p47 phox Participates in Activation of RelA in Endothelial Cells*

Activation of endothelial cell NF- (cid:1) B by interleukin (IL)-1 constitutes an event critical to the progression of the innate immune response. In this context, oxidants have been associated with NF- (cid:1) B activation, although the molecular source and mechanism of targeting have remained obscure. We found that RelA, essential for NF- (cid:1) B activation by IL-1, was associated with the NADPH oxidase adapter protein p47 phox in yeast two-hybrid, coprecipitation, and in vitro binding studies. RelA and p47-GFP also colocalized in endothelial cells in focal submembranous dorsoventral protrusions. Overexpression of p47 phox synergized with IL-1 (cid:2) in the activation of an artificial (cid:1) B-luciferase reporter and specifically augmented IL-1 (cid:2) -induced RelA transactivation activity. p47 phox overexpression also greatly increased IL-1 (cid:2) -stimulated RelA phosphorylation, whereas it had no effect on I- (cid:1) B degradation or on RelA nuclear translocation or (cid:1) B binding. The tandem SH3 domains of p47 phox were found to associate with a proline-rich mid-region of RelA (RelA-PR) located between the Rel homology and transactivation and both interaction reporters (histidine and adenine), in addition to lacZ expression (filter lift assay). Positive col- onies were restreaked, and single clones were retested. Candidate library plasmids were passaged through Escherichia coli , stably trans- formed into AH109, tested for autonomous transactivation, and mated to Y187 yeast harboring the bait vector, and the diploids were retested for auxotrophy and lacZ expression. Interactions between RelA and different p47 phox domains were tested by mating Gal4-AD-RelA-trans- formed AH109 with Y187 expressing Gal4-BD fusions with various p47 phox truncations and assessing quantitative (cid:2) -galactosidase expres- sion. Briefly, the diploids were selected with leucine- and tryptophan-deficient media and replated under selection. Overnight liquid cultures of diploid colonies were diluted and grown to mid-log phase in selection medium, adjusted to identical A 600 readings, then washed twice, and resuspended in 0.3 ml of Z buffer (60 m M Na 2 HPO 4 , 40 m M NaH 2 PO 4 , pH 7.0, 10 m M KCl, and 1 m M MgSO 4 ). 0.1 ml of cell suspension was then freeze-thawed three times using liquid nitrogen and vortexed with glass beads, and the supernatant was added to 0.7 ml of Z buffer with 0.27% (cid:2) -mercaptoethanol. Color was developed using 0.16 ml of o nitrophenyl (cid:2) - D -galactopyranoside (4 mg/ml in Z buffer) and stopped with 0.4 ml of 1 M Na 2 CO 3 . The tubes were centrifuged at 14,000 (cid:3) g , and the A 420 was measured. Negative controls included Gal4-BD or Gal4-AD without fusions.

Activation of vascular endothelial NF-B by IL-1␤ 1 during the innate immune response drives the production of a number of proteins, such as tumor necrosis factor, intercellular adhesion molecule 1, and E-selectin, necessary for the full expression of acute inflammation. A linear activation sequence following engagement of the type I IL-1 receptor (IL-1RI) has been intensely studied and involves binding of the MyD88 and Tollip adapters to the IL-1 receptor chains (1,2) with subsequent recruitment of IL-1 receptor-associated kinase (1,3). IL-1 receptor-associated kinase appears to leave the receptor complex to bind TRAF6, which associates with the mitogen-activated protein 3-kinase TAK1 through the adapter TAB2 (4,5). TAK1 in turn activates NF-B-inducing kinase, which acts upstream of the I-B kinase complex (6). This latter complex phosphorylates I-B (7)(8)(9)(10)(11), marking it for ubiquitination and degradation, with consequent unmasking of the nuclear localization sequence of Rel family members, nuclear translocation of the active NF-B complex, and transactivation of NF-B-responsive genes.
Despite articulation of this universal pathway, a number of details concerning NF-B activation remain unresolved. In particular, numerous studies suggest an important role for endogenous oxidants in the activation of NF-B by IL-1. IL-1 increases oxidant production, and strategies designed to diminish oxidant production or augment oxidant scavenging cause decreased IL-1-induced NF-B activity (12)(13)(14). However, exogenous H 2 O 2 does not replicate the time course of cytokine-induced NF-B translocation (15), and the involvement of oxidants in IL-1 signaling may be cell-specific (15,16). In addition, the mechanism of endogenous oxidant involvement is unclear, because none of the signaling elements listed above present obvious redox-responsive switches. Further, the source of oxidants and the mechanism by which oxidants target the NF-B pathway have not been identified.
