The human ICAM-2 promoter is endothelial cell-specific in vitro and in vivo and contains critical Sp1 and GATA binding sites.

The expression of intercellular adhesion molecule 2 (ICAM-2) in adult tissues is restricted to vascular endothelial cells and megakaryocytes. We have previously shown that the endothelial-specific in vivo activity of the human ICAM-2 promoter is contained within a small (0.33-kilobase (kbp)) 5'-flanking region of the gene. Here we describe the in vitro characterization of this region. The ICAM-2 promoter is TATA-less, and transcription in endothelial cells initiates at four sites. Reporter gene expression directed by the promoter was 125-fold greater than vector alone in bovine aortic endothelial cells but less than 2-fold vector alone in non-endothelial (COS) cells, confirming that specificity in vivo was paralleled in vitro. The addition of 2.7 kbp of 5'-flanking region to the 0.33-kbp fragment had no effect on promoter activity or specificity. The mutation of an Sp1 motif centered at base pair -194 or an eight-base pair palindrome at -268 each reduced promoter activity by 70%. Mutation of GATA motifs at -145 and -53 reduced promoter activity by 78 and 61%, respectively. Specific binding of bovine aortic endothelial cells nuclear proteins to the Sp1 and GATA sites was demonstrated by gel shift analysis. Promoter activity in COS cells was transactivated 3-4-fold by overexpression of GATA-2. The results presented here suggest that transcription from the ICAM-2 promoter in endothelial cells is regulated by the interplay of several positive-acting factors and provide the basis for further analysis of endothelial-specific gene expression.


From the Immunology Research Center, St Vincent's Hospital, Fitzroy 3065, Victoria, Australia
The expression of intercellular adhesion molecule 2 (ICAM-2) in adult tissues is restricted to vascular endothelial cells and megakaryocytes. We have previously shown that the endothelial-specific in vivo activity of the human ICAM-2 promoter is contained within a small (0.33-kilobase (kbp)) 5-flanking region of the gene. Here we describe the in vitro characterization of this region. The ICAM-2 promoter is TATA-less, and transcription in endothelial cells initiates at four sites. Reporter gene expression directed by the promoter was 125-fold greater than vector alone in bovine aortic endothelial cells but less than 2-fold vector alone in non-endothelial (COS) cells, confirming that specificity in vivo was paralleled in vitro. The addition of 2.7 kbp of 5-flanking region to the 0.33-kbp fragment had no effect on promoter activity or specificity. The mutation of an Sp1 motif centered at base pair ؊194 or an eight-base pair palindrome at ؊268 each reduced promoter activity by 70%. Mutation of GATA motifs at ؊145 and ؊53 reduced promoter activity by 78 and 61%, respectively. Specific binding of bovine aortic endothelial cells nuclear proteins to the Sp1 and GATA sites was demonstrated by gel shift analysis. Promoter activity in COS cells was transactivated 3-4-fold by overexpression of GATA-2. The results presented here suggest that transcription from the ICAM-2 promoter in endothelial cells is regulated by the interplay of several positive-acting factors and provide the basis for further analysis of endothelial-specific gene expression.
The endothelial cells that form the lining of all blood vessels perform a wide range of functions. In addition to providing a selective barrier between the bloodstream and tissues, vascular endothelial cells are critical for processes including thrombosis, angiogenesis, leukocyte trafficking, and the maintenance of vascular tone (1). The vascular endothelium also plays an important role in cancer metastasis and in the pathogenesis of non-neoplastic diseases such as rheumatoid arthritis and atherosclerosis (2,3). To gain an understanding of the regulation of endothelium-specific gene expression and thus provide insights into these processes and conditions, a number of studies have focused on the characterization of endothelial cell-specific promoters (4 -14). Promoters capable of targeting gene expression to endothelial cells in vivo as well as in vitro may be particularly useful, not only in mechanistic studies but also in medical applications ranging from transgenic modification of donor organs in xenotransplantation (15,16) to gene therapy for diseases affecting the cardiovascular system (17).
The promoters of several genes that are expressed predominantly or exclusively in endothelial cells have been examined both in vitro in transient transfection assays and in vivo in transgenic mouse studies. The human and mouse tie1 promoters demonstrated endothelial cell specificity in vivo, with expression in the vasculature varying from organ to organ but surprisingly were not specific in vitro (7). In contrast, the human KDR/flk-1 and von Willebrand factor promoters were endothelial cell-specific in vitro, but their activity in vivo was undetectable (18) or restricted to a subset of endothelial cells (12), respectively. Unlike these promoters, which presumably lack important regulatory elements, the mouse tie2 (tek) promoter/enhancer was capable of driving endothelial cell-specific transgene expression in vitro and in vivo (4), although the strong expression in all vascular endothelium in adult tie2/lacZ transgenic mice was somewhat unexpected because tie2 expression is downregulated in adult animals (19).
