Transcriptional regulation of a Xenopus embryonic epidermal keratin gene

Embryonic development is a process whereby a totipotent cell, the fertilized egg, divides and differentiates into discrete tissues. This process is thought to be governed by a combination of cell interactions (‘regulative’ development) and cell-autonomous developmental programs dictated by materials inherited from the egg cytoplasm (‘mosaic’ development). One approach to the study of these control mechanisms is to identify and clone genes whose expression is closely linked to changes in embryonic commitment and analyze the trans-acting factors involved in the regulation of this expression.

In developing Xenopus, the earliest known instance of tissue differentiation is the activation of a family of epidermal keratin genes in the ectoderm of late-blastula-stage embryos (Jonas et al. 1985; Winkles et al. 1985). These genes, including XK81A1, are activated in a cell-autonomous manner, but can be rapidly deactivated by exposure of competent ectoderm to inducers that convert these cells into mesoderm (Dawid et al. 1988; Symes et al. 1988), neural tissue (Jamrich et al. 1987), or cement gland (M. Jamrich & S. Sato, unpublished). Embryonic epidermal keratin gene expression is thus a system which can be used to study both cell-autonomous and inductive modes of embryonic development.

Several lines of evidence suggest that the injection of DNA into fertilized Xenopus eggs can be a suitable system for studying the expression of genes active during early embryogenesis. DNA injected into Xenopus eggs has been shown to be rapidly complexed with proteins and organized into structures resembling nuclei (Forbes et al. 1983). Injected DNA may also be replicated to varying degrees (Marini et al. 1988), and usually persists as an extrachromosomal element up to the tailbud stage before largely disappearing from the embryo. Some introduced sequences appear to be integrated into the Xenopus genome and transmitted through the germ line (Etkin & Pearman, 1987). For purposes of studying gene regulation, a more useful observation is that in some cases DNA injected into fertilized eggs appears to be correctly expressed in the embryo. This was first demonstrated in Xenopus for the gene GS17, which is normally expressed only during gastrulation. Cloned copies of this gene, injected into cleaving eggs, were found to be transcribed at the correct time (Krieg & Melton, 1985), and this was shown to be dependent upon a stage-specific enhancerlike element located in the 5’ flanking DNA (Krieg & Melton, 1987). It has also been shown that microinjected copies of the gene encoding muscle-specific actin are active primarily in the correct tissue, i.e. the myotome region of somites (Mohun et al. 1986; Wilson et al. 1986). The tissue-specific expression of this gene also appeared to be dependent upon sequence elements in the 5’ flanking DNA.

In this paper, we report the results of injecting various modified versions of the XK81A1 keratin gene into fertilized Xenopus eggs. These DNAs were transcribed from the correct initiation site, and when sufficient 5’ flanking DNA was present, their transcripts accumulated exclusively in the epidermis. Analysis of gene fusion and 5’-deletion constructs have made it possible to localize the major regulatory elements to a region between -487 and +26 bp relative to the transcription initiation site.

To generate K1310, a 6·4 kb EcoRI fragment, containing the complete coding sequence of the XK81A1 keratin gene, 1310bp of 5’ flanking and about 1·4 kb of 3’-flanking sequence, was isolated from the genomic clone λG8132 and subcloned in pUC18. To distinguish between transcripts of the injected and the endogenous genes, a 36bp marker sequence encoding bovine vasopressin was inserted, without disrupting the keratin open reading frame, at a unique Ncol site at a position in the first exon corresponding to the nonhelical amino terminal domain of the keratin protein. Derivatives of K1310 with reduced 5’ flanking regions were generated by cleaving this plasmid with two restriction endonucleases, followed by progressive digestion with exonuclease in (Boehringer) and SI nuclease (BRL) as described by Henik-off (1984). Suitable deletion clones were selected after restriction enzyme analysis and the precise endpoint of each deletion was determined by sequence analysis. These deletion derivatives were named according to the remaining 5’ flanking DNA, e.g. K1310, K487, etc.

