Mechanisms Identified in the Transcriptional Control of Epithelial Gene Expression*

Epithelium-specific gene expression is fundamental in both embryogenesis and the maintenance of adult tissues, and impairment of epithelial characteristics contributes to diseases such as cancer. We have here analyzed the 5 (cid:42) -region of the epithelial (E-) cadherin gene in order to understand mechanisms of epithelium- specific transcription and loss of expression during ep-ithelial-mesenchymal transitions. The regulatory region of the mouse epithelial cadherin gene is composed of a promoter (from position (cid:50) 94 to the transcription start site) and a 150-base pair enhancer located in the first intron. The 5 (cid:42) -promoter consists of positive regulatory elements (a CCAAT-box and two AP-2 binding sites in a GC-rich region) and the palindromic element E-Pal that activates and represses transcription in epithelial and mesenchymal cells, respectively. The enhancer of the first intron stimulates the activity of heterologous pro- moters exclusively in epithelial cells. This epithelium-specific enhancer consists of three elements (E I to E III; E II and E III bind AP-2) that are necessary and sufficient for activity. We thus propose two regulatory mech- anisms by which epithelial specificity of epithelial cadherin expression is determined: cells by E-Pal activity in epithelial cells both E-Pal and the epithelium-specific enhancer.

Epithelia are essential and abundant tissues in most eukaryotic organs. Epithelial cells are the first identifiable embryonic cell type, which appears during compaction of the morula early in development (Fleming et al., 1993). During gastrulation, epithelial-mesenchymal transitions take place, and in this process epithelium-specific genes are repressed, and genes of the mesenchymal (and neuronal) lineages are activated (Cunningham and Edelman, 1990;Jessell and Melton, 1992). New epithelia usually derive from existing ones, i.e. from the ectoderm or endoderm, but can also be formed from the mesoderm by mesenchymal-epithelial transitions (e.g. during development of the kidney) (Saxen, 1987). In development, epithelialmesenchymal and mesenchymal-epithelial transitions take place in a temporally and spatially controlled manner (Valles et al., 1991;Boyer and Thiery, 1993;Birchmeier and Birchmeier, 1993), whereas in tumors these changes are highly uncontrolled; loss of epithelial character is typically observed late in progression of carcinomas and correlates there with the acquisition of invasive and metastatic potential (Birchmeier and Behrens, 1994;Reichmann, 1994).
Epithelial cells form continuous cell layers, and they are generally polar. In single-layered epithelia (e.g. the mature intestine), apical and basolateral cell surface are separated by tight junctions (Citi, 1993). An example of a multilayered epithelium is the skin, where basal cells (stem cells) are covered by layers of gradually differentiating cells (Fusenig et al., 1994). Thus, epithelia are extremely complex tissues, and they are highly variable in type and degree of differentiation. Typical structures in epithelia are adherens junctions and desmosomes (Buxton and Magee, 1992;Tsukita et al., 1993;Hü lsken et al., 1994a), which are organelles responsible for strong intercellular adhesion; epithelial cells also form hemidesmosomes to the basement membranes at their basal side (Timpl, 1989;Sonnenberg et al., 1991). Polar epithelial cells developed special mechanisms that allow the transport of membrane proteins to either the apical or basolateral surface (Eaton and Simons, 1995). Epithelial cells express characteristic genes that are responsible for the maintenance of the epithelial phenotype; for example, components of junctions or keratins (Birchmeier and Behrens, 1994;Kouklis et al., 1994;Buxton et al., 1993), specific epithelial products (e.g. albumin in the liver) (Cereghini et al., 1987), and specific transcription factors (e.g. LFB-3 in the liver or kidney) (De Simone et al., 1991).