The first promoter shown to target uniform, high-level transgene expression to the vasculature of transgenic mice was that of the human intercellular adhesion molecule 2 gene (ICAM-2) 1 (20). Human ICAM-2 is a counter receptor for lymphocyte function-associated antigen-1 (21) and may provide a costimulatory signal for T cell stimulation by allogeneic class II major histocompatibility complex (22). In human tissues, the expression of ICAM-2 is restricted largely to endothelial cells and megakaryocytes (23). Unlike the mouse tie2 promoter/enhancer, which comprises 2.1 kbp of promoter and 10 kbp of the first intron (4), the human ICAM-2 promoter contains all of the signals necessary for endothelium-specific transgene expression in vivo within a relatively small (Ͻ350 bp) region, presenting a novel opportunity for analysis. In this study, we report the first in vitro characterization of the ICAM-2 promoter. We demonstrate that the ICAM-2 promoter is endothelial cellspecific in vitro and identify four elements, an Sp1 site, two GATA sites, and an 8-bp palindromic sequence also present in the tie2 enhancer, that are necessary for full promoter activity. We also show that the ICAM-2 promoter is transactivated in non-endothelial cells by overexpression of two members of the GATA family of transcription factors, GATA-1 and GATA-2.

EXPERIMENTAL PROCEDURES
DNA Preparation, Sequencing, and Analysis-Plasmid DNA was prepared from bacterial cultures using Qiagen Plasmid Midi or Maxi Kits. DNA sequencing reactions were performed using Thermo Sequenase (Amersham Pharmacia Biotech). T3 and T7 primers were used to sequence inserts cloned into pBluescript II (Stratagene), and the prim-* 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  ers GL3-1 and GL3-2 (corresponding to bp 51-68 and to the complement of bp 4785-4803, respectively, in the vector) were used for inserts cloned into pGL3-Basic (Promega). Reactions were electrophoresed at either the Nucleotide Sequencing Facility, Monash University, Victoria, or the Molecular Biology Facility, Griffith University, Queensland, Australia. The DNA sequence was scanned for putative transcription factor binding sites using the MatInspector program (24) and the TRANSFAC data base (25). Comparison of ICAM-2 sequences was carried out using the Clustal W program (26).
Isolation of Human ICAM-2 5Ј-Flanking Region-The Human GenomeWalker kit (CLONTECH) was used to walk upstream of known ICAM-2 5Ј-flanking sequence. A primer complementary to the sequence ϩ27/ϩ50 relative to the ICAM-2 translation start site (21) was used with an outer adaptor primer in primary PCRs using five adaptorligated human genomic libraries as templates. Samples of the primary PCRs were used in secondary PCRs with a nested ICAM-2-specific primer (Ϫ17/ϩ5 relative to the translation start site) and a nested adaptor primer. Products of the secondary PCRs were resolved by agarose gel electrophoresis, purified using the QIAquick Gel extraction kit (Qiagen), and cloned into pBluescript II.
Generation and Testing of Transgenic Mice-The 5Ј end of the 0.33kbp ICAM-2 promoter fragment in an ICAM-2 promoter/human CD59 microinjection construct (20) is delineated by an NcoI site. This site was employed in the precise fusion of 2.7 kbp of upstream flanking sequence (isolated as described above) to the 0.33-kbp region. The resulting construct, designated PK, was used to generate transgenic mice as described (20). Peripheral blood leukocytes from the transgenic mice were examined for the expression of CD59 by flow cytometric analysis as described (20). Fresh frozen tissue sections prepared from mouse organs were stained to detect the pattern of CD59 expression as described (20).
Rapid Amplification of cDNA Ends (RACE)-The method used to identify the 5Ј end of ICAM-2 transcripts was based on the 5Ј RACE system (Life Technologies, Inc.). In brief, total RNA (2.5 g) prepared from human umbilical vein endothelial cells was reverse-transcribed from an antisense primer (ϩ27/ϩ50 relative to the ICAM-2 translation start site) using SuperScript II (Life Technologies). The cDNA was purified and tailed with dCTP using terminal deoxynucleotide transferase (Promega). An aliquot of the tailed cDNA was used as template in a PCR with primers AAP (5Ј-GGCCACGCGTCGACTAGTACGGGI-IGGGIIGGGIIG-3Ј) and a nested antisense primer (Ϫ17/ϩ5 relative to the ICAM-2 translation start site) using the Advantage Tth polymerase mix (CLONTECH) to maximize fidelity. PCR was carried out in a DNA engine thermal cycler (MJ Research) using the following parameters: 94°C for 2 min; 5 cycles of 92°C/10 s, 55°C/10 s, 72°C/30 s; and 25 cycles of 92°C/5 s, 55°C/5 s, 72°C/30 s. An aliquot (0.1%) of the first PCR was used directly in a second PCR with primers AUAP (5Ј-GGC-CACGCGTCGACTAGTAC-3Ј) and the nested antisense (Ϫ17/ϩ5) primer under the same conditions except that the annealing temperature was increased to 60°C. Products of the second PCR were purified using the High Pure PCR product purification kit (Boehringer Mannheim) and cloned into pBluescript II.
