Identification of Control Elements 3’ to the Human Keratin 1 Gene That Regulate Cell Type and Differentiation-specific Expression*

To define DNA regulatory elements that mediate the response of the keratin 1 (Kl) gene to Ca2+-induced differentiation, regions spanning the 5’- and 3”flanking sequences, coding regions, and introns from the human K1 gene were cloned into vectors containing the chloramphenicol acetyltransferase (CAT) reporter gene and transfected into cultured mouse keratinocytes. A 4.3-kilobase (kb) region located 3’ to the K1 gene stimulated CAT activity in response to increasing Ca2+ concentrations from 0.05 mM (basal cells) to 1.2 mM (differentiated cells). The 4.3-kb fragment was also active in human epidermal cells but inactive in NIH 3T3 cells and primary mouse fibroblasts. Deletion analysis localized the activity to the terminal 1682 base pairs (bp) of the flanking sequence which retained CaB+ sensitivity in epidermal cells but was not active in mesenchymal cells. Removal of a 207-base pair ele- ment created an enhancer which was active in both epidermal and mesenchumal cells but was still Ca2+-inducible. Further deletions identified two elements which functioned synergistically to give maximal Ca2+-sensitive activity. Stably transfected epidermal cell lines expressed CAT under the direction of these elements when grafted onto nude mice to reconstitute an intact epidermis. Previously reported keratin regulatory motifs were not contained in the 1682-bp frag- ment, but an AP-1 site was identified in one of the synergistic subunits.

To define DNA regulatory elements that mediate the response of the keratin 1 ( K l ) gene to Ca2+-induced differentiation, regions spanning the 5'-and 3"flanking sequences, coding regions, and introns from the human K1 gene were cloned into vectors containing the chloramphenicol acetyltransferase (CAT) reporter gene and transfected into cultured mouse keratinocytes. A

CaB+ sensitivity in epidermal cells but was not active in mesenchymal cells. Removal of a 207-base pair element created an enhancer which was active in both epidermal and mesenchumal cells but was still Ca2+inducible. Further deletions identified
two elements which functioned synergistically to give maximal Ca2+sensitive activity. Stably transfected epidermal cell lines expressed CAT under the direction of these elements when grafted onto nude mice to reconstitute an intact epidermis. Previously reported keratin regulatory motifs were not contained in the 1682-bp fragment, but an AP-1 site was identified in one of the synergistic subunits.
Epidermal differentiation is a complex process that encompasses the expression and organization of unique cytoskeletal proteins and the cessation of proliferation. Although a number of the molecular changes associated with this process have been characterized, the factor or factors which regulate the process remain unknown. One of the earliest and most clearly inducible alterations in differentiating keratinocytes is a change in the pattern of keratin gene expression. Proliferating keratinocytes in the basal layer express keratins K5 and K14, but these keratins are repressed in differentiating keratinocytes and a new pair of keratins, K1 and K10, are induced (1)(2)(3).
Keratins are regulated primarily at the level of transcription, although posttranscriptional modification has been identified (1,2,(4)(5)(6)(7). In cultured keratinocytes the expression of * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "(lcluertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. mRNA and protein for keratins 1 and 10 can be induced by specific concentrations of extracellular Ca2+ (8). The induction of keratin 1 expression requires protein synthesis, as treatment with cycloheximide inhibits keratin 1 transcription in the appropriate Ca2+ medium (8). This suggests that an inducible protein is involved in the regulation by Ca2+. Many eukaryotic genes exhibiting tissue-, developmental-, or differentiation-specific patterns of expression have been shown to be regulated by cis-acting elements. These elements often function as binding sites for trans-acting factors which either increase or decrease the expression of their associated genes. Such elements have been identified either 5' (9)(10)(11)(12)(13)(14), within introns (15), or 3' to keratin genes (16). In light of the epidermal and differentiation-specific pattern of keratin 1 expression, it seemed likely that similar elements would be involved in its regulation.
