The Human a2 Integrin Gene Promoter IDENTIFICATION OF POSITIVE AND NEGATIVE REGULATORY ELEMENTS IMPORTANT FOR CELL-TYPE AND DEVELOPMENTALLY RESTRICTED GENE EXPRESSION*

The az& integrin serves as a collagen receptor or a collagedaminin receptor, depending upon cell type. Expression of the integrin is regulated during normal cel- lular differentiation and is altered during carcinogen-esis. We have previously demonstrated that increased expression of the az& integrin during megakaryocytic differentiation is a consequence of increased a2 mRNA due to transcriptional activation of the a2 integrin gene and that the decreased expression of the integrin in breast adenocarcinoma is due to decreased steady-state levels of az mRNA. We now report the identification and characterization of the 5‘-flanking region of the az integrin gene. The S’-untranslated region of the a2 mRNA extends 129 base pairs 5’ to the site of translation initiation. The promoter region lacks TATA and CAAT boxes but contains an abbreviated initiator sequence and six Spl binding sites. Consensus binding sites for AP-1 and AP-2 complexes, a GATA box, a Pu.1 box, and two palin-dromic motifs with potential to bind the estrogen recep- tor are also present. A 961-base pair fragment of the S’-flanking region directs both cell type- and differentia- tion-specific expression of a reporter gene in T47-D epithelial cells and in pluripotent hematopoietic KS62 cells upon megakaryocytic differentiation. The integrin


8950.
a2P1 integrin by the proliferative layers of squamous epithelium with the absence of a2 expression by the more differentiated superficial epithelial cell layers is one striking example of differentiation-dependent regulation of the a2P1 integrin (14). The normal high-level expression of the a2Pl integrin by columnar epithelium of ducts and ductules of normal human breast tissue is reduced in moderately differentiated adenocarcinomas of the breast and is more markedly diminished in poorly differentiated carcinomas (15).
To begin to discern the molecular mechanisms by which both the cell type and differentiation-dependent expression of the a2P1 integrin are regulated, we have employed a model of hematopoietic differentiation. The K562 cell, a leukemic cell line derived from a patient with chronic myelogenous leukemia in blast crisis, represents a pluripotent hematopoietic cell (16). K562 cells cultured in the presence of 40 I~M phorbol dibutyrate acquire megakaryocytic characteristics, including expression of the aIIbP3 and a2Pl integrins (17,18). We have recently demonstrated that the increased surface expression of the a2Pl integrin is a consequence of increased steady-state levels of a2 mRNA due to transcriptional activation of the a2 gene. The long-lived PI mRNAwas not altered during differentiation (19).
We now report the identification of the 5"flanking region of the a2 integrin gene, compare its structure to the other known integrin promoters, and demonstrate that cell type-specific promoter activity resides within the 5"flanking region of the gene.

MATERIALS AND METHODS Cell Cultures and Dansfection
Assays"T47-D and K562 cell lines obtained from the ATCC were propagated in RPMI 1640 medium either with or without 0.2 unitdml insulin, respectively. Megakaryocytic differentiation of K562 cells was induced by the addition of 40 m phorbol dibutyrate in dimethyl sulfoxide, as previously described (18).
