Cloning and Characterization of the Human b 4 -Integrin Gene Promoter and Enhancers*

The cell-surface adhesion molecule a 6 b 4 -integrin is a receptor for laminins and a component of hemidesmosomes. b 4 -Integrin expression is restricted to proliferat- ing basal keratinocytes in the epidermis and is suppressed when differentiation commences. Altered b 4 - integrin expression levels correlate significantly with the aggressive behavior of cancers. In order to clarify the mechanisms that regulate transcription of the b 4 - integrin gene, we cloned its 5 * -flanking region. This 5 * flanking region was found to have a high G 1 C content and not to contain either TATA or CAAT boxes. Nested delimitation and reporter analyses mapped a basal promoter to nucleotides 2 106 to 1 105, surrounding the most proximal transcription initiation site. Gel retardation and mutational analyses revealed that cooperation between AP1 and Ets, interacting with other factors, mediated the promoter activity. In addition to the promoter element, enhancer activity was found in the first intron ( 1 1905/ 1 3933) and in a sequence upstream of the promoter region ( 2 414/ 2 107). These findings should facilitate our understanding of the regulation of b 4 -inte- grin gene expression in processes such as cell growth and differentiation, apoptosis, and cancer development and metastasis.

Integrins comprise a superfamily of cell-cell and cell-matrix adhesion molecules, each consisting of a dimer of specific ␣and ␤-subunits (1). The ␤4-subunit specifically associates with the ␣6-subunit and functions as a receptor for laminins 1, 2, 4, and 5 (2). ␣ 6 ␤ 4 -Integrin is expressed at the basal surface of most epithelia and binds to the anchoring filaments as a component of the hemidesmosome, a rigid adhesive microstructure (3). The functional importance of ␤ 4 -integrin in cell adhesion was demonstrated by a recent study on ␤4-knock-out mice, in which the hemidesmosomes were disrupted, resulting in erosion and blistering of the skin (4,5). Patients with a hereditary skin blistering disorder called junctional epidermolysis bullosa with pyloric atresia show reduced numbers, or even absence, of hemidesmosomes in the skin due to mutation of this gene (6). ␤ 4 -Integrin expression has been found to be restricted to proliferating basal keratinocytes and to be suppressed when differentiation commences (7). ␤ 4 -Integrin was initially identified as a tumor-associated antigen, defined by the monoclonal antibody A9 raised against human squamous cell carcinoma (8,9), in which ␤ 4 -integrin is no longer restricted to the epidermal basal cell layer but extends to the suprabasal and upper layers (8,10,11). Altered ␤ 4 -integrin expression during pulmonary, pancreatic, and uterine cervical carcinogenesis has also been observed (12)(13)(14). An extreme example of altered expression is seen in thyroid carcinomas; normal thyroid follicular cells do not express ␤ 4 -integrin at all, but thyroid cancer cells contain high levels (15). An association between enhanced ␤ 4 -integrin expression and early recurrence of squamous cell carcinoma has been demonstrated (8), and ␤ 4 -integrin expression levels have been found to correlate with the degree of invasiveness of colon carcinoma cells (16) and with high metastatic potential of lung cancer and melanoma cell lines (17). Introduction of ␤ 4 -integrin into colon carcinoma cells enhances their ability to invade laminin matrices (18).
In contrast to other ␤-subunits, ␤ 4 -integrin has an exceptionally large cytoplasmic domain, which associates with cytoskeletal and signaling molecules (19), and it regulates cell proliferation through Ras-mitogen-activated protein kinase signaling pathways (20). Recently, ␣ 6 ␤ 4 -integrin was found to activate phosphatidylinositol-3-OH kinase pathways upon binding to laminin-1 (21). The enhanced cell migration resulting from formation and stabilization of actin-containing structures may account for the aggressive behavior of ␤ 4 -integrinexpressing cancers (22).