Human endothelial cells express a functional NADPH oxidase that appears to be essential for transmitting signals initiated by mechanical shear, human immunodeficiency virus type 1 Tat, tumor necrosis factor, phorbol ester, and growth factors (17)(18)(19)(20)(21). Because of the diverse phenotypic responses invoked by these various stimuli, we recently suggested that one basis for signal specificity of reactive oxidants may be through spatial site direction of the NADPH oxidase to relevant signaling complexes. As an example, a two-hybrid screen using the full-length oxidase adapter p47 phox yielded the signaling protein TRAF4 as a functional binding partner (22). In this latter screen, TRAF4 was the only cDNA recovered, perhaps because of the intramolecular masking of the core binding domains of p47 in the intact molecule. In the present study, we used a truncated p47 phox lacking a purported autoinhibitory C terminus as a screening bait and recovered the NF-B family member p65/RelA. Further evidence suggests the participation of p47 phox in RelA phosphorylation and transactivation but not I-B degradation or translocation of RelA into the nucleus.

EXPERIMENTAL PROCEDURES
Plasmid Construction-All of the PCR amplifications for subcloning or mutagenesis were performed with either Pfu Turbo (Stratagene) or Pfx Platinum (Invitrogen). All of the new constructs were confirmed by sequencing. pGBKT7-p47, pGBKT7-p47(205-390), pGBKT7-p47(1-205), pGBKT7-p47(1-298), pCINF-p47, pGEX-p47, and p47-GFP were constructed as previously described (18,22). pGBKT7-p47(153-286), pGBKT7-p47(153-219), pGBKT7-p47(223-286), and pGBKT7-p47(153-219;R193) were constructed by PCR amplification of the corresponding domains between EcoRI and SalI sites with the addition of stop codons and ligation into the Gal4-BD shuttle vector pGBKT7 (Clontech). pGBKT7-p47(4 -125) was derived by PCR amplification between NcoI and EcoRI sites and ligation into pGBKT7. pGEX-p47(1-298) was made by PCR amplification of p47  between EcoRI sites with addition of a stop codon and ligation into pGEX-2TK. Full-length RelA was amplified between XhoI and XbaI sites from a HUVEC library and ligated directly into pCI-neo (Promega) to create pCIN-RelA. pNF-B-luciferase was obtained from Clontech and consisted of four sequential B consensus sites (GGGAATTTCC) upstream of the herpes simplex virus-TK TATA-like promoter. pRL-TK and pFR-luc were obtained from Promega and Stratagene, respectively. A plasmid expressing Gal4-RelA was the kind gift of Dr. M. Lienhard Schmitz. All of the RelA truncations were derived by PCR amplification of RelA segments between XhoI and XbaI sites with insertion of stop codons, cloning into pCR4-TOPO blunt (Invitrogen), and then subcloning into pCI-neo. pGEX-RelA-PR was constructed by amplification of RelA (301-431) between BamHI and EcoRI sites and ligation into pGEX-2TK. pELAMluc was previously described (23). p67 phox cDNA was obtained from the American Type Culture Collection, and a V204A mutation was introduced with PCR mutagenesis. The mutant was then subcloned into the XbaI and KpnI sites of pShuttle (Clontech) to create pSh-p67(V204A). Adenoviruses harboring wild type p47 phox and lacZ were previously described (22), and a similar technique was used to generate Ad-p47(W193R). Equivalent expression of wild type p47 phox and p47(W193R) was demonstrated by Western blot at an equal multiplicity of infection (not shown). lacZ transgene was expressed in Ͼ95% of HUVEC.
