Complex NF-κB Interactions at the Distal Tumor Necrosis Factor Promoter Region in Human Monocytes*

We describe a dense cluster of DNA-protein interactions located 600 nucleotides upstream of the transcriptional start site of the human tumor necrosis factor (TNF) gene. This area was identified as being of potential importance for lipopolysaccharide-inducible TNF expression in the human monocyte cell line Mono Mac 6, based on reporter gene analysis of point mutations at a number of nuclear factor κB (NF-κB)-like motifs within the human TNF promoter region. The area contains two NF-κB sites, which are here shown by DNase I and methylation interference footprinting to flank a novel binding site. UV cross-linking studies reveal that the novel site can also bind NF-κB as well as an unknown protein(s) of approximately 40 kDa. We show that these three adjacent κB-binding sites differ markedly in their relative affinities for p50/p50, p65/p65, and p65/p50, yet this 39-nucleotide segment of DNA appears capable of binding up to three NF-κB heterodimers simultaneously. Reporter gene studies indicate that each element of the cluster contributes to lipopolysaccharide-induced transcriptional activation in Mono Mac 6 cells. These findings suggest that NF-κB acts in a complex manner to activate TNF transcription in human monocytes.

A fundamental problem in infectious disease is to understand the mechanisms by which microbial stimuli such as bacterial lipopolysaccharide (LPS) 1 regulate the expression of human inflammatory mediators such as tumor necrosis factor-␣ (TNF). Production of TNF by macrophages and monocytes occurs early in infection and orchestrates much of the subsequent inflammatory response (1). Because excessive production of TNF is responsible for many of the pathological consequences of severe infection, its regulation is critical. This occurs at both the transcriptional and post-transcriptional levels (2), with transcriptional regulation showing specificity for both stimulus and cell type (3). Reporter gene studies of the human TNF promoter using a range of stimuli and cell types have identified a number of regulatory elements upstream of the transcriptional start site, mostly within the proximal 200 base pairs (4 -13). However, we still have a poor understanding of the precise regulatory events that are responsible for the rapid induction of high level TNF expression in human monocytes and macrophages during acute bacterial infection.
Much debate has centered on the role of nuclear factor-B (NF-B) in transcriptional activation of the human TNF gene in monocytes. Although there is strong evidence that murine TNF expression is influenced by a number of NF-B binding sites (14 -16), the situation in humans is less clear. Several lines of circumstantial evidence suggest that NF-B could play an important role. A number of NF-B-like binding motifs are located between the start of transcription of the TNF gene and the 3Ј terminus of the lymphotoxin-␣ gene: they are here denoted B1 (Ϫ873 to Ϫ864 nt), B2 (Ϫ627 to Ϫ618 nt), B2a (Ϫ598 to Ϫ589 nt), CK-1 (Ϫ213 to Ϫ204 nt (17)), and B3 (Ϫ98 to Ϫ89 nt). It is worth noting that site B2 is 100% conserved between mouse, rabbit, and human TNF promoters (18). Moreover, LPS induces nuclear translocation of NF-B in human monocytes, and LPS-inducible TNF expression is suppressed by specific inhibitors of NF-B mobilization (19,20). However, studies of the role of NF-B in the transcriptional activation of the human TNF gene have yielded a conflicting picture. Initial gene reporter studies suggested that NF-B binding in the human TNF promoter region is of little functional importance when tested in the murine macrophage cell line P388D1 (5) or when analyzed in other contexts such as phorbol 12-myristate 13-acetate stimulation of U937 (myeloblastoid) cells (4) and K562 (erythroblastoid) cells (21). However, recent studies indicate that site B3 contributes to transcriptional activation of the TNF gene by superantigen (6) and cooperates with an adjacent cAMP-responsive element site to provide an enhancement of TNF transcription in LPS-stimulated human monocyte cell line THP-1 (13).