By contrast, ␤ 4 -integrin expression has been shown to be reduced in basal cell carcinoma of the skin (23) and adenocarcinoma of the breast (24) and prostate (25). Invasive types of urinary bladder carcinoma have been found to have lost colocalized ␤ 4 -integrin and collagen-VII (26), and transfection of ␤ 4 -integrin into human urinary bladder carcinoma cells has been found to inhibit cell growth and lead to apoptosis (27) by inducing a cell cycle inhibitor, p21 (28). These findings indicate that ␤ 4 -integrin has complicated dual functions.
Although several studies have shown the significance of altered ␤ 4 -integrin expression, the mechanism responsible for regulating ␤ 4 -integrin expression has not been elucidated. ␤ 4 -Integrin expression seems to be regulated, at least partly, at the transcription level (7). In order to elucidate the mechanism responsible for ␤ 4 -integrin gene expression and the transcriptional basis for cancer invasion and metastasis, we cloned the 5Ј-flanking region of the ␤ 4 -integrin gene and identified several cis-acting elements. Below, we report that cooperative transactivation by AP1, Ets, MyoD, and other factors is involved in the complicated transcriptional regulation of the ␤ 4 -integrin gene.

EXPERIMENTAL PROCEDURES
Cloning and Nucleotide Sequencing of the ␤ 4 -Integrin Gene 5Ј-Flanking Region-A human placental genomic DNA library constructed in the phage vector -FIX II (Stratagene, La Jolla, CA) was screened by plaque hybridization using a 623-bp 1 fragment corresponding to nucleotides 287-909 of ␤ 4 -integrin cDNA (29). One positive plaque was identified, and its DNA insert was subjected to characterization by restriction endonuclease mapping and Southern blot analysis. Digestion with HindIII restriction endonuclease and Southern blot hybridization with a synthetic oligonucleotide specific to the 5Ј-end of ␤ 4integrin cDNA (5Ј-ATCAGCGTCAGCCTCTCTGG-3Ј) revealed that an 8.8-kb fragment contained the 5Ј-flanking region of the ␤ 4 -integrin gene. This 8.8-kb fragment was subcloned into the HindIII site of pUC118 (the resulting plasmid was designated p8.8Hind) and sequenced with an ABI 377 autosequencer (Perkin-Elmer).
Northern Blot Analysis-Total RNA (15 g/lane) was fractionated by electrophoresis and transferred to Hybond N (Amersham Corp., UK) by capillary blotting. Hybridization was performed by using a radiolabeled human ␤ 4 -integrin cDNA fragment (nucleotides 287-909) (29). The quality and quantity of the electrophoresed RNA were determined by rehybridization of the same blot with a 1.0-kb human glyceraldehyde-3-phosphate dehydrogenase (G3PDH) cDNA (30). The signal intensity of both the ␤ 4 -integrin and G3PDH autoradiograms was measured with an image densitometer (model GS-700; Bio-Rad). The relative expression level of ␤ 4 -integrin was calculated as the ratio to G3PDH, after subtracting the background signals.
Primer Extension Analysis-Primer extension was carried out as described previously (31). The antisense oligonucleotide PE-0 (5Ј-GCGCGGGCGAGCAGGGACTGTCCGAGGGCGGGCGC-3Ј) was hybridized with 3 g of mRNA from ␤ 4 -integrin-expressing human epidermoid carcinoma cell line A431 (29). The first strand cDNA was synthesized with SuperScript II reverse transcriptase (Life Technologies, Inc.) at 50°C for 30 min in the presence of 0.5 Ci/l [␣-32 P]dCTP, and the fragments produced were analyzed by 6% denaturing sequencing gel (Long Ranger, FMC BioProducts, Rockland, ME) electrophoresis.
Cell Lines-Five human urinary bladder carcinoma cell lines, DAB1, UMUC6-dox, EJ, KU7, and UMUC2 (32)(33)(34), were used. The ␤ 4 -integrin-expressing human epidermoid carcinoma cell line A431 (29) and three human fibroblast cell lines, WI38, HFL-1, and NB1RGD, which do not express ␤ 4 -integrin, were obtained from Riken Cell Bank (Tsukuba, Japan) (35). All of the cells were maintained at 37°C under a humidified atmosphere of 5% CO 2 and 95% air. The urinary bladder carcinoma cells and A431 were maintained in Dulbecco's minimal essential medium supplemented with 10% fetal calf serum, and the fibroblast cell lines were maintained in minimal essential medium supplemented with 10% fetal calf serum.
