Post-induction, Stimulus-specific Regulation of Tumor Necrosis Factor mRNA Expression*

The tumor necrosis factor (TNF) gene is activated by multiple extracellular signals in a stimulus- and cell type-specific fashion. Based on the presence of κB-like DNA motifs in the region upstream of the TNF gene, some have proposed a direct role for NF-κB in lipopolysaccharide (LPS)-induced TNF gene transcription in cells of the monocyte/macrophage lineage. However, we have previously demonstrated a general and critical role for a minimal TNF promoter region bearing only one of the κB-like motifs, κ3, which is bound by nuclear factor of activated T cell proteins in lymphocytes and fibroblasts in response to multiple stimuli and Ets proteins in LPS-stimulated macrophages. Here, in an effort to resolve these contrasting findings, we used a combination of site-directed mutagenesis of the TNF promoter, quantitative DNase I footprinting, and analysis of endogenous TNF mRNA production in response to multiple stimuli under conditions that inhibit NF-κB activation (using the proteasome inhibitor lactacystin and using cells lacking either functional NF-κB essential modulator, which is the IκB kinase regulatory subunit, or the Nemo gene itself). We find that TNF mRNA production in response to ionophore is NF-κB-independent, but inhibition of NF-κB activation attenuates virus- and LPS-induced TNF mRNA levels after initial induction. We conclude that induction of TNF gene transcription by virus or LPS does not depend upon NF-κB binding to the proximal promoter; rather, a stimulus-specific post-induction mechanism involving NF-κB, yet to be characterized, is involved in the maintenance of maximal TNF mRNA levels.

Tumor necrosis factor (TNF, 3 originally known as tumor necrosis factor-␣ or cachectin) is a cytokine with multiple roles in the innate and adaptive immune response. TNF expression is rapidly induced in response to multiple extracellular stimuli, including engagement of antigen receptor on T and B cells and infection by certain viruses and bacteria (1); indeed, TNF is one of the first genes expressed in activated T cells (2). TNF is also a potent activator of the NF-B family of proteins, which activate transcription by binding to DNA motifs with the consensus sequence 5Ј-GGGRNYYYCC-3Ј, where R is a purine (typically A), Y is a pyrimidine (typically T), and N is any nucleotide (3,4).
Shortly after the identification of NF-B as an inducible DNA-binding factor (5,6), four DNA motifs resembling NF-B-binding sites were described in the promoter of the murine TNF gene (7,8), eventually designated B1, B2, B2a, and B3 (9,10), and three in the promoter of the human TNF gene, 1, 2, and 3 (11) (Fig. 1). A role for NF-B in LPSinduced TNF gene transcription was initially postulated based upon the murine TNF promoter sites (7,8,12). We note that of these murine promoter elements, B3, the site with the highest affinity for NF-B (7,8,10,12), is not conserved in the human promoter or in primate promoters in general (13). Furthermore, the three B-like motifs in the human TNF promoter do not function as LPS-or virus-inducible enhancers when multimerized and fused to a heterologous promoter, as would be expected for a true NF-B site (11).
Later studies identified three additional NF-B-like sites in the human TNF promoter, B1, B2, and (Fig. 1). Of these B-like sites, B1, which lies 500 -510 bp upstream of the TNF transcription start site, has the highest affinity for NF-B (9,10,14), but mutation of the site has no effect on LPS-mediated induction of TNF gene expression (9,10). The 2 site (11), a cytokine-1-like motif (15), has consistently been shown to have little or no affinity for NF-B and no role in TNF gene expression (9 -11, 14, 16, 17). The closely spaced B2, , and 1 sites (corresponding to murine B2a) have been reported to be weak NF-B p50/p65-binding sites (9,10). Although the 1 site at Ϫ597 to Ϫ588 bp relative to the TNF mRNA cap site was reported to be a stronger p50/p65-binding site than B2 in one report (16), another study stated that 1 fails to bind NF-B (14). It has been reported that point mutations of these sites reduce LPS induction of the TNF gene roughly 2-fold (9,10); however, original studies demonstrated that specific deletion of 1 or 2 had no effect upon LPS or virus inducibility of the TNF in reporter assays (11). Consistent with these results, multiple studies have concluded that the 200 proximal base pairs of the human TNF promoter, which lack 1 and 2 ( Fig. 1), are sufficient for its induction by LPS in multiple monocytic cell lines (11,14,17,18). We have also shown that this minimal TNF promoter region is sufficient for activation of TNF gene transcription and formation of specific enhancer complexes in response to several other stimuli in a variety of cell types (18 -20).