In previous studies, a 10.8-kb' fragment encoding human keratin 1 (HK1) had been isolated from a h genomic library (17). This fragment included the entire coding region as well as 1200 bp of the 5'-and 4300 bp of 3"flanking sequences (17). When this genomic clone was introduced into transgenic mice, HK1 was expressed only in the epidermis (3). Expression of HK1 was first detected in the developing epidermis of 15-day embryos, suggesting that the clone contained sequences which determined developmental specificity (3). Within the epidermis of these transgenic mice, endogenous mouse keratin 1 was confined to the suprabasal cell compartment, whereas HK1 was expressed in most suprabasal cells, but also was detected in some basal cells. This suggested that the transgene was not as tightly regulated as the endogenous mouse keratin 1. When primary keratinocyte cultures were prepared from these transgenic mice, MK1 and HK1 could be induced by culturing in Ca2+ > 0.10 mM (18). Although both genes were Ca2+-inducible, they had different optima, with mouse K1 being maximally induced in 0.12 mM Ca2+ and human K1 in 0.6 mM Ca2+. These studies indicated that the 10.8-kb genomic fragment included sequences which targeted HK1 expression primarily to the suprabasal epidermis and which responded to Ca2+-induced differentiation in vitro.
We now describe the localization of a tissue-specific and Ca2+-inducible DNA regulatory region within this HK1 genomic fragment. Transient transfection of primary mouse and human keratinocytes with fusion constructs of different regions of HK1 and the chloramphenicol acetyltransferase (CAT) gene has shown that this regulatory segment is located within a 4.3-kb fragment located 3' to the HK1 gene. Deletion studies reveal an array of regulatory elements in the terminal The abbreviations used are: kb, kilobase(s); HK1, human keratin 1; MK1, mouse keratin 1; SV40, simian virus 40; bp, base pair(s); CAT, chloramphenicol acetyltransferase; SVE, simian virus 40 enhancer; KGM, keratinocyte growth medium; EMEM, eagle's minimal essential medium; EGF, epidermal growth factor. 377 1682 bp of this 3"flanking sequence. There appear to be two enhancers which function synergistically, one of which confers Ca2+ inducibility, and a negative element that is important in cell type specificity as its removal leads to activity in both epidermal and mesenchymal cells.

MATERIALS AND METHODS
Plasmid Constructions-Molecular cloning followed established protocols as described (19). pAlOCAT, containing the SV40 early promoter, T n 9 chloramphenicol acetyltransferase (CAT) gene, and small t antigen splice site and polyadenylation signal was used as the origin for all cloning vectors (20).
pGem3CAT(+ and -) and pGem7ZCAT(+ and -) contain the 1.8-kb BglII-BamHI fragment of pAlOCAT, which includes the CAT gene and all necessary SV40 sequences for RNA transcription and processing, cloned in both orientations in pGem3 (Promega, Madison, WI), or pGem-7Zf(+) (Promega). For each experiment, either pAlOCAT, pGEM3CAT, or pGEM7ZCAT served as a negative control (labeled SV(-)). pSVECAT contains the SV40 HindIII-PuuII enhancer fragment cloned 3' to the BglII-BamHI fragment of pGEM7ZCAT and served as a positive control for CAT RNA and protein stability, since pSVECAT contains the identical signals for RNA transcription and processing as the HKI-CAT clones. The HK1 10.8-kb gene was derived from a human h Charon 4A library (17) (Fig. 1). The 5' HK1-CAT fusion constructs were generated by cloning either the 5' EcoRI-PstI HK1 fragment (containing 1.1 kb of the 5"flanking sequence) or the 5' EcoRI-KpnI HK1 fragment (containing 1.2 kb of the 5'flanking sequence), adjacent to the 1.8-kb BglII-BamHI CAT fragment in pGemBCAT(+) (Fig. 1, A and B). Both HK1 5' sequences were cloned 5' to CAT in the sense orientation.
The coding region HK1-CAT fusion construct was generated by cloning the 6.4-kb EcoRI-BamHI HK1 fragment into the BamHI site of pGem7ZCAT, placing HK1 3' to the 1.8-kb CAT fragment in the sense orientation (Fig. 1C).
Digestion of p1700 with XbaI and SnaBI generated a fragment containing the CAT cassette and 1475 base pairs of the human K1 flanking region. This was cloned into pGem7Z at the XbaI and SmaI sites to create p1475. Partial digestion with NsiI created p580, whereas complete digestion with the same enzyme and religation generated p422. The 158-bp NsiI-SnaBI fragment of p1700 was cloned into the NsiI and SmaI sites of pGem7-CAT(+) creating p158. The internal NsiI fragment was cloned into the NsiI site of Gem7Z to create p895. The HincII-SnaBI fragment of p580 was cloned into the HincII and SmaI sites of pGem7CAT(+) making p460. Deletion of the BstXI-BstXI fragment from p580 created p514. Last, deletion of the AuaI-NsiI region of p580 created p429.
