|Home | About | Journals | Submit | Contact Us | Français|
The same morphogenetic signals are often involved in the development of different organs. For developing skin appendages, a model for tissue-specific regulation of signaling is provided by the EDA pathway, which accesses the otherwise ubiquitous NFκB transcription factors. EDA signaling is mediated by ectodysplasin, EDAR and EDARADD, which form a new TNF ligand-receptor-adaptor family that is restricted to skin appendages in vertebrates from fish to human. The critical function of the pathway was demonstrated in the hereditary genetic disorder Anhidrotic Ectodermal Dysplasia (EDA), which is characterized by defective formation of hair follicles, sweat glands and teeth. The pathway does not appear to initiate the development of the appendages, but is regulated by and regulates the course of further morphogenesis. In mice, transgenic and knockout strains have increasingly revealed features of the mechanism, and suggest possible non-invasive interventions to alleviate EDA deficiency, especially in sweat glands and eyes.
Skin appendages include hair follicles, nails, teeth and exocrine glands in mammals, feathers in birds, and scales in fish. Their exquisite adaptations for diverse physiological functions have attracted scientists to investigate the differential morphogenetic mechanisms. Follicle transplantation experiments, gene engineering, and genome-wide transcription profiling of skin and skin appendages have progressively discerned molecular mechanisms involved in skin appendage development.1,2 General signaling pathways such as Wnt, Shh, BMP and Notch all participate; the EDA signaling pathway rather modulates more general signals to promote the formation of specific nascent appendages.3–5
Anhidrotic/hypohidrotic ectodermal dysplasia (EDA) is the most frequent of a group of ectodermal dysplasias (EDs) that affect morphogenesis of two or more skin appendages (OMIM 34500). EDA affects development of hair, teeth and several exocrine glands, including sweat, meibomian and preputial glands.6
As a hereditary genetic disorder, EDA occurs in X-linked recessive, autosomal dominant, and autosomal recessive forms.7–9 Notably, the X-linked form was first described by Charles Darwin in 1875 in his book, “The variations of animals and plants under domestication”, in which he recorded precisely the skin appendage phenotypes and male-specific inheritance pattern of an affected Indian family.10 More than 100 years later, three genes mutated in various cases have been identified from patients. The X-linked EDA gene is mutated in more than 90% of cases, and EDAR and EDARADD genes account for most of the rarer instances of autosomal inheritances.7–11 Unanticipated were the findings that the EDA, EDAR and EDARADD genes encode a new TNF ligand-receptor-adaptor family - the first to be involved in skin appendage development (Fig. 1 and see below).8,9,12
It has turned out that mouse models for EDA have been available since early in the last century. Isolated as spontaneous mutant strains, Tabby, Downless/Sleek and Crinkled have been demonstrated to be mutated in the corresponding Eda, Edar, and Edaradd genes.8,9,13 Confirming their positioning in the same signaling pathway, the mutant strains show identical “all or none” skin appendage phenotypes. Sweat glands and two of the three types of mouse hair, “Guard” and “Zigzag”, are lacking; but a normal number of hair follicles is still formed, all of them producing an abnormal variant of the “Medium”, third type of hair.14,15 In addition, aberrant features of EDA in patients and a mouse model were reported recently in the ectodermally-derived cornea and conjunctiva.16,17 Thus patients and animal models have revealed the first actors in an EDA pathway, providing a novel system to study the development of ectoderm-derived tissues.
As comparative genomics and genetics has progressed, the EDA, EDAR and EDARADD genes initially discovered from patients and mouse models have been found to be highly conserved in other vertebrates, including fish, xenopus, chicken, dogs, cattle, and chimpanzee (e.g., EDAR in Fig. 2). Corresponding spontaneous mutant animal models for EDA pathway genes have thus far been identified in four species: Edar mutant medaka fish; Eda mutant dogs and cattle; and the Tabby, Downless/Sleek and Crinkled mouse mutants.8,9,13,18–20 The skin appendages are often species-specific, so that in contrast to the effects on hair and sweat glands in mammals, fish show a loss of scales.18 In a recent study, the EDA gene was shown to be involved in adaptive evolution of armor plate patterning in three-spine stickleback fishes during the transition from marine to freshwater environment.21 A mouse Eda-A1 isoform restored marine-like armor plates to freshwater sticklebacks, a further indication of the adaptability of the gene to parallel evolution of skin appendages.21,22
Understanding EDA function in skin appendage development requires information about how EDA is regulated and how it regulates downstream targets.