Recently, much progress has been made in the elucidation of the molecular basis of epithelial junction formation (Tsukita et al., 1993;Garrod, 1993;Citi, 1993;Hü lsken et al., 1994b). Adherens junctions are specialized structures containing the transmembrane cell adhesion molecule epithelial cadherin Ecadherin, 1 that recognizes and binds E-cadherin present on the neighboring cells in a Ca 2ϩ -dependent manner. The cDNA of E-cadherin codes for a signal peptide and a presequence at the amino terminus, a large extracellular domain with four repeated domains important in Ca 2ϩ -binding, a single transmembrane sequence, and a short cytoplasmic domain (Takeichi, 1991;Kemler, 1993). E-cadherin is the prototype of a family of Ca 2ϩ -dependent cell adhesion molecules and is expressed in all embryonal and adult epithelial tissues. In development, E-cadherin expression is down-regulated during epithelial-mesenchymal transitions and reappears during reversion to the epithelial phenotype. For example, E-cadherin disappears during differentiation of the dorsal ectoderm into the neural tube (Thiery et al., 1982;Nose and Takeichi, 1986), and it is induced in epithelial cells that develop from mesenchyme during morphogenesis of kidney tubules (Vestweber et al., 1985). Overall, E-cadherin is thus a faithful component in all epithelia and plays a functional role that is essential for the * The work was funded by the Deutsche Forschungsgemeinschaft (DFG), Bonn. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. ‡ Supported in part by the Verband der Chemischen Industrie (Frankfurt).
§ To whom correspondence should be addressed. maintenance of the epithelial phenotype (Imhof et al., 1983;Behrens et al., 1989). Accordingly, homozygous mutations of E-cadherin introduced into mice by homologous recombination disturbed early embryogenesis: The individual cells of the morulae lose their morphologic polarization and do not form a blastocoel. The mutant embryos cannot leave the zona pellucida and, therefore, do not implant into the uterus (Larue et al., 1994;Riethmacher et al., 1995).
Since down-regulation of E-cadherin expression is a frequent event late in progression of human carcinomas and since modulation of E-cadherin expression plays a major role during development, we and others have begun to analyze the Ecadherin promoter and have found epithelial specificity in a fragment 178 bp upstream of the transcription start site Ringwald et al., 1991;Bussemakers et al., 1994). This promoter fragment contains a GC-rich region, a CCAAT-box, and a 12-bp palindromic element, which we named E-Pal. We have furthermore found that epitheliumspecific transcription correlates with factor binding to these elements in vivo and to a loosening of chromatin structure in the promoter region (Hennig et al., 1995). Other epitheliumspecific promoters have recently also been examined: The upstream regulatory regions of the epithelium-specific human papilloma viruses (HPV) 16 and 18 contain several binding sequences for ubiquitous cellular transcription factors (cf. Cripe et al., 1990;Hoppe-Seyler and Butz, 1994;Bernard and Apt, 1994). Epithelial specificity thus appears to be achieved by different combinations of these cellular factors. Specific factors have also been characterized that contribute to epithelial specificity; for instance, a mesenchyme-specific member of the NF-1 family represses transcription in fibroblasts but not in epithelial cells (Apt et al., 1993). Other epithelium-specific activators and repressors such as KRF-1, a coactivator of transcriptional enhancer factor-1 and YY1 have recently been described (Mack and Laimins, 1991;Ishiji et al., 1992, Bauknecht et al., 1992. Here we report that epithelium-specific expression of the E-cadherin gene is achieved by two different mechanisms; E-Pal in the upstream promoter activates or suppresses transcription in epithelial or mesenchymal cells, respectively. A new epithelium-specific enhancer (ESE) was discovered in the first intron of the gene, that enhances transcription in a tissuespecific manner and binds nuclear factors specifically in epithelial cells.

MATERIALS AND METHODS
Plasmid Constructs-The E-cadherin promoter-CAT constructs are derivatives of the deletion Ϫ178/ϩ92 cloned in the pCAT basic vector . To generate the Ϫ94 or Ϫ78 deletion constructs, a HindIII/PstI fragment (Ϫ178 to Ϫ79) was replaced by a doublestranded oligonucleotide, or the blunt-ended vector was directly religated, respectively. To obtain mutants of E-Pal, the CCAAT-box, or the GC-rich region, restriction fragments of the Ϫ178/ϩ92 construct were replaced by corresponding mutant oligonucleotides. The restriction sites PstI (Ϫ99), NarI (Ϫ54), EclXI (Ϫ22), or KspI (ϩ18), were used for the exchange. The CCAAT sequence was mutated to TCCG, and the AP-2 binding sites in subregions GCI and GCII were mutated to GC-CGTTTTC and TTCTGCGGG, respectively.
