A Sp1 Binding Site of the Tumor Necrosis Factor α Promoter Functions as a Nitric Oxide Response Element*

Regulation of gene transcription is an incompletely understood function of nitric oxide (NO). Human leukocytes produce increased amounts of tumor necrosis factor α (TNF-α) in response to NO. This effect is associated with decreases in intracellular cAMP, suggesting that NO might regulate gene transcription through promoter sequences sensitive to cAMP such as cAMP response elements (CRE) and Sp1 binding sites. Here we report that a Sp1 binding site in the TNF-α promoter conveys NO responsiveness. Human U937 cells were differentiated for TNF-α production with phorbol 12-myristate 13-acetate. NO donors and H89, an inhibitor of cAMP-dependent protein kinase increased, while dibutyryl cAMP (Bt2cAMP) decreased TNF-α promoter activity. Deletion or mutation of the proximal Sp1 site, but not the CRE site, abolished the activating effects of NO donors and H89. Further, NO- and H89-mediated increases in TNF-α promoter activity were associated with decreased Sp1 binding. The insertion of Sp1 sites into a minimal cytomegalovirus promoter conferred NO responsiveness, an effect blocked by Bt2cAMP. Mutation of these inserted Sp1 sites prevented this heterologous promoter from responding to NO, H89 and Bt2cAMP. These results identify the Sp1 binding site as a promoter motif that allows NO to control gene transcription.


pathways. Previous experiments have shown that NO donors
increase tumor necrosis factor ␣ (TNF-␣) synthesis in human neutrophil (2) and peripheral blood mononuclear cell preparations (3) through a cGMP-independent mechanism. Recently, we further demonstrated that endogenously produced NO also up-regulates TNF-␣ production by a cGMP-independent mechanism in phorbol 12-myristate 13-acetate (PMA)-differentiated U937 cells transfected with murine, inducible NO synthase (8).
U937 cells lack soluble guanylate cyclase and do not respond to NO with a cGMP signal (8 -9), suggesting that these cells might be useful for exploring cGMP-independent mechanisms by which NO regulates gene transcription. Experiments with U937 cells have demonstrated that NO augments TNF-␣ production through a signaling pathway dependent on decreases in intracellular cAMP (9). This finding suggests that NO might regulate TNF-␣ at the level of gene transcription through effects on cAMP sensitive promoter sites.
cAMP is known to regulate gene transcription through effects on cAMP response elements (CRE) (10). Furthermore, for some genes such as CYP11A and urokinase, cAMP has been shown to alter promoter activity through incompletely defined effects on Sp1 binding sites (11,12). Similar to CRE-binding proteins, cAMP-dependent protein kinase (PKA) phosphorylation of Sp1 has been shown to enhance its DNA binding activity (13). Although Sp1 binding can activate transcription (14,15), for some genes Sp1 functions as a repressor (16,17). The promoter motifs or cell characteristics that determine these divergent effects of Sp1 binding are not fully understood. Therefore, CRE or Sp1 binding sites in the TNF-␣ promoter (18 -20) may be important in mediating cAMP-dependent effects of NO on TNF-␣ transcription. To investigate this question, we examined the effects of NO on the human TNF-␣ promoter in differentiated U937 cells.
Plasmid Construction-The two-plasmid reporter gene system chosen for these experiments lacks cryptic CRE sites that might lead to spurious cAMP-dependent effects (21). The first plasmid was made from plasmid pTet-off (CLONTECH), designated here as pCMV-tTA because it uses the strong immediate cytomegalovirus (CMV) promoter to express tetracyclin (tet)-responsive transcriptional activator (tTA). tTA is a fusion protein of the first 207 amino acids of the tet repressor and of the C-terminal activation domain of the herpes simplex virus VP16 protein. The tTA protein transactivates expression of a reporter gene, chloramphenicol acetyltransferase (CAT), by binding to a tetresponsive element in the promoter of a second plasmid, pUHG10.3CAT (21). Promoterless ptTA vector was generated from pCMV-tTA by removing the whole CMV promoter region using a XhoI/EcoRI digest. The 1311-bp human TNF-␣ promoter, cut from pXP1(-1311TNF) by XhoI/ HindIII (22), was subcloned into the XhoI/EcoRI site of ptTA to generate pTNF-tTA.
