Lipopolysaccharide induction of tissue factor expression in THP-1 monocytic cells. Protein-DNA interactions with the promoter.

Tissue factor, the cellular receptor for factor VII/VIIa, activates both the intrinsic and extrinsic pathways of blood coagulation. In this analysis we have used DNase I footprinting to map the sites of protein-DNA interaction along the promoter (-383 to +8) using nuclear extracts prepared from uninduced and lipopolysaccharide-induced THP-1 cells. We have identified six regions that interact with nuclear factors in both uninduced and induced extracts. Four footprints are contained within a region reported to confer base-line high level expression and lipopolysaccharide and serum induction. Two additional footprints map to a region reported to reduce basal transcription by 50%. The only qualitative change in the footprint pattern with uninduced and induced extracts is the appearance of two hypersensitive sites with uninduced extracts. In addition, changes in the level of protein- DNA binding are detected with only one probe by DNA mobility shift analysis. A combination of well characterized transcription factors (AP1), primarily lymphoid cell specific regulatory proteins (NF-kappa B- and/or Ets-1-related proteins), as well as additional, uncharacterized proteins appear to interact with these sequences. Our data suggest that post-translational modification of existing transcription factors, and not induction of new DNA-binding activity, mediates the lipopolysaccharide induction of tissue factor synthesis in THP-1 cells.


Lipopolysaccharide Induction of Tissue Factor Expression in THP-1 Monocytic Cells
PROTEIN-DNA INTERACTIONS WITH THE PROMOTER* (Received for publication, July 23, 1993, and in revised form, September 8, 1993)

Maryann Donovan-PelusoS, Lisa Dawn George, and Andrea Cortese Hassett
From the Department of Pathology, University of Pittsburgh School of Medicine, Pittsburgh, Pennsylvania 15261 Tissue factor, the cellular receptor for factor VIYVIIa, activates both the intrinsic and extrinsic pathways of blood coagulation. In this analysis we have used DNase I footprinting to map the sites of protein-DNA interaction along the promoter (-383 to +8) using nuclear extracts prepared from uninduced and lipopolysaccharide-induced THP-1 cells. W e have identified six regions that interact with nuclear factors in both uninduced and induced extracts. Four footprints are contained within a region reported to confer base-line high level expression and lipopolysaccharide and serum induction. T w o additional footprints map to a region reported to reduce basal transcription by 50%. The only qualitative change in the footprint pattern with uninduced and induced extracts is the appearance of two hypersensitive sites with uninduced extracts. In addition, changes in the level of protein-DNA binding are detected with only one probe by DNA mobility shift analysis. A combination of well characterized transcription factors (APl), primarily lymphoid cell specific regulatory proteins (NF-KB-and/or Ets-1-related proteins), as well as additional, uncharacterized proteins appear to interact with these sequences. Our data suggest that post-translational modification of existing transcription factors, and not induction of new DNA-binding activity, mediates the lipopolysaccharide induction of tissue factor synthesis in THP-1 cells.
Tissue factor (TF),l the cellular receptor for factor VIZNIIa, binds its ligand and, in the context of phospholipid, activates both the intrinsic pathway of coagulation through the activation of factor M and the extrinsic pathway through activation of factor X (1). Although synthesis of TF by vascular monocytes is tightly regulated, expression can be induced by a variety of agonists, including cytokines (colony-stimulating factor-1 (2) and tissue necrosis factor-a (3)); lipopolysaccharide (LPS), a component of the Escherichia coli cell wall (4,5); and phorbol * This work was supported in part by National Institutes of Health  7). Inappropriate induction of TF synthesis is associated with a number of diseases, including acute myelogenous leukemia (8), fibrosarcoma (91, macrophages associated with ulcerated atherosclerotic plaques (lo), and herpes simplex virusinfected endothelial cells (11). Induction of TF expression by monocytes and endothelial cells is associated with disseminated intravascular coagulation, a rapidly fatal disease in humans. The causative role of TF expression in human disseminated intravascular coagulation is strengthened by experiments, using a baboon model, that block the lethal effects of bacterial infusion by administration of either anticoagulants (12) or blocking antibodies to TF (13).