Cell Culture and Transient Transfection-Bovine aortic endothelial cells (BAEC) were isolated using a modification of a previously described procedure (27). Freshly isolated bovine aortae were ligated, rinsed 3-5 times with phosphate-buffered saline, pH 7.4, and filled with growth medium (see below) lacking fetal calf serum and containing 1 mg/ml collagenase (Boehringer Mannheim). After a 10 -15-min incubation in 5% CO 2 at 37°C, the contents of the aortae were removed and centrifuged at 1500 ϫ g for 10 min. The cell pellet was resuspended in growth medium consisting of Dulbecco's modified Eagle's medium (Life Technologies) supplemented with 15% heat-inactivated fetal calf serum (Life Technologies), 15 mM HEPES (Trace Scientific, Australia), 60 g/ml penicillin (CSL, Australia), 100 g/ml streptomycin (CSL), and 20 g/ml cis-4-hydroxy-L-proline (Sigma) and dispensed into gelatincoated flasks. BAEC were used at passage 3-8. COS-7 (African green monkey fibroblast) cells were obtained from the American Type Culture Collection and cultured in the same medium as used for BAEC except that the concentration of heat-inactivated fetal calf serum was 10%, and no cis-hydroxyproline was added.
Transient transfections were performed at least twice, in triplicate. For transfection of BAEC, approximately 10 5 cells were plated/well in 6-well trays and incubated in 5% CO 2 at 37°C until 40 -50% confluent (16 -20 h). For each well, 4 l of LipofectAMINE (Life Technologies) and 1.5 g of plasmid DNA (1 g of luciferase construct plus 0.5 g of pSV␤Gal (Promega) were preincubated separately in 100 l of Opti-MEM I (Life Technologies) for 45 min at room temperature, mixed together with 800 l Opti-MEM I, and used to overlay the cells. After a 5-h incubation at 37°C, the overlay was replaced with 2 ml of growth medium, and the cells were incubated for a further 64 -72 h before harvesting and analysis. COS cells were transfected similarly except that 2 ϫ 10 5 cells were plated/well, 6 l of LipofectAMINE was used, and the cells were harvested after a 40 -48 h incubation.
For transactivation experiments, the transfected DNA comprised 0.9 g of luciferase construct, 0.2 g of M6␣-SEAP, 0.2 g of pSV␤Gal, and 0.2 g of either pXM-mGATA-1 (28), pMT2-hGATA-2 (29), or the vector pcDNA3 (Invitrogen). M6␣-SEAP was constructed by cloning the GATA-responsive M6␣ promoter from an M6␣-hGH plasmid (Ref. 30; kindly provided by Dr Merlin Crossley, Biochemistry Department, University of Sydney, NSW, Australia) into the secreted alkaline phosphatase reporter plasmid pSEAP2-Basic (CLONTECH). pXM-mGATA-1, which contains the adenovirus major late promoter driving the expression of murine GATA-1, and pMT2-hGATA-2, which contains the same promoter driving the expression of human GATA-2, were also provided by Dr. Merlin Crossley. Luciferase and ␤-galactosidase activities in lysates of transfected cells were measured using the luciferase assay system (Promega) and the chemiluminescent ␤-Gal reporter gene assay kit (Boehringer Mannheim), respectively. Secreted alkaline phosphatase activity in the culture medium of transfected cells was measured using the Great EscAPe SEAP chemiluminescence detection kit (CLONTECH). Light intensity was measured using a MicroLumat LB96P plate luminometer (EG and G Berthold). The protein concentration of lysates was determined using the DC protein assay kit (Bio-Rad) and bovine serum albumin as the standard. Enzyme activities were corrected for background by subtracting the values obtained for mock-transfected cells.
Mutagenesis-A two-step PCR method was used to mutagenize the 0.33-kbp ICAM-2 promoter cloned in pBluescript II such that the 5Ј end is closest to the M13 forward priming site (20). In the first step, PCRs were carried out using an internal mutagenic primer with an external vector-targeted primer. The internal forward primers (mutated bases underlined) were P 8 -F mut , 5Ј-TTCCCCAGACGCGGGGCTTGT-3Ј; P 12 -F mut , 5Ј-GTCTCATATTCGCATGCGAACACCCATTG-3Ј; Sp1-F mut , 5Ј-CATTGCCTGCCCTATCCCTTGCACA-3Ј; GATA U-F mut , 5Ј-GCCCTGCTTTAGAGCAGCTTC-3Ј; and GATA D-F mut , 5Ј-CTCT-CAGTATATGAGAGGA-3Ј. The internal reverse primers (P 8 -R mut , etc.) were complementary to the forward primers. The forward and reverse primers were used in independent PCRs with the M13/pUC reverse and forward amplification primers (Life Technologies), respectively. The two products from each set of first-step PCRs (e.g., P 8 -F mut /M13-R and P 8 -R mut /M13-F) were purified, mixed, and used in the second-step PCR with the external primers only. Products from the second-step PCRs were cloned into pGL3-Basic and sequenced to confirm the presence of the desired mutations.