We have previously demonstrated that the 3 site, which lies within the proximal human TNF promoter region (Fig. 1), behaves as a composite element in conjunction with an adjacent cAMP-response element site (which binds ATF-2/c-Jun) and is a functional binding site for an NFATp dimer (2,(21)(22)(23)(24). Proteins of the Ets/Elk family can also bind to the 3 site and are involved in LPS-and Mycobacterium tuberculosis-mediated TNF gene activation in cells of the monocyte/macrophage lineage (18,25). Consistent with this, mutation of the 3 site has been shown to result in a decrease of LPS-mediated TNF gene expression (9,10,14,17,18) and to disrupt Ets/Elk binding to the site (18). Although some studies reported that NF-B p50/ p65 bound to 3 (14,17,26,27), we and others have shown that the 3 site has very low or negligible affinity for NF-B in gel shift assays (9,10,16,23) and that it exhibits no specific binding to p50/p65 in quantitative DNase I footprinting assays (23). Furthermore, whereas multimers of true NF-B sites can confer LPS inducibility when placed upstream of a heterologous promoter, multimers of isolated copies of 3 cannot (11,14).
A role for NF-B in LPS-induced TNF gene regulation was also initially suggested by the observation that inhibition of TNF gene expression was associated with treatment of cells with pyrrolidine dithiocarbamate or sodium salicylate, which can inhibit LPS-induced nuclear localization of NF-B (16,28). However, these studies provided no direct evidence of a functional role for NF-B in TNF gene expression. Later studies showed that ectopic expression of a dominant-negative form of the NF-B inhibitor protein, IB␣, resulted in a decrease in TNF transcription and synthesis in LPS-treated human primary macrophages or monocytic cell lines (10,29,30) also suggesting a role for NF-B in TNF gene expression.
In an effort to resolve the basic discrepancy between the lack of function of upstream B-like motifs in the TNF promoter and the apparent influence of NF-B upon LPS-mediated TNF transcription in the studies described above, we examined the effects of altering the interaction of NF-B with the TNF promoter and the time dependence of NF-B inhibition upon TNF gene expression in human and murine cells in response to multiple extracellular stimuli. We took the experimental approaches of promoter mutagenesis, in vitro footprinting assays, gene reporter assays, and specific inhibition of the NF-B pathway by using both a pharmacological inhibitor, lactacystin, and cells lacking NEMO, a component of the kinase complex required for NF-B activation. Here we show the following: (i) TNF mRNA expression in T cells activated by calcium influx is independent of NF-B; (ii) the initial phase of TNF mRNA expression in virus-stimulated T cells or LPS-stimulated monocytic cells is not blocked by inhibition of NF-B activity; and (iii) a novel late phase inhibition of TNF mRNA expression occurs following TNF induction by virus or LPS when the NF-B activity is inhibited. Thus, we show that although NF-B does not regulate TNF gene expression through the 200 proximal base pairs of the TNF promoter, it does appear to play a role in sustaining TNF mRNA levels following induction of the gene by virus in T cells and by LPS in cells of the monocyte/macrophage lineage, pointing to an indirect or secondary effect of NF-B in TNF induction by these stimuli, perhaps through cross-talk with other signal transduction pathways or transcription factors involved in direct regulation of the gene.

EXPERIMENTAL PROCEDURES
EMSA and Quantitative DNase I Footprinting-Bacterially expressed and purified NF-B p50/p65 heterodimer and NFATp were prepared as described previously (31,32). For EMSAs, recombinant NF-B p50/65 heterodimer (33) was incubated at a final volume of 20 l at increasing concentrations with oligonucleotide probes of the 3 site (5Ј-GATCCT-GGGTTTCTCCA-3Ј and 5Ј-GATCTGGAGAAACCCAG-3Ј) or the PRDII site (5Ј-GATCCTGGGAAATTCCA-3Ј and 5Ј-GATCTGGAATTTCCCAG-3Ј) end-labeled with 32 P using T4 polynucleotide kinase. Protein-DNA complexes and free DNA were then resolved by electrophoresis at 4°C on Tris borate, 4% EDTA 29:1 acrylamide:bisacrylamide gels as described previously (2) and visualized by autoradiography. For quantitative DNase I footprinting, wild-type and mutant TNF promoter fragments (Ϫ200 to ϩ87) and wild-type IFN-␤ promoter fragment (Ϫ105 to ϩ68) were end-labeled at the coding strand and incubated at a final volume of 50 l with recombinant NF-B p50/p65 or NFATp at the concentrations indicated in the figure legends and digested with 0.33 units of DNase I as described previously (23,32). Ethanol-precipitated DNA was resuspended in 3 l of formamide dye and resolved on 8% wedged (0.4 -1.4 mm) denaturing acrylamide gel as described previously (31,32).