The present investigation focuses on the role of the other NF-B binding sites, in particular the conserved site B2 and its adjacent site B2a. Several studies have shown that NF-B-specific complexes are formed when site B2a is tested against nuclear extracts from stimulated monocytes and macrophages (13,20,22), but there has been doubt about their functional importance. For example, reporter gene studies in human THP-1 cells concluded that the region containing B2 and B2a does not contribute to LPS inducibility, even though NF-B-specific complexes were formed at the site B2a with greater affinity than at the apparently functional site B3 (13). However, a different picture emerged when reporter constructs * This work was supported by the Medical Research Council (UK). 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 U.S.C. Section 1734 solely to indicate this fact.
§ These authors contributed equally to this work. ¶ To whom correspondence should be addressed. containing the human TNF promoter region were transiently expressed in the murine macrophage line ANA-1. In this cell line, which resembles mature bone marrow macrophages, LPS inducibility was found to fall 5-10-fold as a result of deleting the region from Ϫ656 to Ϫ442 nt (22,23).
These discrepant observations raise the intriguing possibility that the functional activity of the region around B2 and B2a has complex determinants that are specific not only for stimulus and cell type but for the stage of cell differentiation as discussed in Ref. 6. Because the primary concern of the present study is the mechanism of LPS-induced TNF expression in human monocytes, here we have used Mono Mac 6, the human cell line whose surface markers and functional properties most closely resemble those of a mature monocyte (24). In this cell line, we find that mutations in sites B2 and B2a sites have a more profound effect on LPS-inducible reporter gene expression than mutations at other NF-B sites in the TNF promoter region. We further show that sites B2 and B2a flank a novel binding site, such that this 39-nt segment of DNA is capable of binding up to three NF-B heterodimers plus an unknown constitutive factor(s). Each element of this cluster appears to contribute to transcriptional activation of the TNF gene.
UV Cross-linking and Immunoprecipitation-The binding reaction was performed with radiolabeled oligoduplex corresponding to the site in which three central dT nucleotides were substituted with bromodeoxyuridine. The EMSA gel was UV illuminated at 302 nm for 30 min at 4°C and exposed to autoradiography for 2-4 h at the same temperature. The region corresponding to the DNA-protein complex was excised, and the proteins were eluted in 2ϫ SDS buffer (100 mM Tris-Cl, pH 6.8, 200 mM dithiothreitol, 4% SDS, 20% glycerol) at 37°C overnight. They were further processed either for SDS-PAGE or for immunoprecipitation as described (26).
Methylation Interference Analysis-Oligonucleotide 2-F (or 2a-R) was labeled with [␥-32 P]ATP (Amersham) and used as primer in a PCR along with 2a-R (or 2-F). The PCR fragment obtained (5 ϫ 10 5 cpm) was gel purified and partially methylated in 200 l of buffer containing 200 mM HEPES, pH 7.8, 1 mM EDTA by adding 1 l of dimethyl sulfate at room temperature for 1 min. The reaction was stopped by addition of 40 l of stop solution (1.5 M sodium acetate, pH 5.2, 1 M ␤-mercaptoethanol). DNA was precipitated and washed several times. The conditions of recombinant protein-DNA binding reactions were the same as for EMSA but scaled up 5-10 times. Bound and free DNAs were eluted overnight at 37°C in 50 mM Tris-Cl, pH 8.0, 1 M NaCl, 2 mM EDTA. The DNA was purified and dissolved in 90 l of H 2 O, 10 l of piperidine was added, and the sample was incubated at 90°C for 30 min and cooled on dry ice. Traces of piperidine were removed by three cycles of evaporation. Cleaved DNA fragments were separated by electrophoresis using a 13% sequencing gel.