Yeast Two-hybrid Screening and Mating Analysis-A human endothelial phage Gal4-AD library was constructed as previously reported (22) and dropped out in the shuttle vector pAD-GAL4 -2.1. Saccharomyces cerevisiae AH109, which harbor two auxotrophic reporter genes (ADE2 and HIS3) and lacZ, were stably transformed with pGBKT7-p47(1-298), which was found to lack autonomous transactivation activity. Transformants were selected and secondarily transformed with the endothelial library using lithium acetate. A high stringency screen was performed, requiring auxotrophic selection for bait vector (tryptophan), library vector (leucine), and both interaction reporters (histidine and adenine), in addition to lacZ expression (filter lift assay). Positive colonies were restreaked, and single clones were retested. Candidate library plasmids were passaged through Escherichia coli, stably transformed into AH109, tested for autonomous transactivation, and mated to Y187 yeast harboring the bait vector, and the diploids were retested for auxotrophy and lacZ expression. Interactions between RelA and different p47 phox domains were tested by mating Gal4-AD-RelA-transformed AH109 with Y187 expressing Gal4-BD fusions with various p47 phox truncations and assessing quantitative ␤-galactosidase expression. Briefly, the diploids were selected with leucine-and tryptophandeficient media and replated under selection. Overnight liquid cultures of diploid colonies were diluted and grown to mid-log phase in selection medium, adjusted to identical A 600 readings, then washed twice, and resuspended in 0.3 ml of Z buffer (60 mM Na 2 HPO 4 , 40 mM NaH 2 PO 4 , pH 7.0, 10 mM KCl, and 1 mM MgSO 4 ). 0.1 ml of cell suspension was then freeze-thawed three times using liquid nitrogen and vortexed with glass beads, and the supernatant was added to 0.7 ml of Z buffer with 0.27% ␤-mercaptoethanol. Color was developed using 0.16 ml of onitrophenyl ␤-D-galactopyranoside (4 mg/ml in Z buffer) and stopped with 0.4 ml of 1 M Na 2 CO 3 . The tubes were centrifuged at 14,000 ϫ g, and the A 420 was measured. Negative controls included Gal4-BD or Gal4-AD without fusions.
GST Pull-down-BL21-RP E. coli were transformed with pGEX, pGEX-p47, or pGEX-p47(1-298) and induced at 37°C for 3 h with isopropyl-␤-D-thiogalactopyranoside. The cells were lysed, and GST fusions were captured on glutathione-Sepharose (Amersham Biosciences). Fusion protein concentrations were estimated by Coomassie Blue PAGE. Full-length or truncated RelA was transcribed using the T7 promoter of pCI-neo and translated in vitro in the presence of [ 35 S]methionine using a kit (TNT quick coupled; Promega). Translated products were assessed on a gel using autoradiography and normalized prior to addition to the binding reaction. Binding, washing, and analysis were performed as described (22). Similar studies were performed in the presence of RelA-PR peptide by bacterial expression of GST-RelA (301-431), purification on GSH-Sepharose, and release of the peptide with thrombin cleavage.
Microscopy-HUVEC were electroporated with p47-GFP with or without RelA-PR and then grown on fibronectin-coated slides. Following IL-1␤ stimulation, cells were washed twice with PBS, fixed with 3% formaldehyde in PBS, washed twice, and permeabilized in 50 mM NaCl, 3 mM MgCl 2 , 200 mM sucrose, 10 mM HEPES, pH 7.4, and 0.5% Triton X-100 for 5 min. The cells were blocked with 3% bovine serum albumin in PBS and incubated with anti-RelA (Santa Cruz) at 1:200 in 1% bovine serum albumin at room temperature for 1 h. After three washes in 1% PBS, the cells were incubated with rhodamine-conjugated secondary antibody (1:400) at room temperature for 1 h. After two washes in 1% bovine serum albumin and two washes in PBS, the cells were mounted with ProLong antifade (Molecular Probes), and the images acquired with a Zeiss Axiovert S100TV LSM 410 laser scanning system.
NF-B and RelA Activation-For assessment of NF-B activation, HUVEC were first infected with the appropriate adenovirus for 1 h at a multiplicity of infection of 1:100, recovered overnight and then synchronized at the G 1 -S transition with thymidine. Six hours after thymidine release, the cells were electroporated with pNF-B-luciferase and pRL-TK. 24 h later, the cells were stimulated with 2 ng/ml IL-1␤. The cells were lysed 6 h later, and both firefly and Renilla luciferase were assessed using a dual luciferase kit (Promega). All of the signals were normalized using the Renilla transfection control, which did not significantly change with transgene expression or IL-1␤ stimulation. Activation of RelA was assessed by cotransfection of HUVEC with the Gal4-BD-RelA fusion, the Gal4-luciferase reporter pFR-luc, and pRL-TK. Transactivation of the Gal4 minimal promoter was assessed 6 h after IL-1␤ addition and was normalized to Renilla luciferase. NF-B activity was also assessed by transactivation of the E-selectin promoter. HUVEC were transfected after cell cycle synchronization using Fu-GENE 6 (Roche Applied Science) with pELAM-luc and pRL-TK. Either pCIN-RelA (301-341), pSh-p67(V204A), or pCIneo were cotransfected. After 24 h, the cells were stimulated with IL-1␤ (2 ng/ml) for 6 h, and the luciferase activity was measured. In some studies, HUVEC were treated with diphenylene iodonium (DPI; 10 M) or MnTBAP (100 M) for 30 min prior to IL-␤1 stimulation.