We have shown that a 1.6kB genomic fragment upstream of EDA transcription start site includes most of the promoter and enhancer binding sites.23 By an extension of the approach, we extended the fragment to 5 kB and 15 kB upstream to study candidate promoter fragments, but transcription activity was unchanged, suggesting promoter and enhancer sites are concentrated in the small region (data not shown). EDA lacks basic TATA box, but has two active SP1 sites. Unlike other TATA-free, SP1-positive genes, however, EDA has a single transcription initiation site.23
A binding site for the Wnt pathway transcription factor Lef1 was detected in the 1.6 kB fragment.7,23 We demonstrated that the site is responsible for full activation of EDA transcription in cell transfection experiments (Fig. 1).24 Consistent with these results, Eda was not expressed in Lef1-null mouse skin, and Wnt6 induced Eda expression in a skin culture system.25 Though Wnt/Lef1 is probably not the sole regulator of EDA expression, because unlike Lef1, which shows a patterned focal expression in hair follicles, EDA is broadly transcribed in skin and hair follicles.25
Wnt signaling is thought to initiate hair follicle formation.26 EDA signaling would then occur directly downstream of the induction signal. Unlike Wnt pathway genes, over expression of Eda-A1 did not increase the number of hair follicles in mice, and the loss of Eda left the total number of hair follicles the same.15,26 Rather, its major effect was to change the types of hair produced from the nascent follicles (see below).
Two active candidate enhancer-binding sites, for GATA and Nkx2, were noted in the 1.6kB putative promoter fragment of the EDA gene (ref. 23 and our unpublished data). GATA and Nkx2 family members are known to interact with SP1 to effect tissue-specific and developmentally restricted expression of target genes, however, their role in EDA expression is still conjectural (Fig. 1). In fact, there is no explanation of the paradox that in spite of strong promoter and enhancer activity in vitro, EDA mRNA expression level in vivo is quite low.7,13,21 The full spectrum of regulatory sequences and protein factors affecting EDA transcription thus remain to be elucidated.
Ectodysplasin, encoded by the X-linked EDA gene, is a type II transmembrane protein.12 Ectodysplasin has a short collagen segment and a TNF domain in its extracellular region.12 A furin cleavage site near the collagen domain releases soluble EDA (TNF) ligands that then form active ligand-trimers comparable to those seen for other TNF ligand family members (Fig. 1).27,28 Missense or in-frame deletion mutations that cause the EDA syndrome are found in TNF, collagen, furin cleavage, and transmembrane sites.29,30
The TNF domain is the functional moiety of ectodysplasin, but mutations in patients indicate that the collagen domain also has an important though unknown function. The domain comprises 19 G-X-Y repeats in two neighboring stretches.13 In-frame deletion of 2 or 4 G-X-Y repeats has no effect on the functional TNF domain but causes EDA syndrome.29,30 One might expect that the domain could contribute to formation or stabilization of a ligand-trimer or multimer, as seen for complement protein C1q, which has 27 N-terminal G-X-Y repeats.27,31 However, the collagen domain in ectodysplasin lacks lysine, a critical cross-linking residue in standard collagen triple helices, including those in C1q-like proteins.32 Furthermore, the crystal structure of ectodysplasin suggests that the collagen domain is dispensable for ligand-trimer formation.33 Therefore the function of this domain remains an interesting puzzle.
That soluble EDA trimers initiate EDA signaling by binding to the receptor EDAR is well established.34 EDA-A1 was demonstrated to bind specifically to EDAR, and no affinity was detected for other TNF family members.34 EDAR has extracellular TNFR domains and an intracellular death domain (DD) region.8 The other DD-containing TNFR members, TNFR1, NGFR, DR3, 4, 5, 6 and Fas are all able to induce cell death, but there are contradictory reports about the ability of EDAR to promote cell death.35–38
Its DD region is responsible for EDAR binding to its cytoplasmic signaling adaptor EDARADD.9 EDARADD is one of three DD-containing adaptor proteins identified in the TNF superfamily, and has high homology to the DD of Myd88, a cytoplasmic adaptor protein in the Toll/IL-1 signaling pathway.9 Like Myd88, EDARADD binds to the TRAF6/TAK1/TAB2 complex to activate the IKK complex (Fig. 1 and see below).39 Indeed, Traf6-deficient mice showed characteristic Tabby phenotypes along with immunodeficiency, suggesting that it is a critical mediator of EDA signaling.40
The EDA gene produces many splicing variants.41 Recently we recovered a total of 9 Eda-A isoforms from mouse keratinocytes.42 As predicted, A1 and A2 isoforms predominate, comprising about 80% of the total.42 The A2 isoform is only 2 amino acids shorter than A1, but binds to a different receptor, XEDAR, which also can activate NFκB mediated transcription in vitro.34,43 We also found two new isoforms, A5 and A5′, which bind to EDAR and XEDAR, respectively, and activate NFκBs in vitro.42
Especially because the Eda-A1 isoform does not fully rescue Tabby phenotypes (see below), involvement of alternative EDA signaling in the process might be anticipated, perhaps employing different isoforms and variant signaling cascades.15,44 However, signaling mediated by XEDAR seems dispensable for skin appendage development: Xedar knockout mice showed no skin appendage phenotypes and Eda-A2 transgenic mice showed no correction of Tabby phenotypes; and in contrast to the reported activity in vitro, a recent study suggests that EDA-A2/XEDAR pathway does not activate NFκBs in vivo.15,45,46
Another EDAR family member, Troy, is also highly expressed in skin appendages.46,47 To understand Troy function, we generated a mutant mouse strain by ENU mutagenesis (unpublished results).48 In the mice, in a C3H background, a missense mutation in the Troy coding region replaces a threonine with proline in TNFR domain 2 (at amino acid 109). We confirmed in cell transfection experiments that this mutation abolishes Troy function in NF-kB activation; but in the homozygous mutant mice hair follicles, hair shafts, hair type composition and sweat glands were all normal (work in progress).
