Abstract
This paper identifies a new, developmental role for transcription factor AP-2 in the activation of amphibian embryonic epidermal keratin gene expression. Keratin transcription factor KTF-1 is shown by several criteria to be identical or closely related to AP-2. KTF-1/AP-2 is shown to be tissue-specific from its first transcription in Xenopus embryos, and restricted to a small number of adult tissues, including skin. Epidermis-specific keratin transcription closely follows specification of the embryonic ectoderm in Xenopus, and is subject to regulation by growth factors and embryonic induction. We further show that in mouse basal kératinocytes, a KTF-l/AP-2-like factor is present and binds to a DNA sequence previously shown to be important in the regulation of the keratin K14 gene, which is actively expressed in these cells. Thus, the study of AP-2 and its role in the regulation of keratin gene transcription should enhance our understanding of both amphibian embryonic development and mammalian skin differentiation.
Introduction
Xenopus embryonic epidermis is an ectodermal tissue that derives from the animal cap of the pre-gastrula embryo. Like other epithelia, the epidermis expresses an array of keratins which constitute a major component of the cytoskeleton. The keratin family of intermediate filament proteins contains at least twenty polypeptides, which are differentially expressed in patterns typical of the developmental origin and stage of the tissue. Thus, the study of keratin expression can yield insight into the mechanisms of epithelial differentiation. In Xenopus, the gene for the embryonic keratin XK81A1 is especially interesting, because it is transcribed immediately following the midblastula transition (MBT), specifically in the animal cap (Jamrich et al. 1987). Furthermore, embryonic blastomeres will autonomously turn on XK81A1 at the appropriate time when dispersed in culture (Sargent et al. 1986; Symes et al. 1988), indicating that the determinants necessary to specify ectoderm are already localized in the animal cap at very early stages. Initially, all the cells of the ectoderm express XK81A1, but transcription can be repressed in response to a number of inductive stimuli, including the conversion of ectoderm to mesoderm by growth factors such as activin A (XTC-MIF), and, after gastrulation, the induction of the neural plate by underlying dorsal mesoderm (Symes et al. 1988; Dawid et al. 1988; Jamrich et al. 1987). A knowledge of the factors controlling tissue-specific expression and repression of XK81A1 could, thus, lead to the identification of autonomous ectodermal determinants and a greater understanding of the molecular mechanisms of embryonic induction.
In earlier studies, we showed that 487 base pairs (bp) of 5’ flanking sequence were sufficient to drive epidermis-specific transcription from the XK81A1 gene injected into Xenopus embryos (Jonas et al. 1989). Subsequently, we identified a protein-binding positive-regulatory sequence within this promoter region, mutation of which reduced epidermal expression of the injected gene by up to 90% (Snape et al. 1990). (We recently learned that in these expression studies we erroneously used as a control DNA a construction with 275 rather than 487 bp of 5’ flanking DNA. We have repeated these expression studies using the correct control, with essentially identical results.) Nuclear extracts prepared from Xenopus embryos contained a protein, designated KTF-1 (for Keratin Transcription Factor-1), which bound specifically to this positive-regulatory sequence. We were unable to purify KTF-1 protein directly from embryos in quantities sufficient for protein sequence, determination. However, as described below, we were able to exploit similarity between the KTF-1 binding sequence, and that for the human transcription factor AP-2, to take an alternative approach to the identification of KTF-1.