Either the CRE sequence (5Ј-TGAGCTCA-3Ј at position Ϫ107 to Ϫ100) or the Sp1 binding site (5Ј-CCCCGCCC-3Ј at position Ϫ52 to Ϫ45) was removed from pTNF-tTA and replaced with a 6-bp EcoRV restriction site (5Ј-GATATC-3Ј) using ExSite™ (Stratagene) to generate the mutant constructs pTNF(dCRE)-tTA and pTNF(dSp1)-tTA, respectively. Furthermore, a pTNF(mSp1)-tTA construct was made from pTNF-tTA by mutating the Sp1 binding site (from 5Ј-CCCCGCCC-3Ј to 5Ј-CCCCGTTC-3Ј) using Chameleon™ (Stratagene). This mutation prevents the binding of Sp1 (23). The other Sp1 site in the TNF-␣ promoter * This work was supported by intramural National Institutes of Health Funds. Portions of this work were presented at the Experimental Biology Meeting, San Francisco, CA, April 18 -22, 1998. 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.
(at position Ϫ173 to Ϫ166) was not studied. It overlaps with an Egr-1 motif and has been shown to bind Egr-1, not Sp1 protein in PMA or lipopolysaccharide stimulated U937 and THP1 cells (20,24).
Next, a pCMV min -tTA construct devoid of enhancer elements was created from pCMV-tTA by PCR. A 35-bp double-stranded oligonucleotide containing three copies of the Sp1 sequence (5Ј-CCCCGCCC CCCGCCCCATTGACGCAATCCCCGCCC-3Ј) or the mutated Sp1 sequence (5Ј-CCCCGTTCCCCGTTCCATTGACGCAATCCCCGTTC-3Ј) was inserted into pCMV min -tTA at a XhoI site to create pCMV min (Sp1) 3 -tTA and pCMV min (mSp1) 3 -tTA, respectively. All vectors were partially sequenced to confirm the correct sequences and orientations.
Cell Transfections and CAT Assay-U937 cells (ATCC) in RPMI 1640 complete medium (10 7 cells/250 l), the two-plasmid reporter gene system, and pSV␤-Gal (Promega), a ␤-galactosidase expression plasmid, were added to a cuvette (10 g/plasmid) and an electric pulse (270 V, 960 F) was applied using an Electro Cell Manipulator ® 600 (BTX Inc.). Cells were then diluted (2.5 ϫ 10 6 cells/ml) and transferred to six-well plates. After incubation at 37°C for 24 h with various reagents in the presence of PMA (100 nM), which induces U937 differentiation and TNF-␣ production, the cells were collected, lysed, and assayed for CAT and ␤-galactosidase using enzyme-linked immunosorbent assay kits (Roche Molecular Biochemicals Co.). ␤-Galactosidase expression was used to normalize results for transfection efficiency. Promoter activity was calculated for each promoter construct as fold induction of CAT expression compared with the CAT expression obtained using the promoterless ptTA plasmid. Expressed in these units the pCMV-tTA promoter construct typically resulted in over 70-fold induction.
Nuclear Protein Extracts-U937 cells were incubated with PMA (100 nM) for 24 h in the absence or presence of SNAP (500 M), H89 (30 M), Bt 2 cAMP (100 M), or SNAP (500 M) plus Bt 2 cAMP (100 M). Nuclear protein extracts were then prepared according to the method described previously (25) except that 1 g/ml each of leupeptin, pepstatin, aprotinin, and antipain (Roche Molecular Biochemicals Co.) were added to both buffer A and C. Protein concentrations in nuclear extracts were determined by BCA protein assay (Pierce).
Statistics-For Figs. 1 and 2 comparisons were made using twotailed Student's t tests adjusted with Holm's procedure. Analysis of variance followed by post-hoc tests using Fisher's least significant difference method was employed to analyze the data in Figs. 4 and 5. Differences were considered to be significant at p Ͻ 0.05.

RESULTS AND DISCUSSION
Effect of NO and PKA Activation or Inhibition on TNF-␣ Promoter Activity-TNF-␣ promoter activity was undetectable in naive, undifferentiated U937 cells under all test conditions (data not shown). Two NO donors, SNAP and SNOG, increased wild type TNF-␣ promoter activity in PMA-differentiated U937 cells by 128 Ϯ 36% (Fig. 1A, p Ͻ 0.05) and 92 Ϯ 4% (Fig. 1B, p Ͻ  0.04), respectively. If this increase in TNF-␣ promoter activity (Fig. 1, A and B) is occurring through the cAMP-lowering effect of NO, then Bt 2 cAMP would be expected to decrease TNF-␣ promoter activity and a PKA inhibitor should mimic the effect of NO by increasing TNF-␣ promoter activity. As shown in Fig.  1, A and B, respectively, Bt 2 cAMP decreased TNF-␣ promoter activity by 39 Ϯ 5% (p Ͻ 0.02), and H89, an inhibitor of PKA, increased TNF-␣ promoter activity by 67 Ϯ 6% (p Ͻ 0.04).