To assess the effect of orientation and position on the functionality of these elements, the pGem7-CAT vectors were used to clone p580, 5' and 3' to CAT in the sense and antisense orientations. The BglII-SnaBI fragment of p580 was isolated, blunt-ended with Klenow fragment, and cloned into the SmaI site of pGem7-CAT (+) or (-1 t o generate the four constructs needed to assess orientation and position. Primary Mouse Epidermal Keratinocytes-Primary mouse keratinocytes were prepared as described (21). Cells were cultured in Eagle's minimal essential medium (EMEM; National Institutes of Health Media Unit), containing 0.05 mM Ca2+, 8% fetal calf serum, and 0.25% antibiotic solution (GIBCO). Primary keratinocytes were transfected using Ca2+ phosphate coprecipitation (22) following the modifications of Harper et al. (23). Cells were plated a t 8 X 106/60mm dish in 0.05 mM Ca2+ medium. Fourteen hours after plating, cells were switched to the same medium containing 10 pM K+ for 4 h. Cells were transfected using 3 or 20 pg of CsC1-purified plasmid plus enough calf thymus DNA (sheared to 1 kb) to bring the total to 20 pg/60mm dish. After 4 h, cultures were washed with PBS and treated with a 3-min 25% dimethyl sulfoxide shock, washed three times with PBS, and returned to 10 p~ K+ and 0.05 mM Ca2+ medium. Sixteen hours later, cultures were refed with 0.05 mM Ca2+ medium. Twenty-four hours later, cultures were changed to fresh 0.05 mM Ca2+ medium or challenged with higher Ca2+ medium (0.12-1.2 mM) for a period of 48 h. Preliminary investigations were performed to establish the optimal quantity of DNA for transfection, length of Ca2+ treatment, and age of cells at the time of transfection.
Mouse BK-1 Cells-Cell line BK-1 is a mouse keratinocyte cell line derived from Balb/C mice as reported (24). When grafted to nude mice, BK-1 cells produced normal skin. BK-1 cells were cultured in EMEM with 0.05 mM Ca", 8% fetal calf serum with the addition of 10 ng/ml EGF (culture grade, Collaborative Research, Inc., Lexington, MA). BK-1 cultures were co-transfected with HK1-derived CAT constructs or control plasmids and pSV2Neo and selected with 200 pg/ml G418. G418-resistant clones were pooled by dish, subcloned in selective medium, and grafted to nude mice. Upon healing of the graft, animals were sacrificed and the graft site removed. Tissue homogenates were prepared in 0.1 M Tris, pH 7.7 (at 4 "C), from the graft site in a glass homogenizer followed by microtip sonication. The resulting slurry was centrifuged, and the supernatant was analyzed for CAT activity.
Fibroblasts-Primary mouse fibroblasts were prepared as described (21,25). Cultures were maintained for 7 days in 1.2 mM Ca2+, EMEM medium prior to transfection. Cells were trypsinized and plated at a density of 0.5 X 106/60-mm dish 14 h prior to transfection. Cells were refed 4 h prior to transfection. DNA was introduced by Ca2+ phosphate co-precipitation for 4 h followed by a 2-min 25% dimethyl sulfoxide shock and refeeding with 1.2 mM Ca2+, EMEM. In some experiments, the human fibroblast cell line YDF (26) and NIH 3T3 cells were used as recipients for DNA transfection using the same conditions described for primary mouse fibroblasts.
Normal Human Epidermal Keratinocytes-Cultures of epidermis from breast skin were obtained from Clonetics (San Diego, CA) and cultured in serum-free keratinocyte growth medium (KGM) (0.05 mM Ca2+) until near confluence. Cells were trypsinized and plated at 5 X lo6 cells/60-mm dish. Fourteen hours after plating, cells were switched to EMEM containing 0.05 mM Ca2+, 10 p~ K+, and 8% fetal calf serum. After 4 h, cultures were transfected using the same protocol as for mouse keratinocytes.
Sixteen hours after transfection, cells were switched back to KGM (0.05 mM Ca") for 24 h prior to challenge in KGM with higher Ca2+ medium.