Transfection and Dual Luciferase Assay-Cells were seeded at a density of 3-6 ϫ 10 4 per 16-mm well and transfected 24 h later with complexes containing 1.2-1.9 l of LipofectAMINE, 4 l Plus reagent (Life Technologies, Inc.), 42 ng pRL-TK vector (Promega, Madison, WI), which contained the Renilla luciferase gene as a transfection efficiency control, and 420 ng of firefly luciferase reporter plasmid per well. Lysates were prepared 24 -48 h after transfection by adding 100 l of reporter lysis buffer (dual luciferase reporter system, Promega), and their luciferase activity was determined with an analytical luminometer (model TD-20/20, Turner Designs, Sunnyvale, CA) and expressed as relative luciferase units (RLU), calculated by determining the ratio of the intensity of the light produced by the firefly luciferase reporter plasmid to that produced by the Renilla luciferase pRL-TK plasmid. Triplicate luciferase assays of all of the samples were performed.
DNase I Footprinting Assay-DNase I footprinting was performed with a Core Footprinting System (Promega) according to the manufacturer's instructions. Briefly, 10 -20 fmol of DNA probes (P1, Ϫ137/ϩ114 and P2, Ϫ137/ϩ232) labeled at only one end were incubated with 0 -160 g of nuclear extract or 0.55 g of recombinant c-Jun protein (Promega) for 10 min on ice. Freshly diluted DNase I was added for 1-2 min at room temperature. The reaction was stopped, phenol:chloroform-extracted, and ethanol-precipitated. The dried DNAs were suspended in formamide loading dye and resolved on a 6% denaturing sequencing gel.

Structure of the 5Ј-Flanking Region of the Human ␤ 4 -Integrin
Gene-An 8.8-kb HindIII fragment of a ␤ 4 -integrin genomic clone was isolated from a human placental genomic library and found to contain an 8.5-kb sequence upstream of the translation initiation site and a 5.2-kb 5Ј-flanking region preceding two exons that were separated by the first intron, which was approximately 3.2 kb (Fig. 1). Exon 1 contained 159 untranslated nucleotides, including the previously described 5Ј-end of ␤ 4 -integrin cDNA (29), and exon 2 contained 89 nucleotides, including the translation initiation codon.
The transcription initiation site was determined by primer extension analysis using a primer antisense to the known 5Јend of ␤ 4 -integrin cDNA, PE0 (dotted arrow in Fig. 2). Multiple elongation products were identified from the mRNA of ␤ 4integrin-expressing human epidermoid carcinoma cell line A431 (Fig. 3, lane 1), but they were not produced when the primer extension reaction was carried out without mRNA (lane 2). We also confirmed these results with the mRNA of human lung carcinoma cell line PC10, which expressed a high level of ␤ 4 -integrin (data not shown). Hereafter, we refer to the most proximal transcriptional initiation site at the cytosine residue of the ␤ 4 -integrin gene as ϩ1 (arrow in Fig. 2).
