A T-cell Enhancer Cooperates with NF-KB to Yield Cytokine Induction of E-selectin Gene Transcription in Endothelial Cells*

ELAMl (endothelial leukocyte adhesion molecule 1, also known as E-selectin) is a highly tissue-specific adhesion molecule that is transiently and exclusively expressed on cytokine-induced endothelial cells. We have identified two proximal ELAMl promoter ele- ments and their DNA-binding factors that are, in addition to NF-KB, essential for ELAMl transcription. Mutation of either element in promoter constructs carrying the first 383 nucleotides of the ELAMl promoter markedly diminished the expression of a fused chloramphenicol acetyltransferase reporter gene. Although multimers of either element failed to display enhancer activity on its own, fusion of the most upstream of these to the NF-KB element had a strong stimulatory effect. This site, ACATCAT, is recognized by a factor we have called NF-ELAM1. The site corresponds to NF-ELAMl’s preferential binding sequence (A/ T)CA(G/T)CA(G/T) as determined in a target definition assay. This element is identical to the T-cell 6A en- hancer found in the T-cell receptor-a, -8, and CD36 genes. Our results suggest that the 6A/NF-ELAMl element can function as a modulator of NF-KB in endothe- lial cells both as well as a T-cell


genes. Our results suggest that the 6A/NF-ELAMl element can function as a modulator of NF-KB in endothelial cells both as well as a T-cell enhancer.
The human endothelial leukocyte adhesion molecule (ELAM1' also known as E-selectin) is a surface-expressed glycoprotein found exclusively on cytokine-induced endothelial cells (1). ELAMl mediates adherence and extravasation of a subset of leukocytes (2-7). Expression of ELAMl is correlated with several acute and chronic pathological conditions such as asthma and psoriasis (8). In this manner it plays a n important role in the inflammatory response. Understanding the mechanisms of regulation involved in ELAMl expression should shed light on the means by which the endothelium is activated to attract the appropriate immune cells to a site of injury.
Consistent with its role in leukocyte tissue infiltration, ELAMl is induced transiently on endothelial cells where both its protein and mRNA levels peak about 4 h post-IL-lD or tumor necrosis factor la treatment, returning to near basal levels 20 h thereafter (3, 9). Studies on the mechanism(s) of this regulation have previously shown that both the ELAMl * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
promoter's transcription activity and its mRNA stability are under the control of newly synthesized proteins.' Genomic sequences covering the ELAMl promoter have recently been isolated (10)(11)(12). Within these sequences a binding site for the transcription factor NF-KB was found at -90 base pairs relative to the transcription start site. In endothelial cells cytokine treatment activates NF-KB DNA binding activity. The importance of NF-KB for enhancing ELAMl transcription was shown by mutation of the KB site. The mutant chimeric ELAMl promoter/CAT construct is unable to induce reporter activity in response to cytokine treatment of the transfected cells (10). In addition, cycloheximide plus IL-1 superinduces NF-KB DNA binding activity in endothelial cells. These results provide at least a partial explanation for the enhanced transcription activity of the ELAMl promoter induced by protein synthesis inhibitors. ' In spite of its central role in controlling the ELAMl promoter, NF-KB by itself does not appear to be sufficient to mediate ELAMl activation. The time course of NF-KB activation and ELAMl induction do not correspond late after cytokine treatment. At 24-h post-IL-1 induction NF-KB, DNA binding activity is still significant, whereas run-on transcription of ELAMl mRNA is undetectable.' Moreover, Whelan et al. (10) have demonstrated that sequences upstream of the NF-KB site between -233 and -117 are essential for the activity of ELAMl promoter/CAT constructs, both in endothelial and non-endothelial cells.
We have undertaken to identify the factors which interact with NF-KB in causing cytokine induction of ELAMl gene transcription. In this paper, we characterize the DNA-binding sites for a number of additional factors necessary for cytokine induction. We show that specific mutation of these sites correlates with loss of transcription activity. In addition we show here cooperation between factors binding one of these sequences and NF-KB. This cooperation appears to be a control feature of the mechanism by which cytokines induce ELAMl transcription. The constitutive binding of this factor may also provide a means of repressing ELAMl expression during the quiescent state.

