Regulation of NF-κB RelA Phosphorylation and Transcriptional Activity by p21 ras and Protein Kinase Cζ in Primary Endothelial Cells*

The activity of the transcription factor NF-κB is thought to be regulated mainly through cytoplasmic retention by IκB molecules. Here we present evidence of a second mechanism of regulation acting on NF-κB after release from IκB. In endothelial cells this mechanism involves phosphorylation of the RelA subunit of NF-κB through a pathway involving activation of protein kinase Cζ (PKCζ) and p21 ras . We show that transcriptional activity of RelA is dependent on phosphorylation of the N-terminal Rel homology domain but not the C-terminal transactivation domain. Inhibition of phosphorylation by dominant negative mutants of PKCζ or p21 ras results in loss of RelA transcriptional activity without interfering with DNA binding. Raf/MEK, small GTPases, phosphatidylinositol 3-kinase, and stress-activated protein kinase pathways are not involved in this mechanism of regulation.

The activity of the transcription factor NF-B is thought to be regulated mainly through cytoplasmic retention by IB molecules. Here we present evidence of a second mechanism of regulation acting on NF-B after release from IB. In endothelial cells this mechanism involves phosphorylation of the RelA subunit of NF-B through a pathway involving activation of protein kinase C (PKC) and p21 ras . We show that transcriptional activity of RelA is dependent on phosphorylation of the N-terminal Rel homology domain but not the C-terminal transactivation domain. Inhibition of phosphorylation by dominant negative mutants of PKC or p21 ras results in loss of RelA transcriptional activity without interfering with DNA binding. Raf/MEK, small GTPases, phosphatidylinositol 3-kinase, and stress-activated protein kinase pathways are not involved in this mechanism of regulation.
The NF-B/Rel family of dimeric transcription factors is involved in the immediate early transcription, i.e. independent of protein synthesis, of a large array of genes induced by mitogenic or pathogen-associated stimuli. In its active form, NF-B is a nuclear homo-or heterodimeric complex of a number of different Rel family members. The canonical and most abundant form of NF-B is composed of a 50-kDa (p50, or NFB1) and a 65-kDa (p65, or RelA) subunit. Both subunits can form homodimers as well as heterodimers with other members of the Rel family i.e. c-Rel (Rel), p52 (NFB2), and RelB (1). All members of the Rel family exhibit extensive sequence similarity in their N-terminal region referred to as the Rel homology domain (RHD) 1 responsible for DNA binding and formation of Rel dimers. Only RelA, Rel, and RelB carry a transcription activating domain, and thus only dimers containing one of these proteins activate the transcription of NF-B-dependent genes efficiently. With respect to transcription activation, the RelA subunit appears to have the highest activity.
In most unstimulated cells, NF-B is constitutively retained in the cytoplasm by inhibitory proteins of the IB family, namely IB␣, IB␤, IB␥, p100, p105, and IB⑀ (2). Formation of NF-B⅐IB complexes masks the nuclear localization signal sequence present in NF-B molecules and thus prevents their nuclear translocation. One of the key events in the activation of NF-B is the liberation of functional NF-B dimers from IB, which results in the translocation of NF-B to the nucleus. Cytoplasmic release of NF-B dimers involves site-specific phosphorylation of IB by kinases of the IB signalosome (3)(4)(5)(6), ubiquitination (7), and subsequent proteolytic degradation by the 26 S proteasome pathway (8). Upon nuclear import and binding to specific decameric recognition motifs, which are reflected by the consensus GGGRNNYYCC (where R represents A or G and Y represents C or T), NF-B dimers function as transcriptional activators. IB␣ (9), IB␤, and p105 (10) have been implicated in the inhibition of DNA binding of NF-B complexes. However, there have been several reports showing that NF-B transcriptional activity can be blocked without affecting DNA binding. These include the interactions of NF-B with the glucocorticoid receptor (11,12), the mammalian repressor REP (13), and the interferon-inducible factor p202 (14).
Emerging evidence also suggests a second level of controlling NF-B transcriptional activity that acts directly on NF-B dimers without influencing the degradation of IB molecules. For example, ectopic expression of a dominant negative mutant of the atypical protein kinase C (PKC) or the extracellular signal-regulated kinase 1 inhibit TNF-␣-induced NF-B activity (15). Similarly, inhibition of p38 mitogen-activated protein kinase (p38 MAPK) has been shown to decrease TNF-␣-induced NF-B activity and interleukin-6 expression (16). More recently, tyrosine phosphorylation has been shown to be essential for NF-B activity in bacterial lipopolysaccharide (LPS)-induced monocytic THP1 cells (17). Regulation of NF-B activity by PKC, extracellular signal-regulated kinase 1, p38 MAPK, or tyrosine phosphorylation acts downstream of IB without interfering with NF-B nuclear translocation and DNA binding.
As for other members of the atypical protein kinase C family, PKC is not activated by Ca 2ϩ or diacylglycerol and is insensitive to phorbol esters (18). Unresponsiveness of PKC to Ca 2ϩ and diacylglycerol is consistent with the absence of the Ca 2ϩ binding C2 domain and the presence of only one cysteine-rich zinc finger-like motif in the diacylglycerol binding C1 domain of PKC. PKC is activated by several lipid mediators including phosphatidic acid (19) and phosphatidylinositol 3,4,5-trisphosphate (20). PKC has also been shown to be activated by TNF-␣ and interleukin-1 through sphingomyelin hydrolysis and subsequent generation of ceramide (21)(22)(23). Other pathways leading to PKC activation include the 21-kDa guanine nucleotidebinding p21 ras (24), which in addition to several growth factors is also activated by TNF-␣ as well as LPS (25,26). Ras GTPases have been implicated in the signaling of a variety of extracellular stimuli that control cell proliferation and differentiation. Ras GTPases are activated by members of the guanine nucleotide exchange factor family, which increase Ras GTP loading and are negatively regulated by the GTPase-activating proteins, which enhance the intrinsic rate of hydrolysis of Rasbound GTP. Upon binding to GTP, Ras recruits and activates downstream effectors such as Raf, PI 3-kinase (27) and the kinase suppressor of Ras (28) by a mechanism that is not well understood. p21 ras has also been implicated in controlling NF-B activity in fibroblasts (29,30).