CAT Assays-Cultures were washed three times with phosphatebuffered saline (without Ca2+ or M e ) and frozen on dry ice. Cells were freeze-thawed three times in 0.1 M Tris, pH 7.7, to disrupt the plasma membranes and release the soluble proteins. The soluble fraction was isolated and protein concentration measured using the Bradford Reagent (27) (Bio-Rad); absorbance at 595 nm for samples, as well as for bovine serum albumin standards, were determined in duplicate on a microtiter plate reader. Individual culture extracts were then normalized so that equal amounts of protein were analyzed from each dish. Enzymatic CAT activity was measured using the twophase fluor diffusion assay as described (28) with the following modifications. 1) [3H]Acetyl-CoA (Du Pont-New England Nuclear) was used at a 1:35 (v/v) dilution with unlabeled acetyl-coA (Pharmacia LKB Biotechnology Inc.) to increase the sensitivity, and 2) 10 h was chosen as the reference point for all analyses, since the reaction rate was within the linear range at this time point. Nevertheless, in each experiment a complete time course of activity was performed. All samples were counted on a LKB Scintillation Counter. In all experiments CAT assays were analyzed relative to the values for pSVECAT transfectants in the same experiments. All experiments were performed at least three times and in duplicate for each construct.
Sequence of p1700 and Comparison to Known Regulatory Elements-The nucleotide sequence of p1700 was obtained using dideoxynucleotide sequence analysis (Sequenase, United States Biochemical Corp.) with "S-dATP. B0t.h DNA strands of p1700 were sequenced yielding 1682 neocleotides. Comparison to published sequences in GenBank was performed using the University of Wisconsin Genetics Control Group Software Package.

RESULTS
Enhancer Activity of the 10.8-kb Human Keratin 1 Fragment in Mouse Keratinocytes-In light of the fact that the 10.8-kb human K1 genomic fragment contained sufficient sequence information to respond to Ca2+ in primary transgenic keratinocytes (18), CAT fusion constructs derived from different regions of the human K1 gene were then tested in transient transfections (Fig. 1). These constructs were derived from two overlapping 1.1-and 1.2-kb 5' regions (EcoRI-PstI and EcoRI-KpnI), one 6.4-kb region containing most of the coding sequence and all the introns (EcoRI-BarnHI), and one 4.3-kb 3'-noncoding region (BarnHI-EcoRI). These constructs spanned the entire coding sequence, all introns, 1.2 kb of immediate 5'-flanking sequence, and 4.3 kb of immediate 3"flanking sequence (Fig. 1). None of the HKl 5"flanking region or -coding region constructs showed any significant enhancer activity above background, when compared with either Tris-C1 or the enhancerless constructs. In contrast, the 4.3-kb 3"noncoding region, as well as the positive control pSVECAT, showed strong activity. However, although pSVECAT displayed significant enhancer activity in all Ca2+ concentrations tested, the HK1 3' region was primarily active in higher Ca" concentrations (Table I). For example, the absolute activity of p4.3(+) was significantly lower (30%) than that of pSVECAT in 0.05 mM Ca2+ ( p = 0.007, n = 5; two-tailed Fisher's exact test). Following a Ca2+ shift from 0.05 to 1.2 mM, pSVECAT displays a 2.9-fold induction in activity, whereas p4.3(+) shows nearly a 15-fold increase in activity. The absolute activity of p4.3(+) is 1.5-fold higher  than pSVECAT at this concentration of Ca2+. The same 4.3kb 3' region of HK1 cloned in the reverse orientation 5' to CAT, p4.3(-), also shows a substantial Ca2+-dependent induction, although the absolute level of activity was lower than that of p4.3(+). Fig. 2 shows the Ca2+-sensitive activity of pSVECAT compared to p4.3(+). Both p4.3(+) and pSVECAT increase in activity with increasing Ca2+ concentrations in all experiments, although p4.3(+) always increases to a greater relative and absolute extent. Incrementally, the largest change in expression occurs between 0.12 and 0.6 mM Ca2+, although activity continues to increase between 0.6 and 1.2 mM Ca2+. The low level of activity of p4.3(+) in 0.05 mM Ca2+ medium is not unexpected, since K1 has been shown to be expressed at low levels in 0.05 mM Ca2+ medium (S), and the HK1 transgene is detected in some basal cells in 0.05 mM Ca2' cultures of transgenic epidermis (18).