Plasmids-The Ϫ200 TNF CAT and Ϫ105 IFN-␤ CAT constructs have been described previously (11,22). The 3 site (5Ј-GGGTTTCTCC-3Ј) was converted to the PRDII site (5Ј-GGGAAATTCC-3Ј) in the context of the wild-type Ϫ200 TNF CAT reporter by site-directed mutagenesis (circular mutagenesis chain reaction, QuikChange, Stratagene) and confirmed by sequencing. The BamHI-XbaI fragments of these CAT vectors were used as templates for DNase I footprinting.
Cell Culture-The murine IL-2-dependent T cell clone Ar-5 was maintained as described (2), stimulated with 1 M ionomycin (Calbiochem) or Sendai virus (SPAFAS, Cantrell strain) at 300 hemagglutinating units/ml, as described previously (2), and inhibited with 50 M lactacystin (Calbiochem) added to the medium immediately before the addition of ionomycin or virus. The human epithelial cell line HeLa was maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum, 2 mM L-glutamine, 100 units/ml penicillin, and 100 g/ml streptomycin and stimulated with Sendai virus as described above. The human monocytic cell line Mono-Mac 6 (16) was maintained in RPMI 1640 medium supplemented with 10% fetal calf serum. The murine monocytic cell line J774 was maintained in Dulbecco's modified Eagle's medium supplemented with 12% fetal bovine serum, 12 mM HEPES, 2 mM L-glutamine, 50 M ␤-mercaptoethanol, 60 units/ml penicillin, and 60 g/ml streptomycin. Mono-Mac 6 and J774 cells were activated using LPS (Sigma; Escherichia coli O111:B4) at a concentration of 1 g/ml and inhibited with 50 M lactacystin as above. The murine pre-B cell line 70Z/3 and its derivative 1.3E12 (34), which does not express NEMO (35), were maintained in RPMI 1640 medium supplemented with 10% fetal calf serum and 50 M ␤-mercaptoethanol and activated with virus and LPS, as described above, or with phorbol 12-myristate 13-acetate (PMA; Calbiochem) at 20 ng/ml plus 1 M ionomycin (P ϩ I). The Nemo-deficient murine embryonic fibroblast line (36) and wild-type fibroblasts were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum, 2 mM L-glutamine, and 1 mM sodium pyruvate and activated with 300 hemagglutinating units/ml Sendai virus.
Transfections-Transfections in the Ar-5 T cells were performed using the DEAE-dextran method as described previously (31). At 24 h after transfection, Ar-5 cells were treated with ionomycin or pretreated for 10 min with 1 M cyclosporin A (Sandoz) prior to treatment with ionomycin and harvested ϳ18 h later. Transfections in the HeLa cells were performed using Lipofectamine (Invitrogen) at a 1:1 ratio with DNA according to the manufacturer's protocol. Chloramphenicol acetyltransferase (CAT) assays were performed on cell extracts as described (22,37).

RESULTS
We have demonstrated previously the binding of several transcription factors to the TNF promoter in a cell type-and stimulus-specific fashion using chromatin immunoprecipitation assays (18,20,24,25). The level of resolution of chromatin immunoprecipitation is ϳ300 bp (38), too large to discern binding to a given cognate DNA motif. Given the conflicting results in the literature regarding NF-B p50/p65 heterodimer binding to the 3 site, we sought to examine transcription factor binding at the resolution of individual base pairs using in vitro and in vivo assays, that is, quantitative DNase I footprinting with recombinant proteins and site-directed mutagenesis of the 3 site, which we designed to specifically recruit NF-B to the 3 site in gene reporter assays.
