BMP4-dependent expression of Xenopus Grainyhead-like 1 is essential for epidermal differentiation

The pre-metamorphic tadpole, like the early mammalian fetus, uses an epidermal bi-layer to protect itself against environmental insults(Fuchs, 1998; Nieuwkoop and Farber, 1994). Study of these simple epithelial sheets has led to the delineation of key mechanisms necessary for epidermal specification, differentiation and stratification (Fuchs and Raghavan,2002; Koster et al.,2004; Nieuwkoop and Farber,1994; Porter and Lane,2003; Sasai and De Robertis,1997). Extracellular transforming growth factor β(TGFβ)-like ligands play crucial regulatory roles in these events(Altmann and Brivanlou, 2001; Botchkarev, 2003). For example,BMP4 is essential for epidermal specification of naïve ectodermal cells in the Xenopus embryo. Conversely, low or absent BMP4 activity results in neural specification (Wilson and Hemmati-Brivanlou, 1995). This gradient is established across the developing Xenopus embryo by the Spemann organizer, a small group of cells localized initially in the dorsal lip of the blastopore(Sasai and De Robertis, 1997; Spemann and Mangold, 1924). This region expresses a series of extracellular inhibitors of BMP4 signaling,including chordin or noggin, which are essential for appropriate development of the epidermis and central nervous system, as well as the dorsal mesoderm and gut (Lamb et al., 1993; Piccolo et al.,1996; Sasai et al.,1994; Smith and Harland,1992). Thus, endogenous or ectopic expression of BMP4 ligand-binding antagonists results in neural specification of ectodermal cells(Sasai et al., 1994; Smith and Harland, 1992; Zimmerman et al., 1996). Conversely, inhibition of BMP antagonist activity, or enforced BMP4 expression, results in epidermal differentiation(Munoz-Sanjuan and Brivanlou,2002; Wilson et al.,1997; Wilson and Hemmati-Brivanlou, 1995).

The importance of the BMP4 signaling has resulted in the identification of two major types of effector molecules, downstream of the BMP4 receptor, that modulate epidermal-specific gene expression. Activation of Smad signal transducers, and expression of the `immediate early response' (IER)transacting factors Msx1 and XVent2, occurs with BMP4 binding to its cognate receptor(Onichtchouk et al., 1996; Suzuki et al., 1997b; Wilson et al., 1997). In naïve ectodermal cells, ectopic expression of these factors (like BMP4)induces an epidermal fate with concomitant repression of neurogenesis. IER factors, in turn, modulate transcription of a second set of widely expressed trans-acting factor genes, including XAP-2 and Dlx-3, which have a narrower range of action in ectodermal cells,activating epidermal gene expression with concomitant repression of the neural gene expression (Feledy et al.,1999b; Luo et al.,2001b; Luo et al.,2002; Snape et al.,1991).

Although a plethora of genes activated by this deceptively simple signaling cascade have been identified, it remains unclear how epidermal-specific gene expression is achieved. Examination of the regulatory sequences of several mammalian epidermal-specific structural genes implicate a combinatorial network of ubiquitous factors, such as Ap1 or Sp1, and tissue-restricted proteins, such as Ap2 or Dlx3, that bind promoter/enhancer elements in a context-specific manner to facilitate appropriate high-level expression (Byrne,1997; Crish et al.,2002; Kaufman et al.,2003; Ng et al.,2000; Presland et al.,2001; Sinha et al.,2000; Sinha and Fuchs,2001). However, several tissue-specific binding activities have been reported, suggesting that, as yet, unidentified regulators of vertebrate epidermal differentiation await characterization.

This hypothesis is supported by Drosophila mutagenesis screens,numerous tissue-specific gene loci being identified that influence ectodermal differentiation (Juergens et al.,1984; Nusslein-Volhard et al.,1994; Wieschaus et al.,1984). One of these, grainyhead or grh (also known as NTF-1 or Elf1), has an ectodermal-restricted pattern of expression (Bray et al., 1989; Dynlacht et al., 1989; Uv et al., 1997). Loss of grh function results in ectodermal defects including flimsy cuticles,abnormal head structures and a `blimp' phenotype(Attardi et al., 1993; Bray and Kafatos, 1991; Ostrowski et al., 2002). Recent studies suggest an evolutionary conservation of grh-like function. Cuticular defects occur with disruption of ceGrh, an ectodermal-restricted, C. elegans ortholog of grh(Venkatesan et al., 2003). Moreover, we, and others, have identified three mammalian orthologs of grh, grainyhead-like factors 1, 2 and 3 (Grhl1-3), which share an ectodermally restricted pattern of expression and a high degree of identity in the DNA binding and dimerization domains(Huang and Miller, 2000; Kudryavtseva et al., 2003; Ting et al., 2003b; Wilanowski et al., 2002).