Since the 10.8-kb HK1 fragment is exclusively expressed in the epidermis of transgenic mice (3), we determined whether the 4.3-kb Ca2+-responsive element is cell type-specific. NIH 3T3 cells were transfected with either pSVECAT, p4.3(+), or pA1OCAT. The SV40 enhancer displays significant activity in 3T3 cells (not shown) as it does in a wide variety of cells that have been tested. In contrast, the HK1 enhancer shows no detectable activity above background in NIH 3T3 cells, even though these cells were cultured in medium containing 1.8 mM Ca2+. The SV40 enhancer, but not the HK1 enhancer, was also active in primary mouse fibroblasts in 1.2 mM Ca2+ medium. In contrast, both pSVECAT and p4.3(+) were active in human keratinocytes, and CAT activity directed by p4.3(+) sequences increased 1.6-fold when cells were switched from 0.05 to 0.6 mM Ca2+ (not shown, see below).
Deletion Analysis-To localize the region or regions of enhancer activity within the 4.3-kb sequence, a series of deletions were made in p4.3(+) using known restriction sites. Transfection analysis of the original construct (p4.3(+)), and the initial deletions are shown in Fig. 3 and represent CAT assay data from cells in 0.05 and 0.6 mM Ca2+ medium for 48 h. Analysis of these constructs indicates that the Ca*+-dependent enhancer activity of the 4.3-kb fragment is located in the distal 1682 bp (determined by sequence), as seen in construct p1700 (determined by mobility). p1700 retains both the Ca2+ inducibility (4-&fold) and the absolute level of activity (1.4-fold greater than SVE CAT) seen in the parent construct. Each of the other deletions has lost most or all of the enhancer activity and Ca2+ inducibility. The sequence for construct p1700 is shown in Fig. 4 and will be discussed below.  To determine whether p1700 retains the cell type specificity of p4.3(+), p1700 was transfected into cells of nonepidermal origin, including primary mouse fibroblasts, NIH 3T3 cells, and a human fibroblast cell line YDF. In each case, the activity of p1700 was normalized against the activity of pSVE-CAT. Since each of these cell types has a Ca2+ optimum of >1 mM for growth, fibroblast transfectants were compared with mouse keratinocytes in medium containing 1.2 mM Ca". Fig. 5 shows that whereas p1700 is active in mouse epidermal keratinocytes, this construct is completely inactive in the t,hree mesenchymal cell types examined. Furthermore, p1700, but not pSVECAT, was inactive when mesenchymal cells were cultured in medium containing concentrations of Ca2+ ranging from 0.05 to 1.4 mM (data not shown). Thus, p1700 is Ca2+-inducible only in keratinocytes. Fig. 5 also shows that p1700, like p4.3(+), is active in human as well as mouse keratinocytes. The apparent lower activity in human uersus mouse keratinocytes is due to a lower ratio of p1700/ pSVECAT directed activity, rather than lower absolute activity.
When the 3' region of p1700 is deleted, the resulting construct, p1475, showed increased activity in mouse keratinocytes compared with both p1700 and pSVECAT (Fig. 6), suggesting that the terminal 207-bp region contains a strong negative regulatory element. Fig. 6 also shows that although p1475 has significant activity in 0.05 mM Ca2+ (comparable with that of pSVECAT in 0.6 mM Ca2+), the activity of p1475 still increases 4-5-fold when cells are switched to 0.6 mM Cas+. Since p1475 exhibits significant activity under conditions in which endogenous K1 is not expressed (0.05 mM Ca"), we were interested in determining the activity of this construct in nonepidermal cells. Fig. 7 shows that p1475, but not p1700, is active in NIH 3T3 cells. Similar results were obtained for primary mouse fibroblasts and YDF cells (not shown). In NIH 3T3 cells, p1475 exhibited activity comparable with that of pSVECAT in medium containing zoncentrations of ca2+ ranging from 0.05 to 1.40 mM, suggesting that, like p1700, the Ca2+ inducibility of p1475 is specific for the keratinocytes.
Removal of the internal NsiI-NsiI fragment from p1475 yielded a construct, p580, with activity comparable with p1475 in both epidermal and nonepidermal cells, wherease the reciprocal construct containing only the deleted 895-bp fragment had minimal activity in mouse keratinocytes in all concentrations of Ca2+ (Fig. 8). When the 580-bp recombinant segment was analyzed as its constituent 422-and 158-bp elements, neither was capable of enhancing CAT to the same level as p580, and the segments functioned synergistically rather than additively (Fig. 8). The 422-bp element is responsible for the Ca2+ inducibility of the enhancer array, since p422 is induced 8-10-fold in 0.6 mM Ca2+, whereas p158 maintains essentially constant activity under all conditions.