DNase I Footprinting Assay-Solid phase DNase I footprinting was carried out using the methodology described previously (27). DNA probes were generated by PCR amplification using biotinylated primer TNF-480 (TTGAGTCCTGAGGCCTGTGT) along with [␥-32 P]ATP-labeled primer TNF-680 (GCATTATGAGTCTCCGGGTC) or vice versa. Following gel purification, radiolabeled DNA probe was adsorbed onto magnetic Dynabeads M-280 Streptavidin (Dynal) according to the manufacturer's instructions. 2-4 ϫ 10 4 cpm of bead-DNA probe was incubated with 5-20 g of nuclear extracts in 160 l of binding buffer at room temperature for 20 min. Following addition of 160 l of salt mixture (10 mM MgCl 2 , 5 mM CaCl 2 ) 0.06 -0.25 units DNase I (Boehringer Mannheim) was added to the mixture and incubated at room temperature for 30 s. An equal volume of STOP solution was added (0.5 M NaCl, 40 mM EDTA, 1% SDS), and digested DNA was recovered using the Magnetic Particle Concentrator (Dynal). DNA was washed with 200 l of solution I (2 M NaCl, 20 mM EDTA) followed by 200 l of solution II (10 mM Tris-Cl, pH 8.0, 1 mM EDTA) twice. The reaction was analyzed on 7% sequencing gel. Where indicated autoradiograms were recorded digitally using the UVP Image Store 5000 system (UVP Life Sciences, Cambridge, UK) and analyzed with NIH Image 1.6 software (public domain).
Expression and Purification of Recombinant Proteins-The protein sequences corresponding to amino acids 2-400 of human p50 and amino acids 2-306 of human p65 were recovered by PCR using the appropriate primers. p50 and p65⌬ cDNAs were cloned into the bacterial expression vector pET32a (Novagen) and used to transform BL21(DE3)LysS bacterial strain. Cultures were grown to A 600 ϭ 0.4 -0.6, induced with 1 mM isopropyl-1-thio-␤-D-galactopyranoside (Sigma), and then grown for an additional 3 h before harvesting. Bacterial pellets were resuspended in buffer A (50 mM Tris-Cl, pH 8.0, 0.1 mM EDTA, 0.01% Nonidet P-40, 1 mM dithiothreitol, 10% glycerol) containing 70 mM NaCl and lysed by three sonication cycles. Insoluble material was removed by centrifugation, and cleared lysate was absorbed to biotin-labeled oligoduplex corresponding to consensus NF-B binding sequence bound to streptavidin-agarose (Sigma) for 30 -60 min at room temperature. The agarose beads were washed twice with buffer A containing 100 mM NaCl, and the bound protein was eluted with buffer A containing 500 mM NaCl. To form p65⌬/p50 heterodimer, cleared lysates containing equal amounts of recombinant p50 and p65⌬ were mixed and denatured by adding urea to 8 M, followed by step dialysis against buffer A containing 70 mM NaCl and 6, 4, 2, 1, 0.5, or 0 M urea. The heterodimer was purified from total extract as described above. The purity and concentration of recombinant proteins were assessed on SDS-PAGE gel.
Cell Culture, Transfections, and Luciferase Assay-Mono Mac 6 cells were maintained as described previously (24). Transient transfections were performed on Mono Mac 6 cells by the DEAE-dextran method as described previously (15). Cells were plated in fresh medium at 5 ϫ 10 5 /ml 24 h prior to transfection. The concentration of DEAE-dextran (Amersham Pharmacia Biotech) used was 100 g/ml. After transfection cells were incubated for 24 h prior to LPS activation (1 g/ml) and for a further 6 -12 h before harvesting. Luciferase assay was performed using the Luciferase assay system and Turner Luminometer model 20 (Promega) according to the protocol supplied.

Functional Characterization of the NF-B-like Sites of the Human TNF Promoter Region in Human
Monocytes-To examine whether LPS-inducible specific DNA-protein complexes can be formed at NF-B-like sites within the human TNF promoter region in human monocytes, oligonucleotide duplexes spanning these sites were tested in EMSA with nuclear extracts from unstimulated and LPS-stimulated Mono Mac 6 cells. LPS stimulation resulted in formation of NF-B/Rel-specific complexes at sites B1, B2, B2a, and B3, although they differed considerably in intensity (Fig. 1). Because no NF-B/Rel-specific complexes were formed at site CK-1 (data not shown), this site was excluded from further analysis. The strongest binding was observed at site B1, which formed two NF-B-specific complexes corresponding to p50/p50 homodimer (I) and p65/p50 heterodimer (II), respectively, as defined by using NF-B/Relspecific antibodies (data not shown). The binding at the sites B2 and B2a was about three times weaker, and the complex formed at site B3 was at the limit of detection (Ͼ10 times weaker than that at site B1).