To construct the keratin-globin hybrid gene, clone KG5900, K1310 was truncated by exonuclease III/S1 nuclease digestion (Henikoff, 1984) at position +26 in the nontranslated mRNA leader sequence, and fused at this point with a T6kb EcoRI fragment containing the coding sequence of human βglobin and some 3’ flanking DNA. This EcoRI fragment corresponds to a NcoI-PstI fragment isolated from the genomic subclone pHB9 (Lawn et al. 1980). The Ncol site, which includes the translation initiating ATG, and the PstI site, were converted into EcoRI sites using oligonucleotide adapters. A 4·6 kb Hmdlll fragment from the keratin gene 5’ flanking region was subsequently inserted, yielding a total of 5900bp of upstream DNA. KG5900 does not contain any globin 5’ flanking DNA, and the initiation of transcription of this construct should be directed by the keratin promoter.

A RNase protection probe for human β globin transcripts was prepared from p228-l, which contains the third exon of this gene (Fig. 1A, Karlsson et al. 1988). To generate a probe for RNase protection assays using keratin gene constructs, a 500 bp Xmnl-SphI fragment from K1310 was subcloned in pGEM3 vector (Fig. 1B). The antisense RNA synthesized from this clone, using SP6 polymerase, should protect 315 bp of correctly initiated transcript from K1310 and its derivatives and 236bp from the endogenous keratin gene.

Supercoiled plasmid DNA was purified by the alkaline lysis method (Birnboim & Doly, 1979) followed by CsCl equilibrium centrifugation. RNA traces were removed by passing the DNA through a Sepharose CL-2B column (Pharmacia). Concatamerization of plasmid DNA was carried out by cleaving at a unique restriction endonuclease site in the vector polylinker region. K1310 was linearized at the 3’ end of the gene, while all other constructs were linearized at the 5’ end. Ligation was carried out at l mg ml^-1 DNA in 10 μl with 400 units of T4 ligase (New England Biolabs) at room temperature for 5 min.

Total embryonic RNA was purified from embryos as described previously (Sargent et al. 1986), except that the first phenol/chloroform extraction was carried out in two steps: First, the homogenate was mixed with 0-5 volumes of water-saturated phenol, then 0·5 volumes of chloroform was added, mixed and centrifuged. This change greatly increased the amount of material that could be extracted in a small (0·3-0·5 ml) volume. Most DNA, genomic and plasmid, was removed by precipitation with 2-5 m-LiO at 0°C for 1 h, and remaining traces were destroyed by digestion with 3·3 units ml⁻¹ DNase I (Promega) at room temperature for 15 min in 10mm-Tris pH 7·6, 5mm-MgCl₂.

Radioactively labeled probe RNA was synthesized with SP6 (Promega) or T7 (Stratagene) RNA polymerase in the presence of α-³²P GTP (New England Nuclear) using conditions recommended by the suppliers of the polymerases, and a reaction kit purchased from Promega. The RNA synthesized from 1 μg linearized probe DNA was dissolved in 50μl 10mm-Tris pH7·6, OTmm-EDTA, prepared under RNase-free conditions.

A fraction, usually 1/10, of the purified nucleic acid (prior to LiC1 precipitation) from each batch of injected or control -embryos was set aside for DNA assays. Nicks were introduced in DNA by a 5 min, room temperature treatment with 0·25 m-HC1, followed by 1h at 65°C in 0·2m-NaOH, which also served to hydrolyze RNA. Denatured DNA was bound to Nytran filters (Schleicher & Schuell) and probed with pUC19 DNA, which had been labeled with [α - ³²P] dCTP by nick translation using a kit purchased from BRL.

The hybridization of embryonic and probe RNA was carried out at 37°C for 12 to 16 h in 40μl 80% formamide, 20 mm-NaHPO₄, pH7·0, 400mm-NaCl, 0·1 mm-EDTA. The samples were digested by dilution into 300 pl of 300mm-NaCl, 10 mm-Tris, pH 7·5, 5 mm-EDTA, 40 μg ml⁻¹ RNase A (Boehringer) and 700 units ml⁻¹ RNase T1 (BRL) at 37°C for 30 min. Samples were then adjusted to 1% SDS and digested for 15 min at 37°C with proteinase K (Boehringer, 100 μg ml⁻¹), then extracted with phenol/chloroform and ethanol precipitated with 10 μg carrier tRNA. Pellets were dissolved in formamide and run on 6% acrylamide gels containing 8 m-urea (Maxam & Gilbert, 1980). The markers used for sizing RNA bands were Mspl fragments of pUC19 DNA, labeled with ³²P.