To construct the Sp1/GATA U (upstream) double mutant, the upstream GATA site mutant cloned in pGL3-Basic was used as template. The mutagenic primers were Sp1-F mut and Sp1-R mut , and the external primers were GL3-1 and GL3-2. The second-step PCR product was cloned and sequenced as described above.
Preparation of Nuclear Extracts-BAEC nuclear extracts were prepared as follows. A confluent monolayer of cells in a 25 cm 2 flask were harvested by scraping, washed in cold phosphate-buffered saline (pH 8.0), and resuspended in 50 l of cold Buffer A (10 mM Tris-HCl, pH 9.0, 2 mM MgCl 2 , 5 mM KCl, 10% glycerol, and 1 mM EDTA). 50 l of cold Buffer B (Buffer A plus 1% Nonidet P-40) was added, and the cells were allowed to swell on ice for 15 min then vortexed vigorously. The resulting lysate was centrifuged at 700 ϫ g for 10 min at 4°C. The nuclear pellet was resuspended in 0.5 ml of Buffer A plus 0.5 ml of Buffer B and recentrifuged at 700 ϫ g for 5 min at 4°C. The pellet was resuspended in 50 l of cold Buffer C (20 mM HEPES, 0.4 M NaCl, 1 mM EDTA, and 1 mM EGTA), and nuclear proteins were extracted by vigorous shaking for 30 min at 4°C. Cell debris was pelleted by centrifugation at 18,000 ϫ g for 5 min at 4°C, and the nuclear extract was aliquoted, frozen in liquid nitrogen, and stored at Ϫ70°C. Buffers A, B, and C were adjusted to 1 mM dithiothreitol, 1 mM phenylmethylsulfonyl fluoride, 1 g/ml leupeptin, and 1 g/ml aprotinin before use.
Electrophoretic Mobility Shift Assay-The binding mixture (20 l) contained ϳ15 g of BAEC nuclear extract, 0.5 g poly(dI:dC), and 10 5 cpm of 32 P-labeled double-stranded oligonucleotide probe in 10 mM Tris-HCl, pH 8.0, 5 mM MgCl 2 , 1 mM dithiothreitol, 1 mM EDTA, and 1 mM bovine serum albumin. All of the ingredients except the probe were mixed and incubated for 15 min at room temperature, after which the probe was added, and incubation was continued for 15-30 min. Band shifts were resolved by electrophoresis at 4°C on a non-denaturing 7% polyacrylamide gel. The probes were as follows (coordinates relative to the primary transcription start site): Sp1, bp Ϫ207 to Ϫ183; GATA U (upstream), bp Ϫ155 to Ϫ135; and GATA D (downstream), bp Ϫ61 to Ϫ43. Competition studies were performed by adding a 100-fold molar excess of unlabeled double-stranded oligonucleotides to the binding reactions before the addition of labeled probe. DNA-protein-complexes were judged to be specific if their formation was competitively inhibited by cold unlabeled probe but not by an unrelated oligonucleotide. Mutant competitors (Sp1-F mut /Sp1-R mut etc.) are described in the previous section.

Isolation of 3 kbp of 5Ј-Flanking Region of the Human
ICAM-2 Gene-Human ICAM-2 expression in the tissues is restricted largely to vascular endothelial cells and megakaryocytes (23); in the blood, ICAM-2 is expressed on platelets (31) and at a low level on resting lymphocytes and monocytes but not on granulocytes (23). We have previously shown that 0.33 kbp of 5Ј-flanking region of the human ICAM-2 gene is sufficient to target high-level transgene expression to all vascular endothelium in the heart, kidney, liver, lung, and pancreas of transgenic mice (20), consistent with the tissue distribution of human ICAM-2. Unexpectedly, however, transgenic mouse lymphocytes were negative for transgene expression and granulocytes were very strongly positive. This discrepancy is unlikely to be due to differences in the pattern of expression of human and mouse ICAM-2, because flow cytometric analysis revealed that the hierarchy of ICAM-2 expression levels is similar on human and mouse leukocyte populations i.e. monocytes Ͼ lymphocytes Ͼ Ͼ granulocytes. 2 An alternative explanation is that elements regulating ICAM-2 expression on leukocytes are absent from the 0.33-kbp region. To investigate this possibility and to generate additional material for in vitro analysis of the ICAM-2 promoter, a larger 5Ј-flanking region of human ICAM-2 was isolated. Nested primers targeted to known ICAM-2 promoter sequence were used with adaptor primers to amplify the 5Ј-flanking sequence from five adaptorligated human genomic libraries. The PvuII, SspI, and ScaI libraries yielded products of 1.0, 1.8, and 3.1 kbp, respectively, which were cloned and analyzed by restriction enzyme digestion to produce the map shown in Fig. 1A. The DNA sequence obtained from the 3Ј end of the two larger clones matched that of the 0.33-kbp region (20) except for an additional G at bp Ϫ281 and an additional C at bp Ϫ275 (Fig. 1B). The presence of these bases, confirmed in the original 0.33-kbp clone by resequencing, revealed a consensus NF-B motif (GGGGTTC-2 P. J. Cowan, unpublished results.  (34) using the Clustal W program (26). In the human sequence, potential recognition sites for known transcription factors are indicated by underlining, two palindromes (P 8 and P 12 ) are shown in bold type, and the start sites of transcription identified by 5Ј-RACE are marked by arrows. Numbering of the human sequence is relative to the deduced primary transcription start site (shown in bold type) as ϩ1. Numbering of the mouse sequence is relative to the A of the start codon as ϩ1, since the transcription start site(s) of the mouse ICAM-2 gene is not known. C, schematic representation of putative regulatory elements in the 0.33kbp human ICAM-2 promoter fragment. CCC) at bp Ϫ281 to bp Ϫ272 (Fig. 1B).