Consistent with our previous work and that of others (9, 10, 16, 23), the NF-B p50/p65 heterodimer has very low affinity for the TNF promoter 3 site relative to PRDII, a functional NF-B-binding site from the human interferon-␤ (IFN-␤) promoter (39). As shown in Fig. 2, in an EMSA using 3 and PRDII oligonucleotide probes, we only observe the formation of an NF-B-3 complex at the highest protein concentration, whereas an NF-B-PRDII complex is apparent at much lower protein concentrations (Fig. 2). Using DNase I footprinting, we compared the binding of NF-B p50/p65 and NFATp to both the wild-type human TNF promoter and a mutant TNF promoter in which the 3 site (5Ј-GGGTTTCTCC-3Ј) was replaced with PRDII (5Ј-GGGAAATTCC-3Ј); note that this entails altering only four nucleotides in the TNF promoter (boxed in Fig. 2). In agreement with our previous results using the Ϫ200 TNF promoter fragment (20,23), NF-B does not bind specifically to the 3 site in quantitative DNase I footprinting assays (Fig. 3A, compare lane 2 with lanes 6 -8). By contrast, NFATp readily binds to 3 and other NFAT motifs within the TNF promoter (Fig. 3A, lanes 2-5). When the 3 site is changed to PRDII, however, NF-B clearly binds to the mutant TNF promoter at PRDII (Fig. 3A, compare lanes 10 and 14 -16). Note that NFATp binding patterns to the mutant TNF promoter are unaffected by this four-nucleotide change converting 3 to PRDII (Fig. 3A, lanes 10 -13), consistent with the observation that the recombinant NFATp dimer binds to PRDII. 4 As a control, we performed DNase I footprinting assays of NF-B p50/p65 and NFATp binding to the IFN-␤ promoter. Both NFATp (Fig. 3B, lanes 2-5) and, as expected (32), NF-B p50/ p65 (Fig. 3B, compare lanes 2 and 6 -8) bind to PRDII. Thus, NF-B p50/p65 does not exhibit specific binding to the 3 site of the TNF promoter.
We next sought to determine the functional impact of altering NF-B recruitment to the TNF promoter at the position of the 3 site. We reasoned that if endogenous NF-B is not recruited to the site in the context of the wild-type promoter, forcing the recruitment of NF-B by converting 3 to PRDII should have an obvious effect upon transcription driven by the TNF promoter. We note that infection of HeLa cells with Sendai virus leads to activation of NF-B and induction of the IFN-␤ gene, but HeLa cells do not express NFAT (40), and endogenous TNF is not induced by virus in these cells (41). By contrast, the IFN-␤ promoter is highly virus-inducible in HeLa cells (42).
Consistent with these previous reports, we observed that transcription from the IFN-␤ reporter is highly virus-inducible, and the wild-type TNF gene promoter is not induced by virus infection in HeLa cells (Fig. 4A). Strikingly, changing 3 to PRDII in the context of the wild-type TNF promoter creates a reporter that is virus-inducible in this cell line (Fig. 4A). We next compared the effect of changing 3 to PRDII upon the transcriptional induction of the TNF promoter in the murine T cell clone, Ar-5, in response to treatment with ionomycin, a calcium ionophore. Treatment with ionomycin results in calcineurin-dependent nuclear translocation of NFATp but not of NF-B (3,43,44). This nuclear translocation of NFATp can in turn be strongly inhibited by cyclosporin A (CsA), which blocks the phosphatase activity of calcineurin (43,45); human or murine TNF gene expression in response to calcium ionophore requires NFATp and can be blocked by inhibition of calcineurin activity (2,21,46,47). As predicted by our in vitro binding studies with NFATp above, changing 3 to PRDII in the context of the TNF promoter preserves wild-type levels of ionomycin induction and sensitivity to CsA (Fig. 4B). Note that IFN-␤, an NF-B-dependent gene, is not inducible by ionomycin in T cells (Fig. 4B), consistent with the necessity of the formation of an NF-B-dependent enhanceosome for IFN-␤ gene expression (32). Thus, converting the TNF 3 site to PRDII does not impair NFATp-dependent transcription in response to calcium ionophore in T cells, although it permits NF-B-dependent transcription of TNF in response to virus in a context in which the gene is normally silent, demonstrating a clear functional consequence of forcing the recruitment of NF-B to the TNF promoter.
In resting cells, NF-B proteins typically exist in complex with IB proteins. A wide range of stimuli trigger the dissociation of the NF-B-IB complex and DNA binding of NF-B by activating the IB kinase complex (IKK1, IKK2, and NEMO/ IKK-␥), which phosphorylates IB and targets it for degradation by the proteasome (3). NF-B activation can thus be effectively blocked by specific pharmacological inhibitors of the proteasome, such as lactacystin (48). To determine the role of NF-B in TNF gene regulation by a variety of extracellular stimuli, we utilized both lactacystin and cells that do not express NEMO and thus lack a functional IB kinase complex.