In this study, we analyzed the role of PKC and p21 ras in regulating NF-B activity in endothelial cells. We demonstrate that inhibition of either one of these pathways changes the phosphorylation of the RelA subunit and severely impairs NF-B-mediated transcription without interfering with the ability of NF-B to bind to DNA.

MATERIALS AND METHODS
Cell Culture-Bovine aortic endothelial cells (BAEC) and porcine aortic endothelial cells (PAEC) were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum, L-glutamine (2 mM), penicillin G (100 units/ml), and streptomycin (100 g/ml). Human umbilical vein endothelial cells (HUVEC) were grown in M199 medium supplemented with 15% fetal bovine serum, NaH 2 CO 3 (20 mM), HEPES (25 mM), glutamine (5 mM), heparin (100 g/ml), gentamycin (50 g/ml), and endothelial growth factor (50 g/ml). Primary cultures of PAEC and HUVEC were used between the fourth and the fifth passage. BAEC were used between the fifth and the seventh passage. All cells were grown in culture at 37°C in a 5% humid CO 2 atmosphere. All media and supplements were from Life Technologies, Inc.
Plasmid Constructs-The pcDNA3 vector expressing tagged wildtype Xenopus laevis PKC and rat PKC dominant negative mutant were a kind gift of J. Moscat (Universidad Autónoma, Madrid) and were described elsewhere (15). Expression vectors encoding wild-type p21 ras , a dominant-negative (RasN17) and a constitutively active mutant (RasV12) were a kind gift from G. M. Cooper (Harvard Medical School). The inserts were amplified by polymerase chain reaction with primers carrying appropriate restriction sites and cloned into pcDNA3HA, which is derived from pcDNA3 (Invitrogen, Carlsbad, CA) by inserting a DNA fragment coding for MYPYDVPDYASL, where amino acids 2-12 code for an epitope derived from the hemagglutinin protein of the human influenza virus. RhoA, Rac1, and Cdc42 were amplified from HeLa cDNA by polymerase chain reaction and cloned into pcDNA3HA. Dominant negative mutants of these small GTPases were generated by overlap extension as described elsewhere (31). Cdc42N17 was generated by replacing Thr 17 with Asn employing the overlapping primers 5Ј-GTAAAAACTGTCTCCTGATATCCTAC and 5Ј-GATATCAGGAGA-CAGTTTTTACCAACAGCACC, Rac1N17 was generated by replacing Thr 17 with Asn using the primers 5Ј-CTGTAGGTAAAAACTGCCTACT-GATC and 5Ј-TGATCAGTAGGCAGTTTTTACCTACAGCTCCG, and RhoAN17 was generated by replacing Thr 19 with Asn using the primers 5Ј-GTGGAAAGAACTGCTTGCTCATAGTCTTC and 5Ј-ATGAGCAAG-CAGTTCTTTCCACAGGCTCCATC (the underlined base triplet indicates the mutated amino acid). The Raf-1 dominant negative mutant encompassing the first 259 amino acids encoding the regulatory domain (32) was generated by polymerase chain reaction using full-length human Raf-1 (ATCC 41050) as template. The Src homology 2 (SH2) domain of the 85-kDa regulatory subunit of PI 3-kinase shown to act as a dominant negative mutant (33) was cloned from BAEC cDNA and cloned into the pcDNA3 vector. The different fusion proteins outlined in Fig. 5A were generated by a polymerase chain reaction-based approach and were all expressed from the pcDNA3 vector. The RelA expression plasmid is based on the pcDNA3 vector and comprises the human RelA coding region fused to a N-terminal Myc tag sequence. The RelA/RHD expression vector has been described elsewhere (34). The RelA DNA binding mutant (RelA DNAmut ) that harbors an RF 3 KA mutation at amino acids 33 and 34, respectively (35), and the RelA mutant harboring a Ser 276 3 Ala substitution were generated by primer overlap extension as described above. The sequence of all constructs was confirmed by double-stranded DNA sequencing. The TetO-Luc (pBI5) reporter was a kind gift of H. Bujard (University of Heidelberg). Other reporter constructs used in this study were described previously (12,36).
Transient Transfection and Reporter Assays-Primary BAEC were transfected as described (34). Experiments involving RelA co-transfection were analyzed 20 -24 h after transfection. Where indicated, cells were incubated with human recombinant TNF-␣ (R & D, Systems, Minneapolis, MN; 50 units/ml, 7 h) 40 -44 h after transfection. Cells were washed once in phosphate-buffered saline and disrupted in lysis buffer (0.1 M KH 2 PO 4 , pH 7.6, 1 mM dithiothreitol and 0.05% Triton X-100). Luciferase and ␤-galactosidase activities were assayed as described previously (12). All experiments were done in triplicate as indicated, and luciferase activities were normalized to ␤-galactosidase levels to account for differences in transfection efficiency.