Transgenic mice were generated using a construct (termed PK) in which 3 kbp of human ICAM-2 5Ј-flanking region was used to drive the expression of human CD59. Flow cytometric analysis of leukocytes from 11 PK transgenic mice revealed an identical pattern of transgene expression to that driven by the 0.33-kbp fragment (20), i.e. very strong (Ͼ40-fold human level) expression on granulocytes (Fig. 2) and little or no expression on lymphocytes (data not shown). This suggests that the postulated missing leukocyte-specific regulatory elements, assuming they exist, are not located within the 3 kbp of 5Ј-flanking region of ICAM-2 used in the generation of PK transgenic mice. To compare the level and pattern of transgene expression in tissues driven either by the small (0.33 kbp) or large (3 kbp) ICAM-2 promoter fragment, organs from the original line of transgenic mice (20) and from one of the PK lines were examined by immunocytochemistry. An identical pattern of staining was observed in both lines, with strong transgene expression restricted to all vascular endothelium in the heart, kidney, liver, lung, and pancreas. Representative sections of heart and kidney from transgenic and control mice, stained to detect CD59 expression, are shown in Fig. 3.
Identification of the Transcription Start Sites of ICAM-2-To identify the transcription start site(s) of the human ICAM-2 gene, 5Ј-RACE was performed on total RNA from human umbilical vein endothelial cells, which strongly express ICAM-2 (23). First-strand cDNA prepared using a primer targeted to the first exon was tailed with dC and used in PCR with a G-rich primer and a gene-specific nested primer. After a second PCR with nested primers, the resulting products were cloned into pBluescript II, and five clones were sequenced, revealing four transcription start sites (Fig. 1B). A thymine residue (shown in bold type) located 77 bp upstream of the ICAM-2 start codon was designated the primary transcription initiation site (ϩ1), because the start sites for two independent clones mapped to this point and that for a third mapped immediately upstream (Ϫ1). The remaining start sites were located at Ϫ37 and Ϫ66. Thus like the promoters of several other constitutively ex-pressed cell adhesion molecule genes including PECAM-1 (32) and CD18 (33), the ICAM-2 promoter lacks consensus TATA and CAAT boxes and contains multiple transcription start sites.
Endothelial Cell Specificity of the ICAM-2 Promoter in Vitro-The in vitro promoter activity and specificity of various lengths of ICAM-2 5Ј-flanking region were tested by transfecting different fragments cloned in the luciferase reporter vector pGL3-Basic into BAEC or the simian kidney fibroblast cell line COS. The constructs were cotransfected with pSV␤Gal to correct for differences in transfection efficiency, and luciferase activity was normalized to that of pGL3-Basic alone. The 3.0-, 1.7-, and 0.33-kbp fragments exhibited similar significant promoter activity (100 -130-fold the level of vector alone) in BAEC and only minimal activity (approximately 2-fold vector alone) in COS cells (Fig. 4), indicating that the signals necessary for endothelial cell-specific ICAM-2 promoter activity in vitro are contained within the 0.33-kbp fragment. In vitro analysis of the promoter was therefore performed on this fragment.