Unlike EDAR, XEDAR and Troy both lack a death domain and are 55% homologous in their TNFR domain (and 100% homologous in cysteine residue content and placement). That is remarkably high for TNFR family members, and hints at a possible close functional relation between the two receptors.34,47 For example, the two might represent a redundant function, and it might then be informative to generate a mouse deficient in both.
As described above, the ectodysplasin-EDAR-EDARADD complex interacts with TRAF6 and thereby activates the IKK complex. The IKK complex contains two catalytic subunits, IKK1 and IKK2, and a regulatory subunit, NEMO (Fig. 1). The complex phosphorylates the downstream IkB complex, resulting in NFκB movement into the nucleus to activate transcription of responsive genes.49 Null mutations of the X-linked gene NEMO cause Incontinentia Pigmenti (IP2), a disorder in which males die in utero and heterozygous females show inflammatory and skin pigment changes in human patients and a mouse model.50,51 By contrast, less severe mutations in NEMO retain some NFκB activity and show characteristic EDA phenotypes along with immunodeficiency (EDA-ID).52,53 Thus, though the mechanism of pathological changes in these skin diseases remains largely unknown, NEMO is the mediator of EDA signaling to downstream targets (Fig. 1).50
EDA signaling is also likely mediated in part by IkBα (Fig. 1).54 EDA phenotypes along with immunodeficiency were observed in one patient with a mutant in the gene and in a mouse strain in which IkBα truncated at its N-terminus was inserted into a β-catenin gene.54,55 The mutant forms acted as a ‘super-repressor’ of NFκBs: they were no longer phosphorylatable by the IKK complex, thereby escaping subsequent degradation and maintaining NFκBs in the cytoplasm.54,55
NFκBs include 5 structurally related, evolutionarily conserved members—RelA (p65), p50, c-Rel, RelB and p52. Each can form either homo- or heterodimers, but with somewhat different properties.56 For example, RelA, c-Rel and RelB have a transcription regulatory domain, but p50 and p52 do not.56 Also, RelA, p50, c-Rel and RelB have been shown to regulate keratinocyte growth and differentiation, and to be involved in the pathogenesis of skin disorders including psoriasis, skin cancer, and inflammatory skin diseases.57 Mice deficient in RelA, c-Rel and TNFα lack Guard hair, a primary characteristic of Tabby animals.58 And in cell lines, EDA signaling up-regulates the expression of RelA/p50 but not RelB.46,52,59
Thus, genetic studies demonstrate the primacy of the ectodysplasin-EDAR-EDARADD-TRAF6-NEMO-IkBα-NFκB (RelA) signaling pathway in skin appendage development (Fig. 1), apparently working through the NEMO/IkBα-dependent canonical NFκB cascade.60 Any factors that affect NEMO or RelA activity in skin may result in Tabby-like phenotypes, as shown recently in mice transgenic for a glucocorticoid receptor (GR) under keratin 5 (K5) promoter control.61
The immediate downstream effectors TRAF6, NEMO, IkBα and NFκBs have the general roles usually associated with TNF-NFκB family members in immune or inflammatory processes.62 It is therefore not surprising that mutation in any of them causes severe immunodeficiency along with EDA phenotypes in mice and in human patients.40,52–55 Although EDA is widely expressed in skin, its signaling becomes specific for skin appendage development because EDAR and EDARADD are only expressed in skin appendages.8,9 Thus, EDA can co-opt the function of immune pathway genes through EDARADD-TRAF6 interaction only in skin and at the time of appendage formation, and can thus act without accompanying immunodeficiency.39
General morphogenetic signaling pathways such as Wnt, BMP and Shh also play indispensable roles in skin appendage development.3–5 Their roles are, however, “double-edged”, because failure of fine regulation of the signals often results in skin tumorigenesis or “extreme phenotypes” such as embryonic lethality or excessive formation of skin appendages.26,63–65 In contrast, EDA signaling is specific for skin appendage development, and overexpression or null mutations of the pathway genes only seem to modulate morphogenetic signals that have been otherwise turned on during development of the appendages.15,44,59