AP-2 is an enhancer-binding protein which has been purified and cloned from HeLa cells, and specifically interacts with the consensus sequence GCC(r/c)G(c/G)GGC (Mitchell et al. 1987,1991; Williams et al. 1988). The AP-2 consensus binding sequence matches the central core of the KTF-1 binding site (CCCTGAGG) at 7 out of 9 positions. AP-2 sites are found in the promoters of a number of viral and cellular genes and, in many cases, these sites are important in regulating transcription. There is evidence supporting a role for AP-2 in early developmental decisions; several genes that are activated in response to retinoic acid-induced differentiation of mammalian teratocarcinoma cells contain AP-2 sites (summarized in Lüscher et al. 1989). Recently, Mitchell et al. (1991) have shown tissue-specific distribution of AP-2 RNA in the mouse embryo, including expression in ‘pre-epidermis’. As part of a project to identify transcription factors that might play a role in early amphibian development, the Xenopus cDNA homologous to human AP-2 has recently been cloned (Winning et al. 1991). We now show that the protein encoded by the Xenopus- AP-2 (XAP-2) cDNA shares many properties with KTF-1, and conclude that XAP-2 is, in fact, identical or closely related to KTF-1. We show that KTF-l/XAP-2 RNA and DNA-binding activity are localized in the epidermis of the Xenopus embryo, consistent with a role for this transcription factor in the regulation of keratin expression. We also present evidence supporting the suggestion, made by Leask et al. (1990), that a transcriptional activator of the mammalian basal epidermal cell-specific keratin, K14, is AP-2 or an AP-2-like protein, suggesting evolutionary conservation of the regulation of mammalian K14 and the closely related amphibian keratin gene, XK81A1. Transcriptional activation via mammalian AP-2 can be induced by retinoic acid (Lüscher et al. 1989), cAMP and phorbol esters (Imagawa et al. 1987), or repressed by SV40 large T antigen (Mitchell et al. 1987). Thus, several developmentally important agents may act via KTF-l/XAP-2 in the control of embryonic epidermal gene expression.
Besides showing tissue-specific embryonic expression, KTF-l/XAP-2 RNA and protein were found to be expressed specifically in skin, kidney and brain of the adult frog. By analogy with the human K14 gene, KTF-l/XAP-2 may be involved in transcriptional regulation of certain adult keratin genes in Xenopus skin, but, like mammalian AP-2, it probably also regulates non-keratin genes in Xenopus embryos and adults
Materials and methods
Materials
Xenopus laevis embryos were obtained by artificial fertilization, as in Jonas et al. (1989) and maintained in dechlorinated tap water until they reached the desired stage. Metamorphosing tadpoles and adult frogs were purchased directly from Nasco.
Whole cell extracts
These were prepared by the method of Kumar and Chambon (1988) in buffer containing 0.6 M KC1, with the modification that protease inhibitors were added to the buffer as follows: PMSF (Sigma) at ImM, aprotinin at 28 μ gml-1 and leupeptin at 14 μ gml-1 (both United States Biochemical). Xenopus embryos were homogenized on ice in 300 – 500 μ l buffer per 100 embryos in a Dounce glass homogenizer (B pestle). Metamorphosing tadpoles and adult tissues were blended for 30 s, in approximately 3 ml buffer per tadpole or per gram of tissue, using a Tekmar Tissumizer, and filtered through cheesecloth before homogenizing. To allow more extract to be added to binding reactions without raising the salt concentration above 200 mM (the upper limit for maximal KTF-1 binding), the extract from metamorphosing tadpoles was dialysed against buffer containing 0.2 M KC1. Extracts were prepared from cultured mouse basal epidermal cells (the gift of S.Yuspa; Yuspa et al. 1989) and WI-38 cells (the gift of T. Howard) by the freezing method of Kumar and Chambón (1988), using 1 ml buffer per 150 cm2 flask of cells. The protein concentration of extracts was determined using a Pierce BCA protein assay..
In vitro translated XAP-2
This was prepared as in Winning et al. (1991), except that the protein was not radioactively labeled.
DNA-mobility shift assays
These were carried out as in Snape et al. (1990) with the following modifications. (1) The binding buffer was prepared without NaCl, since salt was supplied by the extract buffer. (2) The reaction volume in some experiments was increased to 40 μl, to allow for the addition of a greater volume of extract. Routinely, 15-20 ul metamorphosing tadpole extract, 4 μ1 tissue culture cell extract and 2 p\ in vitro translated XAP-2 were used per binding reaction. Where necessary, extract buffer was added to the reaction to equalize salt concentrations within an experiment. (3) Poly (di • dC)(dI • dC) nonspecific competitor (Pharmacia) was increased to 7.5 or 10 μg per reaction.
Double-stranded oligonucleotides for probes and competitors were synthesized according to the sequences given in Fig. 4, with 4 bases (TCGA) added to the 5’ end of each one, to create overhanging ends when the complementary strands were annealed. To make labeled probe, the ends were filled in using the Klenow fragment of DNA polymerase I (BRL) and [32P]TTP (Amersham) as in Ausubel et al. (1987). Unincorporated nucleotides were removed by isolating the probe from a 10% acrylamide gel (Ausubel et al. 1987). For the DNA-sequence binding-specificity analysis, the intensity of shifted bands was measured on a Macbeth X-ray film densitometer, using films exposed within the linear response range.