We first stimulated Ar-5 T cells with ionophore or virus in the presence or absence of lactacystin and performed RNase protection assays. As shown in Fig. 5A, pretreatment of T cells with lactacystin does not inhibit ionophore induction of TNF mRNA levels up to 3 h post-ionophore stimulation (lanes 6 and 7). Virus induction of TNF mRNA levels was also not affected at 1 h (Fig. 5A, lanes 2 and 3) but was decreased by ϳ30% at 3 h post-virus induction (Fig. 5A, lanes  4 and 5). Note that the IFN-␤ gene is not inducible by ionophore (Fig. 5A, lanes 8 -10), whereas induction of IFN-␤ by virus, which is dependent upon NF-B, is blocked by treatment with lactacystin (Fig. 5A, lanes 8, 11, and 12).  (Fig. 5B, lanes 2 and 3  and 7 and 8); however, there was a decrease in TNF mRNA levels (ϳ30 -50%) in both cell types at 1 h post-LPS induction (Fig. 5B, compare lanes 4 and 5 and lanes 9 and 10). Thus, although induction of TNF mRNA expression by ionophore in T cells and the early phase of induction of TNF mRNA expression in response to virus in T cells or in response to LPS in monocytic cells was not impaired by lactacystin, there was a late phase stimulus-specific inhibition of mRNA levels following virus and LPS treatment in T cells and monocytic cells, respectively.
Because inhibition of the proteasome by lactacystin may disrupt other cellular processes in addition to activation of NF-B, we next sought to investigate TNF gene regulation by virus and LPS in cells specifically lacking NEMO, the IB kinase regulatory subunit required for activation of NF-B (35). Using the murine pre-B cell line 70Z/3 and its derivative 1.3E2, which does not express functional NEMO (35) and fails to activate NF-B in response to multiple stimuli (49), we investigated TNF gene induction by virus, phorbol ester plus ionophore (P ϩ I), and LPS. We note that we have previously demonstrated NFATpdependent activation of TNF in B cells in response to P ϩ I (23, 24, 50, 51). As shown in Fig. 6, TNF mRNA expression is induced in wild-type 70Z/3 cells by LPS, P ϩ I, and virus (lanes 1-4). Most strikingly, in the corresponding NEMO mutant 1.3E2 cells, P ϩ I induction of TNF was not affected, but induction by LPS and virus was decreased (Fig. 6, lanes [5][6][7][8]. By comparison, as expected, the NF-B-dependent induction of IFN-␤ mRNA expression by virus in wild-type cells (Fig.  6, lanes 9 and 10) is abrogated in the NEMO mutant cells (Fig. 6, lanes 11  and 12).
Given these results in a cell line that fails to express functional NEMO, we next examined the activation of TNF transcription in a more physiological model, embryonic fibroblasts from mice with a targeted disruption of the Nemo gene (36). Because TNF is virus-inducible in fibroblast cells (11,52), we examined the induction of TNF transcription in response to virus; TNF mRNA was first visible in wild-type and Nemo Ϫ/Ϫ murine embryonic fibroblasts (MEFs) 3 h after virus infection (Fig. 7,  lanes 3 and 7); TNF mRNA levels in Nemo Ϫ/Ϫ MEFs, however, are attenuated relative to the wild-type fibroblasts; note that this is particularly evident at the latest time point (Fig. 7, lanes 4  and 8). This additional level of attenuation following initial mRNA induction would not be expected if the primary effect of NF-B upon TNF mRNA expression in response to virus occurred during the initial phase of induction of the TNF gene. . NF-B p50/p65 does not bind specifically to the proximal human TNF promoter. A, quantitative DNase I footprinting using the wild-type human TNF promoter (TNF; Ϫ200 to ϩ87 relative to the start site of transcription) and a mutant promoter (TNF (3 3 PRDII)) in which 3 (5Ј-GGGTTTCTCC-3Ј) has been converted to PRDII (5Ј-GGGAAATTCC-3Ј), a functional NF-B p50/p65-binding site from the IFN-␤ promoter (39), with increasing amounts of recombinant NFATp (4, 20, and 100 ng) in lanes 3-5 and 11-13 or NF-B p50/p65 (4, 20, and 100 ng) in lanes 6 -8 and 14 -16. The positions of 3 (or PRDII) and known NFAT-, Ets-, and Sp1-binding sites (20) are noted. NF-B p50/p65 exhibits no specific binding to 3 but binds to the PRDII site in the context of the TNF promoter, whereas the pattern of NFATp binding to wild-type and mutant promoters is unaffected. B, quantitative DNase I footprinting using the wild-type human IFN-␤ promoter (IFN-␤; Ϫ105 to ϩ69 relative to the start site of transcription) with recombinant NFATp (lanes 3-5) or NF-B p50/p65 (lanes 6 -8) as in A. Note the binding of NFATp and p50/p65 to PRDII.