To examine the role of each site in LPS-inducible transcriptional activation, we generated a set of reporter constructs with 1.2 kilobase pairs of TNF promoter sequence placed in front of a luciferase reporter gene. Point mutations were created at each of the NF-B binding sites, using specific mutations that disrupted the formation of NF-B/Rel-specific complexes when tested by EMSA as above (data not shown). When these constructs were expressed in Mono Mac 6 cells, mutation at site B1 had little effect on LPS-inducible luciferase activity, whereas mutation at site B2 or site B2a resulted in a significant decrease of luciferase activity (approximately 50% reduction of the wild type construct), and mutation at site B3 had a lesser effect (15-20%) (Fig. 2). Thus, despite the fact that B1 site appears to have a higher affinity to NF-B/Rel proteins, sites B2 and B2a appear to have a greater effect on transcriptional activation of the TNF gene in Mono Mac 6 cells.
Inducible Occupancy of a Cluster of Three Sites in the Distal Part of the Human TNF Promoter Region-To explore DNAprotein interactions around the B2 and B2a sites, we performed DNase I footprinting analysis of a segment spanning Ϫ680 to Ϫ480 nt using nuclear extracts prepared from Mono Mac 6 cells before or after LPS stimulation. Nuclear extracts made after LPS stimulation gave protection from DNase I cleavage at sites B2 and B2a and also in the intervening region spanning Ϫ616 to Ϫ599 nt (Fig. 3A). Nuclear extracts made before LPS stimulation also gave protection in this region when tested at high concentrations. A similar pattern of protection was observed using the noncoding strand (Fig. 3B).
This observation raised the possibility of a novel binding site between B2 and B2a, but it was also possible that the area of protection in between B2 and B2a might be because of conformational changes brought about by occupancy of the B2 and B2a sites. To examine this question, the above experiments were repeated with DNA bearing specific mutations in B2, B2a, or the intervening sequence. Each mutation was found to cause a site-specific alteration in the DNase I footprint (Fig. 3C). That is, mutation of site B2 abolished the protection of site B2 (but not of B2a or the intervening sequence), whereas mutation of B2a abolished the protection of B2a (but not of B2 or the intervening sequence), and mutation of the intervening sequence abolished the protection of the intervening sequence (but not of B2 or B2a) (Fig. 3D). We therefore proceeded to characterize the putative novel binding site, FIG. 1. Nuclear factors binding to sites B1, B2, B2a, and B3. Nuclear extracts from Mono Mac 6 cells before and after 1 h of stimulation with LPS were used in EMSA with probes corresponding to B1 (lanes 1 and 2), B2 (lanes 3 and 4), B2a (lanes 5 and 6), or B3 (lanes 7 and 8). Supershift assay (not shown) indicated that complex I consists of p50/p50 homodimer, and complex II consists of p65/p50 heterodimer. Non-NF-B complexes marked as n. here termed , situated in the 19-nt gap between sites B2 and B2a.
Constitutive and Inducible Binding to the Novel Site, -To examine the binding properties of we used an oligonucleotide duplex matching the sequences from Ϫ619 to Ϫ597 nt as a probe for EMSA. Nuclear extracts were obtained from Mono Mac 6 cells that had been stimulated with LPS (Fig. 4A). For comparison, the binding complexes observed with probes specific for sites B2 and B2a are shown. Unstimulated nuclear extracts gave a single major complex with the probe for site . After stimulation this constitutive complex increased in intensity, and on close inspection, a separate sharp band appeared at its lower margin. The latter inducible complex was at maximal intensity between 15 and 120 min after stimulation (data not shown). These data raised the question of whether the novel site might bind two different factors, one constitutively present and the other rapidly inducible in Mono Mac 6 cells.