The nucleotide sequence of the ␤ 4 -integrin gene 5Ј-flanking region did not show either TATA or CAAT boxes, and the sequence between nucleotides Ϫ248 and ϩ1 had a high G ϩ C content (80%). A data base search of the 5Ј-flanking region revealed the presence of several putative binding sites for transcription factors, including AP1, Ets, MyoD, NFkB, and SP1, as shown in Fig. 2.  7), and NB1RGB (lane 8) did not express detectable amounts. These findings agreed with the ␤ 4 -integrin protein expression results (data not shown), suggesting that ␤ 4 -integrin expression in these cells is regulated at the transcription level. The results of the analyses of ␤ 4 -integrin protein expression and the biological behavior of these urinary bladder carcinoma cell lines will be described elsewhere. 2 Transcriptional Activity of the Human ␤ 4 -Integrin Promoter-In order to assess the promoter activity of the ␤ 4 -integrin 5Ј-flanking region, nested deletion mutants were cloned into the firefly luciferase reporter vector pGL3-basic. Each resulting recombinant plasmid DNA was then transiently transfected into ␤ 4 -integrin-expressing DAB1 cells (Fig. 5), and the vector pGL3-basic, which has no promoter, was used as a control. Among the constructs containing 5197-(L5.5K), 2688-(L2), 1805-(L3), 1050-(L6), and 414-(L7) bp upstream sequences, L7 showed the highest relative luciferase activity. Luciferase activity decreased progressively to 25% of L7 as the sequence length increased, suggesting the presence of an upstream suppressor activity in Ϫ5197/Ϫ415. Removal of the upstream region (Ϫ414/Ϫ106) from L10 led to a 36% decrease in luciferase activity, suggesting the presence of a weak enhancer activity in this region. When the sequence as far as ϩ105 bp downstream was deleted (L11), luciferase activity fell to that of pGL3-basic. Deletion of the regions Ϫ106/ϩ105 and Ϫ35/ϩ105 from L5.5K (La) and L10 (Lc), respectively, virtually abolished their luciferase activity. These results indicate the presence of a basal promoter at Ϫ106/ϩ105. As expected, the activity of this promoter (determined by L10/L11 ratios) was high in the ␤ 4integrin-expressing DAB1 and UMUC6-dox cells and low in EJ, K7, UMUC2, WI38, and NB1RGB cells, which do not express ␤ 4 -integrin (Table I). The further delimited construct lacking the 3Ј-half (ϩ4/ϩ105) of L10, designated as F4 (Ϫ96/ϩ3), retained significant promoter activity compared with L11, but it had only about one-fifth the activity of L10 (Fig. 5), indicating that the downstream region from ϩ4 to ϩ105 is also necessary for full promoter activity. On the basis of these reporter assay results, we focused first on the promoter region between Ϫ106 and ϩ105, which was critical for ␤ 4 -integrin transcription.
In an attempt to identify transcription factors with the po-  Fig. 2). The extension reaction was also performed without mRNA, as a negative control (lane 2). The extension products were analyzed by electrophoresis on a 6% denaturing sequencing gel. The arrows indicate anticipated transcription initiation sites, and ϩ1 corresponds to the most proximal site at the cytosine residue of the ␤ 4 -integrin gene (arrows in Figs. 1 and 2). The four lanes to the left contain the M13mp18 sequencing product used as a size marker. tential to interact with this region (Ϫ106/ϩ105), four synthetic double-stranded DNA probes, designated F0 (Ϫ106/Ϫ36), F1 (Ϫ35/ϩ4), F2 (ϩ5/ϩ44), and F3 (ϩ39/ϩ108), were subjected to gel retardation assays. Each probe used contained one or two known consensus sequences, i.e. Ets and MyoD in F0, AP1 in F1, SP1 and Cdx in F2, and NFkB and SP1 in F3. Nuclear extracts of DAB1 and WI38, representative ␤ 4 -integrin-positive and negative cell lines, respectively, were analyzed. Several DNA-protein complexes (ࡗ in Fig. 6) were more abundant in the nuclear extract from DAB1 than from WI38 cells, whereas several rapidly migrating complexes (छ in Fig. 6) with F0 and F1 were more abundant in WI38 than in DAB1 cells.
The above reporter and gel retardation analyses suggest that multiple proteins bind to separate sites on the promoter and that some of these complexes were needed simultaneously for transcriptional activation of the ␤ 4 -integrin gene.
Binding of AP1, Ets, and MyoD to the ␤ 4 -Integrin Promoter Region-Further gel retardation assays were performed using probes D1, D2, and D3, which contained the consensus Ets-, MyoD-, and AP1-binding sites, respectively (Fig. 7). Competition assays using canonical and mutated consensus oligonucleotides were also performed in order to verify the identity of the factors in the DNA-protein complexes.