PKC Immunodetection and Activity Assays-PKC Western blot detection was performed on PVDF membranes using a rabbit polyclonal antibody directed against the C terminus of human PKC (sc-216; Santa Cruz Biotechnology, Inc., Santa Cruz, CA). Bands were visualized using horseradish peroxidase-conjugated donkey anti-rabbit IgG (Pierce) and the ECL assay (Amersham Pharmacia Biotech) according to the manufacturer's instructions. Immunoprecipitation of PKC was carried out as described previously (37) with the following modifications. After preclearing with protein G-Sepharose (Amersham Pharmacia Biotech), extracts were incubated with 3 g of nonimmune rabbit IgG or 3 g of anti-PKC antibody (sc-216; Santa Cruz Biotechnology) for 4 h. Antibodies were captured by adding 20 l of protein G-Sepharose and washed twice in lysis, twice in Tris/LiCl and once in 25 mM Tris-HCl buffer. For autophosphorylation experiments, PAEC were serum-starved for 24 h and metabolically labeled with [ 32 P]orthophosphate (200 Ci/ml, 4 h). Immunoprecipitates were obtained as described above, and captured proteins were eluted by boiling in Laemmli buffer. Proteins were resolved on 10% polyacrylamide gels under denaturing conditions. The gels were dried and subjected to autoradiography. For kinase activity assays, immunoprecipitates obtained from serum-starved PAEC were incubated in reaction buffer (50 mM Tris-HCl, pH 7.5, 5 mM MgCl 2 , 1 mM MnCl 2 , and 100 M ATP) supplemented with 3 g of myelin basic protein and 3 Ci of [␥-32 P]ATP. Reactions were carried out for 20 min at 30°C and stopped by adding Laemmli buffer. Proteins were separated on a 12.5% polyacrylamide gel under denaturing conditions. Gels were dried and quantitated by Phosphor-Imager analysis (Molecular Dynamics, Inc., Sunnyvale, CA). p21 ras Immunodetection and Activity Assay-PAEC and HUVEC were labeled with [ 35 S]Met/Cys for 6 h. Cells (2-3 ϫ 10 6 ) were disrupted (20 min on ice) in lysis buffer (10 mM Tris⅐HCl, pH 7.5, 150 mM NaCl, 1 mM EDTA, 0.2% Triton X-100) supplemented with phosphatase and protease inhibitors. Lysates were cleared by centrifugation, and p21 ras was immunoprecipitated by incubating lysates overnight with a rat monoclonal anti-p21 ras antibody coupled to agarose beads (sc-35AC; Santa Cruz Biotechnology). Beads were washed eight times in wash buffer (50 mM HEPES, pH 7.4, 500 mM NaCl, 5 mM MgCl 2 , 0.1% Triton X-100, 0.005% SDS). Proteins were eluted by boiling in Laemmli buffer and separated on 15% polyacrylamide gels under denaturing conditions. Gels were dried and subjected to autoradiography. To analyze p21 ras GTP loading, serum-starved PAEC were labeled with [ 32 P]orthophosphate for 4 h. Cells were left untreated or stimulated with TNF-␣ (50 units/ml) as indicated. p21 ras immunoprecipitation was carried out as described above. Bound nucleotides were eluted by incubating immunoprecipitates in elution buffer (0.2% SDS, 5 mM dithiothreitol, 1 mM GDP, 1 mM GTP 2 mM EDTA) for 15 min at 68°C. Equal amounts of eluates (500 cpm) were loaded on polyethylenimine cellulose plates (Merck, Darmstadt, Germany), and nucleotides were separated by chromatography in 3 M LiCl, pH 3.4. Plates were dried and quantified by PhosphorImager analysis.
Electrophoretic Mobility Shift Assay-Whole cell extracts were prepared from transfected BAEC as described before (38). All buffers were supplemented with 10 g/ml aprotinin, 25 M leupeptin, 1 M pepstatin, and 1 mM phenylmethylsulfonyl fluoride. Cell extracts were incubated (30 min at room temperature) with 100,000 cpm of doublestranded [␥-32 P]ATP-radiolabeled NF-B oligonucleotide (5Ј-AGTTGA-GGGAATTTCCCAGGC-3Ј), and the resulting DNA-protein complexes were separated on a 5% polyacrylamide gel in Tris/glycine/EDTA buffer at pH 8.5. The amount of cell extracts for each binding reaction was adjusted to ␤-galactosidase activity to compensate for differences in transfection efficiency.
RelA Metabolic Labeling and Immunoprecipitation-PAEC or BAEC cultured in 10-cm dishes were labeled with [ 32 P]orthophosphate (500 Ci/ml) in phosphate-free Dulbecco's modified Eagle's medium for 4 h and stimulated with TNF-␣ for 30 min. LPS stimulation was carried out in the presence of 2% dialyzed fetal bovine serum (Sigma) for 60 min. Cells were washed twice in ice-cold Tris-buffered saline and scraped in 1 ml of lysis buffer (50 mM Tris, pH 7.5, 150 mM NaCl, 1% Nonidet P-40, 0.5% deoxycholate, 0.1% SDS, 20 mM ␤-glycerophosphate, 50 mM NaF, 1 mM orthovanadate, 1 mM EDTA, 1 mM EGTA, 10 g/ml aprotinin, 25 M leupeptin, 1 M pepstatin, and 1 mM phenylmethylsulfonyl fluoride). Extracts were homogenized by passing them five times through a 25gauge needle and cleared by centrifugation. RelA was immunoprecipitated from precleared lysates using an agarose-coupled polyclonal antibody directed against the N terminus of human RelA (sc-109AC; Santa Cruz Biotechnology). Immunoprecipitates were washed four times in lysis buffer and once in 50 mM Tris-HCl, pH 6.8. Proteins were eluted by boiling in Laemmli buffer, separated on 10% polyacrylamide gels under denaturing conditions, and transferred to a PVDF membrane that was subjected to autoradiography. Sequential immunoprecipitations were carried out using RelA (sc-109), NFB1 (sc-114) or IB␣ (sc-371; Santa Cruz Biotechnology) specific antibodies as described elsewhere (39).
Phosphoamino Acid Analysis-RelA was immunoprecipitated from metabolically labeled, TNF-␣-stimulated PAEC and electrophoretically separated as described above. The band corresponding to RelA was cut out, and amino acids were prepared by acidic hydrolysis. Phosphoamino acids were separated by two-dimensional thin layer electrophoresis as described before (40), and plates were subjected to autoradiography.