Electrophoretic Mobility Shift Assay-HUVEC were infected with respective adenoviruses for 1 h as above and the following day were stimulated with IL-1␤ for 30 min prior to lysis in cold lysis buffer (10 mM HEPES, pH 7.8, 10 mM KCl, 0.1 mM EDTA, 2 mM dithiothreitol, 1 mM phenylmethylsulfonyl fluoride, 10 g/ml leupeptin, and 10 g/ml pepstatin A) on ice for 15 min. The cells were disrupted in 0.6% Nonidet P-40 with vortexing, and the nuclei were recovered by centrifugation and extracted at 4°C for 30 min 10% glycerol, 20 mM HEPES, pH 7.8, 420 mM NaCl, 5 mM EDTA, 5 mM dithiothreitol, and 1 mM phenylmethylsulfonyl fluoride. Equal protein equivalents were incubated with 32 Plabeled oligonucleotides from the B3 site of the tumor necrosis factor promoter, as previously described (24). For supershift assays, the nuclear extracts were incubated with respective antibodies for 30 min prior to addition of the labeled probe. All of the supershift antibodies were from Santa Cruz. The samples were fractionated on a 4% acrylamide gel prior to autoradiography.
RelA Translocation and I-B Degradation-After infection with adenoviruses, HUVEC were stimulated with IL-1␤ (2 ng/ml) and lysed at various time intervals. Equal protein equivalents of lysate were loaded onto 12% PAGE gels and immunoblotted for I-B␣ to assess degradation. For RelA translocation, HUVEC nuclei were isolated and extracted as above, and equal protein equivalents were immunoblotted with anti-RelA (Santa Cruz).
RelA Phosphorylation-After infection with adenoviruses, HUVEC were serum-starved overnight and then incubated in phosphate-free Dulbecco's modified Eagle's medium for 30 min. [ 32 P]Orthophosphate (100 Ci/ml) and 1% fetal calf serum were then added, and the cells were incubated for another 4 h. After stimulation with IL-1␤ (2 ng/ml) for 10 min, the cells were lysed on ice for 5 min in lysis buffer containing 0.1% SDS and 0.5% deoxycholate. The cells were sheared through a 23-gauge needle 10 times and pelleted at 13,000 ϫ g for 15 min. After preclearing and normalization for protein, RelA was immunoprecipitated and washed four times in lysis buffer with SDS and deoxycholate, twice in lysis buffer containing 0.5 M NaCl, and once in 50 mM Tris, pH 7.0. One third of the immunoprecipitate was immunoblotted for RelA, and the remainder was analyzed by autoradiography.