A further EMSA investigated the specificity of these complexes by competition with various unlabeled oligonucleotide duplexes. The constitutive complex was abolished by competi-tion with 100ϫ itself but not by 100ϫ excess of oligoduplexes corresponding to site B2 or B2a (data not shown). In contrast the sharp band of the inducible complex appeared to be inhibited by competition with B2 or B2a at least as effectively as by itself. Competition with unlabeled oligonucleotide duplex corresponding to an EGR-1 binding site had little effect on constitutive or inducible complex formation. These results suggested that constitutive and inducible factors interacting with site are of different origin and that the latter might be related to the NF-B/Rel family.
The distinction between constitutive and inducible binding was also illustrated by the different effects of specific alterations to the sequence (Fig. 4B). A substitution of GTGA 3 ATCC at Ϫ611 to Ϫ608 nt disrupted both constitutive and inducible binding. Interestingly, an ATTT 3 GATC substitution at Ϫ608 to Ϫ605 nt had no effect on inducible binding but abolished constitutive binding. Conversely, these data suggest that a CAC 3 GGA substitution at Ϫ604 to Ϫ602 nt might reduce constitutive binding less completely than inducible binding. 2mt is a G to T substitution at Ϫ627 nt within the B2 site (lane 4), 2amt is a GG to TA substitution at Ϫ597 and Ϫ596 nt within the B2a site (lane 5), -mt1 is a GTGA to ATCC substitution at Ϫ611 to Ϫ608 nt in the sequence between sites B2 and B2a (lane 6). D, density analysis of the lanes from panel C. The arrows indicate effects of base substitution, e.g. G to T substitution at Ϫ627 nt (2mt) abolishes protection at site B2 but not at B2a or the intervening region.

An Inducible Complex Bound to is NF-B (p65/p50) Heterodimer-
The rapid kinetics of the inducible complex suggested the involvement of transcription factor(s) that do not need to be synthesized de novo. Because NF-B provides a classical example of this and because includes a sequence (GTGATTTCAC) that has two mismatches with the consensus NF-B site (GGGPuNNPyCCC, where Pu is purine and Py is pyrimidine), we considered the possibility of an inducible complex being formed by NF-B proteins. Two items of circumstantial evidence are given above. First, the LPS-inducible complex at had similar mobility to the complexes at B2 and B2a, which are known to bind p65/p50 (Fig. 4A). Second, the LPSinducible complex at was inhibited by competition with the p65/p50 binding sequences of sites B2 and B2a (data not shown). As a more formal test of this question, we performed an EMSA using specific antibodies against two members of the NF-B/Rel family, p65 and p50. Both antibodies markedly diminished the inducible complex, whereas the antibodies against another member of the family, c-Rel, and against transcription factor EGR-1 had no effect on this complex (Fig. 5). Taken together, these results identify a major inducible complex formed at site as p65/p50 NF-B heterodimer.
p50 and p65 Exhibit Different Binding Patterns across the B2//B2a Cluster-Because the various members of the NF-B/Rel family possess different functional properties, we used recombinant forms of p50 and p65 to examine the possibility that each part of the B2//B2a cluster might preferentially recruit a specific type of NF-B complex. The recombinant p65 was truncated at amino acid 306 to facilitate bacterial expression and is denoted p65⌬; this truncation does not affect the DNA-binding domain.
The results of methylation interference are shown in Fig. 6. It can be seen that p50 homodimers protect site B2a much more strongly than site B2, whereas the opposite is true for p65⌬ homodimers. Site is weakly protected by p50 homodimers but poorly protected by p65⌬ homodimers. All three sites are protected by p65⌬/p50 heterodimers.