RelA Phosphopeptide Mapping-BAEC or BAEC transfected with various RelA constructs as described above were metabolically labeled with 500 Ci/ml [ 32 P]orthophosphate for 4 h. Cell extracts were prepared by detergent lysis as described above and immunoprecipitated with anti-RelA-agarose (Santa Cruz Biotechnology). Immunoprecipitates were boiled in Laemmli buffer, separated by polyacrylamide gel electrophoresis, and transferred to a PVDF membrane. Tryptic digests were obtained as described (40), and equal amounts of radioactivity were loaded on cellulose plates. The first dimensional electrophoretic separation was carried out in ammonium carbonate buffer (pH 8.9). The chromatography was performed in an n-butanol/pyridine/glacial acetic acid/H 2 O (37.5:25:7.5:30) buffer. Plates were exposed to x-ray film or analyzed using a PhosphorImager scanning device (Molecular Dynamics).

Activation of PKC and p21 ras by TNF-␣ in Endothelial
Cells-HUVEC, BAEC, and PAEC expressed similar levels of PKC as assayed by Western blotting (Fig. 1A). In the presence of serum, PKC was constitutively active in these cells (data not shown). However, PKC activity was significantly reduced when endothelial cells were serum-starved. Under serum starvation, PKC was activated by both TNF-␣ and LPS as assayed by PKC autophosphorylation (Fig. 1B) or kinase activity (Fig.  1C). Maximal PKC activity was reached 20 min after TNF-␣ stimulation (Fig. 1C).
As for PKC, p21 ras activity was constitutively high in endothelial cells cultured in the presence of serum, which was also reflected by high MEK1 (mitogen-activated protein and extracellular signal-regulated kinase kinase 1) and extracellular signal-regulated kinase/MAPK activity (data not shown and Ref. 41). Immunoprecipitation of p21 ras from PAEC or HUVEC revealed two closely migrating bands probably corresponding to processed and nonprocessed form of p21 ras (Fig. 1D). Under serum deprivation, both p21 ras and MEK activity were markedly reduced, and TNF-␣ induced p21 ras activity as reflected by an increase in Ras GTP loading (Fig. 1E).
Regulation of RelA Transcriptional Activity by PKC and p21 ras -To test whether PKC is involved in regulation of NF-B activity in endothelial cells, we used a dominant negative mutant of the rat PKC in which Lys 281 was replaced by Trp (PKC mut ; Ref. 15). BAEC were transiently co-transfected with PKC mut and with a NF-B-dependent luciferase reporter (B-Luc), regulated by three NF-B consensus sites derived from the porcine E-selectin promoter (12). TNF-␣-induced luciferase expression was inhibited in a dose-dependent manner by increasing amounts of PKC mut (Fig. 2A). Furthermore, we investigated whether PKC mut would interfere directly with RelA-mediated transcription. When BAEC were co-transfected with RelA and increasing amounts PKC mut together with the B-Luc reporter, luciferase expression was inhibited in a dosedependent manner (Fig. 2B). PKC mut was more efficient in repressing RelA activity than in repressing TNF-␣-mediated NF-B activation, indicating that TNF-␣ may generate additional signals that can partially override the inhibitory effect of PKC mut .
Given that p21 ras has been implicated in the PKC signaling cascade (24,42), we tested whether a dominant negative mutant of p21 ras (RasN17) would interfere with NF-B-mediated transcription. Overexpression of increasing amounts of RasN17 in BAEC abolished TNF-␣-mediated up-regulation of the B-Luc reporter in a dose-dependent manner ( Fig. 2A). This inhibitory effect was more pronounced than the one seen with PKC mut (Fig. 2B). We then analyzed whether RasN17 would interfere directly with RelA activity. Co-transfection of RasN17 with RelA repressed transcription from the B-Luc reporter to a similar extent as PKC mut (Fig. 2B). Both PKC mut and RasN17 also inhibited RelA/NFB1 transcriptional activity to a similar extent as observed for RelA (Fig. 2C). Comparable results were obtained when RelA was co-expressed with a reporter construct under the control of the porcine IB␣ Serum-starved PAEC were labeled with [ 32 P]orthophosphate and stimulated with TNF-␣ (50 units/ml) or LPS (1 g/ml) for 15 min. Cell extracts were prepared by detergent lysis and incubated with nonimmune rabbit IgG or an anti-PKC antibody. Immunoprecipitates were separated on 10% polyacrylamide gels under denaturing conditions. Phosphorylated PKC was revealed by autoradiography. C, PKC immune complex kinase assay. Serum-starved PAEC were stimulated with TNF-␣ (50 units/ml) as indicated. PKC was immunoprecipitated, and kinase reactions were carried using myelin basic protein as a substrate. D, p21 ras expression in endothelial cells. HUVEC or PAEC were labeled with [ 35 S]Met/Cys, and Ras was immunoprecipitated using Y13-259 rat monoclonal antibody in the absence (A) or presence of a 40-fold molar excess of specific peptide (P). Proteins were separated on 15% polyacrylamide gels under denaturing conditions and visualized by autoradiography. E, serum-starved PAEC were labeled with [ 32 P]orthophosphate and stimulated with TNF-␣ (50 units/ml) as indicated. p21 ras was immunoprecipitated, nucleotides were eluted, and equal amounts of radioactive material were separated by thin layer chromatography. GDP and GTP bands were quantified by PhosphorImager analysis, and GTP/GDP ratios were calculated: % GTP ϭ GTP/(GDP ϫ 1.5 ϩ GTP) ϫ 100. Values present mean Ϯ S.D. (n ϭ 2). promoter (Fig. 2D). The observation that a constitutive active mutant of p21 ras (RasV12) did not complement the inhibitory effect of PKC mut suggests that p21 ras does not act downstream of PKC in controlling RelA activity. Moreover, we found that a constitutive active form of PKC, comprising the catalytic region (amino acids 254 -592 of the human PKC) did not overcome the inhibitory effect of RasN17 (data not shown). Taken together, these data suggest that p21 ras and PKC regulate RelA transcriptional activity by separate pathways.