Our group and others have reported that Wnt, Shh, BMP and lymphotoxin-β (LTβ) pathways are located downstream of EDA signaling during Guard hair development (Fig. 1).8,25,59,66 EDA signaling was shown to positively activate Shh and LTβ pathways, but exerts a negative feedback regulation on Wnt and BMP pathways through pathway-specific antagonists.59 In particular, Wnt up-regulates EDA expression, but in turn, EDA signaling activates transcription of Wnt antagonists Dkk4 and Dkk1.24,25,59,67 The possible feedback regulatory action of EDA may provide a balance, helping to ensure proper function of the general morphogenetic signals while avoiding their potential “extreme effects” (Fig. 1).
Several groups have reported that expression of Shh pathway genes is most strikingly affected in Eda-null Tabby mice at various developmental stages, suggesting that this pathway is a proximal target of EDA signaling.8,25,46,59 However, Shh-null mice (and mice ablated for the downstream Shh effector Gli2) arrest hair follicle development only at E15.5, well after Guard hair peg formation; this is a phenotype strikingly different from mice deficient in other EDA pathway genes; and overexpression of Shh causes basal cell carcinoma-like skin tumors rather than augmented hair follicle formation.63,68,69 Therefore, although Shh expression was affected from very early embryonic stages in Tabby mice, its role in EDA signaling is still unclear.59
Lymphotoxin-β is a newly identified EDA target. It had previously been implicated only in immune system development, where LTβ initiates the noncanonical NF-kB pathway for lymphoid organogenesis in cooperation with LTα and LTβR.60,70 For hair follicle formation, however, mice defective in LTβ indicated that it primarily regulated hair shaft formation (again without effect on hair follicle induction); and unexpectedly, LTα and LTβR deficient mice showed normal hair follicles and hair shafts.59 Based on these findings, LTβ may have an unknown receptor active in hair follicle development.
An adhesion molecule, Madcam1 (mucosal addressin cell adhesion molecule 1) has also been suggested to be regulated by EDA signaling (Fig. 1).71 Madcam1 was transiently expressed in Guard hair germs but not in secondary hair follicles, and no expression was seen in Tabby, Downless or Traf6-null mice. Madcam1 has also been inferred as a target of lymphotoxin signaling in the immune system, raising the intriguing possibility that they cooperate in hair follicle development under EDA regulation.72,73
Overall, EDA signaling pathway is thus regulated by general morphogenetic signaling, but in turn regulates these pathways spatio-temporally for skin appendage formation, participating in later stages of specific organ development. A recent study found some premature Guard hair germs in the EDAR-deficient Downless mice at E14.5, again consistent with a role for EDA signaling only after the primary initiation signal.46
Though the EDA molecular pathway is increasingly characterized, its mechanism of action in skin appendage development is only partially known. As one approach to study the role of the gene in further detail, six Tabby mouse derivatives have been studied that are transgenic for EDA ligands under various promoters: CMV, conditional Tet (MMTV), K14 and Involucrin promoters for full length Eda-A1 or A2 transgenes; and K5 and myosin light-chain 2 (MLCH) promoters for the soluble part of EDA-A1 or A2.15,22,45,65,74 The results of studies with various models are concordant. In Tabby mice an A1 transgene restored Guard hair and sweat glands, but not Zigzag hair; whereas in wild-type mice the transgene suppressed Zigzag hair formation.15,22,65 The number of total back hairs was the same in Tabby mice and the transgenic strain, but hair development shifted.15 Guard hair number was restored to wild-type levels in transgenic mice, but some of the abnormal Tabby Awl hairs became structurally normal. This suggests that EDA-A1 is responsible for Guard hair formation and also participates in the patterning of nascent Awl hair.