For analysis of the intronic enhancer, a 1.9-kilobase genomic BamHI fragment containing intron 1 and parts of intron 2 was ligated into the TK-CAT construct (pBLCAT2) (Luckow and Schü tz, 1987). Deletions of the enhancer were made with Exonuclease III (Pharmacia Biotech Inc.), and PCR fragments were ligated into TK-CAT. The AP-2 expression construct SPRSV AP-2 was kindly provided by Dr. T. Williams (New Haven, CT). In the construct AP-2⌬TA (Williams and Tjian, 1991), nucleotides 153-413 were removed. All sequences were confirmed by dideoxy sequencing.
Cells, Transfections, and CAT Assays-Cells were cultured in Dulbecco's modified Eagle's medium containing 10% fetal calf serum. Cell lines are described in Frixen et al. (1991) and Behrens et al. (1991). For transfection experiments, cells were seeded at a density of 10 6 per 10-cm tissue culture dish and transfected with usually 5 g of the various promoter constructs using the calcium phosphate coprecipitation method (Gorman et al., 1982). The amounts of cell extracts for the CAT assays were adjusted according to ␤-galactosidase activity from cotransfected plasmids Rous sarcoma virus lacZ  or pCH110 (Pharmacia).

Characteristics of the E-cadherin Promoter-We and others
have previously analyzed the E-cadherin promoter and found epithelial specificity in a Ϫ178 bp fragment upstream of the transcription start site , Ringwald et al., 1991, Bussemakers et al., 1994. This fragment contains a GC-rich region with subregions GCI and GCII, a CCAAT-box, and a 12-bp palindromic element that we named E-Pal (Fig. 1A). We demonstrate here that deletion or mutation of E-Pal resulted in a 5-fold increase of promoter activity in mesenchymal cells, e.g. fibroblasts or smooth muscle cells (Fig. 1B). Similarly, mutation of E-Pal leads to an increase (2-4-fold) of promoter activity in E-cadherin-negative carcinoma cells (Table I). A slight decrease of promoter activity was observed when E-Pal was mutated or deleted in E-cadherin-expressing carcinoma cell lines. In the mouse mammary epithelial cell line Ep-FosER, which looses E-cadherin expression following activation by estrogen, the E-Pal mutation also leads to an increase of promoter activity (Table I). These data demonstrate that E-Pal is a negative regulatory element in mesenchymal (E-

FIG. 1. Deletion and mutation of the E-Pal element increases E-cadherin promoter activity in mesenchymal cells.
A, schematic representation of the E-cadherin promoter indicating the elements E-Pal, the CCAAT-box, the GC-rich region (with subregions GCI and GCII), and the transcription start site (arrow). The sequences of E-Pal and a mutation involving the two central nucleotides (mut) are indicated. B, activities of promoter fragments (as indicated in panel A) in ras3T3 fibroblasts. Shown are results from duplicate experiments. cadherin-negative) cells but has a weak positive activity in epithelial cells.
We attempted to identify regulatory factors that control the E-cadherin promoter through the E-Pal element, by comparing the functional effects of specific mutations of E-Pal with the capacity of nuclear factor binding in gel retardation assays. Mutations in the center or in the 3Ј-half of E-Pal (mut 1 to mut 4) increased promoter activity in fibroblasts ( Fig. 2A), in contrast to a mutation of the 5Ј-side (mut 5). A specific nuclear factor from fibroblasts was found to bind to the E-Pal element in gel retardation assays (Fig. 2B), which could be competed only by the wild-type oligonucleotide and the one mutated in the 5Ј-half of E-Pal. This suggests that binding of a specific factor (repressor) of fibroblasts to E-Pal correlates with suppression of promoter activity. However, we are aware of the fact that a similar band shift is seen when nuclear extracts from epithelial cells are examined (not shown, but see Behrens et al. (1991)). We next examined the contribution of the CCAAT-box and the GC-rich region to transcriptional activity of the Ecadherin promoter. Mutation of either the core sequence of the CCAAT-box or one of the two consensus binding sites for the transcription factor AP-2 (cf. Williams and Tjian, 1991) strongly reduced promoter activity in epithelial cells (Fig. 3A). These mutations also reduced the activity of the promoter with a mutated E-Pal in fibroblasts (Fig. 3B). Mutation of all three elements completely abolished promoter activity. These data show that both the CCAAT-box and the GC-rich region represent positive regulatory elements in both epithelial and mesenchymal cells.