In this report, we provide the first evidence for a critical role of a vertebrate grainyhead-like factor, Xenopus grainyhead-like 1(XGrhl1) in epidermal ontogeny. This factor is regulated in an epidermal-specific BMP4-dependent manner, modulating expression of epidermal-specific gene expression. Concordant with these observations,disruption of XGrhl1 activity results in defective epidermal differentiation. Moreover, we identify a XGRHL1-dependent target gene,epidermal keratin (XK81A1) (Dawid et al., 1985; Jonas et al.,1985), and show that a crucial 5′ regulatory motif of the XK81A1 gene, which is necessary for high level promoter activity,binds XGRHL1 directly. Together, these observations provide key mechanistic insights into the role of an epidermal-specific transacting factor in vertebrate development.

Xenopus cDNA libraries (stages 11, 17 and 42, a gift from Dr M. King) were screened under low stringency using a ³²P-labeled human Grhl1 cDNA probe (Wilanowski et al.,2002). A full-length Xenopus ortholog of Grhl1 was identified, designated Xenopus grainyhead-like 1 (XGrhl1)(GenBank Accession Number, AY591750).

Wild-type and albino Xenopus (NASCO) eggs were fertilized in vitro, dejellied, cultured using standard methods(Sive et al., 2000) and staged according to Nieuwkoop's normal table of development(Nieuwkoop and Farber, 1994). Microinjection, lineage tracing and animal cap assays were performed as described (Luo et al., 2002; Parvin et al., 1995; Sargent et al., 1986; Smith and Harland, 1991; Wilson and Hemmati-Brivanlou,1995). Isolation of superficial and deep ectodermal layers was performed as described with minor modification(Chalmers et al., 2002).

Whole-mount in situ hybridization was performed using albino Xenopus embryos (Harland,1991; Hemmati-Brivanlou et al., 1990). Digoxigenin-labeled antisense RNA probes were generated from XGrhl1 full-length cDNA. Other probes used were XK81A1, BMP4, XAP2 and Dlx3(Feledy et al., 1999a; Jonas et al., 1985; Luo et al., 2002; Mariani and Harland, 1998).β-Galactosidase activity was used as an injection tracer(Sive et al., 2000) using the substrates XGal (blue) or Red-Gal (Research Organics, Cleveland, OH). For histological analysis, embryos were embedded in paraffin. RNA extractions,first-strand cDNA synthesis and PCR were carried out as previously described(Kelley et al., 1994; Mead et al., 1996). All assays were performed in the linear range.

RNA extractions, first-strand cDNA synthesis and PCR were carried out as previously described (Kelley et al.,1994; Mead et al.,1996). Primers for ODC, XK81A1 keratin,α -actin, NCAM, geminin, XBra, GATA2, Sox3, BMP4, Msx1, XOtx2, XAG1,XZic3, XNrp1, XVent-2, HIS4 and EDD have been reported previously (Kroll et al.,1998; Sasai and de Robertis,1997; Wilson et al.,1997)(http://www.xenbase.org/methods/RT-PCR.html). Other primers for PCR included: XGrhl1(5′-TGACCACCGCCTTCAGTGCT-3′ and 5′-CCTTGGCTGCCCTGACATTG-3′ amplify a 444 bp fragment at 56°C,25 cycles), PCNA (5′-CGATCAGACGGCTTTGACAC-3′ and 5′-CTCCGCTCGCAGAGAACTTT-3′ amplify a 364 bp fragment at 56°C,25 cycles), P63 (5′-CATGCCCAATCCAAATCAAA-3′ and 5′-CATCTGCCTTGCGGTCTCT-3′ amplify a 444 bp fragment at 56°C,25 cycles), ESR-1 (5′-GGATTACAAGCAAGGGTTC-3′ and 5′-TCCCATAGGATAACGTTCAT-3′ amplify a 378 bp fragment at 54°C,28 cycles), XNotch1 (5′-TGCCTTCCAATCTTACGC-3′ and 5′-AGGGCAGTGTTTTAGGTCAA-3′ amplify a 428 bp fragment at 54°C,27 cycles), XDelta1 (5′-CTGTCCCCCTGGCTACATT-3′ and 5′-CCCTCACACAGACAACCACA-3′ amplify a 305 bp fragment at 56°C 25 cycles), Dlx3 (5′-GCTTGTGGGCAACGAG-3′ and 5′-CTGCGTCTGAGTGAGTCCTA-3′ amplify a 292 bp fragment at 56°C,25 cycles), Dlx5 (5′-ATTCTCCCCAGTCTCCAGTG-3′ and 5′-GATAGTGTCCCCAGTTGCGC-3′ amplify a 425 bp fragment at 55°C,25 cycles), XAP2 (5′-CGGGTATGTGTGCGAAACAG-3′ and 5′-GGCGGGAGACCAATAGAGAA-3′ amplify a 445 bp fragment at 56°C,25 cycles) and ESR6e (5′-GGCACAGGGCAATACTGGT-3′ and 5′-CCCCACTTGGCATTATGTTC-3′ amplify a 400 bp fragment at 55°C,27 cycles). PCR conditions were determined for each primer set to ensure that amplification was within a linear range.