Since at least two elements within p1700 are necessary for maximal enhancer activity in mouse keratinocytes, additional 5', 3', and internal deletions in p580 were constructed in an attempt to further localize the enhancer activity (Fig. 9). Neither p514 nor p429 retains the activity of p580. Removal of as few as 66 bp from the 3' end (p514), or 151 bp internally (p429), resulted in a significant reduction in activity, although Ca2+ inducibility was maintained. A 120-bp 5' deletion (p460) from p580 resulted in the loss of basal activity as well as Ca2+ inducibility, demonstrating the importance of this region in the functioning of the enhancer elements. Preliminary findings using DNA footprinting techniques and nuclear extracts from keratinocytes have shown a 30-bp region of strong protection within the 120-bp deleted segment, further supporting the importance of this region in the Ca2+-inducible enhancer activity (not shown).
As p580 contains the smallest region with maximal activity, it was employed to assess the functionality of the element in all orientations and positions. The 580-bp region was cloned 5' and 3' to the CAT gene in both the sense and antisense orientations. These four constructs were transfected into primary mouse keratinocytes and CAT activity measured after culturing cells in 0.05 or 0.6 mM Ca2+ medium. Fig. 10 shows that the basal activity directed by the 580-bp element is greater in the 5' orientation and relative induction by Ca2+ is slightly better in the 5' sense orientation. However, 3' constructs remain active and Ca2+-inducible in both sense and antisense orientations. These experiments show that this element functions in all four orientations and positions and thus fits the classification as an enhancer.
Stably transfected keratinocyte cell lines were created by co-transfection of the cell line BK-1 with either the enhancerless pGem7ZCAT or p4.3 or p580 together with pSV2NEO. Cell lines were selected from each of the plasmids and first analyzed for CAT activity in vitro. These cell lines were then grafted onto nude mice. Twenty-eight days after grafting, the skin at the graft site was harvested, and whole tissue extract was prepared as described and analyzed for CAT activity. CAT activity was not detected in normal nude mouse skin (0 positive grafts/5 graft recipients) or BK-1 cells stably transfected with pGem7ZCAT (0/10 animals). However, constructs p4.3, (7/17 animals) and p580 (11/12 animals) yielded skin grafts with readily detectable CAT activity. This indicates that these elements are capable of functioning as enhancer elements in uiuo in epidermis, although the epidermal specificity cannot be assessed by these studies.

DISCUSSION
We have demonstrated that a 4.3-kb fragment located immediately 3' to the human keratin 1 gene contains genetic elements that direct the synthesis of CAT from the SV40 early promoter in both mouse and human epidermal cells, but not in NIH 3T3 cells nor primary mouse or human fibroblasts. Human Keratin 1 Gene 381 FIG. 4. Nucleotide sequence of the 1700-bp Ca*+-inducible epidermalspecific regulatory region 3' to human keratin 1. The 1700-bp (based on mobility) genomic fragment was sequenced as described under "Materials and Methods" and compared with GenBank sequences for regions of homology. Regions I and I1 (sparsely stippled) confer Ca2+ inducibility and maximal activity, respectively. Asterisks underly sequence 100% homologous to AP-1 binding site, filled circles denote sequence 100% homologous to a TGFP consensus negative response element, and arrows underscore imperfect direct repeats. Region 111 (densely stippled) denotes negative element. Open circles underly sequence 100% homologous to keratinocyte consensus sequence (38).

G G T C A G A A A T G A G A A T A T G C A T T A C A G P I G A fAGTTAG7T.G C T A T A T C A T T
No other DNA regulatory elements were found within a 10.8kb fragment containing the human keratin 1 gene. When transfected into mouse keratinocytes, the 4.3-kb BarnHI-EcoRI human K1 fragment permits Ca2+-sensitive induction of CAT activity when situated either 3' to the gene (in the correct orientation) or 5' to the gene (in the inverse orientation), thus fitting the definition of an enhancer element. However, the 3' construct (p4.3(+)) was more active in mouse epidermal cells in 0.6 and 1.2 mM CaZ+ than both the 5' construct (p4.3(-)) and the SV40 enhancer construct (pSVECAT). Although elevated Caz+ conditions enhance the expression of pSVECAT to a limited extent, the dose-response t o Ca2+, the time course for induction (not shown), and the magnitude of the induction differed between p4.3(+) and pSVECAT. Since pSVECAT and p4.3 contain identical signals for RNA and protein processing, this differential expres-sion most likely occurs at the level of transcription. Thus the HK1 enhancer appears to direct a specific response to Caz+, as well as cell type specificity.