Identification of Potential Cis-acting Elements in the ICAM-2 Promoter-The sequence of the 0.33-kbp promoter fragment was scanned for potential transcription factor binding sites and other motifs and compared with the corresponding sequence from mouse (34) to identify conservation of sites. Two sites for the GATA family of transcription factors, centered at Ϫ145 and Ϫ53 bp relative to the primary start point of transcription in the human promoter, were conserved in position and orientation in the mouse ICAM-2 gene (Fig. 1B). Four conserved sites for the Ets family were located at Ϫ135, Ϫ127, Ϫ70, and Ϫ44. An 8-bp palindrome (CAGATCTG, referred to here as P 8 ) was detected at Ϫ268; this sequence is also present in the first intron of the endothelial-specific tek (tie2) gene and has been shown to be important in conferring endothelial specificity in vivo (4). Seven of the nucleotides in P 8 were conserved in the mouse ICAM-2 sequence, although the distance between this motif and a highly conserved block containing the GATA and Ets sites was 86 bp greater in mouse than human. Another palindrome (CACGCATGCGTG, referred to as P 12 ) present at Ϫ217 was only partly conserved (9/12 bp) in the mouse sequence, as was the NF-B site (8/10 bp) at Ϫ276. An Sp1 motif at Ϫ194 in the human promoter was not conserved within the corresponding region in mouse. The locations of the putative regulatory elements within the human ICAM-2 promoter are indicated schematically in Fig. 1C.
Mutational Analysis of the ICAM-2 Promoter-To identify functionally important elements in the ICAM-2 promoter, several of the motifs described above were mutated, and the effect of each mutation on promoter activity in endothelial cells was determined. Mutated derivatives of the 0.33-kbp promoter fragment were constructed by a two-step PCR method, cloned into pGL3-Basic, and transfected into BAEC. The mutation introduced into the P 8 site (CAGATCTG 3 CAGACGCG), based on that shown to affect tek promoter activity and endothelial cell specificity in vivo (4), reduced ICAM-2 promoter activity in BAEC by 70% (Fig. 5). The upstream GATA site mutation (TTATCA 3 TTTAGA), which had previously been shown to reduce the in vitro activity of the P-selectin (14) and eNOS (11) promoters, caused a 78% reduction in promoter activity (Fig. 5). Mutation of the core sequence of the Sp1 site (CCCCGCCCC 3 CCCTATCCC) or that of the downstream GATA site (AGATAA 3 AGTATA) reduced promoter activity by 70 and 61%, respectively (Fig. 5). The combination of the Sp1 and upstream GATA site mutations had a greater effect than either mutation alone, reducing promoter activity by 87%, although the double mutant retained measurable promoter activity (more than 10-fold the level of vector alone).

FIG. 2. In vivo transgene expression on granulocytes driven by the 3-kbp ICAM-2 promoter fragment.
Leukocytes of a heterozygous mouse from the PK (3-kbp ICAM-2 promoter/human CD59) transgenic line were stained with a fluorescein isothiocyanate-labeled monoclonal antibody to human CD59 and analyzed by flow cytometry. Granulocytes were gated using forward and side scatter. Human and non-transgenic mouse granulocytes stained for CD59 are included as positive and negative controls, respectively.
A mutation was also introduced into the P 12 site (Ϫ217). Although this motif does not represent the binding site for any known transcription factors, it has features suggestive of a tandem binding site for the Ah receptor nuclear translocator (35) complexed to an unknown factor. Ah receptor nuclear translocator recognizes the half-site GTG, and the mutation in P 12 (CACGCATGCGTG 3 TTCGCATGCGAA) was designed to destroy the two potential half-sites. However, this mutation had no effect on ICAM-2 promoter activity in BAEC (Fig. 5).
Specific Interaction of BAEC Nuclear Proteins with Elements in the ICAM-2 Promoter-Specific binding of BAEC nuclear proteins to the functional Sp1 and GATA sites was assessed by electrophoretic mobility shift assay. Several retarded bands were observed when BAEC nuclear extract was incubated with a labeled probe containing the Sp1 site, but only the highest molecular weight DNA-protein complex appeared to be specific (Fig. 6A). Excess cold mutant competitor failed to prevent the formation of this complex, indicating that it is likely to contain a protein involved in the regulation of the ICAM-2 promoter in BAEC.
Incubation of BAEC nuclear extract with the upstream GATA probe resulted in the formation of two specific complexes (Fig. 6B), the smaller of which was competed out by cold wild type competitor but not by cold mutant competitor. The downstream GATA probe generated complexes of similar molecular weight to the two formed with the upstream site, in addition to a third complex of intermediate size (Fig. 6C). Although the results in this case were less clear, it appeared that only the intermediate-sized complex was not competitively inhibited by cold mutant competitor.