One of several DNA-protein complexes that bound to the sequence D1 (ૺ in lanes 2-5 in Fig. 7A) was subjected to competition by the unlabeled canonical Ets oligonucleotide, whereas mutated Ets (GAGAGAGT instead of GAGGAAGT) did not inhibit this binding reaction. When D2 was used as a probe, several complexes were observed, and the binding of one of them (ࡗ in lanes 7-10 in Fig. 7B) was inhibited by canonical Gel retardation assays were performed using F0 (Ϫ106/Ϫ36), F1 (Ϫ35/ ϩ4), F2 (ϩ5/ϩ44), and F3 (ϩ39/ϩ108) probes and nuclear extracts (NE) from DAB1 (D) and WI38 (W) cells. Lanes 1, 6, 11, and 16 contained control samples incubated without NE (denoted by (Ϫ)). 100-Fold molar excesses of competing oligonucleotides were added as indicated (F0 for lanes 3 and 5; F1 for lanes 8 and 10; F2 for lanes 13 and 15; and F3 for lanes 18 and 20). Complexes that were more abundant in DAB1 than in WI38 are indicated by ࡗ, and those that were more abundant in WI38 are indicated by छ.

FIG. 5.
Promoter activity of serial deletion constructs of the ␤ 4 -integrin gene. DAB1 cells were transfected with sequentially deleted reporter constructs. La is a construct of L5.5K lacking the nucleotide sequence from Ϫ106 to ϩ105, and Lc is a construct of L10 lacking the nucleotide sequence from Ϫ35 to ϩ105. F4 is a construct containing the nucleotide sequence from Ϫ96 to ϩ3. The pGL3-basic (denoted by Basic) is a firefly luciferase reporter vector lacking promoter activity. Transient transfection and luciferase assays were performed in triplicate, and the data were normalized to Renilla luciferase activity (RLU indicates relative luciferase units, i.e. the ratio of firefly luciferase activity to Renilla luciferase activity) and are expressed as means Ϯ S.D.

TABLE I Activity of ␤ 4 -integrin promoter in various human cell lines
The activity of each construct in each cell line is presented relative to the activity of the promoterless vector, pGL3-basic. The L10/L11 ratio is designated as promoter activity. The data represent the average of three independent experiments, each performed in triplicate, and corrected for transfection efficiency by the activity of the pRL-TK control plasmid carrying the Renilla luciferase gene. MyoD but not by mutated MyoD (CACGGTAAC instead of CACCTGCTG). When D3 was used as a probe, a major complex (छ in lanes 12-15 in Fig. 7C) was formed, and its binding was inhibited competitively by canonical AP1 but not by mutated AP1 (TGAGTTG instead of TGAGTCA). These results indicate that the binding sites for AP1, Ets, and MyoD were involved in the transcriptional activity of the ␤ 4 -integrin promoter.
The region surrounding the transcription initiation site was also examined by DNase I footprinting analysis. A footprint was clearly visible with recombinant c-Jun protein from Ϫ11 to ϩ7 bp (Fig. 8, lane 5). Although less apparent, a footprint extending from Ϫ10 to ϩ2 bp was also present with the crude nuclear extracts from DAB1 cells (Fig. 8, lanes 2-4). The protected region corresponds to the AP1 site.
In an attempt to identify the Jun and Fos family members that interacted with the AP1 site of the ␤ 4 -integrin promoter, the DAB1 nuclear extract was incubated with the D3 probe in the presence and absence of antibodies against Jun and Fos family members (Fig. 9). An antibody that reacts with the DNA binding domain common to Jun family members specifically In order to further validate the involvement of AP1 and Ets in the regulation of ␤ 4 -integrin transcription, we examined whether mutations in either the Ets-or AP1-binding sites, or both, affected the promoter activity of F4 (Ϫ96/ϩ3), because transcriptional regulation of a variety of genes by AP1 and Ets acting cooperatively has been reported (43)(44)(45)(46)(47). A mutation in Ets alone did not influence the transcriptional activity, whereas a mutation in the AP1 site reduced the activity by 35%, and mutations in both the Ets and AP1 sites reduced it by 56% (Fig. 10). Although complete suppression did not occur, AP1, in cooperation with Ets, is involved at least partly in ␤ 4 -integrin gene expression.