To provide an indication of relative binding affinity, radiolabeled oligoduplexes corresponding to sites B2, B2a, or were titrated against a standard amount of p50/p50, p65⌬/p65⌬, or p65⌬/p50. DNA-protein complexes were resolved by EMSA and quantitated using a PhosphorImager (Fig. 7A). These data confirm that p50 homodimers bind much more strongly to B2a than to B2 or , whereas p65⌬ homodimers bind much more strongly to B2 than to B2a or . However the binding of p65⌬/p50 heterodimers showed less variation between the three sites, being strongest for B2a and weakest for site .
In view of the similar affinity of p65⌬/p50 for each of the three adjacent binding sites, the question arises whether the heterodimer can bind at all three sites simultaneously. When recombinant p65⌬/p50 heterodimer was incubated with an oligoduplex corresponding to the full B2//B2a cluster, EMSA revealed a ladder of three complexes (Fig. 7B) in contrast to  lanes 1 and 2), B2a (lanes 3 and 4), or the intervening sequence (Ϫ619 to Ϫ597 nt) (lanes 5 and  6). The probe shows an discrete inducible complex that migrates slightly ahead of the broader constitutive complex ( lanes  7 and 8 show longer exposure for lanes 5  and 6). B, effect of specific mutations on protein binding to site . Wild type (wt) and different mutant probes (mt1, mt2, and mt3) were incubated with nuclear extracts from unstimulated (lanes 1-4) or stimulated cells (lanes 5-8).
the single complexes seen for the individual binding sites. This appearance is consistent with the hypothesis that the cluster is capable of binding one, two, or three heterodimers simultaneously.
All Three Binding Sites Contribute to Inducible Gene Expression in Human Monocytes-To examine whether binding of protein complexes at the site has any functional effect on transcriptional up-regulation of the TNF gene, we included three different variants of the site whose effects on DNAprotein interactions are illustrated in Fig. 4B, in our set of reporter constructs with 1.2 kilobase pairs of TNF promoter sequence placed in front of a luciferase reporter gene. When expressed in Mono Mac 6 cells variations at the site resulted in a significant decrease of luciferase activity in the range of 40 -55% reduction of the wild type (Fig. 8A). When a construct with the point mutations at the sites B2, B2a, and was expressed in Mono Mac 6 cells, LPS-induced activity of the gene-reporter dropped even further and resulted in about 70% reduction of the wild type (Fig. 8A).
To explore the effect of proximity of the three sites in transactivation of the TNF gene, we introduced into the full TNF promoter a spacer of 6 or 10 nt between site and site B2a, thereby creating an additional half-turn or full turn of the DNA helix. Both insertions resulted in decrease of luciferase activity by about 30% (Fig. 8B).
Thus, each element of this cluster participates in transcriptional regulation of the TNF gene in human monocytes. Interference with the binding of NF-B/Rel proteins to each site of interaction in the region results in 30% of the original TNF gene expression. Repositioning of the sites causes modest re-FIG. 6. Methylation interference footprinting of a probe spanning ؊633 to ؊582 nt of the TNF promoter region using recombinant NF-B proteins. G/A ladder prepared by Maxam-Gilbert sequencing of the same fragment. F, free probe; B p50 , B p65⌬ , and B p65⌬/p50, probes bound to p50/p50 and p65⌬/p65⌬ homodimers and p65⌬/p50 heterodimer, respectively. 7. Binding affinities of recombinant p50/p50 and p65⌬/p65⌬ homodimers and p65⌬/p50 heterodimer to sites B2, B2a, and . A, standard amounts of recombinant proteins were incubated with radioactive probes (diluted as described) corresponding to sites B2, B2a, and . Lanes 1, 7, and 13, nondiluted probe; lanes 2, 8, and 14, 2 times dilution; lanes 3, 9, and 15, 4 times; lanes 4, 10, and 16, 8 times; lanes 5, 11,  and 17, 16 times; lanes 6, 12, and 18, 32 times. Graphs represent quantitative analysis of the autorads. B, EMSA with p65⌬/p50 heterodimer and a probe spanning Ϫ633 to Ϫ582 nt of the TNF promoter region. duction in the level of the gene expression.