Neither PKC mut nor RasN17 altered the levels of overexpressed RelA in BAEC as monitored by Western blotting (data not shown). Additionally, PKC mut or RasN17 did not inhibit DNA binding of RelA as monitored by electrophoretic mobility shift assay (Fig. 3). We conclude therefore that the inhibitory effect of PKC mut or RasN17 is not due to inhibition of RelA DNA binding activity. These data suggest that regulation of NF-B activity by PKC and p21 ras acts downstream of IB directly on RelA.
Role of Small GTPases, Raf-1, and PI 3-Kinase-There are several potential downstream targets of the p21 ras and PKC pathway that may account for the regulatory mechanism of p21 ras or PKC. First, we monitored the effect of p21 ras -related GTPases of the Rho and Rac family, previously shown to regulate NF-B activity in fibroblasts (43), on RelA transcriptional activity. Expression of dominant negative mutants of Cdc42 (Cdc42N17), Rac1 (RacN17), or RhoA (RhoN19) (43) did not inhibit RelA-mediated activation of the B reporter in BAEC (Fig. 4A). Similar results were obtained for TNF-␣-induced NF-B activity (data not shown). Raf-1 and PI 3-kinase have been shown to be involved in p21 ras (44,45) and the latter also in PKC signaling cascades (46) and in regulating NF-B ac-tivity in fibroblasts (29) and hepatocytes (47), respectively. To investigate the role of Raf-1 in modulating NF-B activity in EC, we used a Raf-1 dominant negative mutant (Raf1-259) that has been shown to act as a dominant repressor of Ras-Raf-1 signaling (32). This mutant inhibited a RasV12-induced Elk-1-and c-Jun-dependent reporter system in EC (data not shown). Overexpression of the Raf-1 dominant negative mutant together with RelA led only to an insignificant reduction of B-dependent reporter activity (Fig. 4B). Similar results were obtained expressing a dominant negative mutant of the PI 3-kinase (p85N-SH2; Ref. 33). Moreover, pretreatment of cells with wortmannin, a specific inhibitor of PI 3-kinase, had no effect on RelA-or TNF-␣-induced B reporter activity. Likewise, wortmannin or LY294002, another inhibitor of PI 3-kinase, did not effect TNF-␣-induced IB␣ degradation, NF-B DNA binding, or up-regulation of NF-B-dependent endogenous genes, i.e. IB␣ and E-selectin (data not shown).
RelA RHD Is the Target of PKC and p21 ras -mediated Regulation of Transcriptional Activity-To monitor which domain of RelA is targeted by PKC and p21 ras , we constructed different fusion proteins outlined in Fig. 5A. The first construct was composed of the DNA binding domain derived from the bacterial tetracycline repressor (TET) fused to a transactivation domain derived from the Herpes simplex virus VP16 protein (TET/VP16). The second construct was composed of the TET DNA binding domain fused to the C-terminal region of RelA (amino acids 286 -551) that includes the transactivation domain (TET/RelA286 -551). In addition, we generated a construct composed of the RelA RHD fused to the VP16 transactivation domain (RelA2-320/VP16). Transcriptional activity of constructs harboring the TET DNA binding domain was analyzed by co-transfection with a reporter containing seven tetracycline operons (TetO) fused to a luciferase gene (TetO-Luc). Transcriptional activity of constructs harboring the RelA DNA binding domain was analyzed by co-transfection with the B-Luc reporter. Whereas RelA-mediated transcription was repressed by both PKC mut and RasN17 (Fig. 2B), TET/VP16mediated transcription was not inhibited by these mutants. TET/RelA286 -552-mediated transcription was not inhibited by PKC mut or RasN17, while RelA2-320/VP16 transcriptional activity was inhibited by both mutants in a similar manner to wild type RelA (Fig. 5B). These data suggest that PKC and p21 ras regulate RelA transcriptional activity by targeting the RelA RHD.
Regulation of RelA Transcriptional Activity by PKC and p21 ras Is Dependent on Functional B Consensus Sites-Having established that the regulatory effect of PKC and p21 ras is dependent on RelA RHD, we analyzed whether DNA binding through RHD was necessary for inhibition of RelA transcriptional activity. To test this possibility, we constructed a fusion protein that contains the TET DNA binding domain and the full-length RelA (TET/RelA2-551; Fig. 5A) and carries therefore two DNA binding domains (for TetO and B consensus binding sites). This construct allows one to analyze the effect of PKC mut and RasN17 on its transcriptional activity depending on the binding to two different DNA consensus sites. As shown in Fig. 5C, transcriptional activity of this fusion protein was repressed by PKC mut and RasN17 when co-transfected with the B-Luc reporter, while it was not affected when the TetO-Luc reporter was used. Since the B-Luc reporter harbors the thymidine kinase minimal promoter, while the TetO-Luc harbors the cytomegalovirus minimal promoter, we tested whether the use of these two minimal promoters would account for the differential regulation. To do so, we constructed a B reporter containing the same cytomegalovirus minimal promoter fragment that drives the TetO-Luc construct. When transfected with RelA and PKC mut or RasN17, this B reporter behaved the same way as the reporter construct based on the thymidine kinase minimal promoter (data not shown). We conclude therefore that regulation of RelA transcriptional activity by PKC and p21 ras involves the RelA RHD and is only relevant if RelA binds DNA through a B consensus site.
Phosphorylation of Endogenous RelA-As previously reported, RelA is phosphorylated upon stimulation with TNF-␣

FIG. 4. Role of small GTPases, Raf-1, and PI 3-kinase on NF-B activation in EC.