Proteolytic clipping bandshift assays
These were based on the method of Schreiber et al. (1988). After incubation of binding reactions as above, 1 μl diluted protease (Boehringer Mannheim) was added to each reaction and allowed to react for 10min at room temperature, before electrophoresis as for the standard DNA-mobility shift assay.
Antibody binding reactions
These were carried out as follows: standard DNA-binding reactions were set up as above, without probe DNA or competitor dl-dC, and 2.5 μ 1 (antiserum to DNA-binding region, and anti-Ha-ras) or 1 μl (antiserum to carboxyl terminal region) was added and allowed to react for 1 h on ice. Probe and dl-dC were then added, and after further incubation for 15 min at room temperature, electrophoresis was carried out as usual. Control affinity-purified rabbit antihuman Ha-ras was purchased from Oncor (Gaithersburg, MD).
Tissue specificity analysis of Xenopus embryos
Dissections were carried out in 100mM NaCl,2mM KC1, 1mM MgSO4, 2mM CaCl2, 5mM Hepes (pH 7.0), 0.1 mM EDTA. Blastulae were divided into animal (including all the pigmented cells) and vegetal hemispheres, using watchmakers forceps. Epidermis was removed from tailbud stages by drawing the embryo into and expelling it from a drawn-out Pasteur pipette. The epidermis was then cleaned in dissociating 67 mM sodium phosphate, pH 7.6. All the cells of each embryo were collected, as far as possible. For each stage, 110 embryos were dissected. Nucleic acid was isolated from 10 dissected and 10 whole embryos by the method of Sargent et al. (1986), except that Li Cl precipitation was omitted, so that both RNA and DNA were retained. Whole cell extracts were prepared from 100 dissected and 20 whole embryos. To estimate the number of cells in each fraction, DNA dot blots (Jonas et al. 1989) were hybridized with a nick-translated probe specific for 1723, a highly repeated, interspersed element in the Xenopus genome (Kay and Dawid, 1983) and counted on an Ambis gel scanner. The volume of RNA or extract was adjusted according to this estimation, so that material from the same number of cells was used in each lane or DNA-binding reaction.
RNA gel blots
RNA from adult tissues, isolated as in Sargent et al. (1986), was the gift of V. Agarwal. RNA gels were run, blotted and hybridized with nick translated probes as in Sargent et al. (1986). Integrity of the RNA was confirmed by ethidium bromide staining of the gel before blotting. The probes used were a 1000 bp EcoRI fragment from a XAP-2 cDNA clone, XAP2-6a (Winning et al. 1991), or an EcoRI-NcoI fragment from a human AP-2 cDNA clone, AP-2Rl/Nco, the gift of T.Williams. Probes were labeled using a nick-translation kit from BRL, and [32P]dCTP from Amersham.
Results
KTF-1 and AP-2 are identical or closely related proteins
DNA-mobility shift assay
Previously (Snape et al. 1990), we noted sequence homology between the XK81A1 KTF-1 site, and the binding site for the human transcription factor AP-2. As this laboratory has recently cloned the cDNA corresponding to the Xenopus homologue of human AP-2 (Winning et al. 1991), we decided to investigate the possible relationship between KTF-1 and XAP-2 by comparison of the properties of the two proteins. Winning et al. (1991) have shown that in a DNA-mobility shift assay, in vitro translated XAP-2 binds specifically to an AP-2 site in the promoter of the human metallothionein gene. Fig. 1 shows that both XAP-2 and KTF-1 specifically bind to a 23 base pair (bp) XK81A1 promoter sequence containing the KTF-1 site, but not to a control mutated oligonucleotide. This shows that XAP-2 can recognise the same binding site as KTF-1, and may be a related protein. In the experiment shown in Fig. 1, the mobility of the DNA-protein complex is slightly different for KTF-1 and XAP-2, suggesting that the proteins may not be completely identical. However, the difference might alternatively be due to post-translational modification of the protein, which does not take place in the in vitro translation system, to complex formation between KTF-1 and another protein in the tadpole extract, or to proteolysis of in vitro translated XAP-2. Furthermore, the mobility of the complex obtained using embryo extracts was somewhat variable, and in other experiments (not shown) it appeared more similar to that obtained with XAP-2.