Although the initial localization of transcriptional enhancers was 5' to the genes they regulate, subsequent investigations have identified certain elements within introns and 3' to their associated genes, including genes encoding pglobin, y-globin, a-interferon, T cell receptor 01 and p chains, immunoglobulin heavy chains, CD2, and others (29)(30)(31)(32)(33)(34)(35). Of particular interest to this study is the recent report of a 3' enhancer element in association with the gene encoding EndoA, a developmental-specific type I1 cytokeratin (16).
Unlike the EndoA enhancer element which is comprised almost entirely of six repeated elements with homology to PEA3, there do not appear to be regions of homology to PEA3 in the 3' HK1 enhancer. Nevertheless, both elements are  located 3' to type I1 keratins and appear to function as enhancers. At this time we cannot exclude the possibility that the enhancer we have characterized is a control region for a gene located further downstream. This would imply that the downstream coding sequence would be expressed in an epidermal-specific and Ca2+-inducible fashion. It is known that HK1 is clustered with other type I1 keratin genes on human chromosome 12 (36,37).
Other studies have reported putative regulatory domains that may be involved in the expression of endogenous or viral genes in keratinocytes (10,16,(38)(39)(40)(41)(42)(43)(44)(45). These studies have identified regulatory sequences in conjunction with genes encoding human keratin 14 (5'-GCCTGAGGC-3'), a human K14 pseudogene, and XK81A1, an embryonic epidermal keratin gene in Xenopus luevis which is regulated by a factor called KTF-1 (12,41,46). A frequently occurring epidermal consensus sequence, 5'-AAPuCCAAA-3', has also been re- ported for HPV 16 sequences which appear to confer keratinocyte-specific expression to this virus (38,40). Direct sequence comparison of the HK14 keratinocyte palindrome with the HK1 enhancer elements failed to identify any region with greater than 70% homology. The HK14 keratinocyte palindrome binds AP-2 and was shown to have 3 important guanine residues. In each of the proximal (p422) and distal (p158) 3' HK1 enhancer elements, there are two elements which conserve these guanines and have 60-70% homology to the keratinocyte palindrome (Fig. 4). These are located at nucleotides 66-75, 313-322, 1444-1453, and 1452-1461. A similar degree of homology was identified in assessing the KTF-1 consensus binding sequence (ACCCTGAGGCT), an imperfect palindrome which also has 60% homology to nucleotides 66-75 (14). Although the KTF-1 binding sequence and the keratinocyte HK14 palindrome show homology to each other, they appear to bind different factors as KTF-1 could not compete with AP-2 binding to the HK14 palindrome (47). Sequences similar to the HK14 palindrome have been identified 5' to the gene encoding HK1 and may contribute to the keratinocyte-specific expression pattern of human keratin 1 (46). The smallest fragment of the 3"flanking sequence which retains both Ca2+ inducibility and epidermal specificity is a 1682-bp region. Closer analysis of this region revealed the presence of at least three distinct elements, including a 207bp negative regulatory element at the 3' most end of p4.3. Removal of this region (SnaBI-EcoRI) allowed expression of CAT in all cell types tested. Elements with similar activity, termed "silencers," have been described for several other genes, including those encoding lysozyme (48), several T cell receptors (49, 50), collagen I1 (51), DNA polymerase p (52), alcohol dehydrogenase (53), urokinase plasminogen activator (54), glutathione transferase (55), and a-, p-, and t-globin genes (56)(57)(58). As in the present study, many of these silencers are responsible for cell type-specific expression of their associated genes. In several cases, these silencers had the ability t o suppress heterologous enhancers or promoters (49, 50,56,59). To address the possibility that the 207-bp fragment might represent a silencer, it was cloned into pSVE-CAT. When this fragment was placed 3' to CAT and 3' to the SV40 enhancer, it did not affect the level of SV40-directed CAT activity in nonepidermal cells (data not shown). Although these studies failed to identify silencer activity, the existence of a position-dependent silencer cannot be excluded, since Nakabayashi et al. (60) identified an a-fetoprotein silencer element which suppressed transcription only when it was inserted into the SV40 enhancer 5' to CAT. In this position, the silencer repressed the SV40 promoter in an orientationindependent manner. When the same element was placed 5' to the SV40 enhancer or 3' to CAT, it had no silencer activity in the cell types tested. From the constructs studied, we cannot define a mechanism of action for the observed activity of the 207-bp SnaBI-EcoRI fragment of HK1. It does not appear to alter Ca2+ inducibility as both p1700 and p1475 show the same degree of Ca2+ inducibility. In analyzing the nucleotide sequence of the 207-bp SnaBI-EcoRI fragment, several interesting regions were identified. There is a sequence of 100% homology to the previously identified keratinocyte consensus sequence AAPuCCAAA (Fig. 4) (38). One might postulate that the removal of this element leads to the ubiquitous expression of CAT, even though this element has never been proven to contribute to gene expression. This region also contains sequences with similarity to two previously reported silencer elements (61, 62). Additional investigations are needed to identify the mechanism by which this region is functioning in regulating human keratin 1.