A, BAECs were transfected with the B-Luc reporter (700 ng) and RelA expression plasmid (30 ng) alone or together with Cdc42N17 (500 ng), RacN17 (500 ng), or RhoN19 (500 ng). B, BAECs were transfected with the B-Luc reporter (700 ng) and RelA expression plasmid (30 ng) alone or together with increasing amounts of Raf1-259 or p85N-SH2 (100, 250, and 500 ng). C, BAECs were transfected with the B-Luc reporter (700 ng) and RelA expression plasmid (30 ng). Cells were left untreated or were treated with vehicle (asterisk; Me 2 SO, 1 l/ml) or wortmannin (Wort) at 10, 100, and 1000 nM concentration (16 h, starting 4 h after end of transfection). D, BAECs were transfected with the B-Luc reporter (700 ng). Thirty-six hours after transfection, cells were incubated with vehicle (asterisk; Me 2 SO, 1 l/ ml) or wortmannin (Wort) at 10, 100, and 1000 nM concentration. TNF-␣ (50 units/ml) was added 1 h later, and incubation was continued for 7 h, after which cell extracts were prepared. Luciferase activities were assayed as described under "Materials and Methods." The total amount of DNA in all transfections was kept constant with pcDNA3 plasmid. Luciferase activities were normalized to ␤-galactosidase activities to compensate for differences in transfection efficiency. The error bars represent mean Ϯ S.D. (n ϭ 3).

FIG. 5. RelA RHD is the target for PKC mut -and RasN17-mediated transcriptional repression.
A, schematic representation of plasmid constructs used for transfections. All constructs were cloned into the mammalian expression vector pcDNA3. B, BAEC were transfected with different fusion proteins as outlined in A (all at 30 ng) and with PKC mut (500 ng), RasN17 (500 ng), B-Luc (700 ng), or TETO-Luc (700 ng) reporter as indicated. C, BAEC were transfected with TET/ RelA2-551 (30 ng) along with 700 ng of B-Luc or TETO-Luc and protein kinase expression vectors as indicated (all at 500 ng). 24 h after transfection, cells were lysed, and luciferase activities were assayed as described under "Materials and Methods." The total amount of DNA in all transfections was kept constant with pcDNA3 plasmid. Luciferase activities were normalized to ␤-galactosidase activities to compensate for differences in transfection efficiency. The error bars represent mean Ϯ S.D. (n ϭ 3). (39,48,49). In endothelial cells, TNF-␣ induces RelA phosphorylation as analyzed by immunoprecipitation of RelA from [ 32 P]orthophosphate-labeled cells (Fig. 6A). Several other phosphopeptides were co-immunoprecipitated along with RelA. The identity of the precipitated phosphopeptides was confirmed by Western blot analysis and by sequential immunoprecipitations and identified as NFB1, p105, and IB␣ (data not shown). While RelA phosphorylation was increased upon TNF-␣ stimulation, NFB1 was dephosphorylated (seven independent experiments). The decrease in intensity of the IB␣ corresponding band upon TNF-␣ stimulation was due to IB␣ degradation as assayed by Western blotting. Furthermore, we analyzed the identity of phosphoamino acids derived from RelA isolated from TNF-␣-stimulated PAECs by two-dimensional electrophoresis. As shown in Fig. 6B, these phosphoamino acids were primarily composed of serine and only to a minor extent threonine residues, while no tyrosine phosphorylation was observed.
Phosphorylation of RelA has been attributed to phosphorylation of serines 276 (48) and 529 (50). Both studies could show an exclusive role for the respective serine in RelA phosphorylation. We investigated RelA phosphorylation by two-dimensional separation of tryptic phosphopeptides prepared from nontreated, TNF-␣-(30 min), or LPS-(60 min) treated BAEC. In quiescent EC, RelA is phosphorylated on multiple sites, resulting in at least nine distinct phosphopeptides. Upon TNF-␣ or LPS stimulation, the pattern of RelA phosphorylation changes significantly as reflected by an increase in signal intensity of several phosphopeptides (Fig. 7, spots b, e, f, g, h, and  i). The increase of phosphorylation seems to be strongest on peptide b. TNF-␣ and LPS lead to similar changes in RelA phosphorylation. While peptides b, e, f, g, and h are phosphorylated to a similar extent in TNF-␣-or LPS-stimulated cells, peptide i seems to be more phosphorylated in TNF-␣-treated cells. These data suggest the existence of several constitutive and inducible phosphorylation sites in RelA.
Phosphorylation of Overexpressed RelA-To investigate whether an exogenous stimulus is necessary to trigger RelA phosphorylation or alternatively if free, i.e. non-IB-bound, RelA is sufficient to trigger phosphorylation, we overexpressed RelA in BAEC and analyzed the level of phosphorylation. As shown in Fig. 8A (lane 1), overexpressed RelA is readily phosphorylated in unstimulated cells. TNF-␣ stimulation did not affect the phosphorylation status of overexpressed RelA (data not shown), suggesting that signaling by TNF-␣ might not be essential for RelA phosphorylation. We next examined whether phosphorylation of overexpressed RelA was inhibited by IB␣ co-expression under conditions where all RelA would be com-plexed to IB␣. IB␣ expression resulted in substantial reduction in RelA phosphorylation (Fig. 8A, lane 5). The decrease in RelA phosphorylation was not due to a decrease in protein levels as monitored by immunodetection of RelA on the same membrane used for phosphorylation analysis (Fig. 8B, compare lane 1 with lane 5). These data indicate that RelA is not fully phosphorylated when retained by IB␣. This would suggest three possible scenarios. RelA is phosphorylated (i) in the cytoplasm upon liberation from IB, (ii) in the nucleus before binding to DNA, or (iii) upon binding to DNA. In order to investigate if DNA binding is necessary for RelA phosphorylation, we expressed a previously described RelA DNA bindingdeficient mutant (35). As shown in Fig. 8A (lane 3), the phosphorylation of this DNA binding mutant was strongly reduced as compared with wild type RelA. The most likely explanation for this finding is that phosphorylation of RelA occurs after nuclear translocation and DNA binding.