FIG.
Characterization of the Constitutive Complex at -UV crosslinking experiments were carried out to investigate the nature of the factor(s) that bind to . Nuclear extracts from unstimulated Mono Mac 6 cells were incubated with a bromodeoxyuridine-substituted oligonucleotide probe, and the complexes were separated by EMSA. After UV cross-linking, the complexes were excised and eluted from the gel. When the purified complexes were analyzed by SDS-PAGE, a major band was seen at M r ϳ50 kDa (Fig. 9A). Because the oligoduplex itself migrates at about 10 kDa in this gel system, this would suggest that the constitutive binding factor contains a major component of approximately 40 kDa. UV cross-linking experiments with nuclear extracts from stimulated Mono Mac 6 cells showed the same constitutive component plus two additional bands of ϳ75 and ϳ60 kDa consistent with p65 and p50 (Fig. 9, A, lane 2, and B, lane 1). To provide further confirmation that the inducible complexes contain p65/p50 heterodimer, the cross-linked products were immunoprecipitated with monospecific antibodies and then analyzed on SDS-PAGE. Anti-p65 antibodies immunoprecipitated a complex that would correspond to a protein of 65 kDa, whereas anti-p50 antibodies immunoprecipitated a complex corresponding to a protein of 50 kDa, and control experiments with anti-C/EBP antibodies or anti-biotin antibodies did not yield an immunoprecipitated band (Fig. 9B, lanes  2-5). Taken together, these results indicate that site is constitutively occupied by a protein(s) of about 40 kDa and that LPS stimulation causes p65/p50 NF-B heterodimer to bind at the same site.

DISCUSSION
The aim of the present study was to examine the role of NF-B in the regulation of TNF transcription in Mono Mac 6, a cell line chosen because of its close similarity to well differentiated human monocytes (24). When the functional effects of four different NF-B-like sites in the TNF promoter region were compared by point mutagenesis and reporter gene expression, two neighboring sites in the distal part of the promoter region were identified as making the greatest contribution to LPS-induced gene expression. We show that in between these two sites, denoted B2 and B2a, there is a 19-nt sequence that binds both constitutive and inducible nuclear factors. This intervening sequence is referred to here as , and the cluster as a whole can be summarized as Ϫ627 B2//B2a Ϫ589 , although the functional boundaries of the novel binding sites within remain uncertain.
The capacity of the B2//B2a cluster to bind NF-B has two interesting features. The first is that three NF-B complexes can bind to such a short segment of DNA. Whether they do so simultaneously is a matter for further investigation, but DNase I footprinting patterns suggest that this might be the case. A further suggestion comes from a comparison of the EMSA behavior of the individual binding sites with that of the full  -tailed t test). B, the position of the 6-or 10-nt insertion is marked. Both 6-and 10-nt insertions gave significantly lower inducibility than wild type (wt, p Ͻ 0.01, by paired two-tailed t test). B2//B2a region. Whereas the individual binding sites only form one complex when incubated with recombinant p65⌬/p50, the full B2//B2a region forms three separate complexes that would seem to correspond to the binding of one, two, or three heterodimers (Fig. 7B).
The second potentially important aspect is the wide variation in binding affinity across the cluster and the fact that different members of the NF-B/Rel family preferentially bind to different parts of the cluster. Site B2 has relatively high affinity for p65⌬ homodimer but relatively weak affinity for p50 homodimer, whereas the opposite is true for site B2a. Such preferential binding patterns are likely to be explained by minor sequence differences between the sites (28), yet it is intriguing that we find little variation in binding affinity for p65/p50, the canonical form of NF-B, between all three sites. As shown in Fig. 7A, the difference in p65/p50 binding affinity between site B2a, which is strongest, and site , which is weakest, is in the region of only 3-4-fold. We do not know whether the polarization of the cluster in respect to p65/p65 and p50/p50 binding affinity is of functional significance. But it has been observed that the acquisition of LPS tolerance in Mono Mac 6 cells is associated with a shift in the predominant form of NF-B from p65/p50 heterodimer to p50 homodimer, suggesting that the former activates and the latter tends to suppress TNF expression (29). Thus, it is interesting to speculate that one function of the B2//B2a cluster within human macrophages might be to switch LPS-induced TNF expression on and off by allowing active interaction between different forms of NF-B and the elements of the cluster.