RESULTS
RelA Binds to the Tandem SH3 Domains of p47 phox -To relieve autoinhibitory folding of the p47 phox bait protein, we deleted the C-terminal 92 residues containing two proposed proline-rich motifs and all of the serines known to be phosphorylated in p47 phox of stimulated phagocytes (25,26). This truncation appears to be equivalent to a phosphorylated full-length p47 phox in opening core binding domains (26,27). Upon screening the HUVEC GAL4-AD library, three clones were identified that encoded RelA in frame with the GAL4-AD. All three clones encoded residues 170 -551, encompassing part of the N-terminal Rel homology domain, a mid-protein region of unknown function, and the C-terminal transactivation domains. Using a yeast mating technique, binding of RelA to p47(1-298) was confirmed (Fig. 1a). In contrast, RelA did not bind to the fulllength p47 phox protein, consistent with masking of the RelA binding site by the C terminus of p47 phox . Deletion analysis suggested binding of RelA to the tandem SH3 domains of p47 phox (153-286) but not to the N-terminal PX domain (4 -125) or C-terminal tail containing proline-and arginine-rich domains. Although each SH3 domain bound RelA in isolation, SH3a (153-219) bound its target with 2.3-fold greater avidity compared with SH3b (223-286) as measured by lacZ expression. The binding of SH3a to RelA appeared to depend on its SH3 surface and not some unrelated colocalized motif, because destruction of the tryptophan bridge (p47(153-219;R193)) (18,28,29) abolished binding activity. Unexpectedly, SH3a when accompanied by the N-terminal half of the protein (1-205), and SH3b, when accompanied by the C terminus (205-390), did not bind RelA. Although it is thought that full-length p47 phox in its native configuration involves intramolecular interactions be- FIG. 1. RelA associates with p47 phox . a, p47 phox and various truncations were fused to the Gal4-BD and transformed into Y187 yeast. Interaction with RelA was assessed by mating to AH109 yeast harboring Gal4-AD-RelA and quantification of ␤-galactosidase in solution. p47(1-298) but not full-length p47 phox interacted with RelA. The tandem SH3 domains of p47 phox were necessary and sufficient for this interaction, with the greatest lacZ expression attained with SH3ab, then SH3a, and then SH3b. SH3a containing a W198R mutation displayed minimal interaction. The values are the means Ϯ S.E. of four experiments. b, GST pull-down demonstrating that [ 35 S]methioninelabeled RelA directly bound GST-p47(1-298) but not GST-p47(1-390) (full-length) or GST alone. A more quickly migrating band is likely partially translated or proteolyzed RelA. c, Fx cells were transfected with empty vector or FLAG-p47. Endogenous RelA coprecipitated with FLAG-p47 using anti-FLAG antibodies in transfected but not untransfected cells and did not coprecipitate using isotype control antibodies in FLAG-p47-transfected cells. d, at base line or following stimulation with IL-1␤, HUVEC lysates were immunoprecipitated (IP) with either irrelevant or anti-RelA antibodies and immunoblotted (IB) for p47 phox . Endogenous p47 phox coprecipitated with endogenous RelA in both IL-1␤-stimulated and unstimulated HUVEC. tween SH3a and the C-terminal tail (28) as well as SH3b and the N-terminal PX domain (30), it is likely that binding function in these truncated derivatives is lost because of intramolecular masking of SH3a by the PX domain and SH3b by the C terminus. This would suggest that the binding domains of p47 phox are not entirely functionally independent and may adopt alternate tertiary structures based on the specific truncation.
RelA also bound GST-p47(1-298) in vitro (Fig. 1b), suggesting a direct binding between proteins. The smaller, ϳ54 -58-kDa protein seen in this figure is consistent and specifically binds GST-p47(1-298), likely representing a partially translated or proteolyzed RelA fragment. Lack of RelA binding to the full-length GST-p47(1-390) is consistent with the yeast mating data. In whole cells, endogenous RelA specifically coprecipitated with FLAG-tagged full-length p47 phox (Fig. 1c), suggesting that in mammalian cells the RelA-binding site of p47 phox is at least partially exposed. In support of this interpretation, endogenous p47 phox coprecipitated with endogenous RelA in unstimulated HUVEC, suggesting a preformed complex (Fig.  1d). Binding did not quantitatively increase following IL-1␤ stimulation, suggesting a constitutive nature for this interaction in endothelial cells.
p47 phox Overexpression Increases IL-1␤-induced RelA Activation-Transactivation of an NF-B consensus-luciferase reporter by IL-1␤ was consistently low (122% of control) in Ad-lacZ-infected HUVEC (Fig. 2, a and b). This did not appear to be due to spurious effects of adenoviral infection, interference with the normalizing Renilla reporter, passage number, transfection technique, or low specific activity of IL-1␤ (data not shown) and may suggest that transactivation of this artificial promoter construct in HUVEC by IL-1␤ is fundamentally inefficient. Indeed, in a recent report, fluorocytometric enrichment of high responding HUVEC clones stably transfected with an NF-B reporter was required to demonstrate response to IL-1␤ (31). Despite this low efficiency, overexpression of p47 phox , while having minimal effect alone, increased IL-1␤-stimulated NF-B-luciferase activity 2-fold above control (Fig. 2b). Overexpression of the oxidase-inactive mutant p47(R193), which also would not be expected to bind RelA, had no effect on IL-1␤-induced NF-B-luciferase activity. A similar pattern was observed for the transactivating activity of RelA specifically. The minimal induction of Gal4-RelA by IL-1␤ was potentiated by p47 phox overexpression (Fig. 2c).