The K562 cell line was transfected by eledroporation using a BTX Electro Cell Manipulator 600 (BTX Inc., San Diego, CA) (20). Approximately 1.0 x lo7 cells were transfected in RPMI medium containing 100 pg/ml salmon sperm DNA, 30 pg of plasmid DNA, and 3 pg of RSVluciferase' DNA by electroporation at 275 V and 600 pF. T47-D cells were transfected by calcium phosphate precipitation with 30 pg of plasmid DNA and 3 pg of RSV-luciferase DNA. Cell extracts were harvested after 48 h. Luciferase activity produced by 10 pl of the cell lysate in 190 pl of assay buffer (10 nm Mg[OAclz, 50 nm Tris-MES, pH 7.8, and 2 nm ATP) was analyzed using a Monolight 2010 luminometer (Analytical Luminescence Laboratory, San Diego, CA), as described (21), and was used to normalize for transfection efficiency. Cell extracts containing identical amounts of luciferase activity were then assayed for chloramphenicol acetyltransferase (CAT) activity by the standard method of Gorham et al. (22). The conversion of chloramphenicol to acetylated chloramphenicol was determined by both thin-layer chromatography and differential extraction. FIG. 1. Partial restriction enzyme map of exon-1 and the 6'flanking region of the a, integrin gene. The top line demonstrates the approximately 15-kb Sal1 fragment from the two identical bacteriophage clones (AGa2-l and AGa2-2). The internal 2.0-kb EcoRI genomic fragment hybridized with the 115-bp 5"partial cDNA probe and the oligonucleotide probe (az012). Comparative sequence analysis of the 2.0-kb EcoRI genomic fragment identified the 5'-untranslated region and exon 1, to the first intron. The arrow at +1 denotes the most 5' cluster of transcription start sites. The identified AuaII site which occurs 109 bp downstream in the 5'-untranslated region was utilized for construction of the S1 nuclease protection probe as well as the CAT a2 promoter-reporter gene constructs.
cDNA and Genomic Cloning-Phage clones containing the complete a2 cDNA were obtained from a human placental AGTll cDNA library, graciously provided by Dr. Evan Sadler (23). Genomic clones were isolated from a human lung fibroblast genomic library consisting of MboI partially digested DNA cloned into AFix (Stratagene, La Jolla, CA). Both genomic and cDNA libraries were screened with both a 32P-labeled 118-bp EcoWDmI az cDNA fragment (extending from position -47 to +71) and a 32P-labeled oligonucleotide (extending from position -5 to +16) of the published az cDNA sequence (24). These positions are given in relation to the ATG methionine of the published a2 cDNA sequence. Two identical positive clones (AGa,-l and AGaz-2) were obtained. Partial restriction enzyme mapping of positive phage clones was performed by standard techniques with single and double restriction enzyme digestion and Southern blot analysis (25). Blots were hybridized overnight to radiolabeled cDNA probes prepared by the random-primer method or oligonucleotide probes prepared by end labeling with T4 polynucleotide kinase, washed, and then subjected to autoradiography. The 2.0-kb EcoRI restriction enzyme fragment which hybridized to both cDNA and oligonucleotide probes was subcloned into plasmid SK-Bluescript (Stratagene, La Jolla, CA), and both strands were sequenced by the dideoxynucleotide chain termination method (26). This EcoRI fragment contains all of exon I, part of intron I, and approximately 1 kb of the 5"flanking region. Sections of this genomic fragment were used for S1 nuclease protection assays and to assemble the CAT reporter constructs described below. S1 Nuclease Protection and Primer Extension Analyses-S1 nuclease analysis was performed on total RNA isolated by guanidinium thiocyanate extraction and preparative ultracentrifugation, as described (27,28). A 1-kb 5' EcoRI-AuaII genomic fragment encompassing the 5'flanking sequence and 109 bp of the 5'-untranslated region of exon I was end-labeled with T4 polynucleotide kinase and 1 x lo5 cpm of probe was hybridized to 20 mg of total mRNA from K562 cells induced to differentiate with phorbol dibutyrate for 8 days, uninduced K562 cells, or REH cells for 16 h at 50 "C in a hybridization mixture containing 40 m Pipes, pH 6.4, 4 M NaC1, 1 m EDTA, and 80% formamide. As previously described, K562 cells induced for 8 days with phorbol dibutyrate express high levels of a2 mRNAin comparison to uninduced K562 cells, which express undetectable levels of a2 mRNA (19). REH cells, a B-cell lymphoblastic cell line, fail to express a2 mRNA by the reverse transcription-polymerase chain reaction (data not shown). The hybridization mixture was digested with 500 unitdml S1 nuclease in buffer containing 30 ~l l~ sodium acetate, pH 5.0,300 m NaCl, 4.5 m ZnS04, and 30 m g / d salmon sperm DNA for 1 h at 37 "C. Following ethanol precipitation, protected fragments were analyzed on a denaturing 6% polyacrylamide gel. A 5'-end-labeled probe for p-actin which protects a 170-bp fragment was analyzed in parallel as a control. Standards included 5'-end-labeled MspI-digested plasmid SK-Bluescript as well as a sequencing reaction primed with oligonucleotide az012. The 3'-end of oligo-a2012 differs by 2 bp from the AuaII site (position -20) used to prepare the S1 nuclease probe.