Finally, we addressed the topology of RelA phosphorylation. To test to what degree RelA RHD participates in the overall phosphorylation of RelA, we transfected BAEC with a RelA mutant that encodes the N-terminal 320 amino acids (RelA/ RHD; Ref. 34). Overexpressed RelA/RHD was readily phosphorylated (Fig. 8A, lane 2), and its phosphorylation was completely inhibited by co-expressed IB␣ (Fig. 8A, lane 4). As for RelA, IB␣ co-expression did not change RelA/RHD protein levels as assayed by Western blot analysis (Fig. 8B, lanes 2  and 4).
To investigate the contribution of serine 276 to the overall RelA phosphorylation, RelA was immunoprecipitated from cells transfected with wild-type RelA (RelA wt) or with a mutated form, where Ser 276 was replaced by Ala (RelA S276A). The phosphorylation of RelA S276A was significantly lower as compared with wild type RelA (Fig. 9A). However, the reduction in phosphorylation was not only caused by a decrease in specific phosphorylation but also by reduced protein levels of the RelA S276A mutant. Although EC were transfected with a 3-fold excess of RelA S276A construct as compared with wild type RelA, the level of expression of RelA S276A was always lower than wild type RelA (data not shown). We are currently investigating the causes of this phenomenon.
Comparison of tryptic peptide maps from wild type RelA and RelA S276A revealed the specific loss of a phosphopeptide in RelA S276A (Fig. 9B), corresponding to the phosphopeptide a in the endogenous RelA (Fig. 7), which is constitutively phosphorylated in RelA, and its phosphorylation status is not altered by TNF-␣ or LPS stimulation. These data indicate that Ser 276 is constitutively phosphorylated in EC.
FIG. 6. Phosphorylation of RelA. A, PAEC were metabolically labeled with [ 32 P]orthophosphate and stimulated with TNF-␣ for 30 min. RelA was immunoprecipitated and separated on a 10% polyacrylamide gel under denaturing conditions. B, RelA was immunoprecipitated from TNF-␣-stimulated PAEC, and amino acids were prepared by acidic hydrolysis. Phosphoamino acids were separated by two-dimensional thin layer electrophoresis and revealed by autoradiography. The position of the phosphoamino acid standards is marked by circles.
Inhibition of RelA Phosphorylation by Blockage of PKC and p21 ras Signaling Pathways-We next investigated whether inhibition of PKC or p21 ras signaling pathways would interfere with RelA phosphorylation. Overexpression of RelA together with PKC mut or RasN17 significantly decreased RelA phosphorylation as compared with overexpression of RelA alone (Fig. 10A, top). PKC mut or RasN17 did not decrease RelA protein levels as monitored by immunodetection of RelA (Fig.  10A, bottom). To monitor which domain of RelA was targeted by PKC mut or RasN17, we used different fusion proteins described above and outlined in Fig. 5A. The phosphorylation of RelA2-320/VP16 (containing the RelA RHD) was significantly inhibited by co-expressed PKC mut or RasN17 (Fig. 10B, top). PKC mut or RasN17 did not decrease RelA2-320/VP16 protein Fig. 10B, bottom). Phosphorylation of the TET/RelA286 -551 construct, which contains the C-terminal RelA transactivation domain, was not inhibited by PKC mut or RasN17 (Fig. 10C).
Having established that PKC and p21 ras are involved in the phosphorylation of RelA RHD, we analyzed the phosphorylation pattern of the RHD by tryptic peptide mapping. Phosphopeptides derived from RelA/RHD expressed alone or together with PKC mut or RasN17 were analyzed by twodimensional separation on thin layer cellulose plates. Compared with the phosphopeptide map derived from endogenous or overexpressed full-length RelA (Fig. 7), the most striking difference is the disappearance of the most basic peptide (Figs. 7 and 9, spot a) and the appearance of a very acidic peptide (Fig. 11, spot x). The pattern of RHD phosphorylation Cells were labeled with [ 32 P]orthophosphate, and cell extracts were immunoprecipitated with anti-RelA antibody. Immunoprecipitates were separated on 10% polyacrylamide gels. B, tryptic digests were analyzed by two-dimensional separation on thin layer cellulose plates as described under "Materials and Methods". The sample application point is marked (ϩ). The nomenclature of radioactive spots follows that of Fig. 7 to indicate corresponding spots. was substantially modified when RelA/RHD was co-expressed with PKC mut or RasN17 as compared with overexpression of RelA/RHD alone (Fig. 11). These changes were restricted to three separate peptides and were not equivalent in PKC mutand RasN17-transfected cells. While phosphorylation of peptide b disappeared in cells transfected with PKC mut and RasN17 (Fig. 11), phosphorylation of peptides d and g was only inhibited by PKC mut and not by RasN17 (Fig. 11). These differences in the phosphorylation pattern again suggest that PKC and p21 ras feed into at least partially separated pathways controlling RelA phosphorylation. DISCUSSION It is widely accepted that regulation of NF-B transcriptional activity is controlled mainly by retention of NF-B in the cytoplasm by members of the IB family. In this study, we demonstrate that at least in endothelial cells there is an additional regulatory system that controls the transcriptional activity of nuclear NF-B by targeting the RelA subunit. This regulatory system involves signaling through PKC and p21 ras .
Several kinases have been implicated in the regulation of nuclear RelA transcriptional activity. Protein kinase A is involved in the regulation of RelA transcriptional activity through phosphorylation of Ser 276 in the consensus site (RRPS) located in the RHD (39,48). In addition, p38 MAPK has also been implicated in regulating RelA transcriptional activity. However, contrary to protein kinase A, p38 MAPK may not act directly on RelA (51), as suggested by the observation that inhibition of p38 MAPK does not result in detectable changes in RelA phosphorylation (16). Casein kinase II has also been shown to associate with NF-B in vivo and to phosphorylate the C-terminal transcriptional activation domain of RelA in vitro (52). We now demonstrate that PKC and p21 ras are two additional components in the regulation of RelA transcriptional activity. We show that inhibition of these signaling cascades results in decrease of RelA transcriptional activity that correlates with inhibition of RelA phosphorylation.