Medium size adult female frogs (Nasco) were primed by injecting 500 units of Human Chorionic Gonadotropin (Sigma) into the dorsal lymph sac 12–16 h prior to use. The eggs were obtained by manually squeezing the primed female, fertilized using macerated testes, and were dejellied 15–20min later with 2% cysteine, pH7·8, for 2–3 min.

For injection, micropipettes were drawn from 0·2 mm i.d. R6 glass capillaries (Drummond, Broomall PA), using a Kopf model 700D needle puller (Kopf Instruments, Tujunga CA), to a tip diameter of 5–10μm. Micropipettes were bevelled on a Narishige EG-3 grinder fitted with 0·3 μm abrasive film (Thomas cat. no. 6775-E54) and were held in a Narishige 3MM micromanipulator. DNA was expelled using compressed nitrogen controlled by an Eppendorf 5242 microinjector apparatus. DNA, dissolved in 88mm-NaCl, 10mm-Hepes, pH6·8, was injected deep into the animal hemisphere, during either the first or second cell cycle. Embryos were kept in 100mm-NaCl, 2mm-KCl, Inim-MgSO,", 2mm-CaCl₂, 5mm-Hepes, pH7·0, Olmm-EDTA, 5% Ficoll 400, during injection, transferred after several hours to dechlorinated tap water and cultured at 18°C until they reached the desired stage, usually 22–24 (Nieuwkoop & Faber, 1967). Abnormal embryos were discarded and the remainder either dissected into epidermal and carcass fractions after partial dissociation in 0·9× CMFM/0-l× isethionate buffer (Newport & Kirschner, 1982) or processed without dissection for RNA purification.

Embryos at stages 22–24 were fixed overnight in 4 % paraformaldehyde and embedded in paraffin as described (Jamrich et al. 1987). For immunocytochemistry, 4 μn sections were mounted on poly-L-lysine-coated slides, dewaxed in xylene and rehydrated through an ethanol series. After rinsing in phosphate-buffered saline (PBS; Biofluids) sections were treated for 5 min with 0·25% Triton X-100 in PBS and transferred for 20 min to blocking buffer (10% goat serum, 4% bovine serum albumin, 0 ·02% sodium azide in PBS). Blocked sections were incubated overnight at room temperature with primary antibody (rabbit anti-human hemoglobin, ICN Immunobiologicals, diluted 1·100 in blocking buffer), washed three times for 5 min in 1% goat serum, 0·02% sodium azide in PBS, and finally incubated for 1h at room temperature with rhodamine-conjugated goat anti-rabbit IgG (KPL, inc., Gaithersburg, MD) diluted 1:40 in blocking buffer. After washing as before, stained sections were mounted in 10% glycerol and viewed under a Zeiss fluorescence photomicroscope.

In principle, sequences responsible for the epidermal specificity of keratin gene expression could be located anywhere in the vicinity of the transcription unit. As a first step in localizing these control elements, we fused the 5’ flanking DNA from the keratin gene, from –5900 to +26 bp, to the coding region of a human fi globin gene (Lawn et al. 1980), beginning at the translation-initiating ATG codon (Clone KG59OO, Fig. 1A). Supercoiled KG5900 was injected into fertilized eggs, which were then cultured until stage 22–24 and dissected as described in Materials and methods into fractions designated ‘epidermal’, which is primarily the outer pigmented layer, and ‘carcass’, comprising most of the non-epidermal cells of the embryo. Globin mRNA was assayed by RNase protection in samples representing equal numbers of embryos using an antisense probe corresponding to the third exon (Fig. 1A). Four protected bands were observed, exclusively in epidermal RNA (Fig. 2). The size of the largest band was approximately 260 nt, the size expected from hybridization of the probe to correctly spliced globin mRNA. The reason for the appearance of multiple bands is not known, but the same pattern was observed using RNA from a mouse cell line transfected with the human globin βgene (Karlsson et al. 1988; data not shown). In this experiment, approximately 90% of the KG5900 DNA remaining at stage 22–24 was found in the carcass fraction (Table 1). Thus the transferred keratin promoter was many times more active in epidermal cells than in other regions of the embryo.