The GC-rich region of the E-cadherin promoter binds the transcription factor AP-2, as revealed by footprint analysis (Fig. 4). Mutation of each of the two AP-2 binding sites in the subregion GCI or GCII narrowed the footprint with both recombinant AP-2 and nuclear extracts on the respective sides. Footprint formation at both sites was inhibited by an oligonucleotide containing the AP-2 binding site of the SV40 enhancer (data not shown; cf. Imagawa et al. (1987)). Furthermore, a cotransfected dominant-negative mutant of AP-2 that lacks the transactivation domain inhibited activity of both the Ϫ178 and Ϫ58 bp promoters in a concentration-dependent fashion (Table II). These data indicate that AP-2 or a closely related factor regulate the E-cadherin promoter by binding in a tandem arrangement to the GC-rich region.
Characteristics of the Intronic Enhancer of the E-cadherin Gene-In a recent in vivo analysis, we identified DNase I-hypersensitive sites in the first and second introns of the Ecadherin gene that are specific for E-cadherin-expressing cells (Hennig et al., 1995). The respective genomic regions were here examined for enhancer activity on the TK minimal promoter (Fig. 5). The used 1.9-kilobase BamHI fragment of the first and part of the second intron enhanced promoter activity 4.5-fold in epithelial cells; deletion of intron 2 sequences to position 672 led to a further (12.5-fold) enhancement of activity, which may indicate the presence of a negative regulatory region between positions 1100 and 672. Further deletion abolished activity, thus localizing the enhancer to a region of 200 bp of intron 1 (compare the activity of constructs 672 and 475 in Fig. 5). Similar results were obtained when the deletion fragments  Frixen et al. (1991).
b E-cadherin promoter activity was determined relative to the activity of the SV40 promoter/enhancer as described .
c The E-Pal mutation (mut) is described in Fig. 1. Activity is expressed relative to the Ϫ178 bp construct.
d The T 24 bladder carcinoma cell line was kindly provided by Dr. J. Schalken (Nijmegen). The related EJ 28 bladder carcinoma cell line showed high promoter activity as in differentiated carcinoma cells (not shown). e The mouse mammary epithelial cell lines were kindly provided by Dr. H. Beug (Institute for Molecular Pathology, Vienna, Austria). The cell lines express FosER fusion protein that either was not activated or was activated by estrogen.

FIG. 2. Activity and binding specificity of the E-Pal element.
A, effect of various mutations of E-Pal on the activity of E-cadherin promoter-CAT constructs in ras3T3 fibroblasts. Results are expressed relative to the CAT activity of the wild-type Ϫ178 bp promoter. B, binding of a specific nuclear factor from ras3T3 fibroblasts to the E-Pal oligonucleotide (arrowhead) in a gel retardation assay and competition with an excess of unlabeled oligonucleotides (cf. panel A). Ϫ, no competitor. Gel retardation assays were performed as described in Behrens et al. (1991). were examined with the SV40 minimal promoter. Enhancer activity was also observed when the 1.9-kilobase fragment was tested in the reverse orientation (in the TK and SV40 promoter) or at a distance of 2.7 kilobases from the transcription start site of the SV40 promoter (data not shown).
Footprint analysis of the enhancer in intron 1 revealed bind-ing of nuclear factors to three subregions, E I to E III (Fig. 6).