Effect of Site Deletions on NO Responsiveness of the TNF-␣ Promoter-Substitution of the CRE sequence or the Sp1 binding site with an unrelated 6 bp sequence reduced basal TNF-␣ promoter activity by 43 Ϯ 6% (p Ͻ 0.003) and 48 Ϯ 10% (p Ͻ 0.004), respectively ( Fig. 2A). The NO donors SNAP and SNOG, which reduce cAMP concentrations below basal levels, and H89, which mimics this effect of NO by inhibiting PKA (9), still enhanced activity of the TNF-␣ promoter after replacement of the CRE sequence (Fig. 2, B and C, p Ͻ 0.05 for all). These results demonstrated that NO and H89 similarly altered gene transcription independent of the CRE site. Likewise, Bt 2 cAMP repressed promoter activity despite replacement of the CRE site (Fig. 2B, p Ͻ 0.02). In contrast, Sp1 binding site replacement completely abolished the activating effects of NO donors and H89 (Fig. 2, B and C, p Ͼ 0.8 for all) and partially blunted the inhibitory effect of Bt 2 cAMP (Fig. 2B, p Ͼ 0.1 for inhibition). These data suggested that the Sp1 binding site, not the CRE sequence, mediates induction of the TNF-␣ promoter by both NO and the PKA inhibitor H89. The borderline effect of Bt 2 cAMP on the promoter lacking a Sp1 site suggests that elevated cAMP concentrations may affect TNF-␣ transcription through sites other than Sp1, such as CRE (10) or NF-B (21).
Effect of NO and PKA Activation or Inhibition on Sp1 DNA Binding-As reported previously (26), nuclear extract from PMA-differentiated U937 cells formed two specific complexes (C 1 and C 2 ) with the hot Sp1 binding site of the human TNF-␣ promoter (Fig. 3A). Anti-Sp1 antibody supershifted only the highest order EMSA complex (C 1 ) demonstrating that it contained Sp1 or a closely related protein (Fig. 3A, lane 4). Incubation with either SNAP or H89 similarly decreased formation of the C 1 complex compared with nuclear extract from control cells (Fig. 3B). Conversely, Bt 2 cAMP increased the binding interaction at C 1 . Furthermore, in the presence of Bt 2 cAMP, SNAP had no effect on formation of the C 1 complex (Fig. 3B,  lane 5).
Effect of Sp1 Site-directed Mutation on NO Responsiveness of the TNF-␣ Promoter-If NO and H89 up-regulate the TNF-␣ promoter by reducing Sp1 binding, then the proximal Sp1 site would be functioning as a repressor element. However, deleting the 8-bp Sp1 site and replacing it with a 6-bp EcoRV sequence had decreased overall TNF-␣ promoter activity. To test the possibility that this decrease in promoter activity was created by stearic hindrance between flanking AP1 and AP2 motifs, we mutated the Sp1 site to maintain proper spacing (see "Exper- imental Procedures"). Again, activity of the wild type TNF-␣ promoter (Fig. 4) was up-regulated by NO donors (101 Ϯ 6% for SNAP and 92 Ϯ 5% for SNOG; p Ͻ 0.05) and by H89 (84 Ϯ 14%; p Ͻ 0.05), and down-regulated by Bt 2 cAMP (-33 Ϯ 4%; p Ͻ 0.05). Furthermore, in the presence of Bt 2 cAMP, NO had no effect (SNAP plus Bt 2 cAMP versus Bt 2 cAMP alone, p ϭ not significant). In contrast, the Sp1 mutant construct was unresponsive to NO donors and H89 (Fig. 4, p ϭ not significant). As seen before, Bt 2 cAMP still decreased promoter activity (Fig. 4, p Ͻ 0.05), suggesting that high intracellular concentrations of cAMP can affect additional promoter sites. Notably, mutation of the Sp1 site to prevent Sp1 binding increased activity of the new construct by 43 Ϯ 7% compared with wild type TNF-␣ promoter (Fig. 4, p Ͻ 0.05). Therefore, loss of promoter activity in the previous Sp1 deletion experiment was probably caused by creation of stearic hindrance between flanking regions. Collectively, the results suggest that NO derepresses TNF-␣ promoter activity by decreasing nuclear factor binding to its proximal Sp1 site.