Thus, induction of TNF gene expression occurs in both of the NEMO-deficient cell lines; however, there is a stimulus-specific dependence upon NF-B for achieving wild-type mRNA levels following LPS and virus induction. Taken together, these results indicate that although induction of TNF gene expression does not require recruitment of NF-B to the minimal TNF promoter, there is a secondary role of NF-B in the maintenance of mRNA levels in cells stimulated with virus and LPS.

DISCUSSION
Dysregulation of TNF can lead to a variety of immunopathologies, including septic shock, Crohn disease, and rheumatoid arthritis (1). Thus, control of TNF is critical and is the focus of research in autoimmunity in particular (53). Despite many years of study, however, there remain conflicting results in the literature regarding the regulatory regions and transcription factors involved in TNF gene transcription in LPS-stimulated monocytic cells. Given that LPS is a major source of TNF during sepsis (1), this is an issue of considerable clinical significance.
Some studies have concluded that the distal regions of the human and murine TNF promoter are important for maximal levels of transcriptional induction in response to LPS (9, 10), whereas we and others have concluded that the proximal regions of the human and murine TNF promoters (within 200 bp of the start site of transcription) are sufficient for maximal , treated with ionomycin (Ionophore), or treated with ionomycin and cyclosporin A (Ionophore/CsA) 24 h post-transfection, as indicated, followed by analysis of CAT activity. Both the wild-type and mutant TNF reporters are activated to nearly identical levels by ionophore and inhibited to similar extents by the presence of cyclosporin A, a calcineurin inhibitor that blocks nuclear translocation of NFAT, consistent with the observation that the reporter constructs have identical NFATp binding patterns (Fig. 3A). Note that, as expected, IFN-␤ is not inducible by ionophore, even though NFATp can bind to the IFN-␤ promoter (Fig. 3B).  levels of induction of TNF transcription in response to LPS (11,14,17,18,54). Although we cannot further address this discrepancy here, it should be noted that all of these studies have shown consistently that mutation of the 3 site, which binds NFAT and Ets/Elk proteins and lies within the proximal promoter region, results in a decrease of LPS inducibility of the human TNF gene (9 -11, 14, 17, 18); similarly, mutation of the site corresponding to 3 (see below) in the murine TNF promoter reduces LPS inducibility of the murine TNF gene (10,54). Furthermore, mutation of the Egr-1, cAMP-response element, Ets, and Sp1 elements in the proximal promoter region also reduces LPS inducibility of the human and murine TNF genes (14,18,54,55).
The 3 site of the human TNF promoter (5Ј-GGGTT-TCTCC-3Ј) varies from the consensus B motif (5Ј-GGGRNYYYCC-3Ј) at the 4th position, where 3 has a T instead of a purine; furthermore, the pyrimidine at the 7th position of consensus B motifs is most frequently T, whereas 3 has a C (3,4). It should be noted that at the corresponding position in the murine TNF promoter, the sequence is 5Ј-GGTTTTCTCC-3Ј, which is not an NF-B site but is a strong NFAT-binding motif (45); thus, no B sites are present in the proximal murine TNF promoter.
For the NF-B p50/p65 heterodimer in particular, all available protein-DNA co-crystal structures (56 -58) illustrate contacts within the 5Ј and 3Ј half-sites of B motifs bearing 5Ј-GA-3Ј dinucleotide steps (including PRDII), making the 5Ј-GT-3Ј step in the 5Ј-half of the 3 site distinct from these and most other known p50/p65-binding sites (59). Some have reported the binding of NF-B p50/p65 to the 3 site, using anti-p50 and anti-p65 antibodies and high concentrations of extracts from monocytic cells stimulated with LPS (14,26,27) or the superantigen staphylococcal enterotoxin A (17). As we have shown here, recombinant p50/p65 only binds to an oligonucleotide containing the isolated 3 site at high protein concentrations in an EMSA (Fig. 2), and recombinant p50/p65 exhibits no specific binding to the 3 site or other sites in the TNF proximal promoter in quantitative DNase I footprinting unless 3 is converted to PRDII (Fig. 3A). Furthermore, in other studies using EMSA that compared the relative affinity of NF-B p50/p65 for 3 and for known B sites or other putative B sites from the TNF promoter, 3 was, in nearly every case, classified as either having negligible affinity for p50/p65 (9,10,16) or having weak affinity relative to the other sites, for example 1 (14,27). Finally, we find that converting 3 to a bona fide p50/p65 site, PRDII, by mutating four nucleotides within the context of the proximal promoter confers virus inducibility upon the TNF promoter in HeLa cells (Fig. 4A), consistent with virus-induced recruitment of p50/p65 to PRDII in this context, as is the case for the IFN-␤ promoter (42,60). We conclude that at physiological concentrations, NF-B does not bind to 3 within the proximal TNF promoter.