To confirm the tissue specificity of the injected gene, KG5900-injected embryos were fixed at stages 22-24, embedded in paraffin, sectioned and stained with an antibody against total human hemoglobin (Fig. 3). Consistent with the results of dissection experiments, detectable antibody reaction was confined to epidermal cells. Uninjected embryos showed no staining above background in any tissue. The staining pattern within epidermis was highly mosaic, with only a small percentage of epidermal cells expressing globin protein from the injected gene.

The epidermis-specific expression of the injected keratin-globin fusion gene indicates that sequences in the keratin gene located between – 5900 and +26 carry sufficient information to confer tissue specificity on a reporter gene. Further localization of control elements was carried out by injecting fertilized eggs with derivatives of the XK81A1 keratin gene having reduced 5’ flanking DNA content. Fig. 1B outlines the construction of modified keratin clones. The clones in this series were prepared by subcloning a 6 · 4 kb £coRI fragment derived from the genomic clone λ G8132 (Miyatani et al. 1986) into pUC18. This clone, designated K1310, contains the entire keratin gene, 1310 nucleotides of 5’ flanking sequence and about 1400 nt of 3’ flanking sequence. As described in the Materials and methods section, a 36bp marker element was inserted into the first exon, allowing transcripts from the injected and the endogenous keratin genes to be distinguished by RNase protection assays. Following injection of 100 · 200 pg supercoiled K1310 plasmid DNA into fertilized eggs, transcripts from this DNA were detectable in total RNA prepared from whole embryos at stage 24. The size of the primary additional protected band, 315 nt, indicates that transcription from the injected template initiated primarily at the correct site (Fig. 1B). Two longer transcripts, presumably resulting from initiation at secondary sites, also accumulated in injected embryos. The level of RNA accumulated from supercoiled K1310 tended to be rather low in most experiments. However, when K1310 was cleaved at a unique restriction endonuclease site in the vector DNA and concatamerized by treatment with T4 ligase before injection, its transcriptional activity was increased approximately twenty-fold (Fig. 4). The analysis of tissue-specific expression of deleted derivatives was therefore carried out with concatamerized DNA. Fig. 5 displays the results of the analysis of various deletion clones. Embryos were dissected into epidermal and carcass fractions, and these were subjected to RNase protection analysis as above, using the probe illustrated in Fig. 1B. Each lane in a given experiment (Fig. 5A) corresponds to an equal number of injected embryos. The distribution of exogenous DNA was routinely monitored by dot blot hybridization, and in most cases the majority was found in the carcass fraction (Table 1). We have observed significant variability in the level of RNA accumulated in different batches of embryos injected with the same construct, complicating estimation of transcriptional efficiency of different versions of the keratin promoter. Furthermore, some contamination of carcass samples with epidermis occasionally occurred. In order to avoid these problems, we calculated a ‘specificity index’ (Fig. 5B) to quantify the degree to which each deletion construct exhibits preferential activity in epidermal cells. This value is the ratio of epidermal to carcass signal for the injected gene (the 315 nt band) divided by the same ratio for the endogenous keratin transcript (the 236nt band). If the injected gene is as tissue specific as the endogenous keratin gene (which we know to be completely epidermal specific) the specificity index is 1 · 0.As preferential activity in epidermis is lost, the index declines.

Highly epidermis-specific expression was maintained in the clones which had 487 bp or more 5’ flanking DNA (Fig. 5, K1310 through K487). Further deletion to – 403 resulted in partial failure of this regulation, leading to increased accumulation of keratin mRNA in nonepider-mal tissues. The constructs with 208 or 100 bp of 5’ flanking DNA exhibited little if any epidermal specificity.