Functional Analysis of Sp1 Binding Sites in a Heterologous Promoter-To further explore the ability of the Sp1 binding site to function as a NO response element, we examined whether it could confer NO inducibility to a heterologous, minimal CMV promoter that lacks enhancer sequences. An enhancerless, minimal CMV promoter was unresponsive to SNAP, Bt 2 cAMP, or both (Fig. 5, p ϭ not significant). However, when three copies of the Sp1 binding site were inserted into this promoter, NO induced promoter activity by 115 Ϯ 38% (Fig. 5, p Ͻ 0.05). Although Bt 2 cAMP tended to reduce activity of the promoter construct containing the inserted Sp1 sites, this effect did not reach statistical significance. However, the presence of Bt 2 cAMP prevented the enhancing effect of NO (SNAP plus Bt 2 cAMP versus Bt 2 cAMP alone; Fig. 5, p ϭ not significant). Furthermore, insertion of mutated Sp1 sites failed to confer NO responsiveness. Increases in base-line promoter activity caused  by insertion of both Sp1 and mutated Sp1 binding sites was primarily due to inclusion of a spacer sequence between the second and third Sp1 regions (see "Experimental Procedures") that was later found to unintentionally bind AP1 (data not shown). These results indicate that the addition of Sp1 binding sites into a heterologous promoter can convert it from NO unresponsive to responsive. Collectively, our findings demonstrate that the 8 bp Sp1 sequence (5Ј-CCCCGCCC-3Ј) can function as an essential part of a NO response element in the TNF-␣ promoter.
Modulation of promoter activity by NO has been reported to occur by either cGMP-dependent or -independent mechanisms and can result in either up-regulation or down-regulation of gene transcription (5-7, 27, 28). Work in this area has focused on relatively large promoter regions that appear to be involved in these NO mediated responses (6 -7, 27). Other studies have identified transcription factors including NF-B (3, 28), AP1 (29), heat shock factor 1 (30), and Oct-1 (31) that undergo increased or decreased DNA binding upon exposure of cells or cell extract to exogenous NO. Taken together, these results suggested that NO signaling cascades extend to the level of gene transcription through multiple promoter regulatory sites. A putative NO-responsive promoter element had not been described.
In the present study, we demonstrate that NO increases the activity of the TNF-␣ promoter and identify its Sp1 binding site at position Ϫ52 to Ϫ45 as the target for this cAMP-dependent effect of NO. Deletion of the Sp1, but not the CRE promoter site, resulted in the loss of NO responsiveness. Notably, NO decreased the binding of Sp1 to its promoter site, an effect that was mimicked by a PKA inhibitor and blocked by a PKA activator. Furthermore, mutation of only two base pairs within the Sp1 binding site rendered the TNF-␣ promoter completely unresponsive to both NO donors and a PKA inhibitor. Finally and perhaps most importantly, the insertion of Sp1 sites into a heterologous CMV promoter caused it to become NO responsive.
Recently, Rohlff et al. (13) demonstrated in HL-60 leukemia cells that Sp1 DNA binding could be enhanced by PKA-dependent phosphorylation. This important observation is consistent with our findings that NO, which can reduce PKA activity by decreasing intracellular cAMP concentrations (9) and H89, a direct PKA inhibitor, both decrease Sp1 binding to the TNF-␣ promoter. The association of decreases in Sp1 binding with promoter activation at first seemed paradoxical. However, other investigators have shown that, in contrast to its accepted role as a homeostatic transactivator (14,15), Sp1 binding can sometimes function as a gene repressor (16,17). Our results demonstrate that the proximal Sp1 binding site in the TNF-␣ promoter can function as a repressor element.
The ability of NO to activate the TNF-␣ promoter was blocked by the presence of Bt 2 cAMP, a PKA agonist, and by deletion or mutation of the Sp1 binding site. Presumably, the addition of exogenous cAMP analog kept PKA activated, despite the cAMP lowering effects of NO, and thereby maintained Sp1 in a phosphorylated and bound state (13). It is important to note, however, that NO has also been shown to decrease Sp1 binding activity by a mechanism independent of PKA inactivation. At NO donor concentrations higher than those used in our experiments, NO can disrupt the DNA binding of Sp1, a zinc finger protein, by releasing zinc from thiol groups (32). NO up-regulation of TNF-␣ may also involve effects on the binding of transcription factors to Sp1 flanking sequences which include AP1 and AP2 sites. This latter scenario would suggest that Sp1 is only part of a larger NO response complex.
Transcriptional regulation of TNF-␣ through the Sp1 binding site of its promoter may represent an important mechanism by which NO regulates inflammatory responses. Furthermore, Sp1 binding sites exist in many promoters (11-17, 23, 26). Thus, this Sp1-based regulatory system could account for the effects of NO on the expression of other gene products.