The sequences E II and E III are specifically protected by nuclear extracts from E-cadherin-expressing carcinoma cell lines (MCF-7 and RT 112) but not from E-cadherin-negative carcinoma cells (MDA-MB-231 and T 24); region E I was protected by extracts from both cell types. DNA sequencing of the protected areas revealed that E I to E III are GC-rich; E II and E III contain sequences that match the AP-2 consensus site ( Fig. 7; cf. also Williams and Tjian (1991)). The contribution of the individual elements E I to E III to the function of the intronic enhancer was examined by using various PCR-generated subfragments (Fig. 8): a 149-bp fragment comprising all three elements exhibited strong enhancer activ-  The dominant-negative variant of AP-2 was generated by deleting the cDNA sequences coding for the transactivation domain (Williams and Tjian, 1991). Cotransfections were performed with E-cadherin promoter-CAT constructs, increasing amounts of dominant-negative AP-2, and decreasing amounts of control vector (total transfected cDNA, 5 g).
c Activities of the Ϫ178 bp and Ϫ58 bp E-cadherin promoter-CAT constructs in the presence of control vector without the AP-2 sequence were set to 100%. ity in both orientations on the TK promoter in MCF-7 epithelial cells. Two copies of the enhancer showed 40-fold enhancement of activity (not shown). Interestingly, removal of any of the individual elements abolished enhancer activity, indicating that the integrity of the whole E I to E III cluster is sufficient and necessary to confer enhancer activity. The fragment of 149 bp also confers enhancer activity to the TK promoter in several other E-cadherin-expressing cell lines (Table III). No enhancer activity was found in fibroblasts and dedifferentiated carcinoma cells. We therefore named the 149-bp region ESE, i.e. epithelium-specific enhancer. Detailed footprint analysis of the ESE element with various nuclear extracts revealed general protection of the elements E I to E III in E-cadherin-expressing carcinoma cells; the element E I is protected in E-cadherinnegative carcinoma cells, and the elements E II and E III are protected in fibroblasts (Table III). Region E I contains a consensus binding sequence for the transcription factor H4TF-1 (cf. Dailey et al. (1988); we have not examined this factor any further). Binding of nuclear factors of MCF-7 cells to both regions E II and E III was competed by an AP-2 but not by an SP-1 binding site oligonucleotide ( Fig. 9; Kadonaga et al. (1987)). Oligonucleotides containing the sequences of either E II or E III interfered with factor binding to both regions. Moreover, regions E II and E III were also protected by recombinant AP-2 (Fig. 9). DISCUSSION We report here that epithelium-specific regulation of the E-cadherin gene is controlled by two different mechanisms. First, the E-Pal element in the promoter 5Ј of the transcription start site acts as a positive or negative element in epithelial or mesenchymal cells, respectively. Second, a tissue-specific enhancer (ESE) in the first intron promotes transcription exclusively in epithelial cells. We also demonstrate that both these mechanisms are disturbed in carcinoma cells that have progressed to a less differentiated state and are E-cadherin-negative. We suspect that the E-cadherin gene is activated and repressed by similar mechanisms during mesenchymal-epithelial transitions in development.
Restriction of E-cadherin Promoter Activity to Epithelial Cells Is in Part Due to Cell Type-specific Repression-We describe here three cis-acting elements in the E-cadherin promoter that exhibit positive regulatory activity in epithelial cells: E-Pal, the CCAAT-box, and a GC-rich region. We have previously shown that the E-Pal element also confers positive regulatory activity to a SV40 promoter in epithelial cells . In nonepithelial cells, CCAAT-box and GCrich region exhibit positive transcriptional potential, but this activity is masked by E-Pal, i.e. becomes apparent only when E-Pal is mutated or deleted. We also provide here evidence that a specific factor binds to E-Pal in fibroblasts and that binding specificity of this factor correlates with biological activity. Transcriptional repressors that act via binding to a specific promoter element have been described by others (see . Localization of the intronic E-cadherin enhancer by deletion analysis. Left, deletion fragments in front of the TK promoter. Right, promoter activity of the deletions in MCF-7 epithelial cells; basal activity of the TK promoter is set to 1.0. Numbers in brackets indicate distances in bp from the BamHI site of intron 1. The epithelium-specific enhancer is marked by a shaded ellipse. FIG. 6. DNase I footprinting analysis of the E-cadherin enhancer. A fragment of intron 1 from position 672 to the KpnI site of intron 1 (cf. Fig. 5) was used for footprinting analysis using nuclear extracts from E-cadherin expressing (MCF-7 and RT 112) and nonexpressing cells (MDA-MB-231 and T 24) as indicated. G and G ϩ A, Maxam-Gilbert sequence reactions; Ϫ, absence of extract. The protected regions E I to E III are marked on the right. itz (1990) and Johnson (1995) for reviews). For instance, the Wilms' tumor suppressor gene wt1 suppresses transcription by binding to a GC-rich promoter sequence, and a protein domain of WT1 could be identified that retained inhibitory activity when combined with a heterologous DNA-binding domain (Madden et al., 1991;Drummond et al., 1992). Furthermore, the inhibitory activity of WT1 depends on p53; in its absence, WT1 acts as a transcriptional activator (Maheswaran et al., 1993). Thus, the dual function of WT1 is reminiscent of the opposite activities of E-Pal and its putative binding factor in epithelial and mesenchymal cells.