Analysis of TF gene promoter-luciferase reporter gene constructs following transfection into Cos 7 cells defined the minimal promoter required for high level expression to be contained within 383 bases from the mRNA start site (14). In addition, this region contains at least one and perhaps two elements that mediate serum induction, although no sequence homologous to a conventional serum response element has been identified.
Induction of TF mRNA accumulation in monocytes and in a monocytic cell line, THP-1 cells (151, by LPS results from both transcriptional activation and an increase in mRNA stability (4,5). Transfection of a deletion series of TF promoter-luciferase reporter gene constructs into THP-1 cells identified a 20base pair (bp) region between -192 and -172 relative to the cap site to be sufficient to mediate LPS-induced transcriptional activation (16).
To begin to define the molecular events that modulate TF gene transcription, these studies have characterized the interaction of nuclear factors with the TF promoter. We have prepared nuclear extracts from both uninduced and LPS-induced THP-1 cells in order to compare binding activity with the promoter under conditions where the gene is either expressed at low level (uninduced) or actively transcribed (induced) in these cells. An induction period of 1.5 h was used in these studies since our data, as well as data from other groups, have determined that TF mRNA synthesis is rapidly induced with maximal expression between 1 and 2 h in THP-1 cells (5). We have used a combination of DNase I footprinting to position areas of DNA-protein interaction and DNA mobility shift assays (DMSA) to both quantify and more precisely characterize the nature of the proteins that interact with these sites.

EXPERIMENTAL PROCEDURES
Oligonucleotides and Plasmids-For DNase I footprinting experiments an EcoRI to SacI tissue factor promoter fragment of 391 bp was cloned into the pBluescript I1 polylinker (KS, Strategene) to generate pTFP. pTFP was digested with SacI and the 3' overhang was removed by the 3'6' exonuclease activity of T4 DNA polymerase. XbaI linkers (New England BioLabs) were ligated into the blunted SacI site to generate pTFPX. For DNase I footprinting the following constructs were made. pTFPXASma contains a 167-bp deletion of an internal SmaI fragment that fuses sequences from -383 to -280 to -113 to +8. pTF-PSma contains the 167-bp SmaI fragment with promoter sequences between -280 and -113. Fragments and orientation relative to plasmid polylinker sites are described in Fig. 2. Promoter DNA fragments were digested with EcoRI and XbaI (New England BioLabs), separated by polyacrylamide gel electrophoresis (5% acrylamide, 29:1), and eluted into sterile water. Fragments were 3'-end-labeled on one strand with the large fragment of DNA polymerase I (Klenow) at either the EcoRI site with [a-"PIdATP or the XbaI site with [CY-~~PI~CTP.
Oligonucleotides were synthesized on an Applied Biosystems model 394 DNA synthesizer. Several small fragments used in DNA mobility shift experiments were either purified from TF subclones (Sty 84) or synthesized by polymerase chain reaction (PCR) using promoter fragments as template and standard PCR conditions (PCRl). Oligonucleotides were synthesized to encompass binding sites identified by DNase I footprinting. DNA fragments were annealed in 0.1 M NaCl and were 5'-end-labeled with [y-"P]ATP and T4 polynucleotide kinase (Boehringer Mannheim). Restriction fragments were 3'-end-labeled with d2P-labeled dNTPs and Klenow. Probes used in DMSA are described in Table I.
Cell Lines-THP-1 cells were grown in spinner flasks (Bellco Glass) in RPMI supplemented with 5% heat-inactivated fetal calf serum (low endotoxin, HyClone) and 10 p~ P-mercaptoethanol as described elsewhere (18). For induction experiments, cells were centrifuged, washed in Ca2*/Mg2*-free phosphate-buffered saline (Dulbecco): 137 m~ NaCl; 2.7 m~ KCl; 8 m~ Na2HP04; 1.5 m~ KH2P04; pH 7.2, and grown in serum-free RPMI containing E. coli LPS (Sigma) a t a final concentration of 1 pg/ml for 1.5 h. To ensure induction of TF mRNA under these conditions, mRNA was prepared from both the uninduced and LPSinduced cells (19). and expression of T F mRNA was examined by Northern blotting and hybridization with a radiolabeled TF cDNA probe (20). Preparation of Nuclear Extracts-Nuclear and SlOO (cytoplasmic) extracts were prepared from THP1, WI38, and HEPG2 cells according to the method of Dignam et al. (21). All extracts were dialyzed into sterile Buffer D (20 m~ HEPES, pH 7.9, 20% glycerol, 100 m~ KCl, 0.2 rn EDTA, 0.5 m~ dithiothreitol, 0. 5 mM phenylmethylsulfonyl fluoride) for 5 h a t 4 "C. Extracts were aliquoted, frozen in dry ice-ethanol, and stored a t -80 "C. Protein concentration was determined by the Bradford method (Bio-Rad) (22). As a control for extract quality, all nuclear extracts were incubated with labeled oligonucleotides containing an AP1 consensus binding site, and protein-DNAinteractions were determined. All extracts used in these studies exhibited binding to this control sequence (data not shown).