There have been many reports that inhibition of NF-B, by pharmacological compounds such as pyrrolidine dithiocarbamate or salicylates (16,28) or by expression of dominantnegative forms of either IB␣ (10,29,30) or subunits of the IB kinase complex (30,61), is associated with inhibition of expression of TNF transcription in response to LPS in monocytic cell lines (or primary macrophages, see Ref. 29). This inhibition is typically partial, leading other groups to invoke the influence of factors in addition to NF-B in TNF transcription in response to LPS (9,10). It should be noted, however, that virus induction of IFN-␤, a true NF-B-dependent gene that also depends on coordinate activation of other transcription factors (39), is essentially abrogated when NF-B activity is blocked, as we observe in the case of cells treated with lactacystin ( Fig. 5A) or lacking functional NEMO (Fig. 6). We observe that lactacystin has no effect on TNF mRNA expression induced by ionophore in T cells or upon the initial induction of TNF mRNA expression induced by virus in T cells (Fig. 5A) or by LPS in monocytic cells (Fig. 5B). Similarly, there is no difference in the level of TNF mRNA induced by phorbol ester and ionophore in B cells lacking NEMO relative to wild-type B cells, whereas TNF mRNA levels in response to virus and LPS are reduced in the NEMO-deficient B cells (Fig. 6), and there is a late phase attenuation of TNF mRNA levels in response to virus infection in fibroblasts lacking the Nemo gene relative to wild-type fibroblasts (Fig. 7). Thus, whether NF-B activity in wild-type cells is inhibited or a component of the NF-B signaling pathway is specifically deleted, the cumulative end result is a relative reduction of TNF mRNA. Our results are consistent with a post-induction role for NF-B, independent of binding to the proximal TNF promoter, in maintaining TNF mRNA production following exposure to LPS, in the case of monocytic cells, and virus, in the case of T cells, B cells, and fibroblasts. This is consistent with earlier observations that inhibition of NF-B leads to reduction in TNF mRNA and protein expression in response to LPS.
Mechanisms apart from direct transcriptional regulation have long been postulated for LPS-mediated TNF expression but with conflicting results. For example, an initial study concluded that TNF biosynthesis in the murine macrophage line RAW 264.7 was regulated at the translational level (62), whereas a later study concluded that LPS-mediated synthesis of TNF in RAW 264.7 cells was regulated primarily at the level of transcription rather than translation (63). Although characterization of a mechanism for the NF-B-dependent post-induction effect we observe for TNF expression in response to virus and LPS lies outside the scope of this study, we speculate that it  1-4) and MEFs from mice with targeted disruption of the Nemo gene (Nemo Ϫ/Ϫ , lanes [5][6][7][8], which fail to activate NF-B. Cells were mock-induced or stimulated with Sendai virus (Vir), and RNA was analyzed at 0.5, 3, and 6 h post-stimulation as indicated. Note the attenuation of TNF mRNA production in fibroblasts lacking Nemo is attenuated relative to their wild-type counterparts, particularly at the latest time point. may arise from two principal sources as follows: NF-B binding at distal elements in the TNF locus or cross-talk between NF-B activation and signal transduction pathways regulating TNF expression.
As noted above, with regard to the first scenario, binding of NF-B to the 5Ј distal TNF promoter has been reported (9,10,12), and an NF-B site downstream of the final murine TNF exon has also been described (64), which is a stronger NF-Bbinding motif than the upstream TNF promoter murine B sites (10). These motifs do not function as virus-or LPS-inducible elements in isolation (11,14); furthermore, we have shown, in the same cell lines used by other groups in earlier studies, that deletion of the 5Ј distal NF-B sites from the human or mouse TNF promoters does not impair LPS inducibility of the TNF promoter (18). Thus, although we have not observed a role for NF-B sites in the distal TNF promoter in gene reporter assays, it remains a formal possibility that in the native chromatin context, which we have examined here through RNase protection assays, the binding of NF-B outside of the proximal promoter in response to certain stimuli may contribute to maintenance of TNF mRNA expression.