Proteolytic clipping bandshift assays
Further evidence that KTF-1 and XAP-2 are related proteins was obtained using the proteolytic clipping bandshift assay (PCBA) technique (Schreiber et al. 1988), as shown in Fig. 2. DNA-binding reactions containing the KTF-1 binding site probe, and either in vitro translated XAP-2 or Xenopus whole cell extracts, were treated with increasing concentrations of three proteases, each of which has a different cleavage specificity. When the reactions were run on a nondenaturing acrylamide gel, under conditions routinely used in the mobility-shift assay, each protease gave a characteristic pattern of bands with higher mobility than the undigested protein-DNA complex, resulting from partial digestion of the DNA-binding protein. Only the peptide fragment containing the DNA-binding domain is detected in this assay. With each protease, the pattern obtained for XAP-2 (Fig. 2B) was identical or closely similar to the patterns obtained for Xenopus extracts from either metamorphosing tadpoles (Fig. 2A) or tailbud (stage 19–24) embryos (data not shown).
Antibody-binding reactions
Samples of two rabbit antisera, raised against different regions of human AP-2, were the gift of T. Williams (laboratory of R. Tjian). Evidence that these antibodies cross-react with KTF-1 was obtained by incubating the antisera with extract from metamorphosing Xenopus tadpoles and then adding the KTF-1 binding site probe. When the reactions were run on a mobility-shift assay gel (Fig. 3), there was a small, but reproducible, reduction in the intensity of the shifted band in the reaction with antiserum to the DNA-binding region, suggesting that binding of the antibody to KTF-1 prevented access of the DNA probe to its binding site. In the reaction with antiserum to the carboxyl-terminal region, the migration of the DNA-protein complex was obviously, and reproducibly, retarded, suggesting that this antibody also cross-reacts with KTF-1 and alters the mobility of the DNA-protein complex, but does not prevent KTF-1 from binding DNA. A control, affinity-purified antiserum, anti-human Ha-ras, neither reduced the intensity nor altered the mobility of the KTF-1 shifted band (Fig. 3). Thus, Xenopus KTF-1 shares epitopes with human AP-2.
DNA-binding specificities of KTF-1 and XAP-2
A comparison of the DNA-binding specificities of in vitro translated XAP-2 and KTF-1 from tadpole extracts was carried out. Besides the KTF-1 site at -157 (oligonucleotide X-157; see Fig. 4), the XK81A1 promoter contains a similar sequence at –254 (oligonucleotide X-254), which is conserved in related or coregulated Xenopus embryonic keratin genes (Fig. 4). Moreover, the promoter of K14, the human keratin with the closest known homology to XK81A1, contains two AP-2 like sites (oligonucleotides K14-229 and K14–86), one of which, at –229, can bind a nuclear protein from mouse kératinocytes (see Fig. 9), and act as a positive control element for K14 transcription in human kératinocytes (Leask et al. 1990). Several of the potential KTF-l/AP-2 sites illustrated in Fig. 4 were used as competitors in the mobility-shift assay. The affinities of XAP-2 and tadpole KTF-1 for these competitors were very similar (Table 1 and Fig. 5), suggesting that the two proteins are structurally closely related, at least at the DNA-binding domain. For both factors, the most efficient competitor was X-157, followed by K14-229, MT-175 (the AP-2 site from the human metallothionein gene basal element), and X-254. The K14-86 oligonucleotide competed poorly. None of the three control mutated oligonucleotides competed to any appreciable extent for KTF-l/XAP-2 binding (Fig. 5).
KTF-1 and XAP-2 are tissue-specific, with similar expression profiles
This laboratory has already demonstrated that XAP-2 RNA is absent from Xenopus oocytes and cleavage-stage embryos, with expression beginning shortly after the MBT (Winning et al. 1991). If the KTF-1 DNA-binding activity followed a similar time course, this would be compatible with XAP-2 being KTF-1, and playing an important role in keratin expression in the embryonic epidermis. Whole cell extracts were, therefore, prepared from staged embryos, and mobility-shift assays carried out using the KTF-1 (X-157) probe (Fig. 6A). A shifted band was observed using extracts from pre-MBT embryos, but neither unlabeled X-157 nor control mutated X-157M oligonucleotides at 1000-fold excess competed for binding, indicating that this band represents non-specific binding (Fig. 6B). Specific KTF-1 activity is absent from embryos until after the MBT; it is first detectable at stage 9, corresponding well with the beginning of XAP-2 transcription, which takes place between stages 8 and 9 (Winning et al. 1991). The specificity of the shifted band observed using extract from gastrula-stage embryos is shown in Fig. 6B. The level of KTF-1 activity remains approximately constant from stage 11 throughout embryogenesis (stages 19, 22, 28 and 35/36 were also assayed, but are not shown), followed by an increase at around the feeding tadpole stage (stage 45). Thus the developmental profiles of KTF-1 protein and XAP-2 RNA are similar, consistent with the XAP-2 gene encoding KTF-1.