DNA Mobility Shift Assay-Oligonucleotides and short DNA restriction fragments from regions identified by DNase I footprinting were annealed and end-labeled. Control reactions (no protein) and experimental reactions (10 pg of protein) contained 20 pl of buffer with 6 pg poly(dI-dC).poly(dI-dC) (Pharmacia) as nonspecific competitor. The final concentration of reactants was 15 m~ HEPES, pH 7.9, 2.5 m~ MgC12, 100 mM KCl, 10% glycerol, 100 p~ EDTA, and 250 p~ phenylmethylsulfonyl fluoride. In competition assays, specific competitor was also included at 100-200-fold molar excess relative to labeled probe. Prior to addition of labeled probe, all reactants were incubated on ice for 10-15 min. 3000-5000 cpm of labeled probe were added to each reaction and incubated an additional 10 min on ice. Reaction products were resolved on 4 4 % native acrylamide gels (29:l) in 0.045 M Tris borate, and WI38, fetal lung fibroblast cells electrophoresed on formaldehydeagarose gels and transferred to nylon membranes. In Panel A the blot was hybridized to a radiolabeled TF cDNA probe. In Panel B the same blot was hybridized to a probe for 18 S ribosomal RNA to quantitate mRNA loading. The arrow at the right indicates the TF specific 2.3kilobase band. 0.5 m~ EDTA. Following electrophoresis, gels were transferred to Whatman No. 3MM paper, dried in uacuo, and autoradiographed a t -80 "C.

RESULTS
Analysis of TF mRNA Expression-The increase in steady state levels TF mRNA occurs rapidly following addition of LPS to the growth medium. Fig. 1 compares TF mRNA expression in monocyte and control cell lines. In this analysis, THP-1 and U937 cells were induced for 60 min with 1 pg/ml LPS. WI38 cell mRNA was included as a positive control for TF expression. Within 60 min following incubation of THP-1 cells with LPS the steady state levels of the TF-specific 2.3-kilobase transcript increases.
DNase I Footprint Analysis-DNase I footprinting was used to determine the sites of protein-DNAinteraction between -383 and the mRNA cap site. To detect all sites which may interact with protein we included areas which showed perturbation in intensity of banding pattern when compared with the control lane even when no classical footprint was detected. To resolve protected regions more easily, the 383-bp TF promoter fragment was subdivided into two fragments. Diagrams of the subclones which were used in this analysis are shown in Fig. 2. Footprint analysis of pTFPXASma is shown in Fig. 3. Although nuclear extracts from uninduced and LPS-induced monocytes were incubated with the probes, there is no obvious change in the pattern of protected fragments with induction. However, two hypersensitive sites are identified only with uninduced extract. In Fig. 3 (Panel B ) , a weak but reproducible The changes in hypersensitivity may indicate qualitative modifications in proteins interacting at these sites. Fig. 5 contains the nucleotide sequence of the TF promoter region and summarizes the regions of protein-DNA interaction identified by DNase I footprinting. Computer homology search for reported consensus binding sites indicates that footprint 2 contains a n SP1 motif, footprint 3 14 contains two potentialAP1 binding sites, and footprints 2 and 7 contain a Polyomavirus enhancer activator 3 (PEA3) core consensus binding site. The identification of 2 AP-1 sites in this region by sequence homology (25) and AP-1 binding activity (16) has been reported. Although a computer homology search did not identify a potential NF-KB binding site in footprint 2, there is some homology to the consensus. In addition, this site has been reported to interact with NF-KB (16). Footprint 5 contains an SP1 consensus binding site, suggesting that this might be a candidate binding factor. This information was used to synthesize oligonucleotides to use as competitors in order to determine what factors were binding to these regions of the promoter.