Several downstream effectors may account for the effect of PKC or p21 ras over NF-B. One common feature shared by both PKC and p21 ras is the ability to activate the MEK/extracellular signal-regulated kinase pathway, which has been suggested to control NF-B activity (15,53). However, we found that at least in endothelial cells a dominant negative mutant of Raf1 does not interfere with NF-B-mediated transcription. Another downstream effector of p21 ras and PKC is the c-Jun N-terminal kinase signaling cascade. It is unlikely that this pathway is involved in NF-B regulation in endothelial cells, since a dominant negative c-Jun N-terminal kinase 1 failed to inhibit RelA transcriptional activity (data not shown). Moreover, a dominant negative mutant of Rac1 (RacN17) efficiently blocked p21 ras induced c-Jun N-terminal kinase activation while it failed to inhibit NF-B activity (data not shown and Ref. 54). Inhibition of PI 3-kinase by a dominant negative mutant or by wortmannin failed to have an effect on RelAmediated transcription. Furthermore, inhibition of PI 3-kinase did not impair TNF-␣-induced B-dependent reporter activity.
RelA has been shown to be inducibly phosphorylated upon cytokine stimulation in several cell types, and phosphorylation of the transactivation domain has been proposed to be a major regulatory mechanism by which the activity of several transcription factors is controlled (55). Similarly, phosphorylation of the RelA transactivation domain has been reported (56). In particular, inducible phosphorylation of the TA 2 (amino acids 428 -520) and constitutive phosphorylation of the TA 1 (amino acids 521-551) activation domains have been suggested to control RelA transcriptional activity (56). Recently, phosphorylation of the RelA transactivation domain by RelA-associated casein kinase II has been reported (52), and the importance of phosphorylation of Ser 529 has been revealed (50). In this study, we present evidence that the RHD domain contributes substantially to the overall phosphorylation of RelA. We show that RelA phosphorylation can be inhibited partially by co-expressing IB␣, which suggests that RelA is phosphorylated upon liberation from associated IB molecules. The observation that phosphorylation of full-length RelA is only partially inhibited by IB␣ overexpression, whereas phosphorylation of RelA RHD is completely inhibited, suggests that the C terminus of RelA is constitutively phosphorylated, while inducible phosphorylation occurs mainly on the RHD.
We further show that RelA is constitutively phosphorylated at multiple sites and that phosphorylation of some but not all of this site is increased by TNF-␣ or LPS treatment. While the phosphorylation of some of these sites is regulated by both PKC and p21 ras signaling cascades, phosphorylation of other sites is not altered by these pathways. Furthermore, our data suggest that RelA is constitutively phosphorylated at Ser 276 , and this phosphorylation is not altered by TNF-␣ or LPS treatment. In addition, a RelA S276A mutant retained several phosphorylated sites, which is different from T cells, where the same mutation completely abolished RelA phosphorylation (48). Thus, EC show a similar behavior as fibroblasts, where the RelA S276A mutant can still be phosphorylated (50).
Finally, our data implicate RHD as a central regulator of RelA transcriptional activity and show that the phosphorylation status of RHD can modulate the transcriptional activity of the transactivation domain. This effect seems to be independent of the transactivation domain itself in that it acts on the RelA as well as on a VP16 transactivation domain (Fig. 3B). It is worthwhile to note that both RelA and VP16 belong to the same class of acidic transactivators (57). Whether or not this regulatory effect can be extended to other classes of transactivation domains remains to be established. The RelA RHD may control the activity of the transactivation domain by several mechanisms. For one, RHD phosphorylation could induce conformational changes in the transactivation domain, facilitating interactions with components of the basal transcriptional machinery, essential for RelA transcriptional activity (56). Allosteric control of the DNA binding domain over the transactivation domain has been reported for several transcription factors (58). Therefore, it would be of interest to obtain crystal structure data of full-length RelA bound to DNA in its phosphorylated and nonphosphorylated form.
Second, the phosphorylation status of RHD may regulate interaction of RelA with nuclear cofactors such as cAMP response element-binding protein-binding protein (CBP/p300) (59,60). Although not specifically addressed in this study, it is unlikely that CBP/p300 would be a cofactor involved in the regulation of RelA transcriptional activity by PKC or p21 ras . This hypothesis is supported by the finding that the VP16 transactivation domain, which is thought not to interact with CBP/p300, is repressed by PKC mut or RasN17 when fused to RelA RHD. Furthermore, the TET/RelA2-551 construct that harbors the full-length RelA and should be phosphorylated by protein kinase A and therefore interact with CBP/p300 was only repressed when bound to a B-dependent reporter and not when bound to the TetO reporter. This result favors a model where phosphorylation of RelA RHD by PKC and p21 ras signaling pathways would modulate RelA transcriptional activity through conformational changes of DNA-bound RelA.
Another possible mechanism by which PKC and p21 ras control RelA transcriptional activity is through changes in the DNA binding activity of differently phosphorylated RHDs. It has been shown that DNA binding of RelA can be enhanced by in vitro phosphorylation through protein kinase A and protein kinase C (39). Although our studies do not show changes of in vitro DNA binding as monitored by an electrophoretic mobility shift assay (Fig. 3), there could still be such changes in vivo, since the nuclear environment is poorly reflected by an in vitro binding assay.
In summary, our data show the existence of a second NF-B regulatory system that controls transcriptional activity after liberation of NF-B complexes from their cytoplasmic inhibitors. Although we are only at the beginning of understanding this regulatory mechanism, we show evidence that it might include phosphorylation of NF-B complexes, and we propose p21 ras and PKC signaling molecules as being involved in such a control system.