EXPERIMENTAL PROCEDURES
Cell Lines-Human umbilical vein endothelial cells (HUVEC) were extracted from human umbilical cords by collagenase treatment. The cells were cultured in medium MCDB 131 supplemented with epidermal growth factor (10 ng/ml), hydrocortisone (1 ng/ml), bovine brain extract containing heparin, 2% fetal bovine serum, gentamicin, and amphotericin (Clonetics, CA). IE-7 cells are simian virus 40 (SV40)transformed primary HUVECs. This cell line was originally cloned under the name SGHEC-7 (32) and was grown in the endothelial growth medium described above containing geneticin (300 pg/ml).

T-cell Enhancer and NF-KB Cooperation
HeLa cells were grown in Dulbecco's modified Eagle's medium containing 10% fetal calf serum and 50 pg each of penicillin and streptomycin/ml. IL-1 treatment (40 units/ml) was for 4 h in fresh medium prior to nuclear extract preparation or during 24 h following the day after DNA transfection of CAT constructs. Nuclear Extract Preparation-HUVEC (passage 5-6) and HeLa nuclear extracts were prepared as described (13) with modifications (14). Protein concentrations were adjusted to 3 pg/pl and aliquots stored at -70 "C.
Probe Preparations-The end-labeled probes corresponding to the ELAMl promoter were prepared as described (15). Briefly, 10 pmol of one of the flanking oligonucleotides (18-34 nucleotides) were endlabeled with 10 pl (100 pCi) [y-32P]ATP, in 20 mM Tris-C1, pH 7.75, 10 mM MgC12, 50 mg/ml bovine serum albumin, 10 mM j3-mercaptoethanol, and 10 units of polynucleotide kinase (Biolabs), in 15 pl of total volume. After 45 min a t 37 "C, the enzyme was inactivated at 65 "C (10 min) and other components for the PCR (polymerase chain reaction) added; 10 pl of 10 X PCR buffer (is 0.5 M KCl, 0.1 M Tris-C1, pH 8.4, 1.5 mM MgC12, and 1 mg/ml gelatin), 10 pl of dimethyl sulfoxide, 4 pl of dNTPs (5 mM each, pH 7.5, Pharmacia), 25 pmol of the other primer, -0.5 pg of plasmid DNA containing the ELAMl promoter insert as template, and Hz0 to 100 pl. The sample was heated 5 min at 99 "C in a thermocycler (Ericomp, Twinblock, CA), cooled on ice, spun briefly to collect condensate, and supplemented with 0.5 pl (2.5 units) Taq polymerase (Cetus). Fifty pl mineral oil (Sigma) was overlaid and the sample was cycled 1.5 min at 94 "C, 1.5 min 55 "C, and 30 s at 72 "C (25 cycles), followed by a 5 min extension a t 72 "C. The probe was concentrated by precipitation and either loaded directly on gel or methylated with dimethyl sulfate (5 pl/ml in cacodylate buffer, Ref. 20) and reprecipitated before gel-purification.
Short probes for band shift experiments were generated by endlabeling as described above. After inactivation of the kinase, 25 pmol of the complementary oligonucleotide and 1 pl 5 of M NaCl were added, the mixture was heated 2 min at 65 "C, and allowed to anneal for at least 30 min at room temperature.