Characterization of Proteins That Interact with DNA Sequences-To define the nature of the proteins that interact with promoter elements, DMSA was performed using either oligonucleotides, small DNA restriction fragments, or small DNA fragments that were synthesized by PCR using pTFPX (Fig. 2) as the template. Probes used in this analysis are identified by DNA mobility shift probe (DMSP) numbers (Table I). DMSA corresponding to footprint 2 (DMSP 2 ) and NF-KB binding is shown in Fig. 6; footprint 314 (DMSP 3 and 4 ) are shown in Fig.  7 and footprints 5 , 6 , and 7 (DMSP 5, Sty 84, and PCR 1 ) are shown in Fig. 8. In addition, DMSA of Ets-1 binding activity in THP-1 extracts is shown in Fig. 9. The data are summarized in Table I.
Although a strong footprint is detected between -22 to +7, we were unable to detect a specific shift when DMSP 1, a 29-bp oligonucleotide spanning this region, was incubated with nuclear extracts. This suggests that factors binding to this sequence, in the context of the promoter, were not able to bind to the isolated sequence. DMSP 2 (footprint 2) binds weakly to a factor(s) in uninduced and induced extracts (Fig. 6, Panel A ) . In addition, DMSP 2 binding is competed by preincubation with unlabeled oligonucleotides containing an NF-KB binding site.

Human Tissue Factor Promoter 1363
In order to determine the level of NF-KB binding in THP-1 cell extracts, an oligonucleotide containing the consensus binding site was synthesized, end-labeled, and subjected to DMSA (Fig.  6, Panel B ) . Uninduced THP-1 cells contain proteins that interact with the NF-KB consensus sequence and, following induction, the level of this binding activity increases. DMSP 3 and 4 (footprint 314) bind with different affinities to proteins whose binding is competed by oligonucleotides containing an AP1 binding site. Although AP1 binding to DMSP 3 is unchanged in uninduced and induced extracts (Fig. 7 , Panel  A , lanes 2 and 3 ) there is a substantial increase in AP1 binding to DMSP 4 with induction (Fig. 7, Panel B , lanes 2 and 3 ) . In fact, there is very little protein binding to this site in uninduced extracts (Fig. 7 , Panel B , lane 2). Our data confirm that the promoter sequence between -231 and -198 contains two adjacent AP1 sites. In addition, these results indicate that the affinity of the two sites for AP1 binding activity differs, and the binding kinetics for the distal AP1 binding site change with induction.
DMSP 5 (footprint 5) binds to a nuclear protein that is not competed by AP1, AP2, or SP1 (Fig. 8, Panel A , and Table I). In addition, there is no change in the level of binding for this probe with induction (data not shown). DMSP STY84 (an 84-bp restriction fragment containing footprints 6 and 7 ) generates several specific mobility shift bands (Fig. 8, Panel B 1. These factors are localized t o the nucleus, and none of these binding activities is competed with consensus oligonucleotides containing AP1, AP2, or NF-KB binding sites. In addition, there is no binding activity for these factors in extracts prepared from fetal lung fibroblasts (WI38 cells) or hepatocellular carcinoma (HEPG2) cells (Fig. 8, Panel B , lanes 4 and 5 ) . DMSP PCR 1 (footprints 5,6, and 7 ) binds the same factors as the Sty 84 fragment (Fig.  8, Panel C , open arrows). In addition, this probe binds the DMSP 5-specific factor (dark arrow which is specifically competed by unlabeled DMSP 5 oligonucleotide. The sequences protected by footprints 2 and 7 contain PEA3 core motifs; therefore, we determined the availability of nuclear factors in THP-1 cells to interact with consensus oligonucleotides containing a PEA3 consensus sequence (Fig. 9). THP-1 extracts contain protein(s) that interact with oligonucleotides containing the PEA3 motif, indicated by the two specific mobility shift complexes that are detected. Neither complex is competed by NF-KB or AP1 oligonucleotides (lanes 9-12). By contrast, both of the complexes are competed with excess unlabeled PEA3 oligonucleotide (lanes 3 and 4 ) and with 200-fold excess DMSP 7 (lanes 7 and 81, an oligonucleotide spanning footprint 7 that also contains a PEA3 core motif. Since DMSP 2 contains a PEA3 core motif, this sequence was also used as a competitive inhibitor of Ets-1 binding to its