biol. Chem. 2000; 275: 15336-15342Abstract Full Text Full Text PDF PubMed Scopus (102) Google Scholar). The precise functional roles of XEDAR and TROY and their interplay with the EDAR pathway remain to be determined. We thank the families who participated in these studies and the many clinicians who provided both samples and clinical information. We thank Dr. Teresa Cachero, Dr. Jeffrey Browning, and Dr. Matvey Lukashev (Biogen Inc., Cambridge, MA) for their help with protein sequencing and for the identification of the XEDAR cDNA clone. We are grateful to Dr. Steven R. Wiley (Immunex Corp., Seattle, WA) for the authorization to show the Fn14-TWEAK interaction prior to publication.
Referência(s)