5'CAGAGTTTCTGACATCATTGTAATTTTAAGC
Target Definition Assay-The procedure we developed is a modification of those described previously for purified or cloned transcription factors (16)(17)(18) and is schematically represented in Fig. 4a. The degenerate probe was made by extending 50 pmol of primer 2 (CCAC-GAAGCTTAAAATTA) on 10 pmol of "N7" template (GCTGAC-G/A/T/C) using Klenow DNA polymerase. Ten pmol of the doublestranded oligonucleotide was added to band shift reaction mixtures, which was the maximum amount that did not inhibit the amount of labeled (specific) complex. Such inhibition would suggest nonspecific binding of the N7 oligonucleotide to NF-ELAM1. The retarded NF-ELAMla complex was cut out of the gel, sliced into pieces, and soaked overnight at 37 "C in 500 pl of T E 10 mM Tris-C1, pH 7.5, and 1 mM EDTA plus 0.5% SDS. Twenty pg of tRNA was added to the eluate, and the nucleic acid was recovered by phenol/chloroform extraction and ethanol precipitation, followed by a 70% ethanol wash of the pellet. Half of the sample was used as template in a PCR set up as described above containing 10 pmol of primer 1 and 10 pmol of GCTGACCAAAGCTTCTG as primer 2. In addition, this reaction CAAAGCTTCTG-NNNNNNN-TAATTTTAAGCTTCGTGG, n = was set up with a "hot start," i.e. the Taq enzyme was added only when the reaction mixture had reached 55 "C. This was done to avoid amplification of the labeled probe used in the band shift that shares sequence similarity with the N7 oligo. This cycle of selection/amplification was repeated (the amount of amplified N7 to add in the band shift was again determined empirically), and the purified DNA was cleaved with HindIII, precipitated, and ligated to obtain multimers for 4 h. An aliquot was ligated into M13 (HindIII-cut, phosphatasetreated) and the inserts sequenced using the dideoxy procedure (19).
DNase I Protection Assay-Binding reactions were set up in BCA without EDTA but containing 10 mM MgClz and after preincubation on ice digested 20 min at room temperature with DNase I (Boehringer, 4 units/ml final concentration). The DNA was extracted twice with a mixture of phenol and chloroform (l:l), precipitated, and run on an 8% polyacrylamide (1:9 bisacrylamide), 50% urea, 0.5 X TBE gel.
Methylation Interference Assay-Five-fold scaled-up binding reactions were prepared using the partially methylated probe. The retarded and free probes were separated on a 4% gel and the DNA isolated by electrotransfer to DEAE membranes and further processed as described (20).
Promoter-CAT Constructions and CAT Assay-Constructs were made essentially as described (10) using the oligonucleotides listed in Fig. 7a. In block mutations the Gs were changed into T, the As into C, and vice versa. DNA transfection and CAT assays were performed as in Refs. 21-23. The thin layer chromatography plates were directly scanned in an Ambis (San Diego, CA) radioactivity scanner.
Database Searches-Database searches for transcription factor recognition sites were performed on the Eukaryotic Transcription Factor Recognition Sites Release 3.0 database, as supplied by Dr. D. Ghosh (University of Wisconsin) to the GCG Software.

RESULTS
Two DNA-binding Motifs in Addition to KB Are Found in the ELAMl Promoter Region Responsible for Cytokine Induction-We wished to determine what upstream sequences aside from the NF-KB-binding site might play a role in cytokine induction of ELAM expression. We have previously shown that the region from 233 to 117 nucleotides preceding the transcription start site is necessary for induction (lo), yet contains no binding sequences for known factors (Refs. 10-12 and Fig. 1). To identify nuclear proteins which would recognize binding motifs on this DNA fragment, band shift analyses were carried out with a relatively long probe (-163 to -75) covering this region of the ELAMl promoter. To generate this fragment PCR synthesis was carried out with radiolabeled oligonucleotide primers (see Experimental Procedures). The resultant DNA fragment was specifically recognized by several factors from a nuclear extract of primary human umbilical cord vein endothelial cells (HUVECs) (Fig.  2). One of these bands is induced by IL-1 treatment. Competition with unlabeled KB oligonucleotide (Fig. 2, last lane) indicates that this band corresponds to NF-KB. We have named the other four complexes NF-ELAMla and -1b and NF-ELAM2a and -2b. All complexes were stable in the presence of up to 5 pg of nonspecific competitor DNA (poly(d1-dC), suggesting tight and specific DNA-protein association. A