Primer extension was performed using the antisense oligonucleotide az012 (GACCGTAG'ITGCGCTGGGTT) (corresponding to positions -37 to -18 from the initiation ATG (Fig. 2)) end-labeled with T4 polynucleotide kinase and 2 x lo5 cpm of each nucleotide was hybridized to 5 pg of

A A A T C T C G G A T A T A G A A T G T T C T C T T T C C A G A T G T G A G A C T G A A A T A A T G
I ER I I I

C T C C T A A A C [ T C A G T C A I T T A A A T C A G C T C T T A / \ I T T A T C T C B
I I I I I M I   I  I  I  I  I

I I NFl-Ilk. A C C C G T C C C G T C C A G G C A G G A A A G C C [ T G C C A G G G C G C C Spl AP21
Sol I I I I

T G G T A G C G CTC AGT C A A G G T A A G C G G G G A T T T C
The nucleotide sequence of the 6'-end, W-flanking region, and exon-1 of the a, integrin gene. Exon 1 extends from the most 5' cluster of transcription start sites, denoted by +1, to the intronexon border, denoted by an arrow. The amino acid sequence is shown below the coding region of exon 1. Consensus sequences for binding to transcription factors are bored and labeled. These include binding sites for ubiquitous cell type-specific, and differentiatiomdependent transcription factors, including Spl, AP-1, and AP-2 sites, a GATA and a Pu box, two half-palindromic motifs with potential for estrogen receptor binding, and two partially-conserved CAAT-binding transcription factor-NF1 sites. The conserved TCA sequence of the Inr at the transcription start site is underlined with a solid line. The position of the oligonucleotide (a2012) used for primer extension analysis and sequencing is underlined with a dashed line. poly(A)+ RNA for 16 h at 35 "C in the hybridization buffer described above. The complementary DNA strand was extended with murine mammary tumor virus reverse transcriptase at 42 "C for 2 h. The length of the extended products was determined by sequencing the az genomic fragment using the same primers.
a2 CAT Fusion Constructs-The promoter and enhancer activity of the 5"flanking region was analyzed by inserting the EcoRI-AuaII genomic fragment (-961 to +lo9 in relation to the transcription start site at +1) upstream to CAT structural sequences in the CAT expression vector pCAT-Basic (Promega, Madison, WI), which lacks natural CAT regulatory elements. Nested deletion mutants of the 1.0-kb construct (paz961-CAT) were generated by restriction enzyme digestion of paz961-CAT at the restriction enzyme sites PstI (-301, XmaIII (-92), BglII (-549), and AccI (-776). Cytomegalovirus-CAT containing the cytomegalovirus promoter directing transcription of the CAT gene served as a positive control. Co-transfection in the pRSV-luciferase containing the Rous sarcoma virus promoter served as a control for transfection efficiency in all assays. All transfection experiments were performed at least four times. F'Ic. 3. S1 nuclease protection analysis to identify the transcription initiation site. A 5"end-labeled EcoRI-AuaII genomic fragment (extending from -691 to +log) was annealed to total mRNA (20 pg) from K562 cells induced with phorbol dibutyrate for 8 days, uninduced K562 cells, or REH cells, a lymphoblastic cell line (as described under "Materials and Methods"). The RNA-DNA hybrid was subjected to digestion with SI nuclease and the protected product analyzed on a denaturing 6% polyacrylamide gel. A 5'-end-labeled a-actin control probe which protects a 170-base fragment was analyzed in parallel. Standards include 5'-end-labeled MspI-digested SK-Bluescript plasmid and a sequencing reaction primed with oligonucleotide a2012, which differs from the AuaII site by 2 bp. Two sets of protected fragments, 92-99 and 109-119 nucleotides in length, were identified in the total mRNA from K562 cells induced with phorbol dibutyrate for 8 days.  (24). To determine the length and sequence of the B'-untranslated region, three additional a2 cDNA clones were identified by screening a human placenta X g t l l cDNA library with a 118-bp a2 cDNA, which represented the 5'-end of the published sequence (23,24). The longest of the three clones was partially characterized by restriction enzyme digestion, Southern blot hybridization, and sequence analysis. The 5'-untranslated region of the longest clone extended 129 bp 5' to the initiation methionine; this site was initially designated +1 (Fig. 11, and data is presented below to verify that this site represents one of several heterogeneous 5'-ends of the a2 cDNA. We therefore concluded that the 3'-untranslated region of the a2 mRNA extends approximately 3 kb beyond the end of the published cDNA sequence. Isolation of a 5' a2 Genomic Clone-Current models of gene regulation propose that specific DNA protein interactions regulate developmental and tissue-specific gene transcription. The promoters and enhancers of many genes including those of the integrin receptors aIb, a4, as, PI, CDlla, and CDllc, are lo- Following 16 h of hybridization, the complementary DNA strand was extended with murine mammary tumor virus reverse transcriptase, and the length of the extended products was determined on a 6% denaturing polyacrylamide gel. The exact length of the extended products was determined by running a sequencing gel of the a2 genomic fragment primed with the same oligonucleotide, a2012. The length of the two clusters of extended products agreed with the results of S1 nuclease protection analysis. independent bacteriophage clones from a bacteriophage library derived from human lung fibroblast genomic library in AFix (Stratagene, La Jolla, CA) were screened with the 118-bp a2 cDNA clone and an oligonucleotide probe (a2012). Two bacteriophage clones (AGa2-l + AGa2-2) were shown to be identical by partial restriction enzyme mapping and Southern blot hybridization (Fig. 1). Both AGa2-1 and XGa2-2 contained a 2.0-kb EcoRI restriction fragment which hybridized to both partial cDNA and oligonucleotide probes. Comparative sequence analysis of the 2.0-kb EcoRI genomic fragment and the a2 cDNA clone described above revealed 192 bp of identity beginning at +1 of the cDNA and extending to the border of the first intron; the relationship of the genomic clones to the cDNA is shown in Figs. 1 and 2. The 2.0-kb EcoRI genomic fragment contains 961 bp of 5"flanking sequence, all of exon 1 with 129 bp of 5'untranslated region and 64 bp of coding sequence, and part of the first intron (Fig. 2). The genomic sequence diverged from the cDNA at the intron-exon border of the first intron. The sequence flanking the +1 transcription start site lacked consensus sequences for splice donor or splice acceptor sites.