Our results are a further demonstration of the use of Xenopus embryos as an assay system for identifying regulatory elements in genes expressed in early development. Transferring the keratin gene 5’ flanking DNA (—5900 to +26) to the coding region of the human /3-globin gene resulted in specific accumulation of globin mRNA and protein in the epidermis, leading to the conclusion that sequences residing in the 5’ flanking DNA are sufficient for controlling tissue-specific expression. Since tissue specificity was conferred upon a heterologous coding sequence, this suggests that keratin gene regulation is primarily transcriptional, rather than due to differential RNA stability or other post-transcriptional mechanisms. Deletion analysis showed that DNA downstream from -487 is fully capable of conferring tissue specificity. We have also carried out some experiments with keratin gene constructs with more than 1310 bp of 5’ flanking DNA, up to 5900 bp and have found all of these to be fully epidermis specific (data not shown). Taken together, these results suggest that the region from -487 through +26 includes the critical regulatory sequences for this gene, and it is likely that most, if not all, of the important DNA-protein interactions responsible for the selective transcription of this gene in epidermal cells reside in this small region. However, our experiments do not prove that the other portions of the XK81A1 locus are irrelevant, and it is possible that factors such as mRNA stability or long-range enhancer activities might be important in certain aspects of keratin gene regulation.

As noted in the introduction, expression of XK81A1 as well as other embryonic epidermal keratin genes can be halted by a variety of inductive interactions. Presumably, these negative regulatory events have been correctly imposed on the injected keratin constructs with 487 or more bp of 5’ flanking DNA, since we have not found expression in abnormal tissues with these DNAs. Deletion from -487 to —403 led to the appearance of keratin mRNA in nonepidermal tissues, suggesting the presence of one or more negative regulatory elements in this interval (Fig. 5). The precise location of ectopic keratin expression from the shorter 5’ deletions is not known, since the entire carcass was assayed as a single fraction. The regulation of this gene is not completely negative, however, since further deletion to -208 or -100 resulted in the loss of detectable activity in epidermis. Thus it is reasonable to conclude that control mechanisms are mediated by both positive factors that lead to high level transcription of the keratin gene in epidermal cells and negative controls operating in nonepidermal tissues.

In these experiments, more than 10⁷ copies of cloned genes have been introduced into a rapidly cleaving embryo. This injected DNA may be replicated, distributed, packaged into chromatin, or integrated into the genome in ways that differ from cell to cell. Such nonuniformity might be the basis for the mosaic pattern of epidermal expression of the keratin-globin fusion (Fig. 3). Some, but apparently not all, genes injected into Xenopus embryos exhibit similarly mosaic patterns of expression (see Giebelhaus et al. 1988). Another intriguing but unexplained phenomenon is the observation, also noted by other investigators (Mohun et al. 1986; Wilson et al. 1986), that, for some constructs, linearized DNA is expressed to a much higher level than the same DNA injected in supercoiled form (Fig. 4). This effect might be related to the tendency of injected linearized DNA to replicate more than the equivalent supercoiled form (Marini et al. 1988).

We have cloned and sequenced three other keratin genes whose regulation closely parallels that of XK81A1. These are XK81B1 and XK81B2, other members of the XK81 gene family (Miyatani et al. 1986), and XK70A, which is a distantly related type I keratin (Winkles et al. 1985; Krasner et al. 1988). Comparison of the proximal 600 bp of these genes reveals a single short region common to all four, located in each case between -250 and -270. This is an 8 bp element, located at -258 in XK81A1 with a consensus sequence of GCCTGPuPuG (Fig. 6). A similar sequence has been found in the regulatory region of the mouse c-fos gene (Gilman et al. 1986), where it has been denoted ‘element 2’ and has been shown via methylation interference to bind to unidentified factors from mammalian cells. The function of this element in c-fos expression is not clear, and furthermore it cannot be sufficient for keratin gene regulation since K403, which retains the GCCTGPuPuG, is not fully epidermisspecific. Other features of the regulatory region, shown in Fig. 6, include five copies of a 7 bp sequence with a consensus of GACCTCT, two of which lie within segments of more extensive homology. Further modifications of the 5’ flanking DNA will be required in order to estabfish whether these or other sequence elements are important in the regulation of the keratin gene.

We are grateful to our colleagues Drs Igor Dawid, Der-Hwa Huang, Randy Morse and Alan Wolffe for their helpful comments and discussion of this work. A.M.S. was supported in part by a grant from the Wellcome Trust.