A 3.4 kb XGrhl1 cDNA was subcloned into pRN3 and pCS2+ (kind gifts from P. Lemaire and D. Turner, respectively) to create RN3-XGrhl1 and pCS2+XGrhl1. The plasmid RN3-Δ227XGrhl1 or RN3-EGFPΔ227XGrhl1 were generated from RN3-XGrhl1 by replacing BglII/EcoRI fragment with an adaptor (annealing 5′-GATCTGAGAGCATCATGGCG-3′ and 5′-AATTCGCCATGATGCTCTCA-3′) or PCR amplified fragment of the EGFP cDNA.

Synthetic RNA was made from linearized plasmid DNA with the mMessage mMachine in vitro transcription kits (Ambion, TX). The RNA yield was quantitated by spectrophotometer and its integrity checked by gel electrophoresis. The following plasmids were digested and incubated with the appropriate RNA polymerase: RN3-XGrhl1 (SfiI, T3),RN3-Δ227XGrhl1 (SfiI, T3),RN3-EGFP1Δ227XGrhl1 (SfiI, T3), pSP64T-nucβGal(XhoI, SP6), pSP64T-BMP4 (BamHI, SP6),pSP64T-activin βB (EcoRI, SP6), pCS2+XMAD(NotI, SP6), pSP64T-tBR (EcoRI, SP6),pCS2MT-Ngem (NotI, SP6) and pCS2+noggin(NotI, SP6), RN3-XVent2 (PstI, T3)(Huber et al., 1998; Huber et al., 2001; Kroll et al., 1998; Maeno et al., 1996; Onichtchouk et al., 1996; Smith and Harland, 1992). MOs were obtained from GeneTools, XGrhl1-MO2 having the sequence 5′-GTCGTAGTCTTGTGTCATGATGCTC-3′, and a company-supplied control morpholino (CMO) serving as a control.

Protein:protein interaction assays using GST chimeric factors and electrophoretic mobility shift assays (EMSA) were performed as previously described (Jane et al.,1995).

Blastomeres at the four-cell stage were co-injected with 10 nl of RNA encoding Δ227Grhl1 mutant and the KP487 reporter construct. pRL-TK, a construct in which the thymidine kinase promoter was linked to the renilla luciferase cDNA (30 pg/embryo), was injected to control for injection efficiency. After injection, the embryos were cultured to stage 11. Ten embryos were collected and lysed in 1 × lysis buffer (20 μl/embryo). Luciferase activity was determined in 10 μl of the supernatant, according to the manufacturer's instructions and quantified with a Model TD-20/20 luminometer (Turner Designs). Relative firefly luciferase activity (RLU) was normalized with Renilla luciferase activity in cellular lysates.

To explore the role of grainyhead-like factors in vertebrate embryogenesis,we cloned full-length cDNAs encoding orthologs of grh from staged Xenopus embryo cDNA libraries. Three unique transcripts were identified which share a high degree of identity/similarity with Drosophila grh, a C. elegans grainyhead-like ortholog, and the murine and human Grhl species, particularly in regions defined previously as mediating DNA binding and dimerization(Uv et al., 1994; Venkatesan et al., 2003; Wilanowski et al., 2002). Xenopus Grainyhead-like 1 (XGrhl1), the subject of this report, has the highest level of homology with Drosophila grh (see Fig. S1 in the supplementary material).

Consistent with Drosophila grh expression(Bray et al., 1989; Dynlacht et al., 1989; Uv et al., 1997), XGrhl1 is expressed throughout ontogeny. Maternally derived transcripts (stage 2-8) are replaced by zygotic expression after the mid-blastula transition (stage 9+), which persists throughout embryogenesis(Fig. 1A). Dissection of the germ layers at stage 11 revealed that XGrhl1 transcripts were restricted, like other epidermal-restricted genes, ESR6e and XK81A1 keratin, to the superficial (non-neuronal) layer of the ectoderm at mid-gastrulation (Fig. 1B) (Chalmers et al.,2002; Deblandre et al.,1999; Jonas et al.,1985). By contrast, unlike Drosophila grh, or the neuronal-specific xNotch, xDelta and ESR1 genes, XGrhl1 was neither expressed in the neuroectodermal layer, nor the mesoderm or endoderm at any time point(Fig. 1B; data not shown).

In situ hybridization analysis of staged embryos extended these studies significantly. After fertilization, XGrhl1 transcripts were localized to the blastomeres of the animal pole only(Fig. 2A-D), becoming restricted to the non-neuronal ectoderm of the embryo with progression through gastrulation and neurulation (Fig. 2E-L). Histological sections confirmed this pattern. XGrhl1 expression was observed only in the most superficial cellular layer of the embryo, detectable transcripts being absent from the neural plate(Fig. 2M, arrowhead). At later stages, XGrhl1 transcripts are restricted to the epidermis(Fig. 2P-R), in an identical pattern to XK81A1 (Jonas et al.,1985) (Fig. 2N,O). We conclude that XGrhl1 is a unique marker of superficial non-neuronal ectoderm.