In contrast to activation of RelA, its nuclear translocation and DNA binding did not appear to be affected by p47 phox overexpression. IL-1␤ increased NF-B DNA binding in Ad-lacZ-infected HUVEC, and overexpression of p47 phox did not alter either basal or IL-1␤-stimulated gel shift patterns (Fig.  3a). The nucleoproteins in both Ad-lacZ (not shown) and Ad-p47-infected HUVEC stimulated with IL-1␤ (Fig. 3b) contained both RelA and p50 (NF-B1). As expected, immunoblots confirmed translocation of RelA within 10 min of IL-1␤ stimulation, and this translocation was unaffected by p47 phox overexpression (Fig. 3b). Similarly, p47 phox overexpression did not raise basal RelA levels in either cytosolic (not shown) or nuclear fractions. Further, degradation of I-B␣ was demonstrated within 15-30 min after IL-1␤ stimulation, with partial recovery by 60 min, and this pattern was again unchanged by p47 phox overexpression (Fig. 3c). However, p47 phox greatly affected RelA phosphorylation (Fig. 3d). In Ad-lacZ-infected HUVEC, IL-1␤ treatment increased RelA phosphorylation 1.9fold, whereas in Ad-p47-infected cells, IL-1␤ increased RelA phosphorylation 8.7-fold (averages of three experiments).
Although the most N-proximal of these (DPRPPPR) varies from the consensus by inclusion of an acidic rather than hydrophobic first residue, this motif is well conserved in mouse (EPRPPTR) and chick (EPRPPRR) RelA proteins. In contrast, three of the remaining four polyproline motifs are poorly preserved in other species, with the remaining motif (APGPPQA) being represented in mouse (TPGPPQS) but not chick RelA.
As further evidence that p47 phox binds RelA-PR, we found that the RelA-PR peptide blocked binding of full-length RelA to GST-p47(1-298) in vitro (Fig. 4c). In HUVEC, endogenous RelA also colocalized strongly with p47-GFP in focal peripheral structures in unstimulated (not shown) and IL-1␤-stimulated (Fig. 5, a-c)  the RelA-PR truncation greatly diminished physical colocalization of endogenous RelA within these discreet p47-GFP collections (Fig. 5, d-f).
To further investigate the potential functional importance of an interaction between endogenous p47 phox and RelA, we attempted to decrease IL-1␤-induced activation of endogenous NF-B through ectopic expression of RelA-PR, which we are not aware of as having other binding partners. Because IL-1␤ only weakly activated the consensus NF-B-luciferase reporter, we instead used the native E-selectin promoter, relevant to endothelial cells, to produce a more robust IL-1␤-induced signal. This promoter harbors three NF-B sites, at least two of which cooperate functionally and are required for activity (33). As expected for this native endothelial-specific promoter, IL-1␤ increased ELAM-luciferase activity in HUVEC and in parallel caused a low level of oxidant production (Fig.  6a). Expression of the RelA-PR region had minimal effects on oxidant production but completely blocked IL-1␤-induced Eselectin promoter activation. To further implicate p47 phox , we utilized a separate subunit of the NADPH oxidase, p67 phox harboring a mutation in the purported activation domain (p67(V204A)) (34). This latter mutant has been shown to act as  4. p47 phox binds RelA (301-431). a, GST pull-down shows direct binding of full-length RelA, RelA (301-431), and RelA (170 -431) to GST-p47(1-298). RelA did not bind to GST alone. b, the p47-binding region of RelA (within residues 301-341) comprises a proline-rich segment containing five polyproline sites that are candidate type II SH3binding sequences. The invariant prolines of the PXXP motif are boxed, and the residue numbers are shown. c, GST pull-down of full-length RelA by GST-p47(1-298). Purified RelA-PR decreased direct binding of RelA to GST-p47(1-298). a transdominant inhibitor of the NADPH oxidase in vitro and in recombinant COS-phox cells (34,35). As shown in Fig. 6a, coexpression of p67(V204A) also significantly decreased IL-1␤induced E-selectin promoter activation as well as oxidant production. As further evidence that p47 phox acts on NF-B through its participation in the NADPH oxidase, we found that the oxidase inhibitor DPI and the membrane-permeant superoxide dismutase mimetic MnTBAP also blocked IL-1␤-induced E-selectin promoter transactivation (Fig. 6b). DPI also blocked IL-1␤-induced phosphorylation of endogenous RelA, consistent with the effects of p47 phox -dependent oxidants on RelA phosphorylation (Fig. 6c).