consensus. Although DMSP 2 is not as effective as DMSP 7, there is approximately a 30% reduction in Ets-1-specific complex formation when a 200-fold excess of DMSP 2 is included in the reaction. The fold molar excess of DMSP 2 in this reaction is deceptive in this case because DMSP 2 appears to contain both a weak PEA3 motif and a weak NF-KB motif, and the binding sites overlap (Fig. 5 ) . The presence of overlapping binding sites suggests that occupancy by either factor might preclude binding by the other. Furthermore, it suggests that there may be a competition between these factors for binding and regulation at this site. DISCUSSION The TF gene promoter contains several regions that are resistant to DNase I cleavage when labeled DNA fragments are preincubated with nuclear protein prepared from uninduced and induced THP-1 cells. The results are summarized in Table  I and Fig. 10, which also identifies functional regions of the TF promoter described in the literature (4,14,16). With the exception of footprint 1 (-22 to +8) the additional five footprints are contained within a 200-bp region between -374 and -172 relative to the cap site. Four of the footprints are contained in a region of the promoter reported to be involved in transcriptional activation and induction (footprints 1-5) and two additional footprints map to a region that may be involved in silencing gene expression (footprints 6 and 7). Since footprint 1 (-22 to +7) was very strong we considered that this region, between the TTATAbox and the cap site, might bind a negative factor capable of preventing occupancy by RNA polymerase and its accessory proteins in uninduced THP-1 cell extracts. To explore this possibility DMSP 1, a 29-bp doublestranded oligonucleotide, was synthesized, labeled, and incubated with nuclear extracts. As Table I indicates no specific mobility shift bands were detected even though a range of conditions were employed. A computer search for consensus binding sites in this sequence was also nonproductive. One explanation is that this strongly protected region interacts with RNA polymerase I1 and its accessory factors in the context of the native promoter, preventing digestion by DNase I. However, in the DNA mobility shift assay this complex may be unable to form on the 29-bp fragment due to physical constraints that preclude the generation of protein-DNA complexes.

Protein-DNA Interactions with Human Tissue Factor Promoter
Sequences between -191 and -172 appear to be important for LPS induction of TF expression, since these sequences have been reported to be suficient to mediate LPS induction of reporter gene constructs (16). Using DMSA, we detect weak binding activity to DMSP 2 (footprint 2, -191 to -151) in both uninduced and induced nuclear extracts. This binding is completely abrogated when an unlabeled oligonucleotide containing the NF-KB consensus site is included in the reaction. Other groups have reported a strong induction of NF-KB binding ac- tivity to both its consensus binding sequence and to oligonucleotides spanning -193 to -172 in nuclear extracts prepared from LPS induced (10 pg/ml) THP-1 cells (16). Our data confirm that the level of NF-KB binding activity in THP-1 cells increases with induction; however, we do not see comparable levels of binding to oligonucleotides containing the NF-&-related sequence found in the TF promoter (-191 to -172). In contrast to the strong signal observed when a probe containing the consensus NF-KB binding site is incubated with induced THP-1 cell extracts, DMSP 2 oligonucleotides bind very weakly to factors in both uninduced and induced extracts. One explanation for the reduced binding activity may be differences between the nucleotide sequence contained in DMSP 2 (CG-GAG"TCC) in comparison with the nucleotide sequence of the NF-KB consensus binding site (GGGACTTTCC). In addition, we do not see a dramatic difference in binding to DMSP 2 with uninduced and induced extracts. The two nucleotide sequence differences present in the TF sequence (indicated by bold letters) are not choices that conform to the NF-KB consensus

alternatives GGG(G/A)(APTIC)T(T/C)(T/C)CC).
Since the binding to DMSP 2 is very low level, these two nucleotide differences must be critical for high affinity binding by NF-KB.