Gene reporter expression was approximately halved by mutations that interfered with NF-B binding to any of the three sites in the B2//B2a region. Artificial spacing introduced between the sites also resulted in reduction of gene expression. These data suggest that each element of the cluster contributes to gene transcription and that the structural unity of this region may be important for the maximal levels of expression. In addition, when NF-B-specific interactions were ablated throughout the B2//B2a region, LPS-induced transcriptional activation was reduced by 70%. Although the present study focuses specifically on the question of TNF regulation in human monocytes, it raises the more general question as to whether p65/p50 in isolation can activate TNF transcription and, if so, to what extent this depends on the B2//B2a region. Using COS-7 cells we find that overexpression of p65/ p50 up-regulates luciferase gene expression by 50-fold in the presence of the wild type TNF promoter but by less than 5-fold if the promoter is disabled in the B2//B2a region as depicted in Fig. 8A. 3 Taken together with the present findings, this result highlights the need for further work to understand the context-specific factors that may modulate the functional activity of this region of the promoter in the natural context of human monocytes and other relevant cell types.
An important question in resolving the function of the B// B2a cluster is the nature of the constitutive -binding factor that is found in nuclear extracts of Mono Mac 6 cells. The UV cross-linking data indicate that the constitutive complex includes a protein(s) of around 40 kDa that can be clearly distinguished from the LPS-inducible NF-B/Rel proteins. Our EMSA analysis of mutations within would suggest that the inducible binding motif overlaps with, but is not identical to, the constitutive binding motif. One possibility is that the constitutive binding factor recognizes a B-like motif. Examples of this phenomenon include NF-GMa, whose recognition se-quence is very similar to the NF-B consensus (17), and recombination signal sequence binding protein Jk, which is associated with the B-binding sequence in the interleukin-6 gene promoter (30). However, these are both unlikely candidates for the -binding factor, because they tend to migrate much faster than NF-B on EMSA. A second possibility is that the constitutive factor might recognize another motif embedded within the site, and this raises the further intriguing possibility that it may occupy the site simultaneously with NF-B. The general principle of transcription factor cohabitation is illustrated by a site in the human interferon-␤ promoter, which simultaneously binds NF-B and the high mobility group protein HMG I(Y). NF-B contacts the outer GC-rich bases in the major groove of the DNA helix, whereas HMG I(Y) binds to the central AT-rich section that is exposed in the minor groove, and both factors must be present to activate transcription (31)(32)(33). We have not formally excluded HMG I(Y) as the -binding factor, although the DNA sequence of and the size of the proteins visualized by UV cross-linking would argue against this explanation. We are currently attempting to clone the constitutive -binding factor; however, it is clear from these observations that whatever its molecular identity, its effects on the functional activity of NF-B will be of considerable interest.
Understanding the role of NF-B in human TNF gene regulation has proven difficult, with studies using different cell types and stimuli reaching opposite conclusions. Such context specificity may be of particular importance if, as has been shown for other cytokine genes, the functional activity of the promoter region under consideration requires individual transcription factors to be assembled into higher order enhancer complexes (32). It has been recently demonstrated that cooperation of the B3 site with an adjacent cAMP-responsive element site provides an enhancement of TNF transcription in LPS-stimulated THP-1 cells (13). Our findings support the notion that an assembly of protein complexes at B2//B2a is required for efficient induction of TNF transcription by LPS in Mono Mac 6 cells. Further studies are needed to understand the functional significance of specific interactions that may occur in this region between each of the NF-B binding sites and between NF-B and other nuclear factors.