T -T T G C A T A T A C G A T~T~G G U L T G G A C A R R
-150 -130 -110 All these bands could also be detected in HeLa nuclear extracts (data not shown). Determination of the NF-ELAM1 -binding Site Suggests Identity to a T-cell Enhancer-We next asked whether we could more precisely determine the sequences to which these factors bound. To accomplish this we undertook DNase I footprinting experiments using the same DNA region as in the band shift experiments for the footprinting substrate. A single footprint was observed between nucleotides -153 and -144, suggesting that these nucleotides interacted with DNA binding factor(s) (Fig. 3a). Both endothelial and HeLa cell extracts produced the same footprint (lanes 1-4 versus 6-8). A synthetic oligonucleotide spanning this footprint was tested in band shift assays using nuclear extracts from HUVEC cells (Fig. 3b). The pattern obtained suggests that the -153 to -144 site is recognized by NF-ELAMla and -lb, the slower migrating complexes in Fig. 2. This was confirmed by the observation that this oligonucleotide, when used as unlabeled competitor, inhibited NF-ELAMla and -1b binding to the labeled -163 to -75 fragment in band shift experiments (Fig.  2, EL1 lane). Although Fig. 2 also shows partial competition with NF-ELAM2b, results described below clearly show that this factor binds elsewhere. Since the NF-ELAMla/b bands were competed away by an oligonucleotide carrying a wild type sequence (Fig. 3b, lane 3), whereas a mutant sequence did not (lane 4 ) , NF-ELAMla/b are specific for the -153 to -144 sequence revealed by the DNase I footprint.

MTCCCTC-GGCCTCAGCCGAAGTAGTGTTCAGCTGTTCTTGGCTGACTTCA
The "smear" corresponding to NF-ELAMlb possibly represents a partially degraded NF-ELAMla complex, as it shows identical behavior with respect to the competitors. In addition these extra bands varied somewhat with different nuclear extracts, suggesting variable degradation, of NF-ELAMla. No difference was observed between extracts from noninduced (lune I ) or IL-1 induced (lune 2) cells. We conclude that the sequence between -153 and -144 is the binding site for a factor which is present in both endothelial cells and nonendothelial cells and which binds the element regardless of cytokine treatment of the cells.  1 (lanes 2-4 and lune I , respectively). Lane 3, competition with 20 pmol of the unlabeled NF-ELAM1 recognition sequence (-163 to -133). Lane 4 , competition with the block mutant to the NF-ELAMla band.
(same sequence, but -154 to -144-mutated). The arrowhead points To further define the precise binding sequence for NF-ELAM1, two approaches were undertaken. First, a panel of oligonucleotides was made carrying point mutations in the sequence protected in the DNase I footprint. These oligonucleotides were then tested for their ability to compete in NF-ELAMl band shift assays. Results of these experiments showed that the core of the NF-ELAM1 recognition sequence consists of the ACATCAT heptamer (data not shown). Our second approach allowed us to define the optimal NF-ELAM1-binding site sequence and to detect sequences which bound well but were not in the native promoter (i.e. "up mutations"). To accomplish this we used a modification of the "target definition assay" (see "Experimental Procedures" and Fig. 4u). This method aimed at determining all sequence combinations which are specifically retained by NF-ELAMla. Fig. 4b shows all the sequences thus obtained using HUVEC and HeLa extracts, oriented for the best match to the ACAT-CAT sequence. The histogram in Fig. 4c shows that the consensus target for NF-ELAMla is (A/C)CA(G/T)CA(G/ T). A computer-assisted comparison of this sequence with a database of transcription elements (see "Experimental Procedures") failed to yield perfect matches. However, we did note that the NF-ELAMla-recognized TGACATCAT sequence (-154 to -146) is identical to the critical portion of the 6A element. This 6A element has been identified in both the T-cell receptor-a, -& and CD36 promoters and has been reported to be a T-cell-specific enhancer (24). The sequences for these elements have been aligned in Fig. 4d. Since the NF-ELAM1-binding sequence is identical to part of the 6A element it may bind a factor similar to that bound at this site in the other promoters.