The highly restricted pattern of XGrhl1 expression during ontogeny, coupled with the documented role of BMP4 signaling in Xenopus epidermal differentiation, suggested a potential functional interaction between these factors. To determine whether an active BMP4 signaling pathway was necessary for XGrhl1 expression, we assessed the results of expression of a dominant-negative truncated BMP4-specific receptor (tBR), the BMP antagonist noggin, or the BMP-inhibitory transacting factor geminin, on XGrhl1transcription. Consistent with previously reported effects of neuralization of ectodermal cells expressing these factors ectopically(Kroll et al., 1998; Lamb et al., 1993; Suzuki et al., 1997b), we observed upregulation of NCAM transcripts in explanted animal caps of embryos microinjected with tBR, noggin or geminin(Fig. 3A-B; data not shown). Coincidentally, we observed a dose-dependent reduction in XK81A1 and XGrhl1 expression.

Similar effects were observed in the context of the whole embryo. After co-injection of one blastomere at the four-cell stage with appropriate transcripts and a β-galactosidase (β-gal) mRNA, embryos were allowed to develop to late gastrulation and assayed for XGrhl1 expression by whole-mount in situ hybridization. Epidermal progeny of blastomeres injected with geminin (n=19) or noggin (n=23)showed a complete loss of XGrhl1 expression(Fig. 3C). By contrast, embryos injected with RNA encoding xVent-2 (n=19), an IER component of the BMP4 pathway, or BMP4 (n=20), failed to show any defect in XGrhl1 expression.

To assess whether the specificity of these effects was related to inhibition of BMP4 signaling directly, we examined the outcome of co-injection of tBR and increasing amounts of XMad1 RNA. Enforced expression of XMad1, a BMP4-specific signal transduction molecule, is sufficient to activate the BMP4-dependent signaling cascade, reversing tBR-induced neuroectodermal specification(Wilson et al., 1997), as demonstrated by downregulation of NCAM expression(Fig. 3D). XAGexpression, a marker of the cement gland whose expression is modulated by the relative concentration of BMP4, was also repressed(Gammill and Sive, 2000). By contrast, XK81A1 and XGrhl1 expression was restored to levels observed in caps derived from uninjected embryos. We observed a similar outcome with co-injection of noggin and XMad1 (data not shown). Thus, XGrhl1 expression is dependent on activation of the epidermal differentiation program, is modulated by the BMP4-signaling pathway,and is repressed by regulatory factors that induce neuralization.

These observations raised the possibility that XGrhl1 alone may specify epidermal fate and/or inhibit neural specification. To address this hypothesis, we co-injected increasing amounts of XGrhl- andβ-gal-encoding RNA into animal pole blastomeres, observing no effect on epidermal (top) or neural (bottom) specification at any stage of ontogeny(Fig. 3E; data not shown). Furthermore, co-injection of XGrhl1-expressing RNA with either tBR (Fig. 3F) or noggin (data not shown) at the one-cell stage failed to suppress neural gene expression (NCAM, Sox3 or Nrp1), or rescue known BMP4 epidermal-specific targets including XK81A1 in animal cap explants (Fig. 3F). Taken together, our studies demonstrate that XGrhl1 expression is activated by BMP4 signaling, but in the absence of such a cue, enforced XGrhl1expression is insufficient to specify an epidermal fate.

Establishment of the relative position of XGrhl1 in the BMP4 signaling cascade is complicated by the high concentrations of endogenous BMP4 protein present in ectodermal explants. To circumvent this problem, embryos were isolated at stage 9, and dissociated in Ca²⁺- and Mg²⁺-free buffer (Fig. 4A) (Sargent et al.,1986; Wilson and Hemmati-Brivanlou, 1995). After washing, to remove endogenous extracellular factors, re-aggregated cells were incubated with recombinant BMP4 or media alone until control embryos reached the mid-neurula stage. In the absence of BMP4, cap cells had a neural fate, expressing NCAM,but failing to express either the XK81A1 and ESR6eepidermal-specific markers, or XGrhl1 mRNA, as assayed by semi-quantitative RT-PCR (Fig. 4B). By contrast, increasing amounts of BMP4 induced a significant increase in XK81A1 and ESR6e expression, and a concomitant decrease in NCAM expression. Moreover, epidermal marker expression was associated with a BMP4-dependent induction of XGrhl1transcription. Concomitantly, induction of the Dlx3, Dlx5 and XAP2 transacting factors was observed in BMP4-treated cells. Given that these factors are known downstream components of the BMP4 signaling cascade, these results suggest strongly that XGrhl1 is modulated by BMP4 directly.

New protein synthesis is not required for xVent2 expression, an IER gene transcribed soon after BMP4 binding to its cognate receptor, as shown by its resistance to cycloheximide-mediated (CHX) inhibition(Onichtchouk et al., 1996). By contrast, a second group of BMP4-responsive genes, including Dlx3 and XAP2, cannot be induced in the presence of CHX(Luo et al., 2001a; Luo et al., 2002; Luo et al., 2001b). Studies of XGrhl1 expression demonstrated that, unlike xVent2, this factor had a similar pattern of response to BMP4 as XAP2 and GATA2 in the presence of CHX (Fig. 4B).