To determine whether either factor could activate the ICAM-2 promoter in vitro, expression plasmids for mouse GATA-1 or human GATA-2 were co-transfected with the ICAM-2 promoter-luciferase reporter into COS cells. A synthetic GATA-responsive promoter construct (M6␣-SEAP) consisting of six GATA sites linked to a minimal promoter driving the expression of secreted alkaline phosphatase was co-transfected as a control to confirm expression of the GATA factors. As shown in Fig. 7, expression of GATA-1 or GATA-2 transactivated the  (20) in which CD59 expression is driven by the 0.33-kbp ICAM-2 promoter fragment. c and f, kidney and heart of mouse from transgenic line PK in which CD59 expression is driven by the 3.0-kbp ICAM-2 promoter fragment. All vessels in both organs from the two transgenic lines show strong expression of the transgene. Typical glomerular tufts and intertubular sinusoids in the kidney are indicated by short and long arrows, respectively. Magnification ϫ 50. ICAM-2 promoter by 7.3-and 3.4-fold, respectively. The greater activation of both the ICAM-2 and M6␣ promoters by GATA-1 (Fig. 7) may indicate that GATA-1 is a more potent transactivator in vitro than GATA-2 as previously suggested (37). However, since GATA-1 is not expressed in endothelial cells (28), it is likely that GATA-2 regulates ICAM-2 expression in this cell type in vivo.

DISCUSSION
Human ICAM-2 is constitutively expressed on all vascular endothelial cells (23), and the ICAM-2 promoter is unusual among the endothelial cell-specific promoters studied to date in that the signals necessary for specific expression in vivo reside within a very small (0.33-kbp) region of 5Ј-flanking sequence (20). As the first step toward understanding the mechanism of this specificity, we have characterized the ICAM-2 promoter in vitro. After confirming that the specificity of the 0.33-kbp promoter in vivo was paralleled in vitro, we showed that the addition of up to 2.7 kbp of upstream 5Ј-flanking region did not increase promoter activity in transfected BAEC (Fig. 4) and did not affect the pattern of transgene expression in transgenic mice (Fig. 3), suggesting that most or all of the positive-acting elements regulating ICAM-2 promoter activity in endothelial cells are indeed located in the 336-bp sequence shown in Fig.  1B.
Based on previous reports, the obvious candidates for these elements included consensus motifs for the Sp1 and GATA families of transcription factors (Fig. 1C). Functional Sp1 binding sites are required for maximum in vitro endothelial activity of the promoters of the KDR/flk-1 (18), platelet-derived growth factor (38), and eNOS (11) genes; likewise, GATA sites have been identified as positive elements in the eNOS (11), von Willebrand factor (36), P-selectin (14), and endothelin-1 (39) promoters. Mutational and gel shift analyses were used to demonstrate that the Sp1 site and two GATA sites in the ICAM-2 promoter function as cis-acting positive regulatory elements in endothelial cells in vitro (Figs. 5 and 6). Mutation of the Sp1 site abolished specific binding of a BAEC nuclear protein and reduced ICAM-2 promoter activity in BAEC by 70%. Mutation of the upstream and downstream GATA sites disrupted BAEC nuclear protein binding and reduced promoter activity in BAEC by 78 and 61%, respectively. Although the identity of the protein(s) binding the GATA sites was not determined, we believe that it is likely to be GATA-2, because overexpression of GATA-2 transactivated the ICAM-2 promoter in non-endothelial (COS) cells in vitro (Fig. 7), and GATA-2 has previously been implicated in the endothelial cell expression of several genes (11,14,39). GATA-1 also transactivated the promoter but is not expressed in endothelial cells (28). GATA-1 is believed to play a role in megakaryocyte gene expression (40,41) and thus may be involved in the regulation of ICAM-2 expression in this cell type.
Another positive element identified by mutagenesis was an 8-bp palindrome that has been shown to be an important component of an intronic endothelial cell-specific enhancer in the murine tie2 gene (4). This motif, which we have termed P 8 , is also present in the promoter of the endothelial-specific mouse tie1 gene (7) and (conserved in 7 out of 8 positions) in the promoters of the human tie1 (7) and mouse ICAM-2 (34) genes. Mutation of P 8 reduced ICAM-2 promoter activity in BAEC by 70% (Fig. 5). Mutation of another palindrome (P 12 ) did not affect promoter activity in BAEC, and binding of BAEC nuclear proteins to a P 12 probe (bp Ϫ224 to Ϫ209) could not be clearly demonstrated by electrophoretic mobility shift assay (data not shown). However, it is not possible to conclusively rule out a FIG. 4. In vitro promoter activity of 3.0, 1.7, or 0.33 kbp of ICAM-2 5-flanking region. Fragments cloned into a luciferase reporter vector were transiently transfected with pSV␤Gal (to correct for transfection efficiency) into cultured endothelial cells (BAEC) or a nonendothelial cell line (COS). Promoter activity is expressed as corrected luciferase activity relative to that obtained for the vector alone and is shown in this and subsequent figures as the mean Ϯ S.E. of at least two experiments performed in triplicate. Fragment coordinates are expressed relative to the primary transcription start point. kb, kilobase pairs.
FIG. 5. Identification of important regulatory elements in the ICAM-2 promoter by site-directed mutagenesis. Luciferase constructs containing the wild type 0.33-kbp promoter or derivatives thereof, mutated as indicated by stars, were transfected with pSV␤Gal into BAEC. Corrected luciferase activity was calculated, and promoter activity is expressed as a percentage of that of the wild type promoter. role for this site, because preliminary DNase I footprinting experiments indicated that an 18-bp region encompassing P 12 is protected by a protein(s) present in BAEC nuclear extract. 2 Further mutational analysis and identification of proteins binding to P 12 are in progress.