Enhancer of the ␤ 4 -Integrin Gene in the First Intron-As described above, intron 1 of the ␤ 4 -integrin gene comprised 3.2 kb and was located upstream of the translation start site in exon 2. Transcriptional regulation by their first intron has been reported for several genes (48 -51), and so we examined the enhancer activity of intron 1. The 3.2-kb FspI/HindIII fragment was inserted downstream of the luciferase gene in both forward and reverse orientations (L5.5K-F, L5.5K-R, L7-F, L7-R, L10-F, and L10-R). Transfection of the luciferase gene containing intron 1 in either orientation into DAB1 cells significantly enhanced their luciferase activity (Fig. 11). Six deletion fragments were constructed (Fig. 12) to better localize the region of the downstream enhancers, but only two of them, I3 (ϩ1905/ ϩ3566) and I5 (ϩ1905/ϩ2933), enhanced luciferase activity to  3-5). The specific complex detected and subjected to competitive inhibition by Ets is denoted by ૺ. B, probe D2 (Ϫ56/Ϫ30, containing two MyoD-binding sites) was incubated with (lanes 7-10) or without (lane 6) nuclear extract from DAB1 cells, and an unlabeled probe (D2) or a canonical (M) or mutated (mM) MyoD oligonucleotide was added as a competitor (lanes 8 -10). The specific complex detected and subjected to competitive inhibition by MyoD is denoted by ࡗ. C, probe D3 (Ϫ19/ϩ4, containing an AP1-binding site) was incubated with (lanes [12][13][14][15] or without (lane 11) nuclear extract from DAB1 cells, and an unlabeled probe (D3) or a canonical (A) or mutated AP1 (mA) oligonucleotide was added as a competitor (lanes 13-15). The specific complex detected and subjected to competitive inhibition by AP1 is denoted by छ. The arrowheads indicate the undetermined complexes that were not competitively inhibited by Ets, MyoD, or AP1. The long arrows indicate the nonspecific bands that were not competed for by the excess of unlabeled oligonucleotides the same as the radiolabeled probes. No competitor, indicated by Ϫ, was added to the samples in lanes 2, 7, and 12.
the same extent as the L10-F and L10-R constructs. Almost identical results were obtained with UMUC6-dox cells (Table  II). We therefore concluded that region I5 was responsible for the downstream enhancer activity of intron 1. This region contained various putative consensus sequences, AP1, Ets, SP1, NFkB, and MyoD. Further delimitation of I5, however, abolished its enhancer activity (data not shown), suggesting that, as with the promoter, the cooperation of multiple factors is necessary for enhancer activity of ␤ 4 -integrin. DISCUSSION Accumulated evidence has unveiled important roles of ␤ 4integrin in various physiological and pathological conditions, including cell differentiation, proliferation (11,20), adhesion and migration (3,18,21,52), apoptosis, and signal transduction (27,28,53). The exon-intron organization and boundary sequences flanking 41 exons of the human ␤ 4 -integrin gene have recently been elucidated (54,55), but the promoter region and mechanism that regulates transcription of this intriguing integrin were unknown prior to our study. The 5Ј-region of the human ␤ 4 -integrin gene possesses several features, such as the lack of TATA and CAAT boxes, a high G ϩ C content, and multiple transcription initiation sites, that are commonly seen in a variety of housekeeping genes (56). These features have also been found in several non-housekeeping genes, such as epidermal growth factor receptor (57), ␤ 1 -integrin (58), and DR-nm23, a member of the putative metastatic suppressor nm23-H1 gene family (59).