To localize the relative position of XGrhl1 in the BMP4 signaling pathway, we micro-injected fertilized eggs with XGrhl1 encoding RNA and assessed its effect(s) in the dissociated cap assay. Induction of BMP4, or of the IER genes, Msx1, xVent2 or p63 was not observed(Fig. 4C), consistent with the absence of BMP4 stimulus, and a lack of effect of XGrhl1 on expression of these factors (Bakkers et al., 2002; Wilson and Hemmati-Brivanlou, 1995). By contrast, not only was XK81A1 induced by XGrhl1, but upregulation of Dlx3,Dlx5, XAP2 and ESR6e expression was also observed. Our results strongly support the conclusion that XGrhl1 is induced by BMP4 signaling, and is located downstream of the IER genes, but upstream of structural genes, such as XK81A1, that are necessary for epidermal differentiation.

Our results are consistent with a model in which XGrhl1 expression modulates epidermal differentiation predominantly at a stage subsequent to commitment of ectodermal cells to an epidermal fate. To elucidate further the role of this factor in epidermal differentiation, we identified a specific dominant-negative form of XGrhl1, guided in part by previous observations of a Drosophila dominant-negative mutant. The fly factor, Δ447grh, lacked a 447 amino acid N-terminal activation domain, dimerized with the wild-type protein and blocked GRH function(Attardi et al., 1993). The structurally homologous mutant of XGrhl1 tested here,Δ227XGRHL1, lacks a N-terminal activation domain encoding the first 227 amino acids, dimerizes with wild-type XGRHL1 and has comparable binding affinity to XGRHLl for a consensus grh-binding motif (see Fig. S2 in the supplementary material). In addition, both Δ447grh andΔ 227XGrhl1 transcripts inhibited XGrhl1-induced XK81A1, Dlx3, Dlx5 and XAP2 expression specifically in dissociated cell explant assays (see Fig. S2 in the supplementary material).

Microinjection of Δ227XGrhl1 encoding mRNA (1 ng) into animal pole blastomeres at the two- to 16-cell stage resulted in normal gastrulation and neurulation of Δ227XGrhl1/β-gal injected embryos (stages 1-25; data not shown). By contrast, the effect ofΔ 227XGrhl1 expression on non-neuronal ectoderm at later timepoints was profound, with gross macroscopic distortion of pre-larval epidermal differentiation (Fig. 5A-C; stages 35-40). We observed a consistent loss of specialized surface structures and a failure of evolution of the normal pigmentation pattern in injected ectoderm. Head and neck structures were defective, with absence of eye and otic placodes, failure of formation of the stomatodeal anlage and lack of melanization (Fig. 5A,B). Trunk/tailbud structures were also abnormal, with frequent misshapen embryos and loss of appropriate fin formation(Fig. 5B). Furthermore, there was apparent persistence of embryonic pigment at stage 40 on the injected side alone, which is more characteristic of earlier stages of pre-larval development (Fig. 5A,C;arrows). By contrast, Δ227XGrhl1 transcripts, even at high concentrations, had no macroscopic effect when injected into blastomeres with a neural or mesendodermal fate, specificity being confirmed by rescue ofΔ 227XGrhl1-mediated defects by co-injection of XGrhl1 transcripts (data not shown).

Histological examination confirmed a primaryΔ 227XGrhl1-induced block in epidermal differentiation, most strikingly at stage 40. The outer epithelial layer (OEL) or periderm on the uninjected side consisted of columnar cells with small nuclei and a few apically distributed cytoplasmic yolk platelets(Fig. 5D). Expression ofΔ 227XGrhl1 resulted in poorly differentiated peridermal cells with large nuclei, prominent nucleoli, the presence of cytoplasmic yolk platelets and a pancellular distribution of embryonic pigment granules more characteristic of earlier less differentiated stages of Xenopusontogeny (Nieuwkoop and Farber,1994). By contrast, the underlying flattened inner sensorial layer(ISL) was similar on both sides. Similar changes in the epidermis were observed in Δ227XGrhl1-injected blastomeres, which localized to the trunk and fin regions (Fig. 5E,F), confirming the generalized effects ofΔ 227XGrhl1 expression on epidermal differentiation.

Additional features were evident from histological analysis of embryos in which blastomeres with a non-neuronal ectodermal fate were injected withΔ 227XGrhl1-encoding RNA transcripts. First, disorganization of specialized epidermal structures including otic and optic placode formation was observed with consequent failure of lens structure development in the latter instance (Fig. 5C; data not shown). Coincident with the latter changes, significant regression of the neuroretinal structures was observed, with a reduction of the surrounding pigment layer. Second, structures deep to the epidermis showed significant changes, including disorganized head and trunk mesenchyme, and alteration in the neural tube and the stomodeal anlage. By contrast, histology of embryos in which Δ227XGrhl1 transcripts were injected into blastomeres with a mesoendodermal fate was normal (data not shown). These results indicate that loss of XGRHL1 function results in a specific primary alteration in epidermal differentiation in the maturing Xenopus embryo, with apparent loss of epidermal inductive signals to underlying structures.