IL-1␤ Increases Tyrosine Phosphorylation of IL-1RI-associated Proteins-Because oxidants are known to increase tyrosine phosphorylation events, we searched for an increase in such events in the proximal IL-1 signaling pathway. Several proteins were shown to coprecipitate with the IL-1RI, including a ϳ98-kDa protein that was tyrosine phosphorylated within 2 min of IL-1␤ stimulation (Fig. 7a). In addition, ϳ139-, 106-, 45-, 42-, and 39-kDa proteins displayed increased tyrosine phosphorylation 3-5 min after IL-1␤ stimulation. Expression of p67(V204A) decreased IL-1␤-dependent tyrosine phosphorylation of these proteins (Fig. 7b), consistent with involvement of the NADPH oxidase in IL-1␤-induced protein tyrosine phosphorylation. In addition, the RelA-PR peptide also decreased tyrosine phosphorylation of these proteins (Fig. 7c), suggesting that localization of the oxidase to a RelA complex is necessary for the targeting of oxidants to IL-1 receptor-linked proteins. DISCUSSION Characteristic of most adapter proteins, p47 phox displays multiple binding surfaces that include an N-terminal PX domain, tandem SH3 domains, a variant proline-rich remnant, a basic region, and a C-terminal proline-rich motif. A wealth of studies have demonstrated the importance of most of these domains in self-association or in binding to other oxidase subunits, whereas few studies have documented binding to non-oxidase moieties. Here, we demonstrate binding of the tandem SH3 domains of p47 phox to RelA. Both yeast mating and GST pull-down experiments suggested masking of the RelA-binding domain by the C terminus of p47 phox , consistent with present models of intramolecular folding of this protein (26 -28, 36). The precise mechanism of interaction, however, at this point remains unclear. Physiologic binding was also demonstrated by coprecipitation and colocalization studies in intact endothelial cells, suggesting at least partial unmasking of the RelA binding surface in vivo, even under unstimulated conditions. Thus, the physical state of p47 phox may differ in resting endothelial cells as compared with its state in vitro or in resting phagocytes. In the latter cell type, p47 phox is thought to exist in an unphosphorylated, folded state. In endothelial cells, however, we have previously noted quantitative association of endogenous p47 phox with the cytoskeleton (18), suggesting a binding function not present in resting neutrophils, and have further noted phosphorylation of p47 phox in unstimulated HUVEC. 2 In further support of an open form of p47 phox , a recent study demonstrated coprecipitation of this protein with other NADPH oxidase components, including p22 phox , in unstimulated endothelial cells (37).
Colocalization of the two proteins in endothelial cells was not diffuse but rather concentrated at peripheral dorsolateral protrusions. Importantly, p47 phox localizes to the cortical cytoskeleton of endothelial cells and is highly concentrated in edge ruffles (18,22). RelA similarly has been noted to associate with actin structures (38), and disruption of actin polymerization blocks activation of NF-B by phorbol ester (39). Thus, association of p47 phox with RelA in such focal areas may reflect the ability of the cytoskeleton to function as a dynamic scaffold for signaling complexes.
Overexpression of p47 phox in HUVEC augmented the otherwise weak activation of a B consensus promoter by IL-1␤, suggesting a functional significance for this association. Several observations suggest that this enhancement of NF-B activity was not a nonspecific effect of p47 phox overexpression. First, the anticipated nonspecific effect of p47 phox binding would be to sequester RelA or sterically hinder binding of transcription complexes and therefore decrease, rather than increase, NF-B activation. Second, p47-related enhancement of NF-B was specific for IL-1␤ and was not seen in unstimulated cells or cells stimulated with tumor necrosis factor (not shown). Third, p47 phox overexpression specifically increased IL-1␤-induced RelA phosphorylation and activation but not nuclear importation or I-B␣ degradation. Thus, p47 phox binding also did not appear to simply displace I-B␣ from RelA, consistent with the physically separate binding regions of RelA for I-B␣ and p47 phox .