Mapping the Site of Danscription Initiation-To map the site of transcription initiation, primer extension and S1 nuclease analysis were used in combination. S1 nuclease analysis, performed using a 5'-end-labeled EcoRI-AuaII genomic fragment (extending from -961 to +log) and hybridized to total mRNA from K562 cells induced with phorbol dibutyrate for 8 days, revealed two sets of protected fragments, 92-99 and 109-119 nucleotides in length (Fig. 3). Protected fragments were not  Figs. 1 and 2, with 109 bp of 5'-untranslated region upstream to the CAT structural gene. Constructs pa2776-CAT, pa2549-CAT, pa292-CAT, and pa230-CAT were deletion mutants derived from pa2961-CAT. B, a2 promoter activity in the epithelial cell line, T47-D. The constructs pa2961-CAT, pa2776-CAT, pa2549-CAT, pa292-CAT, and pa230-CAT were transfected in parallel with cytomegalovirus-CAT, which contains the strong viral promoter, and observed with mRNA from uninduced K562 cells which express minimal az mRNAor REH cells, a lymphoblastic cell line which does not express a2 mRNA (Fig. 3). These results suggest the existence of multiple transcription initiation sites, with two major sites located 112 and 129 bp upstream of the initiation methionine. The 5'-end of the full-length cDNA described above, designated +1, terminated at the same site as the 109-bp fragment protected in the S1 nuclease analysis. Primer extension analysis using the antisense oligonucleotide a2012 (Fig. 4) identified multiple extension products whose sizes predicted transcription initiation sites identical to those determined by S1 nuclease analysis. These findings confirmed the presence of multiple transcription initiation sites in two clusters between 112 and 129 bp upstream of the initiation methionine.

The ar2 Integrin Promoter
The 5'-Flanking Region of the a2 Gene-The 5"flanking sequence around the site of transcription initiation of the a2 gene resembles that of other integrin promoters by lacking both TATA and CAAT boxes, as shown in Fig. 2 (29-36). The site of transcription initiation for the other integrin promoters which lack TATA or CAAT boxes is a consensus sequence defined as the initiator sequence (Inr) (29,32,33,36). The Inr consensus sequences for as, aIIb, CDlla, and CDllb resemble closely the original consensus sequence of the terminal deoxynucleotidyl transferase gene (37). The sequence 5'-CCCTGCTC=C-CG-3' at the initiation site of the az integrin gene maintains the consensus TCA at -2 to +1 and conserves pyrimidine-rich stretches on either side of the transcription start site but diverges somewhat from the Inr sequences of other integrins in both the 3' or 5' directions. The Inr has been shown to direct correct transcription initiation in association with nearby Spl sites (38,39). Within 100 bp of the transcription start site, four consensus sequences for Spl are identified, including one Spl site between the two clustered sites of transcription initiation. Additional Spl sites are located at -291 and -307.
In addition to the Spl sites, numerous concensus binding sites for ubiquitous cell type-specific and differentiation-dependent transcription factors are located in the 961-bp segment of the 5'-flanking region. Consensus binding sites for transcription factors include two AP-1 sites at -765 and -702, an AP-2 site at -132, and six AP2-CS4 sites at -311, -294, -222, -216, -55, and +3. Whereas both AP-1 and AP-2 mediate enhanced transcription via the phorbol ester/diacylglycerol-activated protein kinase C pathway, AP-2 also mediates enhanced transcription via a cAMP-dependent protein kinase A pathway. Regulatory regions recognized by AP-2 are found in the SV40 enhancer, human growth hormone, c-myc, and histocompatibility H-2Kb genes (40,41). In addition to consensus sequences for ubiquitous transcription factors, two elements associated with hematopoietic differentiation are adjacent to one another. A consensus site for Pu.1 located at -666 bp of the 5"flanking sequence binds members of the hematopoietic-specific ets family of protooncogenes (42,43). A single GATA site at -679 bp may direct differentiation along the erythroidmegakaryocytic and mast cell lineages (44). The adjacent binding sites for ets and GATA transcription factors suggest the possibility of either cooperative interaction or competitive inhibition. Members of the GATA family of DNA-binding proteins have been shown to function in cooperation with CACCC elements found in globin promoters (44) as well as with other DNA-binding proteins in mediating tissue-specific expression of the rat platelet factor 4 gene (45). However, consensus binding sites for Pu-1 and GATA are adjacent to one another in the CDllb promoter, yet only the Pu-1 site is fimctionally active (46). Two half-palindromic motifs with the potential for estrogen receptor binding are positioned at -514 and -802 (47). We previously observed a correlation between integrin expression and estrogen receptor expression in breast cancer ( E ) , raising the possibility that a2 integrin expression may be hormonally regulated. The identification of two estrogen receptor half-sites supports the role for hormonal regulation of the az integrin gene. Recently, the estrogen receptor gene was shown to be transcriptionally active in megakaryocytes and in the megakaryocytic cell line SO1 (48), an observation consistent with a role for estrogen in regulating expression of a2P1 and other megakaryocytic genes. Sites for steroid receptor binding have not been identified in other integrin promoters.