Both half-sites of E-Pal are similar to the consensus sequence for binding of helix-loop-helix (HLH) transcription factors; it is therefore suggested that HLH factors play a role in the regulation of the E-cadherin promoter by the E-Pal element. Mutation analysis showed that the right HLH binding motif in E-Pal is necessary for function, whereas integrity of the left HLH site is less essential. Interestingly, only the right half of E-Pal is conserved in the human E-cadherin promoter (Bussemakers et al., 1994). We have also found that the HLH proteins Myc and Max bind to E-Pal in vitro; however, no modulation of the activity of the E-cadherin promoter could be demonstrated (data not shown). HLH proteins have frequently been implicated in tissue-specific gene expression (Weintraub et al., 1991;Lee et al., 1995). It is therefore conceivable that yet unknown HLH transcription factors play a role in epitheliumspecific regulation of the E-cadherin promoter. Interestingly, expression of the muscle-specific HLH transcription factor myoD in keratinocytes induces morphological dedifferentiation of the cells and loss of epithelial markers (Boukamp et al., 1992).
We also demonstrate here that the CCAAT-box and the GCrich region are positively acting elements in the E-cadherin promoter and have little cell type specificity of their own. CCAAT-box binding proteins that serve tissue-specific functions have been described previously (Umek et al., 1991;Katz et al., 1993). We found three complexes of factors that bind to the CCAAT-box in gel retardation assays, and these did not appear when the CCAAT sequence was mutated. Competition by specific oligonucleotides revealed that various members of the CCAAT-box family of proteins such as CP-1 and C/EBP (Chodosh et al., 1988;Landschulz et al., 1988) are candidates for activation of the E-cadherin promoter (data not shown). We have not analyzed the CCAAT-box of the E-cadherin promoter any further. Binding and functional data suggest that the transcription factor AP-2 or closely related factors regulate the activity of the GC-rich region of the E-cadherin promoter. (i) In DNase I footprint assays, two subregions GCI and GCII were identified that bound purified AP-2 with similar characteristics as factors present in nuclear extracts. Binding to both sites GCI and GCII could be competed with a single AP-2 specific oligonucleotide. (ii) In transient transfection experiments, AP-2 that lacks the transactivation domain suppressed the activity of the E-cadherin promoter in a concentration-dependent manner. Cis-elements representing AP-2 binding sites have recently been implicated in the regulation of promoters of various epidermal cytokeratins (Leask et al., 1991, Snape et al., 1991, Byrne and Fuchs, 1993. However, it is unlikely that AP-2 has an exclusive role in epithelium-specific gene expression since it is found in ectodermal derivatives but not in several other epithelial tissues that express E-cadherin (Mitchell et al., 1991).