The selective effect of p47 phox overexpression on RelA phosphorylation and activation but not translocation is consistent with a recent series of studies suggesting that NF-B activation requires signaling through both the well described pathway leading to I-B␣ degradation and translocation of NF-B elements into the nucleus and, in addition, a distinct pathway resulting in RelA phosphorylation and activation. For instance, phosphorylation of RelA on Ser 529 of its C-terminal transactivation domain increases its transactivation activity but not its nuclear translocation (40), and genetic ablation of either TRAF2-associated kinase (T2K) or GSK-3␤ decreases IL-1-induced NF-B activation but not p65 translocation or I-B degradation in vivo (41,42). Of note, both IL-1␤ and H 2 O 2 increase RelA phosphorylation (43,44). Activation of RelA appears to involve both phosphatidylinositol 3-kinase and Akt, and interventions that block these two kinases selectively decrease RelA phosphorylation and activation but not translocation (45)(46)(47)(48). Importantly, Rac1, another component of the NADPH oxidase, also appears necessary for IL-1␤-induced RelA activation but not translocation (49,50).
As further evidence that p47 phox participates in endogenous NF-B activation, we decreased IL-1␤-induced transactivation of the highly NF-B-dependent E-selectin promoter through overexpression of the RelA p47-binding domain. This experiment did not involve overexpression of p47 phox , again suggesting specificity of the preceding studies. This latter domain (RelA-PR) appears to be a proline-rich region with no other presently defined function. Therefore, RelA-PR overexpression would not be expected to antagonize homo-or heterodimerization through the N-terminal Rel homology domain nor to interfere with the transactivation function of the C-terminal transactivation domains. Indeed, this RelA mid-region in isolation does not possess basal RelA transactivation squelching activity (32).
Interference with the NADPH oxidase through expression of a trans-dominant interfering mutant of p67 phox , chemical inhibition of the oxidase, or scavenging of O 2 . decreased NF-B-dependent promoter transactivation by IL-1␤, inferring that p47 phox may act through its known function as a constituent of the NADPH oxidase. This result was unexpected, because the p47 phox SH3a domain, which appears to increase the avidity of p47 phox for RelA, is traditionally thought to act instead as a ligand for the C-terminal polyproline motif of p22 phox , thus promoting functional oxidase assembly (28,36). Several models may explain this apparent discrepancy. First, a recent study provides convincing evidence that critical assembly of the oxidase involves non-SH3 regions of p47 phox interacting with a cytosolic loop of p22 phox (residues 51-63) remote from its polyproline motif (51), allowing the possibility for p47-SH3a to interact simultaneously with another protein. Second, because p47-SH3b also binds RelA, it is possible that this binding site tethers RelA to the oxidase or that RelA is "traded" from SH3a to SH3b upon docking with the oxidase. The functional significance of p47-SH3b has not been clearly identified, because it appears to be relatively unimportant for O 2 . production in reconstituted cells (28). Alternatively, RelA may be released from p47 phox upon assembly of the oxidase. Third, it is possible that p47 phox forms homodimers or multimers within the signaling complex, thus binding several proteins at once. Lastly, we cannot exclude the possibility that p47-SH3a binds the two separate target proteins sequentially or that binding to RelA and binding to p22 phox represent unrelated functions of p47 phox . Although the specific targets of oxidants in the IL-1 signaling cascade were not identified in this study, protein-tyrosine phosphatases are known to be reversibly inactivated by O 2 . (52), raising the possibility that focal targeting of even low levels of oxidants may create a microenvironment that favors tyrosine phosphorylation of proteins in the proximal IL-1␤ pathway. Indeed, protein tyrosine phosphorylation events are necessary for NF-B activation but not nuclear translocation (53). Consistent with this hypothesis, we found that IL-1␤ rapidly increased tyrosine phosphorylation of a number of IL-1RI-associated proteins, and this phosphorylation was decreased by suppression of NADPH oxidase activity (p67(V204A)) or delocalization of the oxidase (RelA-PR). IL-1␤ has previously been shown to increase tyrosine phosphorylation of the IL-1RI itself in Saos2 cells (54), although we did not find this effect in endothelial cells. In this latter study, however, tyrosine phosphorylation was essential for phosphatidylinositol 3-kinase recruitment and activation by IL-1␤, suggesting a potential link between oxidants and the phosphatidylinositol 3-kinase/Akt/ RelA phosphorylation pathway. Taken together, our data suggest that the oxidase may be tethered in to a RelA-containing complex through interactions with p47 phox , to direct oxidant production to susceptible proteins acting upstream of RelA phosphorylation.