Architectural changes in the OEL occurring with Δ227XGrhl1expression, coupled with the modulation of expression of XK81A1,ESR6e and other epidermal-restricted factors in explanted cells, support the contention that XGrhl1 activity is necessary for appropriate epidermal differentiation subsequent to commitment. To test this idea further,we asked if XGrhl1 activity was required for epidermal structural gene expression in vivo. We chose to focus our efforts on XK81A1expression, given the central role of keratins in structural and morphogenetic events in vertebrate epidermal cells(Fuchs and Raghavan, 2002; Porter and Lane, 2003). Transcripts encoding Δ227XGrhl1, co-injected with β-gal into one blastomere at the four-cell stage, resulted in complete loss of XK81A1 expression in the progeny of injected blastomeres when assayed at stage 14 (Fig. 6A; data not shown). This effect was reversed with co-injection of wild type XGrhl1, confirming the specificity of Δ227XGrhl1activity (Fig. 6A; Table 1). By contrast,Δ 227XGrhl1 had no effect on XAP2, Dlx3 or BMP4 expression (Fig. 6B). A green fluorescent protein/Δ227XGrhl1 chimera gave a similar result and confirmed that the dominant-negative factor, like the wild-type protein was localized to the nucleus of injected blastomere progeny(Fig. 6B; data not shown).

To corroborate these observations, embryos were injected with a XGrhl1-specific antisense morpholino (XGrhl1MO; see Fig. S3 in the supplementary material). An identical phenotype toΔ 227XGrhl1 was observed with loss of XK81A1 expression with co-injection of a XGrhl1MO and transcripts encoding β-gal into one blastomere of four-cell stage embryos(Fig. 6C). By contrast, a control MO (CMO) had no effect on XK81A1 expression. XK81A1expression was rescued partially by co-injection of XGrhl1MO with M-XGrhl1 transcripts, the latter encoding a synthetic isoform of XGrhl1 which is partially resistant to XGRhl1MO (see Fig. S3 in the supplementary material; Fig. 6D and Table 2). Embryos injected with XGrhl1-specific MOs and followed until later time points (stages 35-40) had a similar phenotypic defect in epidermal differentiation to that observed with Δ227XGrhl1(Fig. 6E). Together, these studies strengthen significantly our model of a key role for XGrhl1in modulating epidermal differentiation and XK81A1 expression specifically.

To determine whether XGrhl1 regulates XK81A1transcription directly, we examined DNA sequences directly upstream of the XK81A1-coding region using a previously defined consensus Drosophila GRH binding motif(Huang et al., 1995; Shirra and Hansen, 1998). A sequence centered at ∼200 bp upstream of the transcriptional start site shares considerable nucleotide identity (56%) with the invertebrate factor binding region and binds recombinant XGRHL1 specifically(Fig. 7A; see Fig. S4 in the supplementary material). Mutagenesis of this sequence identified nucleotides that, when altered, ablated XGRHL1 binding (designated M2; see Fig. S4 in the supplementary material). Interestingly, this region also encodes an adjacent XAP2-binding motif, which has been implicated previously as being essential for XK81A1 expression(Snape et al., 1991). To test the in vivo relevance of in vitro XGRHL1 binding, firefly luciferase reporter cassettes linked to XK81A1 promoter fragments with specific mutations were introduced into Xenopus embryos(Fig. 7B). Similar activity was observed with the full-length promoter, and a promoter truncated to -276 bp relative to the transcriptional initiation site (compare KP487 and KP276). However, deletion of a region between -215 and -101 bp (-113KP487) resulted in a greater than 50% reduction in reporter activity. Ablation of XAP2binding (M157E) or the XGrhl1-binding motif (M2) resulted in a 50%reduction in reporter gene activity. Mutation of both motifs failed to show a further decrease in luciferase expression (M2M157E).

To confirm the importance of XGRHL1 binding for XK81A1transcription, we assessed the functional consequences of inhibition of XGrhl1 activity on XK81A1 promoter (KP487) function. Co-injection of Δ227XGrhl1 transcripts with the KP487 reporter construct resulted in a 50-60% reduction in luciferase reporter activity(Fig. 7C). By contrast,Δ227XGrhl1 transcripts failed to affect M2 reporter gene activity, when compared with embryos injected with the M2-driven luciferase construct alone. Collectively, these results demonstrate that the XGRHL1-binding motif is required for XK81A1 promoter activity.

Intact BMP4 signaling is necessary for differentiation of the embryonic Xenopus epidermis (Baker and Harland, 1997). Our studies demonstrate that XGrhl1 is a downstream epidermal-specific target of this pathway. This observation varies significantly from that reported in Drosophila, in which grhexpression modulates BMP4 activity (Huang et al., 1995). This evolutionary divergence raises three questions: (1) is XGrhl1 necessary for BMP4-dependent epidermal specification; (2) if not, what is the role of XGrhl1 in the pathway;(3) is XGrhl1 involved in other epidermal-specific signaling events?