Ets family members are involved in both endothelial (4,38,42) and megakaryocytic (41) gene expression. The function of four Ets motifs in the human ICAM-2 promoter, which are all largely conserved in the mouse ICAM-2 gene (Fig. 1B), was not determined in this study, but it is interesting to note the similarity between the tandem motifs (GCTTCCCAGCTTCCT) centered at Ϫ132 bp in the ICAM-2 promoter with tandem PEA3 motifs (GCTTCCTCCCTTTCCT) in the promoters of the human and mouse tie1 genes (7).
The presence of an NF-B motif in the ICAM-2 promoter is also intriguing. NF-B is a major regulator of the cytokineinducible expression of various genes including the cell adhesion molecules E-selectin, VCAM-1, ICAM-1, and MadCAM-1 (reviewed in Ref. 43). NF-B is also believed to control the constitutive level of expression of P-selectin in endothelial cells and megakaryocytes by a mechanism involving the interaction of Bcl-3 with promoter-bound p50 or p52 homodimers, and mutation of the P-selectin B site causes a 40% reduction in promoter activity in BAEC (44). ICAM-2 is expressed constitutively at high levels on resting endothelial cells and megakaryocytes in vivo, and the level of expression in human umbilical vein endothelial cells, like that of P-selectin (44), was unchanged by treatment with a variety of inflammatory cytokines (23). Incubation of nuclear extract from resting BAEC with a probe (bp Ϫ285 to Ϫ269) containing the ICAM-2 B site resulted in the formation of a specific DNA/protein complex. 3 Together, these data suggest that the role (if any) of the ICAM-2 B site in endothelial expression is similar to that of the P-selectin site, although this awaits confirmation by mutagenesis and gel supershift analysis.
The small size and tight specificity of the ICAM-2 promoter make it a particularly promising tool for gene therapy and other applications requiring endothelial cell targeting. It is also valuable in the study of the control of gene expression in endothelial cells, as demonstrated in this study. On the basis of the results presented here, we propose that the transcription of ICAM-2 in endothelial cells is regulated primarily by the binding of Sp1 and GATA-2 transcription factors and possibly a novel P 8 binding factor to their cognate sites in the ICAM-2 promoter. Interaction of these DNA-bound proteins via their zinc fingers (45) may induce a conformational change, allowing them to interact more readily with components of the basal transcriptional machinery. For example, as suggested for GATA-1 (45), the interaction of GATA-2 with Sp1 may increase the affinity of the latter for TAF II 110 or some other component of the basal complex. The involvement of other proteins, such as members of the Ets and NF-B families, remains to be determined.
Although Sp1 and GATA-2 may be two of the major determinants of ICAM-2 promoter activity in endothelial cells, their presence alone cannot account for the specificity of the promoter, since Sp1 is ubiquitously expressed, and GATA-2 is expressed relatively widely (37). It is possible that a threshold level of GATA-2 is required for activation of the promoter. Alternatively, GATA-2 may interact with a cell-specific factor to activate transcription of ICAM-2 in endothelial cells. Such an interaction between GATA-1 and the protein FOG (Friend of GATA-1) has been proposed to explain the transcriptional regulation of hematopoietic-specific gene expression (46). It is not known whether FOG binds DNA in a sequence-specific manner (46), but if so, it is tempting to speculate that either the P 8 or P 12 motifs in the ICAM-2 promoter may represent the binding site of an endothelial-restricted FOG-like protein. Finally, it is conceivable that ICAM-2 expression in non-endothelial cells is repressed by a negative-acting factor such as that described for the von Willebrand factor promoter (6). The identification of 3 C. M. Pedic, unpublished results.
FIG. 6. Determination of BAEC nuclear protein binding to the ICAM-2 promoter by electrophoretic mobility shift assay. The indicated radiolabeled oligonucleotide probes were incubated with (ϩ) and without (Ϫ) BAEC nuclear extract and in the presence or absence of a 100-fold molar excess of cold wild type (WT), mutant, or nonspecific (N/s) competitors as indicated. Specific complexes are indicated by arrowheads. A, Sp1 site probe (bp Ϫ207 to Ϫ183); nonspecific competitor, bp Ϫ285 to Ϫ269. B, upstream GATA site probe (bp Ϫ155 to Ϫ135); nonspecific competitor, bp Ϫ207 to Ϫ183. C, downstream GATA site probe (bp Ϫ61 to Ϫ43); nonspecific competitor, bp Ϫ207 to Ϫ183. specific proteins interacting with the ICAM-2 promoter using techniques such as gel supershift and DNA footprinting will be helpful in discriminating between these possibilities.