We mapped the basic promoter region of the ␤ 4 -integrin gene to Ϫ106/ϩ105 by nested deletion constructs. The basic promoter activities, determined by their L10/L11 ratios (Table I), consistently paralleled expression of the ␤ 4 -integrin gene in urinary bladder carcinoma cell lines (Fig. 4). The ␤ 4 -integrinexpressing DAB1 and UMUC6-dox cells showed high promoter activity, whereas the non-expressing EJ, K7, UMUC2, WI38, and NB1RGB cells had low promoter activity (Table I). Competitive gel retardation assays showed that AP1 and Ets bound to this basic promoter region and that introduction of mutations into the AP1-and Ets-binding sites partially reduced promoter activity, confirming that these two factors are involved in ␤ 4 -integrin transactivation to some extent. Cooperation between Ets and AP1 in transcriptional regulation of a variety of genes, including matrix metalloproteinase-1 (MMP) (44,45), MMP-3 (43,44), urokinase (46), and the mammary tumor suppressor gene Maspin (47), has been established. The simultaneous activation of these matrix-degrading proteases and a cell-matrix adhesion molecule, such as ␤ 4 -integrin, may act synergistically to cause cancer invasion and metastasis (21,22). AP1 comprises either a homo-or heterodimer of two Jun proteins or a heterodimer of one Jun and one Fos protein. Gel retardation assays using specific antibodies revealed that c-Jun, JunB, JunD, and a Fos family protein, Fra-2, bound to the ␤ 4 -integrin promoter. Zonal transition of Jun and Fos family members was demonstrated to play an important role in the regulation of epidermal differentiation genes (60), and consistent with our results, basal keratinocytes have been reported to express JunB, JunD, and Fra-2 (61). The Ets family of oncoproteins has been implicated in the generation of several types of cancer, and expression of the protooncogenes c-met (62) and c-erb-B2 (63) has been found to be regulated by transcription factors belonging to the Ets family. Overexpression of Ets-1 in invasive gastric carcinomas has been observed (64), and Ets-1 has been detected in endothelial cells, particularly during tumor vascularization (65). Another Ets family member, E1AF, regulates MMP-9 expression in human breast cancer cells and confers the invasive phenotype on them (66). However, the member of the Ets family that participates in ␤ 4 -integrin promoter activity remains to be identified.
By further delimitation of reporter constructs, we were able to narrow the minimal promoter region down to nucleotides Ϫ96/ϩ3 (F4 in Fig. 5). The fact that further division of this region failed to reproduce transcriptional activity (data not shown) indicates that multiple factors are required for full promoter activity. The construct lacking the 3Ј-half (ϩ4/ϩ105) of L10, designated as F4 (Ϫ96/ϩ3), retained significant promoter activity compared with L11, but its level was only about one-fifth that of L10 (Fig. 5), indicating that the downstream transcribable region from ϩ4 to ϩ105 is also necessary for full promoter activity. Gel retardation assays detected several DNA-protein complexes in this region (ࡗ, lanes 12 and 17 in Fig. 6), suggesting that other, undetermined factors are also present.
Complicated and interdependent interactions between multiple factors in other genes have been reported. For example, AP1, Ets1, and NFkB synergistically activate the transcription of granuclocyte-macrophage colony-stimulating factor (67). How this synergy is accomplished is not understood, but altered binding affinity, formation of new binding sites for additional factors, and a conformational change in DNA have been suggested (67,68). AP1 was demonstrated to mediate DNA bending upon binding (69,70), and the recruitment of a third molecule from outside by binding transcription factors is a fascinating mechanism that may account for the variety of target genes regulated by rather common transcription factors, such as AP1, although a great deal of further research will be needed to verify this hypothesis. Wild-type and mutant promoter constructs of F4 were transfected into DAB1 cells (mE, mutant Ets-binding site; mA, mutant AP1-binding site; mEϩA, mutant Ets-and AP1binding sites). The wild-type and mutant sequences are indicated on the left, the Ets-and AP1-binding sequences are underlined (37,38), and the mutated nucleotides are written in italics. Triplicate luciferase assays were performed, and the data were normalized to Renilla luciferase activity (RLU) and expressed as means Ϯ S.D.
It is possible that ␤ 4 -integrin-negative cells possess molecules that suppress the promoter activity of ␤ 4 -integrin. Our gel retardation assays revealed that certain DNA-protein complexes were more abundant in nuclear extracts of WI38 cells (छ in Fig. 6) than of DAB1 cells, although it should be noted that the different DNA-protein complexes in these cell lines may reflect the protein contents of the nuclear extracts.