To better define the elements responsible for calcium inducibility, additional deletions were made in p1475. These deletions identified two elements, p422 and p158. Neither element alone is sufficient to drive CAT expression to the same level as p1475. However, these elements appear to be interdependent and function synergistically as p580 exhibits the same pattern of expression as p1475. Such synergy has been identified previously for the 5' human keratin 14 regulatory sequences, Drosophila homeobox elements, and skeletal a-actin (46,(63)(64)(65). Unlike the human keratin 14 elements, where neither element in isolation was able to drive expression (46), the human keratin 1 elements function weakly in isolation and are each able to enhance CAT expression to some degree.
Each of the HK1 enhancer elements appears to have a distinct function. The 422-bp proximal element seems to mediate induction by Ca2+, since both p422 and p580 are strongly induced when cells are switched from 0.05 to 0.6 mM Ca2+. The importance of this element was further illustrated and localized by deleting the 5' most 120 bp of p580. This deletion yields a construct, p460, which is essentially inactive. Preliminary DNA footprint analysis has revealed a distinct 30-bp region of protection within the deleted 120-bp segment, strongly supporting the role of this region in trans-activation of the heterologous promoter. Analysis of the nucleotide sequence of p422 reveals the presence of an AP-1 site as indicated in Fig. 4. However, the contribution of this site to the regulation of K1 is unclear, since neither HK1 nor MK1 is induced by phorbol esters. The distal 158-bp element appears to be involved in determining the absolute level of expression since partial or complete deletion of this region from p580 leads to a reduction in activity but not loss of calcium inducibility. Further analysis of the nucleotide sequence of the p158 identified the presence of three repeats of an eight-nucleotide consensus sequence, GGNTGNGG (Fig. 4). The significance of this finding is uncertain. In addition, p158 contains a consensus sequence for the TGFp inhibitory element which is conserved in a number of TGFP-inhibited genes (66). It has been reported that TGFB inhibits HK1 expression at high concentrations (67).
In an attempt to identify regions of homology to known Ca2+-responsive elements, the nucleotide sequence of the HKl enhancer region was compared with known Ca"-responsive consensus sequences. Ca2+-inducible enhancers have been described for the rat prolactin gene and the c-fos gene (68)(69)(70). In the latter case, the cyclic AMP-induced regulatory protein mediates this effect through its interaction with the cAMP response element (71). Within the proximal enhancer element (p422), there are two elements in tandem with six out of eight nucleotide sequence homology to the cAMP response element. However, cAMP is not induced by Ca2+ in mouse keratinocytes nor does cAMP induce differentiation (72). A similar comparison with the 1,25-(OH)2 vitamin DB response element revealed no regions of significant homology within the HK1 enhancer elements (73). The lack of additional homology to calcium-responsive elements may not be surprising as the induction may be related to the Ca2+-induced differentiation rather than to the Ca2+ itself. Although transient transfections provide insight into the functional activity of DNA regulatory elements, they only give information on the in vitro activity of various sequences. That these 3"regulatory elements were also active in vivo was suggested by the detection of CAT activity in skin grafts of keratinocyte cell lines stably transfected with p4.3 or p580. Since the enhancerless pGem7ZCAT construct was inactive in vivo, the studies show that p4.3 and p580 are functionally active within the epidermis in vivo. Unfortunately we are unable to address the questions of specificity or differential expression with the current methods.
In summary, these experiments have identified three regulatory elements 3' to the human keratin 1 gene; two function synergistically in regulating Ca2+ inducibility, whereas the third appears to function in tissue specificity. Additional studies are needed to fully characterize these elements and the trans-acting factors which interact with them. These elements function both in vitro and in vivo and begin to address the question of cis-acting elements which may be involved in gene expression during epidermal differentiation.