Ectopic expression of BMP4 or of IER factors, such as Xmad1,result in epidermal re-specification in cellular progeny of blastomeres with a neural fate (Wilson et al.,1997). Similarly, co-expression of IER factors and BMP antagonists/inhibitors induces epidermal specification in injected ectodermal cells, with coincident repression of neural gene expression(Feledy et al., 1999a; Suzuki et al., 1997a; Suzuki et al., 1997b). Given the temporal pattern of endogenous expression, we expected a similar outcome with enforced expression of XGrhl1. Differing sharply from the effects of IER factors, ectopic expression of XGrhl1 failed to induce epidermal specification. These observations suggest that XGrhl1activity is dispensable for this process, a conclusion supported by the inability of injection of Δ227XGrhl1-encoding transcripts or XGrhl1-specific MOs to affect germ layer specification.

Our studies suggest an alternate model, XGrhl1 functioning downstream of the IER factors in the BMP-signaling cascade(Fig. 8). In this context, AP2 and Dlx-like factors have been shown previously to be essential for appropriate epidermal differentiation(Fuchs and Raghavan, 2002; Luo et al., 2002; Panganiban and Rubenstein,2002). However, it remains unclear how these factors achieve tissue specificity given their wider pattern of gene expression(Luo et al., 2001b; Luo et al., 2002). We show that induction of XK81A1 is dependent on appropriate XGrhl1function. Like Dlx3, expression of XGrhl1 does not induce expression of the epidermal structural gene XK81A1 in the absence of a functional BMP4 pathway, suggesting that morphogen-induced expression of other factors is necessary. One candidate may be AP2, given its ability to rescue the epidermal defect induced by tBR expression in a similar manner to IER regulatory factors (Luo et al., 2001b; Suzuki et al.,1997b; Wilson et al.,1997). Furthermore, like XGrhl1, AP2 fails to repress expression of pan-neural gene markers, a divergence from the effects of IER factor expression (Luo et al.,2002). These observations, together with our studies of the XK81A1 promoter, indicate that XGrhl1 functions predominantly downstream of the IER factors in the BMP4 signaling cascade. Furthermore, our studies of the XK81A1 promoter demonstrate that both AP2 and XGRHL1 are required for XK81A1 expression (see below). Thus, we suggest that characterization of the expression of XGrhl1 and its mechanism of action represents a significant new insight into the regulation of BMP4-responsive epidermal-specific targets,this tissue-specific factor modulating structural gene expression in concert with the more widely expressed regulator AP2 during terminal differentiation.

Intriguingly, XGrhl1 induces AP2 and Dlx3expression in dissociated animal explant cells, consistent with models in which positive feedback loops initiated by downstream regulatory factors stabilize and/or potentiate the decision to undergo a specific cellular fate(Ferrell, 2002; Green, 2002). In addition,this positive `epidermal' loop underscores emerging evidence of a complex crosstalk occurring between regulatory molecules within vertebrate epidermal differentiation programs (Fuchs and Raghavan, 2002). Recent studies have implicated Drosophila GRH as a target of the FGF, Notch and Wnt/Wingless (Wg) signaling pathways(Furriols and Bray, 2001; Hemphala et al., 2003; Lee and Adler, 2004),suggesting that the expression and function of XGrhl1 may represent the sum of these inputs. In this context, the effects of XGRHL1 on the transcription of the bHLH factor ESR6e, an ectodermal target of Notch and BMP4 signaling, may be instructive(Chalmers et al., 2002; Deblandre et al., 1999). ESR6e expression is restricted to the epidermis predominantly,although it is also observed in cells of the neural plate. The coincident expression of ESR6e and XGrhl1 expression, and the induction of ESR6e by ectopic expression of XGrhl1(Fig. 1C; Fig. 4C) suggest a functional link. Indeed, the lack of ciliated cells inΔ 227XGrhl1-expressing epidermis(Fig. 5 and data not shown),coupled with the role of ESR6e in Notch-mediated specification and migration of ciliated cells into the Xenopus periderm(Deblandre et al., 1999)suggest that XGrhl1 may be involved in this process.

Our observations demonstrating a severe defect in end-stage epidermal differentiation induced by perturbation of XGRHL1 function raise the issue of whether this defect is related to loss of XK81A1 expression alone, or to a more generalizable defect in expression of genes of the epidermal program. The latter model is supported by several observations. Although the effect of a knock down of XK81A1 expression on Xenopusdevelopment has not been described, studies of dysregulation of keratin K14,its murine homolog, suggest otherwise. Animals homozygous for K14 disruption exhibit a similar increase in keratinocyte fragility that resembles theΔ 227XGrhl1-induced phenotype(Lloyd et al., 1995). Conversely, these mice are viable at parturition, have appropriate differentiation of embryonic and adult epidermal keratinocytes, and have no evidence of a secondary defect in underlying mesendodermal tissues. A second line of evidence supporting a more generalized epidermal defect is provided by the observation that enforced co-expression of XK81A1 inΔ 227XGrhl1-expressing embryos (or in embryos injected with XGrhl1-MOs) failed to reverse the phenotype (J.T., unpublished). Collectively,these results suggest that XGrhl1 modulates transcription of a range of epidermal-specific genes.