Computer homology analysis of the TF gene promoter indicates that, on the noncoding strand, there are two regions with homology to the PEA3 consensus core sequence (AGGAA(A/C)) (26). One is contained in footprint 2 (-191 to -1721, and an additional site is contained within footprint 7 (-363 to -343). Ets-1 as well as related family members have been reported to bind to the PEA3 motif (27,28). We are currently evaluating Ets-related protein interactions with these sequences. Preliminary data indicate that THP-1 nuclear extracts contain a binding activity that interacts with a consensus oligonucleotide containing the PEA3 enhancer sequence (Fig. 9). Ets-1 binding is reported to be relatively low affinity; however, binding to the consensus oligonucleotide is partially competed by 200-fold excess of DMSP 2 and more effectively competed by 200-fold excess DMSP 7 and Ets-1 consensus sequence (Fig. 9). Although DMSP 2 is a less effective competitor than DMSP 7, the    GCGCGCGGGGCACCGGCTCCCCAAGACTG  GCGCGCCCCGTGGCCGAGGGGTTCTGACG  GTCCCGGAGTTTCCTACCGGG  CAGGGCCTCAAAGGATGGCCC  TGGGGTGAGTCATCCCTT  TGACCCCACTCAGTAGGGAA  GGCGCGGTTGAATCACTGGG  CCGCGCCAACTTAGTGACCC  GATCGGGCAACTAGACCCGCC  CCCGTTGATCTGGGCGGCTAG  GATCCCTTTCCTGCCATAGACCT  GGAAAGGACGGTATCTGGACTAG   GATCGAGGGGACTTTCCCTAG  TAGCTCCCCTGAAAGGGATCG  GATCGATCGCAGGAAGTGAT  CTAGCTAGCGTCCTTCACTA Self, NF-KB Self, AP1 Self, AP1, DMSP 3 Self Self, one DMSP 5 Self Self Self, DMSP 2, DMSP 7 inability of DMSP 2 to completely abrogate Ets-1 binding suggests that the protein interactions with this sequence might be more complex. Our data suggest that there might be a competition for binding to this site by both NF-KB-and Ets-l-related factors. In this case, binding of NF-KB to the DMSP 2 competitor probe would reduce the concentration of probe available to compete for Ets-1 binding to its consensus binding site. This could explain the reduced ability of DMSP 2 to compete for Ets-1 binding. Perhaps the availability of these factors in the nucleus determines which factor is able to bind more efficiently. Unlike the nuclear NF-KB binding which seems to dramatically increase with induction, the Ets-1 binding activity detected in our extracts does not quantitatively change with induction (data not shown). These data suggest that Ets-1 may contribute to the transcriptional regulation of the TF promoter by binding to sequences contained in DMSP 2 and 7. The phosphoprotein Ets-1, or another member of this family of transcriptional activators, interacting with the PEA3 motif, provides an attractive candidate for a key regulator of TF expression in THP-1 cells for several reasons. The PEA3 motif can mediate serum induction as well as phorbol ester induction of gene expression (29). Other groups have described a type I1 proto-enhancer that combines several transactivation domains, each of which is weak in isolation, in sequence to form an enhancer element with stronger activation potential (30). The Polyomavirus enhancer, as well as the collagenase promoter, are included in this subclass (31). In addition, they point out that several genes, including stromelysin, interleukin-2, and Fos, contain linked PEA3 and AP1 motifs. Our data suggest that the TF gene may also be included in each of these groups, since AP1 and PEA3 motifs are also linked, and this association may be important in the function of the promoter. Although TF expression is induced by serum and the sequences between -383 to the cap contains enough information to mediate this induction, there is no consensus serum response element located in this region (14, 17). Since the PEA3 element is serum responsive and there are two reiterations within the TF proximal promoter, factor binding to this region may also be responsible for serum induction.