Properties of the Novel ESE in the E-cadherin Gene-The ESE of the E-cadherin gene was discovered by the analysis of DNase I-hypersensitive sites followed by analysis of deletion constructs and PCR fragments for enhancer activity. The enhancer region corresponds to hypersensitive site 4 in the first intron of the E-cadherin gene, as described previously (Hennig et al., 1995). Transcriptional enhancers are generally associated with loosened chromatin structure, as in the case of the SV40 and polyoma virus enhancer; these enhancers are characterized as nucleosome-free regions and are hypersensitive toward DNase I (Serfling et al., 1985). The minimal region of the ESE of the E-cadherin gene (150 bp) activates transcription from heterologous promoters by a factor of 15 in MCF-7 cells (40-fold when used in a tandem arrangement) that is independent of orientation and distance. Detailed analysis of the ESE revealed three elements, E I to E III, which are all GC-rich. These elements are necessary and sufficient for enhancer activity in several E-cadherin-expressing epithelial cell lines. In contrast, the ESE is inactive in all E-cadherin-negative cell lines such as fibroblasts and dedifferentiated carcinoma cells. The ESE therefore has a modular structure, which is characteristic for several other enhancers, e.g. the CD2 or SV40 enhancers (Serfling et al., 1985;Lake et al., 1990). In the epithelium-specific enhancer of papilloma viruses, numerous elements contribute to enhancer activity, but the involved sequences are generally bound by ubiquitous transcription factors (see Hoppe-Seyler and Butz (1994) for a review). We found that the ESE of the E-cadherin gene is active in cells that FIG. 7. Genomic sequence of the Ecadherin enhancer region. The sequence encompasses the KpnI site of intron 1 (position 311, cf. Fig. 5), the regions E I to E III (boxes), and part of exon 2 (boldface letters). Numbering is from the BamHI site of intron 1. Arrows indicate primers used for PCR cloning of enhancer fragments. The putative AP-2 and H4TF-1 binding sites are marked by dots and a dashed line, respectively. The sequence of the ESE has been submitted to the EMBL data bank, accession number X90561.
FIG. 8. Fine analysis of the E-cadherin enhancer. PCR-generated fragments containing the elements E I to E III (left) were tested for enhancer activity in combination with the TK promoter in epithelial MCF-7 cells. Positions are from the BamHI site of the first intron. Activities are given as -fold induction with respect to the minimal TK promoter (right). express all three factors that bind to E I to E III, but is inactive when one factor is lacking. We also demonstrated that AP-2 or related factors bind to E II and E III; again epithelium-specificity cannot be explained by the action of AP-2 only. It is possible that recently identified splice variants of AP-2 (Meier et al., 1995) play a role in the specific regulation of ESE.
The expression of the E-cadherin gene is thus under the control of an epithelium-specific promoter plus an epitheliumspecific enhancer. We suggest that the combination of these two regulatory mechanisms provides additional specificity and strength of expression of the E-cadherin gene. This is thus different to the regulation of the related L-CAM gene (a chicken homologue of E-cadherin). There, a low activity and nonspecific promoter gains tissue specificity when combined with an enhancer in the second intron (Sorkin et al., 1993). In the Pcadherin gene, an enhancer is also located in the second intron (Hatta et al., 1994). 2 In addition promoter and enhancer of the E-cadherin gene act differently in mesenchymal cells. Here the enhancer is inactive, but the promoter contributes to repression through the E-Pal element. To our knowledge, a combination of an epithelium-specific enhancer with an epitheliumspecific promoter that acts as a repressor in mesenchymal cells has not previously been found. Ϫ Ϫ ϩ ϩ 1.3 a The cell lines and the endogenous E-cadherin expression are as described in Frixen et al. (1991). b Presence (ϩ) and absence (Ϫ) of factor binding to subregions E I to E III as determined by DNase I footprint analysis (cf. Fig. 6). The LX-1 lung cell line showed no enhancer activity and an absence of factor binding to E II and E III.
c Stimulation of TK promoter activity by the E-cadherin enhancer (cf. Fig. 5). The data are from two independent experiments using double values.
FIG. 9. Analysis of AP-2 binding to the E-cadherin enhancer by in vitro footprinting. Footprinting analysis was performed using nuclear extracts from epithelial MCF-7 cells or recombinant AP-2 protein. Competitor oligonucleotides with binding sites for AP-2 and SP-1 as well as E II and E III are indicated. Ϫ, no oligonucleotide or extract was added. Protected regions E I to E III are marked on the left.