It was also interesting to find out whether KTF-1 DNA-binding activity and accumulation of XAP-2 RNA corresponded with respect to tissue specificity. Therefore, whole cell protein extracts and total RNA were prepared from late-blastula-stage Xenopus embryos dissected into animal (pigmented) and vegetal (non-pigmented) halves, tailbud-stage Xenopus embryos dissected into epidermal and non-epidermal fractions, and a selection of adult Xenopus tissues. For each source, RNA gel blots were probed for XAP-2 RNA, while mobility shift assays for KTF-1 were carried out using whole cell extracts (Figs 7 and 8).
When assaying KTF-1 DNA-binding activity in dissected embryos (Fig. 7A), the genomic DNA content of each fraction was compared (see Materials and methods) and the volume of extract added to the DNA-binding reactions was modified accordingly, so that each contained extract from an equivalent number of cells. At the blastula stage there was enrichment of KTF-1 in the animal half, from which the epidermis, plus other tissues, derives and, by the tailbud stage, KTF-1 activity was strikingly localized in the epidermal fraction. When the RNA gel blot (Fig. 7B) was loaded in the equivalent fashion (RNA from an equal number of cells in each lane), XAP-2 transcripts were localized in the blastula animal half, and the tailbud epidermis. Once again, the evidence is consistent with the hypothesis that XAP-2 is KTF-1.
In adult Xenopus tissues, both KTF-1 activity and XAP-2 RNA show similar, highly tissue-specific, expression patterns (Fig. 8). In the KTF-1 assay, the volume of extract in the DNA-binding reactions was modified so that equal amounts of total protein were compared for each tissue. Since many of the extracts appeared negative for KTF-1, the experiment was then repeated with the maximum possible volume of extract in each reaction. Still, KTF-1 activity could not be detected in ovary, liver or lung. Adult skin and kidney, however, both contained significant KTF-1 activity. When total RNA was probed for XAP-2 RNA, transcripts were seen in skin and kidney, but not in oocyte, liver or lung RNA. This is further strong evidence that KTF-1 and XAP-2 are the same factor. In addition, testis RNA is negative and brain RNA is weakly positive for XAP-2 transcripts. Protein extracts were not prepared from these tissues, so there are no data on their KTF-1 activity. Since adult Xenopus tissues (except possibly the esophageal lining; Fouquet et al. 1988) do not express the embryonic keratin genes known to contain KTF-1 binding sites, the function of KTF-l/XAP-2 in adult skin, kidney and brain must be different from that in the embryonic epidermis (see Discussion). Adult tissues expressed only the two longer XAP-2 transcripts, lacking the 1.8 kb band, which thus appears to be specific to the early embryo.
KTF-11XAP-2 is present in mammalian keratin-expressing cells
The data illustrated in Fig. 5 show that a positive transcriptional-regulatory element in the human keratin K14 gene can function as a KTF-1 and XAP-2 binding site, raising the question whether K14-trans-cribing mammalian cells specifically contain KTF-1 activity and AP-2 RNA. Mobility shift, PCBA, DNA-binding specificity and RNA gel blot experiments, equivalent to those described above for Xenopus KTF-l/XAP-2, were, therefore, carried out on cultured mouse basal epidermal cells, which are strongly positive for K14 expression (Yuspa et al. 1989; Vassar et al. 1989), and the human fibroblast cell line, WI-38 (K14 negative; Leask et al. 1990). Fig. 9A shows that whole cell extract from the mouse basal cells gives a strong and specific mobility shift with the standard X-157 probe, while the WI-38 cells gave only a non-specific shift, which was not competed by 1000-fold excess of unlabeled probe. The basal-cell-derived complex migrated with a mobility slightly faster than that of in vitro-synthesized XAP-2 (data not shown), as was also observed for KTF-1 from tadpoles (Fig. 1). Figs 9C and 9D show that the KTF-l/AP-2 DNA-binding protein from mouse cells is virtually indistinguishable, with respect to its PCBA profile, and binding specificity, from Xenopus KTF-l/XAP-2 (compare with Figs 2 and 5 and see Table 1). Moreover, mouse basal cells are positive, and WI-38 cells negative, for AP-2 RNA (Fig. 9B). These results suggest that tissue-specific expression of KTF-l/AP-2 is conserved between amphibians and mammals. The KTF-1 binding site at –229 in the K14 promoter has already been demonstrated by Leask et al. (1990) to act as positive regulator of transcription, and it seems likely that the protein that binds this site, which they call KER-1, is KTF-l/AP-2, or a close relative (see Discussion)
Discussion
KTF-1 and AP-2 are closely related or identical
The XK81A1 gene of Xenopus laevis encodes a type I cytokeratin which is transcribed specifically in the embryonic epidermis. Promoter binding studies using crude embryonic extracts, combined with promoter mutation analysis, led to the identification of an activator of XK81A1 expression: KTF-1 (Keratin Transcription Factor-1; Snape et al. 1990). Tn this paper, further characterization of KTF-1 shows that it is either closely related or identical to XAP-2, the Xenopus homologue of the mammalian tissue-specific transcription factor AP-2.