Additional lines of evidence support this conclusion. First, comparison of Drosophila and Xenopus Grhl-defective phenotypes confirm an evolutionary conservation of Grhl factor function during epidermal ontogeny. Thus, the failure of appropriate epidermal differentiation, abnormal head structures and defective specialized epidermal structures characteristic ofΔ 227XGrhl1 expression mirrors the defects observed with expression of a structurally similar mutant, Δ447grh, in Drosophila. The mutant fly phenotype includes cuticlar defects, a`grainyhead' phenotype and deficits in hooks, mouth pieces and wing structures(Attardi et al., 1993; Bray and Kafatos, 1991; Lee and Adler, 2004). Interestingly, injection of Δ447grh into Xenopusblastomeres with an epidermal fate at the eight- to 16-cell stage results in a similar outcome to that observed with Δ227XGrhl1 (data not shown). Second, we observed XGrhl1-mediated expression of other epidermal-restricted differentiation factors, including AP2, Dlx3 and ESR6e in ectodermal cells (Fig. 4). As discussed above, this suggests a complex requirement for XGrhl1 in the expression of these genes.

What are the identities and functions of other XGrhl1-dependent genes? The failure to resorb cytoplasmic yolk platelets and assume the flattened morphology of the mature peridermal cell suggests that XGrhl1 may modulate repression of the primitive epidermal gene program and/or modulate morphogenetic changes in cell shape. The latter hypothesis is supported by observations demonstrating a role for Drosophila grh in changes in tracheal cell shape and the epidermal-specific failure of murine neural tube closure observed with mouse Grhl3 deficiency (Hemphala et al.,2003; Ting et al.,2003a). Given our demonstration of a conservation of Grhl function, other Grhl targets may be provided by recent characterization of orthologs of Drosophila blimp phenotypes similar to grh(Ostrowski et al., 2002). Preliminary analysis has identified several genes epistatic to GRH,including cadherins and adhesion molecules(Lee and Adler, 2004). Indeed,some of these molecules are also involved in the development of specialized epidermal appendages. Given the apparent evolutionary conservation of Grhl function, it will be important to determine the functional role of vertebrate orthologs of XGrhl1, particularly its role in the stratification of the adult epidermis, as similar mechanisms are operative in the embryonic epidermis and the basal layer of the adult skin(Byrne et al., 1994; Koster et al., 2004; Nieuwkoop and Farber,1994).

Vertebrate promoter/enhancer regulatory elements of type I (and II)keratins, and other structural genes contain functionally important motifs for the non-epidermal specific AP2, AP1 and Sp1 DNA-binding factors, amongst others (Byrne and Fuchs, 1993; Jonas et al., 1989; Kaufman et al., 2002; Leask et al., 1990; Leask et al., 1991; Sinha et al., 2000; Sinha and Fuchs, 2001; Snape et al., 1990; Snape et al., 1991). The expression patterns of these factors suggest that non-epidermal specific factors interact in a combinatorial manner, potentially recruiting keratinocytic-specific co-regulators to modulate appropriate expression(Fuchs and Raghavan, 2002). Several mammalian keratinocyte-specific promoter/enhancer-binding activities have been identified recently, although detailed characterization is awaited(Kaufman et al., 2002; Sinha et al., 2000; Sinha and Fuchs, 2001). Our studies alter this model significantly. We demonstrate clearly the molecular basis by which binding of a novel epidermal-specific factor, XGRHL1, is essential for high-level transcription in epidermal cells. Interestingly,preliminary exploration of murine K14 sequences, as well as Xenopus AP2 and Dlx3 promoters, identified similar Grhl binding motifs(J.T., unpublished).

Adjacent to the XGRHL1 site is a previously defined AP2 motif crucial for maximal promoter activity (Snape et al.,1990; Snape et al.,1991). Our studies suggest a functional interaction between these factors (Fig. 7D,E). Drosophila GRH interacts with dTAF_II110(Attardi and Tjian, 1993; Dynlacht et al., 1989; Dynlacht et al., 1991), a component of the TFIID TATA-binding complex providing a structural basis for the exploration of the molecular mechanism(s) of keratin gene transcription. Interestingly, hTAF_II130, the ortholog of dTAF_II110,interacts with the co-regulator CBP(Nakajima et al., 1998), the latter co-factor being required for AP2-mediated gene activation(Braganca et al., 2003). Together, these observations suggest the existence of a molecular `bridge'between the DNA-bound transacting factors and the transcriptional initiation complex which may explain the requirement for binding of both XGRHL1 and AP2 for maximal promoter function. It will be important to confirm these relationships in future studies.

We thank R. M. Harland, T. D. Sargent, C. Kintner, K. L. Kroll, P. A. Krieg, C. Niehrs, C. Chang, R. S. Bradley, B. M. Gumbiner and R. Tjian for reagents; and members of the Cunningham, Mead and Jane laboratories for helpful discussions. We thank Clair Kelley and Jim Ihlefor critical comments. Support for this work was provided by the NIH, NHMRC, Assisi Foundation of Memphis, The Wellcome Trust and American Lebanese Syrian Associated Charities.