In addition, the Eda-A1 transgene induced hypertrophic sebaceous glands, with both increased sebocyte numbers and size (but no indication of tumorigenesis).15,45,65 Because the Eda deficient Tabby mice still have sebaceous glands, this effect of EDA is inferred to be trophic rather than morphogenetic.15 This is the first indication that EDA can also act in selected targeted proliferative processes.
Notably, a recombinant EDA-A1 ligand on the IgG1 Fc tail showed restoration of skin appendages identical to the transgene when injected into pregnant Tabby mice intravenously—or for sweat glands, in the perinatal period.44 These results suggest a potential for therapeutic use (see below).
Three strains of Edar transgenic mice have also been studied, containing genomic Edar driven by the Edar promoter or a full-length or intracellular segment of Edar formed under K14 promoter control.75–77 A genomic clone containing the gene fully restored Downless phenotypes, whereas the K14-driven full-length Edar unexpectedly suppressed Guard hair formation and the intracellular part of Edar led to disoriented hair follicles in slightly increased numbers.75–77 Apparently, EDAR expression must be finely regulated for normal hair follicle formation.
Fundamental studies must (1) further analyze the initiation of the EDA pathway; (2) determine the downstream mechanism of action; and (3) assess possible therapeutic applications.
The initial regulatory steps in the EDA signaling pathway are still not fully understood. EDAR and EDARADD provide specificity for EDA signaling by restricting its expression to skin appendages. It was reported that recombinant activin A upregulates Edar expression in a skin culture system, but the factors that render EDAR expression tissue-specific have not been studied in detail.25 Furthermore, especially because Eda-A1 transgenes do not rescue all of the EDA phenotype, it remains possible that other EDA isoforms and receptors (including the large amounts of the EDA-A2 isoform and its specific receptor XEDAR, and perhaps the XEDAR-like Troy receptor) participate as well. Although knockout mouse models show dispensability of each of them for skin appendage formation, one can imagine, for example, that Troy and XEDAR have redundant actions in skin appendage formation and must both be disabled.
Downstream of the initiation of the pathway, the fundamental result remains that Eda-deficient mice still can form Medium-type hair. Only the first clues have been found to understand how EDA then participates in the generation of other hair types, and is dispensable for the formation of Medium hair. One extreme possibility is that different types of hair are generated by fundamentally different mechanisms, with EDA required only for Guard and Zigzag hair.8 Based on current data, Medium hair might, for example, be dependent only on Wnt/Lef1 and BMP/noggin.8,78–81 However, direct studies of Guard hair formation show EDA signaling cooperating with Wnt, Shh and BMP pathways from the earliest stages.24,25,59 Furthermore, in humans all hair seems to be dependent on EDA; and in both mouse and man, sweat glands are completely EDA-dependent. In wild-type mice bearing an Eda-A1 transgene and to some extent in LTβ-null mice, Zigzag hairs lost their kinks and became Awl-like.15,59 This suggests that further work will detail a single molecular mechanism for EDA, and that Medium hairs are formed by an unusual default mechanism—possibly based on a downstream feature of the EDA pathway (such as increased endogenous activation of EDAR or an NFκB target). Consistent with this notion, NF-kB is active in Medium hair germs in Tabby and Downless mice.46 EDA pathway has also been suggested to have a comparable role in hair follicle cycling, which recapitulates the morphogenetic process.82
Finally, what are the prospects for eventual effective intervention in the EDA syndrome (or other ectodermal dysplasias)? The Tabby mouse should continue to provide critical tests of principle. Neither EDA transgenes nor injected ligands lead to any findings of lethality or tumorigenesis, providing minimum qualifications for their possible utility in treating the syndrome.15,22,44 One can speculate that the principle of the treatment would be to activate NFκBs at a restricted location and timing. Restoration of Guard hair pegs in cultured Tabby skin by TNFα treatment suggests that a noninvasive or local treatment like topical application of EDA ligands or NFκB activating reagents may be feasible.46 Further encouraging evidence comes from the successful restoration of hair and sweat glands by injection of recombinant EDA-A1 ligands.44 Notably, sweat glands were restored even after birth.44 Sweat gland remediation thus becomes a likely first target for experimental treatment studies. Similarly, the eye is open to topical treatments, and tests can be envisaged of a preventive regimen to prevent ocular surface disease.17
This work was supported by the Intramural Research Program of the NIH, National Institute on Aging. Authors thank Drs. T. Hashimoto, M.C. Durmowicz and R. Nagaraja for their helpful discussions and contributions to the inferences reviewed here.
Previously published online as a Cell Cycle E-publication:http://www.landesbioscience.com/journals/cc/abstract.php?id=3403