To support this conclusion, we have presented several lines of evidence, each of which is circumstantial or correlative in nature, and none alone can be considered conclusive. However, taken together these experiments strongly support a close relationship between these factors. Furthermore, our laboratory isolated Xenopus AP-2 cDNA clones by reduced stringency hybridization with a human AP-2 probe (Winning et al. 1991). Several clones were isolated in this way, and partial DNA sequence data revealed no significant differences (L. Shea and S. Marcus, unpublished data). Therefore, if KTF-1 and AP-2 were not encoded by the same gene, then one would need to postulate two distinct proteins binding the same DNA sequence element with the same affinity, exhibiting similar or identical mobility in band shift experiments, appearing in the same subset of tissues and at the same time in development, sharing similar protease sensitivity and sharing two or more antigenic sites, but nevertheless having insufficient sequence homology to permit isolation of KTF-1 cDNA clones by hybridization with the human AP-2 probe. We regard this proposition as highly improbable. However, our data are consistent with the existence of subtle differences between KTF-1 and XAP-2. The possibility of post-translational modifications, alternative proteins generated from differently processed RNAs, or other minor differences must be considered open.
Developmental significance of the KTF-1 /XAP-2 relationship
We have previously shown (Snape et al. 1990) that the major KTF-1 binding site at —157 is required for efficient expression of the XK81A1 frog keratin gene, supporting an important regulatory role for KTF-1 in this context. Based on the results that we have presented in this paper, this functional significance can presumably be extended to XAP-2. The most obvious mechanism for achieving tissue-specific gene expression is by the localized activity of transcription factors. Appropriately, in the tailbud stage embryo, KTF-1 DNA-binding activity is strikingly concentrated in the epidermis, and this localization appears to be regulated, at least in part, at the level of KTF-l/XAP-2 transcript accumulation. XAP-2 transcripts were first detected shortly after the MBT and, although the signal at this stage was weak, this RNA appeared to be localized in the animal hemisphere as early as the late blastula, with enrichment in the epidermis persisting at the taibud stage. However, there is evidence that KTF-l/XAP-2 activity may not be completely confined to the epidermis; in earlier experiments, insertion of two KTF-1 binding sites 500 bp upstream of a Xenopus β-globin gene, which was then injected into embryos, appeared to enhance transcription in both epidermal and non-epidermal tissues (Snape et al. 1990). In part, this is because globin RNA levels were measured on a per embryo basis, whereas here KTF-l/XAP-2 was evaluated on a per cell basis. Since the non-epidermal fraction contains more cells per embryo than the epidermis, this would exaggerate the level of globin RNA in non-epidermal cells. In addition, although KTF-l/XAP-2 transcripts were detected only in the animal hemisphere of the blastula, this half of the embryo contributes to other tissues besides epidermis, and these non-epidermal tissues may transiently express KTF-l/XAP-2, leading to enhanced expression of the globin construct.
Transcripts from the XK81A1 gene are first detected in the late blastula or early gastrula (stages 9 and 10), peak in the tailbud embryo (stage 35/36), are greatly reduced by stage 44, and are undetectable in the recently metamorphosed frog (Miyatani et al. 1986; P. Mathers, unpublished data). KTF-l/XAP-2 could therefore be involved in establishing high level transcription of XK81A1, but KTF-1 DNA-binding activity and XAP-2 RNA persist after the keratin gene is no longer transcribed, suggesting that they also play a role in the expression of other genes.