In addition to the emphasis placed on the positive regulatory elements required to activate or enhance gene transcription, recent evidence supports the importance of cell type-specific negative transcription factors called silencers. The two partially conserved CTF-NF1 (CCAAT-binding transcription factor-NF1) sites at positions -235 and -181 in a gene which lacks both TATA and C U T box sequences closely resemble the NF1like consensus identified in the rat growth hormone gene, the human retinol-binding protein, the mouse Sparc (osteonectin) gene, the chicken P-globin gene, the mouse Ren-ld negative element, and a silencer site in the mouse a2(I) collagen gene (49). In these genes, the NF1-like consensus sequence serves as a negative regulatory element which suppresses gene expression in a cell type-specific manner.
The sequence of the 5"flanking region between +1 and -400 is composed of 70% guanine (G) and cytosine (C) residues, with 9 HpaII restriction enzyme sites (CCGG) concentrated in the region between +1 and -389. A GC-rich region with numerous HpaII restriction enzyme sites, also observed in the as-integrin promoter, suggests that this may be part of an HpaII tiny island fragment, a region which can be regulated by altering chromatin configuration (50). Distal to the GC-rich region, a short segment of dinucleotide repeats extends from -604 to -647. The dinucleotide repeats of TG or CA which compose the majority of the repetitive sequence have been identified in the introns of human y-globin and cardiac actin genes, and downstream of mouse immunoglobulin genes (51-53). These repetitive elements may induce conformational changes in the DNA helix which result in altered DNA-protein interactions (54,55).
The a2 Promoters Demonstrate Cell Qpe-specific Promoter Activity-To determine if the 5"flanking region of the a2 gene functions to direct transcription in a cell-type and differentiation-dependent manner, the 1070-bp EcoRI-AvaII genomic fragment spanning -961 to +lo9 was placed upstream of CAT structural sequences in a CAT expression vector (paz961-CAT) (Fig. 5A). The promoter activity of the 1070-bp genomic region was determined by transient transfection of the paz-961 CAT construct into T47-D cells and uninduced and induced K562 cells. The pa2961-CAT construct directs CAT enzymatic activity with appropriate cell type specificity when transiently transfected into T47-D and K562 cells. When normalized for transfection efficiency, the paz961-CAT construct produced 34-CAT-BASIC, which lacks promoter activity, into the epithelial cell line, T47-D. Co-transfection with RSV-luciferase was used to control for transfection efficiency. After 48 h of incubation, cell extracts were assayed. Following normalization for transfection efficiency, CAT activity of the eight Constructs was determined by thin-layer chromatography and differential extraction. Relative CAT activity of the constructs is plotted. C, megakaryocyte-specific activity of the a2 promoter. Uninduced K562 cells and K562 cells induced for 2 days with 40 n~ phorbol dibutyrate were transfected by electroporation with the constructs pa2961-CAT, pa2776-CAT, pa2549-CAT, pa292-CAT, pa230-CAT, cytomegalovirus-CAT, and CAT-BASIC and co-transfected with RSV-luciferase. After 48 h, CAT activity was determined as described above. Promoter activity of the different constructs in uninduced K562 cells (hatckd bar) is compared to activity in K562 cells induced to differentiate along the megakaryocytic pathway (solid bar) with 40 n~ phorbol dibutyrate for 4 days.