NF-ELAM
Determination of the NF-ELAM-2-binding Site-While DNase I footprinting defined more precisely the NF-ELAM1binding site (see above), the NF-ELAM2-binding site required a different approach. A methylation interference assay (20) was performed to localize the binding site for NF-ELAM2a which had been demonstrated by the gel mobility assay. Fig.  5a shows the G ladder obtained from unbound DNA (lanes 3 and 5 ) and for DNA retarded by NF-ELAM2a from HUVEC (lane 2) or HeLa (lane 4 ) cells. Careful comparison of the ladders indicates that 3 clustered guanosines at positions -104, -103, and -100 are decreased in intensity, suggesting that NF-ELAM2a binds at this site. To confirm the involvement of these clustered G residues in NF-ELAM2a (and -2b) binding, probes were generated by PCR covering the NF-KB site and its 5' adjacent region (until position -132). As can be seen in Fig. 5b (lane 11, this probe is retarded by NF-KB, NF-ELAM2a, and -2b from IL-1-activated HUVEC nuclear extract. A faint band can be seen between NF-KB and NF-ELAM2a, also visible in Fig. 2, that probably corresponds to an NF-KB degradation product. A second probe was made covering the same ELAM promoter region, in which the 3 G residues at position -104, -103, and -100 were mutated into Ts. This probe was also generated by PCR, using a mutated promoter CAT construct (see Fig. 6) as template. As is clear from Fig. 5b (right lane), this mutated probe is no longer recognized by NF-ELAM2a and -2b.
As with NF-ELAMla and -lb, NF-ELAM2a and -2b showed considerable variation in relative abundance for different batches of nuclear extracts (not shown). The results are consistent with 2b being a breakdown product of 2a, rather than both being members of a family with similar DNAbinding sites.
The "core" of the NF-ELAM2 element, GGATG, was compared with a database of transcription factor-binding sites (see "Experimental Procedures"), but no similarities were found.
Both NF-ELAM1-and -2-binding Motifs Are Required for ELAM Transcription-The functional importance of sequences around both the -150 and -100 nucleotides of the promoter in the cytokine induction of ELAM transcription was further tested in a reporter fusion assay. We have shown previously that both a 383-and 233-base pair ELAMl promoter fragment fused to the CAT reporter gene display cytokine inducibility of the reporter (10). As we show here in Fig.  6a and b, a construct with only 181 base pairs upstream sequence is still IL-1-inducible. However, we also found that when the 10 nucleotides between -153 and -144 were altered in a "blocked mutation, these constructs no longer responded strongly to cytokine induction by reporter expression in the IE-7 endothelial cells (Fig. 6a, panel A and Fig. 6b) or in HeLa cells (Fig. 6a, panel B and Fig. 6c). Similarly, when the 3 Gs at position -104, -103, and -100 were replaced by Ts, CAT induction was markedly reduced (Fig 6). In contrast, mutation of a randomly chosen block of 10 nucleotides within this region (-161 to -170) had no effect on transcription. From these results we conclude that both the NF-ELAM1-and -2binding sites are needed for full cytokine induction of the ELAMl reporter. Thus, at least three DNA elements binding transcription factors (NFKB, NF-ELAM1, and NF-ELAM2)  (lanes 4 and 5). The G residues whose methylation interferes with NF-ELAM-2a binding are indicated by stars. Numbering of the G residues is as in Fig. 1. Panel b, band shift assay using nuclear extract from IL-1-induced HUVEC cells. The probe ( W T ) corresponded to the -132 to -65 ELAM promoter. In the mutated probe ( M U T ) , the G residues at position -104, -103, and -100 had been mutated to Ts. must be occupied to give the expected high level of transcription seen with the ELAMl promoter following cytokine treatment of endothelial cells.