If KTF-l/XAP-2 can function as a transcriptional activator in non-epidermal tissues, then it probably acts in combination with other, positive and/or negative, transcription factors to give completely epidermis-specific keratin expression. There is evidence, from in vivo footprinting, for other factors binding the XK81A1 promoter (P. Mathers, unpublished data). It is expected that these would play a role in tissue-specific transcription, since mutation of the KTF-1 site at –157 leaves residual expression of keratin in the epidermis. However, the keratin promoter used in these experiments retained the site at –254, which, although of lower affinity than the —157 site, is shown here to bind KTF-1 in vitro. Thus it is possible that the remaining epidermal keratin transcription is still mediated by KTF-1. Experiments to test this are currently in progress.
KTF-l/XAP-2 is tissue specific in the adult frog
Since the XK81 embryonic keratins are not transcribed in Xenopus after metamorphosis (except to a low level in esophagus; Fouquet et al. 1988), the finding that KTF-l/XAP-2 is expressed, in a highly tissue-specific pattern, in adult skin, kidney and, to a lesser extent, brain, suggests that, like other developmentally important genes, KTF-l/XAP-2 has an early embryonic function and a separate phase of activity later in development. Considering the high expression levels in skin and kidney, it is tempting to speculate that KTF-l/XAP-2 may play a role in the differentiation of adult epithelia. In addition, AP-2 sites are important in the expression of a variety of mammalian genes, including, for example, tissue plasminogen activator (Rickies et al. 1989), growth hormone (Courtois et al. 1990) and glial fibrillary acidic protein (Miura et al. 1990), so XAP-2 may regulate several different genes in Xenopus as well. Mitchell et al. (1991) have shown recently that AP-2 RNA is localized in fetal and adult mouse tissues, and shows a similar pattern of distribution to the one described here in Xenopus, again suggesting conservation of AP-2 function between amphibians and mammals.
A factor in mouse basal kératinocytes resembles KTF-1/XAP-2
Of the mammalian keratins for which sequence is published, human keratin K14 (Marchuk et al. 1985) shows the greatest similarity to XK81A1 (58.5 % identity at the amino acid level). Sequence homology extends outside the rod region into the globular domains, which are usually less conserved, suggesting a close evolutionary relationship for these genes. K14 is specifically transcribed in the proliferating basal cell layer of the adult skin, and is down-regulated as the cells differentiate (Vassar et al. 1989; Tyner and Fuchs, 1986). The results presented in Fig. 9 strongly suggest that there is a factor in mouse basal kératinocytes that is very similar to KTF-1 and XAP-2. Leask et al. (1990) mention unpublished results suggesting that KER-1, a factor involved in regulating K14 gene expression in human kératinocytes, can bind to the AP-2 site from the human metallothionein gene. Presumably, the AP-2-like mouse kératinocyte factor that we have identified corresponds to KER-1.
In summary, our results support a major developmental role for transcription factor AP-2; the activation of epidermal keratin gene expression in amphibian embryos. The prospect for the study of embryonic development is especially exciting, since it is by the knowledge of factors regulating germ-layer-specific gene expression, and an understanding of the mechanisms that regulate these factors, that we hope to work back towards identification of the primary germ-layer determinants. The epidermal cytokeratin gene is now the earliest Xenopus germ-layer marker for which a transcriptional activator is known, and the regulation of KTF-l/XAP-2, since it is non-maternal and transcribed specifically in the embryo, is itself accessible to further study.
ACKNOWLEDGEMENTS
The authors would like to thank Dr T. Williams (U.C. Berkeley) for making available the human AP-2 cDNA clone, antisera, and unpublished data on human AP-2, Dr S. Yuspa and Dr T. Howard (NIH) for cultured cells, and Dr V. Agarwal (NIH) for RNA samples from adult Xenopus tissues. We also thank Dr P. Mathers, L. Shea and S. Marcus for unpublished data, and other members of the Sargent laboratory for advice and encouragement. A.M.S. is supported in part by a grant from the Wellcome Trust. R.S.W. is supported by a postdoctoral fellowship from the Natural Sciences and Engineering Research Council of Canada.