It is also worth noting that another effect, distinct from direct induction of TNF transcription, has been linked to NF-B binding at sites in the upstream TNF promoter region, LPS tolerance in macrophages, i.e. when macrophages are desensitized by prior treatment with low LPS levels. LPS tolerance and other cases of TNF down-regulation have been attributed to the binding of NF-B p50 homodimer, which plays a general repressive role in gene transcription (3,65,66), to sites in the distal TNF promoter (66 -71).
With regard to the second scenario, cross-talk between NF-B and other signal transduction pathways, the current understanding of signaling events resulting from exposure to LPS has evolved considerably since the time that TNF was first postulated to be an NF-B-dependent gene, although many details remain to be elucidated. It is now known that LPS acts through the transmembrane protein toll-like receptor 4 (TLR4) (72), which activates both NF-B and the mitogen-activated protein kinase (MAPK) pathways through the adapter proteins MyD88 and TIRAP (also called Mal) and interferon regulatory factor (IRF) 3 through the adapter proteins TRAM and TRIF (73)(74)(75). LPS triggers a biphasic activation of NF-B in macrophages; in the early phase (within roughly 30 min of LPS treatment), TLR4 activates the IKK complex through an MyD88-dependent pathway; in the late phase, TNF production activates NF-B via the TNF receptor in an MyD88-independent, autocrine fashion (76,77). LPS induction via TLR4 also activates both the mitogen-activated protein 3-kinase Tpl2 (also called Cot1), which acts as an upstream master regulator of the extracellular signal-regulated kinase (ERK) pathway (78 -80) and Bruton's tyrosine kinase (Btk) (81)(82)(83).
Some studies have implicated adapter proteins proximal to TLR4 in the regulation of TNF; macrophages from mice lacking the MyD88, Tirap, Tram, or Trif genes show decreased levels of TNF biosynthesis in response to LPS relative to their wild-type counterparts (84 -87), although transfection of dominant-negative forms of MyD88 and TIRAP in human macrophages did not reduce production of TNF (88). We and others have shown a role for the ERK pathway in TNF transcription and translation in response to LPS (18, 78, 89 -91); studies in Tpl2 knock-out mice revealed that LPS increased nuclear export of TNF mRNA in murine peritoneal macrophages via a Tpl2-ERK1/ERK2-dependent, NF-B-independent pathway (78). Recent studies have also implicated the Btk pathway in TNF mRNA stability during TNF gene induction by LPS in human and murine monocytes and macrophages, dependent on the p38 MAPK (82,83,92), although another study in human macrophages concluded no role for Btk in LPS-mediated TNF production (93). Studies of early and late phase NF-B activation in response to LPS (76, 77) implicated IRF-3 in the late phase of NF-B activation based on RNA interference experiments in MEFs lacking MyD88 (76), whereas a role for IRF-5 in LPSmediated TNF biosynthesis was also postulated based on experiments in splenic macrophages from mice lacking the irf5 gene (94). Binding of either IRF protein to the TNF promoter was not examined in these studies, however.
Thus, the signal transduction pathways downstream of LPS/ TLR4-mediated signaling are quite complex, and there are several potential points of cross-talk between the activation of NF-B and the ERK or Btk pathways. Notably, Tpl2 exists in a complex with NF-B p105, the precursor protein of the p50 subunit of NF-B (conversely, about one-third of p105 is in complex with Tpl2) (95), which is required for stability and function of Tpl2 in response to LPS (96 -100). The Btk pathway has also been linked to NF-B activation in macrophages, specifically through phosphorylation of the NF-B p65 subunit (101,102), the activating subunit of the p50/p65 heterodimer (3,65). Blockade of NF-B activation may thus impair the late phase effect of NF-B on LPS-induced TNF mRNA expression by influencing other TNF-inducing signal transduction pathways. Although the signal transduction pathways leading to virus-mediated TNF gene expression are less well understood, the TLR family has been implicated in some aspects of the process (75), and as mentioned above, the virus-inducible factor IRF-3 has been implicated in late phase NF-B expression in response to LPS (76). Thus, some of the post-induction NF-Bdependent effects we observe may be shared by LPS-and virusmediated pathways of TNF induction.
Our studies illustrate the complexity of cell-and inducerspecific gene regulation and the difficulty of pinpointing specific pathways based mainly upon gene reporter studies and simple in vitro protein-DNA binding assays. Given that NF-B has been postulated as a therapeutic target for conditions involving TNF dysregulation, such as rheumatoid arthritis (29), our findings regarding an NF-B-independent phase of TNF mRNA expression are of particular clinical relevance as signal transduction pathways regulating TNF biosynthesis are elucidated and therapeutic strategies are developed.