~~
The a2 Integrin Promoter fold greater enzymatic activity than CAT-structural sequences alone in T47-D cells, as shown in Fig. 5B. The a2961-CAT construct was inactive in uninduced and weakly active in induced K562 cells (Fig. 5 0  To further characterize the 5"flanking region required for either epithelial or megakaryocyte-specific promoter activity, 5' deletion mutants of pa2961-CAT, pa2776-CAT (-776 to +log), pa2549-CAT (-549 to +log), pa292-CAT (-92 to +log), and pa230-CAT (-30 to +log) were constructed, as shown in Fig. 5A. The construct pa230-CAT, which retains only 30 bp 5' to the transcription start site, failed to direct expression of the CAT gene in the epithelial cell line T47-D (Fig. 5B). The construct pa292-CAT, containing an additional 62 bp of the 5"flanking region, was capable of directing activity in T47-D cells at 12% the activity of the parent pa2961-CAT construct. The deletion mutants containing 549 or 776 bp of 5"flanking sequence directed high-level activity in T47-D cells at 53% or greater than 100% of the activity of the parent pa2961-CAT construct (Fig.  5B). Based on these deletional analyses, enhancer activity necessary for high-level expression of the a2 gene in breast epithelial cells, as represented by the T47-D cell line, is located between bp -92 and -549 and in the distal 5'-end between bp -549 and -961. One estrogen half-site is located between -92 and -549, suggesting a possible role for estrogen in regulation of a2 expression. The diminished activity of the pa2549-CAT versus pa2961-CAT constructs suggests that additional enhancer elements are also located in the 412 bp between -549 and -961. Another estrogen receptor half-site as well as two AP-1 sites are located within this region.
In contrast to the enhancer elements required for high-level constitutive expression in T47-D epithelial cells, elements controlling megakaryocyte-specific activity of the a2 promoter differ significantly. All deletion mutants including pa230-CAT, pa292-CAT, pa2549-CAT, pa2776-CAT, and the parent construct failed to direct enzymatic activity in uninduced K562 cells. The pa230-CAT construct, however, directs low-level activity in induced K562 cells. In contrast to the results obtained in the breast epithelial cells, the pa292-CAT construct directs the greatest enzymatic activity in induced K562 cells, with 500-fold greater activity in K562 cells induced for 4 days compared to uninduced K562 cells in four separate experiments. In contrast, the longer pa2549-CAT construct directs activity only 100-fold greater than CAT-BASIC or the same construct in uninduced K562 cells, suggesting the role for a silencer which down-regulates a2 expression in this region. The pa2766-CAT and pa2961-CAT constructs fail to direct activity greater than the CAT-BASIC construct alone. Regulatory elements between bp -30 and -92 are necessary for high-level megakaryocytespecific gene activity, and this activity is silenced by an additional upstream sequence in megakaryocytes. Additional regulatory regions in the distal 5'-regulatory region, within introns, or in the long 3'-untranslated region may augment the megakaryocytic differentiation program. Analysis of additional regulatory elements is in progress.
Characterization of the a2 promoter suggests a complex pattern of positive and negative regulatory elements which determines cell type-and differentiation-specific expression. The high-level constitutive expression of the a2 integrin in epithelial cells is directed from the region between bp -92 and -549, with additional enhancer activity located further 5', between -549 and -961. One of the estrogen receptor half-sites is deleted in the pa2549-CAT mutant construct, and both elements may be necessary for high-level a2 expression. In contrast to epithelial cell types, the megakaryocyte-specific promoter activity is present in the -92 to +1 region and is repressed by more 5' sequence. These findings suggest the presence of a megakaryocyte-specific enhancer activity between -30 and -92 and silencer elements in the -92 to -961 bp region. Known silencer elements have not been identified in this region. The dinucleotide repetitive sequence lies just upstream from bp -549 and may act as a silencer by altering chromatin conformation.
In summary, we have identified the promoter and enhancer regions of the a2 integrin gene which confer cell type-specific and differentiation-induced a2 gene expression. The identification of the a2 promoter will now allow dissection of the complex regulation of this integrin gene during normal differentiation, as well as carcinogenesis.