Footprint 314 contains two closely linked binding sites for AP1 (DMSP 3 and 4). DMSP 3 binds with high affinity to AP1 in uninduced and induced nuclear extracts. By contrast, DMSP 4 binds to AP1 with barely detectable affinity with uninduced extracts and this binding increases with LPS induction. The authenticity ofAPl binding to DMSP 4 was confirmed since the binding activity is competed by unlabeled oligonucleotides containing either the AP1 consensus site or by unlabeled DMSP 3 which also binds AP1. Comparison of the DNA sequence in that a novel heterodimer that forms post-induction might rec-c ognize this oligonucleotide. This may be the same heterodimer or a different heterodimer from the complex that recognizes DMSP 3. In addition, the absence of binding activity to DMSP 4 in uninduced extract suggests that a subset of AP1 binding activity is not able to recognize the single nucleotide change from the consensus present in this oligonucleotide. By DNase I footprinting these two binding sites are contiguous along the promoter. In addition, the degree of protection by protein in uninduced and induced extracts is equivalent (Fig. 4). This suggests that AP1 interactions with the two contiguous binding sites may be cooperative in the context of a larger DNA frag-  (Panels A and B , lanes  3-7). Arrow at the right of the panels indicates the specific mobility shift band. Specific competitors: -, no competitor; S, unlabeled probe; A P l , consensus oligonucleotide containing A P 1 binding site; AP2, consensus oligonucleotide containing AP2 binding site; DMSP 3, DMSP 3 oligonucleotide. Unlabeled competitors are indicated and the fold molar excess is given in parentheses.  print pattern is unusual, we have determined that proteins do interact with labeled probes containing this sequence (Fig. 8, PCRl and Sty 84) . We have not been able to compete the binding to either the DMSP 5 oligonucleotide or the Sty 84 fragment with unlabeled oligonucleotides containing several consensus binding sites (Table I). Although footprint 5 contains an SP1 binding motif, the binding to this oligonucleotide is not competed by unlabeled SP1 consensus oligonucleotides. Computer homology search for reported transactivator binding sites did not reveal any additional consensus sites in these two regions.
At least for the Sty 84 fragment, our data indicate that the factors binding to this probe are not present in WI38 cells (fetal lung fibroblasts), which express high levels of TF mRNA, and HEPGB cells (hepatocellular carcinoma cells), which do not express T F mRNA (data not shown). Since footprints 6 and 7 reside in the "silencer" region of the promoter (14), factors binding to these sites could facilitate the repression of TF synthesis in uninduced monocytes. Subsequent modification of these factors following a n induction stimulus could reactivate transcription suggesting that these proteins might be "modulatory" in function. Since WI38 cells constitutively express TF mRNA, and HEPGS cells do not express T F mRNA, modulatory proteins would not be needed. In fact, our analysis suggests that they are not present in extracts prepared from these cells.
At least two of our footprints (5 and 6) appear to interact with nuclear binding proteins although the sequences contained in these regions do not share homology with consensus binding sites in the database. We have prepared nuclear extracts from a number of different cell lines in order to determine if the proteins that bind to these sequences are monocyte specific. By Southwestern blotting we have identified a 150-kDa monocyte- specific protein that binds to probes containing footprints 5 and 6 (data not shown). We are beginning to purify these binding activities in order to characterize them further.
In summary, we have identified six regions between -383 and the cap site that are resistant to DNase I cleavage when DNA is preincubated with nuclear extract. Four of these footprints map to regions reported to be involved in transcriptional activation (footprints 2-5, -277 to -172). Two additional footprints localize to a region that has been reported to reduce base-line expression of the TF promoter by 50% in transfected COS 7 cells (footprints 6 and 7, -383 to -278). With the exception of two single, hypersensitive sites that are detected with probes incubated with uninduced extracts (Figs. 3B and 4 B ) and an increase in binding of AP1 to DMSP 4 oligonucleotides, there are no additional differences in protein-DNA interactions when uninduced or induced extracts are compared. This analysis does not preclude that post-translational modification of pre-existing proteins shifts the promoter from an inactive state to a transcriptionally active state following LPS induction. Alternatively, additional proteins not detected in crude nuclear extracts may contribute to this process.
Although several of the regions seem to bind well characterized factors such as AP1, two additional sites contain a PEA3 core enhancer motif that may interact with a member of the Ets-1 family of transcriptional activators. In addition, the role of S p l binding to the regulation of TF synthesis remains to be evaluated. Finally, we have also identified binding to sequences that appear to be important for constitutive expression (footprint 5) and repression of expression (footprints 6 and 7). Two of these sites (footprints 5 and 6) may bind novel factors since they do not contain homology to reported consensus binding sites by computer sequence analysis. Experiments to further characterize the proteins that interact with these sequences are in progress and will contribute to a better understanding of the molecular events that modulate TF synthesis in monocyte cell lines.