The presence of separate enhancers is an interesting feature of ␤ 4 -integrin gene regulation. We discovered that intron 1 possesses enhancer activity, and we tried to narrow down the enhancer region by preparing delimited constructs. The results showed that further delimitation to less than 1 kb abolished the enhancer activity, which, as observed with the promoter region, seems to be dependent on multiple factors within this 1-kb region. In addition to the downstream enhancer activity, there appears to be a modest upstream enhancer located between nucleotides Ϫ414 and Ϫ107. There are several reports about transcriptional regulation by multiple enhancers (49,51,71,72), and cooperative regulation of certain genes through their physically distant upstream and intronic enhancers has been reported (49,51). The upstream enhancer of the epidermal growth factor receptor gene is necessary for the downstream enhancer to function, whereas the two enhancers of the ␤ 4 -integrin gene seem to function independently. Furthermore, some inhibitors further upstream of the promoter in the region from Ϫ5197 to Ϫ415 seem to be present in the ␤ 4 -integrin gene. The mechanisms that regulate transcription of the extracellular matrix-associated glycoprotein osteonectin and ␤ 4 -integrin are similar (51), as osteonectin has an upstream suppressor, an upstream enhancer, and a downstream enhancer in intron 1.
In addition to the mechanisms described above, the altered expression of ␤ 4 -integrin observed in carcinoma cells may be due to genetic and epigenetic changes. Sequence analyses of the promoter regions have revealed nucleotide alterations (G to C) at position Ϫ57 in DAB1, EJ, and UMUC2 (data not shown), but they showed no correlations with the ␤ 4 -integrin expression level and were suggested to be due to polymorphism. Hypermethylation of the cytosine residues in the promoter was TABLE II Activity of ␤ 4 -integrin upstream and downstream enhancers in DAB1 and UMUC6-dox The activity of each construct in each cell line is presented relative to the activity of the promoterless vector, pGL3-basic. The ratio L5.5K-F or L5.5K-R/L5.5K, L7-F or L7-R/L7, and L10-F or L10-R/L10 are designated as lower enhancer activity. The upstream enhancer activity is represented by the L7/L10 ratio. The data represent the average of three independent experiments, each performed in triplicate, and corrected for transfection efficiency by the activity of the pRL-TK control plasmid carrying the Renilla luciferase gene.  11. Enhancer activity within the first intron of the ␤ 4 -integrin gene. The 3.2-kb first intron of the ␤ 4 -integrin gene was inserted into three representative constructs containing the core promoter (L10), upstream enhancer (L7), or full 5.5-kb upstream (L5.5K) region. The arrows indicate the directions of the intron relative to that of transcription of the ␤ 4 -integrin promoter (R and F, reverse and forward directions, respectively). The transcription initiation site is designated ϩ1, and the shaded boxes represent exons 1 and 2. Triplicate luciferase assays were performed, and the data were normalized to Renilla luciferase activity (RLU) and are expressed as means Ϯ S.D.
FIG. 12. Deletion analysis of the ␤ 4 -integrin downstream enhancer. The following restriction endonucleases were used to make these constructs: F, FspI; S, SacI; N, NdeI; T, Tth111I; and H, HindIII, and the digested fragments are designated I1, I2, I3, I4, I5, and I6, respectively. The forward and reverse directions of the inserted fragments relative to that of transcription of the ␤ 4 -integrin promoter are indicated by F and R, respectively. The transcription initiation site is designated ϩ1, and the shaded boxes represent exons 1 and 2. Triplicate luciferase assays were performed, and the data were normalized to Renilla luciferase activity (RLU) and are expressed as means Ϯ S.D. observed in ␤ 4 -integrin non-expressing KU7 and UMUC2 cells (data not shown), suggesting the involvement of epigenetic mechanisms in silencing the ␤ 4 -integrin gene. Further experiments are necessary to resolve this issue.
In conclusion, we have cloned and characterized the ␤ 4integrin promoter and found that AP1 and Ets, cooperating with other undetermined elements located in the separate promoter regions, regulate ␤ 4 -integrin expression. We have also shown the presence of enhancer elements in intron 1 and upstream of the promoter region, and a suppressor element upstream of the upper enhancer region. These findings should facilitate our understanding of the complicated mechanisms that regulate ␤ 4 -integrin transcription. In view of the important roles of ␤ 4 -integrin, our findings have prompted us to carry out further investigations on this adhesion molecule.