T-cell Enhancer and NF-KB Cooperation
NF-ELAM1 Shows Cooperation with the NFKB Enhancer Activity-We next wished to test whether the -144 to -153 sequence element (the NF-ELAM1-binding site) has enhancer activity of its own. Constructs carrying multimers of only this element linked to the enhancerless SV40 promoter fused to the CAT gene were tested for reporter expression in vector-transfected cells (Fig. 7 A ) . As can be seen in Fig. 7b, constructs carrying one or three NF-ELAM1 elements do not display any constitutive enhancer activity. Moreover, enhancer activity is not found following IL-1 treatment of the cells. These results suggest that NF-ELAM1 does not act independently as a transcription enhancer either constitutively or when in an IL-1 stimulated milieu. We then constructed vectors that carry contiguous NF-ELAM1 and NF-KB elements to test the influence of the two elements together. In these experiments the NF-ELAM1 motif combined with that of the NF-KB element acted to strongly augment the weak IL-1-inducible enhancer activity seen with that of an isolated NF-KB motif (Fig. 7b). As shown in a summary of the data in Fig. 712, introduction of two NF-ELAM1 elements into a construct already carrying two NF-KB elements greatly increased cytokine-inducible reporter expression in both HeLa and HUVEC cells. We conclude that NF-ELAM1 is a DNA-binding factor which acts in concert with NF-KB to enhance cytokine-induced transcription from the ELAMl promoter.
We next tested the isolated NF-ELAM2 element for enhancer activity. As can be seen in Fig. 7d, single or four consecutive copies of this element failed to augment constitutive or IL-1-induced transcription, even when linked to an NF-KB site. Even a construct carrying three copies of the NF-ELAM~-KB tandem yielded only 13-fold induction (not shown), which compares poorly to a construct carrying just two NF-KB elements (Fig. 7b) that shows 9-fold acetylation induction.

T A C G T M C T G T A G T M C M C C C C T~T A G _ " -c Zx(NF ELlINFkB) 3x(NF ELlINFkB) N I l l NF.EUM2 NFrB O I T C A T G C A T l T G W L T G C C T l l C C T
* 'x(NF'EwNFkB' Band shift experiments with the oligonucleotides that were used to make this latter series of constructs showed good binding of NF-ELAM2a and -b (not shown), indicating that lack of NF-ELAMZ enhancer activity of the constructs cannot be ascribed to diminished factor binding.

T f f i T~C C T A f f i G T * U ; C C C T~A C T A G c + +
Interestingly, all constructs carrying both NF-ELAM2 and KB elements (including the one carrying three copies of each) showed reduced background acetylation (0.7% or less) in absence of IL-1 (compare with the last five constructs in Fig.  7b).

DISCUSSION
We have described two sequence elements and their binding factors, NF-ELAM1 and -2, that are both required for ELAMl transcription. Interestingly, the -141 to -153 dement recognized by NF-ELAMla and -l b has no enhancer activity on its own, but is capable of up-regulating a neighboring NF-KB element. Our mutation analysis clearly identifies both NF-ELAMl and -2 elements as essential for ELAMl promoter activity. Both elements have very been well conserved among human, mouse and rabbit ELAMl promoters (10, 25, 26). Moreover, DNA-binding factors comigrating with NF-ELAMla and -b were detected in mouse heart and lung tissue (25).
Surprisingly, fine mapping of the NF-ELAM1 element revealed a sequence with a high degree of similarity to the 6A element, reported to be a T-cell-specific enhancer (24). Our data show that the 6A\NF-ELAM1-binding site is unable to function as an independent enhancer in non-T-cells such as HUVEC, IE-7, and HeLa consistent with the published report. In contrast, we demonstrate a wider role for this element and its binding factor. We have shown that NF-ELAM1 acts as a modulator of NF-KB activity in endothelial cells. The GA\NF-ELAMl element thus appears to be involved in different activation pathways in T-cells compared to endothelial or other cells. Whereas in T-cells the element is active as an enhancer (presumably through a T-cell-specific factor) in other cells the element might require the cooperation of a second transcription factor (such as we have found with NF-KB) to elicit its transcriptional activity.
We have also noted a similarity between the ACATCAT NF-ELAM1-binding site and the ACGTCAT recognition sites for the CREB/ATF family. We initially dismissed the possible binding of these factors to the NF-ELAM1 element because 1) no CREB/ATF-binding sites had been described carrying the G to A transition as seen in the NF-ELAM1-binding sequence, and 2) our target definition assay clearly indicated a preference of NF-ELAM1 for the sequence containing an A. Nevertheless, Georgopoulos et al. (27) recently reported the cDNA cloning of a CRE-BP cDNA, whose encoded protein specifically recognizes the 6A element. We are currently in the process of cloning cDNAs whose encoded proteins bind this element in the ELAMl promoter.
Contrary to NF-ELAM1, we have been unable to detect cooperation between isolated NF-ELAM2 and KB elements. Given our finding that three point mutations that abolish NF-ELAM2a and -2b binding also inhibit transcription, the make the constructs. Panel b, the enhancerless SV40 promoter was tested in IE-7 cells with single or multiple NF-ELAM1 (NF-ELI) or NF-KB (NFKB) elements, or combinations of these (NF-EUINFKB). Panel c, summary of the results for IL-1-inducible activity of constructs carrying two copies of the NF-KB element uers'sus constructs carrying two dimers of NF-KB plus NF-ELAM1 in endothelial and HeLa cell lines. Panel d, enhancer activity of the NF-ELAM2 element. The activity was tested in HUVEC cells treated or not with IL-1.
by guest on March 23, 2020 http://www.jbc.org/ Downloaded from conclusion must be that the NF-ELAM2 element is highly dependent on its location and context in the ELAM promoter for activity. By contrast, basal level transcription from constructs carrying both NF-ELAM2 and KB elements is reduced as compared to constructs carrying one or more KB elements. One might therefore speculate that the NF-ELAM2 element and its binding factors are involved both in repressing leaky activity of the KB enhancer in the quiescent state and increased transcription upon IL-1 induction. Only the second function requires cooperation with other promoter elements. The hierarchy of cooperation that we have found is suggestive of a sequential order of events: DNA binding of NF-ELAM2 stabilizes NF-ELAM1, which in turn augments NF-KB binding and transactivation.
We have looked for cooperative DNA binding by NF-ELAM1, -2, and NF-KB in band shift experiments with DNA fragments carrying all sequence elements. All factors seem to bind independently (Figs. 2,5b,Ref. 15 and data not shown). Moreover, we were unable to show cooperative binding to the isolated DNA fragment. However, due to the very high probeto-complex ratio in band shift assays protein-protein cooperation may not be apparent in these experiments. Since two bands are obtained with the NF-ELAM1-and -2 probe each the slower migrating complexes might consist of NF-ELAM1, -2, and/or NF-KB. However, as these bands are not affected by IL-1 induction, NF-KB is probably not involved. In addition, a complex between NF-ELAM1 and -2 would comigrate whether NF-ELAM1 or NF-ELAM2 oligonucleotide was used as the probe, which is something we do not see. The possibility remains that NF-ELAM1, -2, and NF-KB interact directly or indirectly in vivo or that their transcriptional activity is not a function of DNA binding cooperation but rather a synergistic interaction with the transcription complex.
NF-KB activation has been shown to be involved in cytokine induction of a number of genes (28,29). Yet additional elements and factors would be expected to generate in each case a tissue-specific and temporally regulated response. In spite of high levels of activated NF-KB, ELAMl transcription is essentially shut off 24-h post-IL-1 induction.' This situation contrasts with the prolonged time course of intercellular adhesion molecule 1 induction in endothelial cells (30) that has also been suggested to depend on NF-KB activation (31). For ELAM1, the two elements and their corresponding factors reported here to be essential for transcription may be involved in down-modulation as well, a hypothesis we are currently pursuing.
The definition of the mechanisms involved in ELAMl expression will allow a better understanding of the means by which the inflammatory response is regulated at the gene level. This understanding may in turn provide better means to control and